lkml.org 
[lkml]   [2021]   [Nov]   [27]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
From
Subject[PATCH] openrisc: remove wrappers for clone and fork
Date
The comment here explains that the extra saved registers are clobbered
by _switch. However, looking at switch they are definitely saved, so I
am not sure why these wrappers are needed. This was noticed when
auditing the clone3 syscall path which works fine and does not have the
extra wrapper code.

The patch removes the wrapper code as a cleanup.

Signed-off-by: Stafford Horne <shorne@gmail.com>
---
arch/openrisc/include/asm/syscalls.h | 7 ------
arch/openrisc/kernel/entry.S | 36 ++--------------------------
2 files changed, 2 insertions(+), 41 deletions(-)

diff --git a/arch/openrisc/include/asm/syscalls.h b/arch/openrisc/include/asm/syscalls.h
index 3a7eeae6f56a..c8c8a5072ad9 100644
--- a/arch/openrisc/include/asm/syscalls.h
+++ b/arch/openrisc/include/asm/syscalls.h
@@ -20,11 +20,4 @@ asmlinkage long sys_or1k_atomic(unsigned long type, unsigned long *v1,

#include <asm-generic/syscalls.h>

-asmlinkage long __sys_clone(unsigned long clone_flags, unsigned long newsp,
- void __user *parent_tid, void __user *child_tid, int tls);
-asmlinkage long __sys_fork(void);
-
-#define sys_clone __sys_clone
-#define sys_fork __sys_fork
-
#endif /* __ASM_OPENRISC_SYSCALLS_H */
diff --git a/arch/openrisc/kernel/entry.S b/arch/openrisc/kernel/entry.S
index 59c6d3aa7081..062967e09fbb 100644
--- a/arch/openrisc/kernel/entry.S
+++ b/arch/openrisc/kernel/entry.S
@@ -1139,43 +1139,11 @@ ENTRY(_switch)

/* ==================================================================== */

-/* These all use the delay slot for setting the argument register, so the
+/*
+ * This uses the delay slot for setting the argument register, so the
* jump is always happening after the l.addi instruction.
- *
- * These are all just wrappers that don't touch the link-register r9, so the
- * return from the "real" syscall function will return back to the syscall
- * code that did the l.jal that brought us here.
- */
-
-/* fork requires that we save all the callee-saved registers because they
- * are all effectively clobbered by the call to _switch. Here we store
- * all the registers that aren't touched by the syscall fast path and thus
- * weren't saved there.
*/

-_fork_save_extra_regs_and_call:
- l.sw PT_GPR14(r1),r14
- l.sw PT_GPR16(r1),r16
- l.sw PT_GPR18(r1),r18
- l.sw PT_GPR20(r1),r20
- l.sw PT_GPR22(r1),r22
- l.sw PT_GPR24(r1),r24
- l.sw PT_GPR26(r1),r26
- l.jr r29
- l.sw PT_GPR28(r1),r28
-
-ENTRY(__sys_clone)
- l.movhi r29,hi(sys_clone)
- l.ori r29,r29,lo(sys_clone)
- l.j _fork_save_extra_regs_and_call
- l.nop
-
-ENTRY(__sys_fork)
- l.movhi r29,hi(sys_fork)
- l.ori r29,r29,lo(sys_fork)
- l.j _fork_save_extra_regs_and_call
- l.nop
-
ENTRY(sys_rt_sigreturn)
l.jal _sys_rt_sigreturn
l.addi r3,r1,0
--
2.31.1
\
 
 \ /
  Last update: 2021-11-28 03:31    [W:0.035 / U:0.584 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site