On Tue, Nov 16, 2010 at 07:52:16PM -0600, Jeffrey Hundstad wrote: > Hello, > > I'm not sure if this is interesting at all. Let me know if it is, I > can do bisects if needed, let me know which config additions you > need for debugging. It's very interesting. I put these WARN_ONs in because we believed we could never get a releasepage call with delayed or unwritten extents still on. Previous it could happen during truncate, but with the new truncate sequence that should be gone. Your trace on the other hand comes from kswapd. > [<c1019687>] ? warn_slowpath_common+0x7c/0x8f > [<c10dbdcc>] ? xfs_vm_releasepage+0x32/0x41 > [<c10dbdcc>] ? xfs_vm_releasepage+0x32/0x41 > [<c10196b5>] ? warn_slowpath_null+0x1b/0x1e > [<c10dbdcc>] ? xfs_vm_releasepage+0x32/0x41 > [<c1046672>] ? try_to_release_page+0x2c/0x3c > [<c104fc12>] ? shrink_page_list+0x47d/0x5e3 > [<c104ea7a>] ? isolate_lru_pages+0x64/0x1cf > [<c104fee4>] ? shrink_inactive_list+0x16c/0x1f3 > [<c1050272>] ? shrink_zone+0x307/0x3af > [<c10506c6>] ? kswapd+0x3ac/0x574 > [<c102b335>] ? autoremove_wake_function+0x0/0x31 > [<c105031a>] ? kswapd+0x0/0x574 > [<c102b15f>] ? kthread+0x62/0x67 > [<c102b0fd>] ? kthread+0x0/0x67 > [<c1002b16>] ? kernel_thread_helper+0x6/0x10 This looks like some sort of race between writepage and redirtying. Let me think about it a bit more, but I think it's harmless. What worries me more is that aptitude is apparently able to trigger kswapd writeback easily. _______________________________________________ xfs mailing list xfs@xxxxxxxxxxx http://oss.sgi.com/mailman/listinfo/xfs
