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/edac.h>
31 #include <asm/uaccess.h>
34 #include "edac_core.h"
35 #include "edac_module.h"
38 /* lock to memory controller's control array */
39 static DEFINE_MUTEX(mem_ctls_mutex);
40 static struct list_head mc_devices = LIST_HEAD_INIT(mc_devices);
42 #ifdef CONFIG_EDAC_DEBUG
44 static void edac_mc_dump_channel(struct channel_info *chan)
46 debugf4("\tchannel = %p\n", chan);
47 debugf4("\tchannel->chan_idx = %d\n", chan->chan_idx);
48 debugf4("\tchannel->ce_count = %d\n", chan->ce_count);
49 debugf4("\tchannel->label = '%s'\n", chan->label);
50 debugf4("\tchannel->csrow = %p\n\n", chan->csrow);
53 static void edac_mc_dump_csrow(struct csrow_info *csrow)
55 debugf4("\tcsrow = %p\n", csrow);
56 debugf4("\tcsrow->csrow_idx = %d\n", csrow->csrow_idx);
57 debugf4("\tcsrow->first_page = 0x%lx\n",
59 debugf4("\tcsrow->last_page = 0x%lx\n", csrow->last_page);
60 debugf4("\tcsrow->page_mask = 0x%lx\n", csrow->page_mask);
61 debugf4("\tcsrow->nr_pages = 0x%x\n", csrow->nr_pages);
62 debugf4("\tcsrow->nr_channels = %d\n",
64 debugf4("\tcsrow->channels = %p\n", csrow->channels);
65 debugf4("\tcsrow->mci = %p\n\n", csrow->mci);
68 static void edac_mc_dump_mci(struct mem_ctl_info *mci)
70 debugf3("\tmci = %p\n", mci);
71 debugf3("\tmci->mtype_cap = %lx\n", mci->mtype_cap);
72 debugf3("\tmci->edac_ctl_cap = %lx\n", mci->edac_ctl_cap);
73 debugf3("\tmci->edac_cap = %lx\n", mci->edac_cap);
74 debugf4("\tmci->edac_check = %p\n", mci->edac_check);
75 debugf3("\tmci->nr_csrows = %d, csrows = %p\n",
76 mci->nr_csrows, mci->csrows);
77 debugf3("\tdev = %p\n", mci->dev);
78 debugf3("\tmod_name:ctl_name = %s:%s\n",
79 mci->mod_name, mci->ctl_name);
80 debugf3("\tpvt_info = %p\n\n", mci->pvt_info);
83 #endif /* CONFIG_EDAC_DEBUG */
85 /* 'ptr' points to a possibly unaligned item X such that sizeof(X) is 'size'.
86 * Adjust 'ptr' so that its alignment is at least as stringent as what the
87 * compiler would provide for X and return the aligned result.
89 * If 'size' is a constant, the compiler will optimize this whole function
90 * down to either a no-op or the addition of a constant to the value of 'ptr'.
92 char * edac_align_ptr(void *ptr, unsigned size)
96 /* Here we assume that the alignment of a "long long" is the most
97 * stringent alignment that the compiler will ever provide by default.
98 * As far as I know, this is a reasonable assumption.
100 if (size > sizeof(long))
101 align = sizeof(long long);
102 else if (size > sizeof(int))
103 align = sizeof(long);
104 else if (size > sizeof(short))
106 else if (size > sizeof(char))
107 align = sizeof(short);
116 return (char *) (((unsigned long) ptr) + align - r);
120 * edac_mc_alloc: Allocate a struct mem_ctl_info structure
121 * @size_pvt: size of private storage needed
122 * @nr_csrows: Number of CWROWS needed for this MC
123 * @nr_chans: Number of channels for the MC
125 * Everything is kmalloc'ed as one big chunk - more efficient.
126 * Only can be used if all structures have the same lifetime - otherwise
127 * you have to allocate and initialize your own structures.
129 * Use edac_mc_free() to free mc structures allocated by this function.
132 * NULL allocation failed
133 * struct mem_ctl_info pointer
135 struct mem_ctl_info *edac_mc_alloc(unsigned sz_pvt, unsigned nr_csrows,
138 struct mem_ctl_info *mci;
139 struct csrow_info *csi, *csrow;
140 struct channel_info *chi, *chp, *chan;
145 /* Figure out the offsets of the various items from the start of an mc
146 * structure. We want the alignment of each item to be at least as
147 * stringent as what the compiler would provide if we could simply
148 * hardcode everything into a single struct.
