mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
* 【BUG】NULL pointer dereference at __lookup_swap_cgroup
@ 2022-11-25 10:11 xialonglong
  2022-11-28  1:08 ` Huang, Ying
  0 siblings, 1 reply; 4+ messages in thread
From: xialonglong @ 2022-11-25 10:11 UTC (permalink / raw)
  To: linux-kernel, hannes@cmpxchg.org linux-mm, mhocko,
	roman.gushchin, shakeelb
  Cc: Wangkefeng (OS Kernel Lab), chenwandun, ying.huang, songmuchun

A panic occur in the linux 5.10,we meet it only once,it seems that
there is no special changes between 5.10 and upsteam about swap_cgroup.

The test is based on QEMU with 64GB memory, one 2GB zram device as
swap area.
The test steps:
1.swapoff -a
2.add some memory pressure by stress-ng
3.while (2 minutes) {
   swapoff /dev/zram0
   swapon /dev/zram0
   sleep 3
}
4. swapon -a

Preliminary analysis showed that the swap entry point to a swap area
which have already been swapoff, and no other obvious clues, still
trying to reproduce it.

Any known issue about this feature, or any advise will be appreciated.

Here are the panic log,

Unable to handle kernel NULL pointer dereference at virtual address
0000000000000740
Mem abort info:
ESR = 0x96000004
EC = 0x25: DABT (current EL), IL = 32 bits SET = 0, FnV = 0 EA = 0,
S1PTW = 0 Data abort info:
ISV = 0, ISS = 0x00000004
CM = 0, WnR = 0
user pgtable: 4k pages, 48-bit VAs, pgdp=000000010ae6e000
pgd=0000000000000000, p4d=0000000000000000 Internal error: Oops:
96000004 [#1] SMP Hardware name: QEMU KVM Virtual Machine, BIOS 0.0.0 
02/06/2015
pstate: 00000005 (nzcv daif -PAN -UAO -TCO BTYPE=--)
pc : lookup_swap_cgroup_id+0x38/0x50
lr : mem_cgroup_charge+0x9c/0x424
sp : ffff800102f63bc0
x29: ffff800102f63bc0 x28: ffff0000d0d64d00
x27: 0000000000000000 x26: 0000000000000007
x25: ffff0000018c86a8 x24: ffff0000018c8640
x23: 0000000000000cc0 x22: 0000000000000001
x21: 0000000000000001 x20: ffff800102f63d28
x19: fffffe000373cb40 x18: 0000000000000000
x17: 0000000000000000 x16: ffff8001004715a4
x15: 00000000ffffffff x14: 0000000000003000
x13: 00000000ffffffff x12: 0000000000000040
x11: ffff0000c0403478 x10: ffff0000c040347a
x9 : ffff8001003e957c x8 : 000000000009dddd
x7 : 0000000000000600 x6 : 00000000000000e8
x5 : 0000020000200000 x4 : ffff000000000000
x3 : ffff800101f4c030 x2 : 0000000000000000
x1 : 00000000000001e4 x0 : 0000000000000000

Call trace:
lookup_swap_cgroup_id+0x38/0x50
do_swap_page+0xa64/0xc04
handle_pte_fault+0x1c8/0x214
__handle_mm_fault+0x1b0/0x380
handle_mm_fault+0xf4/0x284
do_page_fault+0x188/0x474
do_translation_fault+0xb8/0xe4
do_mem_abort+0x48/0xb0
el0_da+0x44/0x80
el0_sync_handler+0x88/0xb4
el0_sync+0x160/0x180

