On Thu, 01 Mar 2012 00:31:08 +0100 Milan Broz <mbroz@xxxxxxxxxx> wrote: > Hi Neil, > > I am repeatedly getting deadlock with MD raid5 & running fio check. > > array created just this way > # mdadm -C -l 5 -n 4 -c 64 --assume-clean /dev/md0 /dev/sd[bcde] > > and running this test (on quadcore CPU) > > # fio --name=global --rw=randwrite --size=1G --bsrange=1k-128k --filename=/dev/md0 --name=job1 --name=job2 --name=job3 --name=job4 --end_fsync=1 > > deadlocks in final fsync() > > I can reproduce it on Fedora 3.2.7 kernel (and also 3.3.0-rc5), > below is part of the sysrq trace (full sysrq in attached gz archive) > > I was able to simulate it even when resync is running, it stopped > resync process as well. > > Please let me know if you need more information. > Are you certain it is a deadlock? No forward progress at all? What is in md/stripe_cache_size? Does it change? What happens if you double the number in stripe_cache_size? What if you double it again? NeilBrown
Attachment:
signature.asc
Description: PGP signature
