public inbox for linux-ia64@vger.kernel.org
 help / color / mirror / Atom feed
* [Linux-ia64] setjmp/longjmp redux
@ 2001-06-14 19:15 chas williams
  2001-06-14 20:06 ` David Mosberger
  0 siblings, 1 reply; 2+ messages in thread
From: chas williams @ 2001-06-14 19:15 UTC (permalink / raw)
  To: linux-ia64

hi, new to the list but i did a little reading of the archives and it seems
that setjmp/longjmp cant do what i need -- switch the stack.

why do i want to do this?  i am working on porting openafs which provides
its own thread library (called lwp) that is implemented with setjmp/longjmp.
what needs to be done in order to make this work?  based on my reading
it seems that ar.rnat (???) need to be saved between the context switches?
i dont know much about the ia64 architecture so if someone could give me
some pointers i would be grateful.

the following code works, but only under cc -g.  using cc -O fails.  i assume
that i am getting 'lucky' using -g.

#include <stdio.h>
#include <setjmp.h>
#include <stdlib.h>

static int rc;
static jmp_buf env;

dump_jmp_buf(jmp_buf a)
{
    int i;
    long *ptr = (long *)a;
    for ( i=0; i < 70; i++)
        if (ptr[i] != 0) printf("(%d) 0x%lx\n",i, ptr[i]);
}

main()
{
        char *altstack;
        unsigned long *patcher;
        int stacksize = 16*1024;

        altstack = malloc(192*1024);

        printf("altstack = %p\n", altstack);
        printf("&rc = %p\n", &rc);

        rc = setjmp(env);
        printf("rc = %d\n", rc);
        if (rc = 1)
        {
                printf("return from longjmp\n");
                exit(0);
        }

        ((unsigned long *)env)[0] = (unsigned long *)(altstack + stacksize - 32);

        dump_jmp_buf(env);

        printf("about to longjmp()...\n");
        longjmp(env, 1);
}



^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [Linux-ia64] setjmp/longjmp redux
  2001-06-14 19:15 [Linux-ia64] setjmp/longjmp redux chas williams
@ 2001-06-14 20:06 ` David Mosberger
  0 siblings, 0 replies; 2+ messages in thread
From: David Mosberger @ 2001-06-14 20:06 UTC (permalink / raw)
  To: linux-ia64

>>>>> On Thu, 14 Jun 2001 15:15:09 -0400, chas williams <chas@cmf.nrl.navy.mil> said:

  Chas> hi, new to the list but i did a little reading of the archives
  Chas> and it seems that setjmp/longjmp cant do what i need -- switch
  Chas> the stack.

  Chas> why do i want to do this?  i am working on porting openafs
  Chas> which provides its own thread library (called lwp) that is
  Chas> implemented with setjmp/longjmp.  what needs to be done in
  Chas> order to make this work?  based on my reading it seems that
  Chas> ar.rnat (???) need to be saved between the context switches?
  Chas> i dont know much about the ia64 architecture so if someone
  Chas> could give me some pointers i would be grateful.

For synchronous context switches ("co-routines"), use
makecontext()/setcontext() et al.  They have been added to glibc
recently.  If you're not familiar with these routines, there is an
example in the glibc source tree in stdlib/tst-setcontext.c.
Additional usage notes can be found in in answer 4.10 of the glibc
FAQ.

	--david


^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2001-06-14 20:06 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2001-06-14 19:15 [Linux-ia64] setjmp/longjmp redux chas williams
2001-06-14 20:06 ` David Mosberger

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox