blob: c64ba267902f6aae6145ca8916019fc3de556ef5 [file] [log] [blame]
Mauro Carvalho Chehab52707f92010-05-18 20:43:52 -03001/* Intel i7 core/Nehalem Memory Controller kernel module
2 *
David Sterbae7bf0682010-12-27 16:51:15 +01003 * This driver supports the memory controllers found on the Intel
Mauro Carvalho Chehab52707f92010-05-18 20:43:52 -03004 * 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.
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03007 *
8 * This file may be distributed under the terms of the
9 * GNU General Public License version 2 only.
10 *
Mauro Carvalho Chehab52707f92010-05-18 20:43:52 -030011 * Copyright (c) 2009-2010 by:
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -030012 * 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
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -030028#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>
Randy Dunlap3b918c12009-11-08 01:36:40 -020033#include <linux/delay.h>
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -030034#include <linux/edac.h>
35#include <linux/mmzone.h>
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -030036#include <linux/edac_mce.h>
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -030037#include <linux/smp.h>
Mauro Carvalho Chehab14d2c082009-09-02 23:52:36 -030038#include <asm/processor.h>
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -030039
40#include "edac_core.h"
41
Mauro Carvalho Chehab18c29002010-08-10 18:33:27 -030042/* Static vars */
43static LIST_HEAD(i7core_edac_list);
44static DEFINE_MUTEX(i7core_edac_lock);
45static int probed;
46
Mauro Carvalho Chehab54a08ab2010-08-19 15:51:00 -030047static int use_pci_fixup;
48module_param(use_pci_fixup, int, 0444);
49MODULE_PARM_DESC(use_pci_fixup, "Enable PCI fixup to seek for hidden devices");
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -030050/*
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -030051 * 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
59/*
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -030060 * Alter this version for the module when modifications are made
61 */
Michal Marek152ba392011-04-01 12:41:20 +020062#define I7CORE_REVISION " Ver: 1.0.0"
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -030063#define EDAC_MOD_STR "i7core_edac"
64
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -030065/*
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
Mauro Carvalho Chehabe9bd2e72009-07-09 22:14:35 -030078 /* OFFSETS for Device 0 Function 0 */
79
80#define MC_CFG_CONTROL 0x90
81
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -030082 /* OFFSETS for Device 3 Function 0 */
83
84#define MC_CONTROL 0x48
85#define MC_STATUS 0x4c
86#define MC_MAX_DOD 0x64
87
Mauro Carvalho Chehab442305b2009-06-22 22:48:29 -030088/*
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
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -0300100/* 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
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300112 /* OFFSETS for Devices 4,5 and 6 Function 0 */
113
Mauro Carvalho Chehab0b2b7b72009-06-22 22:48:29 -0300114#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
Mauro Carvalho Chehabf122a892009-06-22 22:48:29 -0300120#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
Mauro Carvalho Chehab0b2b7b72009-06-22 22:48:29 -0300124#define MC_CHANNEL_RANK_PRESENT 0x7c
125 #define RANK_PRESENT_MASK 0xffff
126
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300127#define MC_CHANNEL_ADDR_MATCH 0xf0
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -0300128#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
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300138
Mauro Carvalho Chehab0b2b7b72009-06-22 22:48:29 -0300139 /* OFFSETS for Devices 4,5 and 6 Function 1 */
Mauro Carvalho Chehabb9905382009-08-05 21:36:35 -0300140
Mauro Carvalho Chehab0b2b7b72009-06-22 22:48:29 -0300141#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)
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300148 #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)
Mauro Carvalho Chehab41fcb7f2009-06-22 22:48:31 -0300152 #define MC_DOD_NUMROW_MASK ((1 << 4) | (1 << 3) | (1 << 2))
Mauro Carvalho Chehab5566cb72009-06-22 22:48:31 -0300153 #define MC_DOD_NUMROW(x) (((x) & MC_DOD_NUMROW_MASK) >> 2)
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300154 #define MC_DOD_NUMCOL_MASK 3
155 #define MC_DOD_NUMCOL(x) ((x) & MC_DOD_NUMCOL_MASK)
Mauro Carvalho Chehab0b2b7b72009-06-22 22:48:29 -0300156
Mauro Carvalho Chehabf122a892009-06-22 22:48:29 -0300157#define MC_RANK_PRESENT 0x7c
158
Mauro Carvalho Chehab0b2b7b72009-06-22 22:48:29 -0300159#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
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300182/*
183 * i7core structs
184 */
185
186#define NUM_CHANS 3
Mauro Carvalho Chehab442305b2009-06-22 22:48:29 -0300187#define MAX_DIMMS 3 /* Max DIMMS per channel */
188#define MAX_MCR_FUNC 4
189#define MAX_CHAN_FUNC 3
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300190
191struct i7core_info {
192 u32 mc_control;
193 u32 mc_status;
194 u32 max_dod;
Mauro Carvalho Chehabf122a892009-06-22 22:48:29 -0300195 u32 ch_map;
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300196};
197
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -0300198
199struct 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
Mauro Carvalho Chehab0b2b7b72009-06-22 22:48:29 -0300210struct i7core_channel {
Mauro Carvalho Chehab442305b2009-06-22 22:48:29 -0300211 u32 ranks;
212 u32 dimms;
Mauro Carvalho Chehab0b2b7b72009-06-22 22:48:29 -0300213};
214
Mauro Carvalho Chehab8f331902009-06-22 22:48:29 -0300215struct pci_id_descr {
Mauro Carvalho Chehab66607702009-09-05 00:52:11 -0300216 int dev;
217 int func;
218 int dev_id;
Mauro Carvalho Chehabde06eee2009-10-14 08:02:40 -0300219 int optional;
Mauro Carvalho Chehab8f331902009-06-22 22:48:29 -0300220};
221
Vernon Mauerybd9e19c2010-05-18 19:02:50 -0300222struct pci_id_table {
Mauro Carvalho Chehab1288c182010-08-10 18:57:01 -0300223 const struct pci_id_descr *descr;
224 int n_devs;
Vernon Mauerybd9e19c2010-05-18 19:02:50 -0300225};
226
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -0300227struct i7core_dev {
228 struct list_head list;
229 u8 socket;
230 struct pci_dev **pdev;
Mauro Carvalho Chehabde06eee2009-10-14 08:02:40 -0300231 int n_devs;
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -0300232 struct mem_ctl_info *mci;
233};
234
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300235struct i7core_pvt {
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -0300236 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;
Mauro Carvalho Chehab67166af2009-07-15 06:56:23 -0300241
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300242 struct i7core_info info;
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -0300243 struct i7core_inject inject;
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -0300244 struct i7core_channel channel[NUM_CHANS];
Mauro Carvalho Chehab67166af2009-07-15 06:56:23 -0300245
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -0300246 int ce_count_available;
247 int csrow_map[NUM_CHANS][MAX_DIMMS];
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -0300248
249 /* ECC corrected errors counts per udimm */
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -0300250 unsigned long udimm_ce_count[MAX_DIMMS];
251 int udimm_last_ce_count[MAX_DIMMS];
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -0300252 /* ECC corrected errors counts per rdimm */
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -0300253 unsigned long rdimm_ce_count[NUM_CHANS][MAX_DIMMS];
254 int rdimm_last_ce_count[NUM_CHANS][MAX_DIMMS];
Mauro Carvalho Chehab442305b2009-06-22 22:48:29 -0300255
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -0300256 unsigned int is_registered;
Mauro Carvalho Chehab14d2c082009-09-02 23:52:36 -0300257
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -0300258 /* mcelog glue */
259 struct edac_mce edac_mce;
Mauro Carvalho Chehabca9c90b2009-10-04 10:15:40 -0300260
261 /* Fifo double buffers */
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -0300262 struct mce mce_entry[MCE_LOG_LEN];
Mauro Carvalho Chehabca9c90b2009-10-04 10:15:40 -0300263 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;
Mauro Carvalho Chehab939747bd2010-08-10 11:22:01 -0300270
271 /* Struct to control EDAC polling */
272 struct edac_pci_ctl_info *i7core_pci;
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300273};
274
Mauro Carvalho Chehab8f331902009-06-22 22:48:29 -0300275#define PCI_DESCR(device, function, device_id) \
276 .dev = (device), \
277 .func = (function), \
278 .dev_id = (device_id)
279
Mauro Carvalho Chehab1288c182010-08-10 18:57:01 -0300280static const struct pci_id_descr pci_dev_descr_i7core_nehalem[] = {
Mauro Carvalho Chehab8f331902009-06-22 22:48:29 -0300281 /* 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) },
Mauro Carvalho Chehab224e8712011-03-17 17:02:59 -0300284 /* Exists only for RDIMM */
Mauro Carvalho Chehabde06eee2009-10-14 08:02:40 -0300285 { PCI_DESCR(3, 2, PCI_DEVICE_ID_INTEL_I7_MC_RAS), .optional = 1 },
Mauro Carvalho Chehab8f331902009-06-22 22:48:29 -0300286 { 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) },
Mauro Carvalho Chehab224e8712011-03-17 17:02:59 -0300305
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 */
313 { PCI_DESCR(0, 0, PCI_DEVICE_ID_INTEL_I7_NONCORE) },
314
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300315};
Mauro Carvalho Chehab8f331902009-06-22 22:48:29 -0300316
Mauro Carvalho Chehab1288c182010-08-10 18:57:01 -0300317static const struct pci_id_descr pci_dev_descr_lynnfield[] = {
Mauro Carvalho Chehab52a2e4f2009-10-14 11:21:58 -0300318 { 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
Mauro Carvalho Chehab508fa172009-10-14 13:44:37 -0300327 { 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) },
Mauro Carvalho Chehab224e8712011-03-17 17:02:59 -0300331
332 /*
333 * This is the PCI device has an alternate address on some
334 * processors like Core i7 860
335 */
336 { PCI_DESCR( 0, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE) },
Mauro Carvalho Chehab52a2e4f2009-10-14 11:21:58 -0300337};
338
Mauro Carvalho Chehab1288c182010-08-10 18:57:01 -0300339static const struct pci_id_descr pci_dev_descr_i7core_westmere[] = {
Vernon Mauerybd9e19c2010-05-18 19:02:50 -0300340 /* 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) },
Mauro Carvalho Chehab224e8712011-03-17 17:02:59 -0300364
365 /* Generic Non-core registers */
366 { PCI_DESCR(0, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE_REV2) },
367
Vernon Mauerybd9e19c2010-05-18 19:02:50 -0300368};
369
Mauro Carvalho Chehab1288c182010-08-10 18:57:01 -0300370#define PCI_ID_TABLE_ENTRY(A) { .descr=A, .n_devs = ARRAY_SIZE(A) }
371static const struct pci_id_table pci_dev_table[] = {
Vernon Mauerybd9e19c2010-05-18 19:02:50 -0300372 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),
Mauro Carvalho Chehab3c52cc52010-10-24 11:12:28 -0200375 {0,} /* 0 terminated list. */
Vernon Mauerybd9e19c2010-05-18 19:02:50 -0300376};
377
Mauro Carvalho Chehab8f331902009-06-22 22:48:29 -0300378/*
379 * pci_device_id table for which devices we are looking for
Mauro Carvalho Chehab8f331902009-06-22 22:48:29 -0300380 */
381static const struct pci_device_id i7core_pci_tbl[] __devinitdata = {
Mauro Carvalho Chehabd1fd4fb2009-07-10 18:39:53 -0300382 {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_X58_HUB_MGMT)},
Mauro Carvalho Chehabf05da2f2009-10-14 13:31:06 -0300383 {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_LYNNFIELD_QPI_LINK0)},
Mauro Carvalho Chehab8f331902009-06-22 22:48:29 -0300384 {0,} /* 0 terminated list. */
385};
386
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300387/****************************************************************************
388 Anciliary status routines
389 ****************************************************************************/
390
391 /* MC_CONTROL bits */
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -0300392#define CH_ACTIVE(pvt, ch) ((pvt)->info.mc_control & (1 << (8 + ch)))
393#define ECCx8(pvt) ((pvt)->info.mc_control & (1 << 1))
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300394
395 /* MC_STATUS bits */
Keith Mannthey61053fd2009-09-02 23:46:59 -0300396#define ECC_ENABLED(pvt) ((pvt)->info.mc_status & (1 << 4))
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -0300397#define CH_DISABLED(pvt, ch) ((pvt)->info.mc_status & (1 << ch))
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300398
399 /* MC_MAX_DOD read functions */
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300400static inline int numdimms(u32 dimms)
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300401{
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300402 return (dimms & 0x3) + 1;
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300403}
404
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300405static inline int numrank(u32 rank)
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300406{
407 static int ranks[4] = { 1, 2, 4, -EINVAL };
408
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300409 return ranks[rank & 0x3];
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300410}
411
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300412static inline int numbank(u32 bank)
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300413{
414 static int banks[4] = { 4, 8, 16, -EINVAL };
415
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300416 return banks[bank & 0x3];
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300417}
418
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300419static inline int numrow(u32 row)
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300420{
421 static int rows[8] = {
422 1 << 12, 1 << 13, 1 << 14, 1 << 15,
423 1 << 16, -EINVAL, -EINVAL, -EINVAL,
424 };
425
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300426 return rows[row & 0x7];
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300427}
428
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300429static inline int numcol(u32 col)
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300430{
431 static int cols[8] = {
432 1 << 10, 1 << 11, 1 << 12, -EINVAL,
433 };
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300434 return cols[col & 0x3];
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300435}
436
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -0300437static struct i7core_dev *get_i7core_dev(u8 socket)
Mauro Carvalho Chehab66607702009-09-05 00:52:11 -0300438{
439 struct i7core_dev *i7core_dev;
440
441 list_for_each_entry(i7core_dev, &i7core_edac_list, list) {
442 if (i7core_dev->socket == socket)
443 return i7core_dev;
444 }
445
446 return NULL;
447}
448
Hidetoshi Seto848b2f72010-08-20 04:24:44 -0300449static struct i7core_dev *alloc_i7core_dev(u8 socket,
450 const struct pci_id_table *table)
451{
452 struct i7core_dev *i7core_dev;
453
454 i7core_dev = kzalloc(sizeof(*i7core_dev), GFP_KERNEL);
455 if (!i7core_dev)
456 return NULL;
457
458 i7core_dev->pdev = kzalloc(sizeof(*i7core_dev->pdev) * table->n_devs,
459 GFP_KERNEL);
460 if (!i7core_dev->pdev) {
461 kfree(i7core_dev);
462 return NULL;
463 }
464
465 i7core_dev->socket = socket;
466 i7core_dev->n_devs = table->n_devs;
467 list_add_tail(&i7core_dev->list, &i7core_edac_list);
468
469 return i7core_dev;
470}
471
Hidetoshi Seto2aa9be42010-08-20 04:25:00 -0300472static void free_i7core_dev(struct i7core_dev *i7core_dev)
473{
474 list_del(&i7core_dev->list);
475 kfree(i7core_dev->pdev);
476 kfree(i7core_dev);
477}
478
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300479/****************************************************************************
480 Memory check routines
481 ****************************************************************************/
Mauro Carvalho Chehab67166af2009-07-15 06:56:23 -0300482static struct pci_dev *get_pdev_slot_func(u8 socket, unsigned slot,
483 unsigned func)
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -0300484{
Mauro Carvalho Chehab66607702009-09-05 00:52:11 -0300485 struct i7core_dev *i7core_dev = get_i7core_dev(socket);
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -0300486 int i;
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -0300487
Mauro Carvalho Chehab66607702009-09-05 00:52:11 -0300488 if (!i7core_dev)
489 return NULL;
490
Mauro Carvalho Chehabde06eee2009-10-14 08:02:40 -0300491 for (i = 0; i < i7core_dev->n_devs; i++) {
Mauro Carvalho Chehab66607702009-09-05 00:52:11 -0300492 if (!i7core_dev->pdev[i])
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -0300493 continue;
494
Mauro Carvalho Chehab66607702009-09-05 00:52:11 -0300495 if (PCI_SLOT(i7core_dev->pdev[i]->devfn) == slot &&
496 PCI_FUNC(i7core_dev->pdev[i]->devfn) == func) {
497 return i7core_dev->pdev[i];
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -0300498 }
499 }
500
Mauro Carvalho Chehabeb94fc42009-06-22 22:48:31 -0300501 return NULL;
502}
503
Mauro Carvalho Chehabec6df242009-07-18 10:44:30 -0300504/**
505 * i7core_get_active_channels() - gets the number of channels and csrows
506 * @socket: Quick Path Interconnect socket
507 * @channels: Number of channels that will be returned
508 * @csrows: Number of csrows found
509 *
510 * Since EDAC core needs to know in advance the number of available channels
511 * and csrows, in order to allocate memory for csrows/channels, it is needed
512 * to run two similar steps. At the first step, implemented on this function,
513 * it checks the number of csrows/channels present at one socket.
514 * this is used in order to properly allocate the size of mci components.
515 *
516 * It should be noticed that none of the current available datasheets explain
517 * or even mention how csrows are seen by the memory controller. So, we need
518 * to add a fake description for csrows.
519 * So, this driver is attributing one DIMM memory for one csrow.
520 */
Mauro Carvalho Chehab1288c182010-08-10 18:57:01 -0300521static int i7core_get_active_channels(const u8 socket, unsigned *channels,
Mauro Carvalho Chehab67166af2009-07-15 06:56:23 -0300522 unsigned *csrows)
Mauro Carvalho Chehabeb94fc42009-06-22 22:48:31 -0300523{
524 struct pci_dev *pdev = NULL;
525 int i, j;
526 u32 status, control;
527
528 *channels = 0;
529 *csrows = 0;
530
Mauro Carvalho Chehab67166af2009-07-15 06:56:23 -0300531 pdev = get_pdev_slot_func(socket, 3, 0);
Mauro Carvalho Chehabb7c76152009-06-22 22:48:30 -0300532 if (!pdev) {
Mauro Carvalho Chehab67166af2009-07-15 06:56:23 -0300533 i7core_printk(KERN_ERR, "Couldn't find socket %d fn 3.0!!!\n",
534 socket);
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -0300535 return -ENODEV;
Mauro Carvalho Chehabb7c76152009-06-22 22:48:30 -0300536 }
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -0300537
538 /* Device 3 function 0 reads */
539 pci_read_config_dword(pdev, MC_STATUS, &status);
540 pci_read_config_dword(pdev, MC_CONTROL, &control);
541
542 for (i = 0; i < NUM_CHANS; i++) {
Mauro Carvalho Chehabeb94fc42009-06-22 22:48:31 -0300543 u32 dimm_dod[3];
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -0300544 /* Check if the channel is active */
545 if (!(control & (1 << (8 + i))))
546 continue;
547
548 /* Check if the channel is disabled */
Mauro Carvalho Chehab41fcb7f2009-06-22 22:48:31 -0300549 if (status & (1 << i))
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -0300550 continue;
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -0300551
Mauro Carvalho Chehab67166af2009-07-15 06:56:23 -0300552 pdev = get_pdev_slot_func(socket, i + 4, 1);
Mauro Carvalho Chehabeb94fc42009-06-22 22:48:31 -0300553 if (!pdev) {
Mauro Carvalho Chehab67166af2009-07-15 06:56:23 -0300554 i7core_printk(KERN_ERR, "Couldn't find socket %d "
555 "fn %d.%d!!!\n",
556 socket, i + 4, 1);
Mauro Carvalho Chehabeb94fc42009-06-22 22:48:31 -0300557 return -ENODEV;
558 }
559 /* Devices 4-6 function 1 */
560 pci_read_config_dword(pdev,
561 MC_DOD_CH_DIMM0, &dimm_dod[0]);
562 pci_read_config_dword(pdev,
563 MC_DOD_CH_DIMM1, &dimm_dod[1]);
564 pci_read_config_dword(pdev,
565 MC_DOD_CH_DIMM2, &dimm_dod[2]);
566
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -0300567 (*channels)++;
Mauro Carvalho Chehabeb94fc42009-06-22 22:48:31 -0300568
569 for (j = 0; j < 3; j++) {
570 if (!DIMM_PRESENT(dimm_dod[j]))
571 continue;
572 (*csrows)++;
573 }
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -0300574 }
575
Mauro Carvalho Chehabc77720b2009-07-18 10:43:08 -0300576 debugf0("Number of active channels on socket %d: %d\n",
Mauro Carvalho Chehab67166af2009-07-15 06:56:23 -0300577 socket, *channels);
Mauro Carvalho Chehab1c6fed82009-06-22 22:48:30 -0300578
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -0300579 return 0;
580}
581
Hidetoshi Seto2e5185f2010-08-20 04:32:56 -0300582static int get_dimm_config(const struct mem_ctl_info *mci)
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300583{
584 struct i7core_pvt *pvt = mci->pvt_info;
Mauro Carvalho Chehab1c6fed82009-06-22 22:48:30 -0300585 struct csrow_info *csr;
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300586 struct pci_dev *pdev;
Mauro Carvalho Chehabba6c5c62009-07-15 09:02:32 -0300587 int i, j;
Hidetoshi Seto2e5185f2010-08-20 04:32:56 -0300588 int csrow = 0;
Mauro Carvalho Chehab5566cb72009-06-22 22:48:31 -0300589 unsigned long last_page = 0;
Mauro Carvalho Chehab1c6fed82009-06-22 22:48:30 -0300590 enum edac_type mode;
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300591 enum mem_type mtype;
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300592
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300593 /* Get data from the MC register, function 0 */
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -0300594 pdev = pvt->pci_mcr[0];
Mauro Carvalho Chehab7dd69532009-06-22 22:48:30 -0300595 if (!pdev)
Mauro Carvalho Chehab8f331902009-06-22 22:48:29 -0300596 return -ENODEV;
597
Mauro Carvalho Chehabf122a892009-06-22 22:48:29 -0300598 /* Device 3 function 0 reads */
Mauro Carvalho Chehab7dd69532009-06-22 22:48:30 -0300599 pci_read_config_dword(pdev, MC_CONTROL, &pvt->info.mc_control);
600 pci_read_config_dword(pdev, MC_STATUS, &pvt->info.mc_status);
601 pci_read_config_dword(pdev, MC_MAX_DOD, &pvt->info.max_dod);
602 pci_read_config_dword(pdev, MC_CHANNEL_MAPPER, &pvt->info.ch_map);
Mauro Carvalho Chehabf122a892009-06-22 22:48:29 -0300603
Mauro Carvalho Chehab17cb7b02009-07-20 18:48:18 -0300604 debugf0("QPI %d control=0x%08x status=0x%08x dod=0x%08x map=0x%08x\n",
Mauro Carvalho Chehab4af91882009-09-24 09:58:26 -0300605 pvt->i7core_dev->socket, pvt->info.mc_control, pvt->info.mc_status,
Mauro Carvalho Chehabf122a892009-06-22 22:48:29 -0300606 pvt->info.max_dod, pvt->info.ch_map);
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300607
Mauro Carvalho Chehab1c6fed82009-06-22 22:48:30 -0300608 if (ECC_ENABLED(pvt)) {
Mauro Carvalho Chehab41fcb7f2009-06-22 22:48:31 -0300609 debugf0("ECC enabled with x%d SDCC\n", ECCx8(pvt) ? 8 : 4);
Mauro Carvalho Chehab1c6fed82009-06-22 22:48:30 -0300610 if (ECCx8(pvt))
611 mode = EDAC_S8ECD8ED;
612 else
613 mode = EDAC_S4ECD4ED;
614 } else {
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300615 debugf0("ECC disabled\n");
Mauro Carvalho Chehab1c6fed82009-06-22 22:48:30 -0300616 mode = EDAC_NONE;
617 }
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300618
619 /* FIXME: need to handle the error codes */
Mauro Carvalho Chehab17cb7b02009-07-20 18:48:18 -0300620 debugf0("DOD Max limits: DIMMS: %d, %d-ranked, %d-banked "
621 "x%x x 0x%x\n",
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300622 numdimms(pvt->info.max_dod),
623 numrank(pvt->info.max_dod >> 2),
Mauro Carvalho Chehab276b8242009-07-22 21:45:50 -0300624 numbank(pvt->info.max_dod >> 4),
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300625 numrow(pvt->info.max_dod >> 6),
626 numcol(pvt->info.max_dod >> 9));
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300627
Mauro Carvalho Chehab0b2b7b72009-06-22 22:48:29 -0300628 for (i = 0; i < NUM_CHANS; i++) {
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300629 u32 data, dimm_dod[3], value[8];
Mauro Carvalho Chehab0b2b7b72009-06-22 22:48:29 -0300630
Mauro Carvalho Chehab52a2e4f2009-10-14 11:21:58 -0300631 if (!pvt->pci_ch[i][0])
632 continue;
633
Mauro Carvalho Chehab0b2b7b72009-06-22 22:48:29 -0300634 if (!CH_ACTIVE(pvt, i)) {
635 debugf0("Channel %i is not active\n", i);
636 continue;
637 }
638 if (CH_DISABLED(pvt, i)) {
639 debugf0("Channel %i is disabled\n", i);
640 continue;
641 }
642
Mauro Carvalho Chehabf122a892009-06-22 22:48:29 -0300643 /* Devices 4-6 function 0 */
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -0300644 pci_read_config_dword(pvt->pci_ch[i][0],
Mauro Carvalho Chehab0b2b7b72009-06-22 22:48:29 -0300645 MC_CHANNEL_DIMM_INIT_PARAMS, &data);
646
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -0300647 pvt->channel[i].ranks = (data & QUAD_RANK_PRESENT) ?
Mauro Carvalho Chehab67166af2009-07-15 06:56:23 -0300648 4 : 2;
Mauro Carvalho Chehab0b2b7b72009-06-22 22:48:29 -0300649
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300650 if (data & REGISTERED_DIMM)
651 mtype = MEM_RDDR3;
Mauro Carvalho Chehab14d2c082009-09-02 23:52:36 -0300652 else
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300653 mtype = MEM_DDR3;
654#if 0
Mauro Carvalho Chehab0b2b7b72009-06-22 22:48:29 -0300655 if (data & THREE_DIMMS_PRESENT)
656 pvt->channel[i].dimms = 3;
657 else if (data & SINGLE_QUAD_RANK_PRESENT)
658 pvt->channel[i].dimms = 1;
659 else
660 pvt->channel[i].dimms = 2;
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300661#endif
662
663 /* Devices 4-6 function 1 */
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -0300664 pci_read_config_dword(pvt->pci_ch[i][1],
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300665 MC_DOD_CH_DIMM0, &dimm_dod[0]);
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -0300666 pci_read_config_dword(pvt->pci_ch[i][1],
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300667 MC_DOD_CH_DIMM1, &dimm_dod[1]);
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -0300668 pci_read_config_dword(pvt->pci_ch[i][1],
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300669 MC_DOD_CH_DIMM2, &dimm_dod[2]);
Mauro Carvalho Chehab0b2b7b72009-06-22 22:48:29 -0300670
Mauro Carvalho Chehab1c6fed82009-06-22 22:48:30 -0300671 debugf0("Ch%d phy rd%d, wr%d (0x%08x): "
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300672 "%d ranks, %cDIMMs\n",
Mauro Carvalho Chehab1c6fed82009-06-22 22:48:30 -0300673 i,
674 RDLCH(pvt->info.ch_map, i), WRLCH(pvt->info.ch_map, i),
675 data,
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -0300676 pvt->channel[i].ranks,
Mauro Carvalho Chehab41fcb7f2009-06-22 22:48:31 -0300677 (data & REGISTERED_DIMM) ? 'R' : 'U');
Mauro Carvalho Chehab7dd69532009-06-22 22:48:30 -0300678
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300679 for (j = 0; j < 3; j++) {
680 u32 banks, ranks, rows, cols;
Mauro Carvalho Chehab5566cb72009-06-22 22:48:31 -0300681 u32 size, npages;
Mauro Carvalho Chehab1c6fed82009-06-22 22:48:30 -0300682
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300683 if (!DIMM_PRESENT(dimm_dod[j]))
684 continue;
Mauro Carvalho Chehab1c6fed82009-06-22 22:48:30 -0300685
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300686 banks = numbank(MC_DOD_NUMBANK(dimm_dod[j]));
687 ranks = numrank(MC_DOD_NUMRANK(dimm_dod[j]));
688 rows = numrow(MC_DOD_NUMROW(dimm_dod[j]));
689 cols = numcol(MC_DOD_NUMCOL(dimm_dod[j]));
Mauro Carvalho Chehab1c6fed82009-06-22 22:48:30 -0300690
Mauro Carvalho Chehab5566cb72009-06-22 22:48:31 -0300691 /* DDR3 has 8 I/O banks */
692 size = (rows * cols * banks * ranks) >> (20 - 3);
693
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -0300694 pvt->channel[i].dimms++;
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300695
Mauro Carvalho Chehab17cb7b02009-07-20 18:48:18 -0300696 debugf0("\tdimm %d %d Mb offset: %x, "
697 "bank: %d, rank: %d, row: %#x, col: %#x\n",
698 j, size,
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300699 RANKOFFSET(dimm_dod[j]),
700 banks, ranks, rows, cols);
701
Mauro Carvalho Chehabe9144602010-08-10 20:26:35 -0300702 npages = MiB_TO_PAGES(size);
Mauro Carvalho Chehab5566cb72009-06-22 22:48:31 -0300703
Hidetoshi Seto2e5185f2010-08-20 04:32:56 -0300704 csr = &mci->csrows[csrow];
Mauro Carvalho Chehab5566cb72009-06-22 22:48:31 -0300705 csr->first_page = last_page + 1;
706 last_page += npages;
707 csr->last_page = last_page;
708 csr->nr_pages = npages;
709
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300710 csr->page_mask = 0;
Mauro Carvalho Chehabeb94fc42009-06-22 22:48:31 -0300711 csr->grain = 8;
Hidetoshi Seto2e5185f2010-08-20 04:32:56 -0300712 csr->csrow_idx = csrow;
Mauro Carvalho Chehabeb94fc42009-06-22 22:48:31 -0300713 csr->nr_channels = 1;
714
715 csr->channels[0].chan_idx = i;
716 csr->channels[0].ce_count = 0;
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300717
Hidetoshi Seto2e5185f2010-08-20 04:32:56 -0300718 pvt->csrow_map[i][j] = csrow;
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -0300719
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300720 switch (banks) {
721 case 4:
722 csr->dtype = DEV_X4;
723 break;
724 case 8:
725 csr->dtype = DEV_X8;
726 break;
727 case 16:
728 csr->dtype = DEV_X16;
729 break;
730 default:
731 csr->dtype = DEV_UNKNOWN;
732 }
733
734 csr->edac_mode = mode;
735 csr->mtype = mtype;
736
Hidetoshi Seto2e5185f2010-08-20 04:32:56 -0300737 csrow++;
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300738 }
739
740 pci_read_config_dword(pdev, MC_SAG_CH_0, &value[0]);
741 pci_read_config_dword(pdev, MC_SAG_CH_1, &value[1]);
742 pci_read_config_dword(pdev, MC_SAG_CH_2, &value[2]);
743 pci_read_config_dword(pdev, MC_SAG_CH_3, &value[3]);
744 pci_read_config_dword(pdev, MC_SAG_CH_4, &value[4]);
745 pci_read_config_dword(pdev, MC_SAG_CH_5, &value[5]);
746 pci_read_config_dword(pdev, MC_SAG_CH_6, &value[6]);
747 pci_read_config_dword(pdev, MC_SAG_CH_7, &value[7]);
Mauro Carvalho Chehab17cb7b02009-07-20 18:48:18 -0300748 debugf1("\t[%i] DIVBY3\tREMOVED\tOFFSET\n", i);
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300749 for (j = 0; j < 8; j++)
Mauro Carvalho Chehab17cb7b02009-07-20 18:48:18 -0300750 debugf1("\t\t%#x\t%#x\t%#x\n",
Mauro Carvalho Chehab854d3342009-06-22 22:48:30 -0300751 (value[j] >> 27) & 0x1,
752 (value[j] >> 24) & 0x7,
David Sterba80b8ce82010-12-27 15:39:12 +0000753 (value[j] & ((1 << 24) - 1)));
Mauro Carvalho Chehab0b2b7b72009-06-22 22:48:29 -0300754 }
755
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -0300756 return 0;
757}
758
759/****************************************************************************
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -0300760 Error insertion routines
761 ****************************************************************************/
762
763/* The i7core has independent error injection features per channel.
764 However, to have a simpler code, we don't allow enabling error injection
765 on more than one channel.
766 Also, since a change at an inject parameter will be applied only at enable,
767 we're disabling error injection on all write calls to the sysfs nodes that
768 controls the error code injection.
769 */
Mauro Carvalho Chehab1288c182010-08-10 18:57:01 -0300770static int disable_inject(const struct mem_ctl_info *mci)
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -0300771{
772 struct i7core_pvt *pvt = mci->pvt_info;
773
774 pvt->inject.enable = 0;
775
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -0300776 if (!pvt->pci_ch[pvt->inject.channel][0])
Mauro Carvalho Chehab8f331902009-06-22 22:48:29 -0300777 return -ENODEV;
778
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -0300779 pci_write_config_dword(pvt->pci_ch[pvt->inject.channel][0],
Mauro Carvalho Chehab4157d9f2009-08-05 20:27:15 -0300780 MC_CHANNEL_ERROR_INJECT, 0);
Mauro Carvalho Chehab8f331902009-06-22 22:48:29 -0300781
782 return 0;
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -0300783}
784
785/*
786 * i7core inject inject.section
787 *
788 * accept and store error injection inject.section value
789 * bit 0 - refers to the lower 32-byte half cacheline
790 * bit 1 - refers to the upper 32-byte half cacheline
791 */
792static ssize_t i7core_inject_section_store(struct mem_ctl_info *mci,
793 const char *data, size_t count)
794{
795 struct i7core_pvt *pvt = mci->pvt_info;
796 unsigned long value;
797 int rc;
798
799 if (pvt->inject.enable)
Mauro Carvalho Chehab41fcb7f2009-06-22 22:48:31 -0300800 disable_inject(mci);
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -0300801
802 rc = strict_strtoul(data, 10, &value);
803 if ((rc < 0) || (value > 3))
Mauro Carvalho Chehab2068def2009-08-05 19:28:27 -0300804 return -EIO;
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -0300805
806 pvt->inject.section = (u32) value;
807 return count;
808}
809
810static ssize_t i7core_inject_section_show(struct mem_ctl_info *mci,
811 char *data)
812{
813 struct i7core_pvt *pvt = mci->pvt_info;
814 return sprintf(data, "0x%08x\n", pvt->inject.section);
815}
816
817/*
818 * i7core inject.type
819 *
820 * accept and store error injection inject.section value
821 * bit 0 - repeat enable - Enable error repetition
822 * bit 1 - inject ECC error
823 * bit 2 - inject parity error
824 */
825static ssize_t i7core_inject_type_store(struct mem_ctl_info *mci,
826 const char *data, size_t count)
827{
828 struct i7core_pvt *pvt = mci->pvt_info;
829 unsigned long value;
830 int rc;
831
832 if (pvt->inject.enable)
Mauro Carvalho Chehab41fcb7f2009-06-22 22:48:31 -0300833 disable_inject(mci);
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -0300834
835 rc = strict_strtoul(data, 10, &value);
836 if ((rc < 0) || (value > 7))
Mauro Carvalho Chehab2068def2009-08-05 19:28:27 -0300837 return -EIO;
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -0300838
839 pvt->inject.type = (u32) value;
840 return count;
841}
842
843static ssize_t i7core_inject_type_show(struct mem_ctl_info *mci,
844 char *data)
845{
846 struct i7core_pvt *pvt = mci->pvt_info;
847 return sprintf(data, "0x%08x\n", pvt->inject.type);
848}
849
850/*
851 * i7core_inject_inject.eccmask_store
852 *
853 * The type of error (UE/CE) will depend on the inject.eccmask value:
854 * Any bits set to a 1 will flip the corresponding ECC bit
855 * Correctable errors can be injected by flipping 1 bit or the bits within
856 * a symbol pair (2 consecutive aligned 8-bit pairs - i.e. 7:0 and 15:8 or
857 * 23:16 and 31:24). Flipping bits in two symbol pairs will cause an
858 * uncorrectable error to be injected.
859 */
860static ssize_t i7core_inject_eccmask_store(struct mem_ctl_info *mci,
861 const char *data, size_t count)
862{
863 struct i7core_pvt *pvt = mci->pvt_info;
864 unsigned long value;
865 int rc;
866
867 if (pvt->inject.enable)
Mauro Carvalho Chehab41fcb7f2009-06-22 22:48:31 -0300868 disable_inject(mci);
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -0300869
870 rc = strict_strtoul(data, 10, &value);
871 if (rc < 0)
Mauro Carvalho Chehab2068def2009-08-05 19:28:27 -0300872 return -EIO;
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -0300873
874 pvt->inject.eccmask = (u32) value;
875 return count;
876}
877
878static ssize_t i7core_inject_eccmask_show(struct mem_ctl_info *mci,
879 char *data)
880{
881 struct i7core_pvt *pvt = mci->pvt_info;
882 return sprintf(data, "0x%08x\n", pvt->inject.eccmask);
883}
884
885/*
886 * i7core_addrmatch
887 *
888 * The type of error (UE/CE) will depend on the inject.eccmask value:
889 * Any bits set to a 1 will flip the corresponding ECC bit
890 * Correctable errors can be injected by flipping 1 bit or the bits within
891 * a symbol pair (2 consecutive aligned 8-bit pairs - i.e. 7:0 and 15:8 or
892 * 23:16 and 31:24). Flipping bits in two symbol pairs will cause an
893 * uncorrectable error to be injected.
894 */
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -0300895
Mauro Carvalho Chehaba5538e52009-09-23 18:56:47 -0300896#define DECLARE_ADDR_MATCH(param, limit) \
897static ssize_t i7core_inject_store_##param( \
898 struct mem_ctl_info *mci, \
899 const char *data, size_t count) \
900{ \
Mauro Carvalho Chehabcc301b32009-09-24 16:23:42 -0300901 struct i7core_pvt *pvt; \
Mauro Carvalho Chehaba5538e52009-09-23 18:56:47 -0300902 long value; \
903 int rc; \
904 \
Mauro Carvalho Chehabcc301b32009-09-24 16:23:42 -0300905 debugf1("%s()\n", __func__); \
906 pvt = mci->pvt_info; \
907 \
Mauro Carvalho Chehaba5538e52009-09-23 18:56:47 -0300908 if (pvt->inject.enable) \
909 disable_inject(mci); \
910 \
Mauro Carvalho Chehab4f87fad2009-10-04 11:54:56 -0300911 if (!strcasecmp(data, "any") || !strcasecmp(data, "any\n"))\
Mauro Carvalho Chehaba5538e52009-09-23 18:56:47 -0300912 value = -1; \
913 else { \
914 rc = strict_strtoul(data, 10, &value); \
915 if ((rc < 0) || (value >= limit)) \
916 return -EIO; \
917 } \
918 \
919 pvt->inject.param = value; \
920 \
921 return count; \
922} \
923 \
924static ssize_t i7core_inject_show_##param( \
925 struct mem_ctl_info *mci, \
926 char *data) \
927{ \
Mauro Carvalho Chehabcc301b32009-09-24 16:23:42 -0300928 struct i7core_pvt *pvt; \
929 \
930 pvt = mci->pvt_info; \
931 debugf1("%s() pvt=%p\n", __func__, pvt); \
Mauro Carvalho Chehaba5538e52009-09-23 18:56:47 -0300932 if (pvt->inject.param < 0) \
933 return sprintf(data, "any\n"); \
934 else \
935 return sprintf(data, "%d\n", pvt->inject.param);\
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -0300936}
937
Mauro Carvalho Chehaba5538e52009-09-23 18:56:47 -0300938#define ATTR_ADDR_MATCH(param) \
939 { \
940 .attr = { \
941 .name = #param, \
942 .mode = (S_IRUGO | S_IWUSR) \
943 }, \
944 .show = i7core_inject_show_##param, \
945 .store = i7core_inject_store_##param, \
946 }
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -0300947
Mauro Carvalho Chehaba5538e52009-09-23 18:56:47 -0300948DECLARE_ADDR_MATCH(channel, 3);
949DECLARE_ADDR_MATCH(dimm, 3);
950DECLARE_ADDR_MATCH(rank, 4);
951DECLARE_ADDR_MATCH(bank, 32);
952DECLARE_ADDR_MATCH(page, 0x10000);
953DECLARE_ADDR_MATCH(col, 0x4000);
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -0300954
Mauro Carvalho Chehab1288c182010-08-10 18:57:01 -0300955static int write_and_test(struct pci_dev *dev, const int where, const u32 val)
Mauro Carvalho Chehab276b8242009-07-22 21:45:50 -0300956{
957 u32 read;
958 int count;
959
Mauro Carvalho Chehab4157d9f2009-08-05 20:27:15 -0300960 debugf0("setting pci %02x:%02x.%x reg=%02x value=%08x\n",
961 dev->bus->number, PCI_SLOT(dev->devfn), PCI_FUNC(dev->devfn),
962 where, val);
963
Mauro Carvalho Chehab276b8242009-07-22 21:45:50 -0300964 for (count = 0; count < 10; count++) {
965 if (count)
Mauro Carvalho Chehabb9905382009-08-05 21:36:35 -0300966 msleep(100);
Mauro Carvalho Chehab276b8242009-07-22 21:45:50 -0300967 pci_write_config_dword(dev, where, val);
968 pci_read_config_dword(dev, where, &read);
969
970 if (read == val)
971 return 0;
972 }
973
Mauro Carvalho Chehab4157d9f2009-08-05 20:27:15 -0300974 i7core_printk(KERN_ERR, "Error during set pci %02x:%02x.%x reg=%02x "
975 "write=%08x. Read=%08x\n",
976 dev->bus->number, PCI_SLOT(dev->devfn), PCI_FUNC(dev->devfn),
977 where, val, read);
Mauro Carvalho Chehab276b8242009-07-22 21:45:50 -0300978
979 return -EINVAL;
980}
981
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -0300982/*
983 * This routine prepares the Memory Controller for error injection.
984 * The error will be injected when some process tries to write to the
985 * memory that matches the given criteria.
986 * The criteria can be set in terms of a mask where dimm, rank, bank, page
987 * and col can be specified.
988 * A -1 value for any of the mask items will make the MCU to ignore
989 * that matching criteria for error injection.
990 *
991 * It should be noticed that the error will only happen after a write operation
992 * on a memory that matches the condition. if REPEAT_EN is not enabled at
993 * inject mask, then it will produce just one error. Otherwise, it will repeat
994 * until the injectmask would be cleaned.
995 *
996 * FIXME: This routine assumes that MAXNUMDIMMS value of MC_MAX_DOD
997 * is reliable enough to check if the MC is using the
998 * three channels. However, this is not clear at the datasheet.
999 */
1000static ssize_t i7core_inject_enable_store(struct mem_ctl_info *mci,
1001 const char *data, size_t count)
1002{
1003 struct i7core_pvt *pvt = mci->pvt_info;
1004 u32 injectmask;
1005 u64 mask = 0;
1006 int rc;
1007 long enable;
1008
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001009 if (!pvt->pci_ch[pvt->inject.channel][0])
Mauro Carvalho Chehab8f331902009-06-22 22:48:29 -03001010 return 0;
1011
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001012 rc = strict_strtoul(data, 10, &enable);
1013 if ((rc < 0))
1014 return 0;
1015
1016 if (enable) {
1017 pvt->inject.enable = 1;
1018 } else {
1019 disable_inject(mci);
1020 return count;
1021 }
1022
1023 /* Sets pvt->inject.dimm mask */
1024 if (pvt->inject.dimm < 0)
Alan Cox486dd092009-11-08 01:34:27 -02001025 mask |= 1LL << 41;
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001026 else {
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001027 if (pvt->channel[pvt->inject.channel].dimms > 2)
Alan Cox486dd092009-11-08 01:34:27 -02001028 mask |= (pvt->inject.dimm & 0x3LL) << 35;
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001029 else
Alan Cox486dd092009-11-08 01:34:27 -02001030 mask |= (pvt->inject.dimm & 0x1LL) << 36;
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001031 }
1032
1033 /* Sets pvt->inject.rank mask */
1034 if (pvt->inject.rank < 0)
Alan Cox486dd092009-11-08 01:34:27 -02001035 mask |= 1LL << 40;
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001036 else {
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001037 if (pvt->channel[pvt->inject.channel].dimms > 2)
Alan Cox486dd092009-11-08 01:34:27 -02001038 mask |= (pvt->inject.rank & 0x1LL) << 34;
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001039 else
Alan Cox486dd092009-11-08 01:34:27 -02001040 mask |= (pvt->inject.rank & 0x3LL) << 34;
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001041 }
1042
1043 /* Sets pvt->inject.bank mask */
1044 if (pvt->inject.bank < 0)
Alan Cox486dd092009-11-08 01:34:27 -02001045 mask |= 1LL << 39;
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001046 else
Alan Cox486dd092009-11-08 01:34:27 -02001047 mask |= (pvt->inject.bank & 0x15LL) << 30;
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001048
1049 /* Sets pvt->inject.page mask */
1050 if (pvt->inject.page < 0)
Alan Cox486dd092009-11-08 01:34:27 -02001051 mask |= 1LL << 38;
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001052 else
Alan Cox486dd092009-11-08 01:34:27 -02001053 mask |= (pvt->inject.page & 0xffff) << 14;
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001054
1055 /* Sets pvt->inject.column mask */
1056 if (pvt->inject.col < 0)
Alan Cox486dd092009-11-08 01:34:27 -02001057 mask |= 1LL << 37;
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001058 else
Alan Cox486dd092009-11-08 01:34:27 -02001059 mask |= (pvt->inject.col & 0x3fff);
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001060
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001061 /*
1062 * bit 0: REPEAT_EN
1063 * bits 1-2: MASK_HALF_CACHELINE
1064 * bit 3: INJECT_ECC
1065 * bit 4: INJECT_ADDR_PARITY
1066 */
1067
Mauro Carvalho Chehab7b029d02009-06-22 22:48:29 -03001068 injectmask = (pvt->inject.type & 1) |
1069 (pvt->inject.section & 0x3) << 1 |
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001070 (pvt->inject.type & 0x6) << (3 - 1);
1071
Mauro Carvalho Chehab276b8242009-07-22 21:45:50 -03001072 /* Unlock writes to registers - this register is write only */
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001073 pci_write_config_dword(pvt->pci_noncore,
Mauro Carvalho Chehab276b8242009-07-22 21:45:50 -03001074 MC_CFG_CONTROL, 0x2);
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001075
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001076 write_and_test(pvt->pci_ch[pvt->inject.channel][0],
Mauro Carvalho Chehab276b8242009-07-22 21:45:50 -03001077 MC_CHANNEL_ADDR_MATCH, mask);
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001078 write_and_test(pvt->pci_ch[pvt->inject.channel][0],
Mauro Carvalho Chehab276b8242009-07-22 21:45:50 -03001079 MC_CHANNEL_ADDR_MATCH + 4, mask >> 32L);
1080
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001081 write_and_test(pvt->pci_ch[pvt->inject.channel][0],
Mauro Carvalho Chehab276b8242009-07-22 21:45:50 -03001082 MC_CHANNEL_ERROR_MASK, pvt->inject.eccmask);
1083
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001084 write_and_test(pvt->pci_ch[pvt->inject.channel][0],
Mauro Carvalho Chehab4157d9f2009-08-05 20:27:15 -03001085 MC_CHANNEL_ERROR_INJECT, injectmask);
Mauro Carvalho Chehab276b8242009-07-22 21:45:50 -03001086
1087 /*
1088 * This is something undocumented, based on my tests
1089 * Without writing 8 to this register, errors aren't injected. Not sure
1090 * why.
1091 */
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001092 pci_write_config_dword(pvt->pci_noncore,
Mauro Carvalho Chehab276b8242009-07-22 21:45:50 -03001093 MC_CFG_CONTROL, 8);
1094
Mauro Carvalho Chehab41fcb7f2009-06-22 22:48:31 -03001095 debugf0("Error inject addr match 0x%016llx, ecc 0x%08x,"
1096 " inject 0x%08x\n",
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001097 mask, pvt->inject.eccmask, injectmask);
1098
Mauro Carvalho Chehab7b029d02009-06-22 22:48:29 -03001099
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001100 return count;
1101}
1102
1103static ssize_t i7core_inject_enable_show(struct mem_ctl_info *mci,
1104 char *data)
1105{
1106 struct i7core_pvt *pvt = mci->pvt_info;
Mauro Carvalho Chehab7b029d02009-06-22 22:48:29 -03001107 u32 injectmask;
1108
Mauro Carvalho Chehab52a2e4f2009-10-14 11:21:58 -03001109 if (!pvt->pci_ch[pvt->inject.channel][0])
1110 return 0;
1111
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001112 pci_read_config_dword(pvt->pci_ch[pvt->inject.channel][0],
Mauro Carvalho Chehab4157d9f2009-08-05 20:27:15 -03001113 MC_CHANNEL_ERROR_INJECT, &injectmask);
Mauro Carvalho Chehab7b029d02009-06-22 22:48:29 -03001114
1115 debugf0("Inject error read: 0x%018x\n", injectmask);
1116
1117 if (injectmask & 0x0c)
1118 pvt->inject.enable = 1;
1119
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001120 return sprintf(data, "%d\n", pvt->inject.enable);
1121}
1122
Mauro Carvalho Chehabf338d732009-09-24 17:25:43 -03001123#define DECLARE_COUNTER(param) \
1124static ssize_t i7core_show_counter_##param( \
1125 struct mem_ctl_info *mci, \
1126 char *data) \
1127{ \
1128 struct i7core_pvt *pvt = mci->pvt_info; \
1129 \
1130 debugf1("%s() \n", __func__); \
1131 if (!pvt->ce_count_available || (pvt->is_registered)) \
1132 return sprintf(data, "data unavailable\n"); \
1133 return sprintf(data, "%lu\n", \
1134 pvt->udimm_ce_count[param]); \
Mauro Carvalho Chehab442305b2009-06-22 22:48:29 -03001135}
1136
Mauro Carvalho Chehabf338d732009-09-24 17:25:43 -03001137#define ATTR_COUNTER(param) \
1138 { \
1139 .attr = { \
1140 .name = __stringify(udimm##param), \
1141 .mode = (S_IRUGO | S_IWUSR) \
1142 }, \
1143 .show = i7core_show_counter_##param \
1144 }
1145
1146DECLARE_COUNTER(0);
1147DECLARE_COUNTER(1);
1148DECLARE_COUNTER(2);
1149
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001150/*
1151 * Sysfs struct
1152 */
Mauro Carvalho Chehaba5538e52009-09-23 18:56:47 -03001153
Mauro Carvalho Chehab1288c182010-08-10 18:57:01 -03001154static const struct mcidev_sysfs_attribute i7core_addrmatch_attrs[] = {
Mauro Carvalho Chehaba5538e52009-09-23 18:56:47 -03001155 ATTR_ADDR_MATCH(channel),
1156 ATTR_ADDR_MATCH(dimm),
1157 ATTR_ADDR_MATCH(rank),
1158 ATTR_ADDR_MATCH(bank),
1159 ATTR_ADDR_MATCH(page),
1160 ATTR_ADDR_MATCH(col),
Mauro Carvalho Chehab1288c182010-08-10 18:57:01 -03001161 { } /* End of list */
Mauro Carvalho Chehaba5538e52009-09-23 18:56:47 -03001162};
1163
Mauro Carvalho Chehab1288c182010-08-10 18:57:01 -03001164static const struct mcidev_sysfs_group i7core_inject_addrmatch = {
Mauro Carvalho Chehaba5538e52009-09-23 18:56:47 -03001165 .name = "inject_addrmatch",
1166 .mcidev_attr = i7core_addrmatch_attrs,
1167};
1168
Mauro Carvalho Chehab1288c182010-08-10 18:57:01 -03001169static const struct mcidev_sysfs_attribute i7core_udimm_counters_attrs[] = {
Mauro Carvalho Chehabf338d732009-09-24 17:25:43 -03001170 ATTR_COUNTER(0),
1171 ATTR_COUNTER(1),
1172 ATTR_COUNTER(2),
Marcin Slusarz64aab722010-09-30 15:15:30 -07001173 { .attr = { .name = NULL } }
Mauro Carvalho Chehabf338d732009-09-24 17:25:43 -03001174};
1175
Mauro Carvalho Chehab1288c182010-08-10 18:57:01 -03001176static const struct mcidev_sysfs_group i7core_udimm_counters = {
Mauro Carvalho Chehabf338d732009-09-24 17:25:43 -03001177 .name = "all_channel_counts",
1178 .mcidev_attr = i7core_udimm_counters_attrs,
1179};
1180
Mauro Carvalho Chehab1288c182010-08-10 18:57:01 -03001181static const struct mcidev_sysfs_attribute i7core_sysfs_rdimm_attrs[] = {
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001182 {
1183 .attr = {
1184 .name = "inject_section",
1185 .mode = (S_IRUGO | S_IWUSR)
1186 },
1187 .show = i7core_inject_section_show,
1188 .store = i7core_inject_section_store,
1189 }, {
1190 .attr = {
1191 .name = "inject_type",
1192 .mode = (S_IRUGO | S_IWUSR)
1193 },
1194 .show = i7core_inject_type_show,
1195 .store = i7core_inject_type_store,
1196 }, {
1197 .attr = {
1198 .name = "inject_eccmask",
1199 .mode = (S_IRUGO | S_IWUSR)
1200 },
1201 .show = i7core_inject_eccmask_show,
1202 .store = i7core_inject_eccmask_store,
1203 }, {
Mauro Carvalho Chehaba5538e52009-09-23 18:56:47 -03001204 .grp = &i7core_inject_addrmatch,
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001205 }, {
1206 .attr = {
1207 .name = "inject_enable",
1208 .mode = (S_IRUGO | S_IWUSR)
1209 },
1210 .show = i7core_inject_enable_show,
1211 .store = i7core_inject_enable_store,
1212 },
Mauro Carvalho Chehab1288c182010-08-10 18:57:01 -03001213 { } /* End of list */
1214};
1215
1216static const struct mcidev_sysfs_attribute i7core_sysfs_udimm_attrs[] = {
1217 {
1218 .attr = {
1219 .name = "inject_section",
1220 .mode = (S_IRUGO | S_IWUSR)
1221 },
1222 .show = i7core_inject_section_show,
1223 .store = i7core_inject_section_store,
1224 }, {
1225 .attr = {
1226 .name = "inject_type",
1227 .mode = (S_IRUGO | S_IWUSR)
1228 },
1229 .show = i7core_inject_type_show,
1230 .store = i7core_inject_type_store,
1231 }, {
1232 .attr = {
1233 .name = "inject_eccmask",
1234 .mode = (S_IRUGO | S_IWUSR)
1235 },
1236 .show = i7core_inject_eccmask_show,
1237 .store = i7core_inject_eccmask_store,
1238 }, {
1239 .grp = &i7core_inject_addrmatch,
1240 }, {
1241 .attr = {
1242 .name = "inject_enable",
1243 .mode = (S_IRUGO | S_IWUSR)
1244 },
1245 .show = i7core_inject_enable_show,
1246 .store = i7core_inject_enable_store,
1247 }, {
1248 .grp = &i7core_udimm_counters,
1249 },
1250 { } /* End of list */
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03001251};
1252
1253/****************************************************************************
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03001254 Device initialization routines: put/get, init/exit
1255 ****************************************************************************/
1256
1257/*
Hidetoshi Seto64c10f62010-08-20 04:28:14 -03001258 * i7core_put_all_devices 'put' all the devices that we have
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03001259 * reserved via 'get'
1260 */
Mauro Carvalho Chehab13d6e9b2009-09-05 12:15:20 -03001261static void i7core_put_devices(struct i7core_dev *i7core_dev)
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03001262{
Mauro Carvalho Chehab13d6e9b2009-09-05 12:15:20 -03001263 int i;
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03001264
Mauro Carvalho Chehab22e6bcb2009-09-05 23:06:50 -03001265 debugf0(__FILE__ ": %s()\n", __func__);
Mauro Carvalho Chehabde06eee2009-10-14 08:02:40 -03001266 for (i = 0; i < i7core_dev->n_devs; i++) {
Mauro Carvalho Chehab22e6bcb2009-09-05 23:06:50 -03001267 struct pci_dev *pdev = i7core_dev->pdev[i];
1268 if (!pdev)
1269 continue;
1270 debugf0("Removing dev %02x:%02x.%d\n",
1271 pdev->bus->number,
1272 PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
1273 pci_dev_put(pdev);
1274 }
Mauro Carvalho Chehab13d6e9b2009-09-05 12:15:20 -03001275}
Mauro Carvalho Chehab66607702009-09-05 00:52:11 -03001276
Mauro Carvalho Chehab13d6e9b2009-09-05 12:15:20 -03001277static void i7core_put_all_devices(void)
1278{
Mauro Carvalho Chehab42538682009-09-24 09:59:13 -03001279 struct i7core_dev *i7core_dev, *tmp;
Mauro Carvalho Chehab13d6e9b2009-09-05 12:15:20 -03001280
Mauro Carvalho Chehab39300e72010-08-11 23:40:15 -03001281 list_for_each_entry_safe(i7core_dev, tmp, &i7core_edac_list, list) {
Mauro Carvalho Chehab13d6e9b2009-09-05 12:15:20 -03001282 i7core_put_devices(i7core_dev);
Hidetoshi Seto2aa9be42010-08-20 04:25:00 -03001283 free_i7core_dev(i7core_dev);
Mauro Carvalho Chehab39300e72010-08-11 23:40:15 -03001284 }
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03001285}
1286
Mauro Carvalho Chehab1288c182010-08-10 18:57:01 -03001287static void __init i7core_xeon_pci_fixup(const struct pci_id_table *table)
Keith Manntheybc2d7242009-09-03 00:05:05 -03001288{
1289 struct pci_dev *pdev = NULL;
1290 int i;
Mauro Carvalho Chehab54a08ab2010-08-19 15:51:00 -03001291
Keith Manntheybc2d7242009-09-03 00:05:05 -03001292 /*
David Sterbae7bf0682010-12-27 16:51:15 +01001293 * On Xeon 55xx, the Intel Quick Path Arch Generic Non-core pci buses
Keith Manntheybc2d7242009-09-03 00:05:05 -03001294 * aren't announced by acpi. So, we need to use a legacy scan probing
1295 * to detect them
1296 */
Vernon Mauerybd9e19c2010-05-18 19:02:50 -03001297 while (table && table->descr) {
1298 pdev = pci_get_device(PCI_VENDOR_ID_INTEL, table->descr[0].dev_id, NULL);
1299 if (unlikely(!pdev)) {
1300 for (i = 0; i < MAX_SOCKET_BUSES; i++)
1301 pcibios_scan_specific_bus(255-i);
1302 }
Mauro Carvalho Chehabbda14282010-06-30 01:41:35 -03001303 pci_dev_put(pdev);
Vernon Mauerybd9e19c2010-05-18 19:02:50 -03001304 table++;
Keith Manntheybc2d7242009-09-03 00:05:05 -03001305 }
1306}
1307
Mauro Carvalho Chehabbda14282010-06-30 01:41:35 -03001308static unsigned i7core_pci_lastbus(void)
1309{
1310 int last_bus = 0, bus;
1311 struct pci_bus *b = NULL;
1312
1313 while ((b = pci_find_next_bus(b)) != NULL) {
1314 bus = b->number;
1315 debugf0("Found bus %d\n", bus);
1316 if (bus > last_bus)
1317 last_bus = bus;
1318 }
1319
1320 debugf0("Last bus %d\n", last_bus);
1321
1322 return last_bus;
1323}
1324
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03001325/*
Hidetoshi Seto64c10f62010-08-20 04:28:14 -03001326 * i7core_get_all_devices Find and perform 'get' operation on the MCH's
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03001327 * device/functions we want to reference for this driver
1328 *
1329 * Need to 'get' device 16 func 1 and func 2
1330 */
Hidetoshi Setob197cba2010-08-20 04:24:31 -03001331static int i7core_get_onedevice(struct pci_dev **prev,
1332 const struct pci_id_table *table,
1333 const unsigned devno,
1334 const unsigned last_bus)
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03001335{
Mauro Carvalho Chehab66607702009-09-05 00:52:11 -03001336 struct i7core_dev *i7core_dev;
Hidetoshi Setob197cba2010-08-20 04:24:31 -03001337 const struct pci_id_descr *dev_descr = &table->descr[devno];
Mauro Carvalho Chehab66607702009-09-05 00:52:11 -03001338
Mauro Carvalho Chehab8f331902009-06-22 22:48:29 -03001339 struct pci_dev *pdev = NULL;
Mauro Carvalho Chehab67166af2009-07-15 06:56:23 -03001340 u8 bus = 0;
1341 u8 socket = 0;
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03001342
Mauro Carvalho Chehabc77720b2009-07-18 10:43:08 -03001343 pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
Mauro Carvalho Chehabde06eee2009-10-14 08:02:40 -03001344 dev_descr->dev_id, *prev);
Mauro Carvalho Chehabc77720b2009-07-18 10:43:08 -03001345
Mauro Carvalho Chehab224e8712011-03-17 17:02:59 -03001346 /*
1347 * On Xeon 55xx, the Intel Quckpath Arch Generic Non-core regs
1348 * is at addr 8086:2c40, instead of 8086:2c41. So, we need
1349 * to probe for the alternate address in case of failure
1350 */
1351 if (dev_descr->dev_id == PCI_DEVICE_ID_INTEL_I7_NONCORE && !pdev)
1352 pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
1353 PCI_DEVICE_ID_INTEL_I7_NONCORE_ALT, *prev);
1354
1355 if (dev_descr->dev_id == PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE && !pdev)
1356 pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
1357 PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE_ALT,
1358 *prev);
1359
Mauro Carvalho Chehabc77720b2009-07-18 10:43:08 -03001360 if (!pdev) {
1361 if (*prev) {
1362 *prev = pdev;
1363 return 0;
Mauro Carvalho Chehabd1fd4fb2009-07-10 18:39:53 -03001364 }
1365
Mauro Carvalho Chehabde06eee2009-10-14 08:02:40 -03001366 if (dev_descr->optional)
Mauro Carvalho Chehabc77720b2009-07-18 10:43:08 -03001367 return 0;
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -03001368
Vernon Mauerybd9e19c2010-05-18 19:02:50 -03001369 if (devno == 0)
1370 return -ENODEV;
1371
Daniel J Bluemanab089372010-07-23 23:16:52 +01001372 i7core_printk(KERN_INFO,
Mauro Carvalho Chehabc77720b2009-07-18 10:43:08 -03001373 "Device not found: dev %02x.%d PCI ID %04x:%04x\n",
Mauro Carvalho Chehabde06eee2009-10-14 08:02:40 -03001374 dev_descr->dev, dev_descr->func,
1375 PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -03001376
Mauro Carvalho Chehabc77720b2009-07-18 10:43:08 -03001377 /* End of list, leave */
1378 return -ENODEV;
1379 }
1380 bus = pdev->bus->number;
Mauro Carvalho Chehab8f331902009-06-22 22:48:29 -03001381
Mauro Carvalho Chehabbda14282010-06-30 01:41:35 -03001382 socket = last_bus - bus;
Mauro Carvalho Chehabc77720b2009-07-18 10:43:08 -03001383
Mauro Carvalho Chehab66607702009-09-05 00:52:11 -03001384 i7core_dev = get_i7core_dev(socket);
1385 if (!i7core_dev) {
Hidetoshi Seto848b2f72010-08-20 04:24:44 -03001386 i7core_dev = alloc_i7core_dev(socket, table);
Hidetoshi Seto28966372010-08-20 04:28:51 -03001387 if (!i7core_dev) {
1388 pci_dev_put(pdev);
Mauro Carvalho Chehab66607702009-09-05 00:52:11 -03001389 return -ENOMEM;
Hidetoshi Seto28966372010-08-20 04:28:51 -03001390 }
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03001391 }
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03001392
Mauro Carvalho Chehab66607702009-09-05 00:52:11 -03001393 if (i7core_dev->pdev[devno]) {
Mauro Carvalho Chehabc77720b2009-07-18 10:43:08 -03001394 i7core_printk(KERN_ERR,
1395 "Duplicated device for "
1396 "dev %02x:%02x.%d PCI ID %04x:%04x\n",
Mauro Carvalho Chehabde06eee2009-10-14 08:02:40 -03001397 bus, dev_descr->dev, dev_descr->func,
1398 PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
Mauro Carvalho Chehabc77720b2009-07-18 10:43:08 -03001399 pci_dev_put(pdev);
1400 return -ENODEV;
1401 }
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -03001402
Mauro Carvalho Chehab66607702009-09-05 00:52:11 -03001403 i7core_dev->pdev[devno] = pdev;
Mauro Carvalho Chehabc77720b2009-07-18 10:43:08 -03001404
1405 /* Sanity check */
Mauro Carvalho Chehabde06eee2009-10-14 08:02:40 -03001406 if (unlikely(PCI_SLOT(pdev->devfn) != dev_descr->dev ||
1407 PCI_FUNC(pdev->devfn) != dev_descr->func)) {
Mauro Carvalho Chehabc77720b2009-07-18 10:43:08 -03001408 i7core_printk(KERN_ERR,
1409 "Device PCI ID %04x:%04x "
1410 "has dev %02x:%02x.%d instead of dev %02x:%02x.%d\n",
Mauro Carvalho Chehabde06eee2009-10-14 08:02:40 -03001411 PCI_VENDOR_ID_INTEL, dev_descr->dev_id,
Mauro Carvalho Chehabc77720b2009-07-18 10:43:08 -03001412 bus, PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn),
Mauro Carvalho Chehabde06eee2009-10-14 08:02:40 -03001413 bus, dev_descr->dev, dev_descr->func);
Mauro Carvalho Chehabc77720b2009-07-18 10:43:08 -03001414 return -ENODEV;
1415 }
1416
1417 /* Be sure that the device is enabled */
1418 if (unlikely(pci_enable_device(pdev) < 0)) {
1419 i7core_printk(KERN_ERR,
1420 "Couldn't enable "
1421 "dev %02x:%02x.%d PCI ID %04x:%04x\n",
Mauro Carvalho Chehabde06eee2009-10-14 08:02:40 -03001422 bus, dev_descr->dev, dev_descr->func,
1423 PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
Mauro Carvalho Chehabc77720b2009-07-18 10:43:08 -03001424 return -ENODEV;
1425 }
1426
Mauro Carvalho Chehabd4c27792009-09-05 04:12:02 -03001427 debugf0("Detected socket %d dev %02x:%02x.%d PCI ID %04x:%04x\n",
Mauro Carvalho Chehabde06eee2009-10-14 08:02:40 -03001428 socket, bus, dev_descr->dev,
1429 dev_descr->func,
1430 PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
Mauro Carvalho Chehabc77720b2009-07-18 10:43:08 -03001431
Mauro Carvalho Chehaba3e15412010-08-21 08:52:41 -03001432 /*
1433 * As stated on drivers/pci/search.c, the reference count for
1434 * @from is always decremented if it is not %NULL. So, as we need
1435 * to get all devices up to null, we need to do a get for the device
1436 */
1437 pci_dev_get(pdev);
1438
Mauro Carvalho Chehabc77720b2009-07-18 10:43:08 -03001439 *prev = pdev;
1440
1441 return 0;
1442}
1443
Hidetoshi Seto64c10f62010-08-20 04:28:14 -03001444static int i7core_get_all_devices(void)
Mauro Carvalho Chehabc77720b2009-07-18 10:43:08 -03001445{
Mauro Carvalho Chehab3c52cc52010-10-24 11:12:28 -02001446 int i, rc, last_bus;
Mauro Carvalho Chehabc77720b2009-07-18 10:43:08 -03001447 struct pci_dev *pdev = NULL;
Mauro Carvalho Chehab3c52cc52010-10-24 11:12:28 -02001448 const struct pci_id_table *table = pci_dev_table;
Mauro Carvalho Chehabc77720b2009-07-18 10:43:08 -03001449
Mauro Carvalho Chehabbda14282010-06-30 01:41:35 -03001450 last_bus = i7core_pci_lastbus();
1451
Mauro Carvalho Chehab3c52cc52010-10-24 11:12:28 -02001452 while (table && table->descr) {
Vernon Mauerybd9e19c2010-05-18 19:02:50 -03001453 for (i = 0; i < table->n_devs; i++) {
1454 pdev = NULL;
1455 do {
Hidetoshi Setob197cba2010-08-20 04:24:31 -03001456 rc = i7core_get_onedevice(&pdev, table, i,
Mauro Carvalho Chehabbda14282010-06-30 01:41:35 -03001457 last_bus);
Vernon Mauerybd9e19c2010-05-18 19:02:50 -03001458 if (rc < 0) {
1459 if (i == 0) {
1460 i = table->n_devs;
1461 break;
1462 }
1463 i7core_put_all_devices();
1464 return -ENODEV;
1465 }
1466 } while (pdev);
1467 }
Mauro Carvalho Chehab3c52cc52010-10-24 11:12:28 -02001468 table++;
Mauro Carvalho Chehabc77720b2009-07-18 10:43:08 -03001469 }
Mauro Carvalho Chehab66607702009-09-05 00:52:11 -03001470
Mauro Carvalho Chehabc77720b2009-07-18 10:43:08 -03001471 return 0;
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -03001472}
1473
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001474static int mci_bind_devs(struct mem_ctl_info *mci,
1475 struct i7core_dev *i7core_dev)
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -03001476{
1477 struct i7core_pvt *pvt = mci->pvt_info;
1478 struct pci_dev *pdev;
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001479 int i, func, slot;
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -03001480
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001481 pvt->is_registered = 0;
Mauro Carvalho Chehabde06eee2009-10-14 08:02:40 -03001482 for (i = 0; i < i7core_dev->n_devs; i++) {
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001483 pdev = i7core_dev->pdev[i];
1484 if (!pdev)
Mauro Carvalho Chehab66607702009-09-05 00:52:11 -03001485 continue;
1486
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001487 func = PCI_FUNC(pdev->devfn);
1488 slot = PCI_SLOT(pdev->devfn);
1489 if (slot == 3) {
1490 if (unlikely(func > MAX_MCR_FUNC))
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -03001491 goto error;
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001492 pvt->pci_mcr[func] = pdev;
1493 } else if (likely(slot >= 4 && slot < 4 + NUM_CHANS)) {
1494 if (unlikely(func > MAX_CHAN_FUNC))
1495 goto error;
1496 pvt->pci_ch[slot - 4][func] = pdev;
1497 } else if (!slot && !func)
1498 pvt->pci_noncore = pdev;
1499 else
1500 goto error;
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -03001501
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001502 debugf0("Associated fn %d.%d, dev = %p, socket %d\n",
1503 PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn),
1504 pdev, i7core_dev->socket);
Mauro Carvalho Chehab14d2c082009-09-02 23:52:36 -03001505
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001506 if (PCI_SLOT(pdev->devfn) == 3 &&
1507 PCI_FUNC(pdev->devfn) == 2)
1508 pvt->is_registered = 1;
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -03001509 }
Mauro Carvalho Chehabe9bd2e72009-07-09 22:14:35 -03001510
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -03001511 return 0;
1512
1513error:
1514 i7core_printk(KERN_ERR, "Device %d, function %d "
1515 "is out of the expected range\n",
1516 slot, func);
1517 return -EINVAL;
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03001518}
1519
Mauro Carvalho Chehab442305b2009-06-22 22:48:29 -03001520/****************************************************************************
1521 Error check routines
1522 ****************************************************************************/
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001523static void i7core_rdimm_update_csrow(struct mem_ctl_info *mci,
Mauro Carvalho Chehab1288c182010-08-10 18:57:01 -03001524 const int chan,
1525 const int dimm,
1526 const int add)
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001527{
1528 char *msg;
1529 struct i7core_pvt *pvt = mci->pvt_info;
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001530 int row = pvt->csrow_map[chan][dimm], i;
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001531
1532 for (i = 0; i < add; i++) {
1533 msg = kasprintf(GFP_KERNEL, "Corrected error "
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001534 "(Socket=%d channel=%d dimm=%d)",
1535 pvt->i7core_dev->socket, chan, dimm);
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001536
1537 edac_mc_handle_fbd_ce(mci, row, 0, msg);
1538 kfree (msg);
1539 }
1540}
1541
1542static void i7core_rdimm_update_ce_count(struct mem_ctl_info *mci,
Mauro Carvalho Chehab1288c182010-08-10 18:57:01 -03001543 const int chan,
1544 const int new0,
1545 const int new1,
1546 const int new2)
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001547{
1548 struct i7core_pvt *pvt = mci->pvt_info;
1549 int add0 = 0, add1 = 0, add2 = 0;
1550 /* Updates CE counters if it is not the first time here */
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001551 if (pvt->ce_count_available) {
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001552 /* Updates CE counters */
1553
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001554 add2 = new2 - pvt->rdimm_last_ce_count[chan][2];
1555 add1 = new1 - pvt->rdimm_last_ce_count[chan][1];
1556 add0 = new0 - pvt->rdimm_last_ce_count[chan][0];
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001557
1558 if (add2 < 0)
1559 add2 += 0x7fff;
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001560 pvt->rdimm_ce_count[chan][2] += add2;
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001561
1562 if (add1 < 0)
1563 add1 += 0x7fff;
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001564 pvt->rdimm_ce_count[chan][1] += add1;
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001565
1566 if (add0 < 0)
1567 add0 += 0x7fff;
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001568 pvt->rdimm_ce_count[chan][0] += add0;
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001569 } else
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001570 pvt->ce_count_available = 1;
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001571
1572 /* Store the new values */
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001573 pvt->rdimm_last_ce_count[chan][2] = new2;
1574 pvt->rdimm_last_ce_count[chan][1] = new1;
1575 pvt->rdimm_last_ce_count[chan][0] = new0;
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001576
1577 /*updated the edac core */
1578 if (add0 != 0)
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001579 i7core_rdimm_update_csrow(mci, chan, 0, add0);
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001580 if (add1 != 0)
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001581 i7core_rdimm_update_csrow(mci, chan, 1, add1);
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001582 if (add2 != 0)
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001583 i7core_rdimm_update_csrow(mci, chan, 2, add2);
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001584
1585}
1586
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001587static void i7core_rdimm_check_mc_ecc_err(struct mem_ctl_info *mci)
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001588{
1589 struct i7core_pvt *pvt = mci->pvt_info;
1590 u32 rcv[3][2];
1591 int i, new0, new1, new2;
1592
1593 /*Read DEV 3: FUN 2: MC_COR_ECC_CNT regs directly*/
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001594 pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_0,
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001595 &rcv[0][0]);
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001596 pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_1,
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001597 &rcv[0][1]);
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001598 pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_2,
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001599 &rcv[1][0]);
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001600 pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_3,
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001601 &rcv[1][1]);
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001602 pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_4,
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001603 &rcv[2][0]);
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001604 pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_5,
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001605 &rcv[2][1]);
1606 for (i = 0 ; i < 3; i++) {
1607 debugf3("MC_COR_ECC_CNT%d = 0x%x; MC_COR_ECC_CNT%d = 0x%x\n",
1608 (i * 2), rcv[i][0], (i * 2) + 1, rcv[i][1]);
1609 /*if the channel has 3 dimms*/
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001610 if (pvt->channel[i].dimms > 2) {
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001611 new0 = DIMM_BOT_COR_ERR(rcv[i][0]);
1612 new1 = DIMM_TOP_COR_ERR(rcv[i][0]);
1613 new2 = DIMM_BOT_COR_ERR(rcv[i][1]);
1614 } else {
1615 new0 = DIMM_TOP_COR_ERR(rcv[i][0]) +
1616 DIMM_BOT_COR_ERR(rcv[i][0]);
1617 new1 = DIMM_TOP_COR_ERR(rcv[i][1]) +
1618 DIMM_BOT_COR_ERR(rcv[i][1]);
1619 new2 = 0;
1620 }
1621
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001622 i7core_rdimm_update_ce_count(mci, i, new0, new1, new2);
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001623 }
1624}
Mauro Carvalho Chehab442305b2009-06-22 22:48:29 -03001625
1626/* This function is based on the device 3 function 4 registers as described on:
1627 * Intel Xeon Processor 5500 Series Datasheet Volume 2
1628 * http://www.intel.com/Assets/PDF/datasheet/321322.pdf
1629 * also available at:
1630 * http://www.arrownac.com/manufacturers/intel/s/nehalem/5500-datasheet-v2.pdf
1631 */
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001632static void i7core_udimm_check_mc_ecc_err(struct mem_ctl_info *mci)
Mauro Carvalho Chehab442305b2009-06-22 22:48:29 -03001633{
1634 struct i7core_pvt *pvt = mci->pvt_info;
1635 u32 rcv1, rcv0;
1636 int new0, new1, new2;
1637
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001638 if (!pvt->pci_mcr[4]) {
Mauro Carvalho Chehabb9905382009-08-05 21:36:35 -03001639 debugf0("%s MCR registers not found\n", __func__);
Mauro Carvalho Chehab442305b2009-06-22 22:48:29 -03001640 return;
1641 }
1642
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001643 /* Corrected test errors */
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001644 pci_read_config_dword(pvt->pci_mcr[4], MC_TEST_ERR_RCV1, &rcv1);
1645 pci_read_config_dword(pvt->pci_mcr[4], MC_TEST_ERR_RCV0, &rcv0);
Mauro Carvalho Chehab442305b2009-06-22 22:48:29 -03001646
1647 /* Store the new values */
1648 new2 = DIMM2_COR_ERR(rcv1);
1649 new1 = DIMM1_COR_ERR(rcv0);
1650 new0 = DIMM0_COR_ERR(rcv0);
1651
Mauro Carvalho Chehab442305b2009-06-22 22:48:29 -03001652 /* Updates CE counters if it is not the first time here */
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001653 if (pvt->ce_count_available) {
Mauro Carvalho Chehab442305b2009-06-22 22:48:29 -03001654 /* Updates CE counters */
1655 int add0, add1, add2;
1656
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001657 add2 = new2 - pvt->udimm_last_ce_count[2];
1658 add1 = new1 - pvt->udimm_last_ce_count[1];
1659 add0 = new0 - pvt->udimm_last_ce_count[0];
Mauro Carvalho Chehab442305b2009-06-22 22:48:29 -03001660
1661 if (add2 < 0)
1662 add2 += 0x7fff;
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001663 pvt->udimm_ce_count[2] += add2;
Mauro Carvalho Chehab442305b2009-06-22 22:48:29 -03001664
1665 if (add1 < 0)
1666 add1 += 0x7fff;
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001667 pvt->udimm_ce_count[1] += add1;
Mauro Carvalho Chehab442305b2009-06-22 22:48:29 -03001668
1669 if (add0 < 0)
1670 add0 += 0x7fff;
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001671 pvt->udimm_ce_count[0] += add0;
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001672
1673 if (add0 | add1 | add2)
1674 i7core_printk(KERN_ERR, "New Corrected error(s): "
1675 "dimm0: +%d, dimm1: +%d, dimm2 +%d\n",
1676 add0, add1, add2);
Mauro Carvalho Chehab442305b2009-06-22 22:48:29 -03001677 } else
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001678 pvt->ce_count_available = 1;
Mauro Carvalho Chehab442305b2009-06-22 22:48:29 -03001679
1680 /* Store the new values */
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001681 pvt->udimm_last_ce_count[2] = new2;
1682 pvt->udimm_last_ce_count[1] = new1;
1683 pvt->udimm_last_ce_count[0] = new0;
Mauro Carvalho Chehab442305b2009-06-22 22:48:29 -03001684}
1685
Mauro Carvalho Chehab8a2f1182009-07-15 19:01:08 -03001686/*
1687 * According with tables E-11 and E-12 of chapter E.3.3 of Intel 64 and IA-32
1688 * Architectures Software Developer’s Manual Volume 3B.
Mauro Carvalho Chehabf237fcf2009-07-15 19:53:24 -03001689 * Nehalem are defined as family 0x06, model 0x1a
1690 *
1691 * The MCA registers used here are the following ones:
Mauro Carvalho Chehab8a2f1182009-07-15 19:01:08 -03001692 * struct mce field MCA Register
Mauro Carvalho Chehabf237fcf2009-07-15 19:53:24 -03001693 * m->status MSR_IA32_MC8_STATUS
1694 * m->addr MSR_IA32_MC8_ADDR
1695 * m->misc MSR_IA32_MC8_MISC
Mauro Carvalho Chehab8a2f1182009-07-15 19:01:08 -03001696 * In the case of Nehalem, the error information is masked at .status and .misc
1697 * fields
1698 */
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -03001699static void i7core_mce_output_error(struct mem_ctl_info *mci,
Mauro Carvalho Chehab1288c182010-08-10 18:57:01 -03001700 const struct mce *m)
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -03001701{
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001702 struct i7core_pvt *pvt = mci->pvt_info;
Mauro Carvalho Chehaba6395392009-07-17 10:54:23 -03001703 char *type, *optype, *err, *msg;
Mauro Carvalho Chehab8a2f1182009-07-15 19:01:08 -03001704 unsigned long error = m->status & 0x1ff0000l;
Mauro Carvalho Chehaba6395392009-07-17 10:54:23 -03001705 u32 optypenum = (m->status >> 4) & 0x07;
Mathias Krause8cf2d232011-08-18 09:17:00 +02001706 u32 core_err_cnt = (m->status >> 38) & 0x7fff;
Mauro Carvalho Chehab8a2f1182009-07-15 19:01:08 -03001707 u32 dimm = (m->misc >> 16) & 0x3;
1708 u32 channel = (m->misc >> 18) & 0x3;
1709 u32 syndrome = m->misc >> 32;
1710 u32 errnum = find_first_bit(&error, 32);
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001711 int csrow;
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -03001712
Mauro Carvalho Chehabc5d34522009-07-17 10:28:15 -03001713 if (m->mcgstatus & 1)
1714 type = "FATAL";
1715 else
1716 type = "NON_FATAL";
1717
Mauro Carvalho Chehaba6395392009-07-17 10:54:23 -03001718 switch (optypenum) {
Mauro Carvalho Chehabb9905382009-08-05 21:36:35 -03001719 case 0:
1720 optype = "generic undef request";
1721 break;
1722 case 1:
1723 optype = "read error";
1724 break;
1725 case 2:
1726 optype = "write error";
1727 break;
1728 case 3:
1729 optype = "addr/cmd error";
1730 break;
1731 case 4:
1732 optype = "scrubbing error";
1733 break;
1734 default:
1735 optype = "reserved";
1736 break;
Mauro Carvalho Chehaba6395392009-07-17 10:54:23 -03001737 }
1738
Mauro Carvalho Chehab8a2f1182009-07-15 19:01:08 -03001739 switch (errnum) {
1740 case 16:
1741 err = "read ECC error";
1742 break;
1743 case 17:
1744 err = "RAS ECC error";
1745 break;
1746 case 18:
1747 err = "write parity error";
1748 break;
1749 case 19:
1750 err = "redundacy loss";
1751 break;
1752 case 20:
1753 err = "reserved";
1754 break;
1755 case 21:
1756 err = "memory range error";
1757 break;
1758 case 22:
1759 err = "RTID out of range";
1760 break;
1761 case 23:
1762 err = "address parity error";
1763 break;
1764 case 24:
1765 err = "byte enable parity error";
1766 break;
1767 default:
1768 err = "unknown";
1769 }
1770
Mauro Carvalho Chehabf237fcf2009-07-15 19:53:24 -03001771 /* FIXME: should convert addr into bank and rank information */
Mauro Carvalho Chehab8a2f1182009-07-15 19:01:08 -03001772 msg = kasprintf(GFP_ATOMIC,
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001773 "%s (addr = 0x%08llx, cpu=%d, Dimm=%d, Channel=%d, "
Mauro Carvalho Chehaba6395392009-07-17 10:54:23 -03001774 "syndrome=0x%08x, count=%d, Err=%08llx:%08llx (%s: %s))\n",
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001775 type, (long long) m->addr, m->cpu, dimm, channel,
Mauro Carvalho Chehaba6395392009-07-17 10:54:23 -03001776 syndrome, core_err_cnt, (long long)m->status,
1777 (long long)m->misc, optype, err);
Mauro Carvalho Chehab8a2f1182009-07-15 19:01:08 -03001778
1779 debugf0("%s", msg);
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -03001780
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001781 csrow = pvt->csrow_map[channel][dimm];
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001782
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -03001783 /* Call the helper to output message */
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001784 if (m->mcgstatus & 1)
1785 edac_mc_handle_fbd_ue(mci, csrow, 0,
1786 0 /* FIXME: should be channel here */, msg);
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001787 else if (!pvt->is_registered)
Mauro Carvalho Chehabb4e8f0b2009-09-02 23:49:59 -03001788 edac_mc_handle_fbd_ce(mci, csrow,
1789 0 /* FIXME: should be channel here */, msg);
Mauro Carvalho Chehab8a2f1182009-07-15 19:01:08 -03001790
1791 kfree(msg);
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -03001792}
1793
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03001794/*
Mauro Carvalho Chehab87d1d272009-06-22 22:48:29 -03001795 * i7core_check_error Retrieve and process errors reported by the
1796 * hardware. Called by the Core module.
1797 */
1798static void i7core_check_error(struct mem_ctl_info *mci)
1799{
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -03001800 struct i7core_pvt *pvt = mci->pvt_info;
1801 int i;
1802 unsigned count = 0;
Mauro Carvalho Chehabca9c90b2009-10-04 10:15:40 -03001803 struct mce *m;
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -03001804
Mauro Carvalho Chehabca9c90b2009-10-04 10:15:40 -03001805 /*
1806 * MCE first step: Copy all mce errors into a temporary buffer
1807 * We use a double buffering here, to reduce the risk of
Lucas De Marchi25985ed2011-03-30 22:57:33 -03001808 * losing an error.
Mauro Carvalho Chehabca9c90b2009-10-04 10:15:40 -03001809 */
1810 smp_rmb();
Mauro Carvalho Chehab321ece42009-10-08 13:11:08 -03001811 count = (pvt->mce_out + MCE_LOG_LEN - pvt->mce_in)
1812 % MCE_LOG_LEN;
Mauro Carvalho Chehabca9c90b2009-10-04 10:15:40 -03001813 if (!count)
Vernon Mauery8a311e12010-04-16 19:40:19 -03001814 goto check_ce_error;
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001815
Mauro Carvalho Chehabca9c90b2009-10-04 10:15:40 -03001816 m = pvt->mce_outentry;
Mauro Carvalho Chehab321ece42009-10-08 13:11:08 -03001817 if (pvt->mce_in + count > MCE_LOG_LEN) {
1818 unsigned l = MCE_LOG_LEN - pvt->mce_in;
Mauro Carvalho Chehabca9c90b2009-10-04 10:15:40 -03001819
1820 memcpy(m, &pvt->mce_entry[pvt->mce_in], sizeof(*m) * l);
1821 smp_wmb();
1822 pvt->mce_in = 0;
1823 count -= l;
1824 m += l;
1825 }
1826 memcpy(m, &pvt->mce_entry[pvt->mce_in], sizeof(*m) * count);
1827 smp_wmb();
1828 pvt->mce_in += count;
1829
1830 smp_rmb();
1831 if (pvt->mce_overrun) {
1832 i7core_printk(KERN_ERR, "Lost %d memory errors\n",
1833 pvt->mce_overrun);
1834 smp_wmb();
1835 pvt->mce_overrun = 0;
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -03001836 }
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001837
Mauro Carvalho Chehabca9c90b2009-10-04 10:15:40 -03001838 /*
1839 * MCE second step: parse errors and display
1840 */
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -03001841 for (i = 0; i < count; i++)
Mauro Carvalho Chehabca9c90b2009-10-04 10:15:40 -03001842 i7core_mce_output_error(mci, &pvt->mce_outentry[i]);
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -03001843
Mauro Carvalho Chehabca9c90b2009-10-04 10:15:40 -03001844 /*
1845 * Now, let's increment CE error counts
1846 */
Vernon Mauery8a311e12010-04-16 19:40:19 -03001847check_ce_error:
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001848 if (!pvt->is_registered)
1849 i7core_udimm_check_mc_ecc_err(mci);
1850 else
1851 i7core_rdimm_check_mc_ecc_err(mci);
Mauro Carvalho Chehab87d1d272009-06-22 22:48:29 -03001852}
1853
1854/*
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -03001855 * i7core_mce_check_error Replicates mcelog routine to get errors
1856 * This routine simply queues mcelog errors, and
1857 * return. The error itself should be handled later
1858 * by i7core_check_error.
Mauro Carvalho Chehab6e103be2009-10-05 09:40:09 -03001859 * WARNING: As this routine should be called at NMI time, extra care should
1860 * be taken to avoid deadlocks, and to be as fast as possible.
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -03001861 */
1862static int i7core_mce_check_error(void *priv, struct mce *mce)
1863{
Mauro Carvalho Chehabc5d34522009-07-17 10:28:15 -03001864 struct mem_ctl_info *mci = priv;
1865 struct i7core_pvt *pvt = mci->pvt_info;
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -03001866
Mauro Carvalho Chehab8a2f1182009-07-15 19:01:08 -03001867 /*
1868 * Just let mcelog handle it if the error is
1869 * outside the memory controller
1870 */
1871 if (((mce->status & 0xffff) >> 7) != 1)
1872 return 0;
1873
Mauro Carvalho Chehabf237fcf2009-07-15 19:53:24 -03001874 /* Bank 8 registers are the only ones that we know how to handle */
1875 if (mce->bank != 8)
1876 return 0;
1877
Randy Dunlap3b918c12009-11-08 01:36:40 -02001878#ifdef CONFIG_SMP
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001879 /* Only handle if it is the right mc controller */
Mauro Carvalho Chehab6e103be2009-10-05 09:40:09 -03001880 if (cpu_data(mce->cpu).phys_proc_id != pvt->i7core_dev->socket)
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001881 return 0;
Randy Dunlap3b918c12009-11-08 01:36:40 -02001882#endif
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001883
Mauro Carvalho Chehabca9c90b2009-10-04 10:15:40 -03001884 smp_rmb();
Mauro Carvalho Chehab321ece42009-10-08 13:11:08 -03001885 if ((pvt->mce_out + 1) % MCE_LOG_LEN == pvt->mce_in) {
Mauro Carvalho Chehabca9c90b2009-10-04 10:15:40 -03001886 smp_wmb();
1887 pvt->mce_overrun++;
1888 return 0;
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -03001889 }
Mauro Carvalho Chehab6e103be2009-10-05 09:40:09 -03001890
1891 /* Copy memory error at the ringbuffer */
1892 memcpy(&pvt->mce_entry[pvt->mce_out], mce, sizeof(*mce));
Mauro Carvalho Chehabca9c90b2009-10-04 10:15:40 -03001893 smp_wmb();
Mauro Carvalho Chehab321ece42009-10-08 13:11:08 -03001894 pvt->mce_out = (pvt->mce_out + 1) % MCE_LOG_LEN;
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -03001895
Mauro Carvalho Chehabc5d34522009-07-17 10:28:15 -03001896 /* Handle fatal errors immediately */
1897 if (mce->mcgstatus & 1)
1898 i7core_check_error(mci);
1899
David Sterbae7bf0682010-12-27 16:51:15 +01001900 /* Advise mcelog that the errors were handled */
Mauro Carvalho Chehab8a2f1182009-07-15 19:01:08 -03001901 return 1;
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -03001902}
1903
Hidetoshi Setoa3aa0a42010-08-20 04:25:18 -03001904static void i7core_pci_ctl_create(struct i7core_pvt *pvt)
1905{
1906 pvt->i7core_pci = edac_pci_create_generic_ctl(
1907 &pvt->i7core_dev->pdev[0]->dev,
1908 EDAC_MOD_STR);
1909 if (unlikely(!pvt->i7core_pci))
1910 pr_warn("Unable to setup PCI error report via EDAC\n");
1911}
1912
1913static void i7core_pci_ctl_release(struct i7core_pvt *pvt)
1914{
1915 if (likely(pvt->i7core_pci))
1916 edac_pci_release_generic_ctl(pvt->i7core_pci);
1917 else
1918 i7core_printk(KERN_ERR,
1919 "Couldn't find mem_ctl_info for socket %d\n",
1920 pvt->i7core_dev->socket);
1921 pvt->i7core_pci = NULL;
1922}
1923
Hidetoshi Seto1c6edbb2010-08-20 04:32:33 -03001924static void i7core_unregister_mci(struct i7core_dev *i7core_dev)
1925{
1926 struct mem_ctl_info *mci = i7core_dev->mci;
1927 struct i7core_pvt *pvt;
1928
1929 if (unlikely(!mci || !mci->pvt_info)) {
1930 debugf0("MC: " __FILE__ ": %s(): dev = %p\n",
1931 __func__, &i7core_dev->pdev[0]->dev);
1932
1933 i7core_printk(KERN_ERR, "Couldn't find mci handler\n");
1934 return;
1935 }
1936
1937 pvt = mci->pvt_info;
1938
1939 debugf0("MC: " __FILE__ ": %s(): mci = %p, dev = %p\n",
1940 __func__, mci, &i7core_dev->pdev[0]->dev);
1941
1942 /* Disable MCE NMI handler */
1943 edac_mce_unregister(&pvt->edac_mce);
1944
1945 /* Disable EDAC polling */
1946 i7core_pci_ctl_release(pvt);
1947
1948 /* Remove MC sysfs nodes */
1949 edac_mc_del_mc(mci->dev);
1950
1951 debugf1("%s: free mci struct\n", mci->ctl_name);
1952 kfree(mci->ctl_name);
1953 edac_mc_free(mci);
1954 i7core_dev->mci = NULL;
1955}
1956
Hidetoshi Setoaace4282010-08-20 04:32:45 -03001957static int i7core_register_mci(struct i7core_dev *i7core_dev)
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03001958{
1959 struct mem_ctl_info *mci;
1960 struct i7core_pvt *pvt;
Hidetoshi Setoaace4282010-08-20 04:32:45 -03001961 int rc, channels, csrows;
1962
1963 /* Check the number of active and not disabled channels */
1964 rc = i7core_get_active_channels(i7core_dev->socket, &channels, &csrows);
1965 if (unlikely(rc < 0))
1966 return rc;
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -03001967
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03001968 /* allocate a new MC control structure */
Hidetoshi Setoaace4282010-08-20 04:32:45 -03001969 mci = edac_mc_alloc(sizeof(*pvt), csrows, channels, i7core_dev->socket);
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001970 if (unlikely(!mci))
1971 return -ENOMEM;
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03001972
Mauro Carvalho Chehab3cfd0142010-08-10 23:23:46 -03001973 debugf0("MC: " __FILE__ ": %s(): mci = %p, dev = %p\n",
1974 __func__, mci, &i7core_dev->pdev[0]->dev);
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03001975
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03001976 pvt = mci->pvt_info;
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -03001977 memset(pvt, 0, sizeof(*pvt));
Mauro Carvalho Chehab67166af2009-07-15 06:56:23 -03001978
Mauro Carvalho Chehab6d37d242010-08-20 12:48:26 -03001979 /* Associates i7core_dev and mci for future usage */
1980 pvt->i7core_dev = i7core_dev;
1981 i7core_dev->mci = mci;
1982
Mauro Carvalho Chehab41fcb7f2009-06-22 22:48:31 -03001983 /*
1984 * FIXME: how to handle RDDR3 at MCI level? It is possible to have
1985 * Mixed RDDR3/UDDR3 with Nehalem, provided that they are on different
1986 * memory channels
1987 */
1988 mci->mtype_cap = MEM_FLAG_DDR3;
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03001989 mci->edac_ctl_cap = EDAC_FLAG_NONE;
1990 mci->edac_cap = EDAC_FLAG_NONE;
1991 mci->mod_name = "i7core_edac.c";
1992 mci->mod_ver = I7CORE_REVISION;
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001993 mci->ctl_name = kasprintf(GFP_KERNEL, "i7 core #%d",
1994 i7core_dev->socket);
1995 mci->dev_name = pci_name(i7core_dev->pdev[0]);
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03001996 mci->ctl_page_to_phys = NULL;
Mauro Carvalho Chehab1288c182010-08-10 18:57:01 -03001997
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -03001998 /* Store pci devices at mci for faster access */
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03001999 rc = mci_bind_devs(mci, i7core_dev);
Mauro Carvalho Chehab41fcb7f2009-06-22 22:48:31 -03002000 if (unlikely(rc < 0))
Hidetoshi Seto628c5dd2010-08-20 04:28:40 -03002001 goto fail0;
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -03002002
Hidetoshi Seto59398132010-08-20 04:28:25 -03002003 if (pvt->is_registered)
2004 mci->mc_driver_sysfs_attributes = i7core_sysfs_rdimm_attrs;
2005 else
2006 mci->mc_driver_sysfs_attributes = i7core_sysfs_udimm_attrs;
2007
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -03002008 /* Get dimm basic config */
Hidetoshi Seto2e5185f2010-08-20 04:32:56 -03002009 get_dimm_config(mci);
Hidetoshi Seto59398132010-08-20 04:28:25 -03002010 /* record ptr to the generic device */
2011 mci->dev = &i7core_dev->pdev[0]->dev;
2012 /* Set the function pointer to an actual operation function */
2013 mci->edac_check = i7core_check_error;
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -03002014
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03002015 /* add this new MC control structure to EDAC's list of MCs */
Mauro Carvalho Chehabb7c76152009-06-22 22:48:30 -03002016 if (unlikely(edac_mc_add_mc(mci))) {
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03002017 debugf0("MC: " __FILE__
2018 ": %s(): failed edac_mc_add_mc()\n", __func__);
2019 /* FIXME: perhaps some code should go here that disables error
2020 * reporting if we just enabled it
2021 */
Mauro Carvalho Chehabb7c76152009-06-22 22:48:30 -03002022
2023 rc = -EINVAL;
Hidetoshi Seto628c5dd2010-08-20 04:28:40 -03002024 goto fail0;
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03002025 }
2026
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03002027 /* Default error mask is any memory */
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -03002028 pvt->inject.channel = 0;
Mauro Carvalho Chehab194a40f2009-06-22 22:48:28 -03002029 pvt->inject.dimm = -1;
2030 pvt->inject.rank = -1;
2031 pvt->inject.bank = -1;
2032 pvt->inject.page = -1;
2033 pvt->inject.col = -1;
2034
Hidetoshi Setoa3aa0a42010-08-20 04:25:18 -03002035 /* allocating generic PCI control info */
2036 i7core_pci_ctl_create(pvt);
2037
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -03002038 /* Registers on edac_mce in order to receive memory errors */
Mauro Carvalho Chehabc5d34522009-07-17 10:28:15 -03002039 pvt->edac_mce.priv = mci;
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -03002040 pvt->edac_mce.check_error = i7core_mce_check_error;
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -03002041 rc = edac_mce_register(&pvt->edac_mce);
Mauro Carvalho Chehabb9905382009-08-05 21:36:35 -03002042 if (unlikely(rc < 0)) {
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -03002043 debugf0("MC: " __FILE__
2044 ": %s(): failed edac_mce_register()\n", __func__);
Hidetoshi Seto628c5dd2010-08-20 04:28:40 -03002045 goto fail1;
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03002046 }
2047
Hidetoshi Seto628c5dd2010-08-20 04:28:40 -03002048 return 0;
2049
2050fail1:
2051 i7core_pci_ctl_release(pvt);
2052 edac_mc_del_mc(mci->dev);
2053fail0:
2054 kfree(mci->ctl_name);
2055 edac_mc_free(mci);
Hidetoshi Seto1c6edbb2010-08-20 04:32:33 -03002056 i7core_dev->mci = NULL;
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03002057 return rc;
2058}
2059
2060/*
2061 * i7core_probe Probe for ONE instance of device to see if it is
2062 * present.
2063 * return:
2064 * 0 for FOUND a device
2065 * < 0 for error code
2066 */
Mauro Carvalho Chehab2d95d812010-06-30 01:42:21 -03002067
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03002068static int __devinit i7core_probe(struct pci_dev *pdev,
2069 const struct pci_device_id *id)
2070{
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03002071 int rc;
2072 struct i7core_dev *i7core_dev;
2073
Mauro Carvalho Chehab2d95d812010-06-30 01:42:21 -03002074 /* get the pci devices we want to reserve for our use */
2075 mutex_lock(&i7core_edac_lock);
2076
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03002077 /*
Mauro Carvalho Chehabd4c27792009-09-05 04:12:02 -03002078 * All memory controllers are allocated at the first pass.
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03002079 */
Mauro Carvalho Chehab2d95d812010-06-30 01:42:21 -03002080 if (unlikely(probed >= 1)) {
2081 mutex_unlock(&i7core_edac_lock);
Mauro Carvalho Chehab76a7bd82010-10-24 11:36:19 -02002082 return -ENODEV;
Mauro Carvalho Chehab2d95d812010-06-30 01:42:21 -03002083 }
2084 probed++;
Mauro Carvalho Chehabde06eee2009-10-14 08:02:40 -03002085
Hidetoshi Seto64c10f62010-08-20 04:28:14 -03002086 rc = i7core_get_all_devices();
Mauro Carvalho Chehabf4742942009-09-05 02:35:08 -03002087 if (unlikely(rc < 0))
2088 goto fail0;
2089
2090 list_for_each_entry(i7core_dev, &i7core_edac_list, list) {
Hidetoshi Setoaace4282010-08-20 04:32:45 -03002091 rc = i7core_register_mci(i7core_dev);
Mauro Carvalho Chehabd4c27792009-09-05 04:12:02 -03002092 if (unlikely(rc < 0))
2093 goto fail1;
Mauro Carvalho Chehabd5381642009-07-09 22:06:41 -03002094 }
2095
Mauro Carvalho Chehabef708b52009-06-22 22:48:30 -03002096 i7core_printk(KERN_INFO, "Driver loaded.\n");
Mauro Carvalho Chehab8f331902009-06-22 22:48:29 -03002097
Mauro Carvalho Chehab66607702009-09-05 00:52:11 -03002098 mutex_unlock(&i7core_edac_lock);
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03002099 return 0;
2100
Mauro Carvalho Chehab66607702009-09-05 00:52:11 -03002101fail1:
Mauro Carvalho Chehab88ef5ea2010-08-20 15:39:38 -03002102 list_for_each_entry(i7core_dev, &i7core_edac_list, list)
2103 i7core_unregister_mci(i7core_dev);
2104
Mauro Carvalho Chehab13d6e9b2009-09-05 12:15:20 -03002105 i7core_put_all_devices();
Mauro Carvalho Chehab66607702009-09-05 00:52:11 -03002106fail0:
2107 mutex_unlock(&i7core_edac_lock);
Mauro Carvalho Chehabb7c76152009-06-22 22:48:30 -03002108 return rc;
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03002109}
2110
2111/*
2112 * i7core_remove destructor for one instance of device
2113 *
2114 */
2115static void __devexit i7core_remove(struct pci_dev *pdev)
2116{
Hidetoshi Seto64c10f62010-08-20 04:28:14 -03002117 struct i7core_dev *i7core_dev;
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03002118
2119 debugf0(__FILE__ ": %s()\n", __func__);
2120
Mauro Carvalho Chehab22e6bcb2009-09-05 23:06:50 -03002121 /*
2122 * we have a trouble here: pdev value for removal will be wrong, since
2123 * it will point to the X58 register used to detect that the machine
2124 * is a Nehalem or upper design. However, due to the way several PCI
2125 * devices are grouped together to provide MC functionality, we need
2126 * to use a different method for releasing the devices
2127 */
Mauro Carvalho Chehab87d1d272009-06-22 22:48:29 -03002128
Mauro Carvalho Chehab66607702009-09-05 00:52:11 -03002129 mutex_lock(&i7core_edac_lock);
Hidetoshi Seto71fe0172010-08-20 04:29:47 -03002130
2131 if (unlikely(!probed)) {
2132 mutex_unlock(&i7core_edac_lock);
2133 return;
2134 }
2135
Mauro Carvalho Chehab88ef5ea2010-08-20 15:39:38 -03002136 list_for_each_entry(i7core_dev, &i7core_edac_list, list)
2137 i7core_unregister_mci(i7core_dev);
Hidetoshi Seto64c10f62010-08-20 04:28:14 -03002138
2139 /* Release PCI resources */
2140 i7core_put_all_devices();
2141
Mauro Carvalho Chehab2d95d812010-06-30 01:42:21 -03002142 probed--;
2143
Mauro Carvalho Chehab22e6bcb2009-09-05 23:06:50 -03002144 mutex_unlock(&i7core_edac_lock);
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03002145}
2146
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03002147MODULE_DEVICE_TABLE(pci, i7core_pci_tbl);
2148
2149/*
2150 * i7core_driver pci_driver structure for this module
2151 *
2152 */
2153static struct pci_driver i7core_driver = {
2154 .name = "i7core_edac",
2155 .probe = i7core_probe,
2156 .remove = __devexit_p(i7core_remove),
2157 .id_table = i7core_pci_tbl,
2158};
2159
2160/*
2161 * i7core_init Module entry function
2162 * Try to initialize this module for its devices
2163 */
2164static int __init i7core_init(void)
2165{
2166 int pci_rc;
2167
2168 debugf2("MC: " __FILE__ ": %s()\n", __func__);
2169
2170 /* Ensure that the OPSTATE is set correctly for POLL or NMI */
2171 opstate_init();
2172
Mauro Carvalho Chehab54a08ab2010-08-19 15:51:00 -03002173 if (use_pci_fixup)
2174 i7core_xeon_pci_fixup(pci_dev_table);
Keith Manntheybc2d7242009-09-03 00:05:05 -03002175
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03002176 pci_rc = pci_register_driver(&i7core_driver);
2177
Mauro Carvalho Chehab3ef288a2009-09-02 23:43:33 -03002178 if (pci_rc >= 0)
2179 return 0;
2180
2181 i7core_printk(KERN_ERR, "Failed to register device with error %d.\n",
2182 pci_rc);
2183
2184 return pci_rc;
Mauro Carvalho Chehaba0c36a12009-06-22 22:41:15 -03002185}
2186
2187/*
2188 * i7core_exit() Module exit function
2189 * Unregister the driver
2190 */
2191static void __exit i7core_exit(void)
2192{
2193 debugf2("MC: " __FILE__ ": %s()\n", __func__);
2194 pci_unregister_driver(&i7core_driver);
2195}
2196
2197module_init(i7core_init);
2198module_exit(i7core_exit);
2199
2200MODULE_LICENSE("GPL");
2201MODULE_AUTHOR("Mauro Carvalho Chehab <mchehab@redhat.com>");
2202MODULE_AUTHOR("Red Hat Inc. (http://www.redhat.com)");
2203MODULE_DESCRIPTION("MC Driver for Intel i7 Core memory controllers - "
2204 I7CORE_REVISION);
2205
2206module_param(edac_op_state, int, 0444);
2207MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");