From mboxrd@z Thu Jan 1 00:00:00 1970 From: Shane Shrybman Subject: Re: Oops while rebalancing, now unmountable. Date: Tue, 09 Nov 2010 13:21:32 -0500 Message-ID: <1289326892.4231.2.camel@mars> References: <1289236257.3611.3.camel@mars> <1289310046-sup-839@think> Mime-Version: 1.0 Content-Type: text/plain Cc: linux-btrfs To: Chris Mason Return-path: In-Reply-To: <1289310046-sup-839@think> List-ID: On Tue, 2010-11-09 at 08:42 -0500, Chris Mason wrote: > Excerpts from Shane Shrybman's message of 2010-11-08 12:10:57 -0500: > > Hi, > > > > Got an oops last week while rebalancing that seems to have left me with > > a corrupted btrfs. Kernel was ~2.6.36 + Transparent hugetlb patchset + > > small misc. patchs. > > We have a confirmed and reproducible case where the transparent > hugepages are corrupting btrfs (and only btrfs). I'll work with Andrea > on figuring out the cause. > > So, the first step to trying to fix it is to grab the latest btrfsck and > see if some old copies of the super are working: > > btrfsck -s 1 /dev/xxx > btrfsck -s 2 /dev/xxx > Yeah, I tried that with the latest btrfsck (last commit was: btrfs-debug-tree: add -d option ...) # ./btrfsck -s 1 /dev/sdc1 using SB copy 1, bytenr 67108864 checksum verify failed on 625055924224 wanted C3DFFE41 found FFFFFF88 checksum verify failed on 625055924224 wanted C3DFFE41 found FFFFFF88 checksum verify failed on 625055924224 wanted C3DFFE41 found FFFFFF88 Segmentation fault # ./btrfsck -s 0 /dev/sdc1 using SB copy 0, bytenr 65536 checksum verify failed on 625055924224 wanted C3DFFE41 found FFFFFF88 checksum verify failed on 625055924224 wanted C3DFFE41 found FFFFFF88 checksum verify failed on 625055924224 wanted C3DFFE41 found FFFFFF88 Segmentation fault # ./btrfsck -s 2 /dev/sdc1 using SB copy 2, bytenr 274877906944 checksum verify failed on 625055924224 wanted C3DFFE41 found FFFFFF88 checksum verify failed on 625055924224 wanted C3DFFE41 found FFFFFF88 checksum verify failed on 625055924224 wanted C3DFFE41 found FFFFFF88 Segmentation fault # ./btrfsck -s 3 /dev/sdc1 using SB copy 3, bytenr 1125899906842624 No valid Btrfs found on /dev/sdc1 Hmm, odd that btrfsck -s 0 /dev/sdc1 finds a different checksum than before.