On Fri, Oct 30, 2020 at 7:42 PM Dmitry Vyukov [off-list ref] wrote:

On Fri, Oct 30, 2020 at 2:02 PM Miklos Szeredi [off-list ref] wrote:

quoted

On Mon, Aug 24, 2020 at 11:00 PM Ondrej Mosnacek [off-list ref] wrote:

quoted

On Mon, Aug 24, 2020 at 9:37 PM syzbot
[off-list ref] wrote:

quoted

syzbot has found a reproducer for the following issue on:

Looping in fsdevel and OverlayFS maintainers, as this seems to be
FS/OverlayFS related...

Hmm, the oopsing code is always something like:

All code
========
   0: 1b fe                sbb    %esi,%edi
   2: 49 8d 5e 08          lea    0x8(%r14),%rbx
   6: 48 89 d8              mov    %rbx,%rax
   9: 48 c1 e8 03          shr    $0x3,%rax
   d: 42 80 3c 38 00        cmpb   $0x0,(%rax,%r15,1)
  12: 74 08                je     0x1c
  14: 48 89 df              mov    %rbx,%rdi
  17: e8 bc b4 5b fe        callq  0xfffffffffe5bb4d8
  1c: 48 8b 1b              mov    (%rbx),%rbx
  1f: 48 83 c3 68          add    $0x68,%rbx
  23: 48 89 d8              mov    %rbx,%rax
  26: 48 c1 e8 03          shr    $0x3,%rax
  2a:* 42 80 3c 38 00        cmpb   $0x0,(%rax,%r15,1) <-- trapping instruction
  2f: 74 08                je     0x39
  31: 48 89 df              mov    %rbx,%rdi
  34: e8 9f b4 5b fe        callq  0xfffffffffe5bb4d8
  39: 48 8b 1b              mov    (%rbx),%rbx
  3c: 48 83 c3 0c          add    $0xc,%rbx


And that looks (to me) like the unrolled loop in call_int_hook().  I
don't see how that could be related to overlayfs, though it's
definitely interesting why it only triggers from
overlay->vfs_getattr()->security_inode_getattr()...

quoted

  26: 48 c1 e8 03          shr    $0x3,%rax
  2a:* 42 80 3c 38 00        cmpb   $0x0,(%rax,%r15,1) <-- trapping instruction

This access is part of KASAN check. But the original address kernel
tries to access is NULL, so it's not an issue with KASAN.

The line is this:

int security_inode_getattr(const struct path *path)
{
    if (unlikely(IS_PRIVATE(d_backing_inode(path->dentry))))
        return 0;

So it's either path is NULL, or something in d_backing_inode
dereferences NULL path->dentry.

The reproducer does involve overlayfs:

mkdir(&(0x7f0000000240)='./file1\x00', 0x0)
mkdir(&(0x7f0000000300)='./bus\x00', 0x0)
r0 = creat(&(0x7f00000000c0)='./bus/file1\x00', 0x0)
mkdir(&(0x7f0000000080)='./file0\x00', 0x0)
mount$overlay(0x400002, &(0x7f0000000000)='./bus\x00',
&(0x7f0000000100)='overlay\x00', 0x0,
&(0x7f00000003c0)=ANY=[@ANYBLOB='upperdir=./file1,lowerdir=./bus,workdir=./file0,metacopy=on'])
link(&(0x7f0000000200)='./bus/file1\x00', &(0x7f00000002c0)='./bus/file0\x00')
write$RDMA_USER_CM_CMD_RESOLVE_ADDR(r0, 0x0, 0x0)
acct(&(0x7f0000000040)='./bus/file0\x00')

Though, it may be overlayfs-related, or it may be a generic bug that
requires a tricky reproducer and the only reproducer syzbot come up
with happened to involve overlayfs.
But there are 4 reproducers on syzbot dashboard and all of them
involve overlayfs and they are somewhat different. So my bet would be
on overlayfs.

Seems there's no C reproducer, though.   Can this be reproduced
without KASAN obfuscating the oops?

Thanks,
Miklos