Re: [syzbot] [mm?] WARNING in __folio_rmap_sanity_checks

[David Hildenbrand <david@redhat.com>]

On 05.01.24 18:04, Andrew Morton wrote:
> (cc netdev)
> 
> On Wed, 3 Jan 2024 22:16:13 +0800 "Yin, Fengwei" <fengwei.yin@intel.com> wrote:
> 
>>
>>
>> On 1/3/2024 8:13 PM, David Hildenbrand wrote:
>>> On 03.01.24 12:48, syzbot wrote:
>>>> Hello,
>>>>
>>>> syzbot found the following issue on:
>>>>
>>>> HEAD commit:    ab0b3e6ef50d Add linux-next specific files for 20240102
>>>> git tree:       linux-next
>>>> console+strace: https://syzkaller.appspot.com/x/log.txt?x=17be3e09e80000
>>>> kernel config:
>>>> https://syzkaller.appspot.com/x/.config?x=a14a6350374945f9
>>>> dashboard link:
>>>> https://syzkaller.appspot.com/bug?extid=50ef73537bbc393a25bb
>>>> compiler:       gcc (Debian 12.2.0-14) 12.2.0, GNU ld (GNU Binutils
>>>> for Debian) 2.40
>>>> syz repro:
>>>> https://syzkaller.appspot.com/x/repro.syz?x=14e2256ee80000
>>>> C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=17b57db5e80000
>>>>
>>>> Downloadable assets:
>>>> disk image:
>>>> https://storage.googleapis.com/syzbot-assets/4e6376fe5764/disk-ab0b3e6e.raw.xz
>>>> vmlinux:
>>>> https://storage.googleapis.com/syzbot-assets/7cb9ecbaf001/vmlinux-ab0b3e6e.xz
>>>> kernel image:
>>>> https://storage.googleapis.com/syzbot-assets/2c1a9a6d424f/bzImage-ab0b3e6e.xz
>>>>
>>>> The issue was bisected to:
>>>>
>>>> commit 68f0320824fa59c5429cbc811e6c46e7a30ea32c
>>>> Author: David Hildenbrand <david@redhat.com>
>>>> Date:   Wed Dec 20 22:44:31 2023 +0000
>>>>
>>>>       mm/rmap: convert folio_add_file_rmap_range() into
>>>> folio_add_file_rmap_[pte|ptes|pmd]()
>>>>
>>>> bisection log:
>>>> https://syzkaller.appspot.com/x/bisect.txt?x=10b9e1b1e80000
>>>> final oops:
>>>> https://syzkaller.appspot.com/x/report.txt?x=12b9e1b1e80000
>>>> console output: https://syzkaller.appspot.com/x/log.txt?x=14b9e1b1e80000
>>>>
>>>> IMPORTANT: if you fix the issue, please add the following tag to the
>>>> commit:
>>>> Reported-by: syzbot+50ef73537bbc393a25bb@syzkaller.appspotmail.com
>>>> Fixes: 68f0320824fa ("mm/rmap: convert folio_add_file_rmap_range()
>>>> into folio_add_file_rmap_[pte|ptes|pmd]()")
>>>>
>>>>    kasan_quarantine_reduce+0x18e/0x1d0 mm/kasan/quarantine.c:283
>>>>    __kasan_slab_alloc+0x65/0x90 mm/kasan/common.c:324
>>>>    kasan_slab_alloc include/linux/kasan.h:201 [inline]
>>>>    slab_post_alloc_hook mm/slub.c:3813 [inline]
>>>>    slab_alloc_node mm/slub.c:3860 [inline]
>>>>    kmem_cache_alloc+0x136/0x320 mm/slub.c:3867
>>>>    vm_area_alloc+0x1f/0x220 kernel/fork.c:465
>>>>    mmap_region+0x3ae/0x2a90 mm/mmap.c:2804
>>>>    do_mmap+0x890/0xef0 mm/mmap.c:1379
>>>>    vm_mmap_pgoff+0x1a7/0x3c0 mm/util.c:573
>>>>    ksys_mmap_pgoff+0x421/0x5a0 mm/mmap.c:1425
>>>>    __do_sys_mmap arch/x86/kernel/sys_x86_64.c:93 [inline]
>>>>    __se_sys_mmap arch/x86/kernel/sys_x86_64.c:86 [inline]
>>>>    __x64_sys_mmap+0x125/0x190 arch/x86/kernel/sys_x86_64.c:86
>>>>    do_syscall_x64 arch/x86/entry/common.c:52 [inline]
>>>>    do_syscall_64+0xd0/0x250 arch/x86/entry/common.c:83
>>>>    entry_SYSCALL_64_after_hwframe+0x62/0x6a
>>>> ------------[ cut here ]------------
>>>> WARNING: CPU: 1 PID: 5059 at include/linux/rmap.h:202
>>>> __folio_rmap_sanity_checks+0x4d5/0x630 include/linux/rmap.h:202
>>>> Modules linked in:
>>>> CPU: 1 PID: 5059 Comm: syz-executor115 Not tainted
