Skip to contentSkip to conversation

x86-64: remove unnecessary ptregs call stubs

PATCH Applied Jan Beulich 3 participants 4 messages Jun 18, 2008 – Jun 19, 2008 x86

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 x86-64: remove unnecessary ptregs call stubs

A few of the ptregs call stubs are superfluous these days - the called functions don't have a trailing 'struct pt_regs *' parameter anymore.

---
 arch/x86/ia32/ia32entry.S   |    6 ++----
 arch/x86/kernel/entry_64.S  |    1 -
 include/asm-x86/unistd_64.h |    2 +-
 3 files changed, 3 insertions(+), 6 deletions(-)
arch/x86/ia32/ia32entry.S +2 -4
@@ -370,13 +370,11 @@ quiet_ni_syscall:
 	PTREGSCALL stub32_rt_sigreturn, sys32_rt_sigreturn, %rdi
 	PTREGSCALL stub32_sigreturn, sys32_sigreturn, %rdi
 	PTREGSCALL stub32_sigaltstack, sys32_sigaltstack, %rdx
-	PTREGSCALL stub32_sigsuspend, sys32_sigsuspend, %rcx
 	PTREGSCALL stub32_execve, sys32_execve, %rcx
 	PTREGSCALL stub32_fork, sys_fork, %rdi
 	PTREGSCALL stub32_clone, sys32_clone, %rdx
 	PTREGSCALL stub32_vfork, sys_vfork, %rdi
 	PTREGSCALL stub32_iopl, sys_iopl, %rsi
-	PTREGSCALL stub32_rt_sigsuspend, sys_rt_sigsuspend, %rdx
 
 ENTRY(ia32_ptregs_common)
 	popq %r11
@@ -476,7 +474,7 @@ ia32_sys_call_table:
 	.quad sys_ssetmask
 	.quad sys_setreuid16	/* 70 */
 	.quad sys_setregid16
-	.quad stub32_sigsuspend
+	.quad sys32_sigsuspend
 	.quad compat_sys_sigpending
 	.quad sys_sethostname
 	.quad compat_sys_setrlimit	/* 75 */
@@ -583,7 +581,7 @@ ia32_sys_call_table:
 	.quad sys32_rt_sigpending
 	.quad compat_sys_rt_sigtimedwait
 	.quad sys32_rt_sigqueueinfo
-	.quad stub32_rt_sigsuspend
+	.quad sys_rt_sigsuspend
 	.quad sys32_pread		/* 180 */
 	.quad sys32_pwrite
 	.quad sys_chown16
arch/x86/kernel/entry_64.S +0 -1
@@ -420,7 +420,6 @@ END(\label)
 	PTREGSCALL stub_clone, sys_clone, %r8
 	PTREGSCALL stub_fork, sys_fork, %rdi
 	PTREGSCALL stub_vfork, sys_vfork, %rdi
-	PTREGSCALL stub_rt_sigsuspend, sys_rt_sigsuspend, %rdx
 	PTREGSCALL stub_sigaltstack, sys_sigaltstack, %rdx
 	PTREGSCALL stub_iopl, sys_iopl, %rsi
 
include/asm-x86/unistd_64.h +1 -1
@@ -290,7 +290,7 @@ __SYSCALL(__NR_rt_sigtimedwait, sys_rt_s
 #define __NR_rt_sigqueueinfo			129
 __SYSCALL(__NR_rt_sigqueueinfo, sys_rt_sigqueueinfo)
 #define __NR_rt_sigsuspend			130
-__SYSCALL(__NR_rt_sigsuspend, stub_rt_sigsuspend)
+__SYSCALL(__NR_rt_sigsuspend, sys_rt_sigsuspend)
 #define __NR_sigaltstack			131
 __SYSCALL(__NR_sigaltstack, stub_sigaltstack)
 #define __NR_utime				132
Applied
#1 lore raw
Quoted 2 lines · Jan Beulich

A few of the ptregs call stubs are superfluous these days - the called functions don't have a trailing 'struct pt_regs *' parameter anymore.

applied to tip/x86/cleanups - thanks Jan.

Ingo

#2 lore raw
Quoted 6 lines · Ingo Molnar
Quoted 2 lines · Jan Beulich

A few of the ptregs call stubs are superfluous these days - the called functions don't have a trailing 'struct pt_regs *' parameter anymore.

applied to tip/x86/cleanups - thanks Jan.

Just want to point out that not having a pt_regs * argument is not necessarily sufficient for not needing a ptregs stub. Some code also accesses ptregs directly through the stack top from get_thread_info() and might assume a full one. That said I *think* it's ok for sigsuspend et.al. as in Jan's patch, but I haven't audited the complete code paths in question that they never do that.

Would be good to clarify that in the commit log.

-Andi

#3 lore raw
Quoted 9 lines
Quoted 1 line
Quoted 1 line

Andi Kleen <[email protected]> 19.06.08 14:50 >>>

Quoted 6 lines · Ingo Molnar
Quoted 2 lines · Jan Beulich

A few of the ptregs call stubs are superfluous these days - the called functions don't have a trailing 'struct pt_regs *' parameter anymore.

applied to tip/x86/cleanups - thanks Jan.

Quoted 7 lines

Just want to point out that not having a pt_regs * argument is not necessarily sufficient for not needing a ptregs stub. Some code also accesses ptregs directly through the stack top from get_thread_info() and might assume a full one. That said I *think* it's ok for sigsuspend et.al. as in Jan's patch, but I haven't audited the complete code paths in question that they never do that.

Would be good to clarify that in the commit log.

Looking at both the source and disassembly of sys_rt_sigsuspend() and sys32_sigsuspend(), I cannot see anything pt_regs related. All these functions call are copy_from_user(), spin_lock_irq(), recalc_sigpending(), and schedule(), and besides that they just update some global and task state.

Jan