Re: [LTP] [PATCH v2 2/2] Refactor getegid02 using new LTP API

[Petr Vorel <pvorel@suse.cz>]

> Hi!
> > This code is based on Cyril's suggestion in v1 [1]
> > 	if (GID_SIZE_CHECK(st_egid))
> > 		TST_EXP_EQ_LI(gid, st_egid);
> > 	else
> > 		tst_res(TPASS, "getegid() passed");

> > I wonder which system returns 16 bit gid?

> > man getgid(2) says that originally there was only 16 bit, than kernel 2.4 added
> > support for 32 bit and the glibc getegid() wrapper transparently deal with this.

> > Because even I compile as 32 bit:

> > $ file getegid01_16
> > getegid01_16: ELF 32-bit LSB pie executable, Intel 80386

> > I get comparison, thus 32 bit:
> > $ ./getegid01_16
> > ...
> > getegid01.c:25: TPASS: gid == st_egid (1000)

> > What am I missing?

> This obviously works since 1000 < 32767. As long as the GID fits 16bit
> integer everything works fine with 16bit syscalls. The check returns if
> the egid does fit and returns true if so.

> When the GID does not fit, kernel returns overflow GID, i.e. for all
> values bigger than 16 bit you get the same GID value that does not
> correspond to the actual GID value at all.

> So to make the check false, you have to set up the test user so that it
> has GID > 16bit integer maximum value.

Thanks a lot, Cyril!

Kind regards,
Petr

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp