Bug#642235: linux-2.6: LVM deadlocks and oops
Package: linux-2.6 Version: 3.1.0-rc6 Severity: normal Tags: upstream Hello, While playing with sbuild to build packages on armhf buildd running an upstream patched kernel for i.MX53 LOCO boards (next armhf build daemons). I have hit a deadlock and oops on LVM, so I think it is worth it to document the problem here and maybe it can get fixed. [74093.782238] kjournald starting. Commit interval 5 seconds [74093.821968] EXT3-fs (dm-4): using internal journal [74093.826917] EXT3-fs (dm-4): mounted filesystem with ordered data mode [74397.131757] kjournald starting. Commit interval 5 seconds [74397.169996] EXT3-fs (dm-4): using internal journal [74397.174838] EXT3-fs (dm-4): mounted filesystem with ordered data mode [76601.907689] kjournald starting. Commit interval 5 seconds [76601.945944] EXT3-fs (dm-4): using internal journal [76601.950779] EXT3-fs (dm-4): mounted filesystem with ordered data mode [76656.111857] kjournald starting. Commit interval 5 seconds [76656.150092] EXT3-fs (dm-4): using internal journal [76656.154937] EXT3-fs (dm-4): mounted filesystem with ordered data mode [76793.872136] kjournald starting. Commit interval 5 seconds [76793.910358] EXT3-fs (dm-4): using internal journal [76793.915200] EXT3-fs (dm-4): mounted filesystem with ordered data mode [84651.608193] kjournald starting. Commit interval 5 seconds [84651.646426] EXT3-fs (dm-4): using internal journal [84651.651269] EXT3-fs (dm-4): mounted filesystem with ordered data mode [84738.700610] [84738.700625] === [84738.708389] [ INFO: possible circular locking dependency detected ] [84738.714663] 3.1.0-rc6 #4 [84738.717197] --- [84738.723467] lvcreate/9492 is trying to acquire lock: [84738.728435] (md-suspend_lock){+.+...}, at: [c02c73c4] dm_suspend+0x28/0x29c [84738.735894] [84738.735897] but task is already holding lock: [84738.741740] (journal-j_barrier){+.+...}, at: [c01481c4] journal_lock_updates+0xf0/0xfc [84738.750141] [84738.750145] which lock already depends on the new lock. [84738.750150] [84738.758344] [84738.758346] the existing dependency chain (in reverse order) is: [84738.765839] [84738.765842] - #1 (journal-j_barrier){+.+...}: [84738.771981][c0052114] check_prev_add+0x338/0x680 [84738.777492][c0052e4c] validate_chain+0x4f4/0x674 [84738.782993][c0053714] __lock_acquire+0x748/0x828 [84738.788496][c0053a28] lock_release_non_nested+0x234/0x2bc [84738.794780][c0053b98] __lock_release+0xe8/0x1cc [84738.800195][c0053cdc] lock_release+0x60/0x70 [84738.805348][c04636d4] __mutex_unlock_slowpath+0xd4/0x13c [84738.811551][c046374c] mutex_unlock+0x10/0x14 [84738.816705][c00c9434] freeze_bdev+0x9c/0xa4 [84738.821783][c02c7438] dm_suspend+0x9c/0x29c [84738.826850][c02cc06c] dev_suspend+0x5c/0x90 [84738.831921][c02ccdd4] ctl_ioctl+0x228/0x2c8 [84738.836990][c02cce8c] dm_ctl_ioctl+0x18/0x1c [84738.842145][c00a9fa4] do_vfs_ioctl+0x33c/0x38c [84738.847483][c00aa034] sys_ioctl+0x40/0x64 [84738.852378][c000dfc0] ret_fast_syscall+0x0/0x3c [84738.857801] [84738.857804] - #0 (md-suspend_lock){+.+...}: [84738.863769][c0051c78] print_circular_bug+0x5c/0xfc [84738.869445][c0051f10] check_prev_add+0x134/0x680 [84738.874947][c0052e4c] validate_chain+0x4f4/0x674 [84738.880450][c0053714] __lock_acquire+0x748/0x828 [84738.885952][c0053d78] lock_acquire+0x8c/0xa0 [84738.891107][c04643a8] mutex_lock_nested+0x74/0x30c [84738.896784][c02c73c4] dm_suspend+0x28/0x29c [84738.901851][c02cc06c] dev_suspend+0x5c/0x90 [84738.906920][c02ccdd4] ctl_ioctl+0x228/0x2c8 [84738.911988][c02cce8c] dm_ctl_ioctl+0x18/0x1c [84738.917143][c00a9fa4] do_vfs_ioctl+0x33c/0x38c [84738.922472][c00aa034] sys_ioctl+0x40/0x64 [84738.927367][c000dfc0] ret_fast_syscall+0x0/0x3c [84738.932783] [84738.932786] other info that might help us debug this: [84738.932791] [84738.940814] Possible unsafe locking scenario: [84738.940819] [84738.946747]CPU0CPU1 [84738.951280] [84738.955810] lock(journal-j_barrier); [84738.959757]lock(md-suspend_lock); [84738.966049]lock(journal-j_barrier); [84738.972515] lock(md-suspend_lock); [84738.976287] [84738.976290] *** DEADLOCK *** [84738.976293] [84738.982232] 1 lock held by lvcreate/9492: [84738.986243] #0: (journal-j_barrier){+.+...}, at: [c01481c4] journal_lock_updates+0xf0/0xfc [84738.995085] [84738.995088] stack backtrace: [84738.999455] Backtrace: [84739.001928] [c00114e0] (dump_backtrace+0x0/0x110) from [c04624e0] (dump_stack+0x18/0x1c) [84739.010372] r7:dee6f908 r6:dee6f920 r5: r4: [84739.016090] [c04624c8] (dump_stack+0x0/0x1c)
Bug#642235: linux-2.6: LVM deadlocks and oops
On Tue, Sep 20, 2011 at 05:28:49PM +0100, Hector Oron wrote: Package: linux-2.6 Version: 3.1.0-rc6 Severity: normal Tags: upstream Hello, While playing with sbuild to build packages on armhf buildd running an upstream patched kernel for i.MX53 LOCO boards (next armhf build daemons). I have hit a deadlock and oops on LVM, so I think it is worth it to document the problem here and maybe it can get fixed. [...] Please add the 'forwarded' reference. Ben. -- Ben Hutchings We get into the habit of living before acquiring the habit of thinking. - Albert Camus -- To UNSUBSCRIBE, email to debian-kernel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20110920172703.go2...@decadent.org.uk