Re: [syzbot] KMSAN: uninit-value in generic_bin_search (2)

[Dmitry Vyukov]

_data_ On Mon, Jun 14, 2021 at 8:53 PM Josef Bacik <josef@xxxxxxxxxxxxxx> wrote:
>
> On 6/14/21 1:41 AM, syzbot wrote:
> > Hello,
> >
> > syzbot found the following issue on:
> >
> > HEAD commit:    6099c9da x86: entry: speculatively unpoison pt_regs in do_..
> > git tree:       https://github.com/google/kmsan.git master
> > console output: https://syzkaller.appspot.com/x/log.txt?x=12c7a057d00000
> > kernel config:  https://syzkaller.appspot.com/x/.config?x=4e6842a91012889c
> > dashboard link: https://syzkaller.appspot.com/bug?extid=8aa9678d1cda7a0432b7
> > compiler:       Debian clang version 11.0.1-2
> > 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+8aa9678d1cda7a0432b7@xxxxxxxxxxxxxxxxxxxxxxxxx
> >
> >   tracehook_notify_resume include/linux/tracehook.h:189 [inline]
> >   exit_to_user_mode_loop kernel/entry/common.c:174 [inline]
> >   exit_to_user_mode_prepare+0x486/0x560 kernel/entry/common.c:208
> >   __syscall_exit_to_user_mode_work kernel/entry/common.c:290 [inline]
> >   syscall_exit_to_user_mode+0x35/0x60 kernel/entry/common.c:301
> >   __do_fast_syscall_32+0x14f/0x180 arch/x86/entry/common.c:145
> >   do_fast_syscall_32+0x77/0xd0 arch/x86/entry/common.c:168
> >   do_SYSENTER_32+0x73/0x90 arch/x86/entry/common.c:211
> >   entry_SYSENTER_compat_after_hwframe+0x4d/0x5c
> > =====================================================
> > =====================================================
> > BUG: KMSAN: uninit-value in btrfs_comp_cpu_keys fs/btrfs/ctree.c:1556 [inline]
> > BUG: KMSAN: uninit-value in comp_keys fs/btrfs/ctree.c:1528 [inline]
> > BUG: KMSAN: uninit-value in generic_bin_search+0x799/0xbc0 fs/btrfs/ctree.c:1702
> > CPU: 1 PID: 9364 Comm: syz-executor.2 Tainted: G    B             5.12.0-rc6-syzkaller #0
> > Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
> > Call Trace:
> >   __dump_stack lib/dump_stack.c:79 [inline]
> >   dump_stack+0x24c/0x2e0 lib/dump_stack.c:120
> >   kmsan_report+0xfb/0x1e0 mm/kmsan/kmsan_report.c:118
> >   __msan_warning+0x5c/0xa0 mm/kmsan/kmsan_instr.c:197
> >   btrfs_comp_cpu_keys fs/btrfs/ctree.c:1556 [inline]
> >   comp_keys fs/btrfs/ctree.c:1528 [inline]
> >   generic_bin_search+0x799/0xbc0 fs/btrfs/ctree.c:1702
> >   btrfs_bin_search fs/btrfs/ctree.c:1724 [inline]
>
> This appears to be a bug in KMSAN, the code is doing the correct thing and it
> appears to be complaining about tmp, which is initialized either in the if or
> else part.  The else part may be what's confusing KMSAM here as it's essentially
>
> struct btrfs_disk_key tmp;
> struct btrfs_disk_key unaligned;
>
> else {
>       memcpy(&unaligned, ptr, len); // read_extent_buffer is basically memcpy
>       tmp = &unaligned;
> }

Hi Josef,

Thanks for looking into this.

The mere fact of assigning something to a variable does not make it initialized.
Namely,

int y;
int x = y;
if (x) ...

is not much different from:

int x;
if (x) ...

Are you sure the data used in the code is in fact initialized?
KMSAN says the data originated in btrfs_alloc_tree_block. Was it
initialized? Where?