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
|
/*
* This file is subject to the terms and conditions of the GNU General Public
* License. See the file "COPYING" in the main directory of this archive
* for more details.
*
* Copyright (C) 2002, 2004 by Ralf Baechle
*/
#ifndef _ASM_WAR_H
#define _ASM_WAR_H
/*
* Another R4600 erratum. Due to the lack of errata information the exact
* technical details aren't known. I've experimentally found that disabling
* interrupts during indexed I-cache flushes seems to be sufficient to deal
* with the issue.
*
* #define R4600_V1_INDEX_ICACHEOP_WAR 1
*/
/*
* Pleasures of the R4600 V1.x. Cite from the IDT R4600 V1.7 errata:
*
* 18. The CACHE instructions Hit_Writeback_Invalidate_D, Hit_Writeback_D,
* Hit_Invalidate_D and Create_Dirty_Excl_D should only be
* executed if there is no other dcache activity. If the dcache is
* accessed for another instruction immeidately preceding when these
* cache instructions are executing, it is possible that the dcache
* tag match outputs used by these cache instructions will be
* incorrect. These cache instructions should be preceded by at least
* four instructions that are not any kind of load or store
* instruction.
*
* This is not allowed: lw
* nop
* nop
* nop
* cache Hit_Writeback_Invalidate_D
*
* This is allowed: lw
* nop
* nop
* nop
* nop
* cache Hit_Writeback_Invalidate_D
*
* #define R4600_V1_HIT_CACHEOP_WAR 1
*/
/*
* Writeback and invalidate the primary cache dcache before DMA.
*
* R4600 v2.0 bug: "The CACHE instructions Hit_Writeback_Inv_D,
* Hit_Writeback_D, Hit_Invalidate_D and Create_Dirty_Exclusive_D will only
* operate correctly if the internal data cache refill buffer is empty. These
* CACHE instructions should be separated from any potential data cache miss
* by a load instruction to an uncached address to empty the response buffer."
* (Revision 2.0 device errata from IDT available on http://www.idt.com/
* in .pdf format.)
*
* #define R4600_V2_HIT_CACHEOP_WAR 1
*/
/*
* R4600 CPU modules for the Indy come with both V1.7 and V2.0 processors.
*/
#ifdef CONFIG_SGI_IP22
#define R4600_V1_INDEX_ICACHEOP_WAR 1
#define R4600_V1_HIT_CACHEOP_WAR 1
#define R4600_V2_HIT_CACHEOP_WAR 1
#endif
/*
* But the RM200C seems to have been shipped only with V2.0 R4600s
*/
#ifdef CONFIG_SNI_RM200_PCI
#define R4600_V2_HIT_CACHEOP_WAR 1
#endif
#ifdef CONFIG_CPU_R5432
/*
* When an interrupt happens on a CP0 register read instruction, CPU may
* lock up or read corrupted values of CP0 registers after it enters
* the exception handler.
*
* This workaround makes sure that we read a "safe" CP0 register as the
* first thing in the exception handler, which breaks one of the
* pre-conditions for this problem.
*/
#define R5432_CP0_INTERRUPT_WAR 1
#endif
#if defined(CONFIG_SB1_PASS_1_WORKAROUNDS) || \
defined(CONFIG_SB1_PASS_2_WORKAROUNDS)
/*
* Workaround for the Sibyte M3 errata the text of which can be found at
*
* http://sibyte.broadcom.com/hw/bcm1250/docs/pass2errata.txt
*
* This will enable the use of a special TLB refill handler which does a
* consistency check on the information in c0_badvaddr and c0_entryhi and
* will just return and take the exception again if the information was
* found to be inconsistent.
*/
#define BCM1250_M3_WAR 1
/*
* This is a DUART workaround related to glitches around register accesses
*/
#define SIBYTE_1956_WAR 1
#endif
/*
* Fill buffers not flushed on CACHE instructions
*
* Hit_Invalidate_I cacheops invalidate an icache line but the refill
* for that line can get stale data from the fill buffer instead of
* accessing memory if the previous icache miss was also to that line.
