* makecontext(3)
@ 2003-07-28 19:06 kkonaka
2003-07-29 11:39 ` makecontext(3) Luciano Miguel Ferreira Rocha
0 siblings, 1 reply; 3+ messages in thread
From: kkonaka @ 2003-07-28 19:06 UTC (permalink / raw)
To: linux-c-programming
on the surface(?) calls to makecontext() appear to have to be
preceded by getcontext() to completely initialize its arg: ucontext_t
struct (that is, if I comment out getcontext() calls from a
program like the below, it causes SEGV). why is it the case?
kenji
---
#include <ucontext.h>
#include <unistd.h>
#include <err.h>
static ucontext_t u0;
static ucontext_t u1;
static ucontext_t u2;
#define STACK_SIZE (1024 * 8)
#define NLOOP 7
static int i;
static int j;
static void f()
{
for (i = 0; i < NLOOP; i++) {
printf("f(): i = %d\n", i);
if (swapcontext(&u1, &u2) < 0)
err(1, NULL);
}
}
static void g()
{
for (j = 0; j < NLOOP; j++) {
printf("g(): j = %d\n", j);
if (swapcontext(&u2, &u0) < 0)
err(1, NULL);
}
}
main()
{
static unsigned char _u1_stack[STACK_SIZE];
static unsigned char _u2_stack[STACK_SIZE];
getcontext(&u1);
u1.uc_stack.ss_sp = _u1_stack;
u1.uc_stack.ss_size = sizeof _u1_stack;
u1.uc_stack.ss_flags = 0;
u1.uc_link = NULL;
makecontext(&u1, f, 0);
getcontext(&u2);
u2.uc_stack.ss_sp = _u2_stack;
u2.uc_stack.ss_size = sizeof _u2_stack;
u2.uc_stack.ss_flags = 0;
u2.uc_link = NULL;
makecontext(&u2, g, 0);
while (1) {
if (swapcontext(&u0, &u1) < 0)
err(1, NULL);
printf("main: %d %d\n", i, j);
if (i == NLOOP && j == NLOOP)
exit(0);
}
}
--
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: makecontext(3)
2003-07-28 19:06 makecontext(3) kkonaka
@ 2003-07-29 11:39 ` Luciano Miguel Ferreira Rocha
2003-07-30 13:49 ` makecontext(3) kkonaka
0 siblings, 1 reply; 3+ messages in thread
From: Luciano Miguel Ferreira Rocha @ 2003-07-29 11:39 UTC (permalink / raw)
To: kkonaka; +Cc: linux-c-programming
On Mon, Jul 28, 2003 at 03:06:30PM -0400, kkonaka@mac.com wrote:
> on the surface(?) calls to makecontext() appear to have to be
> preceded by getcontext() to completely initialize its arg: ucontext_t
> struct (that is, if I comment out getcontext() calls from a
> program like the below, it causes SEGV). why is it the case?
> kenji
Hello,
From the manual page:
The makecontext() function modifies the context pointed to by ucp
(which was obtained from a call to getcontext()). Before calling make-
context(), one should allocate a new stack for this context, assigning
to ucp->uc_stack, and define a successor context, assigning to
ucp->uc_link.
I assume that the getcontext() is needed to get values for registers
(cs, ds, es, gs, ...).
Regards,
Luciano Rocha
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: makecontext(3)
2003-07-29 11:39 ` makecontext(3) Luciano Miguel Ferreira Rocha
@ 2003-07-30 13:49 ` kkonaka
0 siblings, 0 replies; 3+ messages in thread
From: kkonaka @ 2003-07-30 13:49 UTC (permalink / raw)
To: luciano, linux-c-programming
howdy thanks,
> > makecontext() appear to have to be
> > preceded by getcontext() to completely initialize its arg
> I assume that the getcontext() is needed to get values for registers
> (cs, ds, es, gs, ...).
hmm., that certainly is part of the answer :) but, why makecontext()
doesn't/cannot fill up some appropriate(??) values automatically into
such fields? -- it looks to me the machine context makecontext() have
to fabricate is simply the one to start calling the given function as
one of its argument (= its second arg: "void *func()"). and it appears
to me that this new machine context doesn't have anything to do with
that of the caller. this make me think getcontext() may not be
necessary. -- ?? (well, obviously I'm missing something)
kenji
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2003-07-30 13:49 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2003-07-28 19:06 makecontext(3) kkonaka
2003-07-29 11:39 ` makecontext(3) Luciano Miguel Ferreira Rocha
2003-07-30 13:49 ` makecontext(3) kkonaka
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).