[PATCH 0/3] be2net: patch set

[PATCH 0/3] be2net: patch set

[Sathya Perla]

Hi David,

This patch set contains 2 driver fixes to a Lancer HW issue and a fix
to a double free bug. Pls apply to the "net" tree. Thanks!

Patch 1 now enables filters only after creating RXQs. This is done as 
HW issues were observed on Lancer adapters if filters
(flags, mac addrs etc) are enabled *before* creating RXQs. This patch
changes the driver design by enabling filters in be_open() --
instead of be_setup() -- after RXQs are created and buffers posted.

Patch 2 fixes an RX stall issue that was seen on Lancer adapters when
RXQs are destroyed while they are in an "out of buffer" state.
This patch fixes this issue by posting 64 buffers to each RXQ before
destroying them in the close path. This is done after ensuring that no
more new packets are selected for transfer to the RXQs by disabling
interface filters.

Patch 3 protects eqo->affinity_mask variable from being freed twice and
resulting in a crash.  It's now freed only when EQs haven't yet been
destroyed.

Kalesh AP (3):
  be2net: enable IFACE filters only after creating RXQs
  be2net: post buffers before destroying RXQs in Lancer
  be2net: protect eqo->affinity_mask from getting freed twice

 drivers/net/ethernet/emulex/benet/be_cmds.h |   5 +
 drivers/net/ethernet/emulex/benet/be_main.c | 187 ++++++++++++++++++----------
 2 files changed, 125 insertions(+), 67 deletions(-)

-- 
2.4.1

[PATCH 1/3] be2net: enable IFACE filters only after creating RXQs

[Sathya Perla]

From: Kalesh AP <kalesh.purayil@avagotech.com>

HW issues were observed on Lancer adapters if IFACE filters
(flags, mac addrs etc) are enabled *before* creating RXQs.  This patch
changes the driver design by enabling filters in be_open() --
instead of be_setup() -- after RXQs are created and buffers posted.
Two new wrapper functions, be_enable_if_filters() and
be_disable_if_filters() are introduced to enable/disable IFACE filters in
be_open()/be_close() respectively. In be_setup() the IFACE is now created
only with the RSS flag.

Signed-off-by: Kalesh AP <kalesh.purayil@avagotech.com>
Signed-off-by: Sathya Perla <sathya.perla@avagotech.com>
---
 drivers/net/ethernet/emulex/benet/be_cmds.h |   5 ++
 drivers/net/ethernet/emulex/benet/be_main.c | 127 ++++++++++++++++++----------
 2 files changed, 85 insertions(+), 47 deletions(-)

diff --git a/drivers/net/ethernet/emulex/benet/be_cmds.h b/drivers/net/ethernet/emulex/benet/be_cmds.h
index 2716e6f..00e3a6b 100644
--- a/drivers/net/ethernet/emulex/benet/be_cmds.h
+++ b/drivers/net/ethernet/emulex/benet/be_cmds.h
@@ -620,6 +620,11 @@ enum be_if_flags {
 					 BE_IF_FLAGS_VLAN_PROMISCUOUS |\
 					 BE_IF_FLAGS_MCAST_PROMISCUOUS)
 
+#define BE_IF_EN_FLAGS	(BE_IF_FLAGS_BROADCAST | BE_IF_FLAGS_PASS_L3L4_ERRORS |\
+			BE_IF_FLAGS_MULTICAST | BE_IF_FLAGS_UNTAGGED)
+
+#define BE_IF_ALL_FILT_FLAGS	(BE_IF_EN_FLAGS | BE_IF_FLAGS_ALL_PROMISCUOUS)
+
 /* An RX interface is an object with one or more MAC addresses and
  * filtering capabilities. */
 struct be_cmd_req_if_create {
diff --git a/drivers/net/ethernet/emulex/benet/be_main.c b/drivers/net/ethernet/emulex/benet/be_main.c
index 6f64242..7730f21 100644
--- a/drivers/net/ethernet/emulex/benet/be_main.c
+++ b/drivers/net/ethernet/emulex/benet/be_main.c
@@ -273,6 +273,10 @@ static int be_mac_addr_set(struct net_device *netdev, void *p)
 	if (ether_addr_equal(addr->sa_data, netdev->dev_addr))
 		return 0;
 
+	/* if device is not running, copy MAC to netdev->dev_addr */
+	if (!netif_running(netdev))
+		goto done;
+
 	/* The PMAC_ADD cmd may fail if the VF doesn't have FILTMGMT
 	 * privilege or if PF did not provision the new MAC address.
 	 * On BE3, this cmd will always fail if the VF doesn't have the
@@ -307,9 +311,9 @@ static int be_mac_addr_set(struct net_device *netdev, void *p)
 		status = -EPERM;
 		goto err;
 	}
-
-	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
-	dev_info(dev, "MAC address changed to %pM\n", mac);
+done:
+	ether_addr_copy(netdev->dev_addr, addr->sa_data);
+	dev_info(dev, "MAC address changed to %pM\n", addr->sa_data);
 	return 0;
 err:
 	dev_warn(dev, "MAC address change to %pM failed\n", addr->sa_data);
@@ -3361,6 +3365,33 @@ static void be_rx_qs_destroy(struct be_adapter *adapter)
 	}
 }
 
+static void be_disable_if_filters(struct be_adapter *adapter)
+{
+	be_cmd_pmac_del(adapter, adapter->if_handle,
+			adapter->pmac_id[0], 0);
+
+	be_clear_uc_list(adapter);
+
+	/* The IFACE flags are enabled in the open path and cleared
+	 * in the close path. When a VF gets detached from the host and
+	 * assigned to a VM the following happens:
+	 *	- VF's IFACE flags get cleared in the detach path
+	 *	- IFACE create is issued by the VF in the attach path
+	 * Due to a bug in the BE3/Skyhawk-R FW
+	 * (Lancer FW doesn't have the bug), the IFACE capability flags
+	 * specified along with the IFACE create cmd issued by a VF are not
+	 * honoured by FW.  As a consequence, if a *new* driver
+	 * (that enables/disables IFACE flags in open/close)
+	 * is loaded in the host and an *old* driver is * used by a VM/VF,
+	 * the IFACE gets created *without* the needed flags.
+	 * To avoid this, disable RX-filter flags only for Lancer.
+	 */
+	if (lancer_chip(adapter)) {
+		be_cmd_rx_filter(adapter, BE_IF_ALL_FILT_FLAGS, OFF);
+		adapter->if_flags &= ~BE_IF_ALL_FILT_FLAGS;
+	}
+}
+
 static int be_close(struct net_device *netdev)
 {
 	struct be_adapter *adapter = netdev_priv(netdev);
@@ -3373,6 +3404,8 @@ static int be_close(struct net_device *netdev)
 	if (!(adapter->flags & BE_FLAGS_SETUP_DONE))
 		return 0;
 
+	be_disable_if_filters(adapter);
+
 	be_roce_dev_close(adapter);
 
 	if (adapter->flags & BE_FLAGS_NAPI_ENABLED) {
@@ -3392,7 +3425,6 @@ static int be_close(struct net_device *netdev)
 	be_tx_compl_clean(adapter);
 
 	be_rx_qs_destroy(adapter);
-	be_clear_uc_list(adapter);
 
 	for_all_evt_queues(adapter, eqo, i) {
 		if (msix_enabled(adapter))
@@ -3477,6 +3509,31 @@ static int be_rx_qs_create(struct be_adapter *adapter)
 	return 0;
 }
 
+static int be_enable_if_filters(struct be_adapter *adapter)
+{
+	int status;
+
+	status = be_cmd_rx_filter(adapter, BE_IF_EN_FLAGS, ON);
+	if (status)
+		return status;
+
+	/* For BE3 VFs, the PF programs the initial MAC address */
+	if (!(BEx_chip(adapter) && be_virtfn(adapter))) {
+		status = be_cmd_pmac_add(adapter, adapter->netdev->dev_addr,
+					 adapter->if_handle,
+					 &adapter->pmac_id[0], 0);
+		if (status)
+			return status;
+	}
+
+	if (adapter->vlans_added)
+		be_vid_config(adapter);
+
+	be_set_rx_mode(adapter->netdev);
+
+	return 0;
+}
+
 static int be_open(struct net_device *netdev)
 {
 	struct be_adapter *adapter = netdev_priv(netdev);
@@ -3490,6 +3547,10 @@ static int be_open(struct net_device *netdev)
 	if (status)
 		goto err;
 
+	status = be_enable_if_filters(adapter);
+	if (status)
+		goto err;
+
 	status = be_irq_register(adapter);
 	if (status)
 		goto err;
@@ -3686,16 +3747,6 @@ static void be_cancel_err_detection(struct be_adapter *adapter)
 	}
 }
 
-static void be_mac_clear(struct be_adapter *adapter)
-{
-	if (adapter->pmac_id) {
-		be_cmd_pmac_del(adapter, adapter->if_handle,
-				adapter->pmac_id[0], 0);
-		kfree(adapter->pmac_id);
-		adapter->pmac_id = NULL;
-	}
-}
-
 #ifdef CONFIG_BE2NET_VXLAN
 static void be_disable_vxlan_offloads(struct be_adapter *adapter)
 {
@@ -3770,8 +3821,8 @@ static int be_clear(struct be_adapter *adapter)
 #ifdef CONFIG_BE2NET_VXLAN
 	be_disable_vxlan_offloads(adapter);
 #endif
-	/* delete the primary mac along with the uc-mac list */
-	be_mac_clear(adapter);
+	kfree(adapter->pmac_id);
+	adapter->pmac_id = NULL;
 
 	be_cmd_if_destroy(adapter, adapter->if_handle,  0);
 
@@ -3782,25 +3833,11 @@ static int be_clear(struct be_adapter *adapter)
 	return 0;
 }
 
-static int be_if_create(struct be_adapter *adapter, u32 *if_handle,
-			u32 cap_flags, u32 vf)
-{
-	u32 en_flags;
-
-	en_flags = BE_IF_FLAGS_UNTAGGED | BE_IF_FLAGS_BROADCAST |
-		   BE_IF_FLAGS_MULTICAST | BE_IF_FLAGS_PASS_L3L4_ERRORS |
-		   BE_IF_FLAGS_RSS | BE_IF_FLAGS_DEFQ_RSS;
-
-	en_flags &= cap_flags;
-
-	return be_cmd_if_create(adapter, cap_flags, en_flags, if_handle, vf);
-}
-
 static int be_vfs_if_create(struct be_adapter *adapter)
 {
 	struct be_resources res = {0};
+	u32 cap_flags, en_flags, vf;
 	struct be_vf_cfg *vf_cfg;
-	u32 cap_flags, vf;
 	int status;
 
 	/* If a FW profile exists, then cap_flags are updated */
@@ -3821,8 +3858,12 @@ static int be_vfs_if_create(struct be_adapter *adapter)
 			}
 		}
 
-		status = be_if_create(adapter, &vf_cfg->if_handle,
-				      cap_flags, vf + 1);
+		en_flags = cap_flags & (BE_IF_FLAGS_UNTAGGED |
+					BE_IF_FLAGS_BROADCAST |
+					BE_IF_FLAGS_MULTICAST |
+					BE_IF_FLAGS_PASS_L3L4_ERRORS);
+		status = be_cmd_if_create(adapter, cap_flags, en_flags,
+					  &vf_cfg->if_handle, vf + 1);
 		if (status)
 			return status;
 	}
@@ -4194,15 +4235,8 @@ static int be_mac_setup(struct be_adapter *adapter)
 
 		memcpy(adapter->netdev->dev_addr, mac, ETH_ALEN);
 		memcpy(adapter->netdev->perm_addr, mac, ETH_ALEN);
-	} else {
-		/* Maybe the HW was reset; dev_addr must be re-programmed */
-		memcpy(mac, adapter->netdev->dev_addr, ETH_ALEN);
 	}
 
-	/* For BE3-R VFs, the PF programs the initial MAC address */
-	if (!(BEx_chip(adapter) && be_virtfn(adapter)))
-		be_cmd_pmac_add(adapter, mac, adapter->if_handle,
-				&adapter->pmac_id[0], 0);
 	return 0;
 }
 
@@ -4342,6 +4376,7 @@ static int be_func_init(struct be_adapter *adapter)
 static int be_setup(struct be_adapter *adapter)
 {
 	struct device *dev = &adapter->pdev->dev;
+	u32 en_flags;
 	int status;
 
 	status = be_func_init(adapter);
@@ -4364,8 +4399,11 @@ static int be_setup(struct be_adapter *adapter)
 	if (status)
 		goto err;
 
-	status = be_if_create(adapter, &adapter->if_handle,
-			      be_if_cap_flags(adapter), 0);
+	/* will enable all the needed filter flags in be_open() */
+	en_flags = BE_IF_FLAGS_RSS | BE_IF_FLAGS_DEFQ_RSS;
+	en_flags = en_flags & be_if_cap_flags(adapter);
+	status = be_cmd_if_create(adapter, be_if_cap_flags(adapter), en_flags,
+				  &adapter->if_handle, 0);
 	if (status)
 		goto err;
 
@@ -4391,11 +4429,6 @@ static int be_setup(struct be_adapter *adapter)
 		dev_err(dev, "Please upgrade firmware to version >= 4.0\n");
 	}
 
-	if (adapter->vlans_added)
-		be_vid_config(adapter);
-
-	be_set_rx_mode(adapter->netdev);
-
 	status = be_cmd_set_flow_control(adapter, adapter->tx_fc,
 					 adapter->rx_fc);
 	if (status)
-- 
2.4.1

[PATCH 2/3] be2net: post buffers before destroying RXQs in Lancer

[Sathya Perla]

From: Kalesh AP <kalesh.purayil@avagotech.com>

An RX stall issue was seen on Lancer adapters, when RXQs are destroyed
while they are in an "out of buffer" state. This patch fixes this issue
by posting 64 buffers to each RXQ before destroying them in the close path.
This is done after ensuring that no more new packets are selected for
transfer to the RXQs by disabling interface filters.

Signed-off-by: Kalesh AP <kalesh.purayil@avagotech.com>
Signed-off-by: Sathya Perla <sathya.perla@avagotech.com>
---
 drivers/net/ethernet/emulex/benet/be_main.c | 42 ++++++++++++++++++++---------
 1 file changed, 30 insertions(+), 12 deletions(-)

diff --git a/drivers/net/ethernet/emulex/benet/be_main.c b/drivers/net/ethernet/emulex/benet/be_main.c
index 7730f21..14ae67a 100644
--- a/drivers/net/ethernet/emulex/benet/be_main.c
+++ b/drivers/net/ethernet/emulex/benet/be_main.c
@@ -2451,10 +2451,24 @@ static void be_eq_clean(struct be_eq_obj *eqo)
 	be_eq_notify(eqo->adapter, eqo->q.id, false, true, num, 0);
 }
 
+/* Free posted rx buffers that were not used */
+static void be_rxq_clean(struct be_rx_obj *rxo)
+{
+	struct be_queue_info *rxq = &rxo->q;
+	struct be_rx_page_info *page_info;
+
+	while (atomic_read(&rxq->used) > 0) {
+		page_info = get_rx_page_info(rxo);
+		put_page(page_info->page);
+		memset(page_info, 0, sizeof(*page_info));
+	}
+	BUG_ON(atomic_read(&rxq->used));
+	rxq->tail = 0;
+	rxq->head = 0;
+}
+
 static void be_rx_cq_clean(struct be_rx_obj *rxo)
 {
-	struct be_rx_page_info *page_info;
-	struct be_queue_info *rxq = &rxo->q;
 	struct be_queue_info *rx_cq = &rxo->cq;
 	struct be_rx_compl_info *rxcp;
 	struct be_adapter *adapter = rxo->adapter;
@@ -2491,16 +2505,6 @@ static void be_rx_cq_clean(struct be_rx_obj *rxo)
 
 	/* After cleanup, leave the CQ in unarmed state */
 	be_cq_notify(adapter, rx_cq->id, false, 0);
-
-	/* Then free posted rx buffers that were not used */
-	while (atomic_read(&rxq->used) > 0) {
-		page_info = get_rx_page_info(rxo);
-		put_page(page_info->page);
-		memset(page_info, 0, sizeof(*page_info));
-	}
-	BUG_ON(atomic_read(&rxq->used));
-	rxq->tail = 0;
-	rxq->head = 0;
 }
 
 static void be_tx_compl_clean(struct be_adapter *adapter)
@@ -3358,8 +3362,22 @@ static void be_rx_qs_destroy(struct be_adapter *adapter)
 	for_all_rx_queues(adapter, rxo, i) {
 		q = &rxo->q;
 		if (q->created) {
+			/* If RXQs are destroyed while in an "out of buffer"
+			 * state, there is a possibility of an HW stall on
+			 * Lancer. So, post 64 buffers to each queue to relieve
+			 * the "out of buffer" condition.
+			 * Make sure there's space in the RXQ before posting.
+			 */
+			if (lancer_chip(adapter)) {
+				be_rx_cq_clean(rxo);
+				if (atomic_read(&q->used) == 0)
+					be_post_rx_frags(rxo, GFP_KERNEL,
+							 MAX_RX_POST);
+			}
+
 			be_cmd_rxq_destroy(adapter, q);
 			be_rx_cq_clean(rxo);
+			be_rxq_clean(rxo);
 		}
 		be_queue_free(adapter, q);
 	}
-- 
2.4.1

[PATCH 3/3] be2net: protect eqo->affinity_mask from getting freed twice

[Sathya Perla]

From: Kalesh AP <kalesh.purayil@avagotech.com>

There are paths in the driver such as an unrecoverable error (UE) detection
followed by a driver unload wherein be_clear() is invoked twice.
Individual data structures are reset so that they are not cleaned/freed
twice. This patch does the same for eqo->affinity_mask. It is freed only
if EQs haven't yet been destroyed. This fixes a possible crash when
affinity_mask is freed twice.

Signed-off-by: Kalesh AP <kalesh.purayil@avagotech.com>
Signed-off-by: Sathya Perla <sathya.perla@avagotech.com>
---
 drivers/net/ethernet/emulex/benet/be_main.c | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)

diff --git a/drivers/net/ethernet/emulex/benet/be_main.c b/drivers/net/ethernet/emulex/benet/be_main.c
index 14ae67a..c28e3bf 100644
--- a/drivers/net/ethernet/emulex/benet/be_main.c
+++ b/drivers/net/ethernet/emulex/benet/be_main.c
@@ -2584,8 +2584,8 @@ static void be_evt_queues_destroy(struct be_adapter *adapter)
 			be_cmd_q_destroy(adapter, &eqo->q, QTYPE_EQ);
 			napi_hash_del(&eqo->napi);
 			netif_napi_del(&eqo->napi);
+			free_cpumask_var(eqo->affinity_mask);
 		}
-		free_cpumask_var(eqo->affinity_mask);
 		be_queue_free(adapter, &eqo->q);
 	}
 }
@@ -2602,13 +2602,7 @@ static int be_evt_queues_create(struct be_adapter *adapter)
 
 	for_all_evt_queues(adapter, eqo, i) {
 		int numa_node = dev_to_node(&adapter->pdev->dev);
-		if (!zalloc_cpumask_var(&eqo->affinity_mask, GFP_KERNEL))
-			return -ENOMEM;
-		cpumask_set_cpu(cpumask_local_spread(i, numa_node),
-				eqo->affinity_mask);
-		netif_napi_add(adapter->netdev, &eqo->napi, be_poll,
-			       BE_NAPI_WEIGHT);
-		napi_hash_add(&eqo->napi);
+
 		aic = &adapter->aic_obj[i];
 		eqo->adapter = adapter;
 		eqo->idx = i;
@@ -2624,6 +2618,14 @@ static int be_evt_queues_create(struct be_adapter *adapter)
 		rc = be_cmd_eq_create(adapter, eqo);
 		if (rc)
 			return rc;
+
+		if (!zalloc_cpumask_var(&eqo->affinity_mask, GFP_KERNEL))
+			return -ENOMEM;
+		cpumask_set_cpu(cpumask_local_spread(i, numa_node),
+				eqo->affinity_mask);
+		netif_napi_add(adapter->netdev, &eqo->napi, be_poll,
+			       BE_NAPI_WEIGHT);
+		napi_hash_add(&eqo->napi);
 	}
 	return 0;
 }
-- 
2.4.1

Re: [PATCH 0/3] be2net: patch set

[David Miller]

From: Sathya Perla <sathya.perla@avagotech.com>
Date: Wed,  5 Aug 2015 03:27:47 -0400

> This patch set contains 2 driver fixes to a Lancer HW issue and a fix
> to a double free bug. Pls apply to the "net" tree. Thanks!
> 
> Patch 1 now enables filters only after creating RXQs. This is done as 
> HW issues were observed on Lancer adapters if filters
> (flags, mac addrs etc) are enabled *before* creating RXQs. This patch
> changes the driver design by enabling filters in be_open() --
> instead of be_setup() -- after RXQs are created and buffers posted.
> 
> Patch 2 fixes an RX stall issue that was seen on Lancer adapters when
> RXQs are destroyed while they are in an "out of buffer" state.
> This patch fixes this issue by posting 64 buffers to each RXQ before
> destroying them in the close path. This is done after ensuring that no
> more new packets are selected for transfer to the RXQs by disabling
> interface filters.
> 
> Patch 3 protects eqo->affinity_mask variable from being freed twice and
> resulting in a crash.  It's now freed only when EQs haven't yet been
> destroyed.

Series applied, thanks.

[PATCH 0/3] be2net: patch set

[Sathya Perla]

Hi David, this patch set includes 3 bug fixes to the be2net driver.

Patch 1 fixes a vlan isolation issue with VFs. When a VF is placed in
promiscous mode, it could receive packets belonging to any vlan, as
the PF driver grants vlan promisc capability to VFs. The PF
driver now disables the vlan promisc capability for VFs to fix this
problem.

Patch 2 fixes the call to MODIFY_EQ_DELAY FW cmd to not include more
than 8 EQs per cmd. The FW is not capable of handling more than 8 EQs
per cmd.

Patch 3 fixes an EEH error detection issue. On Power platforms,
when an EEH error occurs, the slot disconnect state is more reliably
detected via an MMIO read compared to a config read. So, the error
register reads that occur every second are now done via MMIO.

Pls apply this patch set to the "net" tree. Thanks!

Suresh Reddy (2):
  be2net: restrict MODIFY_EQ_DELAY cmd to a max of 8 EQs
  be2net: use PCI MMIO read instead of config read for errors

Vasundhara Volam (1):
  be2net: Prevent VFs from enabling VLAN promiscuous mode

Re: [PATCH 0/3] be2net: patch set

[David Miller]

From: Sathya Perla <sathya.perla@emulex.com>
Date: Fri, 20 Mar 2015 06:28:22 -0400

> Hi David, this patch set includes 3 bug fixes to the be2net driver.

Series applied, thank you.

[PATCH 0/3] be2net: patch set

[Sathya Perla]

Pls apply the following bug fixes to the 'net' tree. Thanks.

Suresh Reddy (2):
  be2net: increase the timeout value for loopback-test FW cmd
  be2net: fix max_evt_qs calculation for BE3 in SR-IOV config

Vasundhara Volam (1):
  be2net: disable RSS when number of RXQs is reduced to 1 via
    set-channels

 drivers/net/ethernet/emulex/benet/be.h      |    3 +-
 drivers/net/ethernet/emulex/benet/be_cmds.c |   33 ++++++++++++++++++--------
 drivers/net/ethernet/emulex/benet/be_main.c |   29 ++++++++++++++---------
 3 files changed, 43 insertions(+), 22 deletions(-)

Re: [PATCH 0/3] be2net: patch set

[David Miller]

From: Sathya Perla <sathya.perla@emulex.com>
Date: Mon, 6 Jan 2014 13:02:22 +0530

> Pls apply the following bug fixes to the 'net' tree. Thanks.

Series applied, thanks.