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"
37 /* lock to memory controller's control array */
38 static DEFINE_MUTEX(mem_ctls_mutex
);
39 static struct list_head mc_devices
= LIST_HEAD_INIT(mc_devices
);
41 #ifdef CONFIG_EDAC_DEBUG
43 static void edac_mc_dump_channel(struct channel_info
*chan
)
45 debugf4("\tchannel = %p\n", chan
);
46 debugf4("\tchannel->chan_idx = %d\n", chan
->chan_idx
);
47 debugf4("\tchannel->ce_count = %d\n", chan
->ce_count
);
48 debugf4("\tchannel->label = '%s'\n", chan
->label
);
49 debugf4("\tchannel->csrow = %p\n\n", chan
->csrow
);
52 static void edac_mc_dump_csrow(struct csrow_info
*csrow
)
54 debugf4("\tcsrow = %p\n", csrow
);
55 debugf4("\tcsrow->csrow_idx = %d\n", csrow
->csrow_idx
);
56 debugf4("\tcsrow->first_page = 0x%lx\n", csrow
->first_page
);
57 debugf4("\tcsrow->last_page = 0x%lx\n", csrow
->last_page
);
58 debugf4("\tcsrow->page_mask = 0x%lx\n", csrow
->page_mask
);
59 debugf4("\tcsrow->nr_pages = 0x%x\n", csrow
->nr_pages
);
60 debugf4("\tcsrow->nr_channels = %d\n", csrow
->nr_channels
);
61 debugf4("\tcsrow->channels = %p\n", csrow
->channels
);
62 debugf4("\tcsrow->mci = %p\n\n", csrow
->mci
);
65 static void edac_mc_dump_mci(struct mem_ctl_info
*mci
)
67 debugf3("\tmci = %p\n", mci
);
68 debugf3("\tmci->mtype_cap = %lx\n", mci
->mtype_cap
);
69 debugf3("\tmci->edac_ctl_cap = %lx\n", mci
->edac_ctl_cap
);
70 debugf3("\tmci->edac_cap = %lx\n", mci
->edac_cap
);
71 debugf4("\tmci->edac_check = %p\n", mci
->edac_check
);
72 debugf3("\tmci->nr_csrows = %d, csrows = %p\n",
73 mci
->nr_csrows
, mci
->csrows
);
74 debugf3("\tdev = %p\n", mci
->dev
);
75 debugf3("\tmod_name:ctl_name = %s:%s\n", mci
->mod_name
, mci
->ctl_name
);
76 debugf3("\tpvt_info = %p\n\n", mci
->pvt_info
);
79 #endif /* CONFIG_EDAC_DEBUG */
81 /* 'ptr' points to a possibly unaligned item X such that sizeof(X) is 'size'.
82 * Adjust 'ptr' so that its alignment is at least as stringent as what the
83 * compiler would provide for X and return the aligned result.
85 * If 'size' is a constant, the compiler will optimize this whole function
86 * down to either a no-op or the addition of a constant to the value of 'ptr'.
88 void *edac_align_ptr(void *ptr
, unsigned size
)
92 /* Here we assume that the alignment of a "long long" is the most
93 * stringent alignment that the compiler will ever provide by default.
94 * As far as I know, this is a reasonable assumption.
96 if (size
> sizeof(long))
97 align
= sizeof(long long);
98 else if (size
> sizeof(int))
100 else if (size
> sizeof(short))
102 else if (size
> sizeof(char))
103 align
= sizeof(short);
112 return (void *)(((unsigned long)ptr
) + align
- r
);
116 * edac_mc_alloc: Allocate a struct mem_ctl_info structure
117 * @size_pvt: size of private storage needed
118 * @nr_csrows: Number of CWROWS needed for this MC
119 * @nr_chans: Number of channels for the MC
121 * Everything is kmalloc'ed as one big chunk - more efficient.
122 * Only can be used if all structures have the same lifetime - otherwise
123 * you have to allocate and initialize your own structures.
125 * Use edac_mc_free() to free mc structures allocated by this function.
128 * NULL allocation failed
129 * struct mem_ctl_info pointer
131 struct mem_ctl_info
*edac_mc_alloc(unsigned sz_pvt
, unsigned nr_csrows
,
134 struct mem_ctl_info
*mci
;
135 struct csrow_info
*csi
, *csrow
;
136 struct channel_info
*chi
, *chp
, *chan
;
141 /* Figure out the offsets of the various items from the start of an mc
142 * structure. We want the alignment of each item to be at least as
143 * stringent as what the compiler would provide if we could simply
144 * hardcode everything into a single struct.
146 mci
= (struct mem_ctl_info
*)0;
147 csi
= edac_align_ptr(&mci
[1], sizeof(*csi
));
148 chi
= edac_align_ptr(&csi
[nr_csrows
], sizeof(*chi
));
149 pvt
= edac_align_ptr(&chi
[nr_chans
* nr_csrows
], sz_pvt
);
150 size
= ((unsigned long)pvt
) + sz_pvt
;
152 if ((mci
= kmalloc(size
, GFP_KERNEL
)) == NULL
)
155 /* Adjust pointers so they point within the memory we just allocated
156 * rather than an imaginary chunk of memory located at address 0.
158 csi
= (struct csrow_info
*)(((char *)mci
) + ((unsigned long)csi
));
159 chi
= (struct channel_info
*)(((char *)mci
) + ((unsigned long)chi
));
160 pvt
= sz_pvt
? (((char *)mci
) + ((unsigned long)pvt
)) : NULL
;
162 memset(mci
, 0, size
); /* clear all fields */
165 mci
->nr_csrows
= nr_csrows
;
167 for (row
= 0; row
< nr_csrows
; row
++) {
169 csrow
->csrow_idx
= row
;
171 csrow
->nr_channels
= nr_chans
;
172 chp
= &chi
[row
* nr_chans
];
173 csrow
->channels
= chp
;
175 for (chn
= 0; chn
< nr_chans
; chn
++) {
177 chan
->chan_idx
= chn
;
182 mci
->op_state
= OP_ALLOC
;
187 EXPORT_SYMBOL_GPL(edac_mc_alloc
);
190 * edac_mc_free: Free a previously allocated 'mci' structure
191 * @mci: pointer to a struct mem_ctl_info structure
193 void edac_mc_free(struct mem_ctl_info
*mci
)
198 EXPORT_SYMBOL_GPL(edac_mc_free
);
200 static struct mem_ctl_info
*find_mci_by_dev(struct device
*dev
)
202 struct mem_ctl_info
*mci
;
203 struct list_head
*item
;
205 debugf3("%s()\n", __func__
);
207 list_for_each(item
, &mc_devices
) {
208 mci
= list_entry(item
, struct mem_ctl_info
, link
);
218 * handler for EDAC to check if NMI type handler has asserted interrupt
220 static int edac_mc_assert_error_check_and_clear(void)
224 if (edac_op_state
== EDAC_OPSTATE_POLL
)
227 old_state
= edac_err_assert
;
234 * edac_mc_workq_function
235 * performs the operation scheduled by a workq request
237 static void edac_mc_workq_function(struct work_struct
*work_req
)
239 struct delayed_work
*d_work
= (struct delayed_work
*)work_req
;
240 struct mem_ctl_info
*mci
= to_edac_mem_ctl_work(d_work
);
242 mutex_lock(&mem_ctls_mutex
);
244 /* Only poll controllers that are running polled and have a check */
245 if (edac_mc_assert_error_check_and_clear() && (mci
->edac_check
!= NULL
))
246 mci
->edac_check(mci
);
249 * FIXME: temp place holder for PCI checks,
250 * goes away when we break out PCI
252 edac_pci_do_parity_check();
254 mutex_unlock(&mem_ctls_mutex
);
257 queue_delayed_work(edac_workqueue
, &mci
->work
,
258 msecs_to_jiffies(edac_mc_get_poll_msec()));
262 * edac_mc_workq_setup
263 * initialize a workq item for this mci
