On Tue, Aug 28, 2012 at 06:06:10PM -0400, Vivek Goyal wrote:

On Tue, Aug 28, 2012 at 10:37:36AM -0700, Kent Overstreet wrote:

quoted

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.

Hi Kent,

So above deadlock possibility arises only if multiple threads are doing
multiple splits from same pool and all the threads get blocked on mempool
and don't return from ->make_request function.

Is there any existing driver in the tree which can cause this deadlock or
it becomes a possibility only when splitting and bcache code shows up?

It is emphatically possible with dm, though you would probably need a
pathalogical configuration of your targets for it to be possible in
practice - at least with the targets currently commonly in use.

With most of the newer dm targets coming down the pipe I expect it
should be possible under real world configurations, with the caveat that
under that kind of memory pressure in practice many other things will
fall over first.