Re: alloc_area_pte: page already exists

[Andrea Arcangeli <andrea@suse.de>]

On Thu, Aug 09, 2001 at 05:33:27PM +0200, Bjorn Wesen wrote:
> I realised I'm not entirely sure on if it's ok to do such "dangerous"
> functions inside, say, tq_immediate using queue_task even ? Doesn't that
> run in the interrupt context also, upon exit of the interrupt before going
> back ?

Yes it does, also you should use tasklets, never tq_immediate, these days
(tasklets can run in parallel and they're serialized only against
themself).

> IOW I want the irq to "trigger" the freeing of the iovecs but it's ok if
> it's done later, as long as it's done without any races :)

Your design looks suspect, but you can do that safely (at least as far
as vfree is concerned) with keventd's schedule_task().

> BTW how does vfree cope with not walking all tasks pgd's to remove the
> relevant pte's ? Doesn't that give exactly this kind of problem ? (pte's

vfree as usual walks the pgd/pmd to reach the pte. It knows the
pgd/pmd/pte cannot go away and it serlializes against vmalloc with the
vmlist_lock, it sounds ok.

Andrea