![Nick Piggin](/assets/img/avatar_default.png)
Implement get_user_pages_fast without locking in the fastpath on x86. Do an optimistic lockless pagetable walk, without taking mmap_sem or any page table locks or even mmap_sem. Page table existence is guaranteed by turning interrupts off (combined with the fact that we're always looking up the current mm, means we can do the lockless page table walk within the constraints of the TLB shootdown design). Basically we can do this lockless pagetable walk in a similar manner to the way the CPU's pagetable walker does not have to take any locks to find present ptes. This patch (combined with the subsequent ones to convert direct IO to use it) was found to give about 10% performance improvement on a 2 socket 8 core Intel Xeon system running an OLTP workload on DB2 v9.5 "To test the effects of the patch, an OLTP workload was run on an IBM x3850 M2 server with 2 processors (quad-core Intel Xeon processors at 2.93 GHz) using IBM DB2 v9.5 running Linux 2.6.24rc7 kernel. Comparing runs with and without the patch resulted in an overall performance benefit of ~9.8%. Correspondingly, oprofiles showed that samples from __up_read and __down_read routines that is seen during thread contention for system resources was reduced from 2.8% down to .05%. Monitoring the /proc/vmstat output from the patched run showed that the counter for fast_gup contained a very high number while the fast_gup_slow value was zero." (fast_gup is the old name for get_user_pages_fast, fast_gup_slow is a counter we had for the number of times the slowpath was invoked). The main reason for the improvement is that DB2 has multiple threads each issuing direct-IO. Direct-IO uses get_user_pages, and thus the threads contend the mmap_sem cacheline, and can also contend on page table locks. I would anticipate larger performance gains on larger systems, however I think DB2 uses an adaptive mix of threads and processes, so it could be that thread contention remains pretty constant as machine size increases. In which case, we stuck with "only" a 10% gain. The downside of using get_user_pages_fast is that if there is not a pte with the correct permissions for the access, we end up falling back to get_user_pages and so the get_user_pages_fast is a bit of extra work. However this should not be the common case in most performance critical code. [akpm@linux-foundation.org: coding-style fixes] [akpm@linux-foundation.org: build fix] [akpm@linux-foundation.org: Kconfig fix] [akpm@linux-foundation.org: Makefile fix/cleanup] [akpm@linux-foundation.org: warning fix] Signed-off-by: Nick Piggin <npiggin@suse.de> Cc: Dave Kleikamp <shaggy@austin.ibm.com> Cc: Andy Whitcroft <apw@shadowen.org> Cc: Ingo Molnar <mingo@elte.hu> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Andi Kleen <andi@firstfloor.org> Cc: Dave Kleikamp <shaggy@austin.ibm.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Zach Brown <zach.brown@oracle.com> Cc: Jens Axboe <jens.axboe@oracle.com> Reviewed-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
455 lines
13 KiB
C
455 lines
13 KiB
C
#ifndef _ASM_UACCES_H_
|
|
#define _ASM_UACCES_H_
|
|
/*
|
|
* User space memory access functions
|
|
*/
|
|
#include <linux/errno.h>
|
|
#include <linux/compiler.h>
|
|
#include <linux/thread_info.h>
|
|
#include <linux/prefetch.h>
|
|
#include <linux/string.h>
|
|
#include <asm/asm.h>
|
|
#include <asm/page.h>
|
|
|
|
#define VERIFY_READ 0
|
|
#define VERIFY_WRITE 1
|
|
|
|
/*
|
|
* The fs value determines whether argument validity checking should be
|
|
* performed or not. If get_fs() == USER_DS, checking is performed, with
|
|
* get_fs() == KERNEL_DS, checking is bypassed.
|
|
*
|
|
* For historical reasons, these macros are grossly misnamed.
|
|
*/
|
|
|
|
#define MAKE_MM_SEG(s) ((mm_segment_t) { (s) })
|
|
|
|
#define KERNEL_DS MAKE_MM_SEG(-1UL)
|
|
#define USER_DS MAKE_MM_SEG(PAGE_OFFSET)
|
|
|
|
#define get_ds() (KERNEL_DS)
|
|
#define get_fs() (current_thread_info()->addr_limit)
|
|
#define set_fs(x) (current_thread_info()->addr_limit = (x))
|
|
|
|
#define segment_eq(a, b) ((a).seg == (b).seg)
|
|
|
|
#define __addr_ok(addr) \
|
|
((unsigned long __force)(addr) < \
|
|
(current_thread_info()->addr_limit.seg))
|
|
|
|
/*
|
|
* Test whether a block of memory is a valid user space address.
|
|
* Returns 0 if the range is valid, nonzero otherwise.
|
|
*
|
|
* This is equivalent to the following test:
|
|
* (u33)addr + (u33)size >= (u33)current->addr_limit.seg (u65 for x86_64)
|
|
*
|
|
* This needs 33-bit (65-bit for x86_64) arithmetic. We have a carry...
|
|
*/
|
|
|
|
#define __range_not_ok(addr, size) \
|
|
({ \
|
|
unsigned long flag, roksum; \
|
|
__chk_user_ptr(addr); \
|
|
asm("add %3,%1 ; sbb %0,%0 ; cmp %1,%4 ; sbb $0,%0" \
|
|
: "=&r" (flag), "=r" (roksum) \
|
|
: "1" (addr), "g" ((long)(size)), \
|
|
"rm" (current_thread_info()->addr_limit.seg)); \
|
|
flag; \
|
|
})
|
|
|
|
/**
|
|
* access_ok: - Checks if a user space pointer is valid
|
|
* @type: Type of access: %VERIFY_READ or %VERIFY_WRITE. Note that
|
|
* %VERIFY_WRITE is a superset of %VERIFY_READ - if it is safe
|
|
* to write to a block, it is always safe to read from it.
|
|
* @addr: User space pointer to start of block to check
|
|
* @size: Size of block to check
|
|
*
|
|
* Context: User context only. This function may sleep.
|
|
*
|
|
* Checks if a pointer to a block of memory in user space is valid.
|
|
*
|
|
* Returns true (nonzero) if the memory block may be valid, false (zero)
|
|
* if it is definitely invalid.
|
|
*
|
|
* Note that, depending on architecture, this function probably just
|
|
* checks that the pointer is in the user space range - after calling
|
|
* this function, memory access functions may still return -EFAULT.
|
|
*/
|
|
#define access_ok(type, addr, size) (likely(__range_not_ok(addr, size) == 0))
|
|
|
|
/*
|
|
* The exception table consists of pairs of addresses: the first is the
|
|
* address of an instruction that is allowed to fault, and the second is
|
|
* the address at which the program should continue. No registers are
|
|
* modified, so it is entirely up to the continuation code to figure out
|
|
* what to do.
|
|
*
|
|
* All the routines below use bits of fixup code that are out of line
|
|
* with the main instruction path. This means when everything is well,
|
|
* we don't even have to jump over them. Further, they do not intrude
|
|
* on our cache or tlb entries.
|
|
*/
|
|
|
|
struct exception_table_entry {
|
|
unsigned long insn, fixup;
|
|
};
|
|
|
|
extern int fixup_exception(struct pt_regs *regs);
|
|
|
|
/*
|
|
* These are the main single-value transfer routines. They automatically
|
|
* use the right size if we just have the right pointer type.
|
|
*
|
|
* This gets kind of ugly. We want to return _two_ values in "get_user()"
|
|
* and yet we don't want to do any pointers, because that is too much
|
|
* of a performance impact. Thus we have a few rather ugly macros here,
|
|
* and hide all the ugliness from the user.
|
|
*
|
|
* The "__xxx" versions of the user access functions are versions that
|
|
* do not verify the address space, that must have been done previously
|
|
* with a separate "access_ok()" call (this is used when we do multiple
|
|
* accesses to the same area of user memory).
|
|
*/
|
|
|
|
extern int __get_user_1(void);
|
|
extern int __get_user_2(void);
|
|
extern int __get_user_4(void);
|
|
extern int __get_user_8(void);
|
|
extern int __get_user_bad(void);
|
|
|
|
#define __get_user_x(size, ret, x, ptr) \
|
|
asm volatile("call __get_user_" #size \
|
|
: "=a" (ret),"=d" (x) \
|
|
: "0" (ptr)) \
|
|
|
|
/* Careful: we have to cast the result to the type of the pointer
|
|
* for sign reasons */
|
|
|
|
/**
|
|
* get_user: - Get a simple variable from user space.
|
|
* @x: Variable to store result.
|
|
* @ptr: Source address, in user space.
|
|
*
|
|
* Context: User context only. This function may sleep.
|
|
*
|
|
* This macro copies a single simple variable from user space to kernel
|
|
* space. It supports simple types like char and int, but not larger
|
|
* data types like structures or arrays.
