Commit | Line | Data |
---|---|---|
e6dfb2de DW |
1 | /* |
2 | * Copyright(c) 2013-2015 Intel Corporation. All rights reserved. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or modify | |
5 | * it under the terms of version 2 of the GNU General Public License as | |
6 | * published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, but | |
9 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
11 | * General Public License for more details. | |
12 | */ | |
13 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt | |
4d88a97a | 14 | #include <linux/vmalloc.h> |
e6dfb2de | 15 | #include <linux/device.h> |
62232e45 | 16 | #include <linux/ndctl.h> |
e6dfb2de DW |
17 | #include <linux/slab.h> |
18 | #include <linux/io.h> | |
19 | #include <linux/fs.h> | |
20 | #include <linux/mm.h> | |
21 | #include "nd-core.h" | |
0ba1c634 | 22 | #include "label.h" |
4d88a97a | 23 | #include "nd.h" |
e6dfb2de DW |
24 | |
25 | static DEFINE_IDA(dimm_ida); | |
26 | ||
4d88a97a DW |
27 | /* |
28 | * Retrieve bus and dimm handle and return if this bus supports | |
29 | * get_config_data commands | |
30 | */ | |
31 | static int __validate_dimm(struct nvdimm_drvdata *ndd) | |
32 | { | |
33 | struct nvdimm *nvdimm; | |
34 | ||
35 | if (!ndd) | |
36 | return -EINVAL; | |
37 | ||
38 | nvdimm = to_nvdimm(ndd->dev); | |
39 | ||
40 | if (!nvdimm->dsm_mask) | |
41 | return -ENXIO; | |
42 | if (!test_bit(ND_CMD_GET_CONFIG_DATA, nvdimm->dsm_mask)) | |
43 | return -ENXIO; | |
44 | ||
45 | return 0; | |
46 | } | |
47 | ||
48 | static int validate_dimm(struct nvdimm_drvdata *ndd) | |
49 | { | |
50 | int rc = __validate_dimm(ndd); | |
51 | ||
52 | if (rc && ndd) | |
53 | dev_dbg(ndd->dev, "%pf: %s error: %d\n", | |
54 | __builtin_return_address(0), __func__, rc); | |
55 | return rc; | |
56 | } | |
57 | ||
58 | /** | |
59 | * nvdimm_init_nsarea - determine the geometry of a dimm's namespace area | |
60 | * @nvdimm: dimm to initialize | |
61 | */ | |
62 | int nvdimm_init_nsarea(struct nvdimm_drvdata *ndd) | |
63 | { | |
64 | struct nd_cmd_get_config_size *cmd = &ndd->nsarea; | |
65 | struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(ndd->dev); | |
66 | struct nvdimm_bus_descriptor *nd_desc; | |
67 | int rc = validate_dimm(ndd); | |
68 | ||
69 | if (rc) | |
70 | return rc; | |
71 | ||
72 | if (cmd->config_size) | |
73 | return 0; /* already valid */ | |
74 | ||
75 | memset(cmd, 0, sizeof(*cmd)); | |
76 | nd_desc = nvdimm_bus->nd_desc; | |
77 | return nd_desc->ndctl(nd_desc, to_nvdimm(ndd->dev), | |
78 | ND_CMD_GET_CONFIG_SIZE, cmd, sizeof(*cmd)); | |
79 | } | |
80 | ||
81 | int nvdimm_init_config_data(struct nvdimm_drvdata *ndd) | |
82 | { | |
83 | struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(ndd->dev); | |
84 | struct nd_cmd_get_config_data_hdr *cmd; | |
85 | struct nvdimm_bus_descriptor *nd_desc; | |
86 | int rc = validate_dimm(ndd); | |
87 | u32 max_cmd_size, config_size; | |
88 | size_t offset; | |
89 | ||
90 | if (rc) | |
91 | return rc; | |
92 | ||
93 | if (ndd->data) | |
94 | return 0; | |
95 | ||
4a826c83 DW |
96 | if (ndd->nsarea.status || ndd->nsarea.max_xfer == 0 |
97 | || ndd->nsarea.config_size < ND_LABEL_MIN_SIZE) { | |
98 | dev_dbg(ndd->dev, "failed to init config data area: (%d:%d)\n", | |
99 | ndd->nsarea.max_xfer, ndd->nsarea.config_size); | |
4d88a97a | 100 | return -ENXIO; |
4a826c83 | 101 | } |
4d88a97a DW |
102 | |
103 | ndd->data = kmalloc(ndd->nsarea.config_size, GFP_KERNEL); | |
104 | if (!ndd->data) | |
105 | ndd->data = vmalloc(ndd->nsarea.config_size); | |
106 | ||
107 | if (!ndd->data) | |
108 | return -ENOMEM; | |
109 | ||
110 | max_cmd_size = min_t(u32, PAGE_SIZE, ndd->nsarea.max_xfer); | |
111 | cmd = kzalloc(max_cmd_size + sizeof(*cmd), GFP_KERNEL); | |
112 | if (!cmd) | |
113 | return -ENOMEM; | |
114 | ||
115 | nd_desc = nvdimm_bus->nd_desc; | |
116 | for (config_size = ndd->nsarea.config_size, offset = 0; | |
117 | config_size; config_size -= cmd->in_length, | |
118 | offset += cmd->in_length) { | |
119 | cmd->in_length = min(config_size, max_cmd_size); | |
120 | cmd->in_offset = offset; | |
121 | rc = nd_desc->ndctl(nd_desc, to_nvdimm(ndd->dev), | |
122 | ND_CMD_GET_CONFIG_DATA, cmd, | |
123 | cmd->in_length + sizeof(*cmd)); | |
124 | if (rc || cmd->status) { | |
125 | rc = -ENXIO; | |
126 | break; | |
127 | } | |
128 | memcpy(ndd->data + offset, cmd->out_buf, cmd->in_length); | |
129 | } | |
130 | dev_dbg(ndd->dev, "%s: len: %zu rc: %d\n", __func__, offset, rc); | |
131 | kfree(cmd); | |
132 | ||
133 | return rc; | |
134 | } | |
135 | ||
f524bf27 DW |
136 | int nvdimm_set_config_data(struct nvdimm_drvdata *ndd, size_t offset, |
137 | void *buf, size_t len) | |
138 | { | |
139 | int rc = validate_dimm(ndd); | |
140 | size_t max_cmd_size, buf_offset; | |
141 | struct nd_cmd_set_config_hdr *cmd; | |
142 | struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(ndd->dev); | |
143 | struct nvdimm_bus_descriptor *nd_desc = nvdimm_bus->nd_desc; | |
144 | ||
145 | if (rc) | |
146 | return rc; | |
147 | ||
148 | if (!ndd->data) | |
149 | return -ENXIO; | |
150 | ||
151 | if (offset + len > ndd->nsarea.config_size) | |
152 | return -ENXIO; | |
153 | ||
154 | max_cmd_size = min_t(u32, PAGE_SIZE, len); | |
155 | max_cmd_size = min_t(u32, max_cmd_size, ndd->nsarea.max_xfer); | |
156 | cmd = kzalloc(max_cmd_size + sizeof(*cmd) + sizeof(u32), GFP_KERNEL); | |
157 | if (!cmd) | |
158 | return -ENOMEM; | |
159 | ||
160 | for (buf_offset = 0; len; len -= cmd->in_length, | |
161 | buf_offset += cmd->in_length) { | |
162 | size_t cmd_size; | |
163 | u32 *status; | |
164 | ||
165 | cmd->in_offset = offset + buf_offset; | |
166 | cmd->in_length = min(max_cmd_size, len); | |
167 | memcpy(cmd->in_buf, buf + buf_offset, cmd->in_length); | |
168 | ||
169 | /* status is output in the last 4-bytes of the command buffer */ | |
170 | cmd_size = sizeof(*cmd) + cmd->in_length + sizeof(u32); | |
171 | status = ((void *) cmd) + cmd_size - sizeof(u32); | |
172 | ||
173 | rc = nd_desc->ndctl(nd_desc, to_nvdimm(ndd->dev), | |
174 | ND_CMD_SET_CONFIG_DATA, cmd, cmd_size); | |
175 | if (rc || *status) { | |
176 | rc = rc ? rc : -ENXIO; | |
177 | break; | |
178 | } | |
179 | } | |
180 | kfree(cmd); | |
181 | ||
182 | return rc; | |
183 | } | |
184 | ||
e6dfb2de DW |
185 | static void nvdimm_release(struct device *dev) |
186 | { | |
187 | struct nvdimm *nvdimm = to_nvdimm(dev); | |
188 | ||
189 | ida_simple_remove(&dimm_ida, nvdimm->id); | |
190 | kfree(nvdimm); | |
191 | } | |
192 | ||
193 | static struct device_type nvdimm_device_type = { | |
194 | .name = "nvdimm", | |
195 | .release = nvdimm_release, | |
196 | }; | |
197 | ||
62232e45 | 198 | bool is_nvdimm(struct device *dev) |
e6dfb2de DW |
199 | { |
200 | return dev->type == &nvdimm_device_type; | |
201 | } | |
202 | ||
203 | struct nvdimm *to_nvdimm(struct device *dev) | |
204 | { | |
205 | struct nvdimm *nvdimm = container_of(dev, struct nvdimm, dev); | |
206 | ||
207 | WARN_ON(!is_nvdimm(dev)); | |
208 | return nvdimm; | |
209 | } | |
210 | EXPORT_SYMBOL_GPL(to_nvdimm); | |
211 | ||
bf9bccc1 DW |
212 | struct nvdimm_drvdata *to_ndd(struct nd_mapping *nd_mapping) |
213 | { | |
214 | struct nvdimm *nvdimm = nd_mapping->nvdimm; | |
215 | ||
216 | WARN_ON_ONCE(!is_nvdimm_bus_locked(&nvdimm->dev)); | |
217 | ||
218 | return dev_get_drvdata(&nvdimm->dev); | |
219 | } | |
220 | EXPORT_SYMBOL(to_ndd); | |
221 | ||
222 | void nvdimm_drvdata_release(struct kref *kref) | |
223 | { | |
224 | struct nvdimm_drvdata *ndd = container_of(kref, typeof(*ndd), kref); | |
225 | struct device *dev = ndd->dev; | |
226 | struct resource *res, *_r; | |
227 | ||
228 | dev_dbg(dev, "%s\n", __func__); | |
229 | ||
230 | nvdimm_bus_lock(dev); | |
231 | for_each_dpa_resource_safe(ndd, res, _r) | |
232 | nvdimm_free_dpa(ndd, res); | |
233 | nvdimm_bus_unlock(dev); | |
234 | ||
235 | if (ndd->data && is_vmalloc_addr(ndd->data)) | |
236 | vfree(ndd->data); | |
237 | else | |
238 | kfree(ndd->data); | |
239 | kfree(ndd); | |
240 | put_device(dev); | |
241 | } | |
242 | ||
243 | void get_ndd(struct nvdimm_drvdata *ndd) | |
244 | { | |
245 | kref_get(&ndd->kref); | |
246 | } | |
247 | ||
248 | void put_ndd(struct nvdimm_drvdata *ndd) | |
249 | { | |
250 | if (ndd) | |
251 | kref_put(&ndd->kref, nvdimm_drvdata_release); | |
252 | } | |
253 | ||
e6dfb2de DW |
254 | const char *nvdimm_name(struct nvdimm *nvdimm) |
255 | { | |
256 | return dev_name(&nvdimm->dev); | |
257 | } | |
258 | EXPORT_SYMBOL_GPL(nvdimm_name); | |
259 | ||
260 | void *nvdimm_provider_data(struct nvdimm *nvdimm) | |
261 | { | |
62232e45 DW |
262 | if (nvdimm) |
263 | return nvdimm->provider_data; | |
264 | return NULL; | |
e6dfb2de DW |
265 | } |
266 | EXPORT_SYMBOL_GPL(nvdimm_provider_data); | |
267 | ||
62232e45 DW |
268 | static ssize_t commands_show(struct device *dev, |
269 | struct device_attribute *attr, char *buf) | |
270 | { | |
271 | struct nvdimm *nvdimm = to_nvdimm(dev); | |
272 | int cmd, len = 0; | |
273 | ||
274 | if (!nvdimm->dsm_mask) | |
275 | return sprintf(buf, "\n"); | |
276 | ||
277 | for_each_set_bit(cmd, nvdimm->dsm_mask, BITS_PER_LONG) | |
278 | len += sprintf(buf + len, "%s ", nvdimm_cmd_name(cmd)); | |
279 | len += sprintf(buf + len, "\n"); | |
280 | return len; | |
281 | } | |
282 | static DEVICE_ATTR_RO(commands); | |
283 | ||
eaf96153 DW |
284 | static ssize_t state_show(struct device *dev, struct device_attribute *attr, |