264 * passing in the new delay period in msec
266 void edac_mc_workq_setup(struct mem_ctl_info
*mci
, unsigned msec
)
268 debugf0("%s()\n", __func__
);
270 INIT_DELAYED_WORK(&mci
->work
, edac_mc_workq_function
);
271 queue_delayed_work(edac_workqueue
, &mci
->work
, msecs_to_jiffies(msec
));
275 * edac_mc_workq_teardown
276 * stop the workq processing on this mci
278 void edac_mc_workq_teardown(struct mem_ctl_info
*mci
)
282 status
= cancel_delayed_work(&mci
->work
);
284 /* workq instance might be running, wait for it */
285 flush_workqueue(edac_workqueue
);
290 * edac_reset_delay_period
293 void edac_reset_delay_period(struct mem_ctl_info
*mci
, unsigned long value
)
295 mutex_lock(&mem_ctls_mutex
);
297 /* cancel the current workq request */
298 edac_mc_workq_teardown(mci
);
300 /* restart the workq request, with new delay value */
301 edac_mc_workq_setup(mci
, value
);
303 mutex_unlock(&mem_ctls_mutex
);
306 /* Return 0 on success, 1 on failure.
307 * Before calling this function, caller must
308 * assign a unique value to mci->mc_idx.
310 static int add_mc_to_global_list(struct mem_ctl_info
*mci
)
312 struct list_head
*item
, *insert_before
;
313 struct mem_ctl_info
*p
;
315 insert_before
= &mc_devices
;
317 if (unlikely((p
= find_mci_by_dev(mci
->dev
)) != NULL
))
320 list_for_each(item
, &mc_devices
) {
321 p
= list_entry(item
, struct mem_ctl_info
, link
);
323 if (p
->mc_idx
>= mci
->mc_idx
) {
324 if (unlikely(p
->mc_idx
== mci
->mc_idx
))
327 insert_before
= item
;
332 list_add_tail_rcu(&mci
->link
, insert_before
);
333 atomic_inc(&edac_handlers
);
337 edac_printk(KERN_WARNING
, EDAC_MC
,
338 "%s (%s) %s %s already assigned %d\n", p
->dev
->bus_id
,
339 dev_name(mci
), p
->mod_name
, p
->ctl_name
, p
->mc_idx
);
343 edac_printk(KERN_WARNING
, EDAC_MC
,
344 "bug in low-level driver: attempt to assign\n"
345 " duplicate mc_idx %d in %s()\n", p
->mc_idx
, __func__
);
349 static void complete_mc_list_del(struct rcu_head
*head
)
351 struct mem_ctl_info
*mci
;
353 mci
= container_of(head
, struct mem_ctl_info
, rcu
);
354 INIT_LIST_HEAD(&mci
->link
);
355 complete(&mci
->complete
);
358 static void del_mc_from_global_list(struct mem_ctl_info
*mci
)
360 atomic_dec(&edac_handlers
);
361 list_del_rcu(&mci
->link
);
362 init_completion(&mci
->complete
);
363 call_rcu(&mci
->rcu
, complete_mc_list_del
);
364 wait_for_completion(&mci
->complete
);
368 * edac_mc_find: Search for a mem_ctl_info structure whose index is 'idx'.
370 * If found, return a pointer to the structure.
373 * Caller must hold mem_ctls_mutex.
375 struct mem_ctl_info
*edac_mc_find(int idx
)
377 struct list_head
*item
;
378 struct mem_ctl_info
*mci
;
380 list_for_each(item
, &mc_devices
) {
381 mci
= list_entry(item
, struct mem_ctl_info
, link
);
383 if (mci
->mc_idx
>= idx
) {
384 if (mci
->mc_idx
== idx
)
394 EXPORT_SYMBOL(edac_mc_find
);
397 * edac_mc_add_mc: Insert the 'mci' structure into the mci global list and
398 * create sysfs entries associated with mci structure
399 * @mci: pointer to the mci structure to be added to the list
400 * @mc_idx: A unique numeric identifier to be assigned to the 'mci' structure.
