On Tue, 26 Apr 2016, Shi, Yang wrote:
> On 4/21/2016 5:38 PM, Shi, Yang wrote:
> > Hi folks,
> >
> > I did the below test with huge tmpfs on linux-next-20160420:
> >
> > # mount -t tmpfs huge=1 tmpfs /mnt
> > # cd /mnt
> > Then clone linux kernel
> >
> > Then I got the below bug, such test works well on non-huge tmpfs.
> >
> > BUG: unable to handle kernel NULL pointer dereference at (null)
> > IP: [<ffffffff8119d2f8>] release_freepages+0x18/0xa0
> > PGD 0
> > Oops: 0000 [#1] PREEMPT SMP
> > Modules linked in:
> > CPU: 6 PID: 110 Comm: kcompactd0 Not tainted
> > 4.6.0-rc4-next-20160420-WR7.0.0.0_standard #4
> > Hardware name: Intel Corporation S5520HC/S5520HC, BIOS
> > S5500.86B.01.10.0025.030220091519 03/02/2009
> > task: ffff880361708040 ti: ffff880361704000 task.ti: ffff880361704000
> > RIP: 0010:[<ffffffff8119d2f8>] [<ffffffff8119d2f8>]
> > release_freepages+0x18/0xa0
> > RSP: 0018:ffff880361707cf8 EFLAGS: 00010282
> > RAX: 0000000000000000 RBX: ffff88036ffde7c0 RCX: 0000000000000009
> > RDX: 0000000000001bf1 RSI: 000000000000000f RDI: ffff880361707dd0
> > RBP: ffff880361707d20 R08: 0000000000000007 R09: 0000160000000000
> > R10: ffff88036ffde7c0 R11: 0000000000000000 R12: 0000000000000000
> > R13: ffff880361707dd0 R14: ffff880361707dc0 R15: ffff880361707db0
> > FS: 0000000000000000(0000) GS:ffff880363cc0000(0000)
> > knlGS:0000000000000000
> > CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> > CR2: 0000000000000000 CR3: 0000000002206000 CR4: 00000000000006e0
> > Stack:
> > ffff88036ffde7c0 0000000000000000 0000000000001a00 ffff880361707dc0
> > ffff880361707db0 ffff880361707da0 ffffffff8119f13d ffffffff81196239
> > 0000000000000000 ffff880361708040 0000000000000001 0000000000100000
> > Call Trace:
> > [<ffffffff8119f13d>] compact_zone+0x55d/0x9f0
> > [<ffffffff81196239>] ? fragmentation_index+0x19/0x70
> > [<ffffffff8119f92f>] kcompactd_do_work+0x10f/0x230
> > [<ffffffff8119fae0>] kcompactd+0x90/0x1e0
> > [<ffffffff810a3a40>] ? wait_woken+0xa0/0xa0
> > [<ffffffff8119fa50>] ? kcompactd_do_work+0x230/0x230
> > [<ffffffff810801ed>] kthread+0xdd/0x100
> > [<ffffffff81be5ee2>] ret_from_fork+0x22/0x40
> > [<ffffffff81080110>] ? kthread_create_on_node+0x180/0x180
> > Code: c1 fa 06 31 f6 e8 a9 9b fd ff eb 98 0f 1f 80 00 00 00 00 66 66 66
> > 66 90 55 48 89 e5 41 57 41 56 41 55 49 89 fd 41 54 53 48 8b 07 <48> 8b
> > 10 48 8d 78 e0 49 39 c5 4c 8d 62 e0 74 70 49 be 00 00 00
> > RIP [<ffffffff8119d2f8>] release_freepages+0x18/0xa0
> > RSP <ffff880361707cf8>
> > CR2: 0000000000000000
> > ---[ end trace 855da7e142f7311f ]---
>
> Did some preliminary investigation on this one.
>
> The problem is the cc->freepages list is empty, but cc->nr_freepages > 0, it
> looks the list and nr_freepages get out-of-sync somewhere.
>
> Any hint is appreciated.
Many thanks for continuing to test huge tmpfs on linux-next-20160420.
Sorry for leaving you hanging on that bug, it was an easy one, and
in fact fixed before 2016-04-16; but your report got buried in my
mailbox, and we just didn't have an mmotm including the fix in
between: we've all been busy.
Here's a little collection of fixes to bugs, mostly in compaction.c,
that had hit me when trying to test an mmotm of that vintage: see
the thread "mmotm woes, mainly compaction" starting 2016-04-12 on
lkml or linux-kernel.
The problem you hit there was that kcompactd_do_work() was added
by one patchset, and the freepages_held list by another patchset,
so an INIT_LIST_HEAD(&cc.freepages_held) was overlooked.
I've updated one of fixes below to match how Vlastimil actually fixed
it (pageblock_end_pfn instead of what I did originally); and thrown in
the removal of a bogus hunk from page_alloc.c that wasted a lot of time.
But Andrew now has all these or equivalent fixes in his mmots tree
(I believe freepages_held has been withdrawn for now), so there's
nothing for him or Stephen or Vlastimil or Michal to pick up below:
merely collected for your testing convenience.
Hugh
----
mm/compaction.c | 26 +++++++++++---------------
mm/page_alloc.c | 8 --------
2 files changed, 11 insertions(+), 23 deletions(-)
--- 4.6-rc2-mm1/mm/compaction.c 2016-04-10 09:43:20.314514944 -0700
+++ linux/mm/compaction.c 2016-04-26 13:48:38.309152732 -0700
@@ -638,7 +638,6 @@ isolate_migratepages_block(struct compac
{
struct zone *zone = cc->zone;
unsigned long nr_scanned = 0, nr_isolated = 0;
- struct list_head *migratelist = &cc->migratepages;
struct lruvec *lruvec;
unsigned long flags = 0;
bool locked = false;
@@ -817,7 +816,7 @@ isolate_migratepages_block(struct compac
del_page_from_lru_list(page, lruvec, page_lru(page));
isolate_success:
- list_add(&page->lru, migratelist);
+ list_add(&page->lru, &cc->migratepages);
cc->nr_migratepages++;
nr_isolated++;
@@ -851,9 +850,11 @@ isolate_fail:
spin_unlock_irqrestore(&zone->lru_lock, flags);
locked = false;
}
- putback_movable_pages(migratelist);
- nr_isolated = 0;
+ acct_isolated(zone, cc);
+ putback_movable_pages(&cc->migratepages);
+ cc->nr_migratepages = 0;
cc->last_migrated_pfn = 0;
+ nr_isolated = 0;
}
if (low_pfn < next_skip_pfn) {
@@ -928,17 +929,8 @@ isolate_migratepages_range(struct compac
pfn = isolate_migratepages_block(cc, pfn, block_end_pfn,
ISOLATE_UNEVICTABLE);
-
- /*
- * In case of fatal failure, release everything that might
- * have been isolated in the previous iteration, and signal
- * the failure back to caller.
- */
- if (!pfn) {
- putback_movable_pages(&cc->migratepages);
- cc->nr_migratepages = 0;
+ if (!pfn)
break;
- }
if (cc->nr_migratepages == COMPACT_CLUSTER_MAX)
break;
@@ -1012,7 +1004,7 @@ static void isolate_freepages(struct com
block_start_pfn = pageblock_start_pfn(cc->free_pfn);
block_end_pfn = min(block_start_pfn + pageblock_nr_pages,
zone_end_pfn(zone));
- low_pfn = pageblock_start_pfn(cc->migrate_pfn);
+ low_pfn = pageblock_end_pfn(cc->migrate_pfn);
/*
* Isolate free pages until enough are available to migrate the
@@ -1617,6 +1609,7 @@ static enum compact_result compact_zone_
VM_BUG_ON(!list_empty(&cc.freepages));
VM_BUG_ON(!list_empty(&cc.migratepages));
+ VM_BUG_ON(!list_empty(&cc.freepages_held));
*contended = cc.contended;
return ret;
@@ -1776,6 +1769,7 @@ static void __compact_pgdat(pg_data_t *p
VM_BUG_ON(!list_empty(&cc->freepages));
VM_BUG_ON(!list_empty(&cc->migratepages));
+ VM_BUG_ON(!list_empty(&cc->freepages_held));
if (is_via_compact_memory(cc->order))
continue;
@@ -1940,6 +1934,7 @@ static void kcompactd_do_work(pg_data_t
cc.zone = zone;
INIT_LIST_HEAD(&cc.freepages);
INIT_LIST_HEAD(&cc.migratepages);
+ INIT_LIST_HEAD(&cc.freepages_held);
status = compact_zone(zone, &cc);
@@ -1957,6 +1952,7 @@ static void kcompactd_do_work(pg_data_t
VM_BUG_ON(!list_empty(&cc.freepages));
VM_BUG_ON(!list_empty(&cc.migratepages));
+ VM_BUG_ON(!list_empty(&cc.freepages_held));
}
/*
--- 4.6-rc2-mm1/mm/page_alloc.c 2016-04-10 09:43:20.342515240 -0700
+++ linux/mm/page_alloc.c 2016-04-26 13:48:38.309152732 -0700
@@ -3424,14 +3424,6 @@ retry:
if (order && compaction_made_progress(compact_result))
compaction_retries++;
- /*
- * It can become very expensive to allocate transparent hugepages at
- * fault, so use asynchronous memory compaction for THP unless it is
- * khugepaged trying to collapse.
- */
- if (!is_thp_gfp_mask(gfp_mask) || (current->flags & PF_KTHREAD))
- migration_mode = MIGRATE_SYNC_LIGHT;
-
/* Try direct reclaim and then allocating */
page = __alloc_pages_direct_reclaim(gfp_mask, order, alloc_flags, ac,
&did_some_progress);
