1 | // SPDX-License-Identifier: GPL-2.0 |
2 | /* |
3 | * Optimized xor_block operation for RAID4/5 |
4 | * |
5 | * Copyright IBM Corp. 2016 |
6 | * Author(s): Martin Schwidefsky <schwidefsky@de.ibm.com> |
7 | */ |
8 | |
9 | #include <linux/types.h> |
10 | #include <linux/export.h> |
11 | #include <linux/raid/xor.h> |
12 | #include <asm/xor.h> |
13 | |
14 | static void xor_xc_2(unsigned long bytes, unsigned long * __restrict p1, |
15 | const unsigned long * __restrict p2) |
16 | { |
17 | asm volatile( |
18 | " larl 1,2f\n" |
19 | " aghi %0,-1\n" |
20 | " jm 3f\n" |
21 | " srlg 0,%0,8\n" |
22 | " ltgr 0,0\n" |
23 | " jz 1f\n" |
24 | "0: xc 0(256,%1),0(%2)\n" |
25 | " la %1,256(%1)\n" |
26 | " la %2,256(%2)\n" |
27 | " brctg 0,0b\n" |
28 | "1: ex %0,0(1)\n" |
29 | " j 3f\n" |
30 | "2: xc 0(1,%1),0(%2)\n" |
31 | "3:\n" |
32 | : : "d" (bytes), "a" (p1), "a" (p2) |
33 | : "0" , "1" , "cc" , "memory" ); |
34 | } |
35 | |
36 | static void xor_xc_3(unsigned long bytes, unsigned long * __restrict p1, |
37 | const unsigned long * __restrict p2, |
38 | const unsigned long * __restrict p3) |
39 | { |
40 | asm volatile( |
41 | " larl 1,2f\n" |
42 | " aghi %0,-1\n" |
43 | " jm 3f\n" |
44 | " srlg 0,%0,8\n" |
45 | " ltgr 0,0\n" |
46 | " jz 1f\n" |
47 | "0: xc 0(256,%1),0(%2)\n" |
48 | " xc 0(256,%1),0(%3)\n" |
49 | " la %1,256(%1)\n" |
50 | " la %2,256(%2)\n" |
51 | " la %3,256(%3)\n" |
52 | " brctg 0,0b\n" |
53 | "1: ex %0,0(1)\n" |
54 | " ex %0,6(1)\n" |
55 | " j 3f\n" |
56 | "2: xc 0(1,%1),0(%2)\n" |
57 | " xc 0(1,%1),0(%3)\n" |
58 | "3:\n" |
59 | : "+d" (bytes), "+a" (p1), "+a" (p2), "+a" (p3) |
60 | : : "0" , "1" , "cc" , "memory" ); |
61 | } |
62 | |
63 | static void xor_xc_4(unsigned long bytes, unsigned long * __restrict p1, |
64 | const unsigned long * __restrict p2, |
65 | const unsigned long * __restrict p3, |
66 | const unsigned long * __restrict p4) |
67 | { |
68 | asm volatile( |
69 | " larl 1,2f\n" |
70 | " aghi %0,-1\n" |
71 | " jm 3f\n" |
72 | " srlg 0,%0,8\n" |
73 | " ltgr 0,0\n" |
74 | " jz 1f\n" |
75 | "0: xc 0(256,%1),0(%2)\n" |
76 | " xc 0(256,%1),0(%3)\n" |
77 | " xc 0(256,%1),0(%4)\n" |
78 | " la %1,256(%1)\n" |
79 | " la %2,256(%2)\n" |
80 | " la %3,256(%3)\n" |
81 | " la %4,256(%4)\n" |
82 | " brctg 0,0b\n" |
83 | "1: ex %0,0(1)\n" |
84 | " ex %0,6(1)\n" |
85 | " ex %0,12(1)\n" |
86 | " j 3f\n" |
87 | "2: xc 0(1,%1),0(%2)\n" |
88 | " xc 0(1,%1),0(%3)\n" |
89 | " xc 0(1,%1),0(%4)\n" |
90 | "3:\n" |
91 | : "+d" (bytes), "+a" (p1), "+a" (p2), "+a" (p3), "+a" (p4) |
92 | : : "0" , "1" , "cc" , "memory" ); |
93 | } |
94 | |
95 | static void xor_xc_5(unsigned long bytes, unsigned long * __restrict p1, |
96 | const unsigned long * __restrict p2, |
97 | const unsigned long * __restrict p3, |
98 | const unsigned long * __restrict p4, |
99 | const unsigned long * __restrict p5) |
100 | { |
101 | asm volatile( |
102 | " larl 1,2f\n" |
103 | " aghi %0,-1\n" |
104 | " jm 3f\n" |
105 | " srlg 0,%0,8\n" |
106 | " ltgr 0,0\n" |
107 | " jz 1f\n" |
108 | "0: xc 0(256,%1),0(%2)\n" |
109 | " xc 0(256,%1),0(%3)\n" |
110 | " xc 0(256,%1),0(%4)\n" |
111 | " xc 0(256,%1),0(%5)\n" |
112 | " la %1,256(%1)\n" |
113 | " la %2,256(%2)\n" |
114 | " la %3,256(%3)\n" |
115 | " la %4,256(%4)\n" |
116 | " la %5,256(%5)\n" |
117 | " brctg 0,0b\n" |
118 | "1: ex %0,0(1)\n" |
119 | " ex %0,6(1)\n" |
120 | " ex %0,12(1)\n" |
121 | " ex %0,18(1)\n" |
122 | " j 3f\n" |
123 | "2: xc 0(1,%1),0(%2)\n" |
124 | " xc 0(1,%1),0(%3)\n" |
125 | " xc 0(1,%1),0(%4)\n" |
126 | " xc 0(1,%1),0(%5)\n" |
127 | "3:\n" |
128 | : "+d" (bytes), "+a" (p1), "+a" (p2), "+a" (p3), "+a" (p4), |
129 | "+a" (p5) |
130 | : : "0" , "1" , "cc" , "memory" ); |
131 | } |
132 | |
133 | struct xor_block_template xor_block_xc = { |
134 | .name = "xc" , |
135 | .do_2 = xor_xc_2, |
136 | .do_3 = xor_xc_3, |
137 | .do_4 = xor_xc_4, |
138 | .do_5 = xor_xc_5, |
139 | }; |
140 | EXPORT_SYMBOL(xor_block_xc); |
141 | |