From b7882b7c65abb00194bdb3d4a22d27d70fcc59ba Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Fri, 3 Jul 2009 13:26:39 +0200 Subject: x86: atomic64: Move the 32-bit atomic64_t implementation to a .c file Linus noted that the atomic64_t primitives are all inlines currently which is crazy because these functions have a large register footprint anyway. Move them to a separate file: arch/x86/lib/atomic64_32.c Also, while at it, rename all uses of 'unsigned long long' to the much shorter u64. This makes the appearance of the prototypes a lot nicer - and it also uncovered a few bugs where (yet unused) API variants had 'long' as their return type instead of u64. [ More intrusive changes are not yet done in this patch. ] Reported-by: Linus Torvalds Cc: Eric Dumazet Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: David Howells Cc: Andrew Morton Cc: Arnd Bergmann LKML-Reference: Signed-off-by: Ingo Molnar --- arch/x86/lib/Makefile | 1 + arch/x86/lib/atomic64_32.c | 216 +++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 217 insertions(+) create mode 100644 arch/x86/lib/atomic64_32.c (limited to 'arch/x86/lib') diff --git a/arch/x86/lib/Makefile b/arch/x86/lib/Makefile index f9d35632666..c3c657c8bb8 100644 --- a/arch/x86/lib/Makefile +++ b/arch/x86/lib/Makefile @@ -10,6 +10,7 @@ lib-y += usercopy_$(BITS).o getuser.o putuser.o lib-y += memcpy_$(BITS).o ifeq ($(CONFIG_X86_32),y) + lib-y += atomic64_32.o lib-y += checksum_32.o lib-y += strstr_32.o lib-y += semaphore_32.o string_32.o diff --git a/arch/x86/lib/atomic64_32.c b/arch/x86/lib/atomic64_32.c new file mode 100644 index 00000000000..d21e725d3d8 --- /dev/null +++ b/arch/x86/lib/atomic64_32.c @@ -0,0 +1,216 @@ +#include +#include +#include +#include +#include + +static inline u64 cmpxchg8b(u64 *ptr, u64 old, u64 new) +{ + asm volatile( + + LOCK_PREFIX "cmpxchg8b (%[ptr])\n" + + : "=A" (old) + + : [ptr] "D" (ptr), + "A" (old), + "b" (ll_low(new)), + "c" (ll_high(new)) + + : "memory"); + + return old; +} + +u64 atomic64_cmpxchg(atomic64_t *ptr, u64 old_val, u64 new_val) +{ + return cmpxchg8b(&ptr->counter, old_val, new_val); +} + +/** + * atomic64_xchg - xchg atomic64 variable + * @ptr: pointer to type atomic64_t + * @new_val: value to assign + * + * Atomically xchgs the value of @ptr to @new_val and returns + * the old value. + */ + +u64 atomic64_xchg(atomic64_t *ptr, u64 new_val) +{ + u64 old_val; + + do { + old_val = atomic_read(ptr); + } while (atomic64_cmpxchg(ptr, old_val, new_val) != old_val); + + return old_val; +} + +/** + * atomic64_set - set atomic64 variable + * @ptr: pointer to type atomic64_t + * @new_val: value to assign + * + * Atomically sets the value of @ptr to @new_val. + */ +void atomic64_set(atomic64_t *ptr, u64 new_val) +{ + atomic64_xchg(ptr, new_val); +} + +/** + * atomic64_read - read atomic64 variable + * @ptr: pointer to type atomic64_t + * + * Atomically reads the value of @ptr and returns it. + */ +u64 atomic64_read(atomic64_t *ptr) +{ + u64 curr_val; + + do { + curr_val = __atomic64_read(ptr); + } while (atomic64_cmpxchg(ptr, curr_val, curr_val) != curr_val); + + return curr_val; +} + +/** + * atomic64_add_return - add and return + * @delta: integer value to add + * @ptr: pointer to type atomic64_t + * + * Atomically adds @delta to @ptr and returns @delta + *@ptr + */ +u64 atomic64_add_return(u64 delta, atomic64_t *ptr) +{ + u64 old_val, new_val; + + do { + old_val = atomic_read(ptr); + new_val = old_val + delta; + + } while (atomic64_cmpxchg(ptr, old_val, new_val) != old_val); + + return new_val; +} + +u64 atomic64_sub_return(u64 delta, atomic64_t *ptr) +{ + return atomic64_add_return(-delta, ptr); +} + +u64 atomic64_inc_return(atomic64_t *ptr) +{ + return atomic64_add_return(1, ptr); +} + +u64 atomic64_dec_return(atomic64_t *ptr) +{ + return atomic64_sub_return(1, ptr); +} + +/** + * atomic64_add - add integer to atomic64 variable + * @delta: integer value to add + * @ptr: pointer to type atomic64_t + * + * Atomically adds @delta to @ptr. + */ +void atomic64_add(u64 delta, atomic64_t *ptr) +{ + atomic64_add_return(delta, ptr); +} + +/** + * atomic64_sub - subtract the atomic64 variable + * @delta: integer value to subtract + * @ptr: pointer to type atomic64_t + * + * Atomically subtracts @delta from @ptr. + */ +void atomic64_sub(u64 delta, atomic64_t *ptr) +{ + atomic64_add(-delta, ptr); +} + +/** + * atomic64_sub_and_test - subtract value from variable and test result + * @delta: integer value to subtract + * @ptr: pointer to type atomic64_t + * + * Atomically subtracts @delta from @ptr and returns + * true if the result is zero, or false for all + * other cases. + */ +int atomic64_sub_and_test(u64 delta, atomic64_t *ptr) +{ + u64 old_val = atomic64_sub_return(delta, ptr); + + return old_val == 0; +} + +/** + * atomic64_inc - increment atomic64 variable + * @ptr: pointer to type atomic64_t + * + * Atomically increments @ptr by 1. + */ +void atomic64_inc(atomic64_t *ptr) +{ + atomic64_add(1, ptr); +} + +/** + * atomic64_dec - decrement atomic64 variable + * @ptr: pointer to type atomic64_t + * + * Atomically decrements @ptr by 1. + */ +void atomic64_dec(atomic64_t *ptr) +{ + atomic64_sub(1, ptr); +} + +/** + * atomic64_dec_and_test - decrement and test + * @ptr: pointer to type atomic64_t + * + * Atomically decrements @ptr by 1 and + * returns true if the result is 0, or false for all other + * cases. + */ +int atomic64_dec_and_test(atomic64_t *ptr) +{ + return atomic64_sub_and_test(1, ptr); +} + +/** + * atomic64_inc_and_test - increment and test + * @ptr: pointer to type atomic64_t + * + * Atomically increments @ptr by 1 + * and returns true if the result is zero, or false for all + * other cases. + */ +int atomic64_inc_and_test(atomic64_t *ptr) +{ + return atomic64_sub_and_test(-1, ptr); +} + +/** + * atomic64_add_negative - add and test if negative + * @delta: integer value to add + * @ptr: pointer to type atomic64_t + * + * Atomically adds @delta to @ptr and returns true + * if the result is negative, or false when + * result is greater than or equal to zero. + */ +int atomic64_add_negative(u64 delta, atomic64_t *ptr) +{ + long long old_val = atomic64_add_return(delta, ptr); + + return old_val < 0; +} -- cgit v1.2.3 From aacf682fd8c66b57383c407eecd9d4a28264ee91 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Fri, 3 Jul 2009 12:14:27 +0200 Subject: x86: atomic64: Improve atomic64_read() Linus noticed that the 32-bit version of atomic64_read() was being overly complex with re-reading the value and doing a retry loop over that. Instead we can just rely on cmpxchg8b returning either the new value or returning the current value. We can use any 'old' value, which will be faster as it can be loaded via immediates. Using some value that is not equal to the real value in memory the instruction gets faster. This also has the advantage that the CPU could avoid dirtying the cacheline. Reported-by: Linus Torvalds Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: David Howells Cc: Andrew Morton Cc: Arnd Bergmann LKML-Reference: Signed-off-by: Ingo Molnar --- arch/x86/lib/atomic64_32.c | 8 ++------ 1 file changed, 2 insertions(+), 6 deletions(-) (limited to 'arch/x86/lib') diff --git a/arch/x86/lib/atomic64_32.c b/arch/x86/lib/atomic64_32.c index d21e725d3d8..afa5d444918 100644 --- a/arch/x86/lib/atomic64_32.c +++ b/arch/x86/lib/atomic64_32.c @@ -67,13 +67,9 @@ void atomic64_set(atomic64_t *ptr, u64 new_val) */ u64 atomic64_read(atomic64_t *ptr) { - u64 curr_val; + u64 old = 1LL << 32; - do { - curr_val = __atomic64_read(ptr); - } while (atomic64_cmpxchg(ptr, curr_val, curr_val) != curr_val); - - return curr_val; + return cmpxchg8b(&ptr->counter, old, old); } /** -- cgit v1.2.3 From 69237f94e65d3d7f539f1adb98ef68685c595004 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Fri, 3 Jul 2009 13:26:41 +0200 Subject: x86: atomic64: Improve cmpxchg8b() Rewrite cmpxchg8b() to not use %edi register but a generic "+m" constraint, to increase compiler freedom in code generation and possibly better code. Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: David Howells Cc: Andrew Morton Cc: Arnd Bergmann LKML-Reference: Signed-off-by: Ingo Molnar --- arch/x86/lib/atomic64_32.c | 19 +++++++------------ 1 file changed, 7 insertions(+), 12 deletions(-) (limited to 'arch/x86/lib') diff --git a/arch/x86/lib/atomic64_32.c b/arch/x86/lib/atomic64_32.c index afa5d444918..5fc1e2caa54 100644 --- a/arch/x86/lib/atomic64_32.c +++ b/arch/x86/lib/atomic64_32.c @@ -6,19 +6,14 @@ static inline u64 cmpxchg8b(u64 *ptr, u64 old, u64 new) { - asm volatile( - - LOCK_PREFIX "cmpxchg8b (%[ptr])\n" - - : "=A" (old) - - : [ptr] "D" (ptr), - "A" (old), - "b" (ll_low(new)), - "c" (ll_high(new)) - - : "memory"); + u32 low = new; + u32 high = new >> 32; + asm volatile( + LOCK_PREFIX "cmpxchg8b %1\n" + : "+A" (old), "+m" (*ptr) + : "b" (low), "c" (high) + ); return old; } -- cgit v1.2.3 From 824975ef190e7dcb77718d1cc2cb53769b16d918 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Fri, 3 Jul 2009 12:39:07 +0200 Subject: x86: atomic64: Improve atomic64_add_return() Linus noted (based on Eric Dumazet's numbers) that we would probably be better off not trying an atomic_read() in atomic64_add_return() but intead intentionally let the first cmpxchg8b fail - to get a cache-friendly 'give me ownership of this cacheline' transaction. That can then be followed by the real cmpxchg8b which sets the value local to the CPU. Reported-by: Linus Torvalds Cc: Eric Dumazet Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: David Howells Cc: Andrew Morton Cc: Arnd Bergmann LKML-Reference: Signed-off-by: Ingo Molnar --- arch/x86/lib/atomic64_32.c | 15 ++++++++++++--- 1 file changed, 12 insertions(+), 3 deletions(-) (limited to 'arch/x86/lib') diff --git a/arch/x86/lib/atomic64_32.c b/arch/x86/lib/atomic64_32.c index 5fc1e2caa54..61959627e1e 100644 --- a/arch/x86/lib/atomic64_32.c +++ b/arch/x86/lib/atomic64_32.c @@ -76,13 +76,22 @@ u64 atomic64_read(atomic64_t *ptr) */ u64 atomic64_add_return(u64 delta, atomic64_t *ptr) { - u64 old_val, new_val; + /* + * Try first with a (probably incorrect) assumption about + * what we have there. We'll do two loops most likely, + * but we'll get an ownership MESI transaction straight away + * instead of a read transaction followed by a + * flush-for-ownership transaction: + */ + u64 old_val, new_val, real_val = 1ULL << 32; do { - old_val = atomic_read(ptr); + old_val = real_val; new_val = old_val + delta; - } while (atomic64_cmpxchg(ptr, old_val, new_val) != old_val); + real_val = atomic64_cmpxchg(ptr, old_val, new_val); + + } while (real_val != old_val); return new_val; } -- cgit v1.2.3 From 3ac805d2afd3fa4a07cb5bcf352fd7fa83f28935 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Fri, 3 Jul 2009 12:51:19 +0200 Subject: x86: atomic64: Reduce size of functions cmpxchg8b is a huge instruction in terms of register footprint, we almost never want to inline it, not even within the same code module. GCC 4.3 still messes up for two functions, under-judging the true cost of this instruction - so annotate two key functions to reduce the bloat: arch/x86/lib/atomic64_32.o: text data bss dec hex filename 1763 0 0 1763 6e3 atomic64_32.o.before 435 0 0 435 1b3 atomic64_32.o.after Cc: Linus Torvalds Cc: Eric Dumazet Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: David Howells Cc: Andrew Morton Cc: Arnd Bergmann LKML-Reference: Signed-off-by: Ingo Molnar --- arch/x86/lib/atomic64_32.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'arch/x86/lib') diff --git a/arch/x86/lib/atomic64_32.c b/arch/x86/lib/atomic64_32.c index 61959627e1e..a910238a776 100644 --- a/arch/x86/lib/atomic64_32.c +++ b/arch/x86/lib/atomic64_32.c @@ -4,7 +4,7 @@ #include #include -static inline u64 cmpxchg8b(u64 *ptr, u64 old, u64 new) +static noinline u64 cmpxchg8b(u64 *ptr, u64 old, u64 new) { u32 low = new; u32 high = new >> 32; @@ -74,7 +74,7 @@ u64 atomic64_read(atomic64_t *ptr) * * Atomically adds @delta to @ptr and returns @delta + *@ptr */ -u64 atomic64_add_return(u64 delta, atomic64_t *ptr) +noinline u64 atomic64_add_return(u64 delta, atomic64_t *ptr) { /* * Try first with a (probably incorrect) assumption about -- cgit v1.2.3 From 199e23780a7e75c63a9e3d1108804e3af450ea3e Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Fri, 3 Jul 2009 13:02:39 +0200 Subject: x86: atomic64: Fix unclean type use in atomic64_xchg() Linus noticed that atomic64_xchg() uses atomic_read(), which happens to work because atomic_read() is a macro so the .counter value gets u64-read on 32-bit too - but this is really bogus and serious bugs are waiting to happen. Fix atomic64_xchg() to use __atomic64_read() instead. No code changed: arch/x86/lib/atomic64_32.o: text data bss dec hex filename 435 0 0 435 1b3 atomic64_32.o.before 435 0 0 435 1b3 atomic64_32.o.after md5: bd8ab95e69c93518578bfaf0ea3be4d9 atomic64_32.o.before.asm bd8ab95e69c93518578bfaf0ea3be4d9 atomic64_32.o.after.asm Reported-by: Linus Torvalds Cc: Eric Dumazet Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: David Howells Cc: Andrew Morton Cc: Arnd Bergmann LKML-Reference: Signed-off-by: Ingo Molnar --- arch/x86/lib/atomic64_32.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/x86/lib') diff --git a/arch/x86/lib/atomic64_32.c b/arch/x86/lib/atomic64_32.c index a910238a776..fd28fd3fb74 100644 --- a/arch/x86/lib/atomic64_32.c +++ b/arch/x86/lib/atomic64_32.c @@ -36,7 +36,7 @@ u64 atomic64_xchg(atomic64_t *ptr, u64 new_val) u64 old_val; do { - old_val = atomic_read(ptr); + old_val = __atomic64_read(ptr); } while (atomic64_cmpxchg(ptr, old_val, new_val) != old_val); return old_val; -- cgit v1.2.3 From 67d7178f8fc64b7f68d7dd8a1b21dfa0d42c220c Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Fri, 3 Jul 2009 13:23:02 +0200 Subject: x86: atomic64: Improve atomic64_read() Optimize atomic64_read() as a special open-coded cmpxchg8b variant. This generates nicer code: arch/x86/lib/atomic64_32.o: text data bss dec hex filename 435 0 0 435 1b3 atomic64_32.o.before 431 0 0 431 1af atomic64_32.o.after md5: bd8ab95e69c93518578bfaf0ea3be4d9 atomic64_32.o.before.asm 2bdfd4bd1f6b7b61b7fc127aef90ce3b atomic64_32.o.after.asm Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: David Howells Cc: Andrew Morton Cc: Arnd Bergmann LKML-Reference: Signed-off-by: Ingo Molnar --- arch/x86/lib/atomic64_32.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-) (limited to 'arch/x86/lib') diff --git a/arch/x86/lib/atomic64_32.c b/arch/x86/lib/atomic64_32.c index fd28fd3fb74..cd11803f944 100644 --- a/arch/x86/lib/atomic64_32.c +++ b/arch/x86/lib/atomic64_32.c @@ -62,9 +62,17 @@ void atomic64_set(atomic64_t *ptr, u64 new_val) */ u64 atomic64_read(atomic64_t *ptr) { - u64 old = 1LL << 32; + u64 res; - return cmpxchg8b(&ptr->counter, old, old); + asm volatile( + "mov %%ebx, %%eax\n\t" + "mov %%ecx, %%edx\n\t" + LOCK_PREFIX "cmpxchg8b %1\n" + : "+A" (res) + : "m" (*ptr) + ); + + return res; } /** -- cgit v1.2.3 From 1fde902d52ee13ab9fab155bbae757fdf7daf0c1 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Fri, 3 Jul 2009 17:28:57 +0200 Subject: x86: atomic64: Export APIs to modules atomic64_t primitives are used by a handful of drivers, so export the APIs consistently. These were inlined before. Also mark atomic64_32.o a core object, so that the symbols are available even if not linked to core kernel pieces. Cc: Eric Dumazet Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: David Howells Cc: Andrew Morton Cc: Arnd Bergmann LKML-Reference: Signed-off-by: Ingo Molnar --- arch/x86/lib/Makefile | 2 +- arch/x86/lib/atomic64_32.c | 18 ++++++++++++++++++ 2 files changed, 19 insertions(+), 1 deletion(-) (limited to 'arch/x86/lib') diff --git a/arch/x86/lib/Makefile b/arch/x86/lib/Makefile index c3c657c8bb8..07c31899c9c 100644 --- a/arch/x86/lib/Makefile +++ b/arch/x86/lib/Makefile @@ -10,7 +10,7 @@ lib-y += usercopy_$(BITS).o getuser.o putuser.o lib-y += memcpy_$(BITS).o ifeq ($(CONFIG_X86_32),y) - lib-y += atomic64_32.o + obj-y += atomic64_32.o lib-y += checksum_32.o lib-y += strstr_32.o lib-y += semaphore_32.o string_32.o diff --git a/arch/x86/lib/atomic64_32.c b/arch/x86/lib/atomic64_32.c index cd11803f944..6722a092e40 100644 --- a/arch/x86/lib/atomic64_32.c +++ b/arch/x86/lib/atomic64_32.c @@ -1,5 +1,7 @@ #include +#include #include + #include #include #include @@ -21,6 +23,7 @@ u64 atomic64_cmpxchg(atomic64_t *ptr, u64 old_val, u64 new_val) { return cmpxchg8b(&ptr->counter, old_val, new_val); } +EXPORT_SYMBOL(atomic64_cmpxchg); /** * atomic64_xchg - xchg atomic64 variable @@ -41,6 +44,7 @@ u64 atomic64_xchg(atomic64_t *ptr, u64 new_val) return old_val; } +EXPORT_SYMBOL(atomic64_xchg); /** * atomic64_set - set atomic64 variable @@ -53,6 +57,7 @@ void atomic64_set(atomic64_t *ptr, u64 new_val) { atomic64_xchg(ptr, new_val); } +EXPORT_SYMBOL(atomic64_read); /** * atomic64_read - read atomic64 variable @@ -74,6 +79,7 @@ u64 atomic64_read(atomic64_t *ptr) return res; } +EXPORT_SYMBOL(atomic64_read); /** * atomic64_add_return - add and return @@ -103,21 +109,25 @@ noinline u64 atomic64_add_return(u64 delta, atomic64_t *ptr) return new_val; } +EXPORT_SYMBOL(atomic64_add_return); u64 atomic64_sub_return(u64 delta, atomic64_t *ptr) { return atomic64_add_return(-delta, ptr); } +EXPORT_SYMBOL(atomic64_sub_return); u64 atomic64_inc_return(atomic64_t *ptr) { return atomic64_add_return(1, ptr); } +EXPORT_SYMBOL(atomic64_inc_return); u64 atomic64_dec_return(atomic64_t *ptr) { return