[syzbot] KASAN: use-after-free Read in __oom_reap_task_mm

[syzbot]

Hello,

syzbot found the following issue on:

HEAD commit: e6251ab4551f Merge tag 'nfs-for-5.17-2' of git://git.linux..
git tree: upstream
console output: https://syzkaller.appspot.com/x/log.txt?x=1229b2f8700000
kernel config: https://syzkaller.appspot.com/x/.config?x=266de9da75c71a45
dashboard link: https://syzkaller.appspot.com/bug?extid=2ccf63a4bd07cf39cab0
compiler: gcc (Debian 10.2.1-6) 10.2.1 20210110, GNU ld (GNU Binutils for Debian) 2.35.2
userspace arch: i386

Unfortunately, I don't have any reproducer for this issue yet.

IMPORTANT: if you fix the issue, please add the following tag to the commit:
Reported-by: syzbot+2ccf63...@syzkaller.appspotmail.com

==================================================================
BUG: KASAN: use-after-free in can_madv_lru_vma mm/internal.h:76 [inline]
BUG: KASAN: use-after-free in __oom_reap_task_mm+0x3a4/0x400 mm/oom_kill.c:529
Read of size 8 at addr ffff88807b868ba8 by task syz-executor.2/12778

CPU: 0 PID: 12778 Comm: syz-executor.2 Not tainted 5.17.0-rc3-syzkaller-00029-ge6251ab4551f #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Call Trace:
<TASK>
__dump_stack lib/dump_stack.c:88 [inline]
dump_stack_lvl+0xcd/0x134 lib/dump_stack.c:106
print_address_description.constprop.0.cold+0x8d/0x336 mm/kasan/report.c:255
__kasan_report mm/kasan/report.c:442 [inline]
kasan_report.cold+0x83/0xdf mm/kasan/report.c:459
can_madv_lru_vma mm/internal.h:76 [inline]
__oom_reap_task_mm+0x3a4/0x400 mm/oom_kill.c:529
__do_sys_process_mrelease+0x3f2/0x450 mm/oom_kill.c:1196
do_syscall_32_irqs_on arch/x86/entry/common.c:112 [inline]
__do_fast_syscall_32+0x65/0xf0 arch/x86/entry/common.c:178
do_fast_syscall_32+0x2f/0x70 arch/x86/entry/common.c:203
entry_SYSENTER_compat_after_hwframe+0x4d/0x5c
RIP: 0023:0xf6e70549
Code: 03 74 c0 01 10 05 03 74 b8 01 10 06 03 74 b4 01 10 07 03 74 b0 01 10 08 03 74 d8 01 00 00 00 00 00 51 52 55 89 e5 0f 34 cd 80 <5d> 5a 59 c3 90 90 90 90 8d b4 26 00 00 00 00 8d b4 26 00 00 00 00
RSP: 002b:00000000f586a5cc EFLAGS: 00000296 ORIG_RAX: 00000000000001c0
RAX: ffffffffffffffda RBX: 0000000000000003 RCX: 0000000000000000
RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000000
RBP: 0000000000000000 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000000
R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000
</TASK>

Allocated by task 12778:
kasan_save_stack+0x1e/0x40 mm/kasan/common.c:38
kasan_set_track mm/kasan/common.c:45 [inline]
set_alloc_info mm/kasan/common.c:436 [inline]
__kasan_slab_alloc+0x90/0xc0 mm/kasan/common.c:469
kasan_slab_alloc include/linux/kasan.h:260 [inline]
slab_post_alloc_hook mm/slab.h:732 [inline]
slab_alloc_node mm/slub.c:3230 [inline]
slab_alloc mm/slub.c:3238 [inline]
kmem_cache_alloc+0x202/0x3a0 mm/slub.c:3243
vm_area_dup+0x88/0x2b0 kernel/fork.c:357
dup_mmap kernel/fork.c:554 [inline]
dup_mm+0x5fa/0x13e0 kernel/fork.c:1451
copy_mm kernel/fork.c:1503 [inline]
copy_process+0x71f8/0x7300 kernel/fork.c:2164
kernel_clone+0xe7/0xab0 kernel/fork.c:2555
__do_compat_sys_ia32_clone+0xac/0xe0 arch/x86/kernel/sys_ia32.c:254
do_syscall_32_irqs_on arch/x86/entry/common.c:112 [inline]
__do_fast_syscall_32+0x65/0xf0 arch/x86/entry/common.c:178
do_fast_syscall_32+0x2f/0x70 arch/x86/entry/common.c:203
entry_SYSENTER_compat_after_hwframe+0x4d/0x5c

