[Bug librt/20489] timer_create in compatibility mode on a 64bit big-endian system segfaults

["fweimer at redhat dot com" <sourceware-bugzilla at sourceware dot org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=20489

Florian Weimer <fweimer at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |fweimer at redhat dot com
              Flags|                            |security-

--- Comment #4 from Florian Weimer <fweimer at redhat dot com> ---
(In reply to ian.merin from comment #3)
> (In reply to joseph@codesourcery.com from comment #2)
> > On Thu, 18 Aug 2016, ian.merin at thalesesec dot com wrote:
> > 
> > > #if SHLIB_COMPAT (librt, GLIBC_2_2, GLIBC_2_3_3)
> > 
> > > timer_create is redirected to
> > > int __timer_create_old (clockid_t clock_id, struct sigevent *evp,
> > >                               int *timerid)
> > > 
> > > however, the signature for timer_create is:
> > > 
> > > 
> > > int timer_create(clockid_t clockid, struct sigevent *sevp,
> > >                         timer_t *timerid);
> > > 
> > > On this system, timer_t is defined as a 64bit type, while int is defined as a
> > > 32 bit type.
> > 
> > And in glibc versions in the relevant interval, timer_t was int.  So this 
> > redirection to a function taking int * is correct; callers of this 
> > function were compiled against glibc versions at least 2.2 and less than 
> > 2.3.3, and will be passing a pointer to int.
> > 
> > If there's a problem here, you'll need to explain it further.
> 
> I see.  It would seem that the problem is not what I thought it was. timer_t
> is defined as a void * on the system, and version 2.19 of glibc is being
> used, yet we are still entering the version of timer_create meant for a much
> older version of glibc.  
> 
> Thanks, I'll have to investigate further.

-- 
You are receiving this mail because:
You are on the CC list for the bug.