aboutsummaryrefslogtreecommitdiffstats
path: root/drivers/edac/edac_mc.h
diff options
context:
space:
mode:
authorDouglas Thompson <dougthompson@xmission.com>2007-07-19 04:49:33 -0400
committerLinus Torvalds <torvalds@woody.linux-foundation.org>2007-07-19 13:04:53 -0400
commit7c9281d76c1c0b130f79d5fc021084e9749959d4 (patch)
tree8e54412e8dc529e8bf755633ebe71e35183353d0 /drivers/edac/edac_mc.h
parentd56933e018b14fc7cad322f413eecc6cb6edf12e (diff)
drivers/edac: split out functions to unique files
This is a large patch to refactor the original EDAC module in the kernel and to break it up into better file granularity, such that each source file contains a given subsystem of the EDAC CORE. Originally, the EDAC 'core' was contained in one source file: edac_mc.c with it corresponding edac_mc.h file. Now, there are the following files: edac_module.c The main module init/exit function and other overhead edac_mc.c Code handling the edac_mc class of object edac_mc_sysfs.c Code handling for sysfs presentation edac_pci_sysfs.c Code handling for PCI sysfs presentation edac_core.h CORE .h include file for 'edac_mc' and 'edac_device' drivers edac_module.h Internal CORE .h include file This forms a foundation upon which a later patch can create the 'edac_device' class of object code in a new file 'edac_device.c'. Signed-off-by: Douglas Thompson <dougthompson@xmission.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'drivers/edac/edac_mc.h')
-rw-r--r--drivers/edac/edac_mc.h471
1 files changed, 2 insertions, 469 deletions
diff --git a/drivers/edac/edac_mc.h b/drivers/edac/edac_mc.h
index fdc811d89679..b92d2720a4de 100644
--- a/drivers/edac/edac_mc.h
+++ b/drivers/edac/edac_mc.h
@@ -1,476 +1,9 @@
1/*
2 * MC kernel module
3 * (C) 2003 Linux Networx (http://lnxi.com)
4 * This file may be distributed under the terms of the
5 * GNU General Public License.
6 *
7 * Written by Thayne Harbaugh
8 * Based on work by Dan Hollis <goemon at anime dot net> and others.
9 * http://www.anime.net/~goemon/linux-ecc/
10 *
11 * NMI handling support added by
12 * Dave Peterson <dsp@llnl.gov> <dave_peterson@pobox.com>
13 *
14 * $Id: edac_mc.h,v 1.4.2.10 2005/10/05 00:43:44 dsp_llnl Exp $
15 *
16 */
17
18#ifndef _EDAC_MC_H_
19#define _EDAC_MC_H_
20
21#include <linux/kernel.h>
22#include <linux/types.h>
23#include <linux/module.h>
24#include <linux/spinlock.h>
25#include <linux/smp.h>
26#include <linux/pci.h>
27#include <linux/time.h>
28#include <linux/nmi.h>
29#include <linux/rcupdate.h>
30#include <linux/completion.h>
31#include <linux/kobject.h>
32#include <linux/platform_device.h>
33
34#define EDAC_MC_LABEL_LEN 31
35#define MC_PROC_NAME_MAX_LEN 7
36
37#if PAGE_SHIFT < 20
38#define PAGES_TO_MiB( pages ) ( ( pages ) >> ( 20 - PAGE_SHIFT ) )
39#else /* PAGE_SHIFT > 20 */
40#define PAGES_TO_MiB( pages ) ( ( pages ) << ( PAGE_SHIFT - 20 ) )
41#endif
42
43#define edac_printk(level, prefix, fmt, arg...) \
44 printk(level "EDAC " prefix ": " fmt, ##arg)
45
46#define edac_mc_printk(mci, level, fmt, arg...) \
47 printk(level "EDAC MC%d: " fmt, mci->mc_idx, ##arg)
48
49#define edac_mc_chipset_printk(mci, level, prefix, fmt, arg...) \
50 printk(level "EDAC " prefix " MC%d: " fmt, mci->mc_idx, ##arg)
51
52/* prefixes for edac_printk() and edac_mc_printk() */
53#define EDAC_MC "MC"
54#define EDAC_PCI "PCI"
55#define EDAC_DEBUG "DEBUG"
56
57#ifdef CONFIG_EDAC_DEBUG
58extern int edac_debug_level;
59
60#define edac_debug_printk(level, fmt, arg...) \
61 do { \
62 if (level <= edac_debug_level) \
63 edac_printk(KERN_DEBUG, EDAC_DEBUG, fmt, ##arg); \
64 } while(0)
65
66#define debugf0( ... ) edac_debug_printk(0, __VA_ARGS__ )
67#define debugf1( ... ) edac_debug_printk(1, __VA_ARGS__ )
68#define debugf2( ... ) edac_debug_printk(2, __VA_ARGS__ )
69#define debugf3( ... ) edac_debug_printk(3, __VA_ARGS__ )
70#define debugf4( ... ) edac_debug_printk(4, __VA_ARGS__ )
71
72#else /* !CONFIG_EDAC_DEBUG */
73
74#define debugf0( ... )
75#define debugf1( ... )
76#define debugf2( ... )
77#define debugf3( ... )
78#define debugf4( ... )
79
80#endif /* !CONFIG_EDAC_DEBUG */
81 1
82#define BIT(x) (1 << (x))
83
84#define PCI_VEND_DEV(vend, dev) PCI_VENDOR_ID_ ## vend, \
85 PCI_DEVICE_ID_ ## vend ## _ ## dev
86
87#if defined(CONFIG_X86) && defined(CONFIG_PCI)
88#define dev_name(dev) pci_name(to_pci_dev(dev))
89#else
90#define dev_name(dev) to_platform_device(dev)->name
91#endif
92
93/* memory devices */
94enum dev_type {
95 DEV_UNKNOWN = 0,
96 DEV_X1,
97 DEV_X2,
98 DEV_X4,
99 DEV_X8,
100 DEV_X16,
101 DEV_X32, /* Do these parts exist? */
102 DEV_X64 /* Do these parts exist? */
103};
104
105#define DEV_FLAG_UNKNOWN BIT(DEV_UNKNOWN)
106#define DEV_FLAG_X1 BIT(DEV_X1)
107#define DEV_FLAG_X2 BIT(DEV_X2)
108#define DEV_FLAG_X4 BIT(DEV_X4)
109#define DEV_FLAG_X8 BIT(DEV_X8)
110#define DEV_FLAG_X16 BIT(DEV_X16)
111#define DEV_FLAG_X32 BIT(DEV_X32)
112#define DEV_FLAG_X64 BIT(DEV_X64)
113
114/* memory types */
115enum mem_type {
116 MEM_EMPTY = 0, /* Empty csrow */
117 MEM_RESERVED, /* Reserved csrow type */
118 MEM_UNKNOWN, /* Unknown csrow type */
119 MEM_FPM, /* Fast page mode */
120 MEM_EDO, /* Extended data out */
121 MEM_BEDO, /* Burst Extended data out */
122 MEM_SDR, /* Single data rate SDRAM */
123 MEM_RDR, /* Registered single data rate SDRAM */
124 MEM_DDR, /* Double data rate SDRAM */
125 MEM_RDDR, /* Registered Double data rate SDRAM */
126 MEM_RMBS, /* Rambus DRAM */
127 MEM_DDR2, /* DDR2 RAM */
128 MEM_FB_DDR2, /* fully buffered DDR2 */
129 MEM_RDDR2, /* Registered DDR2 RAM */
130};
131
132#define MEM_FLAG_EMPTY BIT(MEM_EMPTY)
133#define MEM_FLAG_RESERVED BIT(MEM_RESERVED)
134#define MEM_FLAG_UNKNOWN BIT(MEM_UNKNOWN)
135#define MEM_FLAG_FPM BIT(MEM_FPM)
136#define MEM_FLAG_EDO BIT(MEM_EDO)
137#define MEM_FLAG_BEDO BIT(MEM_BEDO)
138#define MEM_FLAG_SDR BIT(MEM_SDR)
139#define MEM_FLAG_RDR BIT(MEM_RDR)
140#define MEM_FLAG_DDR BIT(MEM_DDR)
141#define MEM_FLAG_RDDR BIT(MEM_RDDR)
142#define MEM_FLAG_RMBS BIT(MEM_RMBS)
143#define MEM_FLAG_DDR2 BIT(MEM_DDR2)
144#define MEM_FLAG_FB_DDR2 BIT(MEM_FB_DDR2)
145#define MEM_FLAG_RDDR2 BIT(MEM_RDDR2)
146
147/* chipset Error Detection and Correction capabilities and mode */
148enum edac_type {
149 EDAC_UNKNOWN = 0, /* Unknown if ECC is available */
150 EDAC_NONE, /* Doesnt support ECC */
151 EDAC_RESERVED, /* Reserved ECC type */
152 EDAC_PARITY, /* Detects parity errors */
153 EDAC_EC, /* Error Checking - no correction */
154 EDAC_SECDED, /* Single bit error correction, Double detection */
155 EDAC_S2ECD2ED, /* Chipkill x2 devices - do these exist? */
156 EDAC_S4ECD4ED, /* Chipkill x4 devices */
157 EDAC_S8ECD8ED, /* Chipkill x8 devices */
158 EDAC_S16ECD16ED, /* Chipkill x16 devices */
159};
160
161#define EDAC_FLAG_UNKNOWN BIT(EDAC_UNKNOWN)
162#define EDAC_FLAG_NONE BIT(EDAC_NONE)
163#define EDAC_FLAG_PARITY BIT(EDAC_PARITY)
164#define EDAC_FLAG_EC BIT(EDAC_EC)
165#define EDAC_FLAG_SECDED BIT(EDAC_SECDED)
166#define EDAC_FLAG_S2ECD2ED BIT(EDAC_S2ECD2ED)
167#define EDAC_FLAG_S4ECD4ED BIT(EDAC_S4ECD4ED)
168#define EDAC_FLAG_S8ECD8ED BIT(EDAC_S8ECD8ED)
169#define EDAC_FLAG_S16ECD16ED BIT(EDAC_S16ECD16ED)
170
171/* scrubbing capabilities */
172enum scrub_type {
173 SCRUB_UNKNOWN = 0, /* Unknown if scrubber is available */
174 SCRUB_NONE, /* No scrubber */
175 SCRUB_SW_PROG, /* SW progressive (sequential) scrubbing */
176 SCRUB_SW_SRC, /* Software scrub only errors */
177 SCRUB_SW_PROG_SRC, /* Progressive software scrub from an error */
178 SCRUB_SW_TUNABLE, /* Software scrub frequency is tunable */
179 SCRUB_HW_PROG, /* HW progressive (sequential) scrubbing */
180 SCRUB_HW_SRC, /* Hardware scrub only errors */
181 SCRUB_HW_PROG_SRC, /* Progressive hardware scrub from an error */
182 SCRUB_HW_TUNABLE /* Hardware scrub frequency is tunable */
183};
184
185#define SCRUB_FLAG_SW_PROG BIT(SCRUB_SW_PROG)
186#define SCRUB_FLAG_SW_SRC BIT(SCRUB_SW_SRC_CORR)
187#define SCRUB_FLAG_SW_PROG_SRC BIT(SCRUB_SW_PROG_SRC_CORR)
188#define SCRUB_FLAG_SW_TUN BIT(SCRUB_SW_SCRUB_TUNABLE)
189#define SCRUB_FLAG_HW_PROG BIT(SCRUB_HW_PROG)
190#define SCRUB_FLAG_HW_SRC BIT(SCRUB_HW_SRC_CORR)
191#define SCRUB_FLAG_HW_PROG_SRC BIT(SCRUB_HW_PROG_SRC_CORR)
192#define SCRUB_FLAG_HW_TUN BIT(SCRUB_HW_TUNABLE)
193
194/* FIXME - should have notify capabilities: NMI, LOG, PROC, etc */
195 2
196/* 3/*
197 * There are several things to be aware of that aren't at all obvious: 4 * Older .h file for edac, until all drivers are modified
198 *
199 *
200 * SOCKETS, SOCKET SETS, BANKS, ROWS, CHIP-SELECT ROWS, CHANNELS, etc..
201 *
202 * These are some of the many terms that are thrown about that don't always
203 * mean what people think they mean (Inconceivable!). In the interest of
204 * creating a common ground for discussion, terms and their definitions
205 * will be established.
206 *
207 * Memory devices: The individual chip on a memory stick. These devices
208 * commonly output 4 and 8 bits each. Grouping several
209 * of these in parallel provides 64 bits which is common
210 * for a memory stick.
211 *
212 * Memory Stick: A printed circuit board that agregates multiple
213 * memory devices in parallel. This is the atomic
214 * memory component that is purchaseable by Joe consumer
215 * and loaded into a memory socket.
216 *
217 * Socket: A physical connector on the motherboard that accepts
218 * a single memory stick.
219 *
220 * Channel: Set of memory devices on a memory stick that must be
221 * grouped in parallel with one or more additional
222 * channels from other memory sticks. This parallel
223 * grouping of the output from multiple channels are
224 * necessary for the smallest granularity of memory access.
225 * Some memory controllers are capable of single channel -
226 * which means that memory sticks can be loaded
227 * individually. Other memory controllers are only
228 * capable of dual channel - which means that memory
229 * sticks must be loaded as pairs (see "socket set").
230 *
231 * Chip-select row: All of the memory devices that are selected together.
232 * for a single, minimum grain of memory access.
233 * This selects all of the parallel memory devices across
234 * all of the parallel channels. Common chip-select rows
235 * for single channel are 64 bits, for dual channel 128
236 * bits.
237 *
238 * Single-Ranked stick: A Single-ranked stick has 1 chip-select row of memmory.
239 * Motherboards commonly drive two chip-select pins to
240 * a memory stick. A single-ranked stick, will occupy
241 * only one of those rows. The other will be unused.
242 * 5 *
243 * Double-Ranked stick: A double-ranked stick has two chip-select rows which
244 * access different sets of memory devices. The two
245 * rows cannot be accessed concurrently.
246 *
247 * Double-sided stick: DEPRECATED TERM, see Double-Ranked stick.
248 * A double-sided stick has two chip-select rows which
249 * access different sets of memory devices. The two
250 * rows cannot be accessed concurrently. "Double-sided"
251 * is irrespective of the memory devices being mounted
252 * on both sides of the memory stick.
253 *
254 * Socket set: All of the memory sticks that are required for for
255 * a single memory access or all of the memory sticks
256 * spanned by a chip-select row. A single socket set
257 * has two chip-select rows and if double-sided sticks
258 * are used these will occupy those chip-select rows.
259 *
260 * Bank: This term is avoided because it is unclear when
261 * needing to distinguish between chip-select rows and
262 * socket sets.
263 *
264 * Controller pages:
265 *
266 * Physical pages:
267 *
268 * Virtual pages:
269 *
270 *
271 * STRUCTURE ORGANIZATION AND CHOICES
272 *
273 *
274 *
275 * PS - I enjoyed writing all that about as much as you enjoyed reading it.
276 */
277
278struct channel_info {
279 int chan_idx; /* channel index */
280 u32 ce_count; /* Correctable Errors for this CHANNEL */
281 char label[EDAC_MC_LABEL_LEN + 1]; /* DIMM label on motherboard */
282 struct csrow_info *csrow; /* the parent */
283};
284
285struct csrow_info {
286 unsigned long first_page; /* first page number in dimm */
287 unsigned long last_page; /* last page number in dimm */
288 unsigned long page_mask; /* used for interleaving -
289 * 0UL for non intlv
290 */
291 u32 nr_pages; /* number of pages in csrow */
292 u32 grain; /* granularity of reported error in bytes */
293 int csrow_idx; /* the chip-select row */
294 enum dev_type dtype; /* memory device type */
295 u32 ue_count; /* Uncorrectable Errors for this csrow */
296 u32 ce_count; /* Correctable Errors for this csrow */
297 enum mem_type mtype; /* memory csrow type */
298 enum edac_type edac_mode; /* EDAC mode for this csrow */
299 struct mem_ctl_info *mci; /* the parent */
300
301 struct kobject kobj; /* sysfs kobject for this csrow */
302 struct completion kobj_complete;
303
304 /* FIXME the number of CHANNELs might need to become dynamic */
305 u32 nr_channels;
306 struct channel_info *channels;
307};
308
309struct mem_ctl_info {
310 struct list_head link; /* for global list of mem_ctl_info structs */
311 unsigned long mtype_cap; /* memory types supported by mc */
312 unsigned long edac_ctl_cap; /* Mem controller EDAC capabilities */
313 unsigned long edac_cap; /* configuration capabilities - this is
314 * closely related to edac_ctl_cap. The
315 * difference is that the controller may be
316 * capable of s4ecd4ed which would be listed
317 * in edac_ctl_cap, but if channels aren't
318 * capable of s4ecd4ed then the edac_cap would
319 * not have that capability.
320 */
321 unsigned long scrub_cap; /* chipset scrub capabilities */
322 enum scrub_type scrub_mode; /* current scrub mode */
323
324 /* Translates sdram memory scrub rate given in bytes/sec to the
325 internal representation and configures whatever else needs
326 to be configured.
327 */
328 int (*set_sdram_scrub_rate) (struct mem_ctl_info *mci, u32 *bw);
329
330 /* Get the current sdram memory scrub rate from the internal
331 representation and converts it to the closest matching
332 bandwith in bytes/sec.
333 */
334 int (*get_sdram_scrub_rate) (struct mem_ctl_info *mci, u32 *bw);
335
336 /* pointer to edac checking routine */
337 void (*edac_check) (struct mem_ctl_info * mci);
338
339 /*
340 * Remaps memory pages: controller pages to physical pages.
341 * For most MC's, this will be NULL.
342 */
343 /* FIXME - why not send the phys page to begin with? */
344 unsigned long (*ctl_page_to_phys) (struct mem_ctl_info * mci,
345 unsigned long page);
346 int mc_idx;
347 int nr_csrows;
348 struct csrow_info *csrows;
349 /*
350 * FIXME - what about controllers on other busses? - IDs must be
351 * unique. dev pointer should be sufficiently unique, but
352 * BUS:SLOT.FUNC numbers may not be unique.
353 */
354 struct device *dev;
355 const char *mod_name;
356 const char *mod_ver;
357 const char *ctl_name;
358 char proc_name[MC_PROC_NAME_MAX_LEN + 1];
359 void *pvt_info;
360 u32 ue_noinfo_count; /* Uncorrectable Errors w/o info */
361 u32 ce_noinfo_count; /* Correctable Errors w/o info */
362 u32 ue_count; /* Total Uncorrectable Errors for this MC */
363 u32 ce_count; /* Total Correctable Errors for this MC */
364 unsigned long start_time; /* mci load start time (in jiffies) */
365
366 /* this stuff is for safe removal of mc devices from global list while
367 * NMI handlers may be traversing list
368 */
369 struct rcu_head rcu;
370 struct completion complete;
371
372 /* edac sysfs device control */
373 struct kobject edac_mci_kobj;
374 struct completion kobj_complete;
375};
376
377#ifdef CONFIG_PCI
378
379/* write all or some bits in a byte-register*/
380static inline void pci_write_bits8(struct pci_dev *pdev, int offset, u8 value,
381 u8 mask)
382{
383 if (mask != 0xff) {
384 u8 buf;
385
386 pci_read_config_byte(pdev, offset, &buf);
387 value &= mask;
388 buf &= ~mask;
389 value |= buf;
390 }
391
392 pci_write_config_byte(pdev, offset, value);
393}
394
395/* write all or some bits in a word-register*/
396static inline void pci_write_bits16(struct pci_dev *pdev, int offset,
397 u16 value, u16 mask)
398{
399 if (mask != 0xffff) {
400 u16 buf;
401
402 pci_read_config_word(pdev, offset, &buf);
403 value &= mask;
404 buf &= ~mask;
405 value |= buf;
406 }
407
408 pci_write_config_word(pdev, offset, value);
409}
410
411/* write all or some bits in a dword-register*/
412static inline void pci_write_bits32(struct pci_dev *pdev, int offset,
413 u32 value, u32 mask)
414{
415 if (mask != 0xffff) {
416 u32 buf;
417
418 pci_read_config_dword(pdev, offset, &buf);
419 value &= mask;
420 buf &= ~mask;
421 value |= buf;
422 }
423
424 pci_write_config_dword(pdev, offset, value);
425}
426
427#endif /* CONFIG_PCI */
428
429extern struct mem_ctl_info * edac_mc_find(int idx);
430extern int edac_mc_add_mc(struct mem_ctl_info *mci,int mc_idx);
431extern struct mem_ctl_info * edac_mc_del_mc(struct device *dev);
432extern int edac_mc_find_csrow_by_page(struct mem_ctl_info *mci,
433 unsigned long page);
434
435/*
436 * The no info errors are used when error overflows are reported.
437 * There are a limited number of error logging registers that can
438 * be exausted. When all registers are exhausted and an additional
439 * error occurs then an error overflow register records that an
440 * error occured and the type of error, but doesn't have any
441 * further information. The ce/ue versions make for cleaner
442 * reporting logic and function interface - reduces conditional
443 * statement clutter and extra function arguments.
444 */
445extern void edac_mc_handle_ce(struct mem_ctl_info *mci,
446 unsigned long page_frame_number, unsigned long offset_in_page,
447 unsigned long syndrome, int row, int channel,
448 const char *msg);
449extern void edac_mc_handle_ce_no_info(struct mem_ctl_info *mci,
450 const char *msg);
451extern void edac_mc_handle_ue(struct mem_ctl_info *mci,
452 unsigned long page_frame_number, unsigned long offset_in_page,
453 int row, const char *msg);
454extern void edac_mc_handle_ue_no_info(struct mem_ctl_info *mci,
455 const char *msg);
456extern void edac_mc_handle_fbd_ue(struct mem_ctl_info *mci,
457 unsigned int csrow,
458 unsigned int channel0,
459 unsigned int channel1,
460 char *msg);
461extern void edac_mc_handle_fbd_ce(struct mem_ctl_info *mci,
462 unsigned int csrow,
463 unsigned int channel,
464 char *msg);
465
466/*
467 * This kmalloc's and initializes all the structures.
468 * Can't be used if all structures don't have the same lifetime.
469 */ 6 */
470extern struct mem_ctl_info *edac_mc_alloc(unsigned sz_pvt, unsigned nr_csrows,
471 unsigned nr_chans);
472 7
473/* Free an mc previously allocated by edac_mc_alloc() */ 8#include "edac_core.h"
474extern void edac_mc_free(struct mem_ctl_info *mci);
475 9
476#endif /* _EDAC_MC_H_ */