This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] __readlink_chk: Assume HAVE_INLINED_SYSCALLS



On 20/08/2018 13:18, Adhemerval Zanella wrote:
> 
> 
> On 20/08/2018 11:40, Florian Weimer wrote:
>> HAVE_INLINED_SYSCALLS is always defined on Linux.
>>
>> 2018-08-20  Florian Weimer  <fweimer@redhat.com>
>>
>> 	* sysdeps/unix/sysv/linux/generic/readlink_chk.c: Remove
>> 	HAVE_INLINED_SYSCALLS conditionals
> 
> LGTM.

In fact, do we really need to have a duplicated Linux implementation
for this micro-optimization? Couldn't we just use default debug
one and call '__readlink' instead?

> 
>>
>> diff --git a/sysdeps/unix/sysv/linux/generic/readlink_chk.c b/sysdeps/unix/sysv/linux/generic/readlink_chk.c
>> index 9240408a6b..5075e06dd1 100644
>> --- a/sysdeps/unix/sysv/linux/generic/readlink_chk.c
>> +++ b/sysdeps/unix/sysv/linux/generic/readlink_chk.c
>> @@ -19,10 +19,8 @@
>>  #include <unistd.h>
>>  #include <fcntl.h>
>>  #include <sys/param.h>
>> -#ifdef HAVE_INLINED_SYSCALLS
>>  # include <errno.h>
>>  # include <sysdep.h>
>> -#endif
>>  
>>  
>>  ssize_t
>> @@ -31,9 +29,5 @@ __readlink_chk (const char *path, void *buf, size_t len, size_t buflen)
>>    if (len > buflen)
>>      __chk_fail ();
>>  
>> -#ifdef HAVE_INLINED_SYSCALLS
>>    return INLINE_SYSCALL (readlinkat, 4, AT_FDCWD, path, buf, len);
> 
> Maybe use INLINE_SYSCALL_CALL ?
> 
>> -#else
>> -  return __readlink (path, buf, len);
>> -#endif
>>  }
>>


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]