Freed by task 12780:
kasan_save_stack+0x1e/0x40 mm/kasan/common.c:38
kasan_set_track+0x21/0x30 mm/kasan/common.c:45
kasan_set_free_info+0x20/0x30 mm/kasan/generic.c:370
____kasan_slab_free mm/kasan/common.c:366 [inline]
____kasan_slab_free+0x130/0x160 mm/kasan/common.c:328
kasan_slab_free include/linux/kasan.h:236 [inline]
slab_free_hook mm/slub.c:1728 [inline]
slab_free_freelist_hook+0x8b/0x1c0 mm/slub.c:1754
slab_free mm/slub.c:3509 [inline]
kmem_cache_free+0xd8/0x340 mm/slub.c:3526
remove_vma+0x135/0x170 mm/mmap.c:189
exit_mmap+0x29a/0x670 mm/mmap.c:3186
__mmput+0x122/0x4b0 kernel/fork.c:1114
mmput+0x56/0x60 kernel/fork.c:1135
exit_mm kernel/exit.c:507 [inline]
do_exit+0xa3c/0x2a30 kernel/exit.c:793
do_group_exit+0xd2/0x2f0 kernel/exit.c:935
get_signal+0x4b0/0x28c0 kernel/signal.c:2862
arch_do_signal_or_restart+0x2a9/0x1c40 arch/x86/kernel/signal.c:868
handle_signal_work kernel/entry/common.c:148 [inline]
exit_to_user_mode_loop kernel/entry/common.c:172 [inline]
exit_to_user_mode_prepare+0x17d/0x290 kernel/entry/common.c:207
__syscall_exit_to_user_mode_work kernel/entry/common.c:289 [inline]
syscall_exit_to_user_mode+0x19/0x60 kernel/entry/common.c:300
__do_fast_syscall_32+0x72/0xf0 arch/x86/entry/common.c:181
do_fast_syscall_32+0x2f/0x70 arch/x86/entry/common.c:203
entry_SYSENTER_compat_after_hwframe+0x4d/0x5c

