Hello,
We found the following issue using syzkaller on Linux v6.2.0.
In the function `dvb_frontend_get_event`, function
`wait_event_interruptible` is called
and the condition is `dvb_frontend_test_event(fepriv, events)`.
In the function `dvb_frontend_test_event`, function
`down(&fepriv->sem);` is called.
However, function `wait_event_interruptible` would put the process to sleep.
And function `down(&fepriv->sem);` may block the process.
So there is the issue with "do not call blocking ops when !TASK_RUNNING".
The full report including the Syzkaller reproducer & C reproducer:
https://gist.github.com/ZHYfeng/4c5f8be6adc63b73dba68230d15ece2c
The brief report is below:
Syzkaller hit 'WARNING in dvb_frontend_get_event' bug.
------------[ cut here ]------------
do not call blocking ops when !TASK_RUNNING; state=1 set at
[<ffffffff8161186d>] prepare_to_wait_event+0x6d/0x690
kernel/sched/wait.c:333
WARNING: CPU: 0 PID: 8017 at kernel/sched/core.c:9968
__might_sleep+0x10a/0x160 kernel/sched/core.c:9968
Modules linked in:
CPU: 0 PID: 8017 Comm: syz-executor303 Not tainted 6.2.0 #1
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS
1.13.0-1ubuntu1.1 04/01/2014
RIP: 0010:__might_sleep+0x10a/0x160 kernel/sched/core.c:9968
Code: 9d 03 00 48 8d bb d8 16 00 00 48 89 fa 48 c1 ea 03 80 3c 02 00
75 34 48 8b 93 d8 16 00 00 48 c7 c7 e0 68 4c 8a e8 38 55 72 08 <0f> 0b
e9 75 ff ff ff e8 1a 7b 7f 00 e9 26 ff ff ff 89 34 24 e8 1d
RSP: 0018:ffffc9000e537ac8 EFLAGS: 00010282
RAX: 0000000000000000 RBX: ffff888018bdba80 RCX: 0000000000000000
RDX: 0000000000000000 RSI: ffff888018bdba80 RDI: fffff52001ca6f4b
RBP: ffffffff8a4cd200 R08: 0000000000000000 R09: ffffed1005944f32
R10: ffff88802ca2798b R11: ffffed1005944f31 R12: 000000000000003a
R13: 0000000000000000 R14: 0000000000000000 R15: ffff888044057260
FS: 0000555555995880(0000) GS:ffff88802ca00000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007fd34db66000 CR3: 000000001f479000 CR4: 0000000000350ef0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
<TASK>
down+0x1e/0xa0 kernel/locking/semaphore.c:58
dvb_frontend_test_event drivers/media/dvb-core/dvb_frontend.c:277 [inline]
dvb_frontend_get_event.isra.0+0x528/0x670
drivers/media/dvb-core/dvb_frontend.c:301
dvb_frontend_handle_ioctl+0x1953/0x2ea0
drivers/media/dvb-core/dvb_frontend.c:2726
dvb_frontend_do_ioctl+0x1c5/0x2f0 drivers/media/dvb-core/dvb_frontend.c:2097
dvb_usercopy+0xbe/0x280 drivers/media/dvb-core/dvbdev.c:961
dvb_frontend_ioctl+0x5a/0x80 drivers/media/dvb-core/dvb_frontend.c:2111
vfs_ioctl fs/ioctl.c:51 [inline]
__do_sys_ioctl fs/ioctl.c:870 [inline]
__se_sys_ioctl fs/ioctl.c:856 [inline]
__x64_sys_ioctl+0x198/0x210 fs/ioctl.c:856
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x63/0xcd
RIP: 0033:0x7f569e9f4a7d
Code: 28 c3 e8 36 29 00 00 66 0f 1f 44 00 00 f3 0f 1e fa 48 89 f8 48
89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d
01 f0 ff ff 73 01 c3 48 c7 c1 c0 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007fff77694948 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007f569e9f4a7d
RDX: 0000000020000000 RSI: 0000000080286f4e RDI: 0000000000000003
RBP: 00007f569e9ae440 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 00007f569e9ae4e0
R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000
</TASK>
On Tue, 18 Apr 2023 06:50:07 +0200,
Yu Hao wrote:
>
> Hello,
>
> We found the following issue using syzkaller on Linux v6.2.0.
>
> In the function `dvb_frontend_get_event`, function
> `wait_event_interruptible` is called
> and the condition is `dvb_frontend_test_event(fepriv, events)`.
