2 * Copyright(c) 2013-2015 Intel Corporation. All rights reserved.
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.
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.
13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
14 #include <linux/vmalloc.h>
15 #include <linux/device.h>
16 #include <linux/ndctl.h>
17 #include <linux/slab.h>
25 static DEFINE_IDA(dimm_ida);
28 * Retrieve bus and dimm handle and return if this bus supports
29 * get_config_data commands
31 int nvdimm_check_config_data(struct device *dev)
33 struct nvdimm *nvdimm = to_nvdimm(dev);
35 if (!nvdimm->cmd_mask ||
36 !test_bit(ND_CMD_GET_CONFIG_DATA, &nvdimm->cmd_mask)) {
37 if (test_bit(NDD_ALIASING, &nvdimm->flags))
46 static int validate_dimm(struct nvdimm_drvdata *ndd)
53 rc = nvdimm_check_config_data(ndd->dev);
55 dev_dbg(ndd->dev, "%pf: %s error: %d\n",
56 __builtin_return_address(0), __func__, rc);
61 * nvdimm_init_nsarea - determine the geometry of a dimm's namespace area
62 * @nvdimm: dimm to initialize
64 int nvdimm_init_nsarea(struct nvdimm_drvdata *ndd)
66 struct nd_cmd_get_config_size *cmd = &ndd->nsarea;
67 struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(ndd->dev);
68 struct nvdimm_bus_descriptor *nd_desc;
69 int rc = validate_dimm(ndd);
76 return 0; /* already valid */
78 memset(cmd, 0, sizeof(*cmd));
79 nd_desc = nvdimm_bus->nd_desc;
80 rc = nd_desc->ndctl(nd_desc, to_nvdimm(ndd->dev),
81 ND_CMD_GET_CONFIG_SIZE, cmd, sizeof(*cmd), &cmd_rc);
87 int nvdimm_init_config_data(struct nvdimm_drvdata *ndd)
89 struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(ndd->dev);
90 struct nd_cmd_get_config_data_hdr *cmd;
91 struct nvdimm_bus_descriptor *nd_desc;
92 int rc = validate_dimm(ndd);
93 u32 max_cmd_size, config_size;
102 if (ndd->nsarea.status || ndd->nsarea.max_xfer == 0
103 || ndd->nsarea.config_size < ND_LABEL_MIN_SIZE) {
104 dev_dbg(ndd->dev, "failed to init config data area: (%d:%d)\n",
105 ndd->nsarea.max_xfer, ndd->nsarea.config_size);
109 ndd->data = kvmalloc(ndd->nsarea.config_size, GFP_KERNEL);
113 max_cmd_size = min_t(u32, PAGE_SIZE, ndd->nsarea.max_xfer);
114 cmd = kzalloc(max_cmd_size + sizeof(*cmd), GFP_KERNEL);
118 nd_desc = nvdimm_bus->nd_desc;
119 for (config_size = ndd->nsarea.config_size, offset = 0;
120 config_size; config_size -= cmd->in_length,
121 offset += cmd->in_length) {
122 cmd->in_length = min(config_size, max_cmd_size);
123 cmd->in_offset = offset;
124 rc = nd_desc->ndctl(nd_desc, to_nvdimm(ndd->dev),
125 ND_CMD_GET_CONFIG_DATA, cmd,
126 cmd->in_length + sizeof(*cmd), NULL);
127 if (rc || cmd->status) {
131 memcpy(ndd->data + offset, cmd->out_buf, cmd->in_length);
133 dev_dbg(ndd->dev, "%s: len: %zu rc: %d\n", __func__, offset, rc);
139 int nvdimm_set_config_data(struct nvdimm_drvdata *ndd, size_t offset,
140 void *buf, size_t len)
142 int rc = validate_dimm(ndd);
143 size_t max_cmd_size, buf_offset;
144 struct nd_cmd_set_config_hdr *cmd;
145 struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(ndd->dev);
146 struct nvdimm_bus_descriptor *nd_desc = nvdimm_bus->nd_desc;
154 if (offset + len > ndd->nsarea.config_size)
157 max_cmd_size = min_t(u32, PAGE_SIZE, len);
158 max_cmd_size = min_t(u32, max_cmd_size, ndd->nsarea.max_xfer);
159 cmd = kzalloc(max_cmd_size + sizeof(*cmd) + sizeof(u32), GFP_KERNEL);
163 for (buf_offset = 0; len; len -= cmd->in_length,
164 buf_offset += cmd->in_length) {
168 cmd->in_offset = offset + buf_offset;
169 cmd->in_length = min(max_cmd_size, len);