407 /* FIXME - should a warning be printed if no error detection? correction? */
408 int edac_mc_add_mc(struct mem_ctl_info
*mci
, int mc_idx
)
410 debugf0("%s()\n", __func__
);
411 mci
->mc_idx
= mc_idx
;
412 #ifdef CONFIG_EDAC_DEBUG
413 if (edac_debug_level
>= 3)
414 edac_mc_dump_mci(mci
);
416 if (edac_debug_level
>= 4) {
419 for (i
= 0; i
< mci
->nr_csrows
; i
++) {
422 edac_mc_dump_csrow(&mci
->csrows
[i
]);
423 for (j
= 0; j
< mci
->csrows
[i
].nr_channels
; j
++)
424 edac_mc_dump_channel(&mci
->csrows
[i
].
429 mutex_lock(&mem_ctls_mutex
);
431 if (add_mc_to_global_list(mci
))
434 /* set load time so that error rate can be tracked */
435 mci
->start_time
= jiffies
;
437 if (edac_create_sysfs_mci_device(mci
)) {
438 edac_mc_printk(mci
, KERN_WARNING
,
439 "failed to create sysfs device\n");
443 /* If there IS a check routine, then we are running POLLED */
444 if (mci
->edac_check
!= NULL
) {
445 /* This instance is NOW RUNNING */
446 mci
->op_state
= OP_RUNNING_POLL
;
448 edac_mc_workq_setup(mci
, edac_mc_get_poll_msec());
450 mci
->op_state
= OP_RUNNING_INTERRUPT
;
453 /* Report action taken */
454 edac_mc_printk(mci
, KERN_INFO
, "Giving out device to %s %s: DEV %s\n",
455 mci
->mod_name
, mci
->ctl_name
, dev_name(mci
));
457 mutex_unlock(&mem_ctls_mutex
);
461 del_mc_from_global_list(mci
);
464 mutex_unlock(&mem_ctls_mutex
);
468 EXPORT_SYMBOL_GPL(edac_mc_add_mc
);
471 * edac_mc_del_mc: Remove sysfs entries for specified mci structure and
472 * remove mci structure from global list
473 * @pdev: Pointer to 'struct device' representing mci structure to remove.
475 * Return pointer to removed mci structure, or NULL if device not found.
477 struct mem_ctl_info
*edac_mc_del_mc(struct device
*dev
)
479 struct mem_ctl_info
*mci
;
481 debugf0("MC: %s()\n", __func__
);
482 mutex_lock(&mem_ctls_mutex
);
484 if ((mci
= find_mci_by_dev(dev
)) == NULL
) {
485 mutex_unlock(&mem_ctls_mutex
);
489 /* marking MCI offline */
490 mci
->op_state
= OP_OFFLINE
;
492 /* flush workq processes */
493 edac_mc_workq_teardown(mci
);
495 edac_remove_sysfs_mci_device(mci
);
496 del_mc_from_global_list(mci
);
497 mutex_unlock(&mem_ctls_mutex
);
498 edac_printk(KERN_INFO
, EDAC_MC
,
499 "Removed device %d for %s %s: DEV %s\n", mci
->mc_idx
,
500 mci
->mod_name
, mci
->ctl_name
, dev_name(mci
));
504 EXPORT_SYMBOL_GPL(edac_mc_del_mc
);
506 static void edac_mc_scrub_block(unsigned long page
, unsigned long offset
,
511 unsigned long flags
= 0;
513 debugf3("%s()\n", __func__
);
515 /* ECC error page was not in our memory. Ignore it. */
516 if (!pfn_valid(page
))
519 /* Find the actual page structure then map it and fix */
520 pg
= pfn_to_page(page
);
523 local_irq_save(flags
);
525 virt_addr
= kmap_atomic(pg
, KM_BOUNCE_READ
);
527 /* Perform architecture specific atomic scrub operation */
528 atomic_scrub(virt_addr
+ offset
, size
);
530 /* Unmap and complete */
531 kunmap_atomic(virt_addr
, KM_BOUNCE_READ
);
534 local_irq_restore(flags
);
537 /* FIXME - should return -1 */
538 int edac_mc_find_csrow_by_page(struct mem_ctl_info
*mci
, unsigned long page
)
540 struct csrow_info
*csrows
= mci
->csrows
;
543 debugf1("MC%d: %s(): 0x%lx\n", mci
->mc_idx
, __func__
, page
);
546 for (i
= 0; i
< mci
->nr_csrows
; i
++) {
547 struct csrow_info
*csrow
= &csrows
[i
];
549 if (csrow
->nr_pages
== 0)
552 debugf3("MC%d: %s(): first(0x%lx) page(0x%lx) last(0x%lx) "
553 "mask(0x%lx)\n", mci
->mc_idx
, __func__
,
554 csrow
->first_page
, page
, csrow
->last_page
,
557 if ((page
>= csrow
->first_page
) &&
558 (page
<= csrow
->last_page
) &&
559 ((page
& csrow
->page_mask
) ==
560 (csrow
->first_page
& csrow
->page_mask
))) {
567 edac_mc_printk(mci
, KERN_ERR
,
568 "could not look up page error address %lx\n",
569 (unsigned long)page
);
574 EXPORT_SYMBOL_GPL(edac_mc_find_csrow_by_page
);
576 /* FIXME - setable log (warning/emerg) levels */
577 /* FIXME - integrate with evlog: http://evlog.sourceforge.net/ */
578 void edac_mc_handle_ce(struct mem_ctl_info
*mci
,
579 unsigned long page_frame_number
,
580 unsigned long offset_in_page
, unsigned long syndrome
,
581 int row
, int channel
, const char *msg
)
583 unsigned long remapped_page
;
585 debugf3("MC%d: %s()\n", mci
->mc_idx
, __func__
);
587 /* FIXME - maybe make panic on INTERNAL ERROR an option */
588 if (row
>= mci
->nr_csrows
|| row
< 0) {
589 /* something is wrong */
590 edac_mc_printk(mci
, KERN_ERR
,
591 "INTERNAL ERROR: row out of range "
592 "(%d >= %d)\n", row
, mci
->nr_csrows
);
593 edac_mc_handle_ce_no_info(mci
, "INTERNAL ERROR");
597 if (channel
>= mci
->csrows
[row
].nr_channels
|| channel
< 0) {
598 /* something is wrong */
599 edac_mc_printk(mci
, KERN_ERR
,
600 "INTERNAL ERROR: channel out of range "
601 "(%d >= %d)\n", channel
,
602 mci
->csrows
[row
].nr_channels
);
603 edac_mc_handle_ce_no_info(mci
, "INTERNAL ERROR");
607 if (edac_mc_get_log_ce())
608 /* FIXME - put in DIMM location */
609 edac_mc_printk(mci
, KERN_WARNING
,
610 "CE page 0x%lx, offset 0x%lx, grain %d, syndrome "
611 "0x%lx, row %d, channel %d, label \"%s\": %s\n",
612 page_frame_number
, offset_in_page
,
613 mci
->csrows
[row
].grain
, syndrome
, row
, channel
,
614 mci
->csrows
[row
].channels
[channel
].label
, msg
);
617 mci
->csrows
[row
].ce_count
++;
618 mci
->csrows
[row
].channels
[channel
].ce_count
++;
620 if (mci
->scrub_mode
& SCRUB_SW_SRC
) {
622 * Some MC's can remap memory so that it is still available
623 * at a different address when PCI devices map into memory.
624 * MC's that can't do this lose the memory where PCI devices
625 * are mapped. This mapping is MC dependant and so we call
626 * back into the MC driver for it to map the MC page to
627 * a physical (CPU) page which can then be mapped to a virtual
628 * page - which can then be scrubbed.
630 remapped_page
= mci
->ctl_page_to_phys
?