atomic64_sub_return(1, ptr); } +EXPORT_SYMBOL(atomic64_dec_return); /** * atomic64_add - add integer to atomic64 variable @@ -130,6 +140,7 @@ void atomic64_add(u64 delta, atomic64_t *ptr) { atomic64_add_return(delta, ptr); } +EXPORT_SYMBOL(atomic64_add); /** * atomic64_sub - subtract the atomic64 variable @@ -142,6 +153,7 @@ void atomic64_sub(u64 delta, atomic64_t *ptr) { atomic64_add(-delta, ptr); } +EXPORT_SYMBOL(atomic64_sub); /** * atomic64_sub_and_test - subtract value from variable and test result @@ -158,6 +170,7 @@ int atomic64_sub_and_test(u64 delta, atomic64_t *ptr) return old_val == 0; } +EXPORT_SYMBOL(atomic64_sub_and_test); /** * atomic64_inc - increment atomic64 variable @@ -169,6 +182,7 @@ void atomic64_inc(atomic64_t *ptr) { atomic64_add(1, ptr); } +EXPORT_SYMBOL(atomic64_inc); /** * atomic64_dec - decrement atomic64 variable @@ -180,6 +194,7 @@ void atomic64_dec(atomic64_t *ptr) { atomic64_sub(1, ptr); } +EXPORT_SYMBOL(atomic64_dec); /** * atomic64_dec_and_test - decrement and test @@ -193,6 +208,7 @@ int atomic64_dec_and_test(atomic64_t *ptr) { return atomic64_sub_and_test(1, ptr); } +EXPORT_SYMBOL(atomic64_dec_and_test); /** * atomic64_inc_and_test - increment and test @@ -206,6 +222,7 @@ int atomic64_inc_and_test(atomic64_t *ptr) { return atomic64_sub_and_test(-1, ptr); } +EXPORT_SYMBOL(atomic64_inc_and_test); /** * atomic64_add_negative - add and test if negative @@ -222,3 +239,4 @@ int atomic64_add_negative(u64 delta, atomic64_t *ptr) return old_val < 0; } +EXPORT_SYMBOL(atomic64_add_negative); -- cgit v1.2.3 From 3a8d1788b37435baf6c296f4ea8beb4fa4955f44 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Fri, 3 Jul 2009 19:56:36 +0200 Subject: x86: atomic64: Improve atomic64_xchg() Remove the read-first logic from atomic64_xchg() and simplify the loop. This function was the last user of __atomic64_read() - remove it. Also, change the 'real_val' assumption from the somewhat quirky 1ULL << 32 value to the (just as arbitrary, but simpler) value of 0. Reported-by: Linus Torvalds Cc: Eric Dumazet Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: David Howells Cc: Andrew Morton Cc: Arnd Bergmann LKML-Reference: Signed-off-by: Ingo Molnar --- arch/x86/lib/atomic64_32.c | 21 +++++++++++++++------ 1 file changed, 15 insertions(+), 6 deletions(-) (limited to 'arch/x86/lib') diff --git a/arch/x86/lib/atomic64_32.c b/arch/x86/lib/atomic64_32.c index 6722a092e40..a804f96e90e 100644 --- a/arch/x86/lib/atomic64_32.c +++ b/arch/x86/lib/atomic64_32.c @@ -33,14 +33,23 @@ EXPORT_SYMBOL(atomic64_cmpxchg); * Atomically xchgs the value of @ptr to @new_val and returns * the old value. */ - u64 atomic64_xchg(atomic64_t *ptr, u64 new_val) { - u64 old_val; + /* + * Try first with a (possibly incorrect) assumption about + * what we have there. We'll do two loops most likely, + * but we'll get an ownership MESI transaction straight away + * instead of a read transaction followed by a + * flush-for-ownership transaction: + */ + u64 old_val, real_val = 0; do { - old_val = __atomic64_read(ptr); - } while (atomic64_cmpxchg(ptr, old_val, new_val) != old_val); + old_val = real_val; + + real_val = atomic64_cmpxchg(ptr, old_val, new_val); + + } while (real_val != old_val); return old_val; } @@ -91,13 +100,13 @@ EXPORT_SYMBOL(atomic64_read); noinline u64 atomic64_add_return(u64 delta, atomic64_t *ptr) { /* - * Try first with a (probably incorrect) assumption about + * Try first with a (possibly incorrect) assumption about * what we have there. We'll do two