<lookup_swap_cgroup_id>:             mov     x9, x30
<lookup_swap_cgroup_id+0x4>:         nop
<lookup_swap_cgroup_id+0x8>:         lsr     x2, x0, #58 SWP_TYPE_SHIFT 
== 58    x2 = swp_type
<lookup_swap_cgroup_id+0xc>:         adrp    x1, 0xffff800101f4c000 
<memcg_sockets_enabled_key+0x8>
<lookup_swap_cgroup_id+0x10>:        add     x3, x1, #0x30          x3 
== swap_cgroup_ctrl
<lookup_swap_cgroup_id+0x14>:        ubfx    x6, x0, #11, #47
<lookup_swap_cgroup_id+0x18>:        add     x2, x2, x2, lsl #1
<lookup_swap_cgroup_id+0x1c>:        ubfiz   x1, x0, #1, #11
<lookup_swap_cgroup_id+0x20>:        mov     x5, 
#0x200000                   // #2097152
<lookup_swap_cgroup_id+0x24>:        mov     x4, 
#0xffff000000000000         // #-281474976710656
<lookup_swap_cgroup_id+0x28>:        movk    x5, #0x200, lsl #32
<lookup_swap_cgroup_id+0x2c>:        hint    #0x19
<lookup_swap_cgroup_id+0x30>:        ldr     x0, [x3,x2,lsl #3] 
x3=ffff800101f4c030, x0 = 0
<lookup_swap_cgroup_id+0x34>:        hint    #0x1d
<lookup_swap_cgroup_id+0x38>:        ldr     x0, [x0,x6,lsl #3]  x0 = 0 
+ 0xe8 * 8 == 0x740
<lookup_swap_cgroup_id+0x3c>:        add     x0, x0, x5
<lookup_swap_cgroup_id+0x40>:        lsr     x0, x0, #6
<lookup_swap_cgroup_id+0x44>:        add     x0, x1, x0, lsl #12
<lookup_swap_cgroup_id+0x48>:        ldrh    w0, [x0,x4]
<lookup_swap_cgroup_id+0x4c>:        ret


^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: 【BUG】NULL pointer dereference at __lookup_swap_cgroup
  2022-11-25 10:11 【BUG】NULL pointer dereference at __lookup_swap_cgroup xialonglong
@ 2022-11-28  1:08 ` Huang, Ying
  2022-11-30  1:31   ` xialonglong
  0 siblings, 1 reply; 4+ messages in thread
From: Huang, Ying @ 2022-11-28  1:08 UTC (permalink / raw)
  To: xialonglong
  Cc: linux-kernel, hannes, linux-mm, mhocko, roman.gushchin, shakeelb,
	Wangkefeng (OS Kernel Lab),
	chenwandun, songmuchun

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=ascii, Size: 4922 bytes --]

Hi,

xialonglong <xialonglong1@huawei.com> writes:

> A panic occur in the linux 5.10\fwe meet it only once\fit seems that
> there is no special changes between 5.10 and upsteam about swap_cgroup.
>
> The test is based on QEMU with 64GB memory, one 2GB zram device as
> swap area.
> The test steps:
> 1.swapoff -a
> 2.add some memory pressure by stress-ng
> 3.while (2 minutes) {
>  swapoff /dev/zram0
>  swapon /dev/zram0
>  sleep 3
> }
> 4. swapon -a
>
> Preliminary analysis showed that the swap entry point to a swap area
> which have already been swapoff, and no other obvious clues, still
> trying to reproduce it.

We have a patch as follows to fix a similar issue,

2799e77529c2a25492a4395db93996e3dacd762d
Author:     Miaohe Lin <linmiaohe@huawei.com>
AuthorDate: Mon Jun 28 19:36:50 2021 -0700
Commit:     Linus Torvalds <torvalds@linux-foundation.org>
CommitDate: Tue Jun 29 10:53:49 2021 -0700

swap: fix do_swap_page() race with swapoff

When I was investigating the swap code, I found the below possible race
window:

CPU 1                                   	CPU 2
-----                                   	-----
do_swap_page
  if (data_race(si->flags & SWP_SYNCHRONOUS_IO)
  swap_readpage
    if (data_race(sis->flags & SWP_FS_OPS)) {
                                        	swapoff
					  	  ..
					  	  p->swap_file = NULL;
					  	  ..
    struct file *swap_file = sis->swap_file;
    struct address_space *mapping = swap_file->f_mapping;[oops!]