150 mci = (struct mem_ctl_info *) 0;
151 csi = (struct csrow_info *)edac_align_ptr(&mci[1], sizeof(*csi));
152 chi = (struct channel_info *)
153 edac_align_ptr(&csi[nr_csrows], sizeof(*chi));
154 pvt = edac_align_ptr(&chi[nr_chans * nr_csrows], sz_pvt);
155 size = ((unsigned long) pvt) + sz_pvt;
157 if ((mci = kmalloc(size, GFP_KERNEL)) == NULL)
160 /* Adjust pointers so they point within the memory we just allocated
161 * rather than an imaginary chunk of memory located at address 0.
163 csi = (struct csrow_info *) (((char *) mci) + ((unsigned long) csi));
164 chi = (struct channel_info *) (((char *) mci) + ((unsigned long) chi));
165 pvt = sz_pvt ? (((char *) mci) + ((unsigned long) pvt)) : NULL;
167 memset(mci, 0, size); /* clear all fields */
170 mci->nr_csrows = nr_csrows;
172 for (row = 0; row < nr_csrows; row++) {
174 csrow->csrow_idx = row;
176 csrow->nr_channels = nr_chans;
177 chp = &chi[row * nr_chans];
178 csrow->channels = chp;
180 for (chn = 0; chn < nr_chans; chn++) {
182 chan->chan_idx = chn;
187 mci->op_state = OP_ALLOC;
191 EXPORT_SYMBOL_GPL(edac_mc_alloc);
194 * edac_mc_free: Free a previously allocated 'mci' structure
195 * @mci: pointer to a struct mem_ctl_info structure
197 void edac_mc_free(struct mem_ctl_info *mci)
201 EXPORT_SYMBOL_GPL(edac_mc_free);
203 static struct mem_ctl_info *find_mci_by_dev(struct device *dev)
205 struct mem_ctl_info *mci;
206 struct list_head *item;
208 debugf3("%s()\n", __func__);
210 list_for_each(item, &mc_devices) {
211 mci = list_entry(item, struct mem_ctl_info, link);
221 * handler for EDAC to check if NMI type handler has asserted interrupt
223 static int edac_mc_assert_error_check_and_clear(void)
227 if(edac_op_state == EDAC_OPSTATE_POLL)
230 old_state = edac_err_assert;
237 * edac_mc_workq_function
238 * performs the operation scheduled by a workq request
240 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,20))
241 static void edac_mc_workq_function(struct work_struct *work_req)
243 struct delayed_work *d_work = (struct delayed_work*) work_req;
244 struct mem_ctl_info *mci = to_edac_mem_ctl_work(d_work);
246 static void edac_mc_workq_function(void *ptr)
248 struct mem_ctl_info *mci = (struct mem_ctl_info *) ptr;
251 mutex_lock(&mem_ctls_mutex);
253 /* Only poll controllers that are running polled and have a check */
254 if (edac_mc_assert_error_check_and_clear() && (mci->edac_check != NULL))
255 mci->edac_check(mci);
258 * FIXME: temp place holder for PCI checks,
259 * goes away when we break out PCI
261 edac_pci_do_parity_check();
263 mutex_unlock(&mem_ctls_mutex);
266 queue_delayed_work(edac_workqueue, &mci->work, edac_mc_get_poll_msec());
270 * edac_mc_workq_setup
271 * initialize a workq item for this mci
272 * passing in the new delay period in msec
274 void edac_mc_workq_setup(struct mem_ctl_info *mci, unsigned msec)
276 debugf0("%s()\n", __func__);
278 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,20))
279 INIT_DELAYED_WORK(&mci->work, edac_mc_workq_function);
281 INIT_WORK(&mci->work, edac_mc_workq_function, mci);
283 queue_delayed_work(edac_workqueue, &mci->work, msecs_to_jiffies(msec));
287 * edac_mc_workq_teardown
288 * stop the workq processing on this mci
290 void edac_mc_workq_teardown(struct mem_ctl_info *mci)
294 status = cancel_delayed_work(&mci->work);
296 /* workq instance might be running, wait for it */
297 flush_workqueue(edac_workqueue);
302 * edac_reset_delay_period
305 void edac_reset_delay_period(struct mem_ctl_info *mci, unsigned long value)
307 mutex_lock(&mem_ctls_mutex);
309 /* cancel the current workq request */
310 edac_mc_workq_teardown(mci);
312 /* restart the workq request, with new delay value */
313 edac_mc_workq_setup(mci, value);
315 mutex_unlock(&mem_ctls_mutex);
318 /* Return 0 on success, 1 on failure.
