On Mon, Apr 16, 2018 at 10:52:47PM +0100, Alan Jenkins wrote: > > Without this fix, I get an IO error in this test: > > > > # dd if=/dev/sda of=/dev/null iflag=direct & \ > > while killall -SIGUSR1 dd; do sleep 0.1; done & \ > > echo mem > /sys/power/state ; \ > > sleep 5; killall dd # stop after 5 seconds > > linux-block insisted they wanted this, based on my reproducer above. > If you start wondering why you wouldn't base it on scsi_debug with a new > "quiesce" option, that makes two of us.
Thanks for doing this: > + # Maybe annoying for Xen dom0, but I'm guessing it's not common. > + if grep -q ^flags.*\ hypervisor /proc/cpuinfo && > + (( !HAVE_BARE_METAL_SCSI )); then > + SKIP_REASON=\ > +"Hypervisor detected, but this test wants bare-metal SCSI timings. > +If you have a pass-through device, you may set HAVE_BARE_METAL_SCSI=1." > + return 1 > + fi I don't think we need this, if people want to shoot themselves in the foot by runnning a possibly destructive test suite in a HV we shouldn't stop them. > + > + _have_fio Not needed anymore as per below comment? > + # I tried using fio with the exact same IO pattern, > + # sadly it was not 100% reliable at reproducing this > + # issue. > + # > + dd iflag=direct if="$TEST_DEV" of=/dev/null bs=$bs status=none & > + dd_pid=$! > + > +# fio --output="$FULL" --loops=65535 \ > +# --thread --exitall_on_error \ > +# --direct=1 \ > +# --bs=$bs --rw=read --name=reads \ > +# --filename="$TEST_DEV" & > +# fio_pid=$! I think we can just zap the commented out fio part and the comment about it. -- Johannes Thumshirn Storage [email protected] +49 911 74053 689 SUSE LINUX GmbH, Maxfeldstr. 5, 90409 Nürnberg GF: Felix Imendörffer, Jane Smithard, Graham Norton HRB 21284 (AG Nürnberg) Key fingerprint = EC38 9CAB C2C4 F25D 8600 D0D0 0393 969D 2D76 0850
