[patch] sgi-xp: nested calls to spin_lock_irqsave()

[patch] sgi-xp: nested calls to spin_lock_irqsave()

[Robin Holt]

The code here has a nested spin_lock_irqsave().  It's not needed since
IRQs are already disabled and it causes a problem because it means that
IRQs won't be enabled again at the end.  The second call to
spin_lock_irqsave() will overwrite the value of irq_flags and we can't
restore the proper settings.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Robin Holt <holt@sgi.com>

diff --git a/drivers/misc/sgi-xp/xpc_uv.c b/drivers/misc/sgi-xp/xpc_uv.c
index 17bbacb..87b251a 100644
--- a/drivers/misc/sgi-xp/xpc_uv.c
+++ b/drivers/misc/sgi-xp/xpc_uv.c
@@ -452,9 +452,9 @@ xpc_handle_activate_mq_msg_uv(struct xpc_partition *part,
 
 		if (msg->activate_gru_mq_desc_gpa !=
 		    part_uv->activate_gru_mq_desc_gpa) {
-			spin_lock_irqsave(&part_uv->flags_lock, irq_flags);
+			spin_lock(&part_uv->flags_lock);
 			part_uv->flags &= ~XPC_P_CACHED_ACTIVATE_GRU_MQ_DESC_UV;
-			spin_unlock_irqrestore(&part_uv->flags_lock, irq_flags);
+			spin_unlock(&part_uv->flags_lock);
 			part_uv->activate_gru_mq_desc_gpa =
 			    msg->activate_gru_mq_desc_gpa;
 		}

Re: [patch] sgi-xp: nested calls to spin_lock_irqsave()

[Andrew Morton]

On Wed, 27 Jun 2012 08:51:55 -0500
Robin Holt <holt@sgi.com> wrote:

> The code here has a nested spin_lock_irqsave().  It's not needed since
> IRQs are already disabled and it causes a problem because it means that
> IRQs won't be enabled again at the end.  The second call to
> spin_lock_irqsave() will overwrite the value of irq_flags and we can't
> restore the proper settings.

Please find large crayon and write on forehead "when fixing a bug,
be sure to describe the end-user impact of that bug".

Then others will have a chance of deciding whether the fix should be
backported.

> Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
> Signed-off-by: Robin Holt <holt@sgi.com>

I'm wondering if there should have been a Reported-by:dan in there. 
He's like that.

Re: [patch] sgi-xp: nested calls to spin_lock_irqsave()

[Dan Carpenter]

This is actually my patch, so the patch description is my fault.
It's a static checker patch and I don't have the hardware, but the
changelog could have been written more clearly as well.

Robin, the trick to resending patches is to put a:

From:  Dan Carpenter <dan.carpenter@oracle.com>

before the patch description and then it shows up as mine.

Also, btw, Robin, I'm looking through my mailbox again and you did
respond to the patch the first time I sent it.  Sorry, that I didn't
see that before, and thanks for your help on this.

regards,
dan carpenter

[patch] sgi-xp: nested calls to spin_lock_irqsave()

[Dan Carpenter]

The code here has a nested spin_lock_irqsave().  It's not needed since
IRQs are already disabled and it causes a problem because it means that
IRQs won't be enabled again at the end.  The second call to
spin_lock_irqsave() will overwrite the value of irq_flags and we can't
restore the proper settings.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
---
This is a static checker fix.  It's possible that no one hits this bug
in real life if the function is always called with IRQs disabled.

diff --git a/drivers/misc/sgi-xp/xpc_uv.c b/drivers/misc/sgi-xp/xpc_uv.c
index 17bbacb..87b251a 100644
--- a/drivers/misc/sgi-xp/xpc_uv.c
+++ b/drivers/misc/sgi-xp/xpc_uv.c
@@ -452,9 +452,9 @@ xpc_handle_activate_mq_msg_uv(struct xpc_partition *part,
 
 		if (msg->activate_gru_mq_desc_gpa !=
 		    part_uv->activate_gru_mq_desc_gpa) {
-			spin_lock_irqsave(&part_uv->flags_lock, irq_flags);
+			spin_lock(&part_uv->flags_lock);
 			part_uv->flags &= ~XPC_P_CACHED_ACTIVATE_GRU_MQ_DESC_UV;
-			spin_unlock_irqrestore(&part_uv->flags_lock, irq_flags);
+			spin_unlock(&part_uv->flags_lock);
 			part_uv->activate_gru_mq_desc_gpa =
 			    msg->activate_gru_mq_desc_gpa;
 		}

Re: [patch] sgi-xp: nested calls to spin_lock_irqsave()

[Robin Holt]

On Thu, Dec 15, 2011 at 09:48:05AM +0300, Dan Carpenter wrote:
> The code here has a nested spin_lock_irqsave().  It's not needed since
> IRQs are already disabled and it causes a problem because it means that
> IRQs won't be enabled again at the end.  The second call to
> spin_lock_irqsave() will overwrite the value of irq_flags and we can't
> restore the proper settings.
> 
> Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>

Acked-by: Robin Holt <holt@sgi.com>