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.
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/
11 * Modified by Dave Peterson and Doug Thompson
15 #include <linux/module.h>
16 #include <linux/proc_fs.h>
17 #include <linux/kernel.h>
18 #include <linux/types.h>
19 #include <linux/smp.h>
20 #include <linux/init.h>
21 #include <linux/sysctl.h>
22 #include <linux/highmem.h>
23 #include <linux/timer.h>
24 #include <linux/slab.h>
25 #include <linux/jiffies.h>
26 #include <linux/spinlock.h>
27 #include <linux/list.h>
28 #include <linux/sysdev.h>
29 #include <linux/ctype.h>
30 #include <linux/kthread.h>
31 #include <linux/freezer.h>
32 #include <asm/uaccess.h>
37 #define EDAC_MC_VERSION "Ver: 2.0.1 " __DATE__
40 #ifdef CONFIG_EDAC_DEBUG
41 /* Values of 0 to 4 will generate output */
42 int edac_debug_level
= 1;
43 EXPORT_SYMBOL_GPL(edac_debug_level
);
46 /* EDAC Controls, setable by module parameter, and sysfs */
47 static int log_ue
= 1;
48 static int log_ce
= 1;
49 static int panic_on_ue
;
50 static int poll_msec
= 1000;
52 /* lock to memory controller's control array */
53 static DECLARE_MUTEX(mem_ctls_mutex
);
54 static struct list_head mc_devices
= LIST_HEAD_INIT(mc_devices
);
56 static struct task_struct
*edac_thread
;
59 static int check_pci_parity
= 0; /* default YES check PCI parity */
60 static int panic_on_pci_parity
; /* default no panic on PCI Parity */
61 static atomic_t pci_parity_count
= ATOMIC_INIT(0);
63 static struct kobject edac_pci_kobj
; /* /sys/devices/system/edac/pci */
64 static struct completion edac_pci_kobj_complete
;
65 #endif /* CONFIG_PCI */
67 /* START sysfs data and methods */
70 static const char *mem_types
[] = {
71 [MEM_EMPTY
] = "Empty",
72 [MEM_RESERVED
] = "Reserved",
73 [MEM_UNKNOWN
] = "Unknown",
77 [MEM_SDR
] = "Unbuffered-SDR",
78 [MEM_RDR
] = "Registered-SDR",
79 [MEM_DDR
] = "Unbuffered-DDR",
80 [MEM_RDDR
] = "Registered-DDR",
84 static const char *dev_types
[] = {
85 [DEV_UNKNOWN
] = "Unknown",
95 static const char *edac_caps
[] = {
96 [EDAC_UNKNOWN
] = "Unknown",
98 [EDAC_RESERVED
] = "Reserved",
99 [EDAC_PARITY
] = "PARITY",
101 [EDAC_SECDED
] = "SECDED",
102 [EDAC_S2ECD2ED
] = "S2ECD2ED",
103 [EDAC_S4ECD4ED
] = "S4ECD4ED",
104 [EDAC_S8ECD8ED
] = "S8ECD8ED",
105 [EDAC_S16ECD16ED
] = "S16ECD16ED"
108 /* sysfs object: /sys/devices/system/edac */
109 static struct sysdev_class edac_class
= {
110 set_kset_name("edac"),
114 * /sys/devices/system/edac/mc
116 static struct kobject edac_memctrl_kobj
;
118 /* We use these to wait for the reference counts on edac_memctrl_kobj and
119 * edac_pci_kobj to reach 0.
121 static struct completion edac_memctrl_kobj_complete
;
124 * /sys/devices/system/edac/mc;
125 * data structures and methods
127 static ssize_t
memctrl_int_show(void *ptr
, char *buffer
)
129 int *value
= (int*) ptr
;
130 return sprintf(buffer
, "%u\n", *value
);
133 static ssize_t
memctrl_int_store(void *ptr
, const char *buffer
, size_t count
)
135 int *value
= (int*) ptr
;
137 if (isdigit(*buffer
))
138 *value
= simple_strtoul(buffer
, NULL
, 0);
143 struct memctrl_dev_attribute
{
144 struct attribute attr
;
146 ssize_t (*show
)(void *,char *);
147 ssize_t (*store
)(void *, const char *, size_t);
150 /* Set of show/store abstract level functions for memory control object */
151 static ssize_t
memctrl_dev_show(struct kobject
*kobj
,
152 struct attribute
*attr
, char *buffer
)
154 struct memctrl_dev_attribute
*memctrl_dev
;
155 memctrl_dev
= (struct memctrl_dev_attribute
*)attr
;
157 if (memctrl_dev
->show
)
158 return memctrl_dev
->show(memctrl_dev
->value
, buffer
);
163 static ssize_t
memctrl_dev_store(struct kobject
*kobj
, struct attribute
*attr
,
164 const char *buffer
, size_t count
)
166 struct memctrl_dev_attribute
*memctrl_dev
;
167 memctrl_dev
= (struct memctrl_dev_attribute
*)attr
;
169 if (memctrl_dev
->store
)
170 return memctrl_dev
->store(memctrl_dev
->value
, buffer
, count
);
175 static struct sysfs_ops memctrlfs_ops
= {
176 .show
= memctrl_dev_show
,
177 .store
= memctrl_dev_store
180 #define MEMCTRL_ATTR(_name,_mode,_show,_store) \
181 struct memctrl_dev_attribute attr_##_name = { \
182 .attr = {.name = __stringify(_name), .mode = _mode }, \
188 #define MEMCTRL_STRING_ATTR(_name,_data,_mode,_show,_store) \
189 struct memctrl_dev_attribute attr_##_name = { \
190 .attr = {.name = __stringify(_name), .mode = _mode }, \
196 /* csrow<id> control files */
197 MEMCTRL_ATTR(panic_on_ue
,S_IRUGO
|S_IWUSR
,memctrl_int_show
,memctrl_int_store
);
198 MEMCTRL_ATTR(log_ue
,S_IRUGO
|S_IWUSR
,memctrl_int_show
,memctrl_int_store
);
199 MEMCTRL_ATTR(log_ce
,S_IRUGO
|S_IWUSR
,memctrl_int_show
,memctrl_int_store
);
200 MEMCTRL_ATTR(poll_msec
,S_IRUGO
|S_IWUSR
,memctrl_int_show
,memctrl_int_store
);
202 /* Base Attributes of the memory ECC object */
203 static struct memctrl_dev_attribute
*memctrl_attr
[] = {
211 /* Main MC kobject release() function */
212 static void edac_memctrl_master_release(struct kobject
*kobj
)
214 debugf1("%s()\n", __func__
);
215 complete(&edac_memctrl_kobj_complete
);
218 static struct kobj_type ktype_memctrl
= {
219 .release
= edac_memctrl_master_release
,
220 .sysfs_ops
= &memctrlfs_ops
,
221 .default_attrs
= (struct attribute
**) memctrl_attr
,
224 /* Initialize the main sysfs entries for edac:
225 * /sys/devices/system/edac
232 static int edac_sysfs_memctrl_setup(void)
236 debugf1("%s()\n", __func__
);
238 /* create the /sys/devices/system/edac directory */
239 err
= sysdev_class_register(&edac_class
);
242 debugf1("%s() error=%d\n", __func__
, err
);
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
;
251 /* generate sysfs "..../edac/mc" */
252 err
= kobject_set_name(&edac_memctrl_kobj
,"mc");
257 /* FIXME: maybe new sysdev_create_subdir() */
258 err
= kobject_register(&edac_memctrl_kobj
);
261 debugf1("Failed to register '.../edac/mc'\n");
265 debugf1("Registered '.../edac/mc' kobject\n");
270 sysdev_class_unregister(&edac_class
);
276 * the '..../edac/mc' kobject followed by '..../edac' itself
278 static void edac_sysfs_memctrl_teardown(void)
280 debugf0("MC: " __FILE__
": %s()\n", __func__
);
282 /* Unregister the MC's kobject and wait for reference count to reach
285 init_completion(&edac_memctrl_kobj_complete
