Re: [PATCH -next v3 1/2] posix-timers: Check timespec64 before call clock_set()

[Jinjie Ruan <ruanjinjie@huawei.com>]

On 2024/9/12 20:04, Thomas Gleixner wrote:
> On Thu, Sep 12 2024 at 10:53, Jinjie Ruan wrote:
> 
>> On 2024/9/9 23:19, Richard Cochran wrote:
>>> On Mon, Sep 09, 2024 at 03:41:23PM +0800, Jinjie Ruan wrote:
>>>> diff --git a/kernel/time/posix-timers.c b/kernel/time/posix-timers.c
>>>> index 1cc830ef93a7..34deec619e17 100644
>>>> --- a/kernel/time/posix-timers.c
>>>> +++ b/kernel/time/posix-timers.c
>>>> @@ -1137,6 +1137,9 @@ SYSCALL_DEFINE2(clock_settime, const clockid_t, which_clock,
>>>>  	if (get_timespec64(&new_tp, tp))
>>>>  		return -EFAULT;
>>>>  
>>>> +	if (!timespec64_valid(&new_tp))
>>>> +		return -ERANGE;
>>>
>>> Why not use timespec64_valid_settod()?
>>
>> There was already checks in following code, so it is not necessary to
>> check NULL or timespec64_valid() in ptp core and its drivers, only the
>> second patch is needed.
>>
>> 169 int do_sys_settimeofday64(const struct timespec64 *tv, const struct
>> timezone *tz)
>>  170 {
>>  171 >-------static int firsttime = 1;
>>  172 >-------int error = 0;
>>  173
>>  174 >-------if (tv && !timespec64_valid_settod(tv))
>>  175 >------->-------return -EINVAL;
> 
> How does this code validate timespecs for clock_settime(clockid) where
> clockid != CLOCK_REALTIME?

According to the man manual of clock_settime(), the other clockids are
not settable.

And in Linux kernel code, except for CLOCK_REALTIME which is defined in
posix_clocks array, the clock_set() hooks are not defined and will
return -EINVAL in SYSCALL_DEFINE2(clock_settime), so the check is not
necessary.

> 
> Thanks,
> 
>         tglx