loops most likely, * but we'll get an ownership MESI transaction straight away * instead of a read transaction followed by a * flush-for-ownership transaction: */ - u64 old_val, new_val, real_val = 1ULL << 32; + u64 old_val, new_val, real_val = 0; do { old_val = real_val; -- cgit v1.2.3 From ddf9a003d32f720805ac30bcc15755e9289073de Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Fri, 3 Jul 2009 20:11:30 +0200 Subject: x86: atomic64: Clean up atomic64_sub_and_test() and atomic64_add_negative() Linus noticed that the variable name 'old_val' is confusingly named in these functions - the correct naming is 'new_val'. Reported-by: Linus Torvalds Cc: Eric Dumazet Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: David Howells Cc: Andrew Morton Cc: Arnd Bergmann LKML-Reference: Signed-off-by: Ingo Molnar --- arch/x86/lib/atomic64_32.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'arch/x86/lib') diff --git a/arch/x86/lib/atomic64_32.c b/arch/x86/lib/atomic64_32.c index a804f96e90e..1d98c9eb6ea 100644 --- a/arch/x86/lib/atomic64_32.c +++ b/arch/x86/lib/atomic64_32.c @@ -175,9 +175,9 @@ EXPORT_SYMBOL(atomic64_sub); */ int atomic64_sub_and_test(u64 delta, atomic64_t *ptr) { - u64 old_val = atomic64_sub_return(delta, ptr); + u64 new_val = atomic64_sub_return(delta, ptr); - return old_val == 0; + return new_val == 0; } EXPORT_SYMBOL(atomic64_sub_and_test); @@ -244,8 +244,8 @@ EXPORT_SYMBOL(atomic64_inc_and_test); */ int atomic64_add_negative(u64 delta, atomic64_t *ptr) { - long long old_val = atomic64_add_return(delta, ptr); + s64 new_val = atomic64_add_return(delta, ptr); - return old_val < 0; + return new_val < 0; } EXPORT_SYMBOL(atomic64_add_negative); -- cgit v1.2.3 From a79f0da80a508448434476b77f9d3d1a469eab67 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Fri, 3 Jul 2009 16:50:10 +0200 Subject: x86: atomic64: Inline atomic64_read() again Now atomic64_read() is light weight (no register pressure and small icache), we can inline it again. Also use "=&A" constraint instead of "+A" to avoid warning about unitialized 'res' variable. (gcc had to force 0 in eax/edx) $ size vmlinux.prev vmlinux.after text data bss dec hex filename 4908667 451676 1684868 7045211 6b805b vmlinux.prev 4908651 451676 1684868 7045195 6b804b vmlinux.after Signed-off-by: Eric Dumazet Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: David Howells Cc: Andrew Morton Cc: Arnd Bergmann LKML-Reference: <4A4E1AA2.30002@gmail.com> [ Also fix typo in atomic64_set() export ] Signed-off-by: Ingo Molnar --- arch/x86/lib/atomic64_32.c | 23 +---------------------- 1 file changed, 1 insertion(+), 22 deletions(-) (limited to 'arch/x86/lib') diff --git a/arch/x86/lib/atomic64_32.c b/arch/x86/lib/atomic64_32.c index 1d98c9eb6ea..824fa0be55a 100644 --- a/arch/x86/lib/atomic64_32.c +++ b/arch/x86/lib/atomic64_32.c @@ -66,31 +66,10 @@ void atomic64_set(atomic64_t *ptr, u64 new_val) { atomic64_xchg(ptr, new_val); } -EXPORT_SYMBOL(atomic64_read); +EXPORT_SYMBOL(atomic64_set); /** - * atomic64_read - read atomic64 variable - * @ptr: pointer to type atomic64_t - * - * Atomically reads the value of @ptr and returns it. - */ -u64 atomic64_read(atomic64_t *ptr) -{ - u64 res; - - asm volatile( - "mov %%ebx, %%eax\n\t" - "mov %%ecx, %%edx\n\t" - LOCK_PREFIX "cmpxchg8b %1\n" - : "+A" (res) - : "m" (*ptr) - ); - - return res; -} EXPORT_SYMBOL(atomic64_read); - -/** * atomic64_add_return - add and return * @delta: integer value to add * @ptr: pointer to type atomic64_t -- cgit v1.2.3