Message ID | 20240904161947.821553-1-adhemerval.zanella@linaro.org |
---|---|
State | Superseded |
Headers | show |
Series | linux: Use GLRO(dl_vdso_gettimeofday) on gettimeofday | expand |
* Adhemerval Zanella: > The BZ#24967 fix (1bdda52fe92fd01b424c) missed the gettimeofday for > architectures that define USE_IFUNC_GETTIMEOFDAY. Although it is not > an issue, since there is no pointer mangling, there is also no need > to call dl_vdso_vsym since the vDSO setup was already done by the > loader. What's the impact? Is this an actually visible bug (like always performing the syscall)? Thanks, Florian
On 04/09/24 14:09, Florian Weimer wrote: > * Adhemerval Zanella: > >> The BZ#24967 fix (1bdda52fe92fd01b424c) missed the gettimeofday for >> architectures that define USE_IFUNC_GETTIMEOFDAY. Although it is not >> an issue, since there is no pointer mangling, there is also no need >> to call dl_vdso_vsym since the vDSO setup was already done by the >> loader. > > What's the impact? Is this an actually visible bug (like always > performing the syscall)? I don't think this is a bug, it will just redo the dl_vdso_vsym already done by the loader (so waste some cycles) when gettimeofday is resolved.
diff --git a/sysdeps/unix/sysv/linux/gettimeofday.c b/sysdeps/unix/sysv/linux/gettimeofday.c index 7ab147c614..94cc433202 100644 --- a/sysdeps/unix/sysv/linux/gettimeofday.c +++ b/sysdeps/unix/sysv/linux/gettimeofday.c @@ -36,12 +36,10 @@ __gettimeofday_syscall (struct timeval *restrict tv, void *restrict tz) return INLINE_SYSCALL_CALL (gettimeofday, tv, tz); } -# undef INIT_ARCH -# define INIT_ARCH() \ - void *vdso_gettimeofday = dl_vdso_vsym (HAVE_GETTIMEOFDAY_VSYSCALL) libc_ifunc (__gettimeofday, - vdso_gettimeofday ? VDSO_IFUNC_RET (vdso_gettimeofday) - : (void *) __gettimeofday_syscall) + GLRO(dl_vdso_gettimeofday) != NULL + ? VDSO_IFUNC_RET (GLRO(dl_vdso_gettimeofday)) + : (void *) __gettimeofday_syscall) # else int