Note that for the pages that are swapped in through swap cache, this isn't
an issue. Because the page is locked, and the swap entry will be marked
with SWAP_HAS_CACHE, so swapoff() can not proceed until the page has been
unlocked.

Fix this race by using get/put_swap_device() to guard against concurrent
swapoff.

Can you check whether that can fix your issue?

Best Regards,
Huang, Ying

> Any known issue about this feature, or any advise will be appreciated.
>
> Here are the panic log,
>
> Unable to handle kernel NULL pointer dereference at virtual address
> 0000000000000740
> Mem abort info:
> ESR = 0x96000004
> EC = 0x25: DABT (current EL), IL = 32 bits SET = 0, FnV = 0 EA = 0,
> S1PTW = 0 Data abort info:
> ISV = 0, ISS = 0x00000004
> CM = 0, WnR = 0
> user pgtable: 4k pages, 48-bit VAs, pgdp=000000010ae6e000
> pgd=0000000000000000, p4d=0000000000000000 Internal error: Oops:
> 96000004 [#1] SMP Hardware name: QEMU KVM Virtual Machine, BIOS 0.0.0
> 02/06/2015
> pstate: 00000005 (nzcv daif -PAN -UAO -TCO BTYPE=--)
> pc : lookup_swap_cgroup_id+0x38/0x50
> lr : mem_cgroup_charge+0x9c/0x424
> sp : ffff800102f63bc0
> x29: ffff800102f63bc0 x28: ffff0000d0d64d00
> x27: 0000000000000000 x26: 0000000000000007
> x25: ffff0000018c86a8 x24: ffff0000018c8640
> x23: 0000000000000cc0 x22: 0000000000000001
> x21: 0000000000000001 x20: ffff800102f63d28
> x19: fffffe000373cb40 x18: 0000000000000000
> x17: 0000000000000000 x16: ffff8001004715a4
> x15: 00000000ffffffff x14: 0000000000003000
> x13: 00000000ffffffff x12: 0000000000000040
> x11: ffff0000c0403478 x10: ffff0000c040347a
> x9 : ffff8001003e957c x8 : 000000000009dddd
> x7 : 0000000000000600 x6 : 00000000000000e8
> x5 : 0000020000200000 x4 : ffff000000000000
> x3 : ffff800101f4c030 x2 : 0000000000000000
> x1 : 00000000000001e4 x0 : 0000000000000000
>
> Call trace:
> lookup_swap_cgroup_id+0x38/0x50
> do_swap_page+0xa64/0xc04
> handle_pte_fault+0x1c8/0x214
> __handle_mm_fault+0x1b0/0x380
> handle_mm_fault+0xf4/0x284
> do_page_fault+0x188/0x474
> do_translation_fault+0xb8/0xe4
> do_mem_abort+0x48/0xb0
> el0_da+0x44/0x80
> el0_sync_handler+0x88/0xb4
> el0_sync+0x160/0x180
>
> <lookup_swap_cgroup_id>:       mov   x9, x30
> <lookup_swap_cgroup_id+0x4>:     nop
> <lookup_swap_cgroup_id+0x8>:    
> lsr   x2, x0, #58 SWP_TYPE_SHIFT == 58  x2 =
> swp_type
> <lookup_swap_cgroup_id+0xc>:    
> adrp  x1, 0xffff800101f4c000
> <memcg_sockets_enabled_key+0x8>
> <lookup_swap_cgroup_id+0x10>:   
> add   x3, x1, #0x30    
> x3 == swap_cgroup_ctrl
> <lookup_swap_cgroup_id+0x14>:    ubfx  x6, x0, #11, #47
> <lookup_swap_cgroup_id+0x18>:    add   x2, x2, x2, lsl #1
> <lookup_swap_cgroup_id+0x1c>:    ubfiz  x1, x0, #1, #11
> <lookup_swap_cgroup_id+0x20>:   
> mov   x5,
> #0x200000         
> // #2097152
> <lookup_swap_cgroup_id+0x24>:   
> mov   x4,
> #0xffff000000000000     //
> #-281474976710656
> <lookup_swap_cgroup_id+0x28>:    movk  x5, #0x200, lsl #32
> <lookup_swap_cgroup_id+0x2c>:    hint  #0x19
> <lookup_swap_cgroup_id+0x30>:   
> ldr   x0, [x3,x2,lsl #3] x3=ffff800101f4c030, x0 = 0
> <lookup_swap_cgroup_id+0x34>:    hint  #0x1d
> <lookup_swap_cgroup_id+0x38>:   
> ldr   x0, [x0,x6,lsl #3] x0 = 0 + 0xe8 * 8 == 0x740
> <lookup_swap_cgroup_id+0x3c>:    add   x0, x0, x5
> <lookup_swap_cgroup_id+0x40>:    lsr   x0, x0, #6
> <lookup_swap_cgroup_id+0x44>:    add   x0, x1, x0, lsl #12
> <lookup_swap_cgroup_id+0x48>:    ldrh  w0, [x0,x4]
> <lookup_swap_cgroup_id+0x4c>:    ret