285 | char *buf) | |
286 | { | |
287 | struct nvdimm *nvdimm = to_nvdimm(dev); | |
288 | ||
289 | /* | |
290 | * The state may be in the process of changing, userspace should | |
291 | * quiesce probing if it wants a static answer | |
292 | */ | |
293 | nvdimm_bus_lock(dev); | |
294 | nvdimm_bus_unlock(dev); | |
295 | return sprintf(buf, "%s\n", atomic_read(&nvdimm->busy) | |
296 | ? "active" : "idle"); | |
297 | } | |
298 | static DEVICE_ATTR_RO(state); | |
299 | ||
0ba1c634 DW |
300 | static ssize_t available_slots_show(struct device *dev, |
301 | struct device_attribute *attr, char *buf) | |
302 | { | |
303 | struct nvdimm_drvdata *ndd = dev_get_drvdata(dev); | |
304 | ssize_t rc; | |
305 | u32 nfree; | |
306 | ||
307 | if (!ndd) | |
308 | return -ENXIO; | |
309 | ||
310 | nvdimm_bus_lock(dev); | |
311 | nfree = nd_label_nfree(ndd); | |
312 | if (nfree - 1 > nfree) { | |
313 | dev_WARN_ONCE(dev, 1, "we ate our last label?\n"); | |
314 | nfree = 0; | |
315 | } else | |
316 | nfree--; | |
317 | rc = sprintf(buf, "%d\n", nfree); | |
318 | nvdimm_bus_unlock(dev); | |
319 | return rc; | |
320 | } | |
321 | static DEVICE_ATTR_RO(available_slots); | |
322 | ||
62232e45 | 323 | static struct attribute *nvdimm_attributes[] = { |
eaf96153 | 324 | &dev_attr_state.attr, |
62232e45 | 325 | &dev_attr_commands.attr, |
0ba1c634 | 326 | &dev_attr_available_slots.attr, |
62232e45 DW |
327 | NULL, |
328 | }; | |
329 | ||
330 | struct attribute_group nvdimm_attribute_group = { | |
331 | .attrs = nvdimm_attributes, | |
332 | }; | |
333 | EXPORT_SYMBOL_GPL(nvdimm_attribute_group); | |
334 | ||
e6dfb2de | 335 | struct nvdimm *nvdimm_create(struct nvdimm_bus *nvdimm_bus, void *provider_data, |
62232e45 DW |
336 | const struct attribute_group **groups, unsigned long flags, |
337 | unsigned long *dsm_mask) | |
e6dfb2de DW |
338 | { |
339 | struct nvdimm *nvdimm = kzalloc(sizeof(*nvdimm), GFP_KERNEL); | |
340 | struct device *dev; | |
341 | ||
342 | if (!nvdimm) | |
343 | return NULL; | |
344 | ||
345 | nvdimm->id = ida_simple_get(&dimm_ida, 0, 0, GFP_KERNEL); | |
346 | if (nvdimm->id < 0) { | |
347 | kfree(nvdimm); | |
348 | return NULL; | |
349 | } | |
350 | nvdimm->provider_data = provider_data; | |
351 | nvdimm->flags = flags; | |
62232e45 | 352 | nvdimm->dsm_mask = dsm_mask; |
eaf96153 | 353 | atomic_set(&nvdimm->busy, 0); |
e6dfb2de DW |
354 | dev = &nvdimm->dev; |
355 | dev_set_name(dev, "nmem%d", nvdimm->id); | |
356 | dev->parent = &nvdimm_bus->dev; | |
357 | dev->type = &nvdimm_device_type; | |
62232e45 | 358 | dev->devt = MKDEV(nvdimm_major, nvdimm->id); |
e6dfb2de | 359 | dev->groups = groups; |
4d88a97a | 360 | nd_device_register(dev); |
e6dfb2de DW |
361 | |
362 | return nvdimm; | |
363 | } | |
364 | EXPORT_SYMBOL_GPL(nvdimm_create); | |
4d88a97a | 365 | |
1b40e09a DW |
366 | /** |
367 | * nd_blk_available_dpa - account the unused dpa of BLK region | |
368 | * @nd_mapping: container of dpa-resource-root + labels | |
369 | * | |
370 | * Unlike PMEM, BLK namespaces can occupy discontiguous DPA ranges. | |
371 | */ | |
372 | resource_size_t nd_blk_available_dpa(struct nd_mapping *nd_mapping) | |
