1 /*
2 * Defines, structures, APIs for edac_device
3 *
4 * (C) 2007 Linux Networx (http://lnxi.com)
5 * This file may be distributed under the terms of the
6 * GNU General Public License.
7 *
8 * Written by Thayne Harbaugh
9 * Based on work by Dan Hollis <goemon at anime dot net> and others.
10 * http://www.anime.net/~goemon/linux-ecc/
11 *
12 * NMI handling support added by
13 * Dave Peterson <dsp@llnl.gov> <dave_peterson@pobox.com>
14 *
15 * Refactored for multi-source files:
16 * Doug Thompson <norsk5@xmission.com>
17 *
18 * Please look at Documentation/driver-api/edac.rst for more info about
19 * EDAC core structs and functions.
20 */
21
22 #ifndef _EDAC_DEVICE_H_
23 #define _EDAC_DEVICE_H_
24
25 #include <linux/device.h>
26 #include <linux/edac.h>
27 #include <linux/kobject.h>
28 #include <linux/list.h>
29 #include <linux/types.h>
30 #include <linux/sysfs.h>
31 #include <linux/workqueue.h>
32
33
34 /*
35 * The following are the structures to provide for a generic
36 * or abstract 'edac_device'. This set of structures and the
37 * code that implements the APIs for the same, provide for
38 * registering EDAC type devices which are NOT standard memory.
39 *
40 * CPU caches (L1 and L2)
41 * DMA engines
42 * Core CPU switches
43 * Fabric switch units
44 * PCIe interface controllers
45 * other EDAC/ECC type devices that can be monitored for
46 * errors, etc.
47 *
48 * It allows for a 2 level set of hierarchy. For example:
49 *
50 * cache could be composed of L1, L2 and L3 levels of cache.
51 * Each CPU core would have its own L1 cache, while sharing
52 * L2 and maybe L3 caches.
53 *
54 * View them arranged, via the sysfs presentation:
55 * /sys/devices/system/edac/..
56 *
57 * mc/ <existing memory device directory>
58 * cpu/cpu0/.. <L1 and L2 block directory>
59 * /L1-cache/ce_count
60 * /ue_count
61 * /L2-cache/ce_count
62 * /ue_count
63 * cpu/cpu1/.. <L1 and L2 block directory>
64 * /L1-cache/ce_count
65 * /ue_count
66 * /L2-cache/ce_count
67 * /ue_count
68 * ...
69 *
70 * the L1 and L2 directories would be "edac_device_block's"
71 */
72
73 struct edac_device_counter {
74 u32 ue_count;
75 u32 ce_count;
76 };
77
78 /* forward reference */
79 struct edac_device_ctl_info;
80 struct edac_device_block;
81
82 /* edac_dev_sysfs_attribute structure
83 * used for driver sysfs attributes in mem_ctl_info
84 * for extra controls and attributes:
85 * like high level error Injection controls
86 */
87 struct edac_dev_sysfs_attribute {
88 struct attribute attr;
89 ssize_t (*show)(struct edac_device_ctl_info *, char *);
90 ssize_t (*store)(struct edac_device_ctl_info *, const char *, size_t);
91 };
92
93 /* edac_dev_sysfs_block_attribute structure
94 *
95 * used in leaf 'block' nodes for adding controls/attributes
96 *
97 * each block in each instance of the containing control structure can
98 * have an array of the following. The show function will be filled in
99 * with the show function in the low level driver.
100 */
101 struct edac_dev_sysfs_block_attribute {
102 struct attribute attr;
103 ssize_t (*show)(struct kobject *, struct attribute *, char *);
104 };
105
106 /* device block control structure */
107 struct edac_device_block {
108 struct edac_device_instance *instance; /* Up Pointer */
109 char name[EDAC_DEVICE_NAME_LEN + 1];
110
111 struct edac_device_counter counters; /* basic UE and CE counters */
112
113 int nr_attribs; /* how many attributes */
114
115 /* this block's attributes, could be NULL */
116 struct edac_dev_sysfs_block_attribute *block_attributes;
117
118 /* edac sysfs device control */
119 struct kobject kobj;
120 };
121
122 /* device instance control structure */
123 struct edac_device_instance {
124 struct edac_device_ctl_info *ctl; /* Up pointer */
125 char name[EDAC_DEVICE_NAME_LEN + 4];
126
127 struct edac_device_counter counters; /* instance counters */
128
129 u32 nr_blocks; /* how many blocks */
130 struct edac_device_block *blocks; /* block array */
131
132 /* edac sysfs device control */
133 struct kobject kobj;
134 };
135
136
137 /*
138 * Abstract edac_device control info structure
139 *
140 */
141 struct edac_device_ctl_info {
142 /* for global list of edac_device_ctl_info structs */
143 struct list_head link;
144
145 struct module *owner; /* Module owner of this control struct */
146
147 int dev_idx;
148
149 /* Per instance controls for this edac_device */
150 int log_ue; /* boolean for logging UEs */
151 int log_ce; /* boolean for logging CEs */
152 int panic_on_ue; /* boolean for panic'ing on an UE */
153 unsigned poll_msec; /* number of milliseconds to poll interval */
154 unsigned long delay; /* number of jiffies for poll_msec */
155
156 /* Additional top controller level attributes, but specified
157 * by the low level driver.
158 *
159 * Set by the low level driver to provide attributes at the
160 * controller level, same level as 'ue_count' and 'ce_count' above.
161 * An array of structures, NULL terminated
162 *
163 * If attributes are desired, then set to array of attributes
164 * If no attributes are desired, leave NULL
165 */
166 struct edac_dev_sysfs_attribute *sysfs_attributes;
167
168 /* pointer to main 'edac' subsys in sysfs */
169 const struct bus_type *edac_subsys;
170
171 /* the internal state of this controller instance */
172 int op_state;
173 /* work struct for this instance */
174 struct delayed_work work;
175
176 /* pointer to edac polling checking routine:
177 * If NOT NULL: points to polling check routine
178 * If NULL: Then assumes INTERRUPT operation, where
179 * MC driver will receive events
180 */
181 void (*edac_check) (struct edac_device_ctl_info * edac_dev);
182
183 struct device *dev; /* pointer to device structure */
184
185 const char *mod_name; /* module name */
186 const char *ctl_name; /* edac controller name */
187 const char *dev_name; /* pci/platform/etc... name */
188
189 void *pvt_info; /* pointer to 'private driver' info */
190
191 unsigned long start_time; /* edac_device load start time (jiffies) */
192
193 /* sysfs top name under 'edac' directory
194 * and instance name:
195 * cpu/cpu0/...
196 * cpu/cpu1/...
197 * cpu/cpu2/...
198 * ...
199 */
200 char name[EDAC_DEVICE_NAME_LEN + 1];
201
202 /* Number of instances supported on this control structure
203 * and the array of those instances
204 */
205 u32 nr_instances;
206 struct edac_device_instance *instances;
207 struct edac_device_block *blocks;
208
209 /* Event counters for the this whole EDAC Device */
210 struct edac_device_counter counters;
211
212 /* edac sysfs device control for the 'name'
213 * device this structure controls
214 */
215 struct kobject kobj;
216 };
217
218 /* To get from the instance's wq to the beginning of the ctl structure */
219 #define to_edac_mem_ctl_work(w) \
220 container_of(w, struct mem_ctl_info, work)
221
222 #define to_edac_device_ctl_work(w) \
223 container_of(w,struct edac_device_ctl_info,work)
224
225 /*
226 * The alloc() and free() functions for the 'edac_device' control info
227 * structure. A MC driver will allocate one of these for each edac_device
228 * it is going to control/register with the EDAC CORE.
229 */
230 extern struct edac_device_ctl_info *edac_device_alloc_ctl_info(
231 unsigned sizeof_private,
232 char *edac_device_name, unsigned nr_instances,
233 char *edac_block_name, unsigned nr_blocks,
234 unsigned offset_value,
235 int device_index);
236
237 /* The offset value can be:
238 * -1 indicating no offset value
239 * 0 for zero-based block numbers
240 * 1 for 1-based block number
241 * other for other-based block number
242 */
243 #define BLOCK_OFFSET_VALUE_OFF ((unsigned) -1)
244
245 extern void edac_device_free_ctl_info(struct edac_device_ctl_info *ctl_info);
246
247 /**
248 * edac_device_add_device - Insert the 'edac_dev' structure into the
249 * edac_device global list and create sysfs entries associated with
250 * edac_device structure.
