[dm-devel] [PATCH 2/2] block: Avoid deadlocks with bio allocation by stacking drivers

[Tejun Heo]

Hello, Kent.

On Mon, Sep 10, 2012 at 02:56:33PM -0700, Kent Overstreet wrote:
> commit df7e63cbffa3065fcc4ba2b9a93418d7c7312243
> Author: Kent Overstreet <koverstreet at google.com>
> Date:   Mon Sep 10 14:33:46 2012 -0700
> 
>     block: Avoid deadlocks with bio allocation by stacking drivers
>     
>     Previously, if we ever try to allocate more than once from the same bio
>     set while running under generic_make_request() (i.e. a stacking block
>     driver), we risk deadlock.
>     
>     This is because of the code in generic_make_request() that converts
>     recursion to iteration; any bios we submit won't actually be submitted
>     (so they can complete and eventually be freed) until after we return -
>     this means if we allocate a second bio, we're blocking the first one
>     from ever being freed.
>     
>     Thus if enough threads call into a stacking block driver at the same
>     time with bios that need multiple splits, and the bio_set's reserve gets
>     used up, we deadlock.
>     
>     This can be worked around in the driver code - we could check if we're
>     running under generic_make_request(), then mask out __GFP_WAIT when we
>     go to allocate a bio, and if the allocation fails punt to workqueue and
>     retry the allocation.
>     
>     But this is tricky and not a generic solution. This patch solves it for
>     all users by inverting the previously described technique. We allocate a
>     rescuer workqueue for each bio_set, and then in the allocation code if
>     there are bios on current->bio_list we would be blocking, we punt them
>     to the rescuer workqueue to be submitted.
>     
>     This guarantees forward progress for bio allocations under
>     generic_make_request() provided each bio is submitted before allocating
>     the next, and provided the bios are freed after they complete.
>     
>     Note that this doesn't do anything for allocation from other mempools.
>     Instead of allocating per bio data structures from a mempool, code
>     should use bio_set's front_pad.
>     
>     Tested it by forcing the rescue codepath to be taken (by disabling the
>     first GFP_NOWAIT) attempt, and then ran it with bcache (which does a lot
>     of arbitrary bio splitting) and verified that the rescuer was being
>     invoked.
>     
>     Signed-off-by: Kent Overstreet <koverstreet at google.com>
>     CC: Jens Axboe <axboe at kernel.dk>

I'm still a bit scared but think this is correct.

 Acked-by: Tejun Heo <tj at kernel.org>

One last thing is that we may want to add @name on bioset creation so
that we can name the workqueue properly but that's for another patch.

Thanks.

-- 
tejun