Patch from me and "Martin J. Bligh" Adds a config option which permits the out-of-line spinlock spinning code to be moved inline. So when a kernel is profiled, the cost of the spin appears in the right function, not in .text.lock.foo. --- 25-akpm/arch/i386/Kconfig | 9 ++ 25-akpm/include/asm-i386/rwlock.h | 135 ++++++++++++++++++++++++------------ 25-akpm/include/asm-i386/spinlock.h | 39 +++++++--- 3 files changed, 128 insertions(+), 55 deletions(-) diff -puN arch/i386/Kconfig~config_spinline arch/i386/Kconfig --- 25/arch/i386/Kconfig~config_spinline Fri Mar 12 11:23:03 2004 +++ 25-akpm/arch/i386/Kconfig Fri Mar 12 11:23:03 2004 @@ -1269,6 +1269,15 @@ config DEBUG_PAGEALLOC This results in a large slowdown, but helps to find certain types of memory corruptions. +config SPINLINE + bool "Spinlock inlining" + depends on DEBUG_KERNEL + help + This will change spinlocks from out of line to inline, making them + account cost to the callers in readprofile, rather than the lock + itself (as ".text.lock.filename"). This can be helpful for finding + the callers of locks. + config DEBUG_HIGHMEM bool "Highmem debugging" depends on DEBUG_KERNEL && HIGHMEM diff -puN include/asm-i386/rwlock.h~config_spinline include/asm-i386/rwlock.h --- 25/include/asm-i386/rwlock.h~config_spinline Fri Mar 12 11:23:03 2004 +++ 25-akpm/include/asm-i386/rwlock.h Fri Mar 12 11:23:03 2004 @@ -20,28 +20,52 @@ #define RW_LOCK_BIAS 0x01000000 #define RW_LOCK_BIAS_STR "0x01000000" -#define __build_read_lock_ptr(rw, helper) \ - asm volatile(LOCK "subl $1,(%0)\n\t" \ - "js 2f\n" \ - "1:\n" \ - LOCK_SECTION_START("") \ - "2:\tcall " helper "\n\t" \ - "jmp 1b\n" \ - LOCK_SECTION_END \ - ::"a" (rw) : "memory") - -#define __build_read_lock_const(rw, helper) \ - asm volatile(LOCK "subl $1,%0\n\t" \ - "js 2f\n" \ - "1:\n" \ - LOCK_SECTION_START("") \ - "2:\tpushl %%eax\n\t" \ - "leal %0,%%eax\n\t" \ - "call " helper "\n\t" \ - "popl %%eax\n\t" \ - "jmp 1b\n" \ - LOCK_SECTION_END \ - :"=m" (*(volatile int *)rw) : : "memory") +#ifdef CONFIG_SPINLINE + + #define __build_read_lock_ptr(rw, helper) \ + asm volatile(LOCK "subl $1,(%0)\n\t" \ + "jns 1f\n\t" \ + "call " helper "\n\t" \ + "1:\t" \ + ::"a" (rw) : "memory") + + #define __build_read_lock_const(rw, helper) \ + asm volatile(LOCK "subl $1,%0\n\t" \ + "jns 1f\n\t" \ + "pushl %%eax\n\t" \ + "leal %0,%%eax\n\t" \ + "call " helper "\n\t" \ + "popl %%eax\n\t" \ + "1:\t" \ + :"=m" (*(volatile int *)rw) : : "memory") + +#else /* !CONFIG_SPINLINE */ + + #define __build_read_lock_ptr(rw, helper) \ + asm volatile(LOCK "subl $1,(%0)\n\t" \ + "js 2f\n" \ + "1:\n" \ + LOCK_SECTION_START("") \ + "2:\tcall " helper "\n\t" \ + "jmp 1b\n" \ + LOCK_SECTION_END \ + ::"a" (rw) : "memory") + + #define __build_read_lock_const(rw, helper) \ + asm volatile(LOCK "subl $1,%0\n\t" \ + "js 2f\n" \ + "1:\n" \ + LOCK_SECTION_START("") \ + "2:\tpushl %%eax\n\t" \ + "leal %0,%%eax\n\t" \ + "call " helper "\n\t" \ + "popl %%eax\n\t" \ + "jmp 1b\n" \ + LOCK_SECTION_END \ + :"=m" (*(volatile int *)rw) : : "memory") + +#endif /* CONFIG_SPINLINE */ + #define __build_read_lock(rw, helper) do { \ if (__builtin_constant_p(rw)) \ @@ -50,28 +74,51 @@ __build_read_lock_ptr(rw, helper); \ } while (0) -#define __build_write_lock_ptr(rw, helper) \ - asm volatile(LOCK "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 "subl $" RW_LOCK_BIAS_STR ",%0\n\t" \ - "jnz 2f\n" \ - "1:\n" \ - LOCK_SECTION_START("") \ - "2:\tpushl %%eax\n\t" \ - "leal %0,%%eax\n\t" \ - "call " helper "\n\t" \ - "popl %%eax\n\t" \ - "jmp 1b\n" \ - LOCK_SECTION_END \ - :"=m" (*(volatile int *)rw) : : "memory") +#ifdef CONFIG_SPINLINE + + #define __build_write_lock_ptr(rw, helper) \ + asm volatile(LOCK "subl $" RW_LOCK_BIAS_STR ",(%0)\n\t" \ + "jz 1f\n\t" \ + "call " helper "\n\t" \ + "1:\n" \ + ::"a" (rw) : "memory") + + #define __build_write_lock_const(rw, helper) \ + asm volatile(LOCK "subl $" RW_LOCK_BIAS_STR ",%0\n\t" \ + "jz 1f\n\t" \ + "pushl %%eax\n\t" \ + "leal %0,%%eax\n\t" \ + "call " helper "\n\t" \ + "popl %%eax\n\t" \ + "1:\n" \ + :"=m" (*(volatile int *)rw) : : "memory") + +#else /* !CONFIG_SPINLINE */ + + #define __build_write_lock_ptr(rw, helper) \ + asm volatile(LOCK "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 "subl $" RW_LOCK_BIAS_STR ",%0\n\t" \ + "jnz 2f\n" \ + "1:\n" \ + LOCK_SECTION_START("") \ + "2:\tpushl %%eax\n\t" \ + "leal %0,%%eax\n\t" \ + "call " helper "\n\t" \ + "popl %%eax\n\t" \ + "jmp 1b\n" \ + LOCK_SECTION_END \ + :"=m" (*(volatile int *)rw) : : "memory") + +#endif /* CONFIG_SPINLINE */ #define __build_write_lock(rw, helper) do { \ if (__builtin_constant_p(rw)) \ diff -puN include/asm-i386/spinlock.h~config_spinline include/asm-i386/spinlock.h --- 25/include/asm-i386/spinlock.h~config_spinline Fri Mar 12 11:23:03 2004 +++ 25-akpm/include/asm-i386/spinlock.h Fri Mar 12 11:23:03 2004 @@ -43,18 +43,35 @@ typedef struct { #define spin_is_locked(x) (*(volatile signed char *)(&(x)->lock) <= 0) #define spin_unlock_wait(x) do { barrier(); } while(spin_is_locked(x)) -#define spin_lock_string \ - "\n1:\t" \ - "lock ; decb %0\n\t" \ - "js 2f\n" \ - LOCK_SECTION_START("") \ - "2:\t" \ - "rep;nop\n\t" \ - "cmpb $0,%0\n\t" \ - "jle 2b\n\t" \ - "jmp 1b\n" \ - LOCK_SECTION_END +#ifdef CONFIG_SPINLINE + #define spin_lock_string \ + "\n1:\t" \ + "lock ; decb %0\n\t" \ + "js 2f\n" \ + "jmp 3f\n" \ + "2:\t" \ + "rep;nop\n\t" \ + "cmpb $0,%0\n\t" \ + "jle 2b\n\t" \ + "jmp 1b\n" \ + "3:\t" + +#else /* !CONFIG_SPINLINE */ + + #define spin_lock_string \ + "\n1:\t" \ + "lock ; decb %0\n\t" \ + "js 2f\n" \ + LOCK_SECTION_START("") \ + "2:\t" \ + "rep;nop\n\t" \ + "cmpb $0,%0\n\t" \ + "jle 2b\n\t" \ + "jmp 1b\n" \ + LOCK_SECTION_END + +#endif /* CONFIG_SPINLINE */ /* * This works. Despite all the confusion. * (except on PPro SMP or if we are using OOSTORE) _