631 mci
->ctl_page_to_phys(mci
, page_frame_number
) :
634 edac_mc_scrub_block(remapped_page
, offset_in_page
,
635 mci
->csrows
[row
].grain
);
639 EXPORT_SYMBOL_GPL(edac_mc_handle_ce
);
641 void edac_mc_handle_ce_no_info(struct mem_ctl_info
*mci
, const char *msg
)
643 if (edac_mc_get_log_ce())
644 edac_mc_printk(mci
, KERN_WARNING
,
645 "CE - no information available: %s\n", msg
);
647 mci
->ce_noinfo_count
++;
651 EXPORT_SYMBOL_GPL(edac_mc_handle_ce_no_info
);
653 void edac_mc_handle_ue(struct mem_ctl_info
*mci
,
654 unsigned long page_frame_number
,
655 unsigned long offset_in_page
, int row
, const char *msg
)
657 int len
= EDAC_MC_LABEL_LEN
* 4;
658 char labels
[len
+ 1];
663 debugf3("MC%d: %s()\n", mci
->mc_idx
, __func__
);
665 /* FIXME - maybe make panic on INTERNAL ERROR an option */
666 if (row
>= mci
->nr_csrows
|| row
< 0) {
667 /* something is wrong */
668 edac_mc_printk(mci
, KERN_ERR
,
669 "INTERNAL ERROR: row out of range "
670 "(%d >= %d)\n", row
, mci
->nr_csrows
);
671 edac_mc_handle_ue_no_info(mci
, "INTERNAL ERROR");
675 chars
= snprintf(pos
, len
+ 1, "%s",
676 mci
->csrows
[row
].channels
[0].label
);
680 for (chan
= 1; (chan
< mci
->csrows
[row
].nr_channels
) && (len
> 0);
682 chars
= snprintf(pos
, len
+ 1, ":%s",
683 mci
->csrows
[row
].channels
[chan
].label
);
688 if (edac_mc_get_log_ue())
689 edac_mc_printk(mci
, KERN_EMERG
,
690 "UE page 0x%lx, offset 0x%lx, grain %d, row %d, "
691 "labels \"%s\": %s\n", page_frame_number
,
692 offset_in_page
, mci
->csrows
[row
].grain
, row
,
695 if (edac_mc_get_panic_on_ue())
696 panic("EDAC MC%d: UE page 0x%lx, offset 0x%lx, grain %d, "
697 "row %d, labels \"%s\": %s\n", mci
->mc_idx
,
698 page_frame_number
, offset_in_page
,
699 mci
->csrows
[row
].grain
, row
, labels
, msg
);
702 mci
->csrows
[row
].ue_count
++;
705 EXPORT_SYMBOL_GPL(edac_mc_handle_ue
);
707 void edac_mc_handle_ue_no_info(struct mem_ctl_info
*mci
, const char *msg
)
709 if (edac_mc_get_panic_on_ue())
710 panic("EDAC MC%d: Uncorrected Error", mci
->mc_idx
);
712 if (edac_mc_get_log_ue())
713 edac_mc_printk(mci
, KERN_WARNING
,
714 "UE - no information available: %s\n", msg
);
715 mci
->ue_noinfo_count
++;
719 EXPORT_SYMBOL_GPL(edac_mc_handle_ue_no_info
);
721 /*************************************************************
722 * On Fully Buffered DIMM modules, this help function is
723 * called to process UE events
725 void edac_mc_handle_fbd_ue(struct mem_ctl_info
*mci
,
727 unsigned int channela
,
728 unsigned int channelb
, char *msg
)
730 int len
= EDAC_MC_LABEL_LEN
* 4;
731 char labels
[len
+ 1];
735 if (csrow
>= mci
->nr_csrows
) {
736 /* something is wrong */
737 edac_mc_printk(mci
, KERN_ERR
,
738 "INTERNAL ERROR: row out of range (%d >= %d)\n",
739 csrow
, mci
->nr_csrows
);
740 edac_mc_handle_ue_no_info(mci
, "INTERNAL ERROR");
744 if (channela