);
286 kobject_unregister(&edac_memctrl_kobj
);
287 wait_for_completion(&edac_memctrl_kobj_complete
);
289 /* Unregister the 'edac' object */
290 sysdev_class_unregister(&edac_class
);
294 static ssize_t
edac_pci_int_show(void *ptr
, char *buffer
)
297 return sprintf(buffer
,"%d\n",*value
);
300 static ssize_t
edac_pci_int_store(void *ptr
, const char *buffer
, size_t count
)
304 if (isdigit(*buffer
))
305 *value
= simple_strtoul(buffer
,NULL
,0);
310 struct edac_pci_dev_attribute
{
311 struct attribute attr
;
313 ssize_t (*show
)(void *,char *);
314 ssize_t (*store
)(void *, const char *,size_t);
317 /* Set of show/store abstract level functions for PCI Parity object */
318 static ssize_t
edac_pci_dev_show(struct kobject
*kobj
, struct attribute
*attr
,
321 struct edac_pci_dev_attribute
*edac_pci_dev
;
322 edac_pci_dev
= (struct edac_pci_dev_attribute
*)attr
;
324 if (edac_pci_dev
->show
)
325 return edac_pci_dev
->show(edac_pci_dev
->value
, buffer
);
329 static ssize_t
edac_pci_dev_store(struct kobject
*kobj
,
330 struct attribute
*attr
, const char *buffer
, size_t count
)
332 struct edac_pci_dev_attribute
*edac_pci_dev
;
333 edac_pci_dev
= (struct edac_pci_dev_attribute
*)attr
;
335 if (edac_pci_dev
->show
)
336 return edac_pci_dev
->store(edac_pci_dev
->value
, buffer
, count
);
340 static struct sysfs_ops edac_pci_sysfs_ops
= {
341 .show
= edac_pci_dev_show
,
342 .store
= edac_pci_dev_store
345 #define EDAC_PCI_ATTR(_name,_mode,_show,_store) \
346 struct edac_pci_dev_attribute edac_pci_attr_##_name = { \
347 .attr = {.name = __stringify(_name), .mode = _mode }, \
353 #define EDAC_PCI_STRING_ATTR(_name,_data,_mode,_show,_store) \
354 struct edac_pci_dev_attribute edac_pci_attr_##_name = { \
355 .attr = {.name = __stringify(_name), .mode = _mode }, \
361 /* PCI Parity control files */
362 EDAC_PCI_ATTR(check_pci_parity
, S_IRUGO
|S_IWUSR
, edac_pci_int_show
,
364 EDAC_PCI_ATTR(panic_on_pci_parity
, S_IRUGO
|S_IWUSR
, edac_pci_int_show
,
366 EDAC_PCI_ATTR(pci_parity_count
, S_IRUGO
, edac_pci_int_show
, NULL
);
368 /* Base Attributes of the memory ECC object */
369 static 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
,
376 /* No memory to release */
377 static void edac_pci_release(struct kobject
*kobj
)
379 debugf1("%s()\n", __func__
);
380 complete(&edac_pci_kobj_complete
);
383 static 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
,
390 * edac_sysfs_pci_setup()
393 static int edac_sysfs_pci_setup(void)
397 debugf1("%s()\n", __func__
);
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");
405 /* Instanstiate the csrow object */
406 /* FIXME: maybe new sysdev_create_subdir() */
407 err
= kobject_register(&edac_pci_kobj
);
410 debugf1("Failed to register '.../edac/pci'\n");
412 debugf1("Registered '.../edac/pci' kobject\n");
418 static void edac_sysfs_pci_teardown(void)
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
);
427 static u16
get_pci_parity_status(struct pci_dev
*dev
, int secondary
)
432 where
= secondary
? PCI_SEC_STATUS
: PCI_STATUS
;
433 pci_read_config_word(dev
, where
, &status
);
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
440 if (status
== 0xFFFF) {
443 pci_read_config_dword(dev
, 0, &sanity
);
445 if (sanity
== 0xFFFFFFFF)
449 status
&= PCI_STATUS_DETECTED_PARITY
| PCI_STATUS_SIG_SYSTEM_ERROR
|
453 /* reset only the bits we are interested in */
454 pci_write_config_word(dev
, where
, status
);
459 typedef void (*pci_parity_check_fn_t
) (struct pci_dev
*dev
);
461 /* Clear any PCI parity errors logged by this device. */
462 static void edac_pci_dev_parity_clear(struct pci_dev
*dev
)
466 get_pci_parity_status(dev
, 0);
468 /* read the device TYPE, looking for bridges */
469 pci_read_config_byte(dev
, PCI_HEADER_TYPE
, &header_type
);
471 if ((header_type
& 0x7F) == PCI_HEADER_TYPE_BRIDGE
)
472 get_pci_parity_status(dev
, 1);
479 static void edac_pci_dev_parity_test(struct pci_dev
*dev
)
484 /* read the STATUS register on this device
486 status
= get_pci_parity_status(dev
, 0);
488 debugf2("PCI STATUS= 0x%04x %s\n", status
, dev
->dev
.bus_id
);
490 /* check the status reg for errors */
492 if (status
& (PCI_STATUS_SIG_SYSTEM_ERROR
))
493 edac_printk(KERN_CRIT
, EDAC_PCI
,
494 "Signaled System Error on %s\n",
497 if (status
& (PCI_STATUS_PARITY
)) {
498 edac_printk(KERN_CRIT
, EDAC_PCI
,
499 "Master Data Parity Error on %s\n",
502 atomic_inc(&pci_parity_count
);
505 if (status
& (PCI_STATUS_DETECTED_PARITY
)) {
506 edac_printk(KERN_CRIT
, EDAC_PCI
,
507 "Detected Parity Error on %s\n",
510 atomic_inc(&pci_parity_count
);
514 /* read the device TYPE, looking for bridges */
515 pci_read_config_byte(dev
, PCI_HEADER_TYPE
, &header_type
);
517 debugf2("PCI HEADER TYPE= 0x%02x %s\n", header_type
, dev
->dev
.bus_id
);
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);
523 debugf2("PCI SEC_STATUS= 0x%04x %s\n",
524 status
, dev
->dev
.bus_id
);
526 /* check the secondary status reg for errors */
528 if (status
& (PCI_STATUS_SIG_SYSTEM_ERROR
))
529 edac_printk(KERN_CRIT
, EDAC_PCI
, "Bridge "
530 "Signaled System Error on %s\n",
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
));
538 atomic_inc(&pci_parity_count
);
541 if (status
& (PCI_STATUS_DETECTED_PARITY
)) {
542 edac_printk(KERN_CRIT
, EDAC_PCI
, "Bridge "
543 "Detected Parity Error on %s\n",
546 atomic_inc(&pci_parity_count
);
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
557 static inline void edac_pci_dev_parity_iterator(pci_parity_check_fn_t fn
)
559 struct pci_dev
*dev
= NULL
;
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
565 while((dev
= pci_get_device(PCI_ANY_ID
, PCI_ANY_ID
, dev
)) != NULL
) {
570 static void do_pci_parity_check(void)
575 debugf3("%s()\n", __func__
);
577 if (!check_pci_parity
)
580 before_count
= atomic_read(&pci_parity_count
);
582 /* scan all PCI devices looking for a Parity Error on devices and
585 local_irq_save(flags
);
586 edac_pci_dev_parity_iterator(edac_pci_dev_parity_test
);
587 local_irq_restore(flags
);
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");
597 static inline void clear_pci_parity_errors(void)
599 /* Clear any PCI bus parity errors that devices initially have logged
600 * in their registers.
602 edac_pci_dev_parity_iterator(edac_pci_dev_parity_clear
);
605 #else /* CONFIG_PCI */
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)
613 #endif /* CONFIG_PCI */
615 /* EDAC sysfs CSROW data structures and methods
618 /* Set of more default csrow<id> attribute show/store functions */
619 static ssize_t
csrow_ue_count_show(struct csrow_info
*csrow
, char *data
, int private)
621 return sprintf(data
,"%u\n", csrow
->ue_count
);
624 static ssize_t
csrow_ce_count_show(struct csrow_info
*csrow
, char *data
, int private)
626 return sprintf(data
,"%u\n", csrow
->ce_count
);
629 static ssize_t
csrow_size_show(struct csrow_info
*csrow
, char *data
, int private)
631 return sprintf(data
,"%u\n", PAGES_TO_MiB(csrow
->nr_pages
));
634 static ssize_t
csrow_mem_type_show(struct csrow_info
*csrow
, char *data
, int private)
636 return sprintf(data
,"%s\n", mem_types
[csrow
->mtype
]);
639 static ssize_t
csrow_dev_type_show(struct csrow_info
*csrow
, char *data
, int private)
641 return sprintf(data
,"%s\n", dev_types
[csrow
->dtype
]);
644 static ssize_t
csrow_edac_mode_show(struct csrow_info
*csrow
, char *data
, int private)
646 return sprintf(data
,"%s\n", edac_caps
[csrow
->edac_mode
]);
649 /* show/store functions for DIMM Label attributes */
650 static ssize_t
channel_dimm_label_show(struct csrow_info
*csrow
,
651 char *data
, int channel
)
653 return snprintf(data
, EDAC_MC_LABEL_LEN
,"%s",
654 csrow
->channels
[channel
].label
);
657 static ssize_t
channel_dimm_label_store(struct csrow_info
*csrow
,
662 ssize_t max_size
= 0;
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';
671 /* show function for dynamic chX_ce_count attribute */
672 static ssize_t
channel_ce_count_show(struct csrow_info
*csrow
,
676 return sprintf(data
, "%u\n", csrow
->channels
[channel
].ce_count
);
679 /* csrow specific attribute structure */
680 struct 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);
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)
690 /* Set of show/store higher level functions for default csrow attributes */
691 static ssize_t
csrowdev_show(struct kobject
*kobj
,
692 struct attribute
*attr
,
695 struct csrow_info
*csrow
= to_csrow(kobj
);
696 struct csrowdev_attribute
*csrowdev_attr
= to_csrowdev_attr(attr
);
698 if (csrowdev_attr
->show
)
699 return csrowdev_attr
->show(csrow
,
701 csrowdev_attr
->private);
705 static ssize_t
csrowdev_store(struct kobject
*kobj
, struct attribute
*attr
,
706 const char *buffer
, size_t count
)
708 struct csrow_info
*csrow
= to_csrow(kobj
);
709 struct csrowdev_attribute
* csrowdev_attr
= to_csrowdev_attr(attr
);
711 if (csrowdev_attr
->store
)
712 return csrowdev_attr
->store(csrow
,
715 csrowdev_attr
->private);
719 static struct sysfs_ops csrowfs_ops
= {
720 .show
= csrowdev_show
,
721 .store
= csrowdev_store
724 #define CSROWDEV_ATTR(_name,_mode,_show,_store,_private) \
725 struct csrowdev_attribute attr_##_name = { \
726 .attr = {.name = __stringify(_name), .mode = _mode }, \
729 .private = _private, \
732 /* default cwrow<id>/attribute files */
733 CSROWDEV_ATTR(size_mb
,S_IRUGO
,csrow_size_show
,NULL
,0);
734 CSROWDEV_ATTR(dev_type
,S_IRUGO
,csrow_dev_type_show
,NULL
,0);
735 CSROWDEV_ATTR(mem_type
,S_IRUGO
,csrow_mem_type_show
,NULL
,0);
736 CSROWDEV_ATTR(edac_mode
,S_IRUGO
,csrow_edac_mode_show
,NULL
,0);
737 CSROWDEV_ATTR(ue_count
,S_IRUGO
,csrow_ue_count_show
,NULL
,0);
738 CSROWDEV_ATTR(ce_count
,S_IRUGO
,csrow_ce_count_show
,NULL
,0);
740 /* default attributes of the CSROW<id> object */
741 static struct csrowdev_attribute
*default_csrow_attr
[] = {
752 /* possible dynamic channel DIMM Label attribute files */
753 CSROWDEV_ATTR(ch0_dimm_label
,S_IRUGO
|S_IWUSR
,
754 channel_dimm_label_show
,
755 channel_dimm_label_store
,
757 CSROWDEV_ATTR(ch1_dimm_label
,S_IRUGO
|S_IWUSR
,
758 channel_dimm_label_show
,
759 channel_dimm_label_store
,
761 CSROWDEV_ATTR(ch2_dimm_label
,S_IRUGO
|S_IWUSR
,
762 channel_dimm_label_show
,
763 channel_dimm_label_store
,
765 CSROWDEV_ATTR(ch3_dimm_label
,S_IRUGO
|S_IWUSR
,
766 channel_dimm_label_show
,
767 channel_dimm_label_store
,
769 CSROWDEV_ATTR(ch4_dimm_label
,S_IRUGO
|S_IWUSR
,
770 channel_dimm_label_show
,
771 channel_dimm_label_store
,
773 CSROWDEV_ATTR(ch5_dimm_label
,S_IRUGO
|S_IWUSR
,
774 channel_dimm_label_show
,
775 channel_dimm_label_store
,
778 /* Total possible dynamic DIMM Label attribute file table */
779 static 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
,
788 /* possible dynamic channel ce_count attribute files */
789 CSROWDEV_ATTR(ch0_ce_count
,S_IRUGO
|S_IWUSR
,
790 channel_ce_count_show
,
793 CSROWDEV_ATTR(ch1_ce_count
,S_IRUGO
|S_IWUSR
,
794 channel_ce_count_show
,
797 CSROWDEV_ATTR(ch2_ce_count
,S_IRUGO
|S_IWUSR
,
798 channel_ce_count_show
,
801 CSROWDEV_ATTR(ch3_ce_count
,S_IRUGO
|S_IWUSR
,
802 channel_ce_count_show
,
805 CSROWDEV_ATTR(ch4_ce_count
,S_IRUGO
|S_IWUSR
,
806 channel_ce_count_show
,
809 CSROWDEV_ATTR(ch5_ce_count
,S_IRUGO
|S_IWUSR
,
810 channel_ce_count_show
,
814 /* Total possible dynamic ce_count attribute file table */
815 static struct csrowdev_attribute
*dynamic_csrow_ce_count_attr
[] = {
825 #define EDAC_NR_CHANNELS 6
827 /* Create dynamic CHANNEL files, indexed by 'chan', under specifed CSROW */
828 static int edac_create_channel_files(struct kobject
*kobj
, int chan
)
832 if (chan
>= EDAC_NR_CHANNELS
)
835 /* create the DIMM label attribute file */
836 err
= sysfs_create_file(kobj
,
837 (struct attribute
*) dynamic_csrow_dimm_attr
[chan
]);
840 /* create the CE Count attribute file */
841 err
= sysfs_create_file(kobj
,
842 (struct attribute
*) dynamic_csrow_ce_count_attr
[chan
]);
844 debugf1("%s() dimm labels and ce_count files created", __func__
);
850 /* No memory to release for this kobj */
851 static void edac_csrow_instance_release(struct kobject
*kobj
)
853 struct csrow_info
*cs
;
855 cs
= container_of(kobj
, struct csrow_info
, kobj
);
856 complete(&cs
->kobj_complete
);
859 /* the kobj_type instance for a CSROW */
860 static struct kobj_type ktype_csrow
= {
861 .release
= edac_csrow_instance_release
,
862 .sysfs_ops
= &csrowfs_ops
,
863 .default_attrs
= (struct attribute
**) default_csrow_attr
,
866 /* Create a CSROW object under specifed edac_mc_device */
867 static int edac_create_csrow_object(
868 struct kobject
*edac_mci_kobj
,
869 struct csrow_info
*csrow
,
875 memset(&csrow
->kobj
, 0, sizeof(csrow
->kobj
));
877 /* generate ..../edac/mc/mc<id>/csrow<index> */
879 csrow
->kobj
.parent
= edac_mci_kobj
;
880 csrow
->kobj
.ktype
= &ktype_csrow
;
882 /* name this instance of csrow<id> */
883 err
= kobject_set_name(&csrow
->kobj
,"csrow%d",index
);
887 /* Instanstiate the csrow object */
888 err
= kobject_register(&csrow
->kobj
);
890 /* Create the dyanmic attribute files on this csrow,
891 * namely, the DIMM labels and the channel ce_count
893 for (chan
= 0; chan
< csrow
->nr_channels
; chan
++) {
894 err
= edac_create_channel_files(&csrow
->kobj
,chan
);
904 /* default sysfs methods and data structures for the main MCI kobject */
906 static ssize_t
mci_reset_counters_store(struct mem_ctl_info
*mci
,
907 const char *data
, size_t count
)
911 mci
->ue_noinfo_count
= 0;
912 mci
->ce_noinfo_count
= 0;
916 for (row
= 0; row
< mci
->nr_csrows
; row
++) {
917 struct csrow_info
*ri
= &mci
->csrows
[row
];
922 for (chan
= 0; chan
< ri
->nr_channels
; chan
++)
923 ri
->channels
[chan
].ce_count
= 0;
926 mci
->start_time
= jiffies
;
930 /* memory scrubbing */
931 static ssize_t
mci_sdram_scrub_rate_store(struct mem_ctl_info
*mci
,
932 const char *data
, size_t count
)
936 if (mci
->set_sdram_scrub_rate
) {
938 memctrl_int_store(&bandwidth
, data
, count
);
940 if (!(*mci
->set_sdram_scrub_rate
)(mci
, &bandwidth
)) {
941 edac_printk(KERN_DEBUG
, EDAC_MC
,
942 "Scrub rate set successfully, applied: %d\n",
945 /* FIXME: error codes maybe? */
946 edac_printk(KERN_DEBUG
, EDAC_MC
,
947 "Scrub rate set FAILED, could not apply: %d\n",
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");
958 static ssize_t
mci_sdram_scrub_rate_show(struct mem_ctl_info
*mci
, char *data
)
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",
968 /* FIXME: error codes maybe? */
969 edac_printk(KERN_DEBUG
, EDAC_MC
,
970 "Scrub rate fetch FAILED, got: %d\n",
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");
978 return sprintf(data
, "%d\n", bandwidth
);
981 /* default attribute files for the MCI object */
982 static ssize_t
mci_ue_count_show(struct mem_ctl_info
*mci
, char *data
)
984 return sprintf(data
,"%d\n", mci
->ue_count
);
987 static ssize_t
mci_ce_count_show(struct mem_ctl_info
*mci
, char *data
)
989 return sprintf(data
,"%d\n", mci
->ce_count
);
992 static ssize_t
mci_ce_noinfo_show(struct mem_ctl_info
*mci
, char *data
)
994 return sprintf(data
,"%d\n", mci
->ce_noinfo_count
);
997 static ssize_t
mci_ue_noinfo_show(struct mem_ctl_info
*mci
, char *data
)
999 return sprintf(data
,"%d\n", mci
->ue_noinfo_count
);
1002 static ssize_t
mci_seconds_show(struct mem_ctl_info
*mci
, char *data
)
1004 return sprintf(data
,"%ld\n", (jiffies
- mci
->start_time
) / HZ
);
1007 static ssize_t
mci_ctl_name_show(struct mem_ctl_info
*mci
, char *data
)
1009 return sprintf(data
,"%s\n", mci
->ctl_name
);
1012 static ssize_t
mci_size_mb_show(struct mem_ctl_info
*mci
, char *data
)
1014 int total_pages
, csrow_idx
;
1016 for (total_pages
= csrow_idx
= 0; csrow_idx
< mci
->nr_csrows
;
1018 struct csrow_info
*csrow
= &mci
->csrows
[csrow_idx
];
1020 if (!csrow
->nr_pages
)
1023 total_pages
+= csrow
->nr_pages
;
1026 return sprintf(data
,"%u\n", PAGES_TO_MiB(total_pages
));
1029 struct 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);
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)
1038 /* MCI show/store functions for top most object */
1039 static ssize_t
mcidev_show(struct kobject
*kobj
, struct attribute
*attr
,
1042 struct mem_ctl_info
*mem_ctl_info
= to_mci(kobj
);
1043 struct mcidev_attribute
* mcidev_attr
= to_mcidev_attr(attr
);
1045 if (mcidev_attr
->show
)
1046 return mcidev_attr
->show(mem_ctl_info
, buffer
);
1051 static ssize_t
mcidev_store(struct kobject
*kobj
, struct attribute
*attr
,
1052 const char *buffer
, size_t count
)
1054 struct mem_ctl_info
*mem_ctl_info
= to_mci(kobj
);
1055 struct mcidev_attribute
* mcidev_attr
= to_mcidev_attr(attr
);
1057 if (mcidev_attr
->store
)
1058 return mcidev_attr
->store(mem_ctl_info
, buffer
, count
);
1063 static struct sysfs_ops mci_ops
= {
1064 .show
= mcidev_show
,
1065 .store
= mcidev_store
1068 #define MCIDEV_ATTR(_name,_mode,_show,_store) \
1069 struct mcidev_attribute mci_attr_##_name = { \
1070 .attr = {.name = __stringify(_name), .mode = _mode }, \
1075 /* default Control file */
1076 MCIDEV_ATTR(reset_counters
,S_IWUSR
,NULL
,mci_reset_counters_store
);
1078 /* default Attribute files */
1079 MCIDEV_ATTR(mc_name
,S_IRUGO
,mci_ctl_name_show
,NULL
);
1080 MCIDEV_ATTR(size_mb
,S_IRUGO
,mci_size_mb_show
,NULL
);
1081 MCIDEV_ATTR(seconds_since_reset
,S_IRUGO
,mci_seconds_show
,NULL
);
1082 MCIDEV_ATTR(ue_noinfo_count
,S_IRUGO
,mci_ue_noinfo_show
,NULL
);
1083 MCIDEV_ATTR(ce_noinfo_count
,S_IRUGO
,mci_ce_noinfo_show
,NULL
);
1084 MCIDEV_ATTR(ue_count
,S_IRUGO
,mci_ue_count_show
,NULL
);
1085 MCIDEV_ATTR(ce_count
,S_IRUGO
,mci_ce_count_show
,NULL
);
1087 /* memory scrubber attribute file */
1088 MCIDEV_ATTR(sdram_scrub_rate
,S_IRUGO
|S_IWUSR
,mci_sdram_scrub_rate_show
,mci_sdram_scrub_rate_store
);
1090 static struct mcidev_attribute
*mci_attr
[] = {
1091 &mci_attr_reset_counters
,
1094 &mci_attr_seconds_since_reset
,
1095 &mci_attr_ue_noinfo_count
,
1096 &mci_attr_ce_noinfo_count
,
1099 &mci_attr_sdram_scrub_rate
,
1104 * Release of a MC controlling instance
1106 static void edac_mci_instance_release(struct kobject
*kobj
)
1108 struct mem_ctl_info
*mci
;
1111 debugf0("%s() idx=%d\n", __func__
, mci
->mc_idx
);
1112 complete(&mci
->kobj_complete
);
1115 static struct kobj_type ktype_mci
= {
1116 .release
= edac_mci_instance_release
,
1117 .sysfs_ops
= &mci_ops
,
1118 .default_attrs
= (struct attribute
**) mci_attr
,
1122 #define EDAC_DEVICE_SYMLINK "device"
1125 * Create a new Memory Controller kobject instance,
1126 * mc<id> under the 'mc' directory
1132 static int edac_create_sysfs_mci_device(struct mem_ctl_info
*mci
)
1136 struct csrow_info
*csrow
;
1137 struct kobject
*edac_mci_kobj
=&mci
->edac_mci_kobj
;
1139 debugf0("%s() idx=%d\n", __func__
, mci
->mc_idx
);
1140 memset(edac_mci_kobj
, 0, sizeof(*edac_mci_kobj
));
1142 /* set the name of the mc<id> object */
1143 err
= kobject_set_name(edac_mci_kobj
,"mc%d",mci
->mc_idx
);
1147 /* link to our parent the '..../edac/mc' object */
1148 edac_mci_kobj
->parent
= &edac_memctrl_kobj
;
1149 edac_mci_kobj
->ktype
= &ktype_mci
;
1151 /* register the mc<id> kobject */
1152 err
= kobject_register(edac_mci_kobj
);
1156 /* create a symlink for the device */
1157 err
= sysfs_create_link(edac_mci_kobj
, &mci
->dev
->kobj
,
1158 EDAC_DEVICE_SYMLINK
);
1162 /* Make directories for each CSROW object
1163 * under the mc<id> kobject
1165 for (i
= 0; i
< mci
->nr_csrows
; i
++) {
1166 csrow
= &mci
->csrows
[i
];
1168 /* Only expose populated CSROWs */
1169 if (csrow
->nr_pages
> 0) {
1170 err
= edac_create_csrow_object(edac_mci_kobj
,csrow
,i
);
1178 /* CSROW error: backout what has already been registered, */
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
);
1189 init_completion(&mci
->kobj_complete
);
1190 kobject_unregister(edac_mci_kobj
);
1191 wait_for_completion(&mci
->kobj_complete
);
1196 * remove a Memory Controller instance
1198 static void edac_remove_sysfs_mci_device(struct mem_ctl_info
*mci
)
1202 debugf0("%s()\n", __func__
);
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
);
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
);
1219 /* END OF sysfs data and methods */
1221 #ifdef CONFIG_EDAC_DEBUG
1223 void edac_mc_dump_channel(struct channel_info
*chan
)
1225 debugf4("\tchannel = %p\n", chan
);
1226 debugf4("\tchannel->chan_idx = %d\n", chan
->chan_idx
);
1227 debugf4("\tchannel->ce_count = %d\n", chan
->ce_count
);
1228 debugf4("\tchannel->label = '%s'\n", chan
->label
);
1229 debugf4("\tchannel->csrow = %p\n\n", chan
->csrow
);
1231 EXPORT_SYMBOL_GPL(edac_mc_dump_channel
);
1233 void edac_mc_dump_csrow(struct csrow_info
*csrow
)
1235 debugf4("\tcsrow = %p\n", csrow
);
1236 debugf4("\tcsrow->csrow_idx = %d\n", csrow
->csrow_idx
);
1237 debugf4("\tcsrow->first_page = 0x%lx\n",
1239 debugf4("\tcsrow->last_page = 0x%lx\n", csrow
->last_page
);
1240 debugf4("\tcsrow->page_mask = 0x%lx\n", csrow
->page_mask
);
1241 debugf4("\tcsrow->nr_pages = 0x%x\n", csrow
->nr_pages
);
1242 debugf4("\tcsrow->nr_channels = %d\n",
1243 csrow
->nr_channels
);
1244 debugf4("\tcsrow->channels = %p\n", csrow
->channels
);
1245 debugf4("\tcsrow->mci = %p\n\n", csrow
->mci
);
1247 EXPORT_SYMBOL_GPL(edac_mc_dump_csrow
);
1249 void edac_mc_dump_mci(struct mem_ctl_info
*mci
)
1251 debugf3("\tmci = %p\n", mci
);
1252 debugf3("\tmci->mtype_cap = %lx\n", mci
->mtype_cap
);
1253 debugf3("\tmci->edac_ctl_cap = %lx\n", mci
->edac_ctl_cap
);
1254 debugf3("\tmci->edac_cap = %lx\n", mci
->edac_cap
);
1255 debugf4("\tmci->edac_check = %p\n", mci
->edac_check
);
1256 debugf3("\tmci->nr_csrows = %d, csrows = %p\n",
1257 mci
->nr_csrows
, mci
->csrows
);
1258 debugf3("\tdev = %p\n", mci
->dev
);
1259 debugf3("\tmod_name:ctl_name = %s:%s\n",
1260 mci
->mod_name
, mci
->ctl_name
);
1261 debugf3("\tpvt_info = %p\n\n", mci
->pvt_info
);
1263 EXPORT_SYMBOL_GPL(edac_mc_dump_mci
);
1265 #endif /* CONFIG_EDAC_DEBUG */
1267 /* 'ptr' points to a possibly unaligned item X such that sizeof(X) is 'size'.
1268 * Adjust 'ptr' so that its alignment is at least as stringent as what the
1269 * compiler would provide for X and return the aligned result.
1271 * If 'size' is a constant, the compiler will optimize this whole function
1272 * down to either a no-op or the addition of a constant to the value of 'ptr'.
1274 static inline char * align_ptr(void *ptr
, unsigned size
)
1278 /* Here we assume that the alignment of a "long long" is the most
1279 * stringent alignment that the compiler will ever provide by default.
1280 * As far as I know, this is a reasonable assumption.
1282 if (size
> sizeof(long))
1283 align
= sizeof(long long);
1284 else if (size
> sizeof(int))
1285 align
= sizeof(long);
1286 else if (size
> sizeof(short))
1287 align
= sizeof(int);
1288 else if (size
> sizeof(char))
1289 align
= sizeof(short);
1291 return (char *) ptr
;
1296 return (char *) ptr
;
1298 return (char *) (((unsigned long) ptr
) + align
- r
);
1302 * edac_mc_alloc: Allocate a struct mem_ctl_info structure
1303 * @size_pvt: size of private storage needed
1304 * @nr_csrows: Number of CWROWS needed for this MC
1305 * @nr_chans: Number of channels for the MC
1307 * Everything is kmalloc'ed as one big chunk - more efficient.
1308 * Only can be used if all structures have the same lifetime - otherwise
1309 * you have to allocate and initialize your own structures.
1311 * Use edac_mc_free() to free mc structures allocated by this function.
1314 * NULL allocation failed
1315 * struct mem_ctl_info pointer
1317 struct mem_ctl_info
*edac_mc_alloc(unsigned sz_pvt
, unsigned nr_csrows
,
1320 struct mem_ctl_info
*mci
;
1321 struct csrow_info
*csi
, *csrow
;
1322 struct channel_info
*chi
, *chp
, *chan
;
1327 /* Figure out the offsets of the various items from the start of an mc
1328 * structure. We want the alignment of each item to be at least as
1329 * stringent as what the compiler would provide if we could simply
1330 * hardcode everything into a single struct.
1332 mci
= (struct mem_ctl_info
*) 0;
1333 csi
= (struct csrow_info
*)align_ptr(&mci
[1], sizeof(*csi
));
1334 chi
= (struct channel_info
*)
1335 align_ptr(&csi
[nr_csrows
], sizeof(*chi
));
1336 pvt
= align_ptr(&chi
[nr_chans
* nr_csrows
], sz_pvt
);
1337 size
= ((unsigned long) pvt
) + sz_pvt
;
1339 if ((mci
= kmalloc(size
, GFP_KERNEL
)) == NULL
)
1342 /* Adjust pointers so they point within the memory we just allocated
1343 * rather than an imaginary chunk of memory located at address 0.
