Commit | Line | Data |
---|---|---|
52707f91 MCC |
1 | /* Intel i7 core/Nehalem Memory Controller kernel module |
2 | * | |
3 | * This driver supports yhe memory controllers found on the Intel | |
4 | * processor families i7core, i7core 7xx/8xx, i5core, Xeon 35xx, | |
5 | * Xeon 55xx and Xeon 56xx also known as Nehalem, Nehalem-EP, Lynnfield | |
6 | * and Westmere-EP. | |
a0c36a1f MCC |
7 | * |
8 | * This file may be distributed under the terms of the | |
9 | * GNU General Public License version 2 only. | |
10 | * | |
52707f91 | 11 | * Copyright (c) 2009-2010 by: |
a0c36a1f MCC |
12 | * Mauro Carvalho Chehab <mchehab@redhat.com> |
13 | * | |
14 | * Red Hat Inc. http://www.redhat.com | |
15 | * | |
16 | * Forked and adapted from the i5400_edac driver | |
17 | * | |
18 | * Based on the following public Intel datasheets: | |
19 | * Intel Core i7 Processor Extreme Edition and Intel Core i7 Processor | |
20 | * Datasheet, Volume 2: | |
21 | * http://download.intel.com/design/processor/datashts/320835.pdf | |
22 | * Intel Xeon Processor 5500 Series Datasheet Volume 2 | |
23 | * http://www.intel.com/Assets/PDF/datasheet/321322.pdf | |
24 | * also available at: | |
25 | * http://www.arrownac.com/manufacturers/intel/s/nehalem/5500-datasheet-v2.pdf | |
26 | */ | |
27 | ||
a0c36a1f MCC |
28 | #include <linux/module.h> |
29 | #include <linux/init.h> | |
30 | #include <linux/pci.h> | |
31 | #include <linux/pci_ids.h> | |
32 | #include <linux/slab.h> | |
3b918c12 | 33 | #include <linux/delay.h> |
a0c36a1f MCC |
34 | #include <linux/edac.h> |
35 | #include <linux/mmzone.h> | |
d5381642 | 36 | #include <linux/edac_mce.h> |
f4742949 | 37 | #include <linux/smp.h> |
14d2c083 | 38 | #include <asm/processor.h> |
a0c36a1f MCC |
39 | |
40 | #include "edac_core.h" | |
41 | ||
18c29002 MCC |
42 | /* Static vars */ |
43 | static LIST_HEAD(i7core_edac_list); | |
44 | static DEFINE_MUTEX(i7core_edac_lock); | |
45 | static int probed; | |
46 | ||
54a08ab1 MCC |
47 | static int use_pci_fixup; |
48 | module_param(use_pci_fixup, int, 0444); | |
49 | MODULE_PARM_DESC(use_pci_fixup, "Enable PCI fixup to seek for hidden devices"); | |
f4742949 MCC |
50 | /* |
51 | * This is used for Nehalem-EP and Nehalem-EX devices, where the non-core | |
52 | * registers start at bus 255, and are not reported by BIOS. | |
53 | * We currently find devices with only 2 sockets. In order to support more QPI | |
54 | * Quick Path Interconnect, just increment this number. | |
55 | */ | |
56 | #define MAX_SOCKET_BUSES 2 | |
57 | ||
58 | ||
a0c36a1f MCC |
59 | /* |
60 | * Alter this version for the module when modifications are made | |
61 | */ | |
62 | #define I7CORE_REVISION " Ver: 1.0.0 " __DATE__ | |
63 | #define EDAC_MOD_STR "i7core_edac" | |
64 | ||
a0c36a1f MCC |
65 | /* |
66 | * Debug macros | |
67 | */ | |
68 | #define i7core_printk(level, fmt, arg...) \ | |
69 | edac_printk(level, "i7core", fmt, ##arg) | |
70 | ||
71 | #define i7core_mc_printk(mci, level, fmt, arg...) \ | |
72 | edac_mc_chipset_printk(mci, level, "i7core", fmt, ##arg) | |
73 | ||
74 | /* | |
75 | * i7core Memory Controller Registers | |
76 | */ | |
77 | ||
e9bd2e73 MCC |
78 | /* OFFSETS for Device 0 Function 0 */ |
79 | ||
80 | #define MC_CFG_CONTROL 0x90 | |
81 | ||
a0c36a1f MCC |
82 | /* OFFSETS for Device 3 Function 0 */ |
83 | ||
84 | #define MC_CONTROL 0x48 | |
85 | #define MC_STATUS 0x4c | |
86 | #define MC_MAX_DOD 0x64 | |
87 | ||
442305b1 MCC |
88 | /* |
89 | * OFFSETS for Device 3 Function 4, as inicated on Xeon 5500 datasheet: | |
90 | * http://www.arrownac.com/manufacturers/intel/s/nehalem/5500-datasheet-v2.pdf | |
91 | */ | |
92 | ||
93 | #define MC_TEST_ERR_RCV1 0x60 | |
94 | #define DIMM2_COR_ERR(r) ((r) & 0x7fff) | |
95 | ||
96 | #define MC_TEST_ERR_RCV0 0x64 | |
97 | #define DIMM1_COR_ERR(r) (((r) >> 16) & 0x7fff) | |
98 | #define DIMM0_COR_ERR(r) ((r) & 0x7fff) | |
99 | ||
b4e8f0b6 MCC |
100 | /* OFFSETS for Device 3 Function 2, as inicated on Xeon 5500 datasheet */ |
101 | #define MC_COR_ECC_CNT_0 0x80 | |
102 | #define MC_COR_ECC_CNT_1 0x84 | |
103 | #define MC_COR_ECC_CNT_2 0x88 | |
104 | #define MC_COR_ECC_CNT_3 0x8c | |
105 | #define MC_COR_ECC_CNT_4 0x90 | |
106 | #define MC_COR_ECC_CNT_5 0x94 | |
107 | ||
108 | #define DIMM_TOP_COR_ERR(r) (((r) >> 16) & 0x7fff) | |
109 | #define DIMM_BOT_COR_ERR(r) ((r) & 0x7fff) | |
110 | ||
111 | ||
a0c36a1f MCC |
112 | /* OFFSETS for Devices 4,5 and 6 Function 0 */ |
113 | ||
0b2b7b7e MCC |
114 | #define MC_CHANNEL_DIMM_INIT_PARAMS 0x58 |
115 | #define THREE_DIMMS_PRESENT (1 << 24) | |
116 | #define SINGLE_QUAD_RANK_PRESENT (1 << 23) | |
117 | #define QUAD_RANK_PRESENT (1 << 22) | |
118 | #define REGISTERED_DIMM (1 << 15) | |
119 | ||
f122a892 MCC |
120 | #define MC_CHANNEL_MAPPER 0x60 |
121 | #define RDLCH(r, ch) ((((r) >> (3 + (ch * 6))) & 0x07) - 1) | |
122 | #define WRLCH(r, ch) ((((r) >> (ch * 6)) & 0x07) - 1) | |
123 | ||
0b2b7b7e MCC |
124 | #define MC_CHANNEL_RANK_PRESENT 0x7c |
125 | #define RANK_PRESENT_MASK 0xffff | |
126 | ||
a0c36a1f | 127 | #define MC_CHANNEL_ADDR_MATCH 0xf0 |
194a40fe MCC |
128 | #define MC_CHANNEL_ERROR_MASK 0xf8 |
129 | #define MC_CHANNEL_ERROR_INJECT 0xfc | |
130 | #define INJECT_ADDR_PARITY 0x10 | |
131 | #define INJECT_ECC 0x08 | |
132 | #define MASK_CACHELINE 0x06 | |
133 | #define MASK_FULL_CACHELINE 0x06 | |
134 | #define MASK_MSB32_CACHELINE 0x04 | |
135 | #define MASK_LSB32_CACHELINE 0x02 | |
136 | #define NO_MASK_CACHELINE 0x00 | |
137 | #define REPEAT_EN 0x01 | |
a0c36a1f | 138 | |
0b2b7b7e | 139 | /* OFFSETS for Devices 4,5 and 6 Function 1 */ |
b990538a | 140 | |
0b2b7b7e MCC |
141 | #define MC_DOD_CH_DIMM0 0x48 |
142 | #define MC_DOD_CH_DIMM1 0x4c | |
143 | #define MC_DOD_CH_DIMM2 0x50 | |
144 | #define RANKOFFSET_MASK ((1 << 12) | (1 << 11) | (1 << 10)) | |
145 | #define RANKOFFSET(x) ((x & RANKOFFSET_MASK) >> 10) | |
146 | #define DIMM_PRESENT_MASK (1 << 9) | |
147 | #define DIMM_PRESENT(x) (((x) & DIMM_PRESENT_MASK) >> 9) | |
854d3349 MCC |
148 | #define MC_DOD_NUMBANK_MASK ((1 << 8) | (1 << 7)) |
149 | #define MC_DOD_NUMBANK(x) (((x) & MC_DOD_NUMBANK_MASK) >> 7) | |
150 | #define MC_DOD_NUMRANK_MASK ((1 << 6) | (1 << 5)) | |
151 | #define MC_DOD_NUMRANK(x) (((x) & MC_DOD_NUMRANK_MASK) >> 5) | |
41fcb7fe | 152 | #define MC_DOD_NUMROW_MASK ((1 << 4) | (1 << 3) | (1 << 2)) |
5566cb7c | 153 | #define MC_DOD_NUMROW(x) (((x) & MC_DOD_NUMROW_MASK) >> 2) |
854d3349 MCC |
154 | #define MC_DOD_NUMCOL_MASK 3 |
155 | #define MC_DOD_NUMCOL(x) ((x) & MC_DOD_NUMCOL_MASK) | |
0b2b7b7e | 156 | |
f122a892 MCC |
157 | #define MC_RANK_PRESENT 0x7c |
158 | ||
0b2b7b7e MCC |
159 | #define MC_SAG_CH_0 0x80 |
160 | #define MC_SAG_CH_1 0x84 | |
161 | #define MC_SAG_CH_2 0x88 | |
162 | #define MC_SAG_CH_3 0x8c | |
163 | #define MC_SAG_CH_4 0x90 | |
164 | #define MC_SAG_CH_5 0x94 | |
165 | #define MC_SAG_CH_6 0x98 | |
166 | #define MC_SAG_CH_7 0x9c | |
167 | ||
168 | #define MC_RIR_LIMIT_CH_0 0x40 | |
169 | #define MC_RIR_LIMIT_CH_1 0x44 | |
170 | #define MC_RIR_LIMIT_CH_2 0x48 | |
171 | #define MC_RIR_LIMIT_CH_3 0x4C | |
172 | #define MC_RIR_LIMIT_CH_4 0x50 | |
173 | #define MC_RIR_LIMIT_CH_5 0x54 | |
174 | #define MC_RIR_LIMIT_CH_6 0x58 | |
175 | #define MC_RIR_LIMIT_CH_7 0x5C | |
176 | #define MC_RIR_LIMIT_MASK ((1 << 10) - 1) | |
177 | ||
178 | #define MC_RIR_WAY_CH 0x80 | |
179 | #define MC_RIR_WAY_OFFSET_MASK (((1 << 14) - 1) & ~0x7) | |
180 | #define MC_RIR_WAY_RANK_MASK 0x7 | |
181 | ||
a0c36a1f MCC |
182 | /* |
183 | * i7core structs | |
184 | */ | |
185 | ||
186 | #define NUM_CHANS 3 | |
442305b1 MCC |
187 | #define MAX_DIMMS 3 /* Max DIMMS per channel */ |
188 | #define MAX_MCR_FUNC 4 | |
189 | #define MAX_CHAN_FUNC 3 | |
a0c36a1f MCC |
190 | |
191 | struct i7core_info { | |
192 | u32 mc_control; | |
193 | u32 mc_status; | |
194 | u32 max_dod; | |
f122a892 | 195 | u32 ch_map; |
a0c36a1f MCC |
196 | }; |
197 | ||
194a40fe MCC |
198 | |
199 | struct i7core_inject { | |
200 | int enable; | |
201 | ||
202 | u32 section; | |
203 | u32 type; | |
204 | u32 eccmask; | |
205 | ||
206 | /* Error address mask */ | |
207 | int channel, dimm, rank, bank, page, col; | |
208 | }; | |
209 | ||
0b2b7b7e | 210 | struct i7core_channel { |
442305b1 MCC |
211 | u32 ranks; |
212 | u32 dimms; | |
0b2b7b7e MCC |
213 | }; |
214 | ||
8f331907 | 215 | struct pci_id_descr { |
66607706 MCC |
216 | int dev; |
217 | int func; | |
218 | int dev_id; | |
de06eeef | 219 | int optional; |
8f331907 MCC |
220 | }; |
221 | ||
bd9e19ca | 222 | struct pci_id_table { |
1288c18f MCC |
223 | const struct pci_id_descr *descr; |
224 | int n_devs; | |
bd9e19ca VM |
225 | }; |
226 | ||
f4742949 MCC |
227 | struct i7core_dev { |
228 | struct list_head list; | |
229 | u8 socket; | |
230 | struct pci_dev **pdev; | |
de06eeef | 231 | int n_devs; |
f4742949 MCC |
232 | struct mem_ctl_info *mci; |
233 | }; | |
234 | ||
a0c36a1f | 235 | struct i7core_pvt { |
f4742949 MCC |
236 | struct pci_dev *pci_noncore; |
237 | struct pci_dev *pci_mcr[MAX_MCR_FUNC + 1]; | |
238 | struct pci_dev *pci_ch[NUM_CHANS][MAX_CHAN_FUNC + 1]; | |
239 | ||
240 | struct i7core_dev *i7core_dev; | |
67166af4 | 241 | |
a0c36a1f | 242 | struct i7core_info info; |
194a40fe | 243 | struct i7core_inject inject; |
f4742949 | 244 | struct i7core_channel channel[NUM_CHANS]; |
67166af4 | 245 | |
f4742949 MCC |
246 | int ce_count_available; |
247 | int csrow_map[NUM_CHANS][MAX_DIMMS]; | |
b4e8f0b6 MCC |
248 | |
249 | /* ECC corrected errors counts per udimm */ | |
f4742949 MCC |
250 | unsigned long udimm_ce_count[MAX_DIMMS]; |
251 | int udimm_last_ce_count[MAX_DIMMS]; | |
b4e8f0b6 | 252 | /* ECC corrected errors counts per rdimm */ |
f4742949 MCC |
253 | unsigned long rdimm_ce_count[NUM_CHANS][MAX_DIMMS]; |
254 | int rdimm_last_ce_count[NUM_CHANS][MAX_DIMMS]; | |
442305b1 | 255 | |
f4742949 | 256 | unsigned int is_registered; |
14d2c083 | 257 | |
d5381642 MCC |
258 | /* mcelog glue */ |
259 | struct edac_mce edac_mce; | |
ca9c90ba MCC |
260 | |
261 | /* Fifo double buffers */ | |
d5381642 | 262 | struct mce mce_entry[MCE_LOG_LEN]; |
ca9c90ba MCC |
263 | struct mce mce_outentry[MCE_LOG_LEN]; |
264 | ||
265 | /* Fifo in/out counters */ | |
266 | unsigned mce_in, mce_out; | |
267 | ||
268 | /* Count indicator to show errors not got */ | |
269 | unsigned mce_overrun; | |
939747bd MCC |
270 | |
271 | /* Struct to control EDAC polling */ | |
272 | struct edac_pci_ctl_info *i7core_pci; | |
a0c36a1f MCC |
273 | }; |
274 | ||
8f331907 MCC |
275 | #define PCI_DESCR(device, function, device_id) \ |
276 | .dev = (device), \ | |
277 | .func = (function), \ | |
278 | .dev_id = (device_id) | |
279 | ||
1288c18f | 280 | static const struct pci_id_descr pci_dev_descr_i7core_nehalem[] = { |