>= mci
->csrows
[csrow
].nr_channels
) {
745 /* something is wrong */
746 edac_mc_printk(mci
, KERN_ERR
,
747 "INTERNAL ERROR: channel-a out of range "
749 channela
, mci
->csrows
[csrow
].nr_channels
);
750 edac_mc_handle_ue_no_info(mci
, "INTERNAL ERROR");
754 if (channelb
>= mci
->csrows
[csrow
].nr_channels
) {
755 /* something is wrong */
756 edac_mc_printk(mci
, KERN_ERR
,
757 "INTERNAL ERROR: channel-b out of range "
759 channelb
, mci
->csrows
[csrow
].nr_channels
);
760 edac_mc_handle_ue_no_info(mci
, "INTERNAL ERROR");
765 mci
->csrows
[csrow
].ue_count
++;
767 /* Generate the DIMM labels from the specified channels */
768 chars
= snprintf(pos
, len
+ 1, "%s",
769 mci
->csrows
[csrow
].channels
[channela
].label
);
772 chars
= snprintf(pos
, len
+ 1, "-%s",
773 mci
->csrows
[csrow
].channels
[channelb
].label
);
775 if (edac_mc_get_log_ue())
776 edac_mc_printk(mci
, KERN_EMERG
,
777 "UE row %d, channel-a= %d channel-b= %d "
778 "labels \"%s\": %s\n", csrow
, channela
, channelb
,
781 if (edac_mc_get_panic_on_ue())
782 panic("UE row %d, channel-a= %d channel-b= %d "
783 "labels \"%s\": %s\n", csrow
, channela
,
784 channelb
, labels
, msg
);
787 EXPORT_SYMBOL(edac_mc_handle_fbd_ue
);
789 /*************************************************************
790 * On Fully Buffered DIMM modules, this help function is
791 * called to process CE events
793 void edac_mc_handle_fbd_ce(struct mem_ctl_info
*mci
,
794 unsigned int csrow
, unsigned int channel
, char *msg
)
797 /* Ensure boundary values */
798 if (csrow
>= mci
->nr_csrows
) {
799 /* something is wrong */
800 edac_mc_printk(mci
, KERN_ERR
,
801 "INTERNAL ERROR: row out of range (%d >= %d)\n",
802 csrow
, mci
->nr_csrows
);
803 edac_mc_handle_ce_no_info(mci
, "INTERNAL ERROR");
806 if (channel
>= mci
->csrows
[csrow
].nr_channels
) {
807 /* something is wrong */
808 edac_mc_printk(mci
, KERN_ERR
,
809 "INTERNAL ERROR: channel out of range (%d >= %d)\n",
810 channel
, mci
->csrows
[csrow
].nr_channels
);
811 edac_mc_handle_ce_no_info(mci
, "INTERNAL ERROR");
815 if (edac_mc_get_log_ce())
816 /* FIXME - put in DIMM location */
817 edac_mc_printk(mci
, KERN_WARNING
,
818 "CE row %d, channel %d, label \"%s\": %s\n",
820 mci
->csrows
[csrow
].channels
[channel
].label
, msg
);
823 mci
->csrows
[csrow
].ce_count
++;
824 mci
->csrows
[csrow
].channels
[channel
].ce_count
++;
827 EXPORT_SYMBOL(edac_mc_handle_fbd_ce
);
830 * Iterate over all MC instances and check for ECC, et al, errors
832 void edac_check_mc_devices(void)
834 struct list_head
*item
;
835 struct mem_ctl_info
*mci
;
837 debugf3("%s()\n", __func__
);
838 mutex_lock(&mem_ctls_mutex
);
840 list_for_each(item
, &mc_devices
) {
841 mci
= list_entry(item
, struct mem_ctl_info
, link
);
843 if (mci
->edac_check
!= NULL
)
844 mci
->edac_check(mci
);
847 mutex_unlock(&mem_ctls_mutex
);