netdev.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: clsoto@linux.vnet.ibm.com
To: eli@mellanox.com, roland@kernel.org, sean.hefty@intel.com,
	hal.rosenstock@gmail.com, linux-rdma@vger.kernel.org,
	netdev@vger.kernel.org
Cc: brking@linux.vnet.ibm.com, Carol Soto <clsoto@linux.vnet.ibm.com>
Subject: [Patch 1/2] IB/mlx5: Implementation of PCI error handler
Date: Tue, 11 Mar 2014 22:42:20 -0500	[thread overview]
Message-ID: <20140312034512.065218504@linux.vnet.ibm.com> (raw)
In-Reply-To: 20140312034219.637916521@linux.vnet.ibm.com

[-- Attachment #1: ib_mlx5_add_pci_error.patch --]
[-- Type: text/plain, Size: 3396 bytes --]

This patch is to add PCI error handler function support for mlx5.
Created the functions for error_detected and slot_rest, plus will
send a port down event to users when the driver error_detected
function is invoked. This is to prevent a hang seeing in
mcast_remove_one at the time ib_unregister_device is called for the
ib_sa module. It will fail hardware commands while the driver is
handling a PCI error. It will reduce the hardware commands timeout
to 10 msecs so it does not hang waiting for an interrupt of the
completion of the hardware command.

Signed-off-by: Carol Soto <clsoto@linux.vnet.ibm.com>
---
 drivers/infiniband/hw/mlx5/main.c             |   32 +++++++++++++++++++++++++-
 drivers/net/ethernet/mellanox/mlx5/core/cmd.c |    7 +++++
 include/linux/mlx5/driver.h                   |    4 +--
 3 files changed, 40 insertions(+), 3 deletions(-)

Index: b/drivers/infiniband/hw/mlx5/main.c
===================================================================
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -1508,11 +1508,41 @@ static DEFINE_PCI_DEVICE_TABLE(mlx5_ib_p
 
 MODULE_DEVICE_TABLE(pci, mlx5_ib_pci_table);
 
+static pci_ers_result_t mlx5_pci_err_detected(struct pci_dev *pdev,
+					      pci_channel_state_t state)
+{
+	struct mlx5_ib_dev *dev = mlx5_pci2ibdev(pdev);
+	struct mlx5_core_dev *mdev = &dev->mdev;
+	u8 port;
+
+	/* To avoid the mcast hang with ipoib up */
+	for (port = 1; port <= dev->mdev.caps.num_ports; port++)
+		mlx5_ib_event(mdev, MLX5_DEV_EVENT_PORT_DOWN, &port);
+
+	remove_one(pdev);
+
+	return state == pci_channel_io_perm_failure ?
+		PCI_ERS_RESULT_DISCONNECT : PCI_ERS_RESULT_NEED_RESET;
+}
+
+static pci_ers_result_t mlx5_pci_slot_reset(struct pci_dev *pdev)
+{
+	int ret = init_one(pdev, 0);
+
+	return ret ? PCI_ERS_RESULT_DISCONNECT : PCI_ERS_RESULT_RECOVERED;
+}
+
+static const struct pci_error_handlers mlx5_err_handler = {
+	.error_detected = mlx5_pci_err_detected,
+	.slot_reset     = mlx5_pci_slot_reset,
+};
+
 static struct pci_driver mlx5_ib_driver = {
 	.name		= DRIVER_NAME,
 	.id_table	= mlx5_ib_pci_table,
 	.probe		= init_one,
-	.remove		= remove_one
+	.remove		= remove_one,
+	.err_handler	= &mlx5_err_handler,
 };
 
 static int __init mlx5_ib_init(void)
Index: b/drivers/net/ethernet/mellanox/mlx5/core/cmd.c
===================================================================
--- a/drivers/net/ethernet/mellanox/mlx5/core/cmd.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/cmd.c
@@ -646,6 +646,13 @@ static int mlx5_cmd_invoke(struct mlx5_c
 	if (callback && page_queue)
 		return -EINVAL;
 
+	if (pci_channel_offline(dev->pdev)) {
+		/* Device is going through error recovery
+		 * and cannot accept commands.
+		 */
+		return -EIO;
+	}
+
 	ent = alloc_cmd(cmd, in, out, uout, uout_size, callback, context,
 			page_queue);
 	if (IS_ERR(ent))
Index: b/include/linux/mlx5/driver.h
===================================================================
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -51,10 +51,10 @@ enum {
 };
 
 enum {
-	/* one minute for the sake of bringup. Generally, commands must always
+	/* 10 msecs for the sake of bringup. Generally, commands must always
 	 * complete and we may need to increase this timeout value
 	 */
-	MLX5_CMD_TIMEOUT_MSEC	= 7200 * 1000,
+	MLX5_CMD_TIMEOUT_MSEC	= 10 * 1000,
 	MLX5_CMD_WQ_MAX_NAME	= 32,
 };
 

-- 

  reply	other threads:[~2014-03-12  3:58 UTC|newest]

Thread overview: 11+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2014-03-12  3:42 [Patch 0/2] IB/mlx5: Add PCI error handler support for mlx5 clsoto-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8
2014-03-12  3:42 ` clsoto [this message]
     [not found]   ` <20140312034512.065218504-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>
2014-03-12 18:34     ` [Patch 1/2] IB/mlx5: Implementation of PCI error handler Ben Hutchings
     [not found]       ` <1394649252.23624.36.camel-nDn/Rdv9kqW9Jme8/bJn5UCKIB8iOfG2tUK59QYPAWc@public.gmane.org>
2014-03-12 22:00         ` Carol Soto
2014-03-13  6:45         ` Eli Cohen
2014-03-13 15:12           ` Carol Soto
2014-03-13 15:40             ` Eli Cohen
2014-03-13 15:51               ` Carol Soto
2014-03-13 16:03                 ` Eli Cohen
2014-03-13 16:26                   ` Carol Soto
2014-03-12  3:42 ` [Patch 2/2] IB/mlx5: Free resources during PCI error clsoto-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20140312034512.065218504@linux.vnet.ibm.com \
    --to=clsoto@linux.vnet.ibm.com \
    --cc=brking@linux.vnet.ibm.com \
    --cc=eli@mellanox.com \
    --cc=hal.rosenstock@gmail.com \
    --cc=linux-rdma@vger.kernel.org \
    --cc=netdev@vger.kernel.org \
    --cc=roland@kernel.org \
    --cc=sean.hefty@intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).