Re: echo 3 > /proc/.../drop_caches goes mad with 3.1-rc6, maybefsnotify related
From: Hugh Dickins
Date: Thu Sep 15 2011 - 17:43:43 EST
On Thu, 15 Sep 2011, Tino Keitel wrote:
>
> "echo 3 > /proc/sys/vm/drop_caches" does not return here, and in the
> kernel log I see the log entries below. In fact, the computer becomes
> partly unusable regarding disk access, and I have to reboot.
>
> I currently use 3.1-rc6, but it also happened with older 3.1-rc
> kernels.
>
> As fsnotify is showing up in the trace: I have an inotify_wait always
> running which triggers a mail queue run if something happens in my mail
> queue directory.
>
> INFO: rcu_sched_state detected stall on CPU 1 (t=18000 jiffies)
> INFO: rcu_sched_state detected stall on CPU 1 (t=72030 jiffies)
> INFO: rcu_sched_state detected stall on CPU 1 (t=126060 jiffies)
> INFO: rcu_sched_state detected stall on CPU 1 (t=180090 jiffies)
> INFO: rcu_sched_state detected stall on CPU 1 (t=234120 jiffies)
> INFO: rcu_sched_state detected stall on CPU 1 (t=288150 jiffies)
> INFO: task fsnotify_mark:491 blocked for more than 120 seconds.
> "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
> fsnotify_mark D ffff88021fb10700 0 491 2 0x00000000
> ffff88021eac20d0 0000000000000046 ffff880200000000 ffff88021e8be0d0
> ffff880216497fd8 ffff880216497fd8 ffff880216497fd8 ffff88021eac20d0
> ffff880216497e4c 0000000181037707 0000000200000086 ffffffff819577b0
> Call Trace:
> [<ffffffff814de368>] ? __mutex_lock_slowpath+0xc8/0x140
> [<ffffffff8108ae20>] ? synchronize_rcu_bh+0x60/0x60
> [<ffffffff814de013>] ? mutex_lock+0x23/0x40
> [<ffffffff8106468c>] ? __synchronize_srcu+0x2c/0xc0
> [<ffffffff81103583>] ? fsnotify_mark_destroy+0x83/0x160
> [<ffffffff8105fca0>] ? add_wait_queue+0x60/0x60
> [<ffffffff81103500>] ? fsnotify_put_mark+0x20/0x20
> [<ffffffff8105f53e>] ? kthread+0x7e/0x90
> [<ffffffff814e0b74>] ? kernel_thread_helper+0x4/0x10
> [<ffffffff8105f4c0>] ? kthread_worker_fn+0x180/0x180
> [<ffffffff814e0b70>] ? gs_change+0xb/0xb
> INFO: task inotifywait:25496 blocked for more than 120 seconds.
> "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
> inotifywait D ffff88021fa10700 0 25496 2060 0x00000000
> ffff88006ef46650 0000000000000046 ffff880200000000 ffffffff81826020
> ffff88011355bfd8 ffff88011355bfd8 ffff88011355bfd8 ffff88006ef46650
> 000000000c800000 0000000100000000 0000000000000002 ffff88011355bd88
> Call Trace:
> [<ffffffff814ddc55>] ? schedule_timeout+0x1c5/0x240
> [<ffffffff814d89dd>] ? cache_alloc_refill+0x84/0x4c5
> [<ffffffff8124e997>] ? idr_remove+0x127/0x1c0
> [<ffffffff814dd61b>] ? wait_for_common+0xcb/0x160
> [<ffffffff8103ef00>] ? try_to_wake_up+0x270/0x270
> [<ffffffff8108ae20>] ? synchronize_rcu_bh+0x60/0x60
> [<ffffffff8108ae6d>] ? synchronize_sched+0x4d/0x60
> [<ffffffff8105ca60>] ? find_ge_pid+0x40/0x40
> [<ffffffff810646c3>] ? __synchronize_srcu+0x63/0xc0
> [<ffffffff81102e41>] ? fsnotify_put_group+0x21/0x40
> [<ffffffff81104838>] ? inotify_release+0x18/0x20
> [<ffffffff810d096a>] ? fput+0xea/0x240
> [<ffffffff810cd1ef>] ? filp_close+0x5f/0x90
> [<ffffffff81047116>] ? put_files_struct+0x76/0xe0
Although these stacktraces don't implicate find_get_pages() at all,
please try Shaohua's fix below (see thread: [BUG] infinite loop in
find_get_pages()), which Linus put in his tree yesterday.
Hugh
Subject: mm: account skipped entries to avoid looping in find_get_pages
The found entries by find_get_pages() could be all swap entries. In
this case we skip the entries, but make sure the skipped entries are
accounted, so we don't keep looping.
Using nr_found > nr_skip to simplify code as suggested by Eric.
Reported-and-tested-by: Eric Dumazet <eric.dumazet@xxxxxxxxx>
Signed-off-by: Shaohua Li <shaohua.li@xxxxxxxxx>
diff --git a/mm/filemap.c b/mm/filemap.c
index 645a080..7771871 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -827,13 +827,14 @@ unsigned find_get_pages(struct address_space *mapping, pgoff_t start,
{
unsigned int i;
unsigned int ret;
- unsigned int nr_found;
+ unsigned int nr_found, nr_skip;
rcu_read_lock();
restart:
nr_found = radix_tree_gang_lookup_slot(&mapping->page_tree,
(void ***)pages, NULL, start, nr_pages);
ret = 0;
+ nr_skip = 0;
for (i = 0; i < nr_found; i++) {
struct page *page;
repeat:
@@ -856,6 +857,7 @@ repeat:
* here as an exceptional entry: so skip over it -
* we only reach this from invalidate_mapping_pages().
*/
+ nr_skip++;
continue;
}
@@ -876,7 +878,7 @@ repeat:
* If all entries were removed before we could secure them,
* try again, because callers stop trying once 0 is returned.
*/
- if (unlikely(!ret && nr_found))
+ if (unlikely(!ret && nr_found > nr_skip))
goto restart;
rcu_read_unlock();
return ret;
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/