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: 【BUG】NULL pointer dereference at __lookup_swap_cgroup
  2022-11-28  1:08 ` Huang, Ying
@ 2022-11-30  1:31   ` xialonglong
  2022-11-30  2:37     ` Huang, Ying
  0 siblings, 1 reply; 4+ messages in thread
From: xialonglong @ 2022-11-30  1:31 UTC (permalink / raw)
  To: Huang, Ying
  Cc: linux-kernel, hannes, linux-mm, mhocko, roman.gushchin, shakeelb,
	Wangkefeng (OS Kernel Lab),
	chenwandun, songmuchun, gregkh

Thank you very much for your reply  :)
Inspired by your reply,we successfully reproduced the bug.

The test steps:
1.swapon  /dev/zram0
2.add some memory pressure by stress-ng
3.calling swapoff /dev/zram0 in the do_swap_page function (this changed 
the source code)
4.bug occured in the same place.

After testing, this patch solves the bug.
Finally, there is a small question. Why linux5.10 revert this patch 
(2799e77529c2)?

We found that to fix this bug, the following patches may be required:
efa33fc7f6e mm/shmem: fix shmem_swapin() race with swapoff
5c046235a826 mm/swap: remove confusing checking for non_swap_entry() in 
swap_ra_info()
2799e77529c2 swap: fix do_swap_page() race with swapoff
63d8620ecf93 mm/swapfile: use percpu_ref to serialize against concurrent 
swapoff
seem like all this patchset is needed except commit 5c046235a826 
("mm/swap: remove confusing checking for non_swap_entry() in 
swap_ra_info()")

