Steve French wrote on Mon, Aug 26, 2024 at 11:27:33PM -0500: > I have also confirmed your theory that the regressions (there are > multiple) were likely caused by the netfs change added between > 6.11-rc3 and 6.11-rc4: > " 9p: Fix DIO read through netfs" > > But reverting the cifs.ko part of that patch alters the error but does > not completely fix the problem, so the netfs change is also related David sent a bunch of cifs fixes including this patch: https://lore.kernel.org/r/20240823200819.532106-8-dhowells@xxxxxxxxxx "netfs, cifs: Fix handling of short DIO read" I don't have any samba server around to try myself, did you have a chance to have a look? That "9p" commit touches all netfs filesystems and definitely shouldn't have been labeled 9p (even if it does fix a 9p regression from an earlier netfs commit...), and this all feels a bit like falling forward but hopefully we can get this all fixed by the time 6.11 is ready... -- Dominique