8f331907 MCC |
281 | /* Memory controller */ |
282 | { PCI_DESCR(3, 0, PCI_DEVICE_ID_INTEL_I7_MCR) }, | |
283 | { PCI_DESCR(3, 1, PCI_DEVICE_ID_INTEL_I7_MC_TAD) }, | |
de06eeef MCC |
284 | /* Exists only for RDIMM */ |
285 | { PCI_DESCR(3, 2, PCI_DEVICE_ID_INTEL_I7_MC_RAS), .optional = 1 }, | |
8f331907 MCC |
286 | { PCI_DESCR(3, 4, PCI_DEVICE_ID_INTEL_I7_MC_TEST) }, |
287 | ||
288 | /* Channel 0 */ | |
289 | { PCI_DESCR(4, 0, PCI_DEVICE_ID_INTEL_I7_MC_CH0_CTRL) }, | |
290 | { PCI_DESCR(4, 1, PCI_DEVICE_ID_INTEL_I7_MC_CH0_ADDR) }, | |
291 | { PCI_DESCR(4, 2, PCI_DEVICE_ID_INTEL_I7_MC_CH0_RANK) }, | |
292 | { PCI_DESCR(4, 3, PCI_DEVICE_ID_INTEL_I7_MC_CH0_TC) }, | |
293 | ||
294 | /* Channel 1 */ | |
295 | { PCI_DESCR(5, 0, PCI_DEVICE_ID_INTEL_I7_MC_CH1_CTRL) }, | |
296 | { PCI_DESCR(5, 1, PCI_DEVICE_ID_INTEL_I7_MC_CH1_ADDR) }, | |
297 | { PCI_DESCR(5, 2, PCI_DEVICE_ID_INTEL_I7_MC_CH1_RANK) }, | |
298 | { PCI_DESCR(5, 3, PCI_DEVICE_ID_INTEL_I7_MC_CH1_TC) }, | |
299 | ||
300 | /* Channel 2 */ | |
301 | { PCI_DESCR(6, 0, PCI_DEVICE_ID_INTEL_I7_MC_CH2_CTRL) }, | |
302 | { PCI_DESCR(6, 1, PCI_DEVICE_ID_INTEL_I7_MC_CH2_ADDR) }, | |
303 | { PCI_DESCR(6, 2, PCI_DEVICE_ID_INTEL_I7_MC_CH2_RANK) }, | |
304 | { PCI_DESCR(6, 3, PCI_DEVICE_ID_INTEL_I7_MC_CH2_TC) }, | |
310cbb72 MCC |
305 | |
306 | /* Generic Non-core registers */ | |
307 | /* | |
308 | * This is the PCI device on i7core and on Xeon 35xx (8086:2c41) | |
309 | * On Xeon 55xx, however, it has a different id (8086:2c40). So, | |
310 | * the probing code needs to test for the other address in case of | |
311 | * failure of this one | |
312 | */ | |
fd382654 | 313 | { PCI_DESCR(0, 0, PCI_DEVICE_ID_INTEL_I7_NONCORE) }, |
310cbb72 | 314 | |
a0c36a1f | 315 | }; |
8f331907 | 316 | |
1288c18f | 317 | static const struct pci_id_descr pci_dev_descr_lynnfield[] = { |
52a2e4fc MCC |
318 | { PCI_DESCR( 3, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MCR) }, |
319 | { PCI_DESCR( 3, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_TAD) }, | |
320 | { PCI_DESCR( 3, 4, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_TEST) }, | |
321 | ||
322 | { PCI_DESCR( 4, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_CTRL) }, | |
323 | { PCI_DESCR( 4, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_ADDR) }, | |
324 | { PCI_DESCR( 4, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_RANK) }, | |
325 | { PCI_DESCR( 4, 3, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_TC) }, | |
326 | ||
508fa179 MCC |
327 | { PCI_DESCR( 5, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_CTRL) }, |
328 | { PCI_DESCR( 5, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_ADDR) }, | |
329 | { PCI_DESCR( 5, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_RANK) }, | |
330 | { PCI_DESCR( 5, 3, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_TC) }, | |
52a2e4fc | 331 | |
f05da2f7 MCC |
332 | /* |
333 | * This is the PCI device has an alternate address on some | |
334 | * processors like Core i7 860 | |
335 | */ | |
52a2e4fc MCC |
336 | { PCI_DESCR( 0, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE) }, |
337 | }; | |
338 | ||
1288c18f | 339 | static const struct pci_id_descr pci_dev_descr_i7core_westmere[] = { |
bd9e19ca VM |
340 | /* Memory controller */ |
341 | { PCI_DESCR(3, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MCR_REV2) }, | |
342 | { PCI_DESCR(3, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_TAD_REV2) }, | |
343 | /* Exists only for RDIMM */ | |
344 | { PCI_DESCR(3, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_RAS_REV2), .optional = 1 }, | |
345 | { PCI_DESCR(3, 4, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_TEST_REV2) }, | |
346 | ||
347 | /* Channel 0 */ | |
348 | { PCI_DESCR(4, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_CTRL_REV2) }, | |
349 | { PCI_DESCR(4, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_ADDR_REV2) }, | |
350 | { PCI_DESCR(4, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_RANK_REV2) }, | |
351 | { PCI_DESCR(4, 3, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_TC_REV2) }, | |
352 | ||
353 | /* Channel 1 */ | |
354 | { PCI_DESCR(5, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_CTRL_REV2) }, | |
355 | { PCI_DESCR(5, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_ADDR_REV2) }, | |
356 | { PCI_DESCR(5, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_RANK_REV2) }, | |
357 | { PCI_DESCR(5, 3, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_TC_REV2) }, | |
358 | ||
359 | /* Channel 2 */ | |
360 | { PCI_DESCR(6, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH2_CTRL_REV2) }, | |
361 | { PCI_DESCR(6, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH2_ADDR_REV2) }, | |
362 | { PCI_DESCR(6, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH2_RANK_REV2) }, | |
363 | { PCI_DESCR(6, 3, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH2_TC_REV2) }, | |
364 | ||
365 | /* Generic Non-core registers */ | |
366 | { PCI_DESCR(0, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE_REV2) }, | |
367 | ||
368 | }; | |
369 | ||
1288c18f MCC |
370 | #define PCI_ID_TABLE_ENTRY(A) { .descr=A, .n_devs = ARRAY_SIZE(A) } |
371 | static const struct pci_id_table pci_dev_table[] = { | |
bd9e19ca VM |
372 | PCI_ID_TABLE_ENTRY(pci_dev_descr_i7core_nehalem), |
373 | PCI_ID_TABLE_ENTRY(pci_dev_descr_lynnfield), | |
374 | PCI_ID_TABLE_ENTRY(pci_dev_descr_i7core_westmere), | |
375 | }; | |
376 | ||
8f331907 MCC |
377 | /* |
378 | * pci_device_id table for which devices we are looking for | |
8f331907 MCC |
379 | */ |
380 | static const struct pci_device_id i7core_pci_tbl[] __devinitdata = { | |
d1fd4fb6 | 381 | {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_X58_HUB_MGMT)}, |
f05da2f7 | 382 | {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_LYNNFIELD_QPI_LINK0)}, |
8f331907 MCC |
383 | {0,} /* 0 terminated list. */ |
384 | }; | |
385 | ||
a0c36a1f MCC |
386 | /**************************************************************************** |
387 | Anciliary status routines | |
388 | ****************************************************************************/ | |
389 | ||
390 | /* MC_CONTROL bits */ | |
ef708b53 MCC |
391 | #define CH_ACTIVE(pvt, ch) ((pvt)->info.mc_control & (1 << (8 + ch))) |
392 | #define ECCx8(pvt) ((pvt)->info.mc_control & (1 << 1)) | |
a0c36a1f MCC |
393 | |
394 | /* MC_STATUS bits */ | |
61053fde | 395 | #define ECC_ENABLED(pvt) ((pvt)->info.mc_status & (1 << 4)) |
ef708b53 | 396 | #define CH_DISABLED(pvt, ch) ((pvt)->info.mc_status & (1 << ch)) |
a0c36a1f MCC |
397 | |
398 | /* MC_MAX_DOD read functions */ | |
854d3349 | 399 | static inline int numdimms(u32 dimms) |
a0c36a1f | 400 | { |
854d3349 | 401 | return (dimms & 0x3) + 1; |
a0c36a1f MCC |
402 | } |
403 | ||
854d3349 | 404 | static inline int numrank(u32 rank) |
a0c36a1f MCC |
405 | { |
406 | static int ranks[4] = { 1, 2, 4, -EINVAL }; | |
407 | ||
854d3349 | 408 | return ranks[rank & 0x3]; |
a0c36a1f MCC |
409 | } |
410 | ||
854d3349 | 411 | static inline int numbank(u32 bank) |
a0c36a1f MCC |
412 | { |
413 | static int banks[4] = { 4, 8, 16, -EINVAL }; | |
414 | ||
854d3349 | 415 | return banks[bank & 0x3]; |
a0c36a1f MCC |
416 | } |
417 | ||
854d3349 | 418 | static inline int numrow(u32 row) |
a0c36a1f MCC |
419 | { |
420 | static int rows[8] = { | |
421 | 1 << 12, 1 << 13, 1 << 14, 1 << 15, | |
422 | 1 << 16, -EINVAL, -EINVAL, -EINVAL, | |
423 | }; | |
424 | ||
854d3349 | 425 | return rows[row & 0x7]; |
a0c36a1f MCC |
426 | } |
427 | ||
854d3349 | 428 | static inline int numcol(u32 col) |
a0c36a1f MCC |
429 | { |
430 | static int cols[8] = { | |
431 | 1 << 10, 1 << 11, 1 << 12, -EINVAL, | |
432 | }; | |
854d3349 | 433 | return cols[col & 0x3]; |
a0c36a1f MCC |
434 | } |
435 | ||
f4742949 | 436 | static struct i7core_dev *get_i7core_dev(u8 socket) |
66607706 MCC |
437 | { |
438 | struct i7core_dev *i7core_dev; | |
439 | ||
440 | list_for_each_entry(i7core_dev, &i7core_edac_list, list) { | |
441 | if (i7core_dev->socket == socket) | |
442 | return i7core_dev; | |
443 | } | |
444 | ||
445 | return NULL; | |
446 | } | |
447 | ||
848b2f7e HS |
448 | static struct i7core_dev *alloc_i7core_dev(u8 socket, |
449 | const struct pci_id_table *table) | |
450 | { | |
451 | struct i7core_dev *i7core_dev; | |
452 | ||
453 | i7core_dev = kzalloc(sizeof(*i7core_dev), GFP_KERNEL); | |
454 | if (!i7core_dev) | |
455 | return NULL; | |
456 | ||
457 | i7core_dev->pdev = kzalloc(sizeof(*i7core_dev->pdev) * table->n_devs, | |
458 | GFP_KERNEL); | |
459 | if (!i7core_dev->pdev) { | |
460 | kfree(i7core_dev); | |
461 | return NULL; | |
462 | } | |
463 | ||
464 | i7core_dev->socket = socket; | |
465 | i7core_dev->n_devs = table->n_devs; | |
466 | list_add_tail(&i7core_dev->list, &i7core_edac_list); | |
467 | ||
468 | return i7core_dev; | |
469 | } | |
470 | ||
2aa9be44 HS |
471 | static void free_i7core_dev(struct i7core_dev *i7core_dev) |
472 | { | |
473 | list_del(&i7core_dev->list); | |
474 | kfree(i7core_dev->pdev); | |
475 | kfree(i7core_dev); | |
476 | } | |
477 | ||
a0c36a1f MCC |
478 | /**************************************************************************** |
479 | Memory check routines | |
480 | ****************************************************************************/ | |
67166af4 MCC |
481 | static struct pci_dev *get_pdev_slot_func(u8 socket, unsigned slot, |
482 | unsigned func) | |
ef708b53 | 483 | { |
66607706 | 484 | struct i7core_dev *i7core_dev = get_i7core_dev(socket); |
ef708b53 | 485 | int i; |
ef708b53 | 486 | |
66607706 MCC |
487 | if (!i7core_dev) |
488 | return NULL; | |
489 | ||
de06eeef | 490 | for (i = 0; i < i7core_dev->n_devs; i++) { |
66607706 | 491 | if (!i7core_dev->pdev[i]) |
ef708b53 MCC |
492 | continue; |
493 | ||
66607706 MCC |
494 | if (PCI_SLOT(i7core_dev->pdev[i]->devfn) == slot && |
495 | PCI_FUNC(i7core_dev->pdev[i]->devfn) == func) { | |
496 | return i7core_dev->pdev[i]; | |
ef708b53 MCC |
497 | } |
498 | } | |
499 | ||
eb94fc40 MCC |
500 | return NULL; |
501 | } | |
502 | ||
ec6df24c MCC |
503 | /** |
504 | * i7core_get_active_channels() - gets the number of channels and csrows | |
505 | * @socket: Quick Path Interconnect socket | |
506 | * @channels: Number of channels that will be returned | |
507 | * @csrows: Number of csrows found | |
508 | * | |
509 | * Since EDAC core needs to know in advance the number of available channels | |
510 | * and csrows, in order to allocate memory for csrows/channels, it is needed | |
511 | * to run two similar steps. At the first step, implemented on this function, | |
512 | * it checks the number of csrows/channels present at one socket. | |
