blob: ff8c0020649c0690deae19462c3d0a6e9f59d8fd [file] [log] [blame]
Alan Coxda9bb1d2006-01-18 17:44:13 -08001/*
2 * edac_mc kernel module
Doug Thompson49c0dab72006-07-10 04:45:19 -07003 * (C) 2005, 2006 Linux Networx (http://lnxi.com)
Alan Coxda9bb1d2006-01-18 17:44:13 -08004 * This file may be distributed under the terms of the
5 * GNU General Public License.
6 *
7 * Written by Thayne Harbaugh
8 * Based on work by Dan Hollis <goemon at anime dot net> and others.
9 * http://www.anime.net/~goemon/linux-ecc/
10 *
11 * Modified by Dave Peterson and Doug Thompson
12 *
13 */
14
Alan Coxda9bb1d2006-01-18 17:44:13 -080015#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>
Alan Coxda9bb1d2006-01-18 17:44:13 -080028#include <linux/ctype.h>
Dave Jiangc0d12172007-07-19 01:49:46 -070029#include <linux/edac.h>
Alan Coxda9bb1d2006-01-18 17:44:13 -080030#include <asm/uaccess.h>
31#include <asm/page.h>
32#include <asm/edac.h>
Douglas Thompson20bcb7a2007-07-19 01:49:47 -070033#include "edac_core.h"
Douglas Thompson7c9281d2007-07-19 01:49:33 -070034#include "edac_module.h"
Alan Coxda9bb1d2006-01-18 17:44:13 -080035
Alan Coxda9bb1d2006-01-18 17:44:13 -080036/* lock to memory controller's control array */
Matthias Kaehlcke63b7df92007-07-19 01:49:38 -070037static DEFINE_MUTEX(mem_ctls_mutex);
Robert P. J. Dayff6ac2a2008-04-29 01:03:17 -070038static LIST_HEAD(mc_devices);
Alan Coxda9bb1d2006-01-18 17:44:13 -080039
Alan Coxda9bb1d2006-01-18 17:44:13 -080040#ifdef CONFIG_EDAC_DEBUG
41
Mauro Carvalho Chehaba4b4be32012-01-27 10:26:13 -030042static void edac_mc_dump_channel(struct rank_info *chan)
Alan Coxda9bb1d2006-01-18 17:44:13 -080043{
44 debugf4("\tchannel = %p\n", chan);
45 debugf4("\tchannel->chan_idx = %d\n", chan->chan_idx);
Alan Coxda9bb1d2006-01-18 17:44:13 -080046 debugf4("\tchannel->csrow = %p\n\n", chan->csrow);
Mauro Carvalho Chehaba895bf82012-01-28 09:09:38 -030047 debugf4("\tdimm->ce_count = %d\n", chan->dimm->ce_count);
48 debugf4("\tdimm->label = '%s'\n", chan->dimm->label);
49 debugf4("\tdimm->nr_pages = 0x%x\n", chan->dimm->nr_pages);
Alan Coxda9bb1d2006-01-18 17:44:13 -080050}
51
Adrian Bunk2da1c112007-07-19 01:49:32 -070052static void edac_mc_dump_csrow(struct csrow_info *csrow)
Alan Coxda9bb1d2006-01-18 17:44:13 -080053{
54 debugf4("\tcsrow = %p\n", csrow);
55 debugf4("\tcsrow->csrow_idx = %d\n", csrow->csrow_idx);
Douglas Thompson079708b2007-07-19 01:49:58 -070056 debugf4("\tcsrow->first_page = 0x%lx\n", csrow->first_page);
Alan Coxda9bb1d2006-01-18 17:44:13 -080057 debugf4("\tcsrow->last_page = 0x%lx\n", csrow->last_page);
58 debugf4("\tcsrow->page_mask = 0x%lx\n", csrow->page_mask);
Douglas Thompson079708b2007-07-19 01:49:58 -070059 debugf4("\tcsrow->nr_channels = %d\n", csrow->nr_channels);
Alan Coxda9bb1d2006-01-18 17:44:13 -080060 debugf4("\tcsrow->channels = %p\n", csrow->channels);
61 debugf4("\tcsrow->mci = %p\n\n", csrow->mci);
62}
63
Adrian Bunk2da1c112007-07-19 01:49:32 -070064static void edac_mc_dump_mci(struct mem_ctl_info *mci)
Alan Coxda9bb1d2006-01-18 17:44:13 -080065{
66 debugf3("\tmci = %p\n", mci);
67 debugf3("\tmci->mtype_cap = %lx\n", mci->mtype_cap);
68 debugf3("\tmci->edac_ctl_cap = %lx\n", mci->edac_ctl_cap);
69 debugf3("\tmci->edac_cap = %lx\n", mci->edac_cap);
70 debugf4("\tmci->edac_check = %p\n", mci->edac_check);
71 debugf3("\tmci->nr_csrows = %d, csrows = %p\n",
72 mci->nr_csrows, mci->csrows);
Doug Thompson37f04582006-06-30 01:56:07 -070073 debugf3("\tdev = %p\n", mci->dev);
Douglas Thompson079708b2007-07-19 01:49:58 -070074 debugf3("\tmod_name:ctl_name = %s:%s\n", mci->mod_name, mci->ctl_name);
Alan Coxda9bb1d2006-01-18 17:44:13 -080075 debugf3("\tpvt_info = %p\n\n", mci->pvt_info);
76}
77
Borislav Petkov24f9a7f2010-10-07 18:29:15 +020078#endif /* CONFIG_EDAC_DEBUG */
79
Borislav Petkov239642f2009-11-12 15:33:16 +010080/*
81 * keep those in sync with the enum mem_type
82 */
83const char *edac_mem_types[] = {
84 "Empty csrow",
85 "Reserved csrow type",
86 "Unknown csrow type",
87 "Fast page mode RAM",
88 "Extended data out RAM",
89 "Burst Extended data out RAM",
90 "Single data rate SDRAM",
91 "Registered single data rate SDRAM",
92 "Double data rate SDRAM",
93 "Registered Double data rate SDRAM",
94 "Rambus DRAM",
95 "Unbuffered DDR2 RAM",
96 "Fully buffered DDR2",
97 "Registered DDR2 RAM",
98 "Rambus XDR",
99 "Unbuffered DDR3 RAM",
100 "Registered DDR3 RAM",
101};
102EXPORT_SYMBOL_GPL(edac_mem_types);
103
Mauro Carvalho Chehab93e4fe62012-04-16 10:18:12 -0300104/**
105 * edac_align_ptr - Prepares the pointer offsets for a single-shot allocation
106 * @p: pointer to a pointer with the memory offset to be used. At
107 * return, this will be incremented to point to the next offset
108 * @size: Size of the data structure to be reserved
109 * @n_elems: Number of elements that should be reserved
Alan Coxda9bb1d2006-01-18 17:44:13 -0800110 *
111 * If 'size' is a constant, the compiler will optimize this whole function
Mauro Carvalho Chehab93e4fe62012-04-16 10:18:12 -0300112 * down to either a no-op or the addition of a constant to the value of '*p'.
