Hi Thomas,

On 8/15/19 8:21 PM, Thomas Gleixner wrote:
> On Thu, 15 Aug 2019, Dmitry Safonov wrote:
>> ---
>>  arch/Kconfig                          |  5 +++
>>  arch/x86/Kconfig                      |  1 +
>>  arch/x86/entry/vdso/vdso-layout.lds.S |  9 ++++-
>>  arch/x86/entry/vdso/vdso2c.c          |  3 ++
>>  arch/x86/entry/vdso/vma.c             | 12 +++++++
>>  arch/x86/include/asm/vdso.h           |  1 +
>>  init/Kconfig                          |  1 +
>>  lib/vdso/gettimeofday.c               | 47 +++++++++++++++++++++++++++
> 
> This needs to be split into the generic lib/vdso part and then x86 making
> use of it.

Ok

>> +#ifdef CONFIG_TIME_NS
> 
> This should be COMPILE_WITH_TIME_NS and not CONFIG_TIME_NS
> 
>> +extern u8 timens_page
>> +    __attribute__((visibility("hidden")));
>> +
>> +notrace static __always_inline void clk_to_ns(clockid_t clk, struct 
>> __kernel_timespec *ts)
> 
> This needs notrace because?

Heh, well it's alive from the time it was in arch/x86.
I believe, functions there had it since commit 23adec554a76 ("x86: add
notrace annotations to vsyscall"). Probably, lib/vdso is compiled
without mcount in the Makefile somewhere.
Will drop.

[..]
>> +    /*
>> +     * The kernel allows to set a negative offset only if the current clock
>> +     * value in a namespace is positive, so the result tv_sec can't be
>> +     * negative here.
>> +     */
>> +    ts->tv_nsec += offset64->tv_nsec;
>> +    ts->tv_sec += offset64->tv_sec;
>> +    if (ts->tv_nsec >= NSEC_PER_SEC) {
>> +            ts->tv_nsec -= NSEC_PER_SEC;
>> +            ts->tv_sec++;
>> +    }
>> +    if (ts->tv_nsec < 0) {
>> +            ts->tv_nsec += NSEC_PER_SEC;
>> +            ts->tv_sec--;
>> +    }
> 
> That's broken for 32bit user space on 64bit hosts. On LE due to
> misalignment and on BE because 32bit will read always 0.

Ugh, will look into that.

Thanks,
          Dmitry

Reply via email to