The buggy address belongs to the object at ffff88807b868b58
which belongs to the cache vm_area_struct of size 200
The buggy address is located 80 bytes inside of
200-byte region [ffff88807b868b58, ffff88807b868c20)
The buggy address belongs to the page:
page:ffffea0001ee1a00 refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x7b868
memcg:ffff88804fe94a01
flags: 0xfff00000000200(slab|node=0|zone=1|lastcpupid=0x7ff)
raw: 00fff00000000200 dead000000000100 dead000000000122 ffff888140006a00
raw: 0000000000000000 00000000000f000f 00000001ffffffff ffff88804fe94a01
page dumped because: kasan: bad access detected
page_owner tracks the page as allocated
page last allocated via order 0, migratetype Unmovable, gfp_mask 0x112cc0(GFP_USER|__GFP_NOWARN|__GFP_NORETRY), pid 3175, ts 344170603641, free_ts 344164018467
prep_new_page mm/page_alloc.c:2434 [inline]
get_page_from_freelist+0xa72/0x2f50 mm/page_alloc.c:4165
__alloc_pages+0x1b2/0x500 mm/page_alloc.c:5389
alloc_pages+0x1aa/0x310 mm/mempolicy.c:2271
alloc_slab_page mm/slub.c:1799 [inline]
allocate_slab mm/slub.c:1944 [inline]
new_slab+0x28a/0x3b0 mm/slub.c:2004
___slab_alloc+0x87c/0xe90 mm/slub.c:3018
__slab_alloc.constprop.0+0x4d/0xa0 mm/slub.c:3105
slab_alloc_node mm/slub.c:3196 [inline]
slab_alloc mm/slub.c:3238 [inline]
kmem_cache_alloc+0x35c/0x3a0 mm/slub.c:3243
vm_area_dup+0x88/0x2b0 kernel/fork.c:357
dup_mmap kernel/fork.c:554 [inline]
dup_mm+0x5fa/0x13e0 kernel/fork.c:1451
copy_mm kernel/fork.c:1503 [inline]
copy_process+0x71f8/0x7300 kernel/fork.c:2164
kernel_clone+0xe7/0xab0 kernel/fork.c:2555
__do_sys_clone+0xc8/0x110 kernel/fork.c:2672
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x44/0xae
page last free stack trace:
reset_page_owner include/linux/page_owner.h:24 [inline]
free_pages_prepare mm/page_alloc.c:1352 [inline]
free_pcp_prepare+0x374/0x870 mm/page_alloc.c:1404
free_unref_page_prepare mm/page_alloc.c:3325 [inline]
free_unref_page_list+0x1a9/0xfa0 mm/page_alloc.c:3441
release_pages+0x317/0x1220 mm/swap.c:980
tlb_batch_pages_flush mm/mmu_gather.c:50 [inline]
tlb_flush_mmu_free mm/mmu_gather.c:243 [inline]
tlb_flush_mmu mm/mmu_gather.c:250 [inline]
tlb_finish_mmu+0x165/0x8c0 mm/mmu_gather.c:341
exit_mmap+0x21b/0x670 mm/mmap.c:3180
__mmput+0x122/0x4b0 kernel/fork.c:1114
mmput+0x56/0x60 kernel/fork.c:1135
exit_mm kernel/exit.c:507 [inline]
do_exit+0xa3c/0x2a30 kernel/exit.c:793
do_group_exit+0xd2/0x2f0 kernel/exit.c:935
get_signal+0x4b0/0x28c0 kernel/signal.c:2862
arch_do_signal_or_restart+0x2a9/0x1c40 arch/x86/kernel/signal.c:868
handle_signal_work kernel/entry/common.c:148 [inline]
exit_to_user_mode_loop kernel/entry/common.c:172 [inline]
exit_to_user_mode_prepare+0x17d/0x290 kernel/entry/common.c:207
__syscall_exit_to_user_mode_work kernel/entry/common.c:289 [inline]
syscall_exit_to_user_mode+0x19/0x60 kernel/entry/common.c:300
__do_fast_syscall_32+0x72/0xf0 arch/x86/entry/common.c:181
do_fast_syscall_32+0x2f/0x70 arch/x86/entry/common.c:203
entry_SYSENTER_compat_after_hwframe+0x4d/0x5c

Memory state around the buggy address:
ffff88807b868a80: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
ffff88807b868b00: fb fb fb fc fc fc fc fc fc fc fc fa fb fb fb fb
>ffff88807b868b80: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
^
ffff88807b868c00: fb fb fb fb fc fc fc fc fc fc fc fc fa fb fb fb
ffff88807b868c80: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
==================================================================
----------------
Code disassembly (best guess):
0: 03 74 c0 01 add 0x1(%rax,%rax,8),%esi
4: 10 05 03 74 b8 01 adc %al,0x1b87403(%rip) # 0x1b8740d
a: 10 06 adc %al,(%rsi)
c: 03 74 b4 01 add 0x1(%rsp,%rsi,4),%esi
10: 10 07 adc %al,(%rdi)
12: 03 74 b0 01 add 0x1(%rax,%rsi,4),%esi
16: 10 08 adc %cl,(%rax)
18: 03 74 d8 01 add 0x1(%rax,%rbx,8),%esi
1c: 00 00 add %al,(%rax)
1e: 00 00 add %al,(%rax)
20: 00 51 52 add %dl,0x52(%rcx)
23: 55 push %rbp
24: 89 e5 mov %esp,%ebp
26: 0f 34 sysenter
28: cd 80 int $0x80
* 2a: 5d pop %rbp <-- trapping instruction
2b: 5a pop %rdx
2c: 59 pop %rcx
2d: c3 retq
2e: 90 nop
2f: 90 nop
30: 90 nop
31: 90 nop
32: 8d b4 26 00 00 00 00 lea 0x0(%rsi,%riz,1),%esi
39: 8d b4 26 00 00 00 00 lea 0x0(%rsi,%riz,1),%esi


---
This report is generated by a bot. It may contain errors.
See https://goo.gl/tpsmEJ for more information about syzbot.
syzbot engineers can be reached at syzk...@googlegroups.com.

syzbot will keep track of this issue. See:
https://goo.gl/tpsmEJ#status for how to communicate with syzbot.

