2 * edac_mc kernel module
3 * (C) 2005 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
16 #include <linux/config.h>
17 #include <linux/module.h>
18 #include <linux/proc_fs.h>
19 #include <linux/kernel.h>
20 #include <linux/types.h>
21 #include <linux/smp.h>
22 #include <linux/init.h>
23 #include <linux/sysctl.h>
24 #include <linux/highmem.h>
25 #include <linux/timer.h>
26 #include <linux/slab.h>
27 #include <linux/jiffies.h>
28 #include <linux/spinlock.h>
29 #include <linux/list.h>
30 #include <linux/sysdev.h>
31 #include <linux/ctype.h>
32 #include <linux/kthread.h>
34 #include <asm/uaccess.h>
40 #define EDAC_MC_VERSION "Ver: 2.0.0 " __DATE__
42 /* For now, disable the EDAC sysfs code. The sysfs interface that EDAC
43 * presents to user space needs more thought, and is likely to change
46 #define DISABLE_EDAC_SYSFS
48 #ifdef CONFIG_EDAC_DEBUG
49 /* Values of 0 to 4 will generate output */
50 int edac_debug_level = 1;
51 EXPORT_SYMBOL(edac_debug_level);
54 /* EDAC Controls, setable by module parameter, and sysfs */
55 static int log_ue = 1;
56 static int log_ce = 1;
57 static int panic_on_ue;
58 static int poll_msec = 1000;
60 static int check_pci_parity = 0; /* default YES check PCI parity */
61 static int panic_on_pci_parity; /* default no panic on PCI Parity */
62 static atomic_t pci_parity_count = ATOMIC_INIT(0);
64 /* lock to memory controller's control array */
65 static DECLARE_MUTEX(mem_ctls_mutex);
66 static struct list_head mc_devices = LIST_HEAD_INIT(mc_devices);
68 static struct task_struct *edac_thread;
70 /* Structure of the whitelist and blacklist arrays */
71 struct edac_pci_device_list {
72 unsigned int vendor; /* Vendor ID */
73 unsigned int device; /* Deviice ID */
77 #define MAX_LISTED_PCI_DEVICES 32
79 /* List of PCI devices (vendor-id:device-id) that should be skipped */
80 static struct edac_pci_device_list pci_blacklist[MAX_LISTED_PCI_DEVICES];
81 static int pci_blacklist_count;
83 /* List of PCI devices (vendor-id:device-id) that should be scanned */
84 static struct edac_pci_device_list pci_whitelist[MAX_LISTED_PCI_DEVICES];
85 static int pci_whitelist_count ;
87 /* START sysfs data and methods */
89 #ifndef DISABLE_EDAC_SYSFS
91 static const char *mem_types[] = {
92 [MEM_EMPTY] = "Empty",
93 [MEM_RESERVED] = "Reserved",
94 [MEM_UNKNOWN] = "Unknown",
98 [MEM_SDR] = "Unbuffered-SDR",
99 [MEM_RDR] = "Registered-SDR",
100 [MEM_DDR] = "Unbuffered-DDR",
101 [MEM_RDDR] = "Registered-DDR",
105 static const char *dev_types[] = {
106 [DEV_UNKNOWN] = "Unknown",
116 static const char *edac_caps[] = {
117 [EDAC_UNKNOWN] = "Unknown",
118 [EDAC_NONE] = "None",
119 [EDAC_RESERVED] = "Reserved",
120 [EDAC_PARITY] = "PARITY",
122 [EDAC_SECDED] = "SECDED",
123 [EDAC_S2ECD2ED] = "S2ECD2ED",
124 [EDAC_S4ECD4ED] = "S4ECD4ED",
125 [EDAC_S8ECD8ED] = "S8ECD8ED",
126 [EDAC_S16ECD16ED] = "S16ECD16ED"
130 /* sysfs object: /sys/devices/system/edac */
131 static struct sysdev_class edac_class = {
132 set_kset_name("edac"),
136 * /sys/devices/system/edac/mc
137 * /sys/devices/system/edac/pci
139 static struct kobject edac_memctrl_kobj;
140 static struct kobject edac_pci_kobj;
143 * /sys/devices/system/edac/mc;
144 * data structures and methods
147 static ssize_t memctrl_string_show(void *ptr, char *buffer)
149 char *value = (char*) ptr;
150 return sprintf(buffer, "%s\n", value);
154 static ssize_t memctrl_int_show(void *ptr, char *buffer)
156 int *value = (int*) ptr;
157 return sprintf(buffer, "%d\n", *value);
160 static ssize_t memctrl_int_store(void *ptr, const char *buffer, size_t count)
162 int *value = (int*) ptr;
164 if (isdigit(*buffer))
165 *value = simple_strtoul(buffer, NULL, 0);
170 struct memctrl_dev_attribute {
171 struct attribute attr;
173 ssize_t (*show)(void *,char *);
174 ssize_t (*store)(void *, const char *, size_t);
177 /* Set of show/store abstract level functions for memory control object */
179 memctrl_dev_show(struct kobject *kobj, struct attribute *attr, char *buffer)
181 struct memctrl_dev_attribute *memctrl_dev;
182 memctrl_dev = (struct memctrl_dev_attribute*)attr;
184 if (memctrl_dev->show)
185 return memctrl_dev->show(memctrl_dev->value, buffer);
190 memctrl_dev_store(struct kobject *kobj, struct attribute *attr,
191 const char *buffer, size_t count)
193 struct memctrl_dev_attribute *memctrl_dev;
194 memctrl_dev = (struct memctrl_dev_attribute*)attr;
196 if (memctrl_dev->store)
197 return memctrl_dev->store(memctrl_dev->value, buffer, count);
201 static struct sysfs_ops memctrlfs_ops = {
202 .show = memctrl_dev_show,
203 .store = memctrl_dev_store
206 #define MEMCTRL_ATTR(_name,_mode,_show,_store) \
207 struct memctrl_dev_attribute attr_##_name = { \
208 .attr = {.name = __stringify(_name), .mode = _mode }, \
214 #define MEMCTRL_STRING_ATTR(_name,_data,_mode,_show,_store) \
215 struct memctrl_dev_attribute attr_##_name = { \
216 .attr = {.name = __stringify(_name), .mode = _mode }, \
222 /* cwrow<id> attribute f*/
224 MEMCTRL_STRING_ATTR(mc_version,EDAC_MC_VERSION,S_IRUGO,memctrl_string_show,NULL);
227 /* csrow<id> control files */
228 MEMCTRL_ATTR(panic_on_ue,S_IRUGO|S_IWUSR,memctrl_int_show,memctrl_int_store);
229 MEMCTRL_ATTR(log_ue,S_IRUGO|S_IWUSR,memctrl_int_show,memctrl_int_store);
230 MEMCTRL_ATTR(log_ce,S_IRUGO|S_IWUSR,memctrl_int_show,memctrl_int_store);
231 MEMCTRL_ATTR(poll_msec,S_IRUGO|S_IWUSR,memctrl_int_show,memctrl_int_store);
234 /* Base Attributes of the memory ECC object */
235 static struct memctrl_dev_attribute *memctrl_attr[] = {
243 /* Main MC kobject release() function */
244 static void edac_memctrl_master_release(struct kobject *kobj)
246 debugf1("%s()\n", __func__);
249 static struct kobj_type ktype_memctrl = {
250 .release = edac_memctrl_master_release,
251 .sysfs_ops = &memctrlfs_ops,
252 .default_attrs = (struct attribute **) memctrl_attr,
255 #endif /* DISABLE_EDAC_SYSFS */
257 /* Initialize the main sysfs entries for edac:
258 * /sys/devices/system/edac
265 static int edac_sysfs_memctrl_setup(void)
266 #ifdef DISABLE_EDAC_SYSFS
274 debugf1("%s()\n", __func__);
276 /* create the /sys/devices/system/edac directory */
277 err = sysdev_class_register(&edac_class);
