The patch titled
Subject: mm-migrate-do-not-touch-page-mem_cgroup-of-live-pages-fix-2
has been added to the -mm tree. Its filename is
mm-migrate-do-not-touch-page-mem_cgroup-of-live-pages-fix-2.patch
This patch should soon appear at
http://ozlabs.org/~akpm/mmots/broken-out/mm-migrate-do-not-touch-page-mem_cgroup-of-live-pages-fix-2.patch
and later at
http://ozlabs.org/~akpm/mmotm/broken-out/mm-migrate-do-not-touch-page-mem_cgroup-of-live-pages-fix-2.patch
Before you just go and hit "reply", please:
a) Consider who else should be cc'ed
b) Prefer to cc a suitable mailing list as well
c) Ideally: find the original patch on the mailing list and do a
reply-to-all to that, adding suitable additional cc's
*** Remember to use Documentation/SubmitChecklist when testing your code ***
The -mm tree is included into linux-next and is updated
there every 3-4 working days
------------------------------------------------------
From: Vladimir Davydov <vdavydov@xxxxxxxxxxxxx>
Subject: mm-migrate-do-not-touch-page-mem_cgroup-of-live-pages-fix-2
It must be safe to move these calls outside tree_lock:
Will hopefully fix the lockdep splat:
[ 3587.997451] =================================
[ 3587.997453] [ INFO: inconsistent lock state ]
[ 3587.997456] 4.5.0-rc2-next-20160203-dbg-00007-g37a0a9d-dirty #377 Not tainted
[ 3587.997457] ---------------------------------
[ 3587.997459] inconsistent {IN-SOFTIRQ-W} -> {SOFTIRQ-ON-W} usage.
[ 3587.997462] cc1plus/22766 [HC0[0]:SC0[0]:HE1:SE1] takes:
[ 3587.997464] (&(&mapping->tree_lock)->rlock){+.?...}, at: [<ffffffff8113aaee>] migrate_page_move_mapping+0xbd/0x33f
[ 3587.997474] {IN-SOFTIRQ-W} state was registered at:
[ 3587.997476] [<ffffffff81082515>] __lock_acquire+0x973/0x18fc
[ 3587.997481] [<ffffffff81083c77>] lock_acquire+0x10d/0x1a8
[ 3587.997484] [<ffffffff813c3486>] _raw_spin_lock_irqsave+0x3d/0x51
[ 3587.997489] [<ffffffff81100552>] test_clear_page_writeback+0x75/0x1b4
[ 3587.997493] [<ffffffff810f53cb>] end_page_writeback+0x29/0x4a
[ 3587.997497] [<ffffffff8117ffd5>] end_buffer_async_write+0xfb/0x176
[ 3587.997501] [<ffffffff8117fb53>] end_bio_bh_io_sync+0x2c/0x37
[ 3587.997503] [<ffffffff811d0a43>] bio_endio+0x53/0x5b
[ 3587.997508] [<ffffffff811d8548>] blk_update_request+0x1fb/0x34d
[ 3587.997512] [<ffffffffa00ed512>] scsi_end_request+0x31/0x182 [scsi_mod]
[ 3587.997522] [<ffffffffa00eec2d>] scsi_io_completion+0x186/0x46e [scsi_mod]
[ 3587.997530] [<ffffffffa00e7aa2>] scsi_finish_command+0xd4/0xdd [scsi_mod]
[ 3587.997537] [<ffffffffa00ee51c>] scsi_softirq_done+0xe0/0xe7 [scsi_mod]
[ 3587.997544] [<ffffffff811df3ef>] blk_done_softirq+0x84/0x8b
[ 3587.997548] [<ffffffff8104625c>] __do_softirq+0x196/0x3f5
[ 3587.997552] [<ffffffff810466aa>] irq_exit+0x40/0x94
[ 3587.997554] [<ffffffff813c6041>] do_IRQ+0x101/0x119
[ 3587.997558] [<ffffffff813c4689>] ret_from_intr+0x0/0x19
[ 3587.997561] [<ffffffff812d9906>] cpuidle_enter+0x17/0x19
