On Wed, 2018-02-14 at 06:58 +0100, Mike Galbraith wrote: > On Tue, 2018-02-13 at 15:30 +0100, Mike Galbraith wrote: > > On Tue, 2018-02-13 at 09:15 +0100, Paolo Valente wrote: > > > > > > > > > No bfq in the trace, so probably yet another problem stemming from the > > > interaction between bfq and blk-mq. Hung tasks seem to wait > > > indefinitely for a tag, and the only bfq patch that affects tag > > > distribution is: > > > a52a69ea89dc block, bfq: limit tags for writes and async I/O > > > > > > So, for the next step (and as you already decided to do), please retry > > > after adding the above commit to your initial pair: > > > patches/block-bfq:-fix-occurrences-of-request-finish-method-s-old-name.patch > > > patches/block-bfq:-add-requeue-request-hook.patch > > > > I did it a bit differently, revert that commit in master, and skip it > > in 4.15.3 with the otherwise full set first, and saw no hangs testing > > either. I then added it to the 2 patch series, and was rewarded > > with... no bloody hang. So it's not as deterministic as it had > > appeared to be, there's a phase-of-moon factor. Oh. Joy. > > Bah. I'm back on 4.15.3 with the full set, and it's holding up fine > too, so file this report under "elves 'n' gremlins". And _of course_, roughly two minutes later, IO stalled. Ok gremlin, back to deadline I go, scoot, go wreck somebody else's day ;-) -Mike