2009-04-17 23:04:27

by Divy Le ray

Subject: [PATCH 2.6.30 2/4] cxgb3: fix workqueue flush issues

From: Divy Le Ray <[email protected]>

The fatal error task can be scheduled while processing an offload packet
in NAPI context when the connection handle is bogus. this can race
with the ports being brought down and the cxgb3 workqueue being flushed.
Stop napi processing before flushing the work queue.

The ULP drivers (iSCSI, iWARP) might also schedule a task on keventd_wk
while releasing a connection handle (cxgb3_offload.c::cxgb3_queue_tid_release()).
The driver however does not flush any work on keventd_wq while being unloaded.
This patch also fixes this.

Also call cancel_delayed_work_sync in place of the the deprecated
cancel_rearming_delayed_workqueue.

Signed-off-by: Divy Le Ray <[email protected]>
---

drivers/net/cxgb3/cxgb3_main.c | 8 +++++---
1 files changed, 5 insertions(+), 3 deletions(-)

diff --git a/drivers/net/cxgb3/cxgb3_main.c b/drivers/net/cxgb3/cxgb3_main.c
index 9fdfe0b..99b5032 100644
--- a/drivers/net/cxgb3/cxgb3_main.c
+++ b/drivers/net/cxgb3/cxgb3_main.c
@@ -1117,8 +1117,8 @@ static void cxgb_down(struct adapter *adapter)
spin_unlock_irq(&adapter->work_lock);

free_irq_resources(adapter);
- flush_workqueue(cxgb3_wq); /* wait for external IRQ handler */
quiesce_rx(adapter);
+ flush_workqueue(cxgb3_wq); /* wait for external IRQ handler */
}

static void schedule_chk_task(struct adapter *adap)
@@ -1187,6 +1187,9 @@ static int offload_close(struct t3cdev *tdev)

sysfs_remove_group(&tdev->lldev->dev.kobj, &offload_attr_group);

+ /* Flush work scheduled while releasing TIDs */
+ flush_scheduled_work();
+
tdev->lldev = NULL;
cxgb3_set_dummy_ops(tdev);
t3_tp_set_offload_mode(adapter, 0);
@@ -1247,8 +1250,7 @@ static int cxgb_close(struct net_device *dev)
spin_unlock_irq(&adapter->work_lock);

if (!(adapter->open_device_map & PORT_MASK))
- cancel_rearming_delayed_workqueue(cxgb3_wq,
- &adapter->adap_check_task);
+ cancel_delayed_work_sync(&adapter->adap_check_task);

if (!adapter->open_device_map)
cxgb_down(adapter);

2009-04-20 09:08:55

by David Miller

Subject: Re: [PATCH 2.6.30 2/4] cxgb3: fix workqueue flush issues

From: Divy Le Ray <[email protected]>
Date: Fri, 17 Apr 2009 15:21:17 -0700

> The fatal error task can be scheduled while processing an offload packet
> in NAPI context when the connection handle is bogus. this can race
> with the ports being brought down and the cxgb3 workqueue being flushed.
> Stop napi processing before flushing the work queue.
>
> The ULP drivers (iSCSI, iWARP) might also schedule a task on keventd_wk
> while releasing a connection handle (cxgb3_offload.c::cxgb3_queue_tid_release()).
> The driver however does not flush any work on keventd_wq while being unloaded.
> This patch also fixes this.
>
> Also call cancel_delayed_work_sync in place of the the deprecated
> cancel_rearming_delayed_workqueue.
>
> Signed-off-by: Divy Le Ray <[email protected]>

Applied.