2024-04-12 03:00:22

by Miaohe Lin

[permalink] [raw]
Subject: [PATCH v2] mm/memory-failure: fix deadlock when hugetlb_optimize_vmemmap is enabled

When I did hard offline test with hugetlb pages, below deadlock occurs:

======================================================
WARNING: possible circular locking dependency detected
6.8.0-11409-gf6cef5f8c37f #1 Not tainted
------------------------------------------------------
bash/46904 is trying to acquire lock:
ffffffffabe68910 (cpu_hotplug_lock){++++}-{0:0}, at: static_key_slow_dec+0x16/0x60

but task is already holding lock:
ffffffffabf92ea8 (pcp_batch_high_lock){+.+.}-{3:3}, at: zone_pcp_disable+0x16/0x40

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:

-> #1 (pcp_batch_high_lock){+.+.}-{3:3}:
__mutex_lock+0x6c/0x770
page_alloc_cpu_online+0x3c/0x70
cpuhp_invoke_callback+0x397/0x5f0
__cpuhp_invoke_callback_range+0x71/0xe0
_cpu_up+0xeb/0x210
cpu_up+0x91/0xe0
cpuhp_bringup_mask+0x49/0xb0
bringup_nonboot_cpus+0xb7/0xe0
smp_init+0x25/0xa0
kernel_init_freeable+0x15f/0x3e0
kernel_init+0x15/0x1b0
ret_from_fork+0x2f/0x50
ret_from_fork_asm+0x1a/0x30

-> #0 (cpu_hotplug_lock){++++}-{0:0}:
__lock_acquire+0x1298/0x1cd0
lock_acquire+0xc0/0x2b0
cpus_read_lock+0x2a/0xc0
static_key_slow_dec+0x16/0x60
__hugetlb_vmemmap_restore_folio+0x1b9/0x200
dissolve_free_huge_page+0x211/0x260
__page_handle_poison+0x45/0xc0
memory_failure+0x65e/0xc70
hard_offline_page_store+0x55/0xa0
kernfs_fop_write_iter+0x12c/0x1d0
vfs_write+0x387/0x550
ksys_write+0x64/0xe0
do_syscall_64+0xca/0x1e0
entry_SYSCALL_64_after_hwframe+0x6d/0x75

other info that might help us debug this:

Possible unsafe locking scenario:

CPU0 CPU1
---- ----
lock(pcp_batch_high_lock);
lock(cpu_hotplug_lock);
lock(pcp_batch_high_lock);
rlock(cpu_hotplug_lock);

*** DEADLOCK ***

5 locks held by bash/46904:
#0: ffff98f6c3bb23f0 (sb_writers#5){.+.+}-{0:0}, at: ksys_write+0x64/0xe0
#1: ffff98f6c328e488 (&of->mutex){+.+.}-{3:3}, at: kernfs_fop_write_iter+0xf8/0x1d0
#2: ffff98ef83b31890 (kn->active#113){.+.+}-{0:0}, at: kernfs_fop_write_iter+0x100/0x1d0
#3: ffffffffabf9db48 (mf_mutex){+.+.}-{3:3}, at: memory_failure+0x44/0xc70
#4: ffffffffabf92ea8 (pcp_batch_high_lock){+.+.}-{3:3}, at: zone_pcp_disable+0x16/0x40

stack backtrace:
CPU: 10 PID: 46904 Comm: bash Kdump: loaded Not tainted 6.8.0-11409-gf6cef5f8c37f #1
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.14.0-0-g155821a1990b-prebuilt.qemu.org 04/01/2014
Call Trace:
<TASK>
dump_stack_lvl+0x68/0xa0
check_noncircular+0x129/0x140
__lock_acquire+0x1298/0x1cd0
lock_acquire+0xc0/0x2b0
cpus_read_lock+0x2a/0xc0
static_key_slow_dec+0x16/0x60
__hugetlb_vmemmap_restore_folio+0x1b9/0x200
dissolve_free_huge_page+0x211/0x260
__page_handle_poison+0x45/0xc0
memory_failure+0x65e/0xc70
hard_offline_page_store+0x55/0xa0
kernfs_fop_write_iter+0x12c/0x1d0
vfs_write+0x387/0x550
ksys_write+0x64/0xe0
do_syscall_64+0xca/0x1e0
entry_SYSCALL_64_after_hwframe+0x6d/0x75
RIP: 0033:0x7fc862314887
Code: 10 00 f7 d8 64 89 02 48 c7 c0 ff ff ff ff eb b7 0f 1f 00 f3 0f 1e fa 64 8b 04 25 18 00 00 00 85 c0 75 10 b8 01 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 51 c3 48 83 ec 28 48 89 54 24 18 48 89 74 24
RSP: 002b:00007fff19311268 EFLAGS: 00000246 ORIG_RAX: 0000000000000001
RAX: ffffffffffffffda RBX: 000000000000000c RCX: 00007fc862314887
RDX: 000000000000000c RSI: 000056405645fe10 RDI: 0000000000000001
RBP: 000056405645fe10 R08: 00007fc8623d1460 R09: 000000007fffffff
R10: 0000000000000000 R11: 0000000000000246 R12: 000000000000000c
R13: 00007fc86241b780 R14: 00007fc862417600 R15: 00007fc862416a00

