1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * xor.c : Multiple Devices driver for Linux
5 * Copyright (C) 1996, 1997, 1998, 1999, 2000,
6 * Ingo Molnar, Matti Aarnio, Jakub Jelinek, Richard Henderson.
8 * Dispatch optimized RAID-5 checksumming functions.
12 #include <linux/module.h>
13 #include <linux/gfp.h>
14 #include <linux/raid/xor.h>
15 #include <linux/jiffies.h>
16 #include <linux/preempt.h>
19 #ifndef XOR_SELECT_TEMPLATE
20 #define XOR_SELECT_TEMPLATE(x) (x)
23 /* The xor routines to use. */
24 static struct xor_block_template *active_template;
27 xor_blocks(unsigned int src_count, unsigned int bytes, void *dest, void **srcs)
29 unsigned long *p1, *p2, *p3, *p4;
31 p1 = (unsigned long *) srcs[0];
33 active_template->do_2(bytes, dest, p1);
37 p2 = (unsigned long *) srcs[1];
39 active_template->do_3(bytes, dest, p1, p2);
43 p3 = (unsigned long *) srcs[2];
45 active_template->do_4(bytes, dest, p1, p2, p3);
49 p4 = (unsigned long *) srcs[3];
50 active_template->do_5(bytes, dest, p1, p2, p3, p4);
52 EXPORT_SYMBOL(xor_blocks);
54 /* Set of all registered templates. */
55 static struct xor_block_template *__initdata template_list;
58 static void __init do_xor_register(struct xor_block_template *tmpl)
60 tmpl->next = template_list;
64 static int __init register_xor_blocks(void)
66 active_template = XOR_SELECT_TEMPLATE(NULL);
68 if (!active_template) {
69 #define xor_speed do_xor_register
70 // register all the templates and pick the first as the default
73 active_template = template_list;
79 #define BENCH_SIZE 4096
83 do_xor_speed(struct xor_block_template *tmpl, void *b1, void *b2)
87 ktime_t min, start, diff;
89 tmpl->next = template_list;
94 min = (ktime_t)S64_MAX;
95 for (i = 0; i < 3; i++) {
97 for (j = 0; j < REPS; j++) {
98 mb(); /* prevent loop optimzation */
99 tmpl->do_2(BENCH_SIZE, b1, b2);
102 diff = ktime_sub(ktime_get(), start);
109 // bytes/ns == GB/s, multiply by 1000 to get MB/s [not MiB/s]
110 speed = (1000 * REPS * BENCH_SIZE) / (unsigned int)ktime_to_ns(min);
113 pr_info(" %-16s: %5d MB/sec\n", tmpl->name, speed);
117 calibrate_xor_blocks(void)
120 struct xor_block_template *f, *fastest;
122 fastest = XOR_SELECT_TEMPLATE(NULL);
125 printk(KERN_INFO "xor: automatically using best "
126 "checksumming function %-10s\n",
131 b1 = (void *) __get_free_pages(GFP_KERNEL, 2);
133 printk(KERN_WARNING "xor: Yikes! No memory available.\n");
136 b2 = b1 + 2*PAGE_SIZE + BENCH_SIZE;
139 * If this arch/cpu has a short-circuited selection, don't loop through
140 * all the possible functions, just test the best one
143 #define xor_speed(templ) do_xor_speed((templ), b1, b2)
145 printk(KERN_INFO "xor: measuring software checksum speed\n");
146 template_list = NULL;
148 fastest = template_list;
149 for (f = fastest; f; f = f->next)
150 if (f->speed > fastest->speed)
153 pr_info("xor: using function: %s (%d MB/sec)\n",
154 fastest->name, fastest->speed);
158 free_pages((unsigned long)b1, 2);
160 active_template = fastest;
164 static __exit void xor_exit(void) { }
166 MODULE_LICENSE("GPL");
169 /* when built-in xor.o must initialize before drivers/md/md.o */
170 core_initcall(register_xor_blocks);
173 module_init(calibrate_xor_blocks);
174 module_exit(xor_exit);