> On Fri, Jun 03, 2016 at 11:17:16AM +0200, Philipp Hahn wrote: > > Hello, > > > > Am 02.06.2016 um 22:02 schrieb J. Bruce Fields: > > > On Thu, Jun 02, 2016 at 09:51:27AM +0200, Philipp Hahn wrote: > > >> probably during heavy IO our file server crashed on a BUG_ON in dache.c, > > >> probably triggered by NFS: > > >> > > >>> ------------[ cut here ]------------ > > >>> kernel BUG at /var/build/temp/tmp.BPql4ErveJ/pbuilder/linux-3.16.7-ckt25/fs/dcache.c:2373! > > ... > > >> Our kernel is 3.16.7-ckt25 from Debian + ckt26 and ckt27 on top. > > > > > > So this is the BUG_ON(!d_unhashed(..)) at the top of __d_rehash? I > > > don't know the -ckt kernels, and I don't have a BUG at line 2373 of > > > dcache.c, or something else? > > > > Yes, exactly that first line. > Oh, from googling... it looks like that's what they're calling the > Ubuntu kernels these days? It'd be best to have them triage this first. > > > There have been a lot of changes in the area since 3.16.7. > > > > Too bad. > > Any concrete hint or should I better spent my time convincing management > > to install a newer Linux kernel? > I don't know if a newer kernel's better, but if you were able to > reproduce this frequently then experimenting with different kernel > versions might be one way to work this out. > I suppose 75a2352d0110960aeee1a28ddc09a55f97c99100 might in theory be a > fix? Not a tall sure of that, though. I've been seeing this bug regularly as well and raised a Debian bug report[1] - is it likely that changing the underlying filesystem may avoid the bug? The system where I see the bug has both ext4 on LVM and some btrfs - are there any other possible workarounds, such as disabling SMP or changing filesystem mount options on the server or the NFS clients? - Debian has kernel 4.7.8-1~bpo8+1 in the backports for stable/jessie users[1], is it confirmed that kernel fixes the problem? - if somebody updates their NFS server's kernel from 3.16 to 4.7.8-1~bpo8+1, will the NFS userland stuff be compatible? It looks like the version[2] in stable is 1.2.8-9 while testing and unstable have 1.2.8-9.2. The difference between the two is summarized in the changelog[3], are any of these recent fixes serious enough that there should be a stable-update or backport of 1.2.8-9.2? Regards, Daniel 1. http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=847549 2. https://packages.qa.debian.org/n/nfs-utils.html 3. http://metadata.ftp-master.debian.org/changelogs/main/n/nfs-utils/nfs-utils_1.2.8-9.2_changelog -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