*
* Workaround: generate an icache refill from a different line
*
* Affects:
* MIPS 4K RTL revision <3.0, PRID revision <4
*/
#if defined(CONFIG_MIPS_MALTA) || defined(CONFIG_MIPS_ATLAS) || \
defined(CONFIG_MIPS_SEAD)
#define MIPS4K_ICACHE_REFILL_WAR 1
#endif
/*
* Missing implicit forced flush of evictions caused by CACHE
* instruction
*
* Evictions caused by a CACHE instructions are not forced on to the
* bus. The BIU gives higher priority to fetches than to the data from
* the eviction buffer and no collision detection is performed between
* fetches and pending data from the eviction buffer.
*
* Workaround: Execute a SYNC instruction after the cache instruction
*
* Affects:
* MIPS 5Kc,5Kf RTL revision <2.3, PRID revision <8
* MIPS 20Kc RTL revision <4.0, PRID revision <?
*/
#if defined(CONFIG_MIPS_MALTA) || defined(CONFIG_MIPS_ATLAS) || \
defined(CONFIG_MIPS_SEAD)
#define MIPS_CACHE_SYNC_WAR 1
#endif
/*
* From TX49/H2 manual: "If the instruction (i.e. CACHE) is issued for
* the line which this instruction itself exists, the following
* operation is not guaranteed."
*
* Workaround: do two phase flushing for Index_Invalidate_I
*/
#ifdef CONFIG_CPU_TX49XX
#define TX49XX_ICACHE_INDEX_INV_WAR 1
#endif
/*
* On the RM9000 there is a problem which makes the CreateDirtyExclusive
* cache operation unusable on SMP systems.
*/
#if defined(CONFIG_MOMENCO_JAGUAR_ATX) || defined(CONFIG_PMC_YOSEMITE)
#define RM9000_CDEX_SMP_WAR 1
#endif
/*
* The RM9000 has a bug (though PMC-Sierra opposes it being called that)
* where invalid instructions in the same I-cache line worth of instructions
* being fetched may case spurious exceptions.
*/
#if defined(CONFIG_MOMENCO_JAGUAR_ATX) || defined(CONFIG_MOMENCO_OCELOT_3) || \
defined(CONFIG_PMC_YOSEMITE)
#define ICACHE_REFILLS_WORKAROUND_WAR 1
#endif
/*
* ON the R10000 upto version 2.6 (not sure about 2.7) there is a bug that
* may cause ll / sc and lld / scd sequences to execute non-atomically.
*/
#ifdef CONFIG_SGI_IP27
#define R10000_LLSC_WAR 1
#endif
/*
* Workarounds default to off
*/
#ifndef ICACHE_REFILLS_WORKAROUND_WAR
#define ICACHE_REFILLS_WORKAROUND_WAR 0
#endif
#ifndef R4600_V1_INDEX_ICACHEOP_WAR
#define R4600_V1_INDEX_ICACHEOP_WAR 0
#endif
#ifndef R4600_V1_HIT_CACHEOP_WAR
#define R4600_V1_HIT_CACHEOP_WAR 0
#endif
#ifndef R4600_V2_HIT_CACHEOP_WAR
#define R4600_V2_HIT_CACHEOP_WAR 0
#endif
#ifndef R5432_CP0_INTERRUPT_WAR
#define R5432_CP0_INTERRUPT_WAR 0
#endif
#ifndef BCM1250_M3_WAR
#define BCM1250_M3_WAR 0
#endif
#ifndef SIBYTE_1956_WAR
#define SIBYTE_1956_WAR 0
#endif
#ifndef MIPS4K_ICACHE_REFILL_WAR
#define MIPS4K_ICACHE_REFILL_WAR 0
#endif
#ifndef MIPS_CACHE_SYNC_WAR
#define MIPS_CACHE_SYNC_WAR 0
#endif
#ifndef TX49XX_ICACHE_INDEX_INV_WAR
#define TX49XX_ICACHE_INDEX_INV_WAR 0
#endif
#ifndef RM9000_CDEX_SMP_WAR
#define RM9000_CDEX_SMP_WAR 0
#endif
#ifndef R10000_LLSC_WAR
#define R10000_LLSC_WAR 0
#endif
#endif /* _ASM_WAR_H */
|
