On 2017-04-10 at 11:49 "Dan Cross (Gerrit)"
<[email protected]> wrote:
> spin_lock -> spin_lock_irqsave and matching unlock
>
> I discovered a kernel panic when the mlx driver tried
> to spinlock an IRQ-saving lock. Usage elsewhere in this
> file seems to be OK. Spinlock debugging FTW.
i doubt this will be the last of these. can you look for other
violations in the mlx4 code? for instance, a quick grep for spin_lock
in k/d/n/mlx4 and mlx4u turn up a few that look bad. e.g.
kern/drivers/net/mlx4u/srq.c: spin_lock_init(&srq->lock);
kern/drivers/net/mlx4u/srq.c: spin_lock(&srq->lock);
that's a linux-style init (meaning, irqsave) matched with a non-irqsave
lock.
or this one:
kern/drivers/net/mlx4/srq.c: spin_lock(&srq_table->lock);
kern/drivers/net/mlx4/srq.c: spin_lock_irqsave(&srq_table->lock);
same lock, two different mechanisms for locking it. possibly wrong
(though it might not be if IRQs are disabled already for the first
spin_lock).
the whole #define spinlock stuff in mlx4u and udrvr is a source for
future bugs of this sort.
incidentally, gerrit gives me this error:
Code Review - Error
The page you requested was not found, or you do not have
permission to view this page.
i can see other pages on gerrit, such as ron's recent
https://akaros-review.googlesource.com/#/c/3500/
barret
--
You received this message because you are subscribed to the Google Groups
"Akaros" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
For more options, visit https://groups.google.com/d/optout.
