[PATCH] Clean up read write lock assembly
- Move the slow path fallbacks to their own assembly files
This makes them much easier to read and is needed for the next change.
- Add CFI annotations for unwinding (XXX need review)
- Remove constant case which can never happen with out of line spinlocks
- Use patchable LOCK prefixes
- Don't use lock sections anymore for inline code because they can't
be expressed by the unwinder (this adds one taken jump to the lock
fast path)
Cc: jbeulich@novell.com
Signed-off-by: Andi Kleen <ak@suse.de>
diff --git a/arch/x86_64/lib/Makefile b/arch/x86_64/lib/Makefile
index ccef6ae..b78d417 100644
--- a/arch/x86_64/lib/Makefile
+++ b/arch/x86_64/lib/Makefile
@@ -9,4 +9,4 @@
lib-y := csum-partial.o csum-copy.o csum-wrappers.o delay.o \
usercopy.o getuser.o putuser.o \
thunk.o clear_page.o copy_page.o bitstr.o bitops.o
-lib-y += memcpy.o memmove.o memset.o copy_user.o
+lib-y += memcpy.o memmove.o memset.o copy_user.o rwlock.o
diff --git a/arch/x86_64/lib/rwlock.S b/arch/x86_64/lib/rwlock.S
new file mode 100644
index 0000000..0cde1f8
--- /dev/null
+++ b/arch/x86_64/lib/rwlock.S
@@ -0,0 +1,38 @@
+/* Slow paths of read/write spinlocks. */
+
+#include <linux/linkage.h>
+#include <asm/rwlock.h>
+#include <asm/alternative-asm.i>
+#include <asm/dwarf2.h>
+
+/* rdi: pointer to rwlock_t */
+ENTRY(__write_lock_failed)
+ CFI_STARTPROC
+ LOCK_PREFIX
+ addl $RW_LOCK_BIAS,(%rdi)
+1: rep
+ nop
+ cmpl $RW_LOCK_BIAS,(%rdi)
+ jne 1b
+ LOCK_PREFIX
+ subl $RW_LOCK_BIAS,(%rdi)
+ jnz __write_lock_failed
+ ret
+ CFI_ENDPROC
+END(__write_lock_failed)
+
+/* rdi: pointer to rwlock_t */
+ENTRY(__read_lock_failed)
+ CFI_STARTPROC
+ LOCK_PREFIX
+ incl (%rdi)
+1: rep
+ nop
+ cmpl $1,(%rdi)
+ js 1b
+ LOCK_PREFIX
+ decl (%rdi)
+ js __read_lock_failed
+ ret
+ CFI_ENDPROC
+END(__read_lock_failed)
diff --git a/arch/x86_64/lib/thunk.S b/arch/x86_64/lib/thunk.S
index 332ea5d..6cff27c 100644
--- a/arch/x86_64/lib/thunk.S
+++ b/arch/x86_64/lib/thunk.S
@@ -67,33 +67,3 @@
RESTORE_ARGS 1
ret
CFI_ENDPROC
-
-#ifdef CONFIG_SMP
-/* Support for read/write spinlocks. */
- .text
-/* rax: pointer to rwlock_t */
-ENTRY(__write_lock_failed)
- lock
- addl $RW_LOCK_BIAS,(%rax)
-1: rep
- nop
- cmpl $RW_LOCK_BIAS,(%rax)
- jne 1b
- lock
- subl $RW_LOCK_BIAS,(%rax)
- jnz __write_lock_failed
- ret
-
-/* rax: pointer to rwlock_t */
-ENTRY(__read_lock_failed)
- lock
- incl (%rax)
-1: rep
- nop
- cmpl $1,(%rax)
- js 1b
- lock
- decl (%rax)
- js __read_lock_failed
- ret
-#endif
diff --git a/include/asm-x86_64/rwlock.h b/include/asm-x86_64/rwlock.h
index dea0e94..28a080d 100644
--- a/include/asm-x86_64/rwlock.h
+++ b/include/asm-x86_64/rwlock.h
@@ -18,69 +18,21 @@
#ifndef _ASM_X86_64_RWLOCK_H
#define _ASM_X86_64_RWLOCK_H
-#include <linux/stringify.h>
-
#define RW_LOCK_BIAS 0x01000000