[ 3587.997565] [<ffffffff8107c32a>] call_cpuidle+0x3e/0x40
[ 3587.997569] [<ffffffff8107c65b>] cpu_startup_entry+0x242/0x35e
[ 3587.997572] [<ffffffff813bd4fa>] rest_init+0x131/0x137
[ 3587.997575] [<ffffffff816d4ec1>] start_kernel+0x3dd/0x3ea
[ 3587.997579] [<ffffffff816d42f1>] x86_64_start_reservations+0x2a/0x2c
[ 3587.997582] [<ffffffff816d445d>] x86_64_start_kernel+0x16a/0x178
[ 3587.997586] irq event stamp: 191930
[ 3587.997587] hardirqs last enabled at (191929): [<ffffffff810fb0da>] free_hot_cold_page+0x166/0x179
[ 3587.997591] hardirqs last disabled at (191930): [<ffffffff813c34ad>] _raw_spin_lock_irq+0x13/0x47
[ 3587.997594] softirqs last enabled at (191758): [<ffffffff810463a5>] __do_softirq+0x2df/0x3f5
[ 3587.997597] softirqs last disabled at (191741): [<ffffffff810466aa>] irq_exit+0x40/0x94
[ 3587.997600]
other info that might help us debug this:
[ 3587.997602] Possible unsafe locking scenario:
[ 3587.997604] CPU0
[ 3587.997605] ----
[ 3587.997607] lock(&(&mapping->tree_lock)->rlock);
[ 3587.997610] <Interrupt>
[ 3587.997611] lock(&(&mapping->tree_lock)->rlock);
[ 3587.997614]
*** DEADLOCK ***
[ 3587.997617] 2 locks held by cc1plus/22766:
[ 3587.997618] #0: (&mm->mmap_sem){++++++}, at: [<ffffffff81036ee0>] __do_page_fault+0x140/0x35a
[ 3587.997626] #1: (&(&mapping->tree_lock)->rlock){+.?...}, at: [<ffffffff8113aaee>] migrate_page_move_mapping+0xbd/0x33f
[ 3587.997633]
stack backtrace:
[ 3587.997636] CPU: 7 PID: 22766 Comm: cc1plus Not tainted 4.5.0-rc2-next-20160203-dbg-00007-g37a0a9d-dirty #377
[ 3587.997638] 0000000000000000 ffff88010f73f818 ffffffff811f1b02 ffffffff81f35160
[ 3587.997643] ffff88013813e900 ffff88010f73f850 ffffffff810f3000 0000000000000006
[ 3587.997647] ffff88013813f058 ffff88013813e900 ffffffff8107f99d 0000000000000006
[ 3587.997651] Call Trace:
[ 3587.997656] [<ffffffff811f1b02>] dump_stack+0x67/0x90
[ 3587.997660] [<ffffffff810f3000>] print_usage_bug.part.24+0x259/0x268
[ 3587.997663] [<ffffffff8107f99d>] ? check_usage_forwards+0x11c/0x11c
[ 3587.997666] [<ffffffff81080579>] mark_lock+0x381/0x567
[ 3587.997670] [<ffffffff810807bd>] mark_held_locks+0x5e/0x74
[ 3587.997673] [<ffffffff813c363f>] ? _raw_spin_unlock_irq+0x2c/0x4a
[ 3587.997676] [<ffffffff8108093f>] trace_hardirqs_on_caller+0x16c/0x188
[ 3587.997679] [<ffffffff81080968>] trace_hardirqs_on+0xd/0xf
[ 3587.997682] [<ffffffff813c363f>] _raw_spin_unlock_irq+0x2c/0x4a
[ 3587.997686] [<ffffffff81148bef>] unlock_page_lru+0x11f/0x12a
[ 3587.997689] [<ffffffff8114a90f>] mem_cgroup_migrate+0x196/0x1d9
[ 3587.997692] [<ffffffff8113abf3>] migrate_page_move_mapping+0x1c2/0x33f
[ 3587.997696] [<ffffffff8113b669>] buffer_migrate_page+0x47/0x102
[ 3587.997699] [<ffffffff8113b4e4>] move_to_new_page+0x56/0x194
[ 3587.997702] [<ffffffff8113bb6b>] migrate_pages+0x447/0x978
[ 3587.997705] [<ffffffff811171f0>] ? isolate_freepages_block+0x353/0x353
[ 3587.997708] [<ffffffff81115cd6>] ? pageblock_pfn_to_page+0xbf/0xbf
[ 3587.997711] [<ffffffff81118834>] compact_zone+0x690/0x92e
[ 3587.997714] [<ffffffff81118b40>] compact_zone_order+0x6e/0x8a
[ 3587.997717] [<ffffffff81118dc3>] try_to_compact_pages+0x151/0x28f
[ 3587.997720] [<ffffffff81118dc3>] ? try_to_compact_pages+0x151/0x28f
[ 3587.997723] [<ffffffff8108111a>] ? __lock_is_held+0x3c/0x57
[ 3587.997726] [<ffffffff810fc471>] __alloc_pages_direct_compact+0x3e/0xeb
[ 3587.997729] [<ffffffff810fc965>] __alloc_pages_nodemask+0x447/0xb8b
[ 3587.997732] [<ffffffff81120a18>] ? handle_mm_fault+0x8b4/0x16bf
[ 3587.997737] [<ffffffff8106409f>] ? __might_sleep+0x75/0x7c
[ 3587.997740] [<ffffffff81140704>] do_huge_pmd_anonymous_page+0x1d1/0x3fc
[ 3587.997744] [<ffffffff811202ae>] ? handle_mm_fault+0x14a/0x16bf
[ 3587.997747] [<ffffffff81120623>] handle_mm_fault+0x4bf/0x16bf
[ 3587.997750] [<ffffffff8108111a>] ? __lock_is_held+0x3c/0x57
[ 3587.997754] [<ffffffff81036f7f>] __do_page_fault+0x1df/0x35a
[ 3587.997757] [<ffffffff8103712c>] do_page_fault+0xc/0xe
[ 3587.997760] [<ffffffff813c5892>] page_fault+0x22/0x30
Reported-by: Sergey Senozhatsky <sergey.senozhatsky@xxxxxxxxx>
Cc: Johannes Weiner <hannes@xxxxxxxxxxx>
Cc: Michal Hocko <mhocko@xxxxxxx>
Cc: Vladimir Davydov <vdavydov@xxxxxxxxxxxxx>
Cc: Greg Thelen <gthelen@xxxxxxxxxx>
Cc: Hugh Dickins <hughd@xxxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
mm/migrate.c | 9 +++++----
1 file changed, 5 insertions(+), 4 deletions(-)
diff -puN mm/migrate.c~mm-migrate-do-not-touch-page-mem_cgroup-of-live-pages-fix-2 mm/migrate.c
--- a/mm/migrate.c~mm-migrate-do-not-touch-page-mem_cgroup-of-live-pages-fix-2
+++ a/mm/migrate.c
@@ -379,8 +379,6 @@ int migrate_page_move_mapping(struct add
if (PageSwapBacked(page))
SetPageSwapBacked(newpage);
- mem_cgroup_migrate(page, newpage);
-
get_page(newpage); /* add cache reference */
if (PageSwapCache(page)) {
SetPageSwapCache(newpage);
@@ -430,6 +428,8 @@ int migrate_page_move_mapping(struct add
}
local_irq_enable();
+ mem_cgroup_migrate(page, newpage);
+
return MIGRATEPAGE_SUCCESS;
}
@@ -463,8 +463,6 @@ int migrate_huge_page_move_mapping(struc
newpage->index = page->index;
newpage->mapping = page->mapping;
- mem_cgroup_migrate(page, newpage);
-
get_page(newpage);
radix_tree_replace_slot(pslot, newpage);
@@ -472,6 +470,9 @@ int migrate_huge_page_move_mapping(struc
page_unfreeze_refs(page, expected_count - 1);
spin_unlock_irq(&mapping->tree_lock);
+
+ mem_cgroup_migrate(page, newpage);
+
return MIGRATEPAGE_SUCCESS;
}
_
Patches currently in -mm which might be from vdavydov@xxxxxxxxxxxxx are
mm-vmscan-do-not-clear-shrinker_numa_aware-if-nr_node_ids-==-1.patch
mm-migrate-do-not-touch-page-mem_cgroup-of-live-pages-fix-2.patch
--
To unsubscribe from this list: send the line "unsubscribe mm-commits" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
