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 goto err_request_region
;
106 nvdimm_map
->mem
= memremap(offset
, size
, flags
);
108 nvdimm_map
->iomem
= ioremap(offset
, size
);
110 if (!nvdimm_map
->mem
)
113 dev_WARN_ONCE(dev
, !is_nvdimm_bus_locked(dev
), "%s: bus unlocked!",
115 list_add(&nvdimm_map
->list
, &nvdimm_bus
->mapping_list
);
120 release_mem_region(offset
, size
);
126 static void nvdimm_map_release(struct kref
*kref
)
128 struct nvdimm_bus
*nvdimm_bus
;
129 struct nvdimm_map
*nvdimm_map
;
131 nvdimm_map
= container_of(kref
, struct nvdimm_map
, kref
);
132 nvdimm_bus
= nvdimm_map
->nvdimm_bus
;
134 dev_dbg(&nvdimm_bus
->dev
, "%s: %pa\n", __func__
, &nvdimm_map
->offset
);
135 list_del(&nvdimm_map
->list
);
136 if (nvdimm_map
->flags
)
137 memunmap(nvdimm_map
->mem
);
139 iounmap(nvdimm_map
->iomem
);
140 release_mem_region(nvdimm_map
->offset
, nvdimm_map
->size
);
144 static void nvdimm_map_put(void *data
)
146 struct nvdimm_map
*nvdimm_map
= data
;
147 struct nvdimm_bus
*nvdimm_bus
= nvdimm_map
->nvdimm_bus
;
149 nvdimm_bus_lock(&nvdimm_bus
->dev
);
150 kref_put(&nvdimm_map
->kref
, nvdimm_map_release
);
151 nvdimm_bus_unlock(&nvdimm_bus
->dev
);
155 * devm_nvdimm_memremap - map a resource that is shared across regions
156 * @dev: device that will own a reference to the shared mapping
157 * @offset: physical base address of the mapping
158 * @size: mapping size
159 * @flags: memremap flags, or, if zero, perform an ioremap instead
161 void *devm_nvdimm_memremap(struct device
*dev
, resource_size_t offset
,
162 size_t size
, unsigned long flags
)
164 struct nvdimm_map
*nvdimm_map
;
166 nvdimm_bus_lock(dev
);
167 nvdimm_map
= find_nvdimm_map(dev
, offset
);
169 nvdimm_map
= alloc_nvdimm_map(dev
, offset
, size
, flags
);
171 kref_get(&nvdimm_map
->kref
);
172 nvdimm_bus_unlock(dev
);
174 if (devm_add_action_or_reset(dev
, nvdimm_map_put
, nvdimm_map
))
177 return nvdimm_map
->mem
;
179 EXPORT_SYMBOL_GPL(devm_nvdimm_memremap
);
181 u64
nd_fletcher64(void *addr
, size_t len
, bool le
)
188 for (i
= 0; i
< len
/ sizeof(u32
); i
++) {
189 lo32
+= le
? le32_to_cpu((__le32
) buf
[i
]) : buf
[i
];
193 return hi32
<< 32 | lo32
;
195 EXPORT_SYMBOL_GPL(nd_fletcher64
);
197 struct nvdