* Re: [PATCH net-next] packet: fix broken build.
From: David Miller @ 2012-08-23 16:29 UTC (permalink / raw)
To: rosenr; +Cc: netdev, xemul, edumazet
In-Reply-To: <309002C0DA137042828828FC53D7A934012B21B8CCCE@IL-MB01.marvell.com>
From: Rami Rosen <rosenr@marvell.com>
Date: Thu, 23 Aug 2012 15:55:41 +0300
> This patch fixes a broken build due to a missing header:
> ...
> CC net/ipv4/proc.o
> In file included from include/net/net_namespace.h:15,
> from net/ipv4/proc.c:35:
> include/net/netns/packet.h:11: error: field 'sklist_lock' has incomplete type
> ...
>
> The lock of netns_packet has been replaced by a recent patch to be a mutex instead of a spinlock,
> but we need to replace the header file to be linux/mutex.h instead of linux/spinlock.h as well.
>
> See commit 0fa7fa98dbcc2789409ed24e885485e645803d7f:
> packet: Protect packet sk list with mutex (v2) patch,
>
> Signed-off-by: Rami Rosen <rosenr@marvell.com>
Applied, thanks.
^ permalink raw reply
* RE: [net-next 12/13] igb: Correct PTP support query from ethtool.
From: Vick, Matthew @ 2012-08-23 16:38 UTC (permalink / raw)
To: Richard Cochran, Kirsher, Jeffrey T
Cc: davem@davemloft.net, netdev@vger.kernel.org, gospo@redhat.com,
sassmann@redhat.com
In-Reply-To: <20120823112426.GD2238@netboy.at.omicron.at>
> -----Original Message-----
> From: netdev-owner@vger.kernel.org [mailto:netdev-
> owner@vger.kernel.org] On Behalf Of Richard Cochran
> Sent: Thursday, August 23, 2012 4:24 AM
> To: Kirsher, Jeffrey T
> Cc: davem@davemloft.net; Vick, Matthew; netdev@vger.kernel.org;
> gospo@redhat.com; sassmann@redhat.com
> Subject: Re: [net-next 12/13] igb: Correct PTP support query from
> ethtool.
>
> On Thu, Aug 23, 2012 at 02:56:52AM -0700, Jeff Kirsher wrote:
> > From: Matthew Vick <matthew.vick@intel.com>
> >
> > Update ethtool_get_ts_info to not report any supported functionality
> > on
> > 82575 and add support for V2 Sync and V2 Delay packets. In the case
> > where CONFIG_IGB_PTP is not defined, we should be reporting default
> > values.
>
> > diff --git a/drivers/net/ethernet/intel/igb/igb_ethtool.c
> > b/drivers/net/ethernet/intel/igb/igb_ethtool.c
>
> ...
>
> > +#endif /* CONFIG_IGB_PTP */
> > + default:
> > + return ethtool_op_get_ts_info(dev, info);
>
> This is wrong. Using this function advertises that you support SW TX
> time stamps, which you do not.
>
> Thanks,
> Richard
Thanks--good catch! I thought we supported SW timestamps already. Since we don't, I'll spin a v2 of this patch.
Cheers,
Matthew
^ permalink raw reply
* Re: [PATCH v2] net-tcp: TCP/IP stack bypass for loopback connections
From: Stephen Clark @ 2012-08-23 16:41 UTC (permalink / raw)
To: David Miller; +Cc: billfink, brutus, edumazet, netdev
In-Reply-To: <20120814.003732.714549073635415560.davem@davemloft.net>
On 08/14/2012 03:37 AM, David Miller wrote:
> From: Bill Fink<billfink@mindspring.com>
> Date: Tue, 14 Aug 2012 02:31:55 -0400
>
>
>> I also believe it should be disabled by default, as that is
>> the current behavior, and those who would gain an advantage
>> from using it can easily enable it.
>>
> It benefits basically everyone, it makes things orders of
> magnitude faster.
>
Hmm... Looking at the stats provided I didn't see any orders of
magnitude faster.
>
>> While this could be a very useful feature in some environments,
>> it seems to me it would be safest to have it disabled by default.
>>
> I violently disagree, and there is no way I'm having this
> thing off by default.
> --
> To unsubscribe from this list: send the line "unsubscribe netdev" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
>
--
"They that give up essential liberty to obtain temporary safety,
deserve neither liberty nor safety." (Ben Franklin)
"The course of history shows that as a government grows, liberty
decreases." (Thomas Jefferson)
^ permalink raw reply
* Re: Best way to set kernel thread affinity for handling a socket?
From: Roland Dreier @ 2012-08-23 17:04 UTC (permalink / raw)
To: Ben Hutchings; +Cc: netdev
In-Reply-To: <1345652493.2709.36.camel@bwh-desktop.uk.solarflarecom.com>
On Wed, Aug 22, 2012 at 9:21 AM, Ben Hutchings
<bhutchings@solarflare.com> wrote:
> With RFS we try to do the reverse: move the packets to match the socket
> user. But it's not (yet) turned on by default. See
> Documentation/networking/scaling.txt
Fair enough. However I think at least in this case it sounds like extra
overhead: it should be easy for us to do everything on the CPU where
the packets are being received.
>> I'm thinking about this in the context of the kernel's iSCSI target
>> code (drivers/target/iscsi), which creates threads to handle each
>> iSCSI connection and sets their CPU affinity pretty much randomly
>> (well, based on some "thread id", cf iscsit_thread_get_cpumask()).
>
> Why set the affinity at all?
It's quite possible that, like a lot of other drivers/target code, this
doesn't actually make any sense and the right answer is to rip it
out completely.
Is the affinity due to waking up from a network receive event
enough to keep the threads local to the right CPU?
- R.
^ permalink raw reply
* [PATCH] net-tcp: fix cwnd reduction for non-sack recovery
From: Yuchung Cheng @ 2012-08-23 17:05 UTC (permalink / raw)
To: davem, ncardwell, nanditad; +Cc: mattmathis, netdev, Yuchung Cheng
The cwnd reduction in fast recovery is based on the number of packets
newly delivered per ACK. For non-sack connections every DUPACK
signifies a packet has been delivered, but the sender mistakenly
skips counting them for cwnd reduction.
The fix is to compute newly_acked_sacked after DUPACKs are accounted
in sacked_out for non-sack connections.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
---
net/ipv4/tcp_input.c | 15 +++++++--------
1 files changed, 7 insertions(+), 8 deletions(-)
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index bcfccc5..ce4ffe9 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -2930,13 +2930,14 @@ static void tcp_enter_recovery(struct sock *sk, bool ece_ack)
* tcp_xmit_retransmit_queue().
*/
static void tcp_fastretrans_alert(struct sock *sk, int pkts_acked,
- int newly_acked_sacked, bool is_dupack,
+ int prior_sacked, bool is_dupack,
int flag)
{
struct inet_connection_sock *icsk = inet_csk(sk);
struct tcp_sock *tp = tcp_sk(sk);
int do_lost = is_dupack || ((flag & FLAG_DATA_SACKED) &&
(tcp_fackets_out(tp) > tp->reordering));
+ int newly_acked_sacked = 0;
int fast_rexmit = 0;
if (WARN_ON(!tp->packets_out && tp->sacked_out))
@@ -2996,6 +2997,7 @@ static void tcp_fastretrans_alert(struct sock *sk, int pkts_acked,
tcp_add_reno_sack(sk);
} else
do_lost = tcp_try_undo_partial(sk, pkts_acked);
+ newly_acked_sacked = pkts_acked + tp->sacked_out - prior_sacked;
break;
case TCP_CA_Loss:
if (flag & FLAG_DATA_ACKED)
@@ -3017,6 +3019,7 @@ static void tcp_fastretrans_alert(struct sock *sk, int pkts_acked,
if (is_dupack)
tcp_add_reno_sack(sk);
}
+ newly_acked_sacked = pkts_acked + tp->sacked_out - prior_sacked;
if (icsk->icsk_ca_state <= TCP_CA_Disorder)
tcp_try_undo_dsack(sk);
@@ -3594,7 +3597,6 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
int prior_packets;
int prior_sacked = tp->sacked_out;
int pkts_acked = 0;
- int newly_acked_sacked = 0;
bool frto_cwnd = false;
/* If the ack is older than previous acks
@@ -3670,8 +3672,6 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
flag |= tcp_clean_rtx_queue(sk, prior_fackets, prior_snd_una);
pkts_acked = prior_packets - tp->packets_out;
- newly_acked_sacked = (prior_packets - prior_sacked) -
- (tp->packets_out - tp->sacked_out);
if (tp->frto_counter)
frto_cwnd = tcp_process_frto(sk, flag);
@@ -3685,7 +3685,7 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
tcp_may_raise_cwnd(sk, flag))
tcp_cong_avoid(sk, ack, prior_in_flight);
is_dupack = !(flag & (FLAG_SND_UNA_ADVANCED | FLAG_NOT_DUP));
- tcp_fastretrans_alert(sk, pkts_acked, newly_acked_sacked,
+ tcp_fastretrans_alert(sk, pkts_acked, prior_sacked,
is_dupack, flag);
} else {
if ((flag & FLAG_DATA_ACKED) && !frto_cwnd)
@@ -3702,7 +3702,7 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
no_queue:
/* If data was DSACKed, see if we can undo a cwnd reduction. */
if (flag & FLAG_DSACKING_ACK)
- tcp_fastretrans_alert(sk, pkts_acked, newly_acked_sacked,
+ tcp_fastretrans_alert(sk, pkts_acked, prior_sacked,
is_dupack, flag);
/* If this ack opens up a zero window, clear backoff. It was
* being used to time the probes, and is probably far higher than
@@ -3722,8 +3722,7 @@ old_ack:
*/
if (TCP_SKB_CB(skb)->sacked) {
flag |= tcp_sacktag_write_queue(sk, skb, prior_snd_una);
- newly_acked_sacked = tp->sacked_out - prior_sacked;
- tcp_fastretrans_alert(sk, pkts_acked, newly_acked_sacked,
+ tcp_fastretrans_alert(sk, pkts_acked, prior_sacked,
is_dupack, flag);
}
--
1.7.7.3
^ permalink raw reply related
* Re: [PATCH] net-tcp: fix cwnd reduction for non-sack recovery
From: Nandita Dukkipati @ 2012-08-23 17:21 UTC (permalink / raw)
To: Yuchung Cheng; +Cc: davem, ncardwell, mattmathis, netdev
In-Reply-To: <1345741517-30504-1-git-send-email-ycheng@google.com>
On Thu, Aug 23, 2012 at 10:05 AM, Yuchung Cheng <ycheng@google.com> wrote:
> The cwnd reduction in fast recovery is based on the number of packets
> newly delivered per ACK. For non-sack connections every DUPACK
> signifies a packet has been delivered, but the sender mistakenly
> skips counting them for cwnd reduction.
>
> The fix is to compute newly_acked_sacked after DUPACKs are accounted
> in sacked_out for non-sack connections.
>
> Signed-off-by: Yuchung Cheng <ycheng@google.com>
> ---
> net/ipv4/tcp_input.c | 15 +++++++--------
> 1 files changed, 7 insertions(+), 8 deletions(-)
>
> diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
> index bcfccc5..ce4ffe9 100644
> --- a/net/ipv4/tcp_input.c
> +++ b/net/ipv4/tcp_input.c
> @@ -2930,13 +2930,14 @@ static void tcp_enter_recovery(struct sock *sk, bool ece_ack)
> * tcp_xmit_retransmit_queue().
> */
> static void tcp_fastretrans_alert(struct sock *sk, int pkts_acked,
> - int newly_acked_sacked, bool is_dupack,
> + int prior_sacked, bool is_dupack,
> int flag)
> {
> struct inet_connection_sock *icsk = inet_csk(sk);
> struct tcp_sock *tp = tcp_sk(sk);
> int do_lost = is_dupack || ((flag & FLAG_DATA_SACKED) &&
> (tcp_fackets_out(tp) > tp->reordering));
> + int newly_acked_sacked = 0;
> int fast_rexmit = 0;
>
> if (WARN_ON(!tp->packets_out && tp->sacked_out))
> @@ -2996,6 +2997,7 @@ static void tcp_fastretrans_alert(struct sock *sk, int pkts_acked,
> tcp_add_reno_sack(sk);
> } else
> do_lost = tcp_try_undo_partial(sk, pkts_acked);
> + newly_acked_sacked = pkts_acked + tp->sacked_out - prior_sacked;
> break;
> case TCP_CA_Loss:
> if (flag & FLAG_DATA_ACKED)
> @@ -3017,6 +3019,7 @@ static void tcp_fastretrans_alert(struct sock *sk, int pkts_acked,
> if (is_dupack)
> tcp_add_reno_sack(sk);
> }
> + newly_acked_sacked = pkts_acked + tp->sacked_out - prior_sacked;
>
> if (icsk->icsk_ca_state <= TCP_CA_Disorder)
> tcp_try_undo_dsack(sk);
> @@ -3594,7 +3597,6 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
> int prior_packets;
> int prior_sacked = tp->sacked_out;
> int pkts_acked = 0;
> - int newly_acked_sacked = 0;
> bool frto_cwnd = false;
>
> /* If the ack is older than previous acks
> @@ -3670,8 +3672,6 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
> flag |= tcp_clean_rtx_queue(sk, prior_fackets, prior_snd_una);
>
> pkts_acked = prior_packets - tp->packets_out;
> - newly_acked_sacked = (prior_packets - prior_sacked) -
> - (tp->packets_out - tp->sacked_out);
>
> if (tp->frto_counter)
> frto_cwnd = tcp_process_frto(sk, flag);
> @@ -3685,7 +3685,7 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
> tcp_may_raise_cwnd(sk, flag))
> tcp_cong_avoid(sk, ack, prior_in_flight);
> is_dupack = !(flag & (FLAG_SND_UNA_ADVANCED | FLAG_NOT_DUP));
> - tcp_fastretrans_alert(sk, pkts_acked, newly_acked_sacked,
> + tcp_fastretrans_alert(sk, pkts_acked, prior_sacked,
> is_dupack, flag);
> } else {
> if ((flag & FLAG_DATA_ACKED) && !frto_cwnd)
> @@ -3702,7 +3702,7 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
> no_queue:
> /* If data was DSACKed, see if we can undo a cwnd reduction. */
> if (flag & FLAG_DSACKING_ACK)
> - tcp_fastretrans_alert(sk, pkts_acked, newly_acked_sacked,
> + tcp_fastretrans_alert(sk, pkts_acked, prior_sacked,
> is_dupack, flag);
> /* If this ack opens up a zero window, clear backoff. It was
> * being used to time the probes, and is probably far higher than
> @@ -3722,8 +3722,7 @@ old_ack:
> */
> if (TCP_SKB_CB(skb)->sacked) {
> flag |= tcp_sacktag_write_queue(sk, skb, prior_snd_una);
> - newly_acked_sacked = tp->sacked_out - prior_sacked;
> - tcp_fastretrans_alert(sk, pkts_acked, newly_acked_sacked,
> + tcp_fastretrans_alert(sk, pkts_acked, prior_sacked,
> is_dupack, flag);
> }
>
> --
> 1.7.7.3
>
Acked-by: Nandita Dukkipati <nanditad@google.com>
^ permalink raw reply
* [PATCH 3/8] csiostor: Chelsio FCoE offload driver submission (sources part 3).
From: Naresh Kumar Inna @ 2012-08-23 22:27 UTC (permalink / raw)
To: JBottomley, linux-scsi, dm; +Cc: netdev, naresh, chethan
In-Reply-To: <1345760873-12101-1-git-send-email-naresh@chelsio.com>
This patch contains code for the FC transport template callbacks and the
Mailbox module functionality. The FC transport callbacks include Virtual
Node ports creation and deletion, FC session registration, unregistration and
teardown. The Mailbox module provides services to issue/track/cancel
mailbox commands and wrappers for them.
Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
---
drivers/scsi/csiostor/csio_attr.c | 808 +++++++++++++++++
drivers/scsi/csiostor/csio_mb.c | 1768 +++++++++++++++++++++++++++++++++++++
2 files changed, 2576 insertions(+), 0 deletions(-)
create mode 100644 drivers/scsi/csiostor/csio_attr.c
create mode 100644 drivers/scsi/csiostor/csio_mb.c
diff --git a/drivers/scsi/csiostor/csio_attr.c b/drivers/scsi/csiostor/csio_attr.c
new file mode 100644
index 0000000..fd30798
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_attr.c
@@ -0,0 +1,808 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/kernel.h>
+#include <linux/string.h>
+#include <linux/delay.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/mm.h>
+#include <linux/jiffies.h>
+
+#include "csio_init.h"
+
+static void
+csio_vport_set_state(struct csio_lnode *ln);
+
+/*
+ * csio_reg_rnode - Register a remote port with FC transport.
+ * @rn: Rnode representing remote port.
+ *
+ * Call fc_remote_port_add() to register this remote port with FC transport.
+ * If remote port is Initiator OR Target OR both, change the role appropriately.
+ *
+ */
+void
+csio_reg_rnode(struct csio_rnode *rn)
+{
+ struct csio_lnode *ln = csio_rnode_to_lnode(rn);
+ struct Scsi_Host *shost = csio_ln_to_shost(ln);
+ struct fc_rport_identifiers ids;
+ struct fc_rport *rport;
+ struct csio_service_parms *sp;
+
+ ids.node_name = wwn_to_u64(csio_rn_wwnn(rn));
+ ids.port_name = wwn_to_u64(csio_rn_wwpn(rn));
+ ids.port_id = rn->nport_id;
+ ids.roles = FC_RPORT_ROLE_UNKNOWN;
+
+ if (rn->role & CSIO_RNFR_INITIATOR || rn->role & CSIO_RNFR_TARGET) {
+ rport = rn->rport;
+ CSIO_ASSERT(rport != NULL);
+ goto update_role;
+ }
+
+ rn->rport = fc_remote_port_add(shost, 0, &ids);
+ if (!rn->rport) {
+ csio_ln_err(ln, "Failed to register rport = 0x%x.\n",
+ rn->nport_id);
+ return;
+ }
+
+ ln->num_reg_rnodes++;
+ rport = rn->rport;
+ spin_lock_irq(shost->host_lock);
+ *((struct csio_rnode **)rport->dd_data) = rn;
+ spin_unlock_irq(shost->host_lock);
+
+ sp = &rn->rn_sparm;
+ rport->maxframe_size = sp->csp.rcv_sz;
+ if (G_SP_CLASS_SUPPORT(sp->clsp[2].serv_option))
+ rport->supported_classes = FC_COS_CLASS3;
+ else
+ rport->supported_classes = FC_COS_UNSPECIFIED;
+update_role:
+ if (rn->role & CSIO_RNFR_INITIATOR)
+ ids.roles |= FC_RPORT_ROLE_FCP_INITIATOR;
+ if (rn->role & CSIO_RNFR_TARGET) {
+ ids.roles |= FC_RPORT_ROLE_FCP_TARGET;
+ ln->n_scsi_tgts++;
+ }
+
+ if (ids.roles != FC_RPORT_ROLE_UNKNOWN)
+ fc_remote_port_rolechg(rport, ids.roles);
+
+ rn->scsi_id = rport->scsi_target_id;
+
+ csio_ln_dbg(ln, "Remote port x%x role 0x%x registered\n",
+ rn->nport_id, ids.roles);
+}
+
+/*
+ * csio_unreg_rnode - Unregister a remote port with FC transport.
+ * @rn: Rnode representing remote port.
+ *
+ * Call fc_remote_port_delete() to unregister this remote port with FC
+ * transport.
+ *
+ */
+void
+csio_unreg_rnode(struct csio_rnode *rn)
+{
+ struct csio_lnode *ln = csio_rnode_to_lnode(rn);
+ struct fc_rport *rport = rn->rport;
+
+ rn->role &= ~(CSIO_RNFR_INITIATOR | CSIO_RNFR_TARGET);
+ fc_remote_port_delete(rport);
+ ln->num_reg_rnodes--;
+
+ if (ln->n_scsi_tgts)
+ ln->n_scsi_tgts--;
+
+ if (ln->last_scan_ntgts)
+ ln->last_scan_ntgts--;
+
+ csio_ln_dbg(ln, "Remote port x%x un-registered\n", rn->nport_id);
+}
+
+/*
+ * csio_lnode_async_event - Async events from local port.
+ * @ln: lnode representing local port.
+ *
+ * Async events from local node that FC transport/SCSI ML
+ * should be made aware of (Eg: RSCN).
+ */
+void
+csio_lnode_async_event(struct csio_lnode *ln, enum csio_ln_fc_evt fc_evt)
+{
+ switch (fc_evt) {
+ case CSIO_LN_FC_RSCN:
+ /* Get payload of rscn from ln */
+ /* For each RSCN entry */
+ /*
+ * fc_host_post_event(shost,
+ * fc_get_event_number(),
+ * FCH_EVT_RSCN,
+ * rscn_entry);
+ */
+ break;
+ case CSIO_LN_FC_LINKUP:
+ /* send fc_host_post_event */
+ /* set vport state */
+ if (csio_is_npiv_ln(ln))
+ csio_vport_set_state(ln);
+
+ break;
+ case CSIO_LN_FC_LINKDOWN:
+ /* send fc_host_post_event */
+ /* set vport state */
+ if (csio_is_npiv_ln(ln))
+ csio_vport_set_state(ln);
+
+ break;
+ case CSIO_LN_FC_ATTRIB_UPDATE:
+ csio_fchost_attr_init(ln);
+ break;
+ default:
+ break;
+ }
+}
+
+/*
+ * csio_fchost_attr_init - Initialize FC transport attributes
+ * @ln: Lnode.
+ *
+ */
+void
+csio_fchost_attr_init(struct csio_lnode *ln)
+{
+ struct Scsi_Host *shost = csio_ln_to_shost(ln);
+
+ fc_host_node_name(shost) = wwn_to_u64(csio_ln_wwnn(ln));
+ fc_host_port_name(shost) = wwn_to_u64(csio_ln_wwpn(ln));
+
+ fc_host_supported_classes(shost) = FC_COS_CLASS3;
+ fc_host_max_npiv_vports(shost) =
+ (csio_lnode_to_hw(ln))->fres_info.max_vnps;
+ fc_host_supported_speeds(shost) = FC_PORTSPEED_10GBIT |
+ FC_PORTSPEED_1GBIT;
+
+ fc_host_maxframe_size(shost) = ln->ln_sparm.csp.rcv_sz;
+ memset(fc_host_supported_fc4s(shost), 0,
+ sizeof(fc_host_supported_fc4s(shost)));
+ fc_host_supported_fc4s(shost)[7] = 1;
+
+ memset(fc_host_active_fc4s(shost), 0,
+ sizeof(fc_host_active_fc4s(shost)));
+ fc_host_active_fc4s(shost)[7] = 1;
+}
+
+/*
+ * csio_get_host_port_id - sysfs entries for nport_id is
+ * populated/cached from this function
+ */
+static void
+csio_get_host_port_id(struct Scsi_Host *shost)
+{
+ struct csio_lnode *ln = shost_priv(shost);
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ spin_lock_irq(&hw->lock);
+ fc_host_port_id(shost) = ln->nport_id;
+ spin_unlock_irq(&hw->lock);
+}
+
+/*
+ * csio_get_port_type - Return FC local port type.
+ * @shost: scsi host.
+ *
+ */
+static void
+csio_get_host_port_type(struct Scsi_Host *shost)
+{
+ struct csio_lnode *ln = shost_priv(shost);
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ spin_lock_irq(&hw->lock);
+ if (csio_is_npiv_ln(ln))
+ fc_host_port_type(shost) = FC_PORTTYPE_NPIV;
+ else
+ fc_host_port_type(shost) = FC_PORTTYPE_NPORT;
+ spin_unlock_irq(&hw->lock);
+}
+
+/*
+ * csio_get_port_state - Return FC local port state.
+ * @shost: scsi host.
+ *
+ */
+static void
+csio_get_host_port_state(struct Scsi_Host *shost)
+{
+ struct csio_lnode *ln = shost_priv(shost);
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+ char state[16];
+
+ spin_lock_irq(&hw->lock);
+
+ csio_lnode_state_to_str(ln, state);
+ if (!strcmp(state, "READY"))
+ fc_host_port_state(shost) = FC_PORTSTATE_ONLINE;
+ else if (!strcmp(state, "OFFLINE"))
+ fc_host_port_state(shost) = FC_PORTSTATE_LINKDOWN;
+ else
+ fc_host_port_state(shost) = FC_PORTSTATE_UNKNOWN;
+
+ spin_unlock_irq(&hw->lock);
+}
+
+/*
+ * csio_get_host_speed - Return link speed to FC transport.
+ * @shost: scsi host.
+ *
+ */
+static void
+csio_get_host_speed(struct Scsi_Host *shost)
+{
+ struct csio_lnode *ln = shost_priv(shost);
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ spin_lock_irq(&hw->lock);
+ switch (hw->pport[ln->portid].link_speed) {
+ case FW_PORT_CAP_SPEED_1G:
+ fc_host_speed(shost) = FC_PORTSPEED_1GBIT;
+ break;
+ case FW_PORT_CAP_SPEED_10G:
+ fc_host_speed(shost) = FC_PORTSPEED_10GBIT;
+ break;
+ default:
+ fc_host_speed(shost) = FC_PORTSPEED_UNKNOWN;
+ break;
+ }
+ spin_unlock_irq(&hw->lock);
+}
+
+/*
+ * csio_get_host_fabric_name - Return fabric name
+ * @shost: scsi host.
+ *
+ */
+static void
+csio_get_host_fabric_name(struct Scsi_Host *shost)
+{
+ struct csio_lnode *ln = shost_priv(shost);
+ struct csio_rnode *rn = NULL;
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ spin_lock_irq(&hw->lock);
+ rn = csio_rnode_lookup_portid(ln, FABRIC_DID);
+ if (rn)
+ fc_host_fabric_name(shost) = wwn_to_u64(csio_rn_wwnn(rn));
+ else
+ fc_host_fabric_name(shost) = 0;
+ spin_unlock_irq(&hw->lock);
+}
+
+/*
+ * csio_get_host_speed - Return FC transport statistics.
+ * @ln: Lnode.
+ *
+ */
+static struct fc_host_statistics *
+csio_get_stats(struct Scsi_Host *shost)
+{
+ struct csio_lnode *ln = shost_priv(shost);
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+ struct fc_host_statistics *fhs = &ln->fch_stats;
+ struct fw_fcoe_port_stats fcoe_port_stats;
+ uint64_t seconds;
+
+ memset(&fcoe_port_stats, 0, sizeof(struct fw_fcoe_port_stats));
+ csio_get_phy_port_stats(hw, ln->portid, &fcoe_port_stats);
+
+ fhs->tx_frames += (fcoe_port_stats.tx_bcast_frames +
+ fcoe_port_stats.tx_mcast_frames +
+ fcoe_port_stats.tx_ucast_frames +
+ fcoe_port_stats.tx_offload_frames);
+ fhs->tx_words += (fcoe_port_stats.tx_bcast_bytes +
+ fcoe_port_stats.tx_mcast_bytes +
+ fcoe_port_stats.tx_ucast_bytes +
+ fcoe_port_stats.tx_offload_bytes) /
+ CSIO_WORD_TO_BYTE;
+ fhs->rx_frames += (fcoe_port_stats.rx_bcast_frames +
+ fcoe_port_stats.rx_mcast_frames +
+ fcoe_port_stats.rx_ucast_frames);
+ fhs->rx_words += (fcoe_port_stats.rx_bcast_bytes +
+ fcoe_port_stats.rx_mcast_bytes +
+ fcoe_port_stats.rx_ucast_bytes) /
+ CSIO_WORD_TO_BYTE;
+ fhs->error_frames += fcoe_port_stats.rx_err_frames;
+ fhs->fcp_input_requests += ln->stats.n_input_requests;
+ fhs->fcp_output_requests += ln->stats.n_output_requests;
+ fhs->fcp_control_requests += ln->stats.n_control_requests;
+ fhs->fcp_input_megabytes += ln->stats.n_input_bytes >> 20;
+ fhs->fcp_output_megabytes += ln->stats.n_output_bytes >> 20;
+ fhs->link_failure_count = ln->stats.n_link_down;
+ /* Reset stats for the device */
+ seconds = jiffies_to_msecs(jiffies);
+ fhs->seconds_since_last_reset =
+ (seconds - hw->stats.n_reset_start) / 1000;
+ return fhs;
+}
+
+/*
+ * csio_set_rport_loss_tmo - Set the rport dev loss timeout
+ * @rport: fc rport.
+ * @timeout: new value for dev loss tmo.
+ *
+ * If timeout is non zero set the dev_loss_tmo to timeout, else set
+ * dev_loss_tmo to one.
+ */
+static void
+csio_set_rport_loss_tmo(struct fc_rport *rport, uint32_t timeout)
+{
+ if (timeout)
+ rport->dev_loss_tmo = timeout;
+ else
+ rport->dev_loss_tmo = 1;
+}
+
+static void
+csio_vport_set_state(struct csio_lnode *ln)
+{
+ struct fc_vport *fc_vport = ln->fc_vport;
+ struct csio_lnode *pln = ln->pln;
+ char state[16];
+
+ /* Set fc vport state based on phyiscal lnode */
+ csio_lnode_state_to_str(pln, state);
+ if (strcmp(state, "READY")) {
+ fc_vport_set_state(fc_vport, FC_VPORT_LINKDOWN);
+ return;
+ }
+
+ if (!(pln->flags & CSIO_LNF_NPIVSUPP)) {
+ fc_vport_set_state(fc_vport, FC_VPORT_NO_FABRIC_SUPP);
+ return;
+ }
+
+ /* Set fc vport state based on virtual lnode */
+ csio_lnode_state_to_str(ln, state);
+ if (strcmp(state, "READY")) {
+ fc_vport_set_state(fc_vport, FC_VPORT_LINKDOWN);
+ return;
+ }
+ fc_vport_set_state(fc_vport, FC_VPORT_ACTIVE);
+}
+
+static csio_retval_t
+csio_fcoe_alloc_vnp(struct csio_hw *hw, struct csio_lnode *ln)
+{
+ struct csio_lnode *pln;
+ struct csio_mb *mbp;
+ struct fw_fcoe_vnp_cmd *rsp;
+ csio_retval_t ret;
+ int retry = 0;
+
+ /* Issue VNP cmd to alloc vport */
+ /* Allocate Mbox request */
+ spin_lock_irq(&hw->lock);
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_inc_stats(hw, n_err_nomem);
+ spin_unlock_irq(&hw->lock);
+ return CSIO_NOMEM;
+ }
+
+ pln = ln->pln;
+ ln->fcf_flowid = pln->fcf_flowid;
+ ln->portid = pln->portid;
+
+ csio_fcoe_vnp_alloc_init_mb(ln, mbp, CSIO_MB_DEFAULT_TMO,
+ pln->fcf_flowid, pln->vnp_flowid, 0,
+ csio_ln_wwnn(ln), csio_ln_wwpn(ln),
+ NULL);
+
+ for (retry = 0; retry < 3; retry++) {
+ /* FW is expected to complete vnp cmd in immediate mode
+ * without much delay.
+ * Otherwise, there will be increase in IO latency since HW
+ * lock is held till completion of vnp mbox cmd.
+ */
+ ret = csio_mb_issue(hw, mbp);
+ if (ret != CSIO_BUSY)
+ break;
+
+ /* Retry if mbox returns busy */
+ spin_unlock_irq(&hw->lock);
+ msleep(2000);
+ spin_lock_irq(&hw->lock);
+ }
+
+ if (ret != CSIO_SUCCESS) {
+ csio_ln_err(ln, "Failed to issue mbox FCoE VNP command\n");
+ mempool_free(mbp, hw->mb_mempool);
+ spin_unlock_irq(&hw->lock);
+ return ret;
+ }
+
+ /* Process Mbox response of VNP command */
+ rsp = (struct fw_fcoe_vnp_cmd *)(mbp->mb);
+ if (FW_CMD_RETVAL_GET(ntohl(rsp->alloc_to_len16)) != FW_SUCCESS) {
+ csio_ln_err(ln, "FCOE VNP ALLOC cmd returned 0x%x!\n",
+ FW_CMD_RETVAL_GET(ntohl(rsp->alloc_to_len16)));
+ mempool_free(mbp, hw->mb_mempool);
+ spin_unlock_irq(&hw->lock);
+ return CSIO_NOMEM;
+ }
+
+ ln->vnp_flowid = FW_FCOE_VNP_CMD_VNPI_GET(
+ ntohl(rsp->gen_wwn_to_vnpi));
+ memcpy(csio_ln_wwnn(ln), rsp->vnport_wwnn, 8);
+ memcpy(csio_ln_wwpn(ln), rsp->vnport_wwpn, 8);
+
+ csio_ln_dbg(ln, "FCOE VNPI: 0x%x\n", ln->vnp_flowid);
+ csio_ln_dbg(ln, "\tWWNN: %x%x%x%x%x%x%x%x\n",
+ ln->ln_sparm.wwnn[0], ln->ln_sparm.wwnn[1],
+ ln->ln_sparm.wwnn[2], ln->ln_sparm.wwnn[3],
+ ln->ln_sparm.wwnn[4], ln->ln_sparm.wwnn[5],
+ ln->ln_sparm.wwnn[6], ln->ln_sparm.wwnn[7]);
+ csio_ln_dbg(ln, "\tWWPN: %x%x%x%x%x%x%x%x\n",
+ ln->ln_sparm.wwpn[0], ln->ln_sparm.wwpn[1],
+ ln->ln_sparm.wwpn[2], ln->ln_sparm.wwpn[3],
+ ln->ln_sparm.wwpn[4], ln->ln_sparm.wwpn[5],
+ ln->ln_sparm.wwpn[6], ln->ln_sparm.wwpn[7]);
+
+ mempool_free(mbp, hw->mb_mempool);
+ spin_unlock_irq(&hw->lock);
+
+ return CSIO_SUCCESS;
+}
+
+static csio_retval_t
+csio_fcoe_free_vnp(struct csio_hw *hw, struct csio_lnode *ln)
+{
+ struct csio_lnode *pln;
+ struct csio_mb *mbp;
+ struct fw_fcoe_vnp_cmd *rsp;
+ csio_retval_t ret;
+ int retry = 0;
+
+ /* Issue VNP cmd to free vport */
+ /* Allocate Mbox request */
+
+ spin_lock_irq(&hw->lock);
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_inc_stats(hw, n_err_nomem);
+ spin_unlock_irq(&hw->lock);
+ return CSIO_NOMEM;
+ }
+
+ pln = ln->pln;
+
+ csio_fcoe_vnp_free_init_mb(ln, mbp, CSIO_MB_DEFAULT_TMO,
+ ln->fcf_flowid, ln->vnp_flowid,
+ NULL);
+
+ for (retry = 0; retry < 3; retry++) {
+ ret = csio_mb_issue(hw, mbp);
+ if (ret != CSIO_BUSY)
+ break;
+
+ /* Retry if mbox returns busy */
+ spin_unlock_irq(&hw->lock);
+ msleep(2000);
+ spin_lock_irq(&hw->lock);
+ }
+
+ if (ret) {
+ csio_ln_err(ln, "Failed to issue mbox FCoE VNP command\n");
+ mempool_free(mbp, hw->mb_mempool);
+ spin_unlock_irq(&hw->lock);
+ return CSIO_INVAL;
+ }
+
+ /* Process Mbox response of VNP command */
+ rsp = (struct fw_fcoe_vnp_cmd *)(mbp->mb);
+ if (FW_CMD_RETVAL_GET(ntohl(rsp->alloc_to_len16)) != FW_SUCCESS) {
+ csio_ln_err(ln, "FCOE VNP FREE cmd returned 0x%x!\n",
+ FW_CMD_RETVAL_GET(ntohl(rsp->alloc_to_len16)));
+ mempool_free(mbp, hw->mb_mempool);
+ spin_unlock_irq(&hw->lock);
+ return CSIO_NOMEM;
+ }
+
+ mempool_free(mbp, hw->mb_mempool);
+ spin_unlock_irq(&hw->lock);
+
+ return CSIO_SUCCESS;
+}
+
+static int
+csio_vport_create(struct fc_vport *fc_vport, bool disable)
+{
+ struct Scsi_Host *shost = fc_vport->shost;
+ struct csio_lnode *pln = shost_priv(shost);
+ struct csio_lnode *ln = NULL;
+ struct csio_hw *hw = csio_lnode_to_hw(pln);
+ uint8_t wwn[8];
+ int ret = -1;
+
+ ln = csio_shost_init(hw, &fc_vport->dev, CSIO_FALSE, pln);
+ if (!ln)
+ goto error;
+
+ if (fc_vport->node_name != 0) {
+ u64_to_wwn(fc_vport->node_name, wwn);
+
+ if (!csio_valid_wwn(wwn)) {
+ csio_ln_err(ln,
+ "vport create failed. Invalid wwnn\n");
+ goto error;
+ }
+ memcpy(csio_ln_wwnn(ln), wwn, 8);
+ }
+
+ if (fc_vport->port_name != 0) {
+ u64_to_wwn(fc_vport->port_name, wwn);
+
+ if (!csio_valid_wwn(wwn)) {
+ csio_ln_err(ln,
+ "vport create failed. Invalid wwpn\n");
+ goto error;
+ }
+
+ if (csio_lnode_lookup_by_wwpn(hw, wwn)) {
+ csio_ln_err(ln,
+ "vport create failed. wwpn already exists\n");
+ goto error;
+ }
+ memcpy(csio_ln_wwpn(ln), wwn, 8);
+ }
+
+ fc_vport_set_state(fc_vport, FC_VPORT_INITIALIZING);
+
+ if (csio_fcoe_alloc_vnp(hw, ln))
+ goto error;
+
+ *(struct csio_lnode **)fc_vport->dd_data = ln;
+ ln->fc_vport = fc_vport;
+ if (!fc_vport->node_name)
+ fc_vport->node_name = wwn_to_u64(csio_ln_wwnn(ln));
+ if (!fc_vport->port_name)
+ fc_vport->port_name = wwn_to_u64(csio_ln_wwpn(ln));
+ csio_fchost_attr_init(ln);
+ return 0;
+error:
+ if (ln)
+ csio_shost_exit(ln);
+
+ return ret;
+}
+
+static int
+csio_vport_delete(struct fc_vport *fc_vport)
+{
+ struct csio_lnode *ln = *(struct csio_lnode **)fc_vport->dd_data;
+ struct Scsi_Host *shost = csio_ln_to_shost(ln);
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ spin_lock_irq(&hw->lock);
+ if (csio_is_hw_removing(hw)) {
+ spin_unlock_irq(&hw->lock);
+ csio_shost_exit(ln);
+ return 0;
+ }
+ spin_unlock_irq(&hw->lock);
+
+ /* Quiesce ios and send remove event to lnode */
+ scsi_block_requests(shost);
+ spin_lock_irq(&hw->lock);
+ csio_scsim_cleanup_io_lnode(csio_hw_to_scsim(hw), ln);
+ csio_lnode_close(ln);
+ spin_unlock_irq(&hw->lock);
+ scsi_unblock_requests(shost);
+
+ /* Free vnp */
+ if (fc_vport->vport_state != FC_VPORT_DISABLED)
+ csio_fcoe_free_vnp(hw, ln);
+ csio_ln_err(ln, "vport deleted\n");
+ csio_shost_exit(ln);
+ return 0;
+}
+
+static int
+csio_vport_disable(struct fc_vport *fc_vport, bool disable)
+{
+ struct csio_lnode *ln = *(struct csio_lnode **)fc_vport->dd_data;
+ struct Scsi_Host *shost = csio_ln_to_shost(ln);
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ /* disable vport */
+ if (disable) {
+ /* Quiesce ios and send stop event to lnode */
+ scsi_block_requests(shost);
+ spin_lock_irq(&hw->lock);
+ csio_scsim_cleanup_io_lnode(csio_hw_to_scsim(hw), ln);
+ csio_lnode_stop(ln);
+ spin_unlock_irq(&hw->lock);
+ scsi_unblock_requests(shost);
+
+ /* Free vnp */
+ csio_fcoe_free_vnp(hw, ln);
+ fc_vport_set_state(fc_vport, FC_VPORT_DISABLED);
+ csio_ln_err(ln, "vport disabled\n");
+ return 0;
+ } else {
+ /* enable vport */
+ fc_vport_set_state(fc_vport, FC_VPORT_INITIALIZING);
+ if (csio_fcoe_alloc_vnp(hw, ln)) {
+ csio_ln_err(ln, "vport enabled failed.\n");
+ return -1;
+ }
+ csio_ln_err(ln, "vport enabled\n");
+ return 0;
+ }
+}
+
+static void
+csio_dev_loss_tmo_callbk(struct fc_rport *rport)
+{
+ struct csio_rnode *rn;
+ struct csio_hw *hw;
+ struct csio_lnode *ln;
+
+ rn = *((struct csio_rnode **)rport->dd_data);
+ ln = csio_rnode_to_lnode(rn);
+ hw = csio_lnode_to_hw(ln);
+
+ spin_lock_irq(&hw->lock);
+
+ /* return if driver is being removed or same rnode comes back online */
+ if (csio_is_hw_removing(hw) || csio_is_rnode_ready(rn)) {
+ spin_unlock_irq(&hw->lock);
+ return;
+ }
+
+ csio_ln_dbg(ln, "devloss timeout on rnode:%p portid:x%x flowid:x%x\n",
+ rn, rn->nport_id, csio_rn_flowid(rn));
+
+ csio_inc_stats(ln, n_dev_loss_tmo);
+
+ /*
+ * enqueue devloss event to event worker thread to serialize all
+ * rnode events.
+ */
+ if (csio_enqueue_evt(hw, CSIO_EVT_DEV_LOSS, &rn, sizeof(rn))) {
+ csio_inc_stats(hw, n_evt_drop);
+ spin_unlock_irq(&hw->lock);
+ return;
+ }
+
+ if (!(hw->flags & CSIO_HWF_FWEVT_PENDING)) {
+ hw->flags |= CSIO_HWF_FWEVT_PENDING;
+ spin_unlock_irq(&hw->lock);
+ schedule_work(&hw->evtq_work);
+ return;
+ }
+
+ spin_unlock_irq(&hw->lock);
+}
+
+/* FC transport functions template - Physical port */
+struct fc_function_template csio_fc_transport_funcs = {
+ .show_host_node_name = 1,
+ .show_host_port_name = 1,
+ .show_host_supported_classes = 1,
+ .show_host_supported_fc4s = 1,
+ .show_host_maxframe_size = 1,
+
+ .get_host_port_id = csio_get_host_port_id,
+ .show_host_port_id = 1,
+
+ .get_host_port_type = csio_get_host_port_type,
+ .show_host_port_type = 1,
+
+ .get_host_port_state = csio_get_host_port_state,
+ .show_host_port_state = 1,
+
+ .show_host_active_fc4s = 1,
+ .get_host_speed = csio_get_host_speed,
+ .show_host_speed = 1,
+ .get_host_fabric_name = csio_get_host_fabric_name,
+ .show_host_fabric_name = 1,
+
+ .get_fc_host_stats = csio_get_stats,
+
+ .dd_fcrport_size = sizeof(struct csio_rnode *),
+ .show_rport_maxframe_size = 1,
+ .show_rport_supported_classes = 1,
+
+ .set_rport_dev_loss_tmo = csio_set_rport_loss_tmo,
+ .show_rport_dev_loss_tmo = 1,
+
+ .show_starget_port_id = 1,
+ .show_starget_node_name = 1,
+ .show_starget_port_name = 1,
+
+ .dev_loss_tmo_callbk = csio_dev_loss_tmo_callbk,
+ .dd_fcvport_size = sizeof(struct csio_lnode *),
+
+ .vport_create = csio_vport_create,
+ .vport_disable = csio_vport_disable,
+ .vport_delete = csio_vport_delete,
+};
+
+/* FC transport functions template - Virtual port */
+struct fc_function_template csio_fc_transport_vport_funcs = {
+ .show_host_node_name = 1,
+ .show_host_port_name = 1,
+ .show_host_supported_classes = 1,
+ .show_host_supported_fc4s = 1,
+ .show_host_maxframe_size = 1,
+
+ .get_host_port_id = csio_get_host_port_id,
+ .show_host_port_id = 1,
+
+ .get_host_port_type = csio_get_host_port_type,
+ .show_host_port_type = 1,
+
+ .get_host_port_state = csio_get_host_port_state,
+ .show_host_port_state = 1,
+ .show_host_active_fc4s = 1,
+
+ .get_host_speed = csio_get_host_speed,
+ .show_host_speed = 1,
+
+ .get_host_fabric_name = csio_get_host_fabric_name,
+ .show_host_fabric_name = 1,
+
+ .get_fc_host_stats = csio_get_stats,
+
+ .dd_fcrport_size = sizeof(struct csio_rnode *),
+ .show_rport_maxframe_size = 1,
+ .show_rport_supported_classes = 1,
+
+ .set_rport_dev_loss_tmo = csio_set_rport_loss_tmo,
+ .show_rport_dev_loss_tmo = 1,
+
+ .show_starget_port_id = 1,
+ .show_starget_node_name = 1,
+ .show_starget_port_name = 1,
+
+ .dev_loss_tmo_callbk = csio_dev_loss_tmo_callbk,
+
+};
diff --git a/drivers/scsi/csiostor/csio_mb.c b/drivers/scsi/csiostor/csio_mb.c
new file mode 100644
index 0000000..128fa1b
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_mb.c
@@ -0,0 +1,1768 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/delay.h>
+#include <linux/jiffies.h>
+#include <linux/string.h>
+#include <scsi/scsi_transport_fc.h>
+
+#include "csio_hw.h"
+#include "csio_lnode.h"
+#include "csio_rnode.h"
+#include "csio_mb.h"
+#include "csio_wr.h"
+
+#define csio_mb_is_host_owner(__owner) ((__owner) == CSIO_MBOWNER_PL)
+
+/* MB Command/Response Helpers */
+/*
+ * csio_mb_fw_retval - FW return value from a mailbox response.
+ * @mbp: Mailbox structure
+ *
+ */
+enum fw_retval
+csio_mb_fw_retval(struct csio_mb *mbp)
+{
+ struct fw_cmd_hdr *hdr;
+
+ hdr = (struct fw_cmd_hdr *)(mbp->mb);
+
+ return FW_CMD_RETVAL_GET(ntohl(hdr->lo));
+}
+
+/*
+ * csio_mb_hello - FW HELLO command helper
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @m_mbox: Master mailbox number, if any.
+ * @a_mbox: Mailbox number for asycn notifications.
+ * @master: Device mastership.
+ * @cbfn: Callback, if any.
+ *
+ */
+void
+csio_mb_hello(struct csio_hw *hw, struct csio_mb *mbp, uint32_t tmo,
+ uint32_t m_mbox, uint32_t a_mbox, enum csio_dev_master master,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_hello_cmd *cmdp = (struct fw_hello_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, tmo, hw, cbfn, 1);
+
+ cmdp->op_to_write = htonl(FW_CMD_OP(FW_HELLO_CMD) |
+ FW_CMD_REQUEST | FW_CMD_WRITE);
+ cmdp->retval_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+ cmdp->err_to_clearinit = htonl(
+ FW_HELLO_CMD_MASTERDIS(master == CSIO_MASTER_CANT) |
+ FW_HELLO_CMD_MASTERFORCE(master == CSIO_MASTER_MUST) |
+ FW_HELLO_CMD_MBMASTER(master == CSIO_MASTER_MUST ?
+ m_mbox : FW_HELLO_CMD_MBMASTER_MASK) |
+ FW_HELLO_CMD_MBASYNCNOT(a_mbox) |
+ FW_HELLO_CMD_STAGE(FW_HELLO_CMD_STAGE_OS) |
+ FW_HELLO_CMD_CLEARINIT);
+
+}
+
+/*
+ * csio_mb_process_hello_rsp - FW HELLO response processing helper
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @retval: Mailbox return value from Firmware
+ * @state: State that the function is in.
+ * @mpfn: Master pfn
+ *
+ */
+void
+csio_mb_process_hello_rsp(struct csio_hw *hw, struct csio_mb *mbp,
+ enum fw_retval *retval, enum csio_dev_state *state,
+ uint8_t *mpfn)
+{
+ struct fw_hello_cmd *rsp = (struct fw_hello_cmd *)(mbp->mb);
+ uint32_t value;
+
+ *retval = FW_CMD_RETVAL_GET(ntohl(rsp->retval_len16));
+
+ if (*retval == FW_SUCCESS) {
+ hw->fwrev = ntohl(rsp->fwrev);
+
+ value = ntohl(rsp->err_to_clearinit);
+ *mpfn = FW_HELLO_CMD_MBMASTER_GET(value);
+
+ if (value & FW_HELLO_CMD_INIT)
+ *state = CSIO_DEV_STATE_INIT;
+ else if (value & FW_HELLO_CMD_ERR)
+ *state = CSIO_DEV_STATE_ERR;
+ else
+ *state = CSIO_DEV_STATE_UNINIT;
+ }
+}
+
+/*
+ * csio_mb_bye - FW BYE command helper
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @cbfn: Callback, if any.
+ *
+ */
+void
+csio_mb_bye(struct csio_hw *hw, struct csio_mb *mbp, uint32_t tmo,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_bye_cmd *cmdp = (struct fw_bye_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, tmo, hw, cbfn, 1);
+
+ cmdp->op_to_write = htonl(FW_CMD_OP(FW_BYE_CMD) |
+ FW_CMD_REQUEST | FW_CMD_WRITE);
+ cmdp->retval_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+}
+
+/*
+ * csio_mb_reset - FW RESET command helper
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @reset: Type of reset.
+ * @cbfn: Callback, if any.
+ *
+ */
+void
+csio_mb_reset(struct csio_hw *hw, struct csio_mb *mbp, uint32_t tmo,
+ int reset, int halt,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_reset_cmd *cmdp = (struct fw_reset_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, tmo, hw, cbfn, 1);
+
+ cmdp->op_to_write = htonl(FW_CMD_OP(FW_RESET_CMD) |
+ FW_CMD_REQUEST | FW_CMD_WRITE);
+ cmdp->retval_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+ cmdp->val = htonl(reset);
+ cmdp->halt_pkd = htonl(halt);
+
+}
+
+/*
+ * csio_mb_params - FW PARAMS command helper
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @tmo: Command timeout.
+ * @pf: PF number.
+ * @vf: VF number.
+ * @nparams: Number of paramters
+ * @params: Parameter mnemonic array.
+ * @val: Parameter value array.
+ * @wr: Write/Read PARAMS.
+ * @cbfn: Callback, if any.
+ *
+ */
+void
+csio_mb_params(struct csio_hw *hw, struct csio_mb *mbp, uint32_t tmo,
+ unsigned int pf, unsigned int vf, unsigned int nparams,
+ const u32 *params, u32 *val, bool wr,
+ void (*cbfn)(struct csio_hw *, struct csio_mb *))
+{
+ uint32_t i;
+ uint32_t temp_params = 0, temp_val = 0;
+ struct fw_params_cmd *cmdp = (struct fw_params_cmd *)(mbp->mb);
+ __be32 *p = &cmdp->param[0].mnem;
+
+ CSIO_INIT_MBP(mbp, cmdp, tmo, hw, cbfn, 1);
+
+ cmdp->op_to_vfn = htonl(FW_CMD_OP(FW_PARAMS_CMD) |
+ FW_CMD_REQUEST |
+ (wr ? FW_CMD_WRITE : FW_CMD_READ) |
+ FW_PARAMS_CMD_PFN(pf) |
+ FW_PARAMS_CMD_VFN(vf));
+ cmdp->retval_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+ /* Write Params */
+ if (wr) {
+ while (nparams--) {
+ temp_params = *params++;
+ temp_val = *val++;
+
+ *p++ = htonl(temp_params);
+ *p++ = htonl(temp_val);
+ }
+ } else {
+ for (i = 0; i < nparams; i++, p += 2) {
+ temp_params = *params++;
+ *p = htonl(temp_params);
+ }
+ }
+
+}
+
+/*
+ * csio_mb_process_read_params_rsp - FW PARAMS response processing helper
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @retval: Mailbox return value from Firmware
+ * @nparams: Number of parameters
+ * @val: Parameter value array.
+ *
+ */
+void
+csio_mb_process_read_params_rsp(struct csio_hw *hw, struct csio_mb *mbp,
+ enum fw_retval *retval, unsigned int nparams,
+ u32 *val)
+{
+ struct fw_params_cmd *rsp = (struct fw_params_cmd *)(mbp->mb);
+ uint32_t i;
+ __be32 *p = &rsp->param[0].val;
+
+ *retval = FW_CMD_RETVAL_GET(ntohl(rsp->retval_len16));
+
+ if (*retval == FW_SUCCESS)
+ for (i = 0; i < nparams; i++, p += 2)
+ *val++ = ntohl(*p);
+}
+
+/*
+ * csio_mb_ldst - FW LDST command
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @tmo: timeout
+ * @reg: register
+ *
+ */
+void
+csio_mb_ldst(struct csio_hw *hw, struct csio_mb *mbp, uint32_t tmo, int reg)
+{
+ struct fw_ldst_cmd *ldst_cmd = (struct fw_ldst_cmd *)(mbp->mb);
+ CSIO_INIT_MBP(mbp, ldst_cmd, tmo, hw, NULL, 1);
+
+ /*
+ * Construct and send the Firmware LDST Command to retrieve the
+ * specified PCI-E Configuration Space register.
+ */
+ ldst_cmd->op_to_addrspace =
+ htonl(FW_CMD_OP(FW_LDST_CMD) |
+ FW_CMD_REQUEST |
+ FW_CMD_READ |
+ FW_LDST_CMD_ADDRSPACE(FW_LDST_ADDRSPC_FUNC_PCIE));
+ ldst_cmd->cycles_to_len16 = htonl(FW_LEN16(struct fw_ldst_cmd));
+ ldst_cmd->u.pcie.select_naccess = FW_LDST_CMD_NACCESS(1);
+ ldst_cmd->u.pcie.ctrl_to_fn =
+ (FW_LDST_CMD_LC | FW_LDST_CMD_FN(hw->pfn));
+ ldst_cmd->u.pcie.r = (uint8_t)reg;
+}
+
+/*
+ *
+ * csio_mb_caps_config - FW Read/Write Capabilities command helper
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @wr: Write if 1, Read if 0
+ * @init: Turn on initiator mode.
+ * @tgt: Turn on target mode.
+ * @cofld: If 1, Control Offload for FCoE
+ * @cbfn: Callback, if any.
+ *
+ * This helper assumes that cmdp has MB payload from a previous CAPS
+ * read command.
+ */
+void
+csio_mb_caps_config(struct csio_hw *hw, struct csio_mb *mbp, uint32_t tmo,
+ bool wr, bool init, bool tgt, bool cofld,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_caps_config_cmd *cmdp =
+ (struct fw_caps_config_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, tmo, hw, cbfn, wr ? 0 : 1);
+
+ cmdp->op_to_write = htonl(FW_CMD_OP(FW_CAPS_CONFIG_CMD) |
+ FW_CMD_REQUEST |
+ (wr ? FW_CMD_WRITE : FW_CMD_READ));
+ cmdp->cfvalid_to_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+ /* Read config */
+ if (!wr)
+ return;
+
+ /* Write config */
+ cmdp->fcoecaps = 0;
+
+ if (cofld)
+ cmdp->fcoecaps |= htons(FW_CAPS_CONFIG_FCOE_CTRL_OFLD);
+ if (init)
+ cmdp->fcoecaps |= htons(FW_CAPS_CONFIG_FCOE_INITIATOR);
+ if (tgt)
+ cmdp->fcoecaps |= htons(FW_CAPS_CONFIG_FCOE_TARGET);
+}
+
+void
+csio_rss_glb_config(struct csio_hw *hw, struct csio_mb *mbp,
+ uint32_t tmo, uint8_t mode, unsigned int flags,
+ void (*cbfn)(struct csio_hw *, struct csio_mb *))
+{
+ struct fw_rss_glb_config_cmd *cmdp =
+ (struct fw_rss_glb_config_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, tmo, hw, cbfn, 1);
+
+ cmdp->op_to_write = htonl(FW_CMD_OP(FW_RSS_GLB_CONFIG_CMD) |
+ FW_CMD_REQUEST | FW_CMD_WRITE);
+ cmdp->retval_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+ if (mode == FW_RSS_GLB_CONFIG_CMD_MODE_MANUAL) {
+ cmdp->u.manual.mode_pkd =
+ htonl(FW_RSS_GLB_CONFIG_CMD_MODE(mode));
+ } else if (mode == FW_RSS_GLB_CONFIG_CMD_MODE_BASICVIRTUAL) {
+ cmdp->u.basicvirtual.mode_pkd =
+ htonl(FW_RSS_GLB_CONFIG_CMD_MODE(mode));
+ cmdp->u.basicvirtual.synmapen_to_hashtoeplitz = htonl(flags);
+ }
+}
+
+
+/*
+ * csio_mb_pfvf - FW Write PF/VF capabilities command helper.
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @pf:
+ * @vf:
+ * @txq:
+ * @txq_eht_ctrl:
+ * @rxqi:
+ * @rxq:
+ * @tc:
+ * @vi:
+ * @pmask:
+ * @rcaps:
+ * @wxcaps:
+ * @cbfn: Callback, if any.
+ *
+ */
+void
+csio_mb_pfvf(struct csio_hw *hw, struct csio_mb *mbp, uint32_t tmo,
+ unsigned int pf, unsigned int vf, unsigned int txq,
+ unsigned int txq_eth_ctrl, unsigned int rxqi,
+ unsigned int rxq, unsigned int tc, unsigned int vi,
+ unsigned int cmask, unsigned int pmask, unsigned int nexactf,
+ unsigned int rcaps, unsigned int wxcaps,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_pfvf_cmd *cmdp = (struct fw_pfvf_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, tmo, hw, cbfn, 1);
+
+ cmdp->op_to_vfn = htonl(FW_CMD_OP(FW_PFVF_CMD) |
+ FW_CMD_REQUEST |
+ FW_CMD_WRITE |
+ FW_PFVF_CMD_PFN(pf) |
+ FW_PFVF_CMD_VFN(vf));
+ cmdp->retval_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+ cmdp->niqflint_niq = htonl(FW_PFVF_CMD_NIQFLINT(rxqi) |
+ FW_PFVF_CMD_NIQ(rxq));
+
+ cmdp->type_to_neq = htonl(FW_PFVF_CMD_TYPE |
+ FW_PFVF_CMD_CMASK(cmask) |
+ FW_PFVF_CMD_PMASK(pmask) |
+ FW_PFVF_CMD_NEQ(txq));
+ cmdp->tc_to_nexactf = htonl(FW_PFVF_CMD_TC(tc) |
+ FW_PFVF_CMD_NVI(vi) |
+ FW_PFVF_CMD_NEXACTF(nexactf));
+ cmdp->r_caps_to_nethctrl = htonl(FW_PFVF_CMD_R_CAPS(rcaps) |
+ FW_PFVF_CMD_WX_CAPS(wxcaps) |
+ FW_PFVF_CMD_NETHCTRL(txq_eth_ctrl));
+}
+
+#define CSIO_ADVERT_MASK (FW_PORT_CAP_SPEED_100M | FW_PORT_CAP_SPEED_1G |\
+ FW_PORT_CAP_SPEED_10G | FW_PORT_CAP_ANEG)
+
+/*
+ * csio_mb_port- FW PORT command helper
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @tmo: COmmand timeout
+ * @portid: Port ID to get/set info
+ * @wr: Write/Read PORT information.
+ * @fc: Flow control
+ * @caps: Port capabilites to set.
+ * @cbfn: Callback, if any.
+ *
+ */
+void
+csio_mb_port(struct csio_hw *hw, struct csio_mb *mbp, uint32_t tmo,
+ uint8_t portid, bool wr, uint32_t fc, uint16_t caps,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_port_cmd *cmdp = (struct fw_port_cmd *)(mbp->mb);
+ unsigned int lfc = 0, mdi = FW_PORT_MDI(FW_PORT_MDI_AUTO);
+
+ CSIO_INIT_MBP(mbp, cmdp, tmo, hw, cbfn, 1);
+
+ cmdp->op_to_portid = htonl(FW_CMD_OP(FW_PORT_CMD) |
+ FW_CMD_REQUEST |
+ (wr ? FW_CMD_EXEC : FW_CMD_READ) |
+ FW_PORT_CMD_PORTID(portid));
+ if (!wr) {
+ cmdp->action_to_len16 = htonl(
+ FW_PORT_CMD_ACTION(FW_PORT_ACTION_GET_PORT_INFO) |
+ FW_CMD_LEN16(sizeof(*cmdp) / 16));
+ return;
+ }
+
+ /* Set port */
+ cmdp->action_to_len16 = htonl(
+ FW_PORT_CMD_ACTION(FW_PORT_ACTION_L1_CFG) |
+ FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+ if (fc & PAUSE_RX)
+ lfc |= FW_PORT_CAP_FC_RX;
+ if (fc & PAUSE_TX)
+ lfc |= FW_PORT_CAP_FC_TX;
+
+ if (!(caps & FW_PORT_CAP_ANEG))
+ cmdp->u.l1cfg.rcap = htonl((caps & CSIO_ADVERT_MASK) | lfc);
+ else
+ cmdp->u.l1cfg.rcap = htonl((caps & CSIO_ADVERT_MASK) |
+ lfc | mdi);
+}
+
+/*
+ * csio_mb_process_read_port_rsp - FW PORT command response processing helper
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @retval: Mailbox return value from Firmware
+ * @caps: port capabilities
+ *
+ */
+void
+csio_mb_process_read_port_rsp(struct csio_hw *hw, struct csio_mb *mbp,
+ enum fw_retval *retval, uint16_t *caps)
+{
+ struct fw_port_cmd *rsp = (struct fw_port_cmd *)(mbp->mb);
+
+ *retval = FW_CMD_RETVAL_GET(ntohl(rsp->action_to_len16));
+
+ if (*retval == FW_SUCCESS)
+ *caps = ntohs(rsp->u.info.pcap);
+}
+
+/*
+ * csio_mb_initialize - FW INITIALIZE command helper
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @tmo: COmmand timeout
+ * @cbfn: Callback, if any.
+ *
+ */
+void
+csio_mb_initialize(struct csio_hw *hw, struct csio_mb *mbp, uint32_t tmo,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_initialize_cmd *cmdp = (struct fw_initialize_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, tmo, hw, cbfn, 1);
+
+ cmdp->op_to_write = htonl(FW_CMD_OP(FW_INITIALIZE_CMD) |
+ FW_CMD_REQUEST | FW_CMD_WRITE);
+ cmdp->retval_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+}
+
+/*
+ * csio_mb_iq_alloc - Initializes the mailbox to allocate an
+ * Ingress DMA queue in the firmware.
+ *
+ * @hw: The hw structure
+ * @mbp: Mailbox structure to initialize
+ * @priv: Private object
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @iq_params: Ingress queue params needed for allocation.
+ * @cbfn: The call-back function
+ *
+ *
+ */
+static void
+csio_mb_iq_alloc(struct csio_hw *hw, struct csio_mb *mbp, void *priv,
+ uint32_t mb_tmo, struct csio_iq_params *iq_params,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_iq_cmd *cmdp = (struct fw_iq_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, priv, cbfn, 1);
+
+ cmdp->op_to_vfn = htonl(FW_CMD_OP(FW_IQ_CMD) |
+ FW_CMD_REQUEST | FW_CMD_EXEC |
+ FW_IQ_CMD_PFN(iq_params->pfn) |
+ FW_IQ_CMD_VFN(iq_params->vfn));
+
+ cmdp->alloc_to_len16 = htonl(FW_IQ_CMD_ALLOC |
+ FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+ cmdp->type_to_iqandstindex = htonl(
+ FW_IQ_CMD_VIID(iq_params->viid) |
+ FW_IQ_CMD_TYPE(iq_params->type) |
+ FW_IQ_CMD_IQASYNCH(iq_params->iqasynch));
+
+ cmdp->fl0size = htons(iq_params->fl0size);
+ cmdp->fl0size = htons(iq_params->fl1size);
+
+} /* csio_mb_iq_alloc */
+
+/*
+ * csio_mb_iq_write - Initializes the mailbox for writing into an
+ * Ingress DMA Queue.
+ *
+ * @hw: The HW structure
+ * @mbp: Mailbox structure to initialize
+ * @priv: Private object
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @cascaded_req: TRUE - if this request is cascased with iq-alloc request.
+ * @iq_params: Ingress queue params needed for writing.
+ * @cbfn: The call-back function
+ *
+ * NOTE: We OR relevant bits with cmdp->XXX, instead of just equating,
+ * because this IQ write request can be cascaded with a previous
+ * IQ alloc request, and we dont want to over-write the bits set by
+ * that request. This logic will work even in a non-cascaded case, since the
+ * cmdp structure is zeroed out by CSIO_INIT_MBP.
+ */
+static void
+csio_mb_iq_write(struct csio_hw *hw, struct csio_mb *mbp, void *priv,
+ uint32_t mb_tmo, bool cascaded_req,
+ struct csio_iq_params *iq_params,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_iq_cmd *cmdp = (struct fw_iq_cmd *)(mbp->mb);
+
+ uint32_t iq_start_stop = (iq_params->iq_start) ?
+ FW_IQ_CMD_IQSTART(1) :
+ FW_IQ_CMD_IQSTOP(1);
+
+ /*
+ * If this IQ write is cascaded with IQ alloc request, do not
+ * re-initialize with 0's.
+ *
+ */
+ if (!cascaded_req)
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, priv, cbfn, 1);
+
+ cmdp->op_to_vfn |= htonl(FW_CMD_OP(FW_IQ_CMD) |
+ FW_CMD_REQUEST | FW_CMD_WRITE |
+ FW_IQ_CMD_PFN(iq_params->pfn) |
+ FW_IQ_CMD_VFN(iq_params->vfn));
+ cmdp->alloc_to_len16 |= htonl(iq_start_stop |
+ FW_CMD_LEN16(sizeof(*cmdp) / 16));
+ cmdp->iqid |= htons(iq_params->iqid);
+ cmdp->fl0id |= htons(iq_params->fl0id);
+ cmdp->fl1id |= htons(iq_params->fl1id);
+ cmdp->type_to_iqandstindex |= htonl(
+ FW_IQ_CMD_IQANDST(iq_params->iqandst) |
+ FW_IQ_CMD_IQANUS(iq_params->iqanus) |
+ FW_IQ_CMD_IQANUD(iq_params->iqanud) |
+ FW_IQ_CMD_IQANDSTINDEX(iq_params->iqandstindex));
+ cmdp->iqdroprss_to_iqesize |= htons(
+ FW_IQ_CMD_IQPCIECH(iq_params->iqpciech) |
+ FW_IQ_CMD_IQDCAEN(iq_params->iqdcaen) |
+ FW_IQ_CMD_IQDCACPU(iq_params->iqdcacpu) |
+ FW_IQ_CMD_IQINTCNTTHRESH(iq_params->iqintcntthresh) |
+ FW_IQ_CMD_IQCPRIO(iq_params->iqcprio) |
+ FW_IQ_CMD_IQESIZE(iq_params->iqesize));
+
+ cmdp->iqsize |= htons(iq_params->iqsize);
+ cmdp->iqaddr |= cpu_to_be64(iq_params->iqaddr);
+
+ if (iq_params->type == 0) {
+ cmdp->iqns_to_fl0congen |= htonl(
+ FW_IQ_CMD_IQFLINTIQHSEN(iq_params->iqflintiqhsen)|
+ FW_IQ_CMD_IQFLINTCONGEN(iq_params->iqflintcongen));
+ }
+
+ if (iq_params->fl0size && iq_params->fl0addr &&
+ (iq_params->fl0id != 0xFFFF)) {
+
+ cmdp->iqns_to_fl0congen |= htonl(
+ FW_IQ_CMD_FL0HOSTFCMODE(iq_params->fl0hostfcmode)|
+ FW_IQ_CMD_FL0CPRIO(iq_params->fl0cprio) |
+ FW_IQ_CMD_FL0PADEN(iq_params->fl0paden) |
+ FW_IQ_CMD_FL0PACKEN(iq_params->fl0packen));
+ cmdp->fl0dcaen_to_fl0cidxfthresh |= htons(
+ FW_IQ_CMD_FL0DCAEN(iq_params->fl0dcaen) |
+ FW_IQ_CMD_FL0DCACPU(iq_params->fl0dcacpu) |
+ FW_IQ_CMD_FL0FBMIN(iq_params->fl0fbmin) |
+ FW_IQ_CMD_FL0FBMAX(iq_params->fl0fbmax) |
+ FW_IQ_CMD_FL0CIDXFTHRESH(iq_params->fl0cidxfthresh));
+ cmdp->fl0size |= htons(iq_params->fl0size);
+ cmdp->fl0addr |= cpu_to_be64(iq_params->fl0addr);
+ }
+} /* csio_mb_iq_write */
+
+/*
+ * csio_mb_iq_alloc_write - Initializes the mailbox for allocating an
+ * Ingress DMA Queue.
+ *
+ * @hw: The HW structure
+ * @mbp: Mailbox structure to initialize
+ * @priv: Private data.
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @iq_params: Ingress queue params needed for allocation & writing.
+ * @cbfn: The call-back function
+ *
+ *
+ */
+void
+csio_mb_iq_alloc_write(struct csio_hw *hw, struct csio_mb *mbp, void *priv,
+ uint32_t mb_tmo, struct csio_iq_params *iq_params,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ csio_mb_iq_alloc(hw, mbp, priv, mb_tmo, iq_params, cbfn);
+ csio_mb_iq_write(hw, mbp, priv, mb_tmo, CSIO_TRUE, iq_params, cbfn);
+} /* csio_mb_iq_alloc_write */
+
+/*
+ * csio_mb_iq_alloc_write_rsp - Process the allocation & writing
+ * of ingress DMA queue mailbox's response.
+ *
+ * @hw: The HW structure.
+ * @mbp: Mailbox structure to initialize.
+ * @retval: Firmware return value.
+ * @iq_params: Ingress queue parameters, after allocation and write.
+ *
+ */
+void
+csio_mb_iq_alloc_write_rsp(struct csio_hw *hw, struct csio_mb *mbp,
+ enum fw_retval *ret_val,
+ struct csio_iq_params *iq_params)
+{
+ struct fw_iq_cmd *rsp = (struct fw_iq_cmd *)(mbp->mb);
+
+ *ret_val = FW_CMD_RETVAL_GET(ntohl(rsp->alloc_to_len16));
+ if (*ret_val == FW_SUCCESS) {
+ iq_params->physiqid = ntohs(rsp->physiqid);
+ iq_params->iqid = ntohs(rsp->iqid);
+ iq_params->fl0id = ntohs(rsp->fl0id);
+ iq_params->fl1id = ntohs(rsp->fl1id);
+ } else {
+ iq_params->physiqid = iq_params->iqid =
+ iq_params->fl0id = iq_params->fl1id = 0;
+ }
+} /* csio_mb_iq_alloc_write_rsp */
+
+/*
+ * csio_mb_iq_free - Initializes the mailbox for freeing a
+ * specified Ingress DMA Queue.
+ *
+ * @hw: The HW structure
+ * @mbp: Mailbox structure to initialize
+ * @priv: Private data
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @iq_params: Parameters of ingress queue, that is to be freed.
+ * @cbfn: The call-back function
+ *
+ *
+ */
+void
+csio_mb_iq_free(struct csio_hw *hw, struct csio_mb *mbp, void *priv,
+ uint32_t mb_tmo, struct csio_iq_params *iq_params,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_iq_cmd *cmdp = (struct fw_iq_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, priv, cbfn, 1);
+
+ cmdp->op_to_vfn = htonl(FW_CMD_OP(FW_IQ_CMD) |
+ FW_CMD_REQUEST | FW_CMD_EXEC |
+ FW_IQ_CMD_PFN(iq_params->pfn) |
+ FW_IQ_CMD_VFN(iq_params->vfn));
+ cmdp->alloc_to_len16 = htonl(FW_IQ_CMD_FREE |
+ FW_CMD_LEN16(sizeof(*cmdp) / 16));
+ cmdp->type_to_iqandstindex = htonl(FW_IQ_CMD_TYPE(iq_params->type));
+
+ cmdp->iqid = htons(iq_params->iqid);
+ cmdp->fl0id = htons(iq_params->fl0id);
+ cmdp->fl1id = htons(iq_params->fl1id);
+
+} /* csio_mb_iq_free */
+
+/*
+ * csio_mb_eq_ofld_alloc - Initializes the mailbox for allocating
+ * an offload-egress queue.
+ *
+ * @hw: The HW structure
+ * @mbp: Mailbox structure to initialize
+ * @priv: Private data
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @eq_ofld_params: (Offload) Egress queue paramters.
+ * @cbfn: The call-back function
+ *
+ *
+ */
+static void
+csio_mb_eq_ofld_alloc(struct csio_hw *hw, struct csio_mb *mbp, void *priv,
+ uint32_t mb_tmo, struct csio_eq_params *eq_ofld_params,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_eq_ofld_cmd *cmdp = (struct fw_eq_ofld_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, priv, cbfn, 1);
+ cmdp->op_to_vfn = htonl(FW_CMD_OP(FW_EQ_OFLD_CMD) |
+ FW_CMD_REQUEST | FW_CMD_EXEC |
+ FW_EQ_OFLD_CMD_PFN(eq_ofld_params->pfn) |
+ FW_EQ_OFLD_CMD_VFN(eq_ofld_params->vfn));
+ cmdp->alloc_to_len16 = htonl(FW_EQ_OFLD_CMD_ALLOC |
+ FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+} /* csio_mb_eq_ofld_alloc */
+
+/*
+ * csio_mb_eq_ofld_write - Initializes the mailbox for writing
+ * an alloacted offload-egress queue.
+ *
+ * @hw: The HW structure
+ * @mbp: Mailbox structure to initialize
+ * @priv: Private data
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @cascaded_req: TRUE - if this request is cascased with Eq-alloc request.
+ * @eq_ofld_params: (Offload) Egress queue paramters.
+ * @cbfn: The call-back function
+ *
+ *
+ * NOTE: We OR relevant bits with cmdp->XXX, instead of just equating,
+ * because this EQ write request can be cascaded with a previous
+ * EQ alloc request, and we dont want to over-write the bits set by
+ * that request. This logic will work even in a non-cascaded case, since the
+ * cmdp structure is zeroed out by CSIO_INIT_MBP.
+ */
+static void
+csio_mb_eq_ofld_write(struct csio_hw *hw, struct csio_mb *mbp, void *priv,
+ uint32_t mb_tmo, bool cascaded_req,
+ struct csio_eq_params *eq_ofld_params,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_eq_ofld_cmd *cmdp = (struct fw_eq_ofld_cmd *)(mbp->mb);
+
+ uint32_t eq_start_stop = (eq_ofld_params->eqstart) ?
+ FW_EQ_OFLD_CMD_EQSTART : FW_EQ_OFLD_CMD_EQSTOP;
+
+ /*
+ * If this EQ write is cascaded with EQ alloc request, do not
+ * re-initialize with 0's.
+ *
+ */
+ if (!cascaded_req)
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, priv, cbfn, 1);
+
+ cmdp->op_to_vfn |= htonl(FW_CMD_OP(FW_EQ_OFLD_CMD) |
+ FW_CMD_REQUEST | FW_CMD_WRITE |
+ FW_EQ_OFLD_CMD_PFN(eq_ofld_params->pfn) |
+ FW_EQ_OFLD_CMD_VFN(eq_ofld_params->vfn));
+ cmdp->alloc_to_len16 |= htonl(eq_start_stop |
+ FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+ cmdp->eqid_pkd |= htonl(FW_EQ_OFLD_CMD_EQID(eq_ofld_params->eqid));
+
+ cmdp->fetchszm_to_iqid |= htonl(
+ FW_EQ_OFLD_CMD_HOSTFCMODE(eq_ofld_params->hostfcmode) |
+ FW_EQ_OFLD_CMD_CPRIO(eq_ofld_params->cprio) |
+ FW_EQ_OFLD_CMD_PCIECHN(eq_ofld_params->pciechn) |
+ FW_EQ_OFLD_CMD_IQID(eq_ofld_params->iqid));
+
+ cmdp->dcaen_to_eqsize |= htonl(
+ FW_EQ_OFLD_CMD_DCAEN(eq_ofld_params->dcaen) |
+ FW_EQ_OFLD_CMD_DCACPU(eq_ofld_params->dcacpu) |
+ FW_EQ_OFLD_CMD_FBMIN(eq_ofld_params->fbmin) |
+ FW_EQ_OFLD_CMD_FBMAX(eq_ofld_params->fbmax) |
+ FW_EQ_OFLD_CMD_CIDXFTHRESHO(eq_ofld_params->cidxfthresho) |
+ FW_EQ_OFLD_CMD_CIDXFTHRESH(eq_ofld_params->cidxfthresh) |
+ FW_EQ_OFLD_CMD_EQSIZE(eq_ofld_params->eqsize));
+
+ cmdp->eqaddr |= cpu_to_be64(eq_ofld_params->eqaddr);
+
+} /* csio_mb_eq_ofld_write */
+
+/*
+ * csio_mb_eq_ofld_alloc_write - Initializes the mailbox for allocation
+ * writing into an Engress DMA Queue.
+ *
+ * @hw: The HW structure
+ * @mbp: Mailbox structure to initialize
+ * @priv: Private data.
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @eq_ofld_params: (Offload) Egress queue paramters.
+ * @cbfn: The call-back function
+ *
+ *
+ */
+void
+csio_mb_eq_ofld_alloc_write(struct csio_hw *hw, struct csio_mb *mbp,
+ void *priv, uint32_t mb_tmo,
+ struct csio_eq_params *eq_ofld_params,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ csio_mb_eq_ofld_alloc(hw, mbp, priv, mb_tmo, eq_ofld_params, cbfn);
+ csio_mb_eq_ofld_write(hw, mbp, priv, mb_tmo, CSIO_TRUE,
+ eq_ofld_params, cbfn);
+} /* csio_mb_eq_ofld_alloc_write */
+
+/*
+ * csio_mb_eq_ofld_alloc_write_rsp - Process the allocation
+ * & write egress DMA queue mailbox's response.
+ *
+ * @hw: The HW structure.
+ * @mbp: Mailbox structure to initialize.
+ * @retval: Firmware return value.
+ * @eq_ofld_params: (Offload) Egress queue paramters.
+ *
+ */
+void
+csio_mb_eq_ofld_alloc_write_rsp(struct csio_hw *hw,
+ struct csio_mb *mbp, enum fw_retval *ret_val,
+ struct csio_eq_params *eq_ofld_params)
+{
+ struct fw_eq_ofld_cmd *rsp = (struct fw_eq_ofld_cmd *)(mbp->mb);
+
+ *ret_val = FW_CMD_RETVAL_GET(ntohl(rsp->alloc_to_len16));
+
+ if (*ret_val == FW_SUCCESS) {
+ eq_ofld_params->eqid = FW_EQ_OFLD_CMD_EQID_GET(
+ ntohl(rsp->eqid_pkd));
+ eq_ofld_params->physeqid = FW_EQ_OFLD_CMD_PHYSEQID_GET(
+ ntohl(rsp->physeqid_pkd));
+ } else
+ eq_ofld_params->eqid = 0;
+
+} /* csio_mb_eq_ofld_alloc_write_rsp */
+
+/*
+ * csio_mb_eq_ofld_free - Initializes the mailbox for freeing a
+ * specified Engress DMA Queue.
+ *
+ * @hw: The HW structure
+ * @mbp: Mailbox structure to initialize
+ * @priv: Private data area.
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @eq_ofld_params: (Offload) Egress queue paramters, that is to be freed.
+ * @cbfn: The call-back function
+ *
+ *
+ */
+void
+csio_mb_eq_ofld_free(struct csio_hw *hw, struct csio_mb *mbp, void *priv,
+ uint32_t mb_tmo, struct csio_eq_params *eq_ofld_params,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_eq_ofld_cmd *cmdp = (struct fw_eq_ofld_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, priv, cbfn, 1);
+
+ cmdp->op_to_vfn = htonl(FW_CMD_OP(FW_EQ_OFLD_CMD) |
+ FW_CMD_REQUEST | FW_CMD_EXEC |
+ FW_EQ_OFLD_CMD_PFN(eq_ofld_params->pfn) |
+ FW_EQ_OFLD_CMD_VFN(eq_ofld_params->vfn));
+ cmdp->alloc_to_len16 = htonl(FW_EQ_OFLD_CMD_FREE |
+ FW_CMD_LEN16(sizeof(*cmdp) / 16));
+ cmdp->eqid_pkd = htonl(FW_EQ_OFLD_CMD_EQID(eq_ofld_params->eqid));
+
+} /* csio_mb_eq_ofld_free */
+
+/*
+ * csio_write_fcoe_link_cond_init_mb - Initialize Mailbox to write FCoE link
+ * condition.
+ *
+ * @ln: The Lnode structure
+ * @mbp: Mailbox structure to initialize
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @cbfn: The call back function.
+ *
+ *
+ */
+void
+csio_write_fcoe_link_cond_init_mb(struct csio_lnode *ln, struct csio_mb *mbp,
+ uint32_t mb_tmo, uint8_t port_id, uint32_t sub_opcode,
+ uint8_t cos, bool link_status, uint32_t fcfi,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_fcoe_link_cmd *cmdp =
+ (struct fw_fcoe_link_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, ln, cbfn, 1);
+
+ cmdp->op_to_portid = htonl((
+ FW_CMD_OP(FW_FCOE_LINK_CMD) |
+ FW_CMD_REQUEST |
+ FW_CMD_WRITE |
+ FW_FCOE_LINK_CMD_PORTID(port_id)));
+ cmdp->sub_opcode_fcfi = htonl(
+ FW_FCOE_LINK_CMD_SUB_OPCODE(sub_opcode) |
+ FW_FCOE_LINK_CMD_FCFI(fcfi));
+ cmdp->lstatus = link_status;
+ cmdp->retval_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+} /* csio_write_fcoe_link_cond_init_mb */
+
+/*
+ * csio_fcoe_read_res_info_init_mb - Initializes the mailbox for reading FCoE
+ * resource information(FW_GET_RES_INFO_CMD).
+ *
+ * @hw: The HW structure
+ * @mbp: Mailbox structure to initialize
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @cbfn: The call-back function
+ *
+ *
+ */
+void
+csio_fcoe_read_res_info_init_mb(struct csio_hw *hw, struct csio_mb *mbp,
+ uint32_t mb_tmo,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_fcoe_res_info_cmd *cmdp =
+ (struct fw_fcoe_res_info_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, hw, cbfn, 1);
+
+ cmdp->op_to_read = htonl((FW_CMD_OP(FW_FCOE_RES_INFO_CMD) |
+ FW_CMD_REQUEST |
+ FW_CMD_READ));
+
+ cmdp->retval_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+} /* csio_fcoe_read_res_info_init_mb */
+
+/*
+ * csio_fcoe_vnp_alloc_init_mb - Initializes the mailbox for allocating VNP
+ * in the firmware (FW_FCOE_VNP_CMD).
+ *
+ * @ln: The Lnode structure.
+ * @mbp: Mailbox structure to initialize.
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @fcfi: FCF Index.
+ * @vnpi: vnpi
+ * @iqid: iqid
+ * @vnport_wwnn: vnport WWNN
+ * @vnport_wwpn: vnport WWPN
+ * @cbfn: The call-back function.
+ *
+ *
+ */
+void
+csio_fcoe_vnp_alloc_init_mb(struct csio_lnode *ln, struct csio_mb *mbp,
+ uint32_t mb_tmo, uint32_t fcfi, uint32_t vnpi, uint16_t iqid,
+ uint8_t vnport_wwnn[8], uint8_t vnport_wwpn[8],
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_fcoe_vnp_cmd *cmdp =
+ (struct fw_fcoe_vnp_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, ln, cbfn, 1);
+
+ cmdp->op_to_fcfi = htonl((FW_CMD_OP(FW_FCOE_VNP_CMD) |
+ FW_CMD_REQUEST |
+ FW_CMD_EXEC |
+ FW_FCOE_VNP_CMD_FCFI(fcfi)));
+
+ cmdp->alloc_to_len16 = htonl(FW_FCOE_VNP_CMD_ALLOC |
+ FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+ cmdp->gen_wwn_to_vnpi = htonl(FW_FCOE_VNP_CMD_VNPI(vnpi));
+
+ cmdp->iqid = htons(iqid);
+
+ if (!wwn_to_u64(vnport_wwnn) && !wwn_to_u64(vnport_wwpn))
+ cmdp->gen_wwn_to_vnpi |= htonl(FW_FCOE_VNP_CMD_GEN_WWN);
+
+ if (vnport_wwnn)
+ memcpy(cmdp->vnport_wwnn, vnport_wwnn, 8);
+ if (vnport_wwpn)
+ memcpy(cmdp->vnport_wwpn, vnport_wwpn, 8);
+
+} /* csio_fcoe_vnp_alloc_init_mb */
+
+/*
+ * csio_fcoe_vnp_read_init_mb - Prepares VNP read cmd.
+ * @ln: The Lnode structure.
+ * @mbp: Mailbox structure to initialize.
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @fcfi: FCF Index.
+ * @vnpi: vnpi
+ * @cbfn: The call-back handler.
+ */
+void
+csio_fcoe_vnp_read_init_mb(struct csio_lnode *ln, struct csio_mb *mbp,
+ uint32_t mb_tmo, uint32_t fcfi, uint32_t vnpi,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_fcoe_vnp_cmd *cmdp =
+ (struct fw_fcoe_vnp_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, ln, cbfn, 1);
+ cmdp->op_to_fcfi = htonl(FW_CMD_OP(FW_FCOE_VNP_CMD) |
+ FW_CMD_REQUEST |
+ FW_CMD_READ |
+ FW_FCOE_VNP_CMD_FCFI(fcfi));
+ cmdp->alloc_to_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+ cmdp->gen_wwn_to_vnpi = htonl(FW_FCOE_VNP_CMD_VNPI(vnpi));
+}
+
+/*
+ * csio_fcoe_vnp_free_init_mb - Initializes the mailbox for freeing an
+ * alloacted VNP in the firmware (FW_FCOE_VNP_CMD).
+ *
+ * @ln: The Lnode structure.
+ * @mbp: Mailbox structure to initialize.
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @fcfi: FCF flow id
+ * @vnpi: VNP flow id
+ * @cbfn: The call-back function.
+ * Return: None
+ */
+void
+csio_fcoe_vnp_free_init_mb(struct csio_lnode *ln, struct csio_mb *mbp,
+ uint32_t mb_tmo, uint32_t fcfi, uint32_t vnpi,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_fcoe_vnp_cmd *cmdp =
+ (struct fw_fcoe_vnp_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, ln, cbfn, 1);
+
+ cmdp->op_to_fcfi = htonl(FW_CMD_OP(FW_FCOE_VNP_CMD) |
+ FW_CMD_REQUEST |
+ FW_CMD_EXEC |
+ FW_FCOE_VNP_CMD_FCFI(fcfi));
+ cmdp->alloc_to_len16 = htonl(FW_FCOE_VNP_CMD_FREE |
+ FW_CMD_LEN16(sizeof(*cmdp) / 16));
+ cmdp->gen_wwn_to_vnpi = htonl(FW_FCOE_VNP_CMD_VNPI(vnpi));
+}
+
+/*
+ * csio_fcoe_read_fcf_init_mb - Initializes the mailbox to read the
+ * FCF records.
+ *
+ * @ln: The Lnode structure
+ * @mbp: Mailbox structure to initialize
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @fcf_params: FC-Forwarder parameters.
+ * @cbfn: The call-back function
+ *
+ *
+ */
+void
+csio_fcoe_read_fcf_init_mb(struct csio_lnode *ln, struct csio_mb *mbp,
+ uint32_t mb_tmo, uint32_t portid, uint32_t fcfi,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_fcoe_fcf_cmd *cmdp =
+ (struct fw_fcoe_fcf_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, ln, cbfn, 1);
+
+ cmdp->op_to_fcfi = htonl(FW_CMD_OP(FW_FCOE_FCF_CMD) |
+ FW_CMD_REQUEST |
+ FW_CMD_READ |
+ FW_FCOE_FCF_CMD_FCFI(fcfi));
+ cmdp->retval_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+} /* csio_fcoe_read_fcf_init_mb */
+
+void
+csio_fcoe_read_portparams_init_mb(struct csio_hw *hw, struct csio_mb *mbp,
+ uint32_t mb_tmo,
+ struct fw_fcoe_port_cmd_params *portparams,
+ void (*cbfn)(struct csio_hw *,
+ struct csio_mb *))
+{
+ struct fw_fcoe_stats_cmd *cmdp = (struct fw_fcoe_stats_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, hw, cbfn, 1);
+ mbp->mb_size = 64;
+
+ cmdp->op_to_flowid = htonl(FW_CMD_OP(FW_FCOE_STATS_CMD) |
+ FW_CMD_REQUEST | FW_CMD_READ);
+ cmdp->free_to_len16 = htonl(FW_CMD_LEN16(CSIO_MAX_MB_SIZE/16));
+
+ cmdp->u.ctl.nstats_port = FW_FCOE_STATS_CMD_NSTATS(portparams->nstats) |
+ FW_FCOE_STATS_CMD_PORT(portparams->portid);
+
+ cmdp->u.ctl.port_valid_ix = FW_FCOE_STATS_CMD_IX(portparams->idx) |
+ FW_FCOE_STATS_CMD_PORT_VALID;
+
+} /* csio_fcoe_read_portparams_init_mb */
+
+void
+csio_mb_process_portparams_rsp(
+ struct csio_hw *hw,
+ struct csio_mb *mbp,
+ enum fw_retval *retval,
+ struct fw_fcoe_port_cmd_params *portparams,
+ struct fw_fcoe_port_stats *portstats
+ )
+{
+ struct fw_fcoe_stats_cmd *rsp = (struct fw_fcoe_stats_cmd *)(mbp->mb);
+ struct fw_fcoe_port_stats stats;
+ uint8_t *src;
+ uint8_t *dst;
+
+ *retval = FW_CMD_RETVAL_GET(ntohl(rsp->free_to_len16));
+
+ memset(&stats, 0, sizeof(struct fw_fcoe_port_stats));
+
+ if (*retval == FW_SUCCESS) {
+ dst = (uint8_t *)(&stats) + ((portparams->idx - 1) * 8);
+ src = (uint8_t *)rsp + (CSIO_STATS_OFFSET * 8);
+ memcpy(dst, src, (portparams->nstats * 8));
+ if (portparams->idx == 1) {
+ /* Get the first 6 flits from the Mailbox */
+ portstats->tx_bcast_bytes =
+ be64_to_cpu(stats.tx_bcast_bytes);
+ portstats->tx_bcast_frames =
+ be64_to_cpu(stats.tx_bcast_frames);
+ portstats->tx_mcast_bytes =
+ be64_to_cpu(stats.tx_mcast_bytes);
+ portstats->tx_mcast_frames =
+ be64_to_cpu(stats.tx_mcast_frames);
+ portstats->tx_ucast_bytes =
+ be64_to_cpu(stats.tx_ucast_bytes);
+ portstats->tx_ucast_frames =
+ be64_to_cpu(stats.tx_ucast_frames);
+ }
+ if (portparams->idx == 7) {
+ /* Get the second 6 flits from the Mailbox */
+ portstats->tx_drop_frames =
+ be64_to_cpu(stats.tx_drop_frames);
+ portstats->tx_offload_bytes =
+ be64_to_cpu(stats.tx_offload_bytes);
+ portstats->tx_offload_frames =
+ be64_to_cpu(stats.tx_offload_frames);
+#if 0
+ portstats->rx_pf_bytes =
+ be64_to_cpu(stats.rx_pf_bytes);
+ portstats->rx_pf_frames =
+ be64_to_cpu(stats.rx_pf_frames);
+#endif
+ portstats->rx_bcast_bytes =
+ be64_to_cpu(stats.rx_bcast_bytes);
+ portstats->rx_bcast_frames =
+ be64_to_cpu(stats.rx_bcast_frames);
+ portstats->rx_mcast_bytes =
+ be64_to_cpu(stats.rx_mcast_bytes);
+ }
+ if (portparams->idx == 13) {
+ /* Get the last 4 flits from the Mailbox */
+ portstats->rx_mcast_frames =
+ be64_to_cpu(stats.rx_mcast_frames);
+ portstats->rx_ucast_bytes =
+ be64_to_cpu(stats.rx_ucast_bytes);
+ portstats->rx_ucast_frames =
+ be64_to_cpu(stats.rx_ucast_frames);
+ portstats->rx_err_frames =
+ be64_to_cpu(stats.rx_err_frames);
+ }
+ }
+}
+
+/* Entry points/APIs for MB module */
+/*
+ * csio_mb_intr_enable - Enable Interrupts from mailboxes.
+ * @hw: The HW structure
+ *
+ * Enables CIM interrupt bit in appropriate INT_ENABLE registers.
+ */
+void
+csio_mb_intr_enable(struct csio_hw *hw)
+{
+ csio_wr_reg32(hw, MBMSGRDYINTEN(1), MYPF_REG(CIM_PF_HOST_INT_ENABLE));
+ csio_rd_reg32(hw, MYPF_REG(CIM_PF_HOST_INT_ENABLE));
+}
+
+/*
+ * csio_mb_intr_disable - Disable Interrupts from mailboxes.
+ * @hw: The HW structure
+ *
+ * Disable bit in HostInterruptEnable CIM register.
+ */
+void
+csio_mb_intr_disable(struct csio_hw *hw)
+{
+ csio_wr_reg32(hw, MBMSGRDYINTEN(0), MYPF_REG(CIM_PF_HOST_INT_ENABLE));
+ csio_rd_reg32(hw, MYPF_REG(CIM_PF_HOST_INT_ENABLE));
+}
+
+static void
+csio_mb_dump_fw_dbg(struct csio_hw *hw, __be64 *cmd)
+{
+ struct fw_debug_cmd *dbg = (struct fw_debug_cmd *)cmd;
+
+ if ((FW_DEBUG_CMD_TYPE_GET(ntohl(dbg->op_type))) == 1) {
+ csio_info(hw, "FW print message:\n");
+ csio_info(hw, "\tdebug->dprtstridx = %d\n",
+ ntohs(dbg->u.prt.dprtstridx));
+ csio_info(hw, "\tdebug->dprtstrparam0 = 0x%x\n",
+ ntohl(dbg->u.prt.dprtstrparam0));
+ csio_info(hw, "\tdebug->dprtstrparam1 = 0x%x\n",
+ ntohl(dbg->u.prt.dprtstrparam1));
+ csio_info(hw, "\tdebug->dprtstrparam2 = 0x%x\n",
+ ntohl(dbg->u.prt.dprtstrparam2));
+ csio_info(hw, "\tdebug->dprtstrparam3 = 0x%x\n",
+ ntohl(dbg->u.prt.dprtstrparam3));
+ } else {
+ /* This is a FW assertion */
+ csio_fatal(hw, "FW assertion at %.16s:%u, val0 %#x, val1 %#x\n",
+ dbg->u.assert.filename_0_7,
+ ntohl(dbg->u.assert.line),
+ ntohl(dbg->u.assert.x),
+ ntohl(dbg->u.assert.y));
+ }
+}
+
+static void
+csio_mb_debug_cmd_handler(struct csio_hw *hw)
+{
+ int i;
+ __be64 cmd[CSIO_MB_MAX_REGS];
+ uint32_t ctl_reg = PF_REG(hw->pfn, CIM_PF_MAILBOX_CTRL);
+ uint32_t data_reg = PF_REG(hw->pfn, CIM_PF_MAILBOX_DATA);
+ int size = sizeof(struct fw_debug_cmd);
+
+ /* Copy mailbox data */
+ for (i = 0; i < size; i += 8)
+ cmd[i / 8] = cpu_to_be64(csio_rd_reg64(hw, data_reg + i));
+
+ csio_mb_dump_fw_dbg(hw, cmd);
+
+ /* Notify FW of mailbox by setting owner as UP */
+ csio_wr_reg32(hw, MBMSGVALID | MBINTREQ | MBOWNER(CSIO_MBOWNER_FW),
+ ctl_reg);
+
+ csio_rd_reg32(hw, ctl_reg);
+ wmb();
+}
+
+/*
+ * csio_mb_issue - generic routine for issuing Mailbox commands.
+ * @hw: The HW structure
+ * @mbp: Mailbox command to issue
+ *
+ * Caller should hold hw lock across this call.
+ */
+csio_retval_t
+csio_mb_issue(struct csio_hw *hw, struct csio_mb *mbp)
+{
+ uint32_t owner, ctl;
+ int i;
+ uint32_t ii;
+ __be64 *cmd = mbp->mb;
+ __be64 hdr;
+ struct csio_mbm *mbm = &hw->mbm;
+ uint32_t ctl_reg = PF_REG(hw->pfn, CIM_PF_MAILBOX_CTRL);
+ uint32_t data_reg = PF_REG(hw->pfn, CIM_PF_MAILBOX_DATA);
+ int size = mbp->mb_size;
+ csio_retval_t rv = CSIO_INVAL;
+ struct fw_cmd_hdr *fw_hdr;
+
+ /* Determine mode */
+ if (mbp->mb_cbfn == NULL) {
+ /* Need to issue/get results in the same context */
+ if (mbp->tmo < CSIO_MB_POLL_FREQ) {
+ csio_err(hw, "Invalid tmo: 0x%x\n", mbp->tmo);
+ goto error_out;
+ }
+ } else if (!csio_is_host_intr_enabled(hw) ||
+ !csio_is_hw_intr_enabled(hw)) {
+ csio_err(hw, "Cannot issue mailbox in interrupt mode 0x%x\n",
+ *((uint8_t *)mbp->mb));
+ goto error_out;
+ }
+
+ if (mbm->mcurrent != NULL) {
+ /* Queue mbox cmd, if another mbox cmd is active */
+ if (mbp->mb_cbfn == NULL) {
+ rv = CSIO_BUSY;
+ csio_dbg(hw, "Couldnt own Mailbox %x op:0x%x\n",
+ hw->pfn, *((uint8_t *)mbp->mb));
+
+ goto error_out;
+ } else {
+ list_add_tail(&mbp->list, &mbm->req_q);
+ csio_inc_stats(mbm, n_activeq);
+
+ return CSIO_SUCCESS;
+ }
+ }
+
+ /* Now get ownership of mailbox */
+ owner = MBOWNER_GET(csio_rd_reg32(hw, ctl_reg));
+
+ if (!csio_mb_is_host_owner(owner)) {
+
+ for (i = 0; (owner == CSIO_MBOWNER_NONE) && (i < 3); i++)
+ owner = MBOWNER_GET(csio_rd_reg32(hw, ctl_reg));
+ /*
+ * Mailbox unavailable. In immediate mode, fail the command.
+ * In other modes, enqueue the request.
+ */
+ if (!csio_mb_is_host_owner(owner)) {
+ if (mbp->mb_cbfn == NULL) {
+ rv = owner ? CSIO_BUSY : CSIO_TIMEOUT;
+
+ csio_dbg(hw,
+ "Couldnt own Mailbox %x op:0x%x "
+ "owner:%x\n",
+ hw->pfn, *((uint8_t *)mbp->mb), owner);
+ goto error_out;
+ } else {
+ if (mbm->mcurrent == NULL) {
+ csio_err(hw,
+ "Couldnt own Mailbox %x "
+ "op:0x%x owner:%x\n",
+ hw->pfn, *((uint8_t *)mbp->mb),
+ owner);
+ csio_err(hw,
+ "No outstanding driver"
+ " mailbox as well\n");
+ goto error_out;
+ }
+ }
+ }
+ }
+
+ /* Mailbox is available, copy mailbox data into it */
+ for (i = 0; i < size; i += 8) {
+ csio_wr_reg64(hw, be64_to_cpu(*cmd), data_reg + i);
+ cmd++;
+ }
+
+ CSIO_DUMP_MB(hw, hw->pfn, data_reg);
+
+ /* Start completion timers in non-immediate modes and notify FW */
+ if (mbp->mb_cbfn != NULL) {
+ mbm->mcurrent = mbp;
+ mod_timer(&mbm->timer, jiffies + msecs_to_jiffies(mbp->tmo));
+ csio_wr_reg32(hw, MBMSGVALID | MBINTREQ |
+ MBOWNER(CSIO_MBOWNER_FW), ctl_reg);
+ } else
+ csio_wr_reg32(hw, MBMSGVALID | MBOWNER(CSIO_MBOWNER_FW),
+ ctl_reg);
+
+ /* Flush posted writes */
+ csio_rd_reg32(hw, ctl_reg);
+ wmb();
+
+ csio_inc_stats(mbm, n_req);
+
+ if (mbp->mb_cbfn)
+ return CSIO_SUCCESS;
+
+ /* Poll for completion in immediate mode */
+ cmd = mbp->mb;
+
+ for (ii = 0; ii < mbp->tmo; ii += CSIO_MB_POLL_FREQ) {
+ mdelay(CSIO_MB_POLL_FREQ);
+
+ /* Check for response */
+ ctl = csio_rd_reg32(hw, ctl_reg);
+ if (csio_mb_is_host_owner(MBOWNER_GET(ctl))) {
+
+ if (!(ctl & MBMSGVALID)) {
+ csio_wr_reg32(hw, 0, ctl_reg);
+ continue;
+ }
+
+ CSIO_DUMP_MB(hw, hw->pfn, data_reg);
+
+ hdr = cpu_to_be64(csio_rd_reg64(hw, data_reg));
+ fw_hdr = (struct fw_cmd_hdr *)&hdr;
+
+ switch (FW_CMD_OP_GET(ntohl(fw_hdr->hi))) {
+ case FW_DEBUG_CMD:
+ csio_mb_debug_cmd_handler(hw);
+ continue;
+ }
+
+ /* Copy response */
+ for (i = 0; i < size; i += 8)
+ *cmd++ = cpu_to_be64(csio_rd_reg64
+ (hw, data_reg + i));
+ csio_wr_reg32(hw, 0, ctl_reg);
+
+ if (FW_CMD_RETVAL_GET(*(mbp->mb)))
+ csio_inc_stats(mbm, n_err);
+
+ csio_inc_stats(mbm, n_rsp);
+ return CSIO_SUCCESS;
+ }
+ }
+
+ csio_inc_stats(mbm, n_tmo);
+
+ csio_err(hw, "Mailbox %x op:0x%x timed out!\n",
+ hw->pfn, *((uint8_t *)cmd));
+
+ return CSIO_TIMEOUT;
+
+error_out:
+ csio_inc_stats(mbm, n_err);
+ return rv;
+}
+
+/*
+ * csio_mb_completions - Completion handler for Mailbox commands
+ * @hw: The HW structure
+ * @cbfn_q: Completion queue.
+ *
+ */
+void
+csio_mb_completions(struct csio_hw *hw, struct list_head *cbfn_q)
+{
+ struct csio_mb *mbp;
+ struct csio_mbm *mbm = &hw->mbm;
+ enum fw_retval rv;
+
+ while (!list_empty(cbfn_q)) {
+ csio_deq_from_head(cbfn_q, &mbp);
+
+ rv = csio_mb_fw_retval(mbp);
+ if ((rv != FW_SUCCESS) && (rv != FW_HOSTERROR))
+ csio_inc_stats(mbm, n_err);
+ else if (rv != FW_HOSTERROR)
+ csio_inc_stats(mbm, n_rsp);
+
+ if (mbp->mb_cbfn)
+ mbp->mb_cbfn(hw, mbp);
+ }
+}
+
+static void
+csio_mb_portmod_changed(struct csio_hw *hw, uint8_t port_id)
+{
+ static char *mod_str[] = {
+ NULL, "LR", "SR", "ER", "TWINAX", "active TWINAX", "LRM"
+ };
+
+ struct csio_pport *port = &hw->pport[port_id];
+
+ if (port->mod_type == FW_PORT_MOD_TYPE_NONE)
+ csio_info(hw, "Port:%d - port module unplugged\n", port_id);
+ else if (port->mod_type < ARRAY_SIZE(mod_str))
+ csio_info(hw, "Port:%d - %s port module inserted\n", port_id,
+ mod_str[port->mod_type]);
+ else if (port->mod_type == FW_PORT_MOD_TYPE_NOTSUPPORTED)
+ csio_info(hw,
+ "Port:%d - unsupported optical port module "
+ "inserted\n", port_id);
+ else if (port->mod_type == FW_PORT_MOD_TYPE_UNKNOWN)
+ csio_info(hw,
+ "Port:%d - unknown port module inserted, forcing "
+ "TWINAX\n", port_id);
+ else if (port->mod_type == FW_PORT_MOD_TYPE_ERROR)
+ csio_info(hw, "Port:%d - transceiver module error\n", port_id);
+ else
+ csio_info(hw, "Port:%d - unknown module type %d inserted\n",
+ port_id, port->mod_type);
+}
+
+csio_retval_t
+csio_mb_fwevt_handler(struct csio_hw *hw, __be64 *cmd)
+{
+ uint8_t opcode = *(uint8_t *)cmd;
+ struct fw_port_cmd *pcmd;
+ uint8_t port_id;
+ uint32_t link_status;
+ uint16_t action;
+ uint8_t mod_type;
+
+ if (opcode == FW_PORT_CMD) {
+ pcmd = (struct fw_port_cmd *)cmd;
+ port_id = FW_PORT_CMD_PORTID_GET(
+ ntohl(pcmd->op_to_portid));
+ action = FW_PORT_CMD_ACTION_GET(
+ ntohl(pcmd->action_to_len16));
+ if (action != FW_PORT_ACTION_GET_PORT_INFO) {
+ csio_err(hw, "Unhandled FW_PORT_CMD action: %u\n",
+ action);
+ return CSIO_INVAL;
+ }
+
+ link_status = ntohl(pcmd->u.info.lstatus_to_modtype);
+ mod_type = FW_PORT_CMD_MODTYPE_GET(link_status);
+
+ hw->pport[port_id].link_status =
+ FW_PORT_CMD_LSTATUS_GET(link_status);
+ hw->pport[port_id].link_speed =
+ FW_PORT_CMD_LSPEED_GET(link_status);
+
+ csio_info(hw, "Port:%x - LINK %s\n", port_id,
+ FW_PORT_CMD_LSTATUS_GET(link_status) ? "UP" : "DOWN");
+
+ if (mod_type != hw->pport[port_id].mod_type) {
+ hw->pport[port_id].mod_type = mod_type;
+ csio_mb_portmod_changed(hw, port_id);
+ }
+ } else if (opcode == FW_DEBUG_CMD) {
+ csio_mb_dump_fw_dbg(hw, cmd);
+ } else {
+ csio_dbg(hw, "Gen MB can't handle op:0x%x on evtq.\n", opcode);
+ return CSIO_INVAL;
+ }
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_mb_isr_handler - Handle mailboxes related interrupts.
+ * @hw: The HW structure
+ *
+ * Called from the ISR to handle Mailbox related interrupts.
+ * HW Lock should be held across this call.
+ */
+csio_retval_t
+csio_mb_isr_handler(struct csio_hw *hw)
+{
+ struct csio_mbm *mbm = &hw->mbm;
+ struct csio_mb *mbp = mbm->mcurrent;
+ __be64 *cmd;
+ uint32_t ctl, cim_cause, pl_cause;
+ int i;
+ uint32_t ctl_reg = PF_REG(hw->pfn, CIM_PF_MAILBOX_CTRL);
+ uint32_t data_reg = PF_REG(hw->pfn, CIM_PF_MAILBOX_DATA);
+ int size;
+ __be64 hdr;
+ struct fw_cmd_hdr *fw_hdr;
+
+ pl_cause = csio_rd_reg32(hw, MYPF_REG(PL_PF_INT_CAUSE));
+ cim_cause = csio_rd_reg32(hw, MYPF_REG(CIM_PF_HOST_INT_CAUSE));
+
+ if (!(pl_cause & PFCIM) || !(cim_cause & MBMSGRDYINT)) {
+ csio_inc_stats(hw, n_mbint_unexp);
+ return CSIO_INVAL;
+ }
+
+ /*
+ * The cause registers below HAVE to be cleared in the SAME
+ * order as below: The low level cause register followed by
+ * the upper level cause register. In other words, CIM-cause
+ * first followed by PL-Cause next.
+ */
+ csio_wr_reg32(hw, MBMSGRDYINT, MYPF_REG(CIM_PF_HOST_INT_CAUSE));
+ csio_wr_reg32(hw, PFCIM, MYPF_REG(PL_PF_INT_CAUSE));
+
+ ctl = csio_rd_reg32(hw, ctl_reg);
+
+ if (csio_mb_is_host_owner(MBOWNER_GET(ctl))) {
+
+ CSIO_DUMP_MB(hw, hw->pfn, data_reg);
+
+ if (!(ctl & MBMSGVALID)) {
+ csio_warn(hw,
+ "Stray mailbox interrupt recvd,"
+ " mailbox data not valid\n");
+ csio_wr_reg32(hw, 0, ctl_reg);
+ /* Flush */
+ csio_rd_reg32(hw, ctl_reg);
+ return CSIO_INVAL;
+ }
+
+ hdr = cpu_to_be64(csio_rd_reg64(hw, data_reg));
+ fw_hdr = (struct fw_cmd_hdr *)&hdr;
+
+ switch (FW_CMD_OP_GET(ntohl(fw_hdr->hi))) {
+ case FW_DEBUG_CMD:
+ csio_mb_debug_cmd_handler(hw);
+ return CSIO_INVAL;
+#if 0
+ case FW_ERROR_CMD:
+ case FW_INITIALIZE_CMD: /* When we are not master */
+#endif
+ }
+
+ CSIO_ASSERT(mbp != NULL);
+
+ cmd = mbp->mb;
+ size = mbp->mb_size;
+ /* Get response */
+ for (i = 0; i < size; i += 8)
+ *cmd++ = cpu_to_be64(csio_rd_reg64
+ (hw, data_reg + i));
+
+ csio_wr_reg32(hw, 0, ctl_reg);
+ /* Flush */
+ csio_rd_reg32(hw, ctl_reg);
+
+ mbm->mcurrent = NULL;
+
+ /* Add completion to tail of cbfn queue */
+ list_add_tail(&mbp->list, &mbm->cbfn_q);
+ csio_inc_stats(mbm, n_cbfnq);
+
+ /*
+ * Enqueue event to EventQ. Events processing happens
+ * in Event worker thread context
+ */
+ if (csio_enqueue_evt(hw, CSIO_EVT_MBX, mbp, sizeof(mbp)))
+ csio_inc_stats(hw, n_evt_drop);
+
+ return CSIO_SUCCESS;
+
+ } else {
+ /*
+ * We can get here if mailbox MSIX vector is shared,
+ * or in INTx case. Or a stray interrupt.
+ */
+ csio_dbg(hw, "Host not owner, no mailbox interrupt\n");
+ csio_inc_stats(hw, n_int_stray);
+ return CSIO_INVAL;
+ }
+}
+
+/*
+ * csio_mb_tmo_handler - Timeout handler
+ * @hw: The HW structure
+ *
+ */
+struct csio_mb *
+csio_mb_tmo_handler(struct csio_hw *hw)
+{
+ struct csio_mbm *mbm = &hw->mbm;
+ struct csio_mb *mbp = mbm->mcurrent;
+ struct fw_cmd_hdr *fw_hdr;
+
+ /*
+ * Could be a race b/w the completion handler and the timer
+ * and the completion handler won that race.
+ */
+ if (mbp == NULL) {
+ CSIO_DB_ASSERT(0);
+ return NULL;
+ }
+
+ fw_hdr = (struct fw_cmd_hdr *)(mbp->mb);
+
+ csio_dbg(hw, "Mailbox num:%x op:0x%x timed out\n", hw->pfn,
+ FW_CMD_OP_GET(ntohl(fw_hdr->hi)));
+
+ mbm->mcurrent = NULL;
+ csio_inc_stats(mbm, n_tmo);
+ fw_hdr->lo = htonl(FW_CMD_RETVAL(FW_ETIMEDOUT));
+
+ return mbp;
+}
+
+/*
+ * csio_mb_cancel_all - Cancel all waiting commands.
+ * @hw: The HW structure
+ * @cbfn_q: The callback queue.
+ *
+ * Caller should hold hw lock across this call.
+ */
+void
+csio_mb_cancel_all(struct csio_hw *hw, struct list_head *cbfn_q)
+{
+ struct csio_mb *mbp;
+ struct csio_mbm *mbm = &hw->mbm;
+ struct fw_cmd_hdr *hdr;
+ struct list_head *tmp;
+
+ if (mbm->mcurrent) {
+ mbp = mbm->mcurrent;
+
+ /* Stop mailbox completion timer */
+ del_timer_sync(&mbm->timer);
+
+ /* Add completion to tail of cbfn queue */
+ list_add_tail(&mbp->list, cbfn_q);
+ mbm->mcurrent = NULL;
+ }
+
+ if (!list_empty(&mbm->req_q)) {
+ list_splice_tail_init(&mbm->req_q, cbfn_q);
+ mbm->stats.n_activeq = 0;
+ }
+
+ if (!list_empty(&mbm->cbfn_q)) {
+ list_splice_tail_init(&mbm->cbfn_q, cbfn_q);
+ mbm->stats.n_cbfnq = 0;
+ }
+
+ if (list_empty(cbfn_q))
+ return;
+
+ list_for_each(tmp, cbfn_q) {
+ mbp = (struct csio_mb *)tmp;
+ hdr = (struct fw_cmd_hdr *)(mbp->mb);
+
+ csio_dbg(hw, "Cancelling pending mailbox num %x op:%x\n",
+ hw->pfn, FW_CMD_OP_GET(ntohl(hdr->hi)));
+
+ csio_inc_stats(mbm, n_cancel);
+ hdr->lo = htonl(FW_CMD_RETVAL(FW_HOSTERROR));
+ }
+}
+
+/*
+ * csio_mbm_init - Initialize Mailbox module
+ * @mbm: Mailbox module
+ * @hw: The HW structure
+ * @timer: Timing function for interrupting mailboxes
+ *
+ * Initialize timer and the request/response queues.
+ */
+csio_retval_t
+csio_mbm_init(struct csio_mbm *mbm, struct csio_hw *hw,
+ void (*timer_fn)(uintptr_t))
+{
+ struct timer_list *timer = &mbm->timer;
+
+ init_timer(timer);
+ timer->function = timer_fn;
+ timer->data = (unsigned long)hw;
+
+ INIT_LIST_HEAD(&mbm->req_q);
+ INIT_LIST_HEAD(&mbm->cbfn_q);
+ csio_set_mb_intr_idx(mbm, -1);
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_mbm_exit - Uninitialize mailbox module
+ * @mbm: Mailbox module
+ *
+ * Stop timer.
+ */
+void
+csio_mbm_exit(struct csio_mbm *mbm)
+{
+ del_timer_sync(&mbm->timer);
+
+ CSIO_DB_ASSERT(mbm->mcurrent == NULL);
+ CSIO_DB_ASSERT(list_empty(&mbm->req_q));
+ CSIO_DB_ASSERT(list_empty(&mbm->cbfn_q));
+}
--
1.7.1
^ permalink raw reply related
* [PATCH 4/8] csiostor: Chelsio FCoE offload driver submission (sources part 4).
From: Naresh Kumar Inna @ 2012-08-23 22:27 UTC (permalink / raw)
To: JBottomley, linux-scsi, dm; +Cc: netdev, naresh, chethan
In-Reply-To: <1345760873-12101-1-git-send-email-naresh@chelsio.com>
This patch contains code to implement the local and remote node port
functionality. It includes tracking the firmware events for changes to
the states of these ports.
Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
---
drivers/scsi/csiostor/csio_lnode.c | 2151 ++++++++++++++++++++++++++++++++++++
drivers/scsi/csiostor/csio_rnode.c | 889 +++++++++++++++
2 files changed, 3040 insertions(+), 0 deletions(-)
create mode 100644 drivers/scsi/csiostor/csio_lnode.c
create mode 100644 drivers/scsi/csiostor/csio_rnode.c
diff --git a/drivers/scsi/csiostor/csio_lnode.c b/drivers/scsi/csiostor/csio_lnode.c
new file mode 100644
index 0000000..5d5a1a1
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_lnode.c
@@ -0,0 +1,2151 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/kernel.h>
+#include <linux/delay.h>
+#include <linux/slab.h>
+#include <linux/utsname.h>
+#include <scsi/scsi_transport_fc.h>
+
+#include "csio_hw.h"
+#include "csio_mb.h"
+#include "csio_lnode.h"
+#include "csio_rnode.h"
+#include "csio_fcoe_proto.h"
+
+int csio_fcoe_rnodes = 512;
+int csio_fdmi_enable = 1;
+
+#define PORT_ID_PTR(_x) ((uint8_t *)(&_x) + 1)
+
+/* Lnode SM declarations */
+static void csio_lns_uninit(struct csio_lnode *, enum csio_ln_ev);
+static void csio_lns_online(struct csio_lnode *, enum csio_ln_ev);
+static void csio_lns_ready(struct csio_lnode *, enum csio_ln_ev);
+static void csio_lns_offline(struct csio_lnode *, enum csio_ln_ev);
+
+static csio_retval_t csio_ln_mgmt_submit_req(struct csio_ioreq *,
+ void (*io_cbfn) (struct csio_hw *, struct csio_ioreq *),
+ enum fcoe_cmn_type, struct csio_dma_buf *, uint32_t);
+
+/* LN event mapping */
+static enum csio_ln_ev fwevt_to_lnevt[] = {
+ CSIO_LNE_NONE, /* None */
+ CSIO_LNE_NONE, /* PLOGI_ACC_RCVD */
+ CSIO_LNE_NONE, /* PLOGI_RJT_RCVD */
+ CSIO_LNE_NONE, /* PLOGI_RCVD */
+ CSIO_LNE_NONE, /* PLOGO_RCVD */
+ CSIO_LNE_NONE, /* PRLI_ACC_RCVD */
+ CSIO_LNE_NONE, /* PRLI_RJT_RCVD */
+ CSIO_LNE_NONE, /* PRLI_RCVD */
+ CSIO_LNE_NONE, /* PRLO_RCVD */
+ CSIO_LNE_NONE, /* NPORT_ID_CHGD */
+ CSIO_LNE_LOGO, /* FLOGO_RCVD */
+ CSIO_LNE_LOGO, /* CLR_VIRT_LNK_RCVD */
+ CSIO_LNE_FAB_INIT_DONE,/* FLOGI_ACC_RCVD */
+ CSIO_LNE_NONE, /* FLOGI_RJT_RCVD */
+ CSIO_LNE_FAB_INIT_DONE,/* FDISC_ACC_RCVD */
+ CSIO_LNE_NONE, /* FDISC_RJT_RCVD */
+ CSIO_LNE_NONE, /* FLOGI_TMO_MAX_RETRY */
+ CSIO_LNE_NONE, /* IMPL_LOGO_ADISC_ACC */
+ CSIO_LNE_NONE, /* IMPL_LOGO_ADISC_RJT */
+ CSIO_LNE_NONE, /* IMPL_LOGO_ADISC_CNFLT */
+ CSIO_LNE_NONE, /* PRLI_TMO */
+ CSIO_LNE_NONE, /* ADISC_TMO */
+ CSIO_LNE_NONE, /* RSCN_DEV_LOST */
+ CSIO_LNE_NONE, /* SCR_ACC_RCVD */
+ CSIO_LNE_NONE, /* ADISC_RJT_RCVD */
+ CSIO_LNE_NONE, /* LOGO_SNT */
+ CSIO_LNE_NONE, /* PROTO_ERR_IMPL_LOGO */
+};
+
+#define CSIO_FWE_TO_LNE(_evt) ((_evt > PROTO_ERR_IMPL_LOGO) ? \
+ CSIO_LNE_NONE : \
+ fwevt_to_lnevt[_evt])
+/*
+ * csio_ln_match_by_portid - lookup lnode using given portid.
+ * @hw: HW module
+ * @portid: port-id.
+ *
+ * If found, returns lnode matching given portid otherwise returns NULL.
+ */
+static struct csio_lnode *
+csio_ln_lookup_by_portid(struct csio_hw *hw, uint8_t portid)
+{
+ struct csio_lnode *ln = hw->rln;
+ struct list_head *tmp;
+
+ /* Match siblings lnode with portid */
+ list_for_each(tmp, &hw->sln_head) {
+ ln = (struct csio_lnode *) tmp;
+ if (ln->portid == portid)
+ return ln;
+ }
+
+ return NULL;
+}
+
+/*
+ * csio_ln_lookup_by_vnpi - Lookup lnode using given vnp id.
+ * @hw - HW module
+ * @vnpi - vnp index.
+ * Returns - If found, returns lnode matching given vnp id
+ * otherwise returns NULL.
+ */
+static struct csio_lnode *
+csio_ln_lookup_by_vnpi(struct csio_hw *hw, uint32_t vnp_id)
+{
+ struct list_head *tmp1, *tmp2;
+ struct csio_lnode *sln = NULL, *cln = NULL;
+
+ if (list_empty(&hw->sln_head)) {
+ csio_inc_stats(hw, n_lnlkup_miss);
+ return NULL;
+ }
+ /* Traverse sibling lnodes */
+ list_for_each(tmp1, &hw->sln_head) {
+ sln = (struct csio_lnode *) tmp1;
+
+ /* Match sibling lnode */
+ if (sln->vnp_flowid == vnp_id)
+ return sln;
+
+ if (list_empty(&sln->cln_head))
+ continue;
+
+ /* Traverse children lnodes */
+ list_for_each(tmp2, &sln->cln_head) {
+ cln = (struct csio_lnode *) tmp2;
+
+ if (cln->vnp_flowid == vnp_id)
+ return cln;
+ }
+ }
+ csio_inc_stats(hw, n_lnlkup_miss);
+ return NULL;
+}
+
+/**
+ * csio_lnode_lookup_by_wwpn - Lookup lnode using given wwpn.
+ * @hw: HW module.
+ * @wwpn: WWPN.
+ *
+ * If found, returns lnode matching given wwpn, returns NULL otherwise.
+ */
+struct csio_lnode *
+csio_lnode_lookup_by_wwpn(struct csio_hw *hw, uint8_t *wwpn)
+{
+ struct list_head *tmp1, *tmp2;
+ struct csio_lnode *sln = NULL, *cln = NULL;
+
+ if (list_empty(&hw->sln_head)) {
+ csio_inc_stats(hw, n_lnlkup_miss);
+ return NULL;
+ }
+ /* Traverse sibling lnodes */
+ list_for_each(tmp1, &hw->sln_head) {
+ sln = (struct csio_lnode *) tmp1;
+
+ /* Match sibling lnode */
+ if (!memcmp(csio_ln_wwpn(sln), wwpn, 8))
+ return sln;
+
+ if (list_empty(&sln->cln_head))
+ continue;
+
+ /* Traverse children lnodes */
+ list_for_each(tmp2, &sln->cln_head) {
+ cln = (struct csio_lnode *) tmp2;
+
+ if (!memcmp(csio_ln_wwpn(cln), wwpn, 8))
+ return cln;
+ }
+ }
+ return NULL;
+}
+
+static int
+csio_hostname(uint8_t *buf, size_t buf_len)
+{
+ if (sprintf(buf, "%s", init_utsname()->nodename))
+ return 0;
+ return -1;
+}
+
+static int
+csio_osname(uint8_t *buf, size_t buf_len)
+{
+ uint8_t *ptr = buf;
+
+ strcpy(ptr, init_utsname()->sysname);
+ ptr += strlen(init_utsname()->sysname);
+ *ptr = ' '; /* SPACE */
+ strcpy(ptr, init_utsname()->release);
+ ptr += strlen(init_utsname()->release);
+ *ptr = ' '; /* SPACE */
+ strcpy(ptr, init_utsname()->version);
+ ptr += strlen(init_utsname()->version);
+ *ptr = '\0';
+ return 0;
+}
+
+static inline void
+csio_append_string_attrib(uint8_t **ptr, uint16_t type,
+ uint8_t *val, uint16_t len)
+{
+ struct csio_attrib_entry *ae = (struct csio_attrib_entry *) *ptr;
+ ae->type = htons(type);
+ len += 4; /* includes attribute type and length */
+ len = (len + 3) & ~3; /* should be multiple of 4 bytes */
+ ae->len = htons(len);
+ memset(ae->val.string, 0, len - 4);
+ memcpy(ae->val.string, val, len);
+ *ptr += len;
+}
+
+static inline void
+csio_append_int_attrib(uint8_t **ptr, uint16_t type,
+ uint32_t val, uint16_t len)
+{
+ struct csio_attrib_entry *ae = (struct csio_attrib_entry *) *ptr;
+ ae->type = htons(type);
+ len += 4; /* includes attribute type and length */
+ len = (len + 3) & ~3; /* should be multiple of 4 bytes */
+ ae->len = htons(len);
+ ae->val.integer = htonl(val);
+ *ptr += len;
+}
+
+/*
+ * csio_ln_fdmi_done - FDMI registeration completion
+ * @hw: HW context
+ * @fdmi_req: fdmi request
+ */
+static void
+csio_ln_fdmi_done(struct csio_hw *hw, struct csio_ioreq *fdmi_req)
+{
+ void *cmd;
+ struct csio_lnode *ln = fdmi_req->lnode;
+
+ if (fdmi_req->wr_status != FW_SUCCESS) {
+ csio_ln_err(ln, "WR error:%x in processing fdmi rpa cmd\n",
+ fdmi_req->wr_status);
+ csio_inc_stats(ln, n_fdmi_err);
+ }
+
+ cmd = fdmi_req->dma_buf.vaddr;
+ if (ntohs(csio_ct_rsp(cmd)) != CT_RESPONSE_FS_ACC) {
+ csio_ln_dbg(ln, "fdmi rpa cmd rejected reason %x expl %x\n",
+ csio_ct_reason(cmd), csio_ct_expl(cmd));
+ }
+}
+
+/*
+ * csio_ln_fdmi_rhba_cbfn - RHBA completion
+ * @hw: HW context
+ * @fdmi_req: fdmi request
+ */
+static void
+csio_ln_fdmi_rhba_cbfn(struct csio_hw *hw, struct csio_ioreq *fdmi_req)
+{
+ void *cmd;
+ uint8_t *pld;
+ uint32_t len = 0;
+ struct csio_lnode *ln = fdmi_req->lnode;
+ struct csio_attrib_block *attrib_blk;
+ uint8_t buf[64];
+ uint32_t val;
+ uint8_t *fc4_type;
+
+ if (fdmi_req->wr_status != FW_SUCCESS) {
+ csio_ln_err(ln, "WR error:%x in processing fdmi rhba cmd\n",
+ fdmi_req->wr_status);
+ csio_inc_stats(ln, n_fdmi_err);
+ }
+
+ cmd = fdmi_req->dma_buf.vaddr;
+ if (ntohs(csio_ct_rsp(cmd)) != CT_RESPONSE_FS_ACC) {
+ csio_ln_dbg(ln, "fdmi rhba cmd rejected reason %x expl %x\n",
+ csio_ct_reason(cmd), csio_ct_expl(cmd));
+ }
+
+ if (!csio_is_rnode_ready(fdmi_req->rnode)) {
+ csio_inc_stats(ln, n_fdmi_err);
+ return;
+ }
+
+ /* Prepare CT hdr for RPA cmd */
+ memset(cmd, 0, CT_BASIC_IU_LEN);
+ csio_fill_ct_iu(cmd,
+ CT_GS_MGMT_SERVICE,
+ CT_FDMI_HBA_MGMT_SERVER,
+ htons(CT_FDMI_HBA_RPA));
+
+ /* Prepare RPA payload */
+ pld = (uint8_t *) csio_ct_get_pld(cmd);
+ memcpy(pld, csio_ln_wwpn(ln), 8); /* Port name */
+ pld += 8;
+
+ /* Start appending Port attributes */
+ attrib_blk = (struct csio_attrib_block *) pld;
+ attrib_blk->entry_count = 0;
+ len += sizeof(attrib_blk->entry_count);
+ pld += sizeof(attrib_blk->entry_count);
+
+ fc4_type = &buf[0];
+ memset(fc4_type, 0, 32);
+ fc4_type[2] = 1;
+ fc4_type[7] = 1;
+ csio_append_string_attrib(&pld, SUPPORTED_FC4_TYPES, fc4_type , 32);
+ attrib_blk->entry_count++;
+ val = CSIO_HBA_PORTSPEED_1GBIT | CSIO_HBA_PORTSPEED_10GBIT;
+ csio_append_int_attrib(&pld, SUPPORTED_SPEED, val , 4);
+ attrib_blk->entry_count++;
+
+ if (hw->pport[ln->portid].link_speed == FW_PORT_CAP_SPEED_1G)
+ val = CSIO_HBA_PORTSPEED_1GBIT;
+ else if (hw->pport[ln->portid].link_speed ==
+ FW_PORT_CAP_SPEED_10G)
+ val = CSIO_HBA_PORTSPEED_10GBIT;
+ else
+ val = CSIO_HBA_PORTSPEED_UNKNOWN;
+ csio_append_int_attrib(&pld, PORT_SPEED, val , 4);
+ attrib_blk->entry_count++;
+ csio_append_int_attrib(&pld, MAX_FRAME_LEN, ln->ln_sparm.csp.rcv_sz,
+ 4);
+ attrib_blk->entry_count++;
+
+ strcpy(buf, "csiostor");
+ csio_append_string_attrib(&pld, OS_DEVICE_NAME, buf,
+ (uint16_t)strlen(buf));
+ attrib_blk->entry_count++;
+
+ if (!csio_hostname(buf, sizeof(buf))) {
+ csio_append_string_attrib(&pld, HOST_NAME , buf,
+ (uint16_t)strlen(buf));
+ attrib_blk->entry_count++;
+ }
+ attrib_blk->entry_count = ntohl(attrib_blk->entry_count);
+ len = (uint32_t)(pld - (uint8_t *)cmd);
+
+ /* Submit FDMI RPA request */
+ spin_lock_irq(&hw->lock);
+ if (csio_ln_mgmt_submit_req(fdmi_req, csio_ln_fdmi_done,
+ FCOE_CT, &fdmi_req->dma_buf,
+ len)) {
+ csio_inc_stats(ln, n_fdmi_err);
+ csio_ln_err(ln, "Failed to issue fdmi rpa req\n");
+ }
+ spin_unlock_irq(&hw->lock);
+}
+
+/*
+ * csio_ln_fdmi_dprt_cbfn - DPRT completion
+ * @hw: HW context
+ * @fdmi_req: fdmi request
+ */
+static void
+csio_ln_fdmi_dprt_cbfn(struct csio_hw *hw, struct csio_ioreq *fdmi_req)
+{
+ void *cmd;
+ uint8_t *pld;
+ uint32_t len = 0;
+ struct csio_hba_identifier *id;
+ struct csio_lnode *ln = fdmi_req->lnode;
+ struct csio_reg_port_list *reg_pl;
+ struct csio_attrib_block *attrib_blk;
+ uint8_t buf[64];
+
+ if (fdmi_req->wr_status != FW_SUCCESS) {
+ csio_ln_err(ln, "WR error:%x in processing fdmi dprt cmd\n",
+ fdmi_req->wr_status);
+ csio_inc_stats(ln, n_fdmi_err);
+ }
+
+ if (!csio_is_rnode_ready(fdmi_req->rnode)) {
+ csio_inc_stats(ln, n_fdmi_err);
+ return;
+ }
+ cmd = fdmi_req->dma_buf.vaddr;
+ if (ntohs(csio_ct_rsp(cmd)) != CT_RESPONSE_FS_ACC) {
+ csio_ln_dbg(ln, "fdmi dprt cmd rejected reason %x expl %x\n",
+ csio_ct_reason(cmd), csio_ct_expl(cmd));
+ }
+
+ /* Prepare CT hdr for RHBA cmd */
+ memset(cmd, 0, CT_BASIC_IU_LEN);
+ csio_fill_ct_iu(cmd,
+ CT_GS_MGMT_SERVICE,
+ CT_FDMI_HBA_MGMT_SERVER,
+ htons(CT_FDMI_HBA_RHBA));
+ len = CT_BASIC_IU_LEN;
+
+ /* Prepare RHBA payload */
+ pld = (uint8_t *) csio_ct_get_pld(cmd);
+ id = (struct csio_hba_identifier *) pld;
+ memcpy(id->wwpn, csio_ln_wwpn(ln), 8); /* HBA identifer */
+ pld += 8;
+
+ /* Register one port per hba */
+ reg_pl = (struct csio_reg_port_list *) pld;
+ reg_pl->entry_count = ntohl(1);
+ memcpy(reg_pl->entry.wwpn, csio_ln_wwpn(ln), 8);
+ pld += sizeof(*reg_pl);
+
+ /* Start appending HBA attributes hba */
+ attrib_blk = (struct csio_attrib_block *) pld;
+ attrib_blk->entry_count = 0;
+ len += sizeof(attrib_blk->entry_count);
+ pld += sizeof(attrib_blk->entry_count);
+
+ csio_append_string_attrib(&pld, NODE_NAME, csio_ln_wwnn(ln), 8);
+ attrib_blk->entry_count++;
+
+ memset(buf, 0, sizeof(buf));
+
+ strcpy(buf, "Chelsio Communications");
+ csio_append_string_attrib(&pld, MANUFACTURER, buf,
+ (uint16_t)strlen(buf));
+ attrib_blk->entry_count++;
+ csio_append_string_attrib(&pld, SERIAL_NUMBER, hw->vpd.sn,
+ (uint16_t)sizeof(hw->vpd.sn));
+ attrib_blk->entry_count++;
+ csio_append_string_attrib(&pld, MODEL, hw->vpd.id,
+ (uint16_t)sizeof(hw->vpd.id));
+ attrib_blk->entry_count++;
+ csio_append_string_attrib(&pld, MODEL_DESCRIPTION, hw->model_desc,
+ (uint16_t)strlen(hw->model_desc));
+ attrib_blk->entry_count++;
+ csio_append_string_attrib(&pld, HARDWARE_VERSION, hw->hw_ver,
+ (uint16_t)sizeof(hw->hw_ver));
+ attrib_blk->entry_count++;
+ csio_append_string_attrib(&pld, FIRMWARE_VERSION, hw->fwrev_str,
+ (uint16_t)strlen(hw->fwrev_str));
+ attrib_blk->entry_count++;
+
+ if (!csio_osname(buf, sizeof(buf))) {
+ csio_append_string_attrib(&pld, OS_NAME_VERSION , buf,
+ (uint16_t)strlen(buf));
+ attrib_blk->entry_count++;
+ }
+
+ csio_append_int_attrib(&pld, MAX_CT_PAYLOAD_LEN, 65536, 4);
+ len = (uint32_t)(pld - (uint8_t *)cmd);
+ attrib_blk->entry_count++;
+ attrib_blk->entry_count = ntohl(attrib_blk->entry_count);
+
+ /* Submit FDMI RHBA request */
+ spin_lock_irq(&hw->lock);
+ if (csio_ln_mgmt_submit_req(fdmi_req, csio_ln_fdmi_rhba_cbfn,
+ FCOE_CT, &fdmi_req->dma_buf,
+ len)) {
+ csio_inc_stats(ln, n_fdmi_err);
+ csio_ln_err(ln, "Failed to issue fdmi rhba req\n");
+ }
+ spin_unlock_irq(&hw->lock);
+}
+
+/*
+ * csio_ln_fdmi_dhba_cbfn - DHBA completion
+ * @hw: HW context
+ * @fdmi_req: fdmi request
+ */
+static void
+csio_ln_fdmi_dhba_cbfn(struct csio_hw *hw, struct csio_ioreq *fdmi_req)
+{
+ struct csio_lnode *ln = fdmi_req->lnode;
+ void *cmd;
+ uint8_t *port_name;
+ uint32_t len;
+
+ if (fdmi_req->wr_status != FW_SUCCESS) {
+ csio_ln_err(ln, "WR error:%x in processing fdmi dhba cmd\n",
+ fdmi_req->wr_status);
+ csio_inc_stats(ln, n_fdmi_err);
+ }
+
+ if (!csio_is_rnode_ready(fdmi_req->rnode)) {
+ csio_inc_stats(ln, n_fdmi_err);
+ return;
+ }
+ cmd = fdmi_req->dma_buf.vaddr;
+ if (ntohs(csio_ct_rsp(cmd)) != CT_RESPONSE_FS_ACC) {
+ csio_ln_dbg(ln, "fdmi dhba cmd rejected reason %x expl %x\n",
+ csio_ct_reason(cmd), csio_ct_expl(cmd));
+ }
+
+ /* Send FDMI cmd to de-register any Port attributes if registered
+ * before
+ */
+
+ /* Prepare FDMI DPRT cmd */
+ memset(cmd, 0, CT_BASIC_IU_LEN);
+ csio_fill_ct_iu(cmd,
+ CT_GS_MGMT_SERVICE,
+ CT_FDMI_HBA_MGMT_SERVER,
+ htons(CT_FDMI_HBA_DPRT));
+ len = CT_BASIC_IU_LEN;
+ port_name = (uint8_t *) csio_ct_get_pld(cmd);
+ memcpy(port_name, csio_ln_wwpn(ln), 8);
+ len += 8;
+
+ /* Submit FDMI request */
+ spin_lock_irq(&hw->lock);
+ if (csio_ln_mgmt_submit_req(fdmi_req, csio_ln_fdmi_dprt_cbfn,
+ FCOE_CT, &fdmi_req->dma_buf,
+ len)) {
+ csio_inc_stats(ln, n_fdmi_err);
+ csio_ln_err(ln, "Failed to issue fdmi dprt req\n");
+ }
+ spin_unlock_irq(&hw->lock);
+}
+
+/**
+ * csio_ln_fdmi_start - Start an FDMI request.
+ * @ln: lnode
+ * @context: session context
+ *
+ * Issued with lock held.
+ */
+csio_retval_t
+csio_ln_fdmi_start(struct csio_lnode *ln, void *context)
+{
+ struct csio_ioreq *fdmi_req;
+ struct csio_rnode *fdmi_rn = (struct csio_rnode *) context;
+ void *cmd;
+ struct csio_hba_identifier *id;
+ uint32_t len;
+
+ if (!(ln->flags & CSIO_LNF_FDMI_ENABLE))
+ return CSIO_NOSUPP;
+
+ if (!csio_is_rnode_ready(fdmi_rn))
+ csio_inc_stats(ln, n_fdmi_err);
+
+ /* Send FDMI cmd to de-register any HBA attributes if registered
+ * before
+ */
+
+ fdmi_req = ln->mgmt_req;
+ fdmi_req->lnode = ln;
+ fdmi_req->rnode = fdmi_rn;
+
+ /* Prepare FDMI DHBA cmd */
+ cmd = fdmi_req->dma_buf.vaddr;
+ memset(cmd, 0, CT_BASIC_IU_LEN);
+ csio_fill_ct_iu(cmd,
+ CT_GS_MGMT_SERVICE,
+ CT_FDMI_HBA_MGMT_SERVER,
+ htons(CT_FDMI_HBA_DHBA));
+ len = CT_BASIC_IU_LEN;
+
+ id = (struct csio_hba_identifier *) csio_ct_get_pld(cmd);
+ memcpy(id->wwpn, csio_ln_wwpn(ln), 8);
+ len += sizeof(*id);
+
+ /* Submit FDMI request */
+ if (csio_ln_mgmt_submit_req(fdmi_req, csio_ln_fdmi_dhba_cbfn,
+ FCOE_CT, &fdmi_req->dma_buf,
+ len)) {
+ csio_inc_stats(ln, n_fdmi_err);
+ csio_ln_err(ln, "Failed to issue fdmi dhba req\n");
+ }
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_ln_vnp_read_cbfn - vnp read completion handler.
+ * @hw: HW lnode
+ * @cbfn: Completion handler.
+ *
+ * Reads vnp response and updates ln parameters.
+ */
+static void
+csio_ln_vnp_read_cbfn(struct csio_hw *hw, struct csio_mb *mbp)
+{
+ struct csio_lnode *ln = ((struct csio_lnode *)mbp->priv);
+ struct fw_fcoe_vnp_cmd *rsp = (struct fw_fcoe_vnp_cmd *)(mbp->mb);
+ struct csio_cmn_sp *csp;
+ struct csio_class_sp *clsp;
+ enum fw_retval retval;
+
+ spin_lock_irq(&hw->lock);
+
+ retval = FW_CMD_RETVAL_GET(ntohl(rsp->alloc_to_len16));
+ if (retval != FW_SUCCESS) {
+ csio_err(hw, "FCOE VNP read cmd returned error:0x%x\n", retval);
+ spin_unlock_irq(&hw->lock);
+ mempool_free(mbp, hw->mb_mempool);
+ return;
+ }
+
+ memcpy(ln->mac, rsp->vnport_mac, sizeof(ln->mac));
+ memcpy(&ln->nport_id, &rsp->vnport_mac[3],
+ sizeof(uint8_t)*3);
+ ln->nport_id = ntohl(ln->nport_id);
+ ln->nport_id = ln->nport_id>>8;
+
+ /* Update WWNs */
+ /*
+ * This may look like a duplication of what csio_fcoe_enable_link()
+ * does, but is absolutely necessary if the vnpi changes between
+ * a FCOE LINK UP and FCOE LINK DOWN.
+ */
+ memcpy(csio_ln_wwnn(ln), rsp->vnport_wwnn, 8);
+ memcpy(csio_ln_wwpn(ln), rsp->vnport_wwpn, 8);
+
+ /* Copy common sparam */
+ csp = (struct csio_cmn_sp *) rsp->cmn_srv_parms;
+ ln->ln_sparm.csp.hi_ver = csp->hi_ver;
+ ln->ln_sparm.csp.lo_ver = csp->lo_ver;
+ ln->ln_sparm.csp.bb_credit = ntohs(csp->bb_credit);
+ ln->ln_sparm.csp.word1_flags = ntohs(csp->word1_flags);
+ ln->ln_sparm.csp.rcv_sz = ntohs(csp->rcv_sz);
+ ln->ln_sparm.csp.un1.r_a_tov = ntohl(csp->un1.r_a_tov);
+ ln->ln_sparm.csp.e_d_tov = ntohl(csp->e_d_tov);
+
+ /* Copy word 0 & word 1 of class sparam */
+ clsp = (struct csio_class_sp *) rsp->clsp_word_0_1;
+ ln->ln_sparm.clsp[2].serv_option = clsp->serv_option;
+ ln->ln_sparm.clsp[2].init_ctl_option = clsp->init_ctl_option;
+ ln->ln_sparm.clsp[2].rcv_ctl_option = clsp->rcv_ctl_option;
+ ln->ln_sparm.clsp[2].rcv_data_sz = ntohs(clsp->rcv_data_sz);
+
+ spin_unlock_irq(&hw->lock);
+
+ mempool_free(mbp, hw->mb_mempool);
+
+ /* Send an event to update local attribs */
+ csio_lnode_async_event(ln, CSIO_LN_FC_ATTRIB_UPDATE);
+}
+
+/*
+ * csio_ln_vnp_read - Read vnp params.
+ * @ln: lnode
+ * @cbfn: Completion handler.
+ *
+ * Issued with lock held.
+ */
+static csio_retval_t
+csio_ln_vnp_read(struct csio_lnode *ln,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct csio_hw *hw = ln->hwp;
+ struct csio_mb *mbp;
+
+ /* Allocate Mbox request */
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_inc_stats(hw, n_err_nomem);
+ return CSIO_NOMEM;
+ }
+
+ /* Prepare VNP Command */
+ csio_fcoe_vnp_read_init_mb(ln, mbp,
+ CSIO_MB_DEFAULT_TMO,
+ ln->fcf_flowid,
+ ln->vnp_flowid,
+ cbfn);
+
+ /* Issue MBOX cmd */
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "Failed to issue mbox FCoE VNP command\n");
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_fcoe_enable_link - Enable fcoe link.
+ * @ln: lnode
+ * @enable: enable/disable
+ * Issued with lock held.
+ * Issues mbox cmd to bring up FCOE link on port associated with given ln.
+ */
+static csio_retval_t
+csio_fcoe_enable_link(struct csio_lnode *ln, bool enable)
+{
+ struct csio_hw *hw = ln->hwp;
+ struct csio_mb *mbp;
+ enum fw_retval retval;
+ uint8_t portid;
+ uint8_t sub_op;
+ struct fw_fcoe_link_cmd *lcmd;
+ int i;
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_inc_stats(hw, n_err_nomem);
+ return CSIO_NOMEM;
+ }
+
+ portid = ln->portid;
+ sub_op = enable ? FCOE_LINK_UP : FCOE_LINK_DOWN;
+
+ csio_dbg(hw, "bringing FCOE LINK %s on Port:%d\n",
+ sub_op ? "UP" : "DOWN", portid);
+
+ csio_write_fcoe_link_cond_init_mb(ln, mbp, CSIO_MB_DEFAULT_TMO,
+ portid, sub_op, 0, 0, 0, NULL);
+
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "failed to issue FCOE LINK cmd on port[%d]\n",
+ portid);
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ retval = csio_mb_fw_retval(mbp);
+ if (retval != FW_SUCCESS) {
+ csio_err(hw,
+ "FCOE LINK %s cmd on port[%d] failed with "
+ "ret:x%x\n", sub_op ? "UP" : "DOWN", portid, retval);
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ if (!enable)
+ goto out;
+
+ lcmd = (struct fw_fcoe_link_cmd *)mbp->mb;
+
+ memcpy(csio_ln_wwnn(ln), lcmd->vnport_wwnn, 8);
+ memcpy(csio_ln_wwpn(ln), lcmd->vnport_wwpn, 8);
+
+ for (i = 0; i < CSIO_MAX_PPORTS; i++)
+ if (hw->pport[i].portid == portid)
+ memcpy(hw->pport[i].mac, lcmd->phy_mac, 6);
+
+out:
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_ln_read_fcf_cbfn - Read fcf parameters
+ * @ln: lnode
+ *
+ * read fcf response and Update ln fcf information.
+ */
+static void
+csio_ln_read_fcf_cbfn(struct csio_hw *hw, struct csio_mb *mbp)
+{
+ struct csio_lnode *ln = (struct csio_lnode *)mbp->priv;
+ struct csio_fcf_info *fcf_info;
+ struct fw_fcoe_fcf_cmd *rsp =
+ (struct fw_fcoe_fcf_cmd *)(mbp->mb);
+ enum fw_retval retval;
+
+ spin_lock_irq(&hw->lock);
+
+ retval = FW_CMD_RETVAL_GET(ntohl(rsp->retval_len16));
+ if (retval != FW_SUCCESS) {
+ csio_ln_err(ln, "FCOE FCF cmd failed with ret x%x\n",
+ retval);
+ mempool_free(mbp, hw->mb_mempool);
+ spin_unlock_irq(&hw->lock);
+ return;
+ }
+
+ fcf_info = ln->fcfinfo;
+ fcf_info->priority = FW_FCOE_FCF_CMD_PRIORITY_GET(
+ ntohs(rsp->priority_pkd));
+ fcf_info->vf_id = ntohs(rsp->vf_id);
+ fcf_info->vlan_id = rsp->vlan_id;
+ fcf_info->max_fcoe_size = ntohs(rsp->max_fcoe_size);
+ fcf_info->fka_adv = be32_to_cpu(rsp->fka_adv);
+ fcf_info->fcfi = FW_FCOE_FCF_CMD_FCFI_GET(ntohl(rsp->op_to_fcfi));
+ fcf_info->fpma = FW_FCOE_FCF_CMD_FPMA_GET(rsp->fpma_to_portid);
+ fcf_info->spma = FW_FCOE_FCF_CMD_SPMA_GET(rsp->fpma_to_portid);
+ fcf_info->login = FW_FCOE_FCF_CMD_LOGIN_GET(rsp->fpma_to_portid);
+ fcf_info->portid = FW_FCOE_FCF_CMD_PORTID_GET(rsp->fpma_to_portid);
+ memcpy(fcf_info->fc_map, rsp->fc_map, sizeof(fcf_info->fc_map));
+ memcpy(fcf_info->mac, rsp->mac, sizeof(fcf_info->mac));
+ memcpy(fcf_info->name_id, rsp->name_id, sizeof(fcf_info->name_id));
+ memcpy(fcf_info->fabric, rsp->fabric, sizeof(fcf_info->fabric));
+ memcpy(fcf_info->spma_mac, rsp->spma_mac, sizeof(fcf_info->spma_mac));
+
+ spin_unlock_irq(&hw->lock);
+ mempool_free(mbp, hw->mb_mempool);
+
+}
+
+/*
+ * csio_ln_read_fcf_entry - Read fcf entry.
+ * @ln: lnode
+ * @cbfn: Completion handler.
+ *
+ * Issued with lock held.
+ */
+static csio_retval_t
+csio_ln_read_fcf_entry(struct csio_lnode *ln,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct csio_hw *hw = ln->hwp;
+ struct csio_mb *mbp;
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_inc_stats(hw, n_err_nomem);
+ return CSIO_NOMEM;
+ }
+
+ /* Get FCoE FCF information */
+ csio_fcoe_read_fcf_init_mb(ln, mbp, CSIO_MB_DEFAULT_TMO,
+ ln->portid, ln->fcf_flowid, cbfn);
+
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "failed to issue FCOE FCF cmd\n");
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ if (cbfn == NULL) {
+ spin_unlock_irq(&hw->lock);
+ csio_ln_read_fcf_cbfn(hw, mbp);
+ spin_lock_irq(&hw->lock);
+ }
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_handle_link_up - Logical Linkup event.
+ * @hw - HW module.
+ * @portid - Physical port number
+ * @fcfi - FCF index.
+ * @vnpi - VNP index.
+ * Returns - none.
+ *
+ * This event is received from FW, when virtual link is established between
+ * Physical port[ENode] and FCF. If its new vnpi, then local node object is
+ * created on this FCF and set to [ONLINE] state.
+ * Lnode waits for FW_RDEV_CMD event to be received indicating that
+ * Fabric login is completed and lnode moves to [READY] state.
+ *
+ * This called with hw lock held
+ */
+static void
+csio_handle_link_up(struct csio_hw *hw, uint8_t portid, uint32_t fcfi,
+ uint32_t vnpi)
+{
+ struct csio_lnode *ln = NULL;
+
+ /* Lookup lnode based on vnpi */
+ ln = csio_ln_lookup_by_vnpi(hw, vnpi);
+ if (!ln) {
+ /* Pick lnode based on portid */
+ ln = csio_ln_lookup_by_portid(hw, portid);
+ if (!ln) {
+ csio_err(hw, "failed to lookup fcoe lnode on port:%d\n",
+ portid);
+ CSIO_DB_ASSERT(0);
+ return;
+ }
+
+ /* Check if lnode has valid vnp flowid */
+ if (ln->vnp_flowid != CSIO_INVALID_IDX) {
+ /* New VN-Port */
+ spin_unlock_irq(&hw->lock);
+ csio_lnode_alloc(hw);
+ spin_lock_irq(&hw->lock);
+ if (!ln) {
+ csio_err(hw,
+ "failed to allocate fcoe lnode"
+ "for port:%d vnpi:x%x\n",
+ portid, vnpi);
+ CSIO_DB_ASSERT(0);
+ return;
+ }
+ ln->portid = portid;
+ }
+ ln->vnp_flowid = vnpi;
+ ln->dev_num &= ~0xFFFF;
+ ln->dev_num |= vnpi;
+ }
+
+ /*Initialize fcfi */
+ ln->fcf_flowid = fcfi;
+
+ csio_info(hw, "Port:%d - FCOE LINK UP\n", portid);
+
+ csio_inc_stats(ln, n_link_up);
+
+ /* Send LINKUP event to SM */
+ csio_post_event(&ln->sm, CSIO_LNE_LINKUP);
+}
+
+/*
+ * csio_post_event_rns
+ * @ln - FCOE lnode
+ * @evt - Given rnode event
+ * Returns - none
+ *
+ * Posts given rnode event to all FCOE rnodes connected with given Lnode.
+ * This routine is invoked when lnode receives LINK_DOWN/DOWN_LINK/CLOSE
+ * event.
+ *
+ * This called with hw lock held
+ */
+static void
+csio_post_event_rns(struct csio_lnode *ln, enum csio_rn_ev evt)
+{
+ struct csio_rnode *rnhead = (struct csio_rnode *) &ln->rnhead;
+ struct list_head *tmp, *next;
+ struct csio_rnode *rn;
+
+ list_for_each_safe(tmp, next, &rnhead->sm.sm_list) {
+ rn = (struct csio_rnode *) tmp;
+ csio_post_event(&rn->sm, evt);
+ }
+}
+
+/*
+ * csio_cleanup_rns
+ * @ln - FCOE lnode
+ * Returns - none
+ *
+ * Frees all FCOE rnodes connected with given Lnode.
+ *
+ * This called with hw lock held
+ */
+static void
+csio_cleanup_rns(struct csio_lnode *ln)
+{
+ struct csio_rnode *rnhead = (struct csio_rnode *) &ln->rnhead;
+ struct list_head *tmp, *next_rn;
+ struct csio_rnode *rn;
+
+ list_for_each_safe(tmp, next_rn, &rnhead->sm.sm_list) {
+ rn = (struct csio_rnode *) tmp;
+ csio_put_rnode(ln, rn);
+ }
+
+}
+
+/*
+ * csio_post_event_lns
+ * @ln - FCOE lnode
+ * @evt - Given lnode event
+ * Returns - none
+ *
+ * Posts given lnode event to all FCOE lnodes connected with given Lnode.
+ * This routine is invoked when lnode receives LINK_DOWN/DOWN_LINK/CLOSE
+ * event.
+ *
+ * This called with hw lock held
+ */
+static void
+csio_post_event_lns(struct csio_lnode *ln, enum csio_ln_ev evt)
+{
+ struct list_head *tmp;
+ struct csio_lnode *cln, *sln;
+
+ /* If NPIV lnode, send evt only to that and return */
+ if (csio_is_npiv_ln(ln)) {
+ csio_post_event(&ln->sm, evt);
+ return;
+ }
+
+ sln = ln;
+ /* Traverse children lnodes list and send evt */
+ list_for_each(tmp, &sln->cln_head) {
+ cln = (struct csio_lnode *) tmp;
+ csio_post_event(&cln->sm, evt);
+ }
+
+ /* Send evt to parent lnode */
+ csio_post_event(&ln->sm, evt);
+}
+
+/*
+ * csio_ln_down - Lcoal nport is down
+ * @ln - FCOE Lnode
+ * Returns - none
+ *
+ * Sends LINK_DOWN events to Lnode and its associated NPIVs lnodes.
+ *
+ * This called with hw lock held
+ */
+static void
+csio_ln_down(struct csio_lnode *ln)
+{
+ csio_post_event_lns(ln, CSIO_LNE_LINK_DOWN);
+}
+
+/*
+ * csio_handle_link_down - Logical Linkdown event.
+ * @hw - HW module.
+ * @portid - Physical port number
+ * @fcfi - FCF index.
+ * @vnpi - VNP index.
+ * Returns - none
+ *
+ * This event is received from FW, when virtual link goes down between
+ * Physical port[ENode] and FCF. Lnode and its associated NPIVs lnode hosted on
+ * this vnpi[VN-Port] will be de-instantiated.
+ *
+ * This called with hw lock held
+ */
+static void
+csio_handle_link_down(struct csio_hw *hw, uint8_t portid, uint32_t fcfi,
+ uint32_t vnpi)
+{
+ struct csio_fcf_info *fp;
+ struct csio_lnode *ln;
+
+ /* Lookup lnode based on vnpi */
+ ln = csio_ln_lookup_by_vnpi(hw, vnpi);
+ if (ln) {
+ fp = ln->fcfinfo;
+ csio_inc_stats(ln, n_link_down);
+
+ /*Warn if linkdown received if lnode is not in ready state */
+ if (!csio_is_lnode_ready(ln)) {
+ csio_ln_warn(ln,
+ "warn: FCOE link is already in offline "
+ "Ignoring Fcoe linkdown event on portid %d\n",
+ portid);
+ csio_inc_stats(ln, n_evt_drop);
+ return;
+ }
+
+ /* Verify portid */
+ if (fp->portid != portid) {
+ csio_ln_warn(ln,
+ "warn: FCOE linkdown recv with "
+ "invalid port %d\n", portid);
+ csio_inc_stats(ln, n_evt_drop);
+ return;
+ }
+
+ /* verify fcfi */
+ if (ln->fcf_flowid != fcfi) {
+ csio_ln_warn(ln,
+ "warn: FCOE linkdown recv with "
+ "invalid fcfi x%x\n", fcfi);
+ csio_inc_stats(ln, n_evt_drop);
+ return;
+ }
+
+ csio_info(hw, "Port:%d - FCOE LINK DOWN\n", portid);
+
+ /* Send LINK_DOWN event to lnode s/m */
+ csio_ln_down(ln);
+
+ return;
+ } else {
+ csio_warn(hw,
+ "warn: FCOE linkdown recv with invalid vnpi x%x\n",
+ vnpi);
+ csio_inc_stats(hw, n_evt_drop);
+ }
+}
+
+/*
+ * csio_is_lnode_ready - Checks FCOE lnode is in ready state.
+ * @ln: Lnode module
+ *
+ * Returns True if FCOE lnode is in ready state.
+ */
+int
+csio_is_lnode_ready(struct csio_lnode *ln)
+{
+ return (csio_get_state(ln) == ((csio_sm_state_t)csio_lns_ready));
+}
+
+/*****************************************************************************/
+/* START: Lnode SM */
+/*****************************************************************************/
+/*
+ * csio_lns_uninit - The request in uninit state.
+ * @ln - FCOE lnode.
+ * @evt - Event to be processed.
+ *
+ * Process the given lnode event which is currently in "uninit" state.
+ * Invoked with HW lock held.
+ * Return - none.
+ */
+static void
+csio_lns_uninit(struct csio_lnode *ln, enum csio_ln_ev evt)
+{
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+ struct csio_lnode *rln = hw->rln;
+ csio_retval_t rv;
+
+ csio_inc_stats(ln, n_evt_sm[evt]);
+ switch (evt) {
+
+ case CSIO_LNE_LINKUP:
+ csio_set_state(&ln->sm, csio_lns_online);
+ /* Read FCF only for physical lnode */
+ if (csio_is_phys_ln(ln)) {
+ rv = csio_ln_read_fcf_entry(ln,
+ csio_ln_read_fcf_cbfn);
+ if (rv != CSIO_SUCCESS) {
+ /* TODO: Send HW RESET event */
+ csio_inc_stats(ln, n_err);
+ break;
+ }
+
+ /* Add FCF record */
+ list_add_tail(&ln->fcfinfo->list, &rln->fcf_lsthead);
+ }
+
+ rv = csio_ln_vnp_read(ln, csio_ln_vnp_read_cbfn);
+ if (rv != CSIO_SUCCESS) {
+ /* TODO: Send HW RESET event */
+ csio_inc_stats(ln, n_err);
+ }
+ break;
+
+ case CSIO_LNE_DOWN_LINK:
+ break;
+
+ default:
+ csio_ln_dbg(ln,
+ "unexp ln event %d recv from did:x%x in "
+ "ln state[uninit].\n", evt, ln->nport_id);
+ csio_inc_stats(ln, n_evt_unexp);
+ break;
+ } /* switch event */
+}
+
+/*
+ * csio_lns_online - The request in online state.
+ * @ln - FCOE lnode.
+ * @evt - Event to be processed.
+ *
+ * Process the given lnode event which is currently in "online" state.
+ * Invoked with HW lock held.
+ * Return - none.
+ */
+static void
+csio_lns_online(struct csio_lnode *ln, enum csio_ln_ev evt)
+{
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ csio_inc_stats(ln, n_evt_sm[evt]);
+ switch (evt) {
+
+ case CSIO_LNE_LINKUP:
+ csio_ln_warn(ln,
+ "warn: FCOE link is up already "
+ "Ignoring linkup on port:%d\n", ln->portid);
+ csio_inc_stats(ln, n_evt_drop);
+ break;
+
+ case CSIO_LNE_FAB_INIT_DONE:
+ csio_set_state(&ln->sm, csio_lns_ready);
+
+ spin_unlock_irq(&hw->lock);
+ csio_lnode_async_event(ln, CSIO_LN_FC_LINKUP);
+ spin_lock_irq(&hw->lock);
+
+ break;
+
+ case CSIO_LNE_LINK_DOWN:
+ /* Fall through */
+ case CSIO_LNE_DOWN_LINK:
+ csio_set_state(&ln->sm, csio_lns_uninit);
+ if (csio_is_phys_ln(ln)) {
+ /* Remove FCF entry */
+ list_del_init(&ln->fcfinfo->list);
+ }
+ break;
+
+ default:
+ csio_ln_dbg(ln,
+ "unexp ln event %d recv from did:x%x in "
+ "ln state[uninit].\n", evt, ln->nport_id);
+ csio_inc_stats(ln, n_evt_unexp);
+
+ break;
+ } /* switch event */
+}
+
+/*
+ * csio_lns_ready - The request in ready state.
+ * @ln - FCOE lnode.
+ * @evt - Event to be processed.
+ *
+ * Process the given lnode event which is currently in "ready" state.
+ * Invoked with HW lock held.
+ * Return - none.
+ */
+static void
+csio_lns_ready(struct csio_lnode *ln, enum csio_ln_ev evt)
+{
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ csio_inc_stats(ln, n_evt_sm[evt]);
+ switch (evt) {
+
+ case CSIO_LNE_FAB_INIT_DONE:
+ csio_ln_err(ln,
+ "ignoring event %d recv from did x%x"
+ "in ln state[ready].\n", evt, ln->nport_id);
+ csio_inc_stats(ln, n_evt_drop);
+ break;
+
+ case CSIO_LNE_LINK_DOWN:
+ csio_set_state(&ln->sm, csio_lns_offline);
+ csio_post_event_rns(ln, CSIO_RNFE_DOWN);
+
+ spin_unlock_irq(&hw->lock);
+ csio_lnode_async_event(ln, CSIO_LN_FC_LINKDOWN);
+ spin_lock_irq(&hw->lock);
+
+ if (csio_is_phys_ln(ln)) {
+ /* Remove FCF entry */
+ list_del_init(&ln->fcfinfo->list);
+ }
+ break;
+
+ case CSIO_LNE_DOWN_LINK:
+ csio_set_state(&ln->sm, csio_lns_offline);
+ csio_post_event_rns(ln, CSIO_RNFE_DOWN);
+
+ /* Host need to issue aborts in case if FW has not returned
+ * WRs with status "ABORTED"
+ */
+ spin_unlock_irq(&hw->lock);
+ csio_lnode_async_event(ln, CSIO_LN_FC_LINKDOWN);
+ spin_lock_irq(&hw->lock);
+
+ if (csio_is_phys_ln(ln)) {
+ /* Remove FCF entry */
+ list_del_init(&ln->fcfinfo->list);
+ }
+ break;
+
+ case CSIO_LNE_CLOSE:
+ csio_set_state(&ln->sm, csio_lns_uninit);
+ csio_post_event_rns(ln, CSIO_RNFE_CLOSE);
+ break;
+
+ case CSIO_LNE_LOGO:
+ csio_set_state(&ln->sm, csio_lns_offline);
+ csio_post_event_rns(ln, CSIO_RNFE_DOWN);
+ break;
+
+ default:
+ csio_ln_dbg(ln,
+ "unexp ln event %d recv from did:x%x in "
+ "ln state[uninit].\n", evt, ln->nport_id);
+ csio_inc_stats(ln, n_evt_unexp);
+ CSIO_DB_ASSERT(0);
+ break;
+ } /* switch event */
+}
+
+/*
+ * csio_lns_offline - The request in offline state.
+ * @ln - FCOE lnode.
+ * @evt - Event to be processed.
+ *
+ * Process the given lnode event which is currently in "offline" state.
+ * Invoked with HW lock held.
+ * Return - none.
+ */
+static void
+csio_lns_offline(struct csio_lnode *ln, enum csio_ln_ev evt)
+{
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+ struct csio_lnode *rln = hw->rln;
+ csio_retval_t rv;
+
+ csio_inc_stats(ln, n_evt_sm[evt]);
+ switch (evt) {
+
+ case CSIO_LNE_LINKUP:
+ csio_set_state(&ln->sm, csio_lns_online);
+ /* Read FCF only for physical lnode */
+ if (csio_is_phys_ln(ln)) {
+ rv = csio_ln_read_fcf_entry(ln,
+ csio_ln_read_fcf_cbfn);
+ if (rv != CSIO_SUCCESS) {
+ /* TODO: Send HW RESET event */
+ csio_inc_stats(ln, n_err);
+ break;
+ }
+
+ /* Add FCF record */
+ list_add_tail(&ln->fcfinfo->list, &rln->fcf_lsthead);
+ }
+
+ rv = csio_ln_vnp_read(ln, csio_ln_vnp_read_cbfn);
+ if (rv != CSIO_SUCCESS) {
+ /* TODO: Send HW RESET event */
+ csio_inc_stats(ln, n_err);
+ }
+ break;
+
+ case CSIO_LNE_LINK_DOWN:
+ case CSIO_LNE_DOWN_LINK:
+ case CSIO_LNE_LOGO:
+ csio_ln_err(ln,
+ "ignoring event %d recv from did x%x"
+ "in ln state[offline].\n", evt, ln->nport_id);
+ csio_inc_stats(ln, n_evt_drop);
+ break;
+
+ case CSIO_LNE_CLOSE:
+ csio_set_state(&ln->sm, csio_lns_uninit);
+ csio_post_event_rns(ln, CSIO_RNFE_CLOSE);
+ break;
+
+ default:
+ csio_ln_dbg(ln,
+ "unexp ln event %d recv from did:x%x in "
+ "ln state[offline]\n", evt, ln->nport_id);
+ csio_inc_stats(ln, n_evt_unexp);
+ CSIO_DB_ASSERT(0);
+ break;
+ } /* switch event */
+}
+
+/*****************************************************************************/
+/* END: Lnode SM */
+/*****************************************************************************/
+
+static void
+csio_free_fcfinfo(struct kref *kref)
+{
+ struct csio_fcf_info *fcfinfo = container_of(kref,
+ struct csio_fcf_info, kref);
+ kfree(fcfinfo);
+}
+
+/* Helper routines for attributes */
+/*
+ * csio_lnode_state_to_str - Get current state of FCOE lnode.
+ * @ln - lnode
+ * @str - state of lnode.
+ *
+ */
+void
+csio_lnode_state_to_str(struct csio_lnode *ln, int8_t *str)
+{
+ if (csio_get_state(ln) == ((csio_sm_state_t)csio_lns_uninit)) {
+ strcpy(str, "UNINIT");
+ return;
+ }
+ if (csio_get_state(ln) == ((csio_sm_state_t)csio_lns_ready)) {
+ strcpy(str, "READY");
+ return;
+ }
+ if (csio_get_state(ln) == ((csio_sm_state_t)csio_lns_offline)) {
+ strcpy(str, "OFFLINE");
+ return;
+ }
+ strcpy(str, "UNKNOWN");
+} /* csio_lnode_state_to_str */
+
+
+csio_retval_t
+csio_get_phy_port_stats(struct csio_hw *hw, uint8_t portid,
+ struct fw_fcoe_port_stats *port_stats)
+{
+ struct csio_mb *mbp;
+ struct fw_fcoe_port_cmd_params portparams;
+ enum fw_retval retval;
+ int idx;
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_err(hw, "FCoE FCF PARAMS command out of memory!\n");
+ return CSIO_INVAL;
+ }
+ portparams.portid = portid;
+
+ for (idx = 1; idx <= 3; idx++) {
+ portparams.idx = (idx-1)*6 + 1;
+ portparams.nstats = 6;
+ if (idx == 3)
+ portparams.nstats = 4;
+ csio_fcoe_read_portparams_init_mb(hw, mbp, CSIO_MB_DEFAULT_TMO,
+ &portparams, NULL);
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "Issue of FCoE port params failed!\n");
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+ csio_mb_process_portparams_rsp(hw, mbp, &retval,
+ &portparams, port_stats);
+ }
+
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_ln_mgmt_wr_handler -Mgmt Work Request handler.
+ * @wr - WR.
+ * @len - WR len.
+ * This handler is invoked when an outstanding mgmt WR is completed.
+ * Its invoked in the context of FW event worker thread for every
+ * mgmt event received.
+ * Return - none.
+ */
+
+static void
+csio_ln_mgmt_wr_handler(struct csio_hw *hw, void *wr, uint32_t len)
+{
+ struct csio_mgmtm *mgmtm = csio_hw_to_mgmtm(hw);
+ struct csio_ioreq *io_req = NULL;
+ struct fw_fcoe_els_ct_wr *wr_cmd;
+
+
+ wr_cmd = (struct fw_fcoe_els_ct_wr *) wr;
+
+ if (len < sizeof(struct fw_fcoe_els_ct_wr)) {
+ csio_err(mgmtm->hw,
+ "Invalid ELS CT WR length recvd, len:%x\n", len);
+ mgmtm->stats.n_err++;
+ return;
+ }
+
+ io_req = (struct csio_ioreq *) ((uintptr_t) wr_cmd->cookie);
+ io_req->wr_status = csio_wr_status(wr_cmd);
+
+ /* lookup ioreq exists in our active Q */
+ spin_lock_irq(&hw->lock);
+ if (csio_mgmt_req_lookup(mgmtm, io_req) != CSIO_SUCCESS) {
+ csio_err(mgmtm->hw,
+ "Error- Invalid IO handle recv in WR. handle: %p\n",
+ io_req);
+ mgmtm->stats.n_err++;
+ spin_unlock_irq(&hw->lock);
+ return;
+ }
+
+ mgmtm = csio_hw_to_mgmtm(hw);
+
+ /* Dequeue from active queue */
+ list_del_init(&io_req->sm.sm_list);
+ mgmtm->stats.n_active--;
+ spin_unlock_irq(&hw->lock);
+
+ if (io_req->io_cbfn) {
+ /* io_req will be freed by completion handler */
+ io_req->io_cbfn(hw, io_req);
+ }
+}
+
+/**
+ * csio_fcoe_fwevt_handler - Event handler for Firmware FCoE events.
+ * @hw: HW module
+ * @cpl_op: CPL opcode
+ * @cmd: FW cmd/WR.
+ *
+ * Process received FCoE cmd/WR event from FW.
+ */
+void
+csio_fcoe_fwevt_handler(struct csio_hw *hw, __u8 cpl_op, __be64 *cmd)
+{
+ struct csio_lnode *ln;
+ struct csio_rnode *rn;
+ uint8_t portid, opcode = *(uint8_t *)cmd;
+ struct fw_fcoe_link_cmd *lcmd;
+ struct fw_wr_hdr *wr;
+ struct fw_rdev_wr *rdev_wr;
+ enum fw_fcoe_link_status lstatus;
+ uint32_t fcfi, rdev_flowid, vnpi;
+ enum csio_ln_ev evt;
+
+ if (cpl_op == CPL_FW6_MSG && opcode == FW_FCOE_LINK_CMD) {
+
+ lcmd = (struct fw_fcoe_link_cmd *)cmd;
+ lstatus = lcmd->lstatus;
+ portid = FW_FCOE_LINK_CMD_PORTID_GET(
+ ntohl(lcmd->op_to_portid));
+ fcfi = FW_FCOE_LINK_CMD_FCFI_GET(ntohl(lcmd->sub_opcode_fcfi));
+ vnpi = FW_FCOE_LINK_CMD_VNPI_GET(ntohl(lcmd->vnpi_pkd));
+
+ if (lstatus == FCOE_LINKUP) {
+
+ /* HW lock here */
+ spin_lock_irq(&hw->lock);
+ csio_handle_link_up(hw, portid, fcfi, vnpi);
+ spin_unlock_irq(&hw->lock);
+ /* HW un lock here */
+
+ } else if (lstatus == FCOE_LINKDOWN) {
+
+ /* HW lock here */
+ spin_lock_irq(&hw->lock);
+ csio_handle_link_down(hw, portid, fcfi, vnpi);
+ spin_unlock_irq(&hw->lock);
+ /* HW un lock here */
+ } else {
+ csio_warn(hw, "Unexpected FCOE LINK status:0x%x\n",
+ ntohl(lcmd->lstatus));
+ csio_inc_stats(hw, n_cpl_unexp);
+ }
+ } else if (cpl_op == CPL_FW6_PLD) {
+ wr = (struct fw_wr_hdr *) (cmd + 4);
+ if (FW_WR_OP_GET(be32_to_cpu(wr->hi))
+ == FW_RDEV_WR) {
+
+ rdev_wr = (struct fw_rdev_wr *) (cmd + 4);
+
+ rdev_flowid = FW_RDEV_WR_FLOWID_GET(
+ ntohl(rdev_wr->alloc_to_len16));
+ vnpi = FW_RDEV_WR_ASSOC_FLOWID_GET(
+ ntohl(rdev_wr->flags_to_assoc_flowid));
+
+ csio_dbg(hw,
+ "FW_RDEV_WR: flowid:x%x ev_cause:x%x "
+ "vnpi:0x%x\n", rdev_flowid,
+ rdev_wr->event_cause, vnpi);
+
+ if (rdev_wr->protocol != PROT_FCOE) {
+ csio_err(hw,
+ "FW_RDEV_WR: invalid proto:x%x "
+ "received with flowid:x%x\n",
+ rdev_wr->protocol,
+ rdev_flowid);
+ csio_inc_stats(hw, n_evt_drop);
+ return;
+ }
+
+ /* HW lock here */
+ spin_lock_irq(&hw->lock);
+ ln = csio_ln_lookup_by_vnpi(hw, vnpi);
+ if (!ln) {
+ csio_err(hw,
+ "FW_DEV_WR: invalid vnpi:x%x received "
+ "with flowid:x%x\n", vnpi, rdev_flowid);
+ csio_inc_stats(hw, n_evt_drop);
+ spin_unlock_irq(&hw->lock);
+ return;
+ }
+
+ rn = csio_confirm_rnode(ln, rdev_flowid,
+ &rdev_wr->u.fcoe_rdev);
+ if (!rn) {
+ csio_ln_dbg(ln,
+ "Failed to confirm rnode "
+ "for flowid:x%x\n", rdev_flowid);
+ csio_inc_stats(hw, n_evt_drop);
+ spin_unlock_irq(&hw->lock);
+ return;
+ }
+
+ /* save previous event for debugging */
+ ln->prev_evt = ln->cur_evt;
+ ln->cur_evt = rdev_wr->event_cause;
+ csio_inc_stats(ln, n_evt_fw[rdev_wr->event_cause]);
+
+ /* Translate all the fabric events to lnode SM events */
+ evt = CSIO_FWE_TO_LNE(rdev_wr->event_cause);
+ if (evt) {
+ csio_ln_dbg(ln,
+ "Posting event to lnode event:%d "
+ "cause:%d flowid:x%x\n", evt,
+ rdev_wr->event_cause, rdev_flowid);
+ csio_post_event(&ln->sm, evt);
+ }
+
+ /* Handover event to rn SM here. */
+ csio_rnode_fwevt_handler(rn, rdev_wr->event_cause);
+
+ spin_unlock_irq(&hw->lock);
+ } else {
+ csio_warn(hw, "unexpected WR op(0x%x) recv\n",
+ FW_WR_OP_GET(be32_to_cpu((wr->hi))));
+ csio_inc_stats(hw, n_cpl_unexp);
+ }
+ } else if (cpl_op == CPL_FW6_MSG) {
+ wr = (struct fw_wr_hdr *) (cmd);
+ if (FW_WR_OP_GET(be32_to_cpu(wr->hi))
+ == FW_FCOE_ELS_CT_WR) {
+ csio_ln_mgmt_wr_handler(hw, wr,
+ sizeof(struct fw_fcoe_els_ct_wr));
+ } else {
+ csio_warn(hw, "unexpected WR op(0x%x) recv\n",
+ FW_WR_OP_GET(be32_to_cpu((wr->hi))));
+ csio_inc_stats(hw, n_cpl_unexp);
+ }
+ } else {
+ csio_warn(hw, "unexpected CPL op(0x%x) recv\n", opcode);
+ csio_inc_stats(hw, n_cpl_unexp);
+ }
+}
+
+/**
+ * csio_lnode_start - Kickstart lnode discovery.
+ * @ln: lnode
+ *
+ * This routine kickstarts the discovery by issuing an FCOE_LINK (up) command.
+ */
+csio_retval_t
+csio_lnode_start(struct csio_lnode *ln)
+{
+ csio_retval_t rv = CSIO_SUCCESS;
+ if (csio_is_phys_ln(ln) && !(ln->flags & CSIO_LNF_LINK_ENABLE)) {
+ rv = csio_fcoe_enable_link(ln, 1);
+ ln->flags |= CSIO_LNF_LINK_ENABLE;
+ }
+
+ return rv;
+}
+
+/**
+ * csio_lnode_stop - Stop the lnode.
+ * @ln: lnode
+ *
+ * This routine is invoked by HW module to stop lnode and its associated NPIV
+ * lnodes.
+ */
+void
+csio_lnode_stop(struct csio_lnode *ln)
+{
+ csio_post_event_lns(ln, CSIO_LNE_DOWN_LINK);
+ if (csio_is_phys_ln(ln) && (ln->flags & CSIO_LNF_LINK_ENABLE)) {
+ csio_fcoe_enable_link(ln, 0);
+ ln->flags &= ~CSIO_LNF_LINK_ENABLE;
+ }
+ csio_ln_dbg(ln, "stopping ln :%p\n", ln);
+}
+
+/**
+ * csio_lnode_close - Close an lnode.
+ * @ln: lnode
+ *
+ * This routine is invoked by HW module to close an lnode and its
+ * associated NPIV lnodes. Lnode and its associated NPIV lnodes are
+ * set to uninitialized state.
+ */
+void
+csio_lnode_close(struct csio_lnode *ln)
+{
+ csio_post_event_lns(ln, CSIO_LNE_CLOSE);
+ if (csio_is_phys_ln(ln))
+ ln->vnp_flowid = CSIO_INVALID_IDX;
+
+ csio_ln_dbg(ln, "closed ln :%p\n", ln);
+}
+
+/*
+ * csio_ln_prep_ecwr - Prepare ELS/CT WR.
+ * @io_req - IO request.
+ * @wr_len - WR len
+ * @immd_len - WR immediate data
+ * @sub_op - Sub opcode
+ * @sid - source portid.
+ * @did - destination portid
+ * @flow_id - flowid
+ * @fw_wr - ELS/CT WR to be prepared.
+ * Returns: CSIO_SUCCESS - on success
+ */
+static csio_retval_t
+csio_ln_prep_ecwr(struct csio_ioreq *io_req, uint32_t wr_len,
+ uint32_t immd_len, uint8_t sub_op, uint32_t sid,
+ uint32_t did, uint32_t flow_id, uint8_t *fw_wr)
+{
+ struct fw_fcoe_els_ct_wr *wr;
+ uint32_t port_id;
+
+ wr = (struct fw_fcoe_els_ct_wr *)fw_wr;
+ wr->op_immdlen = cpu_to_be32(FW_WR_OP(FW_FCOE_ELS_CT_WR) |
+ FW_FCOE_ELS_CT_WR_IMMDLEN(immd_len));
+
+ wr_len = CSIO_ROUNDUP(wr_len, 16);
+ wr->flowid_len16 = cpu_to_be32(FW_WR_FLOWID(flow_id) |
+ FW_WR_LEN16(wr_len));
+ wr->els_ct_type = sub_op;
+ wr->ctl_pri = 0;
+ wr->cp_en_class = 0;
+ wr->cookie = io_req->fw_handle;
+ wr->iqid = (uint16_t)cpu_to_be16(csio_q_physiqid(
+ io_req->lnode->hwp, io_req->iq_idx));
+ wr->fl_to_sp = FW_FCOE_ELS_CT_WR_SP(1);
+ wr->tmo_val = (uint8_t) io_req->tmo;
+ port_id = htonl(sid);
+ memcpy(wr->l_id, PORT_ID_PTR(port_id), 3);
+ port_id = htonl(did);
+ memcpy(wr->r_id, PORT_ID_PTR(port_id), 3);
+
+ /* Prepare RSP SGL */
+ wr->rsp_dmalen = cpu_to_be32(io_req->dma_buf.len);
+ wr->rsp_dmaaddr = cpu_to_be64(io_req->dma_buf.paddr);
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_ln_mgmt_submit_wr - Post elsct work request.
+ * @mgmtm - mgmtm
+ * @io_req - io request.
+ * @sub_op - ELS or CT request type
+ * @pld - Dma Payload buffer
+ * @pld_len - Payload len
+ * Prepares ELSCT Work request and sents it to FW.
+ * Returns: CSIO_SUCCESS - on success
+ */
+static csio_retval_t
+csio_ln_mgmt_submit_wr(struct csio_mgmtm *mgmtm, struct csio_ioreq *io_req,
+ uint8_t sub_op, struct csio_dma_buf *pld,
+ uint32_t pld_len)
+{
+ struct csio_wr_pair wrp;
+ struct csio_lnode *ln = io_req->lnode;
+ struct csio_rnode *rn = io_req->rnode;
+ struct csio_hw *hw = mgmtm->hw;
+ uint8_t fw_wr[64];
+ struct ulptx_sgl dsgl;
+ uint32_t wr_size = 0;
+ uint8_t im_len = 0;
+ uint32_t wr_off = 0;
+
+ csio_retval_t ret = CSIO_SUCCESS;
+
+ /* Calculate WR Size for this ELS REQ */
+ wr_size = sizeof(struct fw_fcoe_els_ct_wr);
+
+ /* Send as immediate data if pld < 256 */
+ if (pld_len < 256) {
+ wr_size += ALIGN(pld_len, 8);
+ im_len = (uint8_t)pld_len;
+ } else
+ wr_size += sizeof(struct ulptx_sgl);
+
+ /* Roundup WR size in units of 16 bytes */
+ wr_size = ALIGN(wr_size, 16);
+
+ /* Get WR to send ELS REQ */
+ ret = csio_wr_get(hw, mgmtm->eq_idx, wr_size, &wrp);
+ if (ret != CSIO_SUCCESS) {
+ csio_err(hw, "Failed to get WR for ec_req %p ret:%d\n",
+ io_req, ret);
+ return ret;
+ }
+
+ /* Prepare Generic WR used by all ELS/CT cmd */
+ csio_ln_prep_ecwr(io_req, wr_size, im_len, sub_op,
+ ln->nport_id, rn->nport_id,
+ csio_rn_flowid(rn),
+ &fw_wr[0]);
+
+ /* Copy ELS/CT WR CMD */
+ csio_wr_copy_to_wrp(&fw_wr[0], &wrp, wr_off,
+ sizeof(struct fw_fcoe_els_ct_wr));
+ wr_off += sizeof(struct fw_fcoe_els_ct_wr);
+
+ /* Copy payload to Immediate section of WR */
+ if (im_len)
+ csio_wr_copy_to_wrp(pld->vaddr, &wrp, wr_off, im_len);
+ else {
+ /* Program DSGL to dma payload */
+ dsgl.cmd_nsge = htonl(ULPTX_CMD(ULP_TX_SC_DSGL) |
+ ULPTX_MORE | ULPTX_NSGE(1));
+ dsgl.len0 = cpu_to_be32(pld_len);
+ dsgl.addr0 = cpu_to_be64(pld->paddr);
+ csio_wr_copy_to_wrp(&dsgl, &wrp, ALIGN(wr_off, 8),
+ sizeof(struct ulptx_sgl));
+ }
+
+ /* Issue work request to xmit ELS/CT req to FW */
+ csio_wr_issue(mgmtm->hw, mgmtm->eq_idx, CSIO_FALSE);
+ return ret;
+}
+
+/*
+ * csio_ln_mgmt_submit_req - Submit FCOE Mgmt request.
+ * @io_req - IO Request
+ * @io_cbfn - Completion handler.
+ * @req_type - ELS or CT request type
+ * @pld - Dma Payload buffer
+ * @pld_len - Payload len
+ *
+ *
+ * This API used submit managment ELS/CT request.
+ * This called with hw lock held
+ * Returns: CSIO_SUCCESS - on success
+ * CSIO_NOMEM - on error.
+ */
+static csio_retval_t
+csio_ln_mgmt_submit_req(struct csio_ioreq *io_req,
+ void (*io_cbfn) (struct csio_hw *, struct csio_ioreq *),
+ enum fcoe_cmn_type req_type, struct csio_dma_buf *pld,
+ uint32_t pld_len)
+{
+ struct csio_hw *hw = csio_lnode_to_hw(io_req->lnode);
+ struct csio_mgmtm *mgmtm = csio_hw_to_mgmtm(hw);
+ csio_retval_t rv;
+
+ io_req->io_cbfn = io_cbfn; /* Upper layer callback handler */
+ io_req->fw_handle = (uintptr_t) (io_req);
+ io_req->eq_idx = mgmtm->eq_idx;
+ io_req->iq_idx = mgmtm->iq_idx;
+
+ rv = csio_ln_mgmt_submit_wr(mgmtm, io_req, req_type, pld, pld_len);
+ if (rv == CSIO_SUCCESS) {
+ list_add_tail(&io_req->sm.sm_list, &mgmtm->active_q);
+ mgmtm->stats.n_active++;
+ }
+ return rv;
+}
+
+/*
+ * csio_ln_fdmi_init - FDMI Init entry point.
+ * @ln: lnode
+ */
+static csio_retval_t
+csio_ln_fdmi_init(struct csio_lnode *ln)
+{
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+ struct csio_dma_buf *dma_buf;
+
+ /* Allocate MGMT request required for FDMI */
+ ln->mgmt_req = kzalloc(sizeof(struct csio_ioreq), GFP_KERNEL);
+ if (!ln->mgmt_req) {
+ csio_ln_err(ln, "Failed to alloc ioreq for FDMI\n");
+ csio_inc_stats(hw, n_err_nomem);
+ return CSIO_NOMEM;
+ }
+
+ /* Allocate Dma buffers for FDMI response Payload */
+ dma_buf = &ln->mgmt_req->dma_buf;
+ dma_buf->len = 2048;
+ dma_buf->vaddr = pci_alloc_consistent(hw->pdev, dma_buf->len,
+ &dma_buf->paddr);
+ if (!dma_buf->vaddr) {
+ csio_err(hw, "Failed to alloc DMA buffer for FDMI!\n");
+ kfree(ln->mgmt_req);
+ ln->mgmt_req = NULL;
+ return CSIO_NOMEM;
+ }
+
+ ln->flags |= CSIO_LNF_FDMI_ENABLE;
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_ln_fdmi_exit - FDMI exit entry point.
+ * @ln: lnode
+ */
+static csio_retval_t
+csio_ln_fdmi_exit(struct csio_lnode *ln)
+{
+ struct csio_dma_buf *dma_buf;
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ if (!ln->mgmt_req)
+ return CSIO_SUCCESS;
+
+ dma_buf = &ln->mgmt_req->dma_buf;
+ if (dma_buf->vaddr)
+ pci_free_consistent(hw->pdev, dma_buf->len, dma_buf->vaddr,
+ dma_buf->paddr);
+
+ kfree(ln->mgmt_req);
+ return CSIO_SUCCESS;
+}
+
+int
+csio_scan_done(struct csio_lnode *ln, unsigned long ticks,
+ unsigned long time, unsigned long max_scan_ticks,
+ unsigned long delta_scan_ticks)
+{
+ int rv = 0;
+
+ if (time >= max_scan_ticks)
+ return 1;
+
+ if (!ln->tgt_scan_tick)
+ ln->tgt_scan_tick = ticks;
+
+ if (((ticks - ln->tgt_scan_tick) >= delta_scan_ticks)) {
+ if (!ln->last_scan_ntgts)
+ ln->last_scan_ntgts = ln->n_scsi_tgts;
+ else {
+ if (ln->last_scan_ntgts == ln->n_scsi_tgts)
+ return 1;
+
+ ln->last_scan_ntgts = ln->n_scsi_tgts;
+ }
+ ln->tgt_scan_tick = ticks;
+ }
+ return rv;
+}
+
+/*
+ * csio_notify_lnodes:
+ * @hw: HW module
+ * @note: Notification
+ *
+ * Called from the HW SM to fan out notifications to the
+ * Lnode SM. Since the HW SM is entered with lock held,
+ * there is no need to hold locks here.
+ *
+ */
+void
+csio_notify_lnodes(struct csio_hw *hw, enum csio_ln_notify note)
+{
+ struct list_head *tmp;
+ struct csio_lnode *ln;
+
+ csio_dbg(hw, "Notifying all nodes of event %d\n", note);
+
+ /* Traverse children lnodes list and send evt */
+ list_for_each(tmp, &hw->sln_head) {
+ ln = (struct csio_lnode *) tmp;
+
+ switch (note) {
+ case CSIO_LN_NOTIFY_HWREADY:
+ csio_lnode_start(ln);
+ break;
+
+ case CSIO_LN_NOTIFY_HWRESET:
+ case CSIO_LN_NOTIFY_HWREMOVE:
+ csio_lnode_close(ln);
+ break;
+
+ case CSIO_LN_NOTIFY_HWSTOP:
+ csio_lnode_stop(ln);
+ break;
+
+ default:
+ break;
+
+ }
+ }
+}
+
+/*
+ * csio_disable_lnodes:
+ * @hw: HW module
+ * @portid:port id
+ * @disable: disable/enable flag.
+ * If disable=1, disables all lnode hosted on given physical port.
+ * otherwise enables all the lnodes on given phsysical port.
+ * This routine need to called with hw lock held.
+ */
+void
+csio_disable_lnodes(struct csio_hw *hw, uint8_t portid, bool disable)
+{
+ struct list_head *tmp;
+ struct csio_lnode *ln;
+
+ csio_dbg(hw, "Notifying event to all nodes of port:%d\n", portid);
+
+ /* Traverse sibling lnodes list and send evt */
+ list_for_each(tmp, &hw->sln_head) {
+ ln = (struct csio_lnode *) tmp;
+ if (ln->portid != portid)
+ continue;
+
+ if (disable)
+ csio_lnode_stop(ln);
+ else
+ csio_lnode_start(ln);
+ }
+}
+
+/*
+ * csio_ln_init - Initialize an lnode.
+ * @ln: lnode
+ *
+ */
+static csio_retval_t
+csio_ln_init(struct csio_lnode *ln)
+{
+ int rv = CSIO_INVAL;
+ struct csio_lnode *rln, *pln;
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ csio_init_state(&ln->sm, csio_lns_uninit);
+ ln->vnp_flowid = CSIO_INVALID_IDX;
+ ln->fcf_flowid = CSIO_INVALID_IDX;
+
+ if (csio_is_root_ln(ln)) {
+
+ /* This is the lnode used during initialization */
+
+ ln->fcfinfo = kzalloc(sizeof(struct csio_fcf_info), GFP_KERNEL);
+ if (!ln->fcfinfo) {
+ csio_ln_err(ln, "Failed to alloc FCF record\n");
+ csio_inc_stats(hw, n_err_nomem);
+ goto err;
+ }
+
+ INIT_LIST_HEAD(&ln->fcf_lsthead);
+ kref_init(&ln->fcfinfo->kref);
+
+ if (csio_fdmi_enable && csio_ln_fdmi_init(ln))
+ goto err;
+
+ } else { /* Either a non-root physical or a virtual lnode */
+
+ /*
+ * THe rest is common for non-root physical and NPIV lnodes.
+ * Just get references to all other modules
+ */
+ rln = csio_root_lnode(ln);
+
+ if (csio_is_npiv_ln(ln)) {
+ /* NPIV */
+ pln = csio_parent_lnode(ln);
+ kref_get(&pln->fcfinfo->kref);
+ ln->fcfinfo = pln->fcfinfo;
+ } else {
+ /* Another non-root physical lnode (FCF) */
+ ln->fcfinfo = kzalloc(sizeof(struct csio_fcf_info),
+ GFP_KERNEL);
+ if (!ln->fcfinfo) {
+ csio_ln_err(ln,
+ "Failed to alloc FCF info\n");
+ csio_inc_stats(hw, n_err_nomem);
+ goto err;
+ }
+
+ kref_init(&ln->fcfinfo->kref);
+
+ if (csio_fdmi_enable && csio_ln_fdmi_init(ln))
+ goto err;
+ }
+
+ } /* if (!csio_is_root_ln(ln)) */
+
+ return CSIO_SUCCESS;
+err:
+ return rv;
+}
+
+static void
+csio_ln_exit(struct csio_lnode *ln)
+{
+ struct csio_lnode *pln;
+
+ csio_cleanup_rns(ln);
+ if (csio_is_npiv_ln(ln)) {
+ pln = csio_parent_lnode(ln);
+ kref_put(&pln->fcfinfo->kref, csio_free_fcfinfo);
+ } else {
+ kref_put(&ln->fcfinfo->kref, csio_free_fcfinfo);
+ if (csio_fdmi_enable)
+ csio_ln_fdmi_exit(ln);
+ }
+ ln->fcfinfo = NULL;
+}
+
+/**
+ * csio_lnode_init - Initialize the members of an lnode.
+ * @ln: lnode
+ *
+ */
+csio_retval_t
+csio_lnode_init(struct csio_lnode *ln, struct csio_hw *hw,
+ struct csio_lnode *pln)
+{
+ int rv = -CSIO_INVAL;
+
+ /* Link this lnode to hw */
+ csio_lnode_to_hw(ln) = hw;
+
+ /* Link child to parent if child lnode */
+ if (pln)
+ ln->pln = pln;
+ else
+ ln->pln = NULL;
+
+ /* Initialize scsi_tgt and timers to zero */
+ ln->n_scsi_tgts = 0;
+ ln->last_scan_ntgts = 0;
+ ln->tgt_scan_tick = 0;
+
+ /* Initialize rnode list */
+ INIT_LIST_HEAD(&ln->rnhead);
+ INIT_LIST_HEAD(&ln->cln_head);
+
+ /* Initialize log level for debug */
+ ln->params.log_level = hw->params.log_level;
+
+ if (csio_ln_init(ln))
+ goto err;
+
+ /* Add lnode to list of sibling or children lnodes */
+ spin_lock_irq(&hw->lock);
+ list_add_tail(&ln->sm.sm_list, pln ? &pln->cln_head : &hw->sln_head);
+ if (pln)
+ pln->num_vports++;
+ spin_unlock_irq(&hw->lock);
+
+ hw->num_lns++;
+
+ return CSIO_SUCCESS;
+err:
+ csio_lnode_to_hw(ln) = NULL;
+ return rv;
+}
+
+/**
+ * csio_lnode_exit - De-instantiate an lnode.
+ * @ln: lnode
+ *
+ */
+void
+csio_lnode_exit(struct csio_lnode *ln)
+{
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ csio_ln_exit(ln);
+
+ /* Remove this lnode from hw->sln_head */
+ spin_lock_irq(&hw->lock);
+
+ list_del_init(&ln->sm.sm_list);
+
+ /* If it is children lnode, decrement the
+ * counter in its parent lnode
+ */
+ if (ln->pln)
+ ln->pln->num_vports--;
+
+ /* Update root lnode pointer */
+ if (list_empty(&hw->sln_head))
+ hw->rln = NULL;
+ else
+ hw->rln = (struct csio_lnode *)csio_list_next(&hw->sln_head);
+
+ spin_unlock_irq(&hw->lock);
+
+ csio_lnode_to_hw(ln) = NULL;
+ hw->num_lns--;
+}
diff --git a/drivers/scsi/csiostor/csio_rnode.c b/drivers/scsi/csiostor/csio_rnode.c
new file mode 100644
index 0000000..29e47ac
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_rnode.c
@@ -0,0 +1,889 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/string.h>
+#include <scsi/scsi_transport_fc.h>
+
+#include "csio_hw.h"
+#include "csio_lnode.h"
+#include "csio_rnode.h"
+#include "csio_fcoe_proto.h"
+
+static csio_retval_t csio_rnode_init(struct csio_rnode *, struct csio_lnode *);
+static void csio_rnode_exit(struct csio_rnode *);
+
+/* Static machine forward declarations */
+static void csio_rns_uninit(struct csio_rnode *, enum csio_rn_ev);
+static void csio_rns_ready(struct csio_rnode *, enum csio_rn_ev);
+static void csio_rns_offline(struct csio_rnode *, enum csio_rn_ev);
+static void csio_rns_disappeared(struct csio_rnode *, enum csio_rn_ev);
+
+/* RNF event mapping */
+static enum csio_rn_ev fwevt_to_rnevt[] = {
+ CSIO_RNFE_NONE, /* None */
+ CSIO_RNFE_LOGGED_IN, /* PLOGI_ACC_RCVD */
+ CSIO_RNFE_NONE, /* PLOGI_RJT_RCVD */
+ CSIO_RNFE_PLOGI_RECV, /* PLOGI_RCVD */
+ CSIO_RNFE_LOGO_RECV, /* PLOGO_RCVD */
+ CSIO_RNFE_PRLI_DONE, /* PRLI_ACC_RCVD */
+ CSIO_RNFE_NONE, /* PRLI_RJT_RCVD */
+ CSIO_RNFE_PRLI_RECV, /* PRLI_RCVD */
+ CSIO_RNFE_PRLO_RECV, /* PRLO_RCVD */
+ CSIO_RNFE_NONE, /* NPORT_ID_CHGD */
+ CSIO_RNFE_LOGO_RECV, /* FLOGO_RCVD */
+ CSIO_RNFE_NONE, /* CLR_VIRT_LNK_RCVD */
+ CSIO_RNFE_LOGGED_IN, /* FLOGI_ACC_RCVD */
+ CSIO_RNFE_NONE, /* FLOGI_RJT_RCVD */
+ CSIO_RNFE_LOGGED_IN, /* FDISC_ACC_RCVD */
+ CSIO_RNFE_NONE, /* FDISC_RJT_RCVD */
+ CSIO_RNFE_NONE, /* FLOGI_TMO_MAX_RETRY */
+ CSIO_RNFE_NONE, /* IMPL_LOGO_ADISC_ACC */
+ CSIO_RNFE_NONE, /* IMPL_LOGO_ADISC_RJT */
+ CSIO_RNFE_NONE, /* IMPL_LOGO_ADISC_CNFLT */
+ CSIO_RNFE_NONE, /* PRLI_TMO */
+ CSIO_RNFE_NONE, /* ADISC_TMO */
+ CSIO_RNFE_NAME_MISSING, /* RSCN_DEV_LOST */
+ CSIO_RNFE_NONE, /* SCR_ACC_RCVD */
+ CSIO_RNFE_NONE, /* ADISC_RJT_RCVD */
+ CSIO_RNFE_NONE, /* LOGO_SNT */
+ CSIO_RNFE_LOGO_RECV, /* PROTO_ERR_IMPL_LOGO */
+};
+
+#define CSIO_FWE_TO_RNFE(_evt) ((_evt > PROTO_ERR_IMPL_LOGO) ? \
+ CSIO_RNFE_NONE : \
+ fwevt_to_rnevt[_evt])
+int
+csio_is_rnode_ready(struct csio_rnode *rn)
+{
+ return csio_match_state(rn, csio_rns_ready);
+}
+
+static int
+csio_is_rnode_uninit(struct csio_rnode *rn)
+{
+ return csio_match_state(rn, csio_rns_uninit);
+}
+
+/*
+ * csio_rn_lookup - Finds the rnode with the given flowid
+ * @ln - lnode
+ * @flowid - flowid.
+ *
+ * Does the rnode lookup on the given lnode and flowid.If no matching entry
+ * found, NULL is returned.
+ */
+static struct csio_rnode *
+csio_rn_lookup(struct csio_lnode *ln, uint32_t flowid)
+{
+ struct csio_rnode *rnhead = (struct csio_rnode *) &ln->rnhead;
+ struct list_head *tmp;
+ struct csio_rnode *rn;
+
+ list_for_each(tmp, &rnhead->sm.sm_list) {
+ rn = (struct csio_rnode *) tmp;
+ if (rn->flowid == flowid)
+ return rn;
+ }
+
+ return NULL;
+}
+
+/*
+ * csio_rn_lookup_wwpn - Finds the rnode with the given wwpn
+ * @ln: lnode
+ * @wwpn: wwpn
+ *
+ * Does the rnode lookup on the given lnode and wwpn. If no matching entry
+ * found, NULL is returned.
+ */
+static struct csio_rnode *
+csio_rn_lookup_wwpn(struct csio_lnode *ln, uint8_t *wwpn)
+{
+ struct csio_rnode *rnhead = (struct csio_rnode *) &ln->rnhead;
+ struct list_head *tmp;
+ struct csio_rnode *rn;
+
+ list_for_each(tmp, &rnhead->sm.sm_list) {
+ rn = (struct csio_rnode *) tmp;
+ if (!memcmp(csio_rn_wwpn(rn), wwpn, 8))
+ return rn;
+ }
+
+ return NULL;
+}
+
+/**
+ * csio_rnode_lookup_portid - Finds the rnode with the given portid
+ * @ln: lnode
+ * @portid: port id
+ *
+ * Lookup the rnode list for a given portid. If no matching entry
+ * found, NULL is returned.
+ */
+struct csio_rnode *
+csio_rnode_lookup_portid(struct csio_lnode *ln, uint32_t portid)
+{
+ struct csio_rnode *rnhead = (struct csio_rnode *) &ln->rnhead;
+ struct list_head *tmp;
+ struct csio_rnode *rn;
+
+ list_for_each(tmp, &rnhead->sm.sm_list) {
+ rn = (struct csio_rnode *) tmp;
+ if (rn->nport_id == portid)
+ return rn;
+ }
+
+ return NULL;
+}
+
+static int
+csio_rn_dup_flowid(struct csio_lnode *ln, uint32_t rdev_flowid,
+ uint32_t *vnp_flowid)
+{
+ struct csio_rnode *rnhead;
+ struct list_head *tmp, *tmp1;
+ struct csio_rnode *rn;
+ struct csio_lnode *ln_tmp;
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ list_for_each(tmp1, &hw->sln_head) {
+ ln_tmp = (struct csio_lnode *) tmp1;
+ if (ln_tmp == ln)
+ continue;
+
+ rnhead = (struct csio_rnode *)&ln_tmp->rnhead;
+ list_for_each(tmp, &rnhead->sm.sm_list) {
+
+ rn = (struct csio_rnode *) tmp;
+ if (csio_is_rnode_ready(rn)) {
+ if (rn->flowid == rdev_flowid) {
+ *vnp_flowid = csio_ln_flowid(ln_tmp);
+ return 1;
+ }
+ }
+ }
+ }
+
+ return 0;
+}
+
+static struct csio_rnode *
+csio_alloc_rnode(struct csio_lnode *ln)
+{
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ struct csio_rnode *rn = mempool_alloc(hw->rnode_mempool, GFP_ATOMIC);
+ if (!rn)
+ goto err;
+
+ memset(rn, 0, sizeof(struct csio_rnode));
+ if (csio_rnode_init(rn, ln))
+ goto err_free;
+
+ csio_inc_stats(ln, n_rnode_alloc);
+
+ return rn;
+
+err_free:
+ mempool_free(rn, hw->rnode_mempool);
+err:
+ csio_inc_stats(ln, n_rnode_nomem);
+ return NULL;
+}
+
+static void
+csio_free_rnode(struct csio_rnode *rn)
+{
+ struct csio_hw *hw = csio_lnode_to_hw(csio_rnode_to_lnode(rn));
+
+ csio_rnode_exit(rn);
+ csio_inc_stats(rn->lnp, n_rnode_free);
+ mempool_free(rn, hw->rnode_mempool);
+}
+
+/*
+ * csio_get_rnode - Gets rnode with the given flowid
+ * @ln - lnode
+ * @flowid - flow id.
+ *
+ * Does the rnode lookup on the given lnode and flowid. If no matching
+ * rnode found, then new rnode with given npid is allocated and returned.
+ */
+static struct csio_rnode *
+csio_get_rnode(struct csio_lnode *ln, uint32_t flowid)
+{
+ struct csio_rnode *rn;
+
+ rn = csio_rn_lookup(ln, flowid);
+ if (!rn) {
+ rn = csio_alloc_rnode(ln);
+ if (!rn)
+ return NULL;
+
+ rn->flowid = flowid;
+ }
+
+ return rn;
+}
+
+/*
+ * csio_put_rnode - Frees the given rnode
+ * @ln - lnode
+ * @flowid - flow id.
+ *
+ * Does the rnode lookup on the given lnode and flowid. If no matching
+ * rnode found, then new rnode with given npid is allocated and returned.
+ */
+void
+csio_put_rnode(struct csio_lnode *ln, struct csio_rnode *rn)
+{
+ CSIO_DB_ASSERT(csio_is_rnode_uninit(rn) != 0);
+ csio_free_rnode(rn);
+}
+
+/*
+ * csio_confirm_rnode - confirms rnode based on wwpn.
+ * @ln: lnode
+ * @rdev_flowid: remote device flowid
+ * @rdevp: remote device params
+ * This routines searches other rnode in list having same wwpn of new rnode.
+ * If there is a match, then matched rnode is returned and otherwise new rnode
+ * is returned.
+ * returns rnode.
+ */
+struct csio_rnode *
+csio_confirm_rnode(struct csio_lnode *ln, uint32_t rdev_flowid,
+ struct fcoe_rdev_entry *rdevp)
+{
+ uint8_t rport_type;
+ struct csio_rnode *rn, *match_rn;
+ uint32_t vnp_flowid;
+
+ rport_type =
+ FW_RDEV_WR_RPORT_TYPE_GET(rdevp->rd_xfer_rdy_to_rport_type);
+ /* Drop rdev event for cntrl port */
+ if (rport_type == FAB_CTLR_VNPORT) {
+ csio_ln_dbg(ln,
+ "Unhandled rport_type:%d recv in rdev evt "
+ "ssni:x%x\n", rport_type, rdev_flowid);
+ return NULL;
+ }
+
+ /* Lookup on flowid */
+ rn = csio_rn_lookup(ln, rdev_flowid);
+ if (!rn) {
+
+ /* Drop events with duplicate flowid */
+ if (csio_rn_dup_flowid(ln, rdev_flowid, &vnp_flowid)) {
+ csio_ln_warn(ln,
+ "ssni:%x already active on vnpi:%x",
+ rdev_flowid, vnp_flowid);
+ return NULL;
+ }
+
+ /* skip wwpn lookup for fabric ports, cntrl port */
+ if (rport_type == FLOGI_VFPORT || rport_type == FDISC_VFPORT
+ || rport_type == FAB_CTLR_VNPORT) {
+ goto alloc_rnode;
+ }
+
+ /* Lookup on wwpn for NPORTs */
+ rn = csio_rn_lookup_wwpn(ln, rdevp->wwpn);
+ if (!rn)
+ goto alloc_rnode;
+
+ /* found rn */
+ goto found_rnode;
+ } else {
+ /* verify rnode found for fabric ports, cntrl port */
+ if (rport_type == FLOGI_VFPORT || rport_type == FDISC_VFPORT
+ || rport_type == FAB_CTLR_VNPORT) {
+
+ /* Rnode role mismatch. Allocate new rnode */
+ if (rn->role == CSIO_RNFR_NS ||
+ rn->role == CSIO_RNFR_NPORT) {
+ csio_ln_dbg(ln,
+ "rnode role mismatch found ssni:x%x "
+ "role:%d new_type:%d\n",
+ rdev_flowid, rn->role, rport_type);
+ if (csio_is_rnode_ready(rn)) {
+ csio_ln_warn(ln,
+ "rnode is already"
+ "active ssni:x%x\n",
+ rdev_flowid);
+ CSIO_DB_ASSERT(0);
+ }
+ csio_rn_flowid(rn) = CSIO_INVALID_IDX;
+ goto alloc_rnode;
+ } else
+ goto found_rnode;
+ }
+
+ /* wwpn match */
+ if (!memcmp(csio_rn_wwpn(rn), rdevp->wwpn, 8)) {
+ /* Update rn */
+ goto found_rnode;
+ }
+
+ /* Search for rnode that have same wwpn */
+ match_rn = csio_rn_lookup_wwpn(ln, rdevp->wwpn);
+ if (match_rn != NULL) {
+ csio_ln_dbg(ln,
+ "ssni:x%x changed for rport name(wwpn):%llx "
+ "did:x%x\n", rdev_flowid,
+ wwn_to_u64(rdevp->wwpn),
+ match_rn->nport_id);
+ csio_rn_flowid(rn) = CSIO_INVALID_IDX;
+ rn = match_rn;
+ csio_inc_stats(ln, n_rnode_match);
+ } else {
+ csio_ln_dbg(ln,
+ "rnode wwpn mismatch found ssni:x%x "
+ "name(wwpn):%llx\n",
+ rdev_flowid,
+ wwn_to_u64(csio_rn_wwpn(rn)));
+ if (csio_is_rnode_ready(rn)) {
+ csio_ln_warn(ln,
+ "rnode is already active "
+ "wwpn:%llx ssni:x%x\n",
+ wwn_to_u64(csio_rn_wwpn(rn)),
+ rdev_flowid);
+ CSIO_DB_ASSERT(0);
+ }
+ csio_rn_flowid(rn) = CSIO_INVALID_IDX;
+ goto alloc_rnode;
+ }
+ }
+
+found_rnode:
+ csio_ln_dbg(ln, "found rnode:%p ssni:x%x name(wwpn):%llx\n",
+ rn, rdev_flowid, wwn_to_u64(rdevp->wwpn));
+
+ /* Update flowid */
+ csio_rn_flowid(rn) = rdev_flowid;
+
+ /* update rdev entry */
+ rn->rdev_entry = rdevp;
+ return rn;
+
+alloc_rnode:
+ rn = csio_get_rnode(ln, rdev_flowid);
+ if (!rn)
+ return NULL;
+
+ csio_ln_dbg(ln, "alloc rnode:%p ssni:x%x name(wwpn):%llx\n",
+ rn, rdev_flowid, wwn_to_u64(rdevp->wwpn));
+
+ /* update rdev entry */
+ rn->rdev_entry = rdevp;
+ return rn;
+}
+
+/*
+ * csio_rn_verify_rparams - verify rparams.
+ * @ln: lnode
+ * @rn: rnode
+ * @rdevp: remote device params
+ * returns success if rparams are verified.
+ */
+static csio_retval_t
+csio_rn_verify_rparams(struct csio_lnode *ln, struct csio_rnode *rn,
+ struct fcoe_rdev_entry *rdevp)
+{
+ uint8_t null[8];
+ uint8_t rport_type;
+ uint8_t fc_class;
+ uint32_t *did;
+
+ did = (uint32_t *) &rdevp->r_id[0];
+ rport_type =
+ FW_RDEV_WR_RPORT_TYPE_GET(rdevp->rd_xfer_rdy_to_rport_type);
+ switch (rport_type) {
+ case FLOGI_VFPORT:
+ rn->role = CSIO_RNFR_FABRIC;
+ if (((ntohl(*did) >> 8) & DID_MASK) != FABRIC_DID) {
+ csio_ln_err(ln, "ssni:x%x invalid fabric portid\n",
+ csio_rn_flowid(rn));
+ return CSIO_INVAL;
+ }
+ /* NPIV support */
+ if (FW_RDEV_WR_NPIV_GET(rdevp->vft_to_qos))
+ ln->flags |= CSIO_LNF_NPIVSUPP;
+
+ break;
+
+ case NS_VNPORT:
+ rn->role = CSIO_RNFR_NS;
+ if (((ntohl(*did) >> 8) & DID_MASK) != NS_DID) {
+ csio_ln_err(ln, "ssni:x%x invalid fabric portid\n",
+ csio_rn_flowid(rn));
+ return CSIO_INVAL;
+ }
+ break;
+
+ case REG_FC4_VNPORT:
+ case REG_VNPORT:
+ rn->role = CSIO_RNFR_NPORT;
+ if (rdevp->event_cause == PRLI_ACC_RCVD ||
+ rdevp->event_cause == PRLI_RCVD) {
+ if (FW_RDEV_WR_TASK_RETRY_ID_GET(
+ rdevp->enh_disc_to_tgt))
+ rn->fcp_flags |= V_PRLI_FCP_DATA_OVERLAY(1);
+
+ if (FW_RDEV_WR_RETRY_GET(rdevp->enh_disc_to_tgt))
+ rn->fcp_flags |= V_PRLI_FCP_RETRY(1);
+
+ if (FW_RDEV_WR_CONF_CMPL_GET(rdevp->enh_disc_to_tgt))
+ rn->fcp_flags |=
+ V_PRLI_FCP_CONF_COMPL_ALLOWED(1);
+
+ if (FW_RDEV_WR_TGT_GET(rdevp->enh_disc_to_tgt))
+ rn->role |= CSIO_RNFR_TARGET;
+
+ if (FW_RDEV_WR_INI_GET(rdevp->enh_disc_to_tgt))
+ rn->role |= CSIO_RNFR_INITIATOR;
+ }
+
+ break;
+
+ case FDMI_VNPORT:
+ case FAB_CTLR_VNPORT:
+ rn->role = 0;
+ break;
+
+ default:
+ csio_ln_err(ln, "ssni:x%x invalid rport type recv x%x\n",
+ csio_rn_flowid(rn), rport_type);
+ return CSIO_INVAL;
+ }
+
+ /* validate wwpn/wwnn for Name server/remote port */
+ if (rport_type == REG_VNPORT || rport_type == NS_VNPORT) {
+ memset(null, 0, 8);
+ if (!memcmp(rdevp->wwnn, null, 8)) {
+ csio_ln_err(ln,
+ "ssni:x%x invalid wwnn received from"
+ " rport did:x%x\n",
+ csio_rn_flowid(rn),
+ (ntohl(*did) & DID_MASK));
+ return CSIO_INVAL;
+ }
+
+ if (!memcmp(rdevp->wwpn, null, 8)) {
+ csio_ln_err(ln,
+ "ssni:x%x invalid wwpn received from"
+ " rport did:x%x\n",
+ csio_rn_flowid(rn),
+ (ntohl(*did) & DID_MASK));
+ return CSIO_INVAL;
+ }
+
+ }
+
+ /* Copy wwnn, wwpn and nport id */
+ rn->nport_id = (ntohl(*did) >> 8) & DID_MASK;
+ memcpy(csio_rn_wwnn(rn), rdevp->wwnn, 8);
+ memcpy(csio_rn_wwpn(rn), rdevp->wwpn, 8);
+ rn->rn_sparm.csp.rcv_sz = ntohs(rdevp->rcv_fr_sz);
+ fc_class = FW_RDEV_WR_CLASS_GET(rdevp->vft_to_qos);
+ rn->rn_sparm.clsp[fc_class - 1].serv_option = V_SP_CLASS_SUPPORT(1);
+ return CSIO_SUCCESS;
+}
+
+static void
+__csio_reg_rnode(struct csio_rnode *rn)
+{
+ struct csio_hw *hw = csio_lnode_to_hw(csio_rnode_to_lnode(rn));
+ struct csio_lnode *ln = csio_rnode_to_lnode(rn);
+
+ spin_unlock_irq(&hw->lock);
+ csio_reg_rnode(rn);
+ spin_lock_irq(&hw->lock);
+
+ if (rn->nport_id == FDMI_DID)
+ csio_ln_fdmi_start(ln, (void *) rn);
+}
+
+static void
+__csio_unreg_rnode(struct csio_rnode *rn)
+{
+ struct csio_hw *hw = csio_lnode_to_hw(csio_rnode_to_lnode(rn));
+ LIST_HEAD(tmp_q);
+ int cmpl = 0;
+
+ if (!list_empty(&rn->host_cmpl_q)) {
+ csio_dbg(hw, "Returning completion queue I/Os\n");
+ list_splice_tail_init(&rn->host_cmpl_q, &tmp_q);
+ cmpl = 1;
+ }
+
+ spin_unlock_irq(&hw->lock);
+ csio_unreg_rnode(rn);
+ spin_lock_irq(&hw->lock);
+
+ /* Cleanup I/Os that were waiting for rnode to unregister */
+ if (cmpl)
+ csio_scsi_cleanup_io_q(csio_hw_to_scsim(hw), &tmp_q);
+
+}
+
+/*****************************************************************************/
+/* START: Rnode SM */
+/*****************************************************************************/
+
+/*
+ * csio_rns_uninit -
+ * @rn - rnode
+ * @evt - SM event.
+ *
+ */
+static void
+csio_rns_uninit(struct csio_rnode *rn, enum csio_rn_ev evt)
+{
+ struct csio_lnode *ln = csio_rnode_to_lnode(rn);
+ csio_retval_t ret = CSIO_SUCCESS;
+
+ csio_inc_stats(rn, n_evt_sm[evt]);
+
+ switch (evt) {
+ case CSIO_RNFE_LOGGED_IN:
+ case CSIO_RNFE_PLOGI_RECV:
+ ret = csio_rn_verify_rparams(ln, rn, rn->rdev_entry);
+ if (!ret) {
+ csio_set_state(&rn->sm, csio_rns_ready);
+ __csio_reg_rnode(rn);
+ } else {
+ csio_inc_stats(rn, n_err_inval);
+ }
+ break;
+ case CSIO_RNFE_LOGO_RECV:
+ csio_ln_dbg(ln,
+ "ssni:x%x Ignoring event %d recv "
+ "in rn state[uninit]\n", csio_rn_flowid(rn), evt);
+ csio_inc_stats(rn, n_evt_drop);
+ break;
+ default:
+ csio_ln_dbg(ln,
+ "ssni:x%x unexp event %d recv "
+ "in rn state[uninit]\n", csio_rn_flowid(rn), evt);
+ csio_inc_stats(rn, n_evt_unexp);
+ break;
+ }
+}
+
+/*
+ * csio_rns_ready -
+ * @rn - rnode
+ * @evt - SM event.
+ *
+ */
+static void
+csio_rns_ready(struct csio_rnode *rn, enum csio_rn_ev evt)
+{
+ struct csio_lnode *ln = csio_rnode_to_lnode(rn);
+ csio_retval_t ret = CSIO_SUCCESS;
+
+ csio_inc_stats(rn, n_evt_sm[evt]);
+
+ switch (evt) {
+ case CSIO_RNFE_LOGGED_IN:
+ case CSIO_RNFE_PLOGI_RECV:
+ csio_ln_dbg(ln,
+ "ssni:x%x Ignoring event %d recv from did:x%x "
+ "in rn state[ready]\n", csio_rn_flowid(rn), evt,
+ rn->nport_id);
+ csio_inc_stats(rn, n_evt_drop);
+ break;
+
+ case CSIO_RNFE_PRLI_DONE:
+ case CSIO_RNFE_PRLI_RECV:
+ ret = csio_rn_verify_rparams(ln, rn, rn->rdev_entry);
+ if (!ret)
+ __csio_reg_rnode(rn);
+ else
+ csio_inc_stats(rn, n_err_inval);
+
+ break;
+ case CSIO_RNFE_DOWN:
+ csio_set_state(&rn->sm, csio_rns_offline);
+ __csio_unreg_rnode(rn);
+
+ /* FW expected to internally aborted outstanding SCSI WRs
+ * and return all SCSI WRs to host with status "ABORTED".
+ */
+ break;
+
+ case CSIO_RNFE_LOGO_RECV:
+ csio_set_state(&rn->sm, csio_rns_offline);
+
+ __csio_unreg_rnode(rn);
+
+ /* FW expected to internally aborted outstanding SCSI WRs
+ * and return all SCSI WRs to host with status "ABORTED".
+ */
+ break;
+
+ case CSIO_RNFE_CLOSE:
+ /*
+ * Each rnode receives CLOSE event when driver is removed or
+ * device is reset
+ * Note: All outstanding IOs on remote port need to returned
+ * to uppper layer with appropriate error before sending
+ * CLOSE event
+ */
+ csio_set_state(&rn->sm, csio_rns_uninit);
+ __csio_unreg_rnode(rn);
+ break;
+
+ case CSIO_RNFE_NAME_MISSING:
+ csio_set_state(&rn->sm, csio_rns_disappeared);
+ __csio_unreg_rnode(rn);
+
+ /*
+ * FW expected to internally aborted outstanding SCSI WRs
+ * and return all SCSI WRs to host with status "ABORTED".
+ */
+
+ break;
+
+ default:
+ csio_ln_dbg(ln,
+ "ssni:x%x unexp event %d recv from did:x%x "
+ "in rn state[uninit]\n", csio_rn_flowid(rn), evt,
+ rn->nport_id);
+ csio_inc_stats(rn, n_evt_unexp);
+ break;
+ }
+}
+
+/*
+ * csio_rns_offline -
+ * @rn - rnode
+ * @evt - SM event.
+ *
+ */
+static void
+csio_rns_offline(struct csio_rnode *rn, enum csio_rn_ev evt)
+{
+ struct csio_lnode *ln = csio_rnode_to_lnode(rn);
+ csio_retval_t ret = CSIO_SUCCESS;
+
+ csio_inc_stats(rn, n_evt_sm[evt]);
+
+ switch (evt) {
+ case CSIO_RNFE_LOGGED_IN:
+ case CSIO_RNFE_PLOGI_RECV:
+ ret = csio_rn_verify_rparams(ln, rn, rn->rdev_entry);
+ if (!ret) {
+ csio_set_state(&rn->sm, csio_rns_ready);
+ __csio_reg_rnode(rn);
+ } else {
+ csio_inc_stats(rn, n_err_inval);
+ csio_post_event(&rn->sm, CSIO_RNFE_CLOSE);
+ }
+ break;
+
+ case CSIO_RNFE_DOWN:
+ csio_ln_dbg(ln,
+ "ssni:x%x Ignoring event %d recv from did:x%x "
+ "in rn state[offline]\n", csio_rn_flowid(rn), evt,
+ rn->nport_id);
+ csio_inc_stats(rn, n_evt_drop);
+ break;
+
+ case CSIO_RNFE_CLOSE:
+ /* Each rnode receives CLOSE event when driver is removed or
+ * device is reset
+ * Note: All outstanding IOs on remote port need to returned
+ * to uppper layer with appropriate error before sending
+ * CLOSE event
+ */
+ csio_set_state(&rn->sm, csio_rns_uninit);
+ break;
+
+ case CSIO_RNFE_NAME_MISSING:
+ csio_set_state(&rn->sm, csio_rns_disappeared);
+ break;
+
+ default:
+ csio_ln_dbg(ln,
+ "ssni:x%x unexp event %d recv from did:x%x "
+ "in rn state[offline]\n", csio_rn_flowid(rn), evt,
+ rn->nport_id);
+ csio_inc_stats(rn, n_evt_unexp);
+ break;
+ }
+}
+
+/*
+ * csio_rns_disappeared -
+ * @rn - rnode
+ * @evt - SM event.
+ *
+ */
+static void
+csio_rns_disappeared(struct csio_rnode *rn, enum csio_rn_ev evt)
+{
+ struct csio_lnode *ln = csio_rnode_to_lnode(rn);
+ csio_retval_t ret = CSIO_SUCCESS;
+
+ csio_inc_stats(rn, n_evt_sm[evt]);
+
+ switch (evt) {
+ case CSIO_RNFE_LOGGED_IN:
+ case CSIO_RNFE_PLOGI_RECV:
+ ret = csio_rn_verify_rparams(ln, rn, rn->rdev_entry);
+ if (!ret) {
+ csio_set_state(&rn->sm, csio_rns_ready);
+ __csio_reg_rnode(rn);
+ } else {
+ csio_inc_stats(rn, n_err_inval);
+ csio_post_event(&rn->sm, CSIO_RNFE_CLOSE);
+ }
+ break;
+
+ case CSIO_RNFE_CLOSE:
+ /* Each rnode receives CLOSE event when driver is removed or
+ * device is reset.
+ * Note: All outstanding IOs on remote port need to returned
+ * to uppper layer with appropriate error before sending
+ * CLOSE event
+ */
+ csio_set_state(&rn->sm, csio_rns_uninit);
+ break;
+
+ case CSIO_RNFE_DOWN:
+ case CSIO_RNFE_NAME_MISSING:
+ csio_ln_dbg(ln,
+ "ssni:x%x Ignoring event %d recv from did x%x"
+ "in rn state[disappeared]\n", csio_rn_flowid(rn),
+ evt, rn->nport_id);
+ break;
+
+ default:
+ csio_ln_dbg(ln,
+ "ssni:x%x unexp event %d recv from did x%x"
+ "in rn state[disappeared]\n", csio_rn_flowid(rn),
+ evt, rn->nport_id);
+ csio_inc_stats(rn, n_evt_unexp);
+ break;
+ }
+}
+
+/*****************************************************************************/
+/* END: Rnode SM */
+/*****************************************************************************/
+
+/*
+ * csio_rnode_devloss_handler - Device loss event handler
+ * @rn: rnode
+ *
+ * Post event to close rnode SM and free rnode.
+ */
+void
+csio_rnode_devloss_handler(struct csio_rnode *rn)
+{
+ struct csio_lnode *ln = csio_rnode_to_lnode(rn);
+
+ /* ignore if same rnode came back as online */
+ if (csio_is_rnode_ready(rn))
+ return;
+
+ csio_post_event(&rn->sm, CSIO_RNFE_CLOSE);
+
+ /* Free rn if in uninit state */
+ if (csio_is_rnode_uninit(rn))
+ csio_put_rnode(ln, rn);
+}
+
+/**
+ * csio_rnode_fwevt_handler - Event handler for firmware rnode events.
+ * @rn: rnode
+ *
+ */
+void
+csio_rnode_fwevt_handler(struct csio_rnode *rn, uint8_t fwevt)
+{
+ struct csio_lnode *ln = csio_rnode_to_lnode(rn);
+ enum csio_rn_ev evt;
+
+ evt = CSIO_FWE_TO_RNFE(fwevt);
+ if (!evt) {
+ csio_ln_err(ln, "ssni:x%x Unhandled FW Rdev event: %d\n",
+ csio_rn_flowid(rn), fwevt);
+ csio_inc_stats(rn, n_evt_unexp);
+ return;
+ }
+ csio_inc_stats(rn, n_evt_fw[fwevt]);
+
+ /* Track previous & current events for debugging */
+ rn->prev_evt = rn->cur_evt;
+ rn->cur_evt = fwevt;
+
+ /* Post event to rnode SM */
+ csio_post_event(&rn->sm, evt);
+
+ /* Free rn if in uninit state */
+ if (csio_is_rnode_uninit(rn))
+ csio_put_rnode(ln, rn);
+}
+
+/*
+ * csio_rnode_init - Initialize rnode.
+ * @rn: RNode
+ * @ln: Associated lnode
+ *
+ * Caller is responsible for holding the lock. The lock is required
+ * to be held for inserting the rnode in ln->rnhead list.
+ */
+static csio_retval_t
+csio_rnode_init(struct csio_rnode *rn, struct csio_lnode *ln)
+{
+ csio_rnode_to_lnode(rn) = ln;
+ csio_init_state(&rn->sm, csio_rns_uninit);
+ INIT_LIST_HEAD(&rn->host_cmpl_q);
+ csio_rn_flowid(rn) = CSIO_INVALID_IDX;
+
+ /* Add rnode to list of lnodes->rnhead */
+ list_add_tail(&rn->sm.sm_list, &ln->rnhead);
+
+ return CSIO_SUCCESS;
+}
+
+static void
+csio_rnode_exit(struct csio_rnode *rn)
+{
+ list_del_init(&rn->sm.sm_list);
+ CSIO_DB_ASSERT(list_empty(&rn->host_cmpl_q));
+}
--
1.7.1
^ permalink raw reply related
* [PATCH 2/8] csiostor: Chelsio FCoE offload driver submission (sources part 2).
From: Naresh Kumar Inna @ 2012-08-23 22:27 UTC (permalink / raw)
To: JBottomley, linux-scsi, dm; +Cc: netdev, naresh, chethan
In-Reply-To: <1345760873-12101-1-git-send-email-naresh@chelsio.com>
This patch contains code for driver initialization, driver resource
allocation, module parameters and the Work Request module functionality.
Driver initialization includes module entry/exit points, registration with
PCI, FC transport and SCSI mid layer subsystems. The Work Request module
provides services for allocation of DMA queues, posting Work Requests on
them and processing completions.
Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
---
drivers/scsi/csiostor/csio_init.c | 1392 +++++++++++++++++++++++++++++++
drivers/scsi/csiostor/csio_wr.c | 1634 +++++++++++++++++++++++++++++++++++++
2 files changed, 3026 insertions(+), 0 deletions(-)
create mode 100644 drivers/scsi/csiostor/csio_init.c
create mode 100644 drivers/scsi/csiostor/csio_wr.c
diff --git a/drivers/scsi/csiostor/csio_init.c b/drivers/scsi/csiostor/csio_init.c
new file mode 100644
index 0000000..9be85fe
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_init.c
@@ -0,0 +1,1392 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/aer.h>
+#include <linux/mm.h>
+#include <linux/notifier.h>
+#include <linux/kdebug.h>
+#include <linux/version.h>
+#include <linux/seq_file.h>
+#include <linux/debugfs.h>
+#include <linux/string.h>
+#include <linux/export.h>
+
+#include "csio_init.h"
+#include "csio_defs.h"
+
+#define CSIO_MIN_MEMPOOL_SZ 64
+
+static int pf_counter;
+
+#define csio_param_check(name, def, min, max, typ) \
+static int \
+csio_##name##_check(typ val) \
+{ \
+ if (val >= min && val <= max) { \
+ csio_##name = val; \
+ return 0; \
+ } \
+ \
+ pr_warn("Setting csio_"#name" to default of "#def"\n"); \
+ csio_##name = def; \
+ return -EINVAL; \
+}
+
+#define CSIO_MODULE_PARAM(name, def, min, max, desc, typ) \
+module_param(csio_##name, typ, S_IRUGO|S_IWUSR); \
+MODULE_PARM_DESC(csio_##name, desc); \
+csio_param_check(name, def, min, max, typ)
+
+/*
+ * Normally we're willing to become the firmware's Master PF but will be happy
+ * if another PF has already become the Master. Setting "force_master" will
+ * cause this driver to forcibly establish itself as the Master PF.
+ */
+CSIO_MODULE_PARAM(force_master, 0, 0, 1,
+ " Forcibly become Master PF. 0 - No force, 1 - Force", int);
+
+/* Module parameter - Port vector mask for linkup */
+CSIO_MODULE_PARAM(port_mask, 0xf, 0x1, 0xf,
+ " Port vector mask. (4bits) Default - 0xf (all ports)", uint);
+
+/* Module parameter - Default Log level */
+CSIO_MODULE_PARAM(dbg_level, 0xFEFF, 0, 0xFFFF,
+ " Debug level. Default - 0xFEFF. Module[B15:B8]:Level[B7:B0]",
+ int);
+
+/* Module parameter - lun qdepth */
+CSIO_MODULE_PARAM(lun_qdepth, CSIO_MAX_CMD_PER_LUN, 1, 32,
+ " Lun queue depth (1-32), Default - 32", int);
+
+/* Module parameter - Total number of FCoE rnodes */
+CSIO_MODULE_PARAM(fcoe_rnodes, 512, 8, CSIO_FCOE_MAX_RNODES,
+ " Supported number of FCoE rnodes."
+ " Default - 512, Max - 2048.", int);
+
+/*
+ * Module parameter - Interrupt mode
+ * The driver uses the best interrupt scheme available on a platform in the
+ * order MSI-X, MSI, legacy INTx interrupts. This parameter determines
+ * which
+ * of these schemes the driver may consider as follows:
+ *
+ * csio_msi = 2: choose from among all three options
+ * csio_msi = 1: only consider MSI and INTx interrupts
+ * csio_msi = 0: force INTx interrupts
+ */
+CSIO_MODULE_PARAM(msi, 2, 0, 2, " whether to use MSI-X, MSI or INTx", int);
+
+/* Module parameter - Interrupt coalesce count */
+CSIO_MODULE_PARAM(intr_coalesce_cnt, 0, 0, 16,
+ " Interrupt coalesce count, Disable:0, Range:1-16, Default:0",
+ int);
+
+/* Module parameter - Interrupt coalesce time */
+CSIO_MODULE_PARAM(intr_coalesce_time, 10, 5, 200,
+ " Interrupt coalesce time (us), Min:5, Max:200,"
+ " Default: 10", int);
+
+/* Module parameter - Max scan timeout*/
+CSIO_MODULE_PARAM(max_scan_tmo, 0, 0, 30,
+ " Maximum time (seconds) to wait before declaring scan done.",
+ uint);
+
+/* Module parameter - Delta scan timeout*/
+CSIO_MODULE_PARAM(delta_scan_tmo, 5, 2, 5,
+ " Time (seconds) to wait between two FW scan events."
+ " Default - 5 seconds .", uint);
+
+/* Module parameter - Max SCSI IO request */
+CSIO_MODULE_PARAM(scsi_ioreqs, 2048, 64, 2048,
+ " Max IO req elements per function. Default - 2048.", int);
+
+/* Module parameter - FDMI support */
+CSIO_MODULE_PARAM(fdmi_enable, 1, 0, 1,
+ " fdmi support 1:enable 0:disable Default - 1", int);
+
+static struct dentry *csio_debugfs_root;
+
+static struct scsi_transport_template *csio_fcoe_transport;
+static struct scsi_transport_template *csio_fcoe_transport_vport;
+
+static void
+csio_module_params_check(void)
+{
+ csio_dbg_level_check(csio_dbg_level);
+ csio_force_master_check(csio_force_master);
+ csio_port_mask_check(csio_port_mask);
+ csio_lun_qdepth_check(csio_lun_qdepth);
+ csio_fcoe_rnodes_check(csio_fcoe_rnodes);
+ csio_msi_check(csio_msi);
+ csio_intr_coalesce_cnt_check(csio_intr_coalesce_cnt);
+ csio_intr_coalesce_time_check(csio_intr_coalesce_time);
+ csio_max_scan_tmo_check(csio_max_scan_tmo);
+ csio_delta_scan_tmo_check(csio_delta_scan_tmo);
+ csio_scsi_ioreqs_check(csio_scsi_ioreqs);
+ csio_fdmi_enable_check(csio_fdmi_enable);
+}
+
+/*
+ * debugfs support
+ */
+static int
+csio_mem_open(struct inode *inode, struct file *file)
+{
+ file->private_data = inode->i_private;
+ return 0;
+}
+
+static ssize_t
+csio_mem_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
+{
+ loff_t pos = *ppos;
+ loff_t avail = file->f_path.dentry->d_inode->i_size;
+ unsigned int mem = (uintptr_t)file->private_data & 3;
+ struct csio_hw *hw = file->private_data - mem;
+
+ if (pos < 0)
+ return -EINVAL;
+ if (pos >= avail)
+ return 0;
+ if (count > avail - pos)
+ count = avail - pos;
+
+ while (count) {
+ size_t len;
+ int ret, ofst;
+ __be32 data[16];
+
+ if (mem == MEM_MC)
+ ret = csio_hw_mc_read(hw, pos, data, NULL);
+ else
+ ret = csio_hw_edc_read(hw, mem, pos, data, NULL);
+ if (ret)
+ return ret;
+
+ ofst = pos % sizeof(data);
+ len = min(count, sizeof(data) - ofst);
+ if (copy_to_user(buf, (u8 *)data + ofst, len))
+ return -EFAULT;
+
+ buf += len;
+ pos += len;
+ count -= len;
+ }
+ count = pos - *ppos;
+ *ppos = pos;
+ return count;
+}
+
+static const struct file_operations csio_mem_debugfs_fops = {
+ .owner = THIS_MODULE,
+ .open = csio_mem_open,
+ .read = csio_mem_read,
+ .llseek = default_llseek,
+};
+
+static void __devinit
+csio_add_debugfs_mem(struct csio_hw *hw, const char *name,
+ unsigned int idx, unsigned int size_mb)
+{
+ struct dentry *de;
+
+ de = debugfs_create_file(name, S_IRUSR, hw->debugfs_root,
+ (void *)hw + idx, &csio_mem_debugfs_fops);
+ if (de && de->d_inode)
+ de->d_inode->i_size = size_mb << 20;
+}
+
+static int __devinit
+csio_setup_debugfs(struct csio_hw *hw)
+{
+ int i;
+
+ if (IS_ERR_OR_NULL(hw->debugfs_root))
+ return -1;
+
+ i = csio_rd_reg32(hw, MA_TARGET_MEM_ENABLE);
+ if (i & EDRAM0_ENABLE)
+ csio_add_debugfs_mem(hw, "edc0", MEM_EDC0, 5);
+ if (i & EDRAM1_ENABLE)
+ csio_add_debugfs_mem(hw, "edc1", MEM_EDC1, 5);
+ if (i & EXT_MEM_ENABLE)
+ csio_add_debugfs_mem(hw, "mc", MEM_MC,
+ EXT_MEM_SIZE_GET(csio_rd_reg32(hw, MA_EXT_MEMORY_BAR)));
+ return 0;
+}
+
+/*
+ * csio_dfs_create - Creates debug filesystem and proc fs for the given HW.
+ *
+ */
+static int
+csio_dfs_create(struct csio_hw *hw)
+{
+ if (csio_debugfs_root) {
+ hw->debugfs_root = debugfs_create_dir(pci_name(hw->pdev),
+ csio_debugfs_root);
+ csio_setup_debugfs(hw);
+ }
+
+ return 0;
+}
+
+/*
+ * csio_dfs_destroy - Deletes debugfs and procfs entries for the given HW.
+ *
+ */
+static int
+csio_dfs_destroy(struct csio_hw *hw)
+{
+ if (hw->debugfs_root)
+ debugfs_remove_recursive(hw->debugfs_root);
+
+ return 0;
+}
+
+/*
+ * csio_dfs_init - Debug filesystem initialization.
+ *
+ * This is function is called during driver load to initialize debugfs, procfs
+ * used for debugging.
+ */
+static int
+csio_dfs_init(void)
+{
+ csio_debugfs_root = debugfs_create_dir(KBUILD_MODNAME, NULL);
+ if (!csio_debugfs_root)
+ pr_warn("Could not create debugfs entry, continuing\n");
+
+ return 0;
+}
+
+/*
+ * csio_dfs_exit - Cleans up debugfs and procfs created during driver load.
+ * Function that gets called in the unload path.
+ */
+static void
+csio_dfs_exit(void)
+{
+ debugfs_remove(csio_debugfs_root);
+}
+
+/*
+ * csio_pci_init - PCI initialization.
+ * @pdev: PCI device.
+ * @bars: Bitmask of bars to be requested.
+ *
+ * Initializes the PCI function by enabling MMIO, setting bus
+ * mastership and setting DMA mask.
+ */
+static int
+csio_pci_init(struct pci_dev *pdev, int *bars)
+{
+ int rv = -ENODEV;
+
+ *bars = pci_select_bars(pdev, IORESOURCE_MEM);
+
+ if (pci_enable_device_mem(pdev))
+ goto err;
+
+ if (pci_request_selected_regions(pdev, *bars, KBUILD_MODNAME))
+ goto err_disable_device;
+
+ pci_set_master(pdev);
+ pci_try_set_mwi(pdev);
+
+ if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(64))) {
+ pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(64));
+ } else if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(32))) {
+ pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(32));
+ } else {
+ dev_err(&pdev->dev, "No suitable DMA available.\n");
+ goto err_release_regions;
+ }
+
+ return 0;
+
+err_release_regions:
+ pci_release_selected_regions(pdev, *bars);
+err_disable_device:
+ pci_disable_device(pdev);
+err:
+ return rv;
+
+}
+
+/*
+ * csio_pci_exit - PCI unitialization.
+ * @pdev: PCI device.
+ * @bars: Bars to be released.
+ *
+ */
+static void
+csio_pci_exit(struct pci_dev *pdev, int *bars)
+{
+ pci_release_selected_regions(pdev, *bars);
+ pci_disable_device(pdev);
+}
+
+/*
+ * csio_hw_init_workers - Initialize the HW module's worker threads.
+ * @hw: HW module.
+ *
+ */
+static void
+csio_hw_init_workers(struct csio_hw *hw)
+{
+ INIT_WORK(&hw->evtq_work, csio_evtq_worker);
+}
+
+static void
+csio_hw_exit_workers(struct csio_hw *hw)
+{
+ cancel_work_sync(&hw->evtq_work);
+ flush_scheduled_work();
+}
+
+static csio_retval_t
+csio_create_queues(struct csio_hw *hw)
+{
+ int i, j;
+ struct csio_mgmtm *mgmtm = csio_hw_to_mgmtm(hw);
+ csio_retval_t rv;
+ struct csio_scsi_cpu_info *info;
+
+ if (hw->flags & CSIO_HWF_Q_FW_ALLOCED)
+ return CSIO_SUCCESS;
+
+ if (hw->intr_mode != CSIO_IM_MSIX) {
+ rv = csio_wr_iq_create(hw, NULL, hw->intr_iq_idx,
+ 0, hw->pport[0].portid,
+ CSIO_FALSE, NULL);
+ if (rv != CSIO_SUCCESS) {
+ csio_err(hw, " Forward Interrupt IQ failed!: %d\n", rv);
+ return rv;
+ }
+ }
+
+ /* FW event queue */
+ rv = csio_wr_iq_create(hw, NULL, hw->fwevt_iq_idx,
+ csio_get_fwevt_intr_idx(hw),
+ hw->pport[0].portid, CSIO_TRUE, NULL);
+ if (rv != CSIO_SUCCESS) {
+ csio_err(hw, "FW event IQ config failed!: %d\n", rv);
+ return rv;
+ }
+
+ /* Create mgmt queue */
+ rv = csio_wr_eq_create(hw, NULL, mgmtm->eq_idx,
+ mgmtm->iq_idx, hw->pport[0].portid, NULL);
+
+ if (rv != CSIO_SUCCESS) {
+ csio_err(hw, "Mgmt EQ create failed!: %d\n", rv);
+ goto err;
+ }
+
+ /* Create SCSI queues */
+ for (i = 0; i < hw->num_pports; i++) {
+ info = &hw->scsi_cpu_info[i];
+
+ for (j = 0; j < info->max_cpus; j++) {
+ struct csio_scsi_qset *sqset = &hw->sqset[i][j];
+
+ rv = csio_wr_iq_create(hw, NULL, sqset->iq_idx,
+ sqset->intr_idx, i,
+ CSIO_FALSE, NULL);
+ if (rv != CSIO_SUCCESS) {
+ csio_err(hw,
+ "SCSI module IQ config failed [%d][%d]:%d\n",
+ i, j, rv);
+ goto err;
+ }
+ rv = csio_wr_eq_create(hw, NULL, sqset->eq_idx,
+ sqset->iq_idx, i, NULL);
+ if (rv != CSIO_SUCCESS) {
+ csio_err(hw,
+ "SCSI module EQ config failed [%d][%d]:%d\n",
+ i, j, rv);
+ goto err;
+ }
+ } /* for all CPUs */
+ } /* For all ports */
+
+ hw->flags |= CSIO_HWF_Q_FW_ALLOCED;
+ return CSIO_SUCCESS;
+err:
+ csio_wr_destroy_queues(hw, CSIO_TRUE);
+ return CSIO_INVAL;
+}
+
+/*
+ * csio_config_queues - Configure the DMA queues.
+ * @hw: HW module.
+ *
+ * Allocates memory for queues are registers them with FW.
+ */
+csio_retval_t
+csio_config_queues(struct csio_hw *hw)
+{
+ int i, j, idx, k = 0;
+ csio_retval_t rv;
+ struct csio_scsi_qset *sqset;
+ struct csio_mgmtm *mgmtm = csio_hw_to_mgmtm(hw);
+ struct csio_scsi_qset *orig;
+ struct csio_scsi_cpu_info *info;
+
+ if (hw->flags & CSIO_HWF_Q_MEM_ALLOCED)
+ return csio_create_queues(hw);
+
+ /* Calculate number of SCSI queues for MSIX we would like */
+ hw->num_scsi_msix_cpus = num_online_cpus();
+ hw->num_sqsets = num_online_cpus() * hw->num_pports;
+
+ if (hw->num_sqsets > CSIO_MAX_SCSI_QSETS) {
+ hw->num_sqsets = CSIO_MAX_SCSI_QSETS;
+ hw->num_scsi_msix_cpus = CSIO_MAX_SCSI_CPU;
+ }
+
+ /* Initialize max_cpus, may get reduced during msix allocations */
+ for (i = 0; i < hw->num_pports; i++)
+ hw->scsi_cpu_info[i].max_cpus = hw->num_scsi_msix_cpus;
+
+ csio_dbg(hw, "nsqsets:%d scpus:%d\n",
+ hw->num_sqsets, hw->num_scsi_msix_cpus);
+
+ csio_intr_enable(hw);
+
+ if (hw->intr_mode != CSIO_IM_MSIX) {
+
+ /* Allocate Forward interrupt iq. */
+ hw->intr_iq_idx = csio_wr_alloc_q(hw, CSIO_INTR_IQSIZE,
+ CSIO_INTR_WRSIZE, CSIO_INGRESS,
+ (void *)hw, 0, 0, NULL);
+ if (hw->intr_iq_idx == -1) {
+ csio_err(hw,
+ "Forward interrupt queue creation failed\n");
+ goto intr_disable;
+ }
+ }
+
+ /* Allocate the FW evt queue */
+ hw->fwevt_iq_idx = csio_wr_alloc_q(hw, CSIO_FWEVT_IQSIZE,
+ CSIO_FWEVT_WRSIZE,
+ CSIO_INGRESS, (void *)hw,
+ CSIO_FWEVT_FLBUFS, 0,
+ csio_fwevt_intx_handler);
+ if (hw->fwevt_iq_idx == -1) {
+ csio_err(hw, "FW evt queue creation failed\n");
+ goto intr_disable;
+ }
+
+ /* Allocate the mgmt queue */
+ mgmtm->eq_idx = csio_wr_alloc_q(hw, CSIO_MGMT_EQSIZE,
+ CSIO_MGMT_EQ_WRSIZE,
+ CSIO_EGRESS, (void *)hw, 0, 0, NULL);
+ if (mgmtm->eq_idx == -1) {
+ csio_err(hw, "Failed to alloc egress queue for mgmt module\n");
+ goto intr_disable;
+ }
+
+ /* Use FW IQ for MGMT req completion */
+ mgmtm->iq_idx = hw->fwevt_iq_idx;
+
+ /* Allocate SCSI queues */
+ for (i = 0; i < hw->num_pports; i++) {
+ info = &hw->scsi_cpu_info[i];
+
+ for (j = 0; j < hw->num_scsi_msix_cpus; j++) {
+ sqset = &hw->sqset[i][j];
+
+ if (j >= info->max_cpus) {
+ k = j % info->max_cpus;
+ orig = &hw->sqset[i][k];
+ sqset->eq_idx = orig->eq_idx;
+ sqset->iq_idx = orig->iq_idx;
+ continue;
+ }
+
+ idx = csio_wr_alloc_q(hw, csio_scsi_eqsize, 0,
+ CSIO_EGRESS, (void *)hw, 0, 0,
+ NULL);
+ if (idx == -1) {
+ csio_err(hw, "EQ creation failed for idx:%d\n",
+ idx);
+ goto intr_disable;
+ }
+
+ sqset->eq_idx = idx;
+
+ idx = csio_wr_alloc_q(hw, CSIO_SCSI_IQSIZE,
+ CSIO_SCSI_IQ_WRSZ, CSIO_INGRESS,
+ (void *)hw, 0, 0,
+ csio_scsi_intx_handler);
+ if (idx == -1) {
+ csio_err(hw, "IQ creation failed for idx:%d\n",
+ idx);
+ goto intr_disable;
+ }
+ sqset->iq_idx = idx;
+ } /* for all CPUs */
+ } /* For all ports */
+
+ hw->flags |= CSIO_HWF_Q_MEM_ALLOCED;
+
+ rv = csio_create_queues(hw);
+ if (rv != CSIO_SUCCESS)
+ goto intr_disable;
+
+ /*
+ * Now request IRQs for the vectors. In the event of a failure,
+ * cleanup is handled internally by this function.
+ */
+ rv = csio_request_irqs(hw);
+ if (rv != CSIO_SUCCESS)
+ return CSIO_INVAL;
+
+ return CSIO_SUCCESS;
+
+intr_disable:
+ csio_intr_disable(hw, CSIO_FALSE);
+
+ return CSIO_INVAL;
+}
+
+static int
+csio_resource_alloc(struct csio_hw *hw)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ int rv = -ENOMEM;
+
+ wrm->num_q = ((CSIO_MAX_SCSI_QSETS * 2) + CSIO_HW_NIQ +
+ CSIO_HW_NEQ + CSIO_HW_NFLQ + CSIO_HW_NINTXQ);
+
+ hw->mb_mempool = mempool_create_kmalloc_pool(CSIO_MIN_MEMPOOL_SZ,
+ sizeof(struct csio_mb));
+ if (!hw->mb_mempool)
+ goto err;
+
+ hw->rnode_mempool = mempool_create_kmalloc_pool(CSIO_MIN_MEMPOOL_SZ,
+ sizeof(struct csio_rnode));
+ if (!hw->rnode_mempool)
+ goto err_free_mb_mempool;
+
+ hw->scsi_pci_pool = pci_pool_create("csio_scsi_pci_pool", hw->pdev,
+ sizeof(struct csio_fcp_resp), 8, 0);
+ if (!hw->scsi_pci_pool)
+ goto err_free_rn_pool;
+
+ return 0;
+
+err_free_rn_pool:
+ mempool_destroy(hw->rnode_mempool);
+ hw->rnode_mempool = NULL;
+err_free_mb_mempool:
+ mempool_destroy(hw->mb_mempool);
+ hw->mb_mempool = NULL;
+err:
+ return rv;
+}
+
+static void
+csio_resource_free(struct csio_hw *hw)
+{
+ pci_pool_destroy(hw->scsi_pci_pool);
+ hw->scsi_pci_pool = NULL;
+ mempool_destroy(hw->rnode_mempool);
+ hw->rnode_mempool = NULL;
+ mempool_destroy(hw->mb_mempool);
+ hw->mb_mempool = NULL;
+}
+
+/*
+ * csio_hw_alloc - Allocate and initialize the HW module.
+ * @pdev: PCI device.
+ *
+ * Allocates HW structure, DMA, memory resources, maps BARS to
+ * host memory and initializes HW module.
+ */
+static struct csio_hw * __devinit
+csio_hw_alloc(struct pci_dev *pdev)
+{
+ struct csio_hw *hw;
+
+ hw = kzalloc(sizeof(struct csio_hw), GFP_KERNEL);
+ if (!hw)
+ goto err;
+
+ hw->pdev = pdev;
+ strncpy(hw->drv_version, CSIO_DRV_VERSION, 32);
+
+ /* memory pool/DMA pool allocation */
+ if (csio_resource_alloc(hw))
+ goto err_free_hw;
+
+ /* Get the start address of registers from BAR 0 */
+ hw->regstart = ioremap_nocache(pci_resource_start(pdev, 0),
+ pci_resource_len(pdev, 0));
+ if (!hw->regstart) {
+ csio_err(hw, "Could not map BAR 0, regstart = %p\n",
+ hw->regstart);
+ goto err_resource_free;
+ }
+
+ csio_hw_init_workers(hw);
+
+ if (csio_hw_init(hw))
+ goto err_unmap_bar;
+
+ csio_dfs_create(hw);
+
+ csio_dbg(hw, "hw:%p\n", hw);
+
+ return hw;
+
+err_unmap_bar:
+ csio_hw_exit_workers(hw);
+ iounmap(hw->regstart);
+err_resource_free:
+ csio_resource_free(hw);
+err_free_hw:
+ kfree(hw);
+err:
+ return NULL;
+}
+
+/*
+ * csio_hw_free - Uninitialize and free the HW module.
+ * @hw: The HW module
+ *
+ * Disable interrupts, uninit the HW module, free resources, free hw.
+ */
+static void
+csio_hw_free(struct csio_hw *hw)
+{
+ csio_intr_disable(hw, CSIO_TRUE);
+ csio_hw_exit_workers(hw);
+ csio_hw_exit(hw);
+ iounmap(hw->regstart);
+ csio_dfs_destroy(hw);
+ csio_resource_free(hw);
+ kfree(hw);
+}
+
+/**
+ * csio_shost_init - Create and initialize the lnode module.
+ * @hw: The HW module.
+ * @dev: The device associated with this invocation.
+ * @probe: Called from probe context or not?
+ * @os_pln: Parent lnode if any.
+ *
+ * Allocates lnode structure via scsi_host_alloc, initializes
+ * shost, initializes lnode module and registers with SCSI ML
+ * via scsi_host_add. This function is shared between physical and
+ * virtual node ports.
+ */
+struct csio_lnode *
+csio_shost_init(struct csio_hw *hw, struct device *dev,
+ bool probe, struct csio_lnode *pln)
+{
+ struct Scsi_Host *shost = NULL;
+ struct csio_lnode *ln;
+
+ csio_fcoe_shost_template.cmd_per_lun = csio_lun_qdepth;
+ csio_fcoe_shost_vport_template.cmd_per_lun = csio_lun_qdepth;
+
+ /*
+ * hw->pdev is the physical port's PCI dev structure,
+ * which will be different from the NPIV dev structure.
+ */
+ if (dev == &hw->pdev->dev)
+ shost = scsi_host_alloc(
+ &csio_fcoe_shost_template,
+ sizeof(struct csio_lnode));
+ else
+ shost = scsi_host_alloc(
+ &csio_fcoe_shost_vport_template,
+ sizeof(struct csio_lnode));
+
+ if (!shost)
+ goto err;
+
+ ln = shost_priv(shost);
+ memset(ln, 0, sizeof(struct csio_lnode));
+
+ /* Link common lnode to this lnode */
+ ln->dev_num = (shost->host_no << 16);
+
+ shost->can_queue = CSIO_MAX_QUEUE;
+ shost->this_id = -1;
+ shost->unique_id = shost->host_no;
+ shost->max_cmd_len = 16; /* Max CDB length supported */
+ shost->max_id = min_t(uint32_t, csio_fcoe_rnodes,
+ hw->fres_info.max_ssns);
+ shost->max_lun = CSIO_MAX_LUN;
+ if (dev == &hw->pdev->dev)
+ shost->transportt = csio_fcoe_transport;
+ else
+ shost->transportt = csio_fcoe_transport_vport;
+
+ /* root lnode */
+ if (!hw->rln)
+ hw->rln = ln;
+
+ /* Other initialization here: Common, Transport specific */
+ if (csio_lnode_init(ln, hw, pln))
+ goto err_shost_put;
+
+ if (scsi_add_host(shost, dev))
+ goto err_lnode_exit;
+
+ return ln;
+
+err_lnode_exit:
+ csio_lnode_exit(ln);
+err_shost_put:
+ scsi_host_put(shost);
+err:
+ return NULL;
+}
+
+/**
+ * csio_shost_exit - De-instantiate the shost.
+ * @ln: The lnode module corresponding to the shost.
+ *
+ */
+void
+csio_shost_exit(struct csio_lnode *ln)
+{
+ struct Scsi_Host *shost = csio_ln_to_shost(ln);
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ /* Inform transport */
+ fc_remove_host(shost);
+
+ /* Inform SCSI ML */
+ scsi_remove_host(shost);
+
+ /* Flush all the events, so that any rnode removal events
+ * already queued are all handled, before we remove the lnode.
+ */
+ spin_lock_irq(&hw->lock);
+ csio_evtq_flush(hw);
+ spin_unlock_irq(&hw->lock);
+
+ csio_lnode_exit(ln);
+ scsi_host_put(shost);
+}
+
+struct csio_lnode *
+csio_lnode_alloc(struct csio_hw *hw)
+{
+ return csio_shost_init(hw, &hw->pdev->dev, CSIO_FALSE, NULL);
+}
+
+void
+csio_lnodes_block_request(struct csio_hw *hw)
+{
+ struct Scsi_Host *shost;
+ struct csio_lnode *sln;
+ struct csio_lnode *ln;
+ struct list_head *cur_ln, *cur_cln;
+ struct csio_lnode **lnode_list;
+ int cur_cnt = 0, ii;
+
+ lnode_list = kzalloc((sizeof(struct csio_lnode *) * hw->num_lns),
+ GFP_KERNEL);
+ if (!lnode_list) {
+ csio_err(hw, "Failed to allocate lnodes_list");
+ return;
+ }
+
+ spin_lock_irq(&hw->lock);
+ /* Traverse sibling lnodes */
+ list_for_each(cur_ln, &hw->sln_head) {
+ sln = (struct csio_lnode *) cur_ln;
+ lnode_list[cur_cnt++] = sln;
+
+ /* Traverse children lnodes */
+ list_for_each(cur_cln, &sln->cln_head)
+ lnode_list[cur_cnt++] = (struct csio_lnode *) cur_cln;
+ }
+ spin_unlock_irq(&hw->lock);
+
+ for (ii = 0; ii < cur_cnt; ii++) {
+ csio_dbg(hw, "Blocking IOs on lnode: %p\n", lnode_list[ii]);
+ ln = lnode_list[ii];
+ shost = csio_ln_to_shost(ln);
+ scsi_block_requests(shost);
+
+ }
+ kfree(lnode_list);
+}
+
+void
+csio_lnodes_unblock_request(struct csio_hw *hw)
+{
+ struct csio_lnode *ln;
+ struct Scsi_Host *shost;
+ struct csio_lnode *sln;
+ struct list_head *cur_ln, *cur_cln;
+ struct csio_lnode **lnode_list;
+ int cur_cnt = 0, ii;
+
+ lnode_list = kzalloc((sizeof(struct csio_lnode *) * hw->num_lns),
+ GFP_KERNEL);
+ if (!lnode_list) {
+ csio_err(hw, "Failed to allocate lnodes_list");
+ return;
+ }
+
+ spin_lock_irq(&hw->lock);
+ /* Traverse sibling lnodes */
+ list_for_each(cur_ln, &hw->sln_head) {
+ sln = (struct csio_lnode *) cur_ln;
+ lnode_list[cur_cnt++] = sln;
+
+ /* Traverse children lnodes */
+ list_for_each(cur_cln, &sln->cln_head)
+ lnode_list[cur_cnt++] = (struct csio_lnode *) cur_cln;
+ }
+ spin_unlock_irq(&hw->lock);
+
+ for (ii = 0; ii < cur_cnt; ii++) {
+ csio_dbg(hw, "unblocking IOs on lnode: %p\n", lnode_list[ii]);
+ ln = lnode_list[ii];
+ shost = csio_ln_to_shost(ln);
+ scsi_unblock_requests(shost);
+ }
+ kfree(lnode_list);
+}
+
+void
+csio_lnodes_block_by_port(struct csio_hw *hw, uint8_t portid)
+{
+ struct csio_lnode *ln;
+ struct Scsi_Host *shost;
+ struct csio_lnode *sln;
+ struct list_head *cur_ln, *cur_cln;
+ struct csio_lnode **lnode_list;
+ int cur_cnt = 0, ii;
+
+ lnode_list = kzalloc((sizeof(struct csio_lnode *) * hw->num_lns),
+ GFP_KERNEL);
+ if (!lnode_list) {
+ csio_err(hw, "Failed to allocate lnodes_list");
+ return;
+ }
+
+ spin_lock_irq(&hw->lock);
+ /* Traverse sibling lnodes */
+ list_for_each(cur_ln, &hw->sln_head) {
+ sln = (struct csio_lnode *) cur_ln;
+ if (sln->portid != portid)
+ continue;
+
+ lnode_list[cur_cnt++] = sln;
+
+ /* Traverse children lnodes */
+ list_for_each(cur_cln, &sln->cln_head)
+ lnode_list[cur_cnt++] = (struct csio_lnode *) cur_cln;
+ }
+ spin_unlock_irq(&hw->lock);
+
+ for (ii = 0; ii < cur_cnt; ii++) {
+ csio_dbg(hw, "Blocking IOs on lnode: %p\n", lnode_list[ii]);
+ ln = lnode_list[ii];
+ shost = csio_ln_to_shost(ln);
+ scsi_block_requests(shost);
+ }
+ kfree(lnode_list);
+}
+
+void
+csio_lnodes_unblock_by_port(struct csio_hw *hw, uint8_t portid)
+{
+ struct csio_lnode *ln;
+ struct Scsi_Host *shost;
+ struct csio_lnode *sln;
+ struct list_head *cur_ln, *cur_cln;
+ struct csio_lnode **lnode_list;
+ int cur_cnt = 0, ii;
+
+ lnode_list = kzalloc((sizeof(struct csio_lnode *) * hw->num_lns),
+ GFP_KERNEL);
+ if (!lnode_list) {
+ csio_err(hw, "Failed to allocate lnodes_list");
+ return;
+ }
+
+ spin_lock_irq(&hw->lock);
+ /* Traverse sibling lnodes */
+ list_for_each(cur_ln, &hw->sln_head) {
+ sln = (struct csio_lnode *) cur_ln;
+ if (sln->portid != portid)
+ continue;
+ lnode_list[cur_cnt++] = sln;
+
+ /* Traverse children lnodes */
+ list_for_each(cur_cln, &sln->cln_head)
+ lnode_list[cur_cnt++] = (struct csio_lnode *) cur_cln;
+ }
+ spin_unlock_irq(&hw->lock);
+
+ for (ii = 0; ii < cur_cnt; ii++) {
+ csio_dbg(hw, "unblocking IOs on lnode: %p\n", lnode_list[ii]);
+ ln = lnode_list[ii];
+ shost = csio_ln_to_shost(ln);
+ scsi_unblock_requests(shost);
+ }
+ kfree(lnode_list);
+}
+
+void
+csio_lnodes_exit(struct csio_hw *hw, bool npiv)
+{
+ struct csio_lnode *sln;
+ struct csio_lnode *ln;
+ struct list_head *cur_ln, *cur_cln;
+ struct csio_lnode **lnode_list;
+ int cur_cnt = 0, ii;
+
+ lnode_list = kzalloc((sizeof(struct csio_lnode *) * hw->num_lns),
+ GFP_KERNEL);
+ if (!lnode_list) {
+ csio_err(hw, "lnodes_exit: Failed to allocate lnodes_list.\n");
+ return;
+ }
+
+ /* Get all child lnodes(NPIV ports) */
+ spin_lock_irq(&hw->lock);
+ list_for_each(cur_ln, &hw->sln_head) {
+ sln = (struct csio_lnode *) cur_ln;
+
+ /* Traverse children lnodes */
+ list_for_each(cur_cln, &sln->cln_head)
+ lnode_list[cur_cnt++] = (struct csio_lnode *) cur_cln;
+ }
+ spin_unlock_irq(&hw->lock);
+
+ /* Delete NPIV lnodes */
+ for (ii = 0; ii < cur_cnt; ii++) {
+ csio_dbg(hw, "Deleting child lnode: %p\n", lnode_list[ii]);
+ ln = lnode_list[ii];
+ fc_vport_terminate(ln->fc_vport);
+ }
+
+ /* Delete only npiv lnodes */
+ if (npiv)
+ goto free_lnodes;
+
+ cur_cnt = 0;
+ /* Get all physical lnodes */
+ spin_lock_irq(&hw->lock);
+ /* Traverse sibling lnodes */
+ list_for_each(cur_ln, &hw->sln_head) {
+ sln = (struct csio_lnode *) cur_ln;
+ lnode_list[cur_cnt++] = sln;
+ }
+ spin_unlock_irq(&hw->lock);
+
+ /* Delete physical lnodes */
+ for (ii = 0; ii < cur_cnt; ii++) {
+ csio_dbg(hw, "Deleting parent lnode: %p\n", lnode_list[ii]);
+ csio_shost_exit(lnode_list[ii]);
+ }
+
+free_lnodes:
+ kfree(lnode_list);
+}
+
+/*
+ * csio_lnode_init_post: Set lnode attributes after starting HW.
+ * @ln: lnode.
+ *
+ */
+static void
+csio_lnode_init_post(struct csio_lnode *ln)
+{
+ struct Scsi_Host *shost = csio_ln_to_shost(ln);
+
+ csio_fchost_attr_init(ln);
+
+ scsi_scan_host(shost);
+}
+
+/*
+ * csio_probe_one - Instantiate this function.
+ * @pdev: PCI device
+ * @id: Device ID
+ *
+ * This is the .probe() callback of the driver. This function:
+ * - Initializes the PCI function by enabling MMIO, setting bus
+ * mastership and setting DMA mask.
+ * - Allocates HW structure, DMA, memory resources, maps BARS to
+ * host memory and initializes HW module.
+ * - Allocates lnode structure via scsi_host_alloc, initializes
+ * shost, initialized lnode module and registers with SCSI ML
+ * via scsi_host_add.
+ * - Enables interrupts, and starts the chip by kicking off the
+ * HW state machine.
+ * - Once hardware is ready, initiated scan of the host via
+ * scsi_scan_host.
+ */
+static int __devinit
+csio_probe_one(struct pci_dev *pdev, const struct pci_device_id *id)
+{
+ int rv;
+ int bars;
+ int i;
+ struct csio_hw *hw;
+ struct csio_lnode *ln;
+
+ if (pf_counter > 0 && (CSIO_IS_T4_FPGA(pdev->device))) {
+ dev_err(&pdev->dev, "Driver claims only one FPGA function\n");
+ return -ENODEV;
+ }
+
+ rv = csio_pci_init(pdev, &bars);
+ if (rv)
+ goto err;
+
+ hw = csio_hw_alloc(pdev);
+ if (!hw) {
+ rv = -ENODEV;
+ goto err_pci_exit;
+ }
+
+ pci_set_drvdata(pdev, hw);
+
+ if (csio_hw_start(hw) != CSIO_SUCCESS) {
+ dev_err(&pdev->dev,
+ "Failed to start FW, continuing in debug mode.\n");
+ return 0;
+ }
+
+ sprintf(hw->fwrev_str, "%u.%u.%u.%u\n",
+ FW_HDR_FW_VER_MAJOR_GET(hw->fwrev),
+ FW_HDR_FW_VER_MINOR_GET(hw->fwrev),
+ FW_HDR_FW_VER_MICRO_GET(hw->fwrev),
+ FW_HDR_FW_VER_BUILD_GET(hw->fwrev));
+
+ for (i = 0; i < hw->num_pports; i++) {
+ ln = csio_shost_init(hw, &pdev->dev, CSIO_TRUE, NULL);
+ if (!ln) {
+ rv = -ENODEV;
+ break;
+ }
+ /* Initialize portid */
+ ln->portid = hw->pport[i].portid;
+
+ spin_lock_irq(&hw->lock);
+ if (csio_lnode_start(ln) != CSIO_SUCCESS)
+ rv = -ENODEV;
+ spin_unlock_irq(&hw->lock);
+
+ if (rv)
+ break;
+
+ csio_lnode_init_post(ln);
+ }
+
+ if (rv)
+ goto err_lnode_exit;
+
+ pf_counter++;
+
+ return 0;
+
+err_lnode_exit:
+ csio_lnodes_block_request(hw);
+ spin_lock_irq(&hw->lock);
+ csio_hw_stop(hw);
+ spin_unlock_irq(&hw->lock);
+ csio_lnodes_unblock_request(hw);
+ pci_set_drvdata(hw->pdev, NULL);
+ csio_lnodes_exit(hw, 0);
+ csio_hw_free(hw);
+err_pci_exit:
+ csio_pci_exit(pdev, &bars);
+err:
+ dev_err(&pdev->dev, "probe of device failed: %d\n", rv);
+ return rv;
+}
+
+/*
+ * csio_remove_one - Remove one instance of the driver at this PCI function.
+ * @pdev: PCI device
+ *
+ * Used during hotplug operation.
+ */
+static void __devexit
+csio_remove_one(struct pci_dev *pdev)
+{
+ struct csio_hw *hw = pci_get_drvdata(pdev);
+ int bars = pci_select_bars(pdev, IORESOURCE_MEM);
+
+ csio_lnodes_block_request(hw);
+ spin_lock_irq(&hw->lock);
+
+ /* Stops lnode, Rnode s/m
+ * Quiesce IOs.
+ * All sessions with remote ports are unregistered.
+ */
+ csio_hw_stop(hw);
+ spin_unlock_irq(&hw->lock);
+ csio_lnodes_unblock_request(hw);
+
+ csio_lnodes_exit(hw, 0);
+ csio_hw_free(hw);
+ pci_set_drvdata(pdev, NULL);
+ csio_pci_exit(pdev, &bars);
+}
+
+/*
+ * csio_pci_error_detected - PCI error was detected
+ * @pdev: PCI device
+ *
+ */
+static pci_ers_result_t
+csio_pci_error_detected(struct pci_dev *pdev, pci_channel_state_t state)
+{
+ struct csio_hw *hw = pci_get_drvdata(pdev);
+
+ csio_lnodes_block_request(hw);
+ spin_lock_irq(&hw->lock);
+
+ /* Post PCI error detected evt to HW s/m
+ * HW s/m handles this evt by quiescing IOs, unregisters rports
+ * and finally takes the device to offline.
+ */
+ csio_post_event(&hw->sm, CSIO_HWE_PCIERR_DETECTED);
+ spin_unlock_irq(&hw->lock);
+ csio_lnodes_unblock_request(hw);
+ csio_lnodes_exit(hw, 0);
+ csio_intr_disable(hw, CSIO_TRUE);
+ pci_disable_device(pdev);
+ return state == pci_channel_io_perm_failure ?
+ PCI_ERS_RESULT_DISCONNECT : PCI_ERS_RESULT_NEED_RESET;
+}
+
+/*
+ * csio_pci_slot_reset - PCI slot has been reset.
+ * @pdev: PCI device
+ *
+ */
+static pci_ers_result_t
+csio_pci_slot_reset(struct pci_dev *pdev)
+{
+ struct csio_hw *hw = pci_get_drvdata(pdev);
+
+ if (pci_enable_device(pdev)) {
+ dev_err(&pdev->dev, "cannot re-enable device in slot reset\n");
+ return PCI_ERS_RESULT_DISCONNECT;
+ }
+
+ pci_set_master(pdev);
+ pci_restore_state(pdev);
+ pci_save_state(pdev);
+ pci_cleanup_aer_uncorrect_error_status(pdev);
+
+ /* Bring HW s/m to ready state.
+ * but don't resume IOs.
+ */
+ spin_lock_irq(&hw->lock);
+ csio_post_event(&hw->sm, CSIO_HWE_PCIERR_SLOT_RESET);
+ if (!csio_is_hw_ready(hw)) {
+ spin_unlock_irq(&hw->lock);
+ dev_err(&pdev->dev, "Can't initialize HW when in slot reset\n");
+ return PCI_ERS_RESULT_DISCONNECT;
+ }
+ spin_unlock_irq(&hw->lock);
+ return PCI_ERS_RESULT_RECOVERED;
+}
+
+/*
+ * csio_pci_resume - Resume normal operations
+ * @pdev: PCI device
+ *
+ */
+static void
+csio_pci_resume(struct pci_dev *pdev)
+{
+ struct csio_hw *hw = pci_get_drvdata(pdev);
+ struct csio_lnode *ln;
+ int rv = 0;
+ int i;
+
+ /* Bring the LINK UP and Resume IO */
+
+ for (i = 0; i < hw->num_pports; i++) {
+ ln = csio_shost_init(hw, &pdev->dev, CSIO_TRUE, NULL);
+ if (!ln) {
+ rv = -ENODEV;
+ break;
+ }
+ /* Initialize portid */
+ ln->portid = hw->pport[i].portid;
+
+ spin_lock_irq(&hw->lock);
+ if (csio_lnode_start(ln) != CSIO_SUCCESS)
+ rv = -ENODEV;
+ spin_unlock_irq(&hw->lock);
+
+ if (rv)
+ break;
+
+ csio_lnode_init_post(ln);
+ }
+
+ if (rv)
+ goto err_resume_exit;
+
+ return;
+
+err_resume_exit:
+ csio_lnodes_block_request(hw);
+ spin_lock_irq(&hw->lock);
+ csio_hw_stop(hw);
+ spin_unlock_irq(&hw->lock);
+ csio_lnodes_unblock_request(hw);
+ csio_lnodes_exit(hw, 0);
+ csio_hw_free(hw);
+ dev_err(&pdev->dev, "resume of device failed: %d\n", rv);
+}
+
+static struct pci_error_handlers csio_err_handler = {
+ .error_detected = csio_pci_error_detected,
+ .slot_reset = csio_pci_slot_reset,
+ .resume = csio_pci_resume,
+};
+
+static DEFINE_PCI_DEVICE_TABLE(csio_pci_tbl) = {
+ CSIO_DEVICE(CSIO_DEVID_T440DBG_FCOE, 0), /* T440DBG FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T420CR_FCOE, 0), /* T420CR FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T422CR_FCOE, 0), /* T422CR FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T440CR_FCOE, 0), /* T440CR FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T420BCH_FCOE, 0), /* T420BCH FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T440BCH_FCOE, 0), /* T440BCH FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T440CH_FCOE, 0), /* T440CH FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T420SO_FCOE, 0), /* T420SO FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T420CX_FCOE, 0), /* T420CX FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T420BT_FCOE, 0), /* T420BT FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T404BT_FCOE, 0), /* T404BT FCOE */
+ CSIO_DEVICE(CSIO_DEVID_B420_FCOE, 0), /* B420 FCOE */
+ CSIO_DEVICE(CSIO_DEVID_B404_FCOE, 0), /* B404 FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T480CR_FCOE, 0), /* T480 CR FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T440LPCR_FCOE, 0), /* T440 LP-CR FCOE */
+ CSIO_DEVICE(CSIO_DEVID_PE10K, 0), /* PE10K FCOE */
+ CSIO_DEVICE(CSIO_DEVID_PE10K_PF1, 0), /* PE10K FCOE on PF1 */
+ { 0, 0, 0, 0, 0, 0, 0 }
+};
+
+
+static struct pci_driver csio_pci_driver = {
+ .name = KBUILD_MODNAME,
+ .driver = {
+ .owner = THIS_MODULE,
+ },
+ .id_table = csio_pci_tbl,
+ .probe = csio_probe_one,
+ .remove = csio_remove_one,
+ .err_handler = &csio_err_handler,
+};
+
+/*
+ * csio_init - Chelsio storage driver initialization function.
+ *
+ */
+static int __init
+csio_init(void)
+{
+ int rv = -ENOMEM;
+
+ pr_info("%s %s\n", CSIO_DRV_DESC, CSIO_DRV_VERSION);
+
+ csio_module_params_check();
+
+ csio_dfs_init();
+
+ csio_fcoe_transport = fc_attach_transport(&csio_fc_transport_funcs);
+ if (!csio_fcoe_transport)
+ goto err;
+
+ csio_fcoe_transport_vport =
+ fc_attach_transport(&csio_fc_transport_vport_funcs);
+ if (!csio_fcoe_transport_vport)
+ goto err_vport;
+
+ rv = pci_register_driver(&csio_pci_driver);
+ if (rv)
+ goto err_pci;
+
+ return 0;
+
+err_pci:
+ fc_release_transport(csio_fcoe_transport_vport);
+err_vport:
+ fc_release_transport(csio_fcoe_transport);
+err:
+ csio_dfs_exit();
+ return rv;
+}
+
+/*
+ * csio_exit - Chelsio storage driver uninitialization .
+ *
+ * Function that gets called in the unload path.
+ */
+static void __exit
+csio_exit(void)
+{
+ pci_unregister_driver(&csio_pci_driver);
+ csio_dfs_exit();
+ fc_release_transport(csio_fcoe_transport_vport);
+ fc_release_transport(csio_fcoe_transport);
+}
+
+module_init(csio_init);
+module_exit(csio_exit);
+MODULE_AUTHOR(CSIO_DRV_AUTHOR);
+MODULE_DESCRIPTION(CSIO_DRV_DESC);
+MODULE_LICENSE(CSIO_DRV_LICENSE);
+MODULE_DEVICE_TABLE(pci, csio_pci_tbl);
+MODULE_VERSION(CSIO_DRV_VERSION);
+MODULE_FIRMWARE(CSIO_FW_FNAME);
diff --git a/drivers/scsi/csiostor/csio_wr.c b/drivers/scsi/csiostor/csio_wr.c
new file mode 100644
index 0000000..dadd486
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_wr.c
@@ -0,0 +1,1634 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/kernel.h>
+#include <linux/string.h>
+#include <linux/compiler.h>
+#include <linux/slab.h>
+#include <asm/page.h>
+#include <linux/cache.h>
+
+#include "csio_hw.h"
+#include "csio_wr.h"
+#include "csio_mb.h"
+#include "csio_defs.h"
+
+int csio_intr_coalesce_cnt; /* value:SGE_INGRESS_RX_THRESHOLD[0] */
+static int csio_sge_thresh_reg; /* SGE_INGRESS_RX_THRESHOLD[0] */
+
+int csio_intr_coalesce_time = 10; /* value:SGE_TIMER_VALUE_1 */
+static int csio_sge_timer_reg = 1;
+
+#define csio_set_flbuf_size(_hw, _reg, _val) \
+ csio_wr_reg32((_hw), (_val), SGE_FL_BUFFER_SIZE##_reg)
+
+static void
+csio_get_flbuf_size(struct csio_hw *hw, struct csio_sge *sge, uint32_t reg)
+{
+ sge->sge_fl_buf_size[reg] = csio_rd_reg32(hw, SGE_FL_BUFFER_SIZE0 +
+ reg * sizeof(uint32_t));
+}
+
+/* Free list buffer size */
+static inline uint32_t
+csio_wr_fl_bufsz(struct csio_sge *sge, struct csio_dma_buf *buf)
+{
+ return sge->sge_fl_buf_size[buf->paddr & 0xF];
+}
+
+/* Size of the egress queue status page */
+static inline uint32_t
+csio_wr_qstat_pgsz(struct csio_hw *hw)
+{
+ return (hw->wrm.sge.sge_control & EGRSTATUSPAGESIZE(1)) ? 128 : 64;
+}
+
+/* Ring freelist doorbell */
+static inline void
+csio_wr_ring_fldb(struct csio_hw *hw, struct csio_q *flq)
+{
+ /*
+ * Ring the doorbell only when we have atleast CSIO_QCREDIT_SZ
+ * number of bytes in the freelist queue. This translates to atleast
+ * 8 freelist buffer pointers (since each pointer is 8 bytes).
+ */
+ if (flq->inc_idx >= 8) {
+ csio_wr_reg32(hw, DBPRIO(1) | QID(flq->un.fl.flid) |
+ PIDX(flq->inc_idx / 8),
+ MYPF_REG(SGE_PF_KDOORBELL));
+ flq->inc_idx &= 7;
+ }
+}
+
+/* Write a 0 cidx increment value to enable SGE interrupts for this queue */
+static void
+csio_wr_sge_intr_enable(struct csio_hw *hw, uint16_t iqid)
+{
+ csio_wr_reg32(hw, CIDXINC(0) |
+ INGRESSQID(iqid) |
+ TIMERREG(X_TIMERREG_RESTART_COUNTER),
+ MYPF_REG(SGE_PF_GTS));
+}
+
+/*
+ * csio_wr_fill_fl - Populate the FL buffers of a FL queue.
+ * @hw: HW module.
+ * @flq: Freelist queue.
+ *
+ * Fill up freelist buffer entries with buffers of size specified
+ * in the size register.
+ *
+ */
+static csio_retval_t
+csio_wr_fill_fl(struct csio_hw *hw, struct csio_q *flq)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_sge *sge = &wrm->sge;
+ __be64 *d = (__be64 *)(flq->vstart);
+ struct csio_dma_buf *buf = &flq->un.fl.bufs[0];
+ uint64_t paddr;
+ int sreg = flq->un.fl.sreg;
+ int n = flq->credits;
+
+ while (n--) {
+ buf->len = sge->sge_fl_buf_size[sreg];
+ buf->vaddr = pci_alloc_consistent(hw->pdev, buf->len,
+ &buf->paddr);
+ if (!buf->vaddr) {
+ csio_err(hw, "Could only fill %d buffers!\n", n + 1);
+ return CSIO_NOMEM;
+ }
+
+ paddr = buf->paddr | (sreg & 0xF);
+
+ *d++ = cpu_to_be64(paddr);
+ buf++;
+ }
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_wr_update_fl -
+ * @hw: HW module.
+ * @flq: Freelist queue.
+ *
+ *
+ */
+static inline void
+csio_wr_update_fl(struct csio_hw *hw, struct csio_q *flq, uint16_t n)
+{
+
+ flq->inc_idx += n;
+ flq->pidx += n;
+ if (unlikely(flq->pidx >= flq->credits))
+ flq->pidx -= (uint16_t)flq->credits;
+
+ csio_inc_stats(flq, n_flq_refill);
+}
+
+/*
+ * csio_wr_alloc_q - Allocate a WR queue and initialize it.
+ * @hw: HW module
+ * @qsize: Size of the queue in bytes
+ * @wrsize: Since of WR in this queue, if fixed.
+ * @type: Type of queue (Ingress/Egress/Freelist)
+ * @owner: Module that owns this queue.
+ * @nflb: Number of freelist buffers for FL.
+ * @sreg: What is the FL buffer size register?
+ * @iq_int_handler: Ingress queue handler in INTx mode.
+ *
+ * This function allocates and sets up a queue for the caller
+ * of size qsize, aligned at the required boundary. This is subject to
+ * be free entries being available in the queue array. If one is found,
+ * it is initialized with the allocated queue, marked as being used (owner),
+ * and a handle returned to the caller in form of the queue's index
+ * into the q_arr array.
+ * If user has indicated a freelist (by specifying nflb > 0), create
+ * another queue (with its own index into q_arr) for the freelist. Allocate
+ * memory for DMA buffer metadata (vaddr, len etc). Save off the freelist
+ * idx in the ingress queue's flq.idx. This is how a Freelist is associated
+ * with its owning ingress queue.
+ */
+int
+csio_wr_alloc_q(struct csio_hw *hw, uint32_t qsize, uint32_t wrsize,
+ uint16_t type, void *owner, uint32_t nflb, int sreg,
+ iq_handler_t iq_intx_handler)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_q *q, *flq;
+ int free_idx = wrm->free_qidx;
+ int ret_idx = free_idx;
+ uint32_t qsz;
+ int flq_idx;
+
+ if (free_idx >= wrm->num_q) {
+ csio_err(hw, "No more free queues.\n");
+ return -1;
+ }
+
+ switch (type) {
+ case CSIO_EGRESS:
+ qsz = ALIGN(qsize, CSIO_QCREDIT_SZ) +
+ csio_wr_qstat_pgsz(hw);
+ break;
+ case CSIO_INGRESS:
+ switch (wrsize) {
+ case 16:
+ case 32:
+ case 64:
+ case 128:
+ break;
+ default:
+ csio_err(hw, "Invalid Ingress queue WR size:%d\n",
+ wrsize);
+ return -1;
+ }
+
+ /*
+ * Number of elements must be a multiple of 16
+ * So this includes status page size
+ */
+ qsz = ALIGN(qsize/wrsize, 16) * wrsize;
+
+ break;
+ case CSIO_FREELIST:
+ qsz = ALIGN(qsize/wrsize, 8) * wrsize +
+ csio_wr_qstat_pgsz(hw);
+ break;
+ default:
+ csio_err(hw, "Invalid queue type: 0x%x\n", type);
+ return -1;
+ }
+
+ q = wrm->q_arr[free_idx];
+
+ q->vstart = pci_alloc_consistent(hw->pdev, qsz, &q->pstart);
+ if (!q->vstart) {
+ csio_err(hw,
+ "Failed to allocate DMA memory for "
+ "queue at id: %d size: %d\n", free_idx, qsize);
+ return -1;
+ }
+
+ /*
+ * We need to zero out the contents, importantly for ingress,
+ * since we start with a generatiom bit of 1 for ingress.
+ */
+ memset(q->vstart, 0, qsz);
+
+ q->type = type;
+ q->owner = owner;
+ q->pidx = q->cidx = q->inc_idx = 0;
+ q->size = qsz;
+ q->wr_sz = wrsize; /* If using fixed size WRs */
+
+ wrm->free_qidx++;
+
+ if (type == CSIO_INGRESS) {
+ /* Since queue area is set to zero */
+ q->un.iq.genbit = 1;
+
+ /*
+ * Ingress queue status page size is always the size of
+ * the ingress queue entry.
+ */
+ q->credits = (qsz - q->wr_sz) / q->wr_sz;
+ q->vwrap = (void *)((uintptr_t)(q->vstart) + qsz
+ - q->wr_sz);
+
+ /* Allocate memory for FL if requested */
+ if (nflb > 0) {
+ flq_idx = csio_wr_alloc_q(hw, nflb * sizeof(__be64),
+ sizeof(__be64), CSIO_FREELIST,
+ owner, 0, sreg, NULL);
+ if (flq_idx == -1) {
+ csio_err(hw,
+ "Failed to allocate FL queue"
+ " for IQ idx:%d\n", free_idx);
+ return -1;
+ }
+
+ /* Associate the new FL with the Ingress quue */
+ q->un.iq.flq_idx = flq_idx;
+
+ flq = wrm->q_arr[q->un.iq.flq_idx];
+ flq->un.fl.bufs = kzalloc(flq->credits *
+ sizeof(struct csio_dma_buf),
+ GFP_KERNEL);
+ if (!flq->un.fl.bufs) {
+ csio_err(hw,
+ "Failed to allocate FL queue bufs"
+ " for IQ idx:%d\n", free_idx);
+ return -1;
+ }
+
+ flq->un.fl.packen = 0;
+ flq->un.fl.offset = 0;
+ flq->un.fl.sreg = sreg;
+
+ /* Fill up the free list buffers */
+ if (csio_wr_fill_fl(hw, flq))
+ return -1;
+
+ /*
+ * Make sure in a FLQ, atleast 1 credit (8 FL buffers)
+ * remains unpopulated,otherwise HW thinks
+ * FLQ is empty.
+ */
+ flq->pidx = flq->inc_idx = flq->credits - 8;
+ } else {
+ q->un.iq.flq_idx = -1;
+ }
+
+ /* Associate the IQ INTx handler. */
+ q->un.iq.iq_intx_handler = iq_intx_handler;
+
+ csio_q_iqid(hw, ret_idx) = CSIO_MAX_QID;
+
+ } else if (type == CSIO_EGRESS) {
+ q->credits = (qsz - csio_wr_qstat_pgsz(hw)) / CSIO_QCREDIT_SZ;
+ q->vwrap = (void *)((uintptr_t)(q->vstart) + qsz
+ - csio_wr_qstat_pgsz(hw));
+ csio_q_eqid(hw, ret_idx) = CSIO_MAX_QID;
+ } else { /* Freelist */
+ q->credits = (qsz - csio_wr_qstat_pgsz(hw)) / sizeof(__be64);
+ q->vwrap = (void *)((uintptr_t)(q->vstart) + qsz
+ - csio_wr_qstat_pgsz(hw));
+ csio_q_flid(hw, ret_idx) = CSIO_MAX_QID;
+ }
+
+ return ret_idx;
+}
+
+/*
+ * csio_wr_iq_create_rsp - Response handler for IQ creation.
+ * @hw: The HW module.
+ * @mbp: Mailbox.
+ * @iq_idx: Ingress queue that got created.
+ *
+ * Handle FW_IQ_CMD mailbox completion. Save off the assigned IQ/FL ids.
+ */
+static csio_retval_t
+csio_wr_iq_create_rsp(struct csio_hw *hw, struct csio_mb *mbp, int iq_idx)
+{
+ struct csio_iq_params iqp;
+ enum fw_retval retval;
+ uint32_t iq_id;
+ int flq_idx;
+
+ memset(&iqp, 0, sizeof(struct csio_iq_params));
+
+ csio_mb_iq_alloc_write_rsp(hw, mbp, &retval, &iqp);
+
+ if (retval != FW_SUCCESS) {
+ csio_err(hw, "IQ cmd returned 0x%x!\n", retval);
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ csio_q_iqid(hw, iq_idx) = iqp.iqid;
+ csio_q_physiqid(hw, iq_idx) = iqp.physiqid;
+ csio_q_pidx(hw, iq_idx) = csio_q_cidx(hw, iq_idx) = 0;
+ csio_q_inc_idx(hw, iq_idx) = 0;
+
+ /* Actual iq-id. */
+ iq_id = iqp.iqid - hw->wrm.fw_iq_start;
+
+ /* Set the iq-id to iq map table. */
+ if (iq_id >= CSIO_MAX_IQ) {
+ csio_err(hw,
+ "Exceeding MAX_IQ(%d) supported!"
+ " iqid:%d rel_iqid:%d FW iq_start:%d\n",
+ CSIO_MAX_IQ, iq_id, iqp.iqid, hw->wrm.fw_iq_start);
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+ csio_q_set_intr_map(hw, iq_idx, iq_id);
+
+ /*
+ * During FW_IQ_CMD, FW sets interrupt_sent bit to 1 in the SGE
+ * ingress context of this queue. This will block interrupts to
+ * this queue until the next GTS write. Therefore, we do a
+ * 0-cidx increment GTS write for this queue just to clear the
+ * interrupt_sent bit. This will re-enable interrupts to this
+ * queue.
+ */
+ csio_wr_sge_intr_enable(hw, iqp.physiqid);
+
+ flq_idx = csio_q_iq_flq_idx(hw, iq_idx);
+ if (flq_idx != -1) {
+ struct csio_q *flq = hw->wrm.q_arr[flq_idx];
+
+ csio_q_flid(hw, flq_idx) = iqp.fl0id;
+ csio_q_cidx(hw, flq_idx) = 0;
+ csio_q_pidx(hw, flq_idx) = csio_q_credits(hw, flq_idx) - 8;
+ csio_q_inc_idx(hw, flq_idx) = csio_q_credits(hw, flq_idx) - 8;
+
+ /* Now update SGE about the buffers allocated during init */
+ csio_wr_ring_fldb(hw, flq);
+ }
+
+ mempool_free(mbp, hw->mb_mempool);
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_wr_iq_create - Configure an Ingress queue with FW.
+ * @hw: The HW module.
+ * @priv: Private data object.
+ * @iq_idx: Ingress queue index in the WR module.
+ * @vec: MSIX vector.
+ * @portid: PCIE Channel to be associated with this queue.
+ * @async: Is this a FW asynchronous message handling queue?
+ * @cbfn: Completion callback.
+ *
+ * This API configures an ingress queue with FW by issuing a FW_IQ_CMD mailbox
+ * with alloc/write bits set.
+ */
+csio_retval_t
+csio_wr_iq_create(struct csio_hw *hw, void *priv, int iq_idx,
+ uint32_t vec, uint8_t portid, bool async,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct csio_mb *mbp;
+ struct csio_iq_params iqp;
+ int flq_idx;
+
+ memset(&iqp, 0, sizeof(struct csio_iq_params));
+ csio_q_portid(hw, iq_idx) = portid;
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_err(hw, "IQ command out of memory!\n");
+ return CSIO_NOMEM;
+ }
+
+ switch (hw->intr_mode) {
+ case CSIO_IM_INTX:
+ case CSIO_IM_MSI:
+ /* For interrupt forwarding queue only */
+ if (hw->intr_iq_idx == iq_idx)
+ iqp.iqandst = X_INTERRUPTDESTINATION_PCIE;
+ else
+ iqp.iqandst = X_INTERRUPTDESTINATION_IQ;
+ iqp.iqandstindex =
+ csio_q_physiqid(hw, hw->intr_iq_idx);
+ break;
+ case CSIO_IM_MSIX:
+ iqp.iqandst = X_INTERRUPTDESTINATION_PCIE;
+ iqp.iqandstindex = (uint16_t)vec;
+ break;
+ case CSIO_IM_NONE:
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ /* Pass in the ingress queue cmd parameters */
+ iqp.pfn = hw->pfn;
+ iqp.vfn = 0;
+ iqp.iq_start = 1;
+ iqp.viid = 0;
+ iqp.type = FW_IQ_TYPE_FL_INT_CAP;
+ iqp.iqasynch = async;
+ if (csio_intr_coalesce_cnt)
+ iqp.iqanus = X_UPDATESCHEDULING_COUNTER_OPTTIMER;
+ else
+ iqp.iqanus = X_UPDATESCHEDULING_TIMER;
+ iqp.iqanud = X_UPDATEDELIVERY_INTERRUPT;
+ iqp.iqpciech = portid;
+ iqp.iqintcntthresh = (uint8_t)csio_sge_thresh_reg;
+
+ switch (csio_q_wr_sz(hw, iq_idx)) {
+ case 16:
+ iqp.iqesize = 0; break;
+ case 32:
+ iqp.iqesize = 1; break;
+ case 64:
+ iqp.iqesize = 2; break;
+ case 128:
+ iqp.iqesize = 3; break;
+ }
+
+ iqp.iqsize = csio_q_size(hw, iq_idx) /
+ csio_q_wr_sz(hw, iq_idx);
+ iqp.iqaddr = csio_q_pstart(hw, iq_idx);
+
+ flq_idx = csio_q_iq_flq_idx(hw, iq_idx);
+ if (flq_idx != -1) {
+ struct csio_q *flq = hw->wrm.q_arr[flq_idx];
+
+ iqp.fl0paden = 1;
+ iqp.fl0packen = flq->un.fl.packen ? 1 : 0;
+ iqp.fl0fbmin = X_FETCHBURSTMIN_64B;
+ iqp.fl0fbmax = X_FETCHBURSTMAX_512B;
+ iqp.fl0size = csio_q_size(hw, flq_idx) / CSIO_QCREDIT_SZ;
+ iqp.fl0addr = csio_q_pstart(hw, flq_idx);
+ }
+
+ csio_mb_iq_alloc_write(hw, mbp, priv, CSIO_MB_DEFAULT_TMO, &iqp, cbfn);
+
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "Issue of IQ cmd failed!\n");
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ if (cbfn != NULL)
+ return CSIO_SUCCESS;
+
+ return csio_wr_iq_create_rsp(hw, mbp, iq_idx);
+}
+
+/*
+ * csio_wr_eq_create_rsp - Response handler for EQ creation.
+ * @hw: The HW module.
+ * @mbp: Mailbox.
+ * @eq_idx: Egress queue that got created.
+ *
+ * Handle FW_EQ_OFLD_CMD mailbox completion. Save off the assigned EQ ids.
+ */
+static csio_retval_t
+csio_wr_eq_cfg_rsp(struct csio_hw *hw, struct csio_mb *mbp, int eq_idx)
+{
+ struct csio_eq_params eqp;
+ enum fw_retval retval;
+
+ memset(&eqp, 0, sizeof(struct csio_eq_params));
+
+ csio_mb_eq_ofld_alloc_write_rsp(hw, mbp, &retval, &eqp);
+
+ if (retval != FW_SUCCESS) {
+ csio_err(hw, "EQ OFLD cmd returned 0x%x!\n", retval);
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ csio_q_eqid(hw, eq_idx) = (uint16_t)eqp.eqid;
+ csio_q_physeqid(hw, eq_idx) = (uint16_t)eqp.physeqid;
+ csio_q_pidx(hw, eq_idx) = csio_q_cidx(hw, eq_idx) = 0;
+ csio_q_inc_idx(hw, eq_idx) = 0;
+
+ mempool_free(mbp, hw->mb_mempool);
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_wr_eq_create - Configure an Egress queue with FW.
+ * @hw: HW module.
+ * @priv: Private data.
+ * @eq_idx: Egress queue index in the WR module.
+ * @iq_idx: Associated ingress queue index.
+ * @cbfn: Completion callback.
+ *
+ * This API configures a offload egress queue with FW by issuing a
+ * FW_EQ_OFLD_CMD (with alloc + write ) mailbox.
+ */
+csio_retval_t
+csio_wr_eq_create(struct csio_hw *hw, void *priv, int eq_idx,
+ int iq_idx, uint8_t portid,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct csio_mb *mbp;
+ struct csio_eq_params eqp;
+
+ memset(&eqp, 0, sizeof(struct csio_eq_params));
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_err(hw, "EQ command out of memory!\n");
+ return CSIO_NOMEM;
+ }
+
+ eqp.pfn = hw->pfn;
+ eqp.vfn = 0;
+ eqp.eqstart = 1;
+ eqp.hostfcmode = X_HOSTFCMODE_STATUS_PAGE;
+ eqp.iqid = csio_q_iqid(hw, iq_idx);
+ eqp.fbmin = X_FETCHBURSTMIN_64B;
+ eqp.fbmax = X_FETCHBURSTMAX_512B;
+ eqp.cidxfthresh = 0;
+ eqp.pciechn = portid;
+ eqp.eqsize = csio_q_size(hw, eq_idx) / CSIO_QCREDIT_SZ;
+ eqp.eqaddr = csio_q_pstart(hw, eq_idx);
+
+ csio_mb_eq_ofld_alloc_write(hw, mbp, priv, CSIO_MB_DEFAULT_TMO,
+ &eqp, cbfn);
+
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "Issue of EQ OFLD cmd failed!\n");
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ if (cbfn != NULL)
+ return CSIO_SUCCESS;
+
+ return csio_wr_eq_cfg_rsp(hw, mbp, eq_idx);
+}
+
+/*
+ * csio_wr_iq_destroy_rsp - Response handler for IQ removal.
+ * @hw: The HW module.
+ * @mbp: Mailbox.
+ * @iq_idx: Ingress queue that was freed.
+ *
+ * Handle FW_IQ_CMD (free) mailbox completion.
+ */
+static csio_retval_t
+csio_wr_iq_destroy_rsp(struct csio_hw *hw, struct csio_mb *mbp, int iq_idx)
+{
+ enum fw_retval retval = csio_mb_fw_retval(mbp);
+ csio_retval_t rv = CSIO_SUCCESS;
+
+ if (retval != FW_SUCCESS)
+ rv = CSIO_INVAL;
+
+ mempool_free(mbp, hw->mb_mempool);
+
+ return rv;
+}
+
+/*
+ * csio_wr_iq_destroy - Free an ingress queue.
+ * @hw: The HW module.
+ * @priv: Private data object.
+ * @iq_idx: Ingress queue index to destroy
+ * @cbfn: Completion callback.
+ *
+ * This API frees an ingress queue by issuing the FW_IQ_CMD
+ * with the free bit set.
+ */
+static csio_retval_t
+csio_wr_iq_destroy(struct csio_hw *hw, void *priv, int iq_idx,
+ void (*cbfn)(struct csio_hw *, struct csio_mb *))
+{
+ csio_retval_t rv = CSIO_SUCCESS;
+ struct csio_mb *mbp;
+ struct csio_iq_params iqp;
+ int flq_idx;
+
+ memset(&iqp, 0, sizeof(struct csio_iq_params));
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp)
+ return CSIO_NOMEM;
+
+ iqp.pfn = hw->pfn;
+ iqp.vfn = 0;
+ iqp.iqid = csio_q_iqid(hw, iq_idx);
+ iqp.type = FW_IQ_TYPE_FL_INT_CAP;
+
+ flq_idx = csio_q_iq_flq_idx(hw, iq_idx);
+ if (flq_idx != -1)
+ iqp.fl0id = csio_q_flid(hw, flq_idx);
+ else
+ iqp.fl0id = 0xFFFF;
+
+ iqp.fl1id = 0xFFFF;
+
+ csio_mb_iq_free(hw, mbp, priv, CSIO_MB_DEFAULT_TMO, &iqp, cbfn);
+
+ rv = csio_mb_issue(hw, mbp);
+ if (rv != CSIO_SUCCESS) {
+ mempool_free(mbp, hw->mb_mempool);
+ return rv;
+ }
+
+ if (cbfn != NULL)
+ return CSIO_SUCCESS;
+
+ return csio_wr_iq_destroy_rsp(hw, mbp, iq_idx);
+}
+
+/*
+ * csio_wr_eq_destroy_rsp - Response handler for OFLD EQ creation.
+ * @hw: The HW module.
+ * @mbp: Mailbox.
+ * @eq_idx: Egress queue that was freed.
+ *
+ * Handle FW_OFLD_EQ_CMD (free) mailbox completion.
+ */
+static csio_retval_t
+csio_wr_eq_destroy_rsp(struct csio_hw *hw, struct csio_mb *mbp, int eq_idx)
+{
+ enum fw_retval retval = csio_mb_fw_retval(mbp);
+ csio_retval_t rv = CSIO_SUCCESS;
+
+ if (retval != FW_SUCCESS)
+ rv = CSIO_INVAL;
+
+ mempool_free(mbp, hw->mb_mempool);
+
+ return rv;
+}
+
+/*
+ * csio_wr_eq_destroy - Free an Egress queue.
+ * @hw: The HW module.
+ * @priv: Private data object.
+ * @eq_idx: Egress queue index to destroy
+ * @cbfn: Completion callback.
+ *
+ * This API frees an Egress queue by issuing the FW_EQ_OFLD_CMD
+ * with the free bit set.
+ */
+static csio_retval_t
+csio_wr_eq_destroy(struct csio_hw *hw, void *priv, int eq_idx,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ csio_retval_t rv = CSIO_SUCCESS;
+ struct csio_mb *mbp;
+ struct csio_eq_params eqp;
+
+ memset(&eqp, 0, sizeof(struct csio_eq_params));
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp)
+ return CSIO_NOMEM;
+
+ eqp.pfn = hw->pfn;
+ eqp.vfn = 0;
+ eqp.eqid = csio_q_eqid(hw, eq_idx);
+
+ csio_mb_eq_ofld_free(hw, mbp, priv, CSIO_MB_DEFAULT_TMO, &eqp, cbfn);
+
+ rv = csio_mb_issue(hw, mbp);
+ if (rv != CSIO_SUCCESS) {
+ mempool_free(mbp, hw->mb_mempool);
+ return rv;
+ }
+
+ if (cbfn != NULL)
+ return CSIO_SUCCESS;
+
+ return csio_wr_eq_destroy_rsp(hw, mbp, eq_idx);
+}
+
+/*
+ * csio_wr_cleanup_eq_stpg - Cleanup Egress queue status page
+ * @hw: HW module
+ * @qidx: Egress queue index
+ *
+ * Cleanup the Egress queue status page.
+ */
+static void
+csio_wr_cleanup_eq_stpg(struct csio_hw *hw, int qidx)
+{
+ struct csio_q *q = csio_hw_to_wrm(hw)->q_arr[qidx];
+ struct csio_qstatus_page *stp = (struct csio_qstatus_page *)q->vwrap;
+
+ memset(stp, 0, sizeof(*stp));
+}
+
+/*
+ * csio_wr_cleanup_iq_ftr - Cleanup Footer entries in IQ
+ * @hw: HW module
+ * @qidx: Ingress queue index
+ *
+ * Cleanup the footer entries in the given ingress queue,
+ * set to 1 the internal copy of genbit.
+ */
+static void
+csio_wr_cleanup_iq_ftr(struct csio_hw *hw, int qidx)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_q *q = wrm->q_arr[qidx];
+ void *wr;
+ struct csio_iqwr_footer *ftr;
+ uint32_t i = 0;
+
+ /* set to 1 since we are just about zero out genbit */
+ q->un.iq.genbit = 1;
+
+ for (i = 0; i < q->credits; i++) {
+ /* Get the WR */
+ wr = (void *)((uintptr_t)q->vstart +
+ (i * q->wr_sz));
+ /* Get the footer */
+ ftr = (struct csio_iqwr_footer *)((uintptr_t)wr +
+ (q->wr_sz - sizeof(*ftr)));
+ /* Zero out footer */
+ memset(ftr, 0, sizeof(*ftr));
+ }
+}
+
+csio_retval_t
+csio_wr_destroy_queues(struct csio_hw *hw, bool cmd)
+{
+ int i, flq_idx;
+ struct csio_q *q;
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ csio_retval_t rv;
+
+ for (i = 0; i < wrm->free_qidx; i++) {
+ q = wrm->q_arr[i];
+
+ switch (q->type) {
+ case CSIO_EGRESS:
+ if (csio_q_eqid(hw, i) != CSIO_MAX_QID) {
+ csio_wr_cleanup_eq_stpg(hw, i);
+ if (!cmd) {
+ csio_q_eqid(hw, i) = CSIO_MAX_QID;
+ continue;
+ }
+
+ rv = csio_wr_eq_destroy(hw, NULL, i, NULL);
+ if ((rv == CSIO_BUSY) || (rv == CSIO_TIMEOUT))
+ cmd = CSIO_FALSE;
+
+ csio_q_eqid(hw, i) = CSIO_MAX_QID;
+ }
+ case CSIO_INGRESS:
+ if (csio_q_iqid(hw, i) != CSIO_MAX_QID) {
+ csio_wr_cleanup_iq_ftr(hw, i);
+ if (!cmd) {
+ csio_q_iqid(hw, i) = CSIO_MAX_QID;
+ flq_idx = csio_q_iq_flq_idx(hw, i);
+ if (flq_idx != -1)
+ csio_q_flid(hw, flq_idx) =
+ CSIO_MAX_QID;
+ continue;
+ }
+
+ rv = csio_wr_iq_destroy(hw, NULL, i, NULL);
+ if ((rv == CSIO_BUSY) || (rv == CSIO_TIMEOUT))
+ cmd = CSIO_FALSE;
+
+ csio_q_iqid(hw, i) = CSIO_MAX_QID;
+ flq_idx = csio_q_iq_flq_idx(hw, i);
+ if (flq_idx != -1)
+ csio_q_flid(hw, flq_idx) = CSIO_MAX_QID;
+ }
+ default:
+ break;
+ }
+ }
+
+ hw->flags &= ~CSIO_HWF_Q_FW_ALLOCED;
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_wr_get - Get requested size of WR entry/entries from queue.
+ * @hw: HW module.
+ * @qidx: Index of queue.
+ * @size: Cumulative size of Work request(s).
+ * @wrp: Work request pair.
+ *
+ * If requested credits are available, return the start address of the
+ * work request in the work request pair. Set pidx accordingly and
+ * return.
+ *
+ * NOTE about WR pair:
+ * ==================
+ * A WR can start towards the end of a queue, and then continue at the
+ * beginning, since the queue is considered to be circular. This will
+ * require a pair of address/size to be passed back to the caller -
+ * hence Work request pair format.
+ */
+csio_retval_t
+csio_wr_get(struct csio_hw *hw, int qidx, uint32_t size,
+ struct csio_wr_pair *wrp)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_q *q = wrm->q_arr[qidx];
+ void *cwr = (void *)((uintptr_t)(q->vstart) +
+ (q->pidx * CSIO_QCREDIT_SZ));
+ struct csio_qstatus_page *stp = (struct csio_qstatus_page *)q->vwrap;
+ uint16_t cidx = q->cidx = ntohs(stp->cidx);
+ uint16_t pidx = q->pidx;
+ uint32_t req_sz = ALIGN(size, CSIO_QCREDIT_SZ);
+ int req_credits = req_sz / CSIO_QCREDIT_SZ;
+ int credits;
+
+ CSIO_DB_ASSERT(q->owner != NULL);
+ CSIO_DB_ASSERT((qidx >= 0) && (qidx < wrm->free_qidx));
+ CSIO_DB_ASSERT(cidx <= q->credits);
+
+ /* Calculate credits */
+ if (pidx > cidx) {
+ credits = q->credits - (pidx - cidx) - 1;
+ } else if (cidx > pidx) {
+ credits = cidx - pidx - 1;
+ } else {
+ /* cidx == pidx, empty queue */
+ credits = q->credits;
+ csio_inc_stats(q, n_qempty);
+ }
+
+ /*
+ * Check if we have enough credits.
+ * credits = 1 implies queue is full.
+ */
+ if (!credits || (req_credits > credits)) {
+ csio_inc_stats(q, n_qfull);
+ return CSIO_BUSY;
+ }
+
+ /*
+ * If we are here, we have enough credits to satisfy the
+ * request. Check if we are near the end of q, and if WR spills over.
+ * If it does, use the first addr/size to cover the queue until
+ * the end. Fit the remainder portion of the request at the top
+ * of queue and return it in the second addr/len. Set pidx
+ * accordingly.
+ */
+ if (unlikely(((uintptr_t)cwr + req_sz) > (uintptr_t)(q->vwrap))) {
+ wrp->addr1 = cwr;
+ wrp->size1 = (uint32_t)((uintptr_t)q->vwrap - (uintptr_t)cwr);
+ wrp->addr2 = q->vstart;
+ wrp->size2 = req_sz - wrp->size1;
+ q->pidx = (uint16_t)(ALIGN(wrp->size2, CSIO_QCREDIT_SZ) /
+ CSIO_QCREDIT_SZ);
+ csio_inc_stats(q, n_qwrap);
+ csio_inc_stats(q, n_eq_wr_split);
+ } else {
+ wrp->addr1 = cwr;
+ wrp->size1 = req_sz;
+ wrp->addr2 = NULL;
+ wrp->size2 = 0;
+ q->pidx += (uint16_t)req_credits;
+
+ /* We are the end of queue, roll back pidx to top of queue */
+ if (unlikely(q->pidx == q->credits)) {
+ q->pidx = 0;
+ csio_inc_stats(q, n_qwrap);
+ }
+ }
+
+ q->inc_idx = (uint16_t)req_credits;
+
+ csio_inc_stats(q, n_tot_reqs);
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_wr_copy_to_wrp - Copies given data into WR.
+ * @data_buf - Data buffer
+ * @wrp - Work request pair.
+ * @wr_off - Work request offset.
+ * @data_len - Data length.
+ *
+ * Copies the given data in Work Request. Work request pair(wrp) specifies
+ * address information of Work request.
+ * Returns: none
+ */
+void
+csio_wr_copy_to_wrp(void *data_buf, struct csio_wr_pair *wrp,
+ uint32_t wr_off, uint32_t data_len)
+{
+ uint32_t nbytes;
+
+ /* Number of space available in buffer addr1 of WRP */
+ nbytes = ((wrp->size1 - wr_off) >= data_len) ?
+ data_len : (wrp->size1 - wr_off);
+
+ memcpy((uint8_t *) wrp->addr1 + wr_off, data_buf, nbytes);
+ data_len -= nbytes;
+
+ /* Write the remaining data from the begining of circular buffer */
+ if (data_len) {
+ CSIO_DB_ASSERT(data_len <= wrp->size2);
+ CSIO_DB_ASSERT(wrp->addr2 != NULL);
+ memcpy(wrp->addr2, (uint8_t *) data_buf + nbytes, data_len);
+ }
+}
+
+/*
+ * csio_wr_issue - Notify chip of Work request.
+ * @hw: HW module.
+ * @qidx: Index of queue.
+ * @prio: 0: Low priority, 1: High priority
+ *
+ * Rings the SGE Doorbell by writing the current producer index of the passed
+ * in queue into the register.
+ *
+ */
+csio_retval_t
+csio_wr_issue(struct csio_hw *hw, int qidx, bool prio)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_q *q = wrm->q_arr[qidx];
+
+ CSIO_DB_ASSERT((qidx >= 0) && (qidx < wrm->free_qidx));
+
+ wmb();
+ /* Ring SGE Doorbell writing q->pidx into it */
+ csio_wr_reg32(hw, DBPRIO(prio) | QID(q->un.eq.physeqid) |
+ PIDX(q->inc_idx), MYPF_REG(SGE_PF_KDOORBELL));
+ q->inc_idx = 0;
+
+ return CSIO_SUCCESS;
+}
+
+static inline uint32_t
+csio_wr_avail_qcredits(struct csio_q *q)
+{
+ if (q->pidx > q->cidx)
+ return q->pidx - q->cidx;
+ else if (q->cidx > q->pidx)
+ return q->credits - (q->cidx - q->pidx);
+ else
+ return 0; /* cidx == pidx, empty queue */
+}
+
+/*
+ * csio_wr_inval_flq_buf - Invalidate a free list buffer entry.
+ * @hw: HW module.
+ * @flq: The freelist queue.
+ *
+ * Invalidate the driver's version of a freelist buffer entry,
+ * without freeing the associated the DMA memory. The entry
+ * to be invalidated is picked up from the current Free list
+ * queue cidx.
+ *
+ */
+static inline void
+csio_wr_inval_flq_buf(struct csio_hw *hw, struct csio_q *flq)
+{
+ flq->cidx++;
+ if (flq->cidx == flq->credits) {
+ flq->cidx = 0;
+ csio_inc_stats(flq, n_qwrap);
+ }
+}
+
+/*
+ * csio_wr_process_fl - Process a freelist completion.
+ * @hw: HW module.
+ * @q: The ingress queue attached to the Freelist.
+ * @wr: The freelist completion WR in the ingress queue.
+ * @len_to_qid: The lower 32-bits of the first flit of the RSP footer
+ * @iq_handler: Caller's handler for this completion.
+ * @priv: Private pointer of caller
+ *
+ */
+static inline void
+csio_wr_process_fl(struct csio_hw *hw, struct csio_q *q,
+ void *wr, uint32_t len_to_qid,
+ void (*iq_handler)(struct csio_hw *, void *,
+ uint32_t, struct csio_fl_dma_buf *,
+ void *),
+ void *priv)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_sge *sge = &wrm->sge;
+ struct csio_fl_dma_buf flb;
+ struct csio_dma_buf *buf, *fbuf;
+ uint32_t bufsz, len, lastlen = 0;
+ int n;
+ struct csio_q *flq = hw->wrm.q_arr[q->un.iq.flq_idx];
+
+ CSIO_DB_ASSERT(flq != NULL);
+
+ len = len_to_qid;
+
+ if (len & IQWRF_NEWBUF) {
+ if (flq->un.fl.offset > 0) {
+ csio_wr_inval_flq_buf(hw, flq);
+ flq->un.fl.offset = 0;
+ }
+ len = IQWRF_LEN_GET(len);
+ }
+
+ CSIO_DB_ASSERT(len != 0);
+
+ flb.totlen = len;
+
+ /* Consume all freelist buffers used for len bytes */
+ for (n = 0, fbuf = flb.flbufs; ; n++, fbuf++) {
+ buf = &flq->un.fl.bufs[flq->cidx];
+ bufsz = csio_wr_fl_bufsz(sge, buf);
+
+ fbuf->paddr = buf->paddr;
+ fbuf->vaddr = buf->vaddr;
+
+ flb.offset = flq->un.fl.offset;
+ lastlen = min(bufsz, len);
+ fbuf->len = lastlen;
+
+ len -= lastlen;
+ if (!len)
+ break;
+ csio_wr_inval_flq_buf(hw, flq);
+ }
+
+ flb.defer_free = flq->un.fl.packen ? 0 : 1;
+
+ iq_handler(hw, wr, q->wr_sz - sizeof(struct csio_iqwr_footer),
+ &flb, priv);
+
+ if (flq->un.fl.packen)
+ flq->un.fl.offset += ALIGN(lastlen, sge->csio_fl_align);
+ else
+ csio_wr_inval_flq_buf(hw, flq);
+
+}
+
+/*
+ * csio_is_new_iqwr - Is this a new Ingress queue entry ?
+ * @q: Ingress quueue.
+ * @ftr: Ingress queue WR SGE footer.
+ *
+ * The entry is new if our generation bit matches the corresponding
+ * bit in the footer of the current WR.
+ */
+static inline bool
+csio_is_new_iqwr(struct csio_q *q, struct csio_iqwr_footer *ftr)
+{
+ return (q->un.iq.genbit == (ftr->u.type_gen >> IQWRF_GEN_SHIFT));
+}
+
+/*
+ * csio_wr_process_iq - Process elements in Ingress queue.
+ * @hw: HW pointer
+ * @qidx: Index of queue
+ * @iq_handler: Handler for this queue
+ * @priv: Caller's private pointer
+ *
+ * This routine walks through every entry of the ingress queue, calling
+ * the provided iq_handler with the entry, until the generation bit
+ * flips.
+ */
+csio_retval_t
+csio_wr_process_iq(struct csio_hw *hw, struct csio_q *q,
+ void (*iq_handler)(struct csio_hw *, void *,
+ uint32_t, struct csio_fl_dma_buf *,
+ void *),
+ void *priv)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ void *wr = (void *)((uintptr_t)q->vstart + (q->cidx * q->wr_sz));
+ struct csio_iqwr_footer *ftr;
+ uint32_t wr_type, fw_qid, qid;
+ struct csio_q *q_completed;
+ struct csio_q *flq = csio_iq_has_fl(q) ?
+ wrm->q_arr[q->un.iq.flq_idx] : NULL;
+ csio_retval_t rv = CSIO_SUCCESS;
+
+ /* Get the footer */
+ ftr = (struct csio_iqwr_footer *)((uintptr_t)wr +
+ (q->wr_sz - sizeof(*ftr)));
+
+ /*
+ * When q wrapped around last time, driver should have inverted
+ * ic.genbit as well.
+ */
+ while (csio_is_new_iqwr(q, ftr)) {
+
+ CSIO_DB_ASSERT(((uintptr_t)wr + q->wr_sz) <=
+ (uintptr_t)q->vwrap);
+ rmb();
+ wr_type = IQWRF_TYPE_GET(ftr->u.type_gen);
+
+ switch (wr_type) {
+ case X_RSPD_TYPE_CPL:
+ /* Subtract footer from WR len */
+ iq_handler(hw, wr, q->wr_sz - sizeof(*ftr), NULL, priv);
+ break;
+ case X_RSPD_TYPE_FLBUF:
+ csio_wr_process_fl(hw, q, wr,
+ ntohl(ftr->pldbuflen_qid),
+ iq_handler, priv);
+ break;
+ case X_RSPD_TYPE_INTR:
+ fw_qid = ntohl(ftr->pldbuflen_qid);
+ qid = fw_qid - wrm->fw_iq_start;
+ q_completed = hw->wrm.intr_map[qid];
+
+ if (unlikely(qid ==
+ csio_q_physiqid(hw, hw->intr_iq_idx))) {
+ /*
+ * We are already in the Forward Interrupt
+ * Interrupt Queue Service! Do-not service
+ * again!
+ *
+ */
+ } else {
+ CSIO_DB_ASSERT(q_completed);
+ CSIO_DB_ASSERT(
+ q_completed->un.iq.iq_intx_handler);
+
+ /* Call the queue handler. */
+ q_completed->un.iq.iq_intx_handler(hw, NULL,
+ 0, NULL, (void *)q_completed);
+ }
+ break;
+ default:
+ csio_warn(hw, "Unknown resp type 0x%x received\n",
+ wr_type);
+ csio_inc_stats(q, n_rsp_unknown);
+ break;
+ }
+
+ /*
+ * Ingress *always* has fixed size WR entries. Therefore,
+ * there should always be complete WRs towards the end of
+ * queue.
+ */
+ if (((uintptr_t)wr + q->wr_sz) == (uintptr_t)q->vwrap) {
+
+ /* Roll over to start of queue */
+ q->cidx = 0;
+ wr = q->vstart;
+
+ /* Toggle genbit */
+ q->un.iq.genbit ^= 0x1;
+
+ csio_inc_stats(q, n_qwrap);
+ } else {
+ q->cidx++;
+ wr = (void *)((uintptr_t)(q->vstart) +
+ (q->cidx * q->wr_sz));
+ }
+
+ ftr = (struct csio_iqwr_footer *)((uintptr_t)wr +
+ (q->wr_sz - sizeof(*ftr)));
+ q->inc_idx++;
+
+ } /* while (q->un.iq.genbit == hdr->genbit) */
+
+ /*
+ * We need to re-arm SGE interrupts in case we got a stray interrupt,
+ * especially in msix mode. With INTx, this may be a common occurence.
+ */
+ if (unlikely(!q->inc_idx)) {
+ csio_inc_stats(q, n_stray_comp);
+ rv = CSIO_INVAL;
+ goto restart;
+ }
+
+ /* Replenish free list buffers if pending falls below low water mark */
+ if (flq) {
+ uint32_t avail = csio_wr_avail_qcredits(flq);
+ if (avail <= 16) {
+ /* Make sure in FLQ, atleast 1 credit (8 FL buffers)
+ * remains unpopulated otherwise HW thinks
+ * FLQ is empty.
+ */
+ csio_wr_update_fl(hw, flq, (flq->credits - 8) - avail);
+ csio_wr_ring_fldb(hw, flq);
+ }
+ }
+
+restart:
+ /* Now inform SGE about our incremental index value */
+ csio_wr_reg32(hw, CIDXINC(q->inc_idx) |
+ INGRESSQID(q->un.iq.physiqid) |
+ TIMERREG(csio_sge_timer_reg),
+ MYPF_REG(SGE_PF_GTS));
+ q->stats.n_tot_rsps += q->inc_idx;
+
+ q->inc_idx = 0;
+
+ return rv;
+}
+
+csio_retval_t
+csio_wr_process_iq_idx(struct csio_hw *hw, int qidx,
+ void (*iq_handler)(struct csio_hw *, void *,
+ uint32_t, struct csio_fl_dma_buf *,
+ void *),
+ void *priv)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_q *iq = wrm->q_arr[qidx];
+
+ return csio_wr_process_iq(hw, iq, iq_handler, priv);
+}
+
+static int
+csio_closest_timer(struct csio_sge *s, int time)
+{
+ int i, delta, match = 0, min_delta = INT_MAX;
+
+ for (i = 0; i < ARRAY_SIZE(s->timer_val); i++) {
+ delta = time - s->timer_val[i];
+ if (delta < 0)
+ delta = -delta;
+ if (delta < min_delta) {
+ min_delta = delta;
+ match = i;
+ }
+ }
+ return match;
+}
+
+static int
+csio_closest_thresh(struct csio_sge *s, int cnt)
+{
+ int i, delta, match = 0, min_delta = INT_MAX;
+
+ for (i = 0; i < ARRAY_SIZE(s->counter_val); i++) {
+ delta = cnt - s->counter_val[i];
+ if (delta < 0)
+ delta = -delta;
+ if (delta < min_delta) {
+ min_delta = delta;
+ match = i;
+ }
+ }
+ return match;
+}
+
+static void
+csio_wr_fixup_host_params(struct csio_hw *hw)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_sge *sge = &wrm->sge;
+ uint32_t clsz = L1_CACHE_BYTES;
+ uint32_t s_hps = PAGE_SHIFT - 10;
+ uint32_t ingpad = 0;
+ uint32_t stat_len = clsz > 64 ? 128 : 64;
+
+ csio_wr_reg32(hw, HOSTPAGESIZEPF0(s_hps) | HOSTPAGESIZEPF1(s_hps) |
+ HOSTPAGESIZEPF2(s_hps) | HOSTPAGESIZEPF3(s_hps) |
+ HOSTPAGESIZEPF4(s_hps) | HOSTPAGESIZEPF5(s_hps) |
+ HOSTPAGESIZEPF6(s_hps) | HOSTPAGESIZEPF7(s_hps),
+ SGE_HOST_PAGE_SIZE);
+
+ sge->csio_fl_align = clsz < 32 ? 32 : clsz;
+ ingpad = ilog2(sge->csio_fl_align) - 5;
+
+ csio_set_reg_field(hw, SGE_CONTROL, INGPADBOUNDARY_MASK |
+ EGRSTATUSPAGESIZE(1),
+ INGPADBOUNDARY(ingpad) |
+ EGRSTATUSPAGESIZE(stat_len != 64));
+
+ /* FL BUFFER SIZE#0 is Page size i,e already aligned to cache line */
+ csio_wr_reg32(hw, PAGE_SIZE, SGE_FL_BUFFER_SIZE0);
+ csio_wr_reg32(hw,
+ (csio_rd_reg32(hw, SGE_FL_BUFFER_SIZE2) +
+ sge->csio_fl_align - 1) & ~(sge->csio_fl_align - 1),
+ SGE_FL_BUFFER_SIZE2);
+ csio_wr_reg32(hw,
+ (csio_rd_reg32(hw, SGE_FL_BUFFER_SIZE3) +
+ sge->csio_fl_align - 1) & ~(sge->csio_fl_align - 1),
+ SGE_FL_BUFFER_SIZE3);
+
+ csio_wr_reg32(hw, HPZ0(PAGE_SHIFT - 12), ULP_RX_TDDP_PSZ);
+
+ /* default value of rx_dma_offset of the NIC driver */
+ csio_set_reg_field(hw, SGE_CONTROL, PKTSHIFT_MASK,
+ PKTSHIFT(CSIO_SGE_RX_DMA_OFFSET));
+}
+
+static void
+csio_init_intr_coalesce_parms(struct csio_hw *hw)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_sge *sge = &wrm->sge;
+
+ csio_sge_thresh_reg = csio_closest_thresh(sge, csio_intr_coalesce_cnt);
+ if (csio_intr_coalesce_cnt) {
+ csio_sge_thresh_reg = 0;
+ csio_sge_timer_reg = X_TIMERREG_RESTART_COUNTER;
+ return;
+ }
+
+ csio_sge_timer_reg = csio_closest_timer(sge, csio_intr_coalesce_time);
+}
+
+/*
+ * csio_wr_get_sge - Get SGE register values.
+ * @hw: HW module.
+ *
+ * Used by non-master functions and by master-functions relying on config file.
+ */
+static void
+csio_wr_get_sge(struct csio_hw *hw)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_sge *sge = &wrm->sge;
+ uint32_t ingpad;
+ int i;
+ u32 timer_value_0_and_1, timer_value_2_and_3, timer_value_4_and_5;
+ u32 ingress_rx_threshold;
+
+ sge->sge_control = csio_rd_reg32(hw, SGE_CONTROL);
+
+ ingpad = INGPADBOUNDARY_GET(sge->sge_control);
+
+ switch (ingpad) {
+ case X_INGPCIEBOUNDARY_32B:
+ sge->csio_fl_align = 32; break;
+ case X_INGPCIEBOUNDARY_64B:
+ sge->csio_fl_align = 64; break;
+ case X_INGPCIEBOUNDARY_128B:
+ sge->csio_fl_align = 128; break;
+ case X_INGPCIEBOUNDARY_256B:
+ sge->csio_fl_align = 256; break;
+ case X_INGPCIEBOUNDARY_512B:
+ sge->csio_fl_align = 512; break;
+ case X_INGPCIEBOUNDARY_1024B:
+ sge->csio_fl_align = 1024; break;
+ case X_INGPCIEBOUNDARY_2048B:
+ sge->csio_fl_align = 2048; break;
+ case X_INGPCIEBOUNDARY_4096B:
+ sge->csio_fl_align = 4096; break;
+ }
+
+ for (i = 0; i < CSIO_SGE_FL_SIZE_REGS; i++)
+ csio_get_flbuf_size(hw, sge, i);
+
+ timer_value_0_and_1 = csio_rd_reg32(hw, SGE_TIMER_VALUE_0_AND_1);
+ timer_value_2_and_3 = csio_rd_reg32(hw, SGE_TIMER_VALUE_2_AND_3);
+ timer_value_4_and_5 = csio_rd_reg32(hw, SGE_TIMER_VALUE_4_AND_5);
+
+ sge->timer_val[0] = (uint16_t)csio_core_ticks_to_us(hw,
+ TIMERVALUE0_GET(timer_value_0_and_1));
+ sge->timer_val[1] = (uint16_t)csio_core_ticks_to_us(hw,
+ TIMERVALUE1_GET(timer_value_0_and_1));
+ sge->timer_val[2] = (uint16_t)csio_core_ticks_to_us(hw,
+ TIMERVALUE2_GET(timer_value_2_and_3));
+ sge->timer_val[3] = (uint16_t)csio_core_ticks_to_us(hw,
+ TIMERVALUE3_GET(timer_value_2_and_3));
+ sge->timer_val[4] = (uint16_t)csio_core_ticks_to_us(hw,
+ TIMERVALUE4_GET(timer_value_4_and_5));
+ sge->timer_val[5] = (uint16_t)csio_core_ticks_to_us(hw,
+ TIMERVALUE5_GET(timer_value_4_and_5));
+
+ ingress_rx_threshold = csio_rd_reg32(hw, SGE_INGRESS_RX_THRESHOLD);
+ sge->counter_val[0] = THRESHOLD_0_GET(ingress_rx_threshold);
+ sge->counter_val[1] = THRESHOLD_1_GET(ingress_rx_threshold);
+ sge->counter_val[2] = THRESHOLD_2_GET(ingress_rx_threshold);
+ sge->counter_val[3] = THRESHOLD_3_GET(ingress_rx_threshold);
+
+ csio_init_intr_coalesce_parms(hw);
+}
+
+/*
+ * csio_wr_set_sge - Initialize SGE registers
+ * @hw: HW module.
+ *
+ * Used by Master function to initialize SGE registers in the absence
+ * of a config file.
+ */
+static void
+csio_wr_set_sge(struct csio_hw *hw)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_sge *sge = &wrm->sge;
+ int i;
+
+ /*
+ * Set up our basic SGE mode to deliver CPL messages to our Ingress
+ * Queue and Packet Date to the Free List.
+ */
+ csio_set_reg_field(hw, SGE_CONTROL, RXPKTCPLMODE, RXPKTCPLMODE);
+
+ sge->sge_control = csio_rd_reg32(hw, SGE_CONTROL);
+
+ /* sge->csio_fl_align is set up by csio_wr_fixup_host_params(). */
+
+ /*
+ * Set up to drop DOORBELL writes when the DOORBELL FIFO overflows
+ * and generate an interrupt when this occurs so we can recover.
+ */
+ csio_set_reg_field(hw, SGE_DBFIFO_STATUS,
+ HP_INT_THRESH(HP_INT_THRESH_MASK) |
+ LP_INT_THRESH(LP_INT_THRESH_MASK),
+ HP_INT_THRESH(CSIO_SGE_DBFIFO_INT_THRESH) |
+ LP_INT_THRESH(CSIO_SGE_DBFIFO_INT_THRESH));
+ csio_set_reg_field(hw, SGE_DOORBELL_CONTROL, ENABLE_DROP,
+ ENABLE_DROP);
+
+ /* SGE_FL_BUFFER_SIZE0 is set up by csio_wr_fixup_host_params(). */
+
+ csio_set_flbuf_size(hw, 1, CSIO_SGE_FLBUF_SIZE1);
+ csio_set_flbuf_size(hw, 2, CSIO_SGE_FLBUF_SIZE2);
+ csio_set_flbuf_size(hw, 3, CSIO_SGE_FLBUF_SIZE3);
+ csio_set_flbuf_size(hw, 4, CSIO_SGE_FLBUF_SIZE4);
+ csio_set_flbuf_size(hw, 5, CSIO_SGE_FLBUF_SIZE5);
+ csio_set_flbuf_size(hw, 6, CSIO_SGE_FLBUF_SIZE6);
+ csio_set_flbuf_size(hw, 7, CSIO_SGE_FLBUF_SIZE7);
+ csio_set_flbuf_size(hw, 8, CSIO_SGE_FLBUF_SIZE8);
+
+ for (i = 0; i < CSIO_SGE_FL_SIZE_REGS; i++)
+ csio_get_flbuf_size(hw, sge, i);
+
+ /* Initialize interrupt coalescing attributes */
+ sge->timer_val[0] = CSIO_SGE_TIMER_VAL_0;
+ sge->timer_val[1] = CSIO_SGE_TIMER_VAL_1;
+ sge->timer_val[2] = CSIO_SGE_TIMER_VAL_2;
+ sge->timer_val[3] = CSIO_SGE_TIMER_VAL_3;
+ sge->timer_val[4] = CSIO_SGE_TIMER_VAL_4;
+ sge->timer_val[5] = CSIO_SGE_TIMER_VAL_5;
+
+ sge->counter_val[0] = CSIO_SGE_INT_CNT_VAL_0;
+ sge->counter_val[1] = CSIO_SGE_INT_CNT_VAL_1;
+ sge->counter_val[2] = CSIO_SGE_INT_CNT_VAL_2;
+ sge->counter_val[3] = CSIO_SGE_INT_CNT_VAL_3;
+
+ csio_wr_reg32(hw, THRESHOLD_0(sge->counter_val[0]) |
+ THRESHOLD_1(sge->counter_val[1]) |
+ THRESHOLD_2(sge->counter_val[2]) |
+ THRESHOLD_3(sge->counter_val[3]),
+ SGE_INGRESS_RX_THRESHOLD);
+
+ csio_wr_reg32(hw,
+ TIMERVALUE0(csio_us_to_core_ticks(hw, sge->timer_val[0])) |
+ TIMERVALUE1(csio_us_to_core_ticks(hw, sge->timer_val[1])),
+ SGE_TIMER_VALUE_0_AND_1);
+
+ csio_wr_reg32(hw,
+ TIMERVALUE2(csio_us_to_core_ticks(hw, sge->timer_val[2])) |
+ TIMERVALUE3(csio_us_to_core_ticks(hw, sge->timer_val[3])),
+ SGE_TIMER_VALUE_2_AND_3);
+
+ csio_wr_reg32(hw,
+ TIMERVALUE4(csio_us_to_core_ticks(hw, sge->timer_val[4])) |
+ TIMERVALUE5(csio_us_to_core_ticks(hw, sge->timer_val[5])),
+ SGE_TIMER_VALUE_4_AND_5);
+
+ csio_init_intr_coalesce_parms(hw);
+}
+
+void
+csio_wr_sge_init(struct csio_hw *hw)
+{
+ /*
+ * If we are master:
+ * - If we plan to use the config file, we need to fixup some
+ * host specific registers, and read the rest of the SGE
+ * configuration.
+ * - If we dont plan to use the config file, we need to initialize
+ * SGE entirely, including fixing the host specific registers.
+ * If we arent the master, we are only allowed to read and work off of
+ * the already initialized SGE values.
+ *
+ * Therefore, before calling this function, we assume that the master-
+ * ship of the card, and whether to use config file or not, have
+ * already been decided. In other words, CSIO_HWF_USING_SOFT_PARAMS and
+ * CSIO_HWF_MASTER should be set/unset.
+ */
+ if (csio_is_hw_master(hw)) {
+ csio_wr_fixup_host_params(hw);
+
+ if (hw->flags & CSIO_HWF_USING_SOFT_PARAMS)
+ csio_wr_get_sge(hw);
+ else
+ csio_wr_set_sge(hw);
+ } else
+ csio_wr_get_sge(hw);
+}
+
+/*
+ * csio_wrm_init - Initialize Work request module.
+ * @wrm: WR module
+ * @hw: HW pointer
+ *
+ * Allocates memory for an array of queue pointers starting at q_arr.
+ */
+csio_retval_t
+csio_wrm_init(struct csio_wrm *wrm, struct csio_hw *hw)
+{
+ int i;
+
+ if (!wrm->num_q) {
+ csio_err(hw, "Num queues is not set\n");
+ return CSIO_INVAL;
+ }
+
+ wrm->q_arr = kzalloc(sizeof(struct csio_q *) * wrm->num_q, GFP_KERNEL);
+ if (!wrm->q_arr)
+ goto err;
+
+ for (i = 0; i < wrm->num_q; i++) {
+ wrm->q_arr[i] = kzalloc(sizeof(struct csio_q), GFP_KERNEL);
+ if (!wrm->q_arr[i]) {
+ while (--i >= 0)
+ kfree(wrm->q_arr[i]);
+ goto err_free_arr;
+ }
+ }
+ wrm->free_qidx = 0;
+
+ return CSIO_SUCCESS;
+
+err_free_arr:
+ kfree(wrm->q_arr);
+err:
+ return CSIO_NOMEM;
+}
+
+/*
+ * csio_wrm_exit - Initialize Work request module.
+ * @wrm: WR module
+ * @hw: HW module
+ *
+ * Uninitialize WR module. Free q_arr and pointers in it.
+ * We have the additional job of freeing the DMA memory associated
+ * with the queues.
+ */
+void
+csio_wrm_exit(struct csio_wrm *wrm, struct csio_hw *hw)
+{
+ int i;
+ uint32_t j;
+ struct csio_q *q;
+ struct csio_dma_buf *buf;
+
+ for (i = 0; i < wrm->num_q; i++) {
+ q = wrm->q_arr[i];
+
+ if (wrm->free_qidx && (i < wrm->free_qidx)) {
+ if (q->type == CSIO_FREELIST) {
+ if (!q->un.fl.bufs)
+ continue;
+ for (j = 0; j < q->credits; j++) {
+ buf = &q->un.fl.bufs[j];
+ if (!buf->vaddr)
+ continue;
+ pci_free_consistent(hw->pdev, buf->len,
+ buf->vaddr,
+ buf->paddr);
+ }
+ kfree(q->un.fl.bufs);
+ }
+ pci_free_consistent(hw->pdev, q->size,
+ q->vstart, q->pstart);
+ }
+ kfree(q);
+ }
+
+ hw->flags &= ~CSIO_HWF_Q_MEM_ALLOCED;
+
+ kfree(wrm->q_arr);
+}
--
1.7.1
^ permalink raw reply related
* Re: [PATCH] net-tcp: fix cwnd reduction for non-sack recovery
From: Neal Cardwell @ 2012-08-23 17:25 UTC (permalink / raw)
To: Nandita Dukkipati; +Cc: Yuchung Cheng, davem, mattmathis, netdev
In-Reply-To: <CAB_+Fg6QhPDPo7o-QUMuOkkAuLWXPZwUoH2uG4q=jmph9Yv9ng@mail.gmail.com>
On Thu, Aug 23, 2012 at 1:21 PM, Nandita Dukkipati <nanditad@google.com> wrote:
> On Thu, Aug 23, 2012 at 10:05 AM, Yuchung Cheng <ycheng@google.com> wrote:
>> The cwnd reduction in fast recovery is based on the number of packets
>> newly delivered per ACK. For non-sack connections every DUPACK
>> signifies a packet has been delivered, but the sender mistakenly
>> skips counting them for cwnd reduction.
>>
>> The fix is to compute newly_acked_sacked after DUPACKs are accounted
>> in sacked_out for non-sack connections.
>>
>> Signed-off-by: Yuchung Cheng <ycheng@google.com>
>> ---
>> net/ipv4/tcp_input.c | 15 +++++++--------
>> 1 files changed, 7 insertions(+), 8 deletions(-)
>>
>> diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
>> index bcfccc5..ce4ffe9 100644
>> --- a/net/ipv4/tcp_input.c
>> +++ b/net/ipv4/tcp_input.c
>> @@ -2930,13 +2930,14 @@ static void tcp_enter_recovery(struct sock *sk, bool ece_ack)
>> * tcp_xmit_retransmit_queue().
>> */
>> static void tcp_fastretrans_alert(struct sock *sk, int pkts_acked,
>> - int newly_acked_sacked, bool is_dupack,
>> + int prior_sacked, bool is_dupack,
>> int flag)
>> {
>> struct inet_connection_sock *icsk = inet_csk(sk);
>> struct tcp_sock *tp = tcp_sk(sk);
>> int do_lost = is_dupack || ((flag & FLAG_DATA_SACKED) &&
>> (tcp_fackets_out(tp) > tp->reordering));
>> + int newly_acked_sacked = 0;
>> int fast_rexmit = 0;
>>
>> if (WARN_ON(!tp->packets_out && tp->sacked_out))
>> @@ -2996,6 +2997,7 @@ static void tcp_fastretrans_alert(struct sock *sk, int pkts_acked,
>> tcp_add_reno_sack(sk);
>> } else
>> do_lost = tcp_try_undo_partial(sk, pkts_acked);
>> + newly_acked_sacked = pkts_acked + tp->sacked_out - prior_sacked;
>> break;
>> case TCP_CA_Loss:
>> if (flag & FLAG_DATA_ACKED)
>> @@ -3017,6 +3019,7 @@ static void tcp_fastretrans_alert(struct sock *sk, int pkts_acked,
>> if (is_dupack)
>> tcp_add_reno_sack(sk);
>> }
>> + newly_acked_sacked = pkts_acked + tp->sacked_out - prior_sacked;
>>
>> if (icsk->icsk_ca_state <= TCP_CA_Disorder)
>> tcp_try_undo_dsack(sk);
>> @@ -3594,7 +3597,6 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
>> int prior_packets;
>> int prior_sacked = tp->sacked_out;
>> int pkts_acked = 0;
>> - int newly_acked_sacked = 0;
>> bool frto_cwnd = false;
>>
>> /* If the ack is older than previous acks
>> @@ -3670,8 +3672,6 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
>> flag |= tcp_clean_rtx_queue(sk, prior_fackets, prior_snd_una);
>>
>> pkts_acked = prior_packets - tp->packets_out;
>> - newly_acked_sacked = (prior_packets - prior_sacked) -
>> - (tp->packets_out - tp->sacked_out);
>>
>> if (tp->frto_counter)
>> frto_cwnd = tcp_process_frto(sk, flag);
>> @@ -3685,7 +3685,7 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
>> tcp_may_raise_cwnd(sk, flag))
>> tcp_cong_avoid(sk, ack, prior_in_flight);
>> is_dupack = !(flag & (FLAG_SND_UNA_ADVANCED | FLAG_NOT_DUP));
>> - tcp_fastretrans_alert(sk, pkts_acked, newly_acked_sacked,
>> + tcp_fastretrans_alert(sk, pkts_acked, prior_sacked,
>> is_dupack, flag);
>> } else {
>> if ((flag & FLAG_DATA_ACKED) && !frto_cwnd)
>> @@ -3702,7 +3702,7 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
>> no_queue:
>> /* If data was DSACKed, see if we can undo a cwnd reduction. */
>> if (flag & FLAG_DSACKING_ACK)
>> - tcp_fastretrans_alert(sk, pkts_acked, newly_acked_sacked,
>> + tcp_fastretrans_alert(sk, pkts_acked, prior_sacked,
>> is_dupack, flag);
>> /* If this ack opens up a zero window, clear backoff. It was
>> * being used to time the probes, and is probably far higher than
>> @@ -3722,8 +3722,7 @@ old_ack:
>> */
>> if (TCP_SKB_CB(skb)->sacked) {
>> flag |= tcp_sacktag_write_queue(sk, skb, prior_snd_una);
>> - newly_acked_sacked = tp->sacked_out - prior_sacked;
>> - tcp_fastretrans_alert(sk, pkts_acked, newly_acked_sacked,
>> + tcp_fastretrans_alert(sk, pkts_acked, prior_sacked,
>> is_dupack, flag);
>> }
>>
>> --
>> 1.7.7.3
>>
> Acked-by: Nandita Dukkipati <nanditad@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
neal
^ permalink raw reply
* Re: [net-next 10/13] igb: Tidy up wrapping for CONFIG_IGB_PTP.
From: Richard Cochran @ 2012-08-23 17:29 UTC (permalink / raw)
To: Vick, Matthew
Cc: Kirsher, Jeffrey T, davem@davemloft.net, netdev@vger.kernel.org,
gospo@redhat.com, sassmann@redhat.com
In-Reply-To: <06DFBC1E25D8024DB214DC7F41A3CD34488DD039@ORSMSX101.amr.corp.intel.com>
On Thu, Aug 23, 2012 at 04:09:30PM +0000, Vick, Matthew wrote:
> > -----Original Message-----
> > From: Richard Cochran [mailto:richardcochran@gmail.com]
> > Sent: Thursday, August 23, 2012 4:04 AM
> > To: Kirsher, Jeffrey T
> > Cc: davem@davemloft.net; Vick, Matthew; netdev@vger.kernel.org;
> > gospo@redhat.com; sassmann@redhat.com
> > Subject: Re: [net-next 10/13] igb: Tidy up wrapping for CONFIG_IGB_PTP.
> >
> > On Thu, Aug 23, 2012 at 02:56:50AM -0700, Jeff Kirsher wrote:
> > > From: Matthew Vick <matthew.vick@intel.com>
> > >
> > > For users without CONFIG_IGB_PTP=y, we should not be compiling any
> > PTP
> > > code into the driver. Tidy up the wrapping in igb to support this.
> >
> > Actually, you are doing more than that. You are adding a bunch of
> > comments onto the already existing #endifs.
>
> Fair enough. Would you like me to update the patch description?
Better would be to drop of the pendantic #endif /*CONFIG_FOO*/ stuff.
It is just churn.
Thanks,
Richard
^ permalink raw reply
* RE: [net-next 10/13] igb: Tidy up wrapping for CONFIG_IGB_PTP.
From: Vick, Matthew @ 2012-08-23 17:40 UTC (permalink / raw)
To: Richard Cochran
Cc: Kirsher, Jeffrey T, davem@davemloft.net, netdev@vger.kernel.org,
gospo@redhat.com, sassmann@redhat.com
In-Reply-To: <20120823172950.GA2192@netboy.at.omicron.at>
> -----Original Message-----
> From: Richard Cochran [mailto:richardcochran@gmail.com]
> Sent: Thursday, August 23, 2012 10:30 AM
> To: Vick, Matthew
> Cc: Kirsher, Jeffrey T; davem@davemloft.net; netdev@vger.kernel.org;
> gospo@redhat.com; sassmann@redhat.com
> Subject: Re: [net-next 10/13] igb: Tidy up wrapping for CONFIG_IGB_PTP.
>
> On Thu, Aug 23, 2012 at 04:09:30PM +0000, Vick, Matthew wrote:
> > > -----Original Message-----
> > > From: Richard Cochran [mailto:richardcochran@gmail.com]
> > > Sent: Thursday, August 23, 2012 4:04 AM
> > > To: Kirsher, Jeffrey T
> > > Cc: davem@davemloft.net; Vick, Matthew; netdev@vger.kernel.org;
> > > gospo@redhat.com; sassmann@redhat.com
> > > Subject: Re: [net-next 10/13] igb: Tidy up wrapping for
> CONFIG_IGB_PTP.
> > >
> > > On Thu, Aug 23, 2012 at 02:56:50AM -0700, Jeff Kirsher wrote:
> > > > From: Matthew Vick <matthew.vick@intel.com>
> > > >
> > > > For users without CONFIG_IGB_PTP=y, we should not be compiling
> any
> > > PTP
> > > > code into the driver. Tidy up the wrapping in igb to support
> this.
> > >
> > > Actually, you are doing more than that. You are adding a bunch of
> > > comments onto the already existing #endifs.
> >
> > Fair enough. Would you like me to update the patch description?
>
> Better would be to drop of the pendantic #endif /*CONFIG_FOO*/ stuff.
> It is just churn.
>
> Thanks,
> Richard
I'm willing to drop it, but I would like to drop it universally in the driver if that's the case. Is that acceptable? There's no overly long or complex wrapping section in the driver that I think merits being sloppy with the end comments.
Cheers,
Matthew
^ permalink raw reply
* Re: Best way to set kernel thread affinity for handling a socket?
From: Ben Hutchings @ 2012-08-23 17:51 UTC (permalink / raw)
To: Roland Dreier; +Cc: netdev
In-Reply-To: <CAG4TOxOdTiahqVo6+cD=mCN5Ov_hLa_m5gBTuj5sUYsM+GgnYw@mail.gmail.com>
On Thu, 2012-08-23 at 10:04 -0700, Roland Dreier wrote:
> On Wed, Aug 22, 2012 at 9:21 AM, Ben Hutchings
> <bhutchings@solarflare.com> wrote:
> > With RFS we try to do the reverse: move the packets to match the socket
> > user. But it's not (yet) turned on by default. See
> > Documentation/networking/scaling.txt
>
> Fair enough. However I think at least in this case it sounds like extra
> overhead: it should be easy for us to do everything on the CPU where
> the packets are being received.
>
> >> I'm thinking about this in the context of the kernel's iSCSI target
> >> code (drivers/target/iscsi), which creates threads to handle each
> >> iSCSI connection and sets their CPU affinity pretty much randomly
> >> (well, based on some "thread id", cf iscsit_thread_get_cpumask()).
> >
> > Why set the affinity at all?
>
> It's quite possible that, like a lot of other drivers/target code, this
> doesn't actually make any sense and the right answer is to rip it
> out completely.
>
> Is the affinity due to waking up from a network receive event
> enough to keep the threads local to the right CPU?
I expect it to have some influence on where the receiving task is
scheduled, but it doesn't necessarily outweigh other factors that the
scheduler takes into account. So you would really have to test this
yourself.
Ben.
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
^ permalink raw reply
* Re: [net-next 11/13] igb: Update PTP function names/variables and locations.
From: Richard Cochran @ 2012-08-23 17:53 UTC (permalink / raw)
To: Vick, Matthew
Cc: Kirsher, Jeffrey T, davem@davemloft.net, netdev@vger.kernel.org,
gospo@redhat.com, sassmann@redhat.com
In-Reply-To: <06DFBC1E25D8024DB214DC7F41A3CD34488DD079@ORSMSX101.amr.corp.intel.com>
On Thu, Aug 23, 2012 at 04:22:02PM +0000, Vick, Matthew wrote:
> > > #ifdef CONFIG_IGB_PTP
> > > -static int igb_ethtool_get_ts_info(struct net_device *dev,
> > > +static int igb_get_ts_info(struct net_device *dev,
> >
> > I like the old name better.
>
> The old name is out of the coding style of igb. Every other function is igb_get_* or igb_set_*, with the exception of igb_ethtool_begin and igb_ethtool_complete. Are you suggesting we add _ethtool to every ethtool function in igb?
No, just leave the names alone, and keep the functions where they
are. It is just churn.
One of the most useful ways to understand code (at least for me) is to
use git blame. It tells you when code was added, what the reason was,
and how the change looks in context. By moving and renaming willy
nilly, you are obscuring this valuable information.
> > > -#ifdef CONFIG_IGB_PTP
> > > -/**
> > > - * igb_tx_hwtstamp - utility function which checks for TX time stamp
> > > - * @q_vector: pointer to q_vector containing needed info
> > > - * @buffer: pointer to igb_tx_buffer structure
> > > - *
> > > - * If we were asked to do hardware stamping and such a time stamp is
> > > - * available, then it must have been for this skb here because we
> > > only
> > > - * allow only one such packet into the queue.
> > > - */
> > > -static void igb_tx_hwtstamp(struct igb_q_vector *q_vector,
> > > - struct igb_tx_buffer *buffer_info)
> > > -{
> > > - struct igb_adapter *adapter = q_vector->adapter;
> > > - struct e1000_hw *hw = &adapter->hw;
> > > - struct skb_shared_hwtstamps shhwtstamps;
> > > - u64 regval;
> > > -
> > > - /* if skb does not support hw timestamp or TX stamp not valid
> > exit */
> > > - if (likely(!(buffer_info->tx_flags & IGB_TX_FLAGS_TSTAMP)) ||
> > > - !(rd32(E1000_TSYNCTXCTL) & E1000_TSYNCTXCTL_VALID))
> > > - return;
> > > -
> > > - regval = rd32(E1000_TXSTMPL);
> > > - regval |= (u64)rd32(E1000_TXSTMPH) << 32;
> > > -
> > > - igb_systim_to_hwtstamp(adapter, &shhwtstamps, regval);
> > > - skb_tstamp_tx(buffer_info->skb, &shhwtstamps);
> > > -}
> > > -#endif /* CONFIG_IGB_PTP */
> > > -
> >
> > Here you have taken a static local function and made it into a global
> > function. This can have performance impacts.
>
> Which, this function calls igb_systim_to_hwtstamp anyway, which is
> global.
So how does calling two global functions in series improve performance?
> Also, in a follow-on patch I have coming, igb_ptp_tx_hwtstamp won't
> even be called in clean_tx_irq, FWIW.
If this is part of some larger plan, then it would help to see that
plan.
> > > /**
> > > * igb_clean_tx_irq - Reclaim resources after transmit completes
> > > * @q_vector: pointer to q_vector containing needed info @@ -5827,7
> > > +5796,7 @@ static bool igb_clean_tx_irq(struct igb_q_vector
> > *q_vector)
> > >
> > > #ifdef CONFIG_IGB_PTP
> > > /* retrieve hardware timestamp */
> > > - igb_tx_hwtstamp(q_vector, tx_buffer);
> > > + igb_ptp_tx_hwtstamp(q_vector, tx_buffer);
> >
> > This name stinks, too. You know that you can have time stamping all by
> > itself, right? It is logically separate from the ptp clock stuff.
> >
> > This patch doesn't really improve the driver at all, IMHO.
> >
> > Thanks,
> > Richard
>
> Yes, I'm aware. But, as it stands today, we don't use it for anything else. If the function is feature specific, then we should be calling it out as such.
Right now the time stamping is being equated with the clock functions,
but it really should be decoupled. The 82580 can time stamp every
received packet, which can be interesting for performance monitoring,
even without PTP (and adding *that* would be a useful change).
> I'm sorry you feel like this patch doesn't improve the driver. The goal is code cleanup and consistency, both of which I consider to be driver improvements and is why I made the patches.
But the code wasn't dirty in the first place. It doesn't need this
"cleaning." This series undoes the inline-able functions for no good
reason. As far as ixgbe goes, this driver came first, so you might as
well be making *that* driver consistent with this one.
Thanks,
Richard
^ permalink raw reply
* Re: Best way to set kernel thread affinity for handling a socket?
From: Chris Friesen @ 2012-08-23 17:56 UTC (permalink / raw)
To: Roland Dreier; +Cc: Ben Hutchings, netdev
In-Reply-To: <CAG4TOxOdTiahqVo6+cD=mCN5Ov_hLa_m5gBTuj5sUYsM+GgnYw@mail.gmail.com>
On 08/23/2012 11:04 AM, Roland Dreier wrote:
> On Wed, Aug 22, 2012 at 9:21 AM, Ben Hutchings
> <bhutchings@solarflare.com> wrote:
>> With RFS we try to do the reverse: move the packets to match the socket
>> user. But it's not (yet) turned on by default. See
>> Documentation/networking/scaling.txt
>
> Fair enough. However I think at least in this case it sounds like extra
> overhead: it should be easy for us to do everything on the CPU where
> the packets are being received.
In the general case though RFS can be useful. We have a server with one
application instance per core. It's beneficial to be able to steer
packets destined for a given instance to an eth driver queue whose
interrupt is handled by the cpu running that instance.
Depending on the hardware being used, it may be possible to do this more
efficiently. The ixgbe driver (if the feature is enabled) will
periodically look at the outgoing tcp traffic and set up hardware flow
filtering rules to direct incoming traffic for that connection to the
queue that sent out the messages.
Chris
^ permalink raw reply
* RE: [net-next 11/13] igb: Update PTP function names/variables and locations.
From: Keller, Jacob E @ 2012-08-23 18:00 UTC (permalink / raw)
To: Richard Cochran, Vick, Matthew
Cc: Kirsher, Jeffrey T, davem@davemloft.net, netdev@vger.kernel.org,
gospo@redhat.com, sassmann@redhat.com
In-Reply-To: <20120823175335.GB2192@netboy.at.omicron.at>
> -----Original Message-----
> From: netdev-owner@vger.kernel.org [mailto:netdev-owner@vger.kernel.org]
> On Behalf Of Richard Cochran
> Sent: Thursday, August 23, 2012 10:54 AM
> To: Vick, Matthew
> Cc: Kirsher, Jeffrey T; davem@davemloft.net; netdev@vger.kernel.org;
> gospo@redhat.com; sassmann@redhat.com
> Subject: Re: [net-next 11/13] igb: Update PTP function names/variables and
> locations.
>
> On Thu, Aug 23, 2012 at 04:22:02PM +0000, Vick, Matthew wrote:
>
> > > > #ifdef CONFIG_IGB_PTP
> > > > -static int igb_ethtool_get_ts_info(struct net_device *dev,
> > > > +static int igb_get_ts_info(struct net_device *dev,
> > >
> > > I like the old name better.
> >
> > The old name is out of the coding style of igb. Every other function is
> igb_get_* or igb_set_*, with the exception of igb_ethtool_begin and
> igb_ethtool_complete. Are you suggesting we add _ethtool to every ethtool
> function in igb?
>
> No, just leave the names alone, and keep the functions where they are. It
> is just churn.
>
> One of the most useful ways to understand code (at least for me) is to use
> git blame. It tells you when code was added, what the reason was, and how
> the change looks in context. By moving and renaming willy nilly, you are
> obscuring this valuable information.
>
> > > > -#ifdef CONFIG_IGB_PTP
> > > > -/**
> > > > - * igb_tx_hwtstamp - utility function which checks for TX time
> > > > stamp
> > > > - * @q_vector: pointer to q_vector containing needed info
> > > > - * @buffer: pointer to igb_tx_buffer structure
> > > > - *
> > > > - * If we were asked to do hardware stamping and such a time stamp
> > > > is
> > > > - * available, then it must have been for this skb here because we
> > > > only
> > > > - * allow only one such packet into the queue.
> > > > - */
> > > > -static void igb_tx_hwtstamp(struct igb_q_vector *q_vector,
> > > > - struct igb_tx_buffer *buffer_info)
> > > > -{
> > > > - struct igb_adapter *adapter = q_vector->adapter;
> > > > - struct e1000_hw *hw = &adapter->hw;
> > > > - struct skb_shared_hwtstamps shhwtstamps;
> > > > - u64 regval;
> > > > -
> > > > - /* if skb does not support hw timestamp or TX stamp not valid
> > > exit */
> > > > - if (likely(!(buffer_info->tx_flags & IGB_TX_FLAGS_TSTAMP)) ||
> > > > - !(rd32(E1000_TSYNCTXCTL) & E1000_TSYNCTXCTL_VALID))
> > > > - return;
> > > > -
> > > > - regval = rd32(E1000_TXSTMPL);
> > > > - regval |= (u64)rd32(E1000_TXSTMPH) << 32;
> > > > -
> > > > - igb_systim_to_hwtstamp(adapter, &shhwtstamps, regval);
> > > > - skb_tstamp_tx(buffer_info->skb, &shhwtstamps);
> > > > -}
> > > > -#endif /* CONFIG_IGB_PTP */
> > > > -
> > >
> > > Here you have taken a static local function and made it into a
> > > global function. This can have performance impacts.
> >
> > Which, this function calls igb_systim_to_hwtstamp anyway, which is
> > global.
>
> So how does calling two global functions in series improve performance?
>
> > Also, in a follow-on patch I have coming, igb_ptp_tx_hwtstamp won't
> > even be called in clean_tx_irq, FWIW.
>
> If this is part of some larger plan, then it would help to see that plan.
>
> > > > /**
> > > > * igb_clean_tx_irq - Reclaim resources after transmit completes
> > > > * @q_vector: pointer to q_vector containing needed info @@
> > > > -5827,7
> > > > +5796,7 @@ static bool igb_clean_tx_irq(struct igb_q_vector
> > > *q_vector)
> > > >
> > > > #ifdef CONFIG_IGB_PTP
> > > > /* retrieve hardware timestamp */
> > > > - igb_tx_hwtstamp(q_vector, tx_buffer);
> > > > + igb_ptp_tx_hwtstamp(q_vector, tx_buffer);
> > >
> > > This name stinks, too. You know that you can have time stamping all
> > > by itself, right? It is logically separate from the ptp clock stuff.
> > >
> > > This patch doesn't really improve the driver at all, IMHO.
> > >
> > > Thanks,
> > > Richard
> >
> > Yes, I'm aware. But, as it stands today, we don't use it for anything
> else. If the function is feature specific, then we should be calling it
> out as such.
>
> Right now the time stamping is being equated with the clock functions, but
> it really should be decoupled. The 82580 can time stamp every received
> packet, which can be interesting for performance monitoring, even without
> PTP (and adding *that* would be a useful change).
>
The timestamp all does not really work with the ptp clock features gone, because you don't have the clock. You can't equate the time values of the packets when the clock isn't synched to something meaningful. Yes that does not require PTP adjustment functions, but it does require the SYSTIME setup and some method to get the clock correct, which currently is only done in the PTP init sequence. Timestamp all packets also can cause a performance hit when used with certain workloads.
> > I'm sorry you feel like this patch doesn't improve the driver. The goal
> is code cleanup and consistency, both of which I consider to be driver
> improvements and is why I made the patches.
>
> But the code wasn't dirty in the first place. It doesn't need this
> "cleaning." This series undoes the inline-able functions for no good
> reason. As far as ixgbe goes, this driver came first, so you might as well
> be making *that* driver consistent with this one.
ixgbe hardware is (currently) even more closely synched with PTP for the register bits so it does make some sense for ixgbe to remain the way it is. Right now the igb features are partially synched (even before this change) in odd ways. The time values returned when PHC information is disabled are basically only useful for comparing between themselves, not with any meaningful clock on the device.
>
> Thanks,
> Richard
>
> --
> To unsubscribe from this list: send the line "unsubscribe netdev" in the
> body of a message to majordomo@vger.kernel.org More majordomo info at
> http://vger.kernel.org/majordomo-info.html
^ permalink raw reply
* Re: [net-next 10/13] igb: Tidy up wrapping for CONFIG_IGB_PTP.
From: Richard Cochran @ 2012-08-23 18:03 UTC (permalink / raw)
To: Vick, Matthew
Cc: Kirsher, Jeffrey T, davem@davemloft.net, netdev@vger.kernel.org,
gospo@redhat.com, sassmann@redhat.com
In-Reply-To: <06DFBC1E25D8024DB214DC7F41A3CD34488DD257@ORSMSX101.amr.corp.intel.com>
On Thu, Aug 23, 2012 at 05:40:01PM +0000, Vick, Matthew wrote:
> > Better would be to drop of the pendantic #endif /*CONFIG_FOO*/ stuff.
> > It is just churn.
> >
> > Thanks,
> > Richard
>
> I'm willing to drop it, but I would like to drop it universally in the driver if that's the case. Is that acceptable? There's no overly long or complex wrapping section in the driver that I think merits being sloppy with the end comments.
Well, I don't know about the other cases.
drivers/net/ethernet/intel/igb$ grep \#endif *.c | grep -v CONFIG | wc -l
28
drivers/net/ethernet/intel/igb$ grep \#endif *.c | grep CONFIG | wc -l
8
(none of these are for _IGB_PTP)
Having #endif comments is useful when it makes the code more clear. It
is a matter of taste, but I do think having such comments for just two
lines in between is ugly and silly.
Thanks,
Richard
^ permalink raw reply
* RE: [net-next 10/13] igb: Tidy up wrapping for CONFIG_IGB_PTP.
From: Keller, Jacob E @ 2012-08-23 18:03 UTC (permalink / raw)
To: Richard Cochran, Vick, Matthew
Cc: Kirsher, Jeffrey T, davem@davemloft.net, netdev@vger.kernel.org,
gospo@redhat.com, sassmann@redhat.com
In-Reply-To: <20120823172950.GA2192@netboy.at.omicron.at>
> -----Original Message-----
> From: netdev-owner@vger.kernel.org [mailto:netdev-owner@vger.kernel.org]
> On Behalf Of Richard Cochran
> Sent: Thursday, August 23, 2012 10:30 AM
> To: Vick, Matthew
> Cc: Kirsher, Jeffrey T; davem@davemloft.net; netdev@vger.kernel.org;
> gospo@redhat.com; sassmann@redhat.com
> Subject: Re: [net-next 10/13] igb: Tidy up wrapping for CONFIG_IGB_PTP.
>
> On Thu, Aug 23, 2012 at 04:09:30PM +0000, Vick, Matthew wrote:
> > > -----Original Message-----
> > > From: Richard Cochran [mailto:richardcochran@gmail.com]
> > > Sent: Thursday, August 23, 2012 4:04 AM
> > > To: Kirsher, Jeffrey T
> > > Cc: davem@davemloft.net; Vick, Matthew; netdev@vger.kernel.org;
> > > gospo@redhat.com; sassmann@redhat.com
> > > Subject: Re: [net-next 10/13] igb: Tidy up wrapping for
> CONFIG_IGB_PTP.
> > >
> > > On Thu, Aug 23, 2012 at 02:56:50AM -0700, Jeff Kirsher wrote:
> > > > From: Matthew Vick <matthew.vick@intel.com>
> > > >
> > > > For users without CONFIG_IGB_PTP=y, we should not be compiling any
> > > PTP
> > > > code into the driver. Tidy up the wrapping in igb to support this.
> > >
> > > Actually, you are doing more than that. You are adding a bunch of
> > > comments onto the already existing #endifs.
> >
> > Fair enough. Would you like me to update the patch description?
>
> Better would be to drop of the pendantic #endif /*CONFIG_FOO*/ stuff.
> It is just churn.
>
Personally disagree here. I do agree that the churn is annoying with how it breaks git blame, however, in general I prefer tags at the end of #ifdefs even for short ones because it increases my ability to quickly spot matches. The end comment aligns with the starting comment, and even for small blocks makes it easier to process. It is less necessary the smaller the block, but I always prefer to have it than not.
That said, it is nice when git blame points to the right place, and the comments aren't super necessary for such short blocks.
- Jake
> Thanks,
> Richard
> --
> To unsubscribe from this list: send the line "unsubscribe netdev" in the
> body of a message to majordomo@vger.kernel.org More majordomo info at
> http://vger.kernel.org/majordomo-info.html
^ permalink raw reply
* Re: [net-next 11/13] igb: Update PTP function names/variables and locations.
From: Richard Cochran @ 2012-08-23 18:11 UTC (permalink / raw)
To: Keller, Jacob E
Cc: Vick, Matthew, Kirsher, Jeffrey T, davem@davemloft.net,
netdev@vger.kernel.org, gospo@redhat.com, sassmann@redhat.com
In-Reply-To: <02874ECE860811409154E81DA85FBB5807857EEA@ORSMSX105.amr.corp.intel.com>
On Thu, Aug 23, 2012 at 06:00:37PM +0000, Keller, Jacob E wrote:
> >
> > Right now the time stamping is being equated with the clock functions, but
> > it really should be decoupled. The 82580 can time stamp every received
> > packet, which can be interesting for performance monitoring, even without
> > PTP (and adding *that* would be a useful change).
> >
> The timestamp all does not really work with the ptp clock features
> gone, because you don't have the clock. You can't equate the time
> values of the packets when the clock isn't synched to something
> meaningful. Yes that does not require PTP adjustment functions, but
> it does require the SYSTIME setup and some method to get the clock
> correct, which currently is only done in the PTP init sequence.
Relative, high resolution time stamps can be interesting all by
themselves. That is why wireshark has a whole menu of timing choices
including relative since start, inter-packet, and so on.
> Timestamp all packets also can cause a performance hit when used with certain workloads.
Only when enabled.
> ixgbe hardware is (currently) even more closely synched with PTP for the register bits so it does make some sense for ixgbe to remain the way it is. Right now the igb features are partially synched (even before this change) in odd ways. The time values returned when PHC information is disabled are basically only useful for comparing between themselves, not with any meaningful clock on the device.
Yes, I agree that igb is a bit oddly synced WRT clock and time
stamping. I would welcome a change to let it have HW time stamping as
an independent feature.
Thanks,
Richard
^ permalink raw reply
* Re: [PATCH 6/8] csiostor: Chelsio FCoE offload driver submission (headers part 1).
From: Love, Robert W @ 2012-08-23 18:15 UTC (permalink / raw)
To: Naresh Kumar Inna
Cc: JBottomley@parallels.com, linux-scsi@vger.kernel.org,
dm@chelsio.com, netdev@vger.kernel.org, chethan@chelsio.com
In-Reply-To: <1345760873-12101-7-git-send-email-naresh@chelsio.com>
On 8/23/2012 3:27 PM, Naresh Kumar Inna wrote:
> This patch contains the first set of the header files for csiostor driver.
>
> Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
> ---
> drivers/scsi/csiostor/csio_defs.h | 143 ++++++
> drivers/scsi/csiostor/csio_fcoe_proto.h | 843 +++++++++++++++++++++++++++++++
I think most of these protocol definitions can be found in
include/scsi/fc/ and possibly incluse/scsi/scsi_transport_fc.h. I think
any protocol stuff missing should be added to the common header files.
Thanks, //Rob
^ permalink raw reply
* Re: [PATCH net-next v3] net: remove delay at device dismantle
From: Ben Hutchings @ 2012-08-23 18:21 UTC (permalink / raw)
To: Eric Dumazet; +Cc: David Miller, gaofeng, netdev, maheshb, therbert, ebiederm
In-Reply-To: <1345691986.5904.40.camel@edumazet-glaptop>
On Thu, 2012-08-23 at 05:19 +0200, Eric Dumazet wrote:
> From: Eric Dumazet <edumazet@google.com>
>
> I noticed extra one second delay in device dismantle, tracked down to
> a call to dst_dev_event() while some call_rcu() are still in RCU queues.
>
> These call_rcu() were posted by rt_free(struct rtable *rt) calls.
>
> We then wait a little (but one second) in netdev_wait_allrefs() before
> kicking again NETDEV_UNREGISTER.
>
> As the call_rcu() are now completed, dst_dev_event() can do the needed
> device swap on busy dst.
>
> To solve this problem, add a new NETDEV_UNREGISTER_FINAL, called
> after a rcu_barrier(), but outside of RTNL lock.
[...]
So what happens when this races with register_netdevice_notifier()?
Ben.
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
^ permalink raw reply
* RE: [PATCH] mwifiex: use is_zero_ether_addr() instead of memcmp()
From: Bing Zhao @ 2012-08-23 18:28 UTC (permalink / raw)
To: Wei Yongjun, linville-2XuSBdqkA4R54TAoqtyWWQ@public.gmane.org
Cc: yongjun_wei-zrsr2BFq86L20UzCJQGyNP8+0UxHXcjY@public.gmane.org,
linux-wireless-u79uwXL29TY76Z2rM5mHXA@public.gmane.org,
netdev-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
In-Reply-To: <CAPgLHd9iXQf3_W5b4kO1Mv68udqnmfFO8n+R-0ZqzCrhg1Z_WQ-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
Hi Wei,
Thanks for the patch.
> From: Wei Yongjun <yongjun_wei-zrsr2BFq86L20UzCJQGyNP8+0UxHXcjY@public.gmane.org>
>
> Using is_zero_ether_addr() instead of directly use
> memcmp() to determine if the ethernet address is all
> zeros.
>
> spatch with a semantic match is used to found this problem.
> (http://coccinelle.lip6.fr/)
Apparently there are two more occurrences of zero_mac (scan.c:843) and bc_mac (sta_cmd.c:596).
But I can fix them when I get the chance.
>
> Signed-off-by: Wei Yongjun <yongjun_wei-zrsr2BFq86L20UzCJQGyNP8+0UxHXcjY@public.gmane.org>
Acked-by: Bing Zhao <bzhao-eYqpPyKDWXRBDgjK7y7TUQ@public.gmane.org>
Thanks,
Bing
> ---
> drivers/net/wireless/mwifiex/sta_cmdresp.c | 5 +++--
> 1 file changed, 3 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/net/wireless/mwifiex/sta_cmdresp.c b/drivers/net/wireless/mwifiex/sta_cmdresp.c
> index 0b09004..a2a33dc 100644
> --- a/drivers/net/wireless/mwifiex/sta_cmdresp.c
> +++ b/drivers/net/wireless/mwifiex/sta_cmdresp.c
> @@ -17,6 +17,8 @@
> * this warranty disclaimer.
> */
>
> +#include <linux/etherdevice.h>
> +
> #include "decl.h"
> #include "ioctl.h"
> #include "util.h"
> @@ -736,7 +738,6 @@ static int mwifiex_ret_ibss_coalescing_status(struct mwifiex_private *priv,
> {
> struct host_cmd_ds_802_11_ibss_status *ibss_coal_resp =
> &(resp->params.ibss_coalescing);
> - u8 zero_mac[ETH_ALEN] = { 0, 0, 0, 0, 0, 0 };
>
> if (le16_to_cpu(ibss_coal_resp->action) == HostCmd_ACT_GEN_SET)
> return 0;
> @@ -745,7 +746,7 @@ static int mwifiex_ret_ibss_coalescing_status(struct mwifiex_private *priv,
> "info: new BSSID %pM\n", ibss_coal_resp->bssid);
>
> /* If rsp has NULL BSSID, Just return..... No Action */
> - if (!memcmp(ibss_coal_resp->bssid, zero_mac, ETH_ALEN)) {
> + if (is_zero_ether_addr(ibss_coal_resp->bssid)) {
> dev_warn(priv->adapter->dev, "new BSSID is NULL\n");
> return 0;
> }
>
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply
* RE: [net-next 11/13] igb: Update PTP function names/variables and locations.
From: Vick, Matthew @ 2012-08-23 18:35 UTC (permalink / raw)
To: Richard Cochran
Cc: Kirsher, Jeffrey T, davem@davemloft.net, netdev@vger.kernel.org,
gospo@redhat.com, sassmann@redhat.com
In-Reply-To: <20120823175335.GB2192@netboy.at.omicron.at>
> -----Original Message-----
> From: Richard Cochran [mailto:richardcochran@gmail.com]
> Sent: Thursday, August 23, 2012 10:54 AM
> To: Vick, Matthew
> Cc: Kirsher, Jeffrey T; davem@davemloft.net; netdev@vger.kernel.org;
> gospo@redhat.com; sassmann@redhat.com
> Subject: Re: [net-next 11/13] igb: Update PTP function names/variables
> and locations.
>
> On Thu, Aug 23, 2012 at 04:22:02PM +0000, Vick, Matthew wrote:
>
> > > > #ifdef CONFIG_IGB_PTP
> > > > -static int igb_ethtool_get_ts_info(struct net_device *dev,
> > > > +static int igb_get_ts_info(struct net_device *dev,
> > >
> > > I like the old name better.
> >
> > The old name is out of the coding style of igb. Every other function
> is igb_get_* or igb_set_*, with the exception of igb_ethtool_begin and
> igb_ethtool_complete. Are you suggesting we add _ethtool to every
> ethtool function in igb?
>
> No, just leave the names alone, and keep the functions where they are.
> It is just churn.
>
> One of the most useful ways to understand code (at least for me) is to
> use git blame. It tells you when code was added, what the reason was,
> and how the change looks in context. By moving and renaming willy
> nilly, you are obscuring this valuable information.
Repairing is not churn. I think it's better to make the code clearer than require developers use git blame to figure out who named a function in a silly way. I agree, it kind of stinks to lose that history, but I'd rather not skip making the right change because of git blame concerns.
>
> [...]
>
> >
> > Which, this function calls igb_systim_to_hwtstamp anyway, which is
> > global.
>
> So how does calling two global functions in series improve performance?
It isn't calling two global functions. It's calling a global function that calls a static function.
> > Also, in a follow-on patch I have coming, igb_ptp_tx_hwtstamp won't
> > even be called in clean_tx_irq, FWIW.
>
> If this is part of some larger plan, then it would help to see that
> plan.
Definitely fair enough. I will work with Jeff to try and push the whole series next time (at the very least, I'd like to re-spin the series for the ethtool query patch).
> > > > /**
> > > > * igb_clean_tx_irq - Reclaim resources after transmit completes
> > > > * @q_vector: pointer to q_vector containing needed info @@
> > > > -5827,7
> > > > +5796,7 @@ static bool igb_clean_tx_irq(struct igb_q_vector
> > > *q_vector)
> > > >
> > > > #ifdef CONFIG_IGB_PTP
> > > > /* retrieve hardware timestamp */
> > > > - igb_tx_hwtstamp(q_vector, tx_buffer);
> > > > + igb_ptp_tx_hwtstamp(q_vector, tx_buffer);
> > >
> > > This name stinks, too. You know that you can have time stamping all
> > > by itself, right? It is logically separate from the ptp clock
> stuff.
> > >
> > > This patch doesn't really improve the driver at all, IMHO.
> > >
> > > Thanks,
> > > Richard
> >
> > Yes, I'm aware. But, as it stands today, we don't use it for anything
> else. If the function is feature specific, then we should be calling it
> out as such.
>
> Right now the time stamping is being equated with the clock functions,
> but it really should be decoupled. The 82580 can time stamp every
> received packet, which can be interesting for performance monitoring,
> even without PTP (and adding *that* would be a useful change).
As Jake said, there's no support for hardware timestamping without PTP today. As such, if the function is only useful for that feature, then I think it's less ambiguous to leave it the way I've coded it today. If we de-couple hardware timestamping and clock synchronization, which I wouldn't be opposed to, then the function name can reflect becoming more generic. As it stands, I'd rather not call something generic when it isn't. I think it's misleading, personally.
> > I'm sorry you feel like this patch doesn't improve the driver. The
> goal is code cleanup and consistency, both of which I consider to be
> driver improvements and is why I made the patches.
>
> But the code wasn't dirty in the first place. It doesn't need this
> "cleaning." This series undoes the inline-able functions for no good
> reason. As far as ixgbe goes, this driver came first, so you might as
> well be making *that* driver consistent with this one.
>
> Thanks,
> Richard
Actually, yes, the code *was* dirty, because it breaks the coding style of the rest of the driver and isn't as self-documenting as it could be, and the inline-able functions you claim I'm destroying called global functions anyway. Also, just because igb came first doesn't mean we should ignore porting better or clearer solutions from later drivers. If something is better or clearer, I think we should use be using it.
Cheers,
Matthew
^ permalink raw reply
* RE: [net-next 10/13] igb: Tidy up wrapping for CONFIG_IGB_PTP.
From: Vick, Matthew @ 2012-08-23 18:40 UTC (permalink / raw)
To: Keller, Jacob E, Richard Cochran
Cc: Kirsher, Jeffrey T, davem@davemloft.net, netdev@vger.kernel.org,
gospo@redhat.com, sassmann@redhat.com
In-Reply-To: <02874ECE860811409154E81DA85FBB5807857EFE@ORSMSX105.amr.corp.intel.com>
> -----Original Message-----
> From: Keller, Jacob E
> Sent: Thursday, August 23, 2012 11:04 AM
> To: Richard Cochran; Vick, Matthew
> Cc: Kirsher, Jeffrey T; davem@davemloft.net; netdev@vger.kernel.org;
> gospo@redhat.com; sassmann@redhat.com
> Subject: RE: [net-next 10/13] igb: Tidy up wrapping for CONFIG_IGB_PTP.
>
> [...]
>
> Personally disagree here. I do agree that the churn is annoying with
> how it breaks git blame, however, in general I prefer tags at the end
> of #ifdefs even for short ones because it increases my ability to
> quickly spot matches. The end comment aligns with the starting comment,
> and even for small blocks makes it easier to process. It is less
> necessary the smaller the block, but I always prefer to have it than
> not.
>
> That said, it is nice when git blame points to the right place, and the
> comments aren't super necessary for such short blocks.
>
> - Jake
I tend to agree with Jake here--I like having the information. I'm fine removing them, but I'd like to do it for all CONFIG_IGB_PTP wrapping if we're going to do it. What do you think, Richard?
Cheers,
Matthew
^ permalink raw reply
* RE: [net-next 11/13] igb: Update PTP function names/variables and locations.
From: Vick, Matthew @ 2012-08-23 18:44 UTC (permalink / raw)
To: Richard Cochran, Keller, Jacob E
Cc: Kirsher, Jeffrey T, davem@davemloft.net, netdev@vger.kernel.org,
gospo@redhat.com, sassmann@redhat.com
In-Reply-To: <20120823181117.GD2192@netboy.at.omicron.at>
> -----Original Message-----
> From: Richard Cochran [mailto:richardcochran@gmail.com]
> Sent: Thursday, August 23, 2012 11:11 AM
> To: Keller, Jacob E
> Cc: Vick, Matthew; Kirsher, Jeffrey T; davem@davemloft.net;
> netdev@vger.kernel.org; gospo@redhat.com; sassmann@redhat.com
> Subject: Re: [net-next 11/13] igb: Update PTP function names/variables
> and locations.
>
> [...]
>
> Relative, high resolution time stamps can be interesting all by
> themselves. That is why wireshark has a whole menu of timing choices
> including relative since start, inter-packet, and so on.
>
> [...]
>
> Yes, I agree that igb is a bit oddly synced WRT clock and time
> stamping. I would welcome a change to let it have HW time stamping as
> an independent feature.
>
> Thanks,
> Richard
Isn't this discussion creeping outside the scope of this patch? My aim here with this patch is to help tidy up PTP code as it is today. As it is today, igb has no support for hardware timestamping without PTP. If you or someone else writes a follow-on patch to add hardware timestamping without PTP, then they can move and rename the functions accordingly (and I'd personally really appreciate it if they did rename it, since a generic function should be named like a generic function and a PTP function should be named like a PTP function).
Cheers,
Matthew
^ permalink raw reply
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox