]>
Commit | Line | Data |
---|---|---|
7c9281d7 DT |
1 | /* |
2 | * Defines, structures, APIs for edac_core module | |
3 | * | |
4 | * (C) 2007 Linux Networx (http://lnxi.com) | |
5 | * This file may be distributed under the terms of the | |
6 | * GNU General Public License. | |
7 | * | |
8 | * Written by Thayne Harbaugh | |
9 | * Based on work by Dan Hollis <goemon at anime dot net> and others. | |
10 | * http://www.anime.net/~goemon/linux-ecc/ | |
11 | * | |
12 | * NMI handling support added by | |
13 | * Dave Peterson <[email protected]> <[email protected]> | |
14 | * | |
15 | * Refactored for multi-source files: | |
16 | * Doug Thompson <[email protected]> | |
17 | * | |
18 | */ | |
19 | ||
20 | #ifndef _EDAC_CORE_H_ | |
21 | #define _EDAC_CORE_H_ | |
22 | ||
23 | #include <linux/kernel.h> | |
24 | #include <linux/types.h> | |
25 | #include <linux/module.h> | |
26 | #include <linux/spinlock.h> | |
27 | #include <linux/smp.h> | |
28 | #include <linux/pci.h> | |
29 | #include <linux/time.h> | |
30 | #include <linux/nmi.h> | |
31 | #include <linux/rcupdate.h> | |
32 | #include <linux/completion.h> | |
33 | #include <linux/kobject.h> | |
34 | #include <linux/platform_device.h> | |
e27e3dac DT |
35 | #include <linux/sysdev.h> |
36 | #include <linux/workqueue.h> | |
37 | #include <linux/version.h> | |
7c9281d7 DT |
38 | |
39 | #define EDAC_MC_LABEL_LEN 31 | |
e27e3dac DT |
40 | #define EDAC_DEVICE_NAME_LEN 31 |
41 | #define EDAC_ATTRIB_VALUE_LEN 15 | |
42 | #define MC_PROC_NAME_MAX_LEN 7 | |
7c9281d7 DT |
43 | |
44 | #if PAGE_SHIFT < 20 | |
45 | #define PAGES_TO_MiB( pages ) ( ( pages ) >> ( 20 - PAGE_SHIFT ) ) | |
46 | #else /* PAGE_SHIFT > 20 */ | |
47 | #define PAGES_TO_MiB( pages ) ( ( pages ) << ( PAGE_SHIFT - 20 ) ) | |
48 | #endif | |
49 | ||
50 | #define edac_printk(level, prefix, fmt, arg...) \ | |
51 | printk(level "EDAC " prefix ": " fmt, ##arg) | |
52 | ||
53 | #define edac_mc_printk(mci, level, fmt, arg...) \ | |
54 | printk(level "EDAC MC%d: " fmt, mci->mc_idx, ##arg) | |
55 | ||
56 | #define edac_mc_chipset_printk(mci, level, prefix, fmt, arg...) \ | |
57 | printk(level "EDAC " prefix " MC%d: " fmt, mci->mc_idx, ##arg) | |
58 | ||
e27e3dac DT |
59 | /* edac_device printk */ |
60 | #define edac_device_printk(ctl, level, fmt, arg...) \ | |
61 | printk(level "EDAC DEVICE%d: " fmt, ctl->dev_idx, ##arg) | |
62 | ||
7c9281d7 DT |
63 | /* prefixes for edac_printk() and edac_mc_printk() */ |
64 | #define EDAC_MC "MC" | |
65 | #define EDAC_PCI "PCI" | |
66 | #define EDAC_DEBUG "DEBUG" | |
67 | ||
68 | #ifdef CONFIG_EDAC_DEBUG | |
69 | extern int edac_debug_level; | |
70 | ||
71 | #define edac_debug_printk(level, fmt, arg...) \ | |
72 | do { \ | |
73 | if (level <= edac_debug_level) \ | |
e27e3dac | 74 | edac_printk(KERN_EMERG, EDAC_DEBUG, fmt, ##arg); \ |
7c9281d7 DT |
75 | } while(0) |
76 | ||
77 | #define debugf0( ... ) edac_debug_printk(0, __VA_ARGS__ ) | |
78 | #define debugf1( ... ) edac_debug_printk(1, __VA_ARGS__ ) | |
79 | #define debugf2( ... ) edac_debug_printk(2, __VA_ARGS__ ) | |
80 | #define debugf3( ... ) edac_debug_printk(3, __VA_ARGS__ ) | |
81 | #define debugf4( ... ) edac_debug_printk(4, __VA_ARGS__ ) | |
82 | ||
83 | #else /* !CONFIG_EDAC_DEBUG */ | |
84 | ||
85 | #define debugf0( ... ) | |
86 | #define debugf1( ... ) | |
87 | #define debugf2( ... ) | |
88 | #define debugf3( ... ) | |
89 | #define debugf4( ... ) | |
90 | ||
91 | #endif /* !CONFIG_EDAC_DEBUG */ | |
92 | ||
93 | #define BIT(x) (1 << (x)) | |
94 | ||
95 | #define PCI_VEND_DEV(vend, dev) PCI_VENDOR_ID_ ## vend, \ | |
96 | PCI_DEVICE_ID_ ## vend ## _ ## dev | |
97 | ||
c4192705 | 98 | #define dev_name(dev) (dev)->dev_name |
7c9281d7 DT |
99 | |
100 | /* memory devices */ | |
101 | enum dev_type { | |
102 | DEV_UNKNOWN = 0, | |
103 | DEV_X1, | |
104 | DEV_X2, | |
105 | DEV_X4, | |
106 | DEV_X8, | |
107 | DEV_X16, | |
108 | DEV_X32, /* Do these parts exist? */ | |
109 | DEV_X64 /* Do these parts exist? */ | |
110 | }; | |
111 | ||
112 | #define DEV_FLAG_UNKNOWN BIT(DEV_UNKNOWN) | |
113 | #define DEV_FLAG_X1 BIT(DEV_X1) | |
114 | #define DEV_FLAG_X2 BIT(DEV_X2) | |
115 | #define DEV_FLAG_X4 BIT(DEV_X4) | |
116 | #define DEV_FLAG_X8 BIT(DEV_X8) | |
117 | #define DEV_FLAG_X16 BIT(DEV_X16) | |
118 | #define DEV_FLAG_X32 BIT(DEV_X32) | |
119 | #define DEV_FLAG_X64 BIT(DEV_X64) | |
120 | ||
121 | /* memory types */ | |
122 | enum mem_type { | |
123 | MEM_EMPTY = 0, /* Empty csrow */ | |
124 | MEM_RESERVED, /* Reserved csrow type */ | |
125 | MEM_UNKNOWN, /* Unknown csrow type */ | |
126 | MEM_FPM, /* Fast page mode */ | |
127 | MEM_EDO, /* Extended data out */ | |
128 | MEM_BEDO, /* Burst Extended data out */ | |
129 | MEM_SDR, /* Single data rate SDRAM */ | |
130 | MEM_RDR, /* Registered single data rate SDRAM */ | |
131 | MEM_DDR, /* Double data rate SDRAM */ | |
132 | MEM_RDDR, /* Registered Double data rate SDRAM */ | |
133 | MEM_RMBS, /* Rambus DRAM */ | |
134 | MEM_DDR2, /* DDR2 RAM */ | |
135 | MEM_FB_DDR2, /* fully buffered DDR2 */ | |
136 | MEM_RDDR2, /* Registered DDR2 RAM */ | |
137 | }; | |
138 | ||
139 | #define MEM_FLAG_EMPTY BIT(MEM_EMPTY) | |
140 | #define MEM_FLAG_RESERVED BIT(MEM_RESERVED) | |
141 | #define MEM_FLAG_UNKNOWN BIT(MEM_UNKNOWN) | |
142 | #define MEM_FLAG_FPM BIT(MEM_FPM) | |
143 | #define MEM_FLAG_EDO BIT(MEM_EDO) | |
144 | #define MEM_FLAG_BEDO BIT(MEM_BEDO) | |
145 | #define MEM_FLAG_SDR BIT(MEM_SDR) | |
146 | #define MEM_FLAG_RDR BIT(MEM_RDR) | |
147 | #define MEM_FLAG_DDR BIT(MEM_DDR) | |
148 | #define MEM_FLAG_RDDR BIT(MEM_RDDR) | |
149 | #define MEM_FLAG_RMBS BIT(MEM_RMBS) | |
150 | #define MEM_FLAG_DDR2 BIT(MEM_DDR2) | |
151 | #define MEM_FLAG_FB_DDR2 BIT(MEM_FB_DDR2) | |
152 | #define MEM_FLAG_RDDR2 BIT(MEM_RDDR2) | |
153 | ||
154 | /* chipset Error Detection and Correction capabilities and mode */ | |
155 | enum edac_type { | |
156 | EDAC_UNKNOWN = 0, /* Unknown if ECC is available */ | |
157 | EDAC_NONE, /* Doesnt support ECC */ | |
158 | EDAC_RESERVED, /* Reserved ECC type */ | |
159 | EDAC_PARITY, /* Detects parity errors */ | |
160 | EDAC_EC, /* Error Checking - no correction */ | |
161 | EDAC_SECDED, /* Single bit error correction, Double detection */ | |
162 | EDAC_S2ECD2ED, /* Chipkill x2 devices - do these exist? */ | |
163 | EDAC_S4ECD4ED, /* Chipkill x4 devices */ | |
164 | EDAC_S8ECD8ED, /* Chipkill x8 devices */ | |
165 | EDAC_S16ECD16ED, /* Chipkill x16 devices */ | |
166 | }; | |
167 | ||
168 | #define EDAC_FLAG_UNKNOWN BIT(EDAC_UNKNOWN) | |
169 | #define EDAC_FLAG_NONE BIT(EDAC_NONE) | |
170 | #define EDAC_FLAG_PARITY BIT(EDAC_PARITY) | |
171 | #define EDAC_FLAG_EC BIT(EDAC_EC) | |
172 | #define EDAC_FLAG_SECDED BIT(EDAC_SECDED) | |
173 | #define EDAC_FLAG_S2ECD2ED BIT(EDAC_S2ECD2ED) | |
174 | #define EDAC_FLAG_S4ECD4ED BIT(EDAC_S4ECD4ED) | |
175 | #define EDAC_FLAG_S8ECD8ED BIT(EDAC_S8ECD8ED) | |
176 | #define EDAC_FLAG_S16ECD16ED BIT(EDAC_S16ECD16ED) | |
177 | ||
178 | /* scrubbing capabilities */ | |
179 | enum scrub_type { | |
180 | SCRUB_UNKNOWN = 0, /* Unknown if scrubber is available */ | |
181 | SCRUB_NONE, /* No scrubber */ | |
182 | SCRUB_SW_PROG, /* SW progressive (sequential) scrubbing */ | |
183 | SCRUB_SW_SRC, /* Software scrub only errors */ | |
184 | SCRUB_SW_PROG_SRC, /* Progressive software scrub from an error */ | |
185 | SCRUB_SW_TUNABLE, /* Software scrub frequency is tunable */ | |
186 | SCRUB_HW_PROG, /* HW progressive (sequential) scrubbing */ | |
187 | SCRUB_HW_SRC, /* Hardware scrub only errors */ | |
188 | SCRUB_HW_PROG_SRC, /* Progressive hardware scrub from an error */ | |
189 | SCRUB_HW_TUNABLE /* Hardware scrub frequency is tunable */ | |
190 | }; | |
191 | ||
192 | #define SCRUB_FLAG_SW_PROG BIT(SCRUB_SW_PROG) | |
522a94bd DT |
193 | #define SCRUB_FLAG_SW_SRC BIT(SCRUB_SW_SRC) |
194 | #define SCRUB_FLAG_SW_PROG_SRC BIT(SCRUB_SW_PROG_SRC) | |
7c9281d7 DT |
195 | #define SCRUB_FLAG_SW_TUN BIT(SCRUB_SW_SCRUB_TUNABLE) |
196 | #define SCRUB_FLAG_HW_PROG BIT(SCRUB_HW_PROG) | |
522a94bd DT |
197 | #define SCRUB_FLAG_HW_SRC BIT(SCRUB_HW_SRC) |
198 | #define SCRUB_FLAG_HW_PROG_SRC BIT(SCRUB_HW_PROG_SRC) | |
7c9281d7 DT |
199 | #define SCRUB_FLAG_HW_TUN BIT(SCRUB_HW_TUNABLE) |
200 | ||
201 | /* FIXME - should have notify capabilities: NMI, LOG, PROC, etc */ | |
202 | ||
e27e3dac DT |
203 | extern char * edac_align_ptr(void *ptr, unsigned size); |
204 | ||
7c9281d7 DT |
205 | /* |
206 | * There are several things to be aware of that aren't at all obvious: | |
207 | * | |
208 | * | |
209 | * SOCKETS, SOCKET SETS, BANKS, ROWS, CHIP-SELECT ROWS, CHANNELS, etc.. | |
210 | * | |
211 | * These are some of the many terms that are thrown about that don't always | |
212 | * mean what people think they mean (Inconceivable!). In the interest of | |
213 | * creating a common ground for discussion, terms and their definitions | |
214 | * will be established. | |
215 | * | |
216 | * Memory devices: The individual chip on a memory stick. These devices | |
217 | * commonly output 4 and 8 bits each. Grouping several | |
218 | * of these in parallel provides 64 bits which is common | |
219 | * for a memory stick. | |
220 | * | |
221 | * Memory Stick: A printed circuit board that agregates multiple | |
222 | * memory devices in parallel. This is the atomic | |
223 | * memory component that is purchaseable by Joe consumer | |
224 | * and loaded into a memory socket. | |
225 | * | |
226 | * Socket: A physical connector on the motherboard that accepts | |
227 | * a single memory stick. | |
228 | * | |
229 | * Channel: Set of memory devices on a memory stick that must be | |
230 | * grouped in parallel with one or more additional | |
231 | * channels from other memory sticks. This parallel | |
232 | * grouping of the output from multiple channels are | |
233 | * necessary for the smallest granularity of memory access. | |
234 | * Some memory controllers are capable of single channel - | |
235 | * which means that memory sticks can be loaded | |
236 | * individually. Other memory controllers are only | |
237 | * capable of dual channel - which means that memory | |
238 | * sticks must be loaded as pairs (see "socket set"). | |
239 | * | |
240 | * Chip-select row: All of the memory devices that are selected together. | |
241 | * for a single, minimum grain of memory access. | |
242 | * This selects all of the parallel memory devices across | |
243 | * all of the parallel channels. Common chip-select rows | |
244 | * for single channel are 64 bits, for dual channel 128 | |
245 | * bits. | |
246 | * | |
247 | * Single-Ranked stick: A Single-ranked stick has 1 chip-select row of memmory. | |
248 | * Motherboards commonly drive two chip-select pins to | |
249 | * a memory stick. A single-ranked stick, will occupy | |
250 | * only one of those rows. The other will be unused. | |
251 | * | |
252 | * Double-Ranked stick: A double-ranked stick has two chip-select rows which | |
253 | * access different sets of memory devices. The two | |
254 | * rows cannot be accessed concurrently. | |
255 | * | |
256 | * Double-sided stick: DEPRECATED TERM, see Double-Ranked stick. | |
257 | * A double-sided stick has two chip-select rows which | |
258 | * access different sets of memory devices. The two | |
259 | * rows cannot be accessed concurrently. "Double-sided" | |
260 | * is irrespective of the memory devices being mounted | |
261 | * on both sides of the memory stick. | |
262 | * | |
263 | * Socket set: All of the memory sticks that are required for for | |
264 | * a single memory access or all of the memory sticks | |
265 | * spanned by a chip-select row. A single socket set | |
266 | * has two chip-select rows and if double-sided sticks | |
267 | * are used these will occupy those chip-select rows. | |
268 | * | |
269 | * Bank: This term is avoided because it is unclear when | |
270 | * needing to distinguish between chip-select rows and | |
271 | * socket sets. | |
272 | * | |
273 | * Controller pages: | |
274 | * | |
275 | * Physical pages: | |
276 | * | |
277 | * Virtual pages: | |
278 | * | |
279 | * | |
280 | * STRUCTURE ORGANIZATION AND CHOICES | |
281 | * | |
282 | * | |
283 | * | |
284 | * PS - I enjoyed writing all that about as much as you enjoyed reading it. | |
285 | */ | |
286 | ||
287 | struct channel_info { | |
288 | int chan_idx; /* channel index */ | |
289 | u32 ce_count; /* Correctable Errors for this CHANNEL */ | |
290 | char label[EDAC_MC_LABEL_LEN + 1]; /* DIMM label on motherboard */ | |
291 | struct csrow_info *csrow; /* the parent */ | |
292 | }; | |
293 | ||
294 | struct csrow_info { | |
295 | unsigned long first_page; /* first page number in dimm */ | |
296 | unsigned long last_page; /* last page number in dimm */ | |
297 | unsigned long page_mask; /* used for interleaving - | |
298 | * 0UL for non intlv | |
299 | */ | |
300 | u32 nr_pages; /* number of pages in csrow */ | |
301 | u32 grain; /* granularity of reported error in bytes */ | |
302 | int csrow_idx; /* the chip-select row */ | |
303 | enum dev_type dtype; /* memory device type */ | |
304 | u32 ue_count; /* Uncorrectable Errors for this csrow */ | |
305 | u32 ce_count; /* Correctable Errors for this csrow */ | |
306 | enum mem_type mtype; /* memory csrow type */ | |
307 | enum edac_type edac_mode; /* EDAC mode for this csrow */ | |
308 | struct mem_ctl_info *mci; /* the parent */ | |
309 | ||
310 | struct kobject kobj; /* sysfs kobject for this csrow */ | |
311 | struct completion kobj_complete; | |
312 | ||
313 | /* FIXME the number of CHANNELs might need to become dynamic */ | |
314 | u32 nr_channels; | |
315 | struct channel_info *channels; | |
316 | }; | |
317 | ||
318 | struct mem_ctl_info { | |
319 | struct list_head link; /* for global list of mem_ctl_info structs */ | |
320 | unsigned long mtype_cap; /* memory types supported by mc */ | |
321 | unsigned long edac_ctl_cap; /* Mem controller EDAC capabilities */ | |
322 | unsigned long edac_cap; /* configuration capabilities - this is | |
323 | * closely related to edac_ctl_cap. The | |
324 | * difference is that the controller may be | |
325 | * capable of s4ecd4ed which would be listed | |
326 | * in edac_ctl_cap, but if channels aren't | |
327 | * capable of s4ecd4ed then the edac_cap would | |
328 | * not have that capability. | |
329 | */ | |
330 | unsigned long scrub_cap; /* chipset scrub capabilities */ | |
331 | enum scrub_type scrub_mode; /* current scrub mode */ | |
332 | ||
333 | /* Translates sdram memory scrub rate given in bytes/sec to the | |
334 | internal representation and configures whatever else needs | |
335 | to be configured. | |
336 | */ | |
337 | int (*set_sdram_scrub_rate) (struct mem_ctl_info *mci, u32 *bw); | |
338 | ||
339 | /* Get the current sdram memory scrub rate from the internal | |
340 | representation and converts it to the closest matching | |
341 | bandwith in bytes/sec. | |
342 | */ | |
343 | int (*get_sdram_scrub_rate) (struct mem_ctl_info *mci, u32 *bw); | |
344 | ||
345 | /* pointer to edac checking routine */ | |
346 | void (*edac_check) (struct mem_ctl_info * mci); | |
347 | ||
348 | /* | |
349 | * Remaps memory pages: controller pages to physical pages. | |
350 | * For most MC's, this will be NULL. | |
351 | */ | |
352 | /* FIXME - why not send the phys page to begin with? */ | |
353 | unsigned long (*ctl_page_to_phys) (struct mem_ctl_info * mci, | |
354 | unsigned long page); | |
355 | int mc_idx; | |
356 | int nr_csrows; | |
357 | struct csrow_info *csrows; | |
358 | /* | |
359 | * FIXME - what about controllers on other busses? - IDs must be | |
360 | * unique. dev pointer should be sufficiently unique, but | |
361 | * BUS:SLOT.FUNC numbers may not be unique. | |
362 | */ | |
363 | struct device *dev; | |
364 | const char *mod_name; | |
365 | const char *mod_ver; | |
366 | const char *ctl_name; | |
c4192705 | 367 | const char *dev_name; |
7c9281d7 DT |
368 | char proc_name[MC_PROC_NAME_MAX_LEN + 1]; |
369 | void *pvt_info; | |
370 | u32 ue_noinfo_count; /* Uncorrectable Errors w/o info */ | |
371 | u32 ce_noinfo_count; /* Correctable Errors w/o info */ | |
372 | u32 ue_count; /* Total Uncorrectable Errors for this MC */ | |
373 | u32 ce_count; /* Total Correctable Errors for this MC */ | |
374 | unsigned long start_time; /* mci load start time (in jiffies) */ | |
375 | ||
376 | /* this stuff is for safe removal of mc devices from global list while | |
377 | * NMI handlers may be traversing list | |
378 | */ | |
379 | struct rcu_head rcu; | |
380 | struct completion complete; | |
381 | ||
382 | /* edac sysfs device control */ | |
383 | struct kobject edac_mci_kobj; | |
384 | struct completion kobj_complete; | |
81d87cb1 DJ |
385 | |
386 | /* work struct for this MC */ | |
387 | #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,20)) | |
388 | struct delayed_work work; | |
389 | #else | |
390 | struct work_struct work; | |
391 | #endif | |
392 | /* the internal state of this controller instance */ | |
393 | int op_state; | |
7c9281d7 DT |
394 | }; |
395 | ||
e27e3dac DT |
396 | /* |
397 | * The following are the structures to provide for a generice | |
398 | * or abstract 'edac_device'. This set of structures and the | |
399 | * code that implements the APIs for the same, provide for | |
400 | * registering EDAC type devices which are NOT standard memory. | |
401 | * | |
402 | * CPU caches (L1 and L2) | |
403 | * DMA engines | |
404 | * Core CPU swithces | |
405 | * Fabric switch units | |
406 | * PCIe interface controllers | |
407 | * other EDAC/ECC type devices that can be monitored for | |
408 | * errors, etc. | |
409 | * | |
410 | * It allows for a 2 level set of hiearchry. For example: | |
411 | * | |
412 | * cache could be composed of L1, L2 and L3 levels of cache. | |
413 | * Each CPU core would have its own L1 cache, while sharing | |
414 | * L2 and maybe L3 caches. | |
415 | * | |
416 | * View them arranged, via the sysfs presentation: | |
417 | * /sys/devices/system/edac/.. | |
418 | * | |
419 | * mc/ <existing memory device directory> | |
420 | * cpu/cpu0/.. <L1 and L2 block directory> | |
421 | * /L1-cache/ce_count | |
422 | * /ue_count | |
423 | * /L2-cache/ce_count | |
424 | * /ue_count | |
425 | * cpu/cpu1/.. <L1 and L2 block directory> | |
426 | * /L1-cache/ce_count | |
427 | * /ue_count | |
428 | * /L2-cache/ce_count | |
429 | * /ue_count | |
430 | * ... | |
431 | * | |
432 | * the L1 and L2 directories would be "edac_device_block's" | |
433 | */ | |
434 | ||
435 | struct edac_device_counter { | |
436 | u32 ue_count; | |
437 | u32 ce_count; | |
438 | }; | |
439 | ||
440 | #define INC_COUNTER(cnt) (cnt++) | |
441 | ||
442 | /* | |
443 | * An array of these is passed to the alloc() function | |
444 | * to specify attributes of the edac_block | |
445 | */ | |
446 | struct edac_attrib_spec { | |
447 | char name[EDAC_DEVICE_NAME_LEN + 1]; | |
448 | ||
449 | int type; | |
450 | #define EDAC_ATTR_INT 0x01 | |
451 | #define EDAC_ATTR_CHAR 0x02 | |
452 | }; | |
453 | ||
454 | ||
455 | /* Attribute control structure | |
456 | * In this structure is a pointer to the driver's edac_attrib_spec | |
457 | * The life of this pointer is inclusive in the life of the driver's | |
458 | * life cycle. | |
459 | */ | |
460 | struct edac_attrib { | |
461 | struct edac_device_block *block; /* Up Pointer */ | |
462 | ||
463 | struct edac_attrib_spec *spec; /* ptr to module spec entry */ | |
464 | ||
465 | union { /* actual value */ | |
466 | int edac_attrib_int_value; | |
467 | char edac_attrib_char_value[EDAC_ATTRIB_VALUE_LEN + 1]; | |
468 | } edac_attrib_value; | |
469 | }; | |
470 | ||
471 | /* device block control structure */ | |
472 | struct edac_device_block { | |
473 | struct edac_device_instance *instance; /* Up Pointer */ | |
474 | char name[EDAC_DEVICE_NAME_LEN + 1]; | |
475 | ||
476 | struct edac_device_counter counters; /* basic UE and CE counters */ | |
477 | ||
478 | int nr_attribs; /* how many attributes */ | |
479 | struct edac_attrib *attribs; /* this block's attributes */ | |
480 | ||
481 | /* edac sysfs device control */ | |
482 | struct kobject kobj; | |
483 | struct completion kobj_complete; | |
484 | }; | |
485 | ||
486 | /* device instance control structure */ | |
487 | struct edac_device_instance { | |
488 | struct edac_device_ctl_info *ctl; /* Up pointer */ | |
489 | char name[EDAC_DEVICE_NAME_LEN + 4]; | |
490 | ||
491 | struct edac_device_counter counters; /* instance counters */ | |
492 | ||
493 | u32 nr_blocks; /* how many blocks */ | |
494 | struct edac_device_block *blocks; /* block array */ | |
495 | ||
496 | /* edac sysfs device control */ | |
497 | struct kobject kobj; | |
498 | struct completion kobj_complete; | |
499 | }; | |
500 | ||
501 | ||
502 | /* | |
503 | * Abstract edac_device control info structure | |
504 | * | |
505 | */ | |
506 | struct edac_device_ctl_info { | |
507 | /* for global list of edac_device_ctl_info structs */ | |
508 | struct list_head link; | |
509 | ||
510 | int dev_idx; | |
511 | ||
512 | /* Per instance controls for this edac_device */ | |
513 | int log_ue; /* boolean for logging UEs */ | |
514 | int log_ce; /* boolean for logging CEs */ | |
515 | int panic_on_ue; /* boolean for panic'ing on an UE */ | |
516 | unsigned poll_msec; /* number of milliseconds to poll interval */ | |
517 | unsigned long delay; /* number of jiffies for poll_msec */ | |
518 | ||
519 | struct sysdev_class *edac_class; /* pointer to class */ | |
520 | ||
521 | /* the internal state of this controller instance */ | |
522 | int op_state; | |
523 | #define OP_ALLOC 0x100 | |
524 | #define OP_RUNNING_POLL 0x201 | |
525 | #define OP_RUNNING_INTERRUPT 0x202 | |
526 | #define OP_RUNNING_POLL_INTR 0x203 | |
527 | #define OP_OFFLINE 0x300 | |
528 | ||
529 | /* work struct for this instance */ | |
530 | #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,20)) | |
531 | struct delayed_work work; | |
532 | #else | |
533 | struct work_struct work; | |
534 | #endif | |
535 | ||
536 | /* pointer to edac polling checking routine: | |
537 | * If NOT NULL: points to polling check routine | |
538 | * If NULL: Then assumes INTERRUPT operation, where | |
539 | * MC driver will receive events | |
540 | */ | |
541 | void (*edac_check) (struct edac_device_ctl_info * edac_dev); | |
542 | ||
543 | struct device *dev; /* pointer to device structure */ | |
544 | ||
545 | const char *mod_name; /* module name */ | |
546 | const char *ctl_name; /* edac controller name */ | |
c4192705 | 547 | const char *dev_name; /* pci/platform/etc... name */ |
e27e3dac DT |
548 | |
549 | void *pvt_info; /* pointer to 'private driver' info */ | |
550 | ||
551 | unsigned long start_time;/* edac_device load start time (jiffies)*/ | |
552 | ||
553 | /* these are for safe removal of mc devices from global list while | |
554 | * NMI handlers may be traversing list | |
555 | */ | |
556 | struct rcu_head rcu; | |
557 | struct completion complete; | |
558 | ||
559 | /* sysfs top name under 'edac' directory | |
560 | * and instance name: | |
561 | * cpu/cpu0/... | |
562 | * cpu/cpu1/... | |
563 | * cpu/cpu2/... | |
564 | * ... | |
565 | */ | |
566 | char name[EDAC_DEVICE_NAME_LEN + 1]; | |
567 | ||
568 | /* Number of instances supported on this control structure | |
569 | * and the array of those instances | |
570 | */ | |
571 | u32 nr_instances; | |
572 | struct edac_device_instance *instances; | |
573 | ||
574 | /* Event counters for the this whole EDAC Device */ | |
575 | struct edac_device_counter counters; | |
576 | ||
577 | /* edac sysfs device control for the 'name' | |
578 | * device this structure controls | |
579 | */ | |
580 | struct kobject kobj; | |
581 | struct completion kobj_complete; | |
582 | }; | |
583 | ||
584 | /* To get from the instance's wq to the beginning of the ctl structure */ | |
81d87cb1 DJ |
585 | #define to_edac_mem_ctl_work(w) \ |
586 | container_of(w, struct mem_ctl_info, work) | |
587 | ||
e27e3dac DT |
588 | #define to_edac_device_ctl_work(w) \ |
589 | container_of(w,struct edac_device_ctl_info,work) | |
590 | ||
591 | /* Function to calc the number of delay jiffies from poll_msec */ | |
592 | static inline void edac_device_calc_delay( | |
593 | struct edac_device_ctl_info *edac_dev) | |
594 | { | |
595 | /* convert from msec to jiffies */ | |
596 | edac_dev->delay = edac_dev->poll_msec * HZ / 1000; | |
597 | } | |
598 | ||
81d87cb1 DJ |
599 | #define edac_calc_delay(dev) dev->delay = dev->poll_msec * HZ / 1000; |
600 | ||
e27e3dac DT |
601 | /* |
602 | * The alloc() and free() functions for the 'edac_device' control info | |
603 | * structure. A MC driver will allocate one of these for each edac_device | |
604 | * it is going to control/register with the EDAC CORE. | |
605 | */ | |
606 | extern struct edac_device_ctl_info *edac_device_alloc_ctl_info( | |
607 | unsigned sizeof_private, | |
608 | char *edac_device_name, | |
609 | unsigned nr_instances, | |
610 | char *edac_block_name, | |
611 | unsigned nr_blocks, | |
612 | unsigned offset_value, | |
613 | struct edac_attrib_spec *attrib_spec, | |
614 | unsigned nr_attribs | |
615 | ); | |
616 | ||
617 | /* The offset value can be: | |
618 | * -1 indicating no offset value | |
619 | * 0 for zero-based block numbers | |
620 | * 1 for 1-based block number | |
621 | * other for other-based block number | |
622 | */ | |
623 | #define BLOCK_OFFSET_VALUE_OFF ((unsigned) -1) | |
624 | ||
625 | extern void edac_device_free_ctl_info( struct edac_device_ctl_info *ctl_info); | |
626 | ||
7c9281d7 DT |
627 | #ifdef CONFIG_PCI |
628 | ||
629 | /* write all or some bits in a byte-register*/ | |
630 | static inline void pci_write_bits8(struct pci_dev *pdev, int offset, u8 value, | |
631 | u8 mask) | |
632 | { | |
633 | if (mask != 0xff) { | |
634 | u8 buf; | |
635 | ||
636 | pci_read_config_byte(pdev, offset, &buf); | |
637 | value &= mask; | |
638 | buf &= ~mask; | |
639 | value |= buf; | |
640 | } | |
641 | ||
642 | pci_write_config_byte(pdev, offset, value); | |
643 | } | |
644 | ||
645 | /* write all or some bits in a word-register*/ | |
646 | static inline void pci_write_bits16(struct pci_dev *pdev, int offset, | |
647 | u16 value, u16 mask) | |
648 | { | |
649 | if (mask != 0xffff) { | |
650 | u16 buf; | |
651 | ||
652 | pci_read_config_word(pdev, offset, &buf); | |
653 | value &= mask; | |
654 | buf &= ~mask; | |
655 | value |= buf; | |
656 | } | |
657 | ||
658 | pci_write_config_word(pdev, offset, value); | |
659 | } | |
660 | ||
661 | /* write all or some bits in a dword-register*/ | |
662 | static inline void pci_write_bits32(struct pci_dev *pdev, int offset, | |
663 | u32 value, u32 mask) | |
664 | { | |
665 | if (mask != 0xffff) { | |
666 | u32 buf; | |
667 | ||
668 | pci_read_config_dword(pdev, offset, &buf); | |
669 | value &= mask; | |
670 | buf &= ~mask; | |
671 | value |= buf; | |
672 | } | |
673 | ||
674 | pci_write_config_dword(pdev, offset, value); | |
675 | } | |
676 | ||
677 | #endif /* CONFIG_PCI */ | |
678 | ||
679 | extern struct mem_ctl_info * edac_mc_find(int idx); | |
680 | extern int edac_mc_add_mc(struct mem_ctl_info *mci,int mc_idx); | |
681 | extern struct mem_ctl_info * edac_mc_del_mc(struct device *dev); | |
682 | extern int edac_mc_find_csrow_by_page(struct mem_ctl_info *mci, | |
683 | unsigned long page); | |
684 | ||
685 | /* | |
686 | * The no info errors are used when error overflows are reported. | |
687 | * There are a limited number of error logging registers that can | |
688 | * be exausted. When all registers are exhausted and an additional | |
689 | * error occurs then an error overflow register records that an | |
690 | * error occured and the type of error, but doesn't have any | |
691 | * further information. The ce/ue versions make for cleaner | |
692 | * reporting logic and function interface - reduces conditional | |
693 | * statement clutter and extra function arguments. | |
694 | */ | |
695 | extern void edac_mc_handle_ce(struct mem_ctl_info *mci, | |
696 | unsigned long page_frame_number, unsigned long offset_in_page, | |
697 | unsigned long syndrome, int row, int channel, | |
698 | const char *msg); | |
699 | extern void edac_mc_handle_ce_no_info(struct mem_ctl_info *mci, | |
700 | const char *msg); | |
701 | extern void edac_mc_handle_ue(struct mem_ctl_info *mci, | |
702 | unsigned long page_frame_number, unsigned long offset_in_page, | |
703 | int row, const char *msg); | |
704 | extern void edac_mc_handle_ue_no_info(struct mem_ctl_info *mci, | |
705 | const char *msg); | |
706 | extern void edac_mc_handle_fbd_ue(struct mem_ctl_info *mci, | |
707 | unsigned int csrow, | |
708 | unsigned int channel0, | |
709 | unsigned int channel1, | |
710 | char *msg); | |
711 | extern void edac_mc_handle_fbd_ce(struct mem_ctl_info *mci, | |
712 | unsigned int csrow, | |
713 | unsigned int channel, | |
714 | char *msg); | |
715 | ||
716 | /* | |
e27e3dac | 717 | * edac_device APIs |
7c9281d7 DT |
718 | */ |
719 | extern struct mem_ctl_info *edac_mc_alloc(unsigned sz_pvt, unsigned nr_csrows, | |
720 | unsigned nr_chans); | |
7c9281d7 | 721 | extern void edac_mc_free(struct mem_ctl_info *mci); |
e27e3dac DT |
722 | extern int edac_device_add_device(struct edac_device_ctl_info *edac_dev, int edac_idx); |
723 | extern struct edac_device_ctl_info * edac_device_del_device(struct device *dev); | |
724 | extern void edac_device_handle_ue(struct edac_device_ctl_info *edac_dev, | |
725 | int inst_nr, int block_nr, const char *msg); | |
726 | extern void edac_device_handle_ce(struct edac_device_ctl_info *edac_dev, | |
727 | int inst_nr, int block_nr, const char *msg); | |
728 | ||
7c9281d7 DT |
729 | |
730 | #endif /* _EDAC_CORE_H_ */ |