In short, below scene breaks the lock dependency chain:

memory_failure
__page_handle_poison
zone_pcp_disable -- lock(pcp_batch_high_lock)
dissolve_free_huge_page
__hugetlb_vmemmap_restore_folio
static_key_slow_dec
cpus_read_lock -- rlock(cpu_hotplug_lock)

Fix this by calling drain_all_pages() instead.

Fixes: a6b40850c442 ("mm: hugetlb: replace hugetlb_free_vmemmap_enabled with a static_key")
Signed-off-by: Miaohe Lin <[email protected]>
Acked-by: Oscar Salvador <[email protected]>
Cc: <[email protected]>
---
v2:
collect Acked-by tag and extend comment per Oscar. Thanks.
---
mm/memory-failure.c | 16 +++++++++++++---
1 file changed, 13 insertions(+), 3 deletions(-)

diff --git a/mm/memory-failure.c b/mm/memory-failure.c
index edd6e114462f..c6750509d74c 100644
--- a/mm/memory-failure.c
+++ b/mm/memory-failure.c
@@ -153,11 +153,21 @@ static int __page_handle_poison(struct page *page)
{
int ret;

- zone_pcp_disable(page_zone(page));
+ /*
+ * zone_pcp_disable() can't be used here. It will hold pcp_batch_high_lock and
+ * dissolve_free_huge_page() might hold cpu_hotplug_lock via static_key_slow_dec()
+ * when hugetlb vmemmap optimization is enabled. This will break current lock
+ * dependency chain and leads to deadlock.
+ * Disabling pcp before dissolving the page was a deterministic approach because
+ * we made sure that those pages cannot end up in any PCP list. Draining PCP lists
+ * expels those pages to the buddy system, but nothing guarantees that those pages
+ * do not get back to a PCP queue if we need to refill those.
+ */
ret = dissolve_free_huge_page(page);
- if (!ret)
+ if (!ret) {
+ drain_all_pages(page_zone(page));
ret = take_page_off_buddy(page);
- zone_pcp_enable(page_zone(page));
+ }

return ret;
}
--
2.33.0



2024-04-12 03:13:07

by Sidhartha Kumar

[permalink] [raw]
Subject: Re: [PATCH v2] mm/memory-failure: fix deadlock when hugetlb_optimize_vmemmap is enabled

On 4/11/24 7:57 PM, Miaohe Lin wrote:
> When I did hard offline test with hugetlb pages, below deadlock occurs:
>
> ======================================================
> WARNING: possible circular locking dependency detected
> 6.8.0-11409-gf6cef5f8c37f #1 Not tainted
> ------------------------------------------------------
> bash/46904 is trying to acquire lock:
> ffffffffabe68910 (cpu_hotplug_lock){++++}-{0:0}, at: static_key_slow_dec+0x16/0x60
>
> but task is already holding lock:
> ffffffffabf92ea8 (pcp_batch_high_lock){+.+.}-{3:3}, at: zone_pcp_disable+0x16/0x40
>
> which lock already depends on the new lock.
>
> the existing dependency chain (in reverse order) is:
>
> -> #1 (pcp_batch_high_lock){+.+.}-{3:3}:
> __mutex_lock+0x6c/0x770
> page_alloc_cpu_online+0x3c/0x70
> cpuhp_invoke_callback+0x397/0x5f0
> __cpuhp_invoke_callback_range+0x71/0xe0
> _cpu_up+0xeb/0x210
> cpu_up+0x91/0xe0
> cpuhp_bringup_mask+0x49/0xb0
> bringup_nonboot_cpus+0xb7/0xe0
> smp_init+0x25/0xa0
> kernel_init_freeable+0x15f/0x3e0
> kernel_init+0x15/0x1b0
> ret_from_fork+0x2f/0x50
> ret_from_fork_asm+0x1a/0x30
>
> -> #0 (cpu_hotplug_lock){++++}-{0:0}:
> __lock_acquire+0x1298/0x1cd0
> lock_acquire+0xc0/0x2b0
> cpus_read_lock+0x2a/0xc0
> static_key_slow_dec+0x16/0x60
> __hugetlb_vmemmap_restore_folio+0x1b9/0x200
> dissolve_free_huge_page+0x211/0x260
> __page_handle_poison+0x45/0xc0
> memory_failure+0x65e/0xc70
> hard_offline_page_store+0x55/0xa0
> kernfs_fop_write_iter+0x12c/0x1d0
> vfs_write+0x387/0x550
> ksys_write+0x64/0xe0
> do_syscall_64+0xca/0x1e0
> entry_SYSCALL_64_after_hwframe+0x6d/0x75
>
> other info that might help us debug this:
>
> Possible unsafe locking scenario:
>
> CPU0 CPU1
> ---- ----
> lock(pcp_batch_high_lock);
> lock(cpu_hotplug_lock);
> lock(pcp_batch_high_lock);
> rlock(cpu_hotplug_lock);
>
> *** DEADLOCK ***
>
> 5 locks held by bash/46904:
> #0: ffff98f6c3bb23f0 (sb_writers#5){.+.+}-{0:0}, at: ksys_write+0x64/0xe0
> #1: ffff98f6c328e488 (&of->mutex){+.+.}-{3:3}, at: kernfs_fop_write_iter+0xf8/0x1d0
> #2: ffff98ef83b31890 (kn->active#113){.+.+}-{0:0}, at: kernfs_fop_write_iter+0x100/0x1d0
> #3: ffffffffabf9db48 (mf_mutex){+.+.}-{3:3}, at: memory_failure+0x44/0xc70
> #4: ffffffffabf92ea8 (pcp_batch_high_lock){+.+.}-{3:3}, at: zone_pcp_disable+0x16/0x40
>
> stack backtrace:
> CPU: 10 PID: 46904 Comm: bash Kdump: loaded Not tainted 6.8.0-11409-gf6cef5f8c37f #1
> Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.14.0-0-g155821a1990b-prebuilt.qemu.org 04/01/2014
> Call Trace:
> <TASK>
> dump_stack_lvl+0x68/0xa0
> check_noncircular+0x129/0x140
> __lock_acquire+0x1298/0x1cd0
> lock_acquire+0xc0/0x2b0
> cpus_read_lock+0x2a/0xc0
> static_key_slow_dec+0x16/0x60
> __hugetlb_vmemmap_restore_folio+0x1b9/0x200
> dissolve_free_huge_page+0x211/0x260
> __page_handle_poison+0x45/0xc0
> memory_failure+0x65e/0xc70
> hard_offline_page_store+0x55/0xa0
> kernfs_fop_write_iter+0x12c/0x1d0
> vfs_write+0x387/0x550
> ksys_write+0x64/0xe0
> do_syscall_64+0xca/0x1e0
> entry_SYSCALL_64_after_hwframe+0x6d/0x75
> RIP: 0033:0x7fc862314887
> Code: 10 00 f7 d8 64 89 02 48 c7 c0 ff ff ff ff eb b7 0f 1f 00 f3 0f 1e fa 64 8b 04 25 18 00 00 00 85 c0 75 10 b8 01 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 51 c3 48 83 ec 28 48 89 54 24 18 48 89 74 24
> RSP: 002b:00007fff19311268 EFLAGS: 00000246 ORIG_RAX: 0000000000000001
> RAX: ffffffffffffffda RBX: 000000000000000c RCX: 00007fc862314887
> RDX: 000000000000000c RSI: 000056405645fe10 RDI: 0000000000000001
> RBP: 000056405645fe10 R08: 00007fc8623d1460 R09: 000000007fffffff
> R10: 0000000000000000 R11: 0000000000000246 R12: 000000000000000c
> R13: 00007fc86241b780 R14: 00007fc862417600 R15: 00007fc862416a00
>
> In short, below scene breaks the lock dependency chain:
>
> memory_failure
> __page_handle_poison
> zone_pcp_disable -- lock(pcp_batch_high_lock)
> dissolve_free_huge_page
> __hugetlb_vmemmap_restore_folio
> static_key_slow_dec
> cpus_read_lock -- rlock(cpu_hotplug_lock)
>
> Fix this by calling drain_all_pages() instead.
>
> Fixes: a6b40850c442 ("mm: hugetlb: replace hugetlb_free_vmemmap_enabled with a static_key")
> Signed-off-by: Miaohe Lin <[email protected]>
> Acked-by: Oscar Salvador <[email protected]>
> Cc: <[email protected]>
> ---
> v2:
> collect Acked-by tag and extend comment per Oscar. Thanks.
> ---
> mm/memory-failure.c | 16 +++++++++++++---
> 1 file changed, 13 insertions(+), 3 deletions(-)
>
> diff --git a/mm/memory-failure.c b/mm/memory-failure.c
> index edd6e114462f..c6750509d74c 100644
> --- a/mm/memory-failure.c
> +++ b/mm/memory-failure.c
> @@ -153,11 +153,21 @@ static int __page_handle_poison(struct page *page)
> {
> int ret;
>
> - zone_pcp_disable(page_zone(page));
> + /*
> + * zone_pcp_disable() can't be used here. It will hold pcp_batch_high_lock and
> + * dissolve_free_huge_page() might hold cpu_hotplug_lock via static_key_slow_dec()
> + * when hugetlb vmemmap optimization is enabled. This will break current lock
> + * dependency chain and leads to deadlock.
> + * Disabling pcp before dissolving the page was a deterministic approach because
> + * we made sure that those pages cannot end up in any PCP list. Draining PCP lists
> + * expels those pages to the buddy system, but nothing guarantees that those pages
> + * do not get back to a PCP queue if we need to refill those.
> + */
> ret = dissolve_free_huge_page(page);

