Skip to contentSkip to conversation

SPARC and SA_SIGINFO signal handling

DISCUSSION Christophe Rhodes 3 participants 8 messages Oct 29, 2001 – Nov 7, 2001

lore
Share & tools
lore
#0 lore raw

Dear all,

I'm having trouble on SPARC/Linux (of both the 32 and 64 varieties) getting at the third argument of POSIX sa_sigaction signal handlers.

Consider the following code (a simplified version of what I'm actually trying to do):
--- Cut here ---
#include <stdlib.h>
#include <sys/ucontext.h>
#include <signal.h>

void sigsegv_handler (int signo, siginfo_t *info, void *data) {
return;
}

int main () {
int *foo;
struct sigaction sa;

  sa.sa_sigaction = sigsegv_handler;
  sa.sa_flags = SA_SIGINFO | SA_RESTART;
  sigaction(SIGSEGV, &sa, NULL);

foo = NULL;
*foo = 3;
return 0;
}
--- Cut here ---
Running under gdb reveals that, whereas on alpha, x86 and ppc the third argument to sigsegv_handler is a pointer to a ucontext structure, on sparc and sparc64 I get NULL.

Am I doing something wrong, or is the ucontext argument simply not yet implemented on the sparc architecture?

Many thanks,

Christophe

Signature · 4 lines
Jesus College, Cambridge, CB5 8BL                           +44 1223 510 299
http://www-jcsu.jesus.cam.ac.uk/~csr21/                  (defun pling-dollar 
(str schar arg) (first (last +))) (make-dispatch-macro-character #\! t)
(set-dispatch-macro-character #\! #\$ #'pling-dollar)
#1 lore raw

You're doing something really wrong, it works perfectly fine here:

? cat test.c
#include <stdlib.h>
#include <sys/ucontext.h>
#include <signal.h>

void sigsegv_handler (int signo, siginfo_t *info, void *data) {
	if (info != 0)
		exit(1);
	exit(0);
}

int main () {
int *foo;
struct sigaction sa;

  sa.sa_sigaction = sigsegv_handler;
  sa.sa_flags = SA_SIGINFO | SA_RESTART;
  sigaction(SIGSEGV, &sa, NULL);

foo = NULL;
*foo = 3;
return 0;
}
? gcc -o test test.c
? ./test
? echo $?
1
? uname -a
Linux pizda.ninka.net 2.4.14-pre4 #1 SMP Mon Oct 29 18:55:18 PST 2001 sparc64 unknown ?

#2 lore raw
In reply to Christophe Rhodes
Quoted 11 lines · David S. Miller

You're doing something really wrong, it works perfectly fine here:

? cat test.c
#include <stdlib.h>
#include <sys/ucontext.h>
#include <signal.h>

void sigsegv_handler (int signo, siginfo_t *info, void *data) {
	if (info != 0)
        ^^^^

This gets me the siginfo struct; this is fine, and I'm happy with this part.

However, what I don't see to get at is the usercontext/ucontext structure containing register contents and so on, which as far as I am aware should be in the third (data) argument to the sa_sigaction-type sighandler; that's where I'm getting my problems.

Quoted 4 lines
		exit(1);
	exit(0);
}

[...]

Change the info above to data, and...

[ x86 does what I expect... ]

csr21@lambda:~$ uname -a
Linux lambda 2.4.13-ac4 #1 Mon Oct 29 18:26:51 GMT 2001 i686 unknown
csr21@lambda:~$ ./foo
csr21@lambda:~$ echo $?
1

[ sparc doesn't ]

csr21@caligula:~$ uname -a
Linux caligula 2.4.6 #1 SMP Sun Sep 30 16:40:07 BST 2001 sparc64
unknown
csr21@caligula:~$ ./foo
csr21@caligula:~$ echo $?
0

Thanks,

Christophe

Signature · 4 lines
Jesus College, Cambridge, CB5 8BL                           +44 1223 510 299
http://www-jcsu.jesus.cam.ac.uk/~csr21/                  (defun pling-dollar 
(str schar arg) (first (last +))) (make-dispatch-macro-character #\! t)
(set-dispatch-macro-character #\! #\$ #'pling-dollar)
#3 lore raw

From: Christophe Rhodes <[email protected]>
Date: Wed, 31 Oct 2001 09:43:43 +0000

However, what I don't see to get at is the usercontext/ucontext structure containing register contents and so on, which as far as I am aware should be in the third (data) argument to the sa_sigaction-type sighandler; that's where I'm getting my problems.

The "register contents and so on" are in the sigcontext. We don't use ucontext on sparc32.

Franks a lot,
David S. Miller
[email protected]

#5 lore raw
In reply to Christophe Rhodes
Quoted 11 lines · David S. Miller

From: Richard Henderson <[email protected]>
Date: Sat, 3 Nov 2001 11:59:00 -0800

Quoted 2 lines · David S. Miller

The "register contents and so on" are in the sigcontext. We don't use ucontext on sparc32.

In other words, you don't support SA_SIGINFO at all.

Is it required? All the information that thing provides is determinable via other methods.

Sorry for the late response (I've been away); "required" is perhaps putting it strongly, because as you say the information is there, somewhere. However, it would be nice to have, as standardized, that the third argument to the sa_sigaction handler be castable to a ucontext_t, as this would make porting signal-handling code between Linux flavours much easier.

Cheers,

Christophe

Signature · 4 lines
Jesus College, Cambridge, CB5 8BL                           +44 1223 510 299
http://www-jcsu.jesus.cam.ac.uk/~csr21/                  (defun pling-dollar 
(str schar arg) (first (last +))) (make-dispatch-macro-character #\! t)
(set-dispatch-macro-character #\! #\$ #'pling-dollar)
#7 lore raw

From: Richard Henderson <[email protected]>
Date: Sat, 3 Nov 2001 11:59:00 -0800

Quoted 2 lines · David S. Miller

The "register contents and so on" are in the sigcontext. We don't use ucontext on sparc32.

In other words, you don't support SA_SIGINFO at all.

Is it required? All the information that thing provides is determinable via other methods.

Franks a lot,
David S. Miller
[email protected]