On a more cheerful note - although we do have a regression to fix here, since 6.4 we have fixed A LOT of things, and lots of improvements. As some examples: xfstests generic/075, generic/112 and intermittent failures in generic/047 and generic/049 which used to fail, now pass, and multiple tests have been added to the "buildbot" regression testing automation. And in the future it looks like we will catch these earlier. If anyone has regression tests for bugs that have been fixed over the past few years (that aren't already covered by xfstests), I would be happy to add them to our regression testing bucket of tests we run regularly against various server types. On Mon, Apr 8, 2024 at 5:31 PM Steve French <smfrench@xxxxxxxxx> wrote: > > I narrowed down (bisected) the cause of the crash in unmount > generic/046 (to Windows) which causes the hang on following tests > (generic/047 etc). See > http://smb311-linux-testing.southcentralus.cloudapp.azure.com/#/builders/3/builds/61/steps/73/logs/stdio > as an example. > > The regression started when I add the patch from 6.5-rc1 "cifs: Add a > laundromat thread for cached directories" > > The workaround is to simply disable directory caching for that mount - > ie "nohandlecache" mount parameter. This unmount race is only an > issue to servers which support directory leases (e.g. Windows). > > > > > -- > Thanks, > > Steve -- Thanks, Steve