-#define RW_LOCK_BIAS_STR "0x01000000"
+#define RW_LOCK_BIAS_STR "0x01000000"
-#define __build_read_lock_ptr(rw, helper) \
+#define __build_read_lock(rw) \
asm volatile(LOCK_PREFIX "subl $1,(%0)\n\t" \
- "js 2f\n" \
+ "jns 1f\n" \
+ "call __read_lock_failed\n" \
"1:\n" \
- LOCK_SECTION_START("") \
- "2:\tcall " helper "\n\t" \
- "jmp 1b\n" \
- LOCK_SECTION_END \
- ::"a" (rw) : "memory")
+ ::"D" (rw), "i" (RW_LOCK_BIAS) : "memory")
-#define __build_read_lock_const(rw, helper) \
- asm volatile(LOCK_PREFIX "subl $1,%0\n\t" \
- "js 2f\n" \
+#define __build_write_lock(rw) \
+ asm volatile(LOCK_PREFIX "subl %1,(%0)\n\t" \
+ "jz 1f\n" \
+ "\tcall __write_lock_failed\n\t" \
"1:\n" \
- LOCK_SECTION_START("") \
- "2:\tpushq %%rax\n\t" \
- "leaq %0,%%rax\n\t" \
- "call " helper "\n\t" \
- "popq %%rax\n\t" \
- "jmp 1b\n" \
- LOCK_SECTION_END \
- :"=m" (*((volatile int *)rw))::"memory")
-
-#define __build_read_lock(rw, helper) do { \
- if (__builtin_constant_p(rw)) \
- __build_read_lock_const(rw, helper); \
- else \
- __build_read_lock_ptr(rw, helper); \
- } while (0)
-
-#define __build_write_lock_ptr(rw, helper) \
- asm volatile(LOCK_PREFIX "subl $" RW_LOCK_BIAS_STR ",(%0)\n\t" \
- "jnz 2f\n" \
- "1:\n" \
- LOCK_SECTION_START("") \
- "2:\tcall " helper "\n\t" \
- "jmp 1b\n" \
- LOCK_SECTION_END \
- ::"a" (rw) : "memory")
-
-#define __build_write_lock_const(rw, helper) \
- asm volatile(LOCK_PREFIX "subl $" RW_LOCK_BIAS_STR ",%0\n\t" \
- "jnz 2f\n" \
- "1:\n" \
- LOCK_SECTION_START("") \
- "2:\tpushq %%rax\n\t" \
- "leaq %0,%%rax\n\t" \
- "call " helper "\n\t" \
- "popq %%rax\n\t" \
- "jmp 1b\n" \
- LOCK_SECTION_END \
- :"=m" (*((volatile long *)rw))::"memory")
-
-#define __build_write_lock(rw, helper) do { \
- if (__builtin_constant_p(rw)) \
- __build_write_lock_const(rw, helper); \
- else \
- __build_write_lock_ptr(rw, helper); \
- } while (0)
+ ::"D" (rw), "i" (RW_LOCK_BIAS) : "memory")
#endif
diff --git a/include/asm-x86_64/spinlock.h b/include/asm-x86_64/spinlock.h
index 248a79f..a8e3d89 100644
--- a/include/asm-x86_64/spinlock.h
+++ b/include/asm-x86_64/spinlock.h
@@ -79,13 +79,6 @@
*
* On x86, we implement read-write locks as a 32-bit counter
* with the high bit (sign) being the "contended" bit.
- *
- * The inline assembly is non-obvious. Think about it.
- *
- * Changed to use the same technique as rw semaphores. See
- * semaphore.h for details. -ben
- *
- * the helpers are in arch/i386/kernel/semaphore.c
*/
#define __raw_read_can_lock(x) ((int)(x)->lock > 0)
@@ -93,12 +86,12 @@
static inline void __raw_read_lock(raw_rwlock_t *rw)
{
- __build_read_lock(rw, "__read_lock_failed");
+ __build_read_lock(rw);
}
static inline void __raw_write_lock(raw_rwlock_t *rw)
{
- __build_write_lock(rw, "__write_lock_failed");
+ __build_write_lock(rw);
}
static inline int __raw_read_trylock(raw_rwlock_t *lock)