170 memcpy(cmd->in_buf, buf + buf_offset, cmd->in_length);
172 /* status is output in the last 4-bytes of the command buffer */
173 cmd_size = sizeof(*cmd) + cmd->in_length + sizeof(u32);
174 status = ((void *) cmd) + cmd_size - sizeof(u32);
176 rc = nd_desc->ndctl(nd_desc, to_nvdimm(ndd->dev),
177 ND_CMD_SET_CONFIG_DATA, cmd, cmd_size, NULL);
179 rc = rc ? rc : -ENXIO;
188 void nvdimm_set_aliasing(struct device *dev)
190 struct nvdimm *nvdimm = to_nvdimm(dev);
192 set_bit(NDD_ALIASING, &nvdimm->flags);
195 void nvdimm_set_locked(struct device *dev)
197 struct nvdimm *nvdimm = to_nvdimm(dev);
199 set_bit(NDD_LOCKED, &nvdimm->flags);
202 static void nvdimm_release(struct device *dev)
204 struct nvdimm *nvdimm = to_nvdimm(dev);
206 ida_simple_remove(&dimm_ida, nvdimm->id);
210 static struct device_type nvdimm_device_type = {
212 .release = nvdimm_release,
215 bool is_nvdimm(struct device *dev)
217 return dev->type == &nvdimm_device_type;
220 struct nvdimm *to_nvdimm(struct device *dev)
222 struct nvdimm *nvdimm = container_of(dev, struct nvdimm, dev);
224 WARN_ON(!is_nvdimm(dev));
227 EXPORT_SYMBOL_GPL(to_nvdimm);
229 struct nvdimm *nd_blk_region_to_dimm(struct nd_blk_region *ndbr)
231 struct nd_region *nd_region = &ndbr->nd_region;
232 struct nd_mapping *nd_mapping = &nd_region->mapping[0];
234 return nd_mapping->nvdimm;
236 EXPORT_SYMBOL_GPL(nd_blk_region_to_dimm);
238 struct nvdimm_drvdata *to_ndd(struct nd_mapping *nd_mapping)
240 struct nvdimm *nvdimm = nd_mapping->nvdimm;
242 WARN_ON_ONCE(!is_nvdimm_bus_locked(&nvdimm->dev));
244 return dev_get_drvdata(&nvdimm->dev);
246 EXPORT_SYMBOL(to_ndd);
248 void nvdimm_drvdata_release(struct kref *kref)
250 struct nvdimm_drvdata *ndd = container_of(kref, typeof(*ndd), kref);
251 struct device *dev = ndd->dev;
252 struct resource *res, *_r;
254 dev_dbg(dev, "%s\n", __func__);
256 nvdimm_bus_lock(dev);
257 for_each_dpa_resource_safe(ndd, res, _r)
258 nvdimm_free_dpa(ndd, res);
259 nvdimm_bus_unlock(dev);
266 void get_ndd(struct nvdimm_drvdata *ndd)
268 kref_get(&ndd->kref);
271 void put_ndd(struct nvdimm_drvdata *ndd)
274 kref_put(&ndd->kref, nvdimm_drvdata_release);
277 const char *nvdimm_name(struct nvdimm *nvdimm)
279 return dev_name(&nvdimm->dev);
281 EXPORT_SYMBOL_GPL(nvdimm_name);
283 struct kobject *nvdimm_kobj(struct nvdimm *nvdimm)
285 return &nvdimm->dev.kobj;
287 EXPORT_SYMBOL_GPL(nvdimm_kobj);
289 unsigned long nvdimm_cmd_mask(struct nvdimm *nvdimm)
291 return nvdimm->cmd_mask;
293 EXPORT_SYMBOL_GPL(nvdimm_cmd_mask);
295 void *nvdimm_provider_data(struct nvdimm *nvdimm)
298 return nvdimm->provider_data;
301 EXPORT_SYMBOL_GPL(nvdimm_provider_data);
303 static ssize_t commands_show(struct device *dev,
304 struct device_attribute *attr, char *buf)
306 struct nvdimm *nvdimm = to_nvdimm(dev);
309 if (!nvdimm->cmd_mask)
310 return sprintf(buf, "\n");
312 for_each_set_bit(cmd, &nvdimm->cmd_mask, BITS_PER_LONG)
313 len += sprintf(buf + len, "%s ", nvdimm_cmd_name(cmd));
314 len += sprintf(buf + len, "\n");
317 static DEVICE_ATTR_RO(commands);
319 static ssize_t state_show(struct device *dev, struct device_attribute *attr,
322 struct nvdimm *nvdimm = to_nvdimm(dev);
325 * The state may be in the process of changing, userspace should
326 * quiesce probing if it wants a static answer
328 nvdimm_bus_lock(dev);
329 nvdimm_bus_unlock(dev);
330 return sprintf(buf, "%s\n", atomic_read(&nvdimm->busy)
331 ? "active" : "idle");
333 static DEVICE_ATTR_RO(state);