> In the function `dvb_frontend_test_event`, function
> `down(&fepriv->sem);` is called.
> However, function `wait_event_interruptible` would put the process to sleep.
> And function `down(&fepriv->sem);` may block the process.
> So there is the issue with "do not call blocking ops when !TASK_RUNNING".
>
> The full report including the Syzkaller reproducer & C reproducer:
> https://gist.github.com/ZHYfeng/4c5f8be6adc63b73dba68230d15ece2c
FYI, CVE-2023-31084 was assigned to this bug, and I was involved now
though distro's bug report.
So, the use of semaphore together with wait_event*() macro doesn't
look like a good idea. A possible easy workaround would be to
open-code the wait loop like below.
Mauro, let me know if it's an acceptable workaround. Then I'll submit
a proper patch.
thanks,
Takashi
-- 8< --
--- a/drivers/media/dvb-core/dvb_frontend.c
+++ b/drivers/media/dvb-core/dvb_frontend.c
@@ -293,14 +293,22 @@ static int dvb_frontend_get_event(struct dvb_frontend *fe,
}
if (events->eventw == events->eventr) {
- int ret;
+ struct wait_queue_entry wait;
+ int ret = 0;
if (flags & O_NONBLOCK)
return -EWOULDBLOCK;
- ret = wait_event_interruptible(events->wait_queue,
- dvb_frontend_test_event(fepriv, events));
-
+ init_waitqueue_entry(&wait, current);
+ add_wait_queue(&events->wait_queue, &wait);
+ while (!dvb_frontend_test_event(fepriv, events)) {
+ wait_woken(&wait, TASK_INTERRUPTIBLE, 0);
+ if (signal_pending(current)) {
+ ret = -ERESTARTSYS;
+ break;
+ }
+ }
+ remove_wait_queue(&events->wait_queue, &wait);
if (ret < 0)
return ret;
}
On Thu, 27 Apr 2023 09:58:24 +0200,
Takashi Iwai wrote:
>
> On Tue, 18 Apr 2023 06:50:07 +0200,
> Yu Hao wrote:
> >
> > Hello,
> >
> > We found the following issue using syzkaller on Linux v6.2.0.
> >
> > In the function `dvb_frontend_get_event`, function
> > `wait_event_interruptible` is called
> > and the condition is `dvb_frontend_test_event(fepriv, events)`.
> > In the function `dvb_frontend_test_event`, function
> > `down(&fepriv->sem);` is called.
> > However, function `wait_event_interruptible` would put the process to sleep.
> > And function `down(&fepriv->sem);` may block the process.
> > So there is the issue with "do not call blocking ops when !TASK_RUNNING".
> >
> > The full report including the Syzkaller reproducer & C reproducer:
> > https://gist.github.com/ZHYfeng/4c5f8be6adc63b73dba68230d15ece2c
>
> FYI, CVE-2023-31084 was assigned to this bug, and I was involved now
> though distro's bug report.
>
> So, the use of semaphore together with wait_event*() macro doesn't
> look like a good idea. A possible easy workaround would be to
> open-code the wait loop like below.
>
> Mauro, let me know if it's an acceptable workaround. Then I'll submit
> a proper patch.
A gentle ping.
Can anyone please check whether the suggested change is OK or not?
thanks,
Takashi
>
>
> thanks,
>
> Takashi
>
> -- 8< --
> --- a/drivers/media/dvb-core/dvb_frontend.c
> +++ b/drivers/media/dvb-core/dvb_frontend.c
> @@ -293,14 +293,22 @@ static int dvb_frontend_get_event(struct dvb_frontend *fe,
> }
>
> if (events->eventw == events->eventr) {
> - int ret;
> + struct wait_queue_entry wait;
> + int ret = 0;
>
> if (flags & O_NONBLOCK)
> return -EWOULDBLOCK;
>
> - ret = wait_event_interruptible(events->wait_queue,
> - dvb_frontend_test_event(fepriv, events));
> -
> + init_waitqueue_entry(&wait, current);
> + add_wait_queue(&events->wait_queue, &wait);
> + while (!dvb_frontend_test_event(fepriv, events)) {
> + wait_woken(&wait, TASK_INTERRUPTIBLE, 0);
> + if (signal_pending(current)) {
> + ret = -ERESTARTSYS;
> + break;
> + }
> + }
> + remove_wait_queue(&events->wait_queue, &wait);
> if (ret < 0)
> return ret;
> }