Hi Dan,
thank you for testing my patches.
On 6/27/19 8:15 AM, Dan Carpenter wrote:
> Hello Vincenzo Frascino,
>
> This is a semi-automatic email about new static checker warnings.
>
> The patch 00b26474c2f1: "lib/vdso: Provide generic VDSO
> implementation" from Jun 21, 2019, leads to the following Smatch
> complaint:
>
> arch/x86/entry/vdso/vdso32/../../../../../lib/vdso/gettimeofday.c:120 __cvdso_clock_gettime32()
> error: we previously assumed 'res' could be null (see line 107)
>
> lib/vdso/gettimeofday.c
> 101 static __maybe_unused int
> 102 __cvdso_clock_gettime32(clockid_t clock, struct old_timespec32 *res)
> 103 {
> 104 struct __kernel_timespec ts;
> 105 int ret;
> 106
> 107 if (res == NULL)
> ^^^^^^^^^^^
> 108 goto fallback;
> 109
> 110 ret = __cvdso_clock_gettime(clock, &ts);
> 111
> 112 if (ret == 0) {
> 113 res->tv_sec = ts.tv_sec;
> 114 res->tv_nsec = ts.tv_nsec;
> 115 }
> 116
> 117 return ret;
> 118
> 119 fallback:
> 120 return clock_gettime_fallback(clock, (struct __kernel_timespec *)res);
> ^^^
> On x86 this "res" always gets dereferenced.
>
> 121 }
>
I am not sure I understand the details of this bug report. As far as I can see
"res" is never dereferenced in the vDSO library in this case, but it is passed
to the system call unchanged.
static __always_inline
long clock_gettime_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)
{
long ret;
asm ("syscall" : "=a" (ret), "=m" (*_ts) :
"0" (__NR_clock_gettime), "D" (_clkid), "S" (_ts) :
"rcx", "r11");
return ret;
}
This is done to maintain consistency in between the returned error code of the
syscall and of the vDSO library.
Could you please elaborate on why this bug has been reported?
> regards,
> dan carpenter
>
--
Regards,
Vincenzo
