async_tx: use GFP_NOWAIT rather than GFP_IO
authorNeilBrown <neilb@suse.com>
Thu, 7 Jan 2016 00:02:34 +0000 (11:02 +1100)
committerVinod Koul <vinod.koul@intel.com>
Thu, 7 Jan 2016 05:36:18 +0000 (11:06 +0530)
commitb02bab6b0f928d49dbfb03e1e4e9dd43647623d7
tree1d7648d55adc0d47d6da533f7a96a70aa9cf8ec8
parent16605e8d50898ac88b5b504a7fbd63ecdcf37702
async_tx: use GFP_NOWAIT rather than GFP_IO

These async_XX functions are called from md/raid5 in an atomic
section, between get_cpu() and put_cpu(), so they must not sleep.
So use GFP_NOWAIT rather than GFP_IO.

Dan Williams writes: Longer term async_tx needs to be merged into md
directly as we can allocate this unmap data statically per-stripe
rather than per request.

Fixed: 7476bd79fc01 ("async_pq: convert to dmaengine_unmap_data")
Cc: stable@vger.kernel.org (v3.13+)
Reported-and-tested-by: Stanislav Samsonov <slava@annapurnalabs.com>
Acked-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
crypto/async_tx/async_memcpy.c
crypto/async_tx/async_pq.c
crypto/async_tx/async_raid6_recov.c
crypto/async_tx/async_xor.c