When I ran Syzkaller testsuite, I got the following call trace.
================================================================================
UBSAN: Undefined behaviour in ./include/linux/time64.h:120:27
signed integer overflow:
8243129037239968815 * 1000000000 cannot be represented in type 'long long int'
CPU: 5 PID: 28854 Comm: syz-executor.1 Not tainted 4.19.24 #4
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.11.0-0-g63451fca13-prebuilt.qemu-project.org 04/01/2014
Call Trace:
__dump_stack lib/dump_stack.c:77 [inline]
dump_stack+0xca/0x13e lib/dump_stack.c:113
ubsan_epilogue+0xe/0x81 lib/ubsan.c:159
handle_overflow+0x193/0x1e2 lib/ubsan.c:190
timespec64_to_ns include/linux/time64.h:120 [inline]
posix_cpu_timer_set+0x95a/0xb70 kernel/time/posix-cpu-timers.c:687
do_timer_settime+0x198/0x2a0 kernel/time/posix-timers.c:892
__do_sys_timer_settime kernel/time/posix-timers.c:918 [inline]
__se_sys_timer_settime kernel/time/posix-timers.c:904 [inline]
__x64_sys_timer_settime+0x18d/0x260 kernel/time/posix-timers.c:904
do_syscall_64+0xc8/0x580 arch/x86/entry/common.c:290
entry_SYSCALL_64_after_hwframe+0x49/0xbe
RIP: 0033:0x462eb9
Code: f7 d8 64 89 02 b8 ff ff ff ff c3 66 0f 1f 44 00 00 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 bc ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007f14e4127c58 EFLAGS: 00000246 ORIG_RAX: 00000000000000df
RAX: ffffffffffffffda RBX: 000000000073bfa0 RCX: 0000000000462eb9
RDX: 0000000020000080 RSI: 0000000000000000 RDI: 0000000000000000
RBP: 0000000000000004 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 00007f14e41286bc
R13: 00000000004c54cc R14: 0000000000704278 R15: 00000000ffffffff
================================================================================
It is because 'it_interval.tv_sec' is larger than 'KTIME_SEC_MAX' and
'it_interval.tv_sec * NSEC_PER_SEC' overflows in 'timespec64_to_ns()'.
This patch checks whether 'it_interval.tv_sec' is larger than
'KTIME_SEC_MAX' and saturate if that is the case.
Signed-off-by: Xiongfeng Wang <[email protected]>
---
v3:
- add limit check and saturate
v2:
- use 'timespec64_valid_restrict()' to check the input and return
-EINVAL if check fails.
---
kernel/time/posix-timers.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/kernel/time/posix-timers.c b/kernel/time/posix-timers.c
index 0e84bb7..4b57566 100644
--- a/kernel/time/posix-timers.c
+++ b/kernel/time/posix-timers.c
@@ -856,6 +856,10 @@ static int do_timer_settime(timer_t timer_id, int flags,
if (!timespec64_valid(&new_spec64->it_interval) ||
!timespec64_valid(&new_spec64->it_value))
return -EINVAL;
+ if (new_spec64->it_interval.tv_sec > KTIME_SEC_MAX)
+ new_spec64->it_interval.tv_sec = KTIME_SEC_MAX;
+ if (new_spec64->it_value.tv_sec > KTIME_SEC_MAX)
+ new_spec64->it_value.tv_sec = KTIME_SEC_MAX;
if (old_spec64)
memset(old_spec64, 0, sizeof(*old_spec64));
--
1.7.12.4
On Wed, Mar 6, 2019 at 9:41 AM Xiongfeng Wang <[email protected]> wrote:
>
>
> diff --git a/kernel/time/posix-timers.c b/kernel/time/posix-timers.c
> index 0e84bb7..4b57566 100644
> --- a/kernel/time/posix-timers.c
> +++ b/kernel/time/posix-timers.c
> @@ -856,6 +856,10 @@ static int do_timer_settime(timer_t timer_id, int flags,
> if (!timespec64_valid(&new_spec64->it_interval) ||
> !timespec64_valid(&new_spec64->it_value))
> return -EINVAL;
> + if (new_spec64->it_interval.tv_sec > KTIME_SEC_MAX)
> + new_spec64->it_interval.tv_sec = KTIME_SEC_MAX;
> + if (new_spec64->it_value.tv_sec > KTIME_SEC_MAX)
> + new_spec64->it_value.tv_sec = KTIME_SEC_MAX;
>
I looked at the calculation we do later, and I think this can still overflow
if tv_nsec is too large. The largest timespec value we can support is
(struct timespec64) { .tv_sec = 9223372036, .tv_nsec = 854775807 }
Your patch caps the tv_sec value to 9223372036, but it does not
cap the tv_nsec. The easiest fix would be to always set tv_nsec
to 0 if tv_sec>=9223372036, or a more correct calculation would
have to limit tv_nsec if tv_sec==9223372036. I don't know if that
matters or not (it should not, unless we explicitly compare the
ktime_t for equality with KTIME_MAX later).
Arnd
On 2019/3/6 18:38, Arnd Bergmann wrote:
> On Wed, Mar 6, 2019 at 9:41 AM Xiongfeng Wang <[email protected]> wrote:
>>
>>
>> diff --git a/kernel/time/posix-timers.c b/kernel/time/posix-timers.c
>> index 0e84bb7..4b57566 100644
>> --- a/kernel/time/posix-timers.c
>> +++ b/kernel/time/posix-timers.c
>> @@ -856,6 +856,10 @@ static int do_timer_settime(timer_t timer_id, int flags,
>> if (!timespec64_valid(&new_spec64->it_interval) ||
>> !timespec64_valid(&new_spec64->it_value))
>> return -EINVAL;
>> + if (new_spec64->it_interval.tv_sec > KTIME_SEC_MAX)
>> + new_spec64->it_interval.tv_sec = KTIME_SEC_MAX;
>> + if (new_spec64->it_value.tv_sec > KTIME_SEC_MAX)
>> + new_spec64->it_value.tv_sec = KTIME_SEC_MAX;
>>
>
> I looked at the calculation we do later, and I think this can still overflow
> if tv_nsec is too large. The largest timespec value we can support is
>
> (struct timespec64) { .tv_sec = 9223372036, .tv_nsec = 854775807 }
>
> Your patch caps the tv_sec value to 9223372036, but it does not
> cap the tv_nsec. The easiest fix would be to always set tv_nsec
> to 0 if tv_sec>=9223372036, or a more correct calculation would
> have to limit tv_nsec if tv_sec==9223372036. I don't know if that
> matters or not (it should not, unless we explicitly compare the
> ktime_t for equality with KTIME_MAX later).
Thanks for your advice. I will send another one to set tv_nsec
to 0 if tv_sec>=9223372036.
Do need to add a helper to clamp timespec64, such as 'saturate_timespec64_valid()'
to limit it between 0 and KTIME_MAX(or KTIME_SEC_MAX, it's easier.)
Thanks,
Xiongfeng
>
> Arnd
>
> .
>