Best Regards,
Xia, longlong

在 2022/11/28 9:08, Huang, Ying 写道:
> Hi,
>
> xialonglong <xialonglong1@huawei.com> writes:
>
>> A panic occur in the linux 5.10\fwe meet it only once\fit seems that
>> there is no special changes between 5.10 and upsteam about swap_cgroup.
>>
>> The test is based on QEMU with 64GB memory, one 2GB zram device as
>> swap area.
>> The test steps:
>> 1.swapoff -a
>> 2.add some memory pressure by stress-ng
>> 3.while (2 minutes) {
>>   swapoff /dev/zram0
>>   swapon /dev/zram0
>>   sleep 3
>> }
>> 4. swapon -a
>>
>> Preliminary analysis showed that the swap entry point to a swap area
>> which have already been swapoff, and no other obvious clues, still
>> trying to reproduce it.
> We have a patch as follows to fix a similar issue,
>
> 2799e77529c2a25492a4395db93996e3dacd762d
> Author:     Miaohe Lin <linmiaohe@huawei.com>
> AuthorDate: Mon Jun 28 19:36:50 2021 -0700
> Commit:     Linus Torvalds <torvalds@linux-foundation.org>
> CommitDate: Tue Jun 29 10:53:49 2021 -0700
>
> swap: fix do_swap_page() race with swapoff
>
> When I was investigating the swap code, I found the below possible race
> window:
>
> CPU 1                                   	CPU 2
> -----                                   	-----
> do_swap_page
>    if (data_race(si->flags & SWP_SYNCHRONOUS_IO)
>    swap_readpage
>      if (data_race(sis->flags & SWP_FS_OPS)) {
>                                          	swapoff
> 					  	  ..
> 					  	  p->swap_file = NULL;
> 					  	  ..
>      struct file *swap_file = sis->swap_file;
>      struct address_space *mapping = swap_file->f_mapping;[oops!]
>
> Note that for the pages that are swapped in through swap cache, this isn't
> an issue. Because the page is locked, and the swap entry will be marked
> with SWAP_HAS_CACHE, so swapoff() can not proceed until the page has been
> unlocked.
>
> Fix this race by using get/put_swap_device() to guard against concurrent
> swapoff.
>
> Can you check whether that can fix your issue?
>
> Best Regards,
> Huang, Ying
>
>> Any known issue about this feature, or any advise will be appreciated.
>>
>> Here are the panic log,
>>
>> Unable to handle kernel NULL pointer dereference at virtual address
>> 0000000000000740
>> Mem abort info:
>> ESR = 0x96000004
>> EC = 0x25: DABT (current EL), IL = 32 bits SET = 0, FnV = 0 EA = 0,
>> S1PTW = 0 Data abort info:
>> ISV = 0, ISS = 0x00000004
>> CM = 0, WnR = 0
>> user pgtable: 4k pages, 48-bit VAs, pgdp=000000010ae6e000
>> pgd=0000000000000000, p4d=0000000000000000 Internal error: Oops:
>> 96000004 [#1] SMP Hardware name: QEMU KVM Virtual Machine, BIOS 0.0.0
