On Thu, 24 May 2012 11:31:26 +0800 Shaohua Li <shli@xxxxxxxxxx> wrote:
> On Thu, May 24, 2012 at 01:09:12PM +1000, NeilBrown wrote:
> > Not necessarily. It might queue it:
> > if (current->bio_list) {
> > bio_list_add(current->bio_list, bio);
> > return;
> > }
>
> Ah, yes. Looks we can workaround it to increase bioset entry number to
> raid_disks*2. There isn't too much memory wasted with it.
No, that doesn't work. There might be raid_disk*2 separate threads which
have each allocated on entry from the pool, and then they all block waiting
for another.
Adding the pool can change the probability of a deadlock but not the
possibility.
NeilBrown
Attachment:
signature.asc
Description: PGP signature
