linux-kernel - Re: [syzbot] [fs?] kernel BUG in kpagecount

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <6f2e8d0c-8940-4442-a023-477e5e8b9338@redhat.com>
Date: Fri, 10 Jan 2025 16:18:27 +0100
From: David Hildenbrand <david@...hat.com>
To: syzbot <syzbot+3d7dc5eaba6b932f8535@...kaller.appspotmail.com>,
 linux-fsdevel@...r.kernel.org, linux-kernel@...r.kernel.org,
 syzkaller-bugs@...glegroups.com
Subject: Re: [syzbot] [fs?] kernel BUG in kpagecount_read

On 10.01.25 16:09, David Hildenbrand wrote:
> On 10.01.25 15:33, syzbot wrote:
>> Hello,
>>
>> syzbot found the following issue on:
>>
>> HEAD commit:    c061cf420ded Merge tag 'trace-v6.13-rc3' of git://git.kern..
>> git tree:       upstream
>> console output: https://syzkaller.appspot.com/x/log.txt?x=11ee22df980000
>> kernel config:  https://syzkaller.appspot.com/x/.config?x=c22efbd20f8da769
>> dashboard link: https://syzkaller.appspot.com/bug?extid=3d7dc5eaba6b932f8535
>> compiler:       gcc (Debian 12.2.0-14) 12.2.0, GNU ld (GNU Binutils for Debian) 2.40
>>
>> Unfortunately, I don't have any reproducer for this issue yet.
>>
>> Downloadable assets:
>> disk image: https://storage.googleapis.com/syzbot-assets/565ec42c1d1a/disk-c061cf42.raw.xz
>> vmlinux: https://storage.googleapis.com/syzbot-assets/142d1c3a6f99/vmlinux-c061cf42.xz
>> kernel image: https://storage.googleapis.com/syzbot-assets/b21efab0a38b/bzImage-c061cf42.xz
>>
>> IMPORTANT: if you fix the issue, please add the following tag to the commit:
>> Reported-by: syzbot+3d7dc5eaba6b932f8535@...kaller.appspotmail.com
>>
>>    __napi_poll.constprop.0+0xb7/0x550 net/core/dev.c:6883
>>    napi_poll net/core/dev.c:6952 [inline]
>>    net_rx_action+0xa94/0x1010 net/core/dev.c:7074
>>    handle_softirqs+0x213/0x8f0 kernel/softirq.c:561
>>    __do_softirq kernel/softirq.c:595 [inline]
>>    invoke_softirq kernel/softirq.c:435 [inline]
>>    __irq_exit_rcu+0x109/0x170 kernel/softirq.c:662
>>    irq_exit_rcu+0x9/0x30 kernel/softirq.c:678
>>    common_interrupt+0xbf/0xe0 arch/x86/kernel/irq.c:278
>>    asm_common_interrupt+0x26/0x40 arch/x86/include/asm/idtentry.h:693
>> ------------[ cut here ]------------
>> kernel BUG at ./include/linux/mm.h:1221!
>> Oops: invalid opcode: 0000 [#1] PREEMPT SMP KASAN PTI
>> CPU: 1 UID: 0 PID: 11868 Comm: syz.3.1633 Tainted: G     U             6.13.0-rc3-syzkaller-00062-gc061cf420ded #0
>> Tainted: [U]=USER
>> Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 11/25/2024
>> RIP: 0010:folio_entire_mapcount include/linux/mm.h:1221 [inline]
>> RIP: 0010:folio_precise_page_mapcount fs/proc/internal.h:172 [inline]
>> RIP: 0010:kpagecount_read+0x477/0x570 fs/proc/page.c:71
>> Code: 31 ff 49 29 c4 48 8b 44 24 08 4c 01 20 e8 41 77 61 ff eb 92 e8 ca 74 61 ff 48 8b 3c 24 48 c7 c6 20 eb 61 8b e8 6a 34 a8 ff 90 <0f> 0b 4c 89 ff e8 ef de c3 ff e9 5a ff ff ff e8 a5 74 61 ff 48 8b
> 
> Kind of known (at least to me :) ).
> 
> We race with splitting the large folio and end up in
> 	VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
> 
> We could take a speculative reference on the folio to prevent the
> concurrent split ... but I am not really sure it is worth it.
> 

The following should work:

diff --git a/fs/proc/page.c b/fs/proc/page.c
index a55f5acefa974..2868248ffccf6 100644
--- a/fs/proc/page.c
+++ b/fs/proc/page.c
@@ -67,10 +67,17 @@ static ssize_t kpagecount_read(struct file *file, char __user *buf,
                  * memmaps that were actually initialized.
                  */
                 page = pfn_to_online_page(pfn);
-               if (page)
-                       mapcount = folio_precise_page_mapcount(page_folio(page),
-                                                              page);
+               if (!page)
+                       goto write_mapcount;
  
+               folio = page_folio(page);
+               if (!folio_try_get(folio))
+                       goto write_mapcount;
+               if (page_folio(page) == folio)
+                       mapcount = folio_precise_page_mapcount(folio, page);
+               folio_put(folio);
+
+write_mapcount:
                 if (put_user(mapcount, out)) {
                         ret = -EFAULT;
                         break;


In general, I dislike interfaces that can take references
on arbitrary folios. But well, at least this is only accessible to root.

-- 
Cheers,

David / dhildenb