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 #include <linux/libnvdimm.h>
14 #include <linux/badblocks.h>
15 #include <linux/export.h>
16 #include <linux/module.h>
17 #include <linux/blkdev.h>
18 #include <linux/device.h>
19 #include <linux/ctype.h>
20 #include <linux/ndctl.h>
21 #include <linux/mutex.h>
22 #include <linux/slab.h>
27 LIST_HEAD(nvdimm_bus_list);
28 DEFINE_MUTEX(nvdimm_bus_list_mutex);
30 void nvdimm_bus_lock(struct device *dev)
32 struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(dev);
36 mutex_lock(&nvdimm_bus->reconfig_mutex);
38 EXPORT_SYMBOL(nvdimm_bus_lock);
40 void nvdimm_bus_unlock(struct device *dev)
42 struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(dev);
46 mutex_unlock(&nvdimm_bus->reconfig_mutex);
48 EXPORT_SYMBOL(nvdimm_bus_unlock);
50 bool is_nvdimm_bus_locked(struct device *dev)
52 struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(dev);
56 return mutex_is_locked(&nvdimm_bus->reconfig_mutex);
58 EXPORT_SYMBOL(is_nvdimm_bus_locked);
61 struct nvdimm_bus *nvdimm_bus;
62 struct list_head list;
63 resource_size_t offset;
73 static struct nvdimm_map *find_nvdimm_map(struct device *dev,
74 resource_size_t offset)
76 struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(dev);
77 struct nvdimm_map *nvdimm_map;
79 list_for_each_entry(nvdimm_map, &nvdimm_bus->mapping_list, list)
80 if (nvdimm_map->offset == offset)
85 static struct nvdimm_map *alloc_nvdimm_map(struct device *dev,
86 resource_size_t offset, size_t size, unsigned long flags)
88 struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(dev);
89 struct nvdimm_map *nvdimm_map;
91 nvdimm_map = kzalloc(sizeof(*nvdimm_map), GFP_KERNEL);
95 INIT_LIST_HEAD(&nvdimm_map->list);
96 nvdimm_map->nvdimm_bus = nvdimm_bus;
97 nvdimm_map->offset = offset;
98 nvdimm_map->flags = flags;
99 nvdimm_map->size = size;
100 kref_init(&nvdimm_map->kref);
102 if (!request_mem_region(offset, size, dev_name(&nvdimm_bus->dev))) {
103 dev_err(&nvdimm_bus->dev, "failed to request %pa + %zd for %s\n",
104 &offset, size, dev_name(dev));
105 goto err_request_region;
109 nvdimm_map->mem = memremap(offset, size, flags);
111 nvdimm_map->iomem = ioremap(offset, size);
113 if (!nvdimm_map->mem)
116 dev_WARN_ONCE(dev, !is_nvdimm_bus_locked(dev), "%s: bus unlocked!",
118 list_add(&nvdimm_map->list, &nvdimm_bus->mapping_list);
123 release_mem_region(offset, size);
129 static void nvdimm_map_release(struct kref *kref)
131 struct nvdimm_bus *nvdimm_bus;
132 struct nvdimm_map *nvdimm_map;
134 nvdimm_map = container_of(kref, struct nvdimm_map, kref);
135 nvdimm_bus = nvdimm_map->nvdimm_bus;
137 dev_dbg(&nvdimm_bus->dev, "%s: %pa\n", __func__, &nvdimm_map->offset);
138 list_del(&nvdimm_map->list);
139 if (nvdimm_map->flags)
140 memunmap(nvdimm_map->mem);
142 iounmap(nvdimm_map->iomem);
143 release_mem_region(nvdimm_map->offset, nvdimm_map->size);
147 static void nvdimm_map_put(void *data)
149 struct nvdimm_map *nvdimm_map = data;
150 struct nvdimm_bus *nvdimm_bus = nvdimm_map->nvdimm_bus;
152 nvdimm_bus_lock(&nvdimm_bus->dev);
