[PATCH] INTERNAL_SYSCALL_NCS for SH

[PATCH] INTERNAL_SYSCALL_NCS for SH

[Kaz Kojima]

Hi,

The appended patch is an SH version of INTERNAL_SYSCALL_NCS.

BTW, SH port has a few problems with setXid changes.  I'll report
them in the next mail.

Regards,
	kaz
--
2004-09-23  Kaz Kojima  <kkojima@rr.iij4u.or.jp>

	* sysdeps/unix/sysv/linux/sh/sysdep.h (INTERNAL_SYSCALL_NCS): Define.

diff -u3prN ORIG/libc/sysdeps/unix/sysv/linux/sh/sysdep.h LOCAL/libc/sysdeps/unix/sysv/linux/sh/sysdep.h
--- ORIG/libc/sysdeps/unix/sysv/linux/sh/sysdep.h	Mon Jul  5 11:14:22 2004
+++ LOCAL/libc/sysdeps/unix/sysv/linux/sh/sysdep.h	Mon Sep 20 10:10:41 2004
@@ -333,6 +333,20 @@
 									      \
     (int) resultvar; })
 
+/* The _NCS variant allows non-constant syscall numbers.  */
+#define INTERNAL_SYSCALL_NCS(name, err, nr, args...) \
+  ({									      \
+    unsigned long int resultvar;					      \
+    register long int r3 asm ("%r3") = (name);			 	      \
+    SUBSTITUTE_ARGS_##nr(args);						      \
+									      \
+    asm volatile (SYSCALL_INST_STR##nr SYSCALL_INST_PAD			      \
+		  : "=z" (resultvar)					      \
+		  : "r" (r3) ASMFMT_##nr				      \
+		  : "memory");						      \
+									      \
+    (int) resultvar; })
+
 #undef INTERNAL_SYSCALL_DECL
 #define INTERNAL_SYSCALL_DECL(err) do { } while (0)
 

Re: [PATCH] INTERNAL_SYSCALL_NCS for SH

[Kaz Kojima]

> BTW, SH port has a few problems with setXid changes.  I'll report
> them in the next mail.

SH uses the old-fashioned 16-bit id's.  It causes a few problems with
setXid changes.  The first problem is that seteuid.c and setegid.c may
call __setre*[ug]id and causes undefined references.  It seems pthread
cases should use pthread_setre*[ug]id_np instead of __setre*[ug]id.

The second one is that the variable __libc_missing_32bit_uids is defined
in getuid.c but we don't have pthread_getuid.  I've added an SH specific
pthread_getuid as a workaround.  I'd like to append a patch I'm using
for comments.

Regards,
	kaz
--
[nptl/ChangeLog]
2004-09-23  Kaz Kojima  <kkojima@rr.iij4u.or.jp>

	* pthread_setegid.c (__setregid): Set to the pthread version.
	(__setresgid): Likewise.
	* pthread_seteuid.c (__setreuid, __setresuid): Likewise.
	* sysdeps/unix/sysv/linux/sh/Makefile: New file.
	* sysdeps/unix/sysv/linux/sh/pthread_getuid.c: New file.

diff -u3prN ORIG/libc/nptl/pthread_setegid.c LOCAL/libc/nptl/pthread_setegid.c
--- ORIG/libc/nptl/pthread_setegid.c	Mon Sep 20 09:05:06 2004
+++ LOCAL/libc/nptl/pthread_setegid.c	Mon Sep 20 13:23:00 2004
@@ -1,3 +1,5 @@
 #define SINGLE_THREAD
 #define setegid pthread_setegid_np
+#define __setregid pthread_setregid_np
+#define __setresgid pthread_setresgid_np
 #include <setegid.c>
diff -u3prN ORIG/libc/nptl/pthread_seteuid.c LOCAL/libc/nptl/pthread_seteuid.c
--- ORIG/libc/nptl/pthread_seteuid.c	Mon Sep 20 09:05:38 2004
+++ LOCAL/libc/nptl/pthread_seteuid.c	Thu Sep 23 17:04:43 2004
@@ -1,3 +1,5 @@
 #define SINGLE_THREAD
 #define seteuid pthread_seteuid_np
+#define __setreuid pthread_setreuid_np
+#define __setresuid pthread_setresuid_np
 #include <seteuid.c>
diff -u3prN ORIG/libc/nptl/sysdeps/unix/sysv/linux/sh/Makefile LOCAL/libc/nptl/sysdeps/unix/sysv/linux/sh/Makefile
--- ORIG/libc/nptl/sysdeps/unix/sysv/linux/sh/Makefile	Thu Jan  1 09:00:00 1970
+++ LOCAL/libc/nptl/sysdeps/unix/sysv/linux/sh/Makefile	Thu Sep 23 21:31:04 2004
@@ -0,0 +1,2 @@
+# to get __libc_missing_32bit_uids
+libpthread-routines += pthread_getuid
diff -u3prN ORIG/libc/nptl/sysdeps/unix/sysv/linux/sh/pthread_getuid.c LOCAL/libc/nptl/sysdeps/unix/sysv/linux/sh/pthread_getuid.c
--- ORIG/libc/nptl/sysdeps/unix/sysv/linux/sh/pthread_getuid.c	Thu Jan  1 09:00:00 1970
+++ LOCAL/libc/nptl/sysdeps/unix/sysv/linux/sh/pthread_getuid.c	Thu Sep 23 21:32:14 2004
@@ -0,0 +1,3 @@
+#define SINGLE_THREAD
+#define __getuid pthread_getuid_np
+#include <getuid.c>

Re: [PATCH] INTERNAL_SYSCALL_NCS for SH

[Kaz Kojima]

>> BTW, SH port has a few problems with setXid changes.  I'll report
>> them in the next mail.
> 
> SH uses the old-fashioned 16-bit id's.  It causes a few problems with
> setXid changes.  The first problem is that seteuid.c and setegid.c may
> call __setre*[ug]id and causes undefined references.  It seems pthread
> cases should use pthread_setre*[ug]id_np instead of __setre*[ug]id.
> 
> The second one is that the variable __libc_missing_32bit_uids is defined
> in getuid.c but we don't have pthread_getuid.  I've added an SH specific
> pthread_getuid as a workaround.  I'd like to append a patch I'm using
> for comments.

Ugh.  Please ignore this mail.  I didn't noticed that some setXid changes
have been removed.  I'll update my tree and try again.  Sorry for the noise.

Regards,
	kaz

Re: [PATCH] INTERNAL_SYSCALL_NCS for SH

[Kaz Kojima]

> I'll update my tree and try again.  Sorry for the noise.

I've confirmed that the INTERNAL_SYSCALL_NCS patch in

http://sources.redhat.com/ml/libc-hacker/2004-09/msg00088.html

is enough for SH port.

Regards,
	kaz

Re: [PATCH] INTERNAL_SYSCALL_NCS for SH

[Ulrich Drepper]

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

Applied.

- --
➧ Ulrich Drepper ➧ Red Hat, Inc. ➧ 444 Castro St ➧ Mountain View, CA ❖
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.6 (GNU/Linux)

iD8DBQFBU7yN2ijCOnn/RHQRApELAJ9I33+gSqP0xp69EtMIhK6zblmknQCfXVUG
gjVqIqA2aXPr+OyQBjvRsgM=
=iIrY
-----END PGP SIGNATURE-----