On Fri, Jan 29, 2010 at 9:54 AM, John David Anglin
<dave@xxxxxxxxxxxxxxxxxx> wrote:
> Think you missed that I said the application was being linked with
> -static. As a result, pthread_create.o won't be loaded from libpthread.a
> just to resolve __nptl_nthreads.
I linked with -static, and it was resolved to pthread_create.o from
libpthread.a. However, my application called pthread_create *before*
calling pthread_exit.
> This is the application:
>
> // This test only applies to glibc (NPTL) targets.
> // { dg-do run { target *-*-linux* } }
> // { dg-options "-pthread" }
>
> #include <pthread.h>
> #include <cxxabi.h>
> extern "C" int printf (const char *, ...);
>
> int main()
> {
> try
> {
> pthread_exit (0);
> }
> catch (abi::__forced_unwind &)
> {
> printf ("caught forced unwind\n");
> throw;
> }
> catch (...)
> {
> printf ("caught ...\n");
> return 1;
> }
> }
>
> This is the gcc link command:
>
> /home/dave/gnu/gcc/objdir/gcc/testsuite/g++/../../collect-ld -static -o ./forced.xgs /usr/lib/crt1.o /usr/lib/crti.o /home/dave/gnu/gcc/objdir/gcc/testsuite/g++/../../crtbeginT.o -L/home/dave/gnu/gcc/objdir/hppa-linux/./libstdc++-v3/src/.libs -L/home/dave/gnu/gcc/objdir/hppa-linux/./libstdc++-v3/src/.libs -L/home/dave/gnu/gcc/objdir/hppa-linux/./libiberty -L/home/dave/gnu/gcc/objdir/gcc/testsuite/g++/../.. -L/home/dave/gnu/gcc/objdir/hppa-linux/./libstdc++-v3/src/.libs /tmp/ccmLwpAe.o -lstdc++ -lm --start-group -lgcc -lgcc_eh -lpthread -lc --end-group /home/dave/gnu/gcc/objdir/gcc/testsuite/g++/../../crtend.o /usr/lib/crtn.o
>
> Compile with:
>
> g++ -o xx -static -pthread xx.C
>
> Given what you said, the same fault would likely occur on x86.
This isn't supported by glibc. You must pull in pthread_create.o if
you call pthread_exit(). Otherwise nothing defines __nptl_nthreads.
This test crashes on x86.
carlos@crasus:~$ readelf -a test | grep __nptl_nthreads
1863: 00000000 0 NOTYPE WEAK DEFAULT UND __nptl_nthreads
carlos@crasus:~$ ./test
caught forced unwind
Segmentation fault
I don't know why it crashes so late on x86, perhaps during another
operation that manipulate ptr_nthreads?
We blow up in nptl-ini.c during the static initialization that sets
ptr_nthreads to &__nptl_nthreads e.g.
~~~
static const struct pthread_functions pthread_functions =
{
.ptr_pthread_attr_destroy = __pthread_attr_destroy,
...
.ptr_nthreads = &__nptl_nthreads,
.ptr___pthread_unwind = &__pthread_unwind,
.ptr__nptl_deallocate_tsd = __nptl_deallocate_tsd,
.ptr__nptl_setxid = __nptl_setxid,
/* For now only the stack cache needs to be freed. */
.ptr_freeres = nptl_freeres
};
~~~
This is technically a generic glibc bug.
Does linking with `-u pthread_create` fix the testcase? It fixes the
failure on x86.
Cheers,
Carlos.
--
To unsubscribe from this list: send the line "unsubscribe linux-parisc" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