Hi Miaohe,

I recently sent a patch[1] to convert dissolve_free_huge_page() to folios which
changes the function name and the name referenced in the comment so this will
conflict with my patch. It's in mm-unstable now, would you be able to rebase to
that in a new version?

Thanks,
Sid

[1]
https://lore.kernel.org/linux-mm/[email protected]/T/#u


> - if (!ret)
> + if (!ret) {
> + drain_all_pages(page_zone(page));
> ret = take_page_off_buddy(page);
> - zone_pcp_enable(page_zone(page));
> + }
>
> return ret;
> }


2024-04-12 08:12:11

by Miaohe Lin

[permalink] [raw]
Subject: Re: [PATCH v2] mm/memory-failure: fix deadlock when hugetlb_optimize_vmemmap is enabled

On 2024/4/12 11:12, Sidhartha Kumar wrote:
> On 4/11/24 7:57 PM, Miaohe Lin wrote:
>> When I did hard offline test with hugetlb pages, below deadlock occurs:
>>
>> ======================================================
>> WARNING: possible circular locking dependency detected
>> 6.8.0-11409-gf6cef5f8c37f #1 Not tainted
>> ------------------------------------------------------
>> bash/46904 is trying to acquire lock:
>> ffffffffabe68910 (cpu_hotplug_lock){++++}-{0:0}, at: static_key_slow_dec+0x16/0x60
>>
>> but task is already holding lock:
>> ffffffffabf92ea8 (pcp_batch_high_lock){+.+.}-{3:3}, at: zone_pcp_disable+0x16/0x40
>>
>> which lock already depends on the new lock.
>>
>> the existing dependency chain (in reverse order) is:
>>
>> -> #1 (pcp_batch_high_lock){+.+.}-{3:3}:
>>         __mutex_lock+0x6c/0x770
>>         page_alloc_cpu_online+0x3c/0x70
>>         cpuhp_invoke_callback+0x397/0x5f0
>>         __cpuhp_invoke_callback_range+0x71/0xe0
>>         _cpu_up+0xeb/0x210
>>         cpu_up+0x91/0xe0
>>         cpuhp_bringup_mask+0x49/0xb0
>>         bringup_nonboot_cpus+0xb7/0xe0
>>         smp_init+0x25/0xa0
>>         kernel_init_freeable+0x15f/0x3e0
>>         kernel_init+0x15/0x1b0
>>         ret_from_fork+0x2f/0x50
>>         ret_from_fork_asm+0x1a/0x30
>>
>> -> #0 (cpu_hotplug_lock){++++}-{0:0}:
>>         __lock_acquire+0x1298/0x1cd0
>>         lock_acquire+0xc0/0x2b0
>>         cpus_read_lock+0x2a/0xc0
>>         static_key_slow_dec+0x16/0x60
>>         __hugetlb_vmemmap_restore_folio+0x1b9/0x200
>>         dissolve_free_huge_page+0x211/0x260
>>         __page_handle_poison+0x45/0xc0
>>         memory_failure+0x65e/0xc70
>>         hard_offline_page_store+0x55/0xa0
>>         kernfs_fop_write_iter+0x12c/0x1d0
>>         vfs_write+0x387/0x550
>>         ksys_write+0x64/0xe0
>>         do_syscall_64+0xca/0x1e0
>>         entry_SYSCALL_64_after_hwframe+0x6d/0x75
>>
>> other info that might help us debug this:
>>
>>   Possible unsafe locking scenario:
>>
>>         CPU0                    CPU1
>>         ----                    ----
>>    lock(pcp_batch_high_lock);
>>                                 lock(cpu_hotplug_lock);
>>                                 lock(pcp_batch_high_lock);
>>    rlock(cpu_hotplug_lock);
>>
>>   *** DEADLOCK ***
>>
>> 5 locks held by bash/46904:
>>   #0: ffff98f6c3bb23f0 (sb_writers#5){.+.+}-{0:0}, at: ksys_write+0x64/0xe0
>>   #1: ffff98f6c328e488 (&of->mutex){+.+.}-{3:3}, at: kernfs_fop_write_iter+0xf8/0x1d0
>>   #2: ffff98ef83b31890 (kn->active#113){.+.+}-{0:0}, at: kernfs_fop_write_iter+0x100/0x1d0
>>   #3: ffffffffabf9db48 (mf_mutex){+.+.}-{3:3}, at: memory_failure+0x44/0xc70
>>   #4: ffffffffabf92ea8 (pcp_batch_high_lock){+.+.}-{3:3}, at: zone_pcp_disable+0x16/0x40
>>
>> stack backtrace:
>> CPU: 10 PID: 46904 Comm: bash Kdump: loaded Not tainted 6.8.0-11409-gf6cef5f8c37f #1
>> Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.14.0-0-g155821a1990b-prebuilt.qemu.org 04/01/2014
>> Call Trace:
>>   <TASK>
>>   dump_stack_lvl+0x68/0xa0
>>   check_noncircular+0x129/0x140
>>   __lock_acquire+0x1298/0x1cd0
>>   lock_acquire+0xc0/0x2b0
>>   cpus_read_lock+0x2a/0xc0
>>   static_key_slow_dec+0x16/0x60
>>   __hugetlb_vmemmap_restore_folio+0x1b9/0x200
>>   dissolve_free_huge_page+0x211/0x260
>>   __page_handle_poison+0x45/0xc0
>>   memory_failure+0x65e/0xc70
>>   hard_offline_page_store+0x55/0xa0
>>   kernfs_fop_write_iter+0x12c/0x1d0
>>   vfs_write+0x387/0x550
>>   ksys_write+0x64/0xe0
>>   do_syscall_64+0xca/0x1e0
>>   entry_SYSCALL_64_after_hwframe+0x6d/0x75
>> RIP: 0033:0x7fc862314887
>> Code: 10 00 f7 d8 64 89 02 48 c7 c0 ff ff ff ff eb b7 0f 1f 00 f3 0f 1e fa 64 8b 04 25 18 00 00 00 85 c0 75 10 b8 01 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 51 c3 48 83 ec 28 48 89 54 24 18 48 89 74 24
>> RSP: 002b:00007fff19311268 EFLAGS: 00000246 ORIG_RAX: 0000000000000001
>> RAX: ffffffffffffffda RBX: 000000000000000c RCX: 00007fc862314887
>> RDX: 000000000000000c RSI: 000056405645fe10 RDI: 0000000000000001
>> RBP: 000056405645fe10 R08: 00007fc8623d1460 R09: 000000007fffffff
>> R10: 0000000000000000 R11: 0000000000000246 R12: 000000000000000c
>> R13: 00007fc86241b780 R14: 00007fc862417600 R15: 00007fc862416a00
>>
>> In short, below scene breaks the lock dependency chain:
>>
>>   memory_failure
>>    __page_handle_poison
>>     zone_pcp_disable -- lock(pcp_batch_high_lock)
>>     dissolve_free_huge_page
>>      __hugetlb_vmemmap_restore_folio
>>       static_key_slow_dec
>>        cpus_read_lock -- rlock(cpu_hotplug_lock)
>>
>> Fix this by calling drain_all_pages() instead.
>>
>> Fixes: a6b40850c442 ("mm: hugetlb: replace hugetlb_free_vmemmap_enabled with a static_key")
>> Signed-off-by: Miaohe Lin <[email protected]>
>> Acked-by: Oscar Salvador <[email protected]>
>> Cc: <[email protected]>
>> ---
>> v2:
>>   collect Acked-by tag and extend comment per Oscar. Thanks.
>> ---
>>   mm/memory-failure.c | 16 +++++++++++++---
>>   1 file changed, 13 insertions(+), 3 deletions(-)
>>
>> diff --git a/mm/memory-failure.c b/mm/memory-failure.c
>> index edd6e114462f..c6750509d74c 100644
>> --- a/mm/memory-failure.c
>> +++ b/mm/memory-failure.c
>> @@ -153,11 +153,21 @@ static int __page_handle_poison(struct page *page)
>>   {
>>       int ret;
>>   -    zone_pcp_disable(page_zone(page));
>> +    /*
>> +     * zone_pcp_disable() can't be used here. It will hold pcp_batch_high_lock and
>> +     * dissolve_free_huge_page() might hold cpu_hotplug_lock via static_key_slow_dec()
>> +     * when hugetlb vmemmap optimization is enabled. This will break current lock
>> +     * dependency chain and leads to deadlock.
>> +     * Disabling pcp before dissolving the page was a deterministic approach because
>> +     * we made sure that those pages cannot end up in any PCP list. Draining PCP lists
>> +     * expels those pages to the buddy system, but nothing guarantees that those pages
>> +     * do not get back to a PCP queue if we need to refill those.
>> +     */
>>       ret = dissolve_free_huge_page(page);
>
> Hi Miaohe,
>
> I recently sent a patch[1] to convert dissolve_free_huge_page() to folios which changes the function name and the name referenced in the comment so this will conflict with my patch. It's in mm-unstable now, would you be able to rebase to that in a new version?
>

The version 1 of this patch is in mm-unstable too. So it might be better to send a separate patch to extend the comment.