|
|
*
|
|
* @ptr must have pointer-to-simple-variable type, and the result of
|
|
* dereferencing @ptr must be assignable to @x without a cast.
|
|
*
|
|
* Returns zero on success, or -EFAULT on error.
|
|
* On error, the variable @x is set to zero.
|
|
*/
|
|
#ifdef CONFIG_X86_32
|
|
#define __get_user_8(__ret_gu, __val_gu, ptr) \
|
|
__get_user_x(X, __ret_gu, __val_gu, ptr)
|
|
#else
|
|
#define __get_user_8(__ret_gu, __val_gu, ptr) \
|
|
__get_user_x(8, __ret_gu, __val_gu, ptr)
|
|
#endif
|
|
|
|
#define get_user(x, ptr) \
|
|
({ \
|
|
int __ret_gu; \
|
|
unsigned long __val_gu; \
|
|
__chk_user_ptr(ptr); \
|
|
switch (sizeof(*(ptr))) { \
|
|
case 1: \
|
|
__get_user_x(1, __ret_gu, __val_gu, ptr); \
|
|
break; \
|
|
case 2: \
|
|
__get_user_x(2, __ret_gu, __val_gu, ptr); \
|
|
break; \
|
|
case 4: \
|
|
__get_user_x(4, __ret_gu, __val_gu, ptr); \
|
|
break; \
|
|
case 8: \
|
|
__get_user_8(__ret_gu, __val_gu, ptr); \
|
|
break; \
|
|
default: \
|
|
__get_user_x(X, __ret_gu, __val_gu, ptr); \
|
|
break; \
|
|
} \
|
|
(x) = (__typeof__(*(ptr)))__val_gu; \
|
|
__ret_gu; \
|
|
})
|
|
|
|
#define __put_user_x(size, x, ptr, __ret_pu) \
|
|
asm volatile("call __put_user_" #size : "=a" (__ret_pu) \
|
|
:"0" ((typeof(*(ptr)))(x)), "c" (ptr) : "ebx")
|
|
|
|
|
|
|
|
#ifdef CONFIG_X86_32
|
|
#define __put_user_u64(x, addr, err) \
|
|
asm volatile("1: movl %%eax,0(%2)\n" \
|
|
"2: movl %%edx,4(%2)\n" \
|
|
"3:\n" \
|
|
".section .fixup,\"ax\"\n" \
|
|
"4: movl %3,%0\n" \
|
|
" jmp 3b\n" \
|
|
".previous\n" \
|
|
_ASM_EXTABLE(1b, 4b) \
|
|
_ASM_EXTABLE(2b, 4b) \
|
|
: "=r" (err) \
|
|
: "A" (x), "r" (addr), "i" (-EFAULT), "0" (err))
|
|
|
|
#define __put_user_x8(x, ptr, __ret_pu) \
|
|
asm volatile("call __put_user_8" : "=a" (__ret_pu) \
|
|
: "A" ((typeof(*(ptr)))(x)), "c" (ptr) : "ebx")
|
|
#else
|
|
#define __put_user_u64(x, ptr, retval) \
|
|
__put_user_asm(x, ptr, retval, "q", "", "Zr", -EFAULT)
|
|
#define __put_user_x8(x, ptr, __ret_pu) __put_user_x(8, x, ptr, __ret_pu)
|
|
#endif
|
|
|
|
extern void __put_user_bad(void);
|
|
|
|
/*
|
|
* Strange magic calling convention: pointer in %ecx,
|
|
* value in %eax(:%edx), return value in %eax. clobbers %rbx
|
|
*/
|
|
extern void __put_user_1(void);
|
|
extern void __put_user_2(void);
|
|
extern void __put_user_4(void);
|
|
extern void __put_user_8(void);
|
|
|
|
#ifdef CONFIG_X86_WP_WORKS_OK
|
|
|
|
/**
|
|
* put_user: - Write a simple value into user space.
|
|
* @x: Value to copy to user space.
|
|
* @ptr: Destination address, in user space.
|
|
*
|
|
* Context: User context only. This function may sleep.
|
|
*
|
|
* This macro copies a single simple value from kernel space to user
|
|
* space. It supports simple types like char and int, but not larger
|
|
* data types like structures or arrays.
|
|
*
|
|
* @ptr must have pointer-to-simple-variable type, and @x must be assignable
|
|
* to the result of dereferencing @ptr.
|
|
*
|
|
* Returns zero on success, or -EFAULT on error.
|
|
*/
|
|
#define put_user(x, ptr) \
|
|
({ \
|
|
int __ret_pu; \
|
|
__typeof__(*(ptr)) __pu_val; \
|
|
__chk_user_ptr(ptr); \
|
|
__pu_val = x; \
|
|
switch (sizeof(*(ptr))) { \
|
|
case 1: \
|
|
__put_user_x(1, __pu_val, ptr, __ret_pu); \
|
|
break; \
|
|
case 2: \
|
|
__put_user_x(2, __pu_val, ptr, __ret_pu); \
|
|
break; \
|
|
case 4: \
|
|
__put_user_x(4, __pu_val, ptr, __ret_pu); \
|
|
break; \
|
|
case 8: \
|
|
__put_user_x8(__pu_val, ptr, __ret_pu); \
|
|
break; \
|
|
default: \
|
|
__put_user_x(X, __pu_val, ptr, __ret_pu); \
|
|
break; \
|
|
} \
|
|
__ret_pu; \
|
|
})
|
|
|
|
#define __put_user_size(x, ptr, size, retval, errret) \
|
|
do { \
|
|
retval = 0; \
|
|
__chk_user_ptr(ptr); \
|
|
switch (size) { \
|
|
case 1: \
|
|
__put_user_asm(x, ptr, retval, "b", "b", "iq", errret); \
|
|
break; \
|
|
case 2: \
|
|
__put_user_asm(x, ptr, retval, "w", "w", "ir", errret); \
|
|
break; \
|
|
case 4: \
|
|
__put_user_asm(x, ptr, retval, "l", "k", "ir", errret);\
|
|
break; \
|
|
case 8: \
|
|
__put_user_u64((__typeof__(*ptr))(x), ptr, retval); \
|
|
break; \
|
|
default: \
|
|
__put_user_bad(); \
|
|
} \
|
|
} while (0)
|
|
|
|
#else
|
|
|
|
#define __put_user_size(x, ptr, size, retval, errret) \
|
|
do { \
|
|
__typeof__(*(ptr))__pus_tmp = x; \
|
|
retval = 0; \
|
|
\
|
|
if (unlikely(__copy_to_user_ll(ptr, &__pus_tmp, size) != 0)) \
|
|
retval = errret; \
|
|
} while (0)
|
|
|
|
#define put_user(x, ptr) \
|
|
({ \
|
|
int __ret_pu; \
|
|
__typeof__(*(ptr))__pus_tmp = x; \
|
|
__ret_pu = 0; \
|
|
if (unlikely(__copy_to_user_ll(ptr, &__pus_tmp, \
|
|
sizeof(*(ptr))) != 0)) \
|
|
__ret_pu = -EFAULT; \
|
|
__ret_pu; \
|
|
})
|
|
#endif
|
|
|
|
#ifdef CONFIG_X86_32
|
|
#define __get_user_asm_u64(x, ptr, retval, errret) (x) = __get_user_bad()
|
|
#else
|
|
#define __get_user_asm_u64(x, ptr, retval, errret) \
|
|
__get_user_asm(x, ptr, retval, "q", "", "=r", errret)
|
|
#endif
|
|
|
|
#define __get_user_size(x, ptr, size, retval, errret) \
|
|
do { \
|
|
retval = 0; \
|
|
__chk_user_ptr(ptr); \
|
|
switch (size) { \
|
|
case 1: \
|
|
__get_user_asm(x, ptr, retval, "b", "b", "=q", errret); \
|
|
break; \
|
|
case 2: \
|
|
__get_user_asm(x, ptr, retval, "w", "w", "=r", errret); \
|
|
break; \
|
|
case 4: \
|
|
__get_user_asm(x, ptr, retval, "l", "k", "=r", errret); \
|
|
break; \
|
|
case 8: \
|
|
__get_user_asm_u64(x, ptr, retval, errret); \
|
|
break; \
|
|
default: \
|
|
(x) = __get_user_bad(); \
|
|
} \
|
|
} while (0)
|
|
|
|
#define __get_user_asm(x, addr, err, itype, rtype, ltype, errret) \
|
|
asm volatile("1: mov"itype" %2,%"rtype"1\n" \
|
|
"2:\n" \
|
|
".section .fixup,\"ax\"\n" \
|
|
"3: mov %3,%0\n" \
|
|
" xor"itype" %"rtype"1,%"rtype"1\n" \
|
|
" jmp 2b\n" \
|
|
".previous\n" \
|
|
_ASM_EXTABLE(1b, 3b) \
|
|
: "=r" (err), ltype(x) \
|
|
: "m" (__m(addr)), "i" (errret), "0" (err))
|
|
|
|
#define __put_user_nocheck(x, ptr, size) \
|
|
({ \
|
|
long __pu_err; \
|
|
__put_user_size((x), (ptr), (size), __pu_err, -EFAULT); \
|
|
__pu_err; \
|
|
})
|
|
|
|
#define __get_user_nocheck(x, ptr, size) \
|
|
({ \
|
|
long __gu_err; \
|
|
unsigned long __gu_val; \
|
|
__get_user_size(__gu_val, (ptr), (size), __gu_err, -EFAULT); \
|
|
(x) = (__force __typeof__(*(ptr)))__gu_val; \
|
|
__gu_err; \
|
|
})
|
|
|
|
/* FIXME: this hack is definitely wrong -AK */
|
|
struct __large_struct { unsigned long buf[100]; };
|
|
#define __m(x) (*(struct __large_struct __user *)(x))
|
|
|
|
/*
|
|
* Tell gcc we read from memory instead of writing: this is because
|
|
* we do not write to any memory gcc knows about, so there are no
|
|
* aliasing issues.