113 *
114 * The 'p' pointer is absolutely needed to keep the proper advancing
115 * further in memory to the proper offsets when allocating the struct along
116 * with its embedded structs, as edac_device_alloc_ctl_info() does it
117 * above, for example.
118 *
119 * At return, the pointer 'p' will be incremented to be used on a next call
120 * to this function.
Alan Coxda9bb1d2006-01-18 17:44:13 -0800121 */
Mauro Carvalho Chehab93e4fe62012-04-16 10:18:12 -0300122void *edac_align_ptr(void **p, unsigned size, int n_elems)
Alan Coxda9bb1d2006-01-18 17:44:13 -0800123{
124 unsigned align, r;
Mauro Carvalho Chehab93e4fe62012-04-16 10:18:12 -0300125 void *ptr = *p;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800126
Mauro Carvalho Chehab93e4fe62012-04-16 10:18:12 -0300127 *p += size * n_elems;
128
129 /*
130 * 'p' can possibly be an unaligned item X such that sizeof(X) is
131 * 'size'. Adjust 'p' so that its alignment is at least as
132 * stringent as what the compiler would provide for X and return
133 * the aligned result.
134 * Here we assume that the alignment of a "long long" is the most
Alan Coxda9bb1d2006-01-18 17:44:13 -0800135 * stringent alignment that the compiler will ever provide by default.
136 * As far as I know, this is a reasonable assumption.
137 */
138 if (size > sizeof(long))
139 align = sizeof(long long);
140 else if (size > sizeof(int))
141 align = sizeof(long);
142 else if (size > sizeof(short))
143 align = sizeof(int);
144 else if (size > sizeof(char))
145 align = sizeof(short);
146 else
Douglas Thompson079708b2007-07-19 01:49:58 -0700147 return (char *)ptr;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800148
149 r = size % align;
150
151 if (r == 0)
Douglas Thompson079708b2007-07-19 01:49:58 -0700152 return (char *)ptr;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800153
Mauro Carvalho Chehab93e4fe62012-04-16 10:18:12 -0300154 *p += align - r;
155
Douglas Thompson7391c6d2007-07-19 01:50:21 -0700156 return (void *)(((unsigned long)ptr) + align - r);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800157}
158
Alan Coxda9bb1d2006-01-18 17:44:13 -0800159/**
160 * edac_mc_alloc: Allocate a struct mem_ctl_info structure
161 * @size_pvt: size of private storage needed
162 * @nr_csrows: Number of CWROWS needed for this MC
163 * @nr_chans: Number of channels for the MC
164 *
165 * Everything is kmalloc'ed as one big chunk - more efficient.
166 * Only can be used if all structures have the same lifetime - otherwise
167 * you have to allocate and initialize your own structures.
168 *
169 * Use edac_mc_free() to free mc structures allocated by this function.
170 *
171 * Returns:
172 * NULL allocation failed
173 * struct mem_ctl_info pointer
174 */
175struct mem_ctl_info *edac_mc_alloc(unsigned sz_pvt, unsigned nr_csrows,
Doug Thompsonb8f6f972007-07-19 01:50:26 -0700176 unsigned nr_chans, int edac_index)
Alan Coxda9bb1d2006-01-18 17:44:13 -0800177{
Mauro Carvalho Chehab93e4fe62012-04-16 10:18:12 -0300178 void *ptr = NULL;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800179 struct mem_ctl_info *mci;
180 struct csrow_info *csi, *csrow;
Mauro Carvalho Chehaba4b4be32012-01-27 10:26:13 -0300181 struct rank_info *chi, *chp, *chan;
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300182 struct dimm_info *dimm;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800183 void *pvt;
184 unsigned size;
185 int row, chn;
Doug Thompson8096cfa2007-07-19 01:50:27 -0700186 int err;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800187
188 /* Figure out the offsets of the various items from the start of an mc
189 * structure. We want the alignment of each item to be at least as
190 * stringent as what the compiler would provide if we could simply
191 * hardcode everything into a single struct.
192 */
Mauro Carvalho Chehab93e4fe62012-04-16 10:18:12 -0300193 mci = edac_align_ptr(&ptr, sizeof(*mci), 1);
194 csi = edac_align_ptr(&ptr, sizeof(*csi), nr_csrows);
195 chi = edac_align_ptr(&ptr, sizeof(*chi), nr_csrows * nr_chans);
196 dimm = edac_align_ptr(&ptr, sizeof(*dimm), nr_csrows * nr_chans);
197 pvt = edac_align_ptr(&ptr, sz_pvt, 1);
Douglas Thompson079708b2007-07-19 01:49:58 -0700198 size = ((unsigned long)pvt) + sz_pvt;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800199
Doug Thompson8096cfa2007-07-19 01:50:27 -0700200 mci = kzalloc(size, GFP_KERNEL);
201 if (mci == NULL)
Alan Coxda9bb1d2006-01-18 17:44:13 -0800202 return NULL;
203
204 /* Adjust pointers so they point within the memory we just allocated
205 * rather than an imaginary chunk of memory located at address 0.
206 */
Douglas Thompson079708b2007-07-19 01:49:58 -0700207 csi = (struct csrow_info *)(((char *)mci) + ((unsigned long)csi));
Mauro Carvalho Chehaba4b4be32012-01-27 10:26:13 -0300208 chi = (struct rank_info *)(((char *)mci) + ((unsigned long)chi));
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300209 dimm = (struct dimm_info *)(((char *)mci) + ((unsigned long)dimm));
Douglas Thompson079708b2007-07-19 01:49:58 -0700210 pvt = sz_pvt ? (((char *)mci) + ((unsigned long)pvt)) : NULL;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800211
Doug Thompsonb8f6f972007-07-19 01:50:26 -0700212 /* setup index and various internal pointers */
213 mci->mc_idx = edac_index;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800214 mci->csrows = csi;
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300215 mci->dimms = dimm;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800216 mci->pvt_info = pvt;
217 mci->nr_csrows = nr_csrows;
218
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300219 /*
220 * For now, assumes that a per-csrow arrangement for dimms.
221 * This will be latter changed.
222 */
223 dimm = mci->dimms;
224
Alan Coxda9bb1d2006-01-18 17:44:13 -0800225 for (row = 0; row < nr_csrows; row++) {
226 csrow = &csi[row];
227 csrow->csrow_idx = row;
228 csrow->mci = mci;
229 csrow->nr_channels = nr_chans;
230 chp = &chi[row * nr_chans];
231 csrow->channels = chp;
232
233 for (chn = 0; chn < nr_chans; chn++) {
234 chan = &chp[chn];
235 chan->chan_idx = chn;
236 chan->csrow = csrow;
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300237
238 mci->csrows[row].channels[chn].dimm = dimm;
239 dimm->csrow = row;
240 dimm->csrow_channel = chn;
241 dimm++;
242 mci->nr_dimms++;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800243 }
244 }
245
Dave Jiang81d87cb2007-07-19 01:49:52 -0700246 mci->op_state = OP_ALLOC;
Mauro Carvalho Chehab6fe11082010-08-12 00:30:25 -0300247 INIT_LIST_HEAD(&mci->grp_kobj_list);
Dave Jiang81d87cb2007-07-19 01:49:52 -0700248
Doug Thompson8096cfa2007-07-19 01:50:27 -0700249 /*
250 * Initialize the 'root' kobj for the edac_mc controller
251 */
252 err = edac_mc_register_sysfs_main_kobj(mci);
253 if (err) {
254 kfree(mci);
255 return NULL;
256 }
257
258 /* at this point, the root kobj is valid, and in order to
259 * 'free' the object, then the function:
260 * edac_mc_unregister_sysfs_main_kobj() must be called
261 * which will perform kobj unregistration and the actual free
262 * will occur during the kobject callback operation
263 */
Alan Coxda9bb1d2006-01-18 17:44:13 -0800264 return mci;
265}
Dave Peterson91105402006-03-26 01:38:55 -0800266EXPORT_SYMBOL_GPL(edac_mc_alloc);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800267
Alan Coxda9bb1d2006-01-18 17:44:13 -0800268/**
Doug Thompson8096cfa2007-07-19 01:50:27 -0700269 * edac_mc_free
270 * 'Free' a previously allocated 'mci' structure
Alan Coxda9bb1d2006-01-18 17:44:13 -0800271 * @mci: pointer to a struct mem_ctl_info structure
Alan Coxda9bb1d2006-01-18 17:44:13 -0800272 */
273void edac_mc_free(struct mem_ctl_info *mci)
274{
Mauro Carvalho Chehabbbc560a2010-08-16 18:22:43 -0300275 debugf1("%s()\n", __func__);
276
Doug Thompson8096cfa2007-07-19 01:50:27 -0700277 edac_mc_unregister_sysfs_main_kobj(mci);
Mauro Carvalho Chehabaccf74f2010-08-16 18:34:37 -0300278
279 /* free the mci instance memory here */
280 kfree(mci);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800281}
Dave Peterson91105402006-03-26 01:38:55 -0800282EXPORT_SYMBOL_GPL(edac_mc_free);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800283
Doug Thompsonbce19682007-07-26 10:41:14 -0700284
Mauro Carvalho Chehab939747bd2010-08-10 11:22:01 -0300285/**
Doug Thompsonbce19682007-07-26 10:41:14 -0700286 * find_mci_by_dev
287 *
288 * scan list of controllers looking for the one that manages
289 * the 'dev' device
Mauro Carvalho Chehab939747bd2010-08-10 11:22:01 -0300290 * @dev: pointer to a struct device related with the MCI
Doug Thompsonbce19682007-07-26 10:41:14 -0700291 */
Mauro Carvalho Chehab939747bd2010-08-10 11:22:01 -0300292struct mem_ctl_info *find_mci_by_dev(struct device *dev)
Alan Coxda9bb1d2006-01-18 17:44:13 -0800293{
294 struct mem_ctl_info *mci;
295 struct list_head *item;
296
Dave Peterson537fba22006-03-26 01:38:40 -0800297 debugf3("%s()\n", __func__);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800298
299 list_for_each(item, &mc_devices) {
300 mci = list_entry(item, struct mem_ctl_info, link);
301
Doug Thompson37f04582006-06-30 01:56:07 -0700302 if (mci->dev == dev)
Alan Coxda9bb1d2006-01-18 17:44:13 -0800303 return mci;
304 }
305
306 return NULL;
307}
Mauro Carvalho Chehab939747bd2010-08-10 11:22:01 -0300308EXPORT_SYMBOL_GPL(find_mci_by_dev);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800309
Dave Jiang81d87cb2007-07-19 01:49:52 -0700310/*
311 * handler for EDAC to check if NMI type handler has asserted interrupt
312 */
313static int edac_mc_assert_error_check_and_clear(void)
314{
Dave Jiang66ee2f92007-07-19 01:49:54 -0700315 int old_state;
Dave Jiang81d87cb2007-07-19 01:49:52 -0700316
Douglas Thompson079708b2007-07-19 01:49:58 -0700317 if (edac_op_state == EDAC_OPSTATE_POLL)
Dave Jiang81d87cb2007-07-19 01:49:52 -0700318 return 1;
319
Dave Jiang66ee2f92007-07-19 01:49:54 -0700320 old_state = edac_err_assert;
321 edac_err_assert = 0;
Dave Jiang81d87cb2007-07-19 01:49:52 -0700322
Dave Jiang66ee2f92007-07-19 01:49:54 -0700323 return old_state;
Dave Jiang81d87cb2007-07-19 01:49:52 -0700324}
325
326/*
327 * edac_mc_workq_function
328 * performs the operation scheduled by a workq request
329 */
Dave Jiang81d87cb2007-07-19 01:49:52 -0700330static void edac_mc_workq_function(struct work_struct *work_req)
331{
Jean Delvarefbeb4382009-04-13 14:40:21 -0700332 struct delayed_work *d_work = to_delayed_work(work_req);
Dave Jiang81d87cb2007-07-19 01:49:52 -0700333 struct mem_ctl_info *mci = to_edac_mem_ctl_work(d_work);
Dave Jiang81d87cb2007-07-19 01:49:52 -0700334
335 mutex_lock(&mem_ctls_mutex);
336
Doug Thompsonbf52fa42007-07-19 01:50:30 -0700337 /* if this control struct has movd to offline state, we are done */
338 if (mci->op_state == OP_OFFLINE) {
339 mutex_unlock(&mem_ctls_mutex);
340 return;
341 }
342
Dave Jiang81d87cb2007-07-19 01:49:52 -0700343 /* Only poll controllers that are running polled and have a check */
344 if (edac_mc_assert_error_check_and_clear() && (mci->edac_check != NULL))
345 mci->edac_check(mci);
346
Dave Jiang81d87cb2007-07-19 01:49:52 -0700347 mutex_unlock(&mem_ctls_mutex);
348
349 /* Reschedule */
Dave Jiang4de78c62007-07-19 01:49:54 -0700350 queue_delayed_work(edac_workqueue, &mci->work,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700351 msecs_to_jiffies(edac_mc_get_poll_msec()));
Dave Jiang81d87cb2007-07-19 01:49:52 -0700352}
353
354/*
355 * edac_mc_workq_setup
356 * initialize a workq item for this mci
357 * passing in the new delay period in msec
Doug Thompsonbf52fa42007-07-19 01:50:30 -0700358 *
359 * locking model:
360 *
361 * called with the mem_ctls_mutex held
Dave Jiang81d87cb2007-07-19 01:49:52 -0700362 */
Doug Thompsonbf52fa42007-07-19 01:50:30 -0700363static void edac_mc_workq_setup(struct mem_ctl_info *mci, unsigned msec)
Dave Jiang81d87cb2007-07-19 01:49:52 -0700364{
365 debugf0("%s()\n", __func__);
366
Doug Thompsonbf52fa42007-07-19 01:50:30 -0700367 /* if this instance is not in the POLL state, then simply return */
368 if (mci->op_state != OP_RUNNING_POLL)
369 return;
370
Dave Jiang81d87cb2007-07-19 01:49:52 -0700371 INIT_DELAYED_WORK(&mci->work, edac_mc_workq_function);
Dave Jiang81d87cb2007-07-19 01:49:52 -0700372 queue_delayed_work(edac_workqueue, &mci->work, msecs_to_jiffies(msec));
373}
374
375/*
376 * edac_mc_workq_teardown
377 * stop the workq processing on this mci
Doug Thompsonbf52fa42007-07-19 01:50:30 -0700378 *
379 * locking model:
380 *
381 * called WITHOUT lock held
Dave Jiang81d87cb2007-07-19 01:49:52 -0700382 */
Doug Thompsonbf52fa42007-07-19 01:50:30 -0700383static void edac_mc_workq_teardown(struct mem_ctl_info *mci)
Dave Jiang81d87cb2007-07-19 01:49:52 -0700384{
385 int status;
386
Borislav Petkov00740c52010-09-26 12:42:23 +0200387 if (mci->op_state != OP_RUNNING_POLL)
388 return;
389
Doug Thompsonbce19682007-07-26 10:41:14 -0700390 status = cancel_delayed_work(&mci->work);
391 if (status == 0) {
392 debugf0("%s() not canceled, flush the queue\n",
393 __func__);
Doug Thompsonbf52fa42007-07-19 01:50:30 -0700394
Doug Thompsonbce19682007-07-26 10:41:14 -0700395 /* workq instance might be running, wait for it */
396 flush_workqueue(edac_workqueue);
Dave Jiang81d87cb2007-07-19 01:49:52 -0700397 }
398}
399
400/*
Doug Thompsonbce19682007-07-26 10:41:14 -0700401 * edac_mc_reset_delay_period(unsigned long value)
402 *
403 * user space has updated our poll period value, need to
404 * reset our workq delays
Dave Jiang81d87cb2007-07-19 01:49:52 -0700405 */
Doug Thompsonbce19682007-07-26 10:41:14 -0700406void edac_mc_reset_delay_period(int value)
Dave Jiang81d87cb2007-07-19 01:49:52 -0700407{
Doug Thompsonbce19682007-07-26 10:41:14 -0700408 struct mem_ctl_info *mci;
409 struct list_head *item;
Dave Jiang81d87cb2007-07-19 01:49:52 -0700410
Doug Thompsonbf52fa42007-07-19 01:50:30 -0700411 mutex_lock(&mem_ctls_mutex);
412
Doug Thompsonbce19682007-07-26 10:41:14 -0700413 /* scan the list and turn off all workq timers, doing so under lock
414 */
415 list_for_each(item, &mc_devices) {
416 mci = list_entry(item, struct mem_ctl_info, link);
417
418 if (mci->op_state == OP_RUNNING_POLL)
419 cancel_delayed_work(&mci->work);
420 }
421
422 mutex_unlock(&mem_ctls_mutex);
423
424
425 /* re-walk the list, and reset the poll delay */
426 mutex_lock(&mem_ctls_mutex);
427
428 list_for_each(item, &mc_devices) {
429 mci = list_entry(item, struct mem_ctl_info, link);
430
431 edac_mc_workq_setup(mci, (unsigned long) value);
432 }
Dave Jiang81d87cb2007-07-19 01:49:52 -0700433
434 mutex_unlock(&mem_ctls_mutex);
435}
436
Doug Thompsonbce19682007-07-26 10:41:14 -0700437
438
Doug Thompson2d7bbb92006-06-30 01:56:08 -0700439/* Return 0 on success, 1 on failure.
440 * Before calling this function, caller must
441 * assign a unique value to mci->mc_idx.
Doug Thompsonbf52fa42007-07-19 01:50:30 -0700442 *
443 * locking model:
444 *
445 * called with the mem_ctls_mutex lock held
Doug Thompson2d7bbb92006-06-30 01:56:08 -0700446 */
Douglas Thompson079708b2007-07-19 01:49:58 -0700447static int add_mc_to_global_list(struct mem_ctl_info *mci)
Alan Coxda9bb1d2006-01-18 17:44:13 -0800448{
449 struct list_head *item, *insert_before;
450 struct mem_ctl_info *p;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800451
Doug Thompson2d7bbb92006-06-30 01:56:08 -0700452 insert_before = &mc_devices;
453
Doug Thompsonbf52fa42007-07-19 01:50:30 -0700454 p = find_mci_by_dev(mci->dev);
455 if (unlikely(p != NULL))
Doug Thompson2d7bbb92006-06-30 01:56:08 -0700456 goto fail0;
457
458 list_for_each(item, &mc_devices) {
459 p = list_entry(item, struct mem_ctl_info, link);
460
461 if (p->mc_idx >= mci->mc_idx) {
462 if (unlikely(p->mc_idx == mci->mc_idx))
463 goto fail1;
464
465 insert_before = item;
466 break;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800467 }
Alan Coxda9bb1d2006-01-18 17:44:13 -0800468 }
469
470 list_add_tail_rcu(&mci->link, insert_before);
Dave Jiangc0d12172007-07-19 01:49:46 -0700471 atomic_inc(&edac_handlers);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800472 return 0;
Doug Thompson2d7bbb92006-06-30 01:56:08 -0700473
Douglas Thompson052dfb42007-07-19 01:50:13 -0700474fail0:
Doug Thompson2d7bbb92006-06-30 01:56:08 -0700475 edac_printk(KERN_WARNING, EDAC_MC,
Kay Sievers281efb12009-01-06 14:42:57 -0800476 "%s (%s) %s %s already assigned %d\n", dev_name(p->dev),
Stephen Rothwell17aa7e02008-05-05 13:54:19 +1000477 edac_dev_name(mci), p->mod_name, p->ctl_name, p->mc_idx);
Doug Thompson2d7bbb92006-06-30 01:56:08 -0700478 return 1;
479
Douglas Thompson052dfb42007-07-19 01:50:13 -0700480fail1:
Doug Thompson2d7bbb92006-06-30 01:56:08 -0700481 edac_printk(KERN_WARNING, EDAC_MC,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700482 "bug in low-level driver: attempt to assign\n"
483 " duplicate mc_idx %d in %s()\n", p->mc_idx, __func__);
Doug Thompson2d7bbb92006-06-30 01:56:08 -0700484 return 1;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800485}
486
Dave Petersone7ecd892006-03-26 01:38:52 -0800487static void del_mc_from_global_list(struct mem_ctl_info *mci)
Dave Petersona1d03fc2006-03-26 01:38:46 -0800488{
Dave Jiangc0d12172007-07-19 01:49:46 -0700489 atomic_dec(&edac_handlers);
Dave Petersona1d03fc2006-03-26 01:38:46 -0800490 list_del_rcu(&mci->link);
Lai Jiangshane2e77092011-05-26 16:25:58 -0700491
492 /* these are for safe removal of devices from global list while
493 * NMI handlers may be traversing list
494 */
495 synchronize_rcu();
496 INIT_LIST_HEAD(&mci->link);
Dave Petersona1d03fc2006-03-26 01:38:46 -0800497}
498
Alan Coxda9bb1d2006-01-18 17:44:13 -0800499/**
Douglas Thompson5da08312007-07-19 01:49:31 -0700500 * edac_mc_find: Search for a mem_ctl_info structure whose index is 'idx'.
501 *
502 * If found, return a pointer to the structure.
503 * Else return NULL.