Thanks.
.

> Thanks,
> Sid
>
> [1] https://lore.kernel.org/linux-mm/[email protected]/T/#u
>
>
>> -    if (!ret)
>> +    if (!ret) {
>> +        drain_all_pages(page_zone(page));
>>           ret = take_page_off_buddy(page);
>> -    zone_pcp_enable(page_zone(page));
>> +    }
>>         return ret;
>>   }
>
> .


2024-04-12 23:21:20

by Andrew Morton

[permalink] [raw]
Subject: Re: [PATCH v2] mm/memory-failure: fix deadlock when hugetlb_optimize_vmemmap is enabled

On Fri, 12 Apr 2024 16:11:52 +0800 Miaohe Lin <[email protected]> wrote:

> > I recently sent a patch[1] to convert dissolve_free_huge_page() to folios which changes the function name and the name referenced in the comment so this will conflict with my patch. It's in mm-unstable now, would you be able to rebase to that in a new version?
> >

This patch is a hotfixes, cc:stable one so the mm-unstable material will be
based on top of this change.

I've queued this change up as a -fix against v1. And I've retained
this changelog addition:

: This issue won't occur until commit a6b40850c442 ("mm: hugetlb: replace
: hugetlb_free_vmemmap_enabled with a static_key"). As it introduced
: rlock(cpu_hotplug_lock) in dissolve_free_huge_page() code path while
: lock(pcp_batch_high_lock) is already in the __page_handle_poison().

And I've queued another -fix to reflow that block comment to 80 columns.

--- a/mm/memory-failure.c~mm-memory-failure-fix-deadlock-when-hugetlb_optimize_vmemmap-is-enabled-v2-fix
+++ a/mm/memory-failure.c
@@ -155,14 +155,16 @@ static int __page_handle_poison(struct p
int ret;

/*
- * zone_pcp_disable() can't be used here. It will hold pcp_batch_high_lock and
- * dissolve_free_huge_page() might hold cpu_hotplug_lock via static_key_slow_dec()
- * when hugetlb vmemmap optimization is enabled. This will break current lock
- * dependency chain and leads to deadlock.
- * Disabling pcp before dissolving the page was a deterministic approach because
- * we made sure that those pages cannot end up in any PCP list. Draining PCP lists
- * expels those pages to the buddy system, but nothing guarantees that those pages
- * do not get back to a PCP queue if we need to refill those.
+ * zone_pcp_disable() can't be used here. It will
+ * hold pcp_batch_high_lock and dissolve_free_huge_page() might hold
+ * cpu_hotplug_lock via static_key_slow_dec() when hugetlb vmemmap
+ * optimization is enabled. This will break current lock dependency
+ * chain and leads to deadlock.
+ * Disabling pcp before dissolving the page was a deterministic
+ * approach because we made sure that those pages cannot end up in any
+ * PCP list. Draining PCP lists expels those pages to the buddy system,
+ * but nothing guarantees that those pages do not get back to a PCP
+ * queue if we need to refill those.
*/
ret = dissolve_free_huge_page(page);
if (!ret) {
_


2024-04-16 03:26:29

by Miaohe Lin

[permalink] [raw]
Subject: Re: [PATCH v2] mm/memory-failure: fix deadlock when hugetlb_optimize_vmemmap is enabled