>> 02/06/2015
>> pstate: 00000005 (nzcv daif -PAN -UAO -TCO BTYPE=--)
>> pc : lookup_swap_cgroup_id+0x38/0x50
>> lr : mem_cgroup_charge+0x9c/0x424
>> sp : ffff800102f63bc0
>> x29: ffff800102f63bc0 x28: ffff0000d0d64d00
>> x27: 0000000000000000 x26: 0000000000000007
>> x25: ffff0000018c86a8 x24: ffff0000018c8640
>> x23: 0000000000000cc0 x22: 0000000000000001
>> x21: 0000000000000001 x20: ffff800102f63d28
>> x19: fffffe000373cb40 x18: 0000000000000000
>> x17: 0000000000000000 x16: ffff8001004715a4
>> x15: 00000000ffffffff x14: 0000000000003000
>> x13: 00000000ffffffff x12: 0000000000000040
>> x11: ffff0000c0403478 x10: ffff0000c040347a
>> x9 : ffff8001003e957c x8 : 000000000009dddd
>> x7 : 0000000000000600 x6 : 00000000000000e8
>> x5 : 0000020000200000 x4 : ffff000000000000
>> x3 : ffff800101f4c030 x2 : 0000000000000000
>> x1 : 00000000000001e4 x0 : 0000000000000000
>>
>> Call trace:
>> lookup_swap_cgroup_id+0x38/0x50
>> do_swap_page+0xa64/0xc04
>> handle_pte_fault+0x1c8/0x214
>> __handle_mm_fault+0x1b0/0x380
>> handle_mm_fault+0xf4/0x284
>> do_page_fault+0x188/0x474
>> do_translation_fault+0xb8/0xe4
>> do_mem_abort+0x48/0xb0
>> el0_da+0x44/0x80
>> el0_sync_handler+0x88/0xb4
>> el0_sync+0x160/0x180
>>
>> <lookup_swap_cgroup_id>:?????? mov?? x9, x30
>> <lookup_swap_cgroup_id+0x4>:???? nop
>> <lookup_swap_cgroup_id+0x8>:????
>> lsr?? x2, x0, #58 SWP_TYPE_SHIFT == 58? x2 =
>> swp_type
>> <lookup_swap_cgroup_id+0xc>:????
>> adrp? x1, 0xffff800101f4c000
>> <memcg_sockets_enabled_key+0x8>
>> <lookup_swap_cgroup_id+0x10>:???
>> add?? x3, x1, #0x30????
>> x3 == swap_cgroup_ctrl
>> <lookup_swap_cgroup_id+0x14>:??? ubfx? x6, x0, #11, #47
>> <lookup_swap_cgroup_id+0x18>:??? add?? x2, x2, x2, lsl #1
>> <lookup_swap_cgroup_id+0x1c>:??? ubfiz? x1, x0, #1, #11
>> <lookup_swap_cgroup_id+0x20>:???
>> mov?? x5,
>> #0x200000?????????
>> // #2097152
>> <lookup_swap_cgroup_id+0x24>:???
>> mov?? x4,
>> #0xffff000000000000???? //
>> #-281474976710656
>> <lookup_swap_cgroup_id+0x28>:??? movk? x5, #0x200, lsl #32
>> <lookup_swap_cgroup_id+0x2c>:??? hint? #0x19
>> <lookup_swap_cgroup_id+0x30>:???
>> ldr?? x0, [x3,x2,lsl #3] x3=ffff800101f4c030, x0 = 0
>> <lookup_swap_cgroup_id+0x34>:??? hint? #0x1d
>> <lookup_swap_cgroup_id+0x38>:???
>> ldr?? x0, [x0,x6,lsl #3] x0 = 0 + 0xe8 * 8 == 0x740
>> <lookup_swap_cgroup_id+0x3c>:??? add?? x0, x0, x5
>> <lookup_swap_cgroup_id+0x40>:??? lsr?? x0, x0, #6
>> <lookup_swap_cgroup_id+0x44>:??? add?? x0, x1, x0, lsl #12
>> <lookup_swap_cgroup_id+0x48>:??? ldrh? w0, [x0,x4]
>> <lookup_swap_cgroup_id+0x4c>:??? ret

