* [PATCH iwl-next 01/14] virtchnl: create 'include/linux/intel' and move necessary header files
2025-04-08 12:47 [PATCH iwl-next 00/14] Introduce iXD driver Larysa Zaremba
@ 2025-04-08 12:47 ` Larysa Zaremba
2025-04-15 18:53 ` Tony Nguyen
2025-04-08 12:47 ` [PATCH iwl-next 02/14] virtchnl: introduce control plane version fields Larysa Zaremba
` (13 subsequent siblings)
14 siblings, 1 reply; 36+ messages in thread
From: Larysa Zaremba @ 2025-04-08 12:47 UTC (permalink / raw)
To: intel-wired-lan, Tony Nguyen
Cc: David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Simon Horman, Jonathan Corbet, Przemek Kitszel, Jiri Pirko,
Mustafa Ismail, Tatyana Nikolova, Andrew Lunn, Alexander Lobakin,
Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Larysa Zaremba, Sridhar Samudrala,
Jacob Keller, Michal Swiatkowski, Mateusz Polchlopek, Wenjun Wu,
Ahmed Zaki, netdev, linux-doc, linux-kernel, Karlsson, Magnus,
Emil Tantilov, Madhu Chittim, Josh Hay, Milena Olech,
pavan.kumar.linga, Singhai, Anjali
From: Victor Raj <victor.raj@intel.com>
Move intel specific header files into new folder
include/linux/intel.
Suggested-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Reviewed-by: Sridhar Samudrala <sridhar.samudrala@intel.com>
Signed-off-by: Victor Raj <victor.raj@intel.com>
Signed-off-by: Larysa Zaremba <larysa.zaremba@intel.com>
---
drivers/infiniband/hw/irdma/i40iw_if.c | 2 +-
drivers/infiniband/hw/irdma/main.h | 2 +-
drivers/infiniband/hw/irdma/osdep.h | 2 +-
drivers/net/ethernet/intel/i40e/i40e.h | 4 ++--
drivers/net/ethernet/intel/i40e/i40e_client.c | 2 +-
drivers/net/ethernet/intel/i40e/i40e_common.c | 2 +-
drivers/net/ethernet/intel/i40e/i40e_prototype.h | 2 +-
drivers/net/ethernet/intel/i40e/i40e_txrx.c | 2 +-
drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.h | 2 +-
drivers/net/ethernet/intel/iavf/iavf.h | 2 +-
drivers/net/ethernet/intel/iavf/iavf_common.c | 2 +-
drivers/net/ethernet/intel/iavf/iavf_main.c | 2 +-
drivers/net/ethernet/intel/iavf/iavf_prototype.h | 2 +-
drivers/net/ethernet/intel/iavf/iavf_txrx.c | 2 +-
drivers/net/ethernet/intel/iavf/iavf_types.h | 4 +---
drivers/net/ethernet/intel/iavf/iavf_virtchnl.c | 2 +-
drivers/net/ethernet/intel/ice/ice.h | 2 +-
drivers/net/ethernet/intel/ice/ice_common.h | 2 +-
drivers/net/ethernet/intel/ice/ice_idc_int.h | 2 +-
drivers/net/ethernet/intel/ice/ice_txrx_lib.c | 2 +-
drivers/net/ethernet/intel/ice/ice_vf_lib.h | 2 +-
drivers/net/ethernet/intel/ice/ice_virtchnl.h | 2 +-
drivers/net/ethernet/intel/idpf/idpf.h | 2 +-
drivers/net/ethernet/intel/idpf/idpf_txrx.h | 2 +-
drivers/net/ethernet/intel/libie/rx.c | 2 +-
include/linux/{net => }/intel/i40e_client.h | 0
include/linux/{net => }/intel/iidc.h | 0
include/linux/{net => }/intel/libie/rx.h | 0
include/linux/{avf => intel}/virtchnl.h | 0
.../ethernet/intel/idpf => include/linux/intel}/virtchnl2.h | 0
.../intel/idpf => include/linux/intel}/virtchnl2_lan_desc.h | 0
31 files changed, 26 insertions(+), 28 deletions(-)
rename include/linux/{net => }/intel/i40e_client.h (100%)
rename include/linux/{net => }/intel/iidc.h (100%)
rename include/linux/{net => }/intel/libie/rx.h (100%)
rename include/linux/{avf => intel}/virtchnl.h (100%)
rename {drivers/net/ethernet/intel/idpf => include/linux/intel}/virtchnl2.h (100%)
rename {drivers/net/ethernet/intel/idpf => include/linux/intel}/virtchnl2_lan_desc.h (100%)
diff --git a/drivers/infiniband/hw/irdma/i40iw_if.c b/drivers/infiniband/hw/irdma/i40iw_if.c
index cc50a7070371..cfd71fe5bd79 100644
--- a/drivers/infiniband/hw/irdma/i40iw_if.c
+++ b/drivers/infiniband/hw/irdma/i40iw_if.c
@@ -2,7 +2,7 @@
/* Copyright (c) 2015 - 2021 Intel Corporation */
#include "main.h"
#include "i40iw_hw.h"
-#include <linux/net/intel/i40e_client.h>
+#include <linux/intel/i40e_client.h>
static struct i40e_client i40iw_client;
diff --git a/drivers/infiniband/hw/irdma/main.h b/drivers/infiniband/hw/irdma/main.h
index bb0b6494ccb2..d6e446dec2be 100644
--- a/drivers/infiniband/hw/irdma/main.h
+++ b/drivers/infiniband/hw/irdma/main.h
@@ -29,7 +29,7 @@
#include <linux/io-64-nonatomic-lo-hi.h>
#endif
#include <linux/auxiliary_bus.h>
-#include <linux/net/intel/iidc.h>
+#include <linux/intel/iidc.h>
#include <rdma/ib_smi.h>
#include <rdma/ib_verbs.h>
#include <rdma/ib_pack.h>
diff --git a/drivers/infiniband/hw/irdma/osdep.h b/drivers/infiniband/hw/irdma/osdep.h
index 4b4f78288d12..abb10e35d33b 100644
--- a/drivers/infiniband/hw/irdma/osdep.h
+++ b/drivers/infiniband/hw/irdma/osdep.h
@@ -5,7 +5,7 @@
#include <linux/pci.h>
#include <linux/bitfield.h>
-#include <linux/net/intel/iidc.h>
+#include <linux/intel/iidc.h>
#include <rdma/ib_verbs.h>
#define STATS_TIMER_DELAY 60000
diff --git a/drivers/net/ethernet/intel/i40e/i40e.h b/drivers/net/ethernet/intel/i40e/i40e.h
index c67963bfe14e..287ddfec5024 100644
--- a/drivers/net/ethernet/intel/i40e/i40e.h
+++ b/drivers/net/ethernet/intel/i40e/i40e.h
@@ -8,8 +8,8 @@
#include <linux/pci.h>
#include <linux/ptp_clock_kernel.h>
#include <linux/types.h>
-#include <linux/avf/virtchnl.h>
-#include <linux/net/intel/i40e_client.h>
+#include <linux/intel/virtchnl.h>
+#include <linux/intel/i40e_client.h>
#include <net/devlink.h>
#include <net/pkt_cls.h>
#include <net/udp_tunnel.h>
diff --git a/drivers/net/ethernet/intel/i40e/i40e_client.c b/drivers/net/ethernet/intel/i40e/i40e_client.c
index 59263551c383..a6a6d12ec250 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_client.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_client.c
@@ -3,7 +3,7 @@
#include <linux/list.h>
#include <linux/errno.h>
-#include <linux/net/intel/i40e_client.h>
+#include <linux/intel/i40e_client.h>
#include "i40e.h"
diff --git a/drivers/net/ethernet/intel/i40e/i40e_common.c b/drivers/net/ethernet/intel/i40e/i40e_common.c
index b11c35e307ca..cd82d4069861 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_common.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_common.c
@@ -1,7 +1,7 @@
// SPDX-License-Identifier: GPL-2.0
/* Copyright(c) 2013 - 2021 Intel Corporation. */
-#include <linux/avf/virtchnl.h>
+#include <linux/intel/virtchnl.h>
#include <linux/bitfield.h>
#include <linux/delay.h>
#include <linux/etherdevice.h>
diff --git a/drivers/net/ethernet/intel/i40e/i40e_prototype.h b/drivers/net/ethernet/intel/i40e/i40e_prototype.h
index 099bb8ab7d70..eae76404289a 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_prototype.h
+++ b/drivers/net/ethernet/intel/i40e/i40e_prototype.h
@@ -5,7 +5,7 @@
#define _I40E_PROTOTYPE_H_
#include <linux/ethtool.h>
-#include <linux/avf/virtchnl.h>
+#include <linux/intel/virtchnl.h>
#include "i40e_debug.h"
#include "i40e_type.h"
diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
index c006f716a3bd..30df45a146da 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
@@ -2,7 +2,7 @@
/* Copyright(c) 2013 - 2018 Intel Corporation. */
#include <linux/bpf_trace.h>
-#include <linux/net/intel/libie/rx.h>
+#include <linux/intel/libie/rx.h>
#include <linux/prefetch.h>
#include <linux/sctp.h>
#include <net/mpls.h>
diff --git a/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.h b/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.h
index 5cf74f16f433..acaae2f6d00b 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.h
+++ b/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.h
@@ -4,7 +4,7 @@
#ifndef _I40E_VIRTCHNL_PF_H_
#define _I40E_VIRTCHNL_PF_H_
-#include <linux/avf/virtchnl.h>
+#include <linux/intel/virtchnl.h>
#include <linux/netdevice.h>
#include "i40e_type.h"
diff --git a/drivers/net/ethernet/intel/iavf/iavf.h b/drivers/net/ethernet/intel/iavf/iavf.h
index 9de3e0ba3731..cf81c308f92b 100644
--- a/drivers/net/ethernet/intel/iavf/iavf.h
+++ b/drivers/net/ethernet/intel/iavf/iavf.h
@@ -37,7 +37,7 @@
#include <net/net_shaper.h>
#include "iavf_type.h"
-#include <linux/avf/virtchnl.h>
+#include <linux/intel/virtchnl.h>
#include "iavf_txrx.h"
#include "iavf_fdir.h"
#include "iavf_adv_rss.h"
diff --git a/drivers/net/ethernet/intel/iavf/iavf_common.c b/drivers/net/ethernet/intel/iavf/iavf_common.c
index aa751ce3425b..e9968e540110 100644
--- a/drivers/net/ethernet/intel/iavf/iavf_common.c
+++ b/drivers/net/ethernet/intel/iavf/iavf_common.c
@@ -1,7 +1,7 @@
// SPDX-License-Identifier: GPL-2.0
/* Copyright(c) 2013 - 2018 Intel Corporation. */
-#include <linux/avf/virtchnl.h>
+#include <linux/intel/virtchnl.h>
#include <linux/bitfield.h>
#include "iavf_type.h"
#include "iavf_adminq.h"
diff --git a/drivers/net/ethernet/intel/iavf/iavf_main.c b/drivers/net/ethernet/intel/iavf/iavf_main.c
index 6d7ba4d67a19..f396fd1d78f1 100644
--- a/drivers/net/ethernet/intel/iavf/iavf_main.c
+++ b/drivers/net/ethernet/intel/iavf/iavf_main.c
@@ -1,7 +1,7 @@
// SPDX-License-Identifier: GPL-2.0
/* Copyright(c) 2013 - 2018 Intel Corporation. */
-#include <linux/net/intel/libie/rx.h>
+#include <linux/intel/libie/rx.h>
#include <net/netdev_lock.h>
#include "iavf.h"
diff --git a/drivers/net/ethernet/intel/iavf/iavf_prototype.h b/drivers/net/ethernet/intel/iavf/iavf_prototype.h
index cac9d1a35a52..945be3541e8e 100644
--- a/drivers/net/ethernet/intel/iavf/iavf_prototype.h
+++ b/drivers/net/ethernet/intel/iavf/iavf_prototype.h
@@ -6,7 +6,7 @@
#include "iavf_type.h"
#include "iavf_alloc.h"
-#include <linux/avf/virtchnl.h>
+#include <linux/intel/virtchnl.h>
/* Prototypes for shared code functions that are not in
* the standard function pointer structures. These are
diff --git a/drivers/net/ethernet/intel/iavf/iavf_txrx.c b/drivers/net/ethernet/intel/iavf/iavf_txrx.c
index 422312b8b54a..4c8dca868447 100644
--- a/drivers/net/ethernet/intel/iavf/iavf_txrx.c
+++ b/drivers/net/ethernet/intel/iavf/iavf_txrx.c
@@ -2,7 +2,7 @@
/* Copyright(c) 2013 - 2018 Intel Corporation. */
#include <linux/bitfield.h>
-#include <linux/net/intel/libie/rx.h>
+#include <linux/intel/libie/rx.h>
#include <linux/prefetch.h>
#include "iavf.h"
diff --git a/drivers/net/ethernet/intel/iavf/iavf_types.h b/drivers/net/ethernet/intel/iavf/iavf_types.h
index a095855122bf..270bc35f933d 100644
--- a/drivers/net/ethernet/intel/iavf/iavf_types.h
+++ b/drivers/net/ethernet/intel/iavf/iavf_types.h
@@ -4,9 +4,7 @@
#ifndef _IAVF_TYPES_H_
#define _IAVF_TYPES_H_
-#include "iavf_types.h"
-
-#include <linux/avf/virtchnl.h>
+#include <linux/intel/virtchnl.h>
#include <linux/ptp_clock_kernel.h>
/* structure used to queue PTP commands for processing */
diff --git a/drivers/net/ethernet/intel/iavf/iavf_virtchnl.c b/drivers/net/ethernet/intel/iavf/iavf_virtchnl.c
index a6f0e5990be2..c0ba1fd3b98a 100644
--- a/drivers/net/ethernet/intel/iavf/iavf_virtchnl.c
+++ b/drivers/net/ethernet/intel/iavf/iavf_virtchnl.c
@@ -1,7 +1,7 @@
// SPDX-License-Identifier: GPL-2.0
/* Copyright(c) 2013 - 2018 Intel Corporation. */
-#include <linux/net/intel/libie/rx.h>
+#include <linux/intel/libie/rx.h>
#include "iavf.h"
#include "iavf_ptp.h"
diff --git a/drivers/net/ethernet/intel/ice/ice.h b/drivers/net/ethernet/intel/ice/ice.h
index fc127c0f9d66..cacb18a18564 100644
--- a/drivers/net/ethernet/intel/ice/ice.h
+++ b/drivers/net/ethernet/intel/ice/ice.h
@@ -36,7 +36,7 @@
#include <linux/bpf.h>
#include <linux/btf.h>
#include <linux/auxiliary_bus.h>
-#include <linux/avf/virtchnl.h>
+#include <linux/intel/virtchnl.h>
#include <linux/cpu_rmap.h>
#include <linux/dim.h>
#include <linux/gnss.h>
diff --git a/drivers/net/ethernet/intel/ice/ice_common.h b/drivers/net/ethernet/intel/ice/ice_common.h
index aefcf719e460..f7cd636668eb 100644
--- a/drivers/net/ethernet/intel/ice/ice_common.h
+++ b/drivers/net/ethernet/intel/ice/ice_common.h
@@ -11,7 +11,7 @@
#include "ice_nvm.h"
#include "ice_flex_pipe.h"
#include "ice_parser.h"
-#include <linux/avf/virtchnl.h>
+#include <linux/intel/virtchnl.h>
#include "ice_switch.h"
#include "ice_fdir.h"
diff --git a/drivers/net/ethernet/intel/ice/ice_idc_int.h b/drivers/net/ethernet/intel/ice/ice_idc_int.h
index 4b0c86757df9..3d35118dee90 100644
--- a/drivers/net/ethernet/intel/ice/ice_idc_int.h
+++ b/drivers/net/ethernet/intel/ice/ice_idc_int.h
@@ -4,7 +4,7 @@
#ifndef _ICE_IDC_INT_H_
#define _ICE_IDC_INT_H_
-#include <linux/net/intel/iidc.h>
+#include <linux/intel/iidc.h>
struct ice_pf;
diff --git a/drivers/net/ethernet/intel/ice/ice_txrx_lib.c b/drivers/net/ethernet/intel/ice/ice_txrx_lib.c
index 45cfaabc41cb..fb90916e7ef7 100644
--- a/drivers/net/ethernet/intel/ice/ice_txrx_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_txrx_lib.c
@@ -2,7 +2,7 @@
/* Copyright (c) 2019, Intel Corporation. */
#include <linux/filter.h>
-#include <linux/net/intel/libie/rx.h>
+#include <linux/intel/libie/rx.h>
#include "ice_txrx_lib.h"
#include "ice_eswitch.h"
diff --git a/drivers/net/ethernet/intel/ice/ice_vf_lib.h b/drivers/net/ethernet/intel/ice/ice_vf_lib.h
index 482f4285fd35..5cdc98985523 100644
--- a/drivers/net/ethernet/intel/ice/ice_vf_lib.h
+++ b/drivers/net/ethernet/intel/ice/ice_vf_lib.h
@@ -10,7 +10,7 @@
#include <linux/mutex.h>
#include <linux/pci.h>
#include <net/devlink.h>
-#include <linux/avf/virtchnl.h>
+#include <linux/intel/virtchnl.h>
#include "ice_type.h"
#include "ice_flow.h"
#include "ice_virtchnl_fdir.h"
diff --git a/drivers/net/ethernet/intel/ice/ice_virtchnl.h b/drivers/net/ethernet/intel/ice/ice_virtchnl.h
index 222990f229d5..ca3830617a51 100644
--- a/drivers/net/ethernet/intel/ice/ice_virtchnl.h
+++ b/drivers/net/ethernet/intel/ice/ice_virtchnl.h
@@ -7,7 +7,7 @@
#include <linux/types.h>
#include <linux/bitops.h>
#include <linux/if_ether.h>
-#include <linux/avf/virtchnl.h>
+#include <linux/intel/virtchnl.h>
#include "ice_vf_lib.h"
/* Restrict number of MAC Addr and VLAN that non-trusted VF can programmed */
diff --git a/drivers/net/ethernet/intel/idpf/idpf.h b/drivers/net/ethernet/intel/idpf/idpf.h
index 5f73a4cf5161..c42dc828a498 100644
--- a/drivers/net/ethernet/intel/idpf/idpf.h
+++ b/drivers/net/ethernet/intel/idpf/idpf.h
@@ -18,7 +18,7 @@ struct idpf_vport_max_q;
#include <linux/ethtool_netlink.h>
#include <net/gro.h>
-#include "virtchnl2.h"
+#include <linux/intel/virtchnl2.h>
#include "idpf_txrx.h"
#include "idpf_controlq.h"
diff --git a/drivers/net/ethernet/intel/idpf/idpf_txrx.h b/drivers/net/ethernet/intel/idpf/idpf_txrx.h
index b029f566e57c..9f0cbb164054 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_txrx.h
+++ b/drivers/net/ethernet/intel/idpf/idpf_txrx.h
@@ -11,7 +11,7 @@
#include <net/netdev_queues.h>
#include "idpf_lan_txrx.h"
-#include "virtchnl2_lan_desc.h"
+#include <linux/intel/virtchnl2_lan_desc.h>
#define IDPF_LARGE_MAX_Q 256
#define IDPF_MAX_Q 16
diff --git a/drivers/net/ethernet/intel/libie/rx.c b/drivers/net/ethernet/intel/libie/rx.c
index 66a9825fe11f..668785974727 100644
--- a/drivers/net/ethernet/intel/libie/rx.c
+++ b/drivers/net/ethernet/intel/libie/rx.c
@@ -1,7 +1,7 @@
// SPDX-License-Identifier: GPL-2.0-only
/* Copyright (C) 2024 Intel Corporation */
-#include <linux/net/intel/libie/rx.h>
+#include <linux/intel/libie/rx.h>
/* O(1) converting i40e/ice/iavf's 8/10-bit hardware packet type to a parsed
* bitfield struct.
diff --git a/include/linux/net/intel/i40e_client.h b/include/linux/intel/i40e_client.h
similarity index 100%
rename from include/linux/net/intel/i40e_client.h
rename to include/linux/intel/i40e_client.h
diff --git a/include/linux/net/intel/iidc.h b/include/linux/intel/iidc.h
similarity index 100%
rename from include/linux/net/intel/iidc.h
rename to include/linux/intel/iidc.h
diff --git a/include/linux/net/intel/libie/rx.h b/include/linux/intel/libie/rx.h
similarity index 100%
rename from include/linux/net/intel/libie/rx.h
rename to include/linux/intel/libie/rx.h
diff --git a/include/linux/avf/virtchnl.h b/include/linux/intel/virtchnl.h
similarity index 100%
rename from include/linux/avf/virtchnl.h
rename to include/linux/intel/virtchnl.h
diff --git a/drivers/net/ethernet/intel/idpf/virtchnl2.h b/include/linux/intel/virtchnl2.h
similarity index 100%
rename from drivers/net/ethernet/intel/idpf/virtchnl2.h
rename to include/linux/intel/virtchnl2.h
diff --git a/drivers/net/ethernet/intel/idpf/virtchnl2_lan_desc.h b/include/linux/intel/virtchnl2_lan_desc.h
similarity index 100%
rename from drivers/net/ethernet/intel/idpf/virtchnl2_lan_desc.h
rename to include/linux/intel/virtchnl2_lan_desc.h
--
2.47.0
^ permalink raw reply related [flat|nested] 36+ messages in thread* Re: [PATCH iwl-next 01/14] virtchnl: create 'include/linux/intel' and move necessary header files
2025-04-08 12:47 ` [PATCH iwl-next 01/14] virtchnl: create 'include/linux/intel' and move necessary header files Larysa Zaremba
@ 2025-04-15 18:53 ` Tony Nguyen
0 siblings, 0 replies; 36+ messages in thread
From: Tony Nguyen @ 2025-04-15 18:53 UTC (permalink / raw)
To: Larysa Zaremba, intel-wired-lan
Cc: David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Simon Horman, Jonathan Corbet, Przemek Kitszel, Jiri Pirko,
Mustafa Ismail, Tatyana Nikolova, Andrew Lunn, Alexander Lobakin,
Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Sridhar Samudrala, Jacob Keller,
Michal Swiatkowski, Mateusz Polchlopek, Wenjun Wu, Ahmed Zaki,
netdev, linux-doc, linux-kernel, Karlsson, Magnus, Emil Tantilov,
Madhu Chittim, Josh Hay, Milena Olech, pavan.kumar.linga,
Singhai, Anjali
On 4/8/2025 5:47 AM, Larysa Zaremba wrote:
> From: Victor Raj <victor.raj@intel.com>
>
> Move intel specific header files into new folder
> include/linux/intel.
I believe MAINTAINERS needs to be updated since the existing paths don't
cover this.
Thanks,
Tony
> Suggested-by: Alexander Lobakin <aleksander.lobakin@intel.com>
> Reviewed-by: Sridhar Samudrala <sridhar.samudrala@intel.com>
> Signed-off-by: Victor Raj <victor.raj@intel.com>
> Signed-off-by: Larysa Zaremba <larysa.zaremba@intel.com>
> ---
> drivers/infiniband/hw/irdma/i40iw_if.c | 2 +-
> drivers/infiniband/hw/irdma/main.h | 2 +-
> drivers/infiniband/hw/irdma/osdep.h | 2 +-
> drivers/net/ethernet/intel/i40e/i40e.h | 4 ++--
> drivers/net/ethernet/intel/i40e/i40e_client.c | 2 +-
> drivers/net/ethernet/intel/i40e/i40e_common.c | 2 +-
> drivers/net/ethernet/intel/i40e/i40e_prototype.h | 2 +-
> drivers/net/ethernet/intel/i40e/i40e_txrx.c | 2 +-
> drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.h | 2 +-
> drivers/net/ethernet/intel/iavf/iavf.h | 2 +-
> drivers/net/ethernet/intel/iavf/iavf_common.c | 2 +-
> drivers/net/ethernet/intel/iavf/iavf_main.c | 2 +-
> drivers/net/ethernet/intel/iavf/iavf_prototype.h | 2 +-
> drivers/net/ethernet/intel/iavf/iavf_txrx.c | 2 +-
> drivers/net/ethernet/intel/iavf/iavf_types.h | 4 +---
> drivers/net/ethernet/intel/iavf/iavf_virtchnl.c | 2 +-
> drivers/net/ethernet/intel/ice/ice.h | 2 +-
> drivers/net/ethernet/intel/ice/ice_common.h | 2 +-
> drivers/net/ethernet/intel/ice/ice_idc_int.h | 2 +-
> drivers/net/ethernet/intel/ice/ice_txrx_lib.c | 2 +-
> drivers/net/ethernet/intel/ice/ice_vf_lib.h | 2 +-
> drivers/net/ethernet/intel/ice/ice_virtchnl.h | 2 +-
> drivers/net/ethernet/intel/idpf/idpf.h | 2 +-
> drivers/net/ethernet/intel/idpf/idpf_txrx.h | 2 +-
> drivers/net/ethernet/intel/libie/rx.c | 2 +-
> include/linux/{net => }/intel/i40e_client.h | 0
> include/linux/{net => }/intel/iidc.h | 0
> include/linux/{net => }/intel/libie/rx.h | 0
> include/linux/{avf => intel}/virtchnl.h | 0
> .../ethernet/intel/idpf => include/linux/intel}/virtchnl2.h | 0
> .../intel/idpf => include/linux/intel}/virtchnl2_lan_desc.h | 0
> 31 files changed, 26 insertions(+), 28 deletions(-)
> rename include/linux/{net => }/intel/i40e_client.h (100%)
> rename include/linux/{net => }/intel/iidc.h (100%)
> rename include/linux/{net => }/intel/libie/rx.h (100%)
> rename include/linux/{avf => intel}/virtchnl.h (100%)
> rename {drivers/net/ethernet/intel/idpf => include/linux/intel}/virtchnl2.h (100%)
> rename {drivers/net/ethernet/intel/idpf => include/linux/intel}/virtchnl2_lan_desc.h (100%)
^ permalink raw reply [flat|nested] 36+ messages in thread
* [PATCH iwl-next 02/14] virtchnl: introduce control plane version fields
2025-04-08 12:47 [PATCH iwl-next 00/14] Introduce iXD driver Larysa Zaremba
2025-04-08 12:47 ` [PATCH iwl-next 01/14] virtchnl: create 'include/linux/intel' and move necessary header files Larysa Zaremba
@ 2025-04-08 12:47 ` Larysa Zaremba
2025-04-08 12:47 ` [PATCH iwl-next 03/14] libeth: add PCI device initialization helpers to libeth Larysa Zaremba
` (12 subsequent siblings)
14 siblings, 0 replies; 36+ messages in thread
From: Larysa Zaremba @ 2025-04-08 12:47 UTC (permalink / raw)
To: intel-wired-lan, Tony Nguyen
Cc: David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Simon Horman, Jonathan Corbet, Przemek Kitszel, Jiri Pirko,
Mustafa Ismail, Tatyana Nikolova, Andrew Lunn, Alexander Lobakin,
Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Larysa Zaremba, Sridhar Samudrala,
Jacob Keller, Michal Swiatkowski, Mateusz Polchlopek, Wenjun Wu,
Ahmed Zaki, netdev, linux-doc, linux-kernel, Karlsson, Magnus,
Emil Tantilov, Madhu Chittim, Josh Hay, Milena Olech,
pavan.kumar.linga, Singhai, Anjali
From: Victor Raj <victor.raj@intel.com>
In the virtchnl header file, add the Control Plane software
version fields.
Signed-off-by: Victor Raj <victor.raj@intel.com>
Reviewed-by: Sridhar Samudrala <sridhar.samudrala@intel.com>
Signed-off-by: Larysa Zaremba <larysa.zaremba@intel.com>
---
include/linux/intel/virtchnl2.h | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/include/linux/intel/virtchnl2.h b/include/linux/intel/virtchnl2.h
index 63deb120359c..0f31d65823f2 100644
--- a/include/linux/intel/virtchnl2.h
+++ b/include/linux/intel/virtchnl2.h
@@ -467,7 +467,8 @@ VIRTCHNL2_CHECK_STRUCT_LEN(8, virtchnl2_version_info);
* sent per transmit packet without needing to be
* linearized.
* @pad: Padding.
- * @reserved: Reserved.
+ * @cp_ver_major: Control Plane major version number.
+ * @cp_ver_minor: Control Plane minor version number.
* @device_type: See enum virtchl2_device_type.
* @min_sso_packet_len: Min packet length supported by device for single
* segment offload.
@@ -516,7 +517,8 @@ struct virtchnl2_get_capabilities {
__le16 max_tx_hdr_size;
u8 max_sg_bufs_per_tx_pkt;
u8 pad[3];
- u8 reserved[4];
+ __le16 cp_ver_major;
+ __le16 cp_ver_minor;
__le32 device_type;
u8 min_sso_packet_len;
u8 max_hdr_buf_per_lso;
--
2.47.0
^ permalink raw reply related [flat|nested] 36+ messages in thread* [PATCH iwl-next 03/14] libeth: add PCI device initialization helpers to libeth
2025-04-08 12:47 [PATCH iwl-next 00/14] Introduce iXD driver Larysa Zaremba
2025-04-08 12:47 ` [PATCH iwl-next 01/14] virtchnl: create 'include/linux/intel' and move necessary header files Larysa Zaremba
2025-04-08 12:47 ` [PATCH iwl-next 02/14] virtchnl: introduce control plane version fields Larysa Zaremba
@ 2025-04-08 12:47 ` Larysa Zaremba
2025-04-21 14:08 ` Simon Horman
2025-04-08 12:47 ` [PATCH iwl-next 04/14] libeth: allow to create fill queues without NAPI Larysa Zaremba
` (11 subsequent siblings)
14 siblings, 1 reply; 36+ messages in thread
From: Larysa Zaremba @ 2025-04-08 12:47 UTC (permalink / raw)
To: intel-wired-lan, Tony Nguyen
Cc: David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Simon Horman, Jonathan Corbet, Przemek Kitszel, Jiri Pirko,
Mustafa Ismail, Tatyana Nikolova, Andrew Lunn, Alexander Lobakin,
Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Larysa Zaremba, Sridhar Samudrala,
Jacob Keller, Michal Swiatkowski, Mateusz Polchlopek, Wenjun Wu,
Ahmed Zaki, netdev, linux-doc, linux-kernel, Karlsson, Magnus,
Emil Tantilov, Madhu Chittim, Josh Hay, Milena Olech,
pavan.kumar.linga, Singhai, Anjali, Phani R Burra
From: Phani R Burra <phani.r.burra@intel.com>
Add memory related support functions for drivers to access MMIO space and
allocate/free dma buffers.
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Phani R Burra <phani.r.burra@intel.com>
Co-developed-by: Victor Raj <victor.raj@intel.com>
Signed-off-by: Victor Raj <victor.raj@intel.com>
Co-developed-by: Sridhar Samudrala <sridhar.samudrala@intel.com>
Signed-off-by: Sridhar Samudrala <sridhar.samudrala@intel.com>
Co-developed-by: Pavan Kumar Linga <pavan.kumar.linga@intel.com>
Signed-off-by: Pavan Kumar Linga <pavan.kumar.linga@intel.com>
Co-developed-by: Larysa Zaremba <larysa.zaremba@intel.com>
Signed-off-by: Larysa Zaremba <larysa.zaremba@intel.com>
---
drivers/net/ethernet/intel/libeth/Kconfig | 6 +
drivers/net/ethernet/intel/libeth/Makefile | 4 +
drivers/net/ethernet/intel/libeth/pci.c | 184 +++++++++++++++++++++
include/net/libeth/pci.h | 55 ++++++
4 files changed, 249 insertions(+)
create mode 100644 drivers/net/ethernet/intel/libeth/pci.c
create mode 100644 include/net/libeth/pci.h
diff --git a/drivers/net/ethernet/intel/libeth/Kconfig b/drivers/net/ethernet/intel/libeth/Kconfig
index 480293b71dbc..070ad685fb74 100644
--- a/drivers/net/ethernet/intel/libeth/Kconfig
+++ b/drivers/net/ethernet/intel/libeth/Kconfig
@@ -7,3 +7,9 @@ config LIBETH
help
libeth is a common library containing routines shared between several
drivers, but not yet promoted to the generic kernel API.
+
+config LIBETH_PCI
+ tristate
+ help
+ Helper functions for management of PCI resources belonging
+ to networking devices.
diff --git a/drivers/net/ethernet/intel/libeth/Makefile b/drivers/net/ethernet/intel/libeth/Makefile
index 52492b081132..5f0fc9446acb 100644
--- a/drivers/net/ethernet/intel/libeth/Makefile
+++ b/drivers/net/ethernet/intel/libeth/Makefile
@@ -4,3 +4,7 @@
obj-$(CONFIG_LIBETH) += libeth.o
libeth-y := rx.o
+
+obj-$(CONFIG_LIBETH_PCI) += libeth_pci.o
+
+libeth_pci-y := pci.o
diff --git a/drivers/net/ethernet/intel/libeth/pci.c b/drivers/net/ethernet/intel/libeth/pci.c
new file mode 100644
index 000000000000..63cf56a07ce9
--- /dev/null
+++ b/drivers/net/ethernet/intel/libeth/pci.c
@@ -0,0 +1,184 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/* Copyright (C) 2025 Intel Corporation */
+
+#include <net/libeth/pci.h>
+
+/**
+ * libeth_find_mmio_region - find if MMIO region is present in the list
+ * @mmio_list: list that contains MMIO region info
+ * @offset: MMIO region start offset
+ * @bar_idx: BAR index where the offset to search
+ *
+ * Return: MMIO region pointer or NULL if the region info is not present.
+ */
+static struct libeth_pci_mmio_region *
+libeth_find_mmio_region(const struct list_head *mmio_list,
+ resource_size_t offset, int bar_idx)
+{
+ struct libeth_pci_mmio_region *mr;
+
+ list_for_each_entry(mr, mmio_list, list)
+ if (mr->bar_idx == bar_idx && mr->offset == offset)
+ return mr;
+
+ return NULL;
+}
+
+/**
+ * __libeth_pci_get_mmio_addr - get the MMIO virtual address
+ * @mmio_info: contains list of MMIO regions
+ * @offset: register offset of find
+ * @num_args: number of additional arguments present
+ *
+ * This function finds the virtual address of a register offset by iterating
+ * through the non-linear MMIO regions that are mapped by the driver.
+ *
+ * Return: valid MMIO virtual address or NULL.
+ */
+void __iomem *__libeth_pci_get_mmio_addr(struct libeth_mmio_info *mmio_info,
+ resource_size_t offset,
+ int num_args, ...)
+{
+ struct libeth_pci_mmio_region *mr;
+ int bar_idx = 0;
+ va_list args;
+
+ if (num_args) {
+ va_start(args, num_args);
+ bar_idx = va_arg(args, int);
+ va_end(args);
+ }
+
+ list_for_each_entry(mr, &mmio_info->mmio_list, list)
+ if (bar_idx == mr->bar_idx && offset >= mr->offset &&
+ offset < mr->offset + mr->size) {
+ offset -= mr->offset;
+
+ return mr->addr + offset;
+ }
+
+ return NULL;
+}
+EXPORT_SYMBOL_NS_GPL(__libeth_pci_get_mmio_addr, "LIBETH_PCI");
+
+/**
+ * __libeth_pci_map_mmio_region - map PCI device MMIO region
+ * @mmio_info: struct to store the mapped MMIO region
+ * @offset: MMIO region start offset
+ * @size: MMIO region size
+ * @num_args: number of additional arguments present
+ *
+ * Return: true on success, false on memory map failure.
+ */
+bool __libeth_pci_map_mmio_region(struct libeth_mmio_info *mmio_info,
+ resource_size_t offset,
+ resource_size_t size, int num_args, ...)
+{
+ struct pci_dev *pdev = mmio_info->pdev;
+ struct libeth_pci_mmio_region *mr;
+ resource_size_t pa;
+ void __iomem *va;
+ int bar_idx = 0;
+ va_list args;
+
+ if (num_args) {
+ va_start(args, num_args);
+ bar_idx = va_arg(args, int);
+ va_end(args);
+ }
+
+ mr = libeth_find_mmio_region(&mmio_info->mmio_list, offset, bar_idx);
+ if (mr) {
+ pci_warn(pdev, "Mapping of BAR%u with offset %llu already exists\n",
+ bar_idx, offset);
+ return true;
+ }
+
+ pa = pci_resource_start(pdev, bar_idx) + offset;
+ va = ioremap(pa, size);
+ if (!va) {
+ pci_err(pdev, "Failed to allocate BAR%u region\n", bar_idx);
+ return false;
+ }
+
+ mr = kvzalloc(sizeof(*mr), GFP_KERNEL);
+ if (!mr) {
+ iounmap(va);
+ return false;
+ }
+
+ mr->addr = va;
+ mr->offset = offset;
+ mr->size = size;
+ mr->bar_idx = bar_idx;
+
+ list_add_tail(&mr->list, &mmio_info->mmio_list);
+
+ return true;
+}
+EXPORT_SYMBOL_NS_GPL(__libeth_pci_map_mmio_region, "LIBETH_PCI");
+
+/**
+ * libeth_pci_unmap_all_mmio_regions - unmap all PCI device MMIO regions
+ * @mmio_info: contains list of MMIO regions to unmap
+ */
+void libeth_pci_unmap_all_mmio_regions(struct libeth_mmio_info *mmio_info)
+{
+ struct libeth_pci_mmio_region *mr, *tmp;
+
+ list_for_each_entry_safe(mr, tmp, &mmio_info->mmio_list, list) {
+ iounmap(mr->addr);
+ list_del(&mr->list);
+ kfree(mr);
+ }
+}
+EXPORT_SYMBOL_NS_GPL(libeth_pci_unmap_all_mmio_regions, "LIBETH_PCI");
+
+/**
+ * libeth_pci_init_dev - enable and reserve PCI regions of the device
+ * @pdev: PCI device information
+ *
+ * Return: %0 on success, -%errno on failure.
+ */
+int libeth_pci_init_dev(struct pci_dev *pdev)
+{
+ int err;
+
+ err = pci_enable_device(pdev);
+ if (err)
+ return err;
+
+ err = pci_request_mem_regions(pdev, pci_name(pdev));
+ if (err)
+ goto disable_dev;
+
+ err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
+ if (err)
+ goto rel_regions;
+
+ pci_set_master(pdev);
+
+ return 0;
+
+rel_regions:
+ pci_release_regions(pdev);
+disable_dev:
+ pci_disable_device(pdev);
+
+ return err;
+}
+EXPORT_SYMBOL_NS_GPL(libeth_pci_init_dev, "LIBETH_PCI");
+
+/**
+ * libeth_pci_deinit_dev - disable and release the PCI regions of the device
+ * @pdev: PCI device information
+ */
+void libeth_pci_deinit_dev(struct pci_dev *pdev)
+{
+ pci_disable_device(pdev);
+ pci_release_regions(pdev);
+}
+EXPORT_SYMBOL_NS_GPL(libeth_pci_deinit_dev, "LIBETH_PCI");
+
+MODULE_DESCRIPTION("Common Ethernet PCI library");
+MODULE_LICENSE("GPL");
diff --git a/include/net/libeth/pci.h b/include/net/libeth/pci.h
new file mode 100644
index 000000000000..6ae8032c2c8a
--- /dev/null
+++ b/include/net/libeth/pci.h
@@ -0,0 +1,55 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/* Copyright (C) 2025 Intel Corporation */
+
+#ifndef __LIBETH_PCI_H
+#define __LIBETH_PCI_H
+
+#include <linux/pci.h>
+
+/**
+ * struct libeth_pci_mmio_region - structure for MMIO region info
+ * @list: used to add a MMIO region to the list of MMIO regions in
+ * libeth_mmio_info
+ * @addr: virtual address of MMIO region start
+ * @offset: start offset of the MMIO region
+ * @size: size of the MMIO region
+ * @bar_idx: BAR index to which the MMIO region belongs to
+ */
+struct libeth_pci_mmio_region {
+ struct list_head list;
+ void __iomem *addr;
+ resource_size_t offset;
+ resource_size_t size;
+ u16 bar_idx;
+};
+
+/**
+ * struct libeth_mmio_info - contains list of MMIO regions
+ * @pdev: PCI device pointer
+ * @mmio_list: list of MMIO regions
+ */
+struct libeth_mmio_info {
+ struct pci_dev *pdev;
+ struct list_head mmio_list;
+};
+
+#define libeth_pci_map_mmio_region(mmio_info, offset, size, ...) \
+ __libeth_pci_map_mmio_region(mmio_info, offset, size, \
+ COUNT_ARGS(__VA_ARGS__), ##__VA_ARGS__)
+
+#define libeth_pci_get_mmio_addr(mmio_info, offset, ...) \
+ __libeth_pci_get_mmio_addr(mmio_info, offset, \
+ COUNT_ARGS(__VA_ARGS__), ##__VA_ARGS__)
+
+bool __libeth_pci_map_mmio_region(struct libeth_mmio_info *mmio_info,
+ resource_size_t offset,
+ resource_size_t size,
+ int num_args, ...);
+void __iomem *__libeth_pci_get_mmio_addr(struct libeth_mmio_info *mmio_info,
+ resource_size_t region_offset,
+ int num_args, ...);
+void libeth_pci_unmap_all_mmio_regions(struct libeth_mmio_info *mmio_info);
+int libeth_pci_init_dev(struct pci_dev *pdev);
+void libeth_pci_deinit_dev(struct pci_dev *pdev);
+
+#endif /* __LIBETH_PCI_H */
--
2.47.0
^ permalink raw reply related [flat|nested] 36+ messages in thread* Re: [PATCH iwl-next 03/14] libeth: add PCI device initialization helpers to libeth
2025-04-08 12:47 ` [PATCH iwl-next 03/14] libeth: add PCI device initialization helpers to libeth Larysa Zaremba
@ 2025-04-21 14:08 ` Simon Horman
0 siblings, 0 replies; 36+ messages in thread
From: Simon Horman @ 2025-04-21 14:08 UTC (permalink / raw)
To: Larysa Zaremba
Cc: intel-wired-lan, Tony Nguyen, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Jonathan Corbet, Przemek Kitszel,
Jiri Pirko, Mustafa Ismail, Tatyana Nikolova, Andrew Lunn,
Alexander Lobakin, Michael Ellerman, Maciej Fijalkowski,
Lee Trager, Madhavan Srinivasan, Sridhar Samudrala, Jacob Keller,
Michal Swiatkowski, Mateusz Polchlopek, Wenjun Wu, Ahmed Zaki,
netdev, linux-doc, linux-kernel, Karlsson, Magnus, Emil Tantilov,
Madhu Chittim, Josh Hay, Milena Olech, pavan.kumar.linga,
Singhai, Anjali, Phani R Burra
On Tue, Apr 08, 2025 at 02:47:49PM +0200, Larysa Zaremba wrote:
> From: Phani R Burra <phani.r.burra@intel.com>
>
> Add memory related support functions for drivers to access MMIO space and
> allocate/free dma buffers.
>
> Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
> Signed-off-by: Phani R Burra <phani.r.burra@intel.com>
> Co-developed-by: Victor Raj <victor.raj@intel.com>
> Signed-off-by: Victor Raj <victor.raj@intel.com>
> Co-developed-by: Sridhar Samudrala <sridhar.samudrala@intel.com>
> Signed-off-by: Sridhar Samudrala <sridhar.samudrala@intel.com>
> Co-developed-by: Pavan Kumar Linga <pavan.kumar.linga@intel.com>
> Signed-off-by: Pavan Kumar Linga <pavan.kumar.linga@intel.com>
> Co-developed-by: Larysa Zaremba <larysa.zaremba@intel.com>
> Signed-off-by: Larysa Zaremba <larysa.zaremba@intel.com>
...
> diff --git a/drivers/net/ethernet/intel/libeth/pci.c b/drivers/net/ethernet/intel/libeth/pci.c
...
> +/**
> + * __libeth_pci_map_mmio_region - map PCI device MMIO region
> + * @mmio_info: struct to store the mapped MMIO region
> + * @offset: MMIO region start offset
> + * @size: MMIO region size
> + * @num_args: number of additional arguments present
> + *
> + * Return: true on success, false on memory map failure.
> + */
> +bool __libeth_pci_map_mmio_region(struct libeth_mmio_info *mmio_info,
> + resource_size_t offset,
> + resource_size_t size, int num_args, ...)
> +{
> + struct pci_dev *pdev = mmio_info->pdev;
> + struct libeth_pci_mmio_region *mr;
> + resource_size_t pa;
> + void __iomem *va;
> + int bar_idx = 0;
> + va_list args;
> +
> + if (num_args) {
> + va_start(args, num_args);
> + bar_idx = va_arg(args, int);
> + va_end(args);
> + }
> +
> + mr = libeth_find_mmio_region(&mmio_info->mmio_list, offset, bar_idx);
> + if (mr) {
> + pci_warn(pdev, "Mapping of BAR%u with offset %llu already exists\n",
> + bar_idx, offset);
Hi Phani, Larysa, all,
I think that the format specifier here should be %zu rather than %llu.
On ARM W=1 builds gcc 14.2.0 flags this as follows:
CALL scripts/checksyscalls.sh
CC drivers/net/ethernet/intel/libeth/pci.o
In file included from ./include/linux/device.h:15,
from ./include/linux/pci.h:37,
from ./include/net/libeth/pci.h:7,
from drivers/net/ethernet/intel/libeth/pci.c:4:
.../pci.c: In function '__libeth_pci_map_mmio_region':
.../pci.c:92:32: warning: format '%llu' expects argument of type 'long long unsigned int', but argument 4 has type 'resource_size_t' {aka 'unsigned int'} [-Wformat=]
92 | pci_warn(pdev, "Mapping of BAR%u with offset %llu already exists\n",
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.../dev_printk.h:110:30: note: in definition of macro 'dev_printk_index_wrap'
110 | _p_func(dev, fmt, ##__VA_ARGS__); \
| ^~~
.../dev_printk.h:156:61: note: in expansion of macro 'dev_fmt'
156 | dev_printk_index_wrap(_dev_warn, KERN_WARNING, dev, dev_fmt(fmt), ##__VA_ARGS__)
| ^~~~~~~
.../pci.h:2704:41: note: in expansion of macro 'dev_warn'
2704 | #define pci_warn(pdev, fmt, arg...) dev_warn(&(pdev)->dev, fmt, ##arg)
| ^~~~~~~~
.../pci.c:92:17: note: in expansion of macro 'pci_warn'
92 | pci_warn(pdev, "Mapping of BAR%u with offset %llu already exists\n",
| ^~~~~~~~
.../pci.c:92:65: note: format string is defined here
92 | pci_warn(pdev, "Mapping of BAR%u with offset %llu already exists\n",
| ~~~^
| |
| long long unsigned int
| %u
^ permalink raw reply [flat|nested] 36+ messages in thread
* [PATCH iwl-next 04/14] libeth: allow to create fill queues without NAPI
2025-04-08 12:47 [PATCH iwl-next 00/14] Introduce iXD driver Larysa Zaremba
` (2 preceding siblings ...)
2025-04-08 12:47 ` [PATCH iwl-next 03/14] libeth: add PCI device initialization helpers to libeth Larysa Zaremba
@ 2025-04-08 12:47 ` Larysa Zaremba
2025-04-08 12:47 ` [PATCH iwl-next 05/14] libeth: add control queue support Larysa Zaremba
` (10 subsequent siblings)
14 siblings, 0 replies; 36+ messages in thread
From: Larysa Zaremba @ 2025-04-08 12:47 UTC (permalink / raw)
To: intel-wired-lan, Tony Nguyen
Cc: David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Simon Horman, Jonathan Corbet, Przemek Kitszel, Jiri Pirko,
Mustafa Ismail, Tatyana Nikolova, Andrew Lunn, Alexander Lobakin,
Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Larysa Zaremba, Sridhar Samudrala,
Jacob Keller, Michal Swiatkowski, Mateusz Polchlopek, Wenjun Wu,
Ahmed Zaki, netdev, linux-doc, linux-kernel, Karlsson, Magnus,
Emil Tantilov, Madhu Chittim, Josh Hay, Milena Olech,
pavan.kumar.linga, Singhai, Anjali
From: Pavan Kumar Linga <pavan.kumar.linga@intel.com>
Control queues can utilize libeth_rx fill queues, despite working
outside of NAPI context. The only problem is standard fill queues requiring
NAPI that provides them with the device pointer.
Introduce a way to provide the device directly without using NAPI.
Suggested-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Pavan Kumar Linga <pavan.kumar.linga@intel.com>
Signed-off-by: Larysa Zaremba <larysa.zaremba@intel.com>
---
drivers/net/ethernet/intel/libeth/rx.c | 9 +++++----
include/net/libeth/rx.h | 4 +++-
2 files changed, 8 insertions(+), 5 deletions(-)
diff --git a/drivers/net/ethernet/intel/libeth/rx.c b/drivers/net/ethernet/intel/libeth/rx.c
index 66d1d23b8ad2..916e8888959c 100644
--- a/drivers/net/ethernet/intel/libeth/rx.c
+++ b/drivers/net/ethernet/intel/libeth/rx.c
@@ -141,19 +141,20 @@ static bool libeth_rx_page_pool_params_zc(struct libeth_fq *fq,
/**
* libeth_rx_fq_create - create a PP with the default libeth settings
* @fq: buffer queue struct to fill
- * @napi: &napi_struct covering this PP (no usage outside its poll loops)
+ * @napi_dev: &napi_struct for NAPI (data) queues, &device for others
*
* Return: %0 on success, -%errno on failure.
*/
-int libeth_rx_fq_create(struct libeth_fq *fq, struct napi_struct *napi)
+int libeth_rx_fq_create(struct libeth_fq *fq, void *napi_dev)
{
+ struct napi_struct *napi = fq->no_napi ? NULL : napi_dev;
struct page_pool_params pp = {
.flags = PP_FLAG_DMA_MAP | PP_FLAG_DMA_SYNC_DEV,
.order = LIBETH_RX_PAGE_ORDER,
.pool_size = fq->count,
.nid = fq->nid,
- .dev = napi->dev->dev.parent,
- .netdev = napi->dev,
+ .dev = napi ? napi->dev->dev.parent : napi_dev,
+ .netdev = napi ? napi->dev : NULL,
.napi = napi,
.dma_dir = DMA_FROM_DEVICE,
};
diff --git a/include/net/libeth/rx.h b/include/net/libeth/rx.h
index ab05024be518..9b631299fa55 100644
--- a/include/net/libeth/rx.h
+++ b/include/net/libeth/rx.h
@@ -66,6 +66,7 @@ enum libeth_fqe_type {
* @count: number of descriptors/buffers the queue has
* @type: type of the buffers this queue has
* @hsplit: flag whether header split is enabled
+ * @no_napi: the queue is not a data queue and does not have NAPI
* @buf_len: HW-writeable length per each buffer
* @nid: ID of the closest NUMA node with memory
*/
@@ -81,12 +82,13 @@ struct libeth_fq {
/* Cold fields */
enum libeth_fqe_type type:2;
bool hsplit:1;
+ bool no_napi:1;
u32 buf_len;
int nid;
};
-int libeth_rx_fq_create(struct libeth_fq *fq, struct napi_struct *napi);
+int libeth_rx_fq_create(struct libeth_fq *fq, void *napi_dev);
void libeth_rx_fq_destroy(struct libeth_fq *fq);
/**
--
2.47.0
^ permalink raw reply related [flat|nested] 36+ messages in thread* [PATCH iwl-next 05/14] libeth: add control queue support
2025-04-08 12:47 [PATCH iwl-next 00/14] Introduce iXD driver Larysa Zaremba
` (3 preceding siblings ...)
2025-04-08 12:47 ` [PATCH iwl-next 04/14] libeth: allow to create fill queues without NAPI Larysa Zaremba
@ 2025-04-08 12:47 ` Larysa Zaremba
2025-04-10 8:21 ` Leon Romanovsky
2025-04-08 12:47 ` [PATCH iwl-next 06/14] libeth: add bookkeeping support for control queue messages Larysa Zaremba
` (9 subsequent siblings)
14 siblings, 1 reply; 36+ messages in thread
From: Larysa Zaremba @ 2025-04-08 12:47 UTC (permalink / raw)
To: intel-wired-lan, Tony Nguyen
Cc: David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Simon Horman, Jonathan Corbet, Przemek Kitszel, Jiri Pirko,
Mustafa Ismail, Tatyana Nikolova, Andrew Lunn, Alexander Lobakin,
Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Larysa Zaremba, Sridhar Samudrala,
Jacob Keller, Michal Swiatkowski, Mateusz Polchlopek, Wenjun Wu,
Ahmed Zaki, netdev, linux-doc, linux-kernel, Karlsson, Magnus,
Emil Tantilov, Madhu Chittim, Josh Hay, Milena Olech,
pavan.kumar.linga, Singhai, Anjali, Phani R Burra
From: Phani R Burra <phani.r.burra@intel.com>
Libeth will now support control queue setup and configuration APIs.
These are mainly used for mailbox communication between drivers and
control plane.
Make use of the page pool support for managing controlq buffers.
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Phani R Burra <phani.r.burra@intel.com>
Co-developed-by: Victor Raj <victor.raj@intel.com>
Signed-off-by: Victor Raj <victor.raj@intel.com>
Co-developed-by: Sridhar Samudrala <sridhar.samudrala@intel.com>
Signed-off-by: Sridhar Samudrala <sridhar.samudrala@intel.com>
Co-developed-by: Pavan Kumar Linga <pavan.kumar.linga@intel.com>
Signed-off-by: Pavan Kumar Linga <pavan.kumar.linga@intel.com>
Co-developed-by: Larysa Zaremba <larysa.zaremba@intel.com>
Signed-off-by: Larysa Zaremba <larysa.zaremba@intel.com>
---
drivers/net/ethernet/intel/libeth/Kconfig | 8 +
drivers/net/ethernet/intel/libeth/Makefile | 4 +
drivers/net/ethernet/intel/libeth/controlq.c | 608 +++++++++++++++++++
include/net/libeth/controlq.h | 251 ++++++++
4 files changed, 871 insertions(+)
create mode 100644 drivers/net/ethernet/intel/libeth/controlq.c
create mode 100644 include/net/libeth/controlq.h
diff --git a/drivers/net/ethernet/intel/libeth/Kconfig b/drivers/net/ethernet/intel/libeth/Kconfig
index 070ad685fb74..3dad49420663 100644
--- a/drivers/net/ethernet/intel/libeth/Kconfig
+++ b/drivers/net/ethernet/intel/libeth/Kconfig
@@ -8,6 +8,14 @@ config LIBETH
libeth is a common library containing routines shared between several
drivers, but not yet promoted to the generic kernel API.
+config LIBETH_CP
+ tristate
+ select LIBETH
+ select LIBETH_PCI
+ help
+ Common helper routines to communicate with the device Control Plane
+ using virtchnl2 or related mailbox protocols.
+
config LIBETH_PCI
tristate
help
diff --git a/drivers/net/ethernet/intel/libeth/Makefile b/drivers/net/ethernet/intel/libeth/Makefile
index 5f0fc9446acb..861fdabc7be3 100644
--- a/drivers/net/ethernet/intel/libeth/Makefile
+++ b/drivers/net/ethernet/intel/libeth/Makefile
@@ -5,6 +5,10 @@ obj-$(CONFIG_LIBETH) += libeth.o
libeth-y := rx.o
+obj-$(CONFIG_LIBETH_CP) += libeth_cp.o
+
+libeth_cp-y := controlq.o
+
obj-$(CONFIG_LIBETH_PCI) += libeth_pci.o
libeth_pci-y := pci.o
diff --git a/drivers/net/ethernet/intel/libeth/controlq.c b/drivers/net/ethernet/intel/libeth/controlq.c
new file mode 100644
index 000000000000..2ce419591495
--- /dev/null
+++ b/drivers/net/ethernet/intel/libeth/controlq.c
@@ -0,0 +1,608 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/* Copyright (C) 2025 Intel Corporation */
+
+#include <linux/bitfield.h>
+
+#include <net/libeth/controlq.h>
+#include <net/libeth/rx.h>
+
+#define LIBETH_CTLQ_DESC_QWORD0(sz) \
+ (LIBETH_CTLQ_DESC_FLAG_BUF | \
+ LIBETH_CTLQ_DESC_FLAG_RD | \
+ FIELD_PREP(LIBETH_CTLQ_DESC_DATA_LEN, sz))
+
+/**
+ * libeth_ctlq_free_fq - free fill queue resources, including buffers
+ * @ctlq: Rx control queue whose resources need to be freed
+ */
+static void libeth_ctlq_free_fq(struct libeth_ctlq_info *ctlq)
+{
+ struct libeth_fq fq = {
+ .fqes = ctlq->rx_fqes,
+ .pp = ctlq->pp,
+ };
+
+ for (u32 ntc = ctlq->next_to_clean; ntc != ctlq->next_to_post; ) {
+ page_pool_put_full_page(fq.fqes[ntc].page->pp,
+ fq.fqes[ntc].page, false);
+
+ if (++ntc >= ctlq->ring_len)
+ ntc = 0;
+ }
+
+ libeth_rx_fq_destroy(&fq);
+}
+
+/**
+ * libeth_ctlq_init_fq - initialize fill queue for an Rx controlq
+ * @ctlq: control queue that needs a Rx buffer allocation
+ *
+ * Return: %0 on success, -%errno on failure
+ */
+static int libeth_ctlq_init_fq(struct libeth_ctlq_info *ctlq)
+{
+ struct libeth_fq fq = {
+ .count = ctlq->ring_len,
+ .truesize = LIBETH_CTLQ_MAX_BUF_LEN,
+ .nid = NUMA_NO_NODE,
+ .type = LIBETH_FQE_SHORT,
+ .hsplit = true,
+ .no_napi = true,
+ };
+ int err;
+
+ err = libeth_rx_fq_create(&fq, ctlq->dev);
+ if (err)
+ return err;
+
+ ctlq->pp = fq.pp;
+ ctlq->rx_fqes = fq.fqes;
+ ctlq->truesize = fq.truesize;
+
+ return 0;
+}
+
+/**
+ * libeth_ctlq_reset_rx_desc - reset the descriptor with a new address
+ * @desc: descriptor to (re)initialize
+ * @addr: physical address to put into descriptor
+ * @mem_truesize: size of the accessible memory
+ */
+static void libeth_ctlq_reset_rx_desc(struct libeth_ctlq_desc *desc,
+ dma_addr_t addr, u32 mem_truesize)
+{
+ u64 qword;
+
+ *desc = (struct libeth_ctlq_desc) {};
+ qword = LIBETH_CTLQ_DESC_QWORD0(mem_truesize);
+ desc->qword0 = cpu_to_le64(qword);
+
+ qword = FIELD_PREP(LIBETH_CTLQ_DESC_DATA_ADDR_HIGH,
+ upper_32_bits(addr)) |
+ FIELD_PREP(LIBETH_CTLQ_DESC_DATA_ADDR_LOW,
+ lower_32_bits(addr));
+ desc->qword3 = cpu_to_le64(qword);
+}
+
+/**
+ * libeth_ctlq_post_rx_buffs - post buffers to descriptor ring
+ * @ctlq: control queue that requires Rx descriptor ring to be initialized with
+ * new Rx buffers
+ *
+ * The caller must make sure that calls to libeth_ctlq_post_rx_buffs()
+ * and libeth_ctlq_recv() for separate queues are either serialized
+ * or used under ctlq->lock.