319 * Before calling this function, caller must
320 * assign a unique value to mci->mc_idx.
322 static int add_mc_to_global_list (struct mem_ctl_info *mci)
324 struct list_head *item, *insert_before;
325 struct mem_ctl_info *p;
327 insert_before = &mc_devices;
329 if (unlikely((p = find_mci_by_dev(mci->dev)) != NULL))
332 list_for_each(item, &mc_devices) {
333 p = list_entry(item, struct mem_ctl_info, link);
335 if (p->mc_idx >= mci->mc_idx) {
336 if (unlikely(p->mc_idx == mci->mc_idx))
339 insert_before = item;
344 list_add_tail_rcu(&mci->link, insert_before);
345 atomic_inc(&edac_handlers);
349 edac_printk(KERN_WARNING, EDAC_MC,
350 "%s (%s) %s %s already assigned %d\n", p->dev->bus_id,
351 dev_name(mci), p->mod_name, p->ctl_name, p->mc_idx);
355 edac_printk(KERN_WARNING, EDAC_MC,
356 "bug in low-level driver: attempt to assign\n"
357 " duplicate mc_idx %d in %s()\n", p->mc_idx, __func__);
361 static void complete_mc_list_del(struct rcu_head *head)
363 struct mem_ctl_info *mci;
365 mci = container_of(head, struct mem_ctl_info, rcu);
366 INIT_LIST_HEAD(&mci->link);
367 complete(&mci->complete);
370 static void del_mc_from_global_list(struct mem_ctl_info *mci)
372 atomic_dec(&edac_handlers);
373 list_del_rcu(&mci->link);
374 init_completion(&mci->complete);
375 call_rcu(&mci->rcu, complete_mc_list_del);
376 wait_for_completion(&mci->complete);
380 * edac_mc_find: Search for a mem_ctl_info structure whose index is 'idx'.
382 * If found, return a pointer to the structure.
385 * Caller must hold mem_ctls_mutex.
387 struct mem_ctl_info * edac_mc_find(int idx)
389 struct list_head *item;
390 struct mem_ctl_info *mci;
392 list_for_each(item, &mc_devices) {
393 mci = list_entry(item, struct mem_ctl_info, link);
395 if (mci->mc_idx >= idx) {
396 if (mci->mc_idx == idx)
405 EXPORT_SYMBOL(edac_mc_find);
408 * edac_mc_add_mc: Insert the 'mci' structure into the mci global list and
409 * create sysfs entries associated with mci structure
410 * @mci: pointer to the mci structure to be added to the list
411 * @mc_idx: A unique numeric identifier to be assigned to the 'mci' structure.
418 /* FIXME - should a warning be printed if no error detection? correction? */
419 int edac_mc_add_mc(struct mem_ctl_info *mci, int mc_idx)
421 debugf0("%s()\n", __func__);
422 mci->mc_idx = mc_idx;
423 #ifdef CONFIG_EDAC_DEBUG
424 if (edac_debug_level >= 3)
425 edac_mc_dump_mci(mci);
427 if (edac_debug_level >= 4) {
430 for (i = 0; i < mci->nr_csrows; i++) {
433 edac_mc_dump_csrow(&mci->csrows[i]);
434 for (j = 0; j < mci->csrows[i].nr_channels; j++)
435 edac_mc_dump_channel(
436 &mci->csrows[i].channels[j]);
440 mutex_lock(&mem_ctls_mutex);
442 if (add_mc_to_global_list(mci))
445 /* set load time so that error rate can be tracked */
446 mci->start_time = jiffies;
448 if (edac_create_sysfs_mci_device(mci)) {
449 edac_mc_printk(mci, KERN_WARNING,
450 "failed to create sysfs device\n");
454 /* If there IS a check routine, then we are running POLLED */
455 if (mci->edac_check != NULL) {
456 /* This instance is NOW RUNNING */
457 mci->op_state = OP_RUNNING_POLL;
459 edac_mc_workq_setup(mci, edac_mc_get_poll_msec());
461 mci->op_state = OP_RUNNING_INTERRUPT;
464 /* Report action taken */
465 edac_mc_printk(mci, KERN_INFO, "Giving out device to %s %s: DEV %s\n",
466 mci->mod_name, mci->ctl_name, dev_name(mci));
468 mutex_unlock(&mem_ctls_mutex);
472 del_mc_from_global_list(mci);
475 mutex_unlock(&mem_ctls_mutex);
478 EXPORT_SYMBOL_GPL(edac_mc_add_mc);
481 * edac_mc_del_mc: Remove sysfs entries for specified mci structure and
482 * remove mci structure from global list
483 * @pdev: Pointer to 'struct device' representing mci structure to remove.
485 * Return pointer to removed mci structure, or NULL if device not found.
487 struct mem_ctl_info * edac_mc_del_mc(struct device *dev)
489 struct mem_ctl_info *mci;
491 debugf0("MC: %s()\n", __func__);
492 mutex_lock(&mem_ctls_mutex);
494 if ((mci = find_mci_by_dev(dev)) == NULL) {
495 mutex_unlock(&mem_ctls_mutex);
499 /* marking MCI offline */
500 mci->op_state = OP_OFFLINE;
502 /* flush workq processes */
503 edac_mc_workq_teardown(mci);
505 edac_remove_sysfs_mci_device(mci);
506 del_mc_from_global_list(mci);
507 mutex_unlock(&mem_ctls_mutex);
508 edac_printk(KERN_INFO, EDAC_MC,
509 "Removed device %d for %s %s: DEV %s\n", mci->mc_idx,
510 mci->mod_name, mci->ctl_name, dev_name(mci));
513 EXPORT_SYMBOL_GPL(edac_mc_del_mc);
515 static void edac_mc_scrub_block(unsigned long page, unsigned long offset,
520 unsigned long flags = 0;
522 debugf3("%s()\n", __func__);
524 /* ECC error page was not in our memory. Ignore it. */
528 /* Find the actual page structure then map it and fix */
529 pg = pfn_to_page(page);
532 local_irq_save(flags);
534 virt_addr = kmap_atomic(pg, KM_BOUNCE_READ);
536 /* Perform architecture specific atomic scrub operation */
537 atomic_scrub(virt_addr + offset, size);
539 /* Unmap and complete */
540 kunmap_atomic(virt_addr, KM_BOUNCE_READ);
543 local_irq_restore(flags);
546 /* FIXME - should return -1 */
547 int edac_mc_find_csrow_by_page(struct mem_ctl_info *mci, unsigned long page)
549 struct csrow_info *csrows = mci->csrows;
552 debugf1("MC%d: %s(): 0x%lx\n", mci->mc_idx, __func__, page);
555 for (i = 0; i < mci->nr_csrows; i++) {
556 struct csrow_info *csrow = &csrows[i];
558 if (csrow->nr_pages == 0)
561 debugf3("MC%d: %s(): first(0x%lx) page(0x%lx) last(0x%lx) "
562 "mask(0x%lx)\n", mci->mc_idx, __func__,
563 csrow->first_page, page, csrow->last_page,
566 if ((page >= csrow->first_page) &&
567 (page <= csrow->last_page) &&
568 ((page & csrow->page_mask) ==
569 (csrow->first_page & csrow->page_mask))) {
576 edac_mc_printk(mci, KERN_ERR,
577 "could not look up page error address %lx\n",
578 (unsigned long) page);
582 EXPORT_SYMBOL_GPL(edac_mc_find_csrow_by_page);
584 /* FIXME - setable log (warning/emerg) levels */
585 /* FIXME - integrate with evlog: http://evlog.sourceforge.net/ */
586 void edac_mc_handle_ce(struct mem_ctl_info *mci,
587 unsigned long page_frame_number, unsigned long offset_in_page,
588 unsigned long syndrome, int row, int channel, const char *msg)
590 unsigned long remapped_page;
592 debugf3("MC%d: %s()\n", mci->mc_idx, __func__);
594 /* FIXME - maybe make panic on INTERNAL ERROR an option */
595 if (row >= mci->nr_csrows || row < 0) {
596 /* something is wrong */
597 edac_mc_printk(mci, KERN_ERR,
598 "INTERNAL ERROR: row out of range "
599 "(%d >= %d)\n", row, mci->nr_csrows);
600 edac_mc_handle_ce_no_info(mci, "INTERNAL ERROR");
604 if (channel >= mci->csrows[row].nr_channels || channel < 0) {
605 /* something is wrong */
606 edac_mc_printk(mci, KERN_ERR,
607 "INTERNAL ERROR: channel out of range "
608 "(%d >= %d)\n", channel,
609 mci->csrows[row].nr_channels);
610 edac_mc_handle_ce_no_info(mci, "INTERNAL ERROR");
614 if (edac_get_log_ce())
615 /* FIXME - put in DIMM location */
616 edac_mc_printk(mci, KERN_WARNING,
617 "CE page 0x%lx, offset 0x%lx, grain %d, syndrome "
618 "0x%lx, row %d, channel %d, label \"%s\": %s\n",
619 page_frame_number, offset_in_page,
620 mci->csrows[row].grain, syndrome, row, channel,
621 mci->csrows[row].channels[channel].label, msg);
624 mci->csrows[row].ce_count++;
625 mci->csrows[row].channels[channel].ce_count++;
627 if (mci->scrub_mode & SCRUB_SW_SRC) {
629 * Some MC's can remap memory so that it is still available
630 * at a different address when PCI devices map into memory.
