aboutsummaryrefslogtreecommitdiffstats
path: root/drivers
diff options
context:
space:
mode:
Diffstat (limited to 'drivers')
-rw-r--r--drivers/edac/Makefile7
-rw-r--r--drivers/edac/edac_core.h478
-rw-r--r--drivers/edac/edac_mc.c1317
-rw-r--r--drivers/edac/edac_mc.h471
-rw-r--r--drivers/edac/edac_mc_sysfs.c889
-rw-r--r--drivers/edac/edac_module.c130
-rw-r--r--drivers/edac/edac_module.h55
-rw-r--r--drivers/edac/edac_pci_sysfs.c361
8 files changed, 1932 insertions, 1776 deletions
diff --git a/drivers/edac/Makefile b/drivers/edac/Makefile
index 93137fdab4b3..51f59aa84d30 100644
--- a/drivers/edac/Makefile
+++ b/drivers/edac/Makefile
@@ -8,7 +8,12 @@
8# $Id: Makefile,v 1.4.2.3 2005/07/08 22:05:38 dsp_llnl Exp $ 8# $Id: Makefile,v 1.4.2.3 2005/07/08 22:05:38 dsp_llnl Exp $
9 9
10 10
11obj-$(CONFIG_EDAC_MM_EDAC) += edac_mc.o 11obj-$(CONFIG_EDAC_MM_EDAC) += edac_core.o
12
13edac_core-objs := edac_mc.o edac_mc_sysfs.o edac_pci_sysfs.o
14
15edac_core-objs += edac_module.o
16
12obj-$(CONFIG_EDAC_AMD76X) += amd76x_edac.o 17obj-$(CONFIG_EDAC_AMD76X) += amd76x_edac.o
13obj-$(CONFIG_EDAC_E7XXX) += e7xxx_edac.o 18obj-$(CONFIG_EDAC_E7XXX) += e7xxx_edac.o
14obj-$(CONFIG_EDAC_E752X) += e752x_edac.o 19obj-$(CONFIG_EDAC_E752X) += e752x_edac.o
diff --git a/drivers/edac/edac_core.h b/drivers/edac/edac_core.h
new file mode 100644
index 000000000000..397f144791ec
--- /dev/null
+++ b/drivers/edac/edac_core.h
@@ -0,0 +1,478 @@
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 <dsp@llnl.gov> <dave_peterson@pobox.com>
14 *
15 * Refactored for multi-source files:
16 * Doug Thompson <norsk5@xmission.com>
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>
35
36#define EDAC_MC_LABEL_LEN 31
37#define MC_PROC_NAME_MAX_LEN 7
38
39#if PAGE_SHIFT < 20
40#define PAGES_TO_MiB( pages ) ( ( pages ) >> ( 20 - PAGE_SHIFT ) )
41#else /* PAGE_SHIFT > 20 */
42#define PAGES_TO_MiB( pages ) ( ( pages ) << ( PAGE_SHIFT - 20 ) )
43#endif
44
45#define edac_printk(level, prefix, fmt, arg...) \
46 printk(level "EDAC " prefix ": " fmt, ##arg)
47
48#define edac_mc_printk(mci, level, fmt, arg...) \
49 printk(level "EDAC MC%d: " fmt, mci->mc_idx, ##arg)
50
51#define edac_mc_chipset_printk(mci, level, prefix, fmt, arg...) \
52 printk(level "EDAC " prefix " MC%d: " fmt, mci->mc_idx, ##arg)
53
54/* prefixes for edac_printk() and edac_mc_printk() */
55#define EDAC_MC "MC"
56#define EDAC_PCI "PCI"
57#define EDAC_DEBUG "DEBUG"
58
59#ifdef CONFIG_EDAC_DEBUG
60extern int edac_debug_level;
61
62#define edac_debug_printk(level, fmt, arg...) \
63 do { \
64 if (level <= edac_debug_level) \
65 edac_printk(KERN_DEBUG, EDAC_DEBUG, fmt, ##arg); \
66 } while(0)
67
68#define debugf0( ... ) edac_debug_printk(0, __VA_ARGS__ )
69#define debugf1( ... ) edac_debug_printk(1, __VA_ARGS__ )
70#define debugf2( ... ) edac_debug_printk(2, __VA_ARGS__ )
71#define debugf3( ... ) edac_debug_printk(3, __VA_ARGS__ )
72#define debugf4( ... ) edac_debug_printk(4, __VA_ARGS__ )
73
74#else /* !CONFIG_EDAC_DEBUG */
75
76#define debugf0( ... )
77#define debugf1( ... )
78#define debugf2( ... )
79#define debugf3( ... )
80#define debugf4( ... )
81
82#endif /* !CONFIG_EDAC_DEBUG */
83
84#define BIT(x) (1 << (x))
85
86#define PCI_VEND_DEV(vend, dev) PCI_VENDOR_ID_ ## vend, \
87 PCI_DEVICE_ID_ ## vend ## _ ## dev
88
89#if defined(CONFIG_X86) && defined(CONFIG_PCI)
90#define dev_name(dev) pci_name(to_pci_dev(dev))
91#else
92#define dev_name(dev) to_platform_device(dev)->name
93#endif
94
95/* memory devices */
96enum dev_type {
97 DEV_UNKNOWN = 0,
98 DEV_X1,
99 DEV_X2,
100 DEV_X4,
101 DEV_X8,
102 DEV_X16,
103 DEV_X32, /* Do these parts exist? */
104 DEV_X64 /* Do these parts exist? */
105};
106
107#define DEV_FLAG_UNKNOWN BIT(DEV_UNKNOWN)
108#define DEV_FLAG_X1 BIT(DEV_X1)
109#define DEV_FLAG_X2 BIT(DEV_X2)
110#define DEV_FLAG_X4 BIT(DEV_X4)
111#define DEV_FLAG_X8 BIT(DEV_X8)
112#define DEV_FLAG_X16 BIT(DEV_X16)
113#define DEV_FLAG_X32 BIT(DEV_X32)
114#define DEV_FLAG_X64 BIT(DEV_X64)
115
116/* memory types */
117enum mem_type {
118 MEM_EMPTY = 0, /* Empty csrow */
119 MEM_RESERVED, /* Reserved csrow type */
120 MEM_UNKNOWN, /* Unknown csrow type */
121 MEM_FPM, /* Fast page mode */
122 MEM_EDO, /* Extended data out */
123 MEM_BEDO, /* Burst Extended data out */
124 MEM_SDR, /* Single data rate SDRAM */
125 MEM_RDR, /* Registered single data rate SDRAM */
126 MEM_DDR, /* Double data rate SDRAM */
127 MEM_RDDR, /* Registered Double data rate SDRAM */
128 MEM_RMBS, /* Rambus DRAM */
129 MEM_DDR2, /* DDR2 RAM */
130 MEM_FB_DDR2, /* fully buffered DDR2 */
131 MEM_RDDR2, /* Registered DDR2 RAM */
132};
133
134#define MEM_FLAG_EMPTY BIT(MEM_EMPTY)
135#define MEM_FLAG_RESERVED BIT(MEM_RESERVED)
136#define MEM_FLAG_UNKNOWN BIT(MEM_UNKNOWN)
137#define MEM_FLAG_FPM BIT(MEM_FPM)
138#define MEM_FLAG_EDO BIT(MEM_EDO)
139#define MEM_FLAG_BEDO BIT(MEM_BEDO)
140#define MEM_FLAG_SDR BIT(MEM_SDR)
141#define MEM_FLAG_RDR BIT(MEM_RDR)
142#define MEM_FLAG_DDR BIT(MEM_DDR)
143#define MEM_FLAG_RDDR BIT(MEM_RDDR)
144#define MEM_FLAG_RMBS BIT(MEM_RMBS)
145#define MEM_FLAG_DDR2 BIT(MEM_DDR2)
146#define MEM_FLAG_FB_DDR2 BIT(MEM_FB_DDR2)
147#define MEM_FLAG_RDDR2 BIT(MEM_RDDR2)
148
149/* chipset Error Detection and Correction capabilities and mode */
150enum edac_type {
151 EDAC_UNKNOWN = 0, /* Unknown if ECC is available */
152 EDAC_NONE, /* Doesnt support ECC */
153 EDAC_RESERVED, /* Reserved ECC type */
154 EDAC_PARITY, /* Detects parity errors */
155 EDAC_EC, /* Error Checking - no correction */
156 EDAC_SECDED, /* Single bit error correction, Double detection */
157 EDAC_S2ECD2ED, /* Chipkill x2 devices - do these exist? */
158 EDAC_S4ECD4ED, /* Chipkill x4 devices */
159 EDAC_S8ECD8ED, /* Chipkill x8 devices */
160 EDAC_S16ECD16ED, /* Chipkill x16 devices */
161};
162
163#define EDAC_FLAG_UNKNOWN BIT(EDAC_UNKNOWN)
164#define EDAC_FLAG_NONE BIT(EDAC_NONE)
165#define EDAC_FLAG_PARITY BIT(EDAC_PARITY)
166#define EDAC_FLAG_EC BIT(EDAC_EC)
167#define EDAC_FLAG_SECDED BIT(EDAC_SECDED)
168#define EDAC_FLAG_S2ECD2ED BIT(EDAC_S2ECD2ED)
169#define EDAC_FLAG_S4ECD4ED BIT(EDAC_S4ECD4ED)
170#define EDAC_FLAG_S8ECD8ED BIT(EDAC_S8ECD8ED)
171#define EDAC_FLAG_S16ECD16ED BIT(EDAC_S16ECD16ED)
172
173/* scrubbing capabilities */
174enum scrub_type {
175 SCRUB_UNKNOWN = 0, /* Unknown if scrubber is available */
176 SCRUB_NONE, /* No scrubber */
177 SCRUB_SW_PROG, /* SW progressive (sequential) scrubbing */
178 SCRUB_SW_SRC, /* Software scrub only errors */
179 SCRUB_SW_PROG_SRC, /* Progressive software scrub from an error */
180 SCRUB_SW_TUNABLE, /* Software scrub frequency is tunable */
181 SCRUB_HW_PROG, /* HW progressive (sequential) scrubbing */
182 SCRUB_HW_SRC, /* Hardware scrub only errors */
183 SCRUB_HW_PROG_SRC, /* Progressive hardware scrub from an error */
184 SCRUB_HW_TUNABLE /* Hardware scrub frequency is tunable */
185};
186
187#define SCRUB_FLAG_SW_PROG BIT(SCRUB_SW_PROG)
188#define SCRUB_FLAG_SW_SRC BIT(SCRUB_SW_SRC_CORR)
189#define SCRUB_FLAG_SW_PROG_SRC BIT(SCRUB_SW_PROG_SRC_CORR)
190#define SCRUB_FLAG_SW_TUN BIT(SCRUB_SW_SCRUB_TUNABLE)
191#define SCRUB_FLAG_HW_PROG BIT(SCRUB_HW_PROG)
192#define SCRUB_FLAG_HW_SRC BIT(SCRUB_HW_SRC_CORR)
193#define SCRUB_FLAG_HW_PROG_SRC BIT(SCRUB_HW_PROG_SRC_CORR)
194#define SCRUB_FLAG_HW_TUN BIT(SCRUB_HW_TUNABLE)
195
196/* FIXME - should have notify capabilities: NMI, LOG, PROC, etc */
197
198/*
199 * There are several things to be aware of that aren't at all obvious:
200 *
201 *
202 * SOCKETS, SOCKET SETS, BANKS, ROWS, CHIP-SELECT ROWS, CHANNELS, etc..
203 *
204 * These are some of the many terms that are thrown about that don't always
205 * mean what people think they mean (Inconceivable!). In the interest of
206 * creating a common ground for discussion, terms and their definitions
207 * will be established.
208 *
209 * Memory devices: The individual chip on a memory stick. These devices
210 * commonly output 4 and 8 bits each. Grouping several
211 * of these in parallel provides 64 bits which is common
212 * for a memory stick.
213 *
214 * Memory Stick: A printed circuit board that agregates multiple
215 * memory devices in parallel. This is the atomic
216 * memory component that is purchaseable by Joe consumer
217 * and loaded into a memory socket.
218 *
219 * Socket: A physical connector on the motherboard that accepts
220 * a single memory stick.
221 *
222 * Channel: Set of memory devices on a memory stick that must be
223 * grouped in parallel with one or more additional
224 * channels from other memory sticks. This parallel
225 * grouping of the output from multiple channels are
226 * necessary for the smallest granularity of memory access.
227 * Some memory controllers are capable of single channel -
228 * which means that memory sticks can be loaded
229 * individually. Other memory controllers are only
230 * capable of dual channel - which means that memory
231 * sticks must be loaded as pairs (see "socket set").
232 *
233 * Chip-select row: All of the memory devices that are selected together.
234 * for a single, minimum grain of memory access.
235 * This selects all of the parallel memory devices across
236 * all of the parallel channels. Common chip-select rows
237 * for single channel are 64 bits, for dual channel 128
238 * bits.
239 *
240 * Single-Ranked stick: A Single-ranked stick has 1 chip-select row of memmory.
241 * Motherboards commonly drive two chip-select pins to
242 * a memory stick. A single-ranked stick, will occupy
243 * only one of those rows. The other will be unused.
244 *
245 * Double-Ranked stick: A double-ranked stick has two chip-select rows which
246 * access different sets of memory devices. The two
247 * rows cannot be accessed concurrently.
248 *
249 * Double-sided stick: DEPRECATED TERM, see Double-Ranked stick.
250 * A double-sided stick has two chip-select rows which
251 * access different sets of memory devices. The two
252 * rows cannot be accessed concurrently. "Double-sided"
253 * is irrespective of the memory devices being mounted
254 * on both sides of the memory stick.
255 *
256 * Socket set: All of the memory sticks that are required for for
257 * a single memory access or all of the memory sticks
258 * spanned by a chip-select row. A single socket set
259 * has two chip-select rows and if double-sided sticks
260 * are used these will occupy those chip-select rows.
261 *
262 * Bank: This term is avoided because it is unclear when
263 * needing to distinguish between chip-select rows and
264 * socket sets.
265 *
266 * Controller pages:
267 *
268 * Physical pages:
269 *
270 * Virtual pages:
271 *
272 *
273 * STRUCTURE ORGANIZATION AND CHOICES
274 *
275 *
276 *
277 * PS - I enjoyed writing all that about as much as you enjoyed reading it.
278 */
279
280struct channel_info {
281 int chan_idx; /* channel index */
282 u32 ce_count; /* Correctable Errors for this CHANNEL */
283 char label[EDAC_MC_LABEL_LEN + 1]; /* DIMM label on motherboard */
284 struct csrow_info *csrow; /* the parent */
285};
286
287struct csrow_info {
288 unsigned long first_page; /* first page number in dimm */
289 unsigned long last_page; /* last page number in dimm */
290 unsigned long page_mask; /* used for interleaving -
291 * 0UL for non intlv
292 */
293 u32 nr_pages; /* number of pages in csrow */
294 u32 grain; /* granularity of reported error in bytes */
295 int csrow_idx; /* the chip-select row */
296 enum dev_type dtype; /* memory device type */
297 u32 ue_count; /* Uncorrectable Errors for this csrow */
298 u32 ce_count; /* Correctable Errors for this csrow */
299 enum mem_type mtype; /* memory csrow type */
300 enum edac_type edac_mode; /* EDAC mode for this csrow */
301 struct mem_ctl_info *mci; /* the parent */
302
303 struct kobject kobj; /* sysfs kobject for this csrow */
304 struct completion kobj_complete;
305
306 /* FIXME the number of CHANNELs might need to become dynamic */
307 u32 nr_channels;
308 struct channel_info *channels;
309};
310
311struct mem_ctl_info {
312 struct list_head link; /* for global list of mem_ctl_info structs */
313 unsigned long mtype_cap; /* memory types supported by mc */
314 unsigned long edac_ctl_cap; /* Mem controller EDAC capabilities */
315 unsigned long edac_cap; /* configuration capabilities - this is
316 * closely related to edac_ctl_cap. The
317 * difference is that the controller may be
318 * capable of s4ecd4ed which would be listed
319 * in edac_ctl_cap, but if channels aren't
320 * capable of s4ecd4ed then the edac_cap would
321 * not have that capability.
322 */
323 unsigned long scrub_cap; /* chipset scrub capabilities */
324 enum scrub_type scrub_mode; /* current scrub mode */
325
326 /* Translates sdram memory scrub rate given in bytes/sec to the
327 internal representation and configures whatever else needs
328 to be configured.
329 */
330 int (*set_sdram_scrub_rate) (struct mem_ctl_info *mci, u32 *bw);
331
332 /* Get the current sdram memory scrub rate from the internal
333 representation and converts it to the closest matching
334 bandwith in bytes/sec.
335 */
336 int (*get_sdram_scrub_rate) (struct mem_ctl_info *mci, u32 *bw);
337
338 /* pointer to edac checking routine */
339 void (*edac_check) (struct mem_ctl_info * mci);
340
341 /*
342 * Remaps memory pages: controller pages to physical pages.
343 * For most MC's, this will be NULL.
344 */
345 /* FIXME - why not send the phys page to begin with? */
346 unsigned long (*ctl_page_to_phys) (struct mem_ctl_info * mci,
347 unsigned long page);
348 int mc_idx;
349 int nr_csrows;
350 struct csrow_info *csrows;
351 /*
352 * FIXME - what about controllers on other busses? - IDs must be
353 * unique. dev pointer should be sufficiently unique, but
354 * BUS:SLOT.FUNC numbers may not be unique.
355 */
356 struct device *dev;
357 const char *mod_name;
358 const char *mod_ver;
359 const char *ctl_name;
360 char proc_name[MC_PROC_NAME_MAX_LEN + 1];
361 void *pvt_info;
362 u32 ue_noinfo_count; /* Uncorrectable Errors w/o info */
363 u32 ce_noinfo_count; /* Correctable Errors w/o info */
364 u32 ue_count; /* Total Uncorrectable Errors for this MC */
365 u32 ce_count; /* Total Correctable Errors for this MC */
366 unsigned long start_time; /* mci load start time (in jiffies) */
367
368 /* this stuff is for safe removal of mc devices from global list while
369 * NMI handlers may be traversing list
370 */
371 struct rcu_head rcu;
372 struct completion complete;
373
374 /* edac sysfs device control */
375 struct kobject edac_mci_kobj;
376 struct completion kobj_complete;
377};
378
379#ifdef CONFIG_PCI
380
381/* write all or some bits in a byte-register*/
382static inline void pci_write_bits8(struct pci_dev *pdev, int offset, u8 value,
383 u8 mask)
384{
385 if (mask != 0xff) {
386 u8 buf;
387
388 pci_read_config_byte(pdev, offset, &buf);
389 value &= mask;
390 buf &= ~mask;
391 value |= buf;
392 }
393
394 pci_write_config_byte(pdev, offset, value);
395}
396
397/* write all or some bits in a word-register*/
398static inline void pci_write_bits16(struct pci_dev *pdev, int offset,
399 u16 value, u16 mask)
400{
401 if (mask != 0xffff) {
402 u16 buf;
403
404 pci_read_config_word(pdev, offset, &buf);
405 value &= mask;
406 buf &= ~mask;
407 value |= buf;
408 }
409
410 pci_write_config_word(pdev, offset, value);
411}
412
413/* write all or some bits in a dword-register*/
414static inline void pci_write_bits32(struct pci_dev *pdev, int offset,
415 u32 value, u32 mask)
416{
417 if (mask != 0xffff) {
418 u32 buf;
419
420 pci_read_config_dword(pdev, offset, &buf);
421 value &= mask;
422 buf &= ~mask;
423 value |= buf;
424 }
425
426 pci_write_config_dword(pdev, offset, value);
427}
428
429#endif /* CONFIG_PCI */
430
431extern struct mem_ctl_info * edac_mc_find(int idx);
432extern int edac_mc_add_mc(struct mem_ctl_info *mci,int mc_idx);
433extern struct mem_ctl_info * edac_mc_del_mc(struct device *dev);
434extern int edac_mc_find_csrow_by_page(struct mem_ctl_info *mci,
435 unsigned long page);
436
437/*
438 * The no info errors are used when error overflows are reported.
439 * There are a limited number of error logging registers that can
440 * be exausted. When all registers are exhausted and an additional
441 * error occurs then an error overflow register records that an
442 * error occured and the type of error, but doesn't have any
443 * further information. The ce/ue versions make for cleaner
444 * reporting logic and function interface - reduces conditional
445 * statement clutter and extra function arguments.
446 */
447extern void edac_mc_handle_ce(struct mem_ctl_info *mci,
448 unsigned long page_frame_number, unsigned long offset_in_page,
449 unsigned long syndrome, int row, int channel,
450 const char *msg);
451extern void edac_mc_handle_ce_no_info(struct mem_ctl_info *mci,
452 const char *msg);
453extern void edac_mc_handle_ue(struct mem_ctl_info *mci,
454 unsigned long page_frame_number, unsigned long offset_in_page,
455 int row, const char *msg);
456extern void edac_mc_handle_ue_no_info(struct mem_ctl_info *mci,
457 const char *msg);
458extern void edac_mc_handle_fbd_ue(struct mem_ctl_info *mci,
459 unsigned int csrow,
460 unsigned int channel0,
461 unsigned int channel1,
462 char *msg);
463extern void edac_mc_handle_fbd_ce(struct mem_ctl_info *mci,
464 unsigned int csrow,
465 unsigned int channel,
466 char *msg);
467
468/*
469 * This kmalloc's and initializes all the structures.
470 * Can't be used if all structures don't have the same lifetime.
471 */
472extern struct mem_ctl_info *edac_mc_alloc(unsigned sz_pvt, unsigned nr_csrows,
473 unsigned nr_chans);
474
475/* Free an mc previously allocated by edac_mc_alloc() */
476extern void edac_mc_free(struct mem_ctl_info *mci);
477
478#endif /* _EDAC_CORE_H_ */
diff --git a/drivers/edac/edac_mc.c b/drivers/edac/edac_mc.c
index 88bee33e7ecf..3be5b7fe79cd 100644
--- a/drivers/edac/edac_mc.c
+++ b/drivers/edac/edac_mc.c
@@ -27,1197 +27,17 @@
27#include <linux/list.h> 27#include <linux/list.h>
28#include <linux/sysdev.h> 28#include <linux/sysdev.h>
29#include <linux/ctype.h> 29#include <linux/ctype.h>
30#include <linux/kthread.h>
31#include <linux/freezer.h>
32#include <asm/uaccess.h> 30#include <asm/uaccess.h>
33#include <asm/page.h> 31#include <asm/page.h>
34#include <asm/edac.h> 32#include <asm/edac.h>
35#include "edac_mc.h" 33#include "edac_mc.h"
34#include "edac_module.h"
36 35
37#define EDAC_MC_VERSION "Ver: 2.0.1 " __DATE__
38
39
40#ifdef CONFIG_EDAC_DEBUG
41/* Values of 0 to 4 will generate output */
42int edac_debug_level = 1;
43EXPORT_SYMBOL_GPL(edac_debug_level);
44#endif
45
46/* EDAC Controls, setable by module parameter, and sysfs */
47static int log_ue = 1;
48static int log_ce = 1;
49static int panic_on_ue;
50static int poll_msec = 1000;
51 36
52/* lock to memory controller's control array */ 37/* lock to memory controller's control array */
53static DECLARE_MUTEX(mem_ctls_mutex); 38static DECLARE_MUTEX(mem_ctls_mutex);
54static struct list_head mc_devices = LIST_HEAD_INIT(mc_devices); 39static struct list_head mc_devices = LIST_HEAD_INIT(mc_devices);
55 40
56static struct task_struct *edac_thread;
57
58#ifdef CONFIG_PCI
59static int check_pci_parity = 0; /* default YES check PCI parity */
60static int panic_on_pci_parity; /* default no panic on PCI Parity */
61static atomic_t pci_parity_count = ATOMIC_INIT(0);
62
63static struct kobject edac_pci_kobj; /* /sys/devices/system/edac/pci */
64static struct completion edac_pci_kobj_complete;
65#endif /* CONFIG_PCI */
66
67/* START sysfs data and methods */
68
69
70static const char *mem_types[] = {
71 [MEM_EMPTY] = "Empty",
72 [MEM_RESERVED] = "Reserved",
73 [MEM_UNKNOWN] = "Unknown",
74 [MEM_FPM] = "FPM",
75 [MEM_EDO] = "EDO",
76 [MEM_BEDO] = "BEDO",
77 [MEM_SDR] = "Unbuffered-SDR",
78 [MEM_RDR] = "Registered-SDR",
79 [MEM_DDR] = "Unbuffered-DDR",
80 [MEM_RDDR] = "Registered-DDR",
81 [MEM_RMBS] = "RMBS"
82};
83
84static const char *dev_types[] = {
85 [DEV_UNKNOWN] = "Unknown",
86 [DEV_X1] = "x1",
87 [DEV_X2] = "x2",
88 [DEV_X4] = "x4",
89 [DEV_X8] = "x8",
90 [DEV_X16] = "x16",
91 [DEV_X32] = "x32",
92 [DEV_X64] = "x64"
93};
94
95static const char *edac_caps[] = {
96 [EDAC_UNKNOWN] = "Unknown",
97 [EDAC_NONE] = "None",
98 [EDAC_RESERVED] = "Reserved",
99 [EDAC_PARITY] = "PARITY",
100 [EDAC_EC] = "EC",
101 [EDAC_SECDED] = "SECDED",
102 [EDAC_S2ECD2ED] = "S2ECD2ED",
103 [EDAC_S4ECD4ED] = "S4ECD4ED",
104 [EDAC_S8ECD8ED] = "S8ECD8ED",
105 [EDAC_S16ECD16ED] = "S16ECD16ED"
106};
107
108/* sysfs object: /sys/devices/system/edac */
109static struct sysdev_class edac_class = {
110 set_kset_name("edac"),
111};
112
113/* sysfs object:
114 * /sys/devices/system/edac/mc
115 */
116static struct kobject edac_memctrl_kobj;
117
118/* We use these to wait for the reference counts on edac_memctrl_kobj and
119 * edac_pci_kobj to reach 0.
120 */
121static struct completion edac_memctrl_kobj_complete;
122
123/*
124 * /sys/devices/system/edac/mc;
125 * data structures and methods
126 */
127static ssize_t memctrl_int_show(void *ptr, char *buffer)
128{
129 int *value = (int*) ptr;
130 return sprintf(buffer, "%u\n", *value);
131}
132
133static ssize_t memctrl_int_store(void *ptr, const char *buffer, size_t count)
134{
135 int *value = (int*) ptr;
136
137 if (isdigit(*buffer))
138 *value = simple_strtoul(buffer, NULL, 0);
139
140 return count;
141}
142
143struct memctrl_dev_attribute {
144 struct attribute attr;
145 void *value;
146 ssize_t (*show)(void *,char *);
147 ssize_t (*store)(void *, const char *, size_t);
148};
149
150/* Set of show/store abstract level functions for memory control object */
151static ssize_t memctrl_dev_show(struct kobject *kobj,
152 struct attribute *attr, char *buffer)
153{
154 struct memctrl_dev_attribute *memctrl_dev;
155 memctrl_dev = (struct memctrl_dev_attribute*)attr;
156
157 if (memctrl_dev->show)
158 return memctrl_dev->show(memctrl_dev->value, buffer);
159
160 return -EIO;
161}
162
163static ssize_t memctrl_dev_store(struct kobject *kobj, struct attribute *attr,
164 const char *buffer, size_t count)
165{
166 struct memctrl_dev_attribute *memctrl_dev;
167 memctrl_dev = (struct memctrl_dev_attribute*)attr;
168
169 if (memctrl_dev->store)
170 return memctrl_dev->store(memctrl_dev->value, buffer, count);
171
172 return -EIO;
173}
174
175static struct sysfs_ops memctrlfs_ops = {
176 .show = memctrl_dev_show,
177 .store = memctrl_dev_store
178};
179
180#define MEMCTRL_ATTR(_name,_mode,_show,_store) \
181static struct memctrl_dev_attribute attr_##_name = { \
182 .attr = {.name = __stringify(_name), .mode = _mode }, \
183 .value = &_name, \
184 .show = _show, \
185 .store = _store, \
186};
187
188#define MEMCTRL_STRING_ATTR(_name,_data,_mode,_show,_store) \
189static struct memctrl_dev_attribute attr_##_name = { \
190 .attr = {.name = __stringify(_name), .mode = _mode }, \
191 .value = _data, \
192 .show = _show, \
193 .store = _store, \
194};
195
196/* csrow<id> control files */
197MEMCTRL_ATTR(panic_on_ue,S_IRUGO|S_IWUSR,memctrl_int_show,memctrl_int_store);
198MEMCTRL_ATTR(log_ue,S_IRUGO|S_IWUSR,memctrl_int_show,memctrl_int_store);
199MEMCTRL_ATTR(log_ce,S_IRUGO|S_IWUSR,memctrl_int_show,memctrl_int_store);
200MEMCTRL_ATTR(poll_msec,S_IRUGO|S_IWUSR,memctrl_int_show,memctrl_int_store);
201
202/* Base Attributes of the memory ECC object */
203static struct memctrl_dev_attribute *memctrl_attr[] = {
204 &attr_panic_on_ue,
205 &attr_log_ue,
206 &attr_log_ce,
207 &attr_poll_msec,
208 NULL,
209};
210
211/* Main MC kobject release() function */
212static void edac_memctrl_master_release(struct kobject *kobj)
213{
214 debugf1("%s()\n", __func__);
215 complete(&edac_memctrl_kobj_complete);
216}
217
218static struct kobj_type ktype_memctrl = {
219 .release = edac_memctrl_master_release,
220 .sysfs_ops = &memctrlfs_ops,
221 .default_attrs = (struct attribute **) memctrl_attr,
222};
223
224/* Initialize the main sysfs entries for edac:
225 * /sys/devices/system/edac
226 *
227 * and children
228 *
229 * Return: 0 SUCCESS
230 * !0 FAILURE
231 */
232static int edac_sysfs_memctrl_setup(void)
233{
234 int err = 0;
235
236 debugf1("%s()\n", __func__);
237
238 /* create the /sys/devices/system/edac directory */
239 err = sysdev_class_register(&edac_class);
240
241 if (err) {
242 debugf1("%s() error=%d\n", __func__, err);
243 return err;
244 }
245
246 /* Init the MC's kobject */
247 memset(&edac_memctrl_kobj, 0, sizeof (edac_memctrl_kobj));
248 edac_memctrl_kobj.parent = &edac_class.kset.kobj;
249 edac_memctrl_kobj.ktype = &ktype_memctrl;
250
251 /* generate sysfs "..../edac/mc" */
252 err = kobject_set_name(&edac_memctrl_kobj,"mc");
253
254 if (err)
255 goto fail;
256
257 /* FIXME: maybe new sysdev_create_subdir() */
258 err = kobject_register(&edac_memctrl_kobj);
259
260 if (err) {
261 debugf1("Failed to register '.../edac/mc'\n");
262 goto fail;
263 }
264
265 debugf1("Registered '.../edac/mc' kobject\n");
266
267 return 0;
268
269fail:
270 sysdev_class_unregister(&edac_class);
271 return err;
272}
273
274/*
275 * MC teardown:
276 * the '..../edac/mc' kobject followed by '..../edac' itself
277 */
278static void edac_sysfs_memctrl_teardown(void)
279{
280 debugf0("MC: " __FILE__ ": %s()\n", __func__);
281
282 /* Unregister the MC's kobject and wait for reference count to reach
283 * 0.
284 */
285 init_completion(&edac_memctrl_kobj_complete);
286 kobject_unregister(&edac_memctrl_kobj);
287 wait_for_completion(&edac_memctrl_kobj_complete);
288
289 /* Unregister the 'edac' object */
290 sysdev_class_unregister(&edac_class);
291}
292
293#ifdef CONFIG_PCI
294static ssize_t edac_pci_int_show(void *ptr, char *buffer)
295{
296 int *value = ptr;
297 return sprintf(buffer,"%d\n",*value);
298}
299
300static ssize_t edac_pci_int_store(void *ptr, const char *buffer, size_t count)
301{
302 int *value = ptr;
303
304 if (isdigit(*buffer))
305 *value = simple_strtoul(buffer,NULL,0);
306
307 return count;
308}
309
310struct edac_pci_dev_attribute {
311 struct attribute attr;
312 void *value;
313 ssize_t (*show)(void *,char *);
314 ssize_t (*store)(void *, const char *,size_t);
315};
316
317/* Set of show/store abstract level functions for PCI Parity object */
318static ssize_t edac_pci_dev_show(struct kobject *kobj, struct attribute *attr,
319 char *buffer)
320{
321 struct edac_pci_dev_attribute *edac_pci_dev;
322 edac_pci_dev= (struct edac_pci_dev_attribute*)attr;
323
324 if (edac_pci_dev->show)
325 return edac_pci_dev->show(edac_pci_dev->value, buffer);
326 return -EIO;
327}
328
329static ssize_t edac_pci_dev_store(struct kobject *kobj,
330 struct attribute *attr, const char *buffer, size_t count)
331{
332 struct edac_pci_dev_attribute *edac_pci_dev;
333 edac_pci_dev= (struct edac_pci_dev_attribute*)attr;
334
335 if (edac_pci_dev->show)
336 return edac_pci_dev->store(edac_pci_dev->value, buffer, count);
337 return -EIO;
338}
339
340static struct sysfs_ops edac_pci_sysfs_ops = {
341 .show = edac_pci_dev_show,
342 .store = edac_pci_dev_store
343};
344
345#define EDAC_PCI_ATTR(_name,_mode,_show,_store) \
346static struct edac_pci_dev_attribute edac_pci_attr_##_name = { \
347 .attr = {.name = __stringify(_name), .mode = _mode }, \
348 .value = &_name, \
349 .show = _show, \
350 .store = _store, \
351};
352
353#define EDAC_PCI_STRING_ATTR(_name,_data,_mode,_show,_store) \
354static struct edac_pci_dev_attribute edac_pci_attr_##_name = { \
355 .attr = {.name = __stringify(_name), .mode = _mode }, \
356 .value = _data, \
357 .show = _show, \
358 .store = _store, \
359};
360
361/* PCI Parity control files */
362EDAC_PCI_ATTR(check_pci_parity, S_IRUGO|S_IWUSR, edac_pci_int_show,
363 edac_pci_int_store);
364EDAC_PCI_ATTR(panic_on_pci_parity, S_IRUGO|S_IWUSR, edac_pci_int_show,
365 edac_pci_int_store);
366EDAC_PCI_ATTR(pci_parity_count, S_IRUGO, edac_pci_int_show, NULL);
367
368/* Base Attributes of the memory ECC object */
369static struct edac_pci_dev_attribute *edac_pci_attr[] = {
370 &edac_pci_attr_check_pci_parity,
371 &edac_pci_attr_panic_on_pci_parity,
372 &edac_pci_attr_pci_parity_count,
373 NULL,
374};
375
376/* No memory to release */
377static void edac_pci_release(struct kobject *kobj)
378{
379 debugf1("%s()\n", __func__);
380 complete(&edac_pci_kobj_complete);
381}
382
383static struct kobj_type ktype_edac_pci = {
384 .release = edac_pci_release,
385 .sysfs_ops = &edac_pci_sysfs_ops,
386 .default_attrs = (struct attribute **) edac_pci_attr,
387};
388
389/**
390 * edac_sysfs_pci_setup()
391 *
392 */
393static int edac_sysfs_pci_setup(void)
394{
395 int err;
396
397 debugf1("%s()\n", __func__);
398
399 memset(&edac_pci_kobj, 0, sizeof(edac_pci_kobj));
400 edac_pci_kobj.parent = &edac_class.kset.kobj;
401 edac_pci_kobj.ktype = &ktype_edac_pci;
402 err = kobject_set_name(&edac_pci_kobj, "pci");
403
404 if (!err) {
405 /* Instanstiate the csrow object */
406 /* FIXME: maybe new sysdev_create_subdir() */
407 err = kobject_register(&edac_pci_kobj);
408
409 if (err)
410 debugf1("Failed to register '.../edac/pci'\n");
411 else
412 debugf1("Registered '.../edac/pci' kobject\n");
413 }
414
415 return err;
416}
417
418static void edac_sysfs_pci_teardown(void)
419{
420 debugf0("%s()\n", __func__);
421 init_completion(&edac_pci_kobj_complete);
422 kobject_unregister(&edac_pci_kobj);
423 wait_for_completion(&edac_pci_kobj_complete);
424}
425
426
427static u16 get_pci_parity_status(struct pci_dev *dev, int secondary)
428{
429 int where;
430 u16 status;
431
432 where = secondary ? PCI_SEC_STATUS : PCI_STATUS;
433 pci_read_config_word(dev, where, &status);
434
435 /* If we get back 0xFFFF then we must suspect that the card has been
436 * pulled but the Linux PCI layer has not yet finished cleaning up.
437 * We don't want to report on such devices
438 */
439
440 if (status == 0xFFFF) {
441 u32 sanity;
442
443 pci_read_config_dword(dev, 0, &sanity);
444
445 if (sanity == 0xFFFFFFFF)
446 return 0;
447 }
448
449 status &= PCI_STATUS_DETECTED_PARITY | PCI_STATUS_SIG_SYSTEM_ERROR |
450 PCI_STATUS_PARITY;
451
452 if (status)
453 /* reset only the bits we are interested in */
454 pci_write_config_word(dev, where, status);
455
456 return status;
457}
458
459typedef void (*pci_parity_check_fn_t) (struct pci_dev *dev);
460
461/* Clear any PCI parity errors logged by this device. */
462static void edac_pci_dev_parity_clear(struct pci_dev *dev)
463{
464 u8 header_type;
465
466 get_pci_parity_status(dev, 0);
467
468 /* read the device TYPE, looking for bridges */
469 pci_read_config_byte(dev, PCI_HEADER_TYPE, &header_type);
470
471 if ((header_type & 0x7F) == PCI_HEADER_TYPE_BRIDGE)
472 get_pci_parity_status(dev, 1);
473}
474
475/*
476 * PCI Parity polling
477 *
478 */
479static void edac_pci_dev_parity_test(struct pci_dev *dev)
480{
481 u16 status;
482 u8 header_type;
483
484 /* read the STATUS register on this device
485 */
486 status = get_pci_parity_status(dev, 0);
487
488 debugf2("PCI STATUS= 0x%04x %s\n", status, dev->dev.bus_id );
489
490 /* check the status reg for errors */
491 if (status) {
492 if (status & (PCI_STATUS_SIG_SYSTEM_ERROR))
493 edac_printk(KERN_CRIT, EDAC_PCI,
494 "Signaled System Error on %s\n",
495 pci_name(dev));
496
497 if (status & (PCI_STATUS_PARITY)) {
498 edac_printk(KERN_CRIT, EDAC_PCI,
499 "Master Data Parity Error on %s\n",
500 pci_name(dev));
501
502 atomic_inc(&pci_parity_count);
503 }
504
505 if (status & (PCI_STATUS_DETECTED_PARITY)) {
506 edac_printk(KERN_CRIT, EDAC_PCI,
507 "Detected Parity Error on %s\n",
508 pci_name(dev));
509
510 atomic_inc(&pci_parity_count);
511 }
512 }
513
514 /* read the device TYPE, looking for bridges */
515 pci_read_config_byte(dev, PCI_HEADER_TYPE, &header_type);
516
517 debugf2("PCI HEADER TYPE= 0x%02x %s\n", header_type, dev->dev.bus_id );
518
519 if ((header_type & 0x7F) == PCI_HEADER_TYPE_BRIDGE) {
520 /* On bridges, need to examine secondary status register */
521 status = get_pci_parity_status(dev, 1);
522
523 debugf2("PCI SEC_STATUS= 0x%04x %s\n",
524 status, dev->dev.bus_id );
525
526 /* check the secondary status reg for errors */
527 if (status) {
528 if (status & (PCI_STATUS_SIG_SYSTEM_ERROR))
529 edac_printk(KERN_CRIT, EDAC_PCI, "Bridge "
530 "Signaled System Error on %s\n",
531 pci_name(dev));
532
533 if (status & (PCI_STATUS_PARITY)) {
534 edac_printk(KERN_CRIT, EDAC_PCI, "Bridge "
535 "Master Data Parity Error on "
536 "%s\n", pci_name(dev));
537
538 atomic_inc(&pci_parity_count);
539 }
540
541 if (status & (PCI_STATUS_DETECTED_PARITY)) {
542 edac_printk(KERN_CRIT, EDAC_PCI, "Bridge "
543 "Detected Parity Error on %s\n",
544 pci_name(dev));
545
546 atomic_inc(&pci_parity_count);
547 }
548 }
549 }
550}
551
552/*
553 * pci_dev parity list iterator
554 * Scan the PCI device list for one iteration, looking for SERRORs
555 * Master Parity ERRORS or Parity ERRORs on primary or secondary devices
556 */
557static inline void edac_pci_dev_parity_iterator(pci_parity_check_fn_t fn)
558{
559 struct pci_dev *dev = NULL;
560
561 /* request for kernel access to the next PCI device, if any,
562 * and while we are looking at it have its reference count
563 * bumped until we are done with it
564 */
565 while((dev = pci_get_device(PCI_ANY_ID, PCI_ANY_ID, dev)) != NULL) {
566 fn(dev);
567 }
568}
569
570static void do_pci_parity_check(void)
571{
572 unsigned long flags;
573 int before_count;
574
575 debugf3("%s()\n", __func__);
576
577 if (!check_pci_parity)
578 return;
579
580 before_count = atomic_read(&pci_parity_count);
581
582 /* scan all PCI devices looking for a Parity Error on devices and
583 * bridges
584 */
585 local_irq_save(flags);
586 edac_pci_dev_parity_iterator(edac_pci_dev_parity_test);
587 local_irq_restore(flags);
588
589 /* Only if operator has selected panic on PCI Error */
590 if (panic_on_pci_parity) {
591 /* If the count is different 'after' from 'before' */
592 if (before_count != atomic_read(&pci_parity_count))
593 panic("EDAC: PCI Parity Error");
594 }
595}
596
597static inline void clear_pci_parity_errors(void)
598{
599 /* Clear any PCI bus parity errors that devices initially have logged
600 * in their registers.
601 */
602 edac_pci_dev_parity_iterator(edac_pci_dev_parity_clear);
603}
604
605#else /* CONFIG_PCI */
606
607/* pre-process these away */
608#define do_pci_parity_check()
609#define clear_pci_parity_errors()
610#define edac_sysfs_pci_teardown()
611#define edac_sysfs_pci_setup() (0)
612
613#endif /* CONFIG_PCI */
614
615/* EDAC sysfs CSROW data structures and methods
616 */
617
618/* Set of more default csrow<id> attribute show/store functions */
619static ssize_t csrow_ue_count_show(struct csrow_info *csrow, char *data, int private)
620{
621 return sprintf(data,"%u\n", csrow->ue_count);
622}
623
624static ssize_t csrow_ce_count_show(struct csrow_info *csrow, char *data, int private)
625{
626 return sprintf(data,"%u\n", csrow->ce_count);
627}
628
629static ssize_t csrow_size_show(struct csrow_info *csrow, char *data, int private)
630{
631 return sprintf(data,"%u\n", PAGES_TO_MiB(csrow->nr_pages));
632}
633
634static ssize_t csrow_mem_type_show(struct csrow_info *csrow, char *data, int private)
635{
636 return sprintf(data,"%s\n", mem_types[csrow->mtype]);
637}
638
639static ssize_t csrow_dev_type_show(struct csrow_info *csrow, char *data, int private)
640{
641 return sprintf(data,"%s\n", dev_types[csrow->dtype]);
642}
643
644static ssize_t csrow_edac_mode_show(struct csrow_info *csrow, char *data, int private)
645{
646 return sprintf(data,"%s\n", edac_caps[csrow->edac_mode]);
647}
648
649/* show/store functions for DIMM Label attributes */
650static ssize_t channel_dimm_label_show(struct csrow_info *csrow,
651 char *data, int channel)
652{
653 return snprintf(data, EDAC_MC_LABEL_LEN,"%s",
654 csrow->channels[channel].label);
655}
656
657static ssize_t channel_dimm_label_store(struct csrow_info *csrow,
658 const char *data,
659 size_t count,
660 int channel)
661{
662 ssize_t max_size = 0;
663
664 max_size = min((ssize_t)count,(ssize_t)EDAC_MC_LABEL_LEN-1);
665 strncpy(csrow->channels[channel].label, data, max_size);
666 csrow->channels[channel].label[max_size] = '\0';
667
668 return max_size;
669}
670
671/* show function for dynamic chX_ce_count attribute */
672static ssize_t channel_ce_count_show(struct csrow_info *csrow,
673 char *data,
674 int channel)
675{
676 return sprintf(data, "%u\n", csrow->channels[channel].ce_count);
677}
678
679/* csrow specific attribute structure */
680struct csrowdev_attribute {
681 struct attribute attr;
682 ssize_t (*show)(struct csrow_info *,char *,int);
683 ssize_t (*store)(struct csrow_info *, const char *,size_t,int);
684 int private;
685};
686
687#define to_csrow(k) container_of(k, struct csrow_info, kobj)
688#define to_csrowdev_attr(a) container_of(a, struct csrowdev_attribute, attr)
689
690/* Set of show/store higher level functions for default csrow attributes */
691static ssize_t csrowdev_show(struct kobject *kobj,
692 struct attribute *attr,
693 char *buffer)
694{
695 struct csrow_info *csrow = to_csrow(kobj);
696 struct csrowdev_attribute *csrowdev_attr = to_csrowdev_attr(attr);
697
698 if (csrowdev_attr->show)
699 return csrowdev_attr->show(csrow,
700 buffer,
701 csrowdev_attr->private);
702 return -EIO;
703}
704
705static ssize_t csrowdev_store(struct kobject *kobj, struct attribute *attr,
706 const char *buffer, size_t count)
707{
708 struct csrow_info *csrow = to_csrow(kobj);
709 struct csrowdev_attribute * csrowdev_attr = to_csrowdev_attr(attr);
710
711 if (csrowdev_attr->store)
712 return csrowdev_attr->store(csrow,
713 buffer,
714 count,
715 csrowdev_attr->private);
716 return -EIO;
717}
718
719static struct sysfs_ops csrowfs_ops = {
720 .show = csrowdev_show,
721 .store = csrowdev_store
722};
723
724#define CSROWDEV_ATTR(_name,_mode,_show,_store,_private) \
725static struct csrowdev_attribute attr_##_name = { \
726 .attr = {.name = __stringify(_name), .mode = _mode }, \
727 .show = _show, \
728 .store = _store, \
729 .private = _private, \
730};
731
732/* default cwrow<id>/attribute files */
733CSROWDEV_ATTR(size_mb,S_IRUGO,csrow_size_show,NULL,0);
734CSROWDEV_ATTR(dev_type,S_IRUGO,csrow_dev_type_show,NULL,0);
735CSROWDEV_ATTR(mem_type,S_IRUGO,csrow_mem_type_show,NULL,0);
736CSROWDEV_ATTR(edac_mode,S_IRUGO,csrow_edac_mode_show,NULL,0);
737CSROWDEV_ATTR(ue_count,S_IRUGO,csrow_ue_count_show,NULL,0);
738CSROWDEV_ATTR(ce_count,S_IRUGO,csrow_ce_count_show,NULL,0);
739
740/* default attributes of the CSROW<id> object */
741static struct csrowdev_attribute *default_csrow_attr[] = {
742 &attr_dev_type,
743 &attr_mem_type,
744 &attr_edac_mode,
745 &attr_size_mb,
746 &attr_ue_count,
747 &attr_ce_count,
748 NULL,
749};
750
751
752/* possible dynamic channel DIMM Label attribute files */
753CSROWDEV_ATTR(ch0_dimm_label,S_IRUGO|S_IWUSR,
754 channel_dimm_label_show,
755 channel_dimm_label_store,
756 0 );
757CSROWDEV_ATTR(ch1_dimm_label,S_IRUGO|S_IWUSR,
758 channel_dimm_label_show,
759 channel_dimm_label_store,
760 1 );
761CSROWDEV_ATTR(ch2_dimm_label,S_IRUGO|S_IWUSR,
762 channel_dimm_label_show,
763 channel_dimm_label_store,
764 2 );
765CSROWDEV_ATTR(ch3_dimm_label,S_IRUGO|S_IWUSR,
766 channel_dimm_label_show,
767 channel_dimm_label_store,
768 3 );
769CSROWDEV_ATTR(ch4_dimm_label,S_IRUGO|S_IWUSR,
770 channel_dimm_label_show,
771 channel_dimm_label_store,
772 4 );
773CSROWDEV_ATTR(ch5_dimm_label,S_IRUGO|S_IWUSR,
774 channel_dimm_label_show,
775 channel_dimm_label_store,
776 5 );
777
778/* Total possible dynamic DIMM Label attribute file table */
779static struct csrowdev_attribute *dynamic_csrow_dimm_attr[] = {
780 &attr_ch0_dimm_label,
781 &attr_ch1_dimm_label,
782 &attr_ch2_dimm_label,
783 &attr_ch3_dimm_label,
784 &attr_ch4_dimm_label,
785 &attr_ch5_dimm_label
786};
787
788/* possible dynamic channel ce_count attribute files */
789CSROWDEV_ATTR(ch0_ce_count,S_IRUGO|S_IWUSR,
790 channel_ce_count_show,
791 NULL,
792 0 );
793CSROWDEV_ATTR(ch1_ce_count,S_IRUGO|S_IWUSR,
794 channel_ce_count_show,
795 NULL,
796 1 );
797CSROWDEV_ATTR(ch2_ce_count,S_IRUGO|S_IWUSR,
798 channel_ce_count_show,
799 NULL,
800 2 );
801CSROWDEV_ATTR(ch3_ce_count,S_IRUGO|S_IWUSR,
802 channel_ce_count_show,
803 NULL,
804 3 );
805CSROWDEV_ATTR(ch4_ce_count,S_IRUGO|S_IWUSR,
806 channel_ce_count_show,
807 NULL,
808 4 );
809CSROWDEV_ATTR(ch5_ce_count,S_IRUGO|S_IWUSR,
810 channel_ce_count_show,
811 NULL,
812 5 );
813
814/* Total possible dynamic ce_count attribute file table */
815static struct csrowdev_attribute *dynamic_csrow_ce_count_attr[] = {
816 &attr_ch0_ce_count,
817 &attr_ch1_ce_count,
818 &attr_ch2_ce_count,
819 &attr_ch3_ce_count,
820 &attr_ch4_ce_count,
821 &attr_ch5_ce_count
822};
823
824
825#define EDAC_NR_CHANNELS 6
826
827/* Create dynamic CHANNEL files, indexed by 'chan', under specifed CSROW */
828static int edac_create_channel_files(struct kobject *kobj, int chan)
829{
830 int err=-ENODEV;
831
832 if (chan >= EDAC_NR_CHANNELS)
833 return err;
834
835 /* create the DIMM label attribute file */
836 err = sysfs_create_file(kobj,
837 (struct attribute *) dynamic_csrow_dimm_attr[chan]);
838
839 if (!err) {
840 /* create the CE Count attribute file */
841 err = sysfs_create_file(kobj,
842 (struct attribute *) dynamic_csrow_ce_count_attr[chan]);
843 } else {
844 debugf1("%s() dimm labels and ce_count files created", __func__);
845 }
846
847 return err;
848}
849
850/* No memory to release for this kobj */
851static void edac_csrow_instance_release(struct kobject *kobj)
852{
853 struct csrow_info *cs;
854
855 cs = container_of(kobj, struct csrow_info, kobj);
856 complete(&cs->kobj_complete);
857}
858
859/* the kobj_type instance for a CSROW */
860static struct kobj_type ktype_csrow = {
861 .release = edac_csrow_instance_release,
862 .sysfs_ops = &csrowfs_ops,
863 .default_attrs = (struct attribute **) default_csrow_attr,
864};
865
866/* Create a CSROW object under specifed edac_mc_device */
867static int edac_create_csrow_object(
868 struct kobject *edac_mci_kobj,
869 struct csrow_info *csrow,
870 int index)
871{
872 int err = 0;
873 int chan;
874
875 memset(&csrow->kobj, 0, sizeof(csrow->kobj));
876
877 /* generate ..../edac/mc/mc<id>/csrow<index> */
878
879 csrow->kobj.parent = edac_mci_kobj;
880 csrow->kobj.ktype = &ktype_csrow;
881
882 /* name this instance of csrow<id> */
883 err = kobject_set_name(&csrow->kobj,"csrow%d",index);
884 if (err)
885 goto error_exit;
886
887 /* Instanstiate the csrow object */
888 err = kobject_register(&csrow->kobj);
889 if (!err) {
890 /* Create the dyanmic attribute files on this csrow,
891 * namely, the DIMM labels and the channel ce_count
892 */
893 for (chan = 0; chan < csrow->nr_channels; chan++) {
894 err = edac_create_channel_files(&csrow->kobj,chan);
895 if (err)
896 break;
897 }
898 }
899
900error_exit:
901 return err;
902}
903
904/* default sysfs methods and data structures for the main MCI kobject */
905
906static ssize_t mci_reset_counters_store(struct mem_ctl_info *mci,
907 const char *data, size_t count)
908{
909 int row, chan;
910
911 mci->ue_noinfo_count = 0;
912 mci->ce_noinfo_count = 0;
913 mci->ue_count = 0;
914 mci->ce_count = 0;
915
916 for (row = 0; row < mci->nr_csrows; row++) {
917 struct csrow_info *ri = &mci->csrows[row];
918
919 ri->ue_count = 0;
920 ri->ce_count = 0;
921
922 for (chan = 0; chan < ri->nr_channels; chan++)
923 ri->channels[chan].ce_count = 0;
924 }
925
926 mci->start_time = jiffies;
927 return count;
928}
929
930/* memory scrubbing */
931static ssize_t mci_sdram_scrub_rate_store(struct mem_ctl_info *mci,
932 const char *data, size_t count)
933{
934 u32 bandwidth = -1;
935
936 if (mci->set_sdram_scrub_rate) {
937
938 memctrl_int_store(&bandwidth, data, count);
939
940 if (!(*mci->set_sdram_scrub_rate)(mci, &bandwidth)) {
941 edac_printk(KERN_DEBUG, EDAC_MC,
942 "Scrub rate set successfully, applied: %d\n",
943 bandwidth);
944 } else {
945 /* FIXME: error codes maybe? */
946 edac_printk(KERN_DEBUG, EDAC_MC,
947 "Scrub rate set FAILED, could not apply: %d\n",
948 bandwidth);
949 }
950 } else {
951 /* FIXME: produce "not implemented" ERROR for user-side. */
952 edac_printk(KERN_WARNING, EDAC_MC,
953 "Memory scrubbing 'set'control is not implemented!\n");
954 }
955 return count;
956}
957
958static ssize_t mci_sdram_scrub_rate_show(struct mem_ctl_info *mci, char *data)
959{
960 u32 bandwidth = -1;
961
962 if (mci->get_sdram_scrub_rate) {
963 if (!(*mci->get_sdram_scrub_rate)(mci, &bandwidth)) {
964 edac_printk(KERN_DEBUG, EDAC_MC,
965 "Scrub rate successfully, fetched: %d\n",
966 bandwidth);
967 } else {
968 /* FIXME: error codes maybe? */
969 edac_printk(KERN_DEBUG, EDAC_MC,
970 "Scrub rate fetch FAILED, got: %d\n",
971 bandwidth);
972 }
973 } else {
974 /* FIXME: produce "not implemented" ERROR for user-side. */
975 edac_printk(KERN_WARNING, EDAC_MC,
976 "Memory scrubbing 'get' control is not implemented!\n");
977 }
978 return sprintf(data, "%d\n", bandwidth);
979}
980
981/* default attribute files for the MCI object */
982static ssize_t mci_ue_count_show(struct mem_ctl_info *mci, char *data)
983{
984 return sprintf(data,"%d\n", mci->ue_count);
985}
986
987static ssize_t mci_ce_count_show(struct mem_ctl_info *mci, char *data)
988{
989 return sprintf(data,"%d\n", mci->ce_count);
990}
991
992static ssize_t mci_ce_noinfo_show(struct mem_ctl_info *mci, char *data)
993{
994 return sprintf(data,"%d\n", mci->ce_noinfo_count);
995}
996
997static ssize_t mci_ue_noinfo_show(struct mem_ctl_info *mci, char *data)
998{
999 return sprintf(data,"%d\n", mci->ue_noinfo_count);
1000}
1001
1002static ssize_t mci_seconds_show(struct mem_ctl_info *mci, char *data)
1003{
1004 return sprintf(data,"%ld\n", (jiffies - mci->start_time) / HZ);
1005}
1006
1007static ssize_t mci_ctl_name_show(struct mem_ctl_info *mci, char *data)
1008{
1009 return sprintf(data,"%s\n", mci->ctl_name);
1010}
1011
1012static ssize_t mci_size_mb_show(struct mem_ctl_info *mci, char *data)
1013{
1014 int total_pages, csrow_idx;
1015
1016 for (total_pages = csrow_idx = 0; csrow_idx < mci->nr_csrows;
1017 csrow_idx++) {
1018 struct csrow_info *csrow = &mci->csrows[csrow_idx];
1019
1020 if (!csrow->nr_pages)
1021 continue;
1022
1023 total_pages += csrow->nr_pages;
1024 }
1025
1026 return sprintf(data,"%u\n", PAGES_TO_MiB(total_pages));
1027}
1028
1029struct mcidev_attribute {
1030 struct attribute attr;
1031 ssize_t (*show)(struct mem_ctl_info *,char *);
1032 ssize_t (*store)(struct mem_ctl_info *, const char *,size_t);
1033};
1034
1035#define to_mci(k) container_of(k, struct mem_ctl_info, edac_mci_kobj)
1036#define to_mcidev_attr(a) container_of(a, struct mcidev_attribute, attr)
1037
1038/* MCI show/store functions for top most object */
1039static ssize_t mcidev_show(struct kobject *kobj, struct attribute *attr,
1040 char *buffer)
1041{
1042 struct mem_ctl_info *mem_ctl_info = to_mci(kobj);
1043 struct mcidev_attribute * mcidev_attr = to_mcidev_attr(attr);
1044
1045 if (mcidev_attr->show)
1046 return mcidev_attr->show(mem_ctl_info, buffer);
1047
1048 return -EIO;
1049}
1050
1051static ssize_t mcidev_store(struct kobject *kobj, struct attribute *attr,
1052 const char *buffer, size_t count)
1053{
1054 struct mem_ctl_info *mem_ctl_info = to_mci(kobj);
1055 struct mcidev_attribute * mcidev_attr = to_mcidev_attr(attr);
1056
1057 if (mcidev_attr->store)
1058 return mcidev_attr->store(mem_ctl_info, buffer, count);
1059
1060 return -EIO;
1061}
1062
1063static struct sysfs_ops mci_ops = {
1064 .show = mcidev_show,
1065 .store = mcidev_store
1066};
1067
1068#define MCIDEV_ATTR(_name,_mode,_show,_store) \
1069static struct mcidev_attribute mci_attr_##_name = { \
1070 .attr = {.name = __stringify(_name), .mode = _mode }, \
1071 .show = _show, \
1072 .store = _store, \
1073};
1074
1075/* default Control file */
1076MCIDEV_ATTR(reset_counters,S_IWUSR,NULL,mci_reset_counters_store);
1077
1078/* default Attribute files */
1079MCIDEV_ATTR(mc_name,S_IRUGO,mci_ctl_name_show,NULL);
1080MCIDEV_ATTR(size_mb,S_IRUGO,mci_size_mb_show,NULL);
1081MCIDEV_ATTR(seconds_since_reset,S_IRUGO,mci_seconds_show,NULL);
1082MCIDEV_ATTR(ue_noinfo_count,S_IRUGO,mci_ue_noinfo_show,NULL);
1083MCIDEV_ATTR(ce_noinfo_count,S_IRUGO,mci_ce_noinfo_show,NULL);
1084MCIDEV_ATTR(ue_count,S_IRUGO,mci_ue_count_show,NULL);
1085MCIDEV_ATTR(ce_count,S_IRUGO,mci_ce_count_show,NULL);
1086
1087/* memory scrubber attribute file */
1088MCIDEV_ATTR(sdram_scrub_rate,S_IRUGO|S_IWUSR,mci_sdram_scrub_rate_show,mci_sdram_scrub_rate_store);
1089
1090static struct mcidev_attribute *mci_attr[] = {
1091 &mci_attr_reset_counters,
1092 &mci_attr_mc_name,
1093 &mci_attr_size_mb,
1094 &mci_attr_seconds_since_reset,
1095 &mci_attr_ue_noinfo_count,
1096 &mci_attr_ce_noinfo_count,
1097 &mci_attr_ue_count,
1098 &mci_attr_ce_count,
1099 &mci_attr_sdram_scrub_rate,
1100 NULL
1101};
1102
1103/*
1104 * Release of a MC controlling instance
1105 */
1106static void edac_mci_instance_release(struct kobject *kobj)
1107{
1108 struct mem_ctl_info *mci;
1109
1110 mci = to_mci(kobj);
1111 debugf0("%s() idx=%d\n", __func__, mci->mc_idx);
1112 complete(&mci->kobj_complete);
1113}
1114
1115static struct kobj_type ktype_mci = {
1116 .release = edac_mci_instance_release,
1117 .sysfs_ops = &mci_ops,
1118 .default_attrs = (struct attribute **) mci_attr,
1119};
1120
1121
1122#define EDAC_DEVICE_SYMLINK "device"
1123
1124/*
1125 * Create a new Memory Controller kobject instance,
1126 * mc<id> under the 'mc' directory
1127 *
1128 * Return:
1129 * 0 Success
1130 * !0 Failure
1131 */
1132static int edac_create_sysfs_mci_device(struct mem_ctl_info *mci)
1133{
1134 int i;
1135 int err;
1136 struct csrow_info *csrow;
1137 struct kobject *edac_mci_kobj=&mci->edac_mci_kobj;
1138
1139 debugf0("%s() idx=%d\n", __func__, mci->mc_idx);
1140 memset(edac_mci_kobj, 0, sizeof(*edac_mci_kobj));
1141
1142 /* set the name of the mc<id> object */
1143 err = kobject_set_name(edac_mci_kobj,"mc%d",mci->mc_idx);
1144 if (err)
1145 return err;
1146
1147 /* link to our parent the '..../edac/mc' object */
1148 edac_mci_kobj->parent = &edac_memctrl_kobj;
1149 edac_mci_kobj->ktype = &ktype_mci;
1150
1151 /* register the mc<id> kobject */
1152 err = kobject_register(edac_mci_kobj);
1153 if (err)
1154 return err;
1155
1156 /* create a symlink for the device */
1157 err = sysfs_create_link(edac_mci_kobj, &mci->dev->kobj,
1158 EDAC_DEVICE_SYMLINK);
1159 if (err)
1160 goto fail0;
1161
1162 /* Make directories for each CSROW object
1163 * under the mc<id> kobject
1164 */
1165 for (i = 0; i < mci->nr_csrows; i++) {
1166 csrow = &mci->csrows[i];
1167
1168 /* Only expose populated CSROWs */
1169 if (csrow->nr_pages > 0) {
1170 err = edac_create_csrow_object(edac_mci_kobj,csrow,i);
1171 if (err)
1172 goto fail1;
1173 }
1174 }
1175
1176 return 0;
1177
1178 /* CSROW error: backout what has already been registered, */
1179fail1:
1180 for ( i--; i >= 0; i--) {
1181 if (csrow->nr_pages > 0) {
1182 init_completion(&csrow->kobj_complete);
1183 kobject_unregister(&mci->csrows[i].kobj);
1184 wait_for_completion(&csrow->kobj_complete);
1185 }
1186 }
1187
1188fail0:
1189 init_completion(&mci->kobj_complete);
1190 kobject_unregister(edac_mci_kobj);
1191 wait_for_completion(&mci->kobj_complete);
1192 return err;
1193}
1194
1195/*
1196 * remove a Memory Controller instance
1197 */
1198static void edac_remove_sysfs_mci_device(struct mem_ctl_info *mci)
1199{
1200 int i;
1201
1202 debugf0("%s()\n", __func__);
1203
1204 /* remove all csrow kobjects */
1205 for (i = 0; i < mci->nr_csrows; i++) {
1206 if (mci->csrows[i].nr_pages > 0) {
1207 init_completion(&mci->csrows[i].kobj_complete);
1208 kobject_unregister(&mci->csrows[i].kobj);
1209 wait_for_completion(&mci->csrows[i].kobj_complete);
1210 }
1211 }
1212
1213 sysfs_remove_link(&mci->edac_mci_kobj, EDAC_DEVICE_SYMLINK);
1214 init_completion(&mci->kobj_complete);
1215 kobject_unregister(&mci->edac_mci_kobj);
1216 wait_for_completion(&mci->kobj_complete);
1217}
1218
1219/* END OF sysfs data and methods */
1220
1221#ifdef CONFIG_EDAC_DEBUG 41#ifdef CONFIG_EDAC_DEBUG
1222 42
1223static void edac_mc_dump_channel(struct channel_info *chan) 43static void edac_mc_dump_channel(struct channel_info *chan)
@@ -1672,7 +492,7 @@ void edac_mc_handle_ce(struct mem_ctl_info *mci,
1672 return; 492 return;
1673 } 493 }
1674 494
1675 if (log_ce) 495 if (edac_get_log_ce())
1676 /* FIXME - put in DIMM location */ 496 /* FIXME - put in DIMM location */
1677 edac_mc_printk(mci, KERN_WARNING, 497 edac_mc_printk(mci, KERN_WARNING,
1678 "CE page 0x%lx, offset 0x%lx, grain %d, syndrome " 498 "CE page 0x%lx, offset 0x%lx, grain %d, syndrome "
@@ -1707,7 +527,7 @@ EXPORT_SYMBOL_GPL(edac_mc_handle_ce);
1707 527
1708void edac_mc_handle_ce_no_info(struct mem_ctl_info *mci, const char *msg) 528void edac_mc_handle_ce_no_info(struct mem_ctl_info *mci, const char *msg)
1709{ 529{
1710 if (log_ce) 530 if (edac_get_log_ce())
1711 edac_mc_printk(mci, KERN_WARNING, 531 edac_mc_printk(mci, KERN_WARNING,
1712 "CE - no information available: %s\n", msg); 532 "CE - no information available: %s\n", msg);
1713 533
@@ -1751,14 +571,14 @@ void edac_mc_handle_ue(struct mem_ctl_info *mci,
1751 pos += chars; 571 pos += chars;
1752 } 572 }
1753 573
1754 if (log_ue) 574 if (edac_get_log_ue())
1755 edac_mc_printk(mci, KERN_EMERG, 575 edac_mc_printk(mci, KERN_EMERG,
1756 "UE page 0x%lx, offset 0x%lx, grain %d, row %d, " 576 "UE page 0x%lx, offset 0x%lx, grain %d, row %d, "
1757 "labels \"%s\": %s\n", page_frame_number, 577 "labels \"%s\": %s\n", page_frame_number,
1758 offset_in_page, mci->csrows[row].grain, row, labels, 578 offset_in_page, mci->csrows[row].grain, row, labels,
1759 msg); 579 msg);
1760 580
1761 if (panic_on_ue) 581 if (edac_get_panic_on_ue())
1762 panic("EDAC MC%d: UE page 0x%lx, offset 0x%lx, grain %d, " 582 panic("EDAC MC%d: UE page 0x%lx, offset 0x%lx, grain %d, "
1763 "row %d, labels \"%s\": %s\n", mci->mc_idx, 583 "row %d, labels \"%s\": %s\n", mci->mc_idx,
1764 page_frame_number, offset_in_page, 584 page_frame_number, offset_in_page,
@@ -1771,10 +591,10 @@ EXPORT_SYMBOL_GPL(edac_mc_handle_ue);
1771 591
1772void edac_mc_handle_ue_no_info(struct mem_ctl_info *mci, const char *msg) 592void edac_mc_handle_ue_no_info(struct mem_ctl_info *mci, const char *msg)
1773{ 593{
1774 if (panic_on_ue) 594 if (edac_get_panic_on_ue())
1775 panic("EDAC MC%d: Uncorrected Error", mci->mc_idx); 595 panic("EDAC MC%d: Uncorrected Error", mci->mc_idx);
1776 596
1777 if (log_ue) 597 if (edac_get_log_ue())
1778 edac_mc_printk(mci, KERN_WARNING, 598 edac_mc_printk(mci, KERN_WARNING,
1779 "UE - no information available: %s\n", msg); 599 "UE - no information available: %s\n", msg);
1780 mci->ue_noinfo_count++; 600 mci->ue_noinfo_count++;
@@ -1837,13 +657,13 @@ void edac_mc_handle_fbd_ue(struct mem_ctl_info *mci,
1837 chars = snprintf(pos, len + 1, "-%s", 657 chars = snprintf(pos, len + 1, "-%s",
1838 mci->csrows[csrow].channels[channelb].label); 658 mci->csrows[csrow].channels[channelb].label);
1839 659
1840 if (log_ue) 660 if (edac_get_log_ue())
1841 edac_mc_printk(mci, KERN_EMERG, 661 edac_mc_printk(mci, KERN_EMERG,
1842 "UE row %d, channel-a= %d channel-b= %d " 662 "UE row %d, channel-a= %d channel-b= %d "
1843 "labels \"%s\": %s\n", csrow, channela, channelb, 663 "labels \"%s\": %s\n", csrow, channela, channelb,
1844 labels, msg); 664 labels, msg);
1845 665
1846 if (panic_on_ue) 666 if (edac_get_panic_on_ue())
1847 panic("UE row %d, channel-a= %d channel-b= %d " 667 panic("UE row %d, channel-a= %d channel-b= %d "
1848 "labels \"%s\": %s\n", csrow, channela, 668 "labels \"%s\": %s\n", csrow, channela,
1849 channelb, labels, msg); 669 channelb, labels, msg);
@@ -1878,7 +698,7 @@ void edac_mc_handle_fbd_ce(struct mem_ctl_info *mci,
1878 return; 698 return;
1879 } 699 }
1880 700
1881 if (log_ce) 701 if (edac_get_log_ce())
1882 /* FIXME - put in DIMM location */ 702 /* FIXME - put in DIMM location */
1883 edac_mc_printk(mci, KERN_WARNING, 703 edac_mc_printk(mci, KERN_WARNING,
1884 "CE row %d, channel %d, label \"%s\": %s\n", 704 "CE row %d, channel %d, label \"%s\": %s\n",
@@ -1896,7 +716,7 @@ EXPORT_SYMBOL(edac_mc_handle_fbd_ce);
1896/* 716/*
1897 * Iterate over all MC instances and check for ECC, et al, errors 717 * Iterate over all MC instances and check for ECC, et al, errors
1898 */ 718 */
1899static inline void check_mc_devices(void) 719void edac_check_mc_devices(void)
1900{ 720{
1901 struct list_head *item; 721 struct list_head *item;
1902 struct mem_ctl_info *mci; 722 struct mem_ctl_info *mci;
@@ -1913,118 +733,3 @@ static inline void check_mc_devices(void)
1913 733
1914 up(&mem_ctls_mutex); 734 up(&mem_ctls_mutex);
1915} 735}
1916
1917/*
1918 * Check MC status every poll_msec.
1919 * Check PCI status every poll_msec as well.
1920 *
1921 * This where the work gets done for edac.
1922 *
1923 * SMP safe, doesn't use NMI, and auto-rate-limits.
1924 */
1925static void do_edac_check(void)
1926{
1927 debugf3("%s()\n", __func__);
1928 check_mc_devices();
1929 do_pci_parity_check();
1930}
1931
1932static int edac_kernel_thread(void *arg)
1933{
1934 set_freezable();
1935 while (!kthread_should_stop()) {
1936 do_edac_check();
1937
1938 /* goto sleep for the interval */
1939 schedule_timeout_interruptible((HZ * poll_msec) / 1000);
1940 try_to_freeze();
1941 }
1942
1943 return 0;
1944}
1945
1946/*
1947 * edac_mc_init
1948 * module initialization entry point
1949 */
1950static int __init edac_mc_init(void)
1951{
1952 edac_printk(KERN_INFO, EDAC_MC, EDAC_MC_VERSION "\n");
1953
1954 /*
1955 * Harvest and clear any boot/initialization PCI parity errors
1956 *
1957 * FIXME: This only clears errors logged by devices present at time of
1958 * module initialization. We should also do an initial clear
1959 * of each newly hotplugged device.
1960 */
1961 clear_pci_parity_errors();
1962
1963 /* Create the MC sysfs entries */
1964 if (edac_sysfs_memctrl_setup()) {
1965 edac_printk(KERN_ERR, EDAC_MC,
1966 "Error initializing sysfs code\n");
1967 return -ENODEV;
1968 }
1969
1970 /* Create the PCI parity sysfs entries */
1971 if (edac_sysfs_pci_setup()) {
1972 edac_sysfs_memctrl_teardown();
1973 edac_printk(KERN_ERR, EDAC_MC,
1974 "EDAC PCI: Error initializing sysfs code\n");
1975 return -ENODEV;
1976 }
1977
1978 /* create our kernel thread */
1979 edac_thread = kthread_run(edac_kernel_thread, NULL, "kedac");
1980
1981 if (IS_ERR(edac_thread)) {
1982 /* remove the sysfs entries */
1983 edac_sysfs_memctrl_teardown();
1984 edac_sysfs_pci_teardown();
1985 return PTR_ERR(edac_thread);
1986 }
1987
1988 return 0;
1989}
1990
1991/*
1992 * edac_mc_exit()
1993 * module exit/termination functioni
1994 */
1995static void __exit edac_mc_exit(void)
1996{
1997 debugf0("%s()\n", __func__);
1998 kthread_stop(edac_thread);
1999
2000 /* tear down the sysfs device */
2001 edac_sysfs_memctrl_teardown();
2002 edac_sysfs_pci_teardown();
2003}
2004
2005module_init(edac_mc_init);
2006module_exit(edac_mc_exit);
2007
2008MODULE_LICENSE("GPL");
2009MODULE_AUTHOR("Linux Networx (http://lnxi.com) Thayne Harbaugh et al\n"
2010 "Based on work by Dan Hollis et al");
2011MODULE_DESCRIPTION("Core library routines for MC reporting");
2012
2013module_param(panic_on_ue, int, 0644);
2014MODULE_PARM_DESC(panic_on_ue, "Panic on uncorrected error: 0=off 1=on");
2015#ifdef CONFIG_PCI
2016module_param(check_pci_parity, int, 0644);
2017MODULE_PARM_DESC(check_pci_parity, "Check for PCI bus parity errors: 0=off 1=on");
2018module_param(panic_on_pci_parity, int, 0644);
2019MODULE_PARM_DESC(panic_on_pci_parity, "Panic on PCI Bus Parity error: 0=off 1=on");
2020#endif
2021module_param(log_ue, int, 0644);
2022MODULE_PARM_DESC(log_ue, "Log uncorrectable error to console: 0=off 1=on");
2023module_param(log_ce, int, 0644);
2024MODULE_PARM_DESC(log_ce, "Log correctable error to console: 0=off 1=on");
2025module_param(poll_msec, int, 0644);
2026MODULE_PARM_DESC(poll_msec, "Polling period in milliseconds");
2027#ifdef CONFIG_EDAC_DEBUG
2028module_param(edac_debug_level, int, 0644);
2029MODULE_PARM_DESC(edac_debug_level, "Debug level");
2030#endif
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_ */
diff --git a/drivers/edac/edac_mc_sysfs.c b/drivers/edac/edac_mc_sysfs.c
new file mode 100644
index 000000000000..4a5e335f61d3
--- /dev/null
+++ b/drivers/edac/edac_mc_sysfs.c
@@ -0,0 +1,889 @@
1/*
2 * edac_mc kernel module
3 * (C) 2005, 2006 Linux Networx (http://lnxi.com)
4 * This file may be distributed under the terms of the
5 * GNU General Public License.
6 *
7 * Written Doug Thompson <norsk5@xmission.com>
8 *
9 */
10
11#include <linux/module.h>
12#include <linux/sysdev.h>
13#include <linux/ctype.h>
14
15#include "edac_mc.h"
16#include "edac_module.h"
17
18/* MC EDAC Controls, setable by module parameter, and sysfs */
19static int log_ue = 1;
20static int log_ce = 1;
21static int panic_on_ue;
22static int poll_msec = 1000;
23
24/* Getter functions for above */
25int edac_get_log_ue()
26{
27 return log_ue;
28}
29
30int edac_get_log_ce()
31{
32 return log_ce;
33}
34
35int edac_get_panic_on_ue()
36{
37 return panic_on_ue;
38}
39
40int edac_get_poll_msec()
41{
42 return poll_msec;
43}
44
45/* Parameter declarations for above */
46module_param(panic_on_ue, int, 0644);
47MODULE_PARM_DESC(panic_on_ue, "Panic on uncorrected error: 0=off 1=on");
48module_param(log_ue, int, 0644);
49MODULE_PARM_DESC(log_ue, "Log uncorrectable error to console: 0=off 1=on");
50module_param(log_ce, int, 0644);
51MODULE_PARM_DESC(log_ce, "Log correctable error to console: 0=off 1=on");
52module_param(poll_msec, int, 0644);
53MODULE_PARM_DESC(poll_msec, "Polling period in milliseconds");
54
55
56/*
57 * various constants for Memory Controllers
58 */
59static const char *mem_types[] = {
60 [MEM_EMPTY] = "Empty",
61 [MEM_RESERVED] = "Reserved",
62 [MEM_UNKNOWN] = "Unknown",
63 [MEM_FPM] = "FPM",
64 [MEM_EDO] = "EDO",
65 [MEM_BEDO] = "BEDO",
66 [MEM_SDR] = "Unbuffered-SDR",
67 [MEM_RDR] = "Registered-SDR",
68 [MEM_DDR] = "Unbuffered-DDR",
69 [MEM_RDDR] = "Registered-DDR",
70 [MEM_RMBS] = "RMBS"
71};
72
73static const char *dev_types[] = {
74 [DEV_UNKNOWN] = "Unknown",
75 [DEV_X1] = "x1",
76 [DEV_X2] = "x2",
77 [DEV_X4] = "x4",
78 [DEV_X8] = "x8",
79 [DEV_X16] = "x16",
80 [DEV_X32] = "x32",
81 [DEV_X64] = "x64"
82};
83
84static const char *edac_caps[] = {
85 [EDAC_UNKNOWN] = "Unknown",
86 [EDAC_NONE] = "None",
87 [EDAC_RESERVED] = "Reserved",
88 [EDAC_PARITY] = "PARITY",
89 [EDAC_EC] = "EC",
90 [EDAC_SECDED] = "SECDED",
91 [EDAC_S2ECD2ED] = "S2ECD2ED",
92 [EDAC_S4ECD4ED] = "S4ECD4ED",
93 [EDAC_S8ECD8ED] = "S8ECD8ED",
94 [EDAC_S16ECD16ED] = "S16ECD16ED"
95};
96
97/*
98 * sysfs object: /sys/devices/system/edac
99 * need to export to other files in this modules
100 */
101struct sysdev_class edac_class = {
102 set_kset_name("edac"),
103};
104
105/* sysfs object:
106 * /sys/devices/system/edac/mc
107 */
108static struct kobject edac_memctrl_kobj;
109
110/* We use these to wait for the reference counts on edac_memctrl_kobj and
111 * edac_pci_kobj to reach 0.
112 */
113static struct completion edac_memctrl_kobj_complete;
114
115/*
116 * /sys/devices/system/edac/mc;
117 * data structures and methods
118 */
119static ssize_t memctrl_int_show(void *ptr, char *buffer)
120{
121 int *value = (int*) ptr;
122 return sprintf(buffer, "%u\n", *value);
123}
124
125static ssize_t memctrl_int_store(void *ptr, const char *buffer, size_t count)
126{
127 int *value = (int*) ptr;
128
129 if (isdigit(*buffer))
130 *value = simple_strtoul(buffer, NULL, 0);
131
132 return count;
133}
134
135struct memctrl_dev_attribute {
136 struct attribute attr;
137 void *value;
138 ssize_t (*show)(void *,char *);
139 ssize_t (*store)(void *, const char *, size_t);
140};
141
142/* Set of show/store abstract level functions for memory control object */
143static ssize_t memctrl_dev_show(struct kobject *kobj,
144 struct attribute *attr, char *buffer)
145{
146 struct memctrl_dev_attribute *memctrl_dev;
147 memctrl_dev = (struct memctrl_dev_attribute*)attr;
148
149 if (memctrl_dev->show)
150 return memctrl_dev->show(memctrl_dev->value, buffer);
151
152 return -EIO;
153}
154
155static ssize_t memctrl_dev_store(struct kobject *kobj, struct attribute *attr,
156 const char *buffer, size_t count)
157{
158 struct memctrl_dev_attribute *memctrl_dev;
159 memctrl_dev = (struct memctrl_dev_attribute*)attr;
160
161 if (memctrl_dev->store)
162 return memctrl_dev->store(memctrl_dev->value, buffer, count);
163
164 return -EIO;
165}
166
167static struct sysfs_ops memctrlfs_ops = {
168 .show = memctrl_dev_show,
169 .store = memctrl_dev_store
170};
171
172#define MEMCTRL_ATTR(_name,_mode,_show,_store) \
173static struct memctrl_dev_attribute attr_##_name = { \
174 .attr = {.name = __stringify(_name), .mode = _mode }, \
175 .value = &_name, \
176 .show = _show, \
177 .store = _store, \
178};
179
180#define MEMCTRL_STRING_ATTR(_name,_data,_mode,_show,_store) \
181static struct memctrl_dev_attribute attr_##_name = { \
182 .attr = {.name = __stringify(_name), .mode = _mode }, \
183 .value = _data, \
184 .show = _show, \
185 .store = _store, \
186};
187
188/* csrow<id> control files */
189MEMCTRL_ATTR(panic_on_ue,S_IRUGO|S_IWUSR,memctrl_int_show,memctrl_int_store);
190MEMCTRL_ATTR(log_ue,S_IRUGO|S_IWUSR,memctrl_int_show,memctrl_int_store);
191MEMCTRL_ATTR(log_ce,S_IRUGO|S_IWUSR,memctrl_int_show,memctrl_int_store);
192MEMCTRL_ATTR(poll_msec,S_IRUGO|S_IWUSR,memctrl_int_show,memctrl_int_store);
193
194/* Base Attributes of the memory ECC object */
195static struct memctrl_dev_attribute *memctrl_attr[] = {
196 &attr_panic_on_ue,
197 &attr_log_ue,
198 &attr_log_ce,
199 &attr_poll_msec,
200 NULL,
201};
202
203/* Main MC kobject release() function */
204static void edac_memctrl_master_release(struct kobject *kobj)
205{
206 debugf1("%s()\n", __func__);
207 complete(&edac_memctrl_kobj_complete);
208}
209
210static struct kobj_type ktype_memctrl = {
211 .release = edac_memctrl_master_release,
212 .sysfs_ops = &memctrlfs_ops,
213 .default_attrs = (struct attribute **) memctrl_attr,
214};
215
216/* Initialize the main sysfs entries for edac:
217 * /sys/devices/system/edac
218 *
219 * and children
220 *
221 * Return: 0 SUCCESS
222 * !0 FAILURE
223 */
224int edac_sysfs_memctrl_setup(void)
225{
226 int err = 0;
227
228 debugf1("%s()\n", __func__);
229
230 /* create the /sys/devices/system/edac directory */
231 err = sysdev_class_register(&edac_class);
232
233 if (err) {
234 debugf1("%s() error=%d\n", __func__, err);
235 return err;
236 }
237
238 /* Init the MC's kobject */
239 memset(&edac_memctrl_kobj, 0, sizeof (edac_memctrl_kobj));
240 edac_memctrl_kobj.parent = &edac_class.kset.kobj;
241 edac_memctrl_kobj.ktype = &ktype_memctrl;
242
243 /* generate sysfs "..../edac/mc" */
244 err = kobject_set_name(&edac_memctrl_kobj,"mc");
245
246 if (err)
247 goto fail;
248
249 /* FIXME: maybe new sysdev_create_subdir() */
250 err = kobject_register(&edac_memctrl_kobj);
251
252 if (err) {
253 debugf1("Failed to register '.../edac/mc'\n");
254 goto fail;
255 }
256
257 debugf1("Registered '.../edac/mc' kobject\n");
258
259 return 0;
260
261fail:
262 sysdev_class_unregister(&edac_class);
263 return err;
264}
265
266/*
267 * MC teardown:
268 * the '..../edac/mc' kobject followed by '..../edac' itself
269 */
270void edac_sysfs_memctrl_teardown(void)
271{
272 debugf0("MC: " __FILE__ ": %s()\n", __func__);
273
274 /* Unregister the MC's kobject and wait for reference count to reach 0.
275 */
276 init_completion(&edac_memctrl_kobj_complete);
277 kobject_unregister(&edac_memctrl_kobj);
278 wait_for_completion(&edac_memctrl_kobj_complete);
279
280 /* Unregister the 'edac' object */
281 sysdev_class_unregister(&edac_class);
282}
283
284
285/* EDAC sysfs CSROW data structures and methods
286 */
287
288/* Set of more default csrow<id> attribute show/store functions */
289static ssize_t csrow_ue_count_show(struct csrow_info *csrow, char *data, int private)
290{
291 return sprintf(data,"%u\n", csrow->ue_count);
292}
293
294static ssize_t csrow_ce_count_show(struct csrow_info *csrow, char *data, int private)
295{
296 return sprintf(data,"%u\n", csrow->ce_count);
297}
298
299static ssize_t csrow_size_show(struct csrow_info *csrow, char *data, int private)
300{
301 return sprintf(data,"%u\n", PAGES_TO_MiB(csrow->nr_pages));
302}
303
304static ssize_t csrow_mem_type_show(struct csrow_info *csrow, char *data, int private)
305{
306 return sprintf(data,"%s\n", mem_types[csrow->mtype]);
307}
308
309static ssize_t csrow_dev_type_show(struct csrow_info *csrow, char *data, int private)
310{
311 return sprintf(data,"%s\n", dev_types[csrow->dtype]);
312}
313
314static ssize_t csrow_edac_mode_show(struct csrow_info *csrow, char *data, int private)
315{
316 return sprintf(data,"%s\n", edac_caps[csrow->edac_mode]);
317}
318
319/* show/store functions for DIMM Label attributes */
320static ssize_t channel_dimm_label_show(struct csrow_info *csrow,
321 char *data, int channel)
322{
323 return snprintf(data, EDAC_MC_LABEL_LEN,"%s",
324 csrow->channels[channel].label);
325}
326
327static ssize_t channel_dimm_label_store(struct csrow_info *csrow,
328 const char *data,
329 size_t count,
330 int channel)
331{
332 ssize_t max_size = 0;
333
334 max_size = min((ssize_t)count,(ssize_t)EDAC_MC_LABEL_LEN-1);
335 strncpy(csrow->channels[channel].label, data, max_size);
336 csrow->channels[channel].label[max_size] = '\0';
337
338 return max_size;
339}
340
341/* show function for dynamic chX_ce_count attribute */
342static ssize_t channel_ce_count_show(struct csrow_info *csrow,
343 char *data,
344 int channel)
345{
346 return sprintf(data, "%u\n", csrow->channels[channel].ce_count);
347}
348
349/* csrow specific attribute structure */
350struct csrowdev_attribute {
351 struct attribute attr;
352 ssize_t (*show)(struct csrow_info *,char *,int);
353 ssize_t (*store)(struct csrow_info *, const char *,size_t,int);
354 int private;
355};
356
357#define to_csrow(k) container_of(k, struct csrow_info, kobj)
358#define to_csrowdev_attr(a) container_of(a, struct csrowdev_attribute, attr)
359
360/* Set of show/store higher level functions for default csrow attributes */
361static ssize_t csrowdev_show(struct kobject *kobj,
362 struct attribute *attr,
363 char *buffer)
364{
365 struct csrow_info *csrow = to_csrow(kobj);
366 struct csrowdev_attribute *csrowdev_attr = to_csrowdev_attr(attr);
367
368 if (csrowdev_attr->show)
369 return csrowdev_attr->show(csrow,
370 buffer,
371 csrowdev_attr->private);
372 return -EIO;
373}
374
375static ssize_t csrowdev_store(struct kobject *kobj, struct attribute *attr,
376 const char *buffer, size_t count)
377{
378 struct csrow_info *csrow = to_csrow(kobj);
379 struct csrowdev_attribute * csrowdev_attr = to_csrowdev_attr(attr);
380
381 if (csrowdev_attr->store)
382 return csrowdev_attr->store(csrow,
383 buffer,
384 count,
385 csrowdev_attr->private);
386 return -EIO;
387}
388
389static struct sysfs_ops csrowfs_ops = {
390 .show = csrowdev_show,
391 .store = csrowdev_store
392};
393
394#define CSROWDEV_ATTR(_name,_mode,_show,_store,_private) \
395static struct csrowdev_attribute attr_##_name = { \
396 .attr = {.name = __stringify(_name), .mode = _mode }, \
397 .show = _show, \
398 .store = _store, \
399 .private = _private, \
400};
401
402/* default cwrow<id>/attribute files */
403CSROWDEV_ATTR(size_mb,S_IRUGO,csrow_size_show,NULL,0);
404CSROWDEV_ATTR(dev_type,S_IRUGO,csrow_dev_type_show,NULL,0);
405CSROWDEV_ATTR(mem_type,S_IRUGO,csrow_mem_type_show,NULL,0);
406CSROWDEV_ATTR(edac_mode,S_IRUGO,csrow_edac_mode_show,NULL,0);
407CSROWDEV_ATTR(ue_count,S_IRUGO,csrow_ue_count_show,NULL,0);
408CSROWDEV_ATTR(ce_count,S_IRUGO,csrow_ce_count_show,NULL,0);
409
410/* default attributes of the CSROW<id> object */
411static struct csrowdev_attribute *default_csrow_attr[] = {
412 &attr_dev_type,
413 &attr_mem_type,
414 &attr_edac_mode,
415 &attr_size_mb,
416 &attr_ue_count,
417 &attr_ce_count,
418 NULL,
419};
420
421
422/* possible dynamic channel DIMM Label attribute files */
423CSROWDEV_ATTR(ch0_dimm_label,S_IRUGO|S_IWUSR,
424 channel_dimm_label_show,
425 channel_dimm_label_store,
426 0 );
427CSROWDEV_ATTR(ch1_dimm_label,S_IRUGO|S_IWUSR,
428 channel_dimm_label_show,
429 channel_dimm_label_store,
430 1 );
431CSROWDEV_ATTR(ch2_dimm_label,S_IRUGO|S_IWUSR,
432 channel_dimm_label_show,
433 channel_dimm_label_store,
434 2 );
435CSROWDEV_ATTR(ch3_dimm_label,S_IRUGO|S_IWUSR,
436 channel_dimm_label_show,
437 channel_dimm_label_store,
438 3 );
439CSROWDEV_ATTR(ch4_dimm_label,S_IRUGO|S_IWUSR,
440 channel_dimm_label_show,
441 channel_dimm_label_store,
442 4 );
443CSROWDEV_ATTR(ch5_dimm_label,S_IRUGO|S_IWUSR,
444 channel_dimm_label_show,
445 channel_dimm_label_store,
446 5 );
447
448/* Total possible dynamic DIMM Label attribute file table */
449static struct csrowdev_attribute *dynamic_csrow_dimm_attr[] = {
450 &attr_ch0_dimm_label,
451 &attr_ch1_dimm_label,
452 &attr_ch2_dimm_label,
453 &attr_ch3_dimm_label,
454 &attr_ch4_dimm_label,
455 &attr_ch5_dimm_label
456};
457
458/* possible dynamic channel ce_count attribute files */
459CSROWDEV_ATTR(ch0_ce_count,S_IRUGO|S_IWUSR,
460 channel_ce_count_show,
461 NULL,
462 0 );
463CSROWDEV_ATTR(ch1_ce_count,S_IRUGO|S_IWUSR,
464 channel_ce_count_show,
465 NULL,
466 1 );
467CSROWDEV_ATTR(ch2_ce_count,S_IRUGO|S_IWUSR,
468 channel_ce_count_show,
469 NULL,
470 2 );
471CSROWDEV_ATTR(ch3_ce_count,S_IRUGO|S_IWUSR,
472 channel_ce_count_show,
473 NULL,
474 3 );
475CSROWDEV_ATTR(ch4_ce_count,S_IRUGO|S_IWUSR,
476 channel_ce_count_show,
477 NULL,
478 4 );
479CSROWDEV_ATTR(ch5_ce_count,S_IRUGO|S_IWUSR,
480 channel_ce_count_show,
481 NULL,
482 5 );
483
484/* Total possible dynamic ce_count attribute file table */
485static struct csrowdev_attribute *dynamic_csrow_ce_count_attr[] = {
486 &attr_ch0_ce_count,
487 &attr_ch1_ce_count,
488 &attr_ch2_ce_count,
489 &attr_ch3_ce_count,
490 &attr_ch4_ce_count,
491 &attr_ch5_ce_count
492};
493
494
495#define EDAC_NR_CHANNELS 6
496
497/* Create dynamic CHANNEL files, indexed by 'chan', under specifed CSROW */
498static int edac_create_channel_files(struct kobject *kobj, int chan)
499{
500 int err=-ENODEV;
501
502 if (chan >= EDAC_NR_CHANNELS)
503 return err;
504
505 /* create the DIMM label attribute file */
506 err = sysfs_create_file(kobj,
507 (struct attribute *) dynamic_csrow_dimm_attr[chan]);
508
509 if (!err) {
510 /* create the CE Count attribute file */
511 err = sysfs_create_file(kobj,
512 (struct attribute *) dynamic_csrow_ce_count_attr[chan]);
513 } else {
514 debugf1("%s() dimm labels and ce_count files created", __func__);
515 }
516
517 return err;
518}
519
520/* No memory to release for this kobj */
521static void edac_csrow_instance_release(struct kobject *kobj)
522{
523 struct csrow_info *cs;
524
525 cs = container_of(kobj, struct csrow_info, kobj);
526 complete(&cs->kobj_complete);
527}
528
529/* the kobj_type instance for a CSROW */
530static struct kobj_type ktype_csrow = {
531 .release = edac_csrow_instance_release,
532 .sysfs_ops = &csrowfs_ops,
533 .default_attrs = (struct attribute **) default_csrow_attr,
534};
535
536/* Create a CSROW object under specifed edac_mc_device */
537static int edac_create_csrow_object(
538 struct kobject *edac_mci_kobj,
539 struct csrow_info *csrow,
540 int index)
541{
542 int err = 0;
543 int chan;
544
545 memset(&csrow->kobj, 0, sizeof(csrow->kobj));
546
547 /* generate ..../edac/mc/mc<id>/csrow<index> */
548
549 csrow->kobj.parent = edac_mci_kobj;
550 csrow->kobj.ktype = &ktype_csrow;
551
552 /* name this instance of csrow<id> */
553 err = kobject_set_name(&csrow->kobj,"csrow%d",index);
554 if (err)
555 goto error_exit;
556
557 /* Instanstiate the csrow object */
558 err = kobject_register(&csrow->kobj);
559 if (!err) {
560 /* Create the dyanmic attribute files on this csrow,
561 * namely, the DIMM labels and the channel ce_count
562 */
563 for (chan = 0; chan < csrow->nr_channels; chan++) {
564 err = edac_create_channel_files(&csrow->kobj,chan);
565 if (err)
566 break;
567 }
568 }
569
570error_exit:
571 return err;
572}
573
574/* default sysfs methods and data structures for the main MCI kobject */
575
576static ssize_t mci_reset_counters_store(struct mem_ctl_info *mci,
577 const char *data, size_t count)
578{
579 int row, chan;
580
581 mci->ue_noinfo_count = 0;
582 mci->ce_noinfo_count = 0;
583 mci->ue_count = 0;
584 mci->ce_count = 0;
585
586 for (row = 0; row < mci->nr_csrows; row++) {
587 struct csrow_info *ri = &mci->csrows[row];
588
589 ri->ue_count = 0;
590 ri->ce_count = 0;
591
592 for (chan = 0; chan < ri->nr_channels; chan++)
593 ri->channels[chan].ce_count = 0;
594 }
595
596 mci->start_time = jiffies;
597 return count;
598}
599
600/* memory scrubbing */
601static ssize_t mci_sdram_scrub_rate_store(struct mem_ctl_info *mci,
602 const char *data, size_t count)
603{
604 u32 bandwidth = -1;
605
606 if (mci->set_sdram_scrub_rate) {
607
608 memctrl_int_store(&bandwidth, data, count);
609
610 if (!(*mci->set_sdram_scrub_rate)(mci, &bandwidth)) {
611 edac_printk(KERN_DEBUG, EDAC_MC,
612 "Scrub rate set successfully, applied: %d\n",
613 bandwidth);
614 } else {
615 /* FIXME: error codes maybe? */
616 edac_printk(KERN_DEBUG, EDAC_MC,
617 "Scrub rate set FAILED, could not apply: %d\n",
618 bandwidth);
619 }
620 } else {
621 /* FIXME: produce "not implemented" ERROR for user-side. */
622 edac_printk(KERN_WARNING, EDAC_MC,
623 "Memory scrubbing 'set'control is not implemented!\n");
624 }
625 return count;
626}
627
628static ssize_t mci_sdram_scrub_rate_show(struct mem_ctl_info *mci, char *data)
629{
630 u32 bandwidth = -1;
631
632 if (mci->get_sdram_scrub_rate) {
633 if (!(*mci->get_sdram_scrub_rate)(mci, &bandwidth)) {
634 edac_printk(KERN_DEBUG, EDAC_MC,
635 "Scrub rate successfully, fetched: %d\n",
636 bandwidth);
637 } else {
638 /* FIXME: error codes maybe? */
639 edac_printk(KERN_DEBUG, EDAC_MC,
640 "Scrub rate fetch FAILED, got: %d\n",
641 bandwidth);
642 }
643 } else {
644 /* FIXME: produce "not implemented" ERROR for user-side. */
645 edac_printk(KERN_WARNING, EDAC_MC,
646 "Memory scrubbing 'get' control is not implemented!\n");
647 }
648 return sprintf(data, "%d\n", bandwidth);
649}
650
651/* default attribute files for the MCI object */
652static ssize_t mci_ue_count_show(struct mem_ctl_info *mci, char *data)
653{
654 return sprintf(data,"%d\n", mci->ue_count);
655}
656
657static ssize_t mci_ce_count_show(struct mem_ctl_info *mci, char *data)
658{
659 return sprintf(data,"%d\n", mci->ce_count);
660}
661
662static ssize_t mci_ce_noinfo_show(struct mem_ctl_info *mci, char *data)
663{
664 return sprintf(data,"%d\n", mci->ce_noinfo_count);
665}
666
667static ssize_t mci_ue_noinfo_show(struct mem_ctl_info *mci, char *data)
668{
669 return sprintf(data,"%d\n", mci->ue_noinfo_count);
670}
671
672static ssize_t mci_seconds_show(struct mem_ctl_info *mci, char *data)
673{
674 return sprintf(data,"%ld\n", (jiffies - mci->start_time) / HZ);
675}
676
677static ssize_t mci_ctl_name_show(struct mem_ctl_info *mci, char *data)
678{
679 return sprintf(data,"%s\n", mci->ctl_name);
680}
681
682static ssize_t mci_size_mb_show(struct mem_ctl_info *mci, char *data)
683{
684 int total_pages, csrow_idx;
685
686 for (total_pages = csrow_idx = 0; csrow_idx < mci->nr_csrows;
687 csrow_idx++) {
688 struct csrow_info *csrow = &mci->csrows[csrow_idx];
689
690 if (!csrow->nr_pages)
691 continue;
692
693 total_pages += csrow->nr_pages;
694 }
695
696 return sprintf(data,"%u\n", PAGES_TO_MiB(total_pages));
697}
698
699struct mcidev_attribute {
700 struct attribute attr;
701 ssize_t (*show)(struct mem_ctl_info *,char *);
702 ssize_t (*store)(struct mem_ctl_info *, const char *,size_t);
703};
704
705#define to_mci(k) container_of(k, struct mem_ctl_info, edac_mci_kobj)
706#define to_mcidev_attr(a) container_of(a, struct mcidev_attribute, attr)
707
708/* MCI show/store functions for top most object */
709static ssize_t mcidev_show(struct kobject *kobj, struct attribute *attr,
710 char *buffer)
711{
712 struct mem_ctl_info *mem_ctl_info = to_mci(kobj);
713 struct mcidev_attribute * mcidev_attr = to_mcidev_attr(attr);
714
715 if (mcidev_attr->show)
716 return mcidev_attr->show(mem_ctl_info, buffer);
717
718 return -EIO;
719}
720
721static ssize_t mcidev_store(struct kobject *kobj, struct attribute *attr,
722 const char *buffer, size_t count)
723{
724 struct mem_ctl_info *mem_ctl_info = to_mci(kobj);
725 struct mcidev_attribute * mcidev_attr = to_mcidev_attr(attr);
726
727 if (mcidev_attr->store)
728 return mcidev_attr->store(mem_ctl_info, buffer, count);
729
730 return -EIO;
731}
732
733static struct sysfs_ops mci_ops = {
734 .show = mcidev_show,
735 .store = mcidev_store
736};
737
738#define MCIDEV_ATTR(_name,_mode,_show,_store) \
739static struct mcidev_attribute mci_attr_##_name = { \
740 .attr = {.name = __stringify(_name), .mode = _mode }, \
741 .show = _show, \
742 .store = _store, \
743};
744
745/* default Control file */
746MCIDEV_ATTR(reset_counters,S_IWUSR,NULL,mci_reset_counters_store);
747
748/* default Attribute files */
749MCIDEV_ATTR(mc_name,S_IRUGO,mci_ctl_name_show,NULL);
750MCIDEV_ATTR(size_mb,S_IRUGO,mci_size_mb_show,NULL);
751MCIDEV_ATTR(seconds_since_reset,S_IRUGO,mci_seconds_show,NULL);
752MCIDEV_ATTR(ue_noinfo_count,S_IRUGO,mci_ue_noinfo_show,NULL);
753MCIDEV_ATTR(ce_noinfo_count,S_IRUGO,mci_ce_noinfo_show,NULL);
754MCIDEV_ATTR(ue_count,S_IRUGO,mci_ue_count_show,NULL);
755MCIDEV_ATTR(ce_count,S_IRUGO,mci_ce_count_show,NULL);
756
757/* memory scrubber attribute file */
758MCIDEV_ATTR(sdram_scrub_rate,S_IRUGO|S_IWUSR,mci_sdram_scrub_rate_show,mci_sdram_scrub_rate_store);
759
760static struct mcidev_attribute *mci_attr[] = {
761 &mci_attr_reset_counters,
762 &mci_attr_mc_name,
763 &mci_attr_size_mb,
764 &mci_attr_seconds_since_reset,
765 &mci_attr_ue_noinfo_count,
766 &mci_attr_ce_noinfo_count,
767 &mci_attr_ue_count,
768 &mci_attr_ce_count,
769 &mci_attr_sdram_scrub_rate,
770 NULL
771};
772
773/*
774 * Release of a MC controlling instance
775 */
776static void edac_mci_instance_release(struct kobject *kobj)
777{
778 struct mem_ctl_info *mci;
779
780 mci = to_mci(kobj);
781 debugf0("%s() idx=%d\n", __func__, mci->mc_idx);
782 complete(&mci->kobj_complete);
783}
784
785static struct kobj_type ktype_mci = {
786 .release = edac_mci_instance_release,
787 .sysfs_ops = &mci_ops,
788 .default_attrs = (struct attribute **) mci_attr,
789};
790
791
792#define EDAC_DEVICE_SYMLINK "device"
793
794/*
795 * Create a new Memory Controller kobject instance,
796 * mc<id> under the 'mc' directory
797 *
798 * Return:
799 * 0 Success
800 * !0 Failure
801 */
802int edac_create_sysfs_mci_device(struct mem_ctl_info *mci)
803{
804 int i;
805 int err;
806 struct csrow_info *csrow;
807 struct kobject *edac_mci_kobj=&mci->edac_mci_kobj;
808
809 debugf0("%s() idx=%d\n", __func__, mci->mc_idx);
810 memset(edac_mci_kobj, 0, sizeof(*edac_mci_kobj));
811
812 /* set the name of the mc<id> object */
813 err = kobject_set_name(edac_mci_kobj,"mc%d",mci->mc_idx);
814 if (err)
815 return err;
816
817 /* link to our parent the '..../edac/mc' object */
818 edac_mci_kobj->parent = &edac_memctrl_kobj;
819 edac_mci_kobj->ktype = &ktype_mci;
820
821 /* register the mc<id> kobject */
822 err = kobject_register(edac_mci_kobj);
823 if (err)
824 return err;
825
826 /* create a symlink for the device */
827 err = sysfs_create_link(edac_mci_kobj, &mci->dev->kobj,
828 EDAC_DEVICE_SYMLINK);
829 if (err)
830 goto fail0;
831
832 /* Make directories for each CSROW object
833 * under the mc<id> kobject
834 */
835 for (i = 0; i < mci->nr_csrows; i++) {
836 csrow = &mci->csrows[i];
837
838 /* Only expose populated CSROWs */
839 if (csrow->nr_pages > 0) {
840 err = edac_create_csrow_object(edac_mci_kobj,csrow,i);
841 if (err)
842 goto fail1;
843 }
844 }
845
846 return 0;
847
848 /* CSROW error: backout what has already been registered, */
849fail1:
850 for ( i--; i >= 0; i--) {
851 if (csrow->nr_pages > 0) {
852 init_completion(&csrow->kobj_complete);
853 kobject_unregister(&mci->csrows[i].kobj);
854 wait_for_completion(&csrow->kobj_complete);
855 }
856 }
857
858fail0:
859 init_completion(&mci->kobj_complete);
860 kobject_unregister(edac_mci_kobj);
861 wait_for_completion(&mci->kobj_complete);
862 return err;
863}
864
865/*
866 * remove a Memory Controller instance
867 */
868void edac_remove_sysfs_mci_device(struct mem_ctl_info *mci)
869{
870 int i;
871
872 debugf0("%s()\n", __func__);
873
874 /* remove all csrow kobjects */
875 for (i = 0; i < mci->nr_csrows; i++) {
876 if (mci->csrows[i].nr_pages > 0) {
877 init_completion(&mci->csrows[i].kobj_complete);
878 kobject_unregister(&mci->csrows[i].kobj);
879 wait_for_completion(&mci->csrows[i].kobj_complete);
880 }
881 }
882
883 sysfs_remove_link(&mci->edac_mci_kobj, EDAC_DEVICE_SYMLINK);
884 init_completion(&mci->kobj_complete);
885 kobject_unregister(&mci->edac_mci_kobj);
886 wait_for_completion(&mci->kobj_complete);
887}
888
889
diff --git a/drivers/edac/edac_module.c b/drivers/edac/edac_module.c
new file mode 100644
index 000000000000..8db0471a9476
--- /dev/null
+++ b/drivers/edac/edac_module.c
@@ -0,0 +1,130 @@
1
2#include <linux/freezer.h>
3#include <linux/kthread.h>
4
5#include "edac_mc.h"
6#include "edac_module.h"
7
8#define EDAC_MC_VERSION "Ver: 2.0.3" __DATE__
9
10#ifdef CONFIG_EDAC_DEBUG
11/* Values of 0 to 4 will generate output */
12int edac_debug_level = 1;
13EXPORT_SYMBOL_GPL(edac_debug_level);
14#endif
15
16static struct task_struct *edac_thread;
17
18/*
19 * Check MC status every edac_get_poll_msec().
20 * Check PCI status every edac_get_poll_msec() as well.
21 *
22 * This where the work gets done for edac.
23 *
24 * SMP safe, doesn't use NMI, and auto-rate-limits.
25 */
26static void do_edac_check(void)
27{
28 debugf3("%s()\n", __func__);
29
30 /* perform the poll activities */
31 edac_check_mc_devices();
32 edac_pci_do_parity_check();
33}
34
35/*
36 * Action thread for EDAC to perform the POLL operations
37 */
38static int edac_kernel_thread(void *arg)
39{
40 int msec;
41
42 while (!kthread_should_stop()) {
43
44 do_edac_check();
45
46 /* goto sleep for the interval */
47 msec = (HZ * edac_get_poll_msec()) / 1000;
48 schedule_timeout_interruptible(msec);
49 try_to_freeze();
50 }
51
52 return 0;
53}
54
55/*
56 * edac_init
57 * module initialization entry point
58 */
59static int __init edac_init(void)
60{
61 edac_printk(KERN_INFO, EDAC_MC, EDAC_MC_VERSION "\n");
62
63 /*
64 * Harvest and clear any boot/initialization PCI parity errors
65 *
66 * FIXME: This only clears errors logged by devices present at time of
67 * module initialization. We should also do an initial clear
68 * of each newly hotplugged device.
69 */
70 edac_pci_clear_parity_errors();
71
72 /* Create the MC sysfs entries */
73 if (edac_sysfs_memctrl_setup()) {
74 edac_printk(KERN_ERR, EDAC_MC,
75 "Error initializing sysfs code\n");
76 return -ENODEV;
77 }
78
79 /* Create the PCI parity sysfs entries */
80 if (edac_sysfs_pci_setup()) {
81 edac_sysfs_memctrl_teardown();
82 edac_printk(KERN_ERR, EDAC_MC,
83 "PCI: Error initializing sysfs code\n");
84 return -ENODEV;
85 }
86
87 /* create our kernel thread */
88 edac_thread = kthread_run(edac_kernel_thread, NULL, "kedac");
89
90 if (IS_ERR(edac_thread)) {
91 /* remove the sysfs entries */
92 edac_sysfs_memctrl_teardown();
93 edac_sysfs_pci_teardown();
94 return PTR_ERR(edac_thread);
95 }
96
97 return 0;
98}
99
100/*
101 * edac_exit()
102 * module exit/termination function
103 */
104static void __exit edac_exit(void)
105{
106 debugf0("%s()\n", __func__);
107 kthread_stop(edac_thread);
108
109 /* tear down the sysfs device */
110 edac_sysfs_memctrl_teardown();
111 edac_sysfs_pci_teardown();
112}
113
114/*
115 * Inform the kernel of our entry and exit points
116 */
117module_init(edac_init);
118module_exit(edac_exit);
119
120MODULE_LICENSE("GPL");
121MODULE_AUTHOR("Doug Thompson www.softwarebitmaker.com, et al");
122MODULE_DESCRIPTION("Core library routines for EDAC reporting");
123
124/* refer to *_sysfs.c files for parameters that are exported via sysfs */
125
126#ifdef CONFIG_EDAC_DEBUG
127module_param(edac_debug_level, int, 0644);
128MODULE_PARM_DESC(edac_debug_level, "Debug level");
129#endif
130
diff --git a/drivers/edac/edac_module.h b/drivers/edac/edac_module.h
new file mode 100644
index 000000000000..69c77f85bcd4
--- /dev/null
+++ b/drivers/edac/edac_module.h
@@ -0,0 +1,55 @@
1
2/*
3 * edac_module.h
4 *
5 * For defining functions/data for within the EDAC_CORE module only
6 *
7 * written by doug thompson <norsk5@xmission.h>
8 */
9
10#ifndef __EDAC_MODULE_H__
11#define __EDAC_MODULE_H__
12
13#include <linux/sysdev.h>
14
15#include "edac_core.h"
16
17/*
18 * INTERNAL EDAC MODULE:
19 * EDAC memory controller sysfs create/remove functions
20 * and setup/teardown functions
21 */
22extern int edac_create_sysfs_mci_device(struct mem_ctl_info *mci);
23extern void edac_remove_sysfs_mci_device(struct mem_ctl_info *mci);
24extern int edac_sysfs_memctrl_setup(void);
25extern void edac_sysfs_memctrl_teardown(void);
26extern void edac_check_mc_devices(void);
27extern int edac_get_log_ue(void);
28extern int edac_get_log_ce(void);
29extern int edac_get_panic_on_ue(void);
30extern int edac_get_poll_msec(void);
31
32extern int edac_device_create_sysfs(struct edac_device_ctl_info *edac_dev);
33extern void edac_device_remove_sysfs(struct edac_device_ctl_info *edac_dev);
34extern struct sysdev_class *edac_get_edac_class(void);
35
36
37/*
38 * EDAC PCI functions
39 */
40#ifdef CONFIG_PCI
41extern void edac_pci_do_parity_check(void);
42extern void edac_pci_clear_parity_errors(void);
43extern int edac_sysfs_pci_setup(void);
44extern void edac_sysfs_pci_teardown(void);
45#else /* CONFIG_PCI */
46/* pre-process these away */
47#define edac_pci_do_parity_check()
48#define edac_pci_clear_parity_errors()
49#define edac_sysfs_pci_setup() (0)
50#define edac_sysfs_pci_teardown()
51#endif /* CONFIG_PCI */
52
53
54#endif /* __EDAC_MODULE_H__ */
55
diff --git a/drivers/edac/edac_pci_sysfs.c b/drivers/edac/edac_pci_sysfs.c
new file mode 100644
index 000000000000..db23fec522e2
--- /dev/null
+++ b/drivers/edac/edac_pci_sysfs.c
@@ -0,0 +1,361 @@
1/* edac_mc kernel module
2 * (C) 2005, 2006 Linux Networx (http://lnxi.com)
3 * This file may be distributed under the terms of the
4 * GNU General Public License.
5 *
6 * Written Doug Thompson <norsk5@xmission.com>
7 *
8 */
9#include <linux/module.h>
10#include <linux/sysdev.h>
11#include <linux/ctype.h>
12
13#include "edac_mc.h"
14#include "edac_module.h"
15
16
17#ifdef CONFIG_PCI
18static int check_pci_parity = 0; /* default YES check PCI parity */
19static int panic_on_pci_parity; /* default no panic on PCI Parity */
20static atomic_t pci_parity_count = ATOMIC_INIT(0);
21
22static struct kobject edac_pci_kobj; /* /sys/devices/system/edac/pci */
23static struct completion edac_pci_kobj_complete;
24
25
26static ssize_t edac_pci_int_show(void *ptr, char *buffer)
27{
28 int *value = ptr;
29 return sprintf(buffer,"%d\n",*value);
30}
31
32static ssize_t edac_pci_int_store(void *ptr, const char *buffer, size_t count)
33{
34 int *value = ptr;
35
36 if (isdigit(*buffer))
37 *value = simple_strtoul(buffer,NULL,0);
38
39 return count;
40}
41
42struct edac_pci_dev_attribute {
43 struct attribute attr;
44 void *value;
45 ssize_t (*show)(void *,char *);
46 ssize_t (*store)(void *, const char *,size_t);
47};
48
49/* Set of show/store abstract level functions for PCI Parity object */
50static ssize_t edac_pci_dev_show(struct kobject *kobj, struct attribute *attr,
51 char *buffer)
52{
53 struct edac_pci_dev_attribute *edac_pci_dev;
54 edac_pci_dev= (struct edac_pci_dev_attribute*)attr;
55
56 if (edac_pci_dev->show)
57 return edac_pci_dev->show(edac_pci_dev->value, buffer);
58 return -EIO;
59}
60
61static ssize_t edac_pci_dev_store(struct kobject *kobj,
62 struct attribute *attr, const char *buffer, size_t count)
63{
64 struct edac_pci_dev_attribute *edac_pci_dev;
65 edac_pci_dev= (struct edac_pci_dev_attribute*)attr;
66
67 if (edac_pci_dev->show)
68 return edac_pci_dev->store(edac_pci_dev->value, buffer, count);
69 return -EIO;
70}
71
72static struct sysfs_ops edac_pci_sysfs_ops = {
73 .show = edac_pci_dev_show,
74 .store = edac_pci_dev_store
75};
76
77#define EDAC_PCI_ATTR(_name,_mode,_show,_store) \
78static struct edac_pci_dev_attribute edac_pci_attr_##_name = { \
79 .attr = {.name = __stringify(_name), .mode = _mode }, \
80 .value = &_name, \
81 .show = _show, \
82 .store = _store, \
83};
84
85#define EDAC_PCI_STRING_ATTR(_name,_data,_mode,_show,_store) \
86static struct edac_pci_dev_attribute edac_pci_attr_##_name = { \
87 .attr = {.name = __stringify(_name), .mode = _mode }, \
88 .value = _data, \
89 .show = _show, \
90 .store = _store, \
91};
92
93/* PCI Parity control files */
94EDAC_PCI_ATTR(check_pci_parity, S_IRUGO|S_IWUSR, edac_pci_int_show,
95 edac_pci_int_store);
96EDAC_PCI_ATTR(panic_on_pci_parity, S_IRUGO|S_IWUSR, edac_pci_int_show,
97 edac_pci_int_store);
98EDAC_PCI_ATTR(pci_parity_count, S_IRUGO, edac_pci_int_show, NULL);
99
100/* Base Attributes of the memory ECC object */
101static struct edac_pci_dev_attribute *edac_pci_attr[] = {
102 &edac_pci_attr_check_pci_parity,
103 &edac_pci_attr_panic_on_pci_parity,
104 &edac_pci_attr_pci_parity_count,
105 NULL,
106};
107
108/* No memory to release */
109static void edac_pci_release(struct kobject *kobj)
110{
111 debugf1("%s()\n", __func__);
112 complete(&edac_pci_kobj_complete);
113}
114
115static struct kobj_type ktype_edac_pci = {
116 .release = edac_pci_release,
117 .sysfs_ops = &edac_pci_sysfs_ops,
118 .default_attrs = (struct attribute **) edac_pci_attr,
119};
120
121/**
122 * edac_sysfs_pci_setup()
123 *
124 * setup the sysfs for EDAC PCI attributes
125 * assumes edac_class has already been initialized
126 */
127int edac_sysfs_pci_setup(void)
128{
129 int err;
130 struct sysdev_class *edac_class;
131
132 debugf1("%s()\n", __func__);
133
134 edac_class = edac_get_edac_class();
135
136 memset(&edac_pci_kobj, 0, sizeof(edac_pci_kobj));
137 edac_pci_kobj.parent = &edac_class->kset.kobj;
138 edac_pci_kobj.ktype = &ktype_edac_pci;
139 err = kobject_set_name(&edac_pci_kobj, "pci");
140
141 if (!err) {
142 /* Instanstiate the pci object */
143 /* FIXME: maybe new sysdev_create_subdir() */
144 err = kobject_register(&edac_pci_kobj);
145
146 if (err)
147 debugf1("Failed to register '.../edac/pci'\n");
148 else
149 debugf1("Registered '.../edac/pci' kobject\n");
150 }
151
152 return err;
153}
154
155/*
156 * edac_sysfs_pci_teardown
157 *
158 * perform the sysfs teardown for the PCI attributes
159 */
160void edac_sysfs_pci_teardown(void)
161{
162 debugf0("%s()\n", __func__);
163 init_completion(&edac_pci_kobj_complete);
164 kobject_unregister(&edac_pci_kobj);
165 wait_for_completion(&edac_pci_kobj_complete);
166}
167
168
169static u16 get_pci_parity_status(struct pci_dev *dev, int secondary)
170{
171 int where;
172 u16 status;
173
174 where = secondary ? PCI_SEC_STATUS : PCI_STATUS;
175 pci_read_config_word(dev, where, &status);
176
177 /* If we get back 0xFFFF then we must suspect that the card has been
178 * pulled but the Linux PCI layer has not yet finished cleaning up.
179 * We don't want to report on such devices
180 */
181
182 if (status == 0xFFFF) {
183 u32 sanity;
184
185 pci_read_config_dword(dev, 0, &sanity);
186
187 if (sanity == 0xFFFFFFFF)
188 return 0;
189 }
190
191 status &= PCI_STATUS_DETECTED_PARITY | PCI_STATUS_SIG_SYSTEM_ERROR |
192 PCI_STATUS_PARITY;
193
194 if (status)
195 /* reset only the bits we are interested in */
196 pci_write_config_word(dev, where, status);
197
198 return status;
199}
200
201typedef void (*pci_parity_check_fn_t) (struct pci_dev *dev);
202
203/* Clear any PCI parity errors logged by this device. */
204static void edac_pci_dev_parity_clear(struct pci_dev *dev)
205{
206 u8 header_type;
207
208 get_pci_parity_status(dev, 0);
209
210 /* read the device TYPE, looking for bridges */
211 pci_read_config_byte(dev, PCI_HEADER_TYPE, &header_type);
212
213 if ((header_type & 0x7F) == PCI_HEADER_TYPE_BRIDGE)
214 get_pci_parity_status(dev, 1);
215}
216
217/*
218 * PCI Parity polling
219 *
220 */
221static void edac_pci_dev_parity_test(struct pci_dev *dev)
222{
223 u16 status;
224 u8 header_type;
225
226 /* read the STATUS register on this device
227 */
228 status = get_pci_parity_status(dev, 0);
229
230 debugf2("PCI STATUS= 0x%04x %s\n", status, dev->dev.bus_id );
231
232 /* check the status reg for errors */
233 if (status) {
234 if (status & (PCI_STATUS_SIG_SYSTEM_ERROR))
235 edac_printk(KERN_CRIT, EDAC_PCI,
236 "Signaled System Error on %s\n",
237 pci_name(dev));
238
239 if (status & (PCI_STATUS_PARITY)) {
240 edac_printk(KERN_CRIT, EDAC_PCI,
241 "Master Data Parity Error on %s\n",
242 pci_name(dev));
243
244 atomic_inc(&pci_parity_count);
245 }
246
247 if (status & (PCI_STATUS_DETECTED_PARITY)) {
248 edac_printk(KERN_CRIT, EDAC_PCI,
249 "Detected Parity Error on %s\n",
250 pci_name(dev));
251
252 atomic_inc(&pci_parity_count);
253 }
254 }
255
256 /* read the device TYPE, looking for bridges */
257 pci_read_config_byte(dev, PCI_HEADER_TYPE, &header_type);
258
259 debugf2("PCI HEADER TYPE= 0x%02x %s\n", header_type, dev->dev.bus_id );
260
261 if ((header_type & 0x7F) == PCI_HEADER_TYPE_BRIDGE) {
262 /* On bridges, need to examine secondary status register */
263 status = get_pci_parity_status(dev, 1);
264
265 debugf2("PCI SEC_STATUS= 0x%04x %s\n",
266 status, dev->dev.bus_id );
267
268 /* check the secondary status reg for errors */
269 if (status) {
270 if (status & (PCI_STATUS_SIG_SYSTEM_ERROR))
271 edac_printk(KERN_CRIT, EDAC_PCI, "Bridge "
272 "Signaled System Error on %s\n",
273 pci_name(dev));
274
275 if (status & (PCI_STATUS_PARITY)) {
276 edac_printk(KERN_CRIT, EDAC_PCI, "Bridge "
277 "Master Data Parity Error on "
278 "%s\n", pci_name(dev));
279
280 atomic_inc(&pci_parity_count);
281 }
282
283 if (status & (PCI_STATUS_DETECTED_PARITY)) {
284 edac_printk(KERN_CRIT, EDAC_PCI, "Bridge "
285 "Detected Parity Error on %s\n",
286 pci_name(dev));
287
288 atomic_inc(&pci_parity_count);
289 }
290 }
291 }
292}
293
294/*
295 * pci_dev parity list iterator
296 * Scan the PCI device list for one iteration, looking for SERRORs
297 * Master Parity ERRORS or Parity ERRORs on primary or secondary devices
298 */
299static inline void edac_pci_dev_parity_iterator(pci_parity_check_fn_t fn)
300{
301 struct pci_dev *dev = NULL;
302
303 /* request for kernel access to the next PCI device, if any,
304 * and while we are looking at it have its reference count
305 * bumped until we are done with it
306 */
307 while((dev = pci_get_device(PCI_ANY_ID, PCI_ANY_ID, dev)) != NULL) {
308 fn(dev);
309 }
310}
311
312/*
313 * edac_pci_do_parity_check
314 *
315 * performs the actual PCI parity check operation
316 */
317void edac_pci_do_parity_check(void)
318{
319 unsigned long flags;
320 int before_count;
321
322 debugf3("%s()\n", __func__);
323
324 if (!check_pci_parity)
325 return;
326
327 before_count = atomic_read(&pci_parity_count);
328
329 /* scan all PCI devices looking for a Parity Error on devices and
330 * bridges
331 */
332 local_irq_save(flags);
333 edac_pci_dev_parity_iterator(edac_pci_dev_parity_test);
334 local_irq_restore(flags);
335
336 /* Only if operator has selected panic on PCI Error */
337 if (panic_on_pci_parity) {
338 /* If the count is different 'after' from 'before' */
339 if (before_count != atomic_read(&pci_parity_count))
340 panic("EDAC: PCI Parity Error");
341 }
342}
343
344void edac_pci_clear_parity_errors(void)
345{
346 /* Clear any PCI bus parity errors that devices initially have logged
347 * in their registers.
348 */
349 edac_pci_dev_parity_iterator(edac_pci_dev_parity_clear);
350}
351
352
353/*
354 * Define the PCI parameter to the module
355 */
356module_param(check_pci_parity, int, 0644);
357MODULE_PARM_DESC(check_pci_parity, "Check for PCI bus parity errors: 0=off 1=on");
358module_param(panic_on_pci_parity, int, 0644);
359MODULE_PARM_DESC(panic_on_pci_parity, "Panic on PCI Bus Parity error: 0=off 1=on");
360
361#endif /* CONFIG_PCI */
generated by cgit v1.2.2 (git 2.25.0) at 2024-11-17 23:00:44 -0500