+ *
+ * Return: %0 on success, -%ENOMEM if any buffer could not be allocated
+ */
+int libeth_ctlq_post_rx_buffs(struct libeth_ctlq_info *ctlq)
+{
+ u32 ntp = ctlq->next_to_post, ntc = ctlq->next_to_clean, num_to_post;
+ const struct libeth_fq_fp fq = {
+ .pp = ctlq->pp,
+ .fqes = ctlq->rx_fqes,
+ .truesize = ctlq->truesize,
+ .count = ctlq->ring_len,
+ };
+ int ret = 0;
+
+ num_to_post = (ntc > ntp ? 0 : ctlq->ring_len) + ntc - ntp - 1;
+
+ while (num_to_post--) {
+ dma_addr_t addr;
+
+ addr = libeth_rx_alloc(&fq, ntp);
+ if (unlikely(addr == DMA_MAPPING_ERROR)) {
+ ret = -ENOMEM;
+ goto post_bufs;
+ }
+
+ libeth_ctlq_reset_rx_desc(&ctlq->descs[ntp], addr, fq.truesize);
+
+ if (unlikely(++ntp == ctlq->ring_len))
+ ntp = 0;
+ }
+
+post_bufs:
+ if (likely(ctlq->next_to_post != ntp)) {
+ ctlq->next_to_post = ntp;
+
+ writel(ntp, ctlq->reg.tail);
+ }
+
+ return ret;
+}
+EXPORT_SYMBOL_NS_GPL(libeth_ctlq_post_rx_buffs, "LIBETH_CP");
+
+/**
+ * libeth_ctlq_free_tx_msgs - Free Tx control queue messages
+ * @ctlq: Tx control queue being destroyed
+ * @num_msgs: number of messages allocated so far
+ */
+static void libeth_ctlq_free_tx_msgs(struct libeth_ctlq_info *ctlq,
+ u32 num_msgs)
+{
+ for (u32 i = 0; i < num_msgs; i++)
+ kfree(ctlq->tx_msg[i]);
+
+ kvfree(ctlq->tx_msg);
+}
+
+/**
+ * libeth_ctlq_alloc_tx_msgs - Allocate Tx control queue messages
+ * @ctlq: Tx control queue being created
+ *
+ * Return: %0 on success, -%ENOMEM on allocation error
+ */
+static int libeth_ctlq_alloc_tx_msgs(struct libeth_ctlq_info *ctlq)
+{
+ ctlq->tx_msg = kvcalloc(ctlq->ring_len,
+ sizeof(*ctlq->tx_msg), GFP_KERNEL);
+ if (!ctlq->tx_msg)
+ return -ENOMEM;
+
+ for (u32 i = 0; i < ctlq->ring_len; i++) {
+ ctlq->tx_msg[i] = kzalloc(sizeof(*ctlq->tx_msg[i]), GFP_KERNEL);
+
+ if (!ctlq->tx_msg[i]) {
+ libeth_ctlq_free_tx_msgs(ctlq, i);
+ return -ENOMEM;
+ }
+ }
+
+ return 0;
+}
+
+/**
+ * libeth_cp_free_dma_mem - Free the previously allocated DMA memory
+ * @dev: device information
+ * @mem: DMA memory information
+ */
+static void libeth_cp_free_dma_mem(struct device *dev,
+ struct libeth_cp_dma_mem *mem)
+{
+ dma_free_coherent(dev, mem->size, mem->va, mem->pa);
+ mem->va = NULL;
+}
+
+/**
+ * libeth_ctlq_dealloc_ring_res - Free memory allocated for control queue
+ * @ctlq: control queue that requires its ring memory to be freed
+ *
+ * Free the memory used by the ring, buffers and other related structures.
+ */
+static void libeth_ctlq_dealloc_ring_res(struct libeth_ctlq_info *ctlq)
+{
+ struct libeth_cp_dma_mem *dma = &ctlq->ring_mem;
+
+ if (ctlq->type == LIBETH_CTLQ_TYPE_TX)
+ libeth_ctlq_free_tx_msgs(ctlq, ctlq->ring_len);
+ else
+ libeth_ctlq_free_fq(ctlq);
+
+ libeth_cp_free_dma_mem(ctlq->dev, dma);
+}
+
+/**
+ * libeth_cp_alloc_dma_mem - Allocate a DMA memory
+ * @dev: device information
+ * @mem: memory for DMA information to be stored
+ * @size: size of the memory to allocate
+ *
+ * Return: virtual address of DMA memory or NULL.
+ */
+static void *libeth_cp_alloc_dma_mem(struct device *dev,
+ struct libeth_cp_dma_mem *mem, u32 size)
+{
+ size = ALIGN(size, SZ_4K);
+
+ mem->va = dma_alloc_coherent(dev, size, &mem->pa, GFP_KERNEL);
+ mem->size = size;
+
+ return mem->va;
+}
+
+/**
+ * libeth_ctlq_alloc_queue_res - allocate memory for descriptor ring and bufs
+ * @ctlq: control queue that requires its ring resources to be allocated
+ *
+ * Return: %0 on success, -%errno on failure
+ */
+static int libeth_ctlq_alloc_queue_res(struct libeth_ctlq_info *ctlq)
+{
+ size_t size = array_size(ctlq->ring_len, sizeof(*ctlq->descs));
+ struct libeth_cp_dma_mem *dma = &ctlq->ring_mem;
+ int err = -ENOMEM;
+
+ if (!libeth_cp_alloc_dma_mem(ctlq->dev, dma, size))
+ return -ENOMEM;
+
+ ctlq->descs = dma->va;
+
+ if (ctlq->type == LIBETH_CTLQ_TYPE_TX) {
+ if (libeth_ctlq_alloc_tx_msgs(ctlq))
+ goto free_dma_mem;
+ } else {
+ err = libeth_ctlq_init_fq(ctlq);
+ if (err)
+ goto free_dma_mem;
+
+ err = libeth_ctlq_post_rx_buffs(ctlq);
+ if (err) {
+ libeth_ctlq_free_fq(ctlq);
+ goto free_dma_mem;
+ }
+ }
+
+ return 0;
+
+free_dma_mem:
+ libeth_cp_free_dma_mem(ctlq->dev, dma);
+
+ return err;
+}
+
+/**
+ * libeth_ctlq_init_regs - Initialize control queue registers
+ * @ctlq: control queue that needs to be initialized
+ *
+ * Initialize registers. The caller is expected to have already initialized the
+ * descriptor ring memory and buffer memory.
+ */
+static void libeth_ctlq_init_regs(struct libeth_ctlq_info *ctlq)
+{
+ u32 dword;
+
+ if (ctlq->type == VIRTCHNL2_QUEUE_TYPE_RX)
+ writel(ctlq->ring_len - 1, ctlq->reg.tail);
+
+ writel(0, ctlq->reg.head);
+ writel(lower_32_bits(ctlq->ring_mem.pa), ctlq->reg.addr_low);
+ writel(upper_32_bits(ctlq->ring_mem.pa), ctlq->reg.addr_high);
+
+ dword = FIELD_PREP(LIBETH_CTLQ_MBX_ATQ_LEN, ctlq->ring_len) |
+ ctlq->reg.len_ena_mask;
+ writel(dword, ctlq->reg.len);
+}
+
+/**
+ * libeth_find_ctlq - find the controlq for the given id and type
+ * @ctx: controlq context structure
+ * @type: type of controlq to find
+ * @id: controlq id to find
+ *
+ * Return: control queue info pointer on success, NULL on failure
+ */
+struct libeth_ctlq_info *libeth_find_ctlq(struct libeth_ctlq_ctx *ctx,
+ enum virtchnl2_queue_type type,
+ int id)
+{
+ struct libeth_ctlq_info *cq;
+
+ guard(spinlock)(&ctx->ctlqs_lock);
+
+ list_for_each_entry(cq, &ctx->ctlqs, list)
+ if (cq->qid == id && cq->type == type)
+ return cq;
+
+ return NULL;
+}
+EXPORT_SYMBOL_NS_GPL(libeth_find_ctlq, "LIBETH_CP");
+
+/**
+ * libeth_ctlq_add - add one control queue
+ * @ctx: controlq context information
+ * @qinfo: information that requires for queue creation
+ *
+ * Allocate and initialize a control queue and add it to the control queue list.
+ * The ctlq parameter will be allocated/initialized and passed back to the
+ * caller if no errors occur.
+ *
+ * Note: libeth_ctlq_init must be called prior to any calls to libeth_ctlq_add.
+ *
+ * Return: added control queue info pointer on success, error pointer on failure
+ */
+static struct libeth_ctlq_info *
+libeth_ctlq_add(struct libeth_ctlq_ctx *ctx,
+ const struct libeth_ctlq_create_info *qinfo)
+{
+ struct libeth_ctlq_info *ctlq;
+
+ if (qinfo->id != LIBETH_CTLQ_MBX_ID)
+ return ERR_PTR(-EOPNOTSUPP);
+
+ /* libeth_ctlq_init was not called */
+ scoped_guard(spinlock, &ctx->ctlqs_lock)
+ if (!ctx->ctlqs.next)
+ return ERR_PTR(-EINVAL);
+
+ ctlq = kvzalloc(sizeof(*ctlq), GFP_KERNEL);
+ if (!ctlq)
+ return ERR_PTR(-ENOMEM);
+
+ ctlq->type = qinfo->type;
+ ctlq->qid = qinfo->id;
+ ctlq->ring_len = qinfo->len;
+ ctlq->dev = &ctx->mmio_info.pdev->dev;
+ ctlq->reg = qinfo->reg;
+
+ if (libeth_ctlq_alloc_queue_res(ctlq)) {
+ kvfree(ctlq);
+ return ERR_PTR(-ENOMEM);
+ }
+
+ libeth_ctlq_init_regs(ctlq);
+
+ spin_lock_init(&ctlq->lock);
+
+ scoped_guard(spinlock, &ctx->ctlqs_lock)
+ list_add(&ctlq->list, &ctx->ctlqs);
+
+ return ctlq;
+}
+
+/**
+ * libeth_ctlq_remove - deallocate and remove specified control queue
+ * @ctx: libeth context information
+ * @ctlq: specific control queue that needs to be removed
+ */
+static void libeth_ctlq_remove(struct libeth_ctlq_ctx *ctx,
+ struct libeth_ctlq_info *ctlq)
+{
+ scoped_guard(spinlock, &ctx->ctlqs_lock)
+ list_del(&ctlq->list);
+
+ libeth_ctlq_dealloc_ring_res(ctlq);
+ kvfree(ctlq);
+}
+
+/**
+ * libeth_ctlq_init - main initialization routine for all control queues
+ * @ctx: libeth context information
+ * @qinfo: array of structs containing info for each queue to be initialized
+ * @numq: number of queues to initialize
+ *
+ * This initializes queue list and adds any number and any type of control
+ * queues. This is an all or nothing routine; if one fails, all previously
+ * allocated queues will be destroyed. This must be called prior to using
+ * the individual add/remove APIs.
+ *
+ * Return: %0 on success, -%errno on failure
+ */
+int libeth_ctlq_init(struct libeth_ctlq_ctx *ctx,
+ const struct libeth_ctlq_create_info *qinfo,
+ u32 numq)
+{
+ INIT_LIST_HEAD(&ctx->ctlqs);
+ spin_lock_init(&ctx->ctlqs_lock);
+
+ for (u32 i = 0; i < numq; i++) {
+ struct libeth_ctlq_info *ctlq;
+
+ ctlq = libeth_ctlq_add(ctx, &qinfo[i]);
+ if (IS_ERR(ctlq)) {
+ libeth_ctlq_deinit(ctx);
+ return PTR_ERR(ctlq);
+ }
+ }
+
+ return 0;
+}
+EXPORT_SYMBOL_NS_GPL(libeth_ctlq_init, "LIBETH_CP");
+
+/**
+ * libeth_ctlq_deinit - destroy all control queues
+ * @ctx: libeth context information
+ */
+void libeth_ctlq_deinit(struct libeth_ctlq_ctx *ctx)
+{
+ struct libeth_ctlq_info *ctlq, *tmp;
+
+ list_for_each_entry_safe(ctlq, tmp, &ctx->ctlqs, list)
+ libeth_ctlq_remove(ctx, ctlq);
+}
+EXPORT_SYMBOL_NS_GPL(libeth_ctlq_deinit, "LIBETH_CP");
+
+/**
+ * libeth_ctlq_tx_desc_from_msg - initialize a Tx descriptor from a message
+ * @desc: descriptor to be initialized
+ * @msg: filled control queue message
+ */
+static void libeth_ctlq_tx_desc_from_msg(struct libeth_ctlq_desc *desc,
+ const struct libeth_ctlq_msg *msg)
+{
+ const struct libeth_cp_dma_mem *dma = &msg->send_mem;
+ u64 qword;
+
+ qword = FIELD_PREP(LIBETH_CTLQ_DESC_FLAGS, msg->flags) |
+ FIELD_PREP(LIBETH_CTLQ_DESC_INFRA_OPCODE, msg->opcode) |
+ FIELD_PREP(LIBETH_CTLQ_DESC_PFID_VFID, msg->func_id);
+ desc->qword0 = cpu_to_le64(qword);
+
+ qword = FIELD_PREP(LIBETH_CTLQ_DESC_VIRTCHNL_OPCODE,
+ msg->chnl_opcode) |
+ FIELD_PREP(LIBETH_CTLQ_DESC_VIRTCHNL_MSG_RET_VAL,
+ msg->chnl_retval);
+ desc->qword1 = cpu_to_le64(qword);
+
+ qword = FIELD_PREP(LIBETH_CTLQ_DESC_MSG_PARAM0, msg->param0) |
+ FIELD_PREP(LIBETH_CTLQ_DESC_SW_COOKIE,
+ msg->sw_cookie) |
+ FIELD_PREP(LIBETH_CTLQ_DESC_VIRTCHNL_FLAGS,
+ msg->virt_flags);
+ desc->qword2 = cpu_to_le64(qword);
+
+ if (likely(msg->data_len)) {
+ desc->qword0 |=
+ cpu_to_le64(LIBETH_CTLQ_DESC_QWORD0(msg->data_len));
+ qword = FIELD_PREP(LIBETH_CTLQ_DESC_DATA_ADDR_HIGH,
+ upper_32_bits(dma->pa)) |
+ FIELD_PREP(LIBETH_CTLQ_DESC_DATA_ADDR_LOW,
+ lower_32_bits(dma->pa));
+ } else {
+ qword = msg->addr_param;
+ }
+
+ desc->qword3 = cpu_to_le64(qword);
+}
+
+/**
+ * libeth_ctlq_send - send a message to Control Plane or Peer
+ * @ctlq: specific control queue which is used for sending a message
+ * @q_msg: array of queue messages to be sent
+ * @num_q_msg: number of messages to send on control queue
+ *
+ * The control queue will hold a reference to each send message until
+ * the completion for that message has been cleaned.
+ *
+ * The caller must hold ctlq->lock.
+ *
+ * Return: %0 on success, -%errno on failure.
+ */
+int libeth_ctlq_send(struct libeth_ctlq_info *ctlq,
+ struct libeth_ctlq_msg *q_msg, u32 num_q_msg)
+{
+ u32 num_desc_avail, ntu;
+
+ ntu = ctlq->next_to_use;
+
+ num_desc_avail = (ctlq->next_to_clean > ntu ? 0 : ctlq->ring_len) +
+ ctlq->next_to_clean - ntu - 1;
+
+ if (num_desc_avail < num_q_msg)
+ return -EBUSY;
+
+ for (int i = 0; i < num_q_msg; i++) {
+ struct libeth_ctlq_msg *msg = &q_msg[i];
+ struct libeth_ctlq_desc *desc;
+
+ desc = &ctlq->descs[ntu];
+ libeth_ctlq_tx_desc_from_msg(desc, msg);
+
+ if (unlikely(++ntu == ctlq->ring_len))
+ ntu = 0;
+ }
+ writel(ntu, ctlq->reg.tail);
+ ctlq->next_to_use = ntu;
+
+ return 0;
+}
+EXPORT_SYMBOL_NS_GPL(libeth_ctlq_send, "LIBETH_CP");
+
+/**
+ * libeth_ctlq_fill_rx_msg - fill in a message from Rx descriptor and buffer
+ * @msg: message to be filled in
+ * @desc: received descriptor
+ * @rx_buf: fill queue buffer associated with the descriptor
+ */
+static void libeth_ctlq_fill_rx_msg(struct libeth_ctlq_msg *msg,
+ const struct libeth_ctlq_desc *desc,
+ struct libeth_fqe *rx_buf)
+{
+ u64 qword = le64_to_cpu(desc->qword0);
+
+ msg->flags = FIELD_GET(LIBETH_CTLQ_DESC_FLAGS, qword);
+ msg->opcode = FIELD_GET(LIBETH_CTLQ_DESC_INFRA_OPCODE, qword);
+ msg->data_len = FIELD_GET(LIBETH_CTLQ_DESC_DATA_LEN, qword);
+ msg->hw_retval = FIELD_GET(LIBETH_CTLQ_DESC_HW_RETVAL, qword);
+
+ qword = le64_to_cpu(desc->qword1);
+ msg->chnl_opcode =
+ FIELD_GET(LIBETH_CTLQ_DESC_VIRTCHNL_OPCODE, qword);
+ msg->chnl_retval =
+ FIELD_GET(LIBETH_CTLQ_DESC_VIRTCHNL_MSG_RET_VAL, qword);
+
+ qword = le64_to_cpu(desc->qword2);
+ msg->param0 =
+ FIELD_GET(LIBETH_CTLQ_DESC_MSG_PARAM0, qword);
+ msg->sw_cookie =
+ FIELD_GET(LIBETH_CTLQ_DESC_SW_COOKIE, qword);
+ msg->virt_flags =
+ FIELD_GET(LIBETH_CTLQ_DESC_VIRTCHNL_FLAGS, qword);
+
+ if (likely(msg->data_len)) {
+ msg->recv_mem = (struct kvec) {
+ .iov_base = page_address(rx_buf->page),
+ .iov_len = msg->data_len,
+ };
+ libeth_rx_sync_for_cpu(rx_buf, msg->data_len);
+ } else {
+ msg->recv_mem = (struct kvec) {};
+ msg->addr_param = le64_to_cpu(desc->qword3);
+ page_pool_put_full_page(rx_buf->page->pp, rx_buf->page,
+ false);
+ }
+}
+
+/**
+ * libeth_ctlq_recv - receive control queue message call back
+ * @ctlq: control queue that needs to processed for receive
+ * @msg: array of received control queue messages on this q;
+ * needs to be pre-allocated by caller for as many messages as requested
+ * @num_q_msg: number of messages that can be stored in msg buffer
+ *
+ * Called by interrupt handler or polling mechanism. Caller is expected
+ * to free buffers.
+ *
+ * The caller must make sure that calls to libeth_ctlq_post_rx_buffs()
+ * and libeth_ctlq_recv() for separate queues are either serialized
+ * or used under ctlq->lock.
+ *
+ * Return: number of messages received
+ */
+u32 libeth_ctlq_recv(struct libeth_ctlq_info *ctlq, struct libeth_ctlq_msg *msg,
+ u32 num_q_msg)
+{
+ u32 ntc, i;
+
+ ntc = ctlq->next_to_clean;
+
+ for (i = 0; i < num_q_msg; i++) {
+ const struct libeth_ctlq_desc *desc = &ctlq->descs[ntc];
+ struct libeth_fqe *rx_buf = &ctlq->rx_fqes[ntc];
+ u64 qword;
+
+ qword = le64_to_cpu(desc->qword0);
+ if (!FIELD_GET(LIBETH_CTLQ_DESC_FLAG_DD, qword))
+ break;
+
+ dma_rmb();
+
+ if (unlikely(FIELD_GET(LIBETH_CTLQ_DESC_FLAG_ERR, qword)))
+ break;
+
+ libeth_ctlq_fill_rx_msg(&msg[i], desc, rx_buf);
+
+ if (unlikely(++ntc == ctlq->ring_len))
+ ntc = 0;
+ }
+
+ ctlq->next_to_clean = ntc;
+
+ return i;
+}
+EXPORT_SYMBOL_NS_GPL(libeth_ctlq_recv, "LIBETH_CP");
+
+MODULE_DESCRIPTION("Control Plane communication API");
+MODULE_IMPORT_NS("LIBETH");
+MODULE_LICENSE("GPL");
diff --git a/include/net/libeth/controlq.h b/include/net/libeth/controlq.h
new file mode 100644
index 000000000000..35e69d0f48a4
--- /dev/null
+++ b/include/net/libeth/controlq.h
@@ -0,0 +1,251 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/* Copyright (C) 2025 Intel Corporation */
+
+#ifndef __LIBETH_CONTROLQ_H
+#define __LIBETH_CONTROLQ_H
+
+#include <linux/intel/virtchnl2.h>
+
+#include <net/libeth/pci.h>
+#include <net/libeth/rx.h>
+
+/* Default mailbox control queue */
+#define LIBETH_CTLQ_MBX_ID -1
+#define LIBETH_CTLQ_MAX_BUF_LEN SZ_4K
+
+#define LIBETH_CTLQ_TYPE_TX 0
+#define LIBETH_CTLQ_TYPE_RX 1
+
+/* Opcode used to send controlq message to the control plane */
+#define LIBETH_CTLQ_SEND_MSG_TO_CP 0x801
+#define LIBETH_CTLQ_SEND_MSG_TO_PEER 0x804
+
+/**
+ * struct libeth_ctlq_ctx - contains controlq info and MMIO region info
+ * @mmio_info: MMIO region info structure
+ * @ctlqs: list that stores all the control queues
+ * @ctlqs_lock: lock for control queue list
+ */
+struct libeth_ctlq_ctx {
+ struct libeth_mmio_info mmio_info;
+ struct list_head ctlqs;
+ spinlock_t ctlqs_lock; /* protects the ctlqs list */
+};
+
+/**
+ * struct libeth_ctlq_reg - structure representing virtual addresses of the
+ * controlq registers and masks
+ * @head: controlq head register address
+ * @tail: controlq tail register address
+ * @len: register address to write controlq length and enable bit
+ * @addr_high: register address to write the upper 32b of ring physical address
+ * @addr_low: register address to write the lower 32b of ring physical address
+ * @len_mask: mask to read the controlq length
+ * @len_ena_mask: mask to write the controlq enable bit
+ * @head_mask: mask to read the head value
+ */
+struct libeth_ctlq_reg {
+ void __iomem *head;
+ void __iomem *tail;
+ void __iomem *len;
+ void __iomem *addr_high;
+ void __iomem *addr_low;
+ u32 len_mask;
+ u32 len_ena_mask;
+ u32 head_mask;
+};
+
+/**
+ * struct libeth_cp_dma_mem - structure for DMA memory
+ * @va: virtual address
+ * @pa: physical address
+ * @size: memory size
+ */
+struct libeth_cp_dma_mem {
+ void *va;
+ dma_addr_t pa;
+ size_t size;
+};
+
+/**
+ * struct libeth_ctlq_msg - control queue message data
+ * @flags: refer to 'Flags sub-structure' definitions
+ * @opcode: infrastructure message opcode
+ * @data_len: size of the payload
+ * @func_id: queue id for the secondary mailbox queue, 0 for default mailbox
+ * @hw_retval: execution status from the HW
+ * @chnl_opcode: virtchnl message opcode
+ * @chnl_retval: virtchnl return value
+ * @param0: indirect message raw parameter0
+ * @sw_cookie: used to verify the response of the sent virtchnl message
+ * @virt_flags: virtchnl capability flags
+ * @addr_param: additional parameters in place of the address, given no buffer
+ * @recv_mem: virtual address and size of the buffer that contains
+ * the indirect response
+ * @send_mem: physical and virtual address of the DMA buffer,
+ * used for sending
+ */
+struct libeth_ctlq_msg {
+ u16 flags;
+ u16 opcode;
+ u16 data_len;
+ union {
+ u16 func_id;
+ u16 hw_retval;
+ };
+ u32 chnl_opcode;
+ u32 chnl_retval;
+ u32 param0;
+ u16 sw_cookie;
+ u16 virt_flags;
+ u64 addr_param;
+ union {
+ struct kvec recv_mem;
+ struct libeth_cp_dma_mem send_mem;
+ };
+};
+
+/**
+ * struct libeth_ctlq_create_info - control queue create information
+ * @type: control queue type (Rx or Tx)
+ * @id: queue offset passed as input, -1 for default mailbox
+ * @reg: registers accessed by control queue
+ * @len: controlq length
+ */
+struct libeth_ctlq_create_info {
+ enum virtchnl2_queue_type type;
+ int id;
+ struct libeth_ctlq_reg reg;
+ u16 len;
+};
+
+/**
+ * struct libeth_ctlq_info - control queue information
+ * @list: used to add a controlq to the list of queues in libeth_ctlq_ctx
+ * @type: control queue type
+ * @qid: queue identifier
+ * @lock: control queue lock
+ * @ring_mem: descrtiptor ring DMA memory
+ * @descs: array of descrtiptors
+ * @rx_fqes: array of controlq Rx buffers
+ * @tx_msg: Tx messages sent to hardware
+ * @reg: registers used by control queue
+ * @dev: device that owns this control queue
+ * @pp: page pool for controlq Rx buffers
+ * @truesize: size to allocate per buffer
+ * @next_to_use: next available slot to send buffer
+ * @next_to_clean: next descrtiptor to be cleaned
+ * @next_to_post: next available slot to post buffers to after receive
+ * @ring_len: length of the descriptor ring
+ */
+struct libeth_ctlq_info {
+ struct list_head list;
+ enum virtchnl2_queue_type type;
+ int qid;
+ spinlock_t lock; /* for concurrent processing */
+ struct libeth_cp_dma_mem ring_mem;
+ struct libeth_ctlq_desc *descs;
+ union {
+ struct libeth_fqe *rx_fqes;
+ struct libeth_ctlq_msg **tx_msg;
+ };
+ struct libeth_ctlq_reg reg;
+ struct device *dev;
+ struct page_pool *pp;
+ u32 truesize;
+ u32 next_to_clean;
+ union {
+ u32 next_to_use;
+ u32 next_to_post;
+ };
+ u32 ring_len;
+};
+
+#define LIBETH_CTLQ_MBX_ATQ_LEN GENMASK(9, 0)
+
+/* Flags sub-structure
+ * |0 |1 |2 |3 |4 |5 |6 |7 |8 |9 |10 |11 |12 |13 |14 |15 |
+ * |DD |CMP|ERR| * RSV * |FTYPE | *RSV* |RD |VFC|BUF| HOST_ID |
+ */
+ /* libeth controlq descriptor qword0 details */
+#define LIBETH_CTLQ_DESC_FLAG_DD BIT(0)
+#define LIBETH_CTLQ_DESC_FLAG_CMP BIT(1)
+#define LIBETH_CTLQ_DESC_FLAG_ERR BIT(2)
+#define LIBETH_CTLQ_DESC_FLAG_FTYPE_VM BIT(6)
+#define LIBETH_CTLQ_DESC_FLAG_FTYPE_PF BIT(7)
+#define LIBETH_CTLQ_DESC_FLAG_FTYPE GENMASK(7, 6)
+#define LIBETH_CTLQ_DESC_FLAG_RD BIT(10)
+#define LIBETH_CTLQ_DESC_FLAG_VFC BIT(11)
+#define LIBETH_CTLQ_DESC_FLAG_BUF BIT(12)
+#define LIBETH_CTLQ_DESC_FLAG_HOST_ID GENMASK(15, 13)
+
+#define LIBETH_CTLQ_DESC_FLAGS GENMASK(15, 0)
+#define LIBETH_CTLQ_DESC_INFRA_OPCODE GENMASK_ULL(31, 16)
+#define LIBETH_CTLQ_DESC_DATA_LEN GENMASK_ULL(47, 32)
+#define LIBETH_CTLQ_DESC_HW_RETVAL GENMASK_ULL(63, 48)
+
+#define LIBETH_CTLQ_DESC_PFID_VFID GENMASK_ULL(63, 48)
+
+/* libeth controlq descriptor qword1 details */
+#define LIBETH_CTLQ_DESC_VIRTCHNL_OPCODE GENMASK(27, 0)
+#define LIBETH_CTLQ_DESC_VIRTCHNL_DESC_TYPE GENMASK_ULL(31, 28)
+#define LIBETH_CTLQ_DESC_VIRTCHNL_MSG_RET_VAL GENMASK_ULL(63, 32)
+
+/* libeth controlq descriptor qword2 details */
+#define LIBETH_CTLQ_DESC_MSG_PARAM0 GENMASK_ULL(31, 0)
+#define LIBETH_CTLQ_DESC_SW_COOKIE GENMASK_ULL(47, 32)
+#define LIBETH_CTLQ_DESC_VIRTCHNL_FLAGS GENMASK_ULL(63, 48)
+
+/* libeth controlq descriptor qword3 details */
+#define LIBETH_CTLQ_DESC_DATA_ADDR_HIGH GENMASK_ULL(31, 0)
+#define LIBETH_CTLQ_DESC_DATA_ADDR_LOW GENMASK_ULL(63, 32)
+
+/**
+ * struct libeth_ctlq_desc - control queue descriptor format
+ * @qword0: flags, message opcode, data length etc
+ * @qword1: virtchnl opcode, descriptor type and return value
+ * @qword2: indirect message parameters
+ * @qword3: indirect message buffer address
+ */
+struct libeth_ctlq_desc {
+ __le64 qword0;
+ __le64 qword1;
+ __le64 qword2;
+ __le64 qword3;
+};
+
+/**
+ * libeth_ctlq_release_rx_buf - Release Rx buffer for a specific control queue
+ * @ctlq: specific control queue that requires to access the Rx DMA pool
+ * @rx_buf: Rx buffer to be freed
+ *
+ * Driver uses this function to post back the Rx buffer after the usage.
+ */
+static inline void libeth_ctlq_release_rx_buf(struct libeth_ctlq_info *ctlq,
+ struct kvec *rx_buf)
+{
+ struct page *page;
+
+ if (!rx_buf->iov_base)
+ return;
+
+ page = virt_to_page(rx_buf->iov_base);
+ page_pool_put_full_page(page->pp, page, false);
+}
+
+int libeth_ctlq_init(struct libeth_ctlq_ctx *ctx,
+ const struct libeth_ctlq_create_info *qinfo, u32 numq);
+void libeth_ctlq_deinit(struct libeth_ctlq_ctx *ctx);
+
+struct libeth_ctlq_info *libeth_find_ctlq(struct libeth_ctlq_ctx *ctx,
+ enum virtchnl2_queue_type type,
+ int id);
+
+int libeth_ctlq_send(struct libeth_ctlq_info *ctlq,
+ struct libeth_ctlq_msg *q_msg, u32 num_q_msg);
+u32 libeth_ctlq_recv(struct libeth_ctlq_info *ctlq, struct libeth_ctlq_msg *msg,
+ u32 num_q_msg);
+
+int libeth_ctlq_post_rx_buffs(struct libeth_ctlq_info *ctlq);
+
+#endif /* __LIBETH_CONTROLQ_H */
--
2.47.0
^ permalink raw reply related [flat|nested] 36+ messages in thread* Re: [PATCH iwl-next 05/14] libeth: add control queue support
2025-04-08 12:47 ` [PATCH iwl-next 05/14] libeth: add control queue support Larysa Zaremba
@ 2025-04-10 8:21 ` Leon Romanovsky
2025-04-10 10:44 ` Larysa Zaremba
0 siblings, 1 reply; 36+ messages in thread
From: Leon Romanovsky @ 2025-04-10 8:21 UTC (permalink / raw)
To: Larysa Zaremba
Cc: intel-wired-lan, Tony Nguyen, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Simon Horman, Jonathan Corbet,
Przemek Kitszel, Jiri Pirko, Mustafa Ismail, Tatyana Nikolova,
Andrew Lunn, Alexander Lobakin, Michael Ellerman,
Maciej Fijalkowski, Lee Trager, Madhavan Srinivasan,
Sridhar Samudrala, Jacob Keller, Michal Swiatkowski,
Mateusz Polchlopek, Wenjun Wu, Ahmed Zaki, netdev, linux-doc,
linux-kernel, Karlsson, Magnus, Emil Tantilov, Madhu Chittim,
Josh Hay, Milena Olech, pavan.kumar.linga, Singhai, Anjali,
Phani R Burra
On Tue, Apr 08, 2025 at 02:47:51PM +0200, Larysa Zaremba wrote:
> From: Phani R Burra <phani.r.burra@intel.com>
>
> Libeth will now support control queue setup and configuration APIs.
> These are mainly used for mailbox communication between drivers and
> control plane.
>
> Make use of the page pool support for managing controlq buffers.
<...>
> libeth-y := rx.o
>
> +obj-$(CONFIG_LIBETH_CP) += libeth_cp.o
> +
> +libeth_cp-y := controlq.o
So why did you create separate module for it?
Now you have pci -> libeth -> libeth_cp -> ixd, with the potential races between ixd and libeth, am I right?
Thanks
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH iwl-next 05/14] libeth: add control queue support
2025-04-10 8:21 ` Leon Romanovsky
@ 2025-04-10 10:44 ` Larysa Zaremba
2025-04-10 11:23 ` Leon Romanovsky
0 siblings, 1 reply; 36+ messages in thread
From: Larysa Zaremba @ 2025-04-10 10:44 UTC (permalink / raw)
To: Leon Romanovsky
Cc: intel-wired-lan, Tony Nguyen, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Simon Horman, Jonathan Corbet,
Przemek Kitszel, Jiri Pirko, Tatyana Nikolova, Andrew Lunn,
Alexander Lobakin, Michael Ellerman, Maciej Fijalkowski,
Lee Trager, Madhavan Srinivasan, Sridhar Samudrala, Jacob Keller,
Michal Swiatkowski, Mateusz Polchlopek, Ahmed Zaki, netdev,
linux-doc, linux-kernel, Karlsson, Magnus, Emil Tantilov,
Madhu Chittim, Josh Hay, Milena Olech, pavan.kumar.linga,
Singhai, Anjali, Phani R Burra
On Thu, Apr 10, 2025 at 11:21:37AM +0300, Leon Romanovsky wrote:
> On Tue, Apr 08, 2025 at 02:47:51PM +0200, Larysa Zaremba wrote:
> > From: Phani R Burra <phani.r.burra@intel.com>
> >
> > Libeth will now support control queue setup and configuration APIs.
> > These are mainly used for mailbox communication between drivers and
> > control plane.
> >
> > Make use of the page pool support for managing controlq buffers.
>
> <...>
>
> > libeth-y := rx.o
> >
> > +obj-$(CONFIG_LIBETH_CP) += libeth_cp.o
> > +
> > +libeth_cp-y := controlq.o
>
> So why did you create separate module for it?
> Now you have pci -> libeth -> libeth_cp -> ixd, with the potential races between ixd and libeth, am I right?
>
I am not sure what kind of races do you mean, all libeth modules themselves are
stateless and will stay this way [0], all used data is owned by drivers.
As for the module separation, I think there is no harm in keeping it modular.
We intend to use basic libeth (libeth_rx) in drivers that for sure have no use
for libeth_cp. libeth_pci and libeth_cp separation is more arbitral, as we have
no plans for now to use them separately.
Module dependencies are as follows:
libeth_rx and libeth_pci do not depend on other modules.
libeth_cp depends on both libeth_rx and libeth_pci.
idpf directly uses libeth_pci, libeth_rx and libeth_cp.
ixd directly uses libeth_cp and libeth_pci.
[0] https://lore.kernel.org/netdev/61bfa880-6a88-4eac-bab7-040bf72a11ef@intel.com/
> Thanks
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH iwl-next 05/14] libeth: add control queue support
2025-04-10 10:44 ` Larysa Zaremba
@ 2025-04-10 11:23 ` Leon Romanovsky
2025-04-10 12:58 ` Larysa Zaremba
2025-04-10 13:05 ` Alexander Lobakin
0 siblings, 2 replies; 36+ messages in thread
From: Leon Romanovsky @ 2025-04-10 11:23 UTC (permalink / raw)
To: Larysa Zaremba
Cc: intel-wired-lan, Tony Nguyen, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Simon Horman, Jonathan Corbet,
Przemek Kitszel, Jiri Pirko, Tatyana Nikolova, Andrew Lunn,
Alexander Lobakin, Michael Ellerman, Maciej Fijalkowski,
Lee Trager, Madhavan Srinivasan, Sridhar Samudrala, Jacob Keller,
Michal Swiatkowski, Mateusz Polchlopek, Ahmed Zaki, netdev,
linux-doc, linux-kernel, Karlsson, Magnus, Emil Tantilov,
Madhu Chittim, Josh Hay, Milena Olech, pavan.kumar.linga,
Singhai, Anjali, Phani R Burra
On Thu, Apr 10, 2025 at 12:44:33PM +0200, Larysa Zaremba wrote:
> On Thu, Apr 10, 2025 at 11:21:37AM +0300, Leon Romanovsky wrote:
> > On Tue, Apr 08, 2025 at 02:47:51PM +0200, Larysa Zaremba wrote:
> > > From: Phani R Burra <phani.r.burra@intel.com>
> > >
> > > Libeth will now support control queue setup and configuration APIs.
> > > These are mainly used for mailbox communication between drivers and
> > > control plane.
> > >
> > > Make use of the page pool support for managing controlq buffers.
> >
> > <...>
> >
> > > libeth-y := rx.o
> > >
> > > +obj-$(CONFIG_LIBETH_CP) += libeth_cp.o
> > > +
> > > +libeth_cp-y := controlq.o
> >
> > So why did you create separate module for it?
> > Now you have pci -> libeth -> libeth_cp -> ixd, with the potential races between ixd and libeth, am I right?
> >
>
> I am not sure what kind of races do you mean, all libeth modules themselves are
> stateless and will stay this way [0], all used data is owned by drivers.
Somehow such separation doesn't truly work. There are multiple syzkaller
reports per-cycle where module A tries to access module C, which already
doesn't exist because it was proxied through module B.
>
> As for the module separation, I think there is no harm in keeping it modular.
Syzkaller reports disagree with you.
> We intend to use basic libeth (libeth_rx) in drivers that for sure have no use
> for libeth_cp. libeth_pci and libeth_cp separation is more arbitral, as we have
> no plans for now to use them separately.
So let's not over-engineer it.
>
> Module dependencies are as follows:
>
> libeth_rx and libeth_pci do not depend on other modules.
> libeth_cp depends on both libeth_rx and libeth_pci.
> idpf directly uses libeth_pci, libeth_rx and libeth_cp.
> ixd directly uses libeth_cp and libeth_pci.
You can do whatever module architecture for netdev devices, but if you
plan to expose it to RDMA devices, I will vote against any deep layered
module architecture for the drivers.
BTW, please add some Intel prefix to the modules names, they shouldn't
be called in generic names like libeth, e.t.c
Thanks
>
> [0] https://lore.kernel.org/netdev/61bfa880-6a88-4eac-bab7-040bf72a11ef@intel.com/
>
> > Thanks
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH iwl-next 05/14] libeth: add control queue support
2025-04-10 11:23 ` Leon Romanovsky
@ 2025-04-10 12:58 ` Larysa Zaremba
2025-04-10 13:27 ` Leon Romanovsky
2025-04-10 13:05 ` Alexander Lobakin
1 sibling, 1 reply; 36+ messages in thread
From: Larysa Zaremba @ 2025-04-10 12:58 UTC (permalink / raw)
To: Leon Romanovsky
Cc: intel-wired-lan, Tony Nguyen, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Simon Horman, Jonathan Corbet,
Przemek Kitszel, Jiri Pirko, Tatyana Nikolova, Andrew Lunn,
Alexander Lobakin, Michael Ellerman, Maciej Fijalkowski,
Lee Trager, Madhavan Srinivasan, Sridhar Samudrala, Jacob Keller,
Michal Swiatkowski, Mateusz Polchlopek, Ahmed Zaki, netdev,
linux-doc, linux-kernel, Karlsson, Magnus, Emil Tantilov,
Madhu Chittim, Josh Hay, Milena Olech, pavan.kumar.linga,
Singhai, Anjali, Phani R Burra
On Thu, Apr 10, 2025 at 02:23:49PM +0300, Leon Romanovsky wrote:
> On Thu, Apr 10, 2025 at 12:44:33PM +0200, Larysa Zaremba wrote:
> > On Thu, Apr 10, 2025 at 11:21:37AM +0300, Leon Romanovsky wrote:
> > > On Tue, Apr 08, 2025 at 02:47:51PM +0200, Larysa Zaremba wrote:
> > > > From: Phani R Burra <phani.r.burra@intel.com>
> > > >
> > > > Libeth will now support control queue setup and configuration APIs.
> > > > These are mainly used for mailbox communication between drivers and
> > > > control plane.
> > > >
> > > > Make use of the page pool support for managing controlq buffers.
> > >
> > > <...>
> > >
> > > > libeth-y := rx.o
> > > >
> > > > +obj-$(CONFIG_LIBETH_CP) += libeth_cp.o
> > > > +
> > > > +libeth_cp-y := controlq.o
> > >
> > > So why did you create separate module for it?
> > > Now you have pci -> libeth -> libeth_cp -> ixd, with the potential races between ixd and libeth, am I right?
> > >
> >
> > I am not sure what kind of races do you mean, all libeth modules themselves are
> > stateless and will stay this way [0], all used data is owned by drivers.
>
> Somehow such separation doesn't truly work. There are multiple syzkaller
> reports per-cycle where module A tries to access module C, which already
> doesn't exist because it was proxied through module B.
Are there similar reports for libeth and libie modules when iavf is enabled? It
is basically the same hierarchy. (iavf uses both libeth and libie, libie
depends on libeth).
I am just trying to understand, is this a regular situation or did I just mess
smth up?
>
> >
> > As for the module separation, I think there is no harm in keeping it modular.
>
> Syzkaller reports disagree with you.
>
Could you please share them?
> > We intend to use basic libeth (libeth_rx) in drivers that for sure have no use
> > for libeth_cp. libeth_pci and libeth_cp separation is more arbitral, as we have
> > no plans for now to use them separately.
>
> So let's not over-engineer it.
>
> >
> > Module dependencies are as follows:
> >
> > libeth_rx and libeth_pci do not depend on other modules.
> > libeth_cp depends on both libeth_rx and libeth_pci.
> > idpf directly uses libeth_pci, libeth_rx and libeth_cp.
> > ixd directly uses libeth_cp and libeth_pci.
>
I need to amend this: libeth_cp does not depend on libeth_pci in terms of
module namespace, it only uses the header to access struct device that is
stored in libeth_pci-specific mmio_info.
> You can do whatever module architecture for netdev devices, but if you
> plan to expose it to RDMA devices, I will vote against any deep layered
> module architecture for the drivers.
>
> BTW, please add some Intel prefix to the modules names, they shouldn't
> be called in generic names like libeth, e.t.c
>
We did not think this would be a problem, intel has a tradition of calling the
modules pretty ambiguously.
> Thanks
>
> >
> > [0] https://lore.kernel.org/netdev/61bfa880-6a88-4eac-bab7-040bf72a11ef@intel.com/
> >
> > > Thanks
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH iwl-next 05/14] libeth: add control queue support
2025-04-10 12:58 ` Larysa Zaremba
@ 2025-04-10 13:27 ` Leon Romanovsky
2025-04-10 13:33 ` Alexander Lobakin
0 siblings, 1 reply; 36+ messages in thread
From: Leon Romanovsky @ 2025-04-10 13:27 UTC (permalink / raw)
To: Larysa Zaremba
Cc: intel-wired-lan, Tony Nguyen, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Simon Horman, Jonathan Corbet,
Przemek Kitszel, Jiri Pirko, Tatyana Nikolova, Andrew Lunn,
Alexander Lobakin, Michael Ellerman, Maciej Fijalkowski,
Lee Trager, Madhavan Srinivasan, Sridhar Samudrala, Jacob Keller,
Michal Swiatkowski, Mateusz Polchlopek, Ahmed Zaki, netdev,
linux-doc, linux-kernel, Karlsson, Magnus, Emil Tantilov,
Madhu Chittim, Josh Hay, Milena Olech, pavan.kumar.linga,
Singhai, Anjali, Phani R Burra
On Thu, Apr 10, 2025 at 02:58:28PM +0200, Larysa Zaremba wrote:
> On Thu, Apr 10, 2025 at 02:23:49PM +0300, Leon Romanovsky wrote:
> > On Thu, Apr 10, 2025 at 12:44:33PM +0200, Larysa Zaremba wrote:
> > > On Thu, Apr 10, 2025 at 11:21:37AM +0300, Leon Romanovsky wrote:
> > > > On Tue, Apr 08, 2025 at 02:47:51PM +0200, Larysa Zaremba wrote:
> > > > > From: Phani R Burra <phani.r.burra@intel.com>
> > > > >
> > > > > Libeth will now support control queue setup and configuration APIs.
> > > > > These are mainly used for mailbox communication between drivers and
> > > > > control plane.
> > > > >
> > > > > Make use of the page pool support for managing controlq buffers.
> > > >
> > > > <...>
> > > >
> > > > > libeth-y := rx.o
> > > > >
> > > > > +obj-$(CONFIG_LIBETH_CP) += libeth_cp.o
> > > > > +
> > > > > +libeth_cp-y := controlq.o
> > > >
> > > > So why did you create separate module for it?
> > > > Now you have pci -> libeth -> libeth_cp -> ixd, with the potential races between ixd and libeth, am I right?
> > > >
> > >
> > > I am not sure what kind of races do you mean, all libeth modules themselves are
> > > stateless and will stay this way [0], all used data is owned by drivers.
> >
> > Somehow such separation doesn't truly work. There are multiple syzkaller
> > reports per-cycle where module A tries to access module C, which already
> > doesn't exist because it was proxied through module B.
>
> Are there similar reports for libeth and libie modules when iavf is enabled?
To get such report, syzkaller should run on physical iavf, it looks like it doesn't.
Did I miss it here?
https://syzkaller.appspot.com/upstream/s/net
> It is basically the same hierarchy. (iavf uses both libeth and libie, libie
> depends on libeth).
>
> I am just trying to understand, is this a regular situation or did I just mess
> smth up?
My review comment was general one. It is almost impossible to review
this newly proposed architecture split for correctness.
>
> >
> > >
> > > As for the module separation, I think there is no harm in keeping it modular.
> >
> > Syzkaller reports disagree with you.
> >
>
> Could you please share them?
It is not an easy question to answer, because all these reports are complaining
about some wrong locking order or NULL-pointer access. You will never know if
it is because of programming or design error.
As an approximate example, see commits a27c6f46dcec ("RDMA/bnxt_re: Fix an issue in bnxt_re_async_notifier")
and f0df225d12fc ("RDMA/bnxt_re: Add sanity checks on rdev validity").
At the first glance, they look unrelated to our discussion, however
they can serve as an example or races between deinit/disable paths in
parent module vs. child.
>
> > > We intend to use basic libeth (libeth_rx) in drivers that for sure have no use
> > > for libeth_cp. libeth_pci and libeth_cp separation is more arbitral, as we have
> > > no plans for now to use them separately.
> >
> > So let's not over-engineer it.
> >
> > >
> > > Module dependencies are as follows:
> > >
> > > libeth_rx and libeth_pci do not depend on other modules.
> > > libeth_cp depends on both libeth_rx and libeth_pci.
> > > idpf directly uses libeth_pci, libeth_rx and libeth_cp.
> > > ixd directly uses libeth_cp and libeth_pci.
> >
>
> I need to amend this: libeth_cp does not depend on libeth_pci in terms of
> module namespace, it only uses the header to access struct device that is
> stored in libeth_pci-specific mmio_info.
So why did you add SELECT in kconfig?
>
> > You can do whatever module architecture for netdev devices, but if you
> > plan to expose it to RDMA devices, I will vote against any deep layered
> > module architecture for the drivers.
> >
> > BTW, please add some Intel prefix to the modules names, they shouldn't
> > be called in generic names like libeth, e.t.c
> >
>
> We did not think this would be a problem, intel has a tradition of calling the
> modules pretty ambiguously.
I know and it is worth to be changed.
>
> > Thanks
> >
> > >
> > > [0] https://lore.kernel.org/netdev/61bfa880-6a88-4eac-bab7-040bf72a11ef@intel.com/
> > >
> > > > Thanks
>
^ permalink raw reply [flat|nested] 36+ messages in thread* Re: [PATCH iwl-next 05/14] libeth: add control queue support
2025-04-10 13:27 ` Leon Romanovsky
@ 2025-04-10 13:33 ` Alexander Lobakin
2025-04-10 13:58 ` Leon Romanovsky
0 siblings, 1 reply; 36+ messages in thread
From: Alexander Lobakin @ 2025-04-10 13:33 UTC (permalink / raw)
To: Leon Romanovsky, Larysa Zaremba
Cc: intel-wired-lan, Tony Nguyen, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Simon Horman, Jonathan Corbet,
Przemek Kitszel, Jiri Pirko, Tatyana Nikolova, Andrew Lunn,
Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Sridhar Samudrala, Jacob Keller,
Michal Swiatkowski, Mateusz Polchlopek, Ahmed Zaki, netdev,
linux-doc, linux-kernel, Karlsson, Magnus, Emil Tantilov,
Madhu Chittim, Josh Hay, Milena Olech, pavan.kumar.linga,
Singhai, Anjali, Phani R Burra
From: Leon Romanovsky <leon@kernel.org>
Date: Thu, 10 Apr 2025 16:27:06 +0300
> On Thu, Apr 10, 2025 at 02:58:28PM +0200, Larysa Zaremba wrote:
>> On Thu, Apr 10, 2025 at 02:23:49PM +0300, Leon Romanovsky wrote:
>>> On Thu, Apr 10, 2025 at 12:44:33PM +0200, Larysa Zaremba wrote:
>>>> On Thu, Apr 10, 2025 at 11:21:37AM +0300, Leon Romanovsky wrote:
>>>>> On Tue, Apr 08, 2025 at 02:47:51PM +0200, Larysa Zaremba wrote:
>>>>>> From: Phani R Burra <phani.r.burra@intel.com>
>>>>>>
>>>>>> Libeth will now support control queue setup and configuration APIs.
>>>>>> These are mainly used for mailbox communication between drivers and
>>>>>> control plane.
>>>>>>
>>>>>> Make use of the page pool support for managing controlq buffers.
>>>>>
>>>>> <...>
>>>>>
>>>>>> libeth-y := rx.o
>>>>>>
>>>>>> +obj-$(CONFIG_LIBETH_CP) += libeth_cp.o
>>>>>> +
>>>>>> +libeth_cp-y := controlq.o
>>>>>
>>>>> So why did you create separate module for it?
>>>>> Now you have pci -> libeth -> libeth_cp -> ixd, with the potential races between ixd and libeth, am I right?
>>>>>
>>>>
>>>> I am not sure what kind of races do you mean, all libeth modules themselves are
>>>> stateless and will stay this way [0], all used data is owned by drivers.
>>>
>>> Somehow such separation doesn't truly work. There are multiple syzkaller
>>> reports per-cycle where module A tries to access module C, which already
>>> doesn't exist because it was proxied through module B.
>>
>> Are there similar reports for libeth and libie modules when iavf is enabled?
>
> To get such report, syzkaller should run on physical iavf, it looks like it doesn't.
> Did I miss it here?
> https://syzkaller.appspot.com/upstream/s/net
>
>> It is basically the same hierarchy. (iavf uses both libeth and libie, libie
>> depends on libeth).
>>
>> I am just trying to understand, is this a regular situation or did I just mess
>> smth up?
>
> My review comment was general one. It is almost impossible to review
> this newly proposed architecture split for correctness.
>
>>
>>>
>>>>
>>>> As for the module separation, I think there is no harm in keeping it modular.
>>>
>>> Syzkaller reports disagree with you.
>>>
>>
>> Could you please share them?
>
> It is not an easy question to answer, because all these reports are complaining
> about some wrong locking order or NULL-pointer access. You will never know if
> it is because of programming or design error.
>
> As an approximate example, see commits a27c6f46dcec ("RDMA/bnxt_re: Fix an issue in bnxt_re_async_notifier")
> and f0df225d12fc ("RDMA/bnxt_re: Add sanity checks on rdev validity").
> At the first glance, they look unrelated to our discussion, however
> they can serve as an example or races between deinit/disable paths in
> parent module vs. child.
Unrelated. At first, you were talking about module dependencies, now
you're talking about struct device etc dependencies, which is a
completely different thing.
As already said, libeth is stateless, so the latter one can't happen.
The former one is impossible at all. As long as at least 1 child module
is loaded, you can't unload the parent. And load/unload is serialized,
see module core code.
[...]
>> We did not think this would be a problem, intel has a tradition of calling the
>> modules pretty ambiguously.
>
> I know and it is worth to be changed.
Out of scope of this series.
Thanks,
Olek
^ permalink raw reply [flat|nested] 36+ messages in thread* Re: [PATCH iwl-next 05/14] libeth: add control queue support
2025-04-10 13:33 ` Alexander Lobakin
@ 2025-04-10 13:58 ` Leon Romanovsky
2025-04-10 14:04 ` Alexander Lobakin
0 siblings, 1 reply; 36+ messages in thread
From: Leon Romanovsky @ 2025-04-10 13:58 UTC (permalink / raw)
To: Alexander Lobakin
Cc: Larysa Zaremba, intel-wired-lan, Tony Nguyen, David S. Miller,
Eric Dumazet, Jakub Kicinski, Paolo Abeni, Simon Horman,
Jonathan Corbet, Przemek Kitszel, Jiri Pirko, Tatyana Nikolova,
Andrew Lunn, Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Sridhar Samudrala, Jacob Keller,
Michal Swiatkowski, Mateusz Polchlopek, Ahmed Zaki, netdev,
linux-doc, linux-kernel, Karlsson, Magnus, Emil Tantilov,
Madhu Chittim, Josh Hay, Milena Olech, pavan.kumar.linga,
Singhai, Anjali, Phani R Burra
On Thu, Apr 10, 2025 at 03:33:40PM +0200, Alexander Lobakin wrote:
> From: Leon Romanovsky <leon@kernel.org>
> Date: Thu, 10 Apr 2025 16:27:06 +0300
>
> > On Thu, Apr 10, 2025 at 02:58:28PM +0200, Larysa Zaremba wrote:
> >> On Thu, Apr 10, 2025 at 02:23:49PM +0300, Leon Romanovsky wrote:
> >>> On Thu, Apr 10, 2025 at 12:44:33PM +0200, Larysa Zaremba wrote:
> >>>> On Thu, Apr 10, 2025 at 11:21:37AM +0300, Leon Romanovsky wrote:
> >>>>> On Tue, Apr 08, 2025 at 02:47:51PM +0200, Larysa Zaremba wrote:
> >>>>>> From: Phani R Burra <phani.r.burra@intel.com>
> >>>>>>
> >>>>>> Libeth will now support control queue setup and configuration APIs.
> >>>>>> These are mainly used for mailbox communication between drivers and
> >>>>>> control plane.
> >>>>>>
> >>>>>> Make use of the page pool support for managing controlq buffers.
> >>>>>
> >>>>> <...>
> >>>>>
> >>>>>> libeth-y := rx.o
> >>>>>>
> >>>>>> +obj-$(CONFIG_LIBETH_CP) += libeth_cp.o
> >>>>>> +
> >>>>>> +libeth_cp-y := controlq.o
> >>>>>
> >>>>> So why did you create separate module for it?
> >>>>> Now you have pci -> libeth -> libeth_cp -> ixd, with the potential races between ixd and libeth, am I right?
> >>>>>
> >>>>
> >>>> I am not sure what kind of races do you mean, all libeth modules themselves are
> >>>> stateless and will stay this way [0], all used data is owned by drivers.
> >>>
> >>> Somehow such separation doesn't truly work. There are multiple syzkaller
> >>> reports per-cycle where module A tries to access module C, which already
> >>> doesn't exist because it was proxied through module B.
> >>
> >> Are there similar reports for libeth and libie modules when iavf is enabled?
> >
> > To get such report, syzkaller should run on physical iavf, it looks like it doesn't.
> > Did I miss it here?
> > https://syzkaller.appspot.com/upstream/s/net
> >
> >> It is basically the same hierarchy. (iavf uses both libeth and libie, libie
> >> depends on libeth).
> >>
> >> I am just trying to understand, is this a regular situation or did I just mess
> >> smth up?
> >
> > My review comment was general one. It is almost impossible to review
> > this newly proposed architecture split for correctness.
> >
> >>
> >>>
> >>>>
> >>>> As for the module separation, I think there is no harm in keeping it modular.
> >>>
> >>> Syzkaller reports disagree with you.
> >>>
> >>
> >> Could you please share them?
> >
> > It is not an easy question to answer, because all these reports are complaining
> > about some wrong locking order or NULL-pointer access. You will never know if
> > it is because of programming or design error.
> >
> > As an approximate example, see commits a27c6f46dcec ("RDMA/bnxt_re: Fix an issue in bnxt_re_async_notifier")
> > and f0df225d12fc ("RDMA/bnxt_re: Add sanity checks on rdev validity").
> > At the first glance, they look unrelated to our discussion, however
> > they can serve as an example or races between deinit/disable paths in
> > parent module vs. child.
>
> Unrelated. At first, you were talking about module dependencies, now
> you're talking about struct device etc dependencies, which is a
> completely different thing.
>
> As already said, libeth is stateless, so the latter one can't happen.
> The former one is impossible at all. As long as at least 1 child module
> is loaded, you can't unload the parent. And load/unload is serialized,
> see module core code.
It is not only module load/unload. It is bind/unbind, devlink operations
e.t.c, everything that can cause to reload driver in module C.
Thanks
^ permalink raw reply [flat|nested] 36+ messages in thread* Re: [PATCH iwl-next 05/14] libeth: add control queue support
2025-04-10 13:58 ` Leon Romanovsky
@ 2025-04-10 14:04 ` Alexander Lobakin
0 siblings, 0 replies; 36+ messages in thread
From: Alexander Lobakin @ 2025-04-10 14:04 UTC (permalink / raw)
To: Leon Romanovsky
Cc: Larysa Zaremba, intel-wired-lan, Tony Nguyen, David S. Miller,
Eric Dumazet, Jakub Kicinski, Paolo Abeni, Simon Horman,
Jonathan Corbet, Przemek Kitszel, Jiri Pirko, Tatyana Nikolova,
Andrew Lunn, Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Sridhar Samudrala, Jacob Keller,
Michal Swiatkowski, Mateusz Polchlopek, Ahmed Zaki, netdev,
linux-doc, linux-kernel, Karlsson, Magnus, Emil Tantilov,
Madhu Chittim, Josh Hay, Milena Olech, pavan.kumar.linga,
Singhai, Anjali, Phani R Burra
From: Leon Romanovsky <leon@kernel.org>
Date: Thu, 10 Apr 2025 16:58:43 +0300
> On Thu, Apr 10, 2025 at 03:33:40PM +0200, Alexander Lobakin wrote:
>> From: Leon Romanovsky <leon@kernel.org>
>> Date: Thu, 10 Apr 2025 16:27:06 +0300
>>
>>> On Thu, Apr 10, 2025 at 02:58:28PM +0200, Larysa Zaremba wrote:
>>>> On Thu, Apr 10, 2025 at 02:23:49PM +0300, Leon Romanovsky wrote:
>>>>> On Thu, Apr 10, 2025 at 12:44:33PM +0200, Larysa Zaremba wrote:
>>>>>> On Thu, Apr 10, 2025 at 11:21:37AM +0300, Leon Romanovsky wrote:
>>>>>>> On Tue, Apr 08, 2025 at 02:47:51PM +0200, Larysa Zaremba wrote:
>>>>>>>> From: Phani R Burra <phani.r.burra@intel.com>
>>>>>>>>
>>>>>>>> Libeth will now support control queue setup and configuration APIs.
>>>>>>>> These are mainly used for mailbox communication between drivers and
>>>>>>>> control plane.
>>>>>>>>
>>>>>>>> Make use of the page pool support for managing controlq buffers.
>>>>>>>
>>>>>>> <...>
>>>>>>>
>>>>>>>> libeth-y := rx.o
>>>>>>>>
>>>>>>>> +obj-$(CONFIG_LIBETH_CP) += libeth_cp.o
>>>>>>>> +
>>>>>>>> +libeth_cp-y := controlq.o
>>>>>>>
>>>>>>> So why did you create separate module for it?
>>>>>>> Now you have pci -> libeth -> libeth_cp -> ixd, with the potential races between ixd and libeth, am I right?
>>>>>>>
>>>>>>
>>>>>> I am not sure what kind of races do you mean, all libeth modules themselves are
>>>>>> stateless and will stay this way [0], all used data is owned by drivers.
>>>>>
>>>>> Somehow such separation doesn't truly work. There are multiple syzkaller
>>>>> reports per-cycle where module A tries to access module C, which already
>>>>> doesn't exist because it was proxied through module B.
>>>>
>>>> Are there similar reports for libeth and libie modules when iavf is enabled?
>>>
>>> To get such report, syzkaller should run on physical iavf, it looks like it doesn't.
>>> Did I miss it here?
>>> https://syzkaller.appspot.com/upstream/s/net
>>>
>>>> It is basically the same hierarchy. (iavf uses both libeth and libie, libie
>>>> depends on libeth).
>>>>
>>>> I am just trying to understand, is this a regular situation or did I just mess
>>>> smth up?
>>>
>>> My review comment was general one. It is almost impossible to review
>>> this newly proposed architecture split for correctness.
>>>
>>>>
>>>>>
>>>>>>
>>>>>> As for the module separation, I think there is no harm in keeping it modular.
>>>>>
>>>>> Syzkaller reports disagree with you.
>>>>>
>>>>
>>>> Could you please share them?
>>>
>>> It is not an easy question to answer, because all these reports are complaining
>>> about some wrong locking order or NULL-pointer access. You will never know if
>>> it is because of programming or design error.
>>>
>>> As an approximate example, see commits a27c6f46dcec ("RDMA/bnxt_re: Fix an issue in bnxt_re_async_notifier")
>>> and f0df225d12fc ("RDMA/bnxt_re: Add sanity checks on rdev validity").
>>> At the first glance, they look unrelated to our discussion, however
>>> they can serve as an example or races between deinit/disable paths in
>>> parent module vs. child.
>>
>> Unrelated. At first, you were talking about module dependencies, now
>> you're talking about struct device etc dependencies, which is a
>> completely different thing.
>>
>> As already said, libeth is stateless, so the latter one can't happen.
>> The former one is impossible at all. As long as at least 1 child module
>> is loaded, you can't unload the parent. And load/unload is serialized,
>> see module core code.
>
> It is not only module load/unload. It is bind/unbind, devlink operations
> e.t.c, everything that can cause to reload driver in module C.
You still mix module deps vs struct device/driver deps.
Atomic modules are not anyhow tied to the latter, esp. given that the
whole libeth is stateless and is just a bunch of helpers.
Thanks,
Olek
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH iwl-next 05/14] libeth: add control queue support
2025-04-10 11:23 ` Leon Romanovsky
2025-04-10 12:58 ` Larysa Zaremba
@ 2025-04-10 13:05 ` Alexander Lobakin
2025-04-10 13:44 ` Leon Romanovsky
1 sibling, 1 reply; 36+ messages in thread
From: Alexander Lobakin @ 2025-04-10 13:05 UTC (permalink / raw)
To: Leon Romanovsky, Larysa Zaremba
Cc: intel-wired-lan, Tony Nguyen, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Simon Horman, Jonathan Corbet,
Przemek Kitszel, Jiri Pirko, Tatyana Nikolova, Andrew Lunn,
Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Sridhar Samudrala, Jacob Keller,
Michal Swiatkowski, Mateusz Polchlopek, Ahmed Zaki, netdev,
linux-doc, linux-kernel, Karlsson, Magnus, Emil Tantilov,
Madhu Chittim, Josh Hay, Milena Olech, pavan.kumar.linga,
Singhai, Anjali, Phani R Burra
From: Leon Romanovsky <leon@kernel.org>
Date: Thu, 10 Apr 2025 14:23:49 +0300
> On Thu, Apr 10, 2025 at 12:44:33PM +0200, Larysa Zaremba wrote:
>> On Thu, Apr 10, 2025 at 11:21:37AM +0300, Leon Romanovsky wrote:
>>> On Tue, Apr 08, 2025 at 02:47:51PM +0200, Larysa Zaremba wrote:
>>>> From: Phani R Burra <phani.r.burra@intel.com>
>>>>
>>>> Libeth will now support control queue setup and configuration APIs.
>>>> These are mainly used for mailbox communication between drivers and
>>>> control plane.
>>>>
>>>> Make use of the page pool support for managing controlq buffers.
>>>
>>> <...>
>>>
>>>> libeth-y := rx.o
>>>>
>>>> +obj-$(CONFIG_LIBETH_CP) += libeth_cp.o
>>>> +
>>>> +libeth_cp-y := controlq.o
>>>
>>> So why did you create separate module for it?
>>> Now you have pci -> libeth -> libeth_cp -> ixd, with the potential races between ixd and libeth, am I right?
>>>
>>
>> I am not sure what kind of races do you mean, all libeth modules themselves are
>> stateless and will stay this way [0], all used data is owned by drivers.
>
> Somehow such separation doesn't truly work. There are multiple syzkaller
> reports per-cycle where module A tries to access module C, which already
> doesn't exist because it was proxied through module B.
>
>>
>> As for the module separation, I think there is no harm in keeping it modular.
>
> Syzkaller reports disagree with you.
Zero proofs.
Kernel loads modules under mutex, patches relocs etc. Circular deps are
not supported. I've no idea how that possible what you describe.
>
>> We intend to use basic libeth (libeth_rx) in drivers that for sure have no use
>> for libeth_cp. libeth_pci and libeth_cp separation is more arbitral, as we have
>> no plans for now to use them separately.
>
> So let's not over-engineer it.
Let's not over-engineer Netfilter then and compile it into one module?
I don't want to load _pci/_cp when I load something different than idpf
or ixd, it makes no sense.
Moreover, libeth and libeth_xdp are still generic and can be reused by
any vendor. _pci/_cp are not and won't ever be, they are in the same
folder as generic libeth only to not create more folders. I don't want
to link everything together.
>
>>
>> Module dependencies are as follows:
>>
>> libeth_rx and libeth_pci do not depend on other modules.
>> libeth_cp depends on both libeth_rx and libeth_pci.
>> idpf directly uses libeth_pci, libeth_rx and libeth_cp.
>> ixd directly uses libeth_cp and libeth_pci.
>
> You can do whatever module architecture for netdev devices, but if you
> plan to expose it to RDMA devices, I will vote against any deep layered
> module architecture for the drivers.
No plans for RDMA there.
Maybe link the whole kernel to one vmlinux then?
>
> BTW, please add some Intel prefix to the modules names, they shouldn't
> be called in generic names like libeth, e.t.c
Two modules with the same name can't exist within the kernel. libeth was
available and I haven't seen anyone wanting to take it. It's not common
at all to name a module starting with "lib".
I don't want to send huge patches which would only change the prefixes.
Nobody disagreed I can take this name back when libeth was taken.
+ above, anyone can reuse the lib, it's pretty generic, except for ixd
stuff which is isolated. What if someone takes it tomorrow, send one
more big patch which will revert the suffixes?
+ no offense, but I noted several times during reviews of mlx5 stuff
that lots of functions within the module are not prefixed anyhow and may
conflict anytime with some generic symbol.
Thanks,
Olek
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH iwl-next 05/14] libeth: add control queue support
2025-04-10 13:05 ` Alexander Lobakin
@ 2025-04-10 13:44 ` Leon Romanovsky
2025-04-10 13:59 ` Larysa Zaremba
2025-04-10 14:00 ` [Intel-wired-lan] " Alexander Lobakin
0 siblings, 2 replies; 36+ messages in thread
From: Leon Romanovsky @ 2025-04-10 13:44 UTC (permalink / raw)
To: Alexander Lobakin
Cc: Larysa Zaremba, intel-wired-lan, Tony Nguyen, David S. Miller,
Eric Dumazet, Jakub Kicinski, Paolo Abeni, Simon Horman,
Jonathan Corbet, Przemek Kitszel, Jiri Pirko, Tatyana Nikolova,
Andrew Lunn, Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Sridhar Samudrala, Jacob Keller,
Michal Swiatkowski, Mateusz Polchlopek, Ahmed Zaki, netdev,
linux-doc, linux-kernel, Karlsson, Magnus, Emil Tantilov,
Madhu Chittim, Josh Hay, Milena Olech, pavan.kumar.linga,
Singhai, Anjali, Phani R Burra
On Thu, Apr 10, 2025 at 03:05:19PM +0200, Alexander Lobakin wrote:
> From: Leon Romanovsky <leon@kernel.org>
> Date: Thu, 10 Apr 2025 14:23:49 +0300
>
> > On Thu, Apr 10, 2025 at 12:44:33PM +0200, Larysa Zaremba wrote:
> >> On Thu, Apr 10, 2025 at 11:21:37AM +0300, Leon Romanovsky wrote:
> >>> On Tue, Apr 08, 2025 at 02:47:51PM +0200, Larysa Zaremba wrote:
> >>>> From: Phani R Burra <phani.r.burra@intel.com>
> >>>>
> >>>> Libeth will now support control queue setup and configuration APIs.
> >>>> These are mainly used for mailbox communication between drivers and
> >>>> control plane.
> >>>>
> >>>> Make use of the page pool support for managing controlq buffers.
<...>
> >> Module dependencies are as follows:
> >>
> >> libeth_rx and libeth_pci do not depend on other modules.
> >> libeth_cp depends on both libeth_rx and libeth_pci.
> >> idpf directly uses libeth_pci, libeth_rx and libeth_cp.
> >> ixd directly uses libeth_cp and libeth_pci.
> >
> > You can do whatever module architecture for netdev devices, but if you
> > plan to expose it to RDMA devices, I will vote against any deep layered
> > module architecture for the drivers.
>
> No plans for RDMA there.
>
> Maybe link the whole kernel to one vmlinux then?
It seems that you didn't understand at all about what we are talking
here. Please use the opportunity that you are working for the same
company with Larysa and ask her offline. She understood perfectly about
which modules we are talking.
>
> >
> > BTW, please add some Intel prefix to the modules names, they shouldn't
> > be called in generic names like libeth, e.t.c
>
> Two modules with the same name can't exist within the kernel. libeth was
> available and I haven't seen anyone wanting to take it. It's not common
> at all to name a module starting with "lib".
Again, please talk with Larysa. ETH part is problematic in libeth name
and not LIB.
Thanks
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH iwl-next 05/14] libeth: add control queue support
2025-04-10 13:44 ` Leon Romanovsky
@ 2025-04-10 13:59 ` Larysa Zaremba
2025-04-11 17:18 ` Leon Romanovsky
2025-04-10 14:00 ` [Intel-wired-lan] " Alexander Lobakin
1 sibling, 1 reply; 36+ messages in thread
From: Larysa Zaremba @ 2025-04-10 13:59 UTC (permalink / raw)
To: Leon Romanovsky
Cc: Alexander Lobakin, intel-wired-lan, Tony Nguyen, David S. Miller,
Eric Dumazet, Jakub Kicinski, Paolo Abeni, Simon Horman,
Jonathan Corbet, Przemek Kitszel, Jiri Pirko, Tatyana Nikolova,
Andrew Lunn, Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Sridhar Samudrala, Jacob Keller,
Michal Swiatkowski, Mateusz Polchlopek, Ahmed Zaki, netdev,
linux-doc, linux-kernel, Karlsson, Magnus, Emil Tantilov,
Madhu Chittim, Josh Hay, Milena Olech, pavan.kumar.linga,
Singhai, Anjali, Phani R Burra
On Thu, Apr 10, 2025 at 04:44:43PM +0300, Leon Romanovsky wrote:
> On Thu, Apr 10, 2025 at 03:05:19PM +0200, Alexander Lobakin wrote:
> > From: Leon Romanovsky <leon@kernel.org>
> > Date: Thu, 10 Apr 2025 14:23:49 +0300
> >
> > > On Thu, Apr 10, 2025 at 12:44:33PM +0200, Larysa Zaremba wrote:
> > >> On Thu, Apr 10, 2025 at 11:21:37AM +0300, Leon Romanovsky wrote:
> > >>> On Tue, Apr 08, 2025 at 02:47:51PM +0200, Larysa Zaremba wrote:
> > >>>> From: Phani R Burra <phani.r.burra@intel.com>
> > >>>>
> > >>>> Libeth will now support control queue setup and configuration APIs.
> > >>>> These are mainly used for mailbox communication between drivers and
> > >>>> control plane.
> > >>>>
> > >>>> Make use of the page pool support for managing controlq buffers.
>
> <...>
>
> > >> Module dependencies are as follows:
> > >>
> > >> libeth_rx and libeth_pci do not depend on other modules.
> > >> libeth_cp depends on both libeth_rx and libeth_pci.
> > >> idpf directly uses libeth_pci, libeth_rx and libeth_cp.
> > >> ixd directly uses libeth_cp and libeth_pci.
> > >
> > > You can do whatever module architecture for netdev devices, but if you
> > > plan to expose it to RDMA devices, I will vote against any deep layered
> > > module architecture for the drivers.
> >
> > No plans for RDMA there.
> >
> > Maybe link the whole kernel to one vmlinux then?
>
> It seems that you didn't understand at all about what we are talking
> here. Please use the opportunity that you are working for the same
> company with Larysa and ask her offline. She understood perfectly about
> which modules we are talking.
>
While I do understand what kind of module relationship you consider problematic,
I still struggle to understand why stateless lib hierarchy can be problematic.
The fixes that you linked relate more to problematic resource sharing, of which
libeth has none, it does not have its own memory or its own threads, this is
just collection of data structures and functions.
> >
> > >
> > > BTW, please add some Intel prefix to the modules names, they shouldn't
> > > be called in generic names like libeth, e.t.c
> >
> > Two modules with the same name can't exist within the kernel. libeth was
> > available and I haven't seen anyone wanting to take it. It's not common
> > at all to name a module starting with "lib".
>
> Again, please talk with Larysa. ETH part is problematic in libeth name
> and not LIB.
>
> Thanks
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH iwl-next 05/14] libeth: add control queue support
2025-04-10 13:59 ` Larysa Zaremba
@ 2025-04-11 17:18 ` Leon Romanovsky
0 siblings, 0 replies; 36+ messages in thread
From: Leon Romanovsky @ 2025-04-11 17:18 UTC (permalink / raw)
To: Larysa Zaremba
Cc: Alexander Lobakin, intel-wired-lan, Tony Nguyen, David S. Miller,
Eric Dumazet, Jakub Kicinski, Paolo Abeni, Simon Horman,
Jonathan Corbet, Przemek Kitszel, Jiri Pirko, Tatyana Nikolova,
Andrew Lunn, Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Sridhar Samudrala, Jacob Keller,
Michal Swiatkowski, Mateusz Polchlopek, Ahmed Zaki, netdev,
linux-doc, linux-kernel, Karlsson, Magnus, Emil Tantilov,
Madhu Chittim, Josh Hay, Milena Olech, pavan.kumar.linga,
Singhai, Anjali, Phani R Burra
On Thu, Apr 10, 2025 at 03:59:23PM +0200, Larysa Zaremba wrote:
> On Thu, Apr 10, 2025 at 04:44:43PM +0300, Leon Romanovsky wrote:
> > On Thu, Apr 10, 2025 at 03:05:19PM +0200, Alexander Lobakin wrote:
> > > From: Leon Romanovsky <leon@kernel.org>
> > > Date: Thu, 10 Apr 2025 14:23:49 +0300
> > >
> > > > On Thu, Apr 10, 2025 at 12:44:33PM +0200, Larysa Zaremba wrote:
> > > >> On Thu, Apr 10, 2025 at 11:21:37AM +0300, Leon Romanovsky wrote:
> > > >>> On Tue, Apr 08, 2025 at 02:47:51PM +0200, Larysa Zaremba wrote:
> > > >>>> From: Phani R Burra <phani.r.burra@intel.com>
> > > >>>>
> > > >>>> Libeth will now support control queue setup and configuration APIs.
> > > >>>> These are mainly used for mailbox communication between drivers and
> > > >>>> control plane.
> > > >>>>
> > > >>>> Make use of the page pool support for managing controlq buffers.
> >
> > <...>
> >
> > > >> Module dependencies are as follows:
> > > >>
> > > >> libeth_rx and libeth_pci do not depend on other modules.
> > > >> libeth_cp depends on both libeth_rx and libeth_pci.
> > > >> idpf directly uses libeth_pci, libeth_rx and libeth_cp.
> > > >> ixd directly uses libeth_cp and libeth_pci.
> > > >
> > > > You can do whatever module architecture for netdev devices, but if you
> > > > plan to expose it to RDMA devices, I will vote against any deep layered
> > > > module architecture for the drivers.
> > >
> > > No plans for RDMA there.
> > >
> > > Maybe link the whole kernel to one vmlinux then?
> >
> > It seems that you didn't understand at all about what we are talking
> > here. Please use the opportunity that you are working for the same
> > company with Larysa and ask her offline. She understood perfectly about
> > which modules we are talking.
> >
>
> While I do understand what kind of module relationship you consider problematic,
Awesome, thanks.
> I still struggle to understand why stateless lib hierarchy can be problematic.
As I said already, I wrote my remark as a general comment. It is just
a matter of time when perfectly working system will be changed to less
working one. So when you and Alexander are focused to see what is wrong
now, I see what can be in the future.
To make it clear, even for people who sentimentally attached to libeth code:
I didn't ask to change anything, just tried to understand why
you did it like you did it.
> The fixes that you linked relate more to problematic resource sharing, of which
> libeth has none, it does not have its own memory or its own threads, this is
> just collection of data structures and functions.
It is just a matter of time and you will get same issues like I posted.
Thanks
>
> > >
> > > >
> > > > BTW, please add some Intel prefix to the modules names, they shouldn't
> > > > be called in generic names like libeth, e.t.c
> > >
> > > Two modules with the same name can't exist within the kernel. libeth was
> > > available and I haven't seen anyone wanting to take it. It's not common
> > > at all to name a module starting with "lib".
> >
> > Again, please talk with Larysa. ETH part is problematic in libeth name
> > and not LIB.
> >
> > Thanks
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [Intel-wired-lan] [PATCH iwl-next 05/14] libeth: add control queue support
2025-04-10 13:44 ` Leon Romanovsky
2025-04-10 13:59 ` Larysa Zaremba
@ 2025-04-10 14:00 ` Alexander Lobakin
1 sibling, 0 replies; 36+ messages in thread
From: Alexander Lobakin @ 2025-04-10 14:00 UTC (permalink / raw)
To: Leon Romanovsky
Cc: Larysa Zaremba, intel-wired-lan, Tony Nguyen, David S. Miller,
Eric Dumazet, Jakub Kicinski, Paolo Abeni, Simon Horman,
Jonathan Corbet, Przemek Kitszel, Jiri Pirko, Tatyana Nikolova,
Andrew Lunn, Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Sridhar Samudrala, Jacob Keller,
Michal Swiatkowski, Mateusz Polchlopek, Ahmed Zaki, netdev,
linux-doc, linux-kernel, Karlsson, Magnus, Emil Tantilov,
Madhu Chittim, Josh Hay, Milena Olech, pavan.kumar.linga,
Singhai, Anjali, Phani R Burra
From: Leon Romanovsky <leon@kernel.org>
Date: Thu, 10 Apr 2025 16:44:43 +0300
> On Thu, Apr 10, 2025 at 03:05:19PM +0200, Alexander Lobakin wrote:
>> From: Leon Romanovsky <leon@kernel.org>
>> Date: Thu, 10 Apr 2025 14:23:49 +0300
>>
>>> On Thu, Apr 10, 2025 at 12:44:33PM +0200, Larysa Zaremba wrote:
>>>> On Thu, Apr 10, 2025 at 11:21:37AM +0300, Leon Romanovsky wrote:
>>>>> On Tue, Apr 08, 2025 at 02:47:51PM +0200, Larysa Zaremba wrote:
>>>>>> From: Phani R Burra <phani.r.burra@intel.com>
>>>>>>
>>>>>> Libeth will now support control queue setup and configuration APIs.
>>>>>> These are mainly used for mailbox communication between drivers and
>>>>>> control plane.
>>>>>>
>>>>>> Make use of the page pool support for managing controlq buffers.
>
> <...>
>
>>>> Module dependencies are as follows:
>>>>
>>>> libeth_rx and libeth_pci do not depend on other modules.
>>>> libeth_cp depends on both libeth_rx and libeth_pci.
>>>> idpf directly uses libeth_pci, libeth_rx and libeth_cp.
>>>> ixd directly uses libeth_cp and libeth_pci.
>>>
>>> You can do whatever module architecture for netdev devices, but if you
>>> plan to expose it to RDMA devices, I will vote against any deep layered
>>> module architecture for the drivers.
>>
>> No plans for RDMA there.
>>
>> Maybe link the whole kernel to one vmlinux then?
>
> It seems that you didn't understand at all about what we are talking
> here. Please use the opportunity that you are working for the same
> company with Larysa and ask her offline. She understood perfectly about
> which modules we are talking.
She described to you how this looks like currently. No Larysa didn't
understand how scenarios like you described are possible.
>
>>
>>>
>>> BTW, please add some Intel prefix to the modules names, they shouldn't
>>> be called in generic names like libeth, e.t.c
>>
>> Two modules with the same name can't exist within the kernel. libeth was
>> available and I haven't seen anyone wanting to take it. It's not common
>> at all to name a module starting with "lib".
>
> Again, please talk with Larysa. ETH part is problematic in libeth name
> and not LIB.
Do you think we don't talk internally? I wouldn't write all this if
Larysa agreed with you or understood clearly what the problem is.
Don't make her responsible for your own words.
I've still noticed zero proofs for both topics.
As for libeth_, there are no generic code which would use this exact
prefix. Not "lib", not "eth", but whole "libeth".
There are no clear rules how to name or not name the modules. If it's
free and the subsys maintainers are fine, it's fine.
Thanks,
Olek
^ permalink raw reply [flat|nested] 36+ messages in thread
* [PATCH iwl-next 06/14] libeth: add bookkeeping support for control queue messages
2025-04-08 12:47 [PATCH iwl-next 00/14] Introduce iXD driver Larysa Zaremba
` (4 preceding siblings ...)
2025-04-08 12:47 ` [PATCH iwl-next 05/14] libeth: add control queue support Larysa Zaremba
@ 2025-04-08 12:47 ` Larysa Zaremba
2025-04-15 18:54 ` Tony Nguyen
2025-04-21 14:19 ` Simon Horman
2025-04-08 12:47 ` [PATCH iwl-next 07/14] idpf: remove 'vport_params_reqd' field Larysa Zaremba
` (8 subsequent siblings)
14 siblings, 2 replies; 36+ messages in thread
From: Larysa Zaremba @ 2025-04-08 12:47 UTC (permalink / raw)
To: intel-wired-lan, Tony Nguyen
Cc: David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Simon Horman, Jonathan Corbet, Przemek Kitszel, Jiri Pirko,
Mustafa Ismail, Tatyana Nikolova, Andrew Lunn, Alexander Lobakin,
Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Larysa Zaremba, Sridhar Samudrala,
Jacob Keller, Michal Swiatkowski, Mateusz Polchlopek, Wenjun Wu,
Ahmed Zaki, netdev, linux-doc, linux-kernel, Karlsson, Magnus,
Emil Tantilov, Madhu Chittim, Josh Hay, Milena Olech,
pavan.kumar.linga, Singhai, Anjali, Phani R Burra
From: Phani R Burra <phani.r.burra@intel.com>
All send control queue messages are allocated/freed in libeth itself
and tracked with the unique transaction (Xn) ids until they receive
response or time out. Responses can be received out of order, therefore
transactions are stored in an array and tracked though a bitmap.
Pre-allocated DMA memory is used where possible. It reduces the driver
overhead in handling memory allocation/free and message timeouts.
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Phani R Burra <phani.r.burra@intel.com>
Co-developed-by: Victor Raj <victor.raj@intel.com>
Signed-off-by: Victor Raj <victor.raj@intel.com>
Co-developed-by: Pavan Kumar Linga <pavan.kumar.linga@intel.com>
Signed-off-by: Pavan Kumar Linga <pavan.kumar.linga@intel.com>
Co-developed-by: Larysa Zaremba <larysa.zaremba@intel.com>
Signed-off-by: Larysa Zaremba <larysa.zaremba@intel.com>
---
drivers/net/ethernet/intel/libeth/controlq.c | 578 +++++++++++++++++++
include/net/libeth/controlq.h | 169 ++++++
2 files changed, 747 insertions(+)
diff --git a/drivers/net/ethernet/intel/libeth/controlq.c b/drivers/net/ethernet/intel/libeth/controlq.c
index 2ce419591495..f01666701039 100644
--- a/drivers/net/ethernet/intel/libeth/controlq.c
+++ b/drivers/net/ethernet/intel/libeth/controlq.c
@@ -603,6 +603,584 @@ u32 libeth_ctlq_recv(struct libeth_ctlq_info *ctlq, struct libeth_ctlq_msg *msg,
}
EXPORT_SYMBOL_NS_GPL(libeth_ctlq_recv, "LIBETH_CP");
+/**
+ * libeth_ctlq_xn_pop_free - get a free Xn entry from the free list
+ * @xnm: Xn transaction manager
+ *
+ * Retrieve a free Xn entry from the free list.
+ *
+ * Return: valid Xn entry pointer or NULL if there are no free Xn entries.
+ */
+static struct libeth_ctlq_xn *
+libeth_ctlq_xn_pop_free(struct libeth_ctlq_xn_manager *xnm)
+{
+ struct libeth_ctlq_xn *xn;
+ u32 free_idx;
+
+ guard(spinlock)(&xnm->free_xns_bm_lock);
+
+ if (unlikely(xnm->shutdown))
+ return NULL;
+
+ free_idx = find_next_bit(xnm->free_xns_bm, LIBETH_CTLQ_MAX_XN_ENTRIES,
+ 0);
+ if (free_idx == LIBETH_CTLQ_MAX_XN_ENTRIES)
+ return NULL;
+
+ __clear_bit(free_idx, xnm->free_xns_bm);
+ xn = &xnm->ring[free_idx];
+ xn->cookie = xnm->cookie++;
+
+ return xn;
+}
+
+/**
+ * __libeth_ctlq_xn_push_free - unsafely push a Xn entry into the free list
+ * @xnm: Xn transaction manager
+ * @xn: xn entry to be added into the free list
+ */
+static void __libeth_ctlq_xn_push_free(struct libeth_ctlq_xn_manager *xnm,
+ struct libeth_ctlq_xn *xn)
+{
+ __set_bit(xn->index, xnm->free_xns_bm);
+
+ if (likely(!xnm->shutdown))
+ return;
+
+ if (bitmap_full(xnm->free_xns_bm, LIBETH_CTLQ_MAX_XN_ENTRIES))
+ complete(&xnm->can_destroy);
+}
+
+/**
+ * libeth_ctlq_xn_push_free - push a Xn entry into the free list
+ * @xnm: Xn transaction manager
+ * @xn: xn entry to be added into the free list, not locked
+ *
+ * Safely add a used Xn entry back to the free list.
+ */
+static void libeth_ctlq_xn_push_free(struct libeth_ctlq_xn_manager *xnm,
+ struct libeth_ctlq_xn *xn)
+{
+ guard(spinlock)(&xnm->free_xns_bm_lock);
+
+ __libeth_ctlq_xn_push_free(xnm, xn);
+}
+
+/**
+ * libeth_ctlq_xn_put - put an Xn that will not be used in the current thread
+ * @xnm: Xn transaction manager
+ * @xn: async xn entry to be put for now, not locked
+ *
+ * If the Xn manager is being shutdown, nothing will handle the related
+ * async request.
+ */
+static void libeth_ctlq_xn_put(struct libeth_ctlq_xn_manager *xnm,
+ struct libeth_ctlq_xn *xn)
+{
+ guard(spinlock)(&xnm->free_xns_bm_lock);
+
+ if (unlikely(xnm->shutdown))
+ __libeth_ctlq_xn_push_free(xnm, xn);
+}
+
+/**
+ * libeth_ctlq_xn_deinit_dma - free the DMA memory allocated for send messages
+ * @dev: device pointer
+ * @xnm: pointer to the transaction manager
+ * @num_entries: number of Xn entries to free the DMA for
+ */
+static void libeth_ctlq_xn_deinit_dma(struct device *dev,
+ struct libeth_ctlq_xn_manager *xnm,
+ u32 num_entries)
+{
+ for (u32 i = 0; i < num_entries; i++) {
+ struct libeth_ctlq_xn *xn = &xnm->ring[i];
+
+ libeth_cp_free_dma_mem(dev, xn->dma_mem);
+ kfree(xn->dma_mem);
+ }
+}
+
+/**
+ * libeth_ctlq_xn_init_dma - pre-allocate DMA memory for send messages that use
+ * stack variables
+ * @dev: device pointer
+ * @xnm: pointer to transaction manager
+ *
+ * Return: %0 on success or error if memory allocation fails
+ */
+static int libeth_ctlq_xn_init_dma(struct device *dev,
+ struct libeth_ctlq_xn_manager *xnm)
+{
+ u32 i;
+
+ for (i = 0; i < LIBETH_CTLQ_MAX_XN_ENTRIES; i++) {
+ struct libeth_ctlq_xn *xn = &xnm->ring[i];
+ struct libeth_cp_dma_mem *dma_mem;
+
+ dma_mem = kzalloc(sizeof(*dma_mem), GFP_KERNEL);
+ if (!dma_mem)
+ goto dealloc_dma;
+
+ dma_mem->va = libeth_cp_alloc_dma_mem(dev, dma_mem,
+ LIBETH_CTLQ_MAX_BUF_LEN);
+ if (!dma_mem->va) {
+ kfree(dma_mem);
+ goto dealloc_dma;
+ }
+
+ xn->dma_mem = dma_mem;
+ }
+
+ return 0;
+
+dealloc_dma:
+ libeth_ctlq_xn_deinit_dma(dev, xnm, i);
+
+ return -ENOMEM;
+}
+
+/**
+ * libeth_ctlq_xn_process_recv - process Xn data in receive message
+ * @params: Xn receive param information to handle a receive message
+ * @ctlq_msg: received control queue message
+ *
+ * Process a control queue receive message and send a complete event
+ * notification.
+ *
+ * Return: true if a message has been processed, false otherwise.
+ */
+static bool
+libeth_ctlq_xn_process_recv(struct libeth_ctlq_xn_recv_params *params,
+ struct libeth_ctlq_msg *ctlq_msg)
+{
+ struct libeth_ctlq_xn_manager *xnm = params->xnm;
+ struct libeth_ctlq_xn *xn;
+ u16 msg_cookie, xn_index;
+ struct kvec *response;
+ int status;
+ u16 data;
+
+ data = ctlq_msg->sw_cookie;
+ xn_index = FIELD_GET(LIBETH_CTLQ_XN_INDEX_M, data);
+ msg_cookie = FIELD_GET(LIBETH_CTLQ_XN_COOKIE_M, data);
+ status = ctlq_msg->chnl_retval ? -EFAULT : 0;
+
+ xn = &xnm->ring[xn_index];
+ if (ctlq_msg->chnl_opcode != xn->virtchnl_opcode ||
+ msg_cookie != xn->cookie)
+ return false;
+
+ spin_lock(&xn->xn_lock);
+ if (xn->state != LIBETH_CTLQ_XN_ASYNC &&
+ xn->state != LIBETH_CTLQ_XN_WAITING) {
+ spin_unlock(&xn->xn_lock);
+ return false;
+ }
+
+ response = &ctlq_msg->recv_mem;
+ if (xn->state == LIBETH_CTLQ_XN_ASYNC) {
+ xn->resp_cb(xn->send_ctx, response, status);
+ xn->state = LIBETH_CTLQ_XN_IDLE;
+ spin_unlock(&xn->xn_lock);
+ libeth_ctlq_xn_push_free(xnm, xn);
+
+ return true;
+ }
+
+ xn->recv_mem = *response;
+ xn->state = status ? LIBETH_CTLQ_XN_COMPLETED_FAILED :
+ LIBETH_CTLQ_XN_COMPLETED_SUCCESS;
+
+ complete(&xn->cmd_completion_event);
+ spin_unlock(&xn->xn_lock);
+
+ return true;
+}
+
+/**
+ * libeth_xn_check_async_timeout - Check for asynchronous message timeouts
+ * @xnm: Xn transaction manager
+ *
+ * Call the corresponding callback to notify the caller about the timeout.
+ */
+static void libeth_xn_check_async_timeout(struct libeth_ctlq_xn_manager *xnm)
+{
+ u32 idx;
+
+ for_each_clear_bit(idx, xnm->free_xns_bm, LIBETH_CTLQ_MAX_XN_ENTRIES) {
+ struct libeth_ctlq_xn *xn = &xnm->ring[idx];
+ u64 timeout_ms;
+
+ spin_lock(&xn->xn_lock);
+
+ timeout_ms = ktime_ms_delta(ktime_get(), xn->timestamp);
+ if (xn->state != LIBETH_CTLQ_XN_ASYNC ||
+ timeout_ms < xn->timeout_ms) {
+ spin_unlock(&xn->xn_lock);
+ continue;
+ }
+
+ xn->resp_cb(xn->send_ctx, NULL, -ETIMEDOUT);
+ xn->state = LIBETH_CTLQ_XN_IDLE;
+ spin_unlock(&xn->xn_lock);
+ libeth_ctlq_xn_push_free(xnm, xn);
+ }
+}
+
+/**
+ * libeth_ctlq_xn_recv - process control queue receive message
+ * @params: Xn receive param information to handle a receive message
+ *
+ * Process a receive message and update the receive queue buffer.
+ *
+ * Return: true if a message has been processed, false otherwise.
+ */
+bool libeth_ctlq_xn_recv(struct libeth_ctlq_xn_recv_params *params)
+{
+ struct libeth_ctlq_msg ctlq_msg;
+ u32 num_msgs;
+
+ num_msgs = libeth_ctlq_recv(params->ctlq, &ctlq_msg, 1);
+
+ if (num_msgs && !libeth_ctlq_xn_process_recv(params, &ctlq_msg))
+ params->ctlq_msg_handler(params->xnm->ctx, &ctlq_msg);
+
+ libeth_ctlq_post_rx_buffs(params->ctlq);
+ libeth_xn_check_async_timeout(params->xnm);
+
+ return !!num_msgs;
+}
+EXPORT_SYMBOL_NS_GPL(libeth_ctlq_xn_recv, "LIBETH_CP");
+
+/**
+ * libeth_cp_map_dma_mem - map a given virtual address for DMA
+ * @dev: device information
+ * @va: virtual address to be mapped
+ * @size: size of the memory
+ * @direction: DMA direction either from/to device
+ * @dma_mem: memory for DMA information to be stored
+ *
+ * Return: true on success, false on DMA map failure.
+ */
+static bool libeth_cp_map_dma_mem(struct device *dev, void *va, size_t size,
+ int direction,
+ struct libeth_cp_dma_mem *dma_mem)
+{
+ dma_mem->pa = dma_map_single(dev, va, size, direction);
+
+ return dma_mapping_error(dev, dma_mem->pa) ? false : true;
+}
+
+/**
+ * libeth_cp_unmap_dma_mem - unmap previously mapped DMA address
+ * @dev: device information
+ * @dma_mem: DMA memory information
+ */
+static void libeth_cp_unmap_dma_mem(struct device *dev,
+ const struct libeth_cp_dma_mem *dma_mem)
+{
+ dma_unmap_single(dev, dma_mem->pa, dma_mem->size,
+ dma_mem->direction);
+}
+
+/**
+ * libeth_ctlq_xn_process_send - process and send a control queue message
+ * @params: Xn send param information for sending a control queue message
+ * @xn: Assigned Xn entry for tracking the control queue message
+ *
+ * Return: %0 on success, -%errno on failure.
+ */
+static
+int libeth_ctlq_xn_process_send(struct libeth_ctlq_xn_send_params *params,
+ struct libeth_ctlq_xn *xn)
+{
+ size_t buf_len = params->send_buf.iov_len;
+ struct device *dev = params->ctlq->dev;
+ void *buf = params->send_buf.iov_base;
+ struct libeth_cp_dma_mem *dma_mem;
+ u16 cookie;
+ int ret;
+
+ if (!buf || !buf_len)
+ return -EOPNOTSUPP;
+
+ if (libeth_cp_can_send_onstack(buf_len)) {
+ dma_mem = xn->dma_mem;
+ memcpy(dma_mem->va, buf, buf_len);
+ } else {
+ dma_mem = &xn->send_dma_mem;
+ dma_mem->va = buf;
+ dma_mem->size = buf_len;
+ dma_mem->direction = DMA_TO_DEVICE;
+
+ if (!libeth_cp_map_dma_mem(dev, buf, buf_len, DMA_TO_DEVICE,
+ dma_mem))
+ return -ENOMEM;
+ }
+
+ cookie = FIELD_PREP(LIBETH_CTLQ_XN_COOKIE_M, xn->cookie) |
+ FIELD_PREP(LIBETH_CTLQ_XN_INDEX_M, xn->index);
+
+ scoped_guard(spinlock, ¶ms->ctlq->lock) {
+ if (!params->ctlq_msg || params->resp_cb) {
+ struct libeth_ctlq_info *ctlq = params->ctlq;
+
+ *ctlq->tx_msg[ctlq->next_to_use] =
+ params->ctlq_msg ? *params->ctlq_msg :
+ (struct libeth_ctlq_msg) {
+ .opcode = LIBETH_CTLQ_SEND_MSG_TO_CP
+ };
+ params->ctlq_msg = ctlq->tx_msg[ctlq->next_to_use];
+ }
+
+ params->ctlq_msg->sw_cookie = cookie;
+ params->ctlq_msg->send_mem = *dma_mem;
+ params->ctlq_msg->data_len = buf_len;
+ params->ctlq_msg->chnl_opcode = params->chnl_opcode;
+ ret = libeth_ctlq_send(params->ctlq, params->ctlq_msg, 1);
+ }
+
+ if (ret && !libeth_cp_can_send_onstack(buf_len))
+ libeth_cp_unmap_dma_mem(dev, dma_mem);
+
+ return ret;
+}
+
+/**
+ * libeth_ctlq_xn_send - Function to send a control queue message
+ * @params: Xn send param information for sending a control queue message
+ *
+ * Send a control queue (mailbox or config) message.
+ * Based on the params value, the call can be completed synchronously or
+ * asynchronously.
+ *
+ * Return: %0 on success, -%errno on failure.
+ */
+int libeth_ctlq_xn_send(struct libeth_ctlq_xn_send_params *params)
+{
+ bool free_send = !libeth_cp_can_send_onstack(params->send_buf.iov_len);
+ struct libeth_ctlq_xn *xn;
+ int ret;
+
+ if (params->send_buf.iov_len > LIBETH_CTLQ_MAX_BUF_LEN) {
+ ret = -EINVAL;
+ goto free_buf;
+ }
+
+ xn = libeth_ctlq_xn_pop_free(params->xnm);
+ /* no free transactions available */
+ if (unlikely(!xn)) {
+ ret = -EAGAIN;
+ goto free_buf;
+ }
+
+ spin_lock(&xn->xn_lock);
+
+ xn->state = params->resp_cb ? LIBETH_CTLQ_XN_ASYNC :
+ LIBETH_CTLQ_XN_WAITING;
+ xn->ctlq = params->ctlq;
+ xn->virtchnl_opcode = params->chnl_opcode;
+
+ if (params->resp_cb) {
+ xn->send_ctx = params->send_ctx;
+ xn->resp_cb = params->resp_cb;
+ xn->timeout_ms = params->timeout_ms;
+ xn->timestamp = ktime_get();
+ }
+
+ ret = libeth_ctlq_xn_process_send(params, xn);
+ if (ret)
+ goto release_xn;
+ else
+ free_send = false;
+
+ spin_unlock(&xn->xn_lock);
+
+ if (params->resp_cb) {
+ libeth_ctlq_xn_put(params->xnm, xn);
+ return 0;
+ }
+
+ wait_for_completion_timeout(&xn->cmd_completion_event,
+ msecs_to_jiffies(params->timeout_ms));
+
+ spin_lock(&xn->xn_lock);
+ switch (xn->state) {
+ case LIBETH_CTLQ_XN_WAITING:
+ ret = -ETIMEDOUT;
+ break;
+ case LIBETH_CTLQ_XN_COMPLETED_SUCCESS:
+ params->recv_mem = xn->recv_mem;
+ break;
+ default:
+ ret = -EBADMSG;
+ break;
+ }
+
+ /* Free the receive buffer in case of failure. On timeout, receive
+ * buffer is not allocated.
+ */
+ if (ret && ret != -ETIMEDOUT)
+ libeth_ctlq_release_rx_buf(xn->ctlq, &xn->recv_mem);
+
+release_xn:
+ xn->state = LIBETH_CTLQ_XN_IDLE;
+ reinit_completion(&xn->cmd_completion_event);
+ spin_unlock(&xn->xn_lock);
+ libeth_ctlq_xn_push_free(params->xnm, xn);
+free_buf:
+ if (free_send)
+ params->rel_tx_buf(params->send_buf.iov_base);
+
+ return ret;
+}
+EXPORT_SYMBOL_NS_GPL(libeth_ctlq_xn_send, "LIBETH_CP");
+
+/**
+ * libeth_ctlq_xn_send_clean - cleanup the send control queue message buffers
+ * @params: Xn clean param information for send complete handling
+ *
+ * Cleanup the send buffers for the given control queue, if force is set, then
+ * clear all the outstanding send messages irrrespective their send status.
+ * Force should be used during deinit or reset.
+ *
+ * Return: number of send buffers cleaned.
+ */
+u32 libeth_ctlq_xn_send_clean(const struct libeth_ctlq_xn_clean_params *params)
+{
+ struct libeth_ctlq_info *ctlq = params->ctlq;
+ struct device *dev = ctlq->dev;
+ u32 ntc, i;
+
+ spin_lock(&ctlq->lock);
+ ntc = ctlq->next_to_clean;
+
+ for (i = 0; i < params->num_msgs; i++) {
+ struct libeth_ctlq_msg *msg = ctlq->tx_msg[ntc];
+ struct libeth_ctlq_desc *desc;
+ u64 qword;
+
+ desc = &ctlq->descs[ntc];
+ qword = le64_to_cpu(desc->qword0);
+
+ if (!FIELD_GET(LIBETH_CTLQ_DESC_FLAG_DD, qword))
+ break;
+
+ dma_rmb();
+
+ if (!libeth_cp_can_send_onstack(msg->data_len)) {
+ libeth_cp_unmap_dma_mem(dev, &msg->send_mem);
+ params->rel_tx_buf(msg->send_mem.va);
+ }
+
+ memset(msg, 0, sizeof(*msg));
+ desc->qword0 = 0;
+
+ if (unlikely(++ntc == ctlq->ring_len))
+ ntc = 0;
+ }
+
+ ctlq->next_to_clean = ntc;
+ spin_unlock(&ctlq->lock);
+
+ return i;
+}
+EXPORT_SYMBOL_NS_GPL(libeth_ctlq_xn_send_clean, "LIBETH_CP");
+
+/**
+ * libeth_ctlq_xn_deinit - deallocate and free the transaction manager resources
+ * @xnm: pointer to the transaction manager
+ * @ctx: controlq context structure
+ *
+ * All Rx processing must be stopped beforehand.
+ */
+void libeth_ctlq_xn_deinit(struct libeth_ctlq_xn_manager *xnm,
+ struct libeth_ctlq_ctx *ctx)
+{
+ bool must_wait = false;
+ u32 i;
+
+ /* Should be no new clear bits after this */
+ spin_lock(&xnm->free_xns_bm_lock);
+ xnm->shutdown = true;
+
+ for_each_clear_bit(i, xnm->free_xns_bm, LIBETH_CTLQ_MAX_XN_ENTRIES) {
+ struct libeth_ctlq_xn *xn = &xnm->ring[i];
+
+ spin_lock(&xn->xn_lock);
+
+ if (xn->state == LIBETH_CTLQ_XN_WAITING ||
+ xn->state == LIBETH_CTLQ_XN_IDLE) {
+ complete(&xn->cmd_completion_event);
+ must_wait = true;
+ } else if (xn->state == LIBETH_CTLQ_XN_ASYNC) {
+ __libeth_ctlq_xn_push_free(xnm, xn);
+ }
+
+ spin_unlock(&xn->xn_lock);
+ }
+
+ spin_unlock(&xnm->free_xns_bm_lock);
+
+ if (must_wait)
+ wait_for_completion(&xnm->can_destroy);
+
+ libeth_ctlq_xn_deinit_dma(&ctx->mmio_info.pdev->dev, xnm,
+ LIBETH_CTLQ_MAX_XN_ENTRIES);
+ kfree(xnm);
+ libeth_ctlq_deinit(ctx);
+}
+EXPORT_SYMBOL_NS_GPL(libeth_ctlq_xn_deinit, "LIBETH_CP");
+
+/**
+ * libeth_ctlq_xn_init - initialize the Xn transaction manager
+ * @params: Xn init param information for allocating Xn manager resources
+ *
+ * Return: %0 on success, -%errno on failure.
+ */
+int libeth_ctlq_xn_init(struct libeth_ctlq_xn_init_params *params)
+{
+ struct libeth_ctlq_xn_manager *xnm;
+ int ret;
+
+ ret = libeth_ctlq_init(params->ctx, params->cctlq_info, params->num_qs);
+ if (ret)
+ return ret;
+
+ xnm = kzalloc(sizeof(*xnm), GFP_KERNEL);
+ if (!xnm)
+ goto ctlq_deinit;
+
+ ret = libeth_ctlq_xn_init_dma(¶ms->ctx->mmio_info.pdev->dev, xnm);
+ if (ret)
+ goto free_xnm;
+
+ spin_lock_init(&xnm->free_xns_bm_lock);
+ init_completion(&xnm->can_destroy);
+ bitmap_fill(xnm->free_xns_bm, LIBETH_CTLQ_MAX_XN_ENTRIES);
+
+ for (u32 i = 0; i < LIBETH_CTLQ_MAX_XN_ENTRIES; i++) {
+ struct libeth_ctlq_xn *xn = &xnm->ring[i];
+
+ xn->index = i;
+ init_completion(&xn->cmd_completion_event);
+ spin_lock_init(&xn->xn_lock);
+ }
+ xnm->ctx = params->ctx;
+ params->xnm = xnm;
+
+ return 0;
+
+free_xnm:
+ kfree(xnm);
+ctlq_deinit:
+ libeth_ctlq_deinit(params->ctx);
+
+ return -ENOMEM;
+}
+EXPORT_SYMBOL_NS_GPL(libeth_ctlq_xn_init, "LIBETH_CP");
+
MODULE_DESCRIPTION("Control Plane communication API");
MODULE_IMPORT_NS("LIBETH");
MODULE_LICENSE("GPL");
diff --git a/include/net/libeth/controlq.h b/include/net/libeth/controlq.h
index 35e69d0f48a4..a564a8cccb06 100644
--- a/include/net/libeth/controlq.h
+++ b/include/net/libeth/controlq.h
@@ -4,6 +4,8 @@
#ifndef __LIBETH_CONTROLQ_H
#define __LIBETH_CONTROLQ_H
+#include <linux/ktime.h>
+
#include <linux/intel/virtchnl2.h>
#include <net/libeth/pci.h>
@@ -20,6 +22,8 @@
#define LIBETH_CTLQ_SEND_MSG_TO_CP 0x801
#define LIBETH_CTLQ_SEND_MSG_TO_PEER 0x804
+#define LIBETH_CP_TX_COPYBREAK 128
+
/**
* struct libeth_ctlq_ctx - contains controlq info and MMIO region info
* @mmio_info: MMIO region info structure
@@ -60,11 +64,13 @@ struct libeth_ctlq_reg {
* @va: virtual address
* @pa: physical address
* @size: memory size
+ * @direction: memory to device or device to memory
*/
struct libeth_cp_dma_mem {
void *va;
dma_addr_t pa;
size_t size;
+ int direction;
};
/**
@@ -248,4 +254,167 @@ u32 libeth_ctlq_recv(struct libeth_ctlq_info *ctlq, struct libeth_ctlq_msg *msg,
int libeth_ctlq_post_rx_buffs(struct libeth_ctlq_info *ctlq);
+/* Only 8 bits are available in descriptor for Xn index */
+#define LIBETH_CTLQ_MAX_XN_ENTRIES 256
+#define LIBETH_CTLQ_XN_COOKIE_M GENMASK(15, 8)
+#define LIBETH_CTLQ_XN_INDEX_M GENMASK(7, 0)
+
+/**
+ * enum libeth_ctlq_xn_state - Transaction state of a virtchnl message
+ * @LIBETH_CTLQ_XN_IDLE: transaction is available to use
+ * @LIBETH_CTLQ_XN_WAITING: waiting for transaction to complete
+ * @LIBETH_CTLQ_XN_COMPLETED_SUCCESS: transaction completed with success
+ * @LIBETH_CTLQ_XN_COMPLETED_FAILED: transaction completed with failure
+ * @LIBETH_CTLQ_XN_ASYNC: asynchronous virtchnl message transaction type
+ */
+enum libeth_ctlq_xn_state {
+ LIBETH_CTLQ_XN_IDLE = 0,
+ LIBETH_CTLQ_XN_WAITING,
+ LIBETH_CTLQ_XN_COMPLETED_SUCCESS,
+ LIBETH_CTLQ_XN_COMPLETED_FAILED,
+ LIBETH_CTLQ_XN_ASYNC,
+};
+
+/**
+ * struct libeth_ctlq_xn - structure representing a virtchnl transaction entry
+ * @resp_cb: callback to handle the response of an asynchronous virtchnl message
+ * @xn_lock: lock to protect the transaction entry state
+ * @ctlq: send control queue information
+ * @cmd_completion_event: signal when a reply is available
+ * @dma_mem: DMA memory of send buffer that use stack variable
+ * @send_dma_mem: DMA memory of send buffer
+ * @recv_mem: receive buffer
+ * @send_ctx: context for callback function
+ * @timeout_ms: Xn transaction timeout in msecs
+ * @timestamp: timestamp to record the Xn send
+ * @virtchnl_opcode: virtchnl command opcode used for Xn transaction
+ * @state: transaction state of a virtchnl message
+ * @cookie: unique message identifier
+ * @index: index of the transaction entry
+ */
+struct libeth_ctlq_xn {
+ void (*resp_cb)(void *ctx, struct kvec *mem, int status);
+ spinlock_t xn_lock; /* protects state */
+ struct libeth_ctlq_info *ctlq;
+ struct completion cmd_completion_event;
+ struct libeth_cp_dma_mem *dma_mem;
+ struct libeth_cp_dma_mem send_dma_mem;
+ struct kvec recv_mem;
+ void *send_ctx;
+ u64 timeout_ms;
+ ktime_t timestamp;
+ u32 virtchnl_opcode;
+ enum libeth_ctlq_xn_state state;
+ u8 cookie;
+ u8 index;
+};
+
+/**
+ * struct libeth_ctlq_xn_manager - structure representing the array of virtchnl
+ * transaction entries
+ * @ctx: pointer to controlq context structure
+ * @free_xns_bm_lock: lock to protect the free Xn entries bit map
+ * @free_xns_bm: bitmap that represents the free Xn entries
+ * @ring: array of Xn entries
+ * @cookie: unique message identifier
+ */
+struct libeth_ctlq_xn_manager {
+ struct libeth_ctlq_ctx *ctx;
+ spinlock_t free_xns_bm_lock; /* get/check entries */
+ DECLARE_BITMAP(free_xns_bm, LIBETH_CTLQ_MAX_XN_ENTRIES);
+ struct libeth_ctlq_xn ring[LIBETH_CTLQ_MAX_XN_ENTRIES];
+ struct completion can_destroy;
+ bool shutdown;
+ u8 cookie;
+};
+
+/**
+ * struct libeth_ctlq_xn_send_params - structure representing send Xn entry
+ * @resp_cb: callback to handle the response of an asynchronous virtchnl message
+ * @rel_tx_buf: driver entry point for freeing the send buffer after send
+ * @xnm: Xn manager to process Xn entries
+ * @ctlq: send control queue information
+ * @ctlq_msg: control queue message information
+ * @send_buf: represents the buffer that carries outgoing information
+ * @recv_mem: receive buffer
+ * @send_ctx: context for call back function
+ * @timeout_ms: virtchnl transaction timeout in msecs
+ * @chnl_opcode: virtchnl message opcode
+ */
+struct libeth_ctlq_xn_send_params {
+ void (*resp_cb)(void *ctx, struct kvec *mem, int status);
+ void (*rel_tx_buf)(const void *buf_va);
+ struct libeth_ctlq_xn_manager *xnm;
+ struct libeth_ctlq_info *ctlq;
+ struct libeth_ctlq_msg *ctlq_msg;
+ struct kvec send_buf;
+ struct kvec recv_mem;
+ void *send_ctx;
+ u64 timeout_ms;
+ u32 chnl_opcode;
+};
+
+/**
+ * libeth_cp_can_send_onstack - find if a virtchnl message can be sent using a
+ * stack variable
+ * @size: virtchnl buffer size
+ */
+static inline bool libeth_cp_can_send_onstack(u32 size)
+{
+ return size <= LIBETH_CP_TX_COPYBREAK;
+}
+
+/**
+ * struct libeth_ctlq_xn_recv_params - structure representing receive Xn entry
+ * @ctlq_msg_handler: callback to handle a message originated from the peer
+ * @xnm: Xn manager to process Xn entries
+ * @ctx: pointer to context structure
+ * @ctlq: control queue information
+ */
+struct libeth_ctlq_xn_recv_params {
+ void (*ctlq_msg_handler)(struct libeth_ctlq_ctx *ctx,
+ struct libeth_ctlq_msg *msg);
+ struct libeth_ctlq_xn_manager *xnm;
+ struct libeth_ctlq_info *ctlq;
+};
+
+/**
+ * struct libeth_ctlq_xn_clean_params - Data structure used for cleaning the
+ * control queue messages
+ * @rel_tx_buf: driver entry point for freeing the send buffer after send
+ * @ctx: pointer to context structure
+ * @ctlq: control queue information
+ * @send_ctx: context for call back function
+ * @num_msgs: number of messages to be cleaned
+ */
+struct libeth_ctlq_xn_clean_params {
+ void (*rel_tx_buf)(const void *buf_va);
+ struct libeth_ctlq_ctx *ctx;
+ struct libeth_ctlq_info *ctlq;
+ void *send_ctx;
+ u16 num_msgs;
+};
+
+/**
+ * struct libeth_ctlq_xn_init_params - Data structure used for initializing the
+ * Xn transaction manager
+ * @cctlq_info: control queue information
+ * @ctx: pointer to controlq context structure
+ * @xnm: Xn manager to process Xn entries
+ * @num_qs: number of control queues needs to initialized
+ */
+struct libeth_ctlq_xn_init_params {
+ struct libeth_ctlq_create_info *cctlq_info;
+ struct libeth_ctlq_ctx *ctx;
+ struct libeth_ctlq_xn_manager *xnm;
+ u32 num_qs;
+};
+
+int libeth_ctlq_xn_init(struct libeth_ctlq_xn_init_params *params);
+void libeth_ctlq_xn_deinit(struct libeth_ctlq_xn_manager *xnm,
+ struct libeth_ctlq_ctx *ctx);
+int libeth_ctlq_xn_send(struct libeth_ctlq_xn_send_params *params);
+bool libeth_ctlq_xn_recv(struct libeth_ctlq_xn_recv_params *params);
+u32 libeth_ctlq_xn_send_clean(const struct libeth_ctlq_xn_clean_params *params);
+
#endif /* __LIBETH_CONTROLQ_H */
--
2.47.0
^ permalink raw reply related [flat|nested] 36+ messages in thread* Re: [PATCH iwl-next 06/14] libeth: add bookkeeping support for control queue messages
2025-04-08 12:47 ` [PATCH iwl-next 06/14] libeth: add bookkeeping support for control queue messages Larysa Zaremba
@ 2025-04-15 18:54 ` Tony Nguyen
2025-04-21 14:19 ` Simon Horman
1 sibling, 0 replies; 36+ messages in thread
From: Tony Nguyen @ 2025-04-15 18:54 UTC (permalink / raw)
To: Larysa Zaremba, intel-wired-lan
Cc: David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Simon Horman, Jonathan Corbet, Przemek Kitszel, Jiri Pirko,
Mustafa Ismail, Tatyana Nikolova, Andrew Lunn, Alexander Lobakin,
Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Sridhar Samudrala, Jacob Keller,
Michal Swiatkowski, Mateusz Polchlopek, Wenjun Wu, Ahmed Zaki,
netdev, linux-doc, linux-kernel, Karlsson, Magnus, Emil Tantilov,
Madhu Chittim, Josh Hay, Milena Olech, pavan.kumar.linga,
Singhai, Anjali, Phani R Burra
On 4/8/2025 5:47 AM, Larysa Zaremba wrote:
> From: Phani R Burra <phani.r.burra@intel.com>
...
> +/**
> + * struct libeth_ctlq_xn_manager - structure representing the array of virtchnl
> + * transaction entries
> + * @ctx: pointer to controlq context structure
> + * @free_xns_bm_lock: lock to protect the free Xn entries bit map
> + * @free_xns_bm: bitmap that represents the free Xn entries
> + * @ring: array of Xn entries
> + * @cookie: unique message identifier
> + */
Some kdoc issues:
> include/net/libeth/controlq.h:330: warning: Function parameter or
struct member 'can_destroy' not described in 'libeth_ctlq_xn_manager'
> include/net/libeth/controlq.h:330: warning: Function parameter or
struct member 'shutdown' not described in 'libeth_ctlq_xn_manager'
> +struct libeth_ctlq_xn_manager {
> + struct libeth_ctlq_ctx *ctx;
> + spinlock_t free_xns_bm_lock; /* get/check entries */
> + DECLARE_BITMAP(free_xns_bm, LIBETH_CTLQ_MAX_XN_ENTRIES);
> + struct libeth_ctlq_xn ring[LIBETH_CTLQ_MAX_XN_ENTRIES];
> + struct completion can_destroy;
> + bool shutdown;
> + u8 cookie;
> +};
> +
> +/**
> + * struct libeth_ctlq_xn_send_params - structure representing send Xn entry
> + * @resp_cb: callback to handle the response of an asynchronous virtchnl message
> + * @rel_tx_buf: driver entry point for freeing the send buffer after send
> + * @xnm: Xn manager to process Xn entries
> + * @ctlq: send control queue information
> + * @ctlq_msg: control queue message information
> + * @send_buf: represents the buffer that carries outgoing information
> + * @recv_mem: receive buffer
> + * @send_ctx: context for call back function
> + * @timeout_ms: virtchnl transaction timeout in msecs
> + * @chnl_opcode: virtchnl message opcode
> + */
> +struct libeth_ctlq_xn_send_params {
> + void (*resp_cb)(void *ctx, struct kvec *mem, int status);
> + void (*rel_tx_buf)(const void *buf_va);
> + struct libeth_ctlq_xn_manager *xnm;
> + struct libeth_ctlq_info *ctlq;
> + struct libeth_ctlq_msg *ctlq_msg;
> + struct kvec send_buf;
> + struct kvec recv_mem;
> + void *send_ctx;
> + u64 timeout_ms;
> + u32 chnl_opcode;
> +};
> +
> +/**
> + * libeth_cp_can_send_onstack - find if a virtchnl message can be sent using a
> + * stack variable
> + * @size: virtchnl buffer size
> + */
and here:
> include/net/libeth/controlq.h:364: warning: No description found for
return value of 'libeth_cp_can_send_onstack'
> +static inline bool libeth_cp_can_send_onstack(u32 size)
> +{
> + return size <= LIBETH_CP_TX_COPYBREAK;
> +}
> +
> +/**
> + * struct libeth_ctlq_xn_recv_params - structure representing receive Xn entry
> + * @ctlq_msg_handler: callback to handle a message originated from the peer
> + * @xnm: Xn manager to process Xn entries
> + * @ctx: pointer to context structure
> + * @ctlq: control queue information
> + */
and here:
> include/net/libeth/controlq.h:380: warning: Excess struct member
'ctx' description in 'libeth_ctlq_xn_recv_params'
Thanks,
Tony
> +struct libeth_ctlq_xn_recv_params {
> + void (*ctlq_msg_handler)(struct libeth_ctlq_ctx *ctx,
> + struct libeth_ctlq_msg *msg);
> + struct libeth_ctlq_xn_manager *xnm;
> + struct libeth_ctlq_info *ctlq;
> +};
> +
^ permalink raw reply [flat|nested] 36+ messages in thread* Re: [PATCH iwl-next 06/14] libeth: add bookkeeping support for control queue messages
2025-04-08 12:47 ` [PATCH iwl-next 06/14] libeth: add bookkeeping support for control queue messages Larysa Zaremba
2025-04-15 18:54 ` Tony Nguyen
@ 2025-04-21 14:19 ` Simon Horman
1 sibling, 0 replies; 36+ messages in thread
From: Simon Horman @ 2025-04-21 14:19 UTC (permalink / raw)
To: Larysa Zaremba
Cc: intel-wired-lan, Tony Nguyen, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Jonathan Corbet, Przemek Kitszel,
Jiri Pirko, Mustafa Ismail, Tatyana Nikolova, Andrew Lunn,
Alexander Lobakin, Michael Ellerman, Maciej Fijalkowski,
Lee Trager, Madhavan Srinivasan, Sridhar Samudrala, Jacob Keller,
Michal Swiatkowski, Mateusz Polchlopek, Wenjun Wu, Ahmed Zaki,
netdev, linux-doc, linux-kernel, Karlsson, Magnus, Emil Tantilov,
Madhu Chittim, Josh Hay, Milena Olech, pavan.kumar.linga,
Singhai, Anjali, Phani R Burra
On Tue, Apr 08, 2025 at 02:47:52PM +0200, Larysa Zaremba wrote:
> From: Phani R Burra <phani.r.burra@intel.com>
>
> All send control queue messages are allocated/freed in libeth itself
> and tracked with the unique transaction (Xn) ids until they receive
> response or time out. Responses can be received out of order, therefore
> transactions are stored in an array and tracked though a bitmap.
>
> Pre-allocated DMA memory is used where possible. It reduces the driver
> overhead in handling memory allocation/free and message timeouts.
>
> Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
> Signed-off-by: Phani R Burra <phani.r.burra@intel.com>
> Co-developed-by: Victor Raj <victor.raj@intel.com>
> Signed-off-by: Victor Raj <victor.raj@intel.com>
> Co-developed-by: Pavan Kumar Linga <pavan.kumar.linga@intel.com>
> Signed-off-by: Pavan Kumar Linga <pavan.kumar.linga@intel.com>
> Co-developed-by: Larysa Zaremba <larysa.zaremba@intel.com>
> Signed-off-by: Larysa Zaremba <larysa.zaremba@intel.com>
> ---
> drivers/net/ethernet/intel/libeth/controlq.c | 578 +++++++++++++++++++
> include/net/libeth/controlq.h | 169 ++++++
> 2 files changed, 747 insertions(+)
>
> diff --git a/drivers/net/ethernet/intel/libeth/controlq.c b/drivers/net/ethernet/intel/libeth/controlq.c
...
> +/**
> + * libeth_ctlq_xn_deinit - deallocate and free the transaction manager resources
> + * @xnm: pointer to the transaction manager
> + * @ctx: controlq context structure
> + *
> + * All Rx processing must be stopped beforehand.
> + */
> +void libeth_ctlq_xn_deinit(struct libeth_ctlq_xn_manager *xnm,
> + struct libeth_ctlq_ctx *ctx)
> +{
> + bool must_wait = false;
> + u32 i;
> +
> + /* Should be no new clear bits after this */
> + spin_lock(&xnm->free_xns_bm_lock);
> + xnm->shutdown = true;
nit: The line above is not correctly indented.
Flagged by Smatch.
> +
> + for_each_clear_bit(i, xnm->free_xns_bm, LIBETH_CTLQ_MAX_XN_ENTRIES) {
> + struct libeth_ctlq_xn *xn = &xnm->ring[i];
> +
> + spin_lock(&xn->xn_lock);
> +
> + if (xn->state == LIBETH_CTLQ_XN_WAITING ||
> + xn->state == LIBETH_CTLQ_XN_IDLE) {
> + complete(&xn->cmd_completion_event);
> + must_wait = true;
> + } else if (xn->state == LIBETH_CTLQ_XN_ASYNC) {
> + __libeth_ctlq_xn_push_free(xnm, xn);
> + }
> +
> + spin_unlock(&xn->xn_lock);
> + }
> +
> + spin_unlock(&xnm->free_xns_bm_lock);
> +
> + if (must_wait)
> + wait_for_completion(&xnm->can_destroy);
> +
> + libeth_ctlq_xn_deinit_dma(&ctx->mmio_info.pdev->dev, xnm,
> + LIBETH_CTLQ_MAX_XN_ENTRIES);
> + kfree(xnm);
> + libeth_ctlq_deinit(ctx);
> +}
> +EXPORT_SYMBOL_NS_GPL(libeth_ctlq_xn_deinit, "LIBETH_CP");
...
^ permalink raw reply [flat|nested] 36+ messages in thread
* [PATCH iwl-next 07/14] idpf: remove 'vport_params_reqd' field
2025-04-08 12:47 [PATCH iwl-next 00/14] Introduce iXD driver Larysa Zaremba
` (5 preceding siblings ...)
2025-04-08 12:47 ` [PATCH iwl-next 06/14] libeth: add bookkeeping support for control queue messages Larysa Zaremba
@ 2025-04-08 12:47 ` Larysa Zaremba
2025-04-08 12:47 ` [PATCH iwl-next 08/14] idpf: refactor idpf to use libeth controlq and Xn APIs Larysa Zaremba
` (7 subsequent siblings)
14 siblings, 0 replies; 36+ messages in thread
From: Larysa Zaremba @ 2025-04-08 12:47 UTC (permalink / raw)
To: intel-wired-lan, Tony Nguyen
Cc: David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Simon Horman, Jonathan Corbet, Przemek Kitszel, Jiri Pirko,
Mustafa Ismail, Tatyana Nikolova, Andrew Lunn, Alexander Lobakin,
Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Larysa Zaremba, Sridhar Samudrala,
Jacob Keller, Michal Swiatkowski, Mateusz Polchlopek, Wenjun Wu,
Ahmed Zaki, netdev, linux-doc, linux-kernel, Karlsson, Magnus,
Emil Tantilov, Madhu Chittim, Josh Hay, Milena Olech,
pavan.kumar.linga, Singhai, Anjali
From: Pavan Kumar Linga <pavan.kumar.linga@intel.com>
While sending a create vport message to the device
control plane, a create vport virtchnl message is
prepared with all the required info to initialize
the vport. This info is stored in the adapter struct
but never used thereafter. So, remove the said field.
Signed-off-by: Pavan Kumar Linga <pavan.kumar.linga@intel.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Larysa Zaremba <larysa.zaremba@intel.com>
---
drivers/net/ethernet/intel/idpf/idpf.h | 2 --
drivers/net/ethernet/intel/idpf/idpf_lib.c | 2 --
.../net/ethernet/intel/idpf/idpf_virtchnl.c | 31 ++++++-------------
3 files changed, 10 insertions(+), 25 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf.h b/drivers/net/ethernet/intel/idpf/idpf.h
index c42dc828a498..57c9d233f10e 100644
--- a/drivers/net/ethernet/intel/idpf/idpf.h
+++ b/drivers/net/ethernet/intel/idpf/idpf.h
@@ -501,7 +501,6 @@ struct idpf_vc_xn_manager;
* @avail_queues: Device given queue limits
* @vports: Array to store vports created by the driver
* @netdevs: Associated Vport netdevs
- * @vport_params_reqd: Vport params requested
* @vport_params_recvd: Vport params received
* @vport_ids: Array of device given vport identifiers
* @vport_config: Vport config parameters
@@ -555,7 +554,6 @@ struct idpf_adapter {
struct idpf_avail_queue_info avail_queues;
struct idpf_vport **vports;
struct net_device **netdevs;
- struct virtchnl2_create_vport **vport_params_reqd;
struct virtchnl2_create_vport **vport_params_recvd;
u32 *vport_ids;
diff --git a/drivers/net/ethernet/intel/idpf/idpf_lib.c b/drivers/net/ethernet/intel/idpf/idpf_lib.c
index 730a9c7a59f2..7ee8ad61d77d 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_lib.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_lib.c
@@ -964,8 +964,6 @@ static void idpf_vport_rel(struct idpf_vport *vport)
kfree(adapter->vport_params_recvd[idx]);
adapter->vport_params_recvd[idx] = NULL;
- kfree(adapter->vport_params_reqd[idx]);
- adapter->vport_params_reqd[idx] = NULL;
if (adapter->vport_config[idx]) {
kfree(adapter->vport_config[idx]->req_qs_chunks);
adapter->vport_config[idx]->req_qs_chunks = NULL;
diff --git a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
index 3d2413b8684f..bfa2a3dad3d6 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
@@ -1251,14 +1251,10 @@ int idpf_send_create_vport_msg(struct idpf_adapter *adapter,
ssize_t reply_sz;
buf_size = sizeof(struct virtchnl2_create_vport);
- if (!adapter->vport_params_reqd[idx]) {
- adapter->vport_params_reqd[idx] = kzalloc(buf_size,
- GFP_KERNEL);
- if (!adapter->vport_params_reqd[idx])
- return -ENOMEM;
- }
+ vport_msg = kzalloc(buf_size, GFP_KERNEL);
+ if (!vport_msg)
+ return -ENOMEM;
- vport_msg = adapter->vport_params_reqd[idx];
vport_msg->vport_type = cpu_to_le16(VIRTCHNL2_VPORT_TYPE_DEFAULT);
vport_msg->vport_index = cpu_to_le16(idx);
@@ -1275,8 +1271,7 @@ int idpf_send_create_vport_msg(struct idpf_adapter *adapter,
err = idpf_vport_calc_total_qs(adapter, idx, vport_msg, max_q);
if (err) {
dev_err(&adapter->pdev->dev, "Enough queues are not available");
-
- return err;
+ goto rel_buf;
}
if (!adapter->vport_params_recvd[idx]) {
@@ -1284,7 +1279,7 @@ int idpf_send_create_vport_msg(struct idpf_adapter *adapter,
GFP_KERNEL);
if (!adapter->vport_params_recvd[idx]) {
err = -ENOMEM;
- goto free_vport_params;
+ goto rel_buf;
}
}
@@ -1300,13 +1295,15 @@ int idpf_send_create_vport_msg(struct idpf_adapter *adapter,
goto free_vport_params;
}
+ kfree(vport_msg);
+
return 0;
free_vport_params:
kfree(adapter->vport_params_recvd[idx]);
adapter->vport_params_recvd[idx] = NULL;
- kfree(adapter->vport_params_reqd[idx]);
- adapter->vport_params_reqd[idx] = NULL;
+rel_buf:
+ kfree(vport_msg);
return err;
}
@@ -2875,8 +2872,6 @@ static void idpf_vport_params_buf_rel(struct idpf_adapter *adapter)
{
kfree(adapter->vport_params_recvd);
adapter->vport_params_recvd = NULL;
- kfree(adapter->vport_params_reqd);
- adapter->vport_params_reqd = NULL;
kfree(adapter->vport_ids);
adapter->vport_ids = NULL;
}
@@ -2891,17 +2886,11 @@ static int idpf_vport_params_buf_alloc(struct idpf_adapter *adapter)
{
u16 num_max_vports = idpf_get_max_vports(adapter);
- adapter->vport_params_reqd = kcalloc(num_max_vports,
- sizeof(*adapter->vport_params_reqd),
- GFP_KERNEL);
- if (!adapter->vport_params_reqd)
- return -ENOMEM;
-
adapter->vport_params_recvd = kcalloc(num_max_vports,
sizeof(*adapter->vport_params_recvd),
GFP_KERNEL);
if (!adapter->vport_params_recvd)
- goto err_mem;
+ return -ENOMEM;
adapter->vport_ids = kcalloc(num_max_vports, sizeof(u32), GFP_KERNEL);
if (!adapter->vport_ids)
--
2.47.0
^ permalink raw reply related [flat|nested] 36+ messages in thread* [PATCH iwl-next 08/14] idpf: refactor idpf to use libeth controlq and Xn APIs
2025-04-08 12:47 [PATCH iwl-next 00/14] Introduce iXD driver Larysa Zaremba
` (6 preceding siblings ...)
2025-04-08 12:47 ` [PATCH iwl-next 07/14] idpf: remove 'vport_params_reqd' field Larysa Zaremba
@ 2025-04-08 12:47 ` Larysa Zaremba
2025-04-21 14:25 ` Simon Horman
2025-04-08 12:47 ` [PATCH iwl-next 09/14] idpf: make mbx_task queueing and cancelling more consistent Larysa Zaremba
` (6 subsequent siblings)
14 siblings, 1 reply; 36+ messages in thread
From: Larysa Zaremba @ 2025-04-08 12:47 UTC (permalink / raw)
To: intel-wired-lan, Tony Nguyen
Cc: David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Simon Horman, Jonathan Corbet, Przemek Kitszel, Jiri Pirko,
Mustafa Ismail, Tatyana Nikolova, Andrew Lunn, Alexander Lobakin,
Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Larysa Zaremba, Sridhar Samudrala,
Jacob Keller, Michal Swiatkowski, Mateusz Polchlopek, Wenjun Wu,
Ahmed Zaki, netdev, linux-doc, linux-kernel, Karlsson, Magnus,
Emil Tantilov, Madhu Chittim, Josh Hay, Milena Olech,
pavan.kumar.linga, Singhai, Anjali, Michal Kubiak
From: Pavan Kumar Linga <pavan.kumar.linga@intel.com>
Support to initialize and configure controlq, Xn manager,
MMIO and reset APIs was introduced in libeth. As part of it,
most of the existing controlq structures are renamed and
modified. Use those APIs in idpf and make all the necessary changes.
Previously for the send and receive virtchnl messages, there
used to be a memcpy involved in controlq code to copy the buffer
info passed by the send function into the controlq specific
buffers. There was no restriction to use automatic memory
in that case. The new implementation in libeth removed copying
of the send buffer info and introduced DMA mapping of the
send buffer itself. To accommodate it, use dynamic memory for
the send buffers. In case of receive, idpf receives a page pool
buffer allocated by the libeth and care should be taken to
release it after use in the idpf.
The changes are fairly trivial and localized, with a notable exception
being the consolidation of idpf_vc_xn_shutdown and idpf_deinit_dflt_mbx
under the latter name. This has some additional consequences that are
addressed in the following patches.
Signed-off-by: Pavan Kumar Linga <pavan.kumar.linga@intel.com>
Reviewed-by: Michal Kubiak <michal.kubiak@intel.com>
Signed-off-by: Larysa Zaremba <larysa.zaremba@intel.com>
---
drivers/net/ethernet/intel/idpf/Kconfig | 1 +
drivers/net/ethernet/intel/idpf/Makefile | 2 -
drivers/net/ethernet/intel/idpf/idpf.h | 41 +-
.../net/ethernet/intel/idpf/idpf_controlq.c | 624 ------
.../net/ethernet/intel/idpf/idpf_controlq.h | 130 --
.../ethernet/intel/idpf/idpf_controlq_api.h | 174 --
.../ethernet/intel/idpf/idpf_controlq_setup.c | 171 --
drivers/net/ethernet/intel/idpf/idpf_dev.c | 95 +-
drivers/net/ethernet/intel/idpf/idpf_lib.c | 49 +-
drivers/net/ethernet/intel/idpf/idpf_main.c | 83 +-
drivers/net/ethernet/intel/idpf/idpf_mem.h | 20 -
drivers/net/ethernet/intel/idpf/idpf_txrx.h | 2 +-
drivers/net/ethernet/intel/idpf/idpf_vf_dev.c | 93 +-
.../net/ethernet/intel/idpf/idpf_virtchnl.c | 1689 ++++++-----------
.../net/ethernet/intel/idpf/idpf_virtchnl.h | 8 +-
15 files changed, 734 insertions(+), 2448 deletions(-)
delete mode 100644 drivers/net/ethernet/intel/idpf/idpf_controlq.c
delete mode 100644 drivers/net/ethernet/intel/idpf/idpf_controlq.h
delete mode 100644 drivers/net/ethernet/intel/idpf/idpf_controlq_api.h
delete mode 100644 drivers/net/ethernet/intel/idpf/idpf_controlq_setup.c
delete mode 100644 drivers/net/ethernet/intel/idpf/idpf_mem.h
diff --git a/drivers/net/ethernet/intel/idpf/Kconfig b/drivers/net/ethernet/intel/idpf/Kconfig
index 1addd663acad..226ed911c598 100644
--- a/drivers/net/ethernet/intel/idpf/Kconfig
+++ b/drivers/net/ethernet/intel/idpf/Kconfig
@@ -6,6 +6,7 @@ config IDPF
depends on PCI_MSI
select DIMLIB
select LIBETH
+ select LIBETH_CP
help
This driver supports Intel(R) Infrastructure Data Path Function
devices.
diff --git a/drivers/net/ethernet/intel/idpf/Makefile b/drivers/net/ethernet/intel/idpf/Makefile
index 2ce01a0b5898..3407cd20b498 100644
--- a/drivers/net/ethernet/intel/idpf/Makefile
+++ b/drivers/net/ethernet/intel/idpf/Makefile
@@ -6,8 +6,6 @@
obj-$(CONFIG_IDPF) += idpf.o
idpf-y := \
- idpf_controlq.o \
- idpf_controlq_setup.o \
idpf_dev.o \
idpf_ethtool.o \
idpf_lib.o \
diff --git a/drivers/net/ethernet/intel/idpf/idpf.h b/drivers/net/ethernet/intel/idpf/idpf.h
index 57c9d233f10e..f4197e7a2ef8 100644
--- a/drivers/net/ethernet/intel/idpf/idpf.h
+++ b/drivers/net/ethernet/intel/idpf/idpf.h
@@ -17,10 +17,10 @@ struct idpf_vport_max_q;
#include <linux/sctp.h>
#include <linux/ethtool_netlink.h>
#include <net/gro.h>
+#include <net/libeth/controlq.h>
#include <linux/intel/virtchnl2.h>
#include "idpf_txrx.h"
-#include "idpf_controlq.h"
#define GETMAXVAL(num_bits) GENMASK((num_bits) - 1, 0)
@@ -30,11 +30,10 @@ struct idpf_vport_max_q;
#define IDPF_NUM_FILTERS_PER_MSG 20
#define IDPF_NUM_DFLT_MBX_Q 2 /* includes both TX and RX */
#define IDPF_DFLT_MBX_Q_LEN 64
-#define IDPF_DFLT_MBX_ID -1
/* maximum number of times to try before resetting mailbox */
#define IDPF_MB_MAX_ERR 20
#define IDPF_NUM_CHUNKS_PER_MSG(struct_sz, chunk_sz) \
- ((IDPF_CTLQ_MAX_BUF_LEN - (struct_sz)) / (chunk_sz))
+ ((LIBETH_CTLQ_MAX_BUF_LEN - (struct_sz)) / (chunk_sz))
#define IDPF_MAX_WAIT 500
@@ -191,7 +190,8 @@ struct idpf_vport_max_q {
* @trigger_reset: Trigger a reset to occur
*/
struct idpf_reg_ops {
- void (*ctlq_reg_init)(struct idpf_ctlq_create_info *cq);
+ void (*ctlq_reg_init)(struct libeth_mmio_info *mmio,
+ struct libeth_ctlq_create_info *cctlq_info);
int (*intr_reg_init)(struct idpf_vport *vport);
void (*mb_intr_reg_init)(struct idpf_adapter *adapter);
void (*reset_reg_init)(struct idpf_adapter *adapter);
@@ -476,8 +476,6 @@ struct idpf_vport_config {
DECLARE_BITMAP(flags, IDPF_VPORT_CONFIG_FLAGS_NBITS);
};
-struct idpf_vc_xn_manager;
-
/**
* struct idpf_adapter - Device data struct generated on probe
* @pdev: PCI device struct given on probe
@@ -488,7 +486,10 @@ struct idpf_vc_xn_manager;
* @state: Init state machine
* @flags: See enum idpf_flags
* @reset_reg: See struct idpf_reset_reg
- * @hw: Device access data
+ * @ctlq_ctx: controlq context
+ * @asq: Send control queue info
+ * @arq: Receive control queue info
+ * @xn_init_params: Xn transaction manager parameters
* @num_req_msix: Requested number of MSIX vectors
* @num_avail_msix: Available number of MSIX vectors
* @num_msix_entries: Number of entries in MSIX table
@@ -540,7 +541,10 @@ struct idpf_adapter {
enum idpf_state state;
DECLARE_BITMAP(flags, IDPF_FLAGS_NBITS);
struct idpf_reset_reg reset_reg;
- struct idpf_hw hw;
+ struct libeth_ctlq_ctx ctlq_ctx;
+ struct libeth_ctlq_info *asq;
+ struct libeth_ctlq_info *arq;
+ struct libeth_ctlq_xn_init_params xn_init_params;
u16 num_req_msix;
u16 num_avail_msix;
u16 num_msix_entries;
@@ -573,7 +577,6 @@ struct idpf_adapter {
struct delayed_work stats_task;
struct workqueue_struct *stats_wq;
struct virtchnl2_get_capabilities caps;
- struct idpf_vc_xn_manager *vcxn_mngr;
struct idpf_dev_ops dev_ops;
int num_vfs;
@@ -705,19 +708,6 @@ static inline u8 idpf_get_min_tx_pkt_len(struct idpf_adapter *adapter)
return pkt_len ? pkt_len : IDPF_TX_MIN_PKT_LEN;
}
-/**
- * idpf_get_reg_addr - Get BAR0 register address
- * @adapter: private data struct
- * @reg_offset: register offset value
- *
- * Based on the register offset, return the actual BAR0 register address
- */
-static inline void __iomem *idpf_get_reg_addr(struct idpf_adapter *adapter,
- resource_size_t reg_offset)
-{
- return (void __iomem *)(adapter->hw.hw_addr + reg_offset);
-}
-
/**
* idpf_is_reset_detected - check if we were reset at some point
* @adapter: driver specific private structure
@@ -726,11 +716,12 @@ static inline void __iomem *idpf_get_reg_addr(struct idpf_adapter *adapter,
*/
static inline bool idpf_is_reset_detected(struct idpf_adapter *adapter)
{
- if (!adapter->hw.arq)
+ struct libeth_ctlq_info *arq = adapter->arq;
+
+ if (!arq)
return true;
- return !(readl(idpf_get_reg_addr(adapter, adapter->hw.arq->reg.len)) &
- adapter->hw.arq->reg.len_mask);
+ return !(readl(arq->reg.len) & arq->reg.len_mask);
}
/**
diff --git a/drivers/net/ethernet/intel/idpf/idpf_controlq.c b/drivers/net/ethernet/intel/idpf/idpf_controlq.c
deleted file mode 100644
index b28991dd1870..000000000000
--- a/drivers/net/ethernet/intel/idpf/idpf_controlq.c
+++ /dev/null
@@ -1,624 +0,0 @@
-// SPDX-License-Identifier: GPL-2.0-only
-/* Copyright (C) 2023 Intel Corporation */
-
-#include "idpf_controlq.h"
-
-/**
- * idpf_ctlq_setup_regs - initialize control queue registers
- * @cq: pointer to the specific control queue
- * @q_create_info: structs containing info for each queue to be initialized
- */
-static void idpf_ctlq_setup_regs(struct idpf_ctlq_info *cq,
- struct idpf_ctlq_create_info *q_create_info)
-{
- /* set control queue registers in our local struct */
- cq->reg.head = q_create_info->reg.head;
- cq->reg.tail = q_create_info->reg.tail;
- cq->reg.len = q_create_info->reg.len;
- cq->reg.bah = q_create_info->reg.bah;
- cq->reg.bal = q_create_info->reg.bal;
- cq->reg.len_mask = q_create_info->reg.len_mask;
- cq->reg.len_ena_mask = q_create_info->reg.len_ena_mask;
- cq->reg.head_mask = q_create_info->reg.head_mask;
-}
-
-/**
- * idpf_ctlq_init_regs - Initialize control queue registers
- * @hw: pointer to hw struct
- * @cq: pointer to the specific Control queue
- * @is_rxq: true if receive control queue, false otherwise
- *
- * Initialize registers. The caller is expected to have already initialized the
- * descriptor ring memory and buffer memory
- */
-static void idpf_ctlq_init_regs(struct idpf_hw *hw, struct idpf_ctlq_info *cq,
- bool is_rxq)
-{
- /* Update tail to post pre-allocated buffers for rx queues */
- if (is_rxq)
- wr32(hw, cq->reg.tail, (u32)(cq->ring_size - 1));
-
- /* For non-Mailbox control queues only TAIL need to be set */
- if (cq->q_id != -1)
- return;
-
- /* Clear Head for both send or receive */
- wr32(hw, cq->reg.head, 0);
-
- /* set starting point */
- wr32(hw, cq->reg.bal, lower_32_bits(cq->desc_ring.pa));
- wr32(hw, cq->reg.bah, upper_32_bits(cq->desc_ring.pa));
- wr32(hw, cq->reg.len, (cq->ring_size | cq->reg.len_ena_mask));
-}
-
-/**
- * idpf_ctlq_init_rxq_bufs - populate receive queue descriptors with buf
- * @cq: pointer to the specific Control queue
- *
- * Record the address of the receive queue DMA buffers in the descriptors.
- * The buffers must have been previously allocated.
- */
-static void idpf_ctlq_init_rxq_bufs(struct idpf_ctlq_info *cq)
-{
- int i;
-
- for (i = 0; i < cq->ring_size; i++) {
- struct idpf_ctlq_desc *desc = IDPF_CTLQ_DESC(cq, i);
- struct idpf_dma_mem *bi = cq->bi.rx_buff[i];
-
- /* No buffer to post to descriptor, continue */
- if (!bi)
- continue;
-
- desc->flags =
- cpu_to_le16(IDPF_CTLQ_FLAG_BUF | IDPF_CTLQ_FLAG_RD);
- desc->opcode = 0;
- desc->datalen = cpu_to_le16(bi->size);
- desc->ret_val = 0;
- desc->v_opcode_dtype = 0;
- desc->v_retval = 0;
- desc->params.indirect.addr_high =
- cpu_to_le32(upper_32_bits(bi->pa));
- desc->params.indirect.addr_low =
- cpu_to_le32(lower_32_bits(bi->pa));
- desc->params.indirect.param0 = 0;
- desc->params.indirect.sw_cookie = 0;
- desc->params.indirect.v_flags = 0;
- }
-}
-
-/**
- * idpf_ctlq_shutdown - shutdown the CQ
- * @hw: pointer to hw struct
- * @cq: pointer to the specific Control queue
- *
- * The main shutdown routine for any controq queue
- */
-static void idpf_ctlq_shutdown(struct idpf_hw *hw, struct idpf_ctlq_info *cq)
-{
- mutex_lock(&cq->cq_lock);
-
- /* free ring buffers and the ring itself */
- idpf_ctlq_dealloc_ring_res(hw, cq);
-
- /* Set ring_size to 0 to indicate uninitialized queue */
- cq->ring_size = 0;
-
- mutex_unlock(&cq->cq_lock);
- mutex_destroy(&cq->cq_lock);
-}
-
-/**
- * idpf_ctlq_add - add one control queue
- * @hw: pointer to hardware struct
- * @qinfo: info for queue to be created
- * @cq_out: (output) double pointer to control queue to be created
- *
- * Allocate and initialize a control queue and add it to the control queue list.
- * The cq parameter will be allocated/initialized and passed back to the caller
- * if no errors occur.
- *
- * Note: idpf_ctlq_init must be called prior to any calls to idpf_ctlq_add
- */
-int idpf_ctlq_add(struct idpf_hw *hw,
- struct idpf_ctlq_create_info *qinfo,
- struct idpf_ctlq_info **cq_out)
-{
- struct idpf_ctlq_info *cq;
- bool is_rxq = false;
- int err;
-
- cq = kzalloc(sizeof(*cq), GFP_KERNEL);
- if (!cq)
- return -ENOMEM;
-
- cq->cq_type = qinfo->type;
- cq->q_id = qinfo->id;
- cq->buf_size = qinfo->buf_size;
- cq->ring_size = qinfo->len;
-
- cq->next_to_use = 0;
- cq->next_to_clean = 0;
- cq->next_to_post = cq->ring_size - 1;
-
- switch (qinfo->type) {
- case IDPF_CTLQ_TYPE_MAILBOX_RX:
- is_rxq = true;
- fallthrough;
- case IDPF_CTLQ_TYPE_MAILBOX_TX:
- err = idpf_ctlq_alloc_ring_res(hw, cq);
- break;
- default:
- err = -EBADR;
- break;
- }
-
- if (err)
- goto init_free_q;
-
- if (is_rxq) {
- idpf_ctlq_init_rxq_bufs(cq);
- } else {
- /* Allocate the array of msg pointers for TX queues */
- cq->bi.tx_msg = kcalloc(qinfo->len,
- sizeof(struct idpf_ctlq_msg *),
- GFP_KERNEL);
- if (!cq->bi.tx_msg) {
- err = -ENOMEM;
- goto init_dealloc_q_mem;
- }
- }
-
- idpf_ctlq_setup_regs(cq, qinfo);
-
- idpf_ctlq_init_regs(hw, cq, is_rxq);
-
- mutex_init(&cq->cq_lock);
-
- list_add(&cq->cq_list, &hw->cq_list_head);
-
- *cq_out = cq;
-
- return 0;
-
-init_dealloc_q_mem:
- /* free ring buffers and the ring itself */
- idpf_ctlq_dealloc_ring_res(hw, cq);
-init_free_q:
- kfree(cq);
-
- return err;
-}
-
-/**
- * idpf_ctlq_remove - deallocate and remove specified control queue
- * @hw: pointer to hardware struct
- * @cq: pointer to control queue to be removed
- */
-void idpf_ctlq_remove(struct idpf_hw *hw,
- struct idpf_ctlq_info *cq)
-{
- list_del(&cq->cq_list);
- idpf_ctlq_shutdown(hw, cq);
- kfree(cq);
-}
-
-/**
- * idpf_ctlq_init - main initialization routine for all control queues
- * @hw: pointer to hardware struct
- * @num_q: number of queues to initialize
- * @q_info: array of structs containing info for each queue to be initialized
- *
- * This initializes any number and any type of control queues. This is an all
- * or nothing routine; if one fails, all previously allocated queues will be
- * destroyed. This must be called prior to using the individual add/remove
- * APIs.
- */
-int idpf_ctlq_init(struct idpf_hw *hw, u8 num_q,
- struct idpf_ctlq_create_info *q_info)
-{
- struct idpf_ctlq_info *cq, *tmp;
- int err;
- int i;
-
- INIT_LIST_HEAD(&hw->cq_list_head);
-
- for (i = 0; i < num_q; i++) {
- struct idpf_ctlq_create_info *qinfo = q_info + i;
-
- err = idpf_ctlq_add(hw, qinfo, &cq);
- if (err)
- goto init_destroy_qs;
- }
-
- return 0;
-
-init_destroy_qs:
- list_for_each_entry_safe(cq, tmp, &hw->cq_list_head, cq_list)
- idpf_ctlq_remove(hw, cq);
-
- return err;
-}
-
-/**
- * idpf_ctlq_deinit - destroy all control queues
- * @hw: pointer to hw struct
- */
-void idpf_ctlq_deinit(struct idpf_hw *hw)
-{
- struct idpf_ctlq_info *cq, *tmp;
-
- list_for_each_entry_safe(cq, tmp, &hw->cq_list_head, cq_list)
- idpf_ctlq_remove(hw, cq);
-}
-
-/**
- * idpf_ctlq_send - send command to Control Queue (CTQ)
- * @hw: pointer to hw struct
- * @cq: handle to control queue struct to send on
- * @num_q_msg: number of messages to send on control queue
- * @q_msg: pointer to array of queue messages to be sent
- *
- * The caller is expected to allocate DMAable buffers and pass them to the
- * send routine via the q_msg struct / control queue specific data struct.
- * The control queue will hold a reference to each send message until
- * the completion for that message has been cleaned.
- */
-int idpf_ctlq_send(struct idpf_hw *hw, struct idpf_ctlq_info *cq,
- u16 num_q_msg, struct idpf_ctlq_msg q_msg[])
-{
- struct idpf_ctlq_desc *desc;
- int num_desc_avail;
- int err = 0;
- int i;
-
- mutex_lock(&cq->cq_lock);
-
- /* Ensure there are enough descriptors to send all messages */
- num_desc_avail = IDPF_CTLQ_DESC_UNUSED(cq);
- if (num_desc_avail == 0 || num_desc_avail < num_q_msg) {
- err = -ENOSPC;
- goto err_unlock;
- }
-
- for (i = 0; i < num_q_msg; i++) {
- struct idpf_ctlq_msg *msg = &q_msg[i];
-
- desc = IDPF_CTLQ_DESC(cq, cq->next_to_use);
-
- desc->opcode = cpu_to_le16(msg->opcode);
- desc->pfid_vfid = cpu_to_le16(msg->func_id);
-
- desc->v_opcode_dtype = cpu_to_le32(msg->cookie.mbx.chnl_opcode);
- desc->v_retval = cpu_to_le32(msg->cookie.mbx.chnl_retval);
-
- desc->flags = cpu_to_le16((msg->host_id & IDPF_HOST_ID_MASK) <<
- IDPF_CTLQ_FLAG_HOST_ID_S);
- if (msg->data_len) {
- struct idpf_dma_mem *buff = msg->ctx.indirect.payload;
-
- desc->datalen |= cpu_to_le16(msg->data_len);
- desc->flags |= cpu_to_le16(IDPF_CTLQ_FLAG_BUF);
- desc->flags |= cpu_to_le16(IDPF_CTLQ_FLAG_RD);
-
- /* Update the address values in the desc with the pa
- * value for respective buffer
- */
- desc->params.indirect.addr_high =
- cpu_to_le32(upper_32_bits(buff->pa));
- desc->params.indirect.addr_low =
- cpu_to_le32(lower_32_bits(buff->pa));
-
- memcpy(&desc->params, msg->ctx.indirect.context,
- IDPF_INDIRECT_CTX_SIZE);
- } else {
- memcpy(&desc->params, msg->ctx.direct,
- IDPF_DIRECT_CTX_SIZE);
- }
-
- /* Store buffer info */
- cq->bi.tx_msg[cq->next_to_use] = msg;
-
- (cq->next_to_use)++;
- if (cq->next_to_use == cq->ring_size)
- cq->next_to_use = 0;
- }
-
- /* Force memory write to complete before letting hardware
- * know that there are new descriptors to fetch.
- */
- dma_wmb();
-
- wr32(hw, cq->reg.tail, cq->next_to_use);
-
-err_unlock:
- mutex_unlock(&cq->cq_lock);
-
- return err;
-}
-
-/**
- * idpf_ctlq_clean_sq - reclaim send descriptors on HW write back for the
- * requested queue
- * @cq: pointer to the specific Control queue
- * @clean_count: (input|output) number of descriptors to clean as input, and
- * number of descriptors actually cleaned as output
- * @msg_status: (output) pointer to msg pointer array to be populated; needs
- * to be allocated by caller
- *
- * Returns an array of message pointers associated with the cleaned
- * descriptors. The pointers are to the original ctlq_msgs sent on the cleaned
- * descriptors. The status will be returned for each; any messages that failed
- * to send will have a non-zero status. The caller is expected to free original
- * ctlq_msgs and free or reuse the DMA buffers.
- */
-int idpf_ctlq_clean_sq(struct idpf_ctlq_info *cq, u16 *clean_count,
- struct idpf_ctlq_msg *msg_status[])
-{
- struct idpf_ctlq_desc *desc;
- u16 i, num_to_clean;
- u16 ntc, desc_err;
-
- if (*clean_count == 0)
- return 0;
- if (*clean_count > cq->ring_size)
- return -EBADR;
-
- mutex_lock(&cq->cq_lock);
-
- ntc = cq->next_to_clean;
-
- num_to_clean = *clean_count;
-
- for (i = 0; i < num_to_clean; i++) {
- /* Fetch next descriptor and check if marked as done */
- desc = IDPF_CTLQ_DESC(cq, ntc);
- if (!(le16_to_cpu(desc->flags) & IDPF_CTLQ_FLAG_DD))
- break;
-
- /* Ensure no other fields are read until DD flag is checked */
- dma_rmb();
-
- /* strip off FW internal code */
- desc_err = le16_to_cpu(desc->ret_val) & 0xff;
-
- msg_status[i] = cq->bi.tx_msg[ntc];
- msg_status[i]->status = desc_err;
-
- cq->bi.tx_msg[ntc] = NULL;
-
- /* Zero out any stale data */
- memset(desc, 0, sizeof(*desc));
-
- ntc++;
- if (ntc == cq->ring_size)
- ntc = 0;
- }
-
- cq->next_to_clean = ntc;
-
- mutex_unlock(&cq->cq_lock);
-
- /* Return number of descriptors actually cleaned */
- *clean_count = i;
-
- return 0;
-}
-
-/**
- * idpf_ctlq_post_rx_buffs - post buffers to descriptor ring
- * @hw: pointer to hw struct
- * @cq: pointer to control queue handle
- * @buff_count: (input|output) input is number of buffers caller is trying to
- * return; output is number of buffers that were not posted
- * @buffs: array of pointers to dma mem structs to be given to hardware
- *
- * Caller uses this function to return DMA buffers to the descriptor ring after
- * consuming them; buff_count will be the number of buffers.
- *
- * Note: this function needs to be called after a receive call even
- * if there are no DMA buffers to be returned, i.e. buff_count = 0,
- * buffs = NULL to support direct commands
- */
-int idpf_ctlq_post_rx_buffs(struct idpf_hw *hw, struct idpf_ctlq_info *cq,
- u16 *buff_count, struct idpf_dma_mem **buffs)
-{
- struct idpf_ctlq_desc *desc;
- u16 ntp = cq->next_to_post;
- bool buffs_avail = false;
- u16 tbp = ntp + 1;
- int i = 0;
-
- if (*buff_count > cq->ring_size)
- return -EBADR;
-
- if (*buff_count > 0)
- buffs_avail = true;
-
- mutex_lock(&cq->cq_lock);
-
- if (tbp >= cq->ring_size)
- tbp = 0;
-
- if (tbp == cq->next_to_clean)
- /* Nothing to do */
- goto post_buffs_out;
-
- /* Post buffers for as many as provided or up until the last one used */
- while (ntp != cq->next_to_clean) {
- desc = IDPF_CTLQ_DESC(cq, ntp);
-
- if (cq->bi.rx_buff[ntp])
- goto fill_desc;
- if (!buffs_avail) {
- /* If the caller hasn't given us any buffers or
- * there are none left, search the ring itself
- * for an available buffer to move to this
- * entry starting at the next entry in the ring
- */
- tbp = ntp + 1;
-
- /* Wrap ring if necessary */
- if (tbp >= cq->ring_size)
- tbp = 0;
-
- while (tbp != cq->next_to_clean) {
- if (cq->bi.rx_buff[tbp]) {
- cq->bi.rx_buff[ntp] =
- cq->bi.rx_buff[tbp];
- cq->bi.rx_buff[tbp] = NULL;
-
- /* Found a buffer, no need to
- * search anymore
- */
- break;
- }
-
- /* Wrap ring if necessary */
- tbp++;
- if (tbp >= cq->ring_size)
- tbp = 0;
- }
-
- if (tbp == cq->next_to_clean)
- goto post_buffs_out;
- } else {
- /* Give back pointer to DMA buffer */
- cq->bi.rx_buff[ntp] = buffs[i];
- i++;
-
- if (i >= *buff_count)
- buffs_avail = false;
- }
-
-fill_desc:
- desc->flags =
- cpu_to_le16(IDPF_CTLQ_FLAG_BUF | IDPF_CTLQ_FLAG_RD);
-
- /* Post buffers to descriptor */
- desc->datalen = cpu_to_le16(cq->bi.rx_buff[ntp]->size);
- desc->params.indirect.addr_high =
- cpu_to_le32(upper_32_bits(cq->bi.rx_buff[ntp]->pa));
- desc->params.indirect.addr_low =
- cpu_to_le32(lower_32_bits(cq->bi.rx_buff[ntp]->pa));
-
- ntp++;
- if (ntp == cq->ring_size)
- ntp = 0;
- }
-
-post_buffs_out:
- /* Only update tail if buffers were actually posted */
- if (cq->next_to_post != ntp) {
- if (ntp)
- /* Update next_to_post to ntp - 1 since current ntp
- * will not have a buffer
- */
- cq->next_to_post = ntp - 1;
- else
- /* Wrap to end of end ring since current ntp is 0 */
- cq->next_to_post = cq->ring_size - 1;
-
- dma_wmb();
-
- wr32(hw, cq->reg.tail, cq->next_to_post);
- }
-
- mutex_unlock(&cq->cq_lock);
-
- /* return the number of buffers that were not posted */
- *buff_count = *buff_count - i;
-
- return 0;
-}
-
-/**
- * idpf_ctlq_recv - receive control queue message call back
- * @cq: pointer to control queue handle to receive on
- * @num_q_msg: (input|output) input number of messages that should be received;
- * output number of messages actually received
- * @q_msg: (output) array of received control queue messages on this q;
- * needs to be pre-allocated by caller for as many messages as requested
- *
- * Called by interrupt handler or polling mechanism. Caller is expected
- * to free buffers
- */
-int idpf_ctlq_recv(struct idpf_ctlq_info *cq, u16 *num_q_msg,
- struct idpf_ctlq_msg *q_msg)
-{
- u16 num_to_clean, ntc, flags;
- struct idpf_ctlq_desc *desc;
- int err = 0;
- u16 i;
-
- /* take the lock before we start messing with the ring */
- mutex_lock(&cq->cq_lock);
-
- ntc = cq->next_to_clean;
-
- num_to_clean = *num_q_msg;
-
- for (i = 0; i < num_to_clean; i++) {
- /* Fetch next descriptor and check if marked as done */
- desc = IDPF_CTLQ_DESC(cq, ntc);
- flags = le16_to_cpu(desc->flags);
-
- if (!(flags & IDPF_CTLQ_FLAG_DD))
- break;
-
- /* Ensure no other fields are read until DD flag is checked */
- dma_rmb();
-
- q_msg[i].vmvf_type = (flags &
- (IDPF_CTLQ_FLAG_FTYPE_VM |
- IDPF_CTLQ_FLAG_FTYPE_PF)) >>
- IDPF_CTLQ_FLAG_FTYPE_S;
-
- if (flags & IDPF_CTLQ_FLAG_ERR)
- err = -EBADMSG;
-
- q_msg[i].cookie.mbx.chnl_opcode =
- le32_to_cpu(desc->v_opcode_dtype);
- q_msg[i].cookie.mbx.chnl_retval =
- le32_to_cpu(desc->v_retval);
-
- q_msg[i].opcode = le16_to_cpu(desc->opcode);
- q_msg[i].data_len = le16_to_cpu(desc->datalen);
- q_msg[i].status = le16_to_cpu(desc->ret_val);
-
- if (desc->datalen) {
- memcpy(q_msg[i].ctx.indirect.context,
- &desc->params.indirect, IDPF_INDIRECT_CTX_SIZE);
-
- /* Assign pointer to dma buffer to ctlq_msg array
- * to be given to upper layer
- */
- q_msg[i].ctx.indirect.payload = cq->bi.rx_buff[ntc];
-
- /* Zero out pointer to DMA buffer info;
- * will be repopulated by post buffers API
- */
- cq->bi.rx_buff[ntc] = NULL;
- } else {
- memcpy(q_msg[i].ctx.direct, desc->params.raw,
- IDPF_DIRECT_CTX_SIZE);
- }
-
- /* Zero out stale data in descriptor */
- memset(desc, 0, sizeof(struct idpf_ctlq_desc));
-
- ntc++;
- if (ntc == cq->ring_size)
- ntc = 0;
- }
-
- cq->next_to_clean = ntc;
-
- mutex_unlock(&cq->cq_lock);
-
- *num_q_msg = i;
- if (*num_q_msg == 0)
- err = -ENOMSG;
-
- return err;
-}
diff --git a/drivers/net/ethernet/intel/idpf/idpf_controlq.h b/drivers/net/ethernet/intel/idpf/idpf_controlq.h
deleted file mode 100644
index c1aba09e9856..000000000000
--- a/drivers/net/ethernet/intel/idpf/idpf_controlq.h
+++ /dev/null
@@ -1,130 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0-only */
-/* Copyright (C) 2023 Intel Corporation */
-
-#ifndef _IDPF_CONTROLQ_H_
-#define _IDPF_CONTROLQ_H_
-
-#include <linux/slab.h>
-
-#include "idpf_controlq_api.h"
-
-/* Maximum buffer length for all control queue types */
-#define IDPF_CTLQ_MAX_BUF_LEN 4096
-
-#define IDPF_CTLQ_DESC(R, i) \
- (&(((struct idpf_ctlq_desc *)((R)->desc_ring.va))[i]))
-
-#define IDPF_CTLQ_DESC_UNUSED(R) \
- ((u16)((((R)->next_to_clean > (R)->next_to_use) ? 0 : (R)->ring_size) + \
- (R)->next_to_clean - (R)->next_to_use - 1))
-
-/* Control Queue default settings */
-#define IDPF_CTRL_SQ_CMD_TIMEOUT 250 /* msecs */
-
-struct idpf_ctlq_desc {
- /* Control queue descriptor flags */
- __le16 flags;
- /* Control queue message opcode */
- __le16 opcode;
- __le16 datalen; /* 0 for direct commands */
- union {
- __le16 ret_val;
- __le16 pfid_vfid;
-#define IDPF_CTLQ_DESC_VF_ID_S 0
-#define IDPF_CTLQ_DESC_VF_ID_M (0x7FF << IDPF_CTLQ_DESC_VF_ID_S)
-#define IDPF_CTLQ_DESC_PF_ID_S 11
-#define IDPF_CTLQ_DESC_PF_ID_M (0x1F << IDPF_CTLQ_DESC_PF_ID_S)
- };
-
- /* Virtchnl message opcode and virtchnl descriptor type
- * v_opcode=[27:0], v_dtype=[31:28]
- */
- __le32 v_opcode_dtype;
- /* Virtchnl return value */
- __le32 v_retval;
- union {
- struct {
- __le32 param0;
- __le32 param1;
- __le32 param2;
- __le32 param3;
- } direct;
- struct {
- __le32 param0;
- __le16 sw_cookie;
- /* Virtchnl flags */
- __le16 v_flags;
- __le32 addr_high;
- __le32 addr_low;
- } indirect;
- u8 raw[16];
- } params;
-};
-
-/* Flags sub-structure
- * |0 |1 |2 |3 |4 |5 |6 |7 |8 |9 |10 |11 |12 |13 |14 |15 |
- * |DD |CMP|ERR| * RSV * |FTYPE | *RSV* |RD |VFC|BUF| HOST_ID |
- */
-/* command flags and offsets */
-#define IDPF_CTLQ_FLAG_DD_S 0
-#define IDPF_CTLQ_FLAG_CMP_S 1
-#define IDPF_CTLQ_FLAG_ERR_S 2
-#define IDPF_CTLQ_FLAG_FTYPE_S 6
-#define IDPF_CTLQ_FLAG_RD_S 10
-#define IDPF_CTLQ_FLAG_VFC_S 11
-#define IDPF_CTLQ_FLAG_BUF_S 12
-#define IDPF_CTLQ_FLAG_HOST_ID_S 13
-
-#define IDPF_CTLQ_FLAG_DD BIT(IDPF_CTLQ_FLAG_DD_S) /* 0x1 */
-#define IDPF_CTLQ_FLAG_CMP BIT(IDPF_CTLQ_FLAG_CMP_S) /* 0x2 */
-#define IDPF_CTLQ_FLAG_ERR BIT(IDPF_CTLQ_FLAG_ERR_S) /* 0x4 */
-#define IDPF_CTLQ_FLAG_FTYPE_VM BIT(IDPF_CTLQ_FLAG_FTYPE_S) /* 0x40 */
-#define IDPF_CTLQ_FLAG_FTYPE_PF BIT(IDPF_CTLQ_FLAG_FTYPE_S + 1) /* 0x80 */
-#define IDPF_CTLQ_FLAG_RD BIT(IDPF_CTLQ_FLAG_RD_S) /* 0x400 */
-#define IDPF_CTLQ_FLAG_VFC BIT(IDPF_CTLQ_FLAG_VFC_S) /* 0x800 */
-#define IDPF_CTLQ_FLAG_BUF BIT(IDPF_CTLQ_FLAG_BUF_S) /* 0x1000 */
-
-/* Host ID is a special field that has 3b and not a 1b flag */
-#define IDPF_CTLQ_FLAG_HOST_ID_M MAKE_MASK(0x7000UL, IDPF_CTLQ_FLAG_HOST_ID_S)
-
-struct idpf_mbxq_desc {
- u8 pad[8]; /* CTLQ flags/opcode/len/retval fields */
- u32 chnl_opcode; /* avoid confusion with desc->opcode */
- u32 chnl_retval; /* ditto for desc->retval */
- u32 pf_vf_id; /* used by CP when sending to PF */
-};
-
-/* Define the driver hardware struct to replace other control structs as needed
- * Align to ctlq_hw_info
- */
-struct idpf_hw {
- void __iomem *hw_addr;
- resource_size_t hw_addr_len;
-
- struct idpf_adapter *back;
-
- /* control queue - send and receive */
- struct idpf_ctlq_info *asq;
- struct idpf_ctlq_info *arq;
-
- /* pci info */
- u16 device_id;
- u16 vendor_id;
- u16 subsystem_device_id;
- u16 subsystem_vendor_id;
- u8 revision_id;
- bool adapter_stopped;
-
- struct list_head cq_list_head;
-};
-
-int idpf_ctlq_alloc_ring_res(struct idpf_hw *hw,
- struct idpf_ctlq_info *cq);
-
-void idpf_ctlq_dealloc_ring_res(struct idpf_hw *hw, struct idpf_ctlq_info *cq);
-
-/* prototype for functions used for dynamic memory allocation */
-void *idpf_alloc_dma_mem(struct idpf_hw *hw, struct idpf_dma_mem *mem,
- u64 size);
-void idpf_free_dma_mem(struct idpf_hw *hw, struct idpf_dma_mem *mem);
-#endif /* _IDPF_CONTROLQ_H_ */
diff --git a/drivers/net/ethernet/intel/idpf/idpf_controlq_api.h b/drivers/net/ethernet/intel/idpf/idpf_controlq_api.h
deleted file mode 100644
index e8e046ef2f0d..000000000000
--- a/drivers/net/ethernet/intel/idpf/idpf_controlq_api.h
+++ /dev/null
@@ -1,174 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0-only */
-/* Copyright (C) 2023 Intel Corporation */
-
-#ifndef _IDPF_CONTROLQ_API_H_
-#define _IDPF_CONTROLQ_API_H_
-
-#include "idpf_mem.h"
-
-struct idpf_hw;
-
-/* Used for queue init, response and events */
-enum idpf_ctlq_type {
- IDPF_CTLQ_TYPE_MAILBOX_TX = 0,
- IDPF_CTLQ_TYPE_MAILBOX_RX = 1,
- IDPF_CTLQ_TYPE_CONFIG_TX = 2,
- IDPF_CTLQ_TYPE_CONFIG_RX = 3,
- IDPF_CTLQ_TYPE_EVENT_RX = 4,
- IDPF_CTLQ_TYPE_RDMA_TX = 5,
- IDPF_CTLQ_TYPE_RDMA_RX = 6,
- IDPF_CTLQ_TYPE_RDMA_COMPL = 7
-};
-
-/* Generic Control Queue Structures */
-struct idpf_ctlq_reg {
- /* used for queue tracking */
- u32 head;
- u32 tail;
- /* Below applies only to default mb (if present) */
- u32 len;
- u32 bah;
- u32 bal;
- u32 len_mask;
- u32 len_ena_mask;
- u32 head_mask;
-};
-
-/* Generic queue msg structure */
-struct idpf_ctlq_msg {
- u8 vmvf_type; /* represents the source of the message on recv */
-#define IDPF_VMVF_TYPE_VF 0
-#define IDPF_VMVF_TYPE_VM 1
-#define IDPF_VMVF_TYPE_PF 2
- u8 host_id;
- /* 3b field used only when sending a message to CP - to be used in
- * combination with target func_id to route the message
- */
-#define IDPF_HOST_ID_MASK 0x7
-
- u16 opcode;
- u16 data_len; /* data_len = 0 when no payload is attached */
- union {
- u16 func_id; /* when sending a message */
- u16 status; /* when receiving a message */
- };
- union {
- struct {
- u32 chnl_opcode;
- u32 chnl_retval;
- } mbx;
- } cookie;
- union {
-#define IDPF_DIRECT_CTX_SIZE 16
-#define IDPF_INDIRECT_CTX_SIZE 8
- /* 16 bytes of context can be provided or 8 bytes of context
- * plus the address of a DMA buffer
- */
- u8 direct[IDPF_DIRECT_CTX_SIZE];
- struct {
- u8 context[IDPF_INDIRECT_CTX_SIZE];
- struct idpf_dma_mem *payload;
- } indirect;
- struct {
- u32 rsvd;
- u16 data;
- u16 flags;
- } sw_cookie;
- } ctx;
-};
-
-/* Generic queue info structures */
-/* MB, CONFIG and EVENT q do not have extended info */
-struct idpf_ctlq_create_info {
- enum idpf_ctlq_type type;
- int id; /* absolute queue offset passed as input
- * -1 for default mailbox if present
- */
- u16 len; /* Queue length passed as input */
- u16 buf_size; /* buffer size passed as input */
- u64 base_address; /* output, HPA of the Queue start */
- struct idpf_ctlq_reg reg; /* registers accessed by ctlqs */
-
- int ext_info_size;
- void *ext_info; /* Specific to q type */
-};
-
-/* Control Queue information */
-struct idpf_ctlq_info {
- struct list_head cq_list;
-
- enum idpf_ctlq_type cq_type;
- int q_id;
- struct mutex cq_lock; /* control queue lock */
- /* used for interrupt processing */
- u16 next_to_use;
- u16 next_to_clean;
- u16 next_to_post; /* starting descriptor to post buffers
- * to after recev
- */
-
- struct idpf_dma_mem desc_ring; /* descriptor ring memory
- * idpf_dma_mem is defined in OSdep.h
- */
- union {
- struct idpf_dma_mem **rx_buff;
- struct idpf_ctlq_msg **tx_msg;
- } bi;
-
- u16 buf_size; /* queue buffer size */
- u16 ring_size; /* Number of descriptors */
- struct idpf_ctlq_reg reg; /* registers accessed by ctlqs */
-};
-
-/**
- * enum idpf_mbx_opc - PF/VF mailbox commands
- * @idpf_mbq_opc_send_msg_to_cp: used by PF or VF to send a message to its CP
- */
-enum idpf_mbx_opc {
- idpf_mbq_opc_send_msg_to_cp = 0x0801,
-};
-
-/* API supported for control queue management */
-/* Will init all required q including default mb. "q_info" is an array of
- * create_info structs equal to the number of control queues to be created.
- */
-int idpf_ctlq_init(struct idpf_hw *hw, u8 num_q,
- struct idpf_ctlq_create_info *q_info);
-
-/* Allocate and initialize a single control queue, which will be added to the
- * control queue list; returns a handle to the created control queue
- */
-int idpf_ctlq_add(struct idpf_hw *hw,
- struct idpf_ctlq_create_info *qinfo,
- struct idpf_ctlq_info **cq);
-
-/* Deinitialize and deallocate a single control queue */
-void idpf_ctlq_remove(struct idpf_hw *hw,
- struct idpf_ctlq_info *cq);
-
-/* Sends messages to HW and will also free the buffer*/
-int idpf_ctlq_send(struct idpf_hw *hw,
- struct idpf_ctlq_info *cq,
- u16 num_q_msg,
- struct idpf_ctlq_msg q_msg[]);
-
-/* Receives messages and called by interrupt handler/polling
- * initiated by app/process. Also caller is supposed to free the buffers
- */
-int idpf_ctlq_recv(struct idpf_ctlq_info *cq, u16 *num_q_msg,
- struct idpf_ctlq_msg *q_msg);
-
-/* Reclaims send descriptors on HW write back */
-int idpf_ctlq_clean_sq(struct idpf_ctlq_info *cq, u16 *clean_count,
- struct idpf_ctlq_msg *msg_status[]);
-
-/* Indicate RX buffers are done being processed */
-int idpf_ctlq_post_rx_buffs(struct idpf_hw *hw,
- struct idpf_ctlq_info *cq,
- u16 *buff_count,
- struct idpf_dma_mem **buffs);
-
-/* Will destroy all q including the default mb */
-void idpf_ctlq_deinit(struct idpf_hw *hw);
-
-#endif /* _IDPF_CONTROLQ_API_H_ */
diff --git a/drivers/net/ethernet/intel/idpf/idpf_controlq_setup.c b/drivers/net/ethernet/intel/idpf/idpf_controlq_setup.c
deleted file mode 100644
index a942a6385d06..000000000000
--- a/drivers/net/ethernet/intel/idpf/idpf_controlq_setup.c
+++ /dev/null
@@ -1,171 +0,0 @@
-// SPDX-License-Identifier: GPL-2.0-only
-/* Copyright (C) 2023 Intel Corporation */
-
-#include "idpf_controlq.h"
-
-/**
- * idpf_ctlq_alloc_desc_ring - Allocate Control Queue (CQ) rings
- * @hw: pointer to hw struct
- * @cq: pointer to the specific Control queue
- */
-static int idpf_ctlq_alloc_desc_ring(struct idpf_hw *hw,
- struct idpf_ctlq_info *cq)
-{
- size_t size = cq->ring_size * sizeof(struct idpf_ctlq_desc);
-
- cq->desc_ring.va = idpf_alloc_dma_mem(hw, &cq->desc_ring, size);
- if (!cq->desc_ring.va)
- return -ENOMEM;
-
- return 0;
-}
-
-/**
- * idpf_ctlq_alloc_bufs - Allocate Control Queue (CQ) buffers
- * @hw: pointer to hw struct
- * @cq: pointer to the specific Control queue
- *
- * Allocate the buffer head for all control queues, and if it's a receive
- * queue, allocate DMA buffers
- */
-static int idpf_ctlq_alloc_bufs(struct idpf_hw *hw,
- struct idpf_ctlq_info *cq)
-{
- int i;
-
- /* Do not allocate DMA buffers for transmit queues */
- if (cq->cq_type == IDPF_CTLQ_TYPE_MAILBOX_TX)
- return 0;
-
- /* We'll be allocating the buffer info memory first, then we can
- * allocate the mapped buffers for the event processing
- */
- cq->bi.rx_buff = kcalloc(cq->ring_size, sizeof(struct idpf_dma_mem *),
- GFP_KERNEL);
- if (!cq->bi.rx_buff)
- return -ENOMEM;
-
- /* allocate the mapped buffers (except for the last one) */
- for (i = 0; i < cq->ring_size - 1; i++) {
- struct idpf_dma_mem *bi;
- int num = 1; /* number of idpf_dma_mem to be allocated */
-
- cq->bi.rx_buff[i] = kcalloc(num, sizeof(struct idpf_dma_mem),
- GFP_KERNEL);
- if (!cq->bi.rx_buff[i])
- goto unwind_alloc_cq_bufs;
-
- bi = cq->bi.rx_buff[i];
-
- bi->va = idpf_alloc_dma_mem(hw, bi, cq->buf_size);
- if (!bi->va) {
- /* unwind will not free the failed entry */
- kfree(cq->bi.rx_buff[i]);
- goto unwind_alloc_cq_bufs;
- }
- }
-
- return 0;
-
-unwind_alloc_cq_bufs:
- /* don't try to free the one that failed... */
- i--;
- for (; i >= 0; i--) {
- idpf_free_dma_mem(hw, cq->bi.rx_buff[i]);
- kfree(cq->bi.rx_buff[i]);
- }
- kfree(cq->bi.rx_buff);
-
- return -ENOMEM;
-}
-
-/**
- * idpf_ctlq_free_desc_ring - Free Control Queue (CQ) rings
- * @hw: pointer to hw struct
- * @cq: pointer to the specific Control queue
- *
- * This assumes the posted send buffers have already been cleaned
- * and de-allocated
- */
-static void idpf_ctlq_free_desc_ring(struct idpf_hw *hw,
- struct idpf_ctlq_info *cq)
-{
- idpf_free_dma_mem(hw, &cq->desc_ring);
-}
-
-/**
- * idpf_ctlq_free_bufs - Free CQ buffer info elements
- * @hw: pointer to hw struct
- * @cq: pointer to the specific Control queue
- *
- * Free the DMA buffers for RX queues, and DMA buffer header for both RX and TX
- * queues. The upper layers are expected to manage freeing of TX DMA buffers
- */
-static void idpf_ctlq_free_bufs(struct idpf_hw *hw, struct idpf_ctlq_info *cq)
-{
- void *bi;
-
- if (cq->cq_type == IDPF_CTLQ_TYPE_MAILBOX_RX) {
- int i;
-
- /* free DMA buffers for rx queues*/
- for (i = 0; i < cq->ring_size; i++) {
- if (cq->bi.rx_buff[i]) {
- idpf_free_dma_mem(hw, cq->bi.rx_buff[i]);
- kfree(cq->bi.rx_buff[i]);
- }
- }
-
- bi = (void *)cq->bi.rx_buff;
- } else {
- bi = (void *)cq->bi.tx_msg;
- }
-
- /* free the buffer header */
- kfree(bi);
-}
-
-/**
- * idpf_ctlq_dealloc_ring_res - Free memory allocated for control queue
- * @hw: pointer to hw struct
- * @cq: pointer to the specific Control queue
- *
- * Free the memory used by the ring, buffers and other related structures
- */
-void idpf_ctlq_dealloc_ring_res(struct idpf_hw *hw, struct idpf_ctlq_info *cq)
-{
- /* free ring buffers and the ring itself */
- idpf_ctlq_free_bufs(hw, cq);
- idpf_ctlq_free_desc_ring(hw, cq);
-}
-
-/**
- * idpf_ctlq_alloc_ring_res - allocate memory for descriptor ring and bufs
- * @hw: pointer to hw struct
- * @cq: pointer to control queue struct
- *
- * Do *NOT* hold cq_lock when calling this as the memory allocation routines
- * called are not going to be atomic context safe
- */
-int idpf_ctlq_alloc_ring_res(struct idpf_hw *hw, struct idpf_ctlq_info *cq)
-{
- int err;
-
- /* allocate the ring memory */
- err = idpf_ctlq_alloc_desc_ring(hw, cq);
- if (err)
- return err;
-
- /* allocate buffers in the rings */
- err = idpf_ctlq_alloc_bufs(hw, cq);
- if (err)
- goto idpf_init_cq_free_ring;
-
- /* success! */
- return 0;
-
-idpf_init_cq_free_ring:
- idpf_free_dma_mem(hw, &cq->desc_ring);
-
- return err;
-}
diff --git a/drivers/net/ethernet/intel/idpf/idpf_dev.c b/drivers/net/ethernet/intel/idpf/idpf_dev.c
index 41e4bd49402a..0663faa2dfe1 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_dev.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_dev.c
@@ -9,42 +9,32 @@
/**
* idpf_ctlq_reg_init - initialize default mailbox registers
- * @cq: pointer to the array of create control queues
+ * @mmio: struct that contains MMIO region info
+ * @cci: struct where the register offset pointer to be copied to
*/
-static void idpf_ctlq_reg_init(struct idpf_ctlq_create_info *cq)
+static void idpf_ctlq_reg_init(struct libeth_mmio_info *mmio,
+ struct libeth_ctlq_create_info *cci)
{
- int i;
-
- for (i = 0; i < IDPF_NUM_DFLT_MBX_Q; i++) {
- struct idpf_ctlq_create_info *ccq = cq + i;
-
- switch (ccq->type) {
- case IDPF_CTLQ_TYPE_MAILBOX_TX:
- /* set head and tail registers in our local struct */
- ccq->reg.head = PF_FW_ATQH;
- ccq->reg.tail = PF_FW_ATQT;
- ccq->reg.len = PF_FW_ATQLEN;
- ccq->reg.bah = PF_FW_ATQBAH;
- ccq->reg.bal = PF_FW_ATQBAL;
- ccq->reg.len_mask = PF_FW_ATQLEN_ATQLEN_M;
- ccq->reg.len_ena_mask = PF_FW_ATQLEN_ATQENABLE_M;
- ccq->reg.head_mask = PF_FW_ATQH_ATQH_M;
- break;
- case IDPF_CTLQ_TYPE_MAILBOX_RX:
- /* set head and tail registers in our local struct */
- ccq->reg.head = PF_FW_ARQH;
- ccq->reg.tail = PF_FW_ARQT;
- ccq->reg.len = PF_FW_ARQLEN;
- ccq->reg.bah = PF_FW_ARQBAH;
- ccq->reg.bal = PF_FW_ARQBAL;
- ccq->reg.len_mask = PF_FW_ARQLEN_ARQLEN_M;
- ccq->reg.len_ena_mask = PF_FW_ARQLEN_ARQENABLE_M;
- ccq->reg.head_mask = PF_FW_ARQH_ARQH_M;
- break;
- default:
- break;
- }
- }
+ struct libeth_ctlq_reg *tx_reg = &cci[LIBETH_CTLQ_TYPE_TX].reg;
+ struct libeth_ctlq_reg *rx_reg = &cci[LIBETH_CTLQ_TYPE_RX].reg;
+
+ tx_reg->head = libeth_pci_get_mmio_addr(mmio, PF_FW_ATQH);
+ tx_reg->tail = libeth_pci_get_mmio_addr(mmio, PF_FW_ATQT);
+ tx_reg->len = libeth_pci_get_mmio_addr(mmio, PF_FW_ATQLEN);
+ tx_reg->addr_high = libeth_pci_get_mmio_addr(mmio, PF_FW_ATQBAH);
+ tx_reg->addr_low = libeth_pci_get_mmio_addr(mmio, PF_FW_ATQBAL);
+ tx_reg->len_mask = PF_FW_ATQLEN_ATQLEN_M;
+ tx_reg->len_ena_mask = PF_FW_ATQLEN_ATQENABLE_M;
+ tx_reg->head_mask = PF_FW_ATQH_ATQH_M;
+
+ rx_reg->head = libeth_pci_get_mmio_addr(mmio, PF_FW_ARQH);
+ rx_reg->tail = libeth_pci_get_mmio_addr(mmio, PF_FW_ARQT);
+ rx_reg->len = libeth_pci_get_mmio_addr(mmio, PF_FW_ARQLEN);
+ rx_reg->addr_high = libeth_pci_get_mmio_addr(mmio, PF_FW_ARQBAH);
+ rx_reg->addr_low = libeth_pci_get_mmio_addr(mmio, PF_FW_ARQBAL);
+ rx_reg->len_mask = PF_FW_ARQLEN_ARQLEN_M;
+ rx_reg->len_ena_mask = PF_FW_ARQLEN_ARQENABLE_M;
+ rx_reg->head_mask = PF_FW_ARQH_ARQH_M;
}
/**
@@ -53,13 +43,14 @@ static void idpf_ctlq_reg_init(struct idpf_ctlq_create_info *cq)
*/
static void idpf_mb_intr_reg_init(struct idpf_adapter *adapter)
{
+ struct libeth_mmio_info *mmio = &adapter->ctlq_ctx.mmio_info;
struct idpf_intr_reg *intr = &adapter->mb_vector.intr_reg;
u32 dyn_ctl = le32_to_cpu(adapter->caps.mailbox_dyn_ctl);
- intr->dyn_ctl = idpf_get_reg_addr(adapter, dyn_ctl);
+ intr->dyn_ctl = libeth_pci_get_mmio_addr(mmio, dyn_ctl);
intr->dyn_ctl_intena_m = PF_GLINT_DYN_CTL_INTENA_M;
intr->dyn_ctl_itridx_m = PF_GLINT_DYN_CTL_ITR_INDX_M;
- intr->icr_ena = idpf_get_reg_addr(adapter, PF_INT_DIR_OICR_ENA);
+ intr->icr_ena = libeth_pci_get_mmio_addr(mmio, PF_INT_DIR_OICR_ENA);
intr->icr_ena_ctlq_m = PF_INT_DIR_OICR_ENA_M;
}
@@ -72,6 +63,7 @@ static int idpf_intr_reg_init(struct idpf_vport *vport)
struct idpf_adapter *adapter = vport->adapter;
int num_vecs = vport->num_q_vectors;
struct idpf_vec_regs *reg_vals;
+ struct libeth_mmio_info *mmio;
int num_regs, i, err = 0;
u32 rx_itr, tx_itr;
u16 total_vecs;
@@ -88,14 +80,17 @@ static int idpf_intr_reg_init(struct idpf_vport *vport)
goto free_reg_vals;
}
+ mmio = &adapter->ctlq_ctx.mmio_info;
+
for (i = 0; i < num_vecs; i++) {
struct idpf_q_vector *q_vector = &vport->q_vectors[i];
u16 vec_id = vport->q_vector_idxs[i] - IDPF_MBX_Q_VEC;
struct idpf_intr_reg *intr = &q_vector->intr_reg;
+ struct idpf_vec_regs *reg = ®_vals[vec_id];
u32 spacing;
- intr->dyn_ctl = idpf_get_reg_addr(adapter,
- reg_vals[vec_id].dyn_ctl_reg);
+ intr->dyn_ctl = libeth_pci_get_mmio_addr(mmio,
+ reg->dyn_ctl_reg);
intr->dyn_ctl_intena_m = PF_GLINT_DYN_CTL_INTENA_M;
intr->dyn_ctl_intena_msk_m = PF_GLINT_DYN_CTL_INTENA_MSK_M;
intr->dyn_ctl_itridx_s = PF_GLINT_DYN_CTL_ITR_INDX_S;
@@ -105,16 +100,14 @@ static int idpf_intr_reg_init(struct idpf_vport *vport)
intr->dyn_ctl_sw_itridx_ena_m =
PF_GLINT_DYN_CTL_SW_ITR_INDX_ENA_M;
- spacing = IDPF_ITR_IDX_SPACING(reg_vals[vec_id].itrn_index_spacing,
+ spacing = IDPF_ITR_IDX_SPACING(reg->itrn_index_spacing,
IDPF_PF_ITR_IDX_SPACING);
rx_itr = PF_GLINT_ITR_ADDR(VIRTCHNL2_ITR_IDX_0,
- reg_vals[vec_id].itrn_reg,
- spacing);
+ reg->itrn_reg, spacing);
tx_itr = PF_GLINT_ITR_ADDR(VIRTCHNL2_ITR_IDX_1,
- reg_vals[vec_id].itrn_reg,
- spacing);
- intr->rx_itr = idpf_get_reg_addr(adapter, rx_itr);
- intr->tx_itr = idpf_get_reg_addr(adapter, tx_itr);
+ reg->itrn_reg, spacing);
+ intr->rx_itr = libeth_pci_get_mmio_addr(mmio, rx_itr);
+ intr->tx_itr = libeth_pci_get_mmio_addr(mmio, tx_itr);
}
free_reg_vals:
@@ -129,7 +122,9 @@ static int idpf_intr_reg_init(struct idpf_vport *vport)
*/
static void idpf_reset_reg_init(struct idpf_adapter *adapter)
{
- adapter->reset_reg.rstat = idpf_get_reg_addr(adapter, PFGEN_RSTAT);
+ adapter->reset_reg.rstat =
+ libeth_pci_get_mmio_addr(&adapter->ctlq_ctx.mmio_info,
+ PFGEN_RSTAT);
adapter->reset_reg.rstat_m = PFGEN_RSTAT_PFR_STATE_M;
}
@@ -141,11 +136,11 @@ static void idpf_reset_reg_init(struct idpf_adapter *adapter)
static void idpf_trigger_reset(struct idpf_adapter *adapter,
enum idpf_flags __always_unused trig_cause)
{
- u32 reset_reg;
+ void __iomem *addr;
- reset_reg = readl(idpf_get_reg_addr(adapter, PFGEN_CTRL));
- writel(reset_reg | PFGEN_CTRL_PFSWR,
- idpf_get_reg_addr(adapter, PFGEN_CTRL));
+ addr = libeth_pci_get_mmio_addr(&adapter->ctlq_ctx.mmio_info,
+ PFGEN_CTRL);
+ writel(readl(addr) | PFGEN_CTRL_PFSWR, addr);
}
/**
diff --git a/drivers/net/ethernet/intel/idpf/idpf_lib.c b/drivers/net/ethernet/intel/idpf/idpf_lib.c
index 7ee8ad61d77d..bf403dcf794e 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_lib.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_lib.c
@@ -1187,6 +1187,7 @@ void idpf_statistics_task(struct work_struct *work)
*/
void idpf_mbx_task(struct work_struct *work)
{
+ struct libeth_ctlq_xn_recv_params xn_params = {};
struct idpf_adapter *adapter;
adapter = container_of(work, struct idpf_adapter, mbx_task.work);
@@ -1197,7 +1198,11 @@ void idpf_mbx_task(struct work_struct *work)
queue_delayed_work(adapter->mbx_wq, &adapter->mbx_task,
msecs_to_jiffies(300));
- idpf_recv_mb_msg(adapter);
+ xn_params.xnm = adapter->xn_init_params.xnm;
+ xn_params.ctlq = adapter->arq;
+ xn_params.ctlq_msg_handler = idpf_recv_event_msg;
+
+ libeth_ctlq_xn_recv(&xn_params);
}
/**
@@ -1656,15 +1661,14 @@ void idpf_deinit_task(struct idpf_adapter *adapter)
/**
* idpf_check_reset_complete - check that reset is complete
- * @hw: pointer to hw struct
+ * @adapter: adapter to check
* @reset_reg: struct with reset registers
*
* Returns 0 if device is ready to use, or -EBUSY if it's in reset.
**/
-static int idpf_check_reset_complete(struct idpf_hw *hw,
+static int idpf_check_reset_complete(struct idpf_adapter *adapter,
struct idpf_reset_reg *reset_reg)
{
- struct idpf_adapter *adapter = hw->back;
int i;
for (i = 0; i < 2000; i++) {
@@ -1750,7 +1754,6 @@ static int idpf_init_hard_reset(struct idpf_adapter *adapter)
idpf_vc_core_deinit(adapter);
if (!is_reset)
reg_ops->trigger_reset(adapter, IDPF_HR_FUNC_RESET);
- idpf_deinit_dflt_mbx(adapter);
} else {
dev_err(dev, "Unhandled hard reset cause\n");
err = -EBADRQC;
@@ -1758,7 +1761,7 @@ static int idpf_init_hard_reset(struct idpf_adapter *adapter)
}
/* Wait for reset to complete */
- err = idpf_check_reset_complete(&adapter->hw, &adapter->reset_reg);
+ err = idpf_check_reset_complete(adapter, &adapter->reset_reg);
if (err) {
dev_err(dev, "The driver was unable to contact the device's firmware. Check that the FW is running. Driver state= 0x%x\n",
adapter->state);
@@ -2303,40 +2306,6 @@ static int idpf_set_mac(struct net_device *netdev, void *p)
return err;
}
-/**
- * idpf_alloc_dma_mem - Allocate dma memory
- * @hw: pointer to hw struct
- * @mem: pointer to dma_mem struct
- * @size: size of the memory to allocate
- */
-void *idpf_alloc_dma_mem(struct idpf_hw *hw, struct idpf_dma_mem *mem, u64 size)
-{
- struct idpf_adapter *adapter = hw->back;
- size_t sz = ALIGN(size, 4096);
-
- mem->va = dma_alloc_coherent(&adapter->pdev->dev, sz,
- &mem->pa, GFP_KERNEL);
- mem->size = sz;
-
- return mem->va;
-}
-
-/**
- * idpf_free_dma_mem - Free the allocated dma memory
- * @hw: pointer to hw struct
- * @mem: pointer to dma_mem struct
- */
-void idpf_free_dma_mem(struct idpf_hw *hw, struct idpf_dma_mem *mem)
-{
- struct idpf_adapter *adapter = hw->back;
-
- dma_free_coherent(&adapter->pdev->dev, mem->size,
- mem->va, mem->pa);
- mem->size = 0;
- mem->va = NULL;
- mem->pa = 0;
-}
-
static const struct net_device_ops idpf_netdev_ops = {
.ndo_open = idpf_open,
.ndo_stop = idpf_stop,
diff --git a/drivers/net/ethernet/intel/idpf/idpf_main.c b/drivers/net/ethernet/intel/idpf/idpf_main.c
index bec4a02c5373..0de919413f1e 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_main.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_main.c
@@ -9,8 +9,20 @@
MODULE_DESCRIPTION(DRV_SUMMARY);
MODULE_IMPORT_NS("LIBETH");
+MODULE_IMPORT_NS("LIBETH_CP");
+MODULE_IMPORT_NS("LIBETH_PCI");
MODULE_LICENSE("GPL");
+/**
+ * idpf_decfg_device - deconfigure device and device specific resources
+ * @adapter: driver specific private structure
+ */
+static void idpf_decfg_device(struct idpf_adapter *adapter)
+{
+ libeth_pci_unmap_all_mmio_regions(&adapter->ctlq_ctx.mmio_info);
+ libeth_pci_deinit_dev(adapter->pdev);
+}
+
/**
* idpf_remove - Device removal routine
* @pdev: PCI device information struct
@@ -35,7 +47,6 @@ static void idpf_remove(struct pci_dev *pdev)
/* Be a good citizen and leave the device clean on exit */
adapter->dev_ops.reg_ops.trigger_reset(adapter, IDPF_HR_FUNC_RESET);
- idpf_deinit_dflt_mbx(adapter);
if (!adapter->netdevs)
goto destroy_wqs;
@@ -69,15 +80,13 @@ static void idpf_remove(struct pci_dev *pdev)
adapter->vport_config = NULL;
kfree(adapter->netdevs);
adapter->netdevs = NULL;
- kfree(adapter->vcxn_mngr);
- adapter->vcxn_mngr = NULL;
mutex_destroy(&adapter->vport_ctrl_lock);
mutex_destroy(&adapter->vector_lock);
mutex_destroy(&adapter->queue_lock);
mutex_destroy(&adapter->vc_buf_lock);
- pci_set_drvdata(pdev, NULL);
+ idpf_decfg_device(adapter);
kfree(adapter);
}
@@ -91,31 +100,42 @@ static void idpf_shutdown(struct pci_dev *pdev)
cancel_delayed_work_sync(&adapter->vc_event_task);
idpf_vc_core_deinit(adapter);
- idpf_deinit_dflt_mbx(adapter);
if (system_state == SYSTEM_POWER_OFF)
pci_set_power_state(pdev, PCI_D3hot);
}
/**
- * idpf_cfg_hw - Initialize HW struct
- * @adapter: adapter to setup hw struct for
+ * idpf_cfg_device - configure device and device specific resources
+ * @adapter: driver specific private structure
*
- * Returns 0 on success, negative on failure
+ * Return: %0 on success, -%errno on failure.
*/
-static int idpf_cfg_hw(struct idpf_adapter *adapter)
+static int idpf_cfg_device(struct idpf_adapter *adapter)
{
+ struct libeth_mmio_info *mmio_info = &adapter->ctlq_ctx.mmio_info;
struct pci_dev *pdev = adapter->pdev;
- struct idpf_hw *hw = &adapter->hw;
+ int err;
- hw->hw_addr = pcim_iomap_table(pdev)[0];
- if (!hw->hw_addr) {
- pci_err(pdev, "failed to allocate PCI iomap table\n");
+ err = libeth_pci_init_dev(pdev);
+ if (err)
+ return err;
+ mmio_info->pdev = pdev;
+ INIT_LIST_HEAD(&mmio_info->mmio_list);
+
+ err = libeth_pci_map_mmio_region(mmio_info, 0,
+ pci_resource_len(pdev, 0));
+ if (!err) {
+ libeth_pci_deinit_dev(pdev);
return -ENOMEM;
}
- hw->back = adapter;
+ err = pci_enable_ptm(pdev, NULL);
+ if (err)
+ pci_dbg(pdev, "PCIe PTM is not supported by PCIe bus/controller\n");
+
+ pci_set_drvdata(pdev, adapter);
return 0;
}
@@ -156,28 +176,13 @@ static int idpf_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
}
adapter->pdev = pdev;
- err = pcim_enable_device(pdev);
- if (err)
- goto err_free;
-
- err = pcim_iomap_regions(pdev, BIT(0), pci_name(pdev));
+ err = idpf_cfg_device(adapter);
if (err) {
- pci_err(pdev, "pcim_iomap_regions failed %pe\n", ERR_PTR(err));
-
+ pci_err(pdev, "Failed to configure device specific resources: %pe\n",
+ ERR_PTR(err));
goto err_free;
}
- /* set up for high or low dma */
- err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64));
- if (err) {
- pci_err(pdev, "DMA configuration failed: %pe\n", ERR_PTR(err));
-
- goto err_free;
- }
-
- pci_set_master(pdev);
- pci_set_drvdata(pdev, adapter);
-
adapter->init_wq = alloc_workqueue("%s-%s-init",
WQ_UNBOUND | WQ_MEM_RECLAIM, 0,
dev_driver_string(dev),
@@ -185,7 +190,7 @@ static int idpf_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
if (!adapter->init_wq) {
dev_err(dev, "Failed to allocate init workqueue\n");
err = -ENOMEM;
- goto err_free;
+ goto err_init_wq;
}
adapter->serv_wq = alloc_workqueue("%s-%s-service",
@@ -231,13 +236,6 @@ static int idpf_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
/* setup msglvl */
adapter->msg_enable = netif_msg_init(-1, IDPF_AVAIL_NETIF_M);
- err = idpf_cfg_hw(adapter);
- if (err) {
- dev_err(dev, "Failed to configure HW structure for adapter: %d\n",
- err);
- goto err_cfg_hw;
- }
-
mutex_init(&adapter->vport_ctrl_lock);
mutex_init(&adapter->vector_lock);
mutex_init(&adapter->queue_lock);
@@ -256,8 +254,6 @@ static int idpf_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
return 0;
-err_cfg_hw:
- destroy_workqueue(adapter->vc_event_wq);
err_vc_event_wq_alloc:
destroy_workqueue(adapter->stats_wq);
err_stats_wq_alloc:
@@ -266,8 +262,11 @@ static int idpf_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
destroy_workqueue(adapter->serv_wq);
err_serv_wq_alloc:
destroy_workqueue(adapter->init_wq);
+err_init_wq:
+ idpf_decfg_device(adapter);
err_free:
kfree(adapter);
+
return err;
}
diff --git a/drivers/net/ethernet/intel/idpf/idpf_mem.h b/drivers/net/ethernet/intel/idpf/idpf_mem.h
deleted file mode 100644
index b21a04fccf0f..000000000000
--- a/drivers/net/ethernet/intel/idpf/idpf_mem.h
+++ /dev/null
@@ -1,20 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0-only */
-/* Copyright (C) 2023 Intel Corporation */
-
-#ifndef _IDPF_MEM_H_
-#define _IDPF_MEM_H_
-
-#include <linux/io.h>
-
-struct idpf_dma_mem {
- void *va;
- dma_addr_t pa;
- size_t size;
-};
-
-#define wr32(a, reg, value) writel((value), ((a)->hw_addr + (reg)))
-#define rd32(a, reg) readl((a)->hw_addr + (reg))
-#define wr64(a, reg, value) writeq((value), ((a)->hw_addr + (reg)))
-#define rd64(a, reg) readq((a)->hw_addr + (reg))
-
-#endif /* _IDPF_MEM_H_ */
diff --git a/drivers/net/ethernet/intel/idpf/idpf_txrx.h b/drivers/net/ethernet/intel/idpf/idpf_txrx.h
index 9f0cbb164054..d02ca4ee564c 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_txrx.h
+++ b/drivers/net/ethernet/intel/idpf/idpf_txrx.h
@@ -240,7 +240,7 @@ enum idpf_tx_ctx_desc_eipt_offload {
(sizeof(u16) * IDPF_RX_MAX_PTYPE_PROTO_IDS))
#define IDPF_RX_PTYPE_HDR_SZ sizeof(struct virtchnl2_get_ptype_info)
#define IDPF_RX_MAX_PTYPES_PER_BUF \
- DIV_ROUND_DOWN_ULL((IDPF_CTLQ_MAX_BUF_LEN - IDPF_RX_PTYPE_HDR_SZ), \
+ DIV_ROUND_DOWN_ULL(LIBETH_CTLQ_MAX_BUF_LEN - IDPF_RX_PTYPE_HDR_SZ, \
IDPF_RX_MAX_PTYPE_SZ)
#define IDPF_GET_PTYPE_SIZE(p) struct_size((p), proto_id, (p)->proto_id_count)
diff --git a/drivers/net/ethernet/intel/idpf/idpf_vf_dev.c b/drivers/net/ethernet/intel/idpf/idpf_vf_dev.c
index aba828abcb17..a2c4e41e1797 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_vf_dev.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_vf_dev.c
@@ -9,42 +9,32 @@
/**
* idpf_vf_ctlq_reg_init - initialize default mailbox registers
- * @cq: pointer to the array of create control queues
+ * @mmio: struct that contains MMIO region info
+ * @cci: struct where the register offset pointer to be copied to
*/
-static void idpf_vf_ctlq_reg_init(struct idpf_ctlq_create_info *cq)
+static void idpf_vf_ctlq_reg_init(struct libeth_mmio_info *mmio,
+ struct libeth_ctlq_create_info *cci)
{
- int i;
-
- for (i = 0; i < IDPF_NUM_DFLT_MBX_Q; i++) {
- struct idpf_ctlq_create_info *ccq = cq + i;
-
- switch (ccq->type) {
- case IDPF_CTLQ_TYPE_MAILBOX_TX:
- /* set head and tail registers in our local struct */
- ccq->reg.head = VF_ATQH;
- ccq->reg.tail = VF_ATQT;
- ccq->reg.len = VF_ATQLEN;
- ccq->reg.bah = VF_ATQBAH;
- ccq->reg.bal = VF_ATQBAL;
- ccq->reg.len_mask = VF_ATQLEN_ATQLEN_M;
- ccq->reg.len_ena_mask = VF_ATQLEN_ATQENABLE_M;
- ccq->reg.head_mask = VF_ATQH_ATQH_M;
- break;
- case IDPF_CTLQ_TYPE_MAILBOX_RX:
- /* set head and tail registers in our local struct */
- ccq->reg.head = VF_ARQH;
- ccq->reg.tail = VF_ARQT;
- ccq->reg.len = VF_ARQLEN;
- ccq->reg.bah = VF_ARQBAH;
- ccq->reg.bal = VF_ARQBAL;
- ccq->reg.len_mask = VF_ARQLEN_ARQLEN_M;
- ccq->reg.len_ena_mask = VF_ARQLEN_ARQENABLE_M;
- ccq->reg.head_mask = VF_ARQH_ARQH_M;
- break;
- default:
- break;
- }
- }
+ struct libeth_ctlq_reg *tx_reg = &cci[LIBETH_CTLQ_TYPE_TX].reg;
+ struct libeth_ctlq_reg *rx_reg = &cci[LIBETH_CTLQ_TYPE_RX].reg;
+
+ tx_reg->head = libeth_pci_get_mmio_addr(mmio, VF_ATQH);
+ tx_reg->tail = libeth_pci_get_mmio_addr(mmio, VF_ATQT);
+ tx_reg->len = libeth_pci_get_mmio_addr(mmio, VF_ATQLEN);
+ tx_reg->addr_high = libeth_pci_get_mmio_addr(mmio, VF_ATQBAH);
+ tx_reg->addr_low = libeth_pci_get_mmio_addr(mmio, VF_ATQBAL);
+ tx_reg->len_mask = VF_ATQLEN_ATQLEN_M;
+ tx_reg->len_ena_mask = VF_ATQLEN_ATQENABLE_M;
+ tx_reg->head_mask = VF_ATQH_ATQH_M;
+
+ rx_reg->head = libeth_pci_get_mmio_addr(mmio, VF_ARQH);
+ rx_reg->tail = libeth_pci_get_mmio_addr(mmio, VF_ARQT);
+ rx_reg->len = libeth_pci_get_mmio_addr(mmio, VF_ARQLEN);
+ rx_reg->addr_high = libeth_pci_get_mmio_addr(mmio, VF_ARQBAH);
+ rx_reg->addr_low = libeth_pci_get_mmio_addr(mmio, VF_ARQBAL);
+ rx_reg->len_mask = VF_ARQLEN_ARQLEN_M;
+ rx_reg->len_ena_mask = VF_ARQLEN_ARQENABLE_M;
+ rx_reg->head_mask = VF_ARQH_ARQH_M;
}
/**
@@ -53,13 +43,14 @@ static void idpf_vf_ctlq_reg_init(struct idpf_ctlq_create_info *cq)
*/
static void idpf_vf_mb_intr_reg_init(struct idpf_adapter *adapter)
{
+ struct libeth_mmio_info *mmio = &adapter->ctlq_ctx.mmio_info;
struct idpf_intr_reg *intr = &adapter->mb_vector.intr_reg;
u32 dyn_ctl = le32_to_cpu(adapter->caps.mailbox_dyn_ctl);
- intr->dyn_ctl = idpf_get_reg_addr(adapter, dyn_ctl);
+ intr->dyn_ctl = libeth_pci_get_mmio_addr(mmio, dyn_ctl);
intr->dyn_ctl_intena_m = VF_INT_DYN_CTL0_INTENA_M;
intr->dyn_ctl_itridx_m = VF_INT_DYN_CTL0_ITR_INDX_M;
- intr->icr_ena = idpf_get_reg_addr(adapter, VF_INT_ICR0_ENA1);
+ intr->icr_ena = libeth_pci_get_mmio_addr(mmio, VF_INT_ICR0_ENA1);
intr->icr_ena_ctlq_m = VF_INT_ICR0_ENA1_ADMINQ_M;
}
@@ -72,6 +63,7 @@ static int idpf_vf_intr_reg_init(struct idpf_vport *vport)
struct idpf_adapter *adapter = vport->adapter;
int num_vecs = vport->num_q_vectors;
struct idpf_vec_regs *reg_vals;
+ struct libeth_mmio_info *mmio;
int num_regs, i, err = 0;
u32 rx_itr, tx_itr;
u16 total_vecs;
@@ -88,14 +80,17 @@ static int idpf_vf_intr_reg_init(struct idpf_vport *vport)
goto free_reg_vals;
}
+ mmio = &adapter->ctlq_ctx.mmio_info;
+
for (i = 0; i < num_vecs; i++) {
struct idpf_q_vector *q_vector = &vport->q_vectors[i];
u16 vec_id = vport->q_vector_idxs[i] - IDPF_MBX_Q_VEC;
struct idpf_intr_reg *intr = &q_vector->intr_reg;
+ struct idpf_vec_regs *reg = ®_vals[vec_id];
u32 spacing;
- intr->dyn_ctl = idpf_get_reg_addr(adapter,
- reg_vals[vec_id].dyn_ctl_reg);
+ intr->dyn_ctl = libeth_pci_get_mmio_addr(mmio,
+ reg->dyn_ctl_reg);
intr->dyn_ctl_intena_m = VF_INT_DYN_CTLN_INTENA_M;
intr->dyn_ctl_intena_msk_m = VF_INT_DYN_CTLN_INTENA_MSK_M;
intr->dyn_ctl_itridx_s = VF_INT_DYN_CTLN_ITR_INDX_S;
@@ -105,16 +100,14 @@ static int idpf_vf_intr_reg_init(struct idpf_vport *vport)
intr->dyn_ctl_sw_itridx_ena_m =
VF_INT_DYN_CTLN_SW_ITR_INDX_ENA_M;
- spacing = IDPF_ITR_IDX_SPACING(reg_vals[vec_id].itrn_index_spacing,
+ spacing = IDPF_ITR_IDX_SPACING(reg->itrn_index_spacing,
IDPF_VF_ITR_IDX_SPACING);
rx_itr = VF_INT_ITRN_ADDR(VIRTCHNL2_ITR_IDX_0,
- reg_vals[vec_id].itrn_reg,
- spacing);
+ reg->itrn_reg, spacing);
tx_itr = VF_INT_ITRN_ADDR(VIRTCHNL2_ITR_IDX_1,
- reg_vals[vec_id].itrn_reg,
- spacing);
- intr->rx_itr = idpf_get_reg_addr(adapter, rx_itr);
- intr->tx_itr = idpf_get_reg_addr(adapter, tx_itr);
+ reg->itrn_reg, spacing);
+ intr->rx_itr = libeth_pci_get_mmio_addr(mmio, rx_itr);
+ intr->tx_itr = libeth_pci_get_mmio_addr(mmio, tx_itr);
}
free_reg_vals:
@@ -129,7 +122,9 @@ static int idpf_vf_intr_reg_init(struct idpf_vport *vport)
*/
static void idpf_vf_reset_reg_init(struct idpf_adapter *adapter)
{
- adapter->reset_reg.rstat = idpf_get_reg_addr(adapter, VFGEN_RSTAT);
+ adapter->reset_reg.rstat =
+ libeth_pci_get_mmio_addr(&adapter->ctlq_ctx.mmio_info,
+ VFGEN_RSTAT);
adapter->reset_reg.rstat_m = VFGEN_RSTAT_VFR_STATE_M;
}
@@ -141,10 +136,14 @@ static void idpf_vf_reset_reg_init(struct idpf_adapter *adapter)
static void idpf_vf_trigger_reset(struct idpf_adapter *adapter,
enum idpf_flags trig_cause)
{
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .chnl_opcode = VIRTCHNL2_OP_RESET_VF,
+ };
+
/* Do not send VIRTCHNL2_OP_RESET_VF message on driver unload */
if (trig_cause == IDPF_HR_FUNC_RESET &&
!test_bit(IDPF_REMOVE_IN_PROG, adapter->flags))
- idpf_send_mb_msg(adapter, VIRTCHNL2_OP_RESET_VF, 0, NULL, 0);
+ idpf_send_mb_msg(adapter, &xn_params, NULL, 0);
}
/**
diff --git a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
index bfa2a3dad3d6..2bfb8882ae13 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
@@ -8,99 +8,6 @@
#define IDPF_VC_XN_MIN_TIMEOUT_MSEC 2000
#define IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC (60 * 1000)
-#define IDPF_VC_XN_IDX_M GENMASK(7, 0)
-#define IDPF_VC_XN_SALT_M GENMASK(15, 8)
-#define IDPF_VC_XN_RING_LEN U8_MAX
-
-/**
- * enum idpf_vc_xn_state - Virtchnl transaction status
- * @IDPF_VC_XN_IDLE: not expecting a reply, ready to be used
- * @IDPF_VC_XN_WAITING: expecting a reply, not yet received
- * @IDPF_VC_XN_COMPLETED_SUCCESS: a reply was expected and received,
- * buffer updated
- * @IDPF_VC_XN_COMPLETED_FAILED: a reply was expected and received, but there
- * was an error, buffer not updated
- * @IDPF_VC_XN_SHUTDOWN: transaction object cannot be used, VC torn down
- * @IDPF_VC_XN_ASYNC: transaction sent asynchronously and doesn't have the
- * return context; a callback may be provided to handle
- * return
- */
-enum idpf_vc_xn_state {
- IDPF_VC_XN_IDLE = 1,
- IDPF_VC_XN_WAITING,
- IDPF_VC_XN_COMPLETED_SUCCESS,
- IDPF_VC_XN_COMPLETED_FAILED,
- IDPF_VC_XN_SHUTDOWN,
- IDPF_VC_XN_ASYNC,
-};
-
-struct idpf_vc_xn;
-/* Callback for asynchronous messages */
-typedef int (*async_vc_cb) (struct idpf_adapter *, struct idpf_vc_xn *,
- const struct idpf_ctlq_msg *);
-
-/**
- * struct idpf_vc_xn - Data structure representing virtchnl transactions
- * @completed: virtchnl event loop uses that to signal when a reply is
- * available, uses kernel completion API
- * @state: virtchnl event loop stores the data below, protected by the
- * completion's lock.
- * @reply_sz: Original size of reply, may be > reply_buf.iov_len; it will be
- * truncated on its way to the receiver thread according to
- * reply_buf.iov_len.
- * @reply: Reference to the buffer(s) where the reply data should be written
- * to. May be 0-length (then NULL address permitted) if the reply data
- * should be ignored.
- * @async_handler: if sent asynchronously, a callback can be provided to handle
- * the reply when it's received
- * @vc_op: corresponding opcode sent with this transaction
- * @idx: index used as retrieval on reply receive, used for cookie
- * @salt: changed every message to make unique, used for cookie
- */
-struct idpf_vc_xn {
- struct completion completed;
- enum idpf_vc_xn_state state;
- size_t reply_sz;
- struct kvec reply;
- async_vc_cb async_handler;
- u32 vc_op;
- u8 idx;
- u8 salt;
-};
-
-/**
- * struct idpf_vc_xn_params - Parameters for executing transaction
- * @send_buf: kvec for send buffer
- * @recv_buf: kvec for recv buffer, may be NULL, must then have zero length
- * @timeout_ms: timeout to wait for reply
- * @async: send message asynchronously, will not wait on completion
- * @async_handler: If sent asynchronously, optional callback handler. The user
- * must be careful when using async handlers as the memory for
- * the recv_buf _cannot_ be on stack if this is async.
- * @vc_op: virtchnl op to send
- */
-struct idpf_vc_xn_params {
- struct kvec send_buf;
- struct kvec recv_buf;
- int timeout_ms;
- bool async;
- async_vc_cb async_handler;
- u32 vc_op;
-};
-
-/**
- * struct idpf_vc_xn_manager - Manager for tracking transactions
- * @ring: backing and lookup for transactions
- * @free_xn_bm: bitmap for free transactions
- * @xn_bm_lock: make bitmap access synchronous where necessary
- * @salt: used to make cookie unique every message
- */
-struct idpf_vc_xn_manager {
- struct idpf_vc_xn ring[IDPF_VC_XN_RING_LEN];
- DECLARE_BITMAP(free_xn_bm, IDPF_VC_XN_RING_LEN);
- spinlock_t xn_bm_lock;
- u8 salt;
-};
/**
* idpf_vid_to_vport - Translate vport id to vport pointer
@@ -162,586 +69,100 @@ static void idpf_handle_event_link(struct idpf_adapter *adapter,
/**
* idpf_recv_event_msg - Receive virtchnl event message
- * @adapter: Driver specific private structure
+ * @ctx: control queue context
* @ctlq_msg: message to copy from
*
* Receive virtchnl event message
*/
-static void idpf_recv_event_msg(struct idpf_adapter *adapter,
- struct idpf_ctlq_msg *ctlq_msg)
+void idpf_recv_event_msg(struct libeth_ctlq_ctx *ctx,
+ struct libeth_ctlq_msg *ctlq_msg)
{
- int payload_size = ctlq_msg->ctx.indirect.payload->size;
+ struct kvec *buff = &ctlq_msg->recv_mem;
+ int payload_size = buff->iov_len;
+ struct idpf_adapter *adapter;
struct virtchnl2_event *v2e;
u32 event;
+ adapter = container_of(ctx, struct idpf_adapter, ctlq_ctx);
if (payload_size < sizeof(*v2e)) {
dev_err_ratelimited(&adapter->pdev->dev, "Failed to receive valid payload for event msg (op %d len %d)\n",
- ctlq_msg->cookie.mbx.chnl_opcode,
+ ctlq_msg->chnl_opcode,
payload_size);
- return;
+ goto free_rx_buf;
}
- v2e = (struct virtchnl2_event *)ctlq_msg->ctx.indirect.payload->va;
+ v2e = (struct virtchnl2_event *)buff->iov_base;
event = le32_to_cpu(v2e->event);
switch (event) {
case VIRTCHNL2_EVENT_LINK_CHANGE:
idpf_handle_event_link(adapter, v2e);
- return;
+ break;
default:
dev_err(&adapter->pdev->dev,
"Unknown event %d from PF\n", event);
break;
}
-}
-
-/**
- * idpf_mb_clean - Reclaim the send mailbox queue entries
- * @adapter: Driver specific private structure
- *
- * Reclaim the send mailbox queue entries to be used to send further messages
- *
- * Returns 0 on success, negative on failure
- */
-static int idpf_mb_clean(struct idpf_adapter *adapter)
-{
- u16 i, num_q_msg = IDPF_DFLT_MBX_Q_LEN;
- struct idpf_ctlq_msg **q_msg;
- struct idpf_dma_mem *dma_mem;
- int err;
-
- q_msg = kcalloc(num_q_msg, sizeof(struct idpf_ctlq_msg *), GFP_ATOMIC);
- if (!q_msg)
- return -ENOMEM;
- err = idpf_ctlq_clean_sq(adapter->hw.asq, &num_q_msg, q_msg);
- if (err)
- goto err_kfree;
-
- for (i = 0; i < num_q_msg; i++) {
- if (!q_msg[i])
- continue;
- dma_mem = q_msg[i]->ctx.indirect.payload;
- if (dma_mem)
- dma_free_coherent(&adapter->pdev->dev, dma_mem->size,
- dma_mem->va, dma_mem->pa);
- kfree(q_msg[i]);
- kfree(dma_mem);
- }
-
-err_kfree:
- kfree(q_msg);
-
- return err;
+free_rx_buf:
+ libeth_ctlq_release_rx_buf(adapter->arq, buff);
}
/**
- * idpf_send_mb_msg - Send message over mailbox
+ * idpf_mb_clean - cleanup the send mailbox queue entries
* @adapter: Driver specific private structure
- * @op: virtchnl opcode
- * @msg_size: size of the payload
- * @msg: pointer to buffer holding the payload
- * @cookie: unique SW generated cookie per message
*
- * Will prepare the control queue message and initiates the send api
- *
- * Returns 0 on success, negative on failure
+ * This is a helper function to clean the send mailbox queue entries.
*/
-int idpf_send_mb_msg(struct idpf_adapter *adapter, u32 op,
- u16 msg_size, u8 *msg, u16 cookie)
+static void idpf_mb_clean(struct idpf_adapter *adapter)
{
- struct idpf_ctlq_msg *ctlq_msg;
- struct idpf_dma_mem *dma_mem;
- int err;
-
- /* If we are here and a reset is detected nothing much can be
- * done. This thread should silently abort and expected to
- * be corrected with a new run either by user or driver
- * flows after reset
- */
- if (idpf_is_reset_detected(adapter))
- return 0;
-
- err = idpf_mb_clean(adapter);
- if (err)
- return err;
-
- ctlq_msg = kzalloc(sizeof(*ctlq_msg), GFP_ATOMIC);
- if (!ctlq_msg)
- return -ENOMEM;
-
- dma_mem = kzalloc(sizeof(*dma_mem), GFP_ATOMIC);
- if (!dma_mem) {
- err = -ENOMEM;
- goto dma_mem_error;
- }
-
- ctlq_msg->opcode = idpf_mbq_opc_send_msg_to_cp;
- ctlq_msg->func_id = 0;
- ctlq_msg->data_len = msg_size;
- ctlq_msg->cookie.mbx.chnl_opcode = op;
- ctlq_msg->cookie.mbx.chnl_retval = 0;
- dma_mem->size = IDPF_CTLQ_MAX_BUF_LEN;
- dma_mem->va = dma_alloc_coherent(&adapter->pdev->dev, dma_mem->size,
- &dma_mem->pa, GFP_ATOMIC);
- if (!dma_mem->va) {
- err = -ENOMEM;
- goto dma_alloc_error;
- }
-
- /* It's possible we're just sending an opcode but no buffer */
- if (msg && msg_size)
- memcpy(dma_mem->va, msg, msg_size);
- ctlq_msg->ctx.indirect.payload = dma_mem;
- ctlq_msg->ctx.sw_cookie.data = cookie;
-
- err = idpf_ctlq_send(&adapter->hw, adapter->hw.asq, 1, ctlq_msg);
- if (err)
- goto send_error;
-
- return 0;
-
-send_error:
- dma_free_coherent(&adapter->pdev->dev, dma_mem->size, dma_mem->va,
- dma_mem->pa);
-dma_alloc_error:
- kfree(dma_mem);
-dma_mem_error:
- kfree(ctlq_msg);
-
- return err;
-}
-
-/* API for virtchnl "transaction" support ("xn" for short).
- *
- * We are reusing the completion lock to serialize the accesses to the
- * transaction state for simplicity, but it could be its own separate synchro
- * as well. For now, this API is only used from within a workqueue context;
- * raw_spin_lock() is enough.
- */
-/**
- * idpf_vc_xn_lock - Request exclusive access to vc transaction
- * @xn: struct idpf_vc_xn* to access
- */
-#define idpf_vc_xn_lock(xn) \
- raw_spin_lock(&(xn)->completed.wait.lock)
-
-/**
- * idpf_vc_xn_unlock - Release exclusive access to vc transaction
- * @xn: struct idpf_vc_xn* to access
- */
-#define idpf_vc_xn_unlock(xn) \
- raw_spin_unlock(&(xn)->completed.wait.lock)
-
-/**
- * idpf_vc_xn_release_bufs - Release reference to reply buffer(s) and
- * reset the transaction state.
- * @xn: struct idpf_vc_xn to update
- */
-static void idpf_vc_xn_release_bufs(struct idpf_vc_xn *xn)
-{
- xn->reply.iov_base = NULL;
- xn->reply.iov_len = 0;
-
- if (xn->state != IDPF_VC_XN_SHUTDOWN)
- xn->state = IDPF_VC_XN_IDLE;
-}
-
-/**
- * idpf_vc_xn_init - Initialize virtchnl transaction object
- * @vcxn_mngr: pointer to vc transaction manager struct
- */
-static void idpf_vc_xn_init(struct idpf_vc_xn_manager *vcxn_mngr)
-{
- int i;
-
- spin_lock_init(&vcxn_mngr->xn_bm_lock);
-
- for (i = 0; i < ARRAY_SIZE(vcxn_mngr->ring); i++) {
- struct idpf_vc_xn *xn = &vcxn_mngr->ring[i];
-
- xn->state = IDPF_VC_XN_IDLE;
- xn->idx = i;
- idpf_vc_xn_release_bufs(xn);
- init_completion(&xn->completed);
- }
-
- bitmap_fill(vcxn_mngr->free_xn_bm, IDPF_VC_XN_RING_LEN);
-}
-
-/**
- * idpf_vc_xn_shutdown - Uninitialize virtchnl transaction object
- * @vcxn_mngr: pointer to vc transaction manager struct
- *
- * All waiting threads will be woken-up and their transaction aborted. Further
- * operations on that object will fail.
- */
-static void idpf_vc_xn_shutdown(struct idpf_vc_xn_manager *vcxn_mngr)
-{
- int i;
-
- spin_lock_bh(&vcxn_mngr->xn_bm_lock);
- bitmap_zero(vcxn_mngr->free_xn_bm, IDPF_VC_XN_RING_LEN);
- spin_unlock_bh(&vcxn_mngr->xn_bm_lock);
-
- for (i = 0; i < ARRAY_SIZE(vcxn_mngr->ring); i++) {
- struct idpf_vc_xn *xn = &vcxn_mngr->ring[i];
+ struct libeth_ctlq_xn_clean_params clean_params = {
+ .ctx = &adapter->ctlq_ctx,
+ .ctlq = adapter->asq,
+ .rel_tx_buf = kfree,
+ .num_msgs = IDPF_DFLT_MBX_Q_LEN,
+ };
- idpf_vc_xn_lock(xn);
- xn->state = IDPF_VC_XN_SHUTDOWN;
- idpf_vc_xn_release_bufs(xn);
- idpf_vc_xn_unlock(xn);
- complete_all(&xn->completed);
- }
+ libeth_ctlq_xn_send_clean(&clean_params);
}
/**
- * idpf_vc_xn_pop_free - Pop a free transaction from free list
- * @vcxn_mngr: transaction manager to pop from
+ * idpf_send_mb_msg - send mailbox message to the device control plane
+ * @adapter: driver specific private structure
+ * @xn_params: Xn send parameters to fill
+ * @send_buf: buffer to send
+ * @send_buf_size: size of the send buffer
*
- * Returns NULL if no free transactions
- */
-static
-struct idpf_vc_xn *idpf_vc_xn_pop_free(struct idpf_vc_xn_manager *vcxn_mngr)
-{
- struct idpf_vc_xn *xn = NULL;
- unsigned long free_idx;
-
- spin_lock_bh(&vcxn_mngr->xn_bm_lock);
- free_idx = find_first_bit(vcxn_mngr->free_xn_bm, IDPF_VC_XN_RING_LEN);
- if (free_idx == IDPF_VC_XN_RING_LEN)
- goto do_unlock;
-
- clear_bit(free_idx, vcxn_mngr->free_xn_bm);
- xn = &vcxn_mngr->ring[free_idx];
- xn->salt = vcxn_mngr->salt++;
-
-do_unlock:
- spin_unlock_bh(&vcxn_mngr->xn_bm_lock);
-
- return xn;
-}
-
-/**
- * idpf_vc_xn_push_free - Push a free transaction to free list
- * @vcxn_mngr: transaction manager to push to
- * @xn: transaction to push
- */
-static void idpf_vc_xn_push_free(struct idpf_vc_xn_manager *vcxn_mngr,
- struct idpf_vc_xn *xn)
-{
- idpf_vc_xn_release_bufs(xn);
- set_bit(xn->idx, vcxn_mngr->free_xn_bm);
-}
-
-/**
- * idpf_vc_xn_exec - Perform a send/recv virtchnl transaction
- * @adapter: driver specific private structure with vcxn_mngr
- * @params: parameters for this particular transaction including
- * -vc_op: virtchannel operation to send
- * -send_buf: kvec iov for send buf and len
- * -recv_buf: kvec iov for recv buf and len (ignored if NULL)
- * -timeout_ms: timeout waiting for a reply (milliseconds)
- * -async: don't wait for message reply, will lose caller context
- * -async_handler: callback to handle async replies
+ * Fill the Xn parameters with the required info to send a virtchnl message.
+ * The send buffer is DMA mapped in the libeth to avoid memcpy.
*
- * @returns >= 0 for success, the size of the initial reply (may or may not be
- * >= @recv_buf.iov_len, but we never overflow @@recv_buf_iov_base). < 0 for
- * error.
- */
-static ssize_t idpf_vc_xn_exec(struct idpf_adapter *adapter,
- const struct idpf_vc_xn_params *params)
-{
- const struct kvec *send_buf = ¶ms->send_buf;
- struct idpf_vc_xn *xn;
- ssize_t retval;
- u16 cookie;
-
- xn = idpf_vc_xn_pop_free(adapter->vcxn_mngr);
- /* no free transactions available */
- if (!xn)
- return -ENOSPC;
-
- idpf_vc_xn_lock(xn);
- if (xn->state == IDPF_VC_XN_SHUTDOWN) {
- retval = -ENXIO;
- goto only_unlock;
- } else if (xn->state != IDPF_VC_XN_IDLE) {
- /* We're just going to clobber this transaction even though
- * it's not IDLE. If we don't reuse it we could theoretically
- * eventually leak all the free transactions and not be able to
- * send any messages. At least this way we make an attempt to
- * remain functional even though something really bad is
- * happening that's corrupting what was supposed to be free
- * transactions.
- */
- WARN_ONCE(1, "There should only be idle transactions in free list (idx %d op %d)\n",
- xn->idx, xn->vc_op);
- }
-
- xn->reply = params->recv_buf;
- xn->reply_sz = 0;
- xn->state = params->async ? IDPF_VC_XN_ASYNC : IDPF_VC_XN_WAITING;
- xn->vc_op = params->vc_op;
- xn->async_handler = params->async_handler;
- idpf_vc_xn_unlock(xn);
-
- if (!params->async)
- reinit_completion(&xn->completed);
- cookie = FIELD_PREP(IDPF_VC_XN_SALT_M, xn->salt) |
- FIELD_PREP(IDPF_VC_XN_IDX_M, xn->idx);
-
- retval = idpf_send_mb_msg(adapter, params->vc_op,
- send_buf->iov_len, send_buf->iov_base,
- cookie);
- if (retval) {
- idpf_vc_xn_lock(xn);
- goto release_and_unlock;
- }
-
- if (params->async)
- return 0;
-
- wait_for_completion_timeout(&xn->completed,
- msecs_to_jiffies(params->timeout_ms));
-
- /* No need to check the return value; we check the final state of the
- * transaction below. It's possible the transaction actually gets more
- * timeout than specified if we get preempted here but after
- * wait_for_completion_timeout returns. This should be non-issue
- * however.
- */
- idpf_vc_xn_lock(xn);
- switch (xn->state) {
- case IDPF_VC_XN_SHUTDOWN:
- retval = -ENXIO;
- goto only_unlock;
- case IDPF_VC_XN_WAITING:
- dev_notice_ratelimited(&adapter->pdev->dev,
- "Transaction timed-out (op:%d cookie:%04x vc_op:%d salt:%02x timeout:%dms)\n",
- params->vc_op, cookie, xn->vc_op,
- xn->salt, params->timeout_ms);
- retval = -ETIME;
- break;
- case IDPF_VC_XN_COMPLETED_SUCCESS:
- retval = xn->reply_sz;
- break;
- case IDPF_VC_XN_COMPLETED_FAILED:
- dev_notice_ratelimited(&adapter->pdev->dev, "Transaction failed (op %d)\n",
- params->vc_op);
- retval = -EIO;
- break;
- default:
- /* Invalid state. */
- WARN_ON_ONCE(1);
- retval = -EIO;
- break;
- }
-
-release_and_unlock:
- idpf_vc_xn_push_free(adapter->vcxn_mngr, xn);
- /* If we receive a VC reply after here, it will be dropped. */
-only_unlock:
- idpf_vc_xn_unlock(xn);
-
- return retval;
-}
-
-/**
- * idpf_vc_xn_forward_async - Handle async reply receives
- * @adapter: private data struct
- * @xn: transaction to handle
- * @ctlq_msg: corresponding ctlq_msg
+ * Cleanup the mailbox queue entries of the previously sent message to
+ * unmap and release the buffer.
*
- * For async sends we're going to lose the caller's context so, if an
- * async_handler was provided, it can deal with the reply, otherwise we'll just
- * check and report if there is an error.
+ * Return: 0 if sending was successful or reset in detected,
+ * negative error code on failure.
*/
-static int
-idpf_vc_xn_forward_async(struct idpf_adapter *adapter, struct idpf_vc_xn *xn,
- const struct idpf_ctlq_msg *ctlq_msg)
+int idpf_send_mb_msg(struct idpf_adapter *adapter,
+ struct libeth_ctlq_xn_send_params *xn_params,
+ void *send_buf, size_t send_buf_size)
{
- int err = 0;
+ if (idpf_is_reset_detected(adapter)) {
+ if (!libeth_cp_can_send_onstack(send_buf_size))
+ kfree(send_buf);
- if (ctlq_msg->cookie.mbx.chnl_opcode != xn->vc_op) {
- dev_err_ratelimited(&adapter->pdev->dev, "Async message opcode does not match transaction opcode (msg: %d) (xn: %d)\n",
- ctlq_msg->cookie.mbx.chnl_opcode, xn->vc_op);
- xn->reply_sz = 0;
- err = -EINVAL;
- goto release_bufs;
+ return -EBUSY;
}
- if (xn->async_handler) {
- err = xn->async_handler(adapter, xn, ctlq_msg);
- goto release_bufs;
- }
-
- if (ctlq_msg->cookie.mbx.chnl_retval) {
- xn->reply_sz = 0;
- dev_err_ratelimited(&adapter->pdev->dev, "Async message failure (op %d)\n",
- ctlq_msg->cookie.mbx.chnl_opcode);
- err = -EINVAL;
- }
+ idpf_mb_clean(adapter);
-release_bufs:
- idpf_vc_xn_push_free(adapter->vcxn_mngr, xn);
+ xn_params->send_buf.iov_base = send_buf;
+ xn_params->send_buf.iov_len = send_buf_size;
+ xn_params->xnm = adapter->xn_init_params.xnm;
+ xn_params->ctlq = adapter->asq;
+ xn_params->rel_tx_buf = kfree;
+ xn_params->ctlq_msg = NULL;
- return err;
-}
-
-/**
- * idpf_vc_xn_forward_reply - copy a reply back to receiving thread
- * @adapter: driver specific private structure with vcxn_mngr
- * @ctlq_msg: controlq message to send back to receiving thread
- */
-static int
-idpf_vc_xn_forward_reply(struct idpf_adapter *adapter,
- const struct idpf_ctlq_msg *ctlq_msg)
-{
- const void *payload = NULL;
- size_t payload_size = 0;
- struct idpf_vc_xn *xn;
- u16 msg_info;
- int err = 0;
- u16 xn_idx;
- u16 salt;
-
- msg_info = ctlq_msg->ctx.sw_cookie.data;
- xn_idx = FIELD_GET(IDPF_VC_XN_IDX_M, msg_info);
- if (xn_idx >= ARRAY_SIZE(adapter->vcxn_mngr->ring)) {
- dev_err_ratelimited(&adapter->pdev->dev, "Out of bounds cookie received: %02x\n",
- xn_idx);
- return -EINVAL;
- }
- xn = &adapter->vcxn_mngr->ring[xn_idx];
- idpf_vc_xn_lock(xn);
- salt = FIELD_GET(IDPF_VC_XN_SALT_M, msg_info);
- if (xn->salt != salt) {
- dev_err_ratelimited(&adapter->pdev->dev, "Transaction salt does not match (exp:%d@%02x(%d) != got:%d@%02x)\n",
- xn->vc_op, xn->salt, xn->state,
- ctlq_msg->cookie.mbx.chnl_opcode, salt);
- idpf_vc_xn_unlock(xn);
- return -EINVAL;
- }
-
- switch (xn->state) {
- case IDPF_VC_XN_WAITING:
- /* success */
- break;
- case IDPF_VC_XN_IDLE:
- dev_err_ratelimited(&adapter->pdev->dev, "Unexpected or belated VC reply (op %d)\n",
- ctlq_msg->cookie.mbx.chnl_opcode);
- err = -EINVAL;
- goto out_unlock;
- case IDPF_VC_XN_SHUTDOWN:
- /* ENXIO is a bit special here as the recv msg loop uses that
- * know if it should stop trying to clean the ring if we lost
- * the virtchnl. We need to stop playing with registers and
- * yield.
- */
- err = -ENXIO;
- goto out_unlock;
- case IDPF_VC_XN_ASYNC:
- err = idpf_vc_xn_forward_async(adapter, xn, ctlq_msg);
- idpf_vc_xn_unlock(xn);
- return err;
- default:
- dev_err_ratelimited(&adapter->pdev->dev, "Overwriting VC reply (op %d)\n",
- ctlq_msg->cookie.mbx.chnl_opcode);
- err = -EBUSY;
- goto out_unlock;
- }
-
- if (ctlq_msg->cookie.mbx.chnl_opcode != xn->vc_op) {
- dev_err_ratelimited(&adapter->pdev->dev, "Message opcode does not match transaction opcode (msg: %d) (xn: %d)\n",
- ctlq_msg->cookie.mbx.chnl_opcode, xn->vc_op);
- xn->reply_sz = 0;
- xn->state = IDPF_VC_XN_COMPLETED_FAILED;
- err = -EINVAL;
- goto out_unlock;
- }
-
- if (ctlq_msg->cookie.mbx.chnl_retval) {
- xn->reply_sz = 0;
- xn->state = IDPF_VC_XN_COMPLETED_FAILED;
- err = -EINVAL;
- goto out_unlock;
- }
-
- if (ctlq_msg->data_len) {
- payload = ctlq_msg->ctx.indirect.payload->va;
- payload_size = ctlq_msg->data_len;
- }
-
- xn->reply_sz = payload_size;
- xn->state = IDPF_VC_XN_COMPLETED_SUCCESS;
-
- if (xn->reply.iov_base && xn->reply.iov_len && payload_size)
- memcpy(xn->reply.iov_base, payload,
- min_t(size_t, xn->reply.iov_len, payload_size));
-
-out_unlock:
- idpf_vc_xn_unlock(xn);
- /* we _cannot_ hold lock while calling complete */
- complete(&xn->completed);
-
- return err;
-}
-
-/**
- * idpf_recv_mb_msg - Receive message over mailbox
- * @adapter: Driver specific private structure
- *
- * Will receive control queue message and posts the receive buffer. Returns 0
- * on success and negative on failure.
- */
-int idpf_recv_mb_msg(struct idpf_adapter *adapter)
-{
- struct idpf_ctlq_msg ctlq_msg;
- struct idpf_dma_mem *dma_mem;
- int post_err, err;
- u16 num_recv;
-
- while (1) {
- /* This will get <= num_recv messages and output how many
- * actually received on num_recv.
- */
- num_recv = 1;
- err = idpf_ctlq_recv(adapter->hw.arq, &num_recv, &ctlq_msg);
- if (err || !num_recv)
- break;
-
- if (ctlq_msg.data_len) {
- dma_mem = ctlq_msg.ctx.indirect.payload;
- } else {
- dma_mem = NULL;
- num_recv = 0;
- }
-
- if (ctlq_msg.cookie.mbx.chnl_opcode == VIRTCHNL2_OP_EVENT)
- idpf_recv_event_msg(adapter, &ctlq_msg);
- else
- err = idpf_vc_xn_forward_reply(adapter, &ctlq_msg);
-
- post_err = idpf_ctlq_post_rx_buffs(&adapter->hw,
- adapter->hw.arq,
- &num_recv, &dma_mem);
-
- /* If post failed clear the only buffer we supplied */
- if (post_err) {
- if (dma_mem)
- dmam_free_coherent(&adapter->pdev->dev,
- dma_mem->size, dma_mem->va,
- dma_mem->pa);
- break;
- }
-
- /* virtchnl trying to shutdown, stop cleaning */
- if (err == -ENXIO)
- break;
- }
-
- return err;
+ return libeth_ctlq_xn_send(xn_params);
}
/**
@@ -782,11 +203,14 @@ static int idpf_wait_for_marker_event(struct idpf_vport *vport)
*/
static int idpf_send_ver_msg(struct idpf_adapter *adapter)
{
- struct idpf_vc_xn_params xn_params = {};
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC,
+ .chnl_opcode = VIRTCHNL2_OP_VERSION,
+ };
+ struct virtchnl2_version_info *vvi_recv;
struct virtchnl2_version_info vvi;
- ssize_t reply_sz;
u32 major, minor;
- int err = 0;
+ int err;
if (adapter->virt_ver_maj) {
vvi.major = cpu_to_le32(adapter->virt_ver_maj);
@@ -796,24 +220,18 @@ static int idpf_send_ver_msg(struct idpf_adapter *adapter)
vvi.minor = cpu_to_le32(IDPF_VIRTCHNL_VERSION_MINOR);
}
- xn_params.vc_op = VIRTCHNL2_OP_VERSION;
- xn_params.send_buf.iov_base = &vvi;
- xn_params.send_buf.iov_len = sizeof(vvi);
- xn_params.recv_buf = xn_params.send_buf;
- xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
-
- reply_sz = idpf_vc_xn_exec(adapter, &xn_params);
- if (reply_sz < 0)
- return reply_sz;
- if (reply_sz < sizeof(vvi))
- return -EIO;
+ err = idpf_send_mb_msg(adapter, &xn_params, &vvi, sizeof(vvi));
+ if (err)
+ return err;
- major = le32_to_cpu(vvi.major);
- minor = le32_to_cpu(vvi.minor);
+ vvi_recv = xn_params.recv_mem.iov_base;
+ major = le32_to_cpu(vvi_recv->major);
+ minor = le32_to_cpu(vvi_recv->minor);
if (major > IDPF_VIRTCHNL_VERSION_MAJOR) {
dev_warn(&adapter->pdev->dev, "Virtchnl major version greater than supported\n");
- return -EINVAL;
+ err = -EINVAL;
+ goto free_rx_buf;
}
if (major == IDPF_VIRTCHNL_VERSION_MAJOR &&
@@ -831,6 +249,9 @@ static int idpf_send_ver_msg(struct idpf_adapter *adapter)
adapter->virt_ver_maj = major;
adapter->virt_ver_min = minor;
+free_rx_buf:
+ libeth_ctlq_release_rx_buf(adapter->arq, &xn_params.recv_mem);
+
return err;
}
@@ -843,9 +264,12 @@ static int idpf_send_ver_msg(struct idpf_adapter *adapter)
*/
static int idpf_send_get_caps_msg(struct idpf_adapter *adapter)
{
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC,
+ .chnl_opcode = VIRTCHNL2_OP_GET_CAPS,
+ };
struct virtchnl2_get_capabilities caps = {};
- struct idpf_vc_xn_params xn_params = {};
- ssize_t reply_sz;
+ int err;
caps.csum_caps =
cpu_to_le32(VIRTCHNL2_CAP_TX_CSUM_L3_IPV4 |
@@ -902,18 +326,14 @@ static int idpf_send_get_caps_msg(struct idpf_adapter *adapter)
VIRTCHNL2_CAP_PROMISC |
VIRTCHNL2_CAP_LOOPBACK);
- xn_params.vc_op = VIRTCHNL2_OP_GET_CAPS;
- xn_params.send_buf.iov_base = ∩︀
- xn_params.send_buf.iov_len = sizeof(caps);
- xn_params.recv_buf.iov_base = &adapter->caps;
- xn_params.recv_buf.iov_len = sizeof(adapter->caps);
- xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ err = idpf_send_mb_msg(adapter, &xn_params, &caps, sizeof(caps));
+ if (err)
+ return err;
- reply_sz = idpf_vc_xn_exec(adapter, &xn_params);
- if (reply_sz < 0)
- return reply_sz;
- if (reply_sz < sizeof(adapter->caps))
- return -EIO;
+ memcpy(&adapter->caps, xn_params.recv_mem.iov_base,
+ sizeof(adapter->caps));
+
+ libeth_ctlq_release_rx_buf(adapter->arq, &xn_params.recv_mem);
return 0;
}
@@ -1098,8 +518,10 @@ static int __idpf_queue_reg_init(struct idpf_vport *vport, u32 *reg_vals,
int num_regs, u32 q_type)
{
struct idpf_adapter *adapter = vport->adapter;
+ struct libeth_mmio_info *mmio;
int i, j, k = 0;
+ mmio = &adapter->ctlq_ctx.mmio_info;
switch (q_type) {
case VIRTCHNL2_QUEUE_TYPE_TX:
for (i = 0; i < vport->num_txq_grp; i++) {
@@ -1107,7 +529,8 @@ static int __idpf_queue_reg_init(struct idpf_vport *vport, u32 *reg_vals,
for (j = 0; j < tx_qgrp->num_txq && k < num_regs; j++, k++)
tx_qgrp->txqs[j]->tail =
- idpf_get_reg_addr(adapter, reg_vals[k]);
+ libeth_pci_get_mmio_addr(mmio,
+ reg_vals[k]);
}
break;
case VIRTCHNL2_QUEUE_TYPE_RX:
@@ -1119,8 +542,8 @@ static int __idpf_queue_reg_init(struct idpf_vport *vport, u32 *reg_vals,
struct idpf_rx_queue *q;
q = rx_qgrp->singleq.rxqs[j];
- q->tail = idpf_get_reg_addr(adapter,
- reg_vals[k]);
+ q->tail = libeth_pci_get_mmio_addr(mmio,
+ reg_vals[k]);
}
}
break;
@@ -1133,8 +556,8 @@ static int __idpf_queue_reg_init(struct idpf_vport *vport, u32 *reg_vals,
struct idpf_buf_queue *q;
q = &rx_qgrp->splitq.bufq_sets[j].bufq;
- q->tail = idpf_get_reg_addr(adapter,
- reg_vals[k]);
+ q->tail = libeth_pci_get_mmio_addr(mmio,
+ reg_vals[k]);
}
}
break;
@@ -1244,11 +667,13 @@ int idpf_queue_reg_init(struct idpf_vport *vport)
int idpf_send_create_vport_msg(struct idpf_adapter *adapter,
struct idpf_vport_max_q *max_q)
{
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC,
+ .chnl_opcode = VIRTCHNL2_OP_CREATE_VPORT,
+ };
struct virtchnl2_create_vport *vport_msg;
- struct idpf_vc_xn_params xn_params = {};
u16 idx = adapter->next_vport;
int err, buf_size;
- ssize_t reply_sz;
buf_size = sizeof(struct virtchnl2_create_vport);
vport_msg = kzalloc(buf_size, GFP_KERNEL);
@@ -1275,27 +700,23 @@ int idpf_send_create_vport_msg(struct idpf_adapter *adapter,
}
if (!adapter->vport_params_recvd[idx]) {
- adapter->vport_params_recvd[idx] = kzalloc(IDPF_CTLQ_MAX_BUF_LEN,
- GFP_KERNEL);
+ adapter->vport_params_recvd[idx] =
+ kzalloc(LIBETH_CTLQ_MAX_BUF_LEN, GFP_KERNEL);
if (!adapter->vport_params_recvd[idx]) {
err = -ENOMEM;
goto rel_buf;
}
}
- xn_params.vc_op = VIRTCHNL2_OP_CREATE_VPORT;
- xn_params.send_buf.iov_base = vport_msg;
- xn_params.send_buf.iov_len = buf_size;
- xn_params.recv_buf.iov_base = adapter->vport_params_recvd[idx];
- xn_params.recv_buf.iov_len = IDPF_CTLQ_MAX_BUF_LEN;
- xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
- reply_sz = idpf_vc_xn_exec(adapter, &xn_params);
- if (reply_sz < 0) {
- err = reply_sz;
+ err = idpf_send_mb_msg(adapter, &xn_params, vport_msg,
+ sizeof(*vport_msg));
+ if (err)
goto free_vport_params;
- }
- kfree(vport_msg);
+ memcpy(adapter->vport_params_recvd[idx], xn_params.recv_mem.iov_base,
+ xn_params.recv_mem.iov_len);
+
+ libeth_ctlq_release_rx_buf(adapter->arq, &xn_params.recv_mem);
return 0;
@@ -1362,19 +783,23 @@ int idpf_check_supported_desc_ids(struct idpf_vport *vport)
*/
int idpf_send_destroy_vport_msg(struct idpf_vport *vport)
{
- struct idpf_vc_xn_params xn_params = {};
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .timeout_ms = IDPF_VC_XN_MIN_TIMEOUT_MSEC,
+ .chnl_opcode = VIRTCHNL2_OP_DESTROY_VPORT,
+ };
struct virtchnl2_vport v_id;
- ssize_t reply_sz;
+ int err;
v_id.vport_id = cpu_to_le32(vport->vport_id);
- xn_params.vc_op = VIRTCHNL2_OP_DESTROY_VPORT;
- xn_params.send_buf.iov_base = &v_id;
- xn_params.send_buf.iov_len = sizeof(v_id);
- xn_params.timeout_ms = IDPF_VC_XN_MIN_TIMEOUT_MSEC;
- reply_sz = idpf_vc_xn_exec(vport->adapter, &xn_params);
+ err = idpf_send_mb_msg(vport->adapter, &xn_params, &v_id, sizeof(v_id));
+ if (err)
+ return err;
+
+ libeth_ctlq_release_rx_buf(vport->adapter->arq,
+ &xn_params.recv_mem);
- return reply_sz < 0 ? reply_sz : 0;
+ return 0;
}
/**
@@ -1386,19 +811,23 @@ int idpf_send_destroy_vport_msg(struct idpf_vport *vport)
*/
int idpf_send_enable_vport_msg(struct idpf_vport *vport)
{
- struct idpf_vc_xn_params xn_params = {};
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC,
+ .chnl_opcode = VIRTCHNL2_OP_ENABLE_VPORT,
+ };
struct virtchnl2_vport v_id;
- ssize_t reply_sz;
+ int err;
v_id.vport_id = cpu_to_le32(vport->vport_id);
- xn_params.vc_op = VIRTCHNL2_OP_ENABLE_VPORT;
- xn_params.send_buf.iov_base = &v_id;
- xn_params.send_buf.iov_len = sizeof(v_id);
- xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
- reply_sz = idpf_vc_xn_exec(vport->adapter, &xn_params);
+ err = idpf_send_mb_msg(vport->adapter, &xn_params, &v_id, sizeof(v_id));
+ if (err)
+ return err;
+
+ libeth_ctlq_release_rx_buf(vport->adapter->arq,
+ &xn_params.recv_mem);
- return reply_sz < 0 ? reply_sz : 0;
+ return 0;
}
/**
@@ -1410,19 +839,23 @@ int idpf_send_enable_vport_msg(struct idpf_vport *vport)
*/
int idpf_send_disable_vport_msg(struct idpf_vport *vport)
{
- struct idpf_vc_xn_params xn_params = {};
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .timeout_ms = IDPF_VC_XN_MIN_TIMEOUT_MSEC,
+ .chnl_opcode = VIRTCHNL2_OP_DISABLE_VPORT,
+ };
struct virtchnl2_vport v_id;
- ssize_t reply_sz;
+ int err;
v_id.vport_id = cpu_to_le32(vport->vport_id);
- xn_params.vc_op = VIRTCHNL2_OP_DISABLE_VPORT;
- xn_params.send_buf.iov_base = &v_id;
- xn_params.send_buf.iov_len = sizeof(v_id);
- xn_params.timeout_ms = IDPF_VC_XN_MIN_TIMEOUT_MSEC;
- reply_sz = idpf_vc_xn_exec(vport->adapter, &xn_params);
+ err = idpf_send_mb_msg(vport->adapter, &xn_params, &v_id, sizeof(v_id));
+ if (err)
+ return err;
- return reply_sz < 0 ? reply_sz : 0;
+ libeth_ctlq_release_rx_buf(vport->adapter->arq,
+ &xn_params.recv_mem);
+
+ return 0;
}
/**
@@ -1434,13 +867,15 @@ int idpf_send_disable_vport_msg(struct idpf_vport *vport)
*/
static int idpf_send_config_tx_queues_msg(struct idpf_vport *vport)
{
- struct virtchnl2_config_tx_queues *ctq __free(kfree) = NULL;
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC,
+ .chnl_opcode = VIRTCHNL2_OP_CONFIG_TX_QUEUES,
+ };
struct virtchnl2_txq_info *qi __free(kfree) = NULL;
- struct idpf_vc_xn_params xn_params = {};
+ struct virtchnl2_config_tx_queues *ctq;
u32 config_sz, chunk_sz, buf_sz;
int totqs, num_msgs, num_chunks;
- ssize_t reply_sz;
- int i, k = 0;
+ int i, k = 0, err = 0;
totqs = vport->num_txq + vport->num_complq;
qi = kcalloc(totqs, sizeof(struct virtchnl2_txq_info), GFP_KERNEL);
@@ -1514,34 +949,38 @@ static int idpf_send_config_tx_queues_msg(struct idpf_vport *vport)
totqs);
num_msgs = DIV_ROUND_UP(totqs, num_chunks);
- buf_sz = struct_size(ctq, qinfo, num_chunks);
- ctq = kzalloc(buf_sz, GFP_KERNEL);
- if (!ctq)
- return -ENOMEM;
-
- xn_params.vc_op = VIRTCHNL2_OP_CONFIG_TX_QUEUES;
- xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
-
for (i = 0, k = 0; i < num_msgs; i++) {
- memset(ctq, 0, buf_sz);
+ buf_sz = struct_size(ctq, qinfo, num_chunks);
+ ctq = kzalloc(buf_sz, GFP_KERNEL);
+ if (!ctq)
+ return -ENOMEM;
+
ctq->vport_id = cpu_to_le32(vport->vport_id);
ctq->num_qinfo = cpu_to_le16(num_chunks);
memcpy(ctq->qinfo, &qi[k], chunk_sz * num_chunks);
- xn_params.send_buf.iov_base = ctq;
- xn_params.send_buf.iov_len = buf_sz;
- reply_sz = idpf_vc_xn_exec(vport->adapter, &xn_params);
- if (reply_sz < 0)
- return reply_sz;
+ err = idpf_send_mb_msg(vport->adapter, &xn_params, ctq, buf_sz);
+ if (err)
+ goto rel_last_buf;
+
+ libeth_ctlq_release_rx_buf(vport->adapter->arq,
+ &xn_params.recv_mem);
k += num_chunks;
totqs -= num_chunks;
num_chunks = min(num_chunks, totqs);
- /* Recalculate buffer size */
- buf_sz = struct_size(ctq, qinfo, num_chunks);
}
- return 0;
+rel_last_buf:
+ /* Only the last buffer might be of size LIBETH_CP_TX_COPYBREAK or less.
+ * For buffers larger than LIBETH_CP_TX_COPYBREAK, are DMA mapped
+ * and released on mailbox cleanup. Smaller buffers are memcopied into
+ * the pre-allocated DMA buffers and are released here.
+ */
+ if (num_msgs && libeth_cp_can_send_onstack(buf_sz))
+ kfree(ctq);
+
+ return err;
}
/**
@@ -1553,13 +992,15 @@ static int idpf_send_config_tx_queues_msg(struct idpf_vport *vport)
*/
static int idpf_send_config_rx_queues_msg(struct idpf_vport *vport)
{
- struct virtchnl2_config_rx_queues *crq __free(kfree) = NULL;
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC,
+ .chnl_opcode = VIRTCHNL2_OP_CONFIG_RX_QUEUES,
+ };
struct virtchnl2_rxq_info *qi __free(kfree) = NULL;
- struct idpf_vc_xn_params xn_params = {};
+ struct virtchnl2_config_rx_queues *crq;
u32 config_sz, chunk_sz, buf_sz;
int totqs, num_msgs, num_chunks;
- ssize_t reply_sz;
- int i, k = 0;
+ int i, k = 0, err = 0;
totqs = vport->num_rxq + vport->num_bufq;
qi = kcalloc(totqs, sizeof(struct virtchnl2_rxq_info), GFP_KERNEL);
@@ -1666,34 +1107,33 @@ static int idpf_send_config_rx_queues_msg(struct idpf_vport *vport)
totqs);
num_msgs = DIV_ROUND_UP(totqs, num_chunks);
- buf_sz = struct_size(crq, qinfo, num_chunks);
- crq = kzalloc(buf_sz, GFP_KERNEL);
- if (!crq)
- return -ENOMEM;
-
- xn_params.vc_op = VIRTCHNL2_OP_CONFIG_RX_QUEUES;
- xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
-
for (i = 0, k = 0; i < num_msgs; i++) {
- memset(crq, 0, buf_sz);
+ buf_sz = struct_size(crq, qinfo, num_chunks);
+ crq = kzalloc(buf_sz, GFP_KERNEL);
+ if (!crq)
+ return -ENOMEM;
+
crq->vport_id = cpu_to_le32(vport->vport_id);
crq->num_qinfo = cpu_to_le16(num_chunks);
memcpy(crq->qinfo, &qi[k], chunk_sz * num_chunks);
- xn_params.send_buf.iov_base = crq;
- xn_params.send_buf.iov_len = buf_sz;
- reply_sz = idpf_vc_xn_exec(vport->adapter, &xn_params);
- if (reply_sz < 0)
- return reply_sz;
+ err = idpf_send_mb_msg(vport->adapter, &xn_params, crq, buf_sz);
+ if (err)
+ goto rel_last_buf;
+
+ libeth_ctlq_release_rx_buf(vport->adapter->arq,
+ &xn_params.recv_mem);
k += num_chunks;
totqs -= num_chunks;
num_chunks = min(num_chunks, totqs);
- /* Recalculate buffer size */
- buf_sz = struct_size(crq, qinfo, num_chunks);
}
- return 0;
+rel_last_buf:
+ if (num_msgs && libeth_cp_can_send_onstack(buf_sz))
+ kfree(crq);
+
+ return err;
}
/**
@@ -1707,14 +1147,18 @@ static int idpf_send_config_rx_queues_msg(struct idpf_vport *vport)
*/
static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, bool ena)
{
- struct virtchnl2_del_ena_dis_queues *eq __free(kfree) = NULL;
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .timeout_ms = ena ? IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC :
+ IDPF_VC_XN_MIN_TIMEOUT_MSEC,
+ .chnl_opcode = ena ? VIRTCHNL2_OP_ENABLE_QUEUES :
+ VIRTCHNL2_OP_DISABLE_QUEUES,
+ };
struct virtchnl2_queue_chunk *qc __free(kfree) = NULL;
u32 num_msgs, num_chunks, num_txq, num_rxq, num_q;
- struct idpf_vc_xn_params xn_params = {};
+ struct virtchnl2_del_ena_dis_queues *eq;
struct virtchnl2_queue_chunks *qcs;
u32 config_sz, chunk_sz, buf_sz;
- ssize_t reply_sz;
- int i, j, k = 0;
+ int i, j, k = 0, err = 0;
num_txq = vport->num_txq + vport->num_complq;
num_rxq = vport->num_rxq + vport->num_bufq;
@@ -1806,40 +1250,34 @@ static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, bool ena)
num_q);
num_msgs = DIV_ROUND_UP(num_q, num_chunks);
- buf_sz = struct_size(eq, chunks.chunks, num_chunks);
- eq = kzalloc(buf_sz, GFP_KERNEL);
- if (!eq)
- return -ENOMEM;
-
- if (ena) {
- xn_params.vc_op = VIRTCHNL2_OP_ENABLE_QUEUES;
- xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
- } else {
- xn_params.vc_op = VIRTCHNL2_OP_DISABLE_QUEUES;
- xn_params.timeout_ms = IDPF_VC_XN_MIN_TIMEOUT_MSEC;
- }
-
for (i = 0, k = 0; i < num_msgs; i++) {
- memset(eq, 0, buf_sz);
+ buf_sz = struct_size(eq, chunks.chunks, num_chunks);
+ eq = kzalloc(buf_sz, GFP_KERNEL);
+ if (!eq)
+ return -ENOMEM;
+
eq->vport_id = cpu_to_le32(vport->vport_id);
eq->chunks.num_chunks = cpu_to_le16(num_chunks);
qcs = &eq->chunks;
memcpy(qcs->chunks, &qc[k], chunk_sz * num_chunks);
- xn_params.send_buf.iov_base = eq;
- xn_params.send_buf.iov_len = buf_sz;
- reply_sz = idpf_vc_xn_exec(vport->adapter, &xn_params);
- if (reply_sz < 0)
- return reply_sz;
+ err = idpf_send_mb_msg(vport->adapter, &xn_params, eq, buf_sz);
+ if (err)
+ goto rel_last_buf;
+
+ libeth_ctlq_release_rx_buf(vport->adapter->arq,
+ &xn_params.recv_mem);
k += num_chunks;
num_q -= num_chunks;
num_chunks = min(num_chunks, num_q);
- /* Recalculate buffer size */
- buf_sz = struct_size(eq, chunks.chunks, num_chunks);
}
- return 0;
+rel_last_buf:
+ if (num_msgs && libeth_cp_can_send_onstack(buf_sz))
+ kfree(eq);
+
+ return err;
}
/**
@@ -1853,13 +1291,17 @@ static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, bool ena)
*/
int idpf_send_map_unmap_queue_vector_msg(struct idpf_vport *vport, bool map)
{
- struct virtchnl2_queue_vector_maps *vqvm __free(kfree) = NULL;
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .timeout_ms = map ? IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC :
+ IDPF_VC_XN_MIN_TIMEOUT_MSEC,
+ .chnl_opcode = map ? VIRTCHNL2_OP_MAP_QUEUE_VECTOR :
+ VIRTCHNL2_OP_UNMAP_QUEUE_VECTOR,
+ };
struct virtchnl2_queue_vector *vqv __free(kfree) = NULL;
- struct idpf_vc_xn_params xn_params = {};
+ struct virtchnl2_queue_vector_maps *vqvm;
u32 config_sz, chunk_sz, buf_sz;
u32 num_msgs, num_chunks, num_q;
- ssize_t reply_sz;
- int i, j, k = 0;
+ int i, j, k = 0, err = 0;
num_q = vport->num_txq + vport->num_rxq;
@@ -1934,39 +1376,34 @@ int idpf_send_map_unmap_queue_vector_msg(struct idpf_vport *vport, bool map)
num_q);
num_msgs = DIV_ROUND_UP(num_q, num_chunks);
- buf_sz = struct_size(vqvm, qv_maps, num_chunks);
- vqvm = kzalloc(buf_sz, GFP_KERNEL);
- if (!vqvm)
- return -ENOMEM;
-
- if (map) {
- xn_params.vc_op = VIRTCHNL2_OP_MAP_QUEUE_VECTOR;
- xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
- } else {
- xn_params.vc_op = VIRTCHNL2_OP_UNMAP_QUEUE_VECTOR;
- xn_params.timeout_ms = IDPF_VC_XN_MIN_TIMEOUT_MSEC;
- }
-
for (i = 0, k = 0; i < num_msgs; i++) {
- memset(vqvm, 0, buf_sz);
- xn_params.send_buf.iov_base = vqvm;
- xn_params.send_buf.iov_len = buf_sz;
+ buf_sz = struct_size(vqvm, qv_maps, num_chunks);
+ vqvm = kzalloc(buf_sz, GFP_KERNEL);
+ if (!vqvm)
+ return -ENOMEM;
+
vqvm->vport_id = cpu_to_le32(vport->vport_id);
vqvm->num_qv_maps = cpu_to_le16(num_chunks);
memcpy(vqvm->qv_maps, &vqv[k], chunk_sz * num_chunks);
- reply_sz = idpf_vc_xn_exec(vport->adapter, &xn_params);
- if (reply_sz < 0)
- return reply_sz;
+ err = idpf_send_mb_msg(vport->adapter, &xn_params, vqvm,
+ buf_sz);
+ if (err)
+ goto rel_last_buf;
+
+ libeth_ctlq_release_rx_buf(vport->adapter->arq,
+ &xn_params.recv_mem);
k += num_chunks;
num_q -= num_chunks;
num_chunks = min(num_chunks, num_q);
- /* Recalculate buffer size */
- buf_sz = struct_size(vqvm, qv_maps, num_chunks);
}
- return 0;
+rel_last_buf:
+ if (num_msgs && libeth_cp_can_send_onstack(buf_sz))
+ kfree(vqvm);
+
+ return err;
}
/**
@@ -2040,21 +1477,24 @@ static void idpf_convert_reg_to_queue_chunks(struct virtchnl2_queue_chunk *dchun
*/
int idpf_send_delete_queues_msg(struct idpf_vport *vport)
{
- struct virtchnl2_del_ena_dis_queues *eq __free(kfree) = NULL;
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .timeout_ms = IDPF_VC_XN_MIN_TIMEOUT_MSEC,
+ .chnl_opcode = VIRTCHNL2_OP_DEL_QUEUES,
+ };
+ struct idpf_adapter *adapter = vport->adapter;
struct virtchnl2_create_vport *vport_params;
struct virtchnl2_queue_reg_chunks *chunks;
- struct idpf_vc_xn_params xn_params = {};
+ struct virtchnl2_del_ena_dis_queues *eq;
struct idpf_vport_config *vport_config;
u16 vport_idx = vport->idx;
- ssize_t reply_sz;
+ int buf_size, err;
u16 num_chunks;
- int buf_size;
- vport_config = vport->adapter->vport_config[vport_idx];
+ vport_config = adapter->vport_config[vport_idx];
if (vport_config->req_qs_chunks) {
chunks = &vport_config->req_qs_chunks->chunks;
} else {
- vport_params = vport->adapter->vport_params_recvd[vport_idx];
+ vport_params = adapter->vport_params_recvd[vport_idx];
chunks = &vport_params->chunks;
}
@@ -2071,13 +1511,17 @@ int idpf_send_delete_queues_msg(struct idpf_vport *vport)
idpf_convert_reg_to_queue_chunks(eq->chunks.chunks, chunks->chunks,
num_chunks);
- xn_params.vc_op = VIRTCHNL2_OP_DEL_QUEUES;
- xn_params.timeout_ms = IDPF_VC_XN_MIN_TIMEOUT_MSEC;
- xn_params.send_buf.iov_base = eq;
- xn_params.send_buf.iov_len = buf_size;
- reply_sz = idpf_vc_xn_exec(vport->adapter, &xn_params);
+ err = idpf_send_mb_msg(adapter, &xn_params, eq, buf_size);
+ if (err)
+ goto rel_buf;
+
+ libeth_ctlq_release_rx_buf(adapter->arq, &xn_params.recv_mem);
+
+rel_buf:
+ if (libeth_cp_can_send_onstack(buf_size))
+ kfree(eq);
- return reply_sz < 0 ? reply_sz : 0;
+ return err;
}
/**
@@ -2112,19 +1556,18 @@ int idpf_send_config_queues_msg(struct idpf_vport *vport)
int idpf_send_add_queues_msg(const struct idpf_vport *vport, u16 num_tx_q,
u16 num_complq, u16 num_rx_q, u16 num_rx_bufq)
{
- struct virtchnl2_add_queues *vc_msg __free(kfree) = NULL;
- struct idpf_vc_xn_params xn_params = {};
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC,
+ .chnl_opcode = VIRTCHNL2_OP_ADD_QUEUES,
+ };
+ struct idpf_adapter *adapter = vport->adapter;
struct idpf_vport_config *vport_config;
+ struct virtchnl2_add_queues *vc_msg;
struct virtchnl2_add_queues aq = {};
u16 vport_idx = vport->idx;
- ssize_t reply_sz;
- int size;
+ int size, err;
- vc_msg = kzalloc(IDPF_CTLQ_MAX_BUF_LEN, GFP_KERNEL);
- if (!vc_msg)
- return -ENOMEM;
-
- vport_config = vport->adapter->vport_config[vport_idx];
+ vport_config = adapter->vport_config[vport_idx];
kfree(vport_config->req_qs_chunks);
vport_config->req_qs_chunks = NULL;
@@ -2134,33 +1577,32 @@ int idpf_send_add_queues_msg(const struct idpf_vport *vport, u16 num_tx_q,
aq.num_rx_q = cpu_to_le16(num_rx_q);
aq.num_rx_bufq = cpu_to_le16(num_rx_bufq);
- xn_params.vc_op = VIRTCHNL2_OP_ADD_QUEUES;
- xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
- xn_params.send_buf.iov_base = &aq;
- xn_params.send_buf.iov_len = sizeof(aq);
- xn_params.recv_buf.iov_base = vc_msg;
- xn_params.recv_buf.iov_len = IDPF_CTLQ_MAX_BUF_LEN;
- reply_sz = idpf_vc_xn_exec(vport->adapter, &xn_params);
- if (reply_sz < 0)
- return reply_sz;
+ err = idpf_send_mb_msg(adapter, &xn_params, &aq, sizeof(aq));
+ if (err)
+ return err;
+
+ vc_msg = xn_params.recv_mem.iov_base;
/* compare vc_msg num queues with vport num queues */
if (le16_to_cpu(vc_msg->num_tx_q) != num_tx_q ||
le16_to_cpu(vc_msg->num_rx_q) != num_rx_q ||
le16_to_cpu(vc_msg->num_tx_complq) != num_complq ||
- le16_to_cpu(vc_msg->num_rx_bufq) != num_rx_bufq)
- return -EINVAL;
+ le16_to_cpu(vc_msg->num_rx_bufq) != num_rx_bufq) {
+ err = -EINVAL;
+ goto free_rx_buf;
+ }
size = struct_size(vc_msg, chunks.chunks,
le16_to_cpu(vc_msg->chunks.num_chunks));
- if (reply_sz < size)
- return -EIO;
vport_config->req_qs_chunks = kmemdup(vc_msg, size, GFP_KERNEL);
if (!vport_config->req_qs_chunks)
- return -ENOMEM;
+ err = -ENOMEM;
- return 0;
+free_rx_buf:
+ libeth_ctlq_release_rx_buf(adapter->arq, &xn_params.recv_mem);
+
+ return err;
}
/**
@@ -2172,49 +1614,43 @@ int idpf_send_add_queues_msg(const struct idpf_vport *vport, u16 num_tx_q,
*/
int idpf_send_alloc_vectors_msg(struct idpf_adapter *adapter, u16 num_vectors)
{
- struct virtchnl2_alloc_vectors *rcvd_vec __free(kfree) = NULL;
- struct idpf_vc_xn_params xn_params = {};
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC,
+ .chnl_opcode = VIRTCHNL2_OP_ALLOC_VECTORS,
+ };
+ struct virtchnl2_alloc_vectors *rcvd_vec;
struct virtchnl2_alloc_vectors ac = {};
- ssize_t reply_sz;
u16 num_vchunks;
- int size;
+ int size, err;
ac.num_vectors = cpu_to_le16(num_vectors);
- rcvd_vec = kzalloc(IDPF_CTLQ_MAX_BUF_LEN, GFP_KERNEL);
- if (!rcvd_vec)
- return -ENOMEM;
+ err = idpf_send_mb_msg(adapter, &xn_params, &ac, sizeof(ac));
+ if (err)
+ return err;
- xn_params.vc_op = VIRTCHNL2_OP_ALLOC_VECTORS;
- xn_params.send_buf.iov_base = ∾
- xn_params.send_buf.iov_len = sizeof(ac);
- xn_params.recv_buf.iov_base = rcvd_vec;
- xn_params.recv_buf.iov_len = IDPF_CTLQ_MAX_BUF_LEN;
- xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
- reply_sz = idpf_vc_xn_exec(adapter, &xn_params);
- if (reply_sz < 0)
- return reply_sz;
+ rcvd_vec = xn_params.recv_mem.iov_base;
num_vchunks = le16_to_cpu(rcvd_vec->vchunks.num_vchunks);
size = struct_size(rcvd_vec, vchunks.vchunks, num_vchunks);
- if (reply_sz < size)
- return -EIO;
-
- if (size > IDPF_CTLQ_MAX_BUF_LEN)
- return -EINVAL;
kfree(adapter->req_vec_chunks);
adapter->req_vec_chunks = kmemdup(rcvd_vec, size, GFP_KERNEL);
- if (!adapter->req_vec_chunks)
- return -ENOMEM;
+ if (!adapter->req_vec_chunks) {
+ err = -ENOMEM;
+ goto free_rx_buf;
+ }
if (le16_to_cpu(adapter->req_vec_chunks->num_vectors) < num_vectors) {
kfree(adapter->req_vec_chunks);
adapter->req_vec_chunks = NULL;
- return -EINVAL;
+ err = -EINVAL;
}
- return 0;
+free_rx_buf:
+ libeth_ctlq_release_rx_buf(adapter->arq, &xn_params.recv_mem);
+
+ return err;
}
/**
@@ -2226,25 +1662,33 @@ int idpf_send_alloc_vectors_msg(struct idpf_adapter *adapter, u16 num_vectors)
int idpf_send_dealloc_vectors_msg(struct idpf_adapter *adapter)
{
struct virtchnl2_alloc_vectors *ac = adapter->req_vec_chunks;
- struct virtchnl2_vector_chunks *vcs = &ac->vchunks;
- struct idpf_vc_xn_params xn_params = {};
- ssize_t reply_sz;
- int buf_size;
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .timeout_ms = IDPF_VC_XN_MIN_TIMEOUT_MSEC,
+ .chnl_opcode = VIRTCHNL2_OP_DEALLOC_VECTORS,
+ };
+ struct virtchnl2_vector_chunks *vcs;
+ int buf_size, err;
- buf_size = struct_size(vcs, vchunks, le16_to_cpu(vcs->num_vchunks));
+ buf_size = struct_size(&ac->vchunks, vchunks,
+ le16_to_cpu(ac->vchunks.num_vchunks));
+ vcs = kmemdup(&ac->vchunks, buf_size, GFP_KERNEL);
+ if (!vcs)
+ return -ENOMEM;
- xn_params.vc_op = VIRTCHNL2_OP_DEALLOC_VECTORS;
- xn_params.send_buf.iov_base = vcs;
- xn_params.send_buf.iov_len = buf_size;
- xn_params.timeout_ms = IDPF_VC_XN_MIN_TIMEOUT_MSEC;
- reply_sz = idpf_vc_xn_exec(adapter, &xn_params);
- if (reply_sz < 0)
- return reply_sz;
+ err = idpf_send_mb_msg(adapter, &xn_params, vcs, buf_size);
+ if (err)
+ goto rel_buf;
kfree(adapter->req_vec_chunks);
adapter->req_vec_chunks = NULL;
- return 0;
+ libeth_ctlq_release_rx_buf(adapter->arq, &xn_params.recv_mem);
+
+rel_buf:
+ if (libeth_cp_can_send_onstack(buf_size))
+ kfree(vcs);
+
+ return err;
}
/**
@@ -2267,18 +1711,22 @@ static int idpf_get_max_vfs(struct idpf_adapter *adapter)
*/
int idpf_send_set_sriov_vfs_msg(struct idpf_adapter *adapter, u16 num_vfs)
{
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC,
+ .chnl_opcode = VIRTCHNL2_OP_SET_SRIOV_VFS,
+ };
struct virtchnl2_sriov_vfs_info svi = {};
- struct idpf_vc_xn_params xn_params = {};
- ssize_t reply_sz;
+ int err;
svi.num_vfs = cpu_to_le16(num_vfs);
- xn_params.vc_op = VIRTCHNL2_OP_SET_SRIOV_VFS;
- xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
- xn_params.send_buf.iov_base = &svi;
- xn_params.send_buf.iov_len = sizeof(svi);
- reply_sz = idpf_vc_xn_exec(adapter, &xn_params);
- return reply_sz < 0 ? reply_sz : 0;
+ err = idpf_send_mb_msg(adapter, &xn_params, &svi, sizeof(svi));
+ if (err)
+ return err;
+
+ libeth_ctlq_release_rx_buf(adapter->arq, &xn_params.recv_mem);
+
+ return 0;
}
/**
@@ -2289,50 +1737,51 @@ int idpf_send_set_sriov_vfs_msg(struct idpf_adapter *adapter, u16 num_vfs)
*/
int idpf_send_get_stats_msg(struct idpf_vport *vport)
{
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC,
+ .chnl_opcode = VIRTCHNL2_OP_GET_STATS,
+ };
struct idpf_netdev_priv *np = netdev_priv(vport->netdev);
struct rtnl_link_stats64 *netstats = &np->netstats;
- struct virtchnl2_vport_stats stats_msg = {};
- struct idpf_vc_xn_params xn_params = {};
- ssize_t reply_sz;
-
+ struct virtchnl2_vport_stats *stats_recv;
+ struct virtchnl2_vport_stats stats = {};
+ int err;
/* Don't send get_stats message if the link is down */
if (np->state <= __IDPF_VPORT_DOWN)
return 0;
- stats_msg.vport_id = cpu_to_le32(vport->vport_id);
-
- xn_params.vc_op = VIRTCHNL2_OP_GET_STATS;
- xn_params.send_buf.iov_base = &stats_msg;
- xn_params.send_buf.iov_len = sizeof(stats_msg);
- xn_params.recv_buf = xn_params.send_buf;
- xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ stats.vport_id = cpu_to_le32(vport->vport_id);
- reply_sz = idpf_vc_xn_exec(vport->adapter, &xn_params);
- if (reply_sz < 0)
- return reply_sz;
- if (reply_sz < sizeof(stats_msg))
- return -EIO;
+ err = idpf_send_mb_msg(vport->adapter, &xn_params, &stats,
+ sizeof(stats));
+ if (err)
+ return err;
spin_lock_bh(&np->stats_lock);
- netstats->rx_packets = le64_to_cpu(stats_msg.rx_unicast) +
- le64_to_cpu(stats_msg.rx_multicast) +
- le64_to_cpu(stats_msg.rx_broadcast);
- netstats->tx_packets = le64_to_cpu(stats_msg.tx_unicast) +
- le64_to_cpu(stats_msg.tx_multicast) +
- le64_to_cpu(stats_msg.tx_broadcast);
- netstats->rx_bytes = le64_to_cpu(stats_msg.rx_bytes);
- netstats->tx_bytes = le64_to_cpu(stats_msg.tx_bytes);
- netstats->rx_errors = le64_to_cpu(stats_msg.rx_errors);
- netstats->tx_errors = le64_to_cpu(stats_msg.tx_errors);
- netstats->rx_dropped = le64_to_cpu(stats_msg.rx_discards);
- netstats->tx_dropped = le64_to_cpu(stats_msg.tx_discards);
-
- vport->port_stats.vport_stats = stats_msg;
+ stats_recv = xn_params.recv_mem.iov_base;
+
+ netstats->rx_packets = le64_to_cpu(stats_recv->rx_unicast) +
+ le64_to_cpu(stats_recv->rx_multicast) +
+ le64_to_cpu(stats_recv->rx_broadcast);
+ netstats->tx_packets = le64_to_cpu(stats_recv->tx_unicast) +
+ le64_to_cpu(stats_recv->tx_multicast) +
+ le64_to_cpu(stats_recv->tx_broadcast);
+ netstats->rx_bytes = le64_to_cpu(stats_recv->rx_bytes);
+ netstats->tx_bytes = le64_to_cpu(stats_recv->tx_bytes);
+ netstats->rx_errors = le64_to_cpu(stats_recv->rx_errors);
+ netstats->tx_errors = le64_to_cpu(stats_recv->tx_errors);
+ netstats->rx_dropped = le64_to_cpu(stats_recv->rx_discards);
+ netstats->tx_dropped = le64_to_cpu(stats_recv->tx_discards);
+
+ vport->port_stats.vport_stats = *stats_recv;
spin_unlock_bh(&np->stats_lock);
+ libeth_ctlq_release_rx_buf(vport->adapter->arq,
+ &xn_params.recv_mem);
+
return 0;
}
@@ -2345,52 +1794,44 @@ int idpf_send_get_stats_msg(struct idpf_vport *vport)
*/
int idpf_send_get_set_rss_lut_msg(struct idpf_vport *vport, bool get)
{
- struct virtchnl2_rss_lut *recv_rl __free(kfree) = NULL;
- struct virtchnl2_rss_lut *rl __free(kfree) = NULL;
- struct idpf_vc_xn_params xn_params = {};
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC,
+ .chnl_opcode = get ? VIRTCHNL2_OP_GET_RSS_LUT :
+ VIRTCHNL2_OP_SET_RSS_LUT,
+ };
+ struct idpf_adapter *adapter = vport->adapter;
+ struct virtchnl2_rss_lut *rl, *recv_rl;
struct idpf_rss_data *rss_data;
int buf_size, lut_buf_size;
- ssize_t reply_sz;
- int i;
+ u16 recv_len;
+ int i, err;
- rss_data =
- &vport->adapter->vport_config[vport->idx]->user_config.rss_data;
+ rss_data = &adapter->vport_config[vport->idx]->user_config.rss_data;
buf_size = struct_size(rl, lut, rss_data->rss_lut_size);
rl = kzalloc(buf_size, GFP_KERNEL);
if (!rl)
return -ENOMEM;
rl->vport_id = cpu_to_le32(vport->vport_id);
+ rl->lut_entries = cpu_to_le16(rss_data->rss_lut_size);
+ for (i = 0; i < rss_data->rss_lut_size; i++)
+ rl->lut[i] = cpu_to_le32(rss_data->rss_lut[i]);
- xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
- xn_params.send_buf.iov_base = rl;
- xn_params.send_buf.iov_len = buf_size;
-
- if (get) {
- recv_rl = kzalloc(IDPF_CTLQ_MAX_BUF_LEN, GFP_KERNEL);
- if (!recv_rl)
- return -ENOMEM;
- xn_params.vc_op = VIRTCHNL2_OP_GET_RSS_LUT;
- xn_params.recv_buf.iov_base = recv_rl;
- xn_params.recv_buf.iov_len = IDPF_CTLQ_MAX_BUF_LEN;
- } else {
- rl->lut_entries = cpu_to_le16(rss_data->rss_lut_size);
- for (i = 0; i < rss_data->rss_lut_size; i++)
- rl->lut[i] = cpu_to_le32(rss_data->rss_lut[i]);
+ err = idpf_send_mb_msg(adapter, &xn_params, rl, buf_size);
+ if (err)
+ goto free_tx_buf;
- xn_params.vc_op = VIRTCHNL2_OP_SET_RSS_LUT;
- }
- reply_sz = idpf_vc_xn_exec(vport->adapter, &xn_params);
- if (reply_sz < 0)
- return reply_sz;
if (!get)
- return 0;
- if (reply_sz < sizeof(struct virtchnl2_rss_lut))
- return -EIO;
+ goto free_rx_buf;
+
+ recv_rl = xn_params.recv_mem.iov_base;
+ recv_len = xn_params.recv_mem.iov_len;
lut_buf_size = le16_to_cpu(recv_rl->lut_entries) * sizeof(u32);
- if (reply_sz < lut_buf_size)
- return -EIO;
+ if (recv_len < lut_buf_size) {
+ err = -EIO;
+ goto free_rx_buf;
+ }
/* size didn't change, we can reuse existing lut buf */
if (rss_data->rss_lut_size == le16_to_cpu(recv_rl->lut_entries))
@@ -2402,13 +1843,19 @@ int idpf_send_get_set_rss_lut_msg(struct idpf_vport *vport, bool get)
rss_data->rss_lut = kzalloc(lut_buf_size, GFP_KERNEL);
if (!rss_data->rss_lut) {
rss_data->rss_lut_size = 0;
- return -ENOMEM;
+ err = -ENOMEM;
+ goto free_rx_buf;
}
do_memcpy:
memcpy(rss_data->rss_lut, recv_rl->lut, rss_data->rss_lut_size);
+free_rx_buf:
+ libeth_ctlq_release_rx_buf(adapter->arq, &xn_params.recv_mem);
+free_tx_buf:
+ if (libeth_cp_can_send_onstack(buf_size))
+ kfree(rl);
- return 0;
+ return err;
}
/**
@@ -2420,53 +1867,44 @@ int idpf_send_get_set_rss_lut_msg(struct idpf_vport *vport, bool get)
*/
int idpf_send_get_set_rss_key_msg(struct idpf_vport *vport, bool get)
{
- struct virtchnl2_rss_key *recv_rk __free(kfree) = NULL;
- struct virtchnl2_rss_key *rk __free(kfree) = NULL;
- struct idpf_vc_xn_params xn_params = {};
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC,
+ .chnl_opcode = get ? VIRTCHNL2_OP_GET_RSS_KEY :
+ VIRTCHNL2_OP_SET_RSS_KEY,
+ };
+ struct idpf_adapter *adapter = vport->adapter;
+ struct virtchnl2_rss_key *rk, *recv_rk;
struct idpf_rss_data *rss_data;
- ssize_t reply_sz;
- int i, buf_size;
- u16 key_size;
+ u16 key_size, recv_len;
+ int i, buf_size, err;
- rss_data =
- &vport->adapter->vport_config[vport->idx]->user_config.rss_data;
+ rss_data = &adapter->vport_config[vport->idx]->user_config.rss_data;
buf_size = struct_size(rk, key_flex, rss_data->rss_key_size);
rk = kzalloc(buf_size, GFP_KERNEL);
if (!rk)
return -ENOMEM;
rk->vport_id = cpu_to_le32(vport->vport_id);
- xn_params.send_buf.iov_base = rk;
- xn_params.send_buf.iov_len = buf_size;
- xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
- if (get) {
- recv_rk = kzalloc(IDPF_CTLQ_MAX_BUF_LEN, GFP_KERNEL);
- if (!recv_rk)
- return -ENOMEM;
+ rk->key_len = cpu_to_le16(rss_data->rss_key_size);
+ for (i = 0; i < rss_data->rss_key_size; i++)
+ rk->key_flex[i] = rss_data->rss_key[i];
- xn_params.vc_op = VIRTCHNL2_OP_GET_RSS_KEY;
- xn_params.recv_buf.iov_base = recv_rk;
- xn_params.recv_buf.iov_len = IDPF_CTLQ_MAX_BUF_LEN;
- } else {
- rk->key_len = cpu_to_le16(rss_data->rss_key_size);
- for (i = 0; i < rss_data->rss_key_size; i++)
- rk->key_flex[i] = rss_data->rss_key[i];
-
- xn_params.vc_op = VIRTCHNL2_OP_SET_RSS_KEY;
- }
+ err = idpf_send_mb_msg(adapter, &xn_params, rk, buf_size);
+ if (err)
+ goto free_tx_buf;
- reply_sz = idpf_vc_xn_exec(vport->adapter, &xn_params);
- if (reply_sz < 0)
- return reply_sz;
if (!get)
- return 0;
- if (reply_sz < sizeof(struct virtchnl2_rss_key))
- return -EIO;
+ goto free_rx_buf;
+
+ recv_rk = xn_params.recv_mem.iov_base;
+ recv_len = xn_params.recv_mem.iov_len;
key_size = min_t(u16, NETDEV_RSS_KEY_LEN,
le16_to_cpu(recv_rk->key_len));
- if (reply_sz < key_size)
- return -EIO;
+ if (recv_len < key_size) {
+ err = -EIO;
+ goto free_rx_buf;
+ }
/* key len didn't change, reuse existing buf */
if (rss_data->rss_key_size == key_size)
@@ -2477,13 +1915,19 @@ int idpf_send_get_set_rss_key_msg(struct idpf_vport *vport, bool get)
rss_data->rss_key = kzalloc(key_size, GFP_KERNEL);
if (!rss_data->rss_key) {
rss_data->rss_key_size = 0;
- return -ENOMEM;
+ err = -ENOMEM;
+ goto free_rx_buf;
}
do_memcpy:
memcpy(rss_data->rss_key, recv_rk->key_flex, rss_data->rss_key_size);
+free_rx_buf:
+ libeth_ctlq_release_rx_buf(adapter->arq, &xn_params.recv_mem);
+free_tx_buf:
+ if (libeth_cp_can_send_onstack(buf_size))
+ kfree(rk);
- return 0;
+ return err;
}
/**
@@ -2548,15 +1992,18 @@ static void idpf_finalize_ptype_lookup(struct libeth_rx_pt *ptype)
*/
int idpf_send_get_rx_ptype_msg(struct idpf_vport *vport)
{
- struct virtchnl2_get_ptype_info *get_ptype_info __free(kfree) = NULL;
- struct virtchnl2_get_ptype_info *ptype_info __free(kfree) = NULL;
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC,
+ .chnl_opcode = VIRTCHNL2_OP_GET_PTYPE_INFO,
+ };
struct libeth_rx_pt *ptype_lkup __free(kfree) = NULL;
+ struct virtchnl2_get_ptype_info *get_ptype_info;
int max_ptype, ptypes_recvd = 0, ptype_offset;
struct idpf_adapter *adapter = vport->adapter;
- struct idpf_vc_xn_params xn_params = {};
+ struct virtchnl2_get_ptype_info *ptype_info;
+ int buf_size = sizeof(*get_ptype_info);
u16 next_ptype_id = 0;
- ssize_t reply_sz;
- int i, j, k;
+ int i, j, k, err = 0;
if (vport->rx_ptype_lkup)
return 0;
@@ -2570,22 +2017,11 @@ int idpf_send_get_rx_ptype_msg(struct idpf_vport *vport)
if (!ptype_lkup)
return -ENOMEM;
- get_ptype_info = kzalloc(sizeof(*get_ptype_info), GFP_KERNEL);
- if (!get_ptype_info)
- return -ENOMEM;
-
- ptype_info = kzalloc(IDPF_CTLQ_MAX_BUF_LEN, GFP_KERNEL);
- if (!ptype_info)
- return -ENOMEM;
-
- xn_params.vc_op = VIRTCHNL2_OP_GET_PTYPE_INFO;
- xn_params.send_buf.iov_base = get_ptype_info;
- xn_params.send_buf.iov_len = sizeof(*get_ptype_info);
- xn_params.recv_buf.iov_base = ptype_info;
- xn_params.recv_buf.iov_len = IDPF_CTLQ_MAX_BUF_LEN;
- xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
-
while (next_ptype_id < max_ptype) {
+ get_ptype_info = kzalloc(buf_size, GFP_KERNEL);
+ if (!get_ptype_info)
+ return -ENOMEM;
+
get_ptype_info->start_ptype_id = cpu_to_le16(next_ptype_id);
if ((next_ptype_id + IDPF_RX_MAX_PTYPES_PER_BUF) > max_ptype)
@@ -2595,13 +2031,15 @@ int idpf_send_get_rx_ptype_msg(struct idpf_vport *vport)
get_ptype_info->num_ptypes =
cpu_to_le16(IDPF_RX_MAX_PTYPES_PER_BUF);
- reply_sz = idpf_vc_xn_exec(adapter, &xn_params);
- if (reply_sz < 0)
- return reply_sz;
+ err = idpf_send_mb_msg(adapter, &xn_params, get_ptype_info,
+ buf_size);
+ if (err)
+ goto free_tx_buf;
+ ptype_info = xn_params.recv_mem.iov_base;
ptypes_recvd += le16_to_cpu(ptype_info->num_ptypes);
if (ptypes_recvd > max_ptype)
- return -EINVAL;
+ goto free_rx_buf;
next_ptype_id = le16_to_cpu(get_ptype_info->start_ptype_id) +
le16_to_cpu(get_ptype_info->num_ptypes);
@@ -2617,8 +2055,8 @@ int idpf_send_get_rx_ptype_msg(struct idpf_vport *vport)
((u8 *)ptype_info + ptype_offset);
ptype_offset += IDPF_GET_PTYPE_SIZE(ptype);
- if (ptype_offset > IDPF_CTLQ_MAX_BUF_LEN)
- return -EINVAL;
+ if (ptype_offset > LIBETH_CTLQ_MAX_BUF_LEN)
+ goto free_rx_buf;
/* 0xFFFF indicates end of ptypes */
if (le16_to_cpu(ptype->ptype_id_10) ==
@@ -2748,12 +2186,24 @@ int idpf_send_get_rx_ptype_msg(struct idpf_vport *vport)
idpf_finalize_ptype_lookup(&ptype_lkup[k]);
}
+ libeth_ctlq_release_rx_buf(adapter->arq,
+ &xn_params.recv_mem);
+ if (libeth_cp_can_send_onstack(buf_size))
+ kfree(get_ptype_info);
}
out:
vport->rx_ptype_lkup = no_free_ptr(ptype_lkup);
return 0;
+
+free_rx_buf:
+ libeth_ctlq_release_rx_buf(adapter->arq, &xn_params.recv_mem);
+free_tx_buf:
+ if (libeth_cp_can_send_onstack(buf_size))
+ kfree(get_ptype_info);
+
+ return err;
}
/**
@@ -2765,40 +2215,25 @@ int idpf_send_get_rx_ptype_msg(struct idpf_vport *vport)
*/
int idpf_send_ena_dis_loopback_msg(struct idpf_vport *vport)
{
- struct idpf_vc_xn_params xn_params = {};
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC,
+ .chnl_opcode = VIRTCHNL2_OP_LOOPBACK,
+ };
struct virtchnl2_loopback loopback;
- ssize_t reply_sz;
+ int err;
loopback.vport_id = cpu_to_le32(vport->vport_id);
loopback.enable = idpf_is_feature_ena(vport, NETIF_F_LOOPBACK);
- xn_params.vc_op = VIRTCHNL2_OP_LOOPBACK;
- xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
- xn_params.send_buf.iov_base = &loopback;
- xn_params.send_buf.iov_len = sizeof(loopback);
- reply_sz = idpf_vc_xn_exec(vport->adapter, &xn_params);
-
- return reply_sz < 0 ? reply_sz : 0;
-}
-
-/**
- * idpf_find_ctlq - Given a type and id, find ctlq info
- * @hw: hardware struct
- * @type: type of ctrlq to find
- * @id: ctlq id to find
- *
- * Returns pointer to found ctlq info struct, NULL otherwise.
