Blame view
crypto/xor.c
3.42 KB
af1a8899d
|
1 |
// SPDX-License-Identifier: GPL-2.0-or-later |
1da177e4c
|
2 3 4 5 6 7 8 |
/* * xor.c : Multiple Devices driver for Linux * * Copyright (C) 1996, 1997, 1998, 1999, 2000, * Ingo Molnar, Matti Aarnio, Jakub Jelinek, Richard Henderson. * * Dispatch optimized RAID-5 checksumming functions. |
1da177e4c
|
9 10 11 12 |
*/ #define BH_TRACE 0 #include <linux/module.h> |
5a0e3ad6a
|
13 |
#include <linux/gfp.h> |
1da177e4c
|
14 |
#include <linux/raid/xor.h> |
bff61975b
|
15 |
#include <linux/jiffies.h> |
56a519913
|
16 |
#include <linux/preempt.h> |
1da177e4c
|
17 |
#include <asm/xor.h> |
27c4d548a
|
18 19 20 |
#ifndef XOR_SELECT_TEMPLATE #define XOR_SELECT_TEMPLATE(x) (x) #endif |
1da177e4c
|
21 22 23 24 |
/* The xor routines to use. */ static struct xor_block_template *active_template; void |
9bc89cd82
|
25 |
xor_blocks(unsigned int src_count, unsigned int bytes, void *dest, void **srcs) |
1da177e4c
|
26 |
{ |
9bc89cd82
|
27 |
unsigned long *p1, *p2, *p3, *p4; |
1da177e4c
|
28 |
|
9bc89cd82
|
29 30 31 |
p1 = (unsigned long *) srcs[0]; if (src_count == 1) { active_template->do_2(bytes, dest, p1); |
1da177e4c
|
32 33 |
return; } |
9bc89cd82
|
34 35 36 |
p2 = (unsigned long *) srcs[1]; if (src_count == 2) { active_template->do_3(bytes, dest, p1, p2); |
1da177e4c
|
37 38 |
return; } |
9bc89cd82
|
39 40 41 |
p3 = (unsigned long *) srcs[2]; if (src_count == 3) { active_template->do_4(bytes, dest, p1, p2, p3); |
1da177e4c
|
42 43 |
return; } |
9bc89cd82
|
44 45 |
p4 = (unsigned long *) srcs[3]; active_template->do_5(bytes, dest, p1, p2, p3, p4); |
1da177e4c
|
46 |
} |
685784aaf
|
47 |
EXPORT_SYMBOL(xor_blocks); |
1da177e4c
|
48 49 |
/* Set of all registered templates. */ |
af7cf25dd
|
50 |
static struct xor_block_template *__initdata template_list; |
1da177e4c
|
51 52 |
#define BENCH_SIZE (PAGE_SIZE) |
af7cf25dd
|
53 |
static void __init |
1da177e4c
|
54 55 56 |
do_xor_speed(struct xor_block_template *tmpl, void *b1, void *b2) { int speed; |
6a328475c
|
57 |
unsigned long now, j; |
1da177e4c
|
58 59 60 61 |
int i, count, max; tmpl->next = template_list; template_list = tmpl; |
56a519913
|
62 |
preempt_disable(); |
1da177e4c
|
63 64 65 66 67 68 69 |
/* * Count the number of XORs done during a whole jiffy, and use * this to calculate the speed of checksumming. We use a 2-page * allocation to have guaranteed color L1-cache layout. */ max = 0; for (i = 0; i < 5; i++) { |
6a328475c
|
70 |
j = jiffies; |
1da177e4c
|
71 |
count = 0; |
6a328475c
|
72 73 74 |
while ((now = jiffies) == j) cpu_relax(); while (time_before(jiffies, now + 1)) { |
685784aaf
|
75 |
mb(); /* prevent loop optimzation */ |
1da177e4c
|
76 77 78 79 80 81 82 83 |
tmpl->do_2(BENCH_SIZE, b1, b2); mb(); count++; mb(); } if (count > max) max = count; } |
56a519913
|
84 |
preempt_enable(); |
1da177e4c
|
85 86 |
speed = max * (HZ * BENCH_SIZE / 1024); tmpl->speed = speed; |
685784aaf
|
87 88 |
printk(KERN_INFO " %-10s: %5d.%03d MB/sec ", tmpl->name, |
1da177e4c
|
89 90 |
speed / 1000, speed % 1000); } |
685784aaf
|
91 92 |
static int __init calibrate_xor_blocks(void) |
1da177e4c
|
93 94 95 |
{ void *b1, *b2; struct xor_block_template *f, *fastest; |
27c4d548a
|
96 |
fastest = XOR_SELECT_TEMPLATE(NULL); |
39457acda
|
97 |
|
39457acda
|
98 99 100 101 102 103 104 |
if (fastest) { printk(KERN_INFO "xor: automatically using best " "checksumming function %-10s ", fastest->name); goto out; } |
39457acda
|
105 |
|
75f296d93
|
106 |
b1 = (void *) __get_free_pages(GFP_KERNEL, 2); |
685784aaf
|
107 108 109 |
if (!b1) { printk(KERN_WARNING "xor: Yikes! No memory available. "); |
1da177e4c
|
110 111 112 113 114 |
return -ENOMEM; } b2 = b1 + 2*PAGE_SIZE + BENCH_SIZE; /* |
685784aaf
|
115 116 |
* If this arch/cpu has a short-circuited selection, don't loop through * all the possible functions, just test the best one |
1da177e4c
|
117 |
*/ |
1da177e4c
|
118 |
#define xor_speed(templ) do_xor_speed((templ), b1, b2) |
39457acda
|
119 120 121 122 123 124 125 |
printk(KERN_INFO "xor: measuring software checksum speed "); XOR_TRY_TEMPLATES; fastest = template_list; for (f = fastest; f; f = f->next) if (f->speed > fastest->speed) fastest = f; |
1da177e4c
|
126 |
|
685784aaf
|
127 128 |
printk(KERN_INFO "xor: using function: %s (%d.%03d MB/sec) ", |
1da177e4c
|
129 130 131 132 133 |
fastest->name, fastest->speed / 1000, fastest->speed % 1000); #undef xor_speed free_pages((unsigned long)b1, 2); |
39457acda
|
134 |
out: |
1da177e4c
|
135 136 137 138 139 |
active_template = fastest; return 0; } static __exit void xor_exit(void) { } |
1da177e4c
|
140 |
MODULE_LICENSE("GPL"); |
685784aaf
|
141 142 |
/* when built-in xor.o must initialize before drivers/md/md.o */ core_initcall(calibrate_xor_blocks); |
1da177e4c
|
143 |
module_exit(xor_exit); |