335 static ssize_t available_slots_show(struct device *dev,
336 struct device_attribute *attr, char *buf)
338 struct nvdimm_drvdata *ndd = dev_get_drvdata(dev);
345 nvdimm_bus_lock(dev);
346 nfree = nd_label_nfree(ndd);
347 if (nfree - 1 > nfree) {
348 dev_WARN_ONCE(dev, 1, "we ate our last label?\n");
352 rc = sprintf(buf, "%d\n", nfree);
353 nvdimm_bus_unlock(dev);
356 static DEVICE_ATTR_RO(available_slots);
358 static struct attribute *nvdimm_attributes[] = {
359 &dev_attr_state.attr,
360 &dev_attr_commands.attr,
361 &dev_attr_available_slots.attr,
365 struct attribute_group nvdimm_attribute_group = {
366 .attrs = nvdimm_attributes,
368 EXPORT_SYMBOL_GPL(nvdimm_attribute_group);
370 struct nvdimm *nvdimm_create(struct nvdimm_bus *nvdimm_bus, void *provider_data,
371 const struct attribute_group **groups, unsigned long flags,
372 unsigned long cmd_mask, int num_flush,
373 struct resource *flush_wpq)
375 struct nvdimm *nvdimm = kzalloc(sizeof(*nvdimm), GFP_KERNEL);
381 nvdimm->id = ida_simple_get(&dimm_ida, 0, 0, GFP_KERNEL);
382 if (nvdimm->id < 0) {
386 nvdimm->provider_data = provider_data;
387 nvdimm->flags = flags;
388 nvdimm->cmd_mask = cmd_mask;
389 nvdimm->num_flush = num_flush;
390 nvdimm->flush_wpq = flush_wpq;
391 atomic_set(&nvdimm->busy, 0);
393 dev_set_name(dev, "nmem%d", nvdimm->id);
394 dev->parent = &nvdimm_bus->dev;
395 dev->type = &nvdimm_device_type;
396 dev->devt = MKDEV(nvdimm_major, nvdimm->id);
397 dev->groups = groups;
398 nd_device_register(dev);
402 EXPORT_SYMBOL_GPL(nvdimm_create);
404 int alias_dpa_busy(struct device *dev, void *data)
406 resource_size_t map_end, blk_start, new;
407 struct blk_alloc_info *info = data;
408 struct nd_mapping *nd_mapping;
409 struct nd_region *nd_region;
410 struct nvdimm_drvdata *ndd;
411 struct resource *res;
414 if (!is_nd_pmem(dev))
417 nd_region = to_nd_region(dev);
418 for (i = 0; i < nd_region->ndr_mappings; i++) {
419 nd_mapping = &nd_region->mapping[i];
420 if (nd_mapping->nvdimm == info->nd_mapping->nvdimm)
424 if (i >= nd_region->ndr_mappings)
427 ndd = to_ndd(nd_mapping);
428 map_end = nd_mapping->start + nd_mapping->size - 1;
429 blk_start = nd_mapping->start;
432 * In the allocation case ->res is set to free space that we are
433 * looking to validate against PMEM aliasing collision rules
434 * (i.e. BLK is allocated after all aliased PMEM).
437 if (info->res->start >= nd_mapping->start
438 && info->res->start < map_end)
446 * Find the free dpa from the end of the last pmem allocation to
447 * the end of the interleave-set mapping.
449 for_each_dpa_resource(ndd, res) {
450 if (strncmp(res->name, "pmem", 4) != 0)
452 if ((res->start >= blk_start && res->start < map_end)
453 || (res->end >= blk_start
454 && res->end <= map_end)) {
455 new = max(blk_start, min(map_end + 1, res->end + 1));
456 if (new != blk_start) {
463 /* update the free space range with the probed blk_start */
464 if (info->res && blk_start > info->res->start) {
465 info->res->start = max(info->res->start, blk_start);
466 if (info->res->start > info->res->end)
467 info->res->end = info->res->start - 1;
471 info->available -= blk_start - nd_mapping->start;
477 * nd_blk_available_dpa - account the unused dpa of BLK region
478 * @nd_mapping: container of dpa-resource-root + labels
480 * Unlike PMEM, BLK namespaces can occupy discontiguous DPA ranges, but
481 * we arrange for them to never start at an lower dpa than the last
482 * PMEM allocation in an aliased region.
