FS-Cache: Don't sleep in page release if __GFP_FS is not set
Don't sleep in __fscache_maybe_release_page() if __GFP_FS is not set. This
goes some way towards mitigating fscache deadlocking against ext4 by way of
the allocator, eg:
INFO: task flush-8:0:24427 blocked for more than 120 seconds.
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
flush-8:0 D ffff88003e2b9fd8 0 24427 2 0x00000000
ffff88003e2b9138 0000000000000046 ffff880012e3a040 ffff88003e2b9fd8
0000000000011c80 ffff88003e2b9fd8 ffffffff81a10400 ffff880012e3a040
0000000000000002 ffff880012e3a040 ffff88003e2b9098 ffffffff8106dcf5
Call Trace:
[<ffffffff8106dcf5>] ? __lock_is_held+0x31/0x53
[<ffffffff81219b61>] ? radix_tree_lookup_element+0xf4/0x12a
[<ffffffff81454bed>] schedule+0x60/0x62
[<ffffffffa01d349c>] __fscache_wait_on_page_write+0x8b/0xa5 [fscache]
[<ffffffff810498a8>] ? __init_waitqueue_head+0x4d/0x4d
[<ffffffffa01d393a>] __fscache_maybe_release_page+0x30c/0x324 [fscache]
[<ffffffffa01d369a>] ? __fscache_maybe_release_page+0x6c/0x324 [fscache]
[<ffffffff81071b53>] ? trace_hardirqs_on_caller+0x114/0x170
[<ffffffffa01fd7b2>] nfs_fscache_release_page+0x68/0x94 [nfs]
[<ffffffffa01ef73e>] nfs_release_page+0x7e/0x86 [nfs]
[<ffffffff810aa553>] try_to_release_page+0x32/0x3b
[<ffffffff810b6c70>] shrink_page_list+0x535/0x71a
[<ffffffff81071b53>] ? trace_hardirqs_on_caller+0x114/0x170
[<ffffffff810b7352>] shrink_inactive_list+0x20a/0x2dd
[<ffffffff81071a13>] ? mark_held_locks+0xbe/0xea
[<ffffffff810b7a65>] shrink_lruvec+0x34c/0x3eb
[<ffffffff810b7bd3>] do_try_to_free_pages+0xcf/0x355
[<ffffffff810b7fc8>] try_to_free_pages+0x9a/0xa1
[<ffffffff810b08d2>] __alloc_pages_nodemask+0x494/0x6f7
[<ffffffff810d9a07>] kmem_getpages+0x58/0x155
[<ffffffff810dc002>] fallback_alloc+0x120/0x1f3
[<ffffffff8106db23>] ? trace_hardirqs_off+0xd/0xf
[<ffffffff810dbed3>] ____cache_alloc_node+0x177/0x186
[<ffffffff81162a6c>] ? ext4_init_io_end+0x1c/0x37
[<ffffffff810dc403>] kmem_cache_alloc+0xf1/0x176
[<ffffffff810b17ac>] ? test_set_page_writeback+0x101/0x113
[<ffffffff81162a6c>] ext4_init_io_end+0x1c/0x37
[<ffffffff81162ce4>] ext4_bio_write_page+0x20f/0x3af
[<ffffffff8115cc02>] mpage_da_submit_io+0x26e/0x2f6
[<ffffffff811088e5>] ? __find_get_block_slow+0x38/0x133
[<ffffffff81161348>] mpage_da_map_and_submit+0x3a7/0x3bd
[<ffffffff81161a60>] ext4_da_writepages+0x30d/0x426
[<ffffffff810b3359>] do_writepages+0x1c/0x2a
[<ffffffff81102f4d>] __writeback_single_inode+0x3e/0xe5
[<ffffffff81103995>] writeback_sb_inodes+0x1bd/0x2f4
[<ffffffff81103b3b>] __writeback_inodes_wb+0x6f/0xb4
[<ffffffff81103c81>] wb_writeback+0x101/0x195
[<ffffffff81071b53>] ? trace_hardirqs_on_caller+0x114/0x170
[<ffffffff811043aa>] ? wb_do_writeback+0xaa/0x173
[<ffffffff8110434a>] wb_do_writeback+0x4a/0x173
[<ffffffff81071bbc>] ? trace_hardirqs_on+0xd/0xf
[<ffffffff81038554>] ? del_timer+0x4b/0x5b
[<ffffffff811044e0>] bdi_writeback_thread+0x6d/0x147
[<ffffffff81104473>] ? wb_do_writeback+0x173/0x173
[<ffffffff81048fbc>] kthread+0xd0/0xd8
[<ffffffff81455eb2>] ? _raw_spin_unlock_irq+0x29/0x3e
[<ffffffff81048eec>] ? __init_kthread_worker+0x55/0x55
[<ffffffff81456aac>] ret_from_fork+0x7c/0xb0
[<ffffffff81048eec>] ? __init_kthread_worker+0x55/0x55
2 locks held by flush-8:0/24427:
#0: (&type->s_umount_key#41){.+.+..}, at: [<ffffffff810e3b73>] grab_super_passive+0x4c/0x76
#1: (jbd2_handle){+.+...}, at: [<ffffffff81190d81>] start_this_handle+0x475/0x4ea
The problem here is that another thread, which is attempting to write the
to-be-stored NFS page to the on-ext4 cache file is waiting for the journal
lock, eg:
INFO: task kworker/u:2:24437 blocked for more than 120 seconds.
