2005-04-17 05:20:36 +07:00
|
|
|
#ifndef _ASM_GENERIC_PERCPU_H_
|
|
|
|
#define _ASM_GENERIC_PERCPU_H_
|
2009-04-22 05:00:29 +07:00
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
#include <linux/compiler.h>
|
2007-05-03 00:27:10 +07:00
|
|
|
#include <linux/threads.h>
|
2009-04-22 05:00:29 +07:00
|
|
|
#include <linux/percpu-defs.h>
|
2008-01-30 19:32:52 +07:00
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
#ifdef CONFIG_SMP
|
|
|
|
|
2008-01-30 19:32:52 +07:00
|
|
|
/*
|
|
|
|
* per_cpu_offset() is the offset that has to be added to a
|
|
|
|
* percpu variable to get to the instance for a certain processor.
|
|
|
|
*
|
|
|
|
* Most arches use the __per_cpu_offset array for those offsets but
|
|
|
|
* some arches have their own ways of determining the offset (x86_64, s390).
|
|
|
|
*/
|
|
|
|
#ifndef __per_cpu_offset
|
2005-04-17 05:20:36 +07:00
|
|
|
extern unsigned long __per_cpu_offset[NR_CPUS];
|
|
|
|
|
2006-07-03 14:24:26 +07:00
|
|
|
#define per_cpu_offset(x) (__per_cpu_offset[x])
|
2008-01-30 19:32:52 +07:00
|
|
|
#endif
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Determine the offset for the currently active processor.
|
|
|
|
* An arch may define __my_cpu_offset to provide a more effective
|
|
|
|
* means of obtaining the offset to the per cpu variables of the
|
|
|
|
* current processor.
|
|
|
|
*/
|
|
|
|
#ifndef __my_cpu_offset
|
|
|
|
#define __my_cpu_offset per_cpu_offset(raw_smp_processor_id())
|
2008-02-24 02:40:17 +07:00
|
|
|
#endif
|
|
|
|
#ifdef CONFIG_DEBUG_PREEMPT
|
2008-01-30 19:32:52 +07:00
|
|
|
#define my_cpu_offset per_cpu_offset(smp_processor_id())
|
|
|
|
#else
|
|
|
|
#define my_cpu_offset __my_cpu_offset
|
|
|
|
#endif
|
|
|
|
|
2014-06-18 06:12:34 +07:00
|
|
|
/*
|
|
|
|
* Arch may define arch_raw_cpu_ptr() to provide more efficient address
|
|
|
|
* translations for raw_cpu_ptr().
|
|
|
|
*/
|
|
|
|
#ifndef arch_raw_cpu_ptr
|
|
|
|
#define arch_raw_cpu_ptr(ptr) SHIFT_PERCPU_PTR(ptr, __my_cpu_offset)
|
2010-09-09 23:17:26 +07:00
|
|
|
#endif
|
2014-06-18 06:12:34 +07:00
|
|
|
|
2008-01-30 19:33:32 +07:00
|
|
|
#ifdef CONFIG_HAVE_SETUP_PER_CPU_AREA
|
2008-01-30 19:32:52 +07:00
|
|
|
extern void setup_per_cpu_areas(void);
|
|
|
|
#endif
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
#endif /* SMP */
|
|
|
|
|
2009-04-22 05:00:24 +07:00
|
|
|
#ifndef PER_CPU_BASE_SECTION
|
|
|
|
#ifdef CONFIG_SMP
|
2010-02-20 07:03:43 +07:00
|
|
|
#define PER_CPU_BASE_SECTION ".data..percpu"
|
2009-04-22 05:00:24 +07:00
|
|
|
#else
|
|
|
|
#define PER_CPU_BASE_SECTION ".data"
|
|
|
|
#endif
|
|
|
|
#endif
|
|
|
|
|
2008-01-30 19:32:52 +07:00
|
|
|
#ifndef PER_CPU_ATTRIBUTES
|
|
|
|
#define PER_CPU_ATTRIBUTES
|
|
|
|
#endif
|
|
|
|
|
2009-07-01 01:41:18 +07:00
|
|
|
#ifndef PER_CPU_DEF_ATTRIBUTES
|
|
|
|
#define PER_CPU_DEF_ATTRIBUTES
|
|
|
|
#endif
|
|
|
|
|
percpu: improve generic percpu modify-return implementation
Some architectures require an additional load to find the address of
percpu pointers. In some implemenatations, the C aliasing rules do not
allow the result of that load to be kept over the store that modifies
the percpu variable, which causes additional loads.
Work around this by finding the pointer first, then operating on that.
It's also possible to mark things as restrict and those kind of games,
but that can require larger and arch specific changes.
On powerpc, __this_cpu_inc_return compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
ld 9,48(13)
ldx 3,9,3
With this patch it compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
To: Tejun Heo <tj@kernel.org>
To: Christoph Lameter <cl@linux.com>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arch@vger.kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>
2016-09-22 22:55:54 +07:00
|
|
|
#define raw_cpu_generic_read(pcp) \
|
|
|
|
({ \
|
|
|
|
*raw_cpu_ptr(&(pcp)); \
|
|
|
|
})
|
|
|
|
|
2014-06-18 06:12:39 +07:00
|
|
|
#define raw_cpu_generic_to_op(pcp, val, op) \
|
|
|
|
do { \
|
|
|
|
*raw_cpu_ptr(&(pcp)) op val; \
|
|
|
|
} while (0)
|
|
|
|
|
|
|
|
#define raw_cpu_generic_add_return(pcp, val) \
|
|
|
|
({ \
|
percpu: improve generic percpu modify-return implementation
Some architectures require an additional load to find the address of
percpu pointers. In some implemenatations, the C aliasing rules do not
allow the result of that load to be kept over the store that modifies
the percpu variable, which causes additional loads.
Work around this by finding the pointer first, then operating on that.
It's also possible to mark things as restrict and those kind of games,
but that can require larger and arch specific changes.
On powerpc, __this_cpu_inc_return compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
ld 9,48(13)
ldx 3,9,3
With this patch it compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
To: Tejun Heo <tj@kernel.org>
To: Christoph Lameter <cl@linux.com>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arch@vger.kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>
2016-09-22 22:55:54 +07:00
|
|
|
typeof(&(pcp)) __p = raw_cpu_ptr(&(pcp)); \
|
|
|
|
\
|
|
|
|
*__p += val; \
|
|
|
|
*__p; \
|
2014-06-18 06:12:39 +07:00
|
|
|
})
|
|
|
|
|
|
|
|
#define raw_cpu_generic_xchg(pcp, nval) \
|
2014-06-18 06:12:40 +07:00
|
|
|
({ \
|
percpu: improve generic percpu modify-return implementation
Some architectures require an additional load to find the address of
percpu pointers. In some implemenatations, the C aliasing rules do not
allow the result of that load to be kept over the store that modifies
the percpu variable, which causes additional loads.
Work around this by finding the pointer first, then operating on that.
It's also possible to mark things as restrict and those kind of games,
but that can require larger and arch specific changes.
On powerpc, __this_cpu_inc_return compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
ld 9,48(13)
ldx 3,9,3
With this patch it compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
To: Tejun Heo <tj@kernel.org>
To: Christoph Lameter <cl@linux.com>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arch@vger.kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>
2016-09-22 22:55:54 +07:00
|
|
|
typeof(&(pcp)) __p = raw_cpu_ptr(&(pcp)); \
|
2014-06-18 06:12:40 +07:00
|
|
|
typeof(pcp) __ret; \
|
percpu: improve generic percpu modify-return implementation
Some architectures require an additional load to find the address of
percpu pointers. In some implemenatations, the C aliasing rules do not
allow the result of that load to be kept over the store that modifies
the percpu variable, which causes additional loads.
Work around this by finding the pointer first, then operating on that.
It's also possible to mark things as restrict and those kind of games,
but that can require larger and arch specific changes.
On powerpc, __this_cpu_inc_return compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
ld 9,48(13)
ldx 3,9,3
With this patch it compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
To: Tejun Heo <tj@kernel.org>
To: Christoph Lameter <cl@linux.com>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arch@vger.kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>
2016-09-22 22:55:54 +07:00
|
|
|
__ret = *__p; \
|
|
|
|
*__p = nval; \
|
2014-06-18 06:12:40 +07:00
|
|
|
__ret; \
|
2014-06-18 06:12:39 +07:00
|
|
|
})
|
|
|
|
|
|
|
|
#define raw_cpu_generic_cmpxchg(pcp, oval, nval) \
|
|
|
|
({ \
|
percpu: improve generic percpu modify-return implementation
Some architectures require an additional load to find the address of
percpu pointers. In some implemenatations, the C aliasing rules do not
allow the result of that load to be kept over the store that modifies
the percpu variable, which causes additional loads.
Work around this by finding the pointer first, then operating on that.
It's also possible to mark things as restrict and those kind of games,
but that can require larger and arch specific changes.
On powerpc, __this_cpu_inc_return compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
ld 9,48(13)
ldx 3,9,3
With this patch it compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
To: Tejun Heo <tj@kernel.org>
To: Christoph Lameter <cl@linux.com>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arch@vger.kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>
2016-09-22 22:55:54 +07:00
|
|
|
typeof(&(pcp)) __p = raw_cpu_ptr(&(pcp)); \
|
2014-06-18 06:12:40 +07:00
|
|
|
typeof(pcp) __ret; \
|
percpu: improve generic percpu modify-return implementation
Some architectures require an additional load to find the address of
percpu pointers. In some implemenatations, the C aliasing rules do not
allow the result of that load to be kept over the store that modifies
the percpu variable, which causes additional loads.
Work around this by finding the pointer first, then operating on that.
It's also possible to mark things as restrict and those kind of games,
but that can require larger and arch specific changes.
On powerpc, __this_cpu_inc_return compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
ld 9,48(13)
ldx 3,9,3
With this patch it compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
To: Tejun Heo <tj@kernel.org>
To: Christoph Lameter <cl@linux.com>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arch@vger.kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>
2016-09-22 22:55:54 +07:00
|
|
|
__ret = *__p; \
|
2014-06-18 06:12:40 +07:00
|
|
|
if (__ret == (oval)) \
|
percpu: improve generic percpu modify-return implementation
Some architectures require an additional load to find the address of
percpu pointers. In some implemenatations, the C aliasing rules do not
allow the result of that load to be kept over the store that modifies
the percpu variable, which causes additional loads.
Work around this by finding the pointer first, then operating on that.
It's also possible to mark things as restrict and those kind of games,
but that can require larger and arch specific changes.
On powerpc, __this_cpu_inc_return compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
ld 9,48(13)
ldx 3,9,3
With this patch it compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
To: Tejun Heo <tj@kernel.org>
To: Christoph Lameter <cl@linux.com>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arch@vger.kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>
2016-09-22 22:55:54 +07:00
|
|
|
*__p = nval; \
|
2014-06-18 06:12:40 +07:00
|
|
|
__ret; \
|
2014-06-18 06:12:39 +07:00
|
|
|
})
|
|
|
|
|
2014-06-18 06:12:40 +07:00
|
|
|
#define raw_cpu_generic_cmpxchg_double(pcp1, pcp2, oval1, oval2, nval1, nval2) \
|
2014-06-18 06:12:39 +07:00
|
|
|
({ \
|
percpu: improve generic percpu modify-return implementation
Some architectures require an additional load to find the address of
percpu pointers. In some implemenatations, the C aliasing rules do not
allow the result of that load to be kept over the store that modifies
the percpu variable, which causes additional loads.
Work around this by finding the pointer first, then operating on that.
It's also possible to mark things as restrict and those kind of games,
but that can require larger and arch specific changes.
On powerpc, __this_cpu_inc_return compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
ld 9,48(13)
ldx 3,9,3
With this patch it compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
To: Tejun Heo <tj@kernel.org>
To: Christoph Lameter <cl@linux.com>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arch@vger.kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>
2016-09-22 22:55:54 +07:00
|
|
|
typeof(&(pcp1)) __p1 = raw_cpu_ptr(&(pcp1)); \
|
|
|
|
typeof(&(pcp2)) __p2 = raw_cpu_ptr(&(pcp2)); \
|
2014-06-18 06:12:39 +07:00
|
|
|
int __ret = 0; \
|
percpu: improve generic percpu modify-return implementation
Some architectures require an additional load to find the address of
percpu pointers. In some implemenatations, the C aliasing rules do not
allow the result of that load to be kept over the store that modifies
the percpu variable, which causes additional loads.
Work around this by finding the pointer first, then operating on that.
It's also possible to mark things as restrict and those kind of games,
but that can require larger and arch specific changes.
On powerpc, __this_cpu_inc_return compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
ld 9,48(13)
ldx 3,9,3
With this patch it compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
To: Tejun Heo <tj@kernel.org>
To: Christoph Lameter <cl@linux.com>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arch@vger.kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>
2016-09-22 22:55:54 +07:00
|
|
|
if (*__p1 == (oval1) && *__p2 == (oval2)) { \
|
|
|
|
*__p1 = nval1; \
|
|
|
|
*__p2 = nval2; \
|
2014-06-18 06:12:39 +07:00
|
|
|
__ret = 1; \
|
|
|
|
} \
|
|
|
|
(__ret); \
|
|
|
|
})
|
|
|
|
|
2014-06-18 06:12:40 +07:00
|
|
|
#define this_cpu_generic_read(pcp) \
|
|
|
|
({ \
|
|
|
|
typeof(pcp) __ret; \
|
2016-11-03 19:09:24 +07:00
|
|
|
preempt_disable_notrace(); \
|
percpu: improve generic percpu modify-return implementation
Some architectures require an additional load to find the address of
percpu pointers. In some implemenatations, the C aliasing rules do not
allow the result of that load to be kept over the store that modifies
the percpu variable, which causes additional loads.
Work around this by finding the pointer first, then operating on that.
It's also possible to mark things as restrict and those kind of games,
but that can require larger and arch specific changes.
On powerpc, __this_cpu_inc_return compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
ld 9,48(13)
ldx 3,9,3
With this patch it compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
To: Tejun Heo <tj@kernel.org>
To: Christoph Lameter <cl@linux.com>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arch@vger.kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>
2016-09-22 22:55:54 +07:00
|
|
|
__ret = raw_cpu_generic_read(pcp); \
|
2016-11-03 19:09:24 +07:00
|
|
|
preempt_enable_notrace(); \
|
2014-06-18 06:12:40 +07:00
|
|
|
__ret; \
|
2014-06-18 06:12:39 +07:00
|
|
|
})
|
|
|
|
|
2014-06-18 06:12:40 +07:00
|
|
|
#define this_cpu_generic_to_op(pcp, val, op) \
|
2014-06-18 06:12:39 +07:00
|
|
|
do { \
|
2014-06-18 06:12:40 +07:00
|
|
|
unsigned long __flags; \
|
|
|
|
raw_local_irq_save(__flags); \
|
percpu: improve generic percpu modify-return implementation
Some architectures require an additional load to find the address of
percpu pointers. In some implemenatations, the C aliasing rules do not
allow the result of that load to be kept over the store that modifies
the percpu variable, which causes additional loads.
Work around this by finding the pointer first, then operating on that.
It's also possible to mark things as restrict and those kind of games,
but that can require larger and arch specific changes.
On powerpc, __this_cpu_inc_return compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
ld 9,48(13)
ldx 3,9,3
With this patch it compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
To: Tejun Heo <tj@kernel.org>
To: Christoph Lameter <cl@linux.com>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arch@vger.kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>
2016-09-22 22:55:54 +07:00
|
|
|
raw_cpu_generic_to_op(pcp, val, op); \
|
2014-06-18 06:12:40 +07:00
|
|
|
raw_local_irq_restore(__flags); \
|
2014-06-18 06:12:39 +07:00
|
|
|
} while (0)
|
|
|
|
|
percpu: improve generic percpu modify-return implementation
Some architectures require an additional load to find the address of
percpu pointers. In some implemenatations, the C aliasing rules do not
allow the result of that load to be kept over the store that modifies
the percpu variable, which causes additional loads.
Work around this by finding the pointer first, then operating on that.
It's also possible to mark things as restrict and those kind of games,
but that can require larger and arch specific changes.
On powerpc, __this_cpu_inc_return compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
ld 9,48(13)
ldx 3,9,3
With this patch it compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
To: Tejun Heo <tj@kernel.org>
To: Christoph Lameter <cl@linux.com>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arch@vger.kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>
2016-09-22 22:55:54 +07:00
|
|
|
|
2014-06-18 06:12:40 +07:00
|
|
|
#define this_cpu_generic_add_return(pcp, val) \
|
2014-06-18 06:12:39 +07:00
|
|
|
({ \
|
2014-06-18 06:12:40 +07:00
|
|
|
typeof(pcp) __ret; \
|
|
|
|
unsigned long __flags; \
|
|
|
|
raw_local_irq_save(__flags); \
|
percpu: improve generic percpu modify-return implementation
Some architectures require an additional load to find the address of
percpu pointers. In some implemenatations, the C aliasing rules do not
allow the result of that load to be kept over the store that modifies
the percpu variable, which causes additional loads.
Work around this by finding the pointer first, then operating on that.
It's also possible to mark things as restrict and those kind of games,
but that can require larger and arch specific changes.
On powerpc, __this_cpu_inc_return compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
ld 9,48(13)
ldx 3,9,3
With this patch it compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
To: Tejun Heo <tj@kernel.org>
To: Christoph Lameter <cl@linux.com>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arch@vger.kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>
2016-09-22 22:55:54 +07:00
|
|
|
__ret = raw_cpu_generic_add_return(pcp, val); \
|
2014-06-18 06:12:40 +07:00
|
|
|
raw_local_irq_restore(__flags); \
|
|
|
|
__ret; \
|
2014-06-18 06:12:39 +07:00
|
|
|
})
|
|
|
|
|
2014-06-18 06:12:40 +07:00
|
|
|
#define this_cpu_generic_xchg(pcp, nval) \
|
|
|
|
({ \
|
|
|
|
typeof(pcp) __ret; \
|
|
|
|
unsigned long __flags; \
|
|
|
|
raw_local_irq_save(__flags); \
|
percpu: improve generic percpu modify-return implementation
Some architectures require an additional load to find the address of
percpu pointers. In some implemenatations, the C aliasing rules do not
allow the result of that load to be kept over the store that modifies
the percpu variable, which causes additional loads.
Work around this by finding the pointer first, then operating on that.
It's also possible to mark things as restrict and those kind of games,
but that can require larger and arch specific changes.
On powerpc, __this_cpu_inc_return compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
ld 9,48(13)
ldx 3,9,3
With this patch it compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
To: Tejun Heo <tj@kernel.org>
To: Christoph Lameter <cl@linux.com>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arch@vger.kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>
2016-09-22 22:55:54 +07:00
|
|
|
__ret = raw_cpu_generic_xchg(pcp, nval); \
|
2014-06-18 06:12:40 +07:00
|
|
|
raw_local_irq_restore(__flags); \
|
|
|
|
__ret; \
|
2014-06-18 06:12:39 +07:00
|
|
|
})
|
|
|
|
|
2014-06-18 06:12:40 +07:00
|
|
|
#define this_cpu_generic_cmpxchg(pcp, oval, nval) \
|
2014-06-18 06:12:39 +07:00
|
|
|
({ \
|
2014-06-18 06:12:40 +07:00
|
|
|
typeof(pcp) __ret; \
|
|
|
|
unsigned long __flags; \
|
|
|
|
raw_local_irq_save(__flags); \
|
percpu: improve generic percpu modify-return implementation
Some architectures require an additional load to find the address of
percpu pointers. In some implemenatations, the C aliasing rules do not
allow the result of that load to be kept over the store that modifies
the percpu variable, which causes additional loads.
Work around this by finding the pointer first, then operating on that.
It's also possible to mark things as restrict and those kind of games,
but that can require larger and arch specific changes.
On powerpc, __this_cpu_inc_return compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
ld 9,48(13)
ldx 3,9,3
With this patch it compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
To: Tejun Heo <tj@kernel.org>
To: Christoph Lameter <cl@linux.com>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arch@vger.kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>
2016-09-22 22:55:54 +07:00
|
|
|
__ret = raw_cpu_generic_cmpxchg(pcp, oval, nval); \
|
2014-06-18 06:12:40 +07:00
|
|
|
raw_local_irq_restore(__flags); \
|
|
|
|
__ret; \
|
2014-06-18 06:12:39 +07:00
|
|
|
})
|
|
|
|
|
2014-06-18 06:12:40 +07:00
|
|
|
#define this_cpu_generic_cmpxchg_double(pcp1, pcp2, oval1, oval2, nval1, nval2) \
|
2014-06-18 06:12:39 +07:00
|
|
|
({ \
|
2014-06-18 06:12:40 +07:00
|
|
|
int __ret; \
|
|
|
|
unsigned long __flags; \
|
|
|
|
raw_local_irq_save(__flags); \
|
|
|
|
__ret = raw_cpu_generic_cmpxchg_double(pcp1, pcp2, \
|
2014-06-18 06:12:39 +07:00
|
|
|
oval1, oval2, nval1, nval2); \
|
2014-06-18 06:12:40 +07:00
|
|
|
raw_local_irq_restore(__flags); \
|
|
|
|
__ret; \
|
2014-06-18 06:12:39 +07:00
|
|
|
})
|
|
|
|
|
2014-06-18 06:12:40 +07:00
|
|
|
#ifndef raw_cpu_read_1
|
percpu: improve generic percpu modify-return implementation
Some architectures require an additional load to find the address of
percpu pointers. In some implemenatations, the C aliasing rules do not
allow the result of that load to be kept over the store that modifies
the percpu variable, which causes additional loads.
Work around this by finding the pointer first, then operating on that.
It's also possible to mark things as restrict and those kind of games,
but that can require larger and arch specific changes.
On powerpc, __this_cpu_inc_return compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
ld 9,48(13)
ldx 3,9,3
With this patch it compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
To: Tejun Heo <tj@kernel.org>
To: Christoph Lameter <cl@linux.com>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arch@vger.kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>
2016-09-22 22:55:54 +07:00
|
|
|
#define raw_cpu_read_1(pcp) raw_cpu_generic_read(pcp)
|
2014-06-18 06:12:40 +07:00
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_read_2
|
percpu: improve generic percpu modify-return implementation
Some architectures require an additional load to find the address of
percpu pointers. In some implemenatations, the C aliasing rules do not
allow the result of that load to be kept over the store that modifies
the percpu variable, which causes additional loads.
Work around this by finding the pointer first, then operating on that.
It's also possible to mark things as restrict and those kind of games,
but that can require larger and arch specific changes.
On powerpc, __this_cpu_inc_return compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
ld 9,48(13)
ldx 3,9,3
With this patch it compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
To: Tejun Heo <tj@kernel.org>
To: Christoph Lameter <cl@linux.com>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arch@vger.kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>
2016-09-22 22:55:54 +07:00
|
|
|
#define raw_cpu_read_2(pcp) raw_cpu_generic_read(pcp)
|
2014-06-18 06:12:40 +07:00
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_read_4
|
percpu: improve generic percpu modify-return implementation
Some architectures require an additional load to find the address of
percpu pointers. In some implemenatations, the C aliasing rules do not
allow the result of that load to be kept over the store that modifies
the percpu variable, which causes additional loads.
Work around this by finding the pointer first, then operating on that.
It's also possible to mark things as restrict and those kind of games,
but that can require larger and arch specific changes.
On powerpc, __this_cpu_inc_return compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
ld 9,48(13)
ldx 3,9,3
With this patch it compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
To: Tejun Heo <tj@kernel.org>
To: Christoph Lameter <cl@linux.com>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arch@vger.kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>
2016-09-22 22:55:54 +07:00
|
|
|
#define raw_cpu_read_4(pcp) raw_cpu_generic_read(pcp)
|
2014-06-18 06:12:40 +07:00
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_read_8
|
percpu: improve generic percpu modify-return implementation
Some architectures require an additional load to find the address of
percpu pointers. In some implemenatations, the C aliasing rules do not
allow the result of that load to be kept over the store that modifies
the percpu variable, which causes additional loads.
Work around this by finding the pointer first, then operating on that.
It's also possible to mark things as restrict and those kind of games,
but that can require larger and arch specific changes.
On powerpc, __this_cpu_inc_return compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
ld 9,48(13)
ldx 3,9,3
With this patch it compiles to:
ld 10,48(13)
ldx 9,3,10
addi 9,9,1
stdx 9,3,10
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
To: Tejun Heo <tj@kernel.org>
To: Christoph Lameter <cl@linux.com>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arch@vger.kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>
2016-09-22 22:55:54 +07:00
|
|
|
#define raw_cpu_read_8(pcp) raw_cpu_generic_read(pcp)
|
2014-06-18 06:12:40 +07:00
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef raw_cpu_write_1
|
|
|
|
#define raw_cpu_write_1(pcp, val) raw_cpu_generic_to_op(pcp, val, =)
|
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_write_2
|
|
|
|
#define raw_cpu_write_2(pcp, val) raw_cpu_generic_to_op(pcp, val, =)
|
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_write_4
|
|
|
|
#define raw_cpu_write_4(pcp, val) raw_cpu_generic_to_op(pcp, val, =)
|
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_write_8
|
|
|
|
#define raw_cpu_write_8(pcp, val) raw_cpu_generic_to_op(pcp, val, =)
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef raw_cpu_add_1
|
|
|
|
#define raw_cpu_add_1(pcp, val) raw_cpu_generic_to_op(pcp, val, +=)
|
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_add_2
|
|
|
|
#define raw_cpu_add_2(pcp, val) raw_cpu_generic_to_op(pcp, val, +=)
|
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_add_4
|
|
|
|
#define raw_cpu_add_4(pcp, val) raw_cpu_generic_to_op(pcp, val, +=)
|
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_add_8
|
|
|
|
#define raw_cpu_add_8(pcp, val) raw_cpu_generic_to_op(pcp, val, +=)
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef raw_cpu_and_1
|
|
|
|
#define raw_cpu_and_1(pcp, val) raw_cpu_generic_to_op(pcp, val, &=)
|
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_and_2
|
|
|
|
#define raw_cpu_and_2(pcp, val) raw_cpu_generic_to_op(pcp, val, &=)
|
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_and_4
|
|
|
|
#define raw_cpu_and_4(pcp, val) raw_cpu_generic_to_op(pcp, val, &=)
|
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_and_8
|
|
|
|
#define raw_cpu_and_8(pcp, val) raw_cpu_generic_to_op(pcp, val, &=)
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef raw_cpu_or_1
|
|
|
|
#define raw_cpu_or_1(pcp, val) raw_cpu_generic_to_op(pcp, val, |=)
|
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_or_2
|
|
|
|
#define raw_cpu_or_2(pcp, val) raw_cpu_generic_to_op(pcp, val, |=)
|
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_or_4
|
|
|
|
#define raw_cpu_or_4(pcp, val) raw_cpu_generic_to_op(pcp, val, |=)
|
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_or_8
|
|
|
|
#define raw_cpu_or_8(pcp, val) raw_cpu_generic_to_op(pcp, val, |=)
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef raw_cpu_add_return_1
|
|
|
|
#define raw_cpu_add_return_1(pcp, val) raw_cpu_generic_add_return(pcp, val)
|
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_add_return_2
|
|
|
|
#define raw_cpu_add_return_2(pcp, val) raw_cpu_generic_add_return(pcp, val)
|
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_add_return_4
|
|
|
|
#define raw_cpu_add_return_4(pcp, val) raw_cpu_generic_add_return(pcp, val)
|
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_add_return_8
|
|
|
|
#define raw_cpu_add_return_8(pcp, val) raw_cpu_generic_add_return(pcp, val)
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef raw_cpu_xchg_1
|
|
|
|
#define raw_cpu_xchg_1(pcp, nval) raw_cpu_generic_xchg(pcp, nval)
|
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_xchg_2
|
|
|
|
#define raw_cpu_xchg_2(pcp, nval) raw_cpu_generic_xchg(pcp, nval)
|
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_xchg_4
|
|
|
|
#define raw_cpu_xchg_4(pcp, nval) raw_cpu_generic_xchg(pcp, nval)
|
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_xchg_8
|
|
|
|
#define raw_cpu_xchg_8(pcp, nval) raw_cpu_generic_xchg(pcp, nval)
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef raw_cpu_cmpxchg_1
|
|
|
|
#define raw_cpu_cmpxchg_1(pcp, oval, nval) \
|
|
|
|
raw_cpu_generic_cmpxchg(pcp, oval, nval)
|
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_cmpxchg_2
|
|
|
|
#define raw_cpu_cmpxchg_2(pcp, oval, nval) \
|
|
|
|
raw_cpu_generic_cmpxchg(pcp, oval, nval)
|
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_cmpxchg_4
|
|
|
|
#define raw_cpu_cmpxchg_4(pcp, oval, nval) \
|
|
|
|
raw_cpu_generic_cmpxchg(pcp, oval, nval)
|
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_cmpxchg_8
|
|
|
|
#define raw_cpu_cmpxchg_8(pcp, oval, nval) \
|
|
|
|
raw_cpu_generic_cmpxchg(pcp, oval, nval)
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef raw_cpu_cmpxchg_double_1
|
|
|
|
#define raw_cpu_cmpxchg_double_1(pcp1, pcp2, oval1, oval2, nval1, nval2) \
|
2014-06-18 06:12:39 +07:00
|
|
|
raw_cpu_generic_cmpxchg_double(pcp1, pcp2, oval1, oval2, nval1, nval2)
|
2014-06-18 06:12:40 +07:00
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_cmpxchg_double_2
|
|
|
|
#define raw_cpu_cmpxchg_double_2(pcp1, pcp2, oval1, oval2, nval1, nval2) \
|
2014-06-18 06:12:39 +07:00
|
|
|
raw_cpu_generic_cmpxchg_double(pcp1, pcp2, oval1, oval2, nval1, nval2)
|
2014-06-18 06:12:40 +07:00
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_cmpxchg_double_4
|
|
|
|
#define raw_cpu_cmpxchg_double_4(pcp1, pcp2, oval1, oval2, nval1, nval2) \
|
2014-06-18 06:12:39 +07:00
|
|
|
raw_cpu_generic_cmpxchg_double(pcp1, pcp2, oval1, oval2, nval1, nval2)
|
2014-06-18 06:12:40 +07:00
|
|
|
#endif
|
|
|
|
#ifndef raw_cpu_cmpxchg_double_8
|
|
|
|
#define raw_cpu_cmpxchg_double_8(pcp1, pcp2, oval1, oval2, nval1, nval2) \
|
2014-06-18 06:12:39 +07:00
|
|
|
raw_cpu_generic_cmpxchg_double(pcp1, pcp2, oval1, oval2, nval1, nval2)
|
2014-06-18 06:12:40 +07:00
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef this_cpu_read_1
|
|
|
|
#define this_cpu_read_1(pcp) this_cpu_generic_read(pcp)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_read_2
|
|
|
|
#define this_cpu_read_2(pcp) this_cpu_generic_read(pcp)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_read_4
|
|
|
|
#define this_cpu_read_4(pcp) this_cpu_generic_read(pcp)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_read_8
|
|
|
|
#define this_cpu_read_8(pcp) this_cpu_generic_read(pcp)
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef this_cpu_write_1
|
|
|
|
#define this_cpu_write_1(pcp, val) this_cpu_generic_to_op(pcp, val, =)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_write_2
|
|
|
|
#define this_cpu_write_2(pcp, val) this_cpu_generic_to_op(pcp, val, =)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_write_4
|
|
|
|
#define this_cpu_write_4(pcp, val) this_cpu_generic_to_op(pcp, val, =)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_write_8
|
|
|
|
#define this_cpu_write_8(pcp, val) this_cpu_generic_to_op(pcp, val, =)
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef this_cpu_add_1
|
|
|
|
#define this_cpu_add_1(pcp, val) this_cpu_generic_to_op(pcp, val, +=)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_add_2
|
|
|
|
#define this_cpu_add_2(pcp, val) this_cpu_generic_to_op(pcp, val, +=)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_add_4
|
|
|
