[Top] [All Lists]

Re: Assertion failed: atomic_read(&mp->m_active_trans)

To: Donald Douwsma <donaldd@xxxxxxx>
Subject: Re: Assertion failed: atomic_read(&mp->m_active_trans)
From: Lachlan McIlroy <lachlan@xxxxxxx>
Date: Tue, 02 Dec 2008 14:25:29 +1100
Cc: xfs@xxxxxxxxxxx
In-reply-to: <492BB095.1000104@xxxxxxx>
References: <492BB095.1000104@xxxxxxx>
Reply-to: lachlan@xxxxxxx
User-agent: Thunderbird (X11/20081105)
That looks fine to me.  Just make sure the PV doesn't get closed when
you check in the changes since the real problem is still unresolved.

Donald Douwsma wrote:
We still occasionally see transactions in flight after remounting
read-only. This has come up a few times in the past, but we never
seem to have gotten to the bottom of it.


Most recently we've seen this on 2.6.27, when unmounting the root
filesystem during shutdown/reboot.

Stack traceback for pid 13170
0xffff81024dcd9080    13170    12901  1    1   R  0xffff81024dcd93c0 *mount
rsp                rip                Function (args)
0xffff8101fb977d18 0xffffffff803b8acd assfail+0x1a (invalid, invalid, invalid)
0xffff8101fb977d50 0xffffffff803a57e4 xfs_attr_quiesce+0x4a (0xffff8102211e4b20)
0xffff8101fb977d70 0xffffffff803a589b xfs_mntupdate+0x7c (0xffff8102211e4b20, 
invalid, invalid)
0xffff8101fb977d90 0xffffffff803b7cf6 xfs_fs_remount+0x49 (invalid, 
0xffff8101fb977dd4, invalid)
0xffff8101fb977dc0 0xffffffff802830fe do_remount_sb+0xe9 (0xffff81025c804670, 
invalid, 0xffff8101ee490000, invalid)
0xffff8101fb977e00 0xffffffff8029698d do_remount+0x7d (0xffff8101fb977e58, 
invalid, invalid, 0xffff8101ee490000)
0xffff8101fb977e40 0xffffffff802974fd do_mount+0x13b (0xffff8102079c2000, 
0xffff8102004ea000, 0xffff810219cb0000, invalid, 0xffff8101ee490000)
0xffff8101fb977f20 0xffffffff8029761a sys_mount+0x89 (0x523d90, invalid, 
invalid, 0xffffffffc0ed0021, 0x523e30)
0xffff8101fb977f80 0xffffffff8020b18b system_call_after_swapgs+0x7b (invalid, 
invalid, invalid, invalid, invalid, invalid)

Previously we've discussed changing the ASSERT_ALWAYS to a normal
ASSERT to lessen the impact for users. Any objections to doing this
until we fix the underlying problem?



xfs mailing list

<Prev in Thread] Current Thread [Next in Thread>