251 *
252 * @edac_dev: pointer to edac_device structure to be added to the list
253 * 'edac_device' structure.
254 *
255 * Returns:
256 * 0 on Success, or an error code on failure
257 */
258 extern int edac_device_add_device(struct edac_device_ctl_info *edac_dev);
259
260 /**
261 * edac_device_del_device - Remove sysfs entries for specified edac_device
262 * structure and then remove edac_device structure from global list
263 *
264 * @dev:
265 * Pointer to struct &device representing the edac device
266 * structure to remove.
267 *
268 * Returns:
269 * Pointer to removed edac_device structure,
270 * or %NULL if device not found.
271 */
272 extern struct edac_device_ctl_info *edac_device_del_device(struct device *dev);
273
274 /**
275 * edac_device_handle_ce_count - Log correctable errors.
276 *
277 * @edac_dev: pointer to struct &edac_device_ctl_info
278 * @inst_nr: number of the instance where the CE error happened
279 * @count: Number of errors to log.
280 * @block_nr: number of the block where the CE error happened
281 * @msg: message to be printed
282 */
283 void edac_device_handle_ce_count(struct edac_device_ctl_info *edac_dev,
284 unsigned int count, int inst_nr, int block_nr,
285 const char *msg);
286
287 /**
288 * edac_device_handle_ue_count - Log uncorrectable errors.
289 *
290 * @edac_dev: pointer to struct &edac_device_ctl_info
291 * @inst_nr: number of the instance where the CE error happened
292 * @count: Number of errors to log.
293 * @block_nr: number of the block where the CE error happened
294 * @msg: message to be printed
295 */
296 void edac_device_handle_ue_count(struct edac_device_ctl_info *edac_dev,
297 unsigned int count, int inst_nr, int block_nr,
298 const char *msg);
299
300 /**
301 * edac_device_handle_ce(): Log a single correctable error
302 *
303 * @edac_dev: pointer to struct &edac_device_ctl_info
304 * @inst_nr: number of the instance where the CE error happened
305 * @block_nr: number of the block where the CE error happened
306 * @msg: message to be printed
307 */
308 static inline void
edac_device_handle_ce(struct edac_device_ctl_info * edac_dev,int inst_nr,int block_nr,const char * msg)309 edac_device_handle_ce(struct edac_device_ctl_info *edac_dev, int inst_nr,
310 int block_nr, const char *msg)
311 {
312 edac_device_handle_ce_count(edac_dev, 1, inst_nr, block_nr, msg);
313 }
314
315 /**
316 * edac_device_handle_ue(): Log a single uncorrectable error
317 *
318 * @edac_dev: pointer to struct &edac_device_ctl_info
319 * @inst_nr: number of the instance where the UE error happened
320 * @block_nr: number of the block where the UE error happened
321 * @msg: message to be printed
322 */
323 static inline void
edac_device_handle_ue(struct edac_device_ctl_info * edac_dev,int inst_nr,int block_nr,const char * msg)324 edac_device_handle_ue(struct edac_device_ctl_info *edac_dev, int inst_nr,
325 int block_nr, const char *msg)
326 {
327 edac_device_handle_ue_count(edac_dev, 1, inst_nr, block_nr, msg);
328 }
329
330 /**
331 * edac_device_alloc_index: Allocate a unique device index number
332 *
333 * Returns:
334 * allocated index number
335 */
336 extern int edac_device_alloc_index(void);
337 extern const char *edac_layer_name[];
338
339 /* Free the actual struct */
__edac_device_free_ctl_info(struct edac_device_ctl_info * ci)340 static inline void __edac_device_free_ctl_info(struct edac_device_ctl_info *ci)
341 {
342 if (ci) {
343 kfree(ci->pvt_info);
344 kfree(ci->blocks);
345 kfree(ci->instances);
346 kfree(ci);
347 }
348 }
349 #endif
350