x86: cpa: move clflush_cache_range()
[wrt350n-kernel.git] / crypto / xor.c
blobb2e6db075e4993eac4fcffb9004b82ee4a0bca71
1 /*
2 * xor.c : Multiple Devices driver for Linux
4 * Copyright (C) 1996, 1997, 1998, 1999, 2000,
5 * Ingo Molnar, Matti Aarnio, Jakub Jelinek, Richard Henderson.
7 * Dispatch optimized RAID-5 checksumming functions.
9 * This program is free software; you can redistribute it and/or modify
10 * it under the terms of the GNU General Public License as published by
11 * the Free Software Foundation; either version 2, or (at your option)
12 * any later version.
14 * You should have received a copy of the GNU General Public License
15 * (for example /usr/src/linux/COPYING); if not, write to the Free
16 * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
19 #define BH_TRACE 0
20 #include <linux/module.h>
21 #include <linux/raid/md.h>
22 #include <linux/raid/xor.h>
23 #include <asm/xor.h>
25 /* The xor routines to use. */
26 static struct xor_block_template *active_template;
28 void
29 xor_blocks(unsigned int src_count, unsigned int bytes, void *dest, void **srcs)
31 unsigned long *p1, *p2, *p3, *p4;
33 p1 = (unsigned long *) srcs[0];
34 if (src_count == 1) {
35 active_template->do_2(bytes, dest, p1);
36 return;
39 p2 = (unsigned long *) srcs[1];
40 if (src_count == 2) {
41 active_template->do_3(bytes, dest, p1, p2);
42 return;
45 p3 = (unsigned long *) srcs[2];
46 if (src_count == 3) {
47 active_template->do_4(bytes, dest, p1, p2, p3);
48 return;
51 p4 = (unsigned long *) srcs[3];
52 active_template->do_5(bytes, dest, p1, p2, p3, p4);
54 EXPORT_SYMBOL(xor_blocks);
56 /* Set of all registered templates. */
57 static struct xor_block_template *template_list;
59 #define BENCH_SIZE (PAGE_SIZE)
61 static void
62 do_xor_speed(struct xor_block_template *tmpl, void *b1, void *b2)
64 int speed;
65 unsigned long now;
66 int i, count, max;
68 tmpl->next = template_list;
69 template_list = tmpl;
72 * Count the number of XORs done during a whole jiffy, and use
73 * this to calculate the speed of checksumming. We use a 2-page
74 * allocation to have guaranteed color L1-cache layout.
76 max = 0;
77 for (i = 0; i < 5; i++) {
78 now = jiffies;
79 count = 0;
80 while (jiffies == now) {
81 mb(); /* prevent loop optimzation */
82 tmpl->do_2(BENCH_SIZE, b1, b2);
83 mb();
84 count++;
85 mb();
87 if (count > max)
88 max = count;
91 speed = max * (HZ * BENCH_SIZE / 1024);
92 tmpl->speed = speed;
94 printk(KERN_INFO " %-10s: %5d.%03d MB/sec\n", tmpl->name,
95 speed / 1000, speed % 1000);
98 static int __init
99 calibrate_xor_blocks(void)
101 void *b1, *b2;
102 struct xor_block_template *f, *fastest;
104 b1 = (void *) __get_free_pages(GFP_KERNEL, 2);
105 if (!b1) {
106 printk(KERN_WARNING "xor: Yikes! No memory available.\n");
107 return -ENOMEM;
109 b2 = b1 + 2*PAGE_SIZE + BENCH_SIZE;
112 * If this arch/cpu has a short-circuited selection, don't loop through
113 * all the possible functions, just test the best one
116 fastest = NULL;
118 #ifdef XOR_SELECT_TEMPLATE
119 fastest = XOR_SELECT_TEMPLATE(fastest);
120 #endif
122 #define xor_speed(templ) do_xor_speed((templ), b1, b2)
124 if (fastest) {
125 printk(KERN_INFO "xor: automatically using best "
126 "checksumming function: %s\n",
127 fastest->name);
128 xor_speed(fastest);
129 } else {
130 printk(KERN_INFO "xor: measuring software checksum speed\n");
131 XOR_TRY_TEMPLATES;
132 fastest = template_list;
133 for (f = fastest; f; f = f->next)
134 if (f->speed > fastest->speed)
135 fastest = f;
138 printk(KERN_INFO "xor: using function: %s (%d.%03d MB/sec)\n",
139 fastest->name, fastest->speed / 1000, fastest->speed % 1000);
141 #undef xor_speed
143 free_pages((unsigned long)b1, 2);
145 active_template = fastest;
146 return 0;
149 static __exit void xor_exit(void) { }
151 MODULE_LICENSE("GPL");
153 /* when built-in xor.o must initialize before drivers/md/md.o */
154 core_initcall(calibrate_xor_blocks);
155 module_exit(xor_exit);