[Yang Shi]

It seems process_mrelease() may be racy with exiting. mmgrab() just
pin mm struct, but doesn't guarantee the associated address space
still exists. mmget or mmget_not_zero() may be needed at first glance.

The below untested patch (compiled) may be able to fix it.

diff --git a/mm/oom_kill.c b/mm/oom_kill.c
index 832fb330376e..c389995bcc71 100644
--- a/mm/oom_kill.c
+++ b/mm/oom_kill.c
@@ -1185,9 +1185,11 @@ SYSCALL_DEFINE2(process_mrelease, int, pidfd,
unsigned int, flags)
if (!reap)
goto drop_mm;

+ if (!mmget_not_zero(mm))
+ goto drop_mm;
if (mmap_read_lock_killable(mm)) {
ret = -EINTR;
- goto drop_mm;
+ goto put_mm;
}
/*
* Check MMF_OOM_SKIP again under mmap_read_lock protection to ensure
@@ -1197,6 +1199,8 @@ SYSCALL_DEFINE2(process_mrelease, int, pidfd,
unsigned int, flags)
ret = -EAGAIN;
mmap_read_unlock(mm);

+put_mm:
+ mmput(mm);
drop_mm:
mmdrop(mm);
put_task:

[Michal Hocko]

[Add Suren - and keeping the full message for reference. My comment in
inlined]

This is what process_mrelease used to do before ba535c1caf3e
("mm/oom_kill: allow process_mrelease to run under mmap_lock
protection"). It shouldn't be really necessary because of the
synchronization that happens on the exit_mmap level. There were some
patches in that area and I am not sure whether something is still
hanging in the Andrew's queue. Suren will surely know better.

--
Michal Hocko
SUSE Labs

[Suren Baghdasaryan]

->

Correct, the proposed change is what we were doing before 64591e8605d6
("mm: protect free_pgtables with mmap_lock write lock in exit_mmap")
and ba535c1caf3e ("mm/oom_kill: allow process_mrelease to run under
mmap_lock protection"). The first patch changed exit_mmap to call
unlock_range, free_pgtables and remove_vma under mmap_write_lock and
the second patch calls __oom_reap_task_mm under mmap_read_lock. From
the call stack IIUC __oom_reap_task_mm at
https://elixir.bootlin.com/linux/v5.17-rc4/source/mm/oom_kill.c#L1196
is racing with remove_vma at
https://elixir.bootlin.com/linux/v5.17-rc4/source/mm/mmap.c#L3186. The
mmap_lock should prevent this race and I don't see anything missing in
Andrew's tree. I'll go over the code again and re-check but so far I
don't see how this race could occur.

[Yang Shi]

Isn't the below race possible?

CPU A CPU B
exiting:
mmap_write_lock
remove_vma()
mmap_write_unlock
process_mrelease:
mmap_read_lock
__oom_reap_task_mm
mmap_read_unlock

[Suren Baghdasaryan]

Sure, that sequence (would not call it a race) is possible but in this
case __oom_reap_task_mm will find no vmas in the mm because exit_mmap
freed and removed all of them.

[Michal Hocko]

On Tue 15-02-22 10:10:53, Suren Baghdasaryan wrote:
> On Tue, Feb 15, 2022 at 9:53 AM Yang Shi <shy8...@gmail.com> wrote:

[...]

> > Isn't the below race possible?
> >
> > CPU A CPU B
> > exiting:
> > mmap_write_lock
> > remove_vma()
> > mmap_write_unlock
> > process_mrelease:
> > mmap_read_lock
> > __oom_reap_task_mm
> > mmap_read_unlock
> >
>
> Sure, that sequence (would not call it a race) is possible but in this
> case __oom_reap_task_mm will find no vmas in the mm because exit_mmap
> freed and removed all of them.

I didn't really have chance to have a closer look but I do not see
exit_mmap doing mm->mmap = NULL so the pointer can be a freed vma unless
I am missing something. I thought we've had it in your patches? Has this
got lost somewhere in the process?

[Suren Baghdasaryan]

Doh! Yes, it looks like I completely missed the actual pointer. That
must be it since I don't see any other possibility. Will post a patch
shortly.
Thanks!

[Suren Baghdasaryan]

Fix posted at https://lore.kernel.org/all/20220215201922....@google.com/