504 *
505 * Caller must hold mem_ctls_mutex.
506 */
Douglas Thompson079708b2007-07-19 01:49:58 -0700507struct mem_ctl_info *edac_mc_find(int idx)
Douglas Thompson5da08312007-07-19 01:49:31 -0700508{
509 struct list_head *item;
510 struct mem_ctl_info *mci;
511
512 list_for_each(item, &mc_devices) {
513 mci = list_entry(item, struct mem_ctl_info, link);
514
515 if (mci->mc_idx >= idx) {
516 if (mci->mc_idx == idx)
517 return mci;
518
519 break;
520 }
521 }
522
523 return NULL;
524}
525EXPORT_SYMBOL(edac_mc_find);
526
527/**
Dave Peterson472678e2006-03-26 01:38:49 -0800528 * edac_mc_add_mc: Insert the 'mci' structure into the mci global list and
529 * create sysfs entries associated with mci structure
Alan Coxda9bb1d2006-01-18 17:44:13 -0800530 * @mci: pointer to the mci structure to be added to the list
Doug Thompson2d7bbb92006-06-30 01:56:08 -0700531 * @mc_idx: A unique numeric identifier to be assigned to the 'mci' structure.
Alan Coxda9bb1d2006-01-18 17:44:13 -0800532 *
533 * Return:
534 * 0 Success
535 * !0 Failure
536 */
537
538/* FIXME - should a warning be printed if no error detection? correction? */
Doug Thompsonb8f6f972007-07-19 01:50:26 -0700539int edac_mc_add_mc(struct mem_ctl_info *mci)
Alan Coxda9bb1d2006-01-18 17:44:13 -0800540{
Dave Peterson537fba22006-03-26 01:38:40 -0800541 debugf0("%s()\n", __func__);
Doug Thompsonb8f6f972007-07-19 01:50:26 -0700542
Alan Coxda9bb1d2006-01-18 17:44:13 -0800543#ifdef CONFIG_EDAC_DEBUG
544 if (edac_debug_level >= 3)
545 edac_mc_dump_mci(mci);
Dave Petersone7ecd892006-03-26 01:38:52 -0800546
Alan Coxda9bb1d2006-01-18 17:44:13 -0800547 if (edac_debug_level >= 4) {
548 int i;
549
550 for (i = 0; i < mci->nr_csrows; i++) {
551 int j;
Dave Petersone7ecd892006-03-26 01:38:52 -0800552
Alan Coxda9bb1d2006-01-18 17:44:13 -0800553 edac_mc_dump_csrow(&mci->csrows[i]);
554 for (j = 0; j < mci->csrows[i].nr_channels; j++)
Douglas Thompson079708b2007-07-19 01:49:58 -0700555 edac_mc_dump_channel(&mci->csrows[i].
Douglas Thompson052dfb42007-07-19 01:50:13 -0700556 channels[j]);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800557 }
558 }
559#endif
Matthias Kaehlcke63b7df92007-07-19 01:49:38 -0700560 mutex_lock(&mem_ctls_mutex);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800561
562 if (add_mc_to_global_list(mci))
Dave Peterson028a7b62006-03-26 01:38:47 -0800563 goto fail0;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800564
565 /* set load time so that error rate can be tracked */
566 mci->start_time = jiffies;
567
eric wollesen9794f332007-02-12 00:53:08 -0800568 if (edac_create_sysfs_mci_device(mci)) {
569 edac_mc_printk(mci, KERN_WARNING,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700570 "failed to create sysfs device\n");
eric wollesen9794f332007-02-12 00:53:08 -0800571 goto fail1;
572 }
Alan Coxda9bb1d2006-01-18 17:44:13 -0800573
Dave Jiang81d87cb2007-07-19 01:49:52 -0700574 /* If there IS a check routine, then we are running POLLED */
575 if (mci->edac_check != NULL) {
576 /* This instance is NOW RUNNING */
577 mci->op_state = OP_RUNNING_POLL;
578
579 edac_mc_workq_setup(mci, edac_mc_get_poll_msec());
580 } else {
581 mci->op_state = OP_RUNNING_INTERRUPT;
582 }
583
Alan Coxda9bb1d2006-01-18 17:44:13 -0800584 /* Report action taken */
Doug Thompsonbf52fa42007-07-19 01:50:30 -0700585 edac_mc_printk(mci, KERN_INFO, "Giving out device to '%s' '%s':"
Stephen Rothwell17aa7e02008-05-05 13:54:19 +1000586 " DEV %s\n", mci->mod_name, mci->ctl_name, edac_dev_name(mci));
Alan Coxda9bb1d2006-01-18 17:44:13 -0800587
Matthias Kaehlcke63b7df92007-07-19 01:49:38 -0700588 mutex_unlock(&mem_ctls_mutex);
Dave Peterson028a7b62006-03-26 01:38:47 -0800589 return 0;
590
Douglas Thompson052dfb42007-07-19 01:50:13 -0700591fail1:
Dave Peterson028a7b62006-03-26 01:38:47 -0800592 del_mc_from_global_list(mci);
593
Douglas Thompson052dfb42007-07-19 01:50:13 -0700594fail0:
Matthias Kaehlcke63b7df92007-07-19 01:49:38 -0700595 mutex_unlock(&mem_ctls_mutex);
Dave Peterson028a7b62006-03-26 01:38:47 -0800596 return 1;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800597}
Dave Peterson91105402006-03-26 01:38:55 -0800598EXPORT_SYMBOL_GPL(edac_mc_add_mc);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800599
Alan Coxda9bb1d2006-01-18 17:44:13 -0800600/**
Dave Peterson472678e2006-03-26 01:38:49 -0800601 * edac_mc_del_mc: Remove sysfs entries for specified mci structure and
602 * remove mci structure from global list
Doug Thompson37f04582006-06-30 01:56:07 -0700603 * @pdev: Pointer to 'struct device' representing mci structure to remove.
Alan Coxda9bb1d2006-01-18 17:44:13 -0800604 *
Dave Peterson18dbc332006-03-26 01:38:50 -0800605 * Return pointer to removed mci structure, or NULL if device not found.
Alan Coxda9bb1d2006-01-18 17:44:13 -0800606 */
Douglas Thompson079708b2007-07-19 01:49:58 -0700607struct mem_ctl_info *edac_mc_del_mc(struct device *dev)
Alan Coxda9bb1d2006-01-18 17:44:13 -0800608{
Dave Peterson18dbc332006-03-26 01:38:50 -0800609 struct mem_ctl_info *mci;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800610
Doug Thompsonbf52fa42007-07-19 01:50:30 -0700611 debugf0("%s()\n", __func__);
612
Matthias Kaehlcke63b7df92007-07-19 01:49:38 -0700613 mutex_lock(&mem_ctls_mutex);
Dave Peterson18dbc332006-03-26 01:38:50 -0800614
Doug Thompsonbf52fa42007-07-19 01:50:30 -0700615 /* find the requested mci struct in the global list */
616 mci = find_mci_by_dev(dev);
617 if (mci == NULL) {
Matthias Kaehlcke63b7df92007-07-19 01:49:38 -0700618 mutex_unlock(&mem_ctls_mutex);
Dave Peterson18dbc332006-03-26 01:38:50 -0800619 return NULL;
620 }
621
Alan Coxda9bb1d2006-01-18 17:44:13 -0800622 del_mc_from_global_list(mci);
Matthias Kaehlcke63b7df92007-07-19 01:49:38 -0700623 mutex_unlock(&mem_ctls_mutex);
Doug Thompsonbf52fa42007-07-19 01:50:30 -0700624
Borislav Petkovbb31b3122010-12-02 17:48:35 +0100625 /* flush workq processes */
Doug Thompsonbf52fa42007-07-19 01:50:30 -0700626 edac_mc_workq_teardown(mci);
Borislav Petkovbb31b3122010-12-02 17:48:35 +0100627
628 /* marking MCI offline */
629 mci->op_state = OP_OFFLINE;
630
631 /* remove from sysfs */
Doug Thompsonbf52fa42007-07-19 01:50:30 -0700632 edac_remove_sysfs_mci_device(mci);
633
Dave Peterson537fba22006-03-26 01:38:40 -0800634 edac_printk(KERN_INFO, EDAC_MC,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700635 "Removed device %d for %s %s: DEV %s\n", mci->mc_idx,
Stephen Rothwell17aa7e02008-05-05 13:54:19 +1000636 mci->mod_name, mci->ctl_name, edac_dev_name(mci));
Doug Thompsonbf52fa42007-07-19 01:50:30 -0700637
Dave Peterson18dbc332006-03-26 01:38:50 -0800638 return mci;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800639}
Dave Peterson91105402006-03-26 01:38:55 -0800640EXPORT_SYMBOL_GPL(edac_mc_del_mc);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800641
Adrian Bunk2da1c112007-07-19 01:49:32 -0700642static void edac_mc_scrub_block(unsigned long page, unsigned long offset,
643 u32 size)
Alan Coxda9bb1d2006-01-18 17:44:13 -0800644{
645 struct page *pg;
646 void *virt_addr;
647 unsigned long flags = 0;
648
Dave Peterson537fba22006-03-26 01:38:40 -0800649 debugf3("%s()\n", __func__);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800650
651 /* ECC error page was not in our memory. Ignore it. */
Douglas Thompson079708b2007-07-19 01:49:58 -0700652 if (!pfn_valid(page))
Alan Coxda9bb1d2006-01-18 17:44:13 -0800653 return;
654
655 /* Find the actual page structure then map it and fix */
656 pg = pfn_to_page(page);
657
658 if (PageHighMem(pg))
659 local_irq_save(flags);
660
Cong Wang4e5df7c2011-11-25 23:14:19 +0800661 virt_addr = kmap_atomic(pg);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800662
663 /* Perform architecture specific atomic scrub operation */
664 atomic_scrub(virt_addr + offset, size);
665
666 /* Unmap and complete */
Cong Wang4e5df7c2011-11-25 23:14:19 +0800667 kunmap_atomic(virt_addr);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800668
669 if (PageHighMem(pg))
670 local_irq_restore(flags);
671}
672
Alan Coxda9bb1d2006-01-18 17:44:13 -0800673/* FIXME - should return -1 */
Dave Petersone7ecd892006-03-26 01:38:52 -0800674int edac_mc_find_csrow_by_page(struct mem_ctl_info *mci, unsigned long page)
Alan Coxda9bb1d2006-01-18 17:44:13 -0800675{
676 struct csrow_info *csrows = mci->csrows;
Mauro Carvalho Chehaba895bf82012-01-28 09:09:38 -0300677 int row, i, j, n;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800678
Dave Peterson537fba22006-03-26 01:38:40 -0800679 debugf1("MC%d: %s(): 0x%lx\n", mci->mc_idx, __func__, page);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800680 row = -1;
681
682 for (i = 0; i < mci->nr_csrows; i++) {
683 struct csrow_info *csrow = &csrows[i];
Mauro Carvalho Chehaba895bf82012-01-28 09:09:38 -0300684 n = 0;
685 for (j = 0; j < csrow->nr_channels; j++) {
686 struct dimm_info *dimm = csrow->channels[j].dimm;
687 n += dimm->nr_pages;
688 }
689 if (n == 0)
Alan Coxda9bb1d2006-01-18 17:44:13 -0800690 continue;
691
Dave Peterson537fba22006-03-26 01:38:40 -0800692 debugf3("MC%d: %s(): first(0x%lx) page(0x%lx) last(0x%lx) "
693 "mask(0x%lx)\n", mci->mc_idx, __func__,
694 csrow->first_page, page, csrow->last_page,
695 csrow->page_mask);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800696
697 if ((page >= csrow->first_page) &&
698 (page <= csrow->last_page) &&
699 ((page & csrow->page_mask) ==
700 (csrow->first_page & csrow->page_mask))) {
701 row = i;
702 break;
703 }
704 }
705
706 if (row == -1)
Dave Peterson537fba22006-03-26 01:38:40 -0800707 edac_mc_printk(mci, KERN_ERR,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700708 "could not look up page error address %lx\n",
709 (unsigned long)page);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800710
711 return row;
712}
Dave Peterson91105402006-03-26 01:38:55 -0800713EXPORT_SYMBOL_GPL(edac_mc_find_csrow_by_page);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800714
Alan Coxda9bb1d2006-01-18 17:44:13 -0800715/* FIXME - setable log (warning/emerg) levels */
716/* FIXME - integrate with evlog: http://evlog.sourceforge.net/ */
717void edac_mc_handle_ce(struct mem_ctl_info *mci,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700718 unsigned long page_frame_number,
719 unsigned long offset_in_page, unsigned long syndrome,
720 int row, int channel, const char *msg)
Alan Coxda9bb1d2006-01-18 17:44:13 -0800721{
722 unsigned long remapped_page;
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300723 char *label = NULL;
Mauro Carvalho Chehab084a4fc2012-01-27 18:38:08 -0300724 u32 grain;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800725
Dave Peterson537fba22006-03-26 01:38:40 -0800726 debugf3("MC%d: %s()\n", mci->mc_idx, __func__);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800727
728 /* FIXME - maybe make panic on INTERNAL ERROR an option */
729 if (row >= mci->nr_csrows || row < 0) {
730 /* something is wrong */
Dave Peterson537fba22006-03-26 01:38:40 -0800731 edac_mc_printk(mci, KERN_ERR,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700732 "INTERNAL ERROR: row out of range "
733 "(%d >= %d)\n", row, mci->nr_csrows);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800734 edac_mc_handle_ce_no_info(mci, "INTERNAL ERROR");
735 return;
736 }
Dave Petersone7ecd892006-03-26 01:38:52 -0800737
Alan Coxda9bb1d2006-01-18 17:44:13 -0800738 if (channel >= mci->csrows[row].nr_channels || channel < 0) {
739 /* something is wrong */
Dave Peterson537fba22006-03-26 01:38:40 -0800740 edac_mc_printk(mci, KERN_ERR,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700741 "INTERNAL ERROR: channel out of range "
742 "(%d >= %d)\n", channel,
743 mci->csrows[row].nr_channels);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800744 edac_mc_handle_ce_no_info(mci, "INTERNAL ERROR");
745 return;
746 }
747
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300748 label = mci->csrows[row].channels[channel].dimm->label;
Mauro Carvalho Chehab084a4fc2012-01-27 18:38:08 -0300749 grain = mci->csrows[row].channels[channel].dimm->grain;
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300750
Dave Jiang4de78c62007-07-19 01:49:54 -0700751 if (edac_mc_get_log_ce())
Alan Coxda9bb1d2006-01-18 17:44:13 -0800752 /* FIXME - put in DIMM location */
Dave Peterson537fba22006-03-26 01:38:40 -0800753 edac_mc_printk(mci, KERN_WARNING,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700754 "CE page 0x%lx, offset 0x%lx, grain %d, syndrome "
755 "0x%lx, row %d, channel %d, label \"%s\": %s\n",
756 page_frame_number, offset_in_page,
Mauro Carvalho Chehab084a4fc2012-01-27 18:38:08 -0300757 grain, syndrome, row, channel,
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300758 label, msg);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800759
760 mci->ce_count++;
761 mci->csrows[row].ce_count++;
Mauro Carvalho Chehab084a4fc2012-01-27 18:38:08 -0300762 mci->csrows[row].channels[channel].dimm->ce_count++;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800763 mci->csrows[row].channels[channel].ce_count++;
764
765 if (mci->scrub_mode & SCRUB_SW_SRC) {
766 /*
767 * Some MC's can remap memory so that it is still available
768 * at a different address when PCI devices map into memory.