>>>> 6.7.0-rc8-next-20240102-syzkaller #0
>>>> Hardware name: Google Google Compute Engine/Google Compute Engine,
>>>> BIOS Google 11/17/2023
>>>> RIP: 0010:__folio_rmap_sanity_checks+0x4d5/0x630 include/linux/rmap.h:202
>>>> Code: 41 83 e4 01 44 89 e6 e8 79 bc b7 ff 45 84 e4 0f 85 08 fc ff ff
>>>> e8 3b c1 b7 ff 48 c7 c6 e0 b5 d9 8a 48 89 df e8 5c 12 f7 ff 90 <0f> 0b
>>>> 90 e9 eb fb ff ff e8 1e c1 b7 ff be 01 00 00 00 48 89 df e8
>>>> RSP: 0018:ffffc900038df978 EFLAGS: 00010293
>>>> RAX: 0000000000000000 RBX: ffffea00008cde00 RCX: ffffffff81687419
>>>> RDX: ffff88807becbb80 RSI: ffffffff81d06104 RDI: 0000000000000000
>>>> RBP: ffffea00008cde00 R08: 0000000000000000 R09: fffffbfff1e75f6a
>>>> R10: ffffffff8f3afb57 R11: 0000000000000001 R12: 0000000000000000
>>>> R13: 0000000000000001 R14: 0000000000000000 R15: dffffc0000000000
>>>> FS:  0000555556508380(0000) GS:ffff8880b9900000(0000)
>>>> knlGS:0000000000000000
>>>> CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
>>>> CR2: 00000000200000c0 CR3: 0000000079000000 CR4: 00000000003506f0
>>>> DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
>>>> DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
>>>> Call Trace:
>>>>    <TASK>
>>>>    __folio_add_rmap mm/rmap.c:1167 [inline]
>>>>    __folio_add_file_rmap mm/rmap.c:1452 [inline]
>>>>    folio_add_file_rmap_ptes+0x8e/0x2c0 mm/rmap.c:1478
>>>>    insert_page_into_pte_locked.isra.0+0x34d/0x960 mm/memory.c:1874
>>>>    insert_page mm/memory.c:1900 [inline]
>>>>    vm_insert_page+0x62c/0x8c0 mm/memory.c:2053
>>>>    packet_mmap+0x314/0x570 net/packet/af_packet.c:4594
>>>>    call_mmap include/linux/fs.h:2090 [inline]
>>>>    mmap_region+0x745/0x2a90 mm/mmap.c:2819
>>>>    do_mmap+0x890/0xef0 mm/mmap.c:1379
>>>>    vm_mmap_pgoff+0x1a7/0x3c0 mm/util.c:573
>>>>    ksys_mmap_pgoff+0x421/0x5a0 mm/mmap.c:1425
>>>>    __do_sys_mmap arch/x86/kernel/sys_x86_64.c:93 [inline]
>>>>    __se_sys_mmap arch/x86/kernel/sys_x86_64.c:86 [inline]
>>>>    __x64_sys_mmap+0x125/0x190 arch/x86/kernel/sys_x86_64.c:86
>>>>    do_syscall_x64 arch/x86/entry/common.c:52 [inline]
>>>>    do_syscall_64+0xd0/0x250 arch/x86/entry/common.c:83
>>>>    entry_SYSCALL_64_after_hwframe+0x62/0x6a
>>>
>>> If I am not wrong, that triggers:
>>>
>>> VM_WARN_ON_FOLIO(folio_test_large(folio) &&
>>>            !folio_test_large_rmappable(folio), folio);
>>>
>>> So we are trying to rmap a large folio that did not go through
>>> folio_prep_large_rmappable().
>>>
>>> net/packet/af_packet.c calls vm_insert_page() on some pages/folios stoed
>>> in the "struct packet_ring_buffer". No idea where that comes from, but I
>>> suspect it's simply some compound allocation.
>> Looks like:
>>    alloc_pg_vec
>>      alloc_one_pg_vec_page
>>           gfp_t gfp_flags = GFP_KERNEL | __GFP_COMP |
>>                             __GFP_ZERO | __GFP_NOWARN | __GFP_NORETRY;
>>
>>           buffer = (char *) __get_free_pages(gfp_flags, order);
>> So you are right here... :).
> 
> Can we suggest what af_packet should be doing to avoid this?

I concluded that af_packet is doing the right thing and we have to 
figure out how we want to handle that internally differently at some point.

I just shared a patch to revert that check for now, let me know if you 
want an "officially" sent patch :)

-- 
Cheers,

David / dhildenb