On 2024/4/13 7:21, Andrew Morton wrote:
> On Fri, 12 Apr 2024 16:11:52 +0800 Miaohe Lin <[email protected]> wrote:
>
>>> I recently sent a patch[1] to convert dissolve_free_huge_page() to folios which changes the function name and the name referenced in the comment so this will conflict with my patch. It's in mm-unstable now, would you be able to rebase to that in a new version?
>>>
>
> This patch is a hotfixes, cc:stable one so the mm-unstable material will be
> based on top of this change.
>
> I've queued this change up as a -fix against v1. And I've retained
> this changelog addition:
>
> : This issue won't occur until commit a6b40850c442 ("mm: hugetlb: replace
> : hugetlb_free_vmemmap_enabled with a static_key"). As it introduced
> : rlock(cpu_hotplug_lock) in dissolve_free_huge_page() code path while
> : lock(pcp_batch_high_lock) is already in the __page_handle_poison().
>
> And I've queued another -fix to reflow that block comment to 80 columns.
>
> --- a/mm/memory-failure.c~mm-memory-failure-fix-deadlock-when-hugetlb_optimize_vmemmap-is-enabled-v2-fix
> +++ a/mm/memory-failure.c
> @@ -155,14 +155,16 @@ static int __page_handle_poison(struct p
> int ret;
>
> /*
> - * zone_pcp_disable() can't be used here. It will hold pcp_batch_high_lock and
> - * dissolve_free_huge_page() might hold cpu_hotplug_lock via static_key_slow_dec()
> - * when hugetlb vmemmap optimization is enabled. This will break current lock
> - * dependency chain and leads to deadlock.
> - * Disabling pcp before dissolving the page was a deterministic approach because
> - * we made sure that those pages cannot end up in any PCP list. Draining PCP lists
> - * expels those pages to the buddy system, but nothing guarantees that those pages
> - * do not get back to a PCP queue if we need to refill those.
> + * zone_pcp_disable() can't be used here. It will
> + * hold pcp_batch_high_lock and dissolve_free_huge_page() might hold
> + * cpu_hotplug_lock via static_key_slow_dec() when hugetlb vmemmap
> + * optimization is enabled. This will break current lock dependency
> + * chain and leads to deadlock.
> + * Disabling pcp before dissolving the page was a deterministic
> + * approach because we made sure that those pages cannot end up in any
> + * PCP list. Draining PCP lists expels those pages to the buddy system,
> + * but nothing guarantees that those pages do not get back to a PCP
> + * queue if we need to refill those.
> */
> ret = dissolve_free_huge_page(page);
> if (!ret) {
> _
>

Many thanks for doing this. :)
.

> .
>


2024-04-16 17:54:56

by Jane Chu

[permalink] [raw]
Subject: Re: [PATCH v2] mm/memory-failure: fix deadlock when hugetlb_optimize_vmemmap is enabled

On 4/11/2024 7:57 PM, Miaohe Lin wrote:

> When I did hard offline test with hugetlb pages, below deadlock occurs:
>
> ======================================================
> WARNING: possible circular locking dependency detected
> 6.8.0-11409-gf6cef5f8c37f #1 Not tainted
> ------------------------------------------------------
> bash/46904 is trying to acquire lock:
> ffffffffabe68910 (cpu_hotplug_lock){++++}-{0:0}, at: static_key_slow_dec+0x16/0x60
>
> but task is already holding lock:
> ffffffffabf92ea8 (pcp_batch_high_lock){+.+.}-{3:3}, at: zone_pcp_disable+0x16/0x40
>
> which lock already depends on the new lock.
>
> the existing dependency chain (in reverse order) is:
>
> -> #1 (pcp_batch_high_lock){+.+.}-{3:3}:
> __mutex_lock+0x6c/0x770
> page_alloc_cpu_online+0x3c/0x70
> cpuhp_invoke_callback+0x397/0x5f0
> __cpuhp_invoke_callback_range+0x71/0xe0
> _cpu_up+0xeb/0x210
> cpu_up+0x91/0xe0
> cpuhp_bringup_mask+0x49/0xb0
> bringup_nonboot_cpus+0xb7/0xe0
> smp_init+0x25/0xa0
> kernel_init_freeable+0x15f/0x3e0
> kernel_init+0x15/0x1b0
> ret_from_fork+0x2f/0x50
> ret_from_fork_asm+0x1a/0x30
>
> -> #0 (cpu_hotplug_lock){++++}-{0:0}:
> __lock_acquire+0x1298/0x1cd0
> lock_acquire+0xc0/0x2b0
> cpus_read_lock+0x2a/0xc0
> static_key_slow_dec+0x16/0x60
> __hugetlb_vmemmap_restore_folio+0x1b9/0x200
> dissolve_free_huge_page+0x211/0x260
> __page_handle_poison+0x45/0xc0
> memory_failure+0x65e/0xc70
> hard_offline_page_store+0x55/0xa0
> kernfs_fop_write_iter+0x12c/0x1d0
> vfs_write+0x387/0x550
> ksys_write+0x64/0xe0
> do_syscall_64+0xca/0x1e0
> entry_SYSCALL_64_after_hwframe+0x6d/0x75
>
> other info that might help us debug this:
>
> Possible unsafe locking scenario:
>
> CPU0 CPU1
> ---- ----
> lock(pcp_batch_high_lock);
> lock(cpu_hotplug_lock);
> lock(pcp_batch_high_lock);
> rlock(cpu_hotplug_lock);
>
> *** DEADLOCK ***
>
> 5 locks held by bash/46904:
> #0: ffff98f6c3bb23f0 (sb_writers#5){.+.+}-{0:0}, at: ksys_write+0x64/0xe0
> #1: ffff98f6c328e488 (&of->mutex){+.+.}-{3:3}, at: kernfs_fop_write_iter+0xf8/0x1d0
> #2: ffff98ef83b31890 (kn->active#113){.+.+}-{0:0}, at: kernfs_fop_write_iter+0x100/0x1d0
> #3: ffffffffabf9db48 (mf_mutex){+.+.}-{3:3}, at: memory_failure+0x44/0xc70
> #4: ffffffffabf92ea8 (pcp_batch_high_lock){+.+.}-{3:3}, at: zone_pcp_disable+0x16/0x40
>
> stack backtrace:
> CPU: 10 PID: 46904 Comm: bash Kdump: loaded Not tainted 6.8.0-11409-gf6cef5f8c37f #1
> Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.14.0-0-g155821a1990b-prebuilt.qemu.org 04/01/2014
> Call Trace:
> <TASK>
> dump_stack_lvl+0x68/0xa0
> check_noncircular+0x129/0x140
> __lock_acquire+0x1298/0x1cd0
> lock_acquire+0xc0/0x2b0
> cpus_read_lock+0x2a/0xc0
> static_key_slow_dec+0x16/0x60
> __hugetlb_vmemmap_restore_folio+0x1b9/0x200
> dissolve_free_huge_page+0x211/0x260
> __page_handle_poison+0x45/0xc0
> memory_failure+0x65e/0xc70
> hard_offline_page_store+0x55/0xa0
> kernfs_fop_write_iter+0x12c/0x1d0
> vfs_write+0x387/0x550
> ksys_write+0x64/0xe0
> do_syscall_64+0xca/0x1e0
> entry_SYSCALL_64_after_hwframe+0x6d/0x75
> RIP: 0033:0x7fc862314887
> Code: 10 00 f7 d8 64 89 02 48 c7 c0 ff ff ff ff eb b7 0f 1f 00 f3 0f 1e fa 64 8b 04 25 18 00 00 00 85 c0 75 10 b8 01 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 51 c3 48 83 ec 28 48 89 54 24 18 48 89 74 24
> RSP: 002b:00007fff19311268 EFLAGS: 00000246 ORIG_RAX: 0000000000000001
> RAX: ffffffffffffffda RBX: 000000000000000c RCX: 00007fc862314887
> RDX: 000000000000000c RSI: 000056405645fe10 RDI: 0000000000000001
> RBP: 000056405645fe10 R08: 00007fc8623d1460 R09: 000000007fffffff
> R10: 0000000000000000 R11: 0000000000000246 R12: 000000000000000c
> R13: 00007fc86241b780 R14: 00007fc862417600 R15: 00007fc862416a00
>
> In short, below scene breaks the lock dependency chain:
>
> memory_failure
> __page_handle_poison
> zone_pcp_disable -- lock(pcp_batch_high_lock)
> dissolve_free_huge_page
> __hugetlb_vmemmap_restore_folio
> static_key_slow_dec
> cpus_read_lock -- rlock(cpu_hotplug_lock)
>
> Fix this by calling drain_all_pages() instead.
>
> Fixes: a6b40850c442 ("mm: hugetlb: replace hugetlb_free_vmemmap_enabled with a static_key")
> Signed-off-by: Miaohe Lin <[email protected]>
> Acked-by: Oscar Salvador <[email protected]>
> Cc: <[email protected]>
> ---
> v2:
> collect Acked-by tag and extend comment per Oscar. Thanks.
> ---
> mm/memory-failure.c | 16 +++++++++++++---
> 1 file changed, 13 insertions(+), 3 deletions(-)
>
> diff --git a/mm/memory-failure.c b/mm/memory-failure.c
> index edd6e114462f..c6750509d74c 100644
> --- a/mm/memory-failure.c
> +++ b/mm/memory-failure.c
> @@ -153,11 +153,21 @@ static int __page_handle_poison(struct page *page)
> {
> int ret;
>
> - zone_pcp_disable(page_zone(page));
> + /*
> + * zone_pcp_disable() can't be used here. It will hold pcp_batch_high_lock and
> + * dissolve_free_huge_page() might hold cpu_hotplug_lock via static_key_slow_dec()
> + * when hugetlb vmemmap optimization is enabled. This will break current lock
> + * dependency chain and leads to deadlock.
> + * Disabling pcp before dissolving the page was a deterministic approach because
> + * we made sure that those pages cannot end up in any PCP list. Draining PCP lists
> + * expels those pages to the buddy system, but nothing guarantees that those pages
> + * do not get back to a PCP queue if we need to refill those.
> + */
> ret = dissolve_free_huge_page(page);
> - if (!ret)
> + if (!ret) {
> + drain_all_pages(page_zone(page));
> ret = take_page_off_buddy(page);
> - zone_pcp_enable(page_zone(page));
> + }
>
> return ret;
> }

Looks good.

Reviewed-by: Jane Chu <[email protected]>

-jane