769 * MC's that can't do this lose the memory where PCI devices
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300770 * are mapped. This mapping is MC dependent and so we call
Alan Coxda9bb1d2006-01-18 17:44:13 -0800771 * back into the MC driver for it to map the MC page to
772 * a physical (CPU) page which can then be mapped to a virtual
773 * page - which can then be scrubbed.
774 */
775 remapped_page = mci->ctl_page_to_phys ?
Douglas Thompson052dfb42007-07-19 01:50:13 -0700776 mci->ctl_page_to_phys(mci, page_frame_number) :
777 page_frame_number;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800778
Mauro Carvalho Chehab084a4fc2012-01-27 18:38:08 -0300779 edac_mc_scrub_block(remapped_page, offset_in_page, grain);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800780 }
781}
Dave Peterson91105402006-03-26 01:38:55 -0800782EXPORT_SYMBOL_GPL(edac_mc_handle_ce);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800783
Dave Petersone7ecd892006-03-26 01:38:52 -0800784void edac_mc_handle_ce_no_info(struct mem_ctl_info *mci, const char *msg)
Alan Coxda9bb1d2006-01-18 17:44:13 -0800785{
Dave Jiang4de78c62007-07-19 01:49:54 -0700786 if (edac_mc_get_log_ce())
Dave Peterson537fba22006-03-26 01:38:40 -0800787 edac_mc_printk(mci, KERN_WARNING,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700788 "CE - no information available: %s\n", msg);
Dave Petersone7ecd892006-03-26 01:38:52 -0800789
Alan Coxda9bb1d2006-01-18 17:44:13 -0800790 mci->ce_noinfo_count++;
791 mci->ce_count++;
792}
Dave Peterson91105402006-03-26 01:38:55 -0800793EXPORT_SYMBOL_GPL(edac_mc_handle_ce_no_info);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800794
Alan Coxda9bb1d2006-01-18 17:44:13 -0800795void edac_mc_handle_ue(struct mem_ctl_info *mci,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700796 unsigned long page_frame_number,
797 unsigned long offset_in_page, int row, const char *msg)
Alan Coxda9bb1d2006-01-18 17:44:13 -0800798{
799 int len = EDAC_MC_LABEL_LEN * 4;
800 char labels[len + 1];
801 char *pos = labels;
802 int chan;
803 int chars;
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300804 char *label = NULL;
Mauro Carvalho Chehab084a4fc2012-01-27 18:38:08 -0300805 u32 grain;
Alan Coxda9bb1d2006-01-18 17:44:13 -0800806
Dave Peterson537fba22006-03-26 01:38:40 -0800807 debugf3("MC%d: %s()\n", mci->mc_idx, __func__);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800808
809 /* FIXME - maybe make panic on INTERNAL ERROR an option */
810 if (row >= mci->nr_csrows || row < 0) {
811 /* something is wrong */
Dave Peterson537fba22006-03-26 01:38:40 -0800812 edac_mc_printk(mci, KERN_ERR,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700813 "INTERNAL ERROR: row out of range "
814 "(%d >= %d)\n", row, mci->nr_csrows);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800815 edac_mc_handle_ue_no_info(mci, "INTERNAL ERROR");
816 return;
817 }
818
Mauro Carvalho Chehab084a4fc2012-01-27 18:38:08 -0300819 grain = mci->csrows[row].channels[0].dimm->grain;
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300820 label = mci->csrows[row].channels[0].dimm->label;
821 chars = snprintf(pos, len + 1, "%s", label);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800822 len -= chars;
823 pos += chars;
Dave Petersone7ecd892006-03-26 01:38:52 -0800824
Alan Coxda9bb1d2006-01-18 17:44:13 -0800825 for (chan = 1; (chan < mci->csrows[row].nr_channels) && (len > 0);
Douglas Thompson052dfb42007-07-19 01:50:13 -0700826 chan++) {
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300827 label = mci->csrows[row].channels[chan].dimm->label;
828 chars = snprintf(pos, len + 1, ":%s", label);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800829 len -= chars;
830 pos += chars;
831 }
832
Dave Jiang4de78c62007-07-19 01:49:54 -0700833 if (edac_mc_get_log_ue())
Dave Peterson537fba22006-03-26 01:38:40 -0800834 edac_mc_printk(mci, KERN_EMERG,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700835 "UE page 0x%lx, offset 0x%lx, grain %d, row %d, "
836 "labels \"%s\": %s\n", page_frame_number,
Mauro Carvalho Chehab084a4fc2012-01-27 18:38:08 -0300837 offset_in_page, grain, row, labels, msg);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800838
Dave Jiang4de78c62007-07-19 01:49:54 -0700839 if (edac_mc_get_panic_on_ue())
Dave Petersone7ecd892006-03-26 01:38:52 -0800840 panic("EDAC MC%d: UE page 0x%lx, offset 0x%lx, grain %d, "
Douglas Thompson052dfb42007-07-19 01:50:13 -0700841 "row %d, labels \"%s\": %s\n", mci->mc_idx,
842 page_frame_number, offset_in_page,
Mauro Carvalho Chehab084a4fc2012-01-27 18:38:08 -0300843 grain, row, labels, msg);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800844
845 mci->ue_count++;
846 mci->csrows[row].ue_count++;
847}
Dave Peterson91105402006-03-26 01:38:55 -0800848EXPORT_SYMBOL_GPL(edac_mc_handle_ue);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800849
Dave Petersone7ecd892006-03-26 01:38:52 -0800850void edac_mc_handle_ue_no_info(struct mem_ctl_info *mci, const char *msg)
Alan Coxda9bb1d2006-01-18 17:44:13 -0800851{
Dave Jiang4de78c62007-07-19 01:49:54 -0700852 if (edac_mc_get_panic_on_ue())
Alan Coxda9bb1d2006-01-18 17:44:13 -0800853 panic("EDAC MC%d: Uncorrected Error", mci->mc_idx);
854
Dave Jiang4de78c62007-07-19 01:49:54 -0700855 if (edac_mc_get_log_ue())
Dave Peterson537fba22006-03-26 01:38:40 -0800856 edac_mc_printk(mci, KERN_WARNING,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700857 "UE - no information available: %s\n", msg);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800858 mci->ue_noinfo_count++;
859 mci->ue_count++;
860}