|
#define this_cpu_add_4(pcp, val) this_cpu_generic_to_op(pcp, val, +=)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_add_8
|
|
|
|
#define this_cpu_add_8(pcp, val) this_cpu_generic_to_op(pcp, val, +=)
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef this_cpu_and_1
|
|
|
|
#define this_cpu_and_1(pcp, val) this_cpu_generic_to_op(pcp, val, &=)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_and_2
|
|
|
|
#define this_cpu_and_2(pcp, val) this_cpu_generic_to_op(pcp, val, &=)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_and_4
|
|
|
|
#define this_cpu_and_4(pcp, val) this_cpu_generic_to_op(pcp, val, &=)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_and_8
|
|
|
|
#define this_cpu_and_8(pcp, val) this_cpu_generic_to_op(pcp, val, &=)
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef this_cpu_or_1
|
|
|
|
#define this_cpu_or_1(pcp, val) this_cpu_generic_to_op(pcp, val, |=)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_or_2
|
|
|
|
#define this_cpu_or_2(pcp, val) this_cpu_generic_to_op(pcp, val, |=)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_or_4
|
|
|
|
#define this_cpu_or_4(pcp, val) this_cpu_generic_to_op(pcp, val, |=)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_or_8
|
|
|
|
#define this_cpu_or_8(pcp, val) this_cpu_generic_to_op(pcp, val, |=)
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef this_cpu_add_return_1
|
|
|
|
#define this_cpu_add_return_1(pcp, val) this_cpu_generic_add_return(pcp, val)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_add_return_2
|
|
|
|
#define this_cpu_add_return_2(pcp, val) this_cpu_generic_add_return(pcp, val)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_add_return_4
|
|
|
|
#define this_cpu_add_return_4(pcp, val) this_cpu_generic_add_return(pcp, val)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_add_return_8
|
|
|
|
#define this_cpu_add_return_8(pcp, val) this_cpu_generic_add_return(pcp, val)
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef this_cpu_xchg_1
|
|
|
|
#define this_cpu_xchg_1(pcp, nval) this_cpu_generic_xchg(pcp, nval)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_xchg_2
|
|
|
|
#define this_cpu_xchg_2(pcp, nval) this_cpu_generic_xchg(pcp, nval)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_xchg_4
|
|
|
|
#define this_cpu_xchg_4(pcp, nval) this_cpu_generic_xchg(pcp, nval)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_xchg_8
|
|
|
|
#define this_cpu_xchg_8(pcp, nval) this_cpu_generic_xchg(pcp, nval)
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef this_cpu_cmpxchg_1
|
|
|
|
#define this_cpu_cmpxchg_1(pcp, oval, nval) \
|
|
|
|
this_cpu_generic_cmpxchg(pcp, oval, nval)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_cmpxchg_2
|
|
|
|
#define this_cpu_cmpxchg_2(pcp, oval, nval) \
|
|
|
|
this_cpu_generic_cmpxchg(pcp, oval, nval)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_cmpxchg_4
|
|
|
|
#define this_cpu_cmpxchg_4(pcp, oval, nval) \
|
|
|
|
this_cpu_generic_cmpxchg(pcp, oval, nval)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_cmpxchg_8
|
|
|
|
#define this_cpu_cmpxchg_8(pcp, oval, nval) \
|
|
|
|
this_cpu_generic_cmpxchg(pcp, oval, nval)
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifndef this_cpu_cmpxchg_double_1
|
|
|
|
#define this_cpu_cmpxchg_double_1(pcp1, pcp2, oval1, oval2, nval1, nval2) \
|
|
|
|
this_cpu_generic_cmpxchg_double(pcp1, pcp2, oval1, oval2, nval1, nval2)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_cmpxchg_double_2
|
|
|
|
#define this_cpu_cmpxchg_double_2(pcp1, pcp2, oval1, oval2, nval1, nval2) \
|
|
|
|
this_cpu_generic_cmpxchg_double(pcp1, pcp2, oval1, oval2, nval1, nval2)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_cmpxchg_double_4
|
|
|
|
#define this_cpu_cmpxchg_double_4(pcp1, pcp2, oval1, oval2, nval1, nval2) \
|
|
|
|
this_cpu_generic_cmpxchg_double(pcp1, pcp2, oval1, oval2, nval1, nval2)
|
|
|
|
#endif
|
|
|
|
#ifndef this_cpu_cmpxchg_double_8
|
|
|
|
#define this_cpu_cmpxchg_double_8(pcp1, pcp2, oval1, oval2, nval1, nval2) \
|
|
|
|
this_cpu_generic_cmpxchg_double(pcp1, pcp2, oval1, oval2, nval1, nval2)
|
|
|
|
#endif
|
2014-06-18 06:12:39 +07:00
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
#endif /* _ASM_GENERIC_PERCPU_H_ */
|