[patch] staging: slicoss: restore IRQs correctly after slic_cmdq_reset()

[patch] staging: slicoss: restore IRQs correctly after slic_cmdq_reset()

[Dan Carpenter]

We can't save two different values in "flags" so it means that IRQs are
not enabled properly at the end of this function.  This isn't a problem
in the current code because it's always called with IRQs disabled so we
don't want to enable them at the end.

This bug is old but it's thanks to David Matlack's recent cleanups that
Smatch can detect it.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
---
The code prior to David Matlack's change was way worse because everyone
shared the same IRQ flags.  It's hard to imagine how it worked at all.

diff --git a/drivers/staging/slicoss/slicoss.c b/drivers/staging/slicoss/slicoss.c
index 5f34ebbf..a609f3e 100644
--- a/drivers/staging/slicoss/slicoss.c
+++ b/drivers/staging/slicoss/slicoss.c
@@ -1392,7 +1392,7 @@ static void slic_cmdq_reset(struct adapter *adapter)
 	unsigned long flags;
 
 	spin_lock_irqsave(&adapter->cmdq_free.lock, flags);
-	spin_lock_irqsave(&adapter->cmdq_done.lock, flags);
+	spin_lock(&adapter->cmdq_done.lock);
 	outstanding = adapter->cmdq_all.count - adapter->cmdq_done.count;
 	outstanding -= adapter->cmdq_free.count;
 	hcmd = adapter->cmdq_all.head;
@@ -1423,7 +1423,7 @@ static void slic_cmdq_reset(struct adapter *adapter)
 			"free_count %d != all count %d\n",
 			adapter->cmdq_free.count, adapter->cmdq_all.count);
 	}
-	spin_unlock_irqrestore(&adapter->cmdq_done.lock, flags);
+	spin_unlock(&adapter->cmdq_done.lock);
 	spin_unlock_irqrestore(&adapter->cmdq_free.lock, flags);
 }
 

Re: [patch] staging: slicoss: restore IRQs correctly after slic_cmdq_reset()

[David Matlack]

On Tue, May 19, 2015 at 5:25 AM, Dan Carpenter <dan.carpenter@oracle.com> wrote:
> We can't save two different values in "flags" so it means that IRQs are
> not enabled properly at the end of this function.  This isn't a problem
> in the current code because it's always called with IRQs disabled so we
> don't want to enable them at the end.

Nice find. Thanks for fixing it!

>
> This bug is old but it's thanks to David Matlack's recent cleanups that
> Smatch can detect it.
>
> Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>

Reviewed-by: David Matlack <dmatlack@google.com>

> ---
> The code prior to David Matlack's change was way worse because everyone
> shared the same IRQ flags.  It's hard to imagine how it worked at all.
>
> diff --git a/drivers/staging/slicoss/slicoss.c b/drivers/staging/slicoss/slicoss.c
> index 5f34ebbf..a609f3e 100644
> --- a/drivers/staging/slicoss/slicoss.c
> +++ b/drivers/staging/slicoss/slicoss.c
> @@ -1392,7 +1392,7 @@ static void slic_cmdq_reset(struct adapter *adapter)
>         unsigned long flags;
>
>         spin_lock_irqsave(&adapter->cmdq_free.lock, flags);
> -       spin_lock_irqsave(&adapter->cmdq_done.lock, flags);
> +       spin_lock(&adapter->cmdq_done.lock);
>         outstanding = adapter->cmdq_all.count - adapter->cmdq_done.count;
>         outstanding -= adapter->cmdq_free.count;
>         hcmd = adapter->cmdq_all.head;
> @@ -1423,7 +1423,7 @@ static void slic_cmdq_reset(struct adapter *adapter)
>                         "free_count %d != all count %d\n",
>                         adapter->cmdq_free.count, adapter->cmdq_all.count);
>         }
> -       spin_unlock_irqrestore(&adapter->cmdq_done.lock, flags);
> +       spin_unlock(&adapter->cmdq_done.lock);
>         spin_unlock_irqrestore(&adapter->cmdq_free.lock, flags);
>  }
>