631 * MC's that can't do this lose the memory where PCI devices
632 * are mapped. This mapping is MC dependant and so we call
633 * back into the MC driver for it to map the MC page to
634 * a physical (CPU) page which can then be mapped to a virtual
635 * page - which can then be scrubbed.
637 remapped_page = mci->ctl_page_to_phys ?
638 mci->ctl_page_to_phys(mci, page_frame_number) :
641 edac_mc_scrub_block(remapped_page, offset_in_page,
642 mci->csrows[row].grain);
645 EXPORT_SYMBOL_GPL(edac_mc_handle_ce);
647 void edac_mc_handle_ce_no_info(struct mem_ctl_info *mci, const char *msg)
649 if (edac_get_log_ce())
650 edac_mc_printk(mci, KERN_WARNING,
651 "CE - no information available: %s\n", msg);
653 mci->ce_noinfo_count++;
656 EXPORT_SYMBOL_GPL(edac_mc_handle_ce_no_info);
658 void edac_mc_handle_ue(struct mem_ctl_info *mci,
659 unsigned long page_frame_number, unsigned long offset_in_page,
660 int row, const char *msg)
662 int len = EDAC_MC_LABEL_LEN * 4;
663 char labels[len + 1];
668 debugf3("MC%d: %s()\n", mci->mc_idx, __func__);
670 /* FIXME - maybe make panic on INTERNAL ERROR an option */
671 if (row >= mci->nr_csrows || row < 0) {
672 /* something is wrong */
673 edac_mc_printk(mci, KERN_ERR,
674 "INTERNAL ERROR: row out of range "
675 "(%d >= %d)\n", row, mci->nr_csrows);
676 edac_mc_handle_ue_no_info(mci, "INTERNAL ERROR");
680 chars = snprintf(pos, len + 1, "%s",
681 mci->csrows[row].channels[0].label);
685 for (chan = 1; (chan < mci->csrows[row].nr_channels) && (len > 0);
687 chars = snprintf(pos, len + 1, ":%s",
688 mci->csrows[row].channels[chan].label);
693 if (edac_get_log_ue())
694 edac_mc_printk(mci, KERN_EMERG,
695 "UE page 0x%lx, offset 0x%lx, grain %d, row %d, "
696 "labels \"%s\": %s\n", page_frame_number,
697 offset_in_page, mci->csrows[row].grain, row, labels,
700 if (edac_get_panic_on_ue())
701 panic("EDAC MC%d: UE page 0x%lx, offset 0x%lx, grain %d, "
702 "row %d, labels \"%s\": %s\n", mci->mc_idx,
703 page_frame_number, offset_in_page,
704 mci->csrows[row].grain, row, labels, msg);
707 mci->csrows[row].ue_count++;
709 EXPORT_SYMBOL_GPL(edac_mc_handle_ue);
711 void edac_mc_handle_ue_no_info(struct mem_ctl_info *mci, const char *msg)
713 if (edac_get_panic_on_ue())
714 panic("EDAC MC%d: Uncorrected Error", mci->mc_idx);
716 if (edac_get_log_ue())
717 edac_mc_printk(mci, KERN_WARNING,
718 "UE - no information available: %s\n", msg);
719 mci->ue_noinfo_count++;
722 EXPORT_SYMBOL_GPL(edac_mc_handle_ue_no_info);
725 /*************************************************************
726 * On Fully Buffered DIMM modules, this help function is
727 * called to process UE events
729 void edac_mc_handle_fbd_ue(struct mem_ctl_info *mci,
731 unsigned int channela,
732 unsigned int channelb,
735 int len = EDAC_MC_LABEL_LEN * 4;
736 char labels[len + 1];
740 if (csrow >= mci->nr_csrows) {
741 /* something is wrong */
