[PATCH v6] Implement -e trace=%statfs option for tracing statfs like syscalls

Abhishek Tiwari erabhishektiwarics at gmail.com
Wed Mar 22 02:23:34 UTC 2017


Thank you for reply :)

On Wed, Mar 22, 2017 at 6:54 AM, Eugene Syromyatnikov <evgsyr at gmail.com> wrote:
>> None of stat related syscalls have their corresponding flag set.
>> $  grep -nr '_\(\|f\|l\)stat' linux/*/syscallent*
>>
>> Should these be updated with their corresponding tags ?
>> fstatfs*-> TD
>> lstat*-> TD
>> {,old_}stat*-> TF
>> or are they intensional?
> Most probably these compatibility calls has been overlooked. As they are
> has the same semantics, they indeed should be a part of appropriate
> syscall groups.
>

What is svr4_PYRAMID_statis syscall related to ? what flag should be
assigned to it?


>> Sir,
>> the statfs64.test file calls the statfs.test itself.
>> On adding 17 statfs64 to my test file gives error
>>
>> undefined: __NR_statfs64
> This is quite possible on 64-bit architectures, as they have no need in
> a separate syscall for retrieving some 64-bit values (like st_size and
> st_blocks).
>
>> Is adding following entries correct ?
> Not really.
>
>> In file tests/scno.h:
>>
>> #ifndef __NR_statfs
>> # define __NR_statfs64 (SYSCALL_BIT | 332 )
> Note that this file is generated (as it is stated in its header), as it
> is ABI-dependent.
>
>> Is adding it to last entry fine ?
>>
>> In file tests/ksysent.h:
>>
>> #ifdef __NR_statfs
>> [__NR_statfs64 & 0xffff] = "statfs64",
>>

Sir, then how to overcome error  to add test for statfs64 ?




More information about the Strace-devel mailing list