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: 【BUG】NULL pointer dereference at __lookup_swap_cgroup
  2022-11-30  1:31   ` xialonglong
@ 2022-11-30  2:37     ` Huang, Ying
  0 siblings, 0 replies; 4+ messages in thread
From: Huang, Ying @ 2022-11-30  2:37 UTC (permalink / raw)
  To: xialonglong
  Cc: linux-kernel, hannes, linux-mm, mhocko, roman.gushchin, shakeelb,
	Wangkefeng (OS Kernel Lab),
	chenwandun, songmuchun, gregkh

xialonglong <xialonglong1@huawei.com> writes:

> Thank you very much for your reply :)
> Inspired by your reply\fwe successfully reproduced the bug.
>
> The test steps:
> 1.swapon /dev/zram0
> 2.add some memory pressure by stress-ng
> 3.calling swapoff /dev/zram0 in the do_swap_page function (this
> changed the source code)
> 4.bug occured in the same place.
>
> After testing, this patch solves the bug.
> Finally, there is a small question. Why linux5.10 revert this patch
> (2799e77529c2)?

2799e77529c2 is merged by v5.14.

Best Regards,
Huang, Ying

> We found that to fix this bug, the following patches may be required:
> efa33fc7f6e mm/shmem: fix shmem_swapin() race with swapoff
> 5c046235a826 mm/swap: remove confusing checking for non_swap_entry()
> in swap_ra_info()
> 2799e77529c2 swap: fix do_swap_page() race with swapoff
> 63d8620ecf93 mm/swapfile: use percpu_ref to serialize against
> concurrent swapoff
> seem like all this patchset is needed except commit 5c046235a826
> ("mm/swap: remove confusing checking for non_swap_entry() in 
> swap_ra_info()")
>
> Best Regards,
> Xia, longlong
>
> ( 2022/11/28 9:08, Huang, Ying S:
>> Hi,
>>
>> xialonglong <xialonglong1@huawei.com> writes:
>>
>>> A panic occur in the linux 5.10\fwe meet it only once\fit seems that
>>> there is no special changes between 5.10 and upsteam about swap_cgroup.
>>>
>>> The test is based on QEMU with 64GB memory, one 2GB zram device as
>>> swap area.
>>> The test steps:
>>> 1.swapoff -a
>>> 2.add some memory pressure by stress-ng
>>> 3.while (2 minutes) {
>>>   swapoff /dev/zram0
>>>   swapon /dev/zram0
>>>   sleep 3
>>> }
>>> 4. swapon -a
>>>
>>> Preliminary analysis showed that the swap entry point to a swap area
>>> which have already been swapoff, and no other obvious clues, still
>>> trying to reproduce it.
>> We have a patch as follows to fix a similar issue,
>>
>> 2799e77529c2a25492a4395db93996e3dacd762d
>> Author:     Miaohe Lin <linmiaohe@huawei.com>
>> AuthorDate: Mon Jun 28 19:36:50 2021 -0700
>> Commit:     Linus Torvalds <torvalds@linux-foundation.org>
>> CommitDate: Tue Jun 29 10:53:49 2021 -0700
>>
>> swap: fix do_swap_page() race with swapoff
>>
>> When I was investigating the swap code, I found the below possible race
>> window:
>>
>> CPU 1                                   	CPU 2
>> -----                                   	-----
>> do_swap_page
>>    if (data_race(si->flags & SWP_SYNCHRONOUS_IO)
>>    swap_readpage
>>      if (data_race(sis->flags & SWP_FS_OPS)) {
>>                                          	swapoff
>> 					  	  ..
>> 					  	  p->swap_file = NULL;
>> 					  	  ..
>>      struct file *swap_file = sis->swap_file;
>>      struct address_space *mapping = swap_file->f_mapping;[oops!]
>>
>> Note that for the pages that are swapped in through swap cache, this isn't
>> an issue. Because the page is locked, and the swap entry will be marked
>> with SWAP_HAS_CACHE, so swapoff() can not proceed until the page has been
>> unlocked.
>>
>> Fix this race by using get/put_swap_device() to guard against concurrent
>> swapoff.
>>
>> Can you check whether that can fix your issue?
>>
>> Best Regards,
>> Huang, Ying
>>
>>> Any known issue about this feature, or any advise will be appreciated.
>>>
>>> Here are the panic log,
>>>
>>> Unable to handle kernel NULL pointer dereference at virtual address
>>> 0000000000000740
>>> Mem abort info:
>>> ESR = 0x96000004
>>> EC = 0x25: DABT (current EL), IL = 32 bits SET = 0, FnV = 0 EA = 0,
>>> S1PTW = 0 Data abort info:
>>> ISV = 0, ISS = 0x00000004
>>> CM = 0, WnR = 0
>>> user pgtable: 4k pages, 48-bit VAs, pgdp=000000010ae6e000
>>> pgd=0000000000000000, p4d=0000000000000000 Internal error: Oops:
>>> 96000004 [#1] SMP Hardware name: QEMU KVM Virtual Machine, BIOS 0.0.0
>>> 02/06/2015
>>> pstate: 00000005 (nzcv daif -PAN -UAO -TCO BTYPE=--)
>>> pc : lookup_swap_cgroup_id+0x38/0x50
>>> lr : mem_cgroup_charge+0x9c/0x424
>>> sp : ffff800102f63bc0
>>> x29: ffff800102f63bc0 x28: ffff0000d0d64d00
>>> x27: 0000000000000000 x26: 0000000000000007
>>> x25: ffff0000018c86a8 x24: ffff0000018c8640
>>> x23: 0000000000000cc0 x22: 0000000000000001
>>> x21: 0000000000000001 x20: ffff800102f63d28
>>> x19: fffffe000373cb40 x18: 0000000000000000
>>> x17: 0000000000000000 x16: ffff8001004715a4
>>> x15: 00000000ffffffff x14: 0000000000003000
>>> x13: 00000000ffffffff x12: 0000000000000040
>>> x11: ffff0000c0403478 x10: ffff0000c040347a
>>> x9 : ffff8001003e957c x8 : 000000000009dddd
>>> x7 : 0000000000000600 x6 : 00000000000000e8
>>> x5 : 0000020000200000 x4 : ffff000000000000
>>> x3 : ffff800101f4c030 x2 : 0000000000000000
>>> x1 : 00000000000001e4 x0 : 0000000000000000
>>>
>>> Call trace:
>>> lookup_swap_cgroup_id+0x38/0x50
>>> do_swap_page+0xa64/0xc04
>>> handle_pte_fault+0x1c8/0x214
>>> __handle_mm_fault+0x1b0/0x380
>>> handle_mm_fault+0xf4/0x284
>>> do_page_fault+0x188/0x474
>>> do_translation_fault+0xb8/0xe4
>>> do_mem_abort+0x48/0xb0
>>> el0_da+0x44/0x80
>>> el0_sync_handler+0x88/0xb4
>>> el0_sync+0x160/0x180
>>>
>>> <lookup_swap_cgroup_id>:?????? mov?? x9, x30
>>> <lookup_swap_cgroup_id+0x4>:???? nop
>>> <lookup_swap_cgroup_id+0x8>:????
>>> lsr?? x2, x0, #58 SWP_TYPE_SHIFT == 58? x2 =
>>> swp_type
>>> <lookup_swap_cgroup_id+0xc>:????
>>> adrp? x1, 0xffff800101f4c000
>>> <memcg_sockets_enabled_key+0x8>
>>> <lookup_swap_cgroup_id+0x10>:???
>>> add?? x3, x1, #0x30????
>>> x3 == swap_cgroup_ctrl
>>> <lookup_swap_cgroup_id+0x14>:??? ubfx? x6, x0, #11, #47
>>> <lookup_swap_cgroup_id+0x18>:??? add?? x2, x2, x2, lsl #1
>>> <lookup_swap_cgroup_id+0x1c>:??? ubfiz? x1, x0, #1, #11
>>> <lookup_swap_cgroup_id+0x20>:???
>>> mov?? x5,
>>> #0x200000?????????
>>> // #2097152
>>> <lookup_swap_cgroup_id+0x24>:???
>>> mov?? x4,
>>> #0xffff000000000000???? //
>>> #-281474976710656
>>> <lookup_swap_cgroup_id+0x28>:??? movk? x5, #0x200, lsl #32
>>> <lookup_swap_cgroup_id+0x2c>:??? hint? #0x19
>>> <lookup_swap_cgroup_id+0x30>:???
>>> ldr?? x0, [x3,x2,lsl #3] x3=ffff800101f4c030, x0 = 0
>>> <lookup_swap_cgroup_id+0x34>:??? hint? #0x1d
>>> <lookup_swap_cgroup_id+0x38>:???
>>> ldr?? x0, [x0,x6,lsl #3] x0 = 0 + 0xe8 * 8 == 0x740
>>> <lookup_swap_cgroup_id+0x3c>:??? add?? x0, x0, x5
>>> <lookup_swap_cgroup_id+0x40>:??? lsr?? x0, x0, #6
>>> <lookup_swap_cgroup_id+0x44>:??? add?? x0, x1, x0, lsl #12
>>> <lookup_swap_cgroup_id+0x48>:??? ldrh? w0, [x0,x4]
>>> <lookup_swap_cgroup_id+0x4c>:??? ret

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2022-11-30  2:38 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-11-25 10:11 【BUG】NULL pointer dereference at __lookup_swap_cgroup xialonglong
2022-11-28  1:08 ` Huang, Ying
2022-11-30  1:31   ` xialonglong
2022-11-30  2:37     ` Huang, Ying

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®