742 edac_mc_printk(mci, KERN_ERR,
743 "INTERNAL ERROR: row out of range (%d >= %d)\n",
744 csrow, mci->nr_csrows);
745 edac_mc_handle_ue_no_info(mci, "INTERNAL ERROR");
749 if (channela >= mci->csrows[csrow].nr_channels) {
750 /* something is wrong */
751 edac_mc_printk(mci, KERN_ERR,
752 "INTERNAL ERROR: channel-a out of range "
754 channela, mci->csrows[csrow].nr_channels);
755 edac_mc_handle_ue_no_info(mci, "INTERNAL ERROR");
759 if (channelb >= mci->csrows[csrow].nr_channels) {
760 /* something is wrong */
761 edac_mc_printk(mci, KERN_ERR,
762 "INTERNAL ERROR: channel-b out of range "
764 channelb, mci->csrows[csrow].nr_channels);
765 edac_mc_handle_ue_no_info(mci, "INTERNAL ERROR");
770 mci->csrows[csrow].ue_count++;
772 /* Generate the DIMM labels from the specified channels */
773 chars = snprintf(pos, len + 1, "%s",
774 mci->csrows[csrow].channels[channela].label);
775 len -= chars; pos += chars;
776 chars = snprintf(pos, len + 1, "-%s",
777 mci->csrows[csrow].channels[channelb].label);
779 if (edac_get_log_ue())
780 edac_mc_printk(mci, KERN_EMERG,
781 "UE row %d, channel-a= %d channel-b= %d "
782 "labels \"%s\": %s\n", csrow, channela, channelb,
785 if (edac_get_panic_on_ue())
786 panic("UE row %d, channel-a= %d channel-b= %d "
787 "labels \"%s\": %s\n", csrow, channela,
788 channelb, labels, msg);
790 EXPORT_SYMBOL(edac_mc_handle_fbd_ue);
792 /*************************************************************
793 * On Fully Buffered DIMM modules, this help function is
794 * called to process CE events
796 void edac_mc_handle_fbd_ce(struct mem_ctl_info *mci,
798 unsigned int channel,
802 /* Ensure boundary values */
803 if (csrow >= mci->nr_csrows) {
804 /* something is wrong */
805 edac_mc_printk(mci, KERN_ERR,
806 "INTERNAL ERROR: row out of range (%d >= %d)\n",
807 csrow, mci->nr_csrows);
808 edac_mc_handle_ce_no_info(mci, "INTERNAL ERROR");
811 if (channel >= mci->csrows[csrow].nr_channels) {
812 /* something is wrong */
813 edac_mc_printk(mci, KERN_ERR,
814 "INTERNAL ERROR: channel out of range (%d >= %d)\n",
815 channel, mci->csrows[csrow].nr_channels);
816 edac_mc_handle_ce_no_info(mci, "INTERNAL ERROR");
820 if (edac_get_log_ce())
821 /* FIXME - put in DIMM location */
822 edac_mc_printk(mci, KERN_WARNING,
823 "CE row %d, channel %d, label \"%s\": %s\n",
825 mci->csrows[csrow].channels[channel].label,
829 mci->csrows[csrow].ce_count++;
830 mci->csrows[csrow].channels[channel].ce_count++;
832 EXPORT_SYMBOL(edac_mc_handle_fbd_ce);
836 * Iterate over all MC instances and check for ECC, et al, errors
838 void edac_check_mc_devices(void)
840 struct list_head *item;
841 struct mem_ctl_info *mci;
843 debugf3("%s()\n", __func__);
844 mutex_lock(&mem_ctls_mutex);
846 list_for_each(item, &mc_devices) {
847 mci = list_entry(item, struct mem_ctl_info, link);
849 if (mci->edac_check != NULL)
850 mci->edac_check(mci);
853 mutex_unlock(&mem_ctls_mutex);