- */
-static struct idpf_ctlq_info *idpf_find_ctlq(struct idpf_hw *hw,
- enum idpf_ctlq_type type, int id)
-{
- struct idpf_ctlq_info *cq, *tmp;
+ err = idpf_send_mb_msg(vport->adapter, &xn_params, &loopback,
+ sizeof(loopback));
+ if (err)
+ return err;
- list_for_each_entry_safe(cq, tmp, &hw->cq_list_head, cq_list)
- if (cq->q_id == id && cq->cq_type == type)
- return cq;
+ libeth_ctlq_release_rx_buf(vport->adapter->arq,
+ &xn_params.recv_mem);
- return NULL;
+ return 0;
}
/**
@@ -2809,40 +2244,43 @@ static struct idpf_ctlq_info *idpf_find_ctlq(struct idpf_hw *hw,
*/
int idpf_init_dflt_mbx(struct idpf_adapter *adapter)
{
- struct idpf_ctlq_create_info ctlq_info[] = {
+ struct libeth_ctlq_ctx *ctx = &adapter->ctlq_ctx;
+ struct libeth_ctlq_create_info ctlq_info[] = {
{
- .type = IDPF_CTLQ_TYPE_MAILBOX_TX,
- .id = IDPF_DFLT_MBX_ID,
+ .type = LIBETH_CTLQ_TYPE_TX,
+ .id = LIBETH_CTLQ_MBX_ID,
.len = IDPF_DFLT_MBX_Q_LEN,
- .buf_size = IDPF_CTLQ_MAX_BUF_LEN
},
{
- .type = IDPF_CTLQ_TYPE_MAILBOX_RX,
- .id = IDPF_DFLT_MBX_ID,
+ .type = LIBETH_CTLQ_TYPE_RX,
+ .id = LIBETH_CTLQ_MBX_ID,
.len = IDPF_DFLT_MBX_Q_LEN,
- .buf_size = IDPF_CTLQ_MAX_BUF_LEN
}
};
- struct idpf_hw *hw = &adapter->hw;
+ struct libeth_ctlq_xn_init_params params = {
+ .num_qs = IDPF_NUM_DFLT_MBX_Q,
+ .cctlq_info = ctlq_info,
+ .ctx = ctx,
+ };
int err;
- adapter->dev_ops.reg_ops.ctlq_reg_init(ctlq_info);
+ adapter->dev_ops.reg_ops.ctlq_reg_init(&ctx->mmio_info,
+ params.cctlq_info);
- err = idpf_ctlq_init(hw, IDPF_NUM_DFLT_MBX_Q, ctlq_info);
+ err = libeth_ctlq_xn_init(¶ms);
if (err)
return err;
- hw->asq = idpf_find_ctlq(hw, IDPF_CTLQ_TYPE_MAILBOX_TX,
- IDPF_DFLT_MBX_ID);
- hw->arq = idpf_find_ctlq(hw, IDPF_CTLQ_TYPE_MAILBOX_RX,
- IDPF_DFLT_MBX_ID);
-
- if (!hw->asq || !hw->arq) {
- idpf_ctlq_deinit(hw);
-
+ adapter->asq = libeth_find_ctlq(ctx, LIBETH_CTLQ_TYPE_TX,
+ LIBETH_CTLQ_MBX_ID);
+ adapter->arq = libeth_find_ctlq(ctx, LIBETH_CTLQ_TYPE_RX,
+ LIBETH_CTLQ_MBX_ID);
+ if (!adapter->asq || !adapter->arq) {
+ libeth_ctlq_xn_deinit(params.xnm, ctx);
return -ENOENT;
}
+ adapter->xn_init_params.xnm = params.xnm;
adapter->state = __IDPF_VER_CHECK;
return 0;
@@ -2854,12 +2292,13 @@ int idpf_init_dflt_mbx(struct idpf_adapter *adapter)
*/
void idpf_deinit_dflt_mbx(struct idpf_adapter *adapter)
{
- if (adapter->hw.arq && adapter->hw.asq) {
+ if (adapter->arq && adapter->asq) {
idpf_mb_clean(adapter);
- idpf_ctlq_deinit(&adapter->hw);
+ libeth_ctlq_xn_deinit(adapter->xn_init_params.xnm,
+ &adapter->ctlq_ctx);
}
- adapter->hw.arq = NULL;
- adapter->hw.asq = NULL;
+ adapter->arq = NULL;
+ adapter->asq = NULL;
}
/**
@@ -2932,15 +2371,6 @@ int idpf_vc_core_init(struct idpf_adapter *adapter)
u16 num_max_vports;
int err = 0;
- if (!adapter->vcxn_mngr) {
- adapter->vcxn_mngr = kzalloc(sizeof(*adapter->vcxn_mngr), GFP_KERNEL);
- if (!adapter->vcxn_mngr) {
- err = -ENOMEM;
- goto init_failed;
- }
- }
- idpf_vc_xn_init(adapter->vcxn_mngr);
-
while (adapter->state != __IDPF_INIT_SW) {
switch (adapter->state) {
case __IDPF_VER_CHECK:
@@ -3054,8 +2484,7 @@ int idpf_vc_core_init(struct idpf_adapter *adapter)
* the mailbox again
*/
adapter->state = __IDPF_VER_CHECK;
- if (adapter->vcxn_mngr)
- idpf_vc_xn_shutdown(adapter->vcxn_mngr);
+ idpf_deinit_dflt_mbx(adapter);
set_bit(IDPF_HR_DRV_LOAD, adapter->flags);
queue_delayed_work(adapter->vc_event_wq, &adapter->vc_event_task,
msecs_to_jiffies(task_delay));
@@ -3078,13 +2507,13 @@ void idpf_vc_core_deinit(struct idpf_adapter *adapter)
/* Avoid transaction timeouts when called during reset */
remove_in_prog = test_bit(IDPF_REMOVE_IN_PROG, adapter->flags);
if (!remove_in_prog)
- idpf_vc_xn_shutdown(adapter->vcxn_mngr);
+ idpf_deinit_dflt_mbx(adapter);
idpf_deinit_task(adapter);
idpf_intr_rel(adapter);
if (remove_in_prog)
- idpf_vc_xn_shutdown(adapter->vcxn_mngr);
+ idpf_deinit_dflt_mbx(adapter);
cancel_delayed_work_sync(&adapter->serv_task);
cancel_delayed_work_sync(&adapter->mbx_task);
@@ -3507,9 +2936,9 @@ u32 idpf_get_vport_id(struct idpf_vport *vport)
/**
* idpf_mac_filter_async_handler - Async callback for mac filters
- * @adapter: private data struct
- * @xn: transaction for message
- * @ctlq_msg: received message
+ * @ctx: controlq context structure
+ * @buff: response buffer pointer and size
+ * @status: async call return value
*
* In some scenarios driver can't sleep and wait for a reply (e.g.: stack is
* holding rtnl_lock) when adding a new mac filter. It puts us in a difficult
@@ -3517,13 +2946,14 @@ u32 idpf_get_vport_id(struct idpf_vport *vport)
* ultimately do is remove it from our list of mac filters and report the
* error.
*/
-static int idpf_mac_filter_async_handler(struct idpf_adapter *adapter,
- struct idpf_vc_xn *xn,
- const struct idpf_ctlq_msg *ctlq_msg)
+static void idpf_mac_filter_async_handler(void *ctx,
+ struct kvec *buff,
+ int status)
{
struct virtchnl2_mac_addr_list *ma_list;
struct idpf_vport_config *vport_config;
struct virtchnl2_mac_addr *mac_addr;
+ struct idpf_adapter *adapter = ctx;
struct idpf_mac_filter *f, *tmp;
struct list_head *ma_list_head;
struct idpf_vport *vport;
@@ -3531,18 +2961,18 @@ static int idpf_mac_filter_async_handler(struct idpf_adapter *adapter,
int i;
/* if success we're done, we're only here if something bad happened */
- if (!ctlq_msg->cookie.mbx.chnl_retval)
- return 0;
+ if (!status)
+ goto free_mem;
+ ma_list = buff->iov_base;
/* make sure at least struct is there */
- if (xn->reply_sz < sizeof(*ma_list))
+ if (buff->iov_len < sizeof(*ma_list))
goto invalid_payload;
- ma_list = ctlq_msg->ctx.indirect.payload->va;
mac_addr = ma_list->mac_addr_list;
num_entries = le16_to_cpu(ma_list->num_mac_addr);
/* we should have received a buffer at least this big */
- if (xn->reply_sz < struct_size(ma_list, mac_addr_list, num_entries))
+ if (buff->iov_len < struct_size(ma_list, mac_addr_list, num_entries))
goto invalid_payload;
vport = idpf_vid_to_vport(adapter, le32_to_cpu(ma_list->vport_id));
@@ -3562,16 +2992,15 @@ static int idpf_mac_filter_async_handler(struct idpf_adapter *adapter,
if (ether_addr_equal(mac_addr[i].addr, f->macaddr))
list_del(&f->list);
spin_unlock_bh(&vport_config->mac_filter_list_lock);
- dev_err_ratelimited(&adapter->pdev->dev, "Received error sending MAC filter request (op %d)\n",
- xn->vc_op);
-
- return 0;
+ dev_err_ratelimited(&adapter->pdev->dev, "Received error %d on sending MAC filter request\n",
+ status);
+ goto free_mem;
invalid_payload:
- dev_err_ratelimited(&adapter->pdev->dev, "Received invalid MAC filter payload (op %d) (len %zd)\n",
- xn->vc_op, xn->reply_sz);
-
- return -EINVAL;
+ dev_err_ratelimited(&adapter->pdev->dev, "Received invalid MAC filter payload (len %zd)\n",
+ buff->iov_len);
+free_mem:
+ libeth_ctlq_release_rx_buf(adapter->arq, buff);
}
/**
@@ -3587,21 +3016,24 @@ int idpf_add_del_mac_filters(struct idpf_vport *vport,
struct idpf_netdev_priv *np,
bool add, bool async)
{
- struct virtchnl2_mac_addr_list *ma_list __free(kfree) = NULL;
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC,
+ .chnl_opcode = add ? VIRTCHNL2_OP_ADD_MAC_ADDR :
+ VIRTCHNL2_OP_DEL_MAC_ADDR,
+ };
struct virtchnl2_mac_addr *mac_addr __free(kfree) = NULL;
struct idpf_adapter *adapter = np->adapter;
- struct idpf_vc_xn_params xn_params = {};
+ struct virtchnl2_mac_addr_list *ma_list;
struct idpf_vport_config *vport_config;
u32 num_msgs, total_filters = 0;
struct idpf_mac_filter *f;
- ssize_t reply_sz;
- int i = 0, k;
+ int i = 0, k, err = 0;
+ u32 buf_size;
- xn_params.vc_op = add ? VIRTCHNL2_OP_ADD_MAC_ADDR :
- VIRTCHNL2_OP_DEL_MAC_ADDR;
- xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
- xn_params.async = async;
- xn_params.async_handler = idpf_mac_filter_async_handler;
+ if (async) {
+ xn_params.resp_cb = idpf_mac_filter_async_handler;
+ xn_params.send_ctx = adapter;
+ }
vport_config = adapter->vport_config[np->vport_idx];
spin_lock_bh(&vport_config->mac_filter_list_lock);
@@ -3656,37 +3088,60 @@ int idpf_add_del_mac_filters(struct idpf_vport *vport,
num_msgs = DIV_ROUND_UP(total_filters, IDPF_NUM_FILTERS_PER_MSG);
for (i = 0, k = 0; i < num_msgs; i++) {
- u32 entries_size, buf_size, num_entries;
+ u32 entries_size, num_entries;
num_entries = min_t(u32, total_filters,
IDPF_NUM_FILTERS_PER_MSG);
entries_size = sizeof(struct virtchnl2_mac_addr) * num_entries;
buf_size = struct_size(ma_list, mac_addr_list, num_entries);
- if (!ma_list || num_entries != IDPF_NUM_FILTERS_PER_MSG) {
- kfree(ma_list);
- ma_list = kzalloc(buf_size, GFP_ATOMIC);
- if (!ma_list)
- return -ENOMEM;
- } else {
- memset(ma_list, 0, buf_size);
- }
+ ma_list = kzalloc(buf_size, GFP_KERNEL);
+ if (!ma_list)
+ return -ENOMEM;
ma_list->vport_id = cpu_to_le32(np->vport_id);
ma_list->num_mac_addr = cpu_to_le16(num_entries);
memcpy(ma_list->mac_addr_list, &mac_addr[k], entries_size);
- xn_params.send_buf.iov_base = ma_list;
- xn_params.send_buf.iov_len = buf_size;
- reply_sz = idpf_vc_xn_exec(adapter, &xn_params);
- if (reply_sz < 0)
- return reply_sz;
+ err = idpf_send_mb_msg(adapter, &xn_params, ma_list, buf_size);
+ if (err)
+ goto free_tx_buf;
+
+ if (!async)
+ libeth_ctlq_release_rx_buf(adapter->arq,
+ &xn_params.recv_mem);
k += num_entries;
total_filters -= num_entries;
}
- return 0;
+free_tx_buf:
+ if (num_msgs && libeth_cp_can_send_onstack(buf_size))
+ kfree(ma_list);
+
+ return err;
+}
+
+/**
+ * idpf_promiscuous_async_handler - async callback for promiscuous mode
+ * @ctx: controlq context structure
+ * @buff: response buffer pointer and size
+ * @status: async call return value
+ *
+ * Nobody is waiting for the promiscuous virtchnl message response. Print
+ * an error message if something went wrong and return.
+ */
+static void idpf_promiscuous_async_handler(void *ctx,
+ struct kvec *buff,
+ int status)
+{
+ struct idpf_adapter *adapter = ctx;
+
+ if (status)
+ dev_err_ratelimited(&adapter->pdev->dev, "Failed to set promiscuous mode: %d\n",
+ status);
+
+ libeth_ctlq_release_rx_buf(adapter->arq, buff);
}
/**
@@ -3703,9 +3158,13 @@ int idpf_set_promiscuous(struct idpf_adapter *adapter,
struct idpf_vport_user_config_data *config_data,
u32 vport_id)
{
- struct idpf_vc_xn_params xn_params = {};
+ struct libeth_ctlq_xn_send_params xn_params = {
+ .timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC,
+ .chnl_opcode = VIRTCHNL2_OP_CONFIG_PROMISCUOUS_MODE,
+ .resp_cb = idpf_promiscuous_async_handler,
+ .send_ctx = adapter,
+ };
struct virtchnl2_promisc_info vpi;
- ssize_t reply_sz;
u16 flags = 0;
if (test_bit(__IDPF_PROMISC_UC, config_data->user_flags))
@@ -3716,13 +3175,5 @@ int idpf_set_promiscuous(struct idpf_adapter *adapter,
vpi.vport_id = cpu_to_le32(vport_id);
vpi.flags = cpu_to_le16(flags);
- xn_params.vc_op = VIRTCHNL2_OP_CONFIG_PROMISCUOUS_MODE;
- xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
- xn_params.send_buf.iov_base = &vpi;
- xn_params.send_buf.iov_len = sizeof(vpi);
- /* setting promiscuous is only ever done asynchronously */
- xn_params.async = true;
- reply_sz = idpf_vc_xn_exec(adapter, &xn_params);
-
- return reply_sz < 0 ? reply_sz : 0;
+ return idpf_send_mb_msg(adapter, &xn_params, &vpi, sizeof(vpi));
}
diff --git a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.h b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.h
index 83da5d8da56b..35226aa9f3b1 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.h
+++ b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.h
@@ -21,9 +21,11 @@ int idpf_get_reg_intr_vecs(struct idpf_vport *vport,
int idpf_queue_reg_init(struct idpf_vport *vport);
int idpf_vport_queue_ids_init(struct idpf_vport *vport);
-int idpf_recv_mb_msg(struct idpf_adapter *adapter);
-int idpf_send_mb_msg(struct idpf_adapter *adapter, u32 op,
- u16 msg_size, u8 *msg, u16 cookie);
+void idpf_recv_event_msg(struct libeth_ctlq_ctx *ctx,
+ struct libeth_ctlq_msg *ctlq_msg);
+int idpf_send_mb_msg(struct idpf_adapter *adapter,
+ struct libeth_ctlq_xn_send_params *xn_params,
+ void *send_buf, size_t send_buf_size);
void idpf_vport_init(struct idpf_vport *vport, struct idpf_vport_max_q *max_q);
u32 idpf_get_vport_id(struct idpf_vport *vport);
--
2.47.0
^ permalink raw reply related [flat|nested] 36+ messages in thread* Re: [PATCH iwl-next 08/14] idpf: refactor idpf to use libeth controlq and Xn APIs
2025-04-08 12:47 ` [PATCH iwl-next 08/14] idpf: refactor idpf to use libeth controlq and Xn APIs Larysa Zaremba
@ 2025-04-21 14:25 ` Simon Horman
0 siblings, 0 replies; 36+ messages in thread
From: Simon Horman @ 2025-04-21 14:25 UTC (permalink / raw)
To: Larysa Zaremba
Cc: intel-wired-lan, Tony Nguyen, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Jonathan Corbet, Przemek Kitszel,
Jiri Pirko, Mustafa Ismail, Tatyana Nikolova, Andrew Lunn,
Alexander Lobakin, Michael Ellerman, Maciej Fijalkowski,
Lee Trager, Madhavan Srinivasan, Sridhar Samudrala, Jacob Keller,
Michal Swiatkowski, Mateusz Polchlopek, Wenjun Wu, Ahmed Zaki,
netdev, linux-doc, linux-kernel, Karlsson, Magnus, Emil Tantilov,
Madhu Chittim, Josh Hay, Milena Olech, pavan.kumar.linga,
Singhai, Anjali, Michal Kubiak
On Tue, Apr 08, 2025 at 02:47:54PM +0200, Larysa Zaremba wrote:
> From: Pavan Kumar Linga <pavan.kumar.linga@intel.com>
>
> Support to initialize and configure controlq, Xn manager,
> MMIO and reset APIs was introduced in libeth. As part of it,
> most of the existing controlq structures are renamed and
> modified. Use those APIs in idpf and make all the necessary changes.
>
> Previously for the send and receive virtchnl messages, there
> used to be a memcpy involved in controlq code to copy the buffer
> info passed by the send function into the controlq specific
> buffers. There was no restriction to use automatic memory
> in that case. The new implementation in libeth removed copying
> of the send buffer info and introduced DMA mapping of the
> send buffer itself. To accommodate it, use dynamic memory for
> the send buffers. In case of receive, idpf receives a page pool
> buffer allocated by the libeth and care should be taken to
> release it after use in the idpf.
>
> The changes are fairly trivial and localized, with a notable exception
> being the consolidation of idpf_vc_xn_shutdown and idpf_deinit_dflt_mbx
> under the latter name. This has some additional consequences that are
> addressed in the following patches.
>
> Signed-off-by: Pavan Kumar Linga <pavan.kumar.linga@intel.com>
> Reviewed-by: Michal Kubiak <michal.kubiak@intel.com>
> Signed-off-by: Larysa Zaremba <larysa.zaremba@intel.com>
...
> diff --git a/drivers/net/ethernet/intel/idpf/idpf.h b/drivers/net/ethernet/intel/idpf/idpf.h
...
> @@ -488,7 +486,10 @@ struct idpf_vc_xn_manager;
> * @state: Init state machine
> * @flags: See enum idpf_flags
> * @reset_reg: See struct idpf_reset_reg
> - * @hw: Device access data
> + * @ctlq_ctx: controlq context
> + * @asq: Send control queue info
> + * @arq: Receive control queue info
> + * @xn_init_params: Xn transaction manager parameters
> * @num_req_msix: Requested number of MSIX vectors
> * @num_avail_msix: Available number of MSIX vectors
> * @num_msix_entries: Number of entries in MSIX table
> @@ -540,7 +541,10 @@ struct idpf_adapter {
> enum idpf_state state;
> DECLARE_BITMAP(flags, IDPF_FLAGS_NBITS);
> struct idpf_reset_reg reset_reg;
> - struct idpf_hw hw;
> + struct libeth_ctlq_ctx ctlq_ctx;
> + struct libeth_ctlq_info *asq;
> + struct libeth_ctlq_info *arq;
> + struct libeth_ctlq_xn_init_params xn_init_params;
> u16 num_req_msix;
> u16 num_avail_msix;
> u16 num_msix_entries;
> @@ -573,7 +577,6 @@ struct idpf_adapter {
> struct delayed_work stats_task;
> struct workqueue_struct *stats_wq;
> struct virtchnl2_get_capabilities caps;
> - struct idpf_vc_xn_manager *vcxn_mngr;
nit: Please also drop the vcxn_mngr from the Kernel doc for struct idpf_adapter.
>
> struct idpf_dev_ops dev_ops;
> int num_vfs;
^ permalink raw reply [flat|nested] 36+ messages in thread
* [PATCH iwl-next 09/14] idpf: make mbx_task queueing and cancelling more consistent
2025-04-08 12:47 [PATCH iwl-next 00/14] Introduce iXD driver Larysa Zaremba
` (7 preceding siblings ...)
2025-04-08 12:47 ` [PATCH iwl-next 08/14] idpf: refactor idpf to use libeth controlq and Xn APIs Larysa Zaremba
@ 2025-04-08 12:47 ` Larysa Zaremba
2025-04-08 12:47 ` [PATCH iwl-next 10/14] idpf: print a debug message and bail in case of non-event ctlq message Larysa Zaremba
` (5 subsequent siblings)
14 siblings, 0 replies; 36+ messages in thread
From: Larysa Zaremba @ 2025-04-08 12:47 UTC (permalink / raw)
To: intel-wired-lan, Tony Nguyen
Cc: David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Simon Horman, Jonathan Corbet, Przemek Kitszel, Jiri Pirko,
Mustafa Ismail, Tatyana Nikolova, Andrew Lunn, Alexander Lobakin,
Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Larysa Zaremba, Sridhar Samudrala,
Jacob Keller, Michal Swiatkowski, Mateusz Polchlopek, Wenjun Wu,
Ahmed Zaki, netdev, linux-doc, linux-kernel, Karlsson, Magnus,
Emil Tantilov, Madhu Chittim, Josh Hay, Milena Olech,
pavan.kumar.linga, Singhai, Anjali, Michal Kubiak
As a consequence of refactoring idpf code to use libeth APIs,
idpf_vc_xn_shutdown was merged with and replaced by idpf_deinit_dflt_mbx.
This does not affect the Tx path, as it checked for a presence of an xn
manager anyway. Rx processing is handled by the mbx_task that is not always
cancelled before calling the new consolidated mailbox deinit function.
Moreover, in the reset path idpf_intr_rel() reschedules it after the deinit
is done. This leads to mbx_task referencing the freed mailbox and causing
KASAN warnings.
To remedy this, in the init path, do the first queueing of mbx_task in
idpf_init_dflt_mbx(), in deinit and reset, always cancel the task in
idpf_deinit_dflt_mbx() and in every flow first call idpf_mb_intr_rel_irq().
Reviewed-by: Michal Kubiak <michal.kubiak@intel.com>
Signed-off-by: Larysa Zaremba <larysa.zaremba@intel.com>
---
drivers/net/ethernet/intel/idpf/idpf.h | 1 +
drivers/net/ethernet/intel/idpf/idpf_lib.c | 9 ++++-----
drivers/net/ethernet/intel/idpf/idpf_virtchnl.c | 6 +++++-
3 files changed, 10 insertions(+), 6 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf.h b/drivers/net/ethernet/intel/idpf/idpf.h
index f4197e7a2ef8..3fadba5d9c6d 100644
--- a/drivers/net/ethernet/intel/idpf/idpf.h
+++ b/drivers/net/ethernet/intel/idpf/idpf.h
@@ -814,6 +814,7 @@ void idpf_vc_event_task(struct work_struct *work);
void idpf_dev_ops_init(struct idpf_adapter *adapter);
void idpf_vf_dev_ops_init(struct idpf_adapter *adapter);
int idpf_intr_req(struct idpf_adapter *adapter);
+void idpf_mb_intr_rel_irq(struct idpf_adapter *adapter);
void idpf_intr_rel(struct idpf_adapter *adapter);
u16 idpf_get_max_tx_hdr_size(struct idpf_adapter *adapter);
int idpf_initiate_soft_reset(struct idpf_vport *vport,
diff --git a/drivers/net/ethernet/intel/idpf/idpf_lib.c b/drivers/net/ethernet/intel/idpf/idpf_lib.c
index bf403dcf794e..7f0c21c2560f 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_lib.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_lib.c
@@ -65,9 +65,11 @@ static void idpf_deinit_vector_stack(struct idpf_adapter *adapter)
* This will also disable interrupt mode and queue up mailbox task. Mailbox
* task will reschedule itself if not in interrupt mode.
*/
-static void idpf_mb_intr_rel_irq(struct idpf_adapter *adapter)
+void idpf_mb_intr_rel_irq(struct idpf_adapter *adapter)
{
- clear_bit(IDPF_MB_INTR_MODE, adapter->flags);
+ if (!test_and_clear_bit(IDPF_MB_INTR_MODE, adapter->flags))
+ return;
+
kfree(free_irq(adapter->msix_entries[0].vector, adapter));
queue_delayed_work(adapter->mbx_wq, &adapter->mbx_task, 0);
}
@@ -1775,14 +1777,11 @@ static int idpf_init_hard_reset(struct idpf_adapter *adapter)
goto unlock_mutex;
}
- queue_delayed_work(adapter->mbx_wq, &adapter->mbx_task, 0);
-
/* Initialize the state machine, also allocate memory and request
* resources
*/
err = idpf_vc_core_init(adapter);
if (err) {
- cancel_delayed_work_sync(&adapter->mbx_task);
idpf_deinit_dflt_mbx(adapter);
goto unlock_mutex;
}
diff --git a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
index 2bfb8882ae13..6ff7d5ba3844 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
@@ -2283,6 +2283,8 @@ int idpf_init_dflt_mbx(struct idpf_adapter *adapter)
adapter->xn_init_params.xnm = params.xnm;
adapter->state = __IDPF_VER_CHECK;
+ queue_delayed_work(adapter->mbx_wq, &adapter->mbx_task, 0);
+
return 0;
}
@@ -2292,6 +2294,9 @@ int idpf_init_dflt_mbx(struct idpf_adapter *adapter)
*/
void idpf_deinit_dflt_mbx(struct idpf_adapter *adapter)
{
+ idpf_mb_intr_rel_irq(adapter);
+ cancel_delayed_work_sync(&adapter->mbx_task);
+
if (adapter->arq && adapter->asq) {
idpf_mb_clean(adapter);
libeth_ctlq_xn_deinit(adapter->xn_init_params.xnm,
@@ -2516,7 +2521,6 @@ void idpf_vc_core_deinit(struct idpf_adapter *adapter)
idpf_deinit_dflt_mbx(adapter);
cancel_delayed_work_sync(&adapter->serv_task);
- cancel_delayed_work_sync(&adapter->mbx_task);
idpf_vport_params_buf_rel(adapter);
--
2.47.0
^ permalink raw reply related [flat|nested] 36+ messages in thread* [PATCH iwl-next 10/14] idpf: print a debug message and bail in case of non-event ctlq message
2025-04-08 12:47 [PATCH iwl-next 00/14] Introduce iXD driver Larysa Zaremba
` (8 preceding siblings ...)
2025-04-08 12:47 ` [PATCH iwl-next 09/14] idpf: make mbx_task queueing and cancelling more consistent Larysa Zaremba
@ 2025-04-08 12:47 ` Larysa Zaremba
2025-04-08 12:47 ` [PATCH iwl-next 11/14] ixd: add basic driver framework for Intel(R) Control Plane Function Larysa Zaremba
` (4 subsequent siblings)
14 siblings, 0 replies; 36+ messages in thread
From: Larysa Zaremba @ 2025-04-08 12:47 UTC (permalink / raw)
To: intel-wired-lan, Tony Nguyen
Cc: David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Simon Horman, Jonathan Corbet, Przemek Kitszel, Jiri Pirko,
Mustafa Ismail, Tatyana Nikolova, Andrew Lunn, Alexander Lobakin,
Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Larysa Zaremba, Sridhar Samudrala,
Jacob Keller, Michal Swiatkowski, Mateusz Polchlopek, Wenjun Wu,
Ahmed Zaki, netdev, linux-doc, linux-kernel, Karlsson, Magnus,
Emil Tantilov, Madhu Chittim, Josh Hay, Milena Olech,
pavan.kumar.linga, Singhai, Anjali, Michal Kubiak
Unlike previous internal idpf ctlq implementation, idpf calls the default
message handler for all received messages that do not have a matching xn
transaction, not only for VIRTCHNL2_OP_EVENT. This leads to many error
messages printing garbage, because the parsing expected a valid event
message, but got e.g. a delayed response for a timed-out transaction.
The information about timed-out transactions and otherwise unhandleable
messages can still be valuable for developers, so print the information
with dynamic debug and exit the function, so the following functions can
parse valid events in peace.
Reviewed-by: Michal Kubiak <michal.kubiak@intel.com>
Signed-off-by: Larysa Zaremba <larysa.zaremba@intel.com>
---
drivers/net/ethernet/intel/idpf/idpf_virtchnl.c | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
index 6ff7d5ba3844..d3b4612c3205 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
@@ -84,6 +84,13 @@ void idpf_recv_event_msg(struct libeth_ctlq_ctx *ctx,
u32 event;
adapter = container_of(ctx, struct idpf_adapter, ctlq_ctx);
+ if (ctlq_msg->chnl_opcode != VIRTCHNL2_OP_EVENT) {
+ dev_dbg(&adapter->pdev->dev,
+ "Unhandled message with opcode %u from CP\n",
+ ctlq_msg->chnl_opcode);
+ goto free_rx_buf;
+ }
+
if (payload_size < sizeof(*v2e)) {
dev_err_ratelimited(&adapter->pdev->dev, "Failed to receive valid payload for event msg (op %d len %d)\n",
ctlq_msg->chnl_opcode,
--
2.47.0
^ permalink raw reply related [flat|nested] 36+ messages in thread* [PATCH iwl-next 11/14] ixd: add basic driver framework for Intel(R) Control Plane Function
2025-04-08 12:47 [PATCH iwl-next 00/14] Introduce iXD driver Larysa Zaremba
` (9 preceding siblings ...)
2025-04-08 12:47 ` [PATCH iwl-next 10/14] idpf: print a debug message and bail in case of non-event ctlq message Larysa Zaremba
@ 2025-04-08 12:47 ` Larysa Zaremba
2025-04-08 12:47 ` [PATCH iwl-next 12/14] ixd: add reset checks and initialize the mailbox Larysa Zaremba
` (3 subsequent siblings)
14 siblings, 0 replies; 36+ messages in thread
From: Larysa Zaremba @ 2025-04-08 12:47 UTC (permalink / raw)
To: intel-wired-lan, Tony Nguyen
Cc: David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Simon Horman, Jonathan Corbet, Przemek Kitszel, Jiri Pirko,
Mustafa Ismail, Tatyana Nikolova, Andrew Lunn, Alexander Lobakin,
Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Larysa Zaremba, Sridhar Samudrala,
Jacob Keller, Michal Swiatkowski, Mateusz Polchlopek, Wenjun Wu,
Ahmed Zaki, netdev, linux-doc, linux-kernel, Karlsson, Magnus,
Emil Tantilov, Madhu Chittim, Josh Hay, Milena Olech,
pavan.kumar.linga, Singhai, Anjali
Add module register and probe functionality. Add the required support
to register IXD PCI driver, as well as probe and remove call backs.
Enable the PCI device and request the kernel to reserve the memory
resources that will be used by the driver. Finally map the BAR0
address space.
For now, use devm_alloc() to allocate adapter, as it requires the least
amount of code. In a later commit, it will be replaced with a devlink
alternative.
Co-developed-by: Amritha Nambiar <amritha.nambiar@intel.com>
Signed-off-by: Amritha Nambiar <amritha.nambiar@intel.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Larysa Zaremba <larysa.zaremba@intel.com>
---
.../device_drivers/ethernet/index.rst | 1 +
.../device_drivers/ethernet/intel/ixd.rst | 39 ++++++
drivers/net/ethernet/intel/Kconfig | 2 +
drivers/net/ethernet/intel/Makefile | 1 +
drivers/net/ethernet/intel/ixd/Kconfig | 13 ++
drivers/net/ethernet/intel/ixd/Makefile | 8 ++
drivers/net/ethernet/intel/ixd/ixd.h | 26 ++++
drivers/net/ethernet/intel/ixd/ixd_lan_regs.h | 28 ++++
drivers/net/ethernet/intel/ixd/ixd_main.c | 122 ++++++++++++++++++
9 files changed, 240 insertions(+)
create mode 100644 Documentation/networking/device_drivers/ethernet/intel/ixd.rst
create mode 100644 drivers/net/ethernet/intel/ixd/Kconfig
create mode 100644 drivers/net/ethernet/intel/ixd/Makefile
create mode 100644 drivers/net/ethernet/intel/ixd/ixd.h
create mode 100644 drivers/net/ethernet/intel/ixd/ixd_lan_regs.h
create mode 100644 drivers/net/ethernet/intel/ixd/ixd_main.c
diff --git a/Documentation/networking/device_drivers/ethernet/index.rst b/Documentation/networking/device_drivers/ethernet/index.rst
index 05d822b904b4..c2b1951fec84 100644
--- a/Documentation/networking/device_drivers/ethernet/index.rst
+++ b/Documentation/networking/device_drivers/ethernet/index.rst
@@ -37,6 +37,7 @@ Contents:
intel/igbvf
intel/ixgbe
intel/ixgbevf
+ intel/ixd
intel/i40e
intel/iavf
intel/ice
diff --git a/Documentation/networking/device_drivers/ethernet/intel/ixd.rst b/Documentation/networking/device_drivers/ethernet/intel/ixd.rst
new file mode 100644
index 000000000000..1387626e5d20
--- /dev/null
+++ b/Documentation/networking/device_drivers/ethernet/intel/ixd.rst
@@ -0,0 +1,39 @@
+.. SPDX-License-Identifier: GPL-2.0+
+
+==========================================================================
+iXD Linux* Base Driver for the Intel(R) Control Plane Function
+==========================================================================
+
+Intel iXD Linux driver.
+Copyright(C) 2025 Intel Corporation.
+
+.. contents::
+
+For questions related to hardware requirements, refer to the documentation
+supplied with your Intel adapter. All hardware requirements listed apply to use
+with Linux.
+
+
+Identifying Your Adapter
+========================
+For information on how to identify your adapter, and for the latest Intel
+network drivers, refer to the Intel Support website:
+http://www.intel.com/support
+
+
+Support
+=======
+For general information, go to the Intel support website at:
+http://www.intel.com/support/
+
+If an issue is identified with the released source code on a supported kernel
+with a supported adapter, email the specific information related to the issue
+to intel-wired-lan@lists.osuosl.org.
+
+
+Trademarks
+==========
+Intel is a trademark or registered trademark of Intel Corporation or its
+subsidiaries in the United States and/or other countries.
+
+* Other names and brands may be claimed as the property of others.
diff --git a/drivers/net/ethernet/intel/Kconfig b/drivers/net/ethernet/intel/Kconfig
index 5a331c1c76cb..d92b832a067a 100644
--- a/drivers/net/ethernet/intel/Kconfig
+++ b/drivers/net/ethernet/intel/Kconfig
@@ -392,4 +392,6 @@ config IGC_LEDS
source "drivers/net/ethernet/intel/idpf/Kconfig"
+source "drivers/net/ethernet/intel/ixd/Kconfig"
+
endif # NET_VENDOR_INTEL
diff --git a/drivers/net/ethernet/intel/Makefile b/drivers/net/ethernet/intel/Makefile
index 04c844ef4964..8de2222a0333 100644
--- a/drivers/net/ethernet/intel/Makefile
+++ b/drivers/net/ethernet/intel/Makefile
@@ -19,3 +19,4 @@ obj-$(CONFIG_IAVF) += iavf/
obj-$(CONFIG_FM10K) += fm10k/
obj-$(CONFIG_ICE) += ice/
obj-$(CONFIG_IDPF) += idpf/
+obj-$(CONFIG_IXD) += ixd/
diff --git a/drivers/net/ethernet/intel/ixd/Kconfig b/drivers/net/ethernet/intel/ixd/Kconfig
new file mode 100644
index 000000000000..3bf6e2407af6
--- /dev/null
+++ b/drivers/net/ethernet/intel/ixd/Kconfig
@@ -0,0 +1,13 @@
+# SPDX-License-Identifier: GPL-2.0-only
+# Copyright (C) 2025 Intel Corporation
+
+config IXD
+ tristate "Intel(R) Control Plane Function Support"
+ depends on PCI_MSI
+ select LIBETH
+ select LIBETH_PCI
+ help
+ This driver supports Intel(R) Control Plane PCI Function
+ of Intel E2100 and later IPUs and FNICs.
+ It facilitates a centralized control over multiple IDPF PFs/VFs/SFs
+ exposed by the same card.
diff --git a/drivers/net/ethernet/intel/ixd/Makefile b/drivers/net/ethernet/intel/ixd/Makefile
new file mode 100644
index 000000000000..3849bc240600
--- /dev/null
+++ b/drivers/net/ethernet/intel/ixd/Makefile
@@ -0,0 +1,8 @@
+# SPDX-License-Identifier: GPL-2.0-only
+# Copyright (C) 2025 Intel Corporation
+
+# Intel(R) Control Plane Function Linux Driver
+
+obj-$(CONFIG_IXD) += ixd.o
+
+ixd-y := ixd_main.o
diff --git a/drivers/net/ethernet/intel/ixd/ixd.h b/drivers/net/ethernet/intel/ixd/ixd.h
new file mode 100644
index 000000000000..b558803914f9
--- /dev/null
+++ b/drivers/net/ethernet/intel/ixd/ixd.h
@@ -0,0 +1,26 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/* Copyright (C) 2025 Intel Corporation */
+
+#ifndef _IXD_H_
+#define _IXD_H_
+
+#include <net/libeth/pci.h>
+
+/**
+ * struct ixd_adapter - Data structure representing a CPF
+ * @hw: Device access data
+ */
+struct ixd_adapter {
+ struct libeth_mmio_info hw;
+};
+
+/**
+ * ixd_to_dev - Get the corresponding device struct from an adapter
+ * @adapter: PCI device driver-specific private data
+ */
+static inline struct device *ixd_to_dev(struct ixd_adapter *adapter)
+{
+ return &adapter->hw.pdev->dev;
+}
+
+#endif /* _IXD_H_ */
diff --git a/drivers/net/ethernet/intel/ixd/ixd_lan_regs.h b/drivers/net/ethernet/intel/ixd/ixd_lan_regs.h
new file mode 100644
index 000000000000..a991eaa8a2aa
--- /dev/null
+++ b/drivers/net/ethernet/intel/ixd/ixd_lan_regs.h
@@ -0,0 +1,28 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/* Copyright (C) 2025 Intel Corporation */
+
+#ifndef _IXD_LAN_REGS_H_
+#define _IXD_LAN_REGS_H_
+
+/* Control Plane Function PCI ID */
+#define IXD_DEV_ID_CPF 0x1453
+
+/* Control Queue (Mailbox) */
+#define PF_FW_MBX_REG_LEN 4096
+#define PF_FW_MBX 0x08400000
+
+/* Reset registers */
+#define PFGEN_RTRIG_REG_LEN 2048
+#define PFGEN_RTRIG 0x08407000 /* Device resets */
+
+/**
+ * struct ixd_bar_region - BAR region description
+ * @offset: BAR region offset
+ * @size: BAR region size
+ */
+struct ixd_bar_region {
+ resource_size_t offset;
+ resource_size_t size;
+};
+
+#endif /* _IXD_LAN_REGS_H_ */
diff --git a/drivers/net/ethernet/intel/ixd/ixd_main.c b/drivers/net/ethernet/intel/ixd/ixd_main.c
new file mode 100644
index 000000000000..1f59db0b8fe7
--- /dev/null
+++ b/drivers/net/ethernet/intel/ixd/ixd_main.c
@@ -0,0 +1,122 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/* Copyright (C) 2025 Intel Corporation */
+
+#include "ixd.h"
+#include "ixd_lan_regs.h"
+
+MODULE_DESCRIPTION("Intel(R) Control Plane Function Device Driver");
+MODULE_IMPORT_NS("LIBETH_PCI");
+MODULE_LICENSE("GPL");
+
+/**
+ * ixd_remove - remove a CPF PCI device
+ * @pdev: PCI device being removed
+ */
+static void ixd_remove(struct pci_dev *pdev)
+{
+ struct ixd_adapter *adapter = pci_get_drvdata(pdev);
+
+ libeth_pci_unmap_all_mmio_regions(&adapter->hw);
+ libeth_pci_deinit_dev(pdev);
+}
+
+/**
+ * ixd_shutdown - shut down a CPF PCI device
+ * @pdev: PCI device being shut down
+ */
+static void ixd_shutdown(struct pci_dev *pdev)
+{
+ ixd_remove(pdev);
+
+ if (system_state == SYSTEM_POWER_OFF)
+ pci_set_power_state(pdev, PCI_D3hot);
+}
+
+/**
+ * ixd_iomap_regions - iomap PCI BARs
+ * @adapter: adapter to map memory regions for
+ *
+ * Returns: %0 on success, negative on failure
+ */
+static int ixd_iomap_regions(struct ixd_adapter *adapter)
+{
+ const struct ixd_bar_region regions[] = {
+ {
+ .offset = PFGEN_RTRIG,
+ .size = PFGEN_RTRIG_REG_LEN,
+ },
+ {
+ .offset = PF_FW_MBX,
+ .size = PF_FW_MBX_REG_LEN,
+ },
+ };
+
+ for (int i = 0; i < ARRAY_SIZE(regions); i++) {
+ struct libeth_mmio_info *mmio_info = &adapter->hw;
+ bool map_ok;
+
+ map_ok = libeth_pci_map_mmio_region(mmio_info,
+ regions[i].offset,
+ regions[i].size);
+ if (!map_ok) {
+ dev_err(ixd_to_dev(adapter),
+ "Failed to map PCI device MMIO region\n");
+
+ libeth_pci_unmap_all_mmio_regions(mmio_info);
+ return -EIO;
+ }
+ }
+
+ return 0;
+}
+
+/**
+ * ixd_probe - probe a CPF PCI device
+ * @pdev: corresponding PCI device
+ * @ent: entry in ixd_pci_tbl
+ *
+ * Returns: %0 on success, negative errno code on failure
+ */
+static int ixd_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
+{
+ struct ixd_adapter *adapter;
+ int err;
+
+ adapter = devm_kzalloc(&pdev->dev, sizeof(*adapter), GFP_KERNEL);
+ if (!adapter)
+ return -ENOMEM;
+ adapter->hw.pdev = pdev;
+ INIT_LIST_HEAD(&adapter->hw.mmio_list);
+
+ err = libeth_pci_init_dev(pdev);
+ if (err)
+ return err;
+
+ pci_set_drvdata(pdev, adapter);
+
+ err = ixd_iomap_regions(adapter);
+ if (err)
+ goto deinit_dev;
+
+ return 0;
+
+deinit_dev:
+ libeth_pci_deinit_dev(pdev);
+
+ return err;
+}
+
+static const struct pci_device_id ixd_pci_tbl[] = {
+ { PCI_VDEVICE(INTEL, IXD_DEV_ID_CPF) },
+ { }
+};
+MODULE_DEVICE_TABLE(pci, ixd_pci_tbl);
+
+static struct pci_driver ixd_driver = {
+ .name = KBUILD_MODNAME,
+ .id_table = ixd_pci_tbl,
+ .probe = ixd_probe,
+ .remove = ixd_remove,
+ .shutdown = ixd_shutdown,
+};
+module_pci_driver(ixd_driver);
--
2.47.0
^ permalink raw reply related [flat|nested] 36+ messages in thread* [PATCH iwl-next 12/14] ixd: add reset checks and initialize the mailbox
2025-04-08 12:47 [PATCH iwl-next 00/14] Introduce iXD driver Larysa Zaremba
` (10 preceding siblings ...)
2025-04-08 12:47 ` [PATCH iwl-next 11/14] ixd: add basic driver framework for Intel(R) Control Plane Function Larysa Zaremba
@ 2025-04-08 12:47 ` Larysa Zaremba
2025-04-08 12:47 ` [PATCH iwl-next 13/14] ixd: add the core initialization Larysa Zaremba
` (2 subsequent siblings)
14 siblings, 0 replies; 36+ messages in thread
From: Larysa Zaremba @ 2025-04-08 12:47 UTC (permalink / raw)
To: intel-wired-lan, Tony Nguyen
Cc: David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Simon Horman, Jonathan Corbet, Przemek Kitszel, Jiri Pirko,
Mustafa Ismail, Tatyana Nikolova, Andrew Lunn, Alexander Lobakin,
Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Larysa Zaremba, Sridhar Samudrala,
Jacob Keller, Michal Swiatkowski, Mateusz Polchlopek, Wenjun Wu,
Ahmed Zaki, netdev, linux-doc, linux-kernel, Karlsson, Magnus,
Emil Tantilov, Madhu Chittim, Josh Hay, Milena Olech,
pavan.kumar.linga, Singhai, Anjali
At the end of the probe, trigger hard reset, initialize and schedule the
after-reset task. If the reset is complete in a pre-determined time,
initialize the default mailbox, through which other resources will be
negotiated.
Co-developed-by: Amritha Nambiar <amritha.nambiar@intel.com>
Signed-off-by: Amritha Nambiar <amritha.nambiar@intel.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Larysa Zaremba <larysa.zaremba@intel.com>
---
drivers/net/ethernet/intel/ixd/Kconfig | 1 +
drivers/net/ethernet/intel/ixd/Makefile | 2 +
drivers/net/ethernet/intel/ixd/ixd.h | 28 +++-
drivers/net/ethernet/intel/ixd/ixd_dev.c | 86 +++++++++++
drivers/net/ethernet/intel/ixd/ixd_lan_regs.h | 40 +++++
drivers/net/ethernet/intel/ixd/ixd_lib.c | 143 ++++++++++++++++++
drivers/net/ethernet/intel/ixd/ixd_main.c | 26 +++-
7 files changed, 318 insertions(+), 8 deletions(-)
create mode 100644 drivers/net/ethernet/intel/ixd/ixd_dev.c
create mode 100644 drivers/net/ethernet/intel/ixd/ixd_lib.c
diff --git a/drivers/net/ethernet/intel/ixd/Kconfig b/drivers/net/ethernet/intel/ixd/Kconfig
index 3bf6e2407af6..bcaf06a874aa 100644
--- a/drivers/net/ethernet/intel/ixd/Kconfig
+++ b/drivers/net/ethernet/intel/ixd/Kconfig
@@ -5,6 +5,7 @@ config IXD
tristate "Intel(R) Control Plane Function Support"
depends on PCI_MSI
select LIBETH
+ select LIBETH_CP
select LIBETH_PCI
help
This driver supports Intel(R) Control Plane PCI Function
diff --git a/drivers/net/ethernet/intel/ixd/Makefile b/drivers/net/ethernet/intel/ixd/Makefile
index 3849bc240600..164b2c86952f 100644
--- a/drivers/net/ethernet/intel/ixd/Makefile
+++ b/drivers/net/ethernet/intel/ixd/Makefile
@@ -6,3 +6,5 @@
obj-$(CONFIG_IXD) += ixd.o
ixd-y := ixd_main.o
+ixd-y += ixd_dev.o
+ixd-y += ixd_lib.o
diff --git a/drivers/net/ethernet/intel/ixd/ixd.h b/drivers/net/ethernet/intel/ixd/ixd.h
index b558803914f9..7efa8db8c5ba 100644
--- a/drivers/net/ethernet/intel/ixd/ixd.h
+++ b/drivers/net/ethernet/intel/ixd/ixd.h
@@ -4,14 +4,25 @@
#ifndef _IXD_H_
#define _IXD_H_
-#include <net/libeth/pci.h>
+#include <net/libeth/controlq.h>
/**
* struct ixd_adapter - Data structure representing a CPF
- * @hw: Device access data
+ * @cp_ctx: Control plane communication context
+ * @init_task: Delayed initialization after reset
+ * @xnm: virtchnl transaction manager
+ * @asq: Send control queue info
+ * @arq: Receive control queue info
*/
struct ixd_adapter {
- struct libeth_mmio_info hw;
+ struct libeth_ctlq_ctx cp_ctx;
+ struct {
+ struct delayed_work init_work;
+ u8 reset_retries;
+ } init_task;
+ struct libeth_ctlq_xn_manager *xnm;
+ struct libeth_ctlq_info *asq;
+ struct libeth_ctlq_info *arq;
};
/**
@@ -20,7 +31,16 @@ struct ixd_adapter {
*/
static inline struct device *ixd_to_dev(struct ixd_adapter *adapter)
{
- return &adapter->hw.pdev->dev;
+ return &adapter->cp_ctx.mmio_info.pdev->dev;
}
+void ixd_ctlq_reg_init(struct ixd_adapter *adapter,
+ struct libeth_ctlq_reg *ctlq_reg_tx,
+ struct libeth_ctlq_reg *ctlq_reg_rx);
+void ixd_trigger_reset(struct ixd_adapter *adapter);
+bool ixd_check_reset_complete(struct ixd_adapter *adapter);
+void ixd_init_task(struct work_struct *work);
+int ixd_init_dflt_mbx(struct ixd_adapter *adapter);
+void ixd_deinit_dflt_mbx(struct ixd_adapter *adapter);
+
#endif /* _IXD_H_ */
diff --git a/drivers/net/ethernet/intel/ixd/ixd_dev.c b/drivers/net/ethernet/intel/ixd/ixd_dev.c
new file mode 100644
index 000000000000..9d68c2ce3bab
--- /dev/null
+++ b/drivers/net/ethernet/intel/ixd/ixd_dev.c
@@ -0,0 +1,86 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/* Copyright (C) 2025 Intel Corporation */
+
+#include "ixd.h"
+#include "ixd_lan_regs.h"
+
+/**
+ * ixd_ctlq_reg_init - Initialize default mailbox registers
+ * @adapter: PCI device driver-specific private data
+ * @ctlq_reg_tx: Transmit queue registers info to be filled
+ * @ctlq_reg_rx: Receive queue registers info to be filled
+ */
+void ixd_ctlq_reg_init(struct ixd_adapter *adapter,
+ struct libeth_ctlq_reg *ctlq_reg_tx,
+ struct libeth_ctlq_reg *ctlq_reg_rx)
+{
+ struct libeth_mmio_info *mmio_info = &adapter->cp_ctx.mmio_info;
+ *ctlq_reg_tx = (struct libeth_ctlq_reg) {
+ .head = libeth_pci_get_mmio_addr(mmio_info, PF_FW_ATQH),
+ .tail = libeth_pci_get_mmio_addr(mmio_info, PF_FW_ATQT),
+ .len = libeth_pci_get_mmio_addr(mmio_info, PF_FW_ATQLEN),
+ .addr_high = libeth_pci_get_mmio_addr(mmio_info, PF_FW_ATQBAH),
+ .addr_low = libeth_pci_get_mmio_addr(mmio_info, PF_FW_ATQBAL),
+ .len_mask = PF_FW_ATQLEN_ATQLEN_M,
+ .len_ena_mask = PF_FW_ATQLEN_ATQENABLE_M,
+ .head_mask = PF_FW_ATQH_ATQH_M,
+ };
+
+ *ctlq_reg_rx = (struct libeth_ctlq_reg) {
+ .head = libeth_pci_get_mmio_addr(mmio_info, PF_FW_ARQH),
+ .tail = libeth_pci_get_mmio_addr(mmio_info, PF_FW_ARQT),
+ .len = libeth_pci_get_mmio_addr(mmio_info, PF_FW_ARQLEN),
+ .addr_high = libeth_pci_get_mmio_addr(mmio_info, PF_FW_ARQBAH),
+ .addr_low = libeth_pci_get_mmio_addr(mmio_info, PF_FW_ARQBAL),
+ .len_mask = PF_FW_ARQLEN_ARQLEN_M,
+ .len_ena_mask = PF_FW_ARQLEN_ARQENABLE_M,
+ .head_mask = PF_FW_ARQH_ARQH_M,
+ };
+}
+
+static const struct ixd_reset_reg ixd_reset_reg = {
+ .rstat = PFGEN_RSTAT,
+ .rstat_m = PFGEN_RSTAT_PFR_STATE_M,
+ .rstat_ok_v = 0b01,
+ .rtrigger = PFGEN_CTRL,
+ .rtrigger_m = PFGEN_CTRL_PFSWR,
+};
+
+/**
+ * ixd_trigger_reset - Trigger PFR reset
+ * @adapter: the device with mapped reset register
+ */
+void ixd_trigger_reset(struct ixd_adapter *adapter)
+{
+ void __iomem *addr;
+ u32 reg_val;
+
+ addr = libeth_pci_get_mmio_addr(&adapter->cp_ctx.mmio_info,
+ ixd_reset_reg.rtrigger);
+ reg_val = readl(addr);
+ writel(reg_val | ixd_reset_reg.rtrigger_m, addr);
+}
+
+/**
+ * ixd_check_reset_complete - Check if the PFR reset is completed
+ * @adapter: CPF being reset
+ */
+bool ixd_check_reset_complete(struct ixd_adapter *adapter)
+{
+ u32 reg_val, reset_status;
+ void __iomem *addr;
+
+ addr = libeth_pci_get_mmio_addr(&adapter->cp_ctx.mmio_info,
+ ixd_reset_reg.rstat);
+ reg_val = readl(addr);
+ reset_status = reg_val & ixd_reset_reg.rstat_m;
+
+ /* 0xFFFFFFFF might be read if the other side hasn't cleared
+ * the register for us yet.
+ */
+ if (reg_val != 0xFFFFFFFF &&
+ reset_status == ixd_reset_reg.rstat_ok_v)
+ return true;
+
+ return false;
+}
diff --git a/drivers/net/ethernet/intel/ixd/ixd_lan_regs.h b/drivers/net/ethernet/intel/ixd/ixd_lan_regs.h
index a991eaa8a2aa..26b1e3cfcf20 100644
--- a/drivers/net/ethernet/intel/ixd/ixd_lan_regs.h
+++ b/drivers/net/ethernet/intel/ixd/ixd_lan_regs.h
@@ -11,9 +11,33 @@
#define PF_FW_MBX_REG_LEN 4096
#define PF_FW_MBX 0x08400000
+#define PF_FW_ARQBAL (PF_FW_MBX)
+#define PF_FW_ARQBAH (PF_FW_MBX + 0x4)
+#define PF_FW_ARQLEN (PF_FW_MBX + 0x8)
+#define PF_FW_ARQLEN_ARQLEN_M GENMASK(12, 0)
+#define PF_FW_ARQLEN_ARQENABLE_S 31
+#define PF_FW_ARQLEN_ARQENABLE_M BIT(PF_FW_ARQLEN_ARQENABLE_S)
+#define PF_FW_ARQH_ARQH_M GENMASK(12, 0)
+#define PF_FW_ARQH (PF_FW_MBX + 0xC)
+#define PF_FW_ARQT (PF_FW_MBX + 0x10)
+
+#define PF_FW_ATQBAL (PF_FW_MBX + 0x14)
+#define PF_FW_ATQBAH (PF_FW_MBX + 0x18)
+#define PF_FW_ATQLEN (PF_FW_MBX + 0x1C)
+#define PF_FW_ATQLEN_ATQLEN_M GENMASK(9, 0)
+#define PF_FW_ATQLEN_ATQENABLE_S 31
+#define PF_FW_ATQLEN_ATQENABLE_M BIT(PF_FW_ATQLEN_ATQENABLE_S)
+#define PF_FW_ATQH_ATQH_M GENMASK(9, 0)
+#define PF_FW_ATQH (PF_FW_MBX + 0x20)
+#define PF_FW_ATQT (PF_FW_MBX + 0x24)
+
/* Reset registers */
#define PFGEN_RTRIG_REG_LEN 2048
#define PFGEN_RTRIG 0x08407000 /* Device resets */
+#define PFGEN_RSTAT 0x08407008 /* PFR status */
+#define PFGEN_RSTAT_PFR_STATE_M GENMASK(1, 0)
+#define PFGEN_CTRL 0x0840700C /* PFR trigger */
+#define PFGEN_CTRL_PFSWR BIT(0)
/**
* struct ixd_bar_region - BAR region description
@@ -25,4 +49,20 @@ struct ixd_bar_region {
resource_size_t size;
};
+/**
+ * struct ixd_reset_reg - structure for reset registers
+ * @rstat: offset of status in register
+ * @rstat_m: status mask
+ * @rstat_ok_v: value that indicates PFR completed status
+ * @rtrigger: offset of reset trigger in register
+ * @rtrigger_m: reset trigger mask
+ */
+struct ixd_reset_reg {
+ u32 rstat;
+ u32 rstat_m;
+ u32 rstat_ok_v;
+ u32 rtrigger;
+ u32 rtrigger_m;
+};
+
#endif /* _IXD_LAN_REGS_H_ */
diff --git a/drivers/net/ethernet/intel/ixd/ixd_lib.c b/drivers/net/ethernet/intel/ixd/ixd_lib.c
new file mode 100644
index 000000000000..42ba7e7da35c
--- /dev/null
+++ b/drivers/net/ethernet/intel/ixd/ixd_lib.c
@@ -0,0 +1,143 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/* Copyright (C) 2025 Intel Corporation */
+
+#include "ixd.h"
+
+#define IXD_DFLT_MBX_Q_LEN 64
+
+/**
+ * ixd_init_ctlq_create_info - Initialize control queue info for creation
+ * @info: destination
+ * @type: type of the queue to create
+ * @ctlq_reg: register assigned to the control queue
+ */
+static void ixd_init_ctlq_create_info(struct libeth_ctlq_create_info *info,
+ enum virtchnl2_queue_type type,
+ const struct libeth_ctlq_reg *ctlq_reg)
+{
+ *info = (struct libeth_ctlq_create_info) {
+ .type = type,
+ .id = -1,
+ .reg = *ctlq_reg,
+ .len = IXD_DFLT_MBX_Q_LEN,
+ };
+}
+
+/**
+ * ixd_init_libeth_xn_params - Initialize xn transaction manager creation info
+ * @params: destination
+ * @adapter: adapter info struct
+ * @ctlqs: list of the managed queues to create
+ * @num_queues: length of the queue list
+ */
+static void ixd_init_libeth_xn_params(struct libeth_ctlq_xn_init_params *params,
+ struct ixd_adapter *adapter,
+ struct libeth_ctlq_create_info *ctlqs,
+ uint num_queues)
+{
+ *params = (struct libeth_ctlq_xn_init_params){
+ .cctlq_info = ctlqs,
+ .ctx = &adapter->cp_ctx,
+ .num_qs = num_queues,
+ };
+}
+
+/**
+ * ixd_adapter_fill_dflt_ctlqs - Find default control queues and store them
+ * @adapter: adapter info struct
+ */
+static void ixd_adapter_fill_dflt_ctlqs(struct ixd_adapter *adapter)
+{
+ guard(spinlock)(&adapter->cp_ctx.ctlqs_lock);
+ struct libeth_ctlq_info *cq;
+
+ list_for_each_entry(cq, &adapter->cp_ctx.ctlqs, list) {
+ if (cq->qid != -1)
+ continue;
+ if (cq->type == VIRTCHNL2_QUEUE_TYPE_RX)
+ adapter->arq = cq;
+ else if (cq->type == VIRTCHNL2_QUEUE_TYPE_TX)
+ adapter->asq = cq;
+ }
+}
+
+/**
+ * ixd_init_dflt_mbx - Setup default mailbox parameters and make request
+ * @adapter: adapter info struct
+ *
+ * Return: %0 on success, negative errno code on failure
+ */
+int ixd_init_dflt_mbx(struct ixd_adapter *adapter)
+{
+ struct libeth_ctlq_create_info ctlqs_info[2];
+ struct libeth_ctlq_xn_init_params xn_params;
+ struct libeth_ctlq_reg ctlq_reg_tx;
+ struct libeth_ctlq_reg ctlq_reg_rx;
+ int err;
+
+ ixd_ctlq_reg_init(adapter, &ctlq_reg_tx, &ctlq_reg_rx);
+ ixd_init_ctlq_create_info(&ctlqs_info[0], VIRTCHNL2_QUEUE_TYPE_TX,
+ &ctlq_reg_tx);
+ ixd_init_ctlq_create_info(&ctlqs_info[1], VIRTCHNL2_QUEUE_TYPE_RX,
+ &ctlq_reg_rx);
+ ixd_init_libeth_xn_params(&xn_params, adapter, ctlqs_info,
+ ARRAY_SIZE(ctlqs_info));
+ err = libeth_ctlq_xn_init(&xn_params);
+ if (err)
+ return err;
+ adapter->xnm = xn_params.xnm;
+
+ ixd_adapter_fill_dflt_ctlqs(adapter);
+
+ if (!adapter->asq || !adapter->arq) {
+ libeth_ctlq_xn_deinit(adapter->xnm, &adapter->cp_ctx);
+ return -ENOENT;
+ }
+
+ return 0;
+}
+
+/**
+ * ixd_deinit_dflt_mbx - Deinitialize default mailbox
+ * @adapter: adapter info struct
+ */
+void ixd_deinit_dflt_mbx(struct ixd_adapter *adapter)
+{
+ if (adapter->arq || adapter->asq)
+ libeth_ctlq_xn_deinit(adapter->xnm, &adapter->cp_ctx);
+
+ adapter->arq = NULL;
+ adapter->asq = NULL;
+ adapter->xnm = NULL;
+}
+
+/**
+ * ixd_init_task - Initialize after reset
+ * @work: init work struct
+ */
+void ixd_init_task(struct work_struct *work)
+{
+ struct ixd_adapter *adapter;
+ int err;
+
+ adapter = container_of(work, struct ixd_adapter,
+ init_task.init_work.work);
+
+ if (!ixd_check_reset_complete(adapter)) {
+ if (++adapter->init_task.reset_retries < 10)
+ queue_delayed_work(system_unbound_wq,
+ &adapter->init_task.init_work,
+ msecs_to_jiffies(500));
+ else
+ dev_err(ixd_to_dev(adapter),
+ "Device reset failed. The driver was unable to contact the device's firmware. Check that the FW is running.\n");
+ return;
+ }
+
+ adapter->init_task.reset_retries = 0;
+ err = ixd_init_dflt_mbx(adapter);
+ if (err)
+ dev_err(ixd_to_dev(adapter),
+ "Failed to initialize the default mailbox: %pe\n",
+ ERR_PTR(err));
+}
diff --git a/drivers/net/ethernet/intel/ixd/ixd_main.c b/drivers/net/ethernet/intel/ixd/ixd_main.c
index 1f59db0b8fe7..bba6fa2dda57 100644
--- a/drivers/net/ethernet/intel/ixd/ixd_main.c
+++ b/drivers/net/ethernet/intel/ixd/ixd_main.c
@@ -5,6 +5,7 @@
#include "ixd_lan_regs.h"
MODULE_DESCRIPTION("Intel(R) Control Plane Function Device Driver");
+MODULE_IMPORT_NS("LIBETH_CP");
MODULE_IMPORT_NS("LIBETH_PCI");
MODULE_LICENSE("GPL");
@@ -16,7 +17,13 @@ static void ixd_remove(struct pci_dev *pdev)
{
struct ixd_adapter *adapter = pci_get_drvdata(pdev);
- libeth_pci_unmap_all_mmio_regions(&adapter->hw);
+ /* Do not mix removal with (re)initialization */
+ cancel_delayed_work_sync(&adapter->init_task.init_work);
+ /* Leave the device clean on exit */
+ ixd_trigger_reset(adapter);
+ ixd_deinit_dflt_mbx(adapter);
+
+ libeth_pci_unmap_all_mmio_regions(&adapter->cp_ctx.mmio_info);
libeth_pci_deinit_dev(pdev);
}
@@ -52,7 +59,7 @@ static int ixd_iomap_regions(struct ixd_adapter *adapter)
};
for (int i = 0; i < ARRAY_SIZE(regions); i++) {
- struct libeth_mmio_info *mmio_info = &adapter->hw;
+ struct libeth_mmio_info *mmio_info = &adapter->cp_ctx.mmio_info;
bool map_ok;
map_ok = libeth_pci_map_mmio_region(mmio_info,
@@ -82,11 +89,15 @@ static int ixd_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
struct ixd_adapter *adapter;
int err;
+ if (WARN_ON(ent->device != IXD_DEV_ID_CPF))
+ return -EINVAL;
+
adapter = devm_kzalloc(&pdev->dev, sizeof(*adapter), GFP_KERNEL);
if (!adapter)
return -ENOMEM;
- adapter->hw.pdev = pdev;
- INIT_LIST_HEAD(&adapter->hw.mmio_list);
+
+ adapter->cp_ctx.mmio_info.pdev = pdev;
+ INIT_LIST_HEAD(&adapter->cp_ctx.mmio_info.mmio_list);
err = libeth_pci_init_dev(pdev);
if (err)
@@ -98,6 +109,13 @@ static int ixd_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
if (err)
goto deinit_dev;
+ INIT_DELAYED_WORK(&adapter->init_task.init_work,
+ ixd_init_task);
+
+ ixd_trigger_reset(adapter);
+ queue_delayed_work(system_unbound_wq, &adapter->init_task.init_work,
+ msecs_to_jiffies(500));
+
return 0;
deinit_dev:
--
2.47.0
^ permalink raw reply related [flat|nested] 36+ messages in thread* [PATCH iwl-next 13/14] ixd: add the core initialization
2025-04-08 12:47 [PATCH iwl-next 00/14] Introduce iXD driver Larysa Zaremba
` (11 preceding siblings ...)
2025-04-08 12:47 ` [PATCH iwl-next 12/14] ixd: add reset checks and initialize the mailbox Larysa Zaremba
@ 2025-04-08 12:47 ` Larysa Zaremba
2025-04-08 12:48 ` [PATCH iwl-next 14/14] ixd: add devlink support Larysa Zaremba
2025-04-16 15:49 ` [PATCH iwl-next 00/14] Introduce iXD driver Alexander Lobakin
14 siblings, 0 replies; 36+ messages in thread
From: Larysa Zaremba @ 2025-04-08 12:47 UTC (permalink / raw)
To: intel-wired-lan, Tony Nguyen
Cc: David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Simon Horman, Jonathan Corbet, Przemek Kitszel, Jiri Pirko,
Mustafa Ismail, Tatyana Nikolova, Andrew Lunn, Alexander Lobakin,
Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Larysa Zaremba, Sridhar Samudrala,
Jacob Keller, Michal Swiatkowski, Mateusz Polchlopek, Wenjun Wu,
Ahmed Zaki, netdev, linux-doc, linux-kernel, Karlsson, Magnus,
Emil Tantilov, Madhu Chittim, Josh Hay, Milena Olech,
pavan.kumar.linga, Singhai, Anjali
As the mailbox is setup, initialize the core. This makes use of the
send and receive mailbox message framework for virtchnl communication
between the driver and device Control Plane (CP).
To start with, driver confirms the virtchnl version with the CP. Once
that is done, it requests and gets the required capabilities and
resources needed such as max vectors, queues, vports etc.
Use a unified way of handling the virtchnl messages, where a single
function handles all related memory management and the caller only provides
the callbacks to fill the send buffer and to handle the response.
Place generic control queue message handling separately to facilitate the
addition of protocols other than virtchannel in the future.
Co-developed-by: Amritha Nambiar <amritha.nambiar@intel.com>
Signed-off-by: Amritha Nambiar <amritha.nambiar@intel.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Larysa Zaremba <larysa.zaremba@intel.com>
---
drivers/net/ethernet/intel/ixd/Makefile | 2 +
drivers/net/ethernet/intel/ixd/ixd.h | 10 +
drivers/net/ethernet/intel/ixd/ixd_ctlq.c | 148 +++++++++++++++
drivers/net/ethernet/intel/ixd/ixd_ctlq.h | 33 ++++
drivers/net/ethernet/intel/ixd/ixd_lib.c | 25 ++-
drivers/net/ethernet/intel/ixd/ixd_main.c | 3 +
drivers/net/ethernet/intel/ixd/ixd_virtchnl.c | 178 ++++++++++++++++++
drivers/net/ethernet/intel/ixd/ixd_virtchnl.h | 12 ++
8 files changed, 410 insertions(+), 1 deletion(-)
create mode 100644 drivers/net/ethernet/intel/ixd/ixd_ctlq.c
create mode 100644 drivers/net/ethernet/intel/ixd/ixd_ctlq.h
create mode 100644 drivers/net/ethernet/intel/ixd/ixd_virtchnl.c
create mode 100644 drivers/net/ethernet/intel/ixd/ixd_virtchnl.h
diff --git a/drivers/net/ethernet/intel/ixd/Makefile b/drivers/net/ethernet/intel/ixd/Makefile
index 164b2c86952f..90abf231fb16 100644
--- a/drivers/net/ethernet/intel/ixd/Makefile
+++ b/drivers/net/ethernet/intel/ixd/Makefile
@@ -6,5 +6,7 @@
obj-$(CONFIG_IXD) += ixd.o
ixd-y := ixd_main.o
+ixd-y += ixd_ctlq.o
ixd-y += ixd_dev.o
ixd-y += ixd_lib.o
+ixd-y += ixd_virtchnl.o
diff --git a/drivers/net/ethernet/intel/ixd/ixd.h b/drivers/net/ethernet/intel/ixd/ixd.h
index 7efa8db8c5ba..ca08b5ead3b5 100644
--- a/drivers/net/ethernet/intel/ixd/ixd.h
+++ b/drivers/net/ethernet/intel/ixd/ixd.h
@@ -10,19 +10,29 @@
* struct ixd_adapter - Data structure representing a CPF
* @cp_ctx: Control plane communication context
* @init_task: Delayed initialization after reset
+ * @mbx_task: Control queue Rx handling
* @xnm: virtchnl transaction manager
* @asq: Send control queue info
* @arq: Receive control queue info
+ * @vc_ver: Negotiated virtchnl version
+ * @caps: Negotiated virtchnl capabilities
*/
struct ixd_adapter {
struct libeth_ctlq_ctx cp_ctx;
struct {
struct delayed_work init_work;
u8 reset_retries;
+ u8 vc_retries;
} init_task;
+ struct delayed_work mbx_task;
struct libeth_ctlq_xn_manager *xnm;
struct libeth_ctlq_info *asq;
struct libeth_ctlq_info *arq;
+ struct {
+ u32 major;
+ u32 minor;
+ } vc_ver;
+ struct virtchnl2_get_capabilities caps;
};
/**
diff --git a/drivers/net/ethernet/intel/ixd/ixd_ctlq.c b/drivers/net/ethernet/intel/ixd/ixd_ctlq.c
new file mode 100644
index 000000000000..d17987c71eed
--- /dev/null
+++ b/drivers/net/ethernet/intel/ixd/ixd_ctlq.c
@@ -0,0 +1,148 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/* Copyright (C) 2025 Intel Corporation */
+
+#include "ixd.h"
+#include "ixd_ctlq.h"
+#include "ixd_virtchnl.h"
+
+/**
+ * ixd_ctlq_clean_sq - Clean the send control queue after sending the message
+ * @adapter: The adapter that sent the messages
+ * @num_sent: Number of sent messages to be released
+ *
+ * Free the libeth send resources after sending the message and handling
+ * the response.
+ */
+static void ixd_ctlq_clean_sq(struct ixd_adapter *adapter, u16 num_sent)
+{
+ if (!num_sent)
+ return;
+
+ struct libeth_ctlq_xn_clean_params params = {
+ .ctlq = adapter->asq,
+ .ctx = &adapter->cp_ctx,
+ .num_msgs = num_sent,
+ .rel_tx_buf = kfree,
+ };
+
+ libeth_ctlq_xn_send_clean(¶ms);
+}
+
+/**
+ * ixd_ctlq_init_sparams - Initialize control queue send parameters
+ * @adapter: The adapter with initialized mailbox
+ * @sparams: Parameters to initialize
+ * @msg_buf: DMA-mappable pointer to the message being sent
+ * @msg_size: Message size
+ */
+static void ixd_ctlq_init_sparams(struct ixd_adapter *adapter,
+ struct libeth_ctlq_xn_send_params *sparams,
+ void *msg_buf, size_t msg_size)
+{
+ *sparams = (struct libeth_ctlq_xn_send_params) {
+ .rel_tx_buf = kfree,
+ .xnm = adapter->xnm,
+ .ctlq = adapter->asq,
+ .timeout_ms = IXD_CTLQ_TIMEOUT,
+ .send_buf = (struct kvec) {
+ .iov_base = msg_buf,
+ .iov_len = msg_size,
+ },
+ };
+}
+
+/**
+ * ixd_ctlq_do_req - Perform a standard virtchnl request
+ * @adapter: The adapter with initialized mailbox
+ * @req: virtchnl request description
+ *
+ * Return: %0 if a message was sent and received a response
+ * that was successfully handled by the custom callback,
+ * negative error otherwise.
+ */
+int ixd_ctlq_do_req(struct ixd_adapter *adapter, const struct ixd_ctlq_req *req)
+{
+ struct libeth_ctlq_xn_send_params send_params = {};
+ u8 onstack_send_buff[LIBETH_CP_TX_COPYBREAK];
+ struct kvec *recv_mem;
+ void *send_buff;
+ int err;
+
+ send_buff = libeth_cp_can_send_onstack(req->send_size) ?
+ &onstack_send_buff : kzalloc(req->send_size, GFP_KERNEL);
+ if (!send_buff)
+ return -ENOMEM;
+
+ ixd_ctlq_init_sparams(adapter, &send_params, send_buff,
+ req->send_size);
+
+ send_params.chnl_opcode = req->opcode;
+
+ if (req->send_buff_init)
+ req->send_buff_init(adapter, send_buff, req->ctx);
+
+ err = libeth_ctlq_xn_send(&send_params);
+ if (err)
+ return err;
+
+ recv_mem = &send_params.recv_mem;
+ if (req->recv_process)
+ err = req->recv_process(adapter, recv_mem->iov_base,
+ recv_mem->iov_len, req->ctx);
+
+ ixd_ctlq_clean_sq(adapter, 1);
+ libeth_ctlq_release_rx_buf(adapter->arq, recv_mem);
+
+ return err;
+}
+
+/**
+ * ixd_ctlq_handle_msg - Default control queue message handler
+ * @ctx: Control plane communication context
+ * @msg: Message received
+ */
+static void ixd_ctlq_handle_msg(struct libeth_ctlq_ctx *ctx,
+ struct libeth_ctlq_msg *msg)
+{
+ struct ixd_adapter *adapter = pci_get_drvdata(ctx->mmio_info.pdev);
+
+ if (ixd_vc_can_handle_msg(msg))
+ ixd_vc_recv_event_msg(adapter, msg);
+ else
+ dev_dbg_ratelimited(ixd_to_dev(adapter),
+ "Received an unsupported opcode 0x%x from the CP\n",
+ msg->chnl_opcode);
+
+ libeth_ctlq_release_rx_buf(adapter->arq, &msg->recv_mem);
+}
+
+/**
+ * ixd_ctlq_recv_mb_msg - Receive a potential message over mailbox periodically
+ * @adapter: The adapter with initialized mailbox
+ */
+static void ixd_ctlq_recv_mb_msg(struct ixd_adapter *adapter)
+{
+ struct libeth_ctlq_xn_recv_params xn_params = {
+ .xnm = adapter->xnm,
+ .ctlq = adapter->arq,
+ .ctlq_msg_handler = ixd_ctlq_handle_msg,
+ };
+
+ libeth_ctlq_xn_recv(&xn_params);
+}
+
+/**
+ * ixd_ctlq_rx_task - Periodically check for mailbox responses and events
+ * @work: work handle
+ */
+void ixd_ctlq_rx_task(struct work_struct *work)
+{
+ struct ixd_adapter *adapter;
+
+ adapter = container_of(work, struct ixd_adapter, mbx_task.work);
+
+ queue_delayed_work(system_unbound_wq, &adapter->mbx_task,
+ msecs_to_jiffies(300));
+
+ ixd_ctlq_recv_mb_msg(adapter);
+}
diff --git a/drivers/net/ethernet/intel/ixd/ixd_ctlq.h b/drivers/net/ethernet/intel/ixd/ixd_ctlq.h
new file mode 100644
index 000000000000..f450a3a0828f
--- /dev/null
+++ b/drivers/net/ethernet/intel/ixd/ixd_ctlq.h
@@ -0,0 +1,33 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/* Copyright (C) 2025 Intel Corporation */
+
+#ifndef _IXD_CTLQ_H_
+#define _IXD_CTLQ_H_
+
+#include "linux/intel/virtchnl2.h"
+
+#define IXD_CTLQ_TIMEOUT 2000
+
+/**
+ * struct ixd_ctlq_req - Standard virtchnl request description
+ * @opcode: protocol opcode, only virtchnl2 is needed for now
+ * @send_size: required length of the send buffer
+ * @send_buff_init: function to initialize the allocated send buffer
+ * @recv_process: function to handle the CP response
+ * @ctx: additional context for callbacks
+ */
+struct ixd_ctlq_req {
+ enum virtchnl2_op opcode;
+ size_t send_size;
+ void (*send_buff_init)(struct ixd_adapter *adapter, void *send_buff,
+ void *ctx);
+ int (*recv_process)(struct ixd_adapter *adapter, void *recv_buff,
+ size_t recv_size, void *ctx);
+ void *ctx;
+};
+
+int ixd_ctlq_do_req(struct ixd_adapter *adapter,
+ const struct ixd_ctlq_req *req);
+void ixd_ctlq_rx_task(struct work_struct *work);
+
+#endif /* _IXD_CTLQ_H_ */
diff --git a/drivers/net/ethernet/intel/ixd/ixd_lib.c b/drivers/net/ethernet/intel/ixd/ixd_lib.c
index 42ba7e7da35c..201545a74cd8 100644
--- a/drivers/net/ethernet/intel/ixd/ixd_lib.c
+++ b/drivers/net/ethernet/intel/ixd/ixd_lib.c
@@ -2,6 +2,7 @@
/* Copyright (C) 2025 Intel Corporation */
#include "ixd.h"
+#include "ixd_virtchnl.h"
#define IXD_DFLT_MBX_Q_LEN 64
@@ -94,6 +95,8 @@ int ixd_init_dflt_mbx(struct ixd_adapter *adapter)
return -ENOENT;
}
+ queue_delayed_work(system_unbound_wq, &adapter->mbx_task, 0);
+
return 0;
}
@@ -103,6 +106,8 @@ int ixd_init_dflt_mbx(struct ixd_adapter *adapter)
*/
void ixd_deinit_dflt_mbx(struct ixd_adapter *adapter)
{
+ cancel_delayed_work_sync(&adapter->mbx_task);
+
if (adapter->arq || adapter->asq)
libeth_ctlq_xn_deinit(adapter->xnm, &adapter->cp_ctx);
@@ -136,8 +141,26 @@ void ixd_init_task(struct work_struct *work)
adapter->init_task.reset_retries = 0;
err = ixd_init_dflt_mbx(adapter);
- if (err)
+ if (err) {
dev_err(ixd_to_dev(adapter),
"Failed to initialize the default mailbox: %pe\n",
ERR_PTR(err));
+ return;
+ }
+
+ if (!ixd_vc_dev_init(adapter)) {
+ adapter->init_task.vc_retries = 0;
+ return;
+ }
+
+ ixd_deinit_dflt_mbx(adapter);
+ if (++adapter->init_task.vc_retries > 5) {
+ dev_err(ixd_to_dev(adapter),
+ "Failed to establish mailbox communications with the hardware\n");
+ return;
+ }
+
+ ixd_trigger_reset(adapter);
+ queue_delayed_work(system_unbound_wq, &adapter->init_task.init_work,
+ msecs_to_jiffies(500));
}
diff --git a/drivers/net/ethernet/intel/ixd/ixd_main.c b/drivers/net/ethernet/intel/ixd/ixd_main.c
index bba6fa2dda57..97679cf4527f 100644
--- a/drivers/net/ethernet/intel/ixd/ixd_main.c
+++ b/drivers/net/ethernet/intel/ixd/ixd_main.c
@@ -2,6 +2,7 @@
/* Copyright (C) 2025 Intel Corporation */
#include "ixd.h"
+#include "ixd_ctlq.h"
#include "ixd_lan_regs.h"
MODULE_DESCRIPTION("Intel(R) Control Plane Function Device Driver");
@@ -19,6 +20,7 @@ static void ixd_remove(struct pci_dev *pdev)
/* Do not mix removal with (re)initialization */
cancel_delayed_work_sync(&adapter->init_task.init_work);
+
/* Leave the device clean on exit */
ixd_trigger_reset(adapter);
ixd_deinit_dflt_mbx(adapter);
@@ -111,6 +113,7 @@ static int ixd_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
INIT_DELAYED_WORK(&adapter->init_task.init_work,
ixd_init_task);
+ INIT_DELAYED_WORK(&adapter->mbx_task, ixd_ctlq_rx_task);
ixd_trigger_reset(adapter);
queue_delayed_work(system_unbound_wq, &adapter->init_task.init_work,
diff --git a/drivers/net/ethernet/intel/ixd/ixd_virtchnl.c b/drivers/net/ethernet/intel/ixd/ixd_virtchnl.c
new file mode 100644
index 000000000000..a1c2ccf36175
--- /dev/null
+++ b/drivers/net/ethernet/intel/ixd/ixd_virtchnl.c
@@ -0,0 +1,178 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/* Copyright (C) 2025 Intel Corporation */
+
+#include "ixd.h"
+#include "ixd_ctlq.h"
+#include "ixd_virtchnl.h"
+
+/**
+ * ixd_vc_recv_event_msg - Handle virtchnl event message
+ * @adapter: The adapter handling the message
+ * @ctlq_msg: Message received
+ */
+void ixd_vc_recv_event_msg(struct ixd_adapter *adapter,
+ struct libeth_ctlq_msg *ctlq_msg)
+{
+ int payload_size = ctlq_msg->data_len;
+ struct virtchnl2_event *v2e;
+
+ if (payload_size < sizeof(*v2e)) {
+ dev_warn_ratelimited(ixd_to_dev(adapter),
+ "Failed to receive valid payload for event msg (op 0x%X len %u)\n",
+ ctlq_msg->chnl_opcode,
+ payload_size);
+ return;
+ }
+
+ v2e = (struct virtchnl2_event *)ctlq_msg->recv_mem.iov_base;
+
+ dev_dbg(ixd_to_dev(adapter), "Got event 0x%X from the CP\n",
+ le32_to_cpu(v2e->event));
+}
+
+/**
+ * ixd_vc_can_handle_msg - Decide if an event has to be handled by virtchnl code
+ * @ctlq_msg: Message received
+ *
+ * Return: %true if virtchnl code can handle the event, %false otherwise
+ */
+bool ixd_vc_can_handle_msg(struct libeth_ctlq_msg *ctlq_msg)
+{
+ return ctlq_msg->chnl_opcode == VIRTCHNL2_OP_EVENT;
+}
+
+/**
+ * ixd_handle_caps - Handle VIRTCHNL2_OP_GET_CAPS response
+ * @adapter: The adapter for which the capabilities are being updated
+ * @recv_buff: Buffer containing the response
+ * @recv_size: Response buffer size
+ * @ctx: unused
+ *
+ * Return: %0 if the response format is correct and was handled as expected,
+ * negative error otherwise.
+ */
+static int ixd_handle_caps(struct ixd_adapter *adapter, void *recv_buff,
+ size_t recv_size, void *ctx)
+{
+ if (recv_size < sizeof(adapter->caps))
+ return -EBADMSG;
+
+ adapter->caps = *(typeof(adapter->caps) *)recv_buff;
+
+ return 0;
+}
+
+/**
+ * ixd_req_vc_caps - Request and save device capability
+ * @adapter: The adapter to get the capabilities for
+ *
+ * Return: success or error if sending the get capability message fails
+ */
+static int ixd_req_vc_caps(struct ixd_adapter *adapter)
+{
+ const struct ixd_ctlq_req req = {
+ .opcode = VIRTCHNL2_OP_GET_CAPS,
+ .send_size = sizeof(struct virtchnl2_get_capabilities),
+ .ctx = NULL,
+ .send_buff_init = NULL,
+ .recv_process = ixd_handle_caps,
+ };
+
+ return ixd_ctlq_do_req(adapter, &req);
+}
+
+/**
+ * ixd_get_vc_ver - Get version info from adapter
+ *
+ * Return: filled in virtchannel2 version info, ready for sending
+ */
+static struct virtchnl2_version_info ixd_get_vc_ver(void)
+{
+ return (struct virtchnl2_version_info) {
+ .major = cpu_to_le32(VIRTCHNL2_VERSION_MAJOR_2),
+ .minor = cpu_to_le32(VIRTCHNL2_VERSION_MINOR_0),
+ };
+}
+
+static void ixd_fill_vc_ver(struct ixd_adapter *adapter, void *send_buff,
+ void *ctx)
+{
+ *(struct virtchnl2_version_info *)send_buff = ixd_get_vc_ver();
+}
+
+/**
+ * ixd_handle_vc_ver - Handle VIRTCHNL2_OP_VERSION response
+ * @adapter: The adapter for which the version is being updated
+ * @recv_buff: Buffer containing the response
+ * @recv_size: Response buffer size
+ * @ctx: Unused
+ *
+ * Return: %0 if the response format is correct and was handled as expected,
+ * negative error otherwise.
+ */
+static int ixd_handle_vc_ver(struct ixd_adapter *adapter, void *recv_buff,
+ size_t recv_size, void *ctx)
+{
+ struct virtchnl2_version_info need_ver = ixd_get_vc_ver();
+ struct virtchnl2_version_info *recv_ver;
+
+ if (recv_size < sizeof(need_ver))
+ return -EBADMSG;
+
+ recv_ver = recv_buff;
+ if (le32_to_cpu(need_ver.major) > le32_to_cpu(recv_ver->major))
+ return -EOPNOTSUPP;
+
+ adapter->vc_ver.major = le32_to_cpu(recv_ver->major);
+ adapter->vc_ver.minor = le32_to_cpu(recv_ver->minor);
+
+ return 0;
+}
+
+/**
+ * ixd_req_vc_version - Request and save Virtchannel2 version
+ * @adapter: The adapter to get the version for
+ *
+ * Return: success or error if sending fails or the response was not as expected
+ */
+static int ixd_req_vc_version(struct ixd_adapter *adapter)
+{
+ const struct ixd_ctlq_req req = {
+ .opcode = VIRTCHNL2_OP_VERSION,
+ .send_size = sizeof(struct virtchnl2_version_info),
+ .ctx = NULL,
+ .send_buff_init = ixd_fill_vc_ver,
+ .recv_process = ixd_handle_vc_ver,
+ };
+
+ return ixd_ctlq_do_req(adapter, &req);
+}
+
+/**
+ * ixd_vc_dev_init - virtchnl device core initialization
+ * @adapter: device information
+ *
+ * Return: %0 on success or error if any step of the initialization fails
+ */
+int ixd_vc_dev_init(struct ixd_adapter *adapter)
+{
+ int err;
+
+ err = ixd_req_vc_version(adapter);
+ if (err) {
+ dev_warn(ixd_to_dev(adapter),
+ "Getting virtchnl version failed, error=%pe\n",
+ ERR_PTR(err));
+ return err;
+ }
+
+ err = ixd_req_vc_caps(adapter);
+ if (err) {
+ dev_warn(ixd_to_dev(adapter),
+ "Getting virtchnl capabilities failed, error=%pe\n",
+ ERR_PTR(err));
+ return err;
+ }
+
+ return err;
+}
diff --git a/drivers/net/ethernet/intel/ixd/ixd_virtchnl.h b/drivers/net/ethernet/intel/ixd/ixd_virtchnl.h
new file mode 100644
index 000000000000..65da427e91b7
--- /dev/null
+++ b/drivers/net/ethernet/intel/ixd/ixd_virtchnl.h
@@ -0,0 +1,12 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/* Copyright (C) 2025 Intel Corporation */
+
+#ifndef _IXD_VIRTCHNL_H_
+#define _IXD_VIRTCHNL_H_
+
+int ixd_vc_dev_init(struct ixd_adapter *adapter);
+bool ixd_vc_can_handle_msg(struct libeth_ctlq_msg *ctlq_msg);
+void ixd_vc_recv_event_msg(struct ixd_adapter *adapter,
+ struct libeth_ctlq_msg *ctlq_msg);
+
+#endif /* _IXD_VIRTCHNL_H_ */
--
2.47.0
^ permalink raw reply related [flat|nested] 36+ messages in thread* [PATCH iwl-next 14/14] ixd: add devlink support
2025-04-08 12:47 [PATCH iwl-next 00/14] Introduce iXD driver Larysa Zaremba
` (12 preceding siblings ...)
2025-04-08 12:47 ` [PATCH iwl-next 13/14] ixd: add the core initialization Larysa Zaremba
@ 2025-04-08 12:48 ` Larysa Zaremba
2025-04-15 18:54 ` Tony Nguyen
2025-04-16 15:49 ` [PATCH iwl-next 00/14] Introduce iXD driver Alexander Lobakin
14 siblings, 1 reply; 36+ messages in thread
From: Larysa Zaremba @ 2025-04-08 12:48 UTC (permalink / raw)
To: intel-wired-lan, Tony Nguyen
Cc: David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Simon Horman, Jonathan Corbet, Przemek Kitszel, Jiri Pirko,
Mustafa Ismail, Tatyana Nikolova, Andrew Lunn, Alexander Lobakin,
Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Larysa Zaremba, Sridhar Samudrala,
Jacob Keller, Michal Swiatkowski, Mateusz Polchlopek, Wenjun Wu,
Ahmed Zaki, netdev, linux-doc, linux-kernel, Karlsson, Magnus,
Emil Tantilov, Madhu Chittim, Josh Hay, Milena Olech,
pavan.kumar.linga, Singhai, Anjali
From: Amritha Nambiar <amritha.nambiar@intel.com>
Enable initial support for the devlink interface with the ixd
driver. The ixd hardware is a single function PCIe device. So, the
PCIe adapter gets its own devlink instance to manage device-wide
resources or configuration.
$ devlink dev show
pci/0000:83:00.6
$ devlink dev info pci/0000:83:00.6
pci/0000:83:00.6:
driver ixd
serial_number 00-a0-c9-ff-ff-23-45-67
versions:
fixed:
device.type MEV
running:
cp 0.0
virtchnl 2.0
Signed-off-by: Amritha Nambiar <amritha.nambiar@intel.com>
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Signed-off-by: Larysa Zaremba <larysa.zaremba@intel.com>
---
Documentation/networking/devlink/ixd.rst | 35 +++++++
drivers/net/ethernet/intel/ixd/Kconfig | 1 +
drivers/net/ethernet/intel/ixd/Makefile | 1 +
drivers/net/ethernet/intel/ixd/ixd_devlink.c | 105 +++++++++++++++++++
drivers/net/ethernet/intel/ixd/ixd_devlink.h | 44 ++++++++
drivers/net/ethernet/intel/ixd/ixd_main.c | 13 ++-
6 files changed, 196 insertions(+), 3 deletions(-)
create mode 100644 Documentation/networking/devlink/ixd.rst
create mode 100644 drivers/net/ethernet/intel/ixd/ixd_devlink.c
create mode 100644 drivers/net/ethernet/intel/ixd/ixd_devlink.h
diff --git a/Documentation/networking/devlink/ixd.rst b/Documentation/networking/devlink/ixd.rst
new file mode 100644
index 000000000000..81b28ffb00f6
--- /dev/null
+++ b/Documentation/networking/devlink/ixd.rst
@@ -0,0 +1,35 @@
+.. SPDX-License-Identifier: GPL-2.0
+
+===================
+ixd devlink support
+===================
+
+This document describes the devlink features implemented by the ``ixd``
+device driver.
+
+Info versions
+=============
+
+The ``ixd`` driver reports the following versions
+
+.. list-table:: devlink info versions implemented
+ :widths: 5 5 5 90
+
+ * - Name
+ - Type
+ - Example
+ - Description
+ * - ``device.type``
+ - fixed
+ - MEV
+ - The hardware type for this device
+ * - ``cp``
+ - running
+ - 0.0
+ - Version number (major.minor) of the Control Plane software
+ running on the device.
+ * - ``virtchnl``
+ - running
+ - 2.0
+ - 2-digit version number (major.minor) of the communication channel
+ (virtchnl) used by the device.
diff --git a/drivers/net/ethernet/intel/ixd/Kconfig b/drivers/net/ethernet/intel/ixd/Kconfig
index bcaf06a874aa..d4f4fa4678bd 100644
--- a/drivers/net/ethernet/intel/ixd/Kconfig
+++ b/drivers/net/ethernet/intel/ixd/Kconfig
@@ -7,6 +7,7 @@ config IXD
select LIBETH
select LIBETH_CP
select LIBETH_PCI
+ select NET_DEVLINK
help
This driver supports Intel(R) Control Plane PCI Function
of Intel E2100 and later IPUs and FNICs.
diff --git a/drivers/net/ethernet/intel/ixd/Makefile b/drivers/net/ethernet/intel/ixd/Makefile
index 90abf231fb16..03760a2580b9 100644
--- a/drivers/net/ethernet/intel/ixd/Makefile
+++ b/drivers/net/ethernet/intel/ixd/Makefile
@@ -8,5 +8,6 @@ obj-$(CONFIG_IXD) += ixd.o
ixd-y := ixd_main.o
ixd-y += ixd_ctlq.o
ixd-y += ixd_dev.o
+ixd-y += ixd_devlink.o
ixd-y += ixd_lib.o
ixd-y += ixd_virtchnl.o
diff --git a/drivers/net/ethernet/intel/ixd/ixd_devlink.c b/drivers/net/ethernet/intel/ixd/ixd_devlink.c
new file mode 100644
index 000000000000..6f60cfe4fab2
--- /dev/null
+++ b/drivers/net/ethernet/intel/ixd/ixd_devlink.c
@@ -0,0 +1,105 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2025, Intel Corporation. */
+
+#include "ixd.h"
+#include "ixd_devlink.h"
+
+#define IXD_DEVLINK_INFO_LEN 128
+
+/**
+ * ixd_fill_dsn - Get the serial number for the ixd device
+ * @adapter: adapter to query
+ * @buf: storage buffer for the info request
+ */
+static void ixd_fill_dsn(struct ixd_adapter *adapter, char *buf)
+{
+ u8 dsn[8];
+
+ /* Copy the DSN into an array in Big Endian format */
+ put_unaligned_be64(pci_get_dsn(adapter->cp_ctx.mmio_info.pdev), dsn);
+
+ snprintf(buf, IXD_DEVLINK_INFO_LEN, "%8phD", dsn);
+}
+
+/**
+ * ixd_fill_device_name - Get the name of the underlying hardware
+ * @adapter: adapter to query
+ * @buf: storage buffer for the info request
+ * @buf_size: size of the storage buffer
+ */
+static void ixd_fill_device_name(struct ixd_adapter *adapter, char *buf,
+ size_t buf_size)
+{
+ if (adapter->caps.device_type == VIRTCHNL2_MEV_DEVICE)
+ snprintf(buf, buf_size, "%s", "MEV");
+ else
+ snprintf(buf, buf_size, "%s", "UNKNOWN");
+}
+
+/**
+ * ixd_devlink_info_get - .info_get devlink handler
+ * @devlink: devlink instance structure
+ * @req: the devlink info request
+ * @extack: extended netdev ack structure
+ *
+ * Callback for the devlink .info_get operation. Reports information about the
+ * device.
+ *
+ * Return: zero on success or an error code on failure.
+ */
+static int ixd_devlink_info_get(struct devlink *devlink,
+ struct devlink_info_req *req,
+ struct netlink_ext_ack *extack)
+{
+ struct ixd_adapter *adapter = devlink_priv(devlink);
+ char buf[IXD_DEVLINK_INFO_LEN];
+ int err;
+
+ ixd_fill_dsn(adapter, buf);
+ err = devlink_info_serial_number_put(req, buf);
+ if (err)
+ return err;
+
+ ixd_fill_device_name(adapter, buf, IXD_DEVLINK_INFO_LEN);
+ err = devlink_info_version_fixed_put(req, "device.type", buf);
+ if (err)
+ return err;
+
+ snprintf(buf, sizeof(buf), "%u.%u",
+ le16_to_cpu(adapter->caps.cp_ver_major),
+ le16_to_cpu(adapter->caps.cp_ver_minor));
+
+ err = devlink_info_version_running_put(req, "cp", buf);
+ if (err)
+ return err;
+
+ snprintf(buf, sizeof(buf), "%u.%u",
+ adapter->vc_ver.major, adapter->vc_ver.minor);
+
+ return devlink_info_version_running_put(req, "virtchnl", buf);
+}
+
+static const struct devlink_ops ixd_devlink_ops = {
+ .info_get = ixd_devlink_info_get,
+};
+
+/**
+ * ixd_adapter_alloc - Allocate devlink and return adapter pointer
+ * @dev: the device to allocate for
+ *
+ * Allocate a devlink instance for this device and return the private area as
+ * the adapter structure.
+ *
+ * Return: adapter structure on success, NULL on failure
+ */
+struct ixd_adapter *ixd_adapter_alloc(struct device *dev)
+{
+ struct devlink *devlink;
+
+ devlink = devlink_alloc(&ixd_devlink_ops, sizeof(struct ixd_adapter),
+ dev);
+ if (!devlink)
+ return NULL;
+
+ return devlink_priv(devlink);
+}
diff --git a/drivers/net/ethernet/intel/ixd/ixd_devlink.h b/drivers/net/ethernet/intel/ixd/ixd_devlink.h
new file mode 100644
index 000000000000..c43ce0655de2
--- /dev/null
+++ b/drivers/net/ethernet/intel/ixd/ixd_devlink.h
@@ -0,0 +1,44 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/* Copyright (c) 2025, Intel Corporation. */
+
+#ifndef _IXD_DEVLINK_H_
+#define _IXD_DEVLINK_H_
+#include <net/devlink.h>
+
+struct ixd_adapter *ixd_adapter_alloc(struct device *dev);
+
+/**
+ * ixd_devlink_free - teardown the devlink
+ * @adapter: the adapter structure to free
+ *
+ */
+static inline void ixd_devlink_free(struct ixd_adapter *adapter)
+{
+ struct devlink *devlink = priv_to_devlink(adapter);
+
+ devlink_free(devlink);
+}
+
+/**
+ * ixd_devlink_unregister - Unregister devlink resources for this adapter.
+ * @adapter: the adapter structure to cleanup
+ *
+ * Releases resources used by devlink and cleans up associated memory.
+ */
+static inline void ixd_devlink_unregister(struct ixd_adapter *adapter)
+{
+ devlink_unregister(priv_to_devlink(adapter));
+}
+
+/**
+ * ixd_devlink_register - Register devlink interface for this adapter
+ * @adapter: pointer to ixd adapter structure to be associated with devlink
+ *
+ * Register the devlink instance associated with this adapter
+ */
+static inline void ixd_devlink_register(struct ixd_adapter *adapter)
+{
+ devlink_register(priv_to_devlink(adapter));
+}
+
+#endif /* _IXD_DEVLINK_H_ */
diff --git a/drivers/net/ethernet/intel/ixd/ixd_main.c b/drivers/net/ethernet/intel/ixd/ixd_main.c
index 97679cf4527f..8312c25f21f8 100644
--- a/drivers/net/ethernet/intel/ixd/ixd_main.c
+++ b/drivers/net/ethernet/intel/ixd/ixd_main.c
@@ -4,6 +4,7 @@
#include "ixd.h"
#include "ixd_ctlq.h"
#include "ixd_lan_regs.h"
+#include "ixd_devlink.h"
MODULE_DESCRIPTION("Intel(R) Control Plane Function Device Driver");
MODULE_IMPORT_NS("LIBETH_CP");
@@ -21,12 +22,15 @@ static void ixd_remove(struct pci_dev *pdev)
/* Do not mix removal with (re)initialization */
cancel_delayed_work_sync(&adapter->init_task.init_work);
+ ixd_devlink_unregister(adapter);
+
/* Leave the device clean on exit */
ixd_trigger_reset(adapter);
ixd_deinit_dflt_mbx(adapter);
libeth_pci_unmap_all_mmio_regions(&adapter->cp_ctx.mmio_info);
libeth_pci_deinit_dev(pdev);
+ ixd_devlink_free(adapter);
}
/**
@@ -94,7 +98,7 @@ static int ixd_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
if (WARN_ON(ent->device != IXD_DEV_ID_CPF))
return -EINVAL;
- adapter = devm_kzalloc(&pdev->dev, sizeof(*adapter), GFP_KERNEL);
+ adapter = ixd_adapter_alloc(&pdev->dev);
if (!adapter)
return -ENOMEM;
@@ -103,7 +107,7 @@ static int ixd_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
err = libeth_pci_init_dev(pdev);
if (err)
- return err;
+ goto free_adapter;
pci_set_drvdata(pdev, adapter);
@@ -119,11 +123,14 @@ static int ixd_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
queue_delayed_work(system_unbound_wq, &adapter->init_task.init_work,
msecs_to_jiffies(500));
+ ixd_devlink_register(adapter);
+
return 0;
deinit_dev:
libeth_pci_deinit_dev(pdev);
-
+free_adapter:
+ ixd_devlink_free(adapter);
return err;
}
--
2.47.0
^ permalink raw reply related [flat|nested] 36+ messages in thread* Re: [PATCH iwl-next 14/14] ixd: add devlink support
2025-04-08 12:48 ` [PATCH iwl-next 14/14] ixd: add devlink support Larysa Zaremba
@ 2025-04-15 18:54 ` Tony Nguyen
0 siblings, 0 replies; 36+ messages in thread
From: Tony Nguyen @ 2025-04-15 18:54 UTC (permalink / raw)
To: Larysa Zaremba, intel-wired-lan
Cc: David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Simon Horman, Jonathan Corbet, Przemek Kitszel, Jiri Pirko,
Mustafa Ismail, Tatyana Nikolova, Andrew Lunn, Alexander Lobakin,
Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Sridhar Samudrala, Jacob Keller,
Michal Swiatkowski, Mateusz Polchlopek, Wenjun Wu, Ahmed Zaki,
netdev, linux-doc, linux-kernel, Karlsson, Magnus, Emil Tantilov,
Madhu Chittim, Josh Hay, Milena Olech, pavan.kumar.linga,
Singhai, Anjali
On 4/8/2025 5:48 AM, Larysa Zaremba wrote:
> From: Amritha Nambiar <amritha.nambiar@intel.com>
>
> Enable initial support for the devlink interface with the ixd
> driver. The ixd hardware is a single function PCIe device. So, the
> PCIe adapter gets its own devlink instance to manage device-wide
> resources or configuration.
>
> $ devlink dev show
> pci/0000:83:00.6
>
> $ devlink dev info pci/0000:83:00.6
> pci/0000:83:00.6:
> driver ixd
> serial_number 00-a0-c9-ff-ff-23-45-67
> versions:
> fixed:
> device.type MEV
> running:
> cp 0.0
> virtchnl 2.0
>
> Signed-off-by: Amritha Nambiar <amritha.nambiar@intel.com>
> Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
> Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
> Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
> Signed-off-by: Larysa Zaremba <larysa.zaremba@intel.com>
> ---
> Documentation/networking/devlink/ixd.rst | 35 +++++++
> drivers/net/ethernet/intel/ixd/Kconfig | 1 +
> drivers/net/ethernet/intel/ixd/Makefile | 1 +
> drivers/net/ethernet/intel/ixd/ixd_devlink.c | 105 +++++++++++++++++++
> drivers/net/ethernet/intel/ixd/ixd_devlink.h | 44 ++++++++
> drivers/net/ethernet/intel/ixd/ixd_main.c | 13 ++-
> 6 files changed, 196 insertions(+), 3 deletions(-)
> create mode 100644 Documentation/networking/devlink/ixd.rst
> create mode 100644 drivers/net/ethernet/intel/ixd/ixd_devlink.c
> create mode 100644 drivers/net/ethernet/intel/ixd/ixd_devlink.h
>
> diff --git a/Documentation/networking/devlink/ixd.rst b/Documentation/networking/devlink/ixd.rst
> new file mode 100644
> index 000000000000..81b28ffb00f6
> --- /dev/null
> +++ b/Documentation/networking/devlink/ixd.rst
The index needs to be updated as well:
Documentation/networking/devlink/ixd.rst: WARNING: document isn't
included in any toctree
Thanks,
Tony
> @@ -0,0 +1,35 @@
> +.. SPDX-License-Identifier: GPL-2.0
> +
> +===================
> +ixd devlink support
> +===================
> +
> +This document describes the devlink features implemented by the ``ixd``
> +device driver.
> +
> +Info versions
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH iwl-next 00/14] Introduce iXD driver
2025-04-08 12:47 [PATCH iwl-next 00/14] Introduce iXD driver Larysa Zaremba
` (13 preceding siblings ...)
2025-04-08 12:48 ` [PATCH iwl-next 14/14] ixd: add devlink support Larysa Zaremba
@ 2025-04-16 15:49 ` Alexander Lobakin
2025-04-16 16:23 ` Keller, Jacob E
14 siblings, 1 reply; 36+ messages in thread
From: Alexander Lobakin @ 2025-04-16 15:49 UTC (permalink / raw)
To: Larysa Zaremba
Cc: intel-wired-lan, Tony Nguyen, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Simon Horman, Jonathan Corbet,
Przemek Kitszel, Jiri Pirko, Mustafa Ismail, Tatyana Nikolova,
Andrew Lunn, Michael Ellerman, Maciej Fijalkowski, Lee Trager,
Madhavan Srinivasan, Sridhar Samudrala, Jacob Keller,
Michal Swiatkowski, Mateusz Polchlopek, Wenjun Wu, Ahmed Zaki,
netdev, linux-doc, linux-kernel, Karlsson, Magnus, Emil Tantilov,
Madhu Chittim, Josh Hay, Milena Olech, pavan.kumar.linga,
Singhai, Anjali
From: Larysa Zaremba <larysa.zaremba@intel.com>
Date: Tue, 8 Apr 2025 14:47:46 +0200
> This patch series adds the iXD driver, which supports the Intel(R)
> Control Plane PCI Function on Intel E2100 and later IPUs and FNICs.
> It facilitates a centralized control over multiple IDPF PFs/VFs/SFs
> exposed by the same card. The reason for the separation is to be able
> to offload the control plane to the host different from where the data
> plane is running.
BTW please move everything you're adding to libeth to libie instead.
This PCI/VC/CP functionality is unlikely to be used by other vendors.
Since libie_cp.ko or how you may want to call it won't link with base
libie.ko, it won't have any pre-idpf HW-specific symbols, so that idpf
could link with it.
libeth stuff is purely vendor-agnostic and I'd really like to keep it so.
Thanks,
Olek
^ permalink raw reply [flat|nested] 36+ messages in thread* RE: [PATCH iwl-next 00/14] Introduce iXD driver
2025-04-16 15:49 ` [PATCH iwl-next 00/14] Introduce iXD driver Alexander Lobakin
@ 2025-04-16 16:23 ` Keller, Jacob E
0 siblings, 0 replies; 36+ messages in thread
From: Keller, Jacob E @ 2025-04-16 16:23 UTC (permalink / raw)
To: Lobakin, Aleksander, Zaremba, Larysa
Cc: intel-wired-lan@lists.osuosl.org, Nguyen, Anthony L,
David S. Miller, Dumazet, Eric, Jakub Kicinski, Paolo Abeni,
Simon Horman, Jonathan Corbet, Kitszel, Przemyslaw, Jiri Pirko,
Mustafa Ismail, Nikolova, Tatyana E, Andrew Lunn,
Michael Ellerman, Fijalkowski, Maciej, Lee Trager,
Madhavan Srinivasan, Samudrala, Sridhar, Michal Swiatkowski,
Polchlopek, Mateusz, Wenjun Wu, Zaki, Ahmed,
netdev@vger.kernel.org, linux-doc@vger.kernel.org,
linux-kernel@vger.kernel.org, Karlsson, Magnus, Tantilov, Emil S,
Chittim, Madhu, Hay, Joshua A, Olech, Milena, Linga, Pavan Kumar,
Singhai, Anjali
> -----Original Message-----
> From: Lobakin, Aleksander <aleksander.lobakin@intel.com>
> Sent: Wednesday, April 16, 2025 8:49 AM
> To: Zaremba, Larysa <larysa.zaremba@intel.com>
> Cc: intel-wired-lan@lists.osuosl.org; Nguyen, Anthony L
> <anthony.l.nguyen@intel.com>; David S. Miller <davem@davemloft.net>;
> Dumazet, Eric <edumazet@google.com>; Jakub Kicinski <kuba@kernel.org>;
> Paolo Abeni <pabeni@redhat.com>; Simon Horman <horms@kernel.org>;
> Jonathan Corbet <corbet@lwn.net>; Kitszel, Przemyslaw
> <przemyslaw.kitszel@intel.com>; Jiri Pirko <jiri@resnulli.us>; Mustafa Ismail
> <mustafa.ismail@intel.com>; Nikolova, Tatyana E
> <tatyana.e.nikolova@intel.com>; Andrew Lunn <andrew+netdev@lunn.ch>;
> Michael Ellerman <mpe@ellerman.id.au>; Fijalkowski, Maciej
> <maciej.fijalkowski@intel.com>; Lee Trager <lee@trager.us>; Madhavan
> Srinivasan <maddy@linux.ibm.com>; Samudrala, Sridhar
> <sridhar.samudrala@intel.com>; Keller, Jacob E <jacob.e.keller@intel.com>;
> Michal Swiatkowski <michal.swiatkowski@linux.intel.com>; Polchlopek, Mateusz
> <mateusz.polchlopek@intel.com>; Wenjun Wu <wenjun1.wu@intel.com>; Zaki,
> Ahmed <ahmed.zaki@intel.com>; netdev@vger.kernel.org; linux-
> doc@vger.kernel.org; linux-kernel@vger.kernel.org; Karlsson, Magnus
> <magnus.karlsson@intel.com>; Tantilov, Emil S <emil.s.tantilov@intel.com>;
> Chittim, Madhu <madhu.chittim@intel.com>; Hay, Joshua A
> <joshua.a.hay@intel.com>; Olech, Milena <milena.olech@intel.com>; Linga,
> Pavan Kumar <pavan.kumar.linga@intel.com>; Singhai, Anjali
> <anjali.singhai@intel.com>
> Subject: Re: [PATCH iwl-next 00/14] Introduce iXD driver
>
> From: Larysa Zaremba <larysa.zaremba@intel.com>
> Date: Tue, 8 Apr 2025 14:47:46 +0200
>
> > This patch series adds the iXD driver, which supports the Intel(R)
> > Control Plane PCI Function on Intel E2100 and later IPUs and FNICs.
> > It facilitates a centralized control over multiple IDPF PFs/VFs/SFs
> > exposed by the same card. The reason for the separation is to be able
> > to offload the control plane to the host different from where the data
> > plane is running.
>
> BTW please move everything you're adding to libeth to libie instead.
> This PCI/VC/CP functionality is unlikely to be used by other vendors.
> Since libie_cp.ko or how you may want to call it won't link with base
> libie.ko, it won't have any pre-idpf HW-specific symbols, so that idpf
> could link with it.
> libeth stuff is purely vendor-agnostic and I'd really like to keep it so.
>
> Thanks,
> Olek
I agree with this too.
Thanks,
Jake
^ permalink raw reply [flat|nested] 36+ messages in thread