484 resource_size_t nd_blk_available_dpa(struct nd_region *nd_region)
486 struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(&nd_region->dev);
487 struct nd_mapping *nd_mapping = &nd_region->mapping[0];
488 struct nvdimm_drvdata *ndd = to_ndd(nd_mapping);
489 struct blk_alloc_info info = {
490 .nd_mapping = nd_mapping,
491 .available = nd_mapping->size,
494 struct resource *res;
499 device_for_each_child(&nvdimm_bus->dev, &info, alias_dpa_busy);
501 /* now account for busy blk allocations in unaliased dpa */
502 for_each_dpa_resource(ndd, res) {
503 if (strncmp(res->name, "blk", 3) != 0)
505 info.available -= resource_size(res);
508 return info.available;
512 * nd_pmem_available_dpa - for the given dimm+region account unallocated dpa
513 * @nd_mapping: container of dpa-resource-root + labels
514 * @nd_region: constrain available space check to this reference region
515 * @overlap: calculate available space assuming this level of overlap
517 * Validate that a PMEM label, if present, aligns with the start of an
518 * interleave set and truncate the available size at the lowest BLK
521 * The expectation is that this routine is called multiple times as it
522 * probes for the largest BLK encroachment for any single member DIMM of
523 * the interleave set. Once that value is determined the PMEM-limit for
524 * the set can be established.
526 resource_size_t nd_pmem_available_dpa(struct nd_region *nd_region,
527 struct nd_mapping *nd_mapping, resource_size_t *overlap)
529 resource_size_t map_start, map_end, busy = 0, available, blk_start;
530 struct nvdimm_drvdata *ndd = to_ndd(nd_mapping);
531 struct resource *res;
537 map_start = nd_mapping->start;
538 map_end = map_start + nd_mapping->size - 1;
539 blk_start = max(map_start, map_end + 1 - *overlap);
540 for_each_dpa_resource(ndd, res) {
541 if (res->start >= map_start && res->start < map_end) {
542 if (strncmp(res->name, "blk", 3) == 0)
543 blk_start = min(blk_start,
544 max(map_start, res->start));
545 else if (res->end > map_end) {
546 reason = "misaligned to iset";
549 busy += resource_size(res);
550 } else if (res->end >= map_start && res->end <= map_end) {
551 if (strncmp(res->name, "blk", 3) == 0) {
553 * If a BLK allocation overlaps the start of
554 * PMEM the entire interleave set may now only
557 blk_start = map_start;
559 busy += resource_size(res);
560 } else if (map_start > res->start && map_start < res->end) {
561 /* total eclipse of the mapping */
562 busy += nd_mapping->size;
563 blk_start = map_start;
567 *overlap = map_end + 1 - blk_start;
568 available = blk_start - map_start;
569 if (busy < available)
570 return available - busy;
574 nd_dbg_dpa(nd_region, ndd, res, "%s\n", reason);
578 void nvdimm_free_dpa(struct nvdimm_drvdata *ndd, struct resource *res)
580 WARN_ON_ONCE(!is_nvdimm_bus_locked(ndd->dev));
582 __release_region(&ndd->dpa, res->start, resource_size(res));
585 struct resource *nvdimm_allocate_dpa(struct nvdimm_drvdata *ndd,
586 struct nd_label_id *label_id, resource_size_t start,
589 char *name = kmemdup(label_id, sizeof(*label_id), GFP_KERNEL);
590 struct resource *res;
595 WARN_ON_ONCE(!is_nvdimm_bus_locked(ndd->dev));
596 res = __request_region(&ndd->dpa, start, n, name, 0);
603 * nvdimm_allocated_dpa - sum up the dpa currently allocated to this label_id
604 * @nvdimm: container of dpa-resource-root + labels
605 * @label_id: dpa resource name of the form {pmem|blk}-<human readable uuid>
607 resource_size_t nvdimm_allocated_dpa(struct nvdimm_drvdata *ndd,
608 struct nd_label_id *label_id)
610 resource_size_t allocated = 0;
611 struct resource *res;
613 for_each_dpa_resource(ndd, res)
614 if (strcmp(res->name, label_id->id) == 0)
615 allocated += resource_size(res);
620 static int count_dimms(struct device *dev, void *c)
629 int nvdimm_bus_check_dimm_count(struct nvdimm_bus *nvdimm_bus, int dimm_count)
632 /* Flush any possible dimm registration failures */
635 device_for_each_child(&nvdimm_bus->dev, &count, count_dimms);
636 dev_dbg(&nvdimm_bus->dev, "%s: count: %d\n", __func__, count);
637 if (count != dimm_count)
641 EXPORT_SYMBOL_GPL(nvdimm_bus_check_dimm_count);
643 void __exit nvdimm_devs_exit(void)
645 ida_destroy(&dimm_ida);