153 kref_put(&nvdimm_map->kref, nvdimm_map_release);
154 nvdimm_bus_unlock(&nvdimm_bus->dev);
158 * devm_nvdimm_memremap - map a resource that is shared across regions
159 * @dev: device that will own a reference to the shared mapping
160 * @offset: physical base address of the mapping
161 * @size: mapping size
162 * @flags: memremap flags, or, if zero, perform an ioremap instead
164 void *devm_nvdimm_memremap(struct device *dev, resource_size_t offset,
165 size_t size, unsigned long flags)
167 struct nvdimm_map *nvdimm_map;
169 nvdimm_bus_lock(dev);
170 nvdimm_map = find_nvdimm_map(dev, offset);
172 nvdimm_map = alloc_nvdimm_map(dev, offset, size, flags);
174 kref_get(&nvdimm_map->kref);
175 nvdimm_bus_unlock(dev);
180 if (devm_add_action_or_reset(dev, nvdimm_map_put, nvdimm_map))
183 return nvdimm_map->mem;
185 EXPORT_SYMBOL_GPL(devm_nvdimm_memremap);
187 u64 nd_fletcher64(void *addr, size_t len, bool le)
194 for (i = 0; i < len / sizeof(u32); i++) {
195 lo32 += le ? le32_to_cpu((__le32) buf[i]) : buf[i];
199 return hi32 << 32 | lo32;
201 EXPORT_SYMBOL_GPL(nd_fletcher64);
203 struct nvdimm_bus_descriptor *to_nd_desc(struct nvdimm_bus *nvdimm_bus)
205 /* struct nvdimm_bus definition is private to libnvdimm */
206 return nvdimm_bus->nd_desc;
208 EXPORT_SYMBOL_GPL(to_nd_desc);
210 struct device *to_nvdimm_bus_dev(struct nvdimm_bus *nvdimm_bus)
212 /* struct nvdimm_bus definition is private to libnvdimm */
213 return &nvdimm_bus->dev;
215 EXPORT_SYMBOL_GPL(to_nvdimm_bus_dev);
217 static bool is_uuid_sep(char sep)
219 if (sep == '\n' || sep == '-' || sep == ':' || sep == '\0')
224 static int nd_uuid_parse(struct device *dev, u8 *uuid_out, const char *buf,
227 const char *str = buf;
231 for (i = 0; i < 16; i++) {
232 if (!isxdigit(str[0]) || !isxdigit(str[1])) {
233 dev_dbg(dev, "%s: pos: %d buf[%zd]: %c buf[%zd]: %c\n",
234 __func__, i, str - buf, str[0],
235 str + 1 - buf, str[1]);
239 uuid[i] = (hex_to_bin(str[0]) << 4) | hex_to_bin(str[1]);
241 if (is_uuid_sep(*str))
245 memcpy(uuid_out, uuid, sizeof(uuid));
250 * nd_uuid_store: common implementation for writing 'uuid' sysfs attributes
251 * @dev: container device for the uuid property
252 * @uuid_out: uuid buffer to replace
253 * @buf: raw sysfs buffer to parse
255 * Enforce that uuids can only be changed while the device is disabled
257 * LOCKING: expects device_lock() is held on entry
259 int nd_uuid_store(struct device *dev, u8 **uuid_out, const char *buf,
268 rc = nd_uuid_parse(dev, uuid, buf, len);
273 *uuid_out = kmemdup(uuid, sizeof(uuid), GFP_KERNEL);
280 ssize_t nd_sector_size_show(unsigned long current_lbasize,
281 const unsigned long *supported, char *buf)
286 for (i = 0; supported[i]; i++)
287 if (current_lbasize == supported[i])
288 len += sprintf(buf + len, "[%ld] ", supported[i]);
290 len += sprintf(buf + len, "%ld ", supported[i]);
291 len += sprintf(buf + len, "\n");
295 ssize_t nd_sector_size_store(struct device *dev, const char *buf,
296 unsigned long *current_lbasize, const unsigned long *supported)
298 unsigned long lbasize;
304 rc = kstrtoul(buf, 0, &lbasize);
308 for (i = 0; supported[i]; i++)
309 if (lbasize == supported[i])
313 *current_lbasize = lbasize;
320 static ssize_t commands_show(struct device *dev,
321 struct device_attribute *attr, char *buf)
324 struct nvdimm_bus *nvdimm_bus = to_nvdimm_bus(dev);
325 struct nvdimm_bus_descriptor *nd_desc = nvdimm_bus->nd_desc;
327 for_each_set_bit(cmd, &nd_desc->cmd_mask, BITS_PER_LONG)
328 len += sprintf(buf + len, "%s ", nvdimm_bus_cmd_name(cmd));
329 len += sprintf(buf + len, "\n");
332 static DEVICE_ATTR_RO(commands);
334 static const char *nvdimm_bus_provider(struct nvdimm_bus *nvdimm_bus)
336 struct nvdimm_bus_descriptor *nd_desc = nvdimm_bus->nd_desc;
337 struct device *parent = nvdimm_bus->dev.parent;
339 if (nd_desc->provider_name)
340 return nd_desc->provider_name;
342 return dev_name(parent);
347 static ssize_t provider_show(struct device *dev,
348 struct device_attribute *attr, char *buf)
350 struct nvdimm_bus *nvdimm_bus = to_nvdimm_bus(dev);
352 return sprintf(buf, "%s\n", nvdimm_bus_provider(nvdimm_bus));
354 static DEVICE_ATTR_RO(provider);
356 static int flush_namespaces(struct device *dev, void *data)
363 static int flush_regions_dimms(struct device *dev, void *data)
367 device_for_each_child(dev, NULL, flush_namespaces);
371 static ssize_t wait_probe_show(struct device *dev,
372 struct device_attribute *attr, char *buf)
374 struct nvdimm_bus *nvdimm_bus = to_nvdimm_bus(dev);
375 struct nvdimm_bus_descriptor *nd_desc = nvdimm_bus->nd_desc;
378 if (nd_desc->flush_probe) {
379 rc = nd_desc->flush_probe(nd_desc);
384 device_for_each_child(dev, NULL, flush_regions_dimms);
385 return sprintf(buf, "1\n");
387 static DEVICE_ATTR_RO(wait_probe);
389 static struct attribute *nvdimm_bus_attributes[] = {
390 &dev_attr_commands.attr,
391 &dev_attr_wait_probe.attr,
392 &dev_attr_provider.attr,
396 struct attribute_group nvdimm_bus_attribute_group = {
397 .attrs = nvdimm_bus_attributes,
399 EXPORT_SYMBOL_GPL(nvdimm_bus_attribute_group);
401 static void set_badblock(struct badblocks *bb, sector_t s, int num)
403 dev_dbg(bb->dev, "Found a poison range (0x%llx, 0x%llx)\n",
404 (u64) s * 512, (u64) num * 512);
405 /* this isn't an error as the hardware will still throw an exception */
406 if (badblocks_set(bb, s, num, 1))
407 dev_info_once(bb->dev, "%s: failed for sector %llx\n",
412 * __add_badblock_range() - Convert a physical address range to bad sectors
413 * @bb: badblocks instance to populate
414 * @ns_offset: namespace offset where the error range begins (in bytes)
415 * @len: number of bytes of poison to be added
417 * This assumes that the range provided with (ns_offset, len) is within
418 * the bounds of physical addresses for this namespace, i.e. lies in the
419 * interval [ns_start, ns_start + ns_size)
421 static void __add_badblock_range(struct badblocks *bb, u64 ns_offset, u64 len)
423 const unsigned int sector_size = 512;
424 sector_t start_sector;
428 start_sector = div_u64(ns_offset, sector_size);
429 num_sectors = div_u64_rem(len, sector_size, &rem);