1345 csi
= (struct csrow_info
*) (((char *) mci
) + ((unsigned long) csi
));
1346 chi
= (struct channel_info
*) (((char *) mci
) + ((unsigned long) chi
));
1347 pvt
= sz_pvt
? (((char *) mci
) + ((unsigned long) pvt
)) : NULL
;
1349 memset(mci
, 0, size
); /* clear all fields */
1351 mci
->pvt_info
= pvt
;
1352 mci
->nr_csrows
= nr_csrows
;
1354 for (row
= 0; row
< nr_csrows
; row
++) {
1356 csrow
->csrow_idx
= row
;
1358 csrow
->nr_channels
= nr_chans
;
1359 chp
= &chi
[row
* nr_chans
];
1360 csrow
->channels
= chp
;
1362 for (chn
= 0; chn
< nr_chans
; chn
++) {
1364 chan
->chan_idx
= chn
;
1365 chan
->csrow
= csrow
;
1371 EXPORT_SYMBOL_GPL(edac_mc_alloc
);
1374 * edac_mc_free: Free a previously allocated 'mci' structure
1375 * @mci: pointer to a struct mem_ctl_info structure
1377 void edac_mc_free(struct mem_ctl_info
*mci
)
1381 EXPORT_SYMBOL_GPL(edac_mc_free
);
1383 static struct mem_ctl_info
*find_mci_by_dev(struct device
*dev
)
1385 struct mem_ctl_info
*mci
;
1386 struct list_head
*item
;
1388 debugf3("%s()\n", __func__
);
1390 list_for_each(item
, &mc_devices
) {
1391 mci
= list_entry(item
, struct mem_ctl_info
, link
);
1393 if (mci
->dev
== dev
)
1400 /* Return 0 on success, 1 on failure.
1401 * Before calling this function, caller must
1402 * assign a unique value to mci->mc_idx.
1404 static int add_mc_to_global_list (struct mem_ctl_info
*mci
)
1406 struct list_head
*item
, *insert_before
;
1407 struct mem_ctl_info
*p
;
1409 insert_before
= &mc_devices
;
1411 if (unlikely((p
= find_mci_by_dev(mci
->dev
)) != NULL
))
1414 list_for_each(item
, &mc_devices
) {
1415 p
= list_entry(item
, struct mem_ctl_info
, link
);
1417 if (p
->mc_idx
>= mci
->mc_idx
) {
1418 if (unlikely(p
->mc_idx
== mci
->mc_idx
))
1421 insert_before
= item
;
1426 list_add_tail_rcu(&mci
->link
, insert_before
);
1430 edac_printk(KERN_WARNING
, EDAC_MC
,
1431 "%s (%s) %s %s already assigned %d\n", p
->dev
->bus_id
,
1432 dev_name(p
->dev
), p
->mod_name
, p
->ctl_name
, p
->mc_idx
);
1436 edac_printk(KERN_WARNING
, EDAC_MC
,
1437 "bug in low-level driver: attempt to assign\n"
1438 " duplicate mc_idx %d in %s()\n", p
->mc_idx
, __func__
);
1442 static void complete_mc_list_del(struct rcu_head
*head
)
1444 struct mem_ctl_info
*mci
;
1446 mci
= container_of(head
, struct mem_ctl_info
, rcu
);
1447 INIT_LIST_HEAD(&mci
->link
);
1448 complete(&mci
->complete
);
1451 static void del_mc_from_global_list(struct mem_ctl_info
*mci
)
1453 list_del_rcu(&mci
->link
);
1454 init_completion(&mci
->complete
);
1455 call_rcu(&mci
->rcu
, complete_mc_list_del
);
1456 wait_for_completion(&mci
->complete
);
1460 * edac_mc_add_mc: Insert the 'mci' structure into the mci global list and
1461 * create sysfs entries associated with mci structure
1462 * @mci: pointer to the mci structure to be added to the list
1463 * @mc_idx: A unique numeric identifier to be assigned to the 'mci' structure.
1470 /* FIXME - should a warning be printed if no error detection? correction? */
1471 int edac_mc_add_mc(struct mem_ctl_info
*mci
, int mc_idx
)
1473 debugf0("%s()\n", __func__
);
1474 mci
->mc_idx
= mc_idx
;
1475 #ifdef CONFIG_EDAC_DEBUG
1476 if (edac_debug_level
>= 3)
1477 edac_mc_dump_mci(mci
);
1479 if (edac_debug_level
>= 4) {
1482 for (i
= 0; i
< mci
->nr_csrows
; i
++) {
1485 edac_mc_dump_csrow(&mci
->csrows
[i
]);
1486 for (j
= 0; j
< mci
->csrows
[i
].nr_channels
; j
++)
1487 edac_mc_dump_channel(
1488 &mci
->csrows
[i
].channels
[j
]);
1492 down(&mem_ctls_mutex
);
1494 if (add_mc_to_global_list(mci
))
1497 /* set load time so that error rate can be tracked */
1498 mci
->start_time
= jiffies
;
1500 if (edac_create_sysfs_mci_device(mci
)) {
1501 edac_mc_printk(mci
, KERN_WARNING
,
1502 "failed to create sysfs device\n");
1506 /* Report action taken */
1507 edac_mc_printk(mci
, KERN_INFO
, "Giving out device to %s %s: DEV %s\n",
1508 mci
->mod_name
, mci
->ctl_name
, dev_name(mci
->dev
));
1510 up(&mem_ctls_mutex
);
1514 del_mc_from_global_list(mci
);
1517 up(&mem_ctls_mutex
);
1520 EXPORT_SYMBOL_GPL(edac_mc_add_mc
);
1523 * edac_mc_del_mc: Remove sysfs entries for specified mci structure and
1524 * remove mci structure from global list
1525 * @pdev: Pointer to 'struct device' representing mci structure to remove.
1527 * Return pointer to removed mci structure, or NULL if device not found.
1529 struct mem_ctl_info
* edac_mc_del_mc(struct device
*dev
)
1531 struct mem_ctl_info
*mci
;
1533 debugf0("MC: %s()\n", __func__
);
1534 down(&mem_ctls_mutex
);
1536 if ((mci
= find_mci_by_dev(dev
)) == NULL
) {
1537 up(&mem_ctls_mutex
);
1541 edac_remove_sysfs_mci_device(mci
);
1542 del_mc_from_global_list(mci
);
1543 up(&mem_ctls_mutex
);
1544 edac_printk(KERN_INFO
, EDAC_MC
,
1545 "Removed device %d for %s %s: DEV %s\n", mci
->mc_idx
,
1546 mci
->mod_name
, mci
->ctl_name
, dev_name(mci
->dev
));
1549 EXPORT_SYMBOL_GPL(edac_mc_del_mc
);
1551 void edac_mc_scrub_block(unsigned long page
, unsigned long offset
, u32 size
)
1555 unsigned long flags
= 0;
1557 debugf3("%s()\n", __func__
);
1559 /* ECC error page was not in our memory. Ignore it. */
1560 if(!pfn_valid(page
))
1563 /* Find the actual page structure then map it and fix */
1564 pg
= pfn_to_page(page
);
1566 if (PageHighMem(pg
))
1567 local_irq_save(flags
);
1569 virt_addr
= kmap_atomic(pg
, KM_BOUNCE_READ
);
1571 /* Perform architecture specific atomic scrub operation */
1572 atomic_scrub(virt_addr
+ offset
, size
);
1574 /* Unmap and complete */
1575 kunmap_atomic(virt_addr
, KM_BOUNCE_READ
);
1577 if (PageHighMem(pg
))
1578 local_irq_restore(flags
);
1580 EXPORT_SYMBOL_GPL(edac_mc_scrub_block
);
1582 /* FIXME - should return -1 */
1583 int edac_mc_find_csrow_by_page(struct mem_ctl_info
*mci
, unsigned long page
)
1585 struct csrow_info
*csrows
= mci
->csrows
;
1588 debugf1("MC%d: %s(): 0x%lx\n", mci
->mc_idx
, __func__
, page
);
1591 for (i
= 0; i
< mci
->nr_csrows
; i
++) {
1592 struct csrow_info
*csrow
= &csrows
[i
];
1594 if (csrow
->nr_pages
== 0)
1597 debugf3("MC%d: %s(): first(0x%lx) page(0x%lx) last(0x%lx) "
1598 "mask(0x%lx)\n", mci
->mc_idx
, __func__
,
1599 csrow
->first_page
, page
, csrow
->last_page
,
1602 if ((page
>= csrow
->first_page
) &&
1603 (page
<= csrow
->last_page
) &&
1604 ((page
& csrow
->page_mask
) ==
1605 (csrow
->first_page
& csrow
->page_mask
))) {
1612 edac_mc_printk(mci
, KERN_ERR
,
1613 "could not look up page error address %lx\n",
1614 (unsigned long) page
);
1618 EXPORT_SYMBOL_GPL(edac_mc_find_csrow_by_page
);
1620 /* FIXME - setable log (warning/emerg) levels */
1621 /* FIXME - integrate with evlog: http://evlog.sourceforge.net/ */
1622 void edac_mc_handle_ce(struct mem_ctl_info
*mci
,
1623 unsigned long page_frame_number
, unsigned long offset_in_page
,
1624 unsigned long syndrome
, int row
, int channel
, const char *msg
)
1626 unsigned long remapped_page
;
1628 debugf3("MC%d: %s()\n", mci
->mc_idx
, __func__
);
1630 /* FIXME - maybe make panic on INTERNAL ERROR an option */
1631 if (row
>= mci
->nr_csrows
|| row
< 0) {
1632 /* something is wrong */
1633 edac_mc_printk(mci
, KERN_ERR
,
1634 "INTERNAL ERROR: row out of range "
1635 "(%d >= %d)\n", row
, mci
->nr_csrows
);
1636 edac_mc_handle_ce_no_info(mci
, "INTERNAL ERROR");
1640 if (channel
>= mci
->csrows
[row
].nr_channels
|| channel
< 0) {
1641 /* something is wrong */
1642 edac_mc_printk(mci
, KERN_ERR
,
1643 "INTERNAL ERROR: channel out of range "
1644 "(%d >= %d)\n", channel
,
1645 mci
->csrows
[row
].nr_channels
);
1646 edac_mc_handle_ce_no_info(mci
, "INTERNAL ERROR");
1651 /* FIXME - put in DIMM location */
1652 edac_mc_printk(mci
, KERN_WARNING
,
1653 "CE page 0x%lx, offset 0x%lx, grain %d, syndrome "
1654 "0x%lx, row %d, channel %d, label \"%s\": %s\n",
1655 page_frame_number
, offset_in_page
,
1656 mci
->csrows
[row
].grain
, syndrome
, row
, channel
,
1657 mci
->csrows
[row
].channels
[channel
].label
, msg
);
1660 mci
->csrows
[row
].ce_count
++;
1661 mci
->csrows
[row
].channels
[channel
].ce_count
++;
1663 if (mci
->scrub_mode
& SCRUB_SW_SRC
) {
1665 * Some MC's can remap memory so that it is still available
1666 * at a different address when PCI devices map into memory.
1667 * MC's that can't do this lose the memory where PCI devices
1668 * are mapped. This mapping is MC dependant and so we call
1669 * back into the MC driver for it to map the MC page to
1670 * a physical (CPU) page which can then be mapped to a virtual
1671 * page - which can then be scrubbed.
1673 remapped_page
= mci
->ctl_page_to_phys
?
1674 mci
->ctl_page_to_phys(mci
, page_frame_number
) :
1677 edac_mc_scrub_block(remapped_page
, offset_in_page
,
1678 mci
->csrows
[row
].grain
);
1681 EXPORT_SYMBOL_GPL(edac_mc_handle_ce
);
1683 void edac_mc_handle_ce_no_info(struct mem_ctl_info
*mci
, const char *msg
)
1686 edac_mc_printk(mci
, KERN_WARNING
,
1687 "CE - no information available: %s\n", msg
);
1689 mci
->ce_noinfo_count
++;
1692 EXPORT_SYMBOL_GPL(edac_mc_handle_ce_no_info
);
1694 void edac_mc_handle_ue(struct mem_ctl_info
*mci
,
1695 unsigned long page_frame_number
, unsigned long offset_in_page
,
1696 int row
, const char *msg
)
1698 int len
= EDAC_MC_LABEL_LEN
* 4;
1699 char labels
[len
+ 1];
1704 debugf3("MC%d: %s()\n", mci
->mc_idx
, __func__
);
1706 /* FIXME - maybe make panic on INTERNAL ERROR an option */
1707 if (row
>= mci
->nr_csrows
|| row
< 0) {
1708 /* something is wrong */
1709 edac_mc_printk(mci
, KERN_ERR
,
1710 "INTERNAL ERROR: row out of range "
1711 "(%d >= %d)\n", row
, mci
->nr_csrows
);
1712 edac_mc_handle_ue_no_info(mci
, "INTERNAL ERROR");
1716 chars
= snprintf(pos
, len
+ 1, "%s",
1717 mci
->csrows
[row
].channels
[0].label
);
1721 for (chan
= 1; (chan
< mci
->csrows
[row
].nr_channels
) && (len
> 0);
1723 chars
= snprintf(pos
, len
+ 1, ":%s",
1724 mci
->csrows
[row
].channels
[chan
].label
);
1730 edac_mc_printk(mci
, KERN_EMERG
,
1731 "UE page 0x%lx, offset 0x%lx, grain %d, row %d, "
1732 "labels \"%s\": %s\n", page_frame_number
,
1733 offset_in_page
, mci
->csrows
[row
].grain
, row
, labels
,
1737 panic("EDAC MC%d: UE page 0x%lx, offset 0x%lx, grain %d, "
1738 "row %d, labels \"%s\": %s\n", mci
->mc_idx
,
1739 page_frame_number
, offset_in_page
,
1740 mci
->csrows
[row
].grain
, row
, labels
, msg
);
1743 mci
->csrows
[row
].ue_count
++;
1745 EXPORT_SYMBOL_GPL(edac_mc_handle_ue
);
1747 void edac_mc_handle_ue_no_info(struct mem_ctl_info
*mci
, const char *msg
)
1750 panic("EDAC MC%d: Uncorrected Error", mci
->mc_idx
);
1753 edac_mc_printk(mci
, KERN_WARNING
,
1754 "UE - no information available: %s\n", msg
);
1755 mci
->ue_noinfo_count
++;
1758 EXPORT_SYMBOL_GPL(edac_mc_handle_ue_no_info
);
1761 /*************************************************************
1762 * On Fully Buffered DIMM modules, this help function is
1763 * called to process UE events
1765 void edac_mc_handle_fbd_ue(struct mem_ctl_info
*mci
,
1767 unsigned int channela
,
1768 unsigned int channelb
,
1771 int len
= EDAC_MC_LABEL_LEN
* 4;
1772 char labels
[len
+ 1];
1776 if (csrow
>= mci
->nr_csrows
) {
1777 /* something is wrong */
1778 edac_mc_printk(mci
, KERN_ERR
,
1779 "INTERNAL ERROR: row out of range (%d >= %d)\n",
1780 csrow
, mci
->nr_csrows
);
1781 edac_mc_handle_ue_no_info(mci
, "INTERNAL ERROR");
1785 if (channela
>= mci
->csrows
[csrow
].nr_channels
) {
1786 /* something is wrong */
1787 edac_mc_printk(mci
, KERN_ERR
,
1788 "INTERNAL ERROR: channel-a out of range "
1790 channela
, mci
->csrows
[csrow
].nr_channels
);
1791 edac_mc_handle_ue_no_info(mci
, "INTERNAL ERROR");
1795 if (channelb
>= mci
->csrows
[csrow
].nr_channels
) {
1796 /* something is wrong */
1797 edac_mc_printk(mci
, KERN_ERR
,
1798 "INTERNAL ERROR: channel-b out of range "
1800 channelb
, mci
->csrows
[csrow
].nr_channels
);
1801 edac_mc_handle_ue_no_info(mci
, "INTERNAL ERROR");
1806 mci
->csrows
[csrow
].ue_count
++;
1808 /* Generate the DIMM labels from the specified channels */
1809 chars
= snprintf(pos
, len
+ 1, "%s",
1810 mci
->csrows
[csrow
].channels
[channela
].label
);
1811 len
-= chars
; pos
+= chars
;
1812 chars
= snprintf(pos
, len
+ 1, "-%s",
1813 mci
->csrows
[csrow
].channels
[channelb
].label
);
1816 edac_mc_printk(mci
, KERN_EMERG
,
1817 "UE row %d, channel-a= %d channel-b= %d "
1818 "labels \"%s\": %s\n", csrow
, channela
, channelb
,
1822 panic("UE row %d, channel-a= %d channel-b= %d "
1823 "labels \"%s\": %s\n", csrow
, channela
,
1824 channelb
, labels
, msg
);
1826 EXPORT_SYMBOL(edac_mc_handle_fbd_ue
);
1828 /*************************************************************
1829 * On Fully Buffered DIMM modules, this help function is
1830 * called to process CE events
1832 void edac_mc_handle_fbd_ce(struct mem_ctl_info
*mci
,
1834 unsigned int channel
,
1838 /* Ensure boundary values */
1839 if (csrow
>= mci
->nr_csrows
) {
1840 /* something is wrong */
1841 edac_mc_printk(mci
, KERN_ERR
,
1842 "INTERNAL ERROR: row out of range (%d >= %d)\n",
1843 csrow
, mci
->nr_csrows
);
1844 edac_mc_handle_ce_no_info(mci
, "INTERNAL ERROR");
1847 if (channel
>= mci
->csrows
[csrow
].nr_channels
) {
1848 /* something is wrong */
1849 edac_mc_printk(mci
, KERN_ERR
,
1850 "INTERNAL ERROR: channel out of range (%d >= %d)\n",
1851 channel
, mci
->csrows
[csrow
].nr_channels
);
1852 edac_mc_handle_ce_no_info(mci
, "INTERNAL ERROR");
1857 /* FIXME - put in DIMM location */
1858 edac_mc_printk(mci
, KERN_WARNING
,
1859 "CE row %d, channel %d, label \"%s\": %s\n",
1861 mci
->csrows
[csrow
].channels
[channel
].label
,
1865 mci
->csrows
[csrow
].ce_count
++;
1866 mci
->csrows
[csrow
].channels
[channel
].ce_count
++;
1868 EXPORT_SYMBOL(edac_mc_handle_fbd_ce
);
1872 * Iterate over all MC instances and check for ECC, et al, errors
1874 static inline void check_mc_devices(void)
1876 struct list_head
*item
;
1877 struct mem_ctl_info
*mci
;
1879 debugf3("%s()\n", __func__
);
1880 down(&mem_ctls_mutex
);
1882 list_for_each(item
, &mc_devices
) {
1883 mci
= list_entry(item
, struct mem_ctl_info
, link
);
1885 if (mci
->edac_check
!= NULL
)
1886 mci
->edac_check(mci
);
1889 up(&mem_ctls_mutex
);
1893 * Check MC status every poll_msec.
1894 * Check PCI status every poll_msec as well.
1896 * This where the work gets done for edac.
1898 * SMP safe, doesn't use NMI, and auto-rate-limits.
1900 static void do_edac_check(void)
1902 debugf3("%s()\n", __func__
);
1904 do_pci_parity_check();
1907 static int edac_kernel_thread(void *arg
)
1909 while (!kthread_should_stop()) {
1912 /* goto sleep for the interval */
1913 schedule_timeout_interruptible((HZ
* poll_msec
) / 1000);
1922 * module initialization entry point
1924 static int __init
edac_mc_init(void)
1926 edac_printk(KERN_INFO
, EDAC_MC
, EDAC_MC_VERSION
"\n");
1929 * Harvest and clear any boot/initialization PCI parity errors
1931 * FIXME: This only clears errors logged by devices present at time of
1932 * module initialization. We should also do an initial clear
1933 * of each newly hotplugged device.
1935 clear_pci_parity_errors();
1937 /* Create the MC sysfs entries */
1938 if (edac_sysfs_memctrl_setup()) {
1939 edac_printk(KERN_ERR
, EDAC_MC
,
1940 "Error initializing sysfs code\n");
1944 /* Create the PCI parity sysfs entries */
1945 if (edac_sysfs_pci_setup()) {
1946 edac_sysfs_memctrl_teardown();
1947 edac_printk(KERN_ERR
, EDAC_MC
,
1948 "EDAC PCI: Error initializing sysfs code\n");
1952 /* create our kernel thread */
1953 edac_thread
= kthread_run(edac_kernel_thread
, NULL
, "kedac");
1955 if (IS_ERR(edac_thread
)) {
1956 /* remove the sysfs entries */
1957 edac_sysfs_memctrl_teardown();
1958 edac_sysfs_pci_teardown();
1959 return PTR_ERR(edac_thread
);
1967 * module exit/termination functioni
1969 static void __exit
edac_mc_exit(void)
1971 debugf0("%s()\n", __func__
);
1972 kthread_stop(edac_thread
);
1974 /* tear down the sysfs device */
1975 edac_sysfs_memctrl_teardown();
1976 edac_sysfs_pci_teardown();
1979 module_init(edac_mc_init
);
1980 module_exit(edac_mc_exit
);
1982 MODULE_LICENSE("GPL");
1983 MODULE_AUTHOR("Linux Networx (http://lnxi.com) Thayne Harbaugh et al\n"
1984 "Based on work by Dan Hollis et al");
1985 MODULE_DESCRIPTION("Core library routines for MC reporting");
1987 module_param(panic_on_ue
, int, 0644);
1988 MODULE_PARM_DESC(panic_on_ue
, "Panic on uncorrected error: 0=off 1=on");
1990 module_param(check_pci_parity
, int, 0644);
1991 MODULE_PARM_DESC(check_pci_parity
, "Check for PCI bus parity errors: 0=off 1=on");
1992 module_param(panic_on_pci_parity
, int, 0644);
1993 MODULE_PARM_DESC(panic_on_pci_parity
, "Panic on PCI Bus Parity error: 0=off 1=on");
1995 module_param(log_ue
, int, 0644);
1996 MODULE_PARM_DESC(log_ue
, "Log uncorrectable error to console: 0=off 1=on");
1997 module_param(log_ce
, int, 0644);
1998 MODULE_PARM_DESC(log_ce
, "Log correctable error to console: 0=off 1=on");
1999 module_param(poll_msec
, int, 0644);
2000 MODULE_PARM_DESC(poll_msec
, "Polling period in milliseconds");
2001 #ifdef CONFIG_EDAC_DEBUG
2002 module_param(edac_debug_level
, int, 0644);
2003 MODULE_PARM_DESC(edac_debug_level
, "Debug level");