[Bug nptl/22298] New: x32: lockups on recursive pthread_mutex_lock after upgrade to 2.26

["arekm at maven dot pl" <sourceware-bugzilla at sourceware dot org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=22298

            Bug ID: 22298
           Summary: x32: lockups on recursive pthread_mutex_lock after
                    upgrade to 2.26
           Product: glibc
           Version: 2.26
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: nptl
          Assignee: unassigned at sourceware dot org
          Reporter: arekm at maven dot pl
                CC: drepper.fsp at gmail dot com
  Target Milestone: ---

Program:

#define _GNU_SOURCE 1

#include <stdio.h>
#include <pthread.h>

int main() {

        static pthread_mutex_t rpmsigTbl_lock =
PTHREAD_RECURSIVE_MUTEX_INITIALIZER_NP;
        printf("LOCK1\n");
        pthread_mutex_lock(&rpmsigTbl_lock);
        printf("LOCK2\n");
        pthread_mutex_lock(&rpmsigTbl_lock);
        printf("LOCK3\n");
}

locks when built on Linux x32 glibc 2.25 but run on glibc 2.26
(and vice versa, locks when built on glibc 2.26 but run on 2.25)

$ ./a.out
LOCK1
LOCK2

It was built with "gcc -fopenmp test.c" and -fopenmp seems to be a important
factor. If not using -fopenmp then the problem is gone.

The problem doesn't happen on i686 and x86_64.

From libc-alpha, Andreas Schwab:

"The problem is that sysdeps/nptl/bits/thread-shared-types.h in 2.26 uses
#if __WORDSIZE == 64 where sysdeps/x86/bits/pthreadtypes.h in 2.25 uses
#ifdef __x86_64__.  This isn't the same on x32."

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