433 if (unlikely(num_sectors > (u64)INT_MAX)) {
434 u64 remaining = num_sectors;
435 sector_t s = start_sector;
438 int done = min_t(u64, remaining, INT_MAX);
440 set_badblock(bb, s, done);
445 set_badblock(bb, start_sector, num_sectors);
448 static void badblocks_populate(struct list_head *poison_list,
449 struct badblocks *bb, const struct resource *res)
451 struct nd_poison *pl;
453 if (list_empty(poison_list))
456 list_for_each_entry(pl, poison_list, list) {
457 u64 pl_end = pl->start + pl->length - 1;
459 /* Discard intervals with no intersection */
460 if (pl_end < res->start)
462 if (pl->start > res->end)
464 /* Deal with any overlap after start of the namespace */
465 if (pl->start >= res->start) {
466 u64 start = pl->start;
469 if (pl_end <= res->end)
472 len = res->start + resource_size(res)
474 __add_badblock_range(bb, start - res->start, len);
477 /* Deal with overlap for poison starting before the namespace */
478 if (pl->start < res->start) {
481 if (pl_end < res->end)
482 len = pl->start + pl->length - res->start;
484 len = resource_size(res);
485 __add_badblock_range(bb, 0, len);
491 * nvdimm_badblocks_populate() - Convert a list of poison ranges to badblocks
492 * @region: parent region of the range to interrogate
493 * @bb: badblocks instance to populate
494 * @res: resource range to consider
496 * The poison list generated during bus initialization may contain
497 * multiple, possibly overlapping physical address ranges. Compare each
498 * of these ranges to the resource range currently being initialized,
499 * and add badblocks entries for all matching sub-ranges
501 void nvdimm_badblocks_populate(struct nd_region *nd_region,
502 struct badblocks *bb, const struct resource *res)
504 struct nvdimm_bus *nvdimm_bus;
505 struct list_head *poison_list;
507 if (!is_memory(&nd_region->dev)) {
508 dev_WARN_ONCE(&nd_region->dev, 1,
509 "%s only valid for pmem regions\n", __func__);
512 nvdimm_bus = walk_to_nvdimm_bus(&nd_region->dev);
513 poison_list = &nvdimm_bus->poison_list;
515 nvdimm_bus_lock(&nvdimm_bus->dev);
516 badblocks_populate(poison_list, bb, res);
517 nvdimm_bus_unlock(&nvdimm_bus->dev);
519 EXPORT_SYMBOL_GPL(nvdimm_badblocks_populate);
521 static void append_poison_entry(struct nvdimm_bus *nvdimm_bus,
522 struct nd_poison *pl, u64 addr, u64 length)
524 lockdep_assert_held(&nvdimm_bus->poison_lock);
527 list_add_tail(&pl->list, &nvdimm_bus->poison_list);
530 static int add_poison(struct nvdimm_bus *nvdimm_bus, u64 addr, u64 length,
533 struct nd_poison *pl;
535 pl = kzalloc(sizeof(*pl), flags);
539 append_poison_entry(nvdimm_bus, pl, addr, length);
543 static int bus_add_poison(struct nvdimm_bus *nvdimm_bus, u64 addr, u64 length)
545 struct nd_poison *pl, *pl_new;
547 spin_unlock(&nvdimm_bus->poison_lock);
548 pl_new = kzalloc(sizeof(*pl_new), GFP_KERNEL);
549 spin_lock(&nvdimm_bus->poison_lock);
551 if (list_empty(&nvdimm_bus->poison_list)) {
554 append_poison_entry(nvdimm_bus, pl_new, addr, length);
559 * There is a chance this is a duplicate, check for those first.
560 * This will be the common case as ARS_STATUS returns all known
561 * errors in the SPA space, and we can't query it per region
563 list_for_each_entry(pl, &nvdimm_bus->poison_list, list)
564 if (pl->start == addr) {
565 /* If length has changed, update this list entry */
566 if (pl->length != length)
573 * If not a duplicate or a simple length update, add the entry as is,
574 * as any overlapping ranges will get resolved when the list is consumed
575 * and converted to badblocks
579 append_poison_entry(nvdimm_bus, pl_new, addr, length);
584 int nvdimm_bus_add_poison(struct nvdimm_bus *nvdimm_bus, u64 addr, u64 length)
588 spin_lock(&nvdimm_bus->poison_lock);
589 rc = bus_add_poison(nvdimm_bus, addr, length);
590 spin_unlock(&nvdimm_bus->poison_lock);
594 EXPORT_SYMBOL_GPL(nvdimm_bus_add_poison);
596 void nvdimm_forget_poison(struct nvdimm_bus *nvdimm_bus, phys_addr_t start,
599 struct list_head *poison_list = &nvdimm_bus->poison_list;
600 u64 clr_end = start + len - 1;
601 struct nd_poison *pl, *next;
603 spin_lock(&nvdimm_bus->poison_lock);
604 WARN_ON_ONCE(list_empty(poison_list));
607 * [start, clr_end] is the poison interval being cleared.
608 * [pl->start, pl_end] is the poison_list entry we're comparing
609 * the above interval against. The poison list entry may need
610 * to be modified (update either start or length), deleted, or
611 * split into two based on the overlap characteristics
614 list_for_each_entry_safe(pl, next, poison_list, list) {
615 u64 pl_end = pl->start + pl->length - 1;
617 /* Skip intervals with no intersection */
620 if (pl->start > clr_end)
622 /* Delete completely overlapped poison entries */
623 if ((pl->start >= start) && (pl_end <= clr_end)) {
628 /* Adjust start point of partially cleared entries */
629 if ((start <= pl->start) && (clr_end > pl->start)) {
630 pl->length -= clr_end - pl->start + 1;
631 pl->start = clr_end + 1;
634 /* Adjust pl->length for partial clearing at the tail end */
635 if ((pl->start < start) && (pl_end <= clr_end)) {
636 /* pl->start remains the same */
637 pl->length = start - pl->start;
641 * If clearing in the middle of an entry, we split it into
642 * two by modifying the current entry to represent one half of
643 * the split, and adding a new entry for the second half.
645 if ((pl->start < start) && (pl_end > clr_end)) {
646 u64 new_start = clr_end + 1;
647 u64 new_len = pl_end - new_start + 1;
649 /* Add new entry covering the right half */
650 add_poison(nvdimm_bus, new_start, new_len, GFP_NOWAIT);
651 /* Adjust this entry to cover the left half */
652 pl->length = start - pl->start;
656 spin_unlock(&nvdimm_bus->poison_lock);
658 EXPORT_SYMBOL_GPL(nvdimm_forget_poison);
660 #ifdef CONFIG_BLK_DEV_INTEGRITY
661 int nd_integrity_init(struct gendisk *disk, unsigned long meta_size)
663 struct blk_integrity bi;
668 memset(&bi, 0, sizeof(bi));
670 bi.tuple_size = meta_size;
671 bi.tag_size = meta_size;
673 blk_integrity_register(disk, &bi);
674 blk_queue_max_integrity_segments(disk->queue, 1);
678 EXPORT_SYMBOL(nd_integrity_init);
680 #else /* CONFIG_BLK_DEV_INTEGRITY */
681 int nd_integrity_init(struct gendisk *disk, unsigned long meta_size)
685 EXPORT_SYMBOL(nd_integrity_init);
689 static __init int libnvdimm_init(void)
693 rc = nvdimm_bus_init();
699 rc = nd_region_init();
713 static __exit void libnvdimm_exit(void)
715 WARN_ON(!list_empty(&nvdimm_bus_list));
719 nd_region_devs_exit();
723 MODULE_LICENSE("GPL v2");
724 MODULE_AUTHOR("Intel Corporation");
725 subsys_initcall(libnvdimm_init);
726 module_exit(libnvdimm_exit);