[PATCH] sparc64: exit_group should kill register windows just like plain exit.

[PATCH] sparc64: exit_group should kill register windows just like plain exit.

[David Miller]

Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
 arch/sparc/kernel/sys32.S      |  2 +-
 arch/sparc/kernel/syscalls.S   | 14 ++++++++++++--
 arch/sparc/kernel/systbls_64.S |  2 +-
 3 files changed, 14 insertions(+), 4 deletions(-)

diff --git a/arch/sparc/kernel/sys32.S b/arch/sparc/kernel/sys32.S
index 44025f4..8475a47 100644
--- a/arch/sparc/kernel/sys32.S
+++ b/arch/sparc/kernel/sys32.S
@@ -47,7 +47,7 @@ STUB:	sra	REG1, 0, REG1; \
 	sra	REG4, 0, REG4
 
 SIGN1(sys32_exit, sparc_exit, %o0)
-SIGN1(sys32_exit_group, sys_exit_group, %o0)
+SIGN1(sys32_exit_group, sparc_exit_group, %o0)
 SIGN1(sys32_wait4, compat_sys_wait4, %o2)
 SIGN1(sys32_creat, sys_creat, %o1)
 SIGN1(sys32_mknod, sys_mknod, %o1)
diff --git a/arch/sparc/kernel/syscalls.S b/arch/sparc/kernel/syscalls.S
index 7f5f65d..bf23477 100644
--- a/arch/sparc/kernel/syscalls.S
+++ b/arch/sparc/kernel/syscalls.S
@@ -118,10 +118,20 @@ ret_from_syscall:
 	ba,pt	%xcc, ret_sys_call
 	 ldx	[%sp + PTREGS_OFF + PT_V9_I0], %o0
 
+	.globl	sparc_exit_group
+	.type	sparc_exit_group,#function
+sparc_exit_group:
+	sethi	%hi(sys_exit_group), %g7
+	ba,pt	%xcc, 1f
+	 or	%g7, %lo(sys_exit_group), %g7
+	.size	sparc_exit_group,.-sparc_exit_group
+
 	.globl	sparc_exit
 	.type	sparc_exit,#function
 sparc_exit:
-	rdpr	%pstate, %g2
+	sethi	%hi(sys_exit), %g7
+	or	%g7, %lo(sys_exit), %g7
+1:	rdpr	%pstate, %g2
 	wrpr	%g2, PSTATE_IE, %pstate
 	rdpr	%otherwin, %g1
 	rdpr	%cansave, %g3
@@ -129,7 +139,7 @@ sparc_exit:
 	wrpr	%g3, 0x0, %cansave
 	wrpr	%g0, 0x0, %otherwin
 	wrpr	%g2, 0x0, %pstate
-	ba,pt	%xcc, sys_exit
+	jmpl	%g7, %g0
 	 stb	%g0, [%g6 + TI_WSAVED]
 	.size	sparc_exit,.-sparc_exit
 
diff --git a/arch/sparc/kernel/systbls_64.S b/arch/sparc/kernel/systbls_64.S
index 1c9af9f..017b74a 100644
--- a/arch/sparc/kernel/systbls_64.S
+++ b/arch/sparc/kernel/systbls_64.S
@@ -133,7 +133,7 @@ sys_call_table:
 /*170*/	.word sys_lsetxattr, sys_fsetxattr, sys_getxattr, sys_lgetxattr, sys_getdents
 	.word sys_setsid, sys_fchdir, sys_fgetxattr, sys_listxattr, sys_llistxattr
 /*180*/	.word sys_flistxattr, sys_removexattr, sys_lremovexattr, sys_nis_syscall, sys_ni_syscall
-	.word sys_setpgid, sys_fremovexattr, sys_tkill, sys_exit_group, sys_newuname
+	.word sys_setpgid, sys_fremovexattr, sys_tkill, sparc_exit_group, sys_newuname
 /*190*/	.word sys_init_module, sys_sparc64_personality, sys_remap_file_pages, sys_epoll_create, sys_epoll_ctl
 	.word sys_epoll_wait, sys_ioprio_set, sys_getppid, sys_nis_syscall, sys_sgetmask
 /*200*/	.word sys_ssetmask, sys_nis_syscall, sys_newlstat, sys_uselib, sys_nis_syscall
-- 
1.7.12.2.dirty

Re: [PATCH] sparc64: exit_group should kill register windows just like plain exit.

[Al Viro]

On Mon, Dec 03, 2012 at 02:30:49PM -0500, David Miller wrote:
> 
> Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
> Signed-off-by: David S. Miller <davem@davemloft.net>
> ---
>  arch/sparc/kernel/sys32.S      |  2 +-
>  arch/sparc/kernel/syscalls.S   | 14 ++++++++++++--
>  arch/sparc/kernel/systbls_64.S |  2 +-
>  3 files changed, 14 insertions(+), 4 deletions(-)
> 
> diff --git a/arch/sparc/kernel/sys32.S b/arch/sparc/kernel/sys32.S
> index 44025f4..8475a47 100644
> --- a/arch/sparc/kernel/sys32.S
> +++ b/arch/sparc/kernel/sys32.S
> @@ -47,7 +47,7 @@ STUB:	sra	REG1, 0, REG1; \
>  	sra	REG4, 0, REG4
>  
>  SIGN1(sys32_exit, sparc_exit, %o0)
> -SIGN1(sys32_exit_group, sys_exit_group, %o0)
> +SIGN1(sys32_exit_group, sparc_exit_group, %o0)

No need of that part, really.  sys_exit_group() will have its argument
(int) normalized all by itself.  The same goes for sys_exit(), so
sys32_exit is also pointless; just have sparc_exit/sparc_exit_group in
32bit syscall table, same as you do in 64bit one...

Re: [PATCH] sparc64: exit_group should kill register windows just like plain exit.

[David Miller]

From: Al Viro <viro@ZenIV.linux.org.uk>
Date: Mon, 3 Dec 2012 19:35:58 +0000

> On Mon, Dec 03, 2012 at 02:30:49PM -0500, David Miller wrote:
>> 
>> Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
>> Signed-off-by: David S. Miller <davem@davemloft.net>
>> ---
>>  arch/sparc/kernel/sys32.S      |  2 +-
>>  arch/sparc/kernel/syscalls.S   | 14 ++++++++++++--
>>  arch/sparc/kernel/systbls_64.S |  2 +-
>>  3 files changed, 14 insertions(+), 4 deletions(-)
>> 
>> diff --git a/arch/sparc/kernel/sys32.S b/arch/sparc/kernel/sys32.S
>> index 44025f4..8475a47 100644
>> --- a/arch/sparc/kernel/sys32.S
>> +++ b/arch/sparc/kernel/sys32.S
>> @@ -47,7 +47,7 @@ STUB:	sra	REG1, 0, REG1; \
>>  	sra	REG4, 0, REG4
>>  
>>  SIGN1(sys32_exit, sparc_exit, %o0)
>> -SIGN1(sys32_exit_group, sys_exit_group, %o0)
>> +SIGN1(sys32_exit_group, sparc_exit_group, %o0)
> 
> No need of that part, really.  sys_exit_group() will have its argument
> (int) normalized all by itself.  The same goes for sys_exit(), so
> sys32_exit is also pointless; just have sparc_exit/sparc_exit_group in
> 32bit syscall table, same as you do in 64bit one...

As we've discussed in other thread, much of sys32.S is irrelevant.  I'll
try to find some time to audit the whole file and take care of all of
the extraneous cases in the sparc-next tree.

Thanks.