|
|
*/
|
|
#define __put_user_asm(x, addr, err, itype, rtype, ltype, errret) \
|
|
asm volatile("1: mov"itype" %"rtype"1,%2\n" \
|
|
"2:\n" \
|
|
".section .fixup,\"ax\"\n" \
|
|
"3: mov %3,%0\n" \
|
|
" jmp 2b\n" \
|
|
".previous\n" \
|
|
_ASM_EXTABLE(1b, 3b) \
|
|
: "=r"(err) \
|
|
: ltype(x), "m" (__m(addr)), "i" (errret), "0" (err))
|
|
/**
|
|
* __get_user: - Get a simple variable from user space, with less checking.
|
|
* @x: Variable to store result.
|
|
* @ptr: Source address, in user space.
|
|
*
|
|
* Context: User context only. This function may sleep.
|
|
*
|
|
* This macro copies a single simple variable from user space to kernel
|
|
* space. It supports simple types like char and int, but not larger
|
|
* data types like structures or arrays.
|
|
*
|
|
* @ptr must have pointer-to-simple-variable type, and the result of
|
|
* dereferencing @ptr must be assignable to @x without a cast.
|
|
*
|
|
* Caller must check the pointer with access_ok() before calling this
|
|
* function.
|
|
*
|
|
* Returns zero on success, or -EFAULT on error.
|
|
* On error, the variable @x is set to zero.
|
|
*/
|
|
|
|
#define __get_user(x, ptr) \
|
|
__get_user_nocheck((x), (ptr), sizeof(*(ptr)))
|
|
/**
|
|
* __put_user: - Write a simple value into user space, with less checking.
|
|
* @x: Value to copy to user space.
|
|
* @ptr: Destination address, in user space.
|
|
*
|
|
* Context: User context only. This function may sleep.
|
|
*
|
|
* This macro copies a single simple value from kernel space to user
|
|
* space. It supports simple types like char and int, but not larger
|
|
* data types like structures or arrays.
|
|
*
|
|
* @ptr must have pointer-to-simple-variable type, and @x must be assignable
|
|
* to the result of dereferencing @ptr.
|
|
*
|
|
* Caller must check the pointer with access_ok() before calling this
|
|
* function.
|
|
*
|
|
* Returns zero on success, or -EFAULT on error.
|
|
*/
|
|
|
|
#define __put_user(x, ptr) \
|
|
__put_user_nocheck((__typeof__(*(ptr)))(x), (ptr), sizeof(*(ptr)))
|
|
|
|
#define __get_user_unaligned __get_user
|
|
#define __put_user_unaligned __put_user
|
|
|
|
/*
|
|
* movsl can be slow when source and dest are not both 8-byte aligned
|
|
*/
|
|
#ifdef CONFIG_X86_INTEL_USERCOPY
|
|
extern struct movsl_mask {
|
|
int mask;
|
|
} ____cacheline_aligned_in_smp movsl_mask;
|
|
#endif
|
|
|
|
#define ARCH_HAS_NOCACHE_UACCESS 1
|
|
|
|
#ifdef CONFIG_X86_32
|
|
# include "uaccess_32.h"
|
|
#else
|
|
# define ARCH_HAS_SEARCH_EXTABLE
|
|
# include "uaccess_64.h"
|
|
#endif
|
|
|
|
#endif
|
|
|