1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
|
/*
* Instruction-patching support.
*
* Copyright (C) 2003 Hewlett-Packard Co
* David Mosberger-Tang <davidm@hpl.hp.com>
*/
#include <linux/init.h>
#include <linux/string.h>
#include <asm/paravirt.h>
#include <asm/patch.h>
#include <asm/processor.h>
#include <asm/sections.h>
#include <asm/system.h>
#include <asm/unistd.h>
/*
* This was adapted from code written by Tony Luck:
*
* The 64-bit value in a "movl reg=value" is scattered between the two words of the bundle
* like this:
*
* 6 6 5 4 3 2 1
* 3210987654321098765432109876543210987654321098765432109876543210
* ABBBBBBBBBBBBBBBBBBBBBBBCCCCCCCCCCCCCCCCCCDEEEEEFFFFFFFFFGGGGGGG
*
* CCCCCCCCCCCCCCCCCCxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
* xxxxAFFFFFFFFFEEEEEDxGGGGGGGxxxxxxxxxxxxxBBBBBBBBBBBBBBBBBBBBBBB
*/
static u64
get_imm64 (u64 insn_addr)
{
u64 *p = (u64 *) (insn_addr & -16); /* mask out slot number */
return ( (p[1] & 0x0800000000000000UL) << 4) | /*A*/
((p[1] & 0x00000000007fffffUL) << 40) | /*B*/
((p[0] & 0xffffc00000000000UL) >> 24) | /*C*/
((p[1] & 0x0000100000000000UL) >> 23) | /*D*/
((p[1] & 0x0003e00000000000UL) >> 29) | /*E*/
((p[1] & 0x07fc000000000000UL) >> 43) | /*F*/
((p[1] & 0x000007f000000000UL) >> 36); /*G*/
}
/* Patch instruction with "val" where "mask" has 1 bits. */
void
ia64_patch (u64 insn_addr, u64 mask, u64 val)
{
u64 m0, m1, v0, v1, b0, b1, *b = (u64 *) (insn_addr & -16);
# define insn_mask ((1UL << 41) - 1)
unsigned long shift;
b0 = b[0]; b1 = b[1];
shift = 5 + 41 * (insn_addr % 16); /* 5 bits of template, then 3 x 41-bit instructions */
if (shift >= 64) {
m1 = mask << (shift - 64);
v1 = val << (shift - 64);
} else {
m0 = mask << shift; m1 = mask >> (64 - shift);
v0 = val << shift; v1 = val >> (64 - shift);
b[0] = (b0 & ~m0) | (v0 & m0);
}
b[1] = (b1 & ~m1) | (v1 & m1);
}
void
ia64_patch_imm64 (u64 insn_addr, u64 val)
{
/* The assembler may generate offset pointing to either slot 1
or slot 2 for a long (2-slot) instruction, occupying slots 1
and 2. */
insn_addr &= -16UL;
ia64_patch(insn_addr + 2,
0x01fffefe000UL, ( ((val & 0x8000000000000000UL) >> 27) /* bit 63 -> 36 */
| ((val & 0x0000000000200000UL) << 0) /* bit 21 -> 21 */
| ((val & 0x00000000001f0000UL) << 6) /* bit 16 -> 22 */
| ((val & 0x000000000000ff80UL) << 20) /* bit 7 -> 27 */
| ((val & 0x000000000000007fUL) << 13) /* bit 0 -> 13 */));
ia64_patch(insn_addr + 1, 0x1ffffffffffUL, val >> 22);
}
void
ia64_patch_imm60 (u64 insn_addr, u64 val)
{
/* The assembler may generate offset pointing to either slot 1
or slot 2 for a long (2-slot) instruction, occupying slots 1
and 2. */
insn_addr &= -16UL;
ia64_patch(insn_addr + 2,
0x011ffffe000UL, ( ((val & 0x0800000000000000UL) >> 23) /* bit 59 -> 36 */
| ((val & 0x00000000000fffffUL) << 13) /* bit 0 -> 13 */));
ia64_patch(insn_addr + 1, 0x1fffffffffcUL, val >> 18);
}
/*
* We need sometimes to load the physical address of a kernel
* object. Often we can convert the virtual address to physical
* at execution time, but sometimes (either for performance reasons
* or during error recovery) we cannot to this. Patch the marked
* bundles to load the physical address.
*/
void __init
ia64_patch_vtop (unsigned long start, unsigned long end)
{
s32 *offp = (s32 *) start;
u64 ip;
while (offp < (s32 *) end) {
ip = (u64) offp + *offp;
/* replace virtual address with corresponding physical address: */
ia64_patch_imm64(ip, ia64_tpa(get_imm64(ip)));
ia64_fc((void *) ip);
++offp;
}
ia64_sync_i();
ia64_srlz_i();
}
/*
* Disable the RSE workaround by turning the conditional branch
* that we tagged in each place the workaround was used into an
* unconditional branch.
*/
void __init
ia64_patch_rse (unsigned long start, unsigned long end)
{
s32 *offp = (s32 *) start;
u64 ip, *b;
while (offp < (s32 *) end) {
ip = (u64) offp + *offp;
b = (u64 *)(ip & -16);
b[1] &= ~0xf800000L;
ia64_fc((void *) ip);
++offp;
}
ia64_sync_i();
ia64_srlz_i();
}
void __init
ia64_patch_mckinley_e9 (unsigned long start, unsigned long end)
{
static int first_time = 1;
int need_workaround;
s32 *offp = (s32 *) start;
u64 *wp;
need_workaround = (local_cpu_data->family == 0x1f && local_cpu_data->model == 0);
if (first_time) {
first_time = 0;
if (need_workaround)
printk(KERN_INFO "Leaving McKinley Errata 9 workaround enabled\n");
}
if (need_workaround)
return;
while (offp < (s32 *) end) {
wp = (u64 *) ia64_imva((char *) offp + *offp);
wp[0] = 0x0000000100000011UL; /* nop.m 0; nop.i 0; br.ret.sptk.many b6 */
wp[1] = 0x0084006880000200UL;
wp[2] = 0x0000000100000000UL; /* nop.m 0; nop.i 0; nop.i 0 */
wp[3] = 0x0004000000000200UL;
ia64_fc(wp); ia64_fc(wp + 2);
++offp;
}
ia64_sync_i();
ia64_srlz_i();
}
extern unsigned long ia64_native_fsyscall_table[NR_syscalls];
extern char ia64_native_fsys_bubble_down[];
struct pv_fsys_data pv_fsys_data __initdata = {
.fsyscall_table = (unsigned long *)ia64_native_fsyscall_table,
.fsys_bubble_down = (void *)ia64_native_fsys_bubble_down,
};
unsigned long * __init
paravirt_get_fsyscall_table(void)
{
return pv_fsys_data.fsyscall_table;
}
char * __init
paravirt_get_fsys_bubble_down(void)
{
return pv_fsys_data.fsys_bubble_down;
}
static void __init
patch_fsyscall_table (unsigned long start, unsigned long end)
{
u64 fsyscall_table = (u64)paravirt_get_fsyscall_table();
s32 *offp = (s32 *) start;
u64 ip;
while (offp < (s32 *) end) {
ip = (u64) ia64_imva((char *) offp + *offp);
ia64_patch_imm64(ip, fsyscall_table);
ia64_fc((void *) ip);
++offp;
}
ia64_sync_i();
ia64_srlz_i();
}
static void __init
patch_brl_fsys_bubble_down (unsigned long start, unsigned long end)
{
u64 fsys_bubble_down = (u64)paravirt_get_fsys_bubble_down();
s32 *offp = (s32 *) start;
u64 ip;
while (offp < (s32 *) end) {
ip = (u64) offp + *offp;
ia64_patch_imm60((u64) ia64_imva((void *) ip),
(u64) (fsys_bubble_down - (ip & -16)) / 16);
ia64_fc((void *) ip);
++offp;
}
ia64_sync_i();
ia64_srlz_i();
}
void __init
ia64_patch_gate (void)
{
# define START(name) ((unsigned long) __start_gate_##name##_patchlist)
# define END(name) ((unsigned long)__end_gate_##name##_patchlist)
patch_fsyscall_table(START(fsyscall), END(fsyscall));
patch_brl_fsys_bubble_down(START(brl_fsys_bubble_down), END(brl_fsys_bubble_down));
ia64_patch_vtop(START(vtop), END(vtop));
ia64_patch_mckinley_e9(START(mckinley_e9), END(mckinley_e9));
}
void ia64_patch_phys_stack_reg(unsigned long val)
{
s32 * offp = (s32 *) __start___phys_stack_reg_patchlist;
s32 * end = (s32 *) __end___phys_stack_reg_patchlist;
u64 ip, mask, imm;
/* see instruction format A4: adds r1 = imm13, r3 */
mask = (0x3fUL << 27) | (0x7f << 13);
imm = (((val >> 7) & 0x3f) << 27) | (val & 0x7f) << 13;
while (offp < end) {
ip = (u64) offp + *offp;
ia64_patch(ip, mask, imm);
ia64_fc(ip);
++offp;
}
ia64_sync_i();
ia64_srlz_i();
}
|