373 | { | |
374 | struct nvdimm_drvdata *ndd = to_ndd(nd_mapping); | |
375 | resource_size_t map_end, busy = 0, available; | |
376 | struct resource *res; | |
377 | ||
378 | if (!ndd) | |
379 | return 0; | |
380 | ||
381 | map_end = nd_mapping->start + nd_mapping->size - 1; | |
382 | for_each_dpa_resource(ndd, res) | |
383 | if (res->start >= nd_mapping->start && res->start < map_end) { | |
384 | resource_size_t end = min(map_end, res->end); | |
385 | ||
386 | busy += end - res->start + 1; | |
387 | } else if (res->end >= nd_mapping->start | |
388 | && res->end <= map_end) { | |
389 | busy += res->end - nd_mapping->start; | |
390 | } else if (nd_mapping->start > res->start | |
391 | && nd_mapping->start < res->end) { | |
392 | /* total eclipse of the BLK region mapping */ | |
393 | busy += nd_mapping->size; | |
394 | } | |
395 | ||
396 | available = map_end - nd_mapping->start + 1; | |
397 | if (busy < available) | |
398 | return available - busy; | |
399 | return 0; | |
400 | } | |
401 | ||
bf9bccc1 DW |
402 | /** |
403 | * nd_pmem_available_dpa - for the given dimm+region account unallocated dpa | |
404 | * @nd_mapping: container of dpa-resource-root + labels | |
405 | * @nd_region: constrain available space check to this reference region | |
406 | * @overlap: calculate available space assuming this level of overlap | |
407 | * | |
408 | * Validate that a PMEM label, if present, aligns with the start of an | |
409 | * interleave set and truncate the available size at the lowest BLK | |
410 | * overlap point. | |
411 | * | |
412 | * The expectation is that this routine is called multiple times as it | |
413 | * probes for the largest BLK encroachment for any single member DIMM of | |
414 | * the interleave set. Once that value is determined the PMEM-limit for | |
415 | * the set can be established. | |
416 | */ | |
417 | resource_size_t nd_pmem_available_dpa(struct nd_region *nd_region, | |
418 | struct nd_mapping *nd_mapping, resource_size_t *overlap) | |
419 | { | |
420 | resource_size_t map_start, map_end, busy = 0, available, blk_start; | |
421 | struct nvdimm_drvdata *ndd = to_ndd(nd_mapping); | |
422 | struct resource *res; | |
423 | const char *reason; | |
424 | ||
425 | if (!ndd) | |
426 | return 0; | |
427 | ||
428 | map_start = nd_mapping->start; | |
429 | map_end = map_start + nd_mapping->size - 1; | |
430 | blk_start = max(map_start, map_end + 1 - *overlap); | |
431 | for_each_dpa_resource(ndd, res) | |
432 | if (res->start >= map_start && res->start < map_end) { | |
433 | if (strncmp(res->name, "blk", 3) == 0) | |
434 | blk_start = min(blk_start, res->start); | |
435 | else if (res->start != map_start) { | |
436 | reason = "misaligned to iset"; | |
437 | goto err; | |
438 | } else { | |
439 | if (busy) { | |
440 | reason = "duplicate overlapping PMEM reservations?"; | |
441 | goto err; | |
442 | } | |
443 | busy += resource_size(res); | |
444 | continue; | |
445 | } | |
446 | } else if (res->end >= map_start && res->end <= map_end) { | |
447 | if (strncmp(res->name, "blk", 3) == 0) { | |
448 | /* | |
449 | * If a BLK allocation overlaps the start of | |