513 | * this is used in order to properly allocate the size of mci components. | |
514 | * | |
515 | * It should be noticed that none of the current available datasheets explain | |
516 | * or even mention how csrows are seen by the memory controller. So, we need | |
517 | * to add a fake description for csrows. | |
518 | * So, this driver is attributing one DIMM memory for one csrow. | |
519 | */ | |
1288c18f | 520 | static int i7core_get_active_channels(const u8 socket, unsigned *channels, |
67166af4 | 521 | unsigned *csrows) |
eb94fc40 MCC |
522 | { |
523 | struct pci_dev *pdev = NULL; | |
524 | int i, j; | |
525 | u32 status, control; | |
526 | ||
527 | *channels = 0; | |
528 | *csrows = 0; | |
529 | ||
67166af4 | 530 | pdev = get_pdev_slot_func(socket, 3, 0); |
b7c76151 | 531 | if (!pdev) { |
67166af4 MCC |
532 | i7core_printk(KERN_ERR, "Couldn't find socket %d fn 3.0!!!\n", |
533 | socket); | |
ef708b53 | 534 | return -ENODEV; |
b7c76151 | 535 | } |
ef708b53 MCC |
536 | |
537 | /* Device 3 function 0 reads */ | |
538 | pci_read_config_dword(pdev, MC_STATUS, &status); | |
539 | pci_read_config_dword(pdev, MC_CONTROL, &control); | |
540 | ||
541 | for (i = 0; i < NUM_CHANS; i++) { | |
eb94fc40 | 542 | u32 dimm_dod[3]; |
ef708b53 MCC |
543 | /* Check if the channel is active */ |
544 | if (!(control & (1 << (8 + i)))) | |
545 | continue; | |
546 | ||
547 | /* Check if the channel is disabled */ | |
41fcb7fe | 548 | if (status & (1 << i)) |
ef708b53 | 549 | continue; |
ef708b53 | 550 | |
67166af4 | 551 | pdev = get_pdev_slot_func(socket, i + 4, 1); |
eb94fc40 | 552 | if (!pdev) { |
67166af4 MCC |
553 | i7core_printk(KERN_ERR, "Couldn't find socket %d " |
554 | "fn %d.%d!!!\n", | |
555 | socket, i + 4, 1); | |
eb94fc40 MCC |
556 | return -ENODEV; |
557 | } | |
558 | /* Devices 4-6 function 1 */ | |
559 | pci_read_config_dword(pdev, | |
560 | MC_DOD_CH_DIMM0, &dimm_dod[0]); | |
561 | pci_read_config_dword(pdev, | |
562 | MC_DOD_CH_DIMM1, &dimm_dod[1]); | |
563 | pci_read_config_dword(pdev, | |
564 | MC_DOD_CH_DIMM2, &dimm_dod[2]); | |
565 | ||
ef708b53 | 566 | (*channels)++; |
eb94fc40 MCC |
567 | |
568 | for (j = 0; j < 3; j++) { | |
569 | if (!DIMM_PRESENT(dimm_dod[j])) | |
570 | continue; | |
571 | (*csrows)++; | |
572 | } | |
ef708b53 MCC |
573 | } |
574 | ||
c77720b9 | 575 | debugf0("Number of active channels on socket %d: %d\n", |
67166af4 | 576 | socket, *channels); |
1c6fed80 | 577 | |
ef708b53 MCC |
578 | return 0; |
579 | } | |
580 | ||
2e5185f7 | 581 | static int get_dimm_config(const struct mem_ctl_info *mci) |
a0c36a1f MCC |
582 | { |
583 | struct i7core_pvt *pvt = mci->pvt_info; | |
1c6fed80 | 584 | struct csrow_info *csr; |
854d3349 | 585 | struct pci_dev *pdev; |
ba6c5c62 | 586 | int i, j; |
2e5185f7 | 587 | int csrow = 0; |
5566cb7c | 588 | unsigned long last_page = 0; |
1c6fed80 | 589 | enum edac_type mode; |
854d3349 | 590 | enum mem_type mtype; |
a0c36a1f | 591 | |
854d3349 | 592 | /* Get data from the MC register, function 0 */ |
f4742949 | 593 | pdev = pvt->pci_mcr[0]; |
7dd6953c | 594 | if (!pdev) |
8f331907 MCC |
595 | return -ENODEV; |
596 | ||
f122a892 | 597 | /* Device 3 function 0 reads */ |
7dd6953c MCC |
598 | pci_read_config_dword(pdev, MC_CONTROL, &pvt->info.mc_control); |
599 | pci_read_config_dword(pdev, MC_STATUS, &pvt->info.mc_status); | |
600 | pci_read_config_dword(pdev, MC_MAX_DOD, &pvt->info.max_dod); | |
601 | pci_read_config_dword(pdev, MC_CHANNEL_MAPPER, &pvt->info.ch_map); | |
f122a892 | 602 | |
17cb7b0c | 603 | debugf0("QPI %d control=0x%08x status=0x%08x dod=0x%08x map=0x%08x\n", |
4af91889 | 604 | pvt->i7core_dev->socket, pvt->info.mc_control, pvt->info.mc_status, |
f122a892 | 605 | pvt->info.max_dod, pvt->info.ch_map); |
a0c36a1f | 606 | |
1c6fed80 | 607 | if (ECC_ENABLED(pvt)) { |
41fcb7fe | 608 | debugf0("ECC enabled with x%d SDCC\n", ECCx8(pvt) ? 8 : 4); |
1c6fed80 MCC |
609 | if (ECCx8(pvt)) |
610 | mode = EDAC_S8ECD8ED; | |
611 | else | |
612 | mode = EDAC_S4ECD4ED; | |
613 | } else { | |
a0c36a1f | 614 | debugf0("ECC disabled\n"); |
1c6fed80 MCC |
615 | mode = EDAC_NONE; |
616 | } | |
a0c36a1f MCC |
617 | |
618 | /* FIXME: need to handle the error codes */ | |
17cb7b0c MCC |
619 | debugf0("DOD Max limits: DIMMS: %d, %d-ranked, %d-banked " |
620 | "x%x x 0x%x\n", | |
854d3349 MCC |
621 | numdimms(pvt->info.max_dod), |
622 | numrank(pvt->info.max_dod >> 2), | |
276b824c | 623 | numbank(pvt->info.max_dod >> 4), |
854d3349 MCC |
624 | numrow(pvt->info.max_dod >> 6), |
625 | numcol(pvt->info.max_dod >> 9)); | |
a0c36a1f | 626 | |
0b2b7b7e | 627 | for (i = 0; i < NUM_CHANS; i++) { |
854d3349 | 628 | u32 data, dimm_dod[3], value[8]; |
0b2b7b7e | 629 | |
52a2e4fc MCC |
630 | if (!pvt->pci_ch[i][0]) |
631 | continue; | |
632 | ||
0b2b7b7e MCC |
633 | if (!CH_ACTIVE(pvt, i)) { |
634 | debugf0("Channel %i is not active\n", i); | |
635 | continue; | |
636 | } | |
637 | if (CH_DISABLED(pvt, i)) { | |
638 | debugf0("Channel %i is disabled\n", i); | |
639 | continue; | |
640 | } | |
641 | ||
f122a892 | 642 | /* Devices 4-6 function 0 */ |
f4742949 | 643 | pci_read_config_dword(pvt->pci_ch[i][0], |
0b2b7b7e MCC |
644 | MC_CHANNEL_DIMM_INIT_PARAMS, &data); |
645 | ||
f4742949 | 646 | pvt->channel[i].ranks = (data & QUAD_RANK_PRESENT) ? |
67166af4 | 647 | 4 : 2; |
0b2b7b7e | 648 | |
854d3349 MCC |
649 | if (data & REGISTERED_DIMM) |
650 | mtype = MEM_RDDR3; | |
14d2c083 | 651 | else |
854d3349 MCC |
652 | mtype = MEM_DDR3; |
653 | #if 0 | |
0b2b7b7e MCC |
654 | if (data & THREE_DIMMS_PRESENT) |
655 | pvt->channel[i].dimms = 3; | |
656 | else if (data & SINGLE_QUAD_RANK_PRESENT) | |
657 | pvt->channel[i].dimms = 1; | |
658 | else | |
659 | pvt->channel[i].dimms = 2; | |
854d3349 MCC |
660 | #endif |
661 | ||
662 | /* Devices 4-6 function 1 */ | |
f4742949 | 663 | pci_read_config_dword(pvt->pci_ch[i][1], |
854d3349 | 664 | MC_DOD_CH_DIMM0, &dimm_dod[0]); |
f4742949 | 665 | pci_read_config_dword(pvt->pci_ch[i][1], |
854d3349 | 666 | MC_DOD_CH_DIMM1, &dimm_dod[1]); |
f4742949 | 667 | pci_read_config_dword(pvt->pci_ch[i][1], |
854d3349 | 668 | MC_DOD_CH_DIMM2, &dimm_dod[2]); |
0b2b7b7e | 669 | |
1c6fed80 | 670 | debugf0("Ch%d phy rd%d, wr%d (0x%08x): " |
854d3349 | 671 | "%d ranks, %cDIMMs\n", |
1c6fed80 MCC |
672 | i, |
673 | RDLCH(pvt->info.ch_map, i), WRLCH(pvt->info.ch_map, i), | |
674 | data, | |
f4742949 | 675 | pvt->channel[i].ranks, |
41fcb7fe | 676 | (data & REGISTERED_DIMM) ? 'R' : 'U'); |
854d3349 MCC |
677 | |
678 | for (j = 0; j < 3; j++) { | |
679 | u32 banks, ranks, rows, cols; | |
5566cb7c | 680 | u32 size, npages; |
854d3349 MCC |
681 | |
682 | if (!DIMM_PRESENT(dimm_dod[j])) | |
683 | continue; | |
684 | ||
685 | banks = numbank(MC_DOD_NUMBANK(dimm_dod[j])); | |
686 | ranks = numrank(MC_DOD_NUMRANK(dimm_dod[j])); | |
687 | rows = numrow(MC_DOD_NUMROW(dimm_dod[j])); | |
688 | cols = numcol(MC_DOD_NUMCOL(dimm_dod[j])); | |
689 | ||
5566cb7c MCC |
690 | /* DDR3 has 8 I/O banks */ |
691 | size = (rows * cols * banks * ranks) >> (20 - 3); | |
692 | ||
f4742949 | 693 | pvt->channel[i].dimms++; |
854d3349 | 694 | |
17cb7b0c MCC |
695 | debugf0("\tdimm %d %d Mb offset: %x, " |
696 | "bank: %d, rank: %d, row: %#x, col: %#x\n", | |
697 | j, size, | |
854d3349 MCC |
698 | RANKOFFSET(dimm_dod[j]), |
699 | banks, ranks, rows, cols); | |
700 | ||
e9144601 | 701 | npages = MiB_TO_PAGES(size); |
5566cb7c | 702 | |
2e5185f7 | 703 | csr = &mci->csrows[csrow]; |
5566cb7c MCC |
704 | csr->first_page = last_page + 1; |
705 | last_page += npages; | |
706 | csr->last_page = last_page; | |
707 | csr->nr_pages = npages; | |
708 | ||
854d3349 | 709 | csr->page_mask = 0; |
eb94fc40 | 710 | csr->grain = 8; |
2e5185f7 | 711 | csr->csrow_idx = csrow; |
eb94fc40 MCC |
712 | csr->nr_channels = 1; |
713 | ||
714 | csr->channels[0].chan_idx = i; | |
715 | csr->channels[0].ce_count = 0; | |
854d3349 | 716 | |
2e5185f7 | 717 | pvt->csrow_map[i][j] = csrow; |
b4e8f0b6 | 718 | |
854d3349 MCC |
719 | switch (banks) { |
720 | case 4: | |
721 | csr->dtype = DEV_X4; | |
722 | break; | |
723 | case 8: | |
724 | csr->dtype = DEV_X8; | |
725 | break; | |
726 | case 16: | |
727 | csr->dtype = DEV_X16; | |
728 | break; | |
729 | default: | |
730 | csr->dtype = DEV_UNKNOWN; | |
731 | } | |
732 | ||
733 | csr->edac_mode = mode; | |
734 | csr->mtype = mtype; | |
735 | ||
2e5185f7 | 736 | csrow++; |
854d3349 | 737 | } |
1c6fed80 | 738 | |
854d3349 MCC |
739 | pci_read_config_dword(pdev, MC_SAG_CH_0, &value[0]); |
740 | pci_read_config_dword(pdev, MC_SAG_CH_1, &value[1]); | |
741 | pci_read_config_dword(pdev, MC_SAG_CH_2, &value[2]); | |
742 | pci_read_config_dword(pdev, MC_SAG_CH_3, &value[3]); | |
743 | pci_read_config_dword(pdev, MC_SAG_CH_4, &value[4]); | |
744 | pci_read_config_dword(pdev, MC_SAG_CH_5, &value[5]); | |
745 | pci_read_config_dword(pdev, MC_SAG_CH_6, &value[6]); | |
746 | pci_read_config_dword(pdev, MC_SAG_CH_7, &value[7]); | |
17cb7b0c | 747 | debugf1("\t[%i] DIVBY3\tREMOVED\tOFFSET\n", i); |
854d3349 | 748 | for (j = 0; j < 8; j++) |
17cb7b0c | 749 | debugf1("\t\t%#x\t%#x\t%#x\n", |
854d3349 MCC |
750 | (value[j] >> 27) & 0x1, |
751 | (value[j] >> 24) & 0x7, | |
752 | (value[j] && ((1 << 24) - 1))); | |
0b2b7b7e MCC |
753 | } |
754 | ||
a0c36a1f MCC |
755 | return 0; |
756 | } | |
757 | ||
194a40fe MCC |
758 | /**************************************************************************** |
759 | Error insertion routines | |
760 | ****************************************************************************/ | |
761 | ||
762 | /* The i7core has independent error injection features per channel. | |
763 | However, to have a simpler code, we don't allow enabling error injection | |
764 | on more than one channel. | |
765 | Also, since a change at an inject parameter will be applied only at enable, | |
766 | we're disabling error injection on all write calls to the sysfs nodes that | |
767 | controls the error code injection. | |
768 | */ | |
1288c18f | 769 | static int disable_inject(const struct mem_ctl_info *mci) |
194a40fe MCC |
770 | { |
771 | struct i7core_pvt *pvt = mci->pvt_info; | |
772 | ||
773 | pvt->inject.enable = 0; | |
774 | ||
f4742949 | 775 | if (!pvt->pci_ch[pvt->inject.channel][0]) |
8f331907 MCC |
776 | return -ENODEV; |
777 | ||
f4742949 | 778 | pci_write_config_dword(pvt->pci_ch[pvt->inject.channel][0], |
4157d9f5 | 779 | MC_CHANNEL_ERROR_INJECT, 0); |
8f331907 MCC |
780 | |
781 | return 0; | |
194a40fe MCC |
782 | } |
783 | ||
784 | /* | |
785 | * i7core inject inject.section | |
786 | * | |
787 | * accept and store error injection inject.section value | |
788 | * bit 0 - refers to the lower 32-byte half cacheline | |
789 | * bit 1 - refers to the upper 32-byte half cacheline | |
790 | */ | |
791 | static ssize_t i7core_inject_section_store(struct mem_ctl_info *mci, | |
792 | const char *data, size_t count) | |
793 | { | |
794 | struct i7core_pvt *pvt = mci->pvt_info; | |
795 | unsigned long value; | |
796 | int rc; | |
797 | ||
798 | if (pvt->inject.enable) | |
41fcb7fe | 799 | disable_inject(mci); |
194a40fe MCC |
800 | |
801 | rc = strict_strtoul(data, 10, &value); | |
802 | if ((rc < 0) || (value > 3)) | |
2068def5 | 803 | return -EIO; |
194a40fe MCC |
804 | |
805 | pvt->inject.section = (u32) value; | |
806 | return count; | |
807 | } | |
808 | ||
809 | static ssize_t i7core_inject_section_show(struct mem_ctl_info *mci, | |
810 | char *data) | |
811 | { | |
812 | struct i7core_pvt *pvt = mci->pvt_info; | |
813 | return sprintf(data, "0x%08x\n", pvt->inject.section); | |
814 | } | |
815 | ||
816 | /* | |
817 | * i7core inject.type | |
818 | * | |
819 | * accept and store error injection inject.section value | |
820 | * bit 0 - repeat enable - Enable error repetition | |
821 | * bit 1 - inject ECC error | |
822 | * bit 2 - inject parity error | |
823 | */ | |
824 | static ssize_t i7core_inject_type_store(struct mem_ctl_info *mci, | |
825 | const char *data, size_t count) | |
826 | { | |
827 | struct i7core_pvt *pvt = mci->pvt_info; | |
828 | unsigned long value; | |
829 | int rc; | |
830 | ||
831 | if (pvt->inject.enable) | |
41fcb7fe | 832 | disable_inject(mci); |
194a40fe MCC |
833 | |
834 | rc = strict_strtoul(data, 10, &value); | |
835 | if ((rc < 0) || (value > 7)) | |
2068def5 | 836 | return -EIO; |
194a40fe MCC |
837 | |
838 | pvt->inject.type = (u32) value; | |
839 | return count; | |
840 | } | |
841 | ||
842 | static ssize_t i7core_inject_type_show(struct mem_ctl_info *mci, | |
843 | char *data) | |
844 | { | |
845 | struct i7core_pvt *pvt = mci->pvt_info; | |
846 | return sprintf(data, "0x%08x\n", pvt->inject.type); | |
847 | } | |
848 | ||
849 | /* | |
850 | * i7core_inject_inject.eccmask_store | |
851 | * | |
852 | * The type of error (UE/CE) will depend on the inject.eccmask value: | |
853 | * Any bits set to a 1 will flip the corresponding ECC bit | |
854 | * Correctable errors can be injected by flipping 1 bit or the bits within | |
855 | * a symbol pair (2 consecutive aligned 8-bit pairs - i.e. 7:0 and 15:8 or | |
856 | * 23:16 and 31:24). Flipping bits in two symbol pairs will cause an | |
857 | * uncorrectable error to be injected. | |
858 | */ | |
859 | static ssize_t i7core_inject_eccmask_store(struct mem_ctl_info *mci, | |
860 | const char *data, size_t count) | |
861 | { | |
862 | struct i7core_pvt *pvt = mci->pvt_info; | |
863 | unsigned long value; | |
864 | int rc; | |
865 | ||
866 | if (pvt->inject.enable) | |
41fcb7fe | 867 | disable_inject(mci); |
194a40fe MCC |
868 | |
869 | rc = strict_strtoul(data, 10, &value); | |
870 | if (rc < 0) | |
2068def5 | 871 | return -EIO; |
194a40fe MCC |
872 | |
873 | pvt->inject.eccmask = (u32) value; | |
874 | return count; | |
875 | } | |
876 | ||
877 | static ssize_t i7core_inject_eccmask_show(struct mem_ctl_info *mci, | |
878 | char *data) | |
879 | { | |
880 | struct i7core_pvt *pvt = mci->pvt_info; | |
881 | return sprintf(data, "0x%08x\n", pvt->inject.eccmask); | |
882 | } | |
883 | ||
884 | /* | |
885 | * i7core_addrmatch | |
886 | * | |
887 | * The type of error (UE/CE) will depend on the inject.eccmask value: | |
888 | * Any bits set to a 1 will flip the corresponding ECC bit | |
889 | * Correctable errors can be injected by flipping 1 bit or the bits within | |
890 | * a symbol pair (2 consecutive aligned 8-bit pairs - i.e. 7:0 and 15:8 or | |
891 | * 23:16 and 31:24). Flipping bits in two symbol pairs will cause an | |
892 | * uncorrectable error to be injected. | |
893 | */ | |
194a40fe | 894 | |
a5538e53 MCC |
895 | #define DECLARE_ADDR_MATCH(param, limit) \ |
896 | static ssize_t i7core_inject_store_##param( \ | |
897 | struct mem_ctl_info *mci, \ | |
898 | const char *data, size_t count) \ | |
899 | { \ | |
cc301b3a | 900 | struct i7core_pvt *pvt; \ |
a5538e53 MCC |
901 | long value; \ |
902 | int rc; \ | |
903 | \ | |
cc301b3a MCC |
904 | debugf1("%s()\n", __func__); \ |
905 | pvt = mci->pvt_info; \ | |
906 | \ | |
a5538e53 MCC |
907 | if (pvt->inject.enable) \ |
908 | disable_inject(mci); \ | |
909 | \ | |
4f87fad1 | 910 | if (!strcasecmp(data, "any") || !strcasecmp(data, "any\n"))\ |
a5538e53 MCC |
911 | value = -1; \ |
912 | else { \ | |
913 | rc = strict_strtoul(data, 10, &value); \ | |
914 | if ((rc < 0) || (value >= limit)) \ | |
915 | return -EIO; \ | |
916 | } \ | |
917 | \ | |
918 | pvt->inject.param = value; \ | |
919 | \ | |
920 | return count; \ | |
921 | } \ | |
922 | \ | |
923 | static ssize_t i7core_inject_show_##param( \ | |
924 | struct mem_ctl_info *mci, \ | |
925 | char *data) \ | |
926 | { \ | |
cc301b3a MCC |
927 | struct i7core_pvt *pvt; \ |
928 | \ | |
929 | pvt = mci->pvt_info; \ | |
930 | debugf1("%s() pvt=%p\n", __func__, pvt); \ | |
a5538e53 MCC |
931 | if (pvt->inject.param < 0) \ |
932 | return sprintf(data, "any\n"); \ | |
933 | else \ | |
934 | return sprintf(data, "%d\n", pvt->inject.param);\ | |
194a40fe MCC |
935 | } |
936 | ||
a5538e53 MCC |
937 | #define ATTR_ADDR_MATCH(param) \ |
938 | { \ | |
939 | .attr = { \ | |
940 | .name = #param, \ | |
941 | .mode = (S_IRUGO | S_IWUSR) \ | |
942 | }, \ | |
943 | .show = i7core_inject_show_##param, \ | |
944 | .store = i7core_inject_store_##param, \ | |
945 | } | |
194a40fe | 946 | |
a5538e53 MCC |
947 | DECLARE_ADDR_MATCH(channel, 3); |
948 | DECLARE_ADDR_MATCH(dimm, 3); | |
949 | DECLARE_ADDR_MATCH(rank, 4); | |
950 | DECLARE_ADDR_MATCH(bank, 32); | |
951 | DECLARE_ADDR_MATCH(page, 0x10000); | |
952 | DECLARE_ADDR_MATCH(col, 0x4000); | |
194a40fe | 953 | |
1288c18f | 954 | static int write_and_test(struct pci_dev *dev, const int where, const u32 val) |
276b824c MCC |
955 | { |
956 | u32 read; | |
957 | int count; | |
958 | ||
4157d9f5 MCC |
959 | debugf0("setting pci %02x:%02x.%x reg=%02x value=%08x\n", |
960 | dev->bus->number, PCI_SLOT(dev->devfn), PCI_FUNC(dev->devfn), | |
961 | where, val); | |
962 | ||
276b824c MCC |
963 | for (count = 0; count < 10; count++) { |
964 | if (count) | |
b990538a | 965 | msleep(100); |
276b824c MCC |
966 | pci_write_config_dword(dev, where, val); |
967 | pci_read_config_dword(dev, where, &read); | |
968 | ||
969 | if (read == val) | |
970 | return 0; | |
971 | } | |
972 | ||
4157d9f5 MCC |
973 | i7core_printk(KERN_ERR, "Error during set pci %02x:%02x.%x reg=%02x " |
974 | "write=%08x. Read=%08x\n", | |
975 | dev->bus->number, PCI_SLOT(dev->devfn), PCI_FUNC(dev->devfn), | |
976 | where, val, read); | |
276b824c MCC |
977 | |
978 | return -EINVAL; | |
979 | } | |
980 | ||
194a40fe MCC |
981 | /* |
982 | * This routine prepares the Memory Controller for error injection. | |
983 | * The error will be injected when some process tries to write to the | |
984 | * memory that matches the given criteria. | |
985 | * The criteria can be set in terms of a mask where dimm, rank, bank, page | |
986 | * and col can be specified. | |
987 | * A -1 value for any of the mask items will make the MCU to ignore | |
988 | * that matching criteria for error injection. | |
989 | * | |
990 | * It should be noticed that the error will only happen after a write operation | |
991 | * on a memory that matches the condition. if REPEAT_EN is not enabled at | |
992 | * inject mask, then it will produce just one error. Otherwise, it will repeat | |
993 | * until the injectmask would be cleaned. | |
994 | * | |
995 | * FIXME: This routine assumes that MAXNUMDIMMS value of MC_MAX_DOD | |
996 | * is reliable enough to check if the MC is using the | |
997 | * three channels. However, this is not clear at the datasheet. | |
998 | */ | |
999 | static ssize_t i7core_inject_enable_store(struct mem_ctl_info *mci, | |
1000 | const char *data, size_t count) | |
1001 | { | |
1002 | struct i7core_pvt *pvt = mci->pvt_info; | |
1003 | u32 injectmask; | |
1004 | u64 mask = 0; | |
1005 | int rc; | |
1006 | long enable; | |
1007 | ||
f4742949 | 1008 | if (!pvt->pci_ch[pvt->inject.channel][0]) |
8f331907 MCC |
1009 | return 0; |
1010 | ||
194a40fe MCC |
1011 | rc = strict_strtoul(data, 10, &enable); |
1012 | if ((rc < 0)) | |
1013 | return 0; | |
1014 | ||
1015 | if (enable) { | |
1016 | pvt->inject.enable = 1; | |
1017 | } else { | |
1018 | disable_inject(mci); | |
1019 | return count; | |
1020 | } | |
1021 | ||
1022 | /* Sets pvt->inject.dimm mask */ | |
1023 | if (pvt->inject.dimm < 0) | |
486dd09f | 1024 | mask |= 1LL << 41; |
194a40fe | 1025 | else { |
f4742949 | 1026 | if (pvt->channel[pvt->inject.channel].dimms > 2) |
486dd09f | 1027 | mask |= (pvt->inject.dimm & 0x3LL) << 35; |
194a40fe | 1028 | else |
486dd09f | 1029 | mask |= (pvt->inject.dimm & 0x1LL) << 36; |
194a40fe MCC |
1030 | } |
1031 | ||
1032 | /* Sets pvt->inject.rank mask */ | |
1033 | if (pvt->inject.rank < 0) | |
486dd09f | 1034 | mask |= 1LL << 40; |
194a40fe | 1035 | else { |
f4742949 | 1036 | if (pvt->channel[pvt->inject.channel].dimms > 2) |
486dd09f | 1037 | mask |= (pvt->inject.rank & 0x1LL) << 34; |
194a40fe | 1038 | else |
486dd09f | 1039 | mask |= (pvt->inject.rank & 0x3LL) << 34; |
194a40fe MCC |
1040 | } |
1041 | ||
1042 | /* Sets pvt->inject.bank mask */ | |
1043 | if (pvt->inject.bank < 0) | |
486dd09f | 1044 | mask |= 1LL << 39; |
194a40fe | 1045 | else |
486dd09f | 1046 | mask |= (pvt->inject.bank & 0x15LL) << 30; |
194a40fe MCC |
1047 | |
1048 | /* Sets pvt->inject.page mask */ | |
1049 | if (pvt->inject.page < 0) | |
486dd09f | 1050 | mask |= 1LL << 38; |
194a40fe | 1051 | else |
486dd09f | 1052 | mask |= (pvt->inject.page & 0xffff) << 14; |
194a40fe MCC |
1053 | |
1054 | /* Sets pvt->inject.column mask */ | |
1055 | if (pvt->inject.col < 0) | |
486dd09f | 1056 | mask |= 1LL << 37; |
194a40fe | 1057 | else |
486dd09f | 1058 | mask |= (pvt->inject.col & 0x3fff); |
194a40fe | 1059 | |
276b824c MCC |
1060 | /* |
1061 | * bit 0: REPEAT_EN | |
1062 | * bits 1-2: MASK_HALF_CACHELINE | |
1063 | * bit 3: INJECT_ECC | |
1064 | * bit 4: INJECT_ADDR_PARITY | |
1065 | */ | |
1066 | ||
1067 | injectmask = (pvt->inject.type & 1) | | |
1068 | (pvt->inject.section & 0x3) << 1 | | |
1069 | (pvt->inject.type & 0x6) << (3 - 1); | |
1070 | ||
1071 | /* Unlock writes to registers - this register is write only */ | |
f4742949 | 1072 | pci_write_config_dword(pvt->pci_noncore, |
67166af4 | 1073 | MC_CFG_CONTROL, 0x2); |
e9bd2e73 | 1074 | |
f4742949 | 1075 | write_and_test(pvt->pci_ch[pvt->inject.channel][0], |
194a40fe | 1076 | MC_CHANNEL_ADDR_MATCH, mask); |
f4742949 | 1077 | write_and_test(pvt->pci_ch[pvt->inject.channel][0], |
