On Sun, 2012-07-15 at 10:59 +0200, Thomas Gleixner wrote: > On Fri, 13 Jul 2012, Jan Kara wrote: > > On Fri 13-07-12 16:25:05, Thomas Gleixner wrote: > > > So the patch below should allow the unplug to take place when blocked > > > on mutexes etc. > > Thanks for the patch! Mike will give it some testing. > > I just found out that this patch will explode nicely when the unplug > code runs into a contended lock. Then we try to block on that lock and > make the rtmutex code unhappy as we are already blocked on something > else. Kinda like so? My x3550 M3 just exploded. Aw poo. [ 6669.133081] Kernel panic - not syncing: rt_mutex_real_waiter(task->pi_blocked_on) lock: 0xffff880175dfd588 waiter: 0xffff880121fc2d58 [ 6669.133083] [ 6669.133086] Pid: 28240, comm: bonnie++ Tainted: G N 3.0.35-rt56-rt #20 [ 6669.133088] Call Trace: [ 6669.133102] [<ffffffff81004562>] dump_trace+0x82/0x2e0 [ 6669.133109] [<ffffffff8154d1ee>] dump_stack+0x69/0x6f [ 6669.133114] [<ffffffff8154d295>] panic+0xa1/0x1e5 [ 6669.133121] [<ffffffff81095289>] task_blocks_on_rt_mutex+0x279/0x2c0 [ 6669.133127] [<ffffffff8154f5d5>] rt_spin_lock_slowlock+0xb5/0x290 [ 6669.133134] [<ffffffff8131d7e4>] blk_flush_plug_list+0x164/0x200 [ 6669.133139] [<ffffffff8154dffe>] schedule+0x5e/0xb0 [ 6669.133143] [<ffffffff8154f1ab>] __rt_mutex_slowlock+0x4b/0xd0 [ 6669.133148] [<ffffffff8154f39b>] rt_mutex_slowlock+0xeb/0x210 [ 6669.133154] [<ffffffff81127bce>] page_referenced_file+0x4e/0x190 [ 6669.133160] [<ffffffff8112954a>] page_referenced+0x6a/0x230 [ 6669.133166] [<ffffffff8110b5e4>] shrink_active_list+0x214/0x3d0 [ 6669.133170] [<ffffffff8110b874>] shrink_list+0xd4/0x120 [ 6669.133176] [<ffffffff8110bc3c>] shrink_zone+0x9c/0x1d0 [ 6669.133180] [<ffffffff8110c07f>] shrink_zones+0x7f/0x1f0 [ 6669.133185] [<ffffffff8110c27d>] do_try_to_free_pages+0x8d/0x370 [ 6669.133189] [<ffffffff8110c8ba>] try_to_free_pages+0xea/0x210 [ 6669.133197] [<ffffffff810ff5e3>] __alloc_pages_nodemask+0x5b3/0x9f0 [ 6669.133205] [<ffffffff81138294>] alloc_pages_current+0xc4/0x150 [ 6669.133211] [<ffffffff810f6296>] find_or_create_page+0x46/0xb0 [ 6669.133217] [<ffffffff81296cc6>] alloc_extent_buffer+0x226/0x4b0 [ 6669.133225] [<ffffffff8126f6b9>] readahead_tree_block+0x19/0x50 [ 6669.133231] [<ffffffff8124f4bf>] reada_for_search+0x1cf/0x230 [ 6669.133237] [<ffffffff81252faa>] read_block_for_search+0x18a/0x200 [ 6669.133242] [<ffffffff8125525a>] btrfs_search_slot+0x25a/0x7e0 [ 6669.133248] [<ffffffff81269144>] btrfs_lookup_csum+0x74/0x180 [ 6669.133254] [<ffffffff8126940f>] __btrfs_lookup_bio_sums+0x1bf/0x3b0 [ 6669.133260] [<ffffffff812775c8>] btrfs_submit_bio_hook+0x158/0x1a0 [ 6669.133270] [<ffffffff81291216>] submit_one_bio+0x66/0xa0 [ 6669.133274] [<ffffffff81295017>] submit_extent_page+0x107/0x220 [ 6669.133278] [<ffffffff81295629>] __extent_read_full_page+0x4b9/0x6e0 [ 6669.133284] [<ffffffff8129669f>] extent_readpages+0xbf/0x100 [ 6669.133289] [<ffffffff811020fe>] __do_page_cache_readahead+0x1ae/0x250 [ 6669.133295] [<ffffffff811024dc>] ra_submit+0x1c/0x30 [ 6669.133299] [<ffffffff810f67eb>] do_generic_file_read.clone.0+0x27b/0x450 [ 6669.133305] [<ffffffff810f7a9b>] generic_file_aio_read+0x1fb/0x2a0 [ 6669.133313] [<ffffffff8115454f>] do_sync_read+0xbf/0x100 [ 6669.133319] [<ffffffff81154e03>] vfs_read+0xc3/0x180 [ 6669.133323] [<ffffffff81154f11>] sys_read+0x51/0xa0 [ 6669.133329] [<ffffffff81557092>] system_call_fastpath+0x16/0x1b [ 6669.133347] [<00007ff8b95bb370>] 0x7ff8b95bb36f > So no, it's not a solution to the problem. Sigh. > > Can you figure out on which lock the stuck thread which did not unplug > due to tsk_is_pi_blocked was blocked? I'll take a peek. -Mike -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