Douglas Thompson079708b2007-07-19 01:49:58 -0700861EXPORT_SYMBOL_GPL(edac_mc_handle_ue_no_info);
Alan Coxda9bb1d2006-01-18 17:44:13 -0800862
eric wollesen9794f332007-02-12 00:53:08 -0800863/*************************************************************
864 * On Fully Buffered DIMM modules, this help function is
865 * called to process UE events
866 */
867void edac_mc_handle_fbd_ue(struct mem_ctl_info *mci,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700868 unsigned int csrow,
869 unsigned int channela,
870 unsigned int channelb, char *msg)
eric wollesen9794f332007-02-12 00:53:08 -0800871{
872 int len = EDAC_MC_LABEL_LEN * 4;
873 char labels[len + 1];
874 char *pos = labels;
875 int chars;
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300876 char *label;
eric wollesen9794f332007-02-12 00:53:08 -0800877
878 if (csrow >= mci->nr_csrows) {
879 /* something is wrong */
880 edac_mc_printk(mci, KERN_ERR,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700881 "INTERNAL ERROR: row out of range (%d >= %d)\n",
882 csrow, mci->nr_csrows);
eric wollesen9794f332007-02-12 00:53:08 -0800883 edac_mc_handle_ue_no_info(mci, "INTERNAL ERROR");
884 return;
885 }
886
887 if (channela >= mci->csrows[csrow].nr_channels) {
888 /* something is wrong */
889 edac_mc_printk(mci, KERN_ERR,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700890 "INTERNAL ERROR: channel-a out of range "
891 "(%d >= %d)\n",
892 channela, mci->csrows[csrow].nr_channels);
eric wollesen9794f332007-02-12 00:53:08 -0800893 edac_mc_handle_ue_no_info(mci, "INTERNAL ERROR");
894 return;
895 }
896
897 if (channelb >= mci->csrows[csrow].nr_channels) {
898 /* something is wrong */
899 edac_mc_printk(mci, KERN_ERR,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700900 "INTERNAL ERROR: channel-b out of range "
901 "(%d >= %d)\n",
902 channelb, mci->csrows[csrow].nr_channels);
eric wollesen9794f332007-02-12 00:53:08 -0800903 edac_mc_handle_ue_no_info(mci, "INTERNAL ERROR");
904 return;
905 }
906
907 mci->ue_count++;
908 mci->csrows[csrow].ue_count++;
909
910 /* Generate the DIMM labels from the specified channels */
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300911 label = mci->csrows[csrow].channels[channela].dimm->label;
912 chars = snprintf(pos, len + 1, "%s", label);
Douglas Thompson079708b2007-07-19 01:49:58 -0700913 len -= chars;
914 pos += chars;
Mauro Carvalho Chehab084a4fc2012-01-27 18:38:08 -0300915
eric wollesen9794f332007-02-12 00:53:08 -0800916 chars = snprintf(pos, len + 1, "-%s",
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300917 mci->csrows[csrow].channels[channelb].dimm->label);
eric wollesen9794f332007-02-12 00:53:08 -0800918
Dave Jiang4de78c62007-07-19 01:49:54 -0700919 if (edac_mc_get_log_ue())
eric wollesen9794f332007-02-12 00:53:08 -0800920 edac_mc_printk(mci, KERN_EMERG,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700921 "UE row %d, channel-a= %d channel-b= %d "
922 "labels \"%s\": %s\n", csrow, channela, channelb,
923 labels, msg);
eric wollesen9794f332007-02-12 00:53:08 -0800924
Dave Jiang4de78c62007-07-19 01:49:54 -0700925 if (edac_mc_get_panic_on_ue())
eric wollesen9794f332007-02-12 00:53:08 -0800926 panic("UE row %d, channel-a= %d channel-b= %d "
Douglas Thompson052dfb42007-07-19 01:50:13 -0700927 "labels \"%s\": %s\n", csrow, channela,
928 channelb, labels, msg);
eric wollesen9794f332007-02-12 00:53:08 -0800929}
930EXPORT_SYMBOL(edac_mc_handle_fbd_ue);
931
932/*************************************************************
933 * On Fully Buffered DIMM modules, this help function is
934 * called to process CE events
935 */
936void edac_mc_handle_fbd_ce(struct mem_ctl_info *mci,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700937 unsigned int csrow, unsigned int channel, char *msg)
eric wollesen9794f332007-02-12 00:53:08 -0800938{
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300939 char *label = NULL;
eric wollesen9794f332007-02-12 00:53:08 -0800940
941 /* Ensure boundary values */
942 if (csrow >= mci->nr_csrows) {
943 /* something is wrong */
944 edac_mc_printk(mci, KERN_ERR,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700945 "INTERNAL ERROR: row out of range (%d >= %d)\n",
946 csrow, mci->nr_csrows);
eric wollesen9794f332007-02-12 00:53:08 -0800947 edac_mc_handle_ce_no_info(mci, "INTERNAL ERROR");
948 return;
949 }
950 if (channel >= mci->csrows[csrow].nr_channels) {
951 /* something is wrong */
952 edac_mc_printk(mci, KERN_ERR,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700953 "INTERNAL ERROR: channel out of range (%d >= %d)\n",
954 channel, mci->csrows[csrow].nr_channels);
eric wollesen9794f332007-02-12 00:53:08 -0800955 edac_mc_handle_ce_no_info(mci, "INTERNAL ERROR");
956 return;
957 }
958
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300959 label = mci->csrows[csrow].channels[channel].dimm->label;
960
Dave Jiang4de78c62007-07-19 01:49:54 -0700961 if (edac_mc_get_log_ce())
eric wollesen9794f332007-02-12 00:53:08 -0800962 /* FIXME - put in DIMM location */
963 edac_mc_printk(mci, KERN_WARNING,
Douglas Thompson052dfb42007-07-19 01:50:13 -0700964 "CE row %d, channel %d, label \"%s\": %s\n",
Mauro Carvalho Chehaba7d7d2e2012-01-27 14:12:32 -0300965 csrow, channel, label, msg);
eric wollesen9794f332007-02-12 00:53:08 -0800966
967 mci->ce_count++;
968 mci->csrows[csrow].ce_count++;
Mauro Carvalho Chehab084a4fc2012-01-27 18:38:08 -0300969 mci->csrows[csrow].channels[channel].dimm->ce_count++;
eric wollesen9794f332007-02-12 00:53:08 -0800970 mci->csrows[csrow].channels[channel].ce_count++;
971}
Douglas Thompson079708b2007-07-19 01:49:58 -0700972EXPORT_SYMBOL(edac_mc_handle_fbd_ce);