450 | * PMEM the entire interleave set may now only | |
451 | * be used for BLK. | |
452 | */ | |
453 | blk_start = map_start; | |
454 | } else { | |
455 | reason = "misaligned to iset"; | |
456 | goto err; | |
457 | } | |
458 | } else if (map_start > res->start && map_start < res->end) { | |
459 | /* total eclipse of the mapping */ | |
460 | busy += nd_mapping->size; | |
461 | blk_start = map_start; | |
462 | } | |
463 | ||
464 | *overlap = map_end + 1 - blk_start; | |
465 | available = blk_start - map_start; | |
466 | if (busy < available) | |
467 | return available - busy; | |
468 | return 0; | |
469 | ||
470 | err: | |
471 | /* | |
472 | * Something is wrong, PMEM must align with the start of the | |
473 | * interleave set, and there can only be one allocation per set. | |
474 | */ | |
475 | nd_dbg_dpa(nd_region, ndd, res, "%s\n", reason); | |
476 | return 0; | |
477 | } | |
478 | ||
4a826c83 DW |
479 | void nvdimm_free_dpa(struct nvdimm_drvdata *ndd, struct resource *res) |
480 | { | |
481 | WARN_ON_ONCE(!is_nvdimm_bus_locked(ndd->dev)); | |
482 | kfree(res->name); | |
483 | __release_region(&ndd->dpa, res->start, resource_size(res)); | |
484 | } | |
485 | ||
486 | struct resource *nvdimm_allocate_dpa(struct nvdimm_drvdata *ndd, | |
487 | struct nd_label_id *label_id, resource_size_t start, | |
488 | resource_size_t n) | |
489 | { | |
490 | char *name = kmemdup(label_id, sizeof(*label_id), GFP_KERNEL); | |
491 | struct resource *res; | |
492 | ||
493 | if (!name) | |
494 | return NULL; | |
495 | ||
496 | WARN_ON_ONCE(!is_nvdimm_bus_locked(ndd->dev)); | |
497 | res = __request_region(&ndd->dpa, start, n, name, 0); | |
498 | if (!res) | |
499 | kfree(name); | |
500 | return res; | |
501 | } | |
502 | ||
bf9bccc1 DW |
503 | /** |
504 | * nvdimm_allocated_dpa - sum up the dpa currently allocated to this label_id | |
505 | * @nvdimm: container of dpa-resource-root + labels | |
506 | * @label_id: dpa resource name of the form {pmem|blk}-<human readable uuid> | |
507 | */ | |
508 | resource_size_t nvdimm_allocated_dpa(struct nvdimm_drvdata *ndd, | |
509 | struct nd_label_id *label_id) | |
510 | { | |
511 | resource_size_t allocated = 0; | |
512 | struct resource *res; | |
513 | ||
514 | for_each_dpa_resource(ndd, res) | |
515 | if (strcmp(res->name, label_id->id) == 0) | |
516 | allocated += resource_size(res); | |
517 | ||
518 | return allocated; | |
519 | } | |
520 | ||
4d88a97a DW |
521 | static int count_dimms(struct device *dev, void *c) |
522 | { | |
523 | int *count = c; | |
524 | ||
525 | if (is_nvdimm(dev)) | |
526 | (*count)++; | |
527 | return 0; | |
528 | } | |
529 | ||
530 | int nvdimm_bus_check_dimm_count(struct nvdimm_bus *nvdimm_bus, int dimm_count) | |
531 | { | |
532 | int count = 0; | |
533 | /* Flush any possible dimm registration failures */ | |
534 | nd_synchronize(); | |
535 | ||
536 | device_for_each_child(&nvdimm_bus->dev, &count, count_dimms); | |
537 | dev_dbg(&nvdimm_bus->dev, "%s: count: %d\n", __func__, count); | |
538 | if (count != dimm_count) | |
539 | return -ENXIO; | |
540 | return 0; | |
541 | } | |
542 | EXPORT_SYMBOL_GPL(nvdimm_bus_check_dimm_count); |