Skip to contentSkip to conversation

Hammer thread fixes

PATCH Ulrich Drepper 2 participants 3 messages Mar 15, 2003

lore
Share & tools
lore
#0 lore raw

The appended two fixes are necessary to get NPTL threads running on hammer. The changes should be obvious. The exit_group syscall isn't present at all so far and the r10 -> r8 register use is necessary because syscall parameter #4 (in r10) is already used for the child_tid parameter.

Linus, please apply.

Signature · 3 lines
--------------.                        ,-.            444 Castro Street
Ulrich Drepper \    ,-----------------'   \ Mountain View, CA 94041 USA
Red Hat         `--' drepper at redhat.com `---------------------------
arch/x86_64/kernel/process.c renamed from arch/x86_64/kernel/process.c.ud +1 -1
@@ -314,7 +314,7 @@
 			err = ia32_child_tls(p, childregs); 
 		else 			
 #endif	 
-			err = do_arch_prctl(p, ARCH_SET_FS, childregs->r10); 
+			err = do_arch_prctl(p, ARCH_SET_FS, childregs->r8); 
 		if (err) 
 			goto out;
 	}
include/asm-x86_64/unistd.h renamed from include/asm-x86_64/unistd.h.ud +3 -1
@@ -520,8 +520,10 @@
 __SYSCALL(__NR_clock_getres, sys_clock_getres)
 #define __NR_clock_nanosleep	230
 __SYSCALL(__NR_clock_nanosleep, sys_clock_nanosleep)
+#define __NR_exit_group		231
+__SYSCALL(__NR_exit_group, sys_exit_group)
 
-#define __NR_syscall_max __NR_clock_nanosleep
+#define __NR_syscall_max __NR_exit_group
 #ifndef __NO_STUBS
 
 /* user-visible error numbers are in the range -1 - -4095 */
#1 lore raw
Quoted 5 lines · Ulrich Drepper

The appended two fixes are necessary to get NPTL threads running on hammer. The changes should be obvious. The exit_group syscall isn't present at all so far and the r10 -> r8 register use is necessary because syscall parameter #4 (in r10) is already used for the child_tid parameter.

It's incorrect like I told you last time. arg 4 is in r10. Linus please don't apply.

The clone prototype is

	int clone(int flags, unsigned long newsp, void *parent_tid, void *child_tid) ;

	rax: __NR_clone
	rdi: flags
	rsi: newsp 
	rdx: parent_tid
	r10: child_tid

See appendix A of the x86-64 ABI for details. The kernel ABI
is different from the user space ABI because of the SYSCALL clobbers.

For exit_group please wait for my next merge.

-Andi

#2 lore raw

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Quoted 2 lines · Andi Kleen

It's incorrect like I told you last time. arg 4 is in r10. Linus please don't apply.

Of course is arg 4 in r10.

Quoted 3 lines

The clone prototype is

	int clone(int flags, unsigned long newsp, void *parent_tid, void *child_tid) ;

That's the prototype in process.c. From the user level it is

int clone(int flags, unsigned long newsp, void *parent_tid, void *child_tid, void *tlsaddr)

Quoted 6 lines
	rax: __NR_clone
	rdi: flags
	rsi: newsp 
	rdx: parent_tid
	r10: child_tid

You don't get it. The parameter which is passed on is the TLS address which does not appear in the parameter list but is passed in the *fifth* parameter. TLS address != child_tid! The patch is correct and regardless of what ak said, please apply it. Unlike ak's claims it is actually tested. I'm running such a kernel and threads actually work (unlike with the original kernel).

- --
- --------------. ,-. 444 Castro Street
Ulrich Drepper \ ,-----------------' \ Mountain View, CA 94041 USA
Red Hat --' drepper at redhat.com ---------------------------
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)

iD8DBQE+c6g82ijCOnn/RHQRApE+AJ9Aus8jJBg81L2A12ghG89HmfPz1wCgyldz NUazXDCnqkvJ3qWAGkNbY9U=
=u/Oe
-----END PGP SIGNATURE-----