blob: dbdffe8d44693afc648d0e0755fa86e0d0d629fe [file] [log] [blame]
Dave Jiangc0d12172007-07-19 01:49:46 -07001/*
2 * Generic EDAC defs
3 *
4 * Author: Dave Jiang <djiang@mvista.com>
5 *
Hitoshi Mitakec3c52bc2008-04-29 01:03:18 -07006 * 2006-2008 (c) MontaVista Software, Inc. This file is licensed under
Dave Jiangc0d12172007-07-19 01:49:46 -07007 * the terms of the GNU General Public License version 2. This program
8 * is licensed "as is" without any warranty of any kind, whether express
9 * or implied.
10 *
11 */
12#ifndef _LINUX_EDAC_H_
13#define _LINUX_EDAC_H_
14
Arun Sharma600634972011-07-26 16:09:06 -070015#include <linux/atomic.h>
Mauro Carvalho Chehab7a623c02012-04-16 16:41:11 -030016#include <linux/device.h>
Paul Gortmaker313162d2012-01-30 11:46:54 -050017#include <linux/completion.h>
18#include <linux/workqueue.h>
Mauro Carvalho Chehab452a6bf2012-03-26 09:35:11 -030019#include <linux/debugfs.h>
Paul Gortmaker313162d2012-01-30 11:46:54 -050020
21struct device;
Dave Jiangc0d12172007-07-19 01:49:46 -070022
23#define EDAC_OPSTATE_INVAL -1
24#define EDAC_OPSTATE_POLL 0
25#define EDAC_OPSTATE_NMI 1
26#define EDAC_OPSTATE_INT 2
27
28extern int edac_op_state;
Dave Jiang66ee2f92007-07-19 01:49:54 -070029extern int edac_err_assert;
Dave Jiangc0d12172007-07-19 01:49:46 -070030extern atomic_t edac_handlers;
Kay Sieversfe5ff8b2011-12-14 15:21:07 -080031extern struct bus_type edac_subsys;
Dave Jiangc0d12172007-07-19 01:49:46 -070032
33extern int edac_handler_set(void);
34extern void edac_atomic_assert_error(void);
Kay Sieversfe5ff8b2011-12-14 15:21:07 -080035extern struct bus_type *edac_get_sysfs_subsys(void);
36extern void edac_put_sysfs_subsys(void);
Dave Jiangc0d12172007-07-19 01:49:46 -070037
Hitoshi Mitakec3c52bc2008-04-29 01:03:18 -070038static inline void opstate_init(void)
39{
40 switch (edac_op_state) {
41 case EDAC_OPSTATE_POLL:
42 case EDAC_OPSTATE_NMI:
43 break;
44 default:
45 edac_op_state = EDAC_OPSTATE_POLL;
46 }
47 return;
48}
49
Mauro Carvalho Chehabc7ef7642013-02-21 13:36:45 -030050/* Max length of a DIMM label*/
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -030051#define EDAC_MC_LABEL_LEN 31
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -030052
Mauro Carvalho Chehabc7ef7642013-02-21 13:36:45 -030053/* Maximum size of the location string */
Chen, Gong56507692013-10-18 14:30:38 -070054#define LOCATION_SIZE 256
Mauro Carvalho Chehabc7ef7642013-02-21 13:36:45 -030055
56/* Defines the maximum number of labels that can be reported */
57#define EDAC_MAX_LABELS 8
58
59/* String used to join two or more labels */
60#define OTHER_LABEL " or "
61
Mauro Carvalho Chehabb0610bb2012-03-21 16:21:07 -030062/**
63 * enum dev_type - describe the type of memory DRAM chips used at the stick
64 * @DEV_UNKNOWN: Can't be determined, or MC doesn't support detect it
65 * @DEV_X1: 1 bit for data
66 * @DEV_X2: 2 bits for data
67 * @DEV_X4: 4 bits for data
68 * @DEV_X8: 8 bits for data
69 * @DEV_X16: 16 bits for data
70 * @DEV_X32: 32 bits for data
71 * @DEV_X64: 64 bits for data
72 *
73 * Typical values are x4 and x8.
74 */
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -030075enum dev_type {
76 DEV_UNKNOWN = 0,
77 DEV_X1,
78 DEV_X2,
79 DEV_X4,
80 DEV_X8,
81 DEV_X16,
82 DEV_X32, /* Do these parts exist? */
83 DEV_X64 /* Do these parts exist? */
84};
85
86#define DEV_FLAG_UNKNOWN BIT(DEV_UNKNOWN)
87#define DEV_FLAG_X1 BIT(DEV_X1)
88#define DEV_FLAG_X2 BIT(DEV_X2)
89#define DEV_FLAG_X4 BIT(DEV_X4)
90#define DEV_FLAG_X8 BIT(DEV_X8)
91#define DEV_FLAG_X16 BIT(DEV_X16)
92#define DEV_FLAG_X32 BIT(DEV_X32)
93#define DEV_FLAG_X64 BIT(DEV_X64)
94
Mauro Carvalho Chehab01a6e282012-02-03 13:17:48 -030095/**
Mauro Carvalho Chehab982216a2012-04-16 13:04:46 -030096 * enum hw_event_mc_err_type - type of the detected error
97 *
98 * @HW_EVENT_ERR_CORRECTED: Corrected Error - Indicates that an ECC
99 * corrected error was detected
100 * @HW_EVENT_ERR_UNCORRECTED: Uncorrected Error - Indicates an error that
101 * can't be corrected by ECC, but it is not
102 * fatal (maybe it is on an unused memory area,
103 * or the memory controller could recover from
104 * it for example, by re-trying the operation).
105 * @HW_EVENT_ERR_FATAL: Fatal Error - Uncorrected error that could not
106 * be recovered.
107 */
108enum hw_event_mc_err_type {
109 HW_EVENT_ERR_CORRECTED,
110 HW_EVENT_ERR_UNCORRECTED,
111 HW_EVENT_ERR_FATAL,
Mauro Carvalho Chehab8dd93d42013-02-19 21:26:22 -0300112 HW_EVENT_ERR_INFO,
Mauro Carvalho Chehab982216a2012-04-16 13:04:46 -0300113};
114
Mauro Carvalho Chehab8dd93d42013-02-19 21:26:22 -0300115static inline char *mc_event_error_type(const unsigned int err_type)
116{
117 switch (err_type) {
118 case HW_EVENT_ERR_CORRECTED:
119 return "Corrected";
120 case HW_EVENT_ERR_UNCORRECTED:
121 return "Uncorrected";
122 case HW_EVENT_ERR_FATAL:
123 return "Fatal";
124 default:
125 case HW_EVENT_ERR_INFO:
126 return "Info";
127 }
128}
129
Mauro Carvalho Chehab982216a2012-04-16 13:04:46 -0300130/**
Mauro Carvalho Chehab01a6e282012-02-03 13:17:48 -0300131 * enum mem_type - memory types. For a more detailed reference, please see
132 * http://en.wikipedia.org/wiki/DRAM
133 *
134 * @MEM_EMPTY Empty csrow
135 * @MEM_RESERVED: Reserved csrow type
136 * @MEM_UNKNOWN: Unknown csrow type
137 * @MEM_FPM: FPM - Fast Page Mode, used on systems up to 1995.
138 * @MEM_EDO: EDO - Extended data out, used on systems up to 1998.
139 * @MEM_BEDO: BEDO - Burst Extended data out, an EDO variant.
140 * @MEM_SDR: SDR - Single data rate SDRAM
141 * http://en.wikipedia.org/wiki/Synchronous_dynamic_random-access_memory
142 * They use 3 pins for chip select: Pins 0 and 2 are
143 * for rank 0; pins 1 and 3 are for rank 1, if the memory
144 * is dual-rank.
145 * @MEM_RDR: Registered SDR SDRAM
146 * @MEM_DDR: Double data rate SDRAM
147 * http://en.wikipedia.org/wiki/DDR_SDRAM
148 * @MEM_RDDR: Registered Double data rate SDRAM
149 * This is a variant of the DDR memories.
150 * A registered memory has a buffer inside it, hiding
151 * part of the memory details to the memory controller.
152 * @MEM_RMBS: Rambus DRAM, used on a few Pentium III/IV controllers.
153 * @MEM_DDR2: DDR2 RAM, as described at JEDEC JESD79-2F.
154 * Those memories are labed as "PC2-" instead of "PC" to
155 * differenciate from DDR.
156 * @MEM_FB_DDR2: Fully-Buffered DDR2, as described at JEDEC Std No. 205
157 * and JESD206.
158 * Those memories are accessed per DIMM slot, and not by
159 * a chip select signal.
160 * @MEM_RDDR2: Registered DDR2 RAM
161 * This is a variant of the DDR2 memories.
162 * @MEM_XDR: Rambus XDR
163 * It is an evolution of the original RAMBUS memories,
164 * created to compete with DDR2. Weren't used on any
165 * x86 arch, but cell_edac PPC memory controller uses it.
166 * @MEM_DDR3: DDR3 RAM
167 * @MEM_RDDR3: Registered DDR3 RAM
168 * This is a variant of the DDR3 memories.
169 */
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300170enum mem_type {
Mauro Carvalho Chehab01a6e282012-02-03 13:17:48 -0300171 MEM_EMPTY = 0,
172 MEM_RESERVED,
173 MEM_UNKNOWN,
174 MEM_FPM,
175 MEM_EDO,
176 MEM_BEDO,
177 MEM_SDR,
178 MEM_RDR,
179 MEM_DDR,
180 MEM_RDDR,
181 MEM_RMBS,
182 MEM_DDR2,
183 MEM_FB_DDR2,
184 MEM_RDDR2,
185 MEM_XDR,
186 MEM_DDR3,
187 MEM_RDDR3,
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300188};
189
190#define MEM_FLAG_EMPTY BIT(MEM_EMPTY)
191#define MEM_FLAG_RESERVED BIT(MEM_RESERVED)
192#define MEM_FLAG_UNKNOWN BIT(MEM_UNKNOWN)
193#define MEM_FLAG_FPM BIT(MEM_FPM)
194#define MEM_FLAG_EDO BIT(MEM_EDO)
195#define MEM_FLAG_BEDO BIT(MEM_BEDO)
196#define MEM_FLAG_SDR BIT(MEM_SDR)
197#define MEM_FLAG_RDR BIT(MEM_RDR)
198#define MEM_FLAG_DDR BIT(MEM_DDR)
199#define MEM_FLAG_RDDR BIT(MEM_RDDR)
200#define MEM_FLAG_RMBS BIT(MEM_RMBS)
201#define MEM_FLAG_DDR2 BIT(MEM_DDR2)
202#define MEM_FLAG_FB_DDR2 BIT(MEM_FB_DDR2)
203#define MEM_FLAG_RDDR2 BIT(MEM_RDDR2)
204#define MEM_FLAG_XDR BIT(MEM_XDR)
205#define MEM_FLAG_DDR3 BIT(MEM_DDR3)
206#define MEM_FLAG_RDDR3 BIT(MEM_RDDR3)
207
Mauro Carvalho Chehabb0610bb2012-03-21 16:21:07 -0300208/**
209 * enum edac-type - Error Detection and Correction capabilities and mode
210 * @EDAC_UNKNOWN: Unknown if ECC is available
211 * @EDAC_NONE: Doesn't support ECC
212 * @EDAC_RESERVED: Reserved ECC type
213 * @EDAC_PARITY: Detects parity errors
214 * @EDAC_EC: Error Checking - no correction
215 * @EDAC_SECDED: Single bit error correction, Double detection
216 * @EDAC_S2ECD2ED: Chipkill x2 devices - do these exist?
217 * @EDAC_S4ECD4ED: Chipkill x4 devices
218 * @EDAC_S8ECD8ED: Chipkill x8 devices
219 * @EDAC_S16ECD16ED: Chipkill x16 devices
220 */
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300221enum edac_type {
Mauro Carvalho Chehabb0610bb2012-03-21 16:21:07 -0300222 EDAC_UNKNOWN = 0,
223 EDAC_NONE,
224 EDAC_RESERVED,
225 EDAC_PARITY,
226 EDAC_EC,
227 EDAC_SECDED,
228 EDAC_S2ECD2ED,
229 EDAC_S4ECD4ED,
230 EDAC_S8ECD8ED,
231 EDAC_S16ECD16ED,
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300232};
233
234#define EDAC_FLAG_UNKNOWN BIT(EDAC_UNKNOWN)
235#define EDAC_FLAG_NONE BIT(EDAC_NONE)
236#define EDAC_FLAG_PARITY BIT(EDAC_PARITY)
237#define EDAC_FLAG_EC BIT(EDAC_EC)
238#define EDAC_FLAG_SECDED BIT(EDAC_SECDED)
239#define EDAC_FLAG_S2ECD2ED BIT(EDAC_S2ECD2ED)
240#define EDAC_FLAG_S4ECD4ED BIT(EDAC_S4ECD4ED)
241#define EDAC_FLAG_S8ECD8ED BIT(EDAC_S8ECD8ED)
242#define EDAC_FLAG_S16ECD16ED BIT(EDAC_S16ECD16ED)
243
Mauro Carvalho Chehabb0610bb2012-03-21 16:21:07 -0300244/**
245 * enum scrub_type - scrubbing capabilities
246 * @SCRUB_UNKNOWN Unknown if scrubber is available
247 * @SCRUB_NONE: No scrubber
248 * @SCRUB_SW_PROG: SW progressive (sequential) scrubbing
249 * @SCRUB_SW_SRC: Software scrub only errors
250 * @SCRUB_SW_PROG_SRC: Progressive software scrub from an error
251 * @SCRUB_SW_TUNABLE: Software scrub frequency is tunable
252 * @SCRUB_HW_PROG: HW progressive (sequential) scrubbing
253 * @SCRUB_HW_SRC: Hardware scrub only errors
254 * @SCRUB_HW_PROG_SRC: Progressive hardware scrub from an error
255 * SCRUB_HW_TUNABLE: Hardware scrub frequency is tunable
256 */
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300257enum scrub_type {
Mauro Carvalho Chehabb0610bb2012-03-21 16:21:07 -0300258 SCRUB_UNKNOWN = 0,
259 SCRUB_NONE,
260 SCRUB_SW_PROG,
261 SCRUB_SW_SRC,
262 SCRUB_SW_PROG_SRC,
263 SCRUB_SW_TUNABLE,
264 SCRUB_HW_PROG,
265 SCRUB_HW_SRC,
266 SCRUB_HW_PROG_SRC,
267 SCRUB_HW_TUNABLE
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300268};
269
270#define SCRUB_FLAG_SW_PROG BIT(SCRUB_SW_PROG)
271#define SCRUB_FLAG_SW_SRC BIT(SCRUB_SW_SRC)
272#define SCRUB_FLAG_SW_PROG_SRC BIT(SCRUB_SW_PROG_SRC)
273#define SCRUB_FLAG_SW_TUN BIT(SCRUB_SW_SCRUB_TUNABLE)
274#define SCRUB_FLAG_HW_PROG BIT(SCRUB_HW_PROG)
275#define SCRUB_FLAG_HW_SRC BIT(SCRUB_HW_SRC)
276#define SCRUB_FLAG_HW_PROG_SRC BIT(SCRUB_HW_PROG_SRC)
277#define SCRUB_FLAG_HW_TUN BIT(SCRUB_HW_TUNABLE)
278
279/* FIXME - should have notify capabilities: NMI, LOG, PROC, etc */
280
281/* EDAC internal operation states */
282#define OP_ALLOC 0x100
283#define OP_RUNNING_POLL 0x201
284#define OP_RUNNING_INTERRUPT 0x202
285#define OP_RUNNING_POLL_INTR 0x203
286#define OP_OFFLINE 0x300
287
288/*
Mauro Carvalho Chehab01a6e282012-02-03 13:17:48 -0300289 * Concepts used at the EDAC subsystem
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300290 *
Mauro Carvalho Chehab01a6e282012-02-03 13:17:48 -0300291 * There are several things to be aware of that aren't at all obvious:
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300292 *
293 * SOCKETS, SOCKET SETS, BANKS, ROWS, CHIP-SELECT ROWS, CHANNELS, etc..
294 *
295 * These are some of the many terms that are thrown about that don't always
296 * mean what people think they mean (Inconceivable!). In the interest of
297 * creating a common ground for discussion, terms and their definitions
298 * will be established.
299 *
Mauro Carvalho Chehab01a6e282012-02-03 13:17:48 -0300300 * Memory devices: The individual DRAM chips on a memory stick. These
301 * devices commonly output 4 and 8 bits each (x4, x8).
302 * Grouping several of these in parallel provides the
303 * number of bits that the memory controller expects:
304 * typically 72 bits, in order to provide 64 bits +
305 * 8 bits of ECC data.
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300306 *
307 * Memory Stick: A printed circuit board that aggregates multiple
Mauro Carvalho Chehab01a6e282012-02-03 13:17:48 -0300308 * memory devices in parallel. In general, this is the
309 * Field Replaceable Unit (FRU) which gets replaced, in
310 * the case of excessive errors. Most often it is also
311 * called DIMM (Dual Inline Memory Module).
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300312 *
Mauro Carvalho Chehab01a6e282012-02-03 13:17:48 -0300313 * Memory Socket: A physical connector on the motherboard that accepts
314 * a single memory stick. Also called as "slot" on several
315 * datasheets.
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300316 *
Mauro Carvalho Chehab01a6e282012-02-03 13:17:48 -0300317 * Channel: A memory controller channel, responsible to communicate
318 * with a group of DIMMs. Each channel has its own
319 * independent control (command) and data bus, and can
320 * be used independently or grouped with other channels.
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300321 *
Mauro Carvalho Chehab01a6e282012-02-03 13:17:48 -0300322 * Branch: It is typically the highest hierarchy on a
323 * Fully-Buffered DIMM memory controller.
324 * Typically, it contains two channels.
325 * Two channels at the same branch can be used in single
326 * mode or in lockstep mode.
327 * When lockstep is enabled, the cacheline is doubled,
328 * but it generally brings some performance penalty.
329 * Also, it is generally not possible to point to just one
330 * memory stick when an error occurs, as the error
331 * correction code is calculated using two DIMMs instead
332 * of one. Due to that, it is capable of correcting more
333 * errors than on single mode.
334 *
335 * Single-channel: The data accessed by the memory controller is contained
336 * into one dimm only. E. g. if the data is 64 bits-wide,
337 * the data flows to the CPU using one 64 bits parallel
338 * access.
339 * Typically used with SDR, DDR, DDR2 and DDR3 memories.
340 * FB-DIMM and RAMBUS use a different concept for channel,
341 * so this concept doesn't apply there.
342 *
343 * Double-channel: The data size accessed by the memory controller is
344 * interlaced into two dimms, accessed at the same time.
345 * E. g. if the DIMM is 64 bits-wide (72 bits with ECC),
346 * the data flows to the CPU using a 128 bits parallel
347 * access.
348 *
349 * Chip-select row: This is the name of the DRAM signal used to select the
350 * DRAM ranks to be accessed. Common chip-select rows for
351 * single channel are 64 bits, for dual channel 128 bits.
352 * It may not be visible by the memory controller, as some
353 * DIMM types have a memory buffer that can hide direct
354 * access to it from the Memory Controller.
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300355 *
356 * Single-Ranked stick: A Single-ranked stick has 1 chip-select row of memory.
357 * Motherboards commonly drive two chip-select pins to
358 * a memory stick. A single-ranked stick, will occupy
359 * only one of those rows. The other will be unused.
360 *
361 * Double-Ranked stick: A double-ranked stick has two chip-select rows which
362 * access different sets of memory devices. The two
363 * rows cannot be accessed concurrently.
364 *
365 * Double-sided stick: DEPRECATED TERM, see Double-Ranked stick.
366 * A double-sided stick has two chip-select rows which
Mauro Carvalho Chehab01a6e282012-02-03 13:17:48 -0300367 * access different sets of memory devices. The two
368 * rows cannot be accessed concurrently. "Double-sided"
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300369 * is irrespective of the memory devices being mounted
370 * on both sides of the memory stick.
371 *
372 * Socket set: All of the memory sticks that are required for
373 * a single memory access or all of the memory sticks
374 * spanned by a chip-select row. A single socket set
375 * has two chip-select rows and if double-sided sticks
376 * are used these will occupy those chip-select rows.
377 *
378 * Bank: This term is avoided because it is unclear when
379 * needing to distinguish between chip-select rows and
380 * socket sets.
381 *
382 * Controller pages:
383 *
384 * Physical pages:
385 *
386 * Virtual pages:
387 *
388 *
389 * STRUCTURE ORGANIZATION AND CHOICES
390 *
391 *
392 *
393 * PS - I enjoyed writing all that about as much as you enjoyed reading it.
394 */
395
Mauro Carvalho Chehab982216a2012-04-16 13:04:46 -0300396/**
397 * enum edac_mc_layer - memory controller hierarchy layer
398 *
399 * @EDAC_MC_LAYER_BRANCH: memory layer is named "branch"
400 * @EDAC_MC_LAYER_CHANNEL: memory layer is named "channel"
401 * @EDAC_MC_LAYER_SLOT: memory layer is named "slot"
402 * @EDAC_MC_LAYER_CHIP_SELECT: memory layer is named "chip select"
Mauro Carvalho Chehabc66b5a72013-02-15 07:21:08 -0300403 * @EDAC_MC_LAYER_ALL_MEM: memory layout is unknown. All memory is mapped
404 * as a single memory area. This is used when
405 * retrieving errors from a firmware driven driver.
Mauro Carvalho Chehab982216a2012-04-16 13:04:46 -0300406 *
407 * This enum is used by the drivers to tell edac_mc_sysfs what name should
408 * be used when describing a memory stick location.
409 */
410enum edac_mc_layer_type {
411 EDAC_MC_LAYER_BRANCH,
412 EDAC_MC_LAYER_CHANNEL,
413 EDAC_MC_LAYER_SLOT,
414 EDAC_MC_LAYER_CHIP_SELECT,
Mauro Carvalho Chehabc66b5a72013-02-15 07:21:08 -0300415 EDAC_MC_LAYER_ALL_MEM,
Mauro Carvalho Chehab982216a2012-04-16 13:04:46 -0300416};
417
418/**
419 * struct edac_mc_layer - describes the memory controller hierarchy
420 * @layer: layer type
421 * @size: number of components per layer. For example,
422 * if the channel layer has two channels, size = 2
423 * @is_virt_csrow: This layer is part of the "csrow" when old API
424 * compatibility mode is enabled. Otherwise, it is
425 * a channel
426 */
427struct edac_mc_layer {
428 enum edac_mc_layer_type type;
429 unsigned size;
430 bool is_virt_csrow;
431};
432
433/*
434 * Maximum number of layers used by the memory controller to uniquely
435 * identify a single memory stick.
436 * NOTE: Changing this constant requires not only to change the constant
437 * below, but also to change the existing code at the core, as there are
438 * some code there that are optimized for 3 layers.
439 */
440#define EDAC_MAX_LAYERS 3
441
442/**
Mauro Carvalho Chehabde3910eb2012-04-24 15:05:43 -0300443 * EDAC_DIMM_OFF - Macro responsible to get a pointer offset inside a pointer array
444 * for the element given by [layer0,layer1,layer2] position
445 *
446 * @layers: a struct edac_mc_layer array, describing how many elements
447 * were allocated for each layer
448 * @n_layers: Number of layers at the @layers array
449 * @layer0: layer0 position
450 * @layer1: layer1 position. Unused if n_layers < 2
451 * @layer2: layer2 position. Unused if n_layers < 3
452 *
453 * For 1 layer, this macro returns &var[layer0] - &var
454 * For 2 layers, this macro is similar to allocate a bi-dimensional array
455 * and to return "&var[layer0][layer1] - &var"
456 * For 3 layers, this macro is similar to allocate a tri-dimensional array
457 * and to return "&var[layer0][layer1][layer2] - &var"
458 *
459 * A loop could be used here to make it more generic, but, as we only have
460 * 3 layers, this is a little faster.
461 * By design, layers can never be 0 or more than 3. If that ever happens,
462 * a NULL is returned, causing an OOPS during the memory allocation routine,
463 * with would point to the developer that he's doing something wrong.
464 */
465#define EDAC_DIMM_OFF(layers, nlayers, layer0, layer1, layer2) ({ \
466 int __i; \
467 if ((nlayers) == 1) \
468 __i = layer0; \
469 else if ((nlayers) == 2) \
470 __i = (layer1) + ((layers[1]).size * (layer0)); \
471 else if ((nlayers) == 3) \
472 __i = (layer2) + ((layers[2]).size * ((layer1) + \
473 ((layers[1]).size * (layer0)))); \
474 else \
475 __i = -EINVAL; \
476 __i; \
477})
478
479/**
480 * EDAC_DIMM_PTR - Macro responsible to get a pointer inside a pointer array
Mauro Carvalho Chehab982216a2012-04-16 13:04:46 -0300481 * for the element given by [layer0,layer1,layer2] position
482 *
483 * @layers: a struct edac_mc_layer array, describing how many elements
484 * were allocated for each layer
485 * @var: name of the var where we want to get the pointer
486 * (like mci->dimms)
487 * @n_layers: Number of layers at the @layers array
488 * @layer0: layer0 position
489 * @layer1: layer1 position. Unused if n_layers < 2
490 * @layer2: layer2 position. Unused if n_layers < 3
491 *
492 * For 1 layer, this macro returns &var[layer0]
493 * For 2 layers, this macro is similar to allocate a bi-dimensional array
494 * and to return "&var[layer0][layer1]"
495 * For 3 layers, this macro is similar to allocate a tri-dimensional array
496 * and to return "&var[layer0][layer1][layer2]"
Mauro Carvalho Chehab982216a2012-04-16 13:04:46 -0300497 */
498#define EDAC_DIMM_PTR(layers, var, nlayers, layer0, layer1, layer2) ({ \
Mauro Carvalho Chehabde3910eb2012-04-24 15:05:43 -0300499 typeof(*var) __p; \
500 int ___i = EDAC_DIMM_OFF(layers, nlayers, layer0, layer1, layer2); \
501 if (___i < 0) \
Mauro Carvalho Chehab982216a2012-04-16 13:04:46 -0300502 __p = NULL; \
Mauro Carvalho Chehabde3910eb2012-04-24 15:05:43 -0300503 else \
504 __p = (var)[___i]; \
Mauro Carvalho Chehab982216a2012-04-16 13:04:46 -0300505 __p; \
506})
507
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300508struct dimm_info {
Mauro Carvalho Chehab7a623c02012-04-16 16:41:11 -0300509 struct device dev;
510
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300511 char label[EDAC_MC_LABEL_LEN + 1]; /* DIMM label on motherboard */
Mauro Carvalho Chehab4275be62012-04-18 15:20:50 -0300512
513 /* Memory location data */
514 unsigned location[EDAC_MAX_LAYERS];
515
516 struct mem_ctl_info *mci; /* the parent */
Mauro Carvalho Chehab084a4fc2012-01-27 18:38:08 -0300517
518 u32 grain; /* granularity of reported error in bytes */
519 enum dev_type dtype; /* memory device type */
520 enum mem_type mtype; /* memory dimm type */
521 enum edac_type edac_mode; /* EDAC mode for this dimm */
522
Mauro Carvalho Chehab4275be62012-04-18 15:20:50 -0300523 u32 nr_pages; /* number of pages on this dimm */
Mauro Carvalho Chehaba895bf82012-01-28 09:09:38 -0300524
Mauro Carvalho Chehab4275be62012-04-18 15:20:50 -0300525 unsigned csrow, cschannel; /* Points to the old API data */
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300526};
527
Mauro Carvalho Chehaba4b4be32012-01-27 10:26:13 -0300528/**
529 * struct rank_info - contains the information for one DIMM rank
530 *
531 * @chan_idx: channel number where the rank is (typically, 0 or 1)
532 * @ce_count: number of correctable errors for this rank
Mauro Carvalho Chehaba4b4be32012-01-27 10:26:13 -0300533 * @csrow: A pointer to the chip select row structure (the parent
534 * structure). The location of the rank is given by
535 * the (csrow->csrow_idx, chan_idx) vector.
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300536 * @dimm: A pointer to the DIMM structure, where the DIMM label
537 * information is stored.
538 *
539 * FIXME: Currently, the EDAC core model will assume one DIMM per rank.
540 * This is a bad assumption, but it makes this patch easier. Later
541 * patches in this series will fix this issue.
Mauro Carvalho Chehaba4b4be32012-01-27 10:26:13 -0300542 */
543struct rank_info {
544 int chan_idx;
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300545 struct csrow_info *csrow;
546 struct dimm_info *dimm;
Mauro Carvalho Chehab4275be62012-04-18 15:20:50 -0300547
548 u32 ce_count; /* Correctable Errors for this csrow */
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300549};
550
551struct csrow_info {
Mauro Carvalho Chehab7a623c02012-04-16 16:41:11 -0300552 struct device dev;
553
Mauro Carvalho Chehaba895bf82012-01-28 09:09:38 -0300554 /* Used only by edac_mc_find_csrow_by_page() */
Mauro Carvalho Chehab084a4fc2012-01-27 18:38:08 -0300555 unsigned long first_page; /* first page number in csrow */
556 unsigned long last_page; /* last page number in csrow */
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300557 unsigned long page_mask; /* used for interleaving -
Mauro Carvalho Chehaba895bf82012-01-28 09:09:38 -0300558 * 0UL for non intlv */
559
Mauro Carvalho Chehab084a4fc2012-01-27 18:38:08 -0300560 int csrow_idx; /* the chip-select row */
561
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300562 u32 ue_count; /* Uncorrectable Errors for this csrow */
563 u32 ce_count; /* Correctable Errors for this csrow */
Mauro Carvalho Chehab084a4fc2012-01-27 18:38:08 -0300564
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300565 struct mem_ctl_info *mci; /* the parent */
566
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300567 /* channel information for this csrow */
568 u32 nr_channels;
Mauro Carvalho Chehabde3910eb2012-04-24 15:05:43 -0300569 struct rank_info **channels;
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300570};
571
Mauro Carvalho Chehab7a623c02012-04-16 16:41:11 -0300572/*
573 * struct errcount_attribute - used to store the several error counts
574 */
575struct errcount_attribute_data {
576 int n_layers;
577 int pos[EDAC_MAX_LAYERS];
578 int layer0, layer1, layer2;
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300579};
580
Mauro Carvalho Chehabc7ef7642013-02-21 13:36:45 -0300581/**
582 * edac_raw_error_desc - Raw error report structure
583 * @grain: minimum granularity for an error report, in bytes
584 * @error_count: number of errors of the same type
585 * @top_layer: top layer of the error (layer[0])
586 * @mid_layer: middle layer of the error (layer[1])
587 * @low_layer: low layer of the error (layer[2])
588 * @page_frame_number: page where the error happened
589 * @offset_in_page: page offset
590 * @syndrome: syndrome of the error (or 0 if unknown or if
591 * the syndrome is not applicable)
592 * @msg: error message
593 * @location: location of the error
594 * @label: label of the affected DIMM(s)
595 * @other_detail: other driver-specific detail about the error
596 * @enable_per_layer_report: if false, the error affects all layers
597 * (typically, a memory controller error)
598 */
599struct edac_raw_error_desc {
600 /*
601 * NOTE: everything before grain won't be cleaned by
602 * edac_raw_error_desc_clean()
603 */
604 char location[LOCATION_SIZE];
605 char label[(EDAC_MC_LABEL_LEN + 1 + sizeof(OTHER_LABEL)) * EDAC_MAX_LABELS];
606 long grain;
607
608 /* the vars below and grain will be cleaned on every new error report */
609 u16 error_count;
610 int top_layer;
611 int mid_layer;
612 int low_layer;
613 unsigned long page_frame_number;
614 unsigned long offset_in_page;
615 unsigned long syndrome;
616 const char *msg;
617 const char *other_detail;
618 bool enable_per_layer_report;
619};
620
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300621/* MEMORY controller information structure
622 */
623struct mem_ctl_info {
Mauro Carvalho Chehab7a623c02012-04-16 16:41:11 -0300624 struct device dev;
Borislav Petkov88d84ac2013-07-19 12:28:25 +0200625 struct bus_type *bus;
Mauro Carvalho Chehab7a623c02012-04-16 16:41:11 -0300626
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300627 struct list_head link; /* for global list of mem_ctl_info structs */
628
629 struct module *owner; /* Module owner of this control struct */
630
631 unsigned long mtype_cap; /* memory types supported by mc */
632 unsigned long edac_ctl_cap; /* Mem controller EDAC capabilities */
633 unsigned long edac_cap; /* configuration capabilities - this is
634 * closely related to edac_ctl_cap. The
635 * difference is that the controller may be
636 * capable of s4ecd4ed which would be listed
637 * in edac_ctl_cap, but if channels aren't
638 * capable of s4ecd4ed then the edac_cap would
639 * not have that capability.
640 */
641 unsigned long scrub_cap; /* chipset scrub capabilities */
642 enum scrub_type scrub_mode; /* current scrub mode */
643
644 /* Translates sdram memory scrub rate given in bytes/sec to the
645 internal representation and configures whatever else needs
646 to be configured.
647 */
648 int (*set_sdram_scrub_rate) (struct mem_ctl_info * mci, u32 bw);
649
650 /* Get the current sdram memory scrub rate from the internal
651 representation and converts it to the closest matching
652 bandwidth in bytes/sec.
653 */
654 int (*get_sdram_scrub_rate) (struct mem_ctl_info * mci);
655
656
657 /* pointer to edac checking routine */
658 void (*edac_check) (struct mem_ctl_info * mci);
659
660 /*
661 * Remaps memory pages: controller pages to physical pages.
662 * For most MC's, this will be NULL.
663 */
664 /* FIXME - why not send the phys page to begin with? */
665 unsigned long (*ctl_page_to_phys) (struct mem_ctl_info * mci,
666 unsigned long page);
667 int mc_idx;
Mauro Carvalho Chehabde3910eb2012-04-24 15:05:43 -0300668 struct csrow_info **csrows;
Mauro Carvalho Chehab4275be62012-04-18 15:20:50 -0300669 unsigned nr_csrows, num_cschannel;
670
Mauro Carvalho Chehab7a623c02012-04-16 16:41:11 -0300671 /*
672 * Memory Controller hierarchy
673 *
674 * There are basically two types of memory controller: the ones that
675 * sees memory sticks ("dimms"), and the ones that sees memory ranks.
676 * All old memory controllers enumerate memories per rank, but most
677 * of the recent drivers enumerate memories per DIMM, instead.
Mauro Carvalho Chehab9713fae2013-03-11 09:28:48 -0300678 * When the memory controller is per rank, csbased is true.
Mauro Carvalho Chehab7a623c02012-04-16 16:41:11 -0300679 */
Mauro Carvalho Chehab4275be62012-04-18 15:20:50 -0300680 unsigned n_layers;
681 struct edac_mc_layer *layers;
Mauro Carvalho Chehab9713fae2013-03-11 09:28:48 -0300682 bool csbased;
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300683
684 /*
685 * DIMM info. Will eventually remove the entire csrows_info some day
686 */
Mauro Carvalho Chehab4275be62012-04-18 15:20:50 -0300687 unsigned tot_dimms;
Mauro Carvalho Chehabde3910eb2012-04-24 15:05:43 -0300688 struct dimm_info **dimms;
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300689
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300690 /*
691 * FIXME - what about controllers on other busses? - IDs must be
692 * unique. dev pointer should be sufficiently unique, but
693 * BUS:SLOT.FUNC numbers may not be unique.
694 */
Mauro Carvalho Chehabfd687502012-03-16 07:44:18 -0300695 struct device *pdev;
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300696 const char *mod_name;
697 const char *mod_ver;
698 const char *ctl_name;
699 const char *dev_name;
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300700 void *pvt_info;
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300701 unsigned long start_time; /* mci load start time (in jiffies) */
702
Mauro Carvalho Chehab4275be62012-04-18 15:20:50 -0300703 /*
704 * drivers shouldn't access those fields directly, as the core
705 * already handles that.
706 */
707 u32 ce_noinfo_count, ue_noinfo_count;
Mauro Carvalho Chehab5926ff52012-02-09 11:05:20 -0300708 u32 ue_mc, ce_mc;
Mauro Carvalho Chehab4275be62012-04-18 15:20:50 -0300709 u32 *ce_per_layer[EDAC_MAX_LAYERS], *ue_per_layer[EDAC_MAX_LAYERS];
710
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300711 struct completion complete;
712
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300713 /* Additional top controller level attributes, but specified
714 * by the low level driver.
715 *
716 * Set by the low level driver to provide attributes at the
Mauro Carvalho Chehab4275be62012-04-18 15:20:50 -0300717 * controller level.
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300718 * An array of structures, NULL terminated
719 *
720 * If attributes are desired, then set to array of attributes
721 * If no attributes are desired, leave NULL
722 */
723 const struct mcidev_sysfs_attribute *mc_driver_sysfs_attributes;
724
725 /* work struct for this MC */
726 struct delayed_work work;
727
Mauro Carvalho Chehabc7ef7642013-02-21 13:36:45 -0300728 /*
729 * Used to report an error - by being at the global struct
730 * makes the memory allocated by the EDAC core
731 */
732 struct edac_raw_error_desc error_desc;
733
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300734 /* the internal state of this controller instance */
735 int op_state;
Mauro Carvalho Chehab452a6bf2012-03-26 09:35:11 -0300736
737#ifdef CONFIG_EDAC_DEBUG
738 struct dentry *debugfs;
739 u8 fake_inject_layer[EDAC_MAX_LAYERS];
740 u32 fake_inject_ue;
Mauro Carvalho Chehab38ced282012-06-12 10:55:57 -0300741 u16 fake_inject_count;
Mauro Carvalho Chehab452a6bf2012-03-26 09:35:11 -0300742#endif
Mauro Carvalho Chehabddeb3542011-03-04 15:11:29 -0300743};
744
Borislav Petkov88d84ac2013-07-19 12:28:25 +0200745/*
746 * Maximum number of memory controllers in the coherent fabric.
747 */
748#define EDAC_MAX_MCS 16
749
Dave Jiangc0d12172007-07-19 01:49:46 -0700750#endif