7b029d03 | 1078 | MC_CHANNEL_ADDR_MATCH + 4, mask >> 32L); |
7b029d03 | 1079 | |
f4742949 | 1080 | write_and_test(pvt->pci_ch[pvt->inject.channel][0], |
194a40fe MCC |
1081 | MC_CHANNEL_ERROR_MASK, pvt->inject.eccmask); |
1082 | ||
f4742949 | 1083 | write_and_test(pvt->pci_ch[pvt->inject.channel][0], |
4157d9f5 | 1084 | MC_CHANNEL_ERROR_INJECT, injectmask); |
276b824c | 1085 | |
194a40fe | 1086 | /* |
276b824c MCC |
1087 | * This is something undocumented, based on my tests |
1088 | * Without writing 8 to this register, errors aren't injected. Not sure | |
1089 | * why. | |
194a40fe | 1090 | */ |
f4742949 | 1091 | pci_write_config_dword(pvt->pci_noncore, |
276b824c | 1092 | MC_CFG_CONTROL, 8); |
194a40fe | 1093 | |
41fcb7fe MCC |
1094 | debugf0("Error inject addr match 0x%016llx, ecc 0x%08x," |
1095 | " inject 0x%08x\n", | |
194a40fe MCC |
1096 | mask, pvt->inject.eccmask, injectmask); |
1097 | ||
7b029d03 | 1098 | |
194a40fe MCC |
1099 | return count; |
1100 | } | |
1101 | ||
1102 | static ssize_t i7core_inject_enable_show(struct mem_ctl_info *mci, | |
1103 | char *data) | |
1104 | { | |
1105 | struct i7core_pvt *pvt = mci->pvt_info; | |
7b029d03 MCC |
1106 | u32 injectmask; |
1107 | ||
52a2e4fc MCC |
1108 | if (!pvt->pci_ch[pvt->inject.channel][0]) |
1109 | return 0; | |
1110 | ||
f4742949 | 1111 | pci_read_config_dword(pvt->pci_ch[pvt->inject.channel][0], |
4157d9f5 | 1112 | MC_CHANNEL_ERROR_INJECT, &injectmask); |
7b029d03 MCC |
1113 | |
1114 | debugf0("Inject error read: 0x%018x\n", injectmask); | |
1115 | ||
1116 | if (injectmask & 0x0c) | |
1117 | pvt->inject.enable = 1; | |
1118 | ||
194a40fe MCC |
1119 | return sprintf(data, "%d\n", pvt->inject.enable); |
1120 | } | |
1121 | ||
f338d736 MCC |
1122 | #define DECLARE_COUNTER(param) \ |
1123 | static ssize_t i7core_show_counter_##param( \ | |
1124 | struct mem_ctl_info *mci, \ | |
1125 | char *data) \ | |
1126 | { \ | |
1127 | struct i7core_pvt *pvt = mci->pvt_info; \ | |
1128 | \ | |
1129 | debugf1("%s() \n", __func__); \ | |
1130 | if (!pvt->ce_count_available || (pvt->is_registered)) \ | |
1131 | return sprintf(data, "data unavailable\n"); \ | |
1132 | return sprintf(data, "%lu\n", \ | |
1133 | pvt->udimm_ce_count[param]); \ | |
1134 | } | |
442305b1 | 1135 | |
f338d736 MCC |
1136 | #define ATTR_COUNTER(param) \ |
1137 | { \ | |
1138 | .attr = { \ | |
1139 | .name = __stringify(udimm##param), \ | |
1140 | .mode = (S_IRUGO | S_IWUSR) \ | |
1141 | }, \ | |
1142 | .show = i7core_show_counter_##param \ | |
d88b8507 | 1143 | } |
442305b1 | 1144 | |
f338d736 MCC |
1145 | DECLARE_COUNTER(0); |
1146 | DECLARE_COUNTER(1); | |
1147 | DECLARE_COUNTER(2); | |
442305b1 | 1148 | |
194a40fe MCC |
1149 | /* |
1150 | * Sysfs struct | |
1151 | */ | |
a5538e53 | 1152 | |
1288c18f | 1153 | static const struct mcidev_sysfs_attribute i7core_addrmatch_attrs[] = { |
a5538e53 MCC |
1154 | ATTR_ADDR_MATCH(channel), |
1155 | ATTR_ADDR_MATCH(dimm), | |
1156 | ATTR_ADDR_MATCH(rank), | |
1157 | ATTR_ADDR_MATCH(bank), | |
1158 | ATTR_ADDR_MATCH(page), | |
1159 | ATTR_ADDR_MATCH(col), | |
1288c18f | 1160 | { } /* End of list */ |
a5538e53 MCC |
1161 | }; |
1162 | ||
1288c18f | 1163 | static const struct mcidev_sysfs_group i7core_inject_addrmatch = { |
a5538e53 MCC |
1164 | .name = "inject_addrmatch", |
1165 | .mcidev_attr = i7core_addrmatch_attrs, | |
1166 | }; | |
1167 | ||
1288c18f | 1168 | static const struct mcidev_sysfs_attribute i7core_udimm_counters_attrs[] = { |
f338d736 MCC |
1169 | ATTR_COUNTER(0), |
1170 | ATTR_COUNTER(1), | |
1171 | ATTR_COUNTER(2), | |
64aab720 | 1172 | { .attr = { .name = NULL } } |
f338d736 MCC |
1173 | }; |
1174 | ||
1288c18f | 1175 | static const struct mcidev_sysfs_group i7core_udimm_counters = { |
f338d736 MCC |
1176 | .name = "all_channel_counts", |
1177 | .mcidev_attr = i7core_udimm_counters_attrs, | |
1178 | }; | |
1179 | ||
1288c18f | 1180 | static const struct mcidev_sysfs_attribute i7core_sysfs_rdimm_attrs[] = { |
194a40fe MCC |
1181 | { |
1182 | .attr = { | |
1183 | .name = "inject_section", | |
1184 | .mode = (S_IRUGO | S_IWUSR) | |
1185 | }, | |
1186 | .show = i7core_inject_section_show, | |
1187 | .store = i7core_inject_section_store, | |
1188 | }, { | |
1189 | .attr = { | |
1190 | .name = "inject_type", | |
1191 | .mode = (S_IRUGO | S_IWUSR) | |
1192 | }, | |
1193 | .show = i7core_inject_type_show, | |
1194 | .store = i7core_inject_type_store, | |
1195 | }, { | |
1196 | .attr = { | |
1197 | .name = "inject_eccmask", | |
1198 | .mode = (S_IRUGO | S_IWUSR) | |
1199 | }, | |
1200 | .show = i7core_inject_eccmask_show, | |
1201 | .store = i7core_inject_eccmask_store, | |
1202 | }, { | |
a5538e53 | 1203 | .grp = &i7core_inject_addrmatch, |
194a40fe MCC |
1204 | }, { |
1205 | .attr = { | |
1206 | .name = "inject_enable", | |
1207 | .mode = (S_IRUGO | S_IWUSR) | |
1208 | }, | |
1209 | .show = i7core_inject_enable_show, | |
1210 | .store = i7core_inject_enable_store, | |
1211 | }, | |
1288c18f MCC |
1212 | { } /* End of list */ |
1213 | }; | |
1214 | ||
1215 | static const struct mcidev_sysfs_attribute i7core_sysfs_udimm_attrs[] = { | |
1216 | { | |
1217 | .attr = { | |
1218 | .name = "inject_section", | |
1219 | .mode = (S_IRUGO | S_IWUSR) | |
1220 | }, | |
1221 | .show = i7core_inject_section_show, | |
1222 | .store = i7core_inject_section_store, | |
1223 | }, { | |
1224 | .attr = { | |
1225 | .name = "inject_type", | |
1226 | .mode = (S_IRUGO | S_IWUSR) | |
1227 | }, | |
1228 | .show = i7core_inject_type_show, | |
1229 | .store = i7core_inject_type_store, | |
1230 | }, { | |
1231 | .attr = { | |
1232 | .name = "inject_eccmask", | |
1233 | .mode = (S_IRUGO | S_IWUSR) | |
1234 | }, | |
1235 | .show = i7core_inject_eccmask_show, | |
1236 | .store = i7core_inject_eccmask_store, | |
1237 | }, { | |
1238 | .grp = &i7core_inject_addrmatch, | |
1239 | }, { | |
1240 | .attr = { | |
1241 | .name = "inject_enable", | |
1242 | .mode = (S_IRUGO | S_IWUSR) | |
1243 | }, | |
1244 | .show = i7core_inject_enable_show, | |
1245 | .store = i7core_inject_enable_store, | |
1246 | }, { | |
1247 | .grp = &i7core_udimm_counters, | |
1248 | }, | |
1249 | { } /* End of list */ | |
194a40fe MCC |
1250 | }; |
1251 | ||
a0c36a1f MCC |
1252 | /**************************************************************************** |
1253 | Device initialization routines: put/get, init/exit | |
1254 | ****************************************************************************/ | |
1255 | ||
1256 | /* | |
64c10f6e | 1257 | * i7core_put_all_devices 'put' all the devices that we have |
a0c36a1f MCC |
1258 | * reserved via 'get' |
1259 | */ | |
13d6e9b6 | 1260 | static void i7core_put_devices(struct i7core_dev *i7core_dev) |
a0c36a1f | 1261 | { |
13d6e9b6 | 1262 | int i; |
a0c36a1f | 1263 | |
22e6bcbd | 1264 | debugf0(__FILE__ ": %s()\n", __func__); |
de06eeef | 1265 | for (i = 0; i < i7core_dev->n_devs; i++) { |
22e6bcbd MCC |
1266 | struct pci_dev *pdev = i7core_dev->pdev[i]; |
1267 | if (!pdev) | |
1268 | continue; | |
1269 | debugf0("Removing dev %02x:%02x.%d\n", | |
1270 | pdev->bus->number, | |
1271 | PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn)); | |
1272 | pci_dev_put(pdev); | |
1273 | } | |
13d6e9b6 | 1274 | } |
66607706 | 1275 | |
13d6e9b6 MCC |
1276 | static void i7core_put_all_devices(void) |
1277 | { | |
42538680 | 1278 | struct i7core_dev *i7core_dev, *tmp; |
13d6e9b6 | 1279 | |
39300e71 | 1280 | list_for_each_entry_safe(i7core_dev, tmp, &i7core_edac_list, list) { |
13d6e9b6 | 1281 | i7core_put_devices(i7core_dev); |
2aa9be44 | 1282 | free_i7core_dev(i7core_dev); |
39300e71 | 1283 | } |
a0c36a1f MCC |
1284 | } |
1285 | ||
1288c18f | 1286 | static void __init i7core_xeon_pci_fixup(const struct pci_id_table *table) |
bc2d7245 KM |
1287 | { |
1288 | struct pci_dev *pdev = NULL; | |
1289 | int i; | |
54a08ab1 | 1290 | |
bc2d7245 KM |
1291 | /* |
1292 | * On Xeon 55xx, the Intel Quckpath Arch Generic Non-core pci buses | |
1293 | * aren't announced by acpi. So, we need to use a legacy scan probing | |
1294 | * to detect them | |
1295 | */ | |
bd9e19ca VM |
1296 | while (table && table->descr) { |
1297 | pdev = pci_get_device(PCI_VENDOR_ID_INTEL, table->descr[0].dev_id, NULL); | |
1298 | if (unlikely(!pdev)) { | |
1299 | for (i = 0; i < MAX_SOCKET_BUSES; i++) | |
1300 | pcibios_scan_specific_bus(255-i); | |
1301 | } | |
bda14289 | 1302 | pci_dev_put(pdev); |
bd9e19ca | 1303 | table++; |
bc2d7245 KM |
1304 | } |
1305 | } | |
1306 | ||
bda14289 MCC |
1307 | static unsigned i7core_pci_lastbus(void) |
1308 | { | |
1309 | int last_bus = 0, bus; | |
1310 | struct pci_bus *b = NULL; | |
1311 | ||
1312 | while ((b = pci_find_next_bus(b)) != NULL) { | |
1313 | bus = b->number; | |
1314 | debugf0("Found bus %d\n", bus); | |
1315 | if (bus > last_bus) | |
1316 | last_bus = bus; | |
1317 | } | |
1318 | ||
1319 | debugf0("Last bus %d\n", last_bus); | |
1320 | ||
1321 | return last_bus; | |
1322 | } | |
1323 | ||
a0c36a1f | 1324 | /* |
64c10f6e | 1325 | * i7core_get_all_devices Find and perform 'get' operation on the MCH's |
a0c36a1f MCC |
1326 | * device/functions we want to reference for this driver |
1327 | * | |
1328 | * Need to 'get' device 16 func 1 and func 2 | |
1329 | */ | |
b197cba0 HS |
1330 | static int i7core_get_onedevice(struct pci_dev **prev, |
1331 | const struct pci_id_table *table, | |
1332 | const unsigned devno, | |
1333 | const unsigned last_bus) | |
a0c36a1f | 1334 | { |
66607706 | 1335 | struct i7core_dev *i7core_dev; |
b197cba0 | 1336 | const struct pci_id_descr *dev_descr = &table->descr[devno]; |
66607706 | 1337 | |
8f331907 | 1338 | struct pci_dev *pdev = NULL; |
67166af4 MCC |
1339 | u8 bus = 0; |
1340 | u8 socket = 0; | |
a0c36a1f | 1341 | |
c77720b9 | 1342 | pdev = pci_get_device(PCI_VENDOR_ID_INTEL, |
de06eeef | 1343 | dev_descr->dev_id, *prev); |
c77720b9 | 1344 | |
c77720b9 MCC |
1345 | /* |
1346 | * On Xeon 55xx, the Intel Quckpath Arch Generic Non-core regs | |
1347 | * is at addr 8086:2c40, instead of 8086:2c41. So, we need | |
1348 | * to probe for the alternate address in case of failure | |
1349 | */ | |
de06eeef | 1350 | if (dev_descr->dev_id == PCI_DEVICE_ID_INTEL_I7_NONCORE && !pdev) |
c77720b9 | 1351 | pdev = pci_get_device(PCI_VENDOR_ID_INTEL, |
fd382654 | 1352 | PCI_DEVICE_ID_INTEL_I7_NONCORE_ALT, *prev); |
d1fd4fb6 | 1353 | |
bd9e19ca | 1354 | if (dev_descr->dev_id == PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE && !pdev) |
f05da2f7 MCC |
1355 | pdev = pci_get_device(PCI_VENDOR_ID_INTEL, |
1356 | PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE_ALT, | |
1357 | *prev); | |
1358 | ||
c77720b9 MCC |
1359 | if (!pdev) { |
1360 | if (*prev) { | |
1361 | *prev = pdev; | |
1362 | return 0; | |
d1fd4fb6 MCC |
1363 | } |
1364 | ||
de06eeef | 1365 | if (dev_descr->optional) |
c77720b9 | 1366 | return 0; |
310cbb72 | 1367 | |
bd9e19ca VM |
1368 | if (devno == 0) |
1369 | return -ENODEV; | |
1370 | ||
ab089374 | 1371 | i7core_printk(KERN_INFO, |
c77720b9 | 1372 | "Device not found: dev %02x.%d PCI ID %04x:%04x\n", |
de06eeef MCC |
1373 | dev_descr->dev, dev_descr->func, |
1374 | PCI_VENDOR_ID_INTEL, dev_descr->dev_id); | |
67166af4 | 1375 | |
c77720b9 MCC |
1376 | /* End of list, leave */ |
1377 | return -ENODEV; | |
1378 | } | |
1379 | bus = pdev->bus->number; | |
67166af4 | 1380 | |
bda14289 | 1381 | socket = last_bus - bus; |
c77720b9 | 1382 | |
66607706 MCC |
1383 | i7core_dev = get_i7core_dev(socket); |
1384 | if (!i7core_dev) { | |
848b2f7e | 1385 | i7core_dev = alloc_i7core_dev(socket, table); |
2896637b HS |
1386 | if (!i7core_dev) { |
1387 | pci_dev_put(pdev); | |
66607706 | 1388 | return -ENOMEM; |
2896637b | 1389 | } |
c77720b9 | 1390 | } |
67166af4 | 1391 | |
66607706 | 1392 | if (i7core_dev->pdev[devno]) { |
c77720b9 MCC |
1393 | i7core_printk(KERN_ERR, |
1394 | "Duplicated device for " | |
1395 | "dev %02x:%02x.%d PCI ID %04x:%04x\n", | |
de06eeef MCC |
1396 | bus, dev_descr->dev, dev_descr->func, |
1397 | PCI_VENDOR_ID_INTEL, dev_descr->dev_id); | |
c77720b9 MCC |
1398 | pci_dev_put(pdev); |
1399 | return -ENODEV; | |
1400 | } | |
67166af4 | 1401 | |
66607706 | 1402 | i7core_dev->pdev[devno] = pdev; |
c77720b9 MCC |
1403 | |
1404 | /* Sanity check */ | |
de06eeef MCC |
1405 | if (unlikely(PCI_SLOT(pdev->devfn) != dev_descr->dev || |
1406 | PCI_FUNC(pdev->devfn) != dev_descr->func)) { | |
c77720b9 MCC |
1407 | i7core_printk(KERN_ERR, |
1408 | "Device PCI ID %04x:%04x " | |
1409 | "has dev %02x:%02x.%d instead of dev %02x:%02x.%d\n", | |
de06eeef | 1410 | PCI_VENDOR_ID_INTEL, dev_descr->dev_id, |
c77720b9 | 1411 | bus, PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn), |
de06eeef | 1412 | bus, dev_descr->dev, dev_descr->func); |
c77720b9 MCC |
1413 | return -ENODEV; |
1414 | } | |
ef708b53 | 1415 | |
c77720b9 MCC |
1416 | /* Be sure that the device is enabled */ |
1417 | if (unlikely(pci_enable_device(pdev) < 0)) { | |
1418 | i7core_printk(KERN_ERR, | |
1419 | "Couldn't enable " | |
1420 | "dev %02x:%02x.%d PCI ID %04x:%04x\n", | |
de06eeef MCC |
1421 | bus, dev_descr->dev, dev_descr->func, |
1422 | PCI_VENDOR_ID_INTEL, dev_descr->dev_id); | |
c77720b9 MCC |
1423 | return -ENODEV; |
1424 | } | |
ef708b53 | 1425 | |
d4c27795 | 1426 | debugf0("Detected socket %d dev %02x:%02x.%d PCI ID %04x:%04x\n", |
de06eeef MCC |
1427 | socket, bus, dev_descr->dev, |
1428 | dev_descr->func, | |
1429 | PCI_VENDOR_ID_INTEL, dev_descr->dev_id); | |
8f331907 | 1430 | |
c77720b9 | 1431 | *prev = pdev; |
ef708b53 | 1432 | |
c77720b9 MCC |
1433 | return 0; |
1434 | } | |
a0c36a1f | 1435 | |
64c10f6e | 1436 | static int i7core_get_all_devices(void) |
c77720b9 | 1437 | { |
64c10f6e | 1438 | int i, j, rc, last_bus; |
c77720b9 | 1439 | struct pci_dev *pdev = NULL; |
64c10f6e | 1440 | const struct pci_id_table *table; |
bd9e19ca | 1441 | |
bda14289 MCC |
1442 | last_bus = i7core_pci_lastbus(); |
1443 | ||
64c10f6e HS |
1444 | for (j = 0; j < ARRAY_SIZE(pci_dev_table); j++) { |
1445 | table = &pci_dev_table[j]; | |
bd9e19ca VM |
1446 | for (i = 0; i < table->n_devs; i++) { |
1447 | pdev = NULL; | |
1448 | do { | |
b197cba0 | 1449 | rc = i7core_get_onedevice(&pdev, table, i, |
bda14289 | 1450 | last_bus); |
bd9e19ca VM |
1451 | if (rc < 0) { |
1452 | if (i == 0) { | |
1453 | i = table->n_devs; | |
1454 | break; | |
1455 | } | |
1456 | i7core_put_all_devices(); | |
1457 | return -ENODEV; | |
1458 | } | |
1459 | } while (pdev); | |
1460 | } | |
c77720b9 | 1461 | } |
66607706 | 1462 | |
ef708b53 | 1463 | return 0; |
ef708b53 MCC |
1464 | } |
1465 | ||
f4742949 MCC |
1466 | static int mci_bind_devs(struct mem_ctl_info *mci, |
1467 | struct i7core_dev *i7core_dev) | |
ef708b53 MCC |
1468 | { |
1469 | struct i7core_pvt *pvt = mci->pvt_info; | |
1470 | struct pci_dev *pdev; | |
f4742949 | 1471 | int i, func, slot; |
ef708b53 | 1472 | |
f4742949 | 1473 | pvt->is_registered = 0; |
de06eeef | 1474 | for (i = 0; i < i7core_dev->n_devs; i++) { |
f4742949 MCC |
1475 | pdev = i7core_dev->pdev[i]; |
1476 | if (!pdev) | |
66607706 MCC |
1477 | continue; |
1478 | ||
f4742949 MCC |
1479 | func = PCI_FUNC(pdev->devfn); |
1480 | slot = PCI_SLOT(pdev->devfn); | |
1481 | if (slot == 3) { | |
1482 | if (unlikely(func > MAX_MCR_FUNC)) | |
1483 | goto error; | |
1484 | pvt->pci_mcr[func] = pdev; | |
1485 | } else if (likely(slot >= 4 && slot < 4 + NUM_CHANS)) { | |
1486 | if (unlikely(func > MAX_CHAN_FUNC)) | |
ef708b53 | 1487 | goto error; |
f4742949 MCC |
1488 | pvt->pci_ch[slot - 4][func] = pdev; |
1489 | } else if (!slot && !func) | |
1490 | pvt->pci_noncore = pdev; | |
1491 | else | |
1492 | goto error; | |
ef708b53 | 1493 | |
f4742949 MCC |
1494 | debugf0("Associated fn %d.%d, dev = %p, socket %d\n", |
1495 | PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn), | |
1496 | pdev, i7core_dev->socket); | |
14d2c083 | 1497 | |
f4742949 MCC |
1498 | if (PCI_SLOT(pdev->devfn) == 3 && |
1499 | PCI_FUNC(pdev->devfn) == 2) | |
1500 | pvt->is_registered = 1; | |
a0c36a1f | 1501 | } |
e9bd2e73 | 1502 | |
a0c36a1f | 1503 | return 0; |
ef708b53 MCC |
1504 | |
1505 | error: | |
1506 | i7core_printk(KERN_ERR, "Device %d, function %d " | |
1507 | "is out of the expected range\n", | |
1508 | slot, func); | |
1509 | return -EINVAL; | |
a0c36a1f MCC |
1510 | } |
1511 | ||
442305b1 MCC |
1512 | /**************************************************************************** |
1513 | Error check routines | |
1514 | ****************************************************************************/ | |
f4742949 | 1515 | static void i7core_rdimm_update_csrow(struct mem_ctl_info *mci, |
1288c18f MCC |
1516 | const int chan, |
1517 | const int dimm, | |
1518 | const int add) | |
b4e8f0b6 MCC |
1519 | { |
1520 | char *msg; | |
1521 | struct i7core_pvt *pvt = mci->pvt_info; | |
f4742949 | 1522 | int row = pvt->csrow_map[chan][dimm], i; |
b4e8f0b6 MCC |
1523 | |
1524 | for (i = 0; i < add; i++) { | |
1525 | msg = kasprintf(GFP_KERNEL, "Corrected error " | |
f4742949 MCC |
1526 | "(Socket=%d channel=%d dimm=%d)", |
1527 | pvt->i7core_dev->socket, chan, dimm); | |
b4e8f0b6 MCC |
1528 | |
1529 | edac_mc_handle_fbd_ce(mci, row, 0, msg); | |
1530 | kfree (msg); | |
1531 | } | |
1532 | } | |
1533 | ||
1534 | static void i7core_rdimm_update_ce_count(struct mem_ctl_info *mci, | |
1288c18f MCC |
1535 | const int chan, |
1536 | const int new0, | |
1537 | const int new1, | |
1538 | const int new2) | |
b4e8f0b6 MCC |
1539 | { |
1540 | struct i7core_pvt *pvt = mci->pvt_info; | |
1541 | int add0 = 0, add1 = 0, add2 = 0; | |
1542 | /* Updates CE counters if it is not the first time here */ | |
f4742949 | 1543 | if (pvt->ce_count_available) { |
b4e8f0b6 MCC |
1544 | /* Updates CE counters */ |
1545 | ||
f4742949 MCC |
1546 | add2 = new2 - pvt->rdimm_last_ce_count[chan][2]; |
1547 | add1 = new1 - pvt->rdimm_last_ce_count[chan][1]; | |
1548 | add0 = new0 - pvt->rdimm_last_ce_count[chan][0]; | |
b4e8f0b6 MCC |
1549 | |
1550 | if (add2 < 0) | |
1551 | add2 += 0x7fff; | |
f4742949 | 1552 | pvt->rdimm_ce_count[chan][2] += add2; |
b4e8f0b6 MCC |
1553 | |
1554 | if (add1 < 0) | |
1555 | add1 += 0x7fff; | |
f4742949 | 1556 | pvt->rdimm_ce_count[chan][1] += add1; |
b4e8f0b6 MCC |
1557 | |
1558 | if (add0 < 0) | |
1559 | add0 += 0x7fff; | |
f4742949 | 1560 | pvt->rdimm_ce_count[chan][0] += add0; |
b4e8f0b6 | 1561 | } else |
f4742949 | 1562 | pvt->ce_count_available = 1; |
b4e8f0b6 MCC |
1563 | |
1564 | /* Store the new values */ | |
f4742949 MCC |
1565 | pvt->rdimm_last_ce_count[chan][2] = new2; |
1566 | pvt->rdimm_last_ce_count[chan][1] = new1; | |
1567 | pvt->rdimm_last_ce_count[chan][0] = new0; | |
b4e8f0b6 MCC |
1568 | |
1569 | /*updated the edac core */ | |
1570 | if (add0 != 0) | |
f4742949 | 1571 | i7core_rdimm_update_csrow(mci, chan, 0, add0); |
b4e8f0b6 | 1572 | if (add1 != 0) |
f4742949 | 1573 | i7core_rdimm_update_csrow(mci, chan, 1, add1); |
b4e8f0b6 | 1574 | if (add2 != 0) |
f4742949 | 1575 | i7core_rdimm_update_csrow(mci, chan, 2, add2); |
b4e8f0b6 MCC |
1576 | |
1577 | } | |
1578 | ||
f4742949 | 1579 | static void i7core_rdimm_check_mc_ecc_err(struct mem_ctl_info *mci) |
b4e8f0b6 MCC |
1580 | { |
1581 | struct i7core_pvt *pvt = mci->pvt_info; | |
1582 | u32 rcv[3][2]; | |
1583 | int i, new0, new1, new2; | |
1584 | ||
1585 | /*Read DEV 3: FUN 2: MC_COR_ECC_CNT regs directly*/ | |
f4742949 | 1586 | pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_0, |
b4e8f0b6 | 1587 | &rcv[0][0]); |
f4742949 | 1588 | pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_1, |
b4e8f0b6 | 1589 | &rcv[0][1]); |
f4742949 | 1590 | pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_2, |
b4e8f0b6 | 1591 | &rcv[1][0]); |
f4742949 | 1592 | pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_3, |
b4e8f0b6 | 1593 | &rcv[1][1]); |
f4742949 | 1594 | pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_4, |
b4e8f0b6 | 1595 | &rcv[2][0]); |
f4742949 | 1596 | pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_5, |
b4e8f0b6 MCC |
1597 | &rcv[2][1]); |
1598 | for (i = 0 ; i < 3; i++) { | |
1599 | debugf3("MC_COR_ECC_CNT%d = 0x%x; MC_COR_ECC_CNT%d = 0x%x\n", | |
1600 | (i * 2), rcv[i][0], (i * 2) + 1, rcv[i][1]); | |
1601 | /*if the channel has 3 dimms*/ | |
f4742949 | 1602 | if (pvt->channel[i].dimms > 2) { |
b4e8f0b6 MCC |
1603 | new0 = DIMM_BOT_COR_ERR(rcv[i][0]); |
1604 | new1 = DIMM_TOP_COR_ERR(rcv[i][0]); | |
1605 | new2 = DIMM_BOT_COR_ERR(rcv[i][1]); | |
1606 | } else { | |
1607 | new0 = DIMM_TOP_COR_ERR(rcv[i][0]) + | |
1608 | DIMM_BOT_COR_ERR(rcv[i][0]); | |
1609 | new1 = DIMM_TOP_COR_ERR(rcv[i][1]) + | |
1610 | DIMM_BOT_COR_ERR(rcv[i][1]); | |
1611 | new2 = 0; | |
1612 | } | |
1613 | ||
f4742949 | 1614 | i7core_rdimm_update_ce_count(mci, i, new0, new1, new2); |
b4e8f0b6 MCC |
1615 | } |
1616 | } | |
442305b1 MCC |
1617 | |
1618 | /* This function is based on the device 3 function 4 registers as described on: | |
1619 | * Intel Xeon Processor 5500 Series Datasheet Volume 2 | |
1620 | * http://www.intel.com/Assets/PDF/datasheet/321322.pdf | |
1621 | * also available at: | |
1622 | * http://www.arrownac.com/manufacturers/intel/s/nehalem/5500-datasheet-v2.pdf | |
1623 | */ | |
f4742949 | 1624 | static void i7core_udimm_check_mc_ecc_err(struct mem_ctl_info *mci) |
442305b1 MCC |
1625 | { |
1626 | struct i7core_pvt *pvt = mci->pvt_info; | |
1627 | u32 rcv1, rcv0; | |
1628 | int new0, new1, new2; | |
1629 | ||
f4742949 | 1630 | if (!pvt->pci_mcr[4]) { |
b990538a | 1631 | debugf0("%s MCR registers not found\n", __func__); |
442305b1 MCC |
1632 | return; |
1633 | } | |
1634 | ||
b4e8f0b6 | 1635 | /* Corrected test errors */ |
f4742949 MCC |
1636 | pci_read_config_dword(pvt->pci_mcr[4], MC_TEST_ERR_RCV1, &rcv1); |
1637 | pci_read_config_dword(pvt->pci_mcr[4], MC_TEST_ERR_RCV0, &rcv0); | |
442305b1 MCC |
1638 | |
1639 | /* Store the new values */ | |
1640 | new2 = DIMM2_COR_ERR(rcv1); | |
1641 | new1 = DIMM1_COR_ERR(rcv0); | |
1642 | new0 = DIMM0_COR_ERR(rcv0); | |
1643 | ||
442305b1 | 1644 | /* Updates CE counters if it is not the first time here */ |
f4742949 | 1645 | if (pvt->ce_count_available) { |
442305b1 MCC |
1646 | /* Updates CE counters */ |
1647 | int add0, add1, add2; | |
1648 | ||
f4742949 MCC |
1649 | add2 = new2 - pvt->udimm_last_ce_count[2]; |
1650 | add1 = new1 - pvt->udimm_last_ce_count[1]; | |
1651 | add0 = new0 - pvt->udimm_last_ce_count[0]; | |
442305b1 MCC |
1652 | |
1653 | if (add2 < 0) | |
1654 | add2 += 0x7fff; | |
f4742949 | 1655 | pvt->udimm_ce_count[2] += add2; |
442305b1 MCC |
1656 | |
1657 | if (add1 < 0) | |
1658 | add1 += 0x7fff; | |
f4742949 | 1659 | pvt->udimm_ce_count[1] += add1; |
442305b1 MCC |
1660 | |
1661 | if (add0 < 0) | |
1662 | add0 += 0x7fff; | |
f4742949 | 1663 | pvt->udimm_ce_count[0] += add0; |
b4e8f0b6 MCC |
1664 | |
1665 | if (add0 | add1 | add2) | |
1666 | i7core_printk(KERN_ERR, "New Corrected error(s): " | |
1667 | "dimm0: +%d, dimm1: +%d, dimm2 +%d\n", | |
1668 | add0, add1, add2); | |
442305b1 | 1669 | } else |
f4742949 | 1670 | pvt->ce_count_available = 1; |
442305b1 MCC |
1671 | |
1672 | /* Store the new values */ | |
f4742949 MCC |
1673 | pvt->udimm_last_ce_count[2] = new2; |
1674 | pvt->udimm_last_ce_count[1] = new1; | |
1675 | pvt->udimm_last_ce_count[0] = new0; | |
442305b1 MCC |
1676 | } |
1677 | ||
8a2f118e MCC |
1678 | /* |
1679 | * According with tables E-11 and E-12 of chapter E.3.3 of Intel 64 and IA-32 | |
1680 | * Architectures Software Developer’s Manual Volume 3B. | |
f237fcf2 MCC |
1681 | * Nehalem are defined as family 0x06, model 0x1a |
1682 | * | |
1683 | * The MCA registers used here are the following ones: | |
8a2f118e | 1684 | * struct mce field MCA Register |
f237fcf2 MCC |
1685 | * m->status MSR_IA32_MC8_STATUS |
1686 | * m->addr MSR_IA32_MC8_ADDR | |
1687 | * m->misc MSR_IA32_MC8_MISC | |
8a2f118e MCC |
1688 | * In the case of Nehalem, the error information is masked at .status and .misc |
1689 | * fields | |
1690 | */ | |
d5381642 | 1691 | static void i7core_mce_output_error(struct mem_ctl_info *mci, |
1288c18f | 1692 | const struct mce *m) |
d5381642 | 1693 | { |
b4e8f0b6 | 1694 | struct i7core_pvt *pvt = mci->pvt_info; |
a639539f | 1695 | char *type, *optype, *err, *msg; |
8a2f118e | 1696 | unsigned long error = m->status & 0x1ff0000l; |
a639539f | 1697 | u32 optypenum = (m->status >> 4) & 0x07; |
8a2f118e MCC |
1698 | u32 core_err_cnt = (m->status >> 38) && 0x7fff; |
1699 | u32 dimm = (m->misc >> 16) & 0x3; | |
1700 | u32 channel = (m->misc >> 18) & 0x3; | |
1701 | u32 syndrome = m->misc >> 32; | |
1702 | u32 errnum = find_first_bit(&error, 32); | |
b4e8f0b6 | 1703 | int csrow; |
8a2f118e | 1704 | |
c5d34528 MCC |
1705 | if (m->mcgstatus & 1) |
1706 | type = "FATAL"; | |
1707 | else | |
1708 | type = "NON_FATAL"; | |
1709 | ||
a639539f | 1710 | switch (optypenum) { |
b990538a MCC |
1711 | case 0: |
1712 | optype = "generic undef request"; | |
1713 | break; | |
1714 | case 1: | |
1715 | optype = "read error"; | |
1716 | break; | |
1717 | case 2: | |
1718 | optype = "write error"; | |
1719 | break; | |
1720 | case 3: | |
1721 | optype = "addr/cmd error"; | |
1722 | break; | |
1723 | case 4: | |
1724 | optype = "scrubbing error"; | |
1725 | break; | |
1726 | default: | |
1727 | optype = "reserved"; | |
1728 | break; | |
a639539f MCC |
1729 | } |
1730 | ||
8a2f118e MCC |
1731 | switch (errnum) { |
1732 | case 16: | |
1733 | err = "read ECC error"; | |
1734 | break; | |
1735 | case 17: | |
1736 | err = "RAS ECC error"; | |
1737 | break; | |
1738 | case 18: | |
1739 | err = "write parity error"; | |
1740 | break; | |
1741 | case 19: | |
1742 | err = "redundacy loss"; | |
1743 | break; | |
1744 | case 20: | |
1745 | err = "reserved"; | |
1746 | break; | |
1747 | case 21: | |
1748 | err = "memory range error"; | |
1749 | break; | |
1750 | case 22: | |
1751 | err = "RTID out of range"; | |
1752 | break; | |
1753 | case 23: | |
1754 | err = "address parity error"; | |
1755 | break; | |
1756 | case 24: | |
1757 | err = "byte enable parity error"; | |
1758 | break; | |
1759 | default: | |
1760 | err = "unknown"; | |
d5381642 | 1761 | } |
d5381642 | 1762 | |
f237fcf2 | 1763 | /* FIXME: should convert addr into bank and rank information */ |
8a2f118e | 1764 | msg = kasprintf(GFP_ATOMIC, |
f4742949 | 1765 | "%s (addr = 0x%08llx, cpu=%d, Dimm=%d, Channel=%d, " |
a639539f | 1766 | "syndrome=0x%08x, count=%d, Err=%08llx:%08llx (%s: %s))\n", |
f4742949 | 1767 | type, (long long) m->addr, m->cpu, dimm, channel, |
a639539f MCC |
1768 | syndrome, core_err_cnt, (long long)m->status, |
1769 | (long long)m->misc, optype, err); | |
8a2f118e MCC |
1770 | |
1771 | debugf0("%s", msg); | |
d5381642 | 1772 | |
f4742949 | 1773 | csrow = pvt->csrow_map[channel][dimm]; |
b4e8f0b6 | 1774 | |
d5381642 | 1775 | /* Call the helper to output message */ |
b4e8f0b6 MCC |
1776 | if (m->mcgstatus & 1) |
1777 | edac_mc_handle_fbd_ue(mci, csrow, 0, | |
1778 | 0 /* FIXME: should be channel here */, msg); | |
f4742949 | 1779 | else if (!pvt->is_registered) |
b4e8f0b6 MCC |
1780 | edac_mc_handle_fbd_ce(mci, csrow, |
1781 | 0 /* FIXME: should be channel here */, msg); | |
8a2f118e MCC |
1782 | |
1783 | kfree(msg); | |
d5381642 MCC |
1784 | } |
1785 | ||
87d1d272 MCC |
1786 | /* |
1787 | * i7core_check_error Retrieve and process errors reported by the | |
1788 | * hardware. Called by the Core module. | |
1789 | */ | |
1790 | static void i7core_check_error(struct mem_ctl_info *mci) | |
1791 | { | |
d5381642 MCC |
1792 | struct i7core_pvt *pvt = mci->pvt_info; |
1793 | int i; | |
1794 | unsigned count = 0; | |
ca9c90ba | 1795 | struct mce *m; |
d5381642 | 1796 | |
ca9c90ba MCC |
1797 | /* |
1798 | * MCE first step: Copy all mce errors into a temporary buffer | |
1799 | * We use a double buffering here, to reduce the risk of | |
1800 | * loosing an error. | |
1801 | */ | |
1802 | smp_rmb(); | |
321ece4d MCC |
1803 | count = (pvt->mce_out + MCE_LOG_LEN - pvt->mce_in) |
1804 | % MCE_LOG_LEN; | |
ca9c90ba | 1805 | if (!count) |
8a311e17 | 1806 | goto check_ce_error; |
f4742949 | 1807 | |
ca9c90ba | 1808 | m = pvt->mce_outentry; |
321ece4d MCC |
1809 | if (pvt->mce_in + count > MCE_LOG_LEN) { |
1810 | unsigned l = MCE_LOG_LEN - pvt->mce_in; | |
f4742949 | 1811 | |
ca9c90ba MCC |
1812 | memcpy(m, &pvt->mce_entry[pvt->mce_in], sizeof(*m) * l); |
1813 | smp_wmb(); | |
1814 | pvt->mce_in = 0; | |
1815 | count -= l; | |
1816 | m += l; | |
1817 | } | |
1818 | memcpy(m, &pvt->mce_entry[pvt->mce_in], sizeof(*m) * count); | |
1819 | smp_wmb(); | |
1820 | pvt->mce_in += count; | |
1821 | ||
1822 | smp_rmb(); | |
1823 | if (pvt->mce_overrun) { | |
1824 | i7core_printk(KERN_ERR, "Lost %d memory errors\n", | |
1825 | pvt->mce_overrun); | |
1826 | smp_wmb(); | |
1827 | pvt->mce_overrun = 0; | |
1828 | } | |
d5381642 | 1829 | |
ca9c90ba MCC |
1830 | /* |
1831 | * MCE second step: parse errors and display | |
1832 | */ | |
d5381642 | 1833 | for (i = 0; i < count; i++) |
ca9c90ba | 1834 | i7core_mce_output_error(mci, &pvt->mce_outentry[i]); |
d5381642 | 1835 | |
ca9c90ba MCC |
1836 | /* |
1837 | * Now, let's increment CE error counts | |
1838 | */ | |
8a311e17 | 1839 | check_ce_error: |
f4742949 MCC |
1840 | if (!pvt->is_registered) |
1841 | i7core_udimm_check_mc_ecc_err(mci); | |
1842 | else | |
1843 | i7core_rdimm_check_mc_ecc_err(mci); | |
87d1d272 MCC |
1844 | } |
1845 | ||
d5381642 MCC |
1846 | /* |
1847 | * i7core_mce_check_error Replicates mcelog routine to get errors | |
1848 | * This routine simply queues mcelog errors, and | |
1849 | * return. The error itself should be handled later | |
1850 | * by i7core_check_error. | |
6e103be1 MCC |
1851 | * WARNING: As this routine should be called at NMI time, extra care should |
1852 | * be taken to avoid deadlocks, and to be as fast as possible. | |
d5381642 MCC |
1853 | */ |
1854 | static int i7core_mce_check_error(void *priv, struct mce *mce) | |
1855 | { | |
c5d34528 MCC |
1856 | struct mem_ctl_info *mci = priv; |
1857 | struct i7core_pvt *pvt = mci->pvt_info; | |
d5381642 | 1858 | |
8a2f118e MCC |
1859 | /* |
1860 | * Just let mcelog handle it if the error is | |
1861 | * outside the memory controller | |
1862 | */ | |
1863 | if (((mce->status & 0xffff) >> 7) != 1) | |
1864 | return 0; | |
1865 | ||
f237fcf2 MCC |
1866 | /* Bank 8 registers are the only ones that we know how to handle */ |
1867 | if (mce->bank != 8) | |
1868 | return 0; | |
1869 | ||
3b918c12 | 1870 | #ifdef CONFIG_SMP |
f4742949 | 1871 | /* Only handle if it is the right mc controller */ |
6e103be1 | 1872 | if (cpu_data(mce->cpu).phys_proc_id != pvt->i7core_dev->socket) |
f4742949 | 1873 | return 0; |
3b918c12 | 1874 | #endif |
f4742949 | 1875 | |
ca9c90ba | 1876 | smp_rmb(); |
321ece4d | 1877 | if ((pvt->mce_out + 1) % MCE_LOG_LEN == pvt->mce_in) { |
ca9c90ba MCC |
1878 | smp_wmb(); |
1879 | pvt->mce_overrun++; | |
1880 | return 0; | |
d5381642 | 1881 | } |
6e103be1 MCC |
1882 | |
1883 | /* Copy memory error at the ringbuffer */ | |
1884 | memcpy(&pvt->mce_entry[pvt->mce_out], mce, sizeof(*mce)); | |
ca9c90ba | 1885 | smp_wmb(); |
321ece4d | 1886 | pvt->mce_out = (pvt->mce_out + 1) % MCE_LOG_LEN; |
d5381642 | 1887 | |
c5d34528 MCC |
1888 | /* Handle fatal errors immediately */ |
1889 | if (mce->mcgstatus & 1) | |
1890 | i7core_check_error(mci); | |
1891 | ||
d5381642 | 1892 | /* Advice mcelog that the error were handled */ |
8a2f118e | 1893 | return 1; |
d5381642 MCC |
1894 | } |
1895 | ||
a3aa0a4a HS |
1896 | static void i7core_pci_ctl_create(struct i7core_pvt *pvt) |
1897 | { | |
1898 | pvt->i7core_pci = edac_pci_create_generic_ctl( | |
1899 | &pvt->i7core_dev->pdev[0]->dev, | |
1900 | EDAC_MOD_STR); | |
1901 | if (unlikely(!pvt->i7core_pci)) | |
1902 | pr_warn("Unable to setup PCI error report via EDAC\n"); | |
1903 | } | |
1904 | ||
1905 | static void i7core_pci_ctl_release(struct i7core_pvt *pvt) | |
1906 | { | |
1907 | if (likely(pvt->i7core_pci)) | |
1908 | edac_pci_release_generic_ctl(pvt->i7core_pci); | |
1909 | else | |
1910 | i7core_printk(KERN_ERR, | |
1911 | "Couldn't find mem_ctl_info for socket %d\n", | |
1912 | pvt->i7core_dev->socket); | |
1913 | pvt->i7core_pci = NULL; | |
1914 | } | |
1915 | ||
1c6edbbe HS |
1916 | static void i7core_unregister_mci(struct i7core_dev *i7core_dev) |
1917 | { | |
1918 | struct mem_ctl_info *mci = i7core_dev->mci; | |
1919 | struct i7core_pvt *pvt; | |
1920 | ||
1921 | if (unlikely(!mci || !mci->pvt_info)) { | |
1922 | debugf0("MC: " __FILE__ ": %s(): dev = %p\n", | |
1923 | __func__, &i7core_dev->pdev[0]->dev); | |
1924 | ||
1925 | i7core_printk(KERN_ERR, "Couldn't find mci handler\n"); | |
1926 | return; | |
1927 | } | |
1928 | ||
1929 | pvt = mci->pvt_info; | |
1930 | ||
1931 | debugf0("MC: " __FILE__ ": %s(): mci = %p, dev = %p\n", | |
1932 | __func__, mci, &i7core_dev->pdev[0]->dev); | |
1933 | ||
1934 | /* Disable MCE NMI handler */ | |
1935 | edac_mce_unregister(&pvt->edac_mce); | |
1936 | ||
1937 | /* Disable EDAC polling */ | |
1938 | i7core_pci_ctl_release(pvt); | |
1939 | ||
1940 | /* Remove MC sysfs nodes */ | |
1941 | edac_mc_del_mc(mci->dev); | |
1942 | ||
1943 | debugf1("%s: free mci struct\n", mci->ctl_name); | |
1944 | kfree(mci->ctl_name); | |
1945 | edac_mc_free(mci); | |
1946 | i7core_dev->mci = NULL; | |
1947 | } | |
1948 | ||
aace4283 | 1949 | static int i7core_register_mci(struct i7core_dev *i7core_dev) |
a0c36a1f MCC |
1950 | { |
1951 | struct mem_ctl_info *mci; | |
1952 | struct i7core_pvt *pvt; | |
aace4283 HS |
1953 | int rc, channels, csrows; |
1954 | ||
1955 | /* Check the number of active and not disabled channels */ | |
1956 | rc = i7core_get_active_channels(i7core_dev->socket, &channels, &csrows); | |
1957 | if (unlikely(rc < 0)) | |
1958 | return rc; | |
a0c36a1f | 1959 | |
a0c36a1f | 1960 | /* allocate a new MC control structure */ |
aace4283 | 1961 | mci = edac_mc_alloc(sizeof(*pvt), csrows, channels, i7core_dev->socket); |
f4742949 MCC |
1962 | if (unlikely(!mci)) |
1963 | return -ENOMEM; | |
a0c36a1f | 1964 | |
3cfd0146 MCC |
1965 | debugf0("MC: " __FILE__ ": %s(): mci = %p, dev = %p\n", |
1966 | __func__, mci, &i7core_dev->pdev[0]->dev); | |
a0c36a1f | 1967 | |
a0c36a1f | 1968 | pvt = mci->pvt_info; |
ef708b53 | 1969 | memset(pvt, 0, sizeof(*pvt)); |
67166af4 | 1970 | |
6d37d240 MCC |
1971 | /* Associates i7core_dev and mci for future usage */ |
1972 | pvt->i7core_dev = i7core_dev; | |
1973 | i7core_dev->mci = mci; | |
1974 | ||
41fcb7fe MCC |
1975 | /* |
1976 | * FIXME: how to handle RDDR3 at MCI level? It is possible to have | |
1977 | * Mixed RDDR3/UDDR3 with Nehalem, provided that they are on different | |
1978 | * memory channels | |
1979 | */ | |
1980 | mci->mtype_cap = MEM_FLAG_DDR3; | |
a0c36a1f MCC |
1981 | mci->edac_ctl_cap = EDAC_FLAG_NONE; |
1982 | mci->edac_cap = EDAC_FLAG_NONE; | |
1983 | mci->mod_name = "i7core_edac.c"; | |
1984 | mci->mod_ver = I7CORE_REVISION; | |
f4742949 MCC |
1985 | mci->ctl_name = kasprintf(GFP_KERNEL, "i7 core #%d", |
1986 | i7core_dev->socket); | |
1987 | mci->dev_name = pci_name(i7core_dev->pdev[0]); | |
a0c36a1f | 1988 | mci->ctl_page_to_phys = NULL; |
1288c18f | 1989 | |
ef708b53 | 1990 | /* Store pci devices at mci for faster access */ |
f4742949 | 1991 | rc = mci_bind_devs(mci, i7core_dev); |
41fcb7fe | 1992 | if (unlikely(rc < 0)) |
628c5ddf | 1993 | goto fail0; |
ef708b53 | 1994 | |
5939813b HS |
1995 | if (pvt->is_registered) |
1996 | mci->mc_driver_sysfs_attributes = i7core_sysfs_rdimm_attrs; | |
1997 | else | |
1998 | mci->mc_driver_sysfs_attributes = i7core_sysfs_udimm_attrs; | |
1999 | ||
ef708b53 | 2000 | /* Get dimm basic config */ |
2e5185f7 | 2001 | get_dimm_config(mci); |
5939813b HS |
2002 | /* record ptr to the generic device */ |
2003 | mci->dev = &i7core_dev->pdev[0]->dev; | |
2004 | /* Set the function pointer to an actual operation function */ | |
2005 | mci->edac_check = i7core_check_error; | |
ef708b53 | 2006 | |
a0c36a1f | 2007 | /* add this new MC control structure to EDAC's list of MCs */ |
b7c76151 | 2008 | if (unlikely(edac_mc_add_mc(mci))) { |
a0c36a1f MCC |
2009 | debugf0("MC: " __FILE__ |
2010 | ": %s(): failed edac_mc_add_mc()\n", __func__); | |
2011 | /* FIXME: perhaps some code should go here that disables error | |
2012 | * reporting if we just enabled it | |
2013 | */ | |
b7c76151 MCC |
2014 | |
2015 | rc = -EINVAL; | |
628c5ddf | 2016 | goto fail0; |
a0c36a1f MCC |
2017 | } |
2018 | ||
194a40fe | 2019 | /* Default error mask is any memory */ |
ef708b53 | 2020 | pvt->inject.channel = 0; |
194a40fe MCC |
2021 | pvt->inject.dimm = -1; |
2022 | pvt->inject.rank = -1; | |
2023 | pvt->inject.bank = -1; | |
2024 | pvt->inject.page = -1; | |
2025 | pvt->inject.col = -1; | |
2026 | ||
a3aa0a4a HS |
2027 | /* allocating generic PCI control info */ |
2028 | i7core_pci_ctl_create(pvt); | |
2029 | ||
d5381642 | 2030 | /* Registers on edac_mce in order to receive memory errors */ |
c5d34528 | 2031 | pvt->edac_mce.priv = mci; |
d5381642 | 2032 | pvt->edac_mce.check_error = i7core_mce_check_error; |
d5381642 | 2033 | rc = edac_mce_register(&pvt->edac_mce); |
b990538a | 2034 | if (unlikely(rc < 0)) { |
d5381642 MCC |
2035 | debugf0("MC: " __FILE__ |
2036 | ": %s(): failed edac_mce_register()\n", __func__); | |
628c5ddf | 2037 | goto fail1; |
f4742949 MCC |
2038 | } |
2039 | ||
628c5ddf HS |
2040 | return 0; |
2041 | ||
2042 | fail1: | |
2043 | i7core_pci_ctl_release(pvt); | |
2044 | edac_mc_del_mc(mci->dev); | |
2045 | fail0: | |
2046 | kfree(mci->ctl_name); | |
2047 | edac_mc_free(mci); | |
1c6edbbe | 2048 | i7core_dev->mci = NULL; |
f4742949 MCC |
2049 | return rc; |
2050 | } | |
2051 | ||
2052 | /* | |
2053 | * i7core_probe Probe for ONE instance of device to see if it is | |
2054 | * present. | |
2055 | * return: | |
2056 | * 0 for FOUND a device | |
2057 | * < 0 for error code | |
2058 | */ | |
2d95d815 | 2059 | |
f4742949 MCC |
2060 | static int __devinit i7core_probe(struct pci_dev *pdev, |
2061 | const struct pci_device_id *id) | |
2062 | { | |
f4742949 MCC |
2063 | int rc; |
2064 | struct i7core_dev *i7core_dev; | |
2065 | ||
2d95d815 MCC |
2066 | /* get the pci devices we want to reserve for our use */ |
2067 | mutex_lock(&i7core_edac_lock); | |
2068 | ||
f4742949 | 2069 | /* |
d4c27795 | 2070 | * All memory controllers are allocated at the first pass. |
f4742949 | 2071 | */ |
2d95d815 MCC |
2072 | if (unlikely(probed >= 1)) { |
2073 | mutex_unlock(&i7core_edac_lock); | |
f4742949 | 2074 | return -EINVAL; |
2d95d815 MCC |
2075 | } |
2076 | probed++; | |
de06eeef | 2077 | |
64c10f6e | 2078 | rc = i7core_get_all_devices(); |
f4742949 MCC |
2079 | if (unlikely(rc < 0)) |
2080 | goto fail0; | |
2081 | ||
2082 | list_for_each_entry(i7core_dev, &i7core_edac_list, list) { | |
aace4283 | 2083 | rc = i7core_register_mci(i7core_dev); |
d4c27795 MCC |
2084 | if (unlikely(rc < 0)) |
2085 | goto fail1; | |
d5381642 MCC |
2086 | } |
2087 | ||
ef708b53 | 2088 | i7core_printk(KERN_INFO, "Driver loaded.\n"); |
8f331907 | 2089 | |
66607706 | 2090 | mutex_unlock(&i7core_edac_lock); |
a0c36a1f MCC |
2091 | return 0; |
2092 | ||
66607706 | 2093 | fail1: |
1c6edbbe HS |
2094 | list_for_each_entry(i7core_dev, &i7core_edac_list, list) { |
2095 | if (i7core_dev->mci) | |
2096 | i7core_unregister_mci(i7core_dev); | |
2097 | } | |
13d6e9b6 | 2098 | i7core_put_all_devices(); |
66607706 MCC |
2099 | fail0: |
2100 | mutex_unlock(&i7core_edac_lock); | |
b7c76151 | 2101 | return rc; |
a0c36a1f MCC |
2102 | } |
2103 | ||
2104 | /* | |
2105 | * i7core_remove destructor for one instance of device | |
2106 | * | |
2107 | */ | |
2108 | static void __devexit i7core_remove(struct pci_dev *pdev) | |
2109 | { | |
64c10f6e | 2110 | struct i7core_dev *i7core_dev; |
a0c36a1f MCC |
2111 | |
2112 | debugf0(__FILE__ ": %s()\n", __func__); | |
2113 | ||
22e6bcbd MCC |
2114 | /* |
2115 | * we have a trouble here: pdev value for removal will be wrong, since | |
2116 | * it will point to the X58 register used to detect that the machine | |
2117 | * is a Nehalem or upper design. However, due to the way several PCI | |
2118 | * devices are grouped together to provide MC functionality, we need | |
2119 | * to use a different method for releasing the devices | |
2120 | */ | |
87d1d272 | 2121 | |
66607706 | 2122 | mutex_lock(&i7core_edac_lock); |
71fe0170 HS |
2123 | |
2124 | if (unlikely(!probed)) { | |
2125 | mutex_unlock(&i7core_edac_lock); | |
2126 | return; | |
2127 | } | |
2128 | ||
64c10f6e | 2129 | list_for_each_entry(i7core_dev, &i7core_edac_list, list) { |
1c6edbbe HS |
2130 | if (i7core_dev->mci) |
2131 | i7core_unregister_mci(i7core_dev); | |
22e6bcbd | 2132 | } |
64c10f6e HS |
2133 | |
2134 | /* Release PCI resources */ | |
2135 | i7core_put_all_devices(); | |
2136 | ||
2d95d815 MCC |
2137 | probed--; |
2138 | ||
66607706 | 2139 | mutex_unlock(&i7core_edac_lock); |
a0c36a1f MCC |
2140 | } |
2141 | ||
a0c36a1f MCC |
2142 | MODULE_DEVICE_TABLE(pci, i7core_pci_tbl); |
2143 | ||
2144 | /* | |
2145 | * i7core_driver pci_driver structure for this module | |
2146 | * | |
2147 | */ | |
2148 | static struct pci_driver i7core_driver = { | |
2149 | .name = "i7core_edac", | |
2150 | .probe = i7core_probe, | |
2151 | .remove = __devexit_p(i7core_remove), | |
2152 | .id_table = i7core_pci_tbl, | |
2153 | }; | |
2154 | ||
2155 | /* | |
2156 | * i7core_init Module entry function | |
2157 | * Try to initialize this module for its devices | |
2158 | */ | |
2159 | static int __init i7core_init(void) | |
2160 | { | |
2161 | int pci_rc; | |
2162 | ||
2163 | debugf2("MC: " __FILE__ ": %s()\n", __func__); | |
2164 | ||
2165 | /* Ensure that the OPSTATE is set correctly for POLL or NMI */ | |
2166 | opstate_init(); | |
2167 | ||
54a08ab1 MCC |
2168 | if (use_pci_fixup) |
2169 | i7core_xeon_pci_fixup(pci_dev_table); | |
bc2d7245 | 2170 | |
a0c36a1f MCC |
2171 | pci_rc = pci_register_driver(&i7core_driver); |
2172 | ||
3ef288a9 MCC |
2173 | if (pci_rc >= 0) |
2174 | return 0; | |
2175 | ||
2176 | i7core_printk(KERN_ERR, "Failed to register device with error %d.\n", | |
2177 | pci_rc); | |
2178 | ||
2179 | return pci_rc; | |
a0c36a1f MCC |
2180 | } |
2181 | ||
2182 | /* | |
2183 | * i7core_exit() Module exit function | |
2184 | * Unregister the driver | |
2185 | */ | |
2186 | static void __exit i7core_exit(void) | |
2187 | { | |
2188 | debugf2("MC: " __FILE__ ": %s()\n", __func__); | |
2189 | pci_unregister_driver(&i7core_driver); | |
2190 | } | |
2191 | ||
2192 | module_init(i7core_init); | |
2193 | module_exit(i7core_exit); | |
2194 | ||
2195 | MODULE_LICENSE("GPL"); | |
2196 | MODULE_AUTHOR("Mauro Carvalho Chehab <mchehab@redhat.com>"); | |
2197 | MODULE_AUTHOR("Red Hat Inc. (http://www.redhat.com)"); | |
2198 | MODULE_DESCRIPTION("MC Driver for Intel i7 Core memory controllers - " | |
2199 | I7CORE_REVISION); | |
2200 | ||
2201 | module_param(edac_op_state, int, 0444); | |
2202 | MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI"); |