Hi On Sat, Oct 5, 2013 at 7:44 AM, Duncan <1i5t5.dun...@cox.net> wrote: > Anatol Pomozov posted on Sat, 05 Oct 2013 04:51:52 -0700 as excerpted: > >> Hi >> >> On Fri, Oct 4, 2013 at 9:42 PM, Duncan <1i5t5.dun...@cox.net> wrote: >>> Anatol Pomozov posted on Fri, 04 Oct 2013 21:03:11 -0700 as excerpted: >>> >>>> Hi, >>>> >>>> I have a home server on Linux Arch (kernel 3.11.2) that uses >>>> multi-device btrfs on root filesystem. >>>> >>>> Until recently it worked completely fine. And yesterday I rebooted it >>>> and the machine did not wake up. >>>> >>>> I booted from a USB (kernel 3.10) and tried to mount the filesystem. >>>> Here is OOPs I see >>> >>>> ------------[ cut here ]------------ >>>> [ 68.126138] kernel BUG at fs/btrfs/inode.c:873! >>>> [ 68.126164] invalid opcode: 0000 [#1] PREEMPT SMP >>> >>>> The only thing that I did recently is defrag /var/log/journal files >>>> (journalctl is very slow because of btrfs COW). Something like this >>>> http://www.mail-archive.com/linux-btrfs@vger.kernel.org/msg24878.html > > I forgot I was going to mention this in the previous reply... > > You can try setting the NOCOW attribute for that file or directory. > There's instructions on the wiki (btrfs.wiki.kernel.org). Virtual > machine images also often work better with NOCOW. > > Meanwhile, thinking about systemd journal files there's also another > patch that is too new to be in the mainline kernel, I think. That bug > was found on systemd journal files specifically, because systemd was > allocating and then writing into them and btrfs wasn't doing the right > thing with that. There's a current thread about it. Since I don't use > systemd, however, I've not followed it that closely.
It is good to know that btrfs developers aware of the journald performance issue. It is really annoying for those who uses btrfs+systemd. > >>> I'm not a dev just a btrfs user and list regular myself, so the traces >>> don't mean much to me. However, the bit I retained in the quote above >>> (especially the 0000 opcode) looks very much like a bug that should be >>> fixed in kernel 3.12 > >>> So the first thing I'd try is either cherrypicking the btrfs patches >>> from 3.12 back to 3.11-stable, or wait for 3.11.5 and check for btrfs >>> patches there, or try 3.12-rcX (rc3 is out and I guess rc4 should be >>> out shortly now as I think it has been nearly a week). >> >> Could you please give me the patch SHA1 you are talking about? > > Sorry, I've not tracked it /that/ closely, as I'm not a dev so the real > technical stuff is over my head, and I run rc kernels from about rc2 > anyway, so I have the fixes reasonably fast already. I simply try to > keep up with the general gist of things well enough to search the list > for that thread I remembered if I need to, and you should be able to do > that as well as I, now that you know the threads and patches are there. Ok, so I still need a fix/workaround for the crash I have. I booted from USB and ran btrfs-zero-log then mounted the filesystems with "-orecovery". The files on the FS look fine. When I try to unmount the fs it hangs in btrfs-transacti (see stacktrace above). When I mount the FS as read-only it unmounts fine. Actually I remembered that I set "chattr +C" on /var/log/journal recursively (even for non-empty files) about a week ago, it might be related to the crash. When I mount "-orw" and try to remove /var/log/journal system hangs in btrfs-transacti thread. -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html