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
kworker/u:2 D ffff880039589768 0 24437 2 0x00000000
ffff8800395896d8 0000000000000046 ffff8800283bf040 ffff880039589fd8
0000000000011c80 ffff880039589fd8 ffff880039f0b040 ffff8800283bf040
0000000000000006 ffff8800283bf6b8 ffff880039589658 ffffffff81071a13
Call Trace:
[<ffffffff81071a13>] ? mark_held_locks+0xbe/0xea
[<ffffffff81455e73>] ? _raw_spin_unlock_irqrestore+0x3a/0x50
[<ffffffff81071b53>] ? trace_hardirqs_on_caller+0x114/0x170
[<ffffffff81071bbc>] ? trace_hardirqs_on+0xd/0xf
[<ffffffff81454bed>] schedule+0x60/0x62
[<ffffffff81190c23>] start_this_handle+0x317/0x4ea
[<ffffffff810498a8>] ? __init_waitqueue_head+0x4d/0x4d
[<ffffffff81190fcc>] jbd2__journal_start+0xb3/0x12e
[<ffffffff81176606>] __ext4_journal_start_sb+0xb2/0xc6
[<ffffffff8115f137>] ext4_da_write_begin+0x109/0x233
[<ffffffff810a964d>] generic_file_buffered_write+0x11a/0x264
[<ffffffff811032cf>] ? __mark_inode_dirty+0x2d/0x1ee
[<ffffffff810ab1ab>] __generic_file_aio_write+0x2a5/0x2d5
[<ffffffff810ab24a>] generic_file_aio_write+0x6f/0xd0
[<ffffffff81159a2c>] ext4_file_write+0x38c/0x3c4
[<ffffffff810e0915>] do_sync_write+0x91/0xd1
[<ffffffffa00a17f0>] cachefiles_write_page+0x26f/0x310 [cachefiles]
[<ffffffffa01d470b>] fscache_write_op+0x21e/0x37a [fscache]
[<ffffffff81455eb2>] ? _raw_spin_unlock_irq+0x29/0x3e
[<ffffffffa01d2479>] fscache_op_work_func+0x78/0xd7 [fscache]
[<ffffffff8104455a>] process_one_work+0x232/0x3a8
[<ffffffff810444ff>] ? process_one_work+0x1d7/0x3a8
[<ffffffff81044ee0>] worker_thread+0x214/0x303
[<ffffffff81044ccc>] ? manage_workers+0x245/0x245
[<ffffffff81048fbc>] kthread+0xd0/0xd8
[<ffffffff81455eb2>] ? _raw_spin_unlock_irq+0x29/0x3e
[<ffffffff81048eec>] ? __init_kthread_worker+0x55/0x55
[<ffffffff81456aac>] ret_from_fork+0x7c/0xb0
[<ffffffff81048eec>] ? __init_kthread_worker+0x55/0x55
4 locks held by kworker/u:2/24437:
#0: (fscache_operation){.+.+.+}, at: [<ffffffff810444ff>] process_one_work+0x1d7/0x3a8
#1: ((&op->work)){+.+.+.}, at: [<ffffffff810444ff>] process_one_work+0x1d7/0x3a8
#2: (sb_writers#14){.+.+.+}, at: [<ffffffff810ab22c>] generic_file_aio_write+0x51/0xd0
#3: (&sb->s_type->i_mutex_key#19){+.+.+.}, at: [<ffffffff810ab236>] generic_file_aio_write+0x5b/0x
fscache already tries to cancel pending stores, but it can't cancel a write
for which I/O is already in progress.
An alternative would be to accept writing garbage to the cache under extreme
circumstances and to kill the afflicted cache object if we have to do this.
However, we really need to know how strapped the allocator is before deciding
to do that.
Signed-off-by: David Howells <dhowells@redhat.com>
Tested-By: Milosz Tanski <milosz@adfin.com>
Acked-by: Jeff Layton <jlayton@redhat.com>
0c59a95d90