[syzbot] KCSAN: data-race in call_rcu / rcu_gp_fqs_loop

[syzbot]

Hello,

syzbot found the following issue on:

HEAD commit: 9c0c4d24ac00 Merge tag 'block-5.15-2021-10-22' of git://gi..
git tree: upstream
console output: https://syzkaller.appspot.com/x/log.txt?x=159c4954b00000
kernel config: https://syzkaller.appspot.com/x/.config?x=6339b6ea86d89fd7
dashboard link: https://syzkaller.appspot.com/bug?extid=4dfb96a94317a78f44d9
compiler: Debian clang version 11.0.1-2, GNU ld (GNU Binutils for Debian) 2.35.2

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+4dfb96...@syzkaller.appspotmail.com

==================================================================
BUG: KCSAN: data-race in call_rcu / rcu_gp_fqs_loop

write to 0xffffffff837342e0 of 8 bytes by task 11 on cpu 1:
rcu_gp_fqs kernel/rcu/tree.c:1910 [inline]
rcu_gp_fqs_loop+0x348/0x470 kernel/rcu/tree.c:1971
rcu_gp_kthread+0x25/0x1a0 kernel/rcu/tree.c:2130
kthread+0x262/0x280 kernel/kthread.c:319
ret_from_fork+0x1f/0x30

read to 0xffffffff837342e0 of 8 bytes by task 379 on cpu 0:
__call_rcu_core kernel/rcu/tree.c:2904 [inline]
__call_rcu kernel/rcu/tree.c:3020 [inline]
call_rcu+0x4c0/0x6d0 kernel/rcu/tree.c:3067
__dentry_kill+0x3ec/0x4e0 fs/dcache.c:596
dput+0xc6/0x360 fs/dcache.c:888
do_unlinkat+0x2a8/0x540 fs/namei.c:4172
__do_sys_unlink fs/namei.c:4217 [inline]
__se_sys_unlink fs/namei.c:4215 [inline]
__x64_sys_unlink+0x2c/0x30 fs/namei.c:4215
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x44/0xa0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x44/0xae

value changed: 0x0000000000005c0d -> 0x0000000000005c0e

Reported by Kernel Concurrency Sanitizer on:
CPU: 0 PID: 379 Comm: udevd Tainted: G W 5.15.0-rc6-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
==================================================================


---
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.

[Marco Elver]

+Cc Paul

data race is in rcu code, presumably not yet discovered by rcutorture?

[Paul E. McKenney]

On Mon, Oct 25, 2021 at 12:31:53PM +0200, Marco Elver wrote:
> +Cc Paul
>
> data race is in rcu code, presumably not yet discovered by rcutorture?

Quite possibly, and I will take a look. Thank you for sending this
along.

Thanx, Paul

[Paul E. McKenney]

On Tue, Oct 26, 2021 at 07:07:30AM -0700, Paul E. McKenney wrote:
> On Mon, Oct 25, 2021 at 12:31:53PM +0200, Marco Elver wrote:
> > +Cc Paul
> >
> > data race is in rcu code, presumably not yet discovered by rcutorture?
>
> Quite possibly, and I will take a look. Thank you for sending this
> along.

And this is (allegedly) fixed by commit 2431774f04d10 ("rcu: Mark accesses
to rcu_state.n_force_qs"), which is in -rcu and slated for the upcoming
merge window. But yes, still a bug in mainline.

[Marco Elver]

On Tue, 26 Oct 2021 at 16:33, Paul E. McKenney <pau...@kernel.org> wrote:
> On Tue, Oct 26, 2021 at 07:07:30AM -0700, Paul E. McKenney wrote:
> > On Mon, Oct 25, 2021 at 12:31:53PM +0200, Marco Elver wrote:
> > > +Cc Paul
> > >
> > > data race is in rcu code, presumably not yet discovered by rcutorture?
> >
> > Quite possibly, and I will take a look. Thank you for sending this
> > along.
>
> And this is (allegedly) fixed by commit 2431774f04d10 ("rcu: Mark accesses
> to rcu_state.n_force_qs"), which is in -rcu and slated for the upcoming
> merge window. But yes, still a bug in mainline.

Thanks for confirming, then the latest incarnation of it must then be
a dup, and I'll mark it fixed:

#syz fix: rcu: Mark accesses to rcu_state.n_force_qs

Apologies I missed there was a previous report and your fix.

Thanks,
-- Marco