279 /* Init the MC's kobject */
280 memset(&edac_memctrl_kobj, 0, sizeof (edac_memctrl_kobj));
281 kobject_init(&edac_memctrl_kobj);
283 edac_memctrl_kobj.parent = &edac_class.kset.kobj;
284 edac_memctrl_kobj.ktype = &ktype_memctrl;
286 /* generate sysfs "..../edac/mc" */
287 err = kobject_set_name(&edac_memctrl_kobj,"mc");
289 /* FIXME: maybe new sysdev_create_subdir() */
290 err = kobject_register(&edac_memctrl_kobj);
292 debugf1("Failed to register '.../edac/mc'\n");
294 debugf1("Registered '.../edac/mc' kobject\n");
298 debugf1("%s() error=%d\n", __func__, err);
303 #endif /* DISABLE_EDAC_SYSFS */
307 * the '..../edac/mc' kobject followed by '..../edac' itself
309 static void edac_sysfs_memctrl_teardown(void)
311 #ifndef DISABLE_EDAC_SYSFS
312 debugf0("MC: " __FILE__ ": %s()\n", __func__);
314 /* Unregister the MC's kobject */
315 kobject_unregister(&edac_memctrl_kobj);
317 /* release the master edac mc kobject */
318 kobject_put(&edac_memctrl_kobj);
320 /* Unregister the 'edac' object */
321 sysdev_class_unregister(&edac_class);
322 #endif /* DISABLE_EDAC_SYSFS */
325 #ifndef DISABLE_EDAC_SYSFS
328 * /sys/devices/system/edac/pci;
329 * data structures and methods
332 struct list_control {
333 struct edac_pci_device_list *list;
339 /* Output the list as: vendor_id:device:id<,vendor_id:device_id> */
340 static ssize_t edac_pci_list_string_show(void *ptr, char *buffer)
342 struct list_control *listctl;
343 struct edac_pci_device_list *list;
349 list = listctl->list;
351 for (i = 0; i < *(listctl->count); i++, list++ ) {
353 len += snprintf(p + len, (PAGE_SIZE-len), ",");
355 len += snprintf(p + len,
358 list->vendor,list->device);
361 len += snprintf(p + len,(PAGE_SIZE-len), "\n");
363 return (ssize_t) len;
368 * Scan string from **s to **e looking for one 'vendor:device' tuple
369 * where each field is a hex value
371 * return 0 if an entry is NOT found
372 * return 1 if an entry is found
373 * fill in *vendor_id and *device_id with values found
375 * In both cases, make sure *s has been moved forward toward *e
377 static int parse_one_device(const char **s,const char **e,
378 unsigned int *vendor_id, unsigned int *device_id)
380 const char *runner, *p;
382 /* if null byte, we are done */
384 (*s)++; /* keep *s moving */
388 /* skip over newlines & whitespace */
389 if ((**s == '\n') || isspace(**s)) {
394 if (!isxdigit(**s)) {
399 /* parse vendor_id */
401 while (runner < *e) {
402 /* scan for vendor:device delimiter */
403 if (*runner == ':') {
404 *vendor_id = simple_strtol((char*) *s, (char**) &p, 16);
411 if (!isxdigit(*runner)) {
416 /* parse device_id */
418 *device_id = simple_strtol((char*)runner, (char**)&p, 16);
427 static ssize_t edac_pci_list_string_store(void *ptr, const char *buffer,
430 struct list_control *listctl;
431 struct edac_pci_device_list *list;
432 unsigned int vendor_id, device_id;
440 list = listctl->list;
441 index = listctl->count;
444 while (*index < MAX_LISTED_PCI_DEVICES) {
446 if (parse_one_device(&s,&e,&vendor_id,&device_id)) {
447 list[ *index ].vendor = vendor_id;
448 list[ *index ].device = device_id;
452 /* check for all data consume */
461 static ssize_t edac_pci_int_show(void *ptr, char *buffer)
464 return sprintf(buffer,"%d\n",*value);
467 static ssize_t edac_pci_int_store(void *ptr, const char *buffer, size_t count)
471 if (isdigit(*buffer))
472 *value = simple_strtoul(buffer,NULL,0);
477 struct edac_pci_dev_attribute {
478 struct attribute attr;
480 ssize_t (*show)(void *,char *);
481 ssize_t (*store)(void *, const char *,size_t);
484 /* Set of show/store abstract level functions for PCI Parity object */
485 static ssize_t edac_pci_dev_show(struct kobject *kobj, struct attribute *attr,
488 struct edac_pci_dev_attribute *edac_pci_dev;
489 edac_pci_dev= (struct edac_pci_dev_attribute*)attr;
491 if (edac_pci_dev->show)
492 return edac_pci_dev->show(edac_pci_dev->value, buffer);
496 static ssize_t edac_pci_dev_store(struct kobject *kobj, struct attribute *attr,
497 const char *buffer, size_t count)
499 struct edac_pci_dev_attribute *edac_pci_dev;
500 edac_pci_dev= (struct edac_pci_dev_attribute*)attr;
502 if (edac_pci_dev->show)
503 return edac_pci_dev->store(edac_pci_dev->value, buffer, count);
507 static struct sysfs_ops edac_pci_sysfs_ops = {
508 .show = edac_pci_dev_show,
509 .store = edac_pci_dev_store
513 #define EDAC_PCI_ATTR(_name,_mode,_show,_store) \
514 struct edac_pci_dev_attribute edac_pci_attr_##_name = { \
515 .attr = {.name = __stringify(_name), .mode = _mode }, \
521 #define EDAC_PCI_STRING_ATTR(_name,_data,_mode,_show,_store) \
522 struct edac_pci_dev_attribute edac_pci_attr_##_name = { \
523 .attr = {.name = __stringify(_name), .mode = _mode }, \
530 static struct list_control pci_whitelist_control = {
531 .list = pci_whitelist,
532 .count = &pci_whitelist_count
535 static struct list_control pci_blacklist_control = {
536 .list = pci_blacklist,
537 .count = &pci_blacklist_count
540 /* whitelist attribute */
541 EDAC_PCI_STRING_ATTR(pci_parity_whitelist,
542 &pci_whitelist_control,
544 edac_pci_list_string_show,
545 edac_pci_list_string_store);
547 EDAC_PCI_STRING_ATTR(pci_parity_blacklist,
548 &pci_blacklist_control,
550 edac_pci_list_string_show,
551 edac_pci_list_string_store);
554 /* PCI Parity control files */
555 EDAC_PCI_ATTR(check_pci_parity,S_IRUGO|S_IWUSR,edac_pci_int_show,edac_pci_int_store);
556 EDAC_PCI_ATTR(panic_on_pci_parity,S_IRUGO|S_IWUSR,edac_pci_int_show,edac_pci_int_store);
557 EDAC_PCI_ATTR(pci_parity_count,S_IRUGO,edac_pci_int_show,NULL);
559 /* Base Attributes of the memory ECC object */
560 static struct edac_pci_dev_attribute *edac_pci_attr[] = {
561 &edac_pci_attr_check_pci_parity,
562 &edac_pci_attr_panic_on_pci_parity,
563 &edac_pci_attr_pci_parity_count,
567 /* No memory to release */
568 static void edac_pci_release(struct kobject *kobj)
570 debugf1("%s()\n", __func__);
573 static struct kobj_type ktype_edac_pci = {
574 .release = edac_pci_release,
575 .sysfs_ops = &edac_pci_sysfs_ops,
576 .default_attrs = (struct attribute **) edac_pci_attr,
579 #endif /* DISABLE_EDAC_SYSFS */
582 * edac_sysfs_pci_setup()
585 static int edac_sysfs_pci_setup(void)
586 #ifdef DISABLE_EDAC_SYSFS
594 debugf1("%s()\n", __func__);
596 memset(&edac_pci_kobj, 0, sizeof(edac_pci_kobj));
598 kobject_init(&edac_pci_kobj);
599 edac_pci_kobj.parent = &edac_class.kset.kobj;
600 edac_pci_kobj.ktype = &ktype_edac_pci;
602 err = kobject_set_name(&edac_pci_kobj, "pci");
604 /* Instanstiate the csrow object */
605 /* FIXME: maybe new sysdev_create_subdir() */
606 err = kobject_register(&edac_pci_kobj);
608 debugf1("Failed to register '.../edac/pci'\n");
610 debugf1("Registered '.../edac/pci' kobject\n");
614 #endif /* DISABLE_EDAC_SYSFS */
616 static void edac_sysfs_pci_teardown(void)
618 #ifndef DISABLE_EDAC_SYSFS
619 debugf0("%s()\n", __func__);
621 kobject_unregister(&edac_pci_kobj);
622 kobject_put(&edac_pci_kobj);
626 #ifndef DISABLE_EDAC_SYSFS
628 /* EDAC sysfs CSROW data structures and methods */
630 /* Set of more detailed csrow<id> attribute show/store functions */
631 static ssize_t csrow_ch0_dimm_label_show(struct csrow_info *csrow, char *data)
635 if (csrow->nr_channels > 0) {
636 size = snprintf(data, EDAC_MC_LABEL_LEN,"%s\n",
637 csrow->channels[0].label);
642 static ssize_t csrow_ch1_dimm_label_show(struct csrow_info *csrow, char *data)
646 if (csrow->nr_channels > 0) {
647 size = snprintf(data, EDAC_MC_LABEL_LEN, "%s\n",
648 csrow->channels[1].label);
653 static ssize_t csrow_ch0_dimm_label_store(struct csrow_info *csrow,
654 const char *data, size_t size)
656 ssize_t max_size = 0;
658 if (csrow->nr_channels > 0) {
659 max_size = min((ssize_t)size,(ssize_t)EDAC_MC_LABEL_LEN-1);
660 strncpy(csrow->channels[0].label, data, max_size);
661 csrow->channels[0].label[max_size] = '\0';
666 static ssize_t csrow_ch1_dimm_label_store(struct csrow_info *csrow,
667 const char *data, size_t size)
669 ssize_t max_size = 0;
671 if (csrow->nr_channels > 1) {
672 max_size = min((ssize_t)size,(ssize_t)EDAC_MC_LABEL_LEN-1);
673 strncpy(csrow->channels[1].label, data, max_size);
674 csrow->channels[1].label[max_size] = '\0';
679 static ssize_t csrow_ue_count_show(struct csrow_info *csrow, char *data)
681 return sprintf(data,"%u\n", csrow->ue_count);
684 static ssize_t csrow_ce_count_show(struct csrow_info *csrow, char *data)
686 return sprintf(data,"%u\n", csrow->ce_count);
689 static ssize_t csrow_ch0_ce_count_show(struct csrow_info *csrow, char *data)
693 if (csrow->nr_channels > 0) {
694 size = sprintf(data,"%u\n", csrow->channels[0].ce_count);
699 static ssize_t csrow_ch1_ce_count_show(struct csrow_info *csrow, char *data)
703 if (csrow->nr_channels > 1) {
704 size = sprintf(data,"%u\n", csrow->channels[1].ce_count);
709 static ssize_t csrow_size_show(struct csrow_info *csrow, char *data)
711 return sprintf(data,"%u\n", PAGES_TO_MiB(csrow->nr_pages));
714 static ssize_t csrow_mem_type_show(struct csrow_info *csrow, char *data)
716 return sprintf(data,"%s\n", mem_types[csrow->mtype]);
719 static ssize_t csrow_dev_type_show(struct csrow_info *csrow, char *data)
721 return sprintf(data,"%s\n", dev_types[csrow->dtype]);
724 static ssize_t csrow_edac_mode_show(struct csrow_info *csrow, char *data)
726 return sprintf(data,"%s\n", edac_caps[csrow->edac_mode]);
729 struct csrowdev_attribute {
730 struct attribute attr;
731 ssize_t (*show)(struct csrow_info *,char *);
732 ssize_t (*store)(struct csrow_info *, const char *,size_t);
735 #define to_csrow(k) container_of(k, struct csrow_info, kobj)
736 #define to_csrowdev_attr(a) container_of(a, struct csrowdev_attribute, attr)
738 /* Set of show/store higher level functions for csrow objects */
739 static ssize_t csrowdev_show(struct kobject *kobj, struct attribute *attr,
742 struct csrow_info *csrow = to_csrow(kobj);
743 struct csrowdev_attribute *csrowdev_attr = to_csrowdev_attr(attr);
745 if (csrowdev_attr->show)
746 return csrowdev_attr->show(csrow, buffer);
750 static ssize_t csrowdev_store(struct kobject *kobj, struct attribute *attr,
751 const char *buffer, size_t count)
753 struct csrow_info *csrow = to_csrow(kobj);
754 struct csrowdev_attribute * csrowdev_attr = to_csrowdev_attr(attr);
756 if (csrowdev_attr->store)
757 return csrowdev_attr->store(csrow, buffer, count);
761 static struct sysfs_ops csrowfs_ops = {
762 .show = csrowdev_show,
763 .store = csrowdev_store
766 #define CSROWDEV_ATTR(_name,_mode,_show,_store) \
767 struct csrowdev_attribute attr_##_name = { \
768 .attr = {.name = __stringify(_name), .mode = _mode }, \
773 /* cwrow<id>/attribute files */
774 CSROWDEV_ATTR(size_mb,S_IRUGO,csrow_size_show,NULL);
775 CSROWDEV_ATTR(dev_type,S_IRUGO,csrow_dev_type_show,NULL);
776 CSROWDEV_ATTR(mem_type,S_IRUGO,csrow_mem_type_show,NULL);
777 CSROWDEV_ATTR(edac_mode,S_IRUGO,csrow_edac_mode_show,NULL);
778 CSROWDEV_ATTR(ue_count,S_IRUGO,csrow_ue_count_show,NULL);
779 CSROWDEV_ATTR(ce_count,S_IRUGO,csrow_ce_count_show,NULL);
780 CSROWDEV_ATTR(ch0_ce_count,S_IRUGO,csrow_ch0_ce_count_show,NULL);
781 CSROWDEV_ATTR(ch1_ce_count,S_IRUGO,csrow_ch1_ce_count_show,NULL);
783 /* control/attribute files */
784 CSROWDEV_ATTR(ch0_dimm_label,S_IRUGO|S_IWUSR,
785 csrow_ch0_dimm_label_show,
786 csrow_ch0_dimm_label_store);
787 CSROWDEV_ATTR(ch1_dimm_label,S_IRUGO|S_IWUSR,
788 csrow_ch1_dimm_label_show,
789 csrow_ch1_dimm_label_store);
792 /* Attributes of the CSROW<id> object */
793 static struct csrowdev_attribute *csrow_attr[] = {
802 &attr_ch0_dimm_label,
803 &attr_ch1_dimm_label,
808 /* No memory to release */
809 static void edac_csrow_instance_release(struct kobject *kobj)
811 debugf1("%s()\n", __func__);
814 static struct kobj_type ktype_csrow = {
815 .release = edac_csrow_instance_release,
816 .sysfs_ops = &csrowfs_ops,
817 .default_attrs = (struct attribute **) csrow_attr,
820 /* Create a CSROW object under specifed edac_mc_device */
821 static int edac_create_csrow_object(struct kobject *edac_mci_kobj,
822 struct csrow_info *csrow, int index )
826 debugf0("%s()\n", __func__);
828 memset(&csrow->kobj, 0, sizeof(csrow->kobj));
830 /* generate ..../edac/mc/mc<id>/csrow<index> */
832 kobject_init(&csrow->kobj);
833 csrow->kobj.parent = edac_mci_kobj;
834 csrow->kobj.ktype = &ktype_csrow;
836 /* name this instance of csrow<id> */
837 err = kobject_set_name(&csrow->kobj,"csrow%d",index);
839 /* Instanstiate the csrow object */
840 err = kobject_register(&csrow->kobj);
842 debugf0("Failed to register CSROW%d\n",index);
844 debugf0("Registered CSROW%d\n",index);
850 /* sysfs data structures and methods for the MCI kobjects */
852 static ssize_t mci_reset_counters_store(struct mem_ctl_info *mci,
853 const char *data, size_t count )
857 mci->ue_noinfo_count = 0;
858 mci->ce_noinfo_count = 0;
861 for (row = 0; row < mci->nr_csrows; row++) {
862 struct csrow_info *ri = &mci->csrows[row];
866 for (chan = 0; chan < ri->nr_channels; chan++)
867 ri->channels[chan].ce_count = 0;
869 mci->start_time = jiffies;
874 static ssize_t mci_ue_count_show(struct mem_ctl_info *mci, char *data)
876 return sprintf(data,"%d\n", mci->ue_count);
879 static ssize_t mci_ce_count_show(struct mem_ctl_info *mci, char *data)
881 return sprintf(data,"%d\n", mci->ce_count);
884 static ssize_t mci_ce_noinfo_show(struct mem_ctl_info *mci, char *data)
886 return sprintf(data,"%d\n", mci->ce_noinfo_count);
889 static ssize_t mci_ue_noinfo_show(struct mem_ctl_info *mci, char *data)
891 return sprintf(data,"%d\n", mci->ue_noinfo_count);
894 static ssize_t mci_seconds_show(struct mem_ctl_info *mci, char *data)
896 return sprintf(data,"%ld\n", (jiffies - mci->start_time) / HZ);
899 static ssize_t mci_mod_name_show(struct mem_ctl_info *mci, char *data)
901 return sprintf(data,"%s %s\n", mci->mod_name, mci->mod_ver);
904 static ssize_t mci_ctl_name_show(struct mem_ctl_info *mci, char *data)
906 return sprintf(data,"%s\n", mci->ctl_name);
909 static int mci_output_edac_cap(char *buf, unsigned long edac_cap)
914 for (bit_idx = 0; bit_idx < 8 * sizeof(edac_cap); bit_idx++) {
915 if ((edac_cap >> bit_idx) & 0x1)
916 p += sprintf(p, "%s ", edac_caps[bit_idx]);
922 static ssize_t mci_edac_capability_show(struct mem_ctl_info *mci, char *data)
926 p += mci_output_edac_cap(p,mci->edac_ctl_cap);
927 p += sprintf(p, "\n");
932 static ssize_t mci_edac_current_capability_show(struct mem_ctl_info *mci,
937 p += mci_output_edac_cap(p,mci->edac_cap);
938 p += sprintf(p, "\n");
943 static int mci_output_mtype_cap(char *buf, unsigned long mtype_cap)
948 for (bit_idx = 0; bit_idx < 8 * sizeof(mtype_cap); bit_idx++) {
949 if ((mtype_cap >> bit_idx) & 0x1)
950 p += sprintf(p, "%s ", mem_types[bit_idx]);
956 static ssize_t mci_supported_mem_type_show(struct mem_ctl_info *mci, char *data)
960 p += mci_output_mtype_cap(p,mci->mtype_cap);
961 p += sprintf(p, "\n");
966 static ssize_t mci_size_mb_show(struct mem_ctl_info *mci, char *data)
968 int total_pages, csrow_idx;
970 for (total_pages = csrow_idx = 0; csrow_idx < mci->nr_csrows;
972 struct csrow_info *csrow = &mci->csrows[csrow_idx];
974 if (!csrow->nr_pages)
976 total_pages += csrow->nr_pages;
979 return sprintf(data,"%u\n", PAGES_TO_MiB(total_pages));
982 struct mcidev_attribute {
983 struct attribute attr;
984 ssize_t (*show)(struct mem_ctl_info *,char *);
985 ssize_t (*store)(struct mem_ctl_info *, const char *,size_t);
988 #define to_mci(k) container_of(k, struct mem_ctl_info, edac_mci_kobj)
989 #define to_mcidev_attr(a) container_of(a, struct mcidev_attribute, attr)
991 static ssize_t mcidev_show(struct kobject *kobj, struct attribute *attr,
994 struct mem_ctl_info *mem_ctl_info = to_mci(kobj);
995 struct mcidev_attribute * mcidev_attr = to_mcidev_attr(attr);
997 if (mcidev_attr->show)
998 return mcidev_attr->show(mem_ctl_info, buffer);
1002 static ssize_t mcidev_store(struct kobject *kobj, struct attribute *attr,
1003 const char *buffer, size_t count)
1005 struct mem_ctl_info *mem_ctl_info = to_mci(kobj);
1006 struct mcidev_attribute * mcidev_attr = to_mcidev_attr(attr);
1008 if (mcidev_attr->store)
1009 return mcidev_attr->store(mem_ctl_info, buffer, count);
1013 static struct sysfs_ops mci_ops = {
1014 .show = mcidev_show,
1015 .store = mcidev_store
1018 #define MCIDEV_ATTR(_name,_mode,_show,_store) \
1019 struct mcidev_attribute mci_attr_##_name = { \
1020 .attr = {.name = __stringify(_name), .mode = _mode }, \
1026 MCIDEV_ATTR(reset_counters,S_IWUSR,NULL,mci_reset_counters_store);
1028 /* Attribute files */
1029 MCIDEV_ATTR(mc_name,S_IRUGO,mci_ctl_name_show,NULL);
1030 MCIDEV_ATTR(module_name,S_IRUGO,mci_mod_name_show,NULL);
1031 MCIDEV_ATTR(edac_capability,S_IRUGO,mci_edac_capability_show,NULL);
1032 MCIDEV_ATTR(size_mb,S_IRUGO,mci_size_mb_show,NULL);
1033 MCIDEV_ATTR(seconds_since_reset,S_IRUGO,mci_seconds_show,NULL);
1034 MCIDEV_ATTR(ue_noinfo_count,S_IRUGO,mci_ue_noinfo_show,NULL);
1035 MCIDEV_ATTR(ce_noinfo_count,S_IRUGO,mci_ce_noinfo_show,NULL);
1036 MCIDEV_ATTR(ue_count,S_IRUGO,mci_ue_count_show,NULL);
1037 MCIDEV_ATTR(ce_count,S_IRUGO,mci_ce_count_show,NULL);
1038 MCIDEV_ATTR(edac_current_capability,S_IRUGO,
1039 mci_edac_current_capability_show,NULL);
1040 MCIDEV_ATTR(supported_mem_type,S_IRUGO,
1041 mci_supported_mem_type_show,NULL);
1044 static struct mcidev_attribute *mci_attr[] = {
1045 &mci_attr_reset_counters,
1046 &mci_attr_module_name,
1048 &mci_attr_edac_capability,
1049 &mci_attr_edac_current_capability,
1050 &mci_attr_supported_mem_type,
1052 &mci_attr_seconds_since_reset,
1053 &mci_attr_ue_noinfo_count,
1054 &mci_attr_ce_noinfo_count,
1062 * Release of a MC controlling instance
1064 static void edac_mci_instance_release(struct kobject *kobj)
1066 struct mem_ctl_info *mci;
1067 mci = container_of(kobj,struct mem_ctl_info,edac_mci_kobj);
1069 debugf0("%s() idx=%d calling kfree\n", __func__, mci->mc_idx);
1074 static struct kobj_type ktype_mci = {
1075 .release = edac_mci_instance_release,
1076 .sysfs_ops = &mci_ops,
1077 .default_attrs = (struct attribute **) mci_attr,
1080 #endif /* DISABLE_EDAC_SYSFS */
1082 #define EDAC_DEVICE_SYMLINK "device"
1085 * Create a new Memory Controller kobject instance,
1086 * mc<id> under the 'mc' directory
1092 static int edac_create_sysfs_mci_device(struct mem_ctl_info *mci)
1093 #ifdef DISABLE_EDAC_SYSFS
1101 struct csrow_info *csrow;
1102 struct kobject *edac_mci_kobj=&mci->edac_mci_kobj;
1104 debugf0("%s() idx=%d\n", __func__, mci->mc_idx);
1106 memset(edac_mci_kobj, 0, sizeof(*edac_mci_kobj));
1107 kobject_init(edac_mci_kobj);
1109 /* set the name of the mc<id> object */
1110 err = kobject_set_name(edac_mci_kobj,"mc%d",mci->mc_idx);
1114 /* link to our parent the '..../edac/mc' object */
1115 edac_mci_kobj->parent = &edac_memctrl_kobj;
1116 edac_mci_kobj->ktype = &ktype_mci;
1118 /* register the mc<id> kobject */
1119 err = kobject_register(edac_mci_kobj);
1123 /* create a symlink for the device */
1124 err = sysfs_create_link(edac_mci_kobj, &mci->pdev->dev.kobj,
1125 EDAC_DEVICE_SYMLINK);
1127 kobject_unregister(edac_mci_kobj);
1131 /* Make directories for each CSROW object
1132 * under the mc<id> kobject
1134 for (i = 0; i < mci->nr_csrows; i++) {
1136 csrow = &mci->csrows[i];
1138 /* Only expose populated CSROWs */
1139 if (csrow->nr_pages > 0) {
1140 err = edac_create_csrow_object(edac_mci_kobj,csrow,i);
1146 /* Mark this MCI instance as having sysfs entries */
1147 mci->sysfs_active = MCI_SYSFS_ACTIVE;
1152 /* CSROW error: backout what has already been registered, */
1154 for ( i--; i >= 0; i--) {
1155 if (csrow->nr_pages > 0) {
1156 kobject_unregister(&mci->csrows[i].kobj);
1157 kobject_put(&mci->csrows[i].kobj);
1161 kobject_unregister(edac_mci_kobj);
1162 kobject_put(edac_mci_kobj);
1166 #endif /* DISABLE_EDAC_SYSFS */
1169 * remove a Memory Controller instance
1171 static void edac_remove_sysfs_mci_device(struct mem_ctl_info *mci)
1173 #ifndef DISABLE_EDAC_SYSFS
1176 debugf0("%s()\n", __func__);
1178 /* remove all csrow kobjects */
1179 for (i = 0; i < mci->nr_csrows; i++) {
1180 if (mci->csrows[i].nr_pages > 0) {
1181 kobject_unregister(&mci->csrows[i].kobj);
1182 kobject_put(&mci->csrows[i].kobj);
1186 sysfs_remove_link(&mci->edac_mci_kobj, EDAC_DEVICE_SYMLINK);
1188 kobject_unregister(&mci->edac_mci_kobj);
1189 kobject_put(&mci->edac_mci_kobj);
1190 #endif /* DISABLE_EDAC_SYSFS */
1193 /* END OF sysfs data and methods */
1195 #ifdef CONFIG_EDAC_DEBUG
1197 EXPORT_SYMBOL(edac_mc_dump_channel);
1199 void edac_mc_dump_channel(struct channel_info *chan)
1201 debugf4("\tchannel = %p\n", chan);
1202 debugf4("\tchannel->chan_idx = %d\n", chan->chan_idx);
1203 debugf4("\tchannel->ce_count = %d\n", chan->ce_count);
1204 debugf4("\tchannel->label = '%s'\n", chan->label);
1205 debugf4("\tchannel->csrow = %p\n\n", chan->csrow);
1209 EXPORT_SYMBOL(edac_mc_dump_csrow);
1211 void edac_mc_dump_csrow(struct csrow_info *csrow)
1213 debugf4("\tcsrow = %p\n", csrow);
1214 debugf4("\tcsrow->csrow_idx = %d\n", csrow->csrow_idx);
1215 debugf4("\tcsrow->first_page = 0x%lx\n",
1217 debugf4("\tcsrow->last_page = 0x%lx\n", csrow->last_page);
1218 debugf4("\tcsrow->page_mask = 0x%lx\n", csrow->page_mask);
1219 debugf4("\tcsrow->nr_pages = 0x%x\n", csrow->nr_pages);
1220 debugf4("\tcsrow->nr_channels = %d\n",
1221 csrow->nr_channels);
1222 debugf4("\tcsrow->channels = %p\n", csrow->channels);
1223 debugf4("\tcsrow->mci = %p\n\n", csrow->mci);
1227 EXPORT_SYMBOL(edac_mc_dump_mci);
1229 void edac_mc_dump_mci(struct mem_ctl_info *mci)
1231 debugf3("\tmci = %p\n", mci);
1232 debugf3("\tmci->mtype_cap = %lx\n", mci->mtype_cap);
1233 debugf3("\tmci->edac_ctl_cap = %lx\n", mci->edac_ctl_cap);
1234 debugf3("\tmci->edac_cap = %lx\n", mci->edac_cap);
1235 debugf4("\tmci->edac_check = %p\n", mci->edac_check);
1236 debugf3("\tmci->nr_csrows = %d, csrows = %p\n",
1237 mci->nr_csrows, mci->csrows);
1238 debugf3("\tpdev = %p\n", mci->pdev);
1239 debugf3("\tmod_name:ctl_name = %s:%s\n",
1240 mci->mod_name, mci->ctl_name);
1241 debugf3("\tpvt_info = %p\n\n", mci->pvt_info);
1245 #endif /* CONFIG_EDAC_DEBUG */
1247 /* 'ptr' points to a possibly unaligned item X such that sizeof(X) is 'size'.
1248 * Adjust 'ptr' so that its alignment is at least as stringent as what the
1249 * compiler would provide for X and return the aligned result.
1251 * If 'size' is a constant, the compiler will optimize this whole function
1252 * down to either a no-op or the addition of a constant to the value of 'ptr'.
1254 static inline char * align_ptr (void *ptr, unsigned size)
1258 /* Here we assume that the alignment of a "long long" is the most
1259 * stringent alignment that the compiler will ever provide by default.
1260 * As far as I know, this is a reasonable assumption.
1262 if (size > sizeof(long))
1263 align = sizeof(long long);
1264 else if (size > sizeof(int))
1265 align = sizeof(long);
1266 else if (size > sizeof(short))
1267 align = sizeof(int);
1268 else if (size > sizeof(char))
1269 align = sizeof(short);
1271 return (char *) ptr;
1276 return (char *) ptr;
1278 return (char *) (((unsigned long) ptr) + align - r);
1282 EXPORT_SYMBOL(edac_mc_alloc);
1285 * edac_mc_alloc: Allocate a struct mem_ctl_info structure
1286 * @size_pvt: size of private storage needed
1287 * @nr_csrows: Number of CWROWS needed for this MC
1288 * @nr_chans: Number of channels for the MC
1290 * Everything is kmalloc'ed as one big chunk - more efficient.
1291 * Only can be used if all structures have the same lifetime - otherwise
1292 * you have to allocate and initialize your own structures.
1294 * Use edac_mc_free() to free mc structures allocated by this function.
1297 * NULL allocation failed
1298 * struct mem_ctl_info pointer
1300 struct mem_ctl_info *edac_mc_alloc(unsigned sz_pvt, unsigned nr_csrows,
1303 struct mem_ctl_info *mci;
1304 struct csrow_info *csi, *csrow;
1305 struct channel_info *chi, *chp, *chan;
1310 /* Figure out the offsets of the various items from the start of an mc
1311 * structure. We want the alignment of each item to be at least as
1312 * stringent as what the compiler would provide if we could simply
1313 * hardcode everything into a single struct.
1315 mci = (struct mem_ctl_info *) 0;
1316 csi = (struct csrow_info *)align_ptr(&mci[1], sizeof(*csi));
1317 chi = (struct channel_info *)
1318 align_ptr(&csi[nr_csrows], sizeof(*chi));
1319 pvt = align_ptr(&chi[nr_chans * nr_csrows], sz_pvt);
1320 size = ((unsigned long) pvt) + sz_pvt;
1322 if ((mci = kmalloc(size, GFP_KERNEL)) == NULL)
1325 /* Adjust pointers so they point within the memory we just allocated
1326 * rather than an imaginary chunk of memory located at address 0.
1328 csi = (struct csrow_info *) (((char *) mci) + ((unsigned long) csi));
1329 chi = (struct channel_info *) (((char *) mci) + ((unsigned long) chi));
1330 pvt = sz_pvt ? (((char *) mci) + ((unsigned long) pvt)) : NULL;
1332 memset(mci, 0, size); /* clear all fields */
1335 mci->pvt_info = pvt;
1336 mci->nr_csrows = nr_csrows;
1338 for (row = 0; row < nr_csrows; row++) {
1340 csrow->csrow_idx = row;
1342 csrow->nr_channels = nr_chans;
1343 chp = &chi[row * nr_chans];
1344 csrow->channels = chp;
1346 for (chn = 0; chn < nr_chans; chn++) {
1348 chan->chan_idx = chn;
1349 chan->csrow = csrow;
1357 EXPORT_SYMBOL(edac_mc_free);
1360 * edac_mc_free: Free a previously allocated 'mci' structure
1361 * @mci: pointer to a struct mem_ctl_info structure
1363 * Free up a previously allocated mci structure
1364 * A MCI structure can be in 2 states after being allocated
1365 * by edac_mc_alloc().
1366 * 1) Allocated in a MC driver's probe, but not yet committed
1367 * 2) Allocated and committed, by a call to edac_mc_add_mc()
1368 * edac_mc_add_mc() is the function that adds the sysfs entries
1369 * thus, this free function must determine which state the 'mci'
1370 * structure is in, then either free it directly or
1371 * perform kobject cleanup by calling edac_remove_sysfs_mci_device().
1375 void edac_mc_free(struct mem_ctl_info *mci)
1377 /* only if sysfs entries for this mci instance exist
1378 * do we remove them and defer the actual kfree via
1379 * the kobject 'release()' callback.
1381 * Otherwise, do a straight kfree now.
1383 if (mci->sysfs_active == MCI_SYSFS_ACTIVE)
1384 edac_remove_sysfs_mci_device(mci);
1391 EXPORT_SYMBOL(edac_mc_find_mci_by_pdev);
1393 struct mem_ctl_info *edac_mc_find_mci_by_pdev(struct pci_dev *pdev)
1395 struct mem_ctl_info *mci;
1396 struct list_head *item;
1398 debugf3("%s()\n", __func__);
1400 list_for_each(item, &mc_devices) {
1401 mci = list_entry(item, struct mem_ctl_info, link);
1403 if (mci->pdev == pdev)
1410 static int add_mc_to_global_list (struct mem_ctl_info *mci)
1412 struct list_head *item, *insert_before;
1413 struct mem_ctl_info *p;
1416 if (list_empty(&mc_devices)) {
1418 insert_before = &mc_devices;
1420 if (edac_mc_find_mci_by_pdev(mci->pdev)) {
1421 edac_printk(KERN_WARNING, EDAC_MC,
1422 "%s (%s) %s %s already assigned %d\n",
1423 mci->pdev->dev.bus_id,
1424 pci_name(mci->pdev), mci->mod_name,
1425 mci->ctl_name, mci->mc_idx);
1429 insert_before = NULL;
1432 list_for_each(item, &mc_devices) {
1433 p = list_entry(item, struct mem_ctl_info, link);
1435 if (p->mc_idx != i) {
1436 insert_before = item;
1445 if (insert_before == NULL)
1446 insert_before = &mc_devices;
1449 list_add_tail_rcu(&mci->link, insert_before);
1455 EXPORT_SYMBOL(edac_mc_add_mc);
1458 * edac_mc_add_mc: Insert the 'mci' structure into the mci global list
1459 * @mci: pointer to the mci structure to be added to the list
1466 /* FIXME - should a warning be printed if no error detection? correction? */
1467 int edac_mc_add_mc(struct mem_ctl_info *mci)
1471 debugf0("%s()\n", __func__);
1472 #ifdef CONFIG_EDAC_DEBUG
1473 if (edac_debug_level >= 3)
1474 edac_mc_dump_mci(mci);
1475 if (edac_debug_level >= 4) {
1478 for (i = 0; i < mci->nr_csrows; i++) {
1480 edac_mc_dump_csrow(&mci->csrows[i]);
1481 for (j = 0; j < mci->csrows[i].nr_channels; j++)
1482 edac_mc_dump_channel(&mci->csrows[i].
1487 down(&mem_ctls_mutex);
1489 if (add_mc_to_global_list(mci))
1492 /* set load time so that error rate can be tracked */
1493 mci->start_time = jiffies;
1495 if (edac_create_sysfs_mci_device(mci)) {
1496 edac_mc_printk(mci, KERN_WARNING,
1497 "failed to create sysfs device\n");
1498 /* FIXME - should there be an error code and unwind? */
1502 /* Report action taken */
1503 edac_mc_printk(mci, KERN_INFO, "Giving out device to %s %s: PCI %s\n",
1504 mci->mod_name, mci->ctl_name, pci_name(mci->pdev));
1509 up(&mem_ctls_mutex);
1515 static void complete_mc_list_del (struct rcu_head *head)
1517 struct mem_ctl_info *mci;
1519 mci = container_of(head, struct mem_ctl_info, rcu);
1520 INIT_LIST_HEAD(&mci->link);
1521 complete(&mci->complete);
1524 static void del_mc_from_global_list (struct mem_ctl_info *mci)
1526 list_del_rcu(&mci->link);
1527 init_completion(&mci->complete);
1528 call_rcu(&mci->rcu, complete_mc_list_del);
1529 wait_for_completion(&mci->complete);
1532 EXPORT_SYMBOL(edac_mc_del_mc);
1535 * edac_mc_del_mc: Remove the specified mci structure from global list
1536 * @mci: Pointer to struct mem_ctl_info structure
1542 int edac_mc_del_mc(struct mem_ctl_info *mci)
1546 debugf0("MC%d: %s()\n", mci->mc_idx, __func__);
1547 down(&mem_ctls_mutex);
1548 del_mc_from_global_list(mci);
1549 edac_printk(KERN_INFO, EDAC_MC,
1550 "Removed device %d for %s %s: PCI %s\n", mci->mc_idx,
1551 mci->mod_name, mci->ctl_name, pci_name(mci->pdev));
1553 up(&mem_ctls_mutex);
1559 EXPORT_SYMBOL(edac_mc_scrub_block);
1561 void edac_mc_scrub_block(unsigned long page, unsigned long offset,
1566 unsigned long flags = 0;
1568 debugf3("%s()\n", __func__);
1570 /* ECC error page was not in our memory. Ignore it. */
1571 if(!pfn_valid(page))
1574 /* Find the actual page structure then map it and fix */
1575 pg = pfn_to_page(page);
1577 if (PageHighMem(pg))
1578 local_irq_save(flags);
1580 virt_addr = kmap_atomic(pg, KM_BOUNCE_READ);
1582 /* Perform architecture specific atomic scrub operation */
1583 atomic_scrub(virt_addr + offset, size);
1585 /* Unmap and complete */
1586 kunmap_atomic(virt_addr, KM_BOUNCE_READ);
1588 if (PageHighMem(pg))
1589 local_irq_restore(flags);
1593 /* FIXME - should return -1 */
1594 EXPORT_SYMBOL(edac_mc_find_csrow_by_page);
1596 int edac_mc_find_csrow_by_page(struct mem_ctl_info *mci,
1599 struct csrow_info *csrows = mci->csrows;
1602 debugf1("MC%d: %s(): 0x%lx\n", mci->mc_idx, __func__, page);
1605 for (i = 0; i < mci->nr_csrows; i++) {
1606 struct csrow_info *csrow = &csrows[i];
1608 if (csrow->nr_pages == 0)
1611 debugf3("MC%d: %s(): first(0x%lx) page(0x%lx) last(0x%lx) "
1612 "mask(0x%lx)\n", mci->mc_idx, __func__,
1613 csrow->first_page, page, csrow->last_page,
1616 if ((page >= csrow->first_page) &&
1617 (page <= csrow->last_page) &&
1618 ((page & csrow->page_mask) ==
1619 (csrow->first_page & csrow->page_mask))) {
1626 edac_mc_printk(mci, KERN_ERR,
1627 "could not look up page error address %lx\n",
1628 (unsigned long) page);
1634 EXPORT_SYMBOL(edac_mc_handle_ce);
1636 /* FIXME - setable log (warning/emerg) levels */
1637 /* FIXME - integrate with evlog: http://evlog.sourceforge.net/ */
1638 void edac_mc_handle_ce(struct mem_ctl_info *mci,
1639 unsigned long page_frame_number,
1640 unsigned long offset_in_page,
1641 unsigned long syndrome, int row, int channel,
1644 unsigned long remapped_page;
1646 debugf3("MC%d: %s()\n", mci->mc_idx, __func__);
1648 /* FIXME - maybe make panic on INTERNAL ERROR an option */
1649 if (row >= mci->nr_csrows || row < 0) {
1650 /* something is wrong */
1651 edac_mc_printk(mci, KERN_ERR,
1652 "INTERNAL ERROR: row out of range "
1653 "(%d >= %d)\n", row, mci->nr_csrows);
1654 edac_mc_handle_ce_no_info(mci, "INTERNAL ERROR");
1657 if (channel >= mci->csrows[row].nr_channels || channel < 0) {
1658 /* something is wrong */
1659 edac_mc_printk(mci, KERN_ERR,
1660 "INTERNAL ERROR: channel out of range "
1661 "(%d >= %d)\n", channel,
1662 mci->csrows[row].nr_channels);
1663 edac_mc_handle_ce_no_info(mci, "INTERNAL ERROR");
1668 /* FIXME - put in DIMM location */
1669 edac_mc_printk(mci, KERN_WARNING,
1670 "CE page 0x%lx, offset 0x%lx, grain %d, syndrome "
1671 "0x%lx, row %d, channel %d, label \"%s\": %s\n",
1672 page_frame_number, offset_in_page,
1673 mci->csrows[row].grain, syndrome, row, channel,
1674 mci->csrows[row].channels[channel].label, msg);
1677 mci->csrows[row].ce_count++;
1678 mci->csrows[row].channels[channel].ce_count++;
1680 if (mci->scrub_mode & SCRUB_SW_SRC) {
1682 * Some MC's can remap memory so that it is still available
1683 * at a different address when PCI devices map into memory.
1684 * MC's that can't do this lose the memory where PCI devices
1685 * are mapped. This mapping is MC dependant and so we call
1686 * back into the MC driver for it to map the MC page to
1687 * a physical (CPU) page which can then be mapped to a virtual
1688 * page - which can then be scrubbed.
1690 remapped_page = mci->ctl_page_to_phys ?
1691 mci->ctl_page_to_phys(mci, page_frame_number) :
1694 edac_mc_scrub_block(remapped_page, offset_in_page,
1695 mci->csrows[row].grain);
1700 EXPORT_SYMBOL(edac_mc_handle_ce_no_info);
1702 void edac_mc_handle_ce_no_info(struct mem_ctl_info *mci,
1706 edac_mc_printk(mci, KERN_WARNING,
1707 "CE - no information available: %s\n", msg);
1708 mci->ce_noinfo_count++;
1713 EXPORT_SYMBOL(edac_mc_handle_ue);
1715 void edac_mc_handle_ue(struct mem_ctl_info *mci,
1716 unsigned long page_frame_number,
1717 unsigned long offset_in_page, int row,
1720 int len = EDAC_MC_LABEL_LEN * 4;
1721 char labels[len + 1];
1726 debugf3("MC%d: %s()\n", mci->mc_idx, __func__);
1728 /* FIXME - maybe make panic on INTERNAL ERROR an option */
1729 if (row >= mci->nr_csrows || row < 0) {
1730 /* something is wrong */
1731 edac_mc_printk(mci, KERN_ERR,
1732 "INTERNAL ERROR: row out of range "
1733 "(%d >= %d)\n", row, mci->nr_csrows);
1734 edac_mc_handle_ue_no_info(mci, "INTERNAL ERROR");
1738 chars = snprintf(pos, len + 1, "%s",
1739 mci->csrows[row].channels[0].label);
1742 for (chan = 1; (chan < mci->csrows[row].nr_channels) && (len > 0);
1744 chars = snprintf(pos, len + 1, ":%s",
1745 mci->csrows[row].channels[chan].label);
1751 edac_mc_printk(mci, KERN_EMERG,
1752 "UE page 0x%lx, offset 0x%lx, grain %d, row %d, "
1753 "labels \"%s\": %s\n", page_frame_number,
1754 offset_in_page, mci->csrows[row].grain, row, labels,
1759 ("EDAC MC%d: UE page 0x%lx, offset 0x%lx, grain %d, row %d,"
1760 " labels \"%s\": %s\n", mci->mc_idx,
1761 page_frame_number, offset_in_page,
1762 mci->csrows[row].grain, row, labels, msg);
1765 mci->csrows[row].ue_count++;
1769 EXPORT_SYMBOL(edac_mc_handle_ue_no_info);
1771 void edac_mc_handle_ue_no_info(struct mem_ctl_info *mci,
1775 panic("EDAC MC%d: Uncorrected Error", mci->mc_idx);
1778 edac_mc_printk(mci, KERN_WARNING,
1779 "UE - no information available: %s\n", msg);
1780 mci->ue_noinfo_count++;
1787 static u16 get_pci_parity_status(struct pci_dev *dev, int secondary)
1792 where = secondary ? PCI_SEC_STATUS : PCI_STATUS;
1793 pci_read_config_word(dev, where, &status);
1795 /* If we get back 0xFFFF then we must suspect that the card has been pulled but
1796 the Linux PCI layer has not yet finished cleaning up. We don't want to report
1799 if (status == 0xFFFF) {
1801 pci_read_config_dword(dev, 0, &sanity);
1802 if (sanity == 0xFFFFFFFF)
1805 status &= PCI_STATUS_DETECTED_PARITY | PCI_STATUS_SIG_SYSTEM_ERROR |
1809 /* reset only the bits we are interested in */
1810 pci_write_config_word(dev, where, status);
1815 typedef void (*pci_parity_check_fn_t) (struct pci_dev *dev);
1817 /* Clear any PCI parity errors logged by this device. */
1818 static void edac_pci_dev_parity_clear( struct pci_dev *dev )
1822 get_pci_parity_status(dev, 0);
1824 /* read the device TYPE, looking for bridges */
1825 pci_read_config_byte(dev, PCI_HEADER_TYPE, &header_type);
1827 if ((header_type & 0x7F) == PCI_HEADER_TYPE_BRIDGE)
1828 get_pci_parity_status(dev, 1);
1832 * PCI Parity polling
1835 static void edac_pci_dev_parity_test(struct pci_dev *dev)
1840 /* read the STATUS register on this device
1842 status = get_pci_parity_status(dev, 0);
1844 debugf2("PCI STATUS= 0x%04x %s\n", status, dev->dev.bus_id );
1846 /* check the status reg for errors */
1848 if (status & (PCI_STATUS_SIG_SYSTEM_ERROR))
1849 edac_printk(KERN_CRIT, EDAC_PCI,
1850 "Signaled System Error on %s\n",
1853 if (status & (PCI_STATUS_PARITY)) {
1854 edac_printk(KERN_CRIT, EDAC_PCI,
1855 "Master Data Parity Error on %s\n",
1858 atomic_inc(&pci_parity_count);
1861 if (status & (PCI_STATUS_DETECTED_PARITY)) {
1862 edac_printk(KERN_CRIT, EDAC_PCI,
1863 "Detected Parity Error on %s\n",
1866 atomic_inc(&pci_parity_count);
1870 /* read the device TYPE, looking for bridges */
1871 pci_read_config_byte(dev, PCI_HEADER_TYPE, &header_type);
1873 debugf2("PCI HEADER TYPE= 0x%02x %s\n", header_type, dev->dev.bus_id );
1875 if ((header_type & 0x7F) == PCI_HEADER_TYPE_BRIDGE) {
1876 /* On bridges, need to examine secondary status register */
1877 status = get_pci_parity_status(dev, 1);
1879 debugf2("PCI SEC_STATUS= 0x%04x %s\n",
1880 status, dev->dev.bus_id );
1882 /* check the secondary status reg for errors */
1884 if (status & (PCI_STATUS_SIG_SYSTEM_ERROR))
1885 edac_printk(KERN_CRIT, EDAC_PCI, "Bridge "
1886 "Signaled System Error on %s\n",
1889 if (status & (PCI_STATUS_PARITY)) {
1890 edac_printk(KERN_CRIT, EDAC_PCI, "Bridge "
1891 "Master Data Parity Error on "
1892 "%s\n", pci_name(dev));
1894 atomic_inc(&pci_parity_count);
1897 if (status & (PCI_STATUS_DETECTED_PARITY)) {
1898 edac_printk(KERN_CRIT, EDAC_PCI, "Bridge "
1899 "Detected Parity Error on %s\n",
1902 atomic_inc(&pci_parity_count);
1909 * check_dev_on_list: Scan for a PCI device on a white/black list
1910 * @list: an EDAC &edac_pci_device_list white/black list pointer
1911 * @free_index: index of next free entry on the list
1912 * @pci_dev: PCI Device pointer
1914 * see if list contains the device.
1916 * Returns: 0 not found
1919 static int check_dev_on_list(struct edac_pci_device_list *list, int free_index,
1920 struct pci_dev *dev)
1923 int rc = 0; /* Assume not found */
1924 unsigned short vendor=dev->vendor;
1925 unsigned short device=dev->device;
1927 /* Scan the list, looking for a vendor/device match
1929 for (i = 0; i < free_index; i++, list++ ) {
1930 if ( (list->vendor == vendor ) &&
1931 (list->device == device )) {
1941 * pci_dev parity list iterator
1942 * Scan the PCI device list for one iteration, looking for SERRORs
1943 * Master Parity ERRORS or Parity ERRORs on primary or secondary devices
1945 static inline void edac_pci_dev_parity_iterator(pci_parity_check_fn_t fn)
1947 struct pci_dev *dev=NULL;
1949 /* request for kernel access to the next PCI device, if any,
1950 * and while we are looking at it have its reference count
1951 * bumped until we are done with it
1953 while((dev = pci_get_device(PCI_ANY_ID, PCI_ANY_ID, dev)) != NULL) {
1955 /* if whitelist exists then it has priority, so only scan those
1956 * devices on the whitelist
1958 if (pci_whitelist_count > 0 ) {
1959 if (check_dev_on_list(pci_whitelist,
1960 pci_whitelist_count, dev))
1964 * if no whitelist, then check if this devices is
1967 if (!check_dev_on_list(pci_blacklist,
1968 pci_blacklist_count, dev))
1974 static void do_pci_parity_check(void)
1976 unsigned long flags;
1979 debugf3("%s()\n", __func__);
1981 if (!check_pci_parity)
1984 before_count = atomic_read(&pci_parity_count);
1986 /* scan all PCI devices looking for a Parity Error on devices and
1989 local_irq_save(flags);
1990 edac_pci_dev_parity_iterator(edac_pci_dev_parity_test);
1991 local_irq_restore(flags);
1993 /* Only if operator has selected panic on PCI Error */
1994 if (panic_on_pci_parity) {
1995 /* If the count is different 'after' from 'before' */
1996 if (before_count != atomic_read(&pci_parity_count))
1997 panic("EDAC: PCI Parity Error");
2002 static inline void clear_pci_parity_errors(void)
2004 /* Clear any PCI bus parity errors that devices initially have logged
2005 * in their registers.
2007 edac_pci_dev_parity_iterator(edac_pci_dev_parity_clear);
2011 #else /* CONFIG_PCI */
2014 static inline void do_pci_parity_check(void)
2020 static inline void clear_pci_parity_errors(void)
2026 #endif /* CONFIG_PCI */
2029 * Iterate over all MC instances and check for ECC, et al, errors
2031 static inline void check_mc_devices (void)
2033 unsigned long flags;
2034 struct list_head *item;
2035 struct mem_ctl_info *mci;
2037 debugf3("%s()\n", __func__);
2039 /* during poll, have interrupts off */
2040 local_irq_save(flags);
2042 list_for_each(item, &mc_devices) {
2043 mci = list_entry(item, struct mem_ctl_info, link);
2045 if (mci->edac_check != NULL)
2046 mci->edac_check(mci);
2049 local_irq_restore(flags);
2054 * Check MC status every poll_msec.
2055 * Check PCI status every poll_msec as well.
2057 * This where the work gets done for edac.
2059 * SMP safe, doesn't use NMI, and auto-rate-limits.
2061 static void do_edac_check(void)
2063 debugf3("%s()\n", __func__);
2065 do_pci_parity_check();
2068 static int edac_kernel_thread(void *arg)
2070 while (!kthread_should_stop()) {
2073 /* goto sleep for the interval */
2074 schedule_timeout_interruptible((HZ * poll_msec) / 1000);
2083 * module initialization entry point
2085 static int __init edac_mc_init(void)
2087 edac_printk(KERN_INFO, EDAC_MC, EDAC_MC_VERSION "\n");
2090 * Harvest and clear any boot/initialization PCI parity errors
2092 * FIXME: This only clears errors logged by devices present at time of
2093 * module initialization. We should also do an initial clear
2094 * of each newly hotplugged device.
2096 clear_pci_parity_errors();
2098 /* Create the MC sysfs entires */
2099 if (edac_sysfs_memctrl_setup()) {
2100 edac_printk(KERN_ERR, EDAC_MC,
2101 "Error initializing sysfs code\n");
2105 /* Create the PCI parity sysfs entries */
2106 if (edac_sysfs_pci_setup()) {
2107 edac_sysfs_memctrl_teardown();
2108 edac_printk(KERN_ERR, EDAC_MC,
2109 "EDAC PCI: Error initializing sysfs code\n");
2113 /* create our kernel thread */
2114 edac_thread = kthread_run(edac_kernel_thread, NULL, "kedac");
2115 if (IS_ERR(edac_thread)) {
2116 /* remove the sysfs entries */
2117 edac_sysfs_memctrl_teardown();
2118 edac_sysfs_pci_teardown();
2119 return PTR_ERR(edac_thread);
2128 * module exit/termination functioni
2130 static void __exit edac_mc_exit(void)
2132 debugf0("%s()\n", __func__);
2134 kthread_stop(edac_thread);
2136 /* tear down the sysfs device */
2137 edac_sysfs_memctrl_teardown();
2138 edac_sysfs_pci_teardown();
2144 module_init(edac_mc_init);
2145 module_exit(edac_mc_exit);
2147 MODULE_LICENSE("GPL");
2148 MODULE_AUTHOR("Linux Networx (http://lnxi.com) Thayne Harbaugh et al\n"
2149 "Based on.work by Dan Hollis et al");
2150 MODULE_DESCRIPTION("Core library routines for MC reporting");
2152 module_param(panic_on_ue, int, 0644);
2153 MODULE_PARM_DESC(panic_on_ue, "Panic on uncorrected error: 0=off 1=on");
2154 module_param(check_pci_parity, int, 0644);
2155 MODULE_PARM_DESC(check_pci_parity, "Check for PCI bus parity errors: 0=off 1=on");
2156 module_param(panic_on_pci_parity, int, 0644);
2157 MODULE_PARM_DESC(panic_on_pci_parity, "Panic on PCI Bus Parity error: 0=off 1=on");
2158 module_param(log_ue, int, 0644);
2159 MODULE_PARM_DESC(log_ue, "Log uncorrectable error to console: 0=off 1=on");
2160 module_param(log_ce, int, 0644);
2161 MODULE_PARM_DESC(log_ce, "Log correctable error to console: 0=off 1=on");
2162 module_param(poll_msec, int, 0644);
2163 MODULE_PARM_DESC(poll_msec, "Polling period in milliseconds");
2164 #ifdef CONFIG_EDAC_DEBUG
2165 module_param(edac_debug_level, int, 0644);
2166 MODULE_PARM_DESC(edac_debug_level, "Debug level");