[PATCH 1/1] Net: qla3xxx, remove sleeping in atomic

[PATCH 1/1] Net: qla3xxx, remove sleeping in atomic

[Jiri Slaby]

Hi, we've found this sleep-inside-spinlock.

The fix below totally untested, please review carefully.

--

We cannot sleep in ql_reset_work under spinlock, unlock before sleep,
relock after.

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
---
 drivers/net/qla3xxx.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/drivers/net/qla3xxx.c b/drivers/net/qla3xxx.c
index 8a823ec..bbc6d4d 100644
--- a/drivers/net/qla3xxx.c
+++ b/drivers/net/qla3xxx.c
@@ -3837,7 +3837,9 @@ static void ql_reset_work(struct work_struct *work)
 						      16) | ISP_CONTROL_RI));
 			}
 
+			spin_unlock_irqrestore(&qdev->hw_lock, hw_flags);
 			ssleep(1);
+			spin_lock_irqsave(&qdev->hw_lock, hw_flags);
 		} while (--max_wait_time);
 		spin_unlock_irqrestore(&qdev->hw_lock, hw_flags);
 
-- 
1.6.3.2
caught by stanse

Re: [PATCH 1/1] Net: qla3xxx, remove sleeping in atomic

[David Miller]

From: Jiri Slaby <jirislaby@gmail.com>
Date: Fri, 19 Jun 2009 17:13:32 +0200

> Hi, we've found this sleep-inside-spinlock.
> 
> The fix below totally untested, please review carefully.
> 
> --
> 
> We cannot sleep in ql_reset_work under spinlock, unlock before sleep,
> relock after.
> 
> Signed-off-by: Jiri Slaby <jirislaby@gmail.com>

This looks good.  The spinlock is just required to be held
around the register accesses.  It's going to drop the lock
as soon as the bits settle post-reset anyways and then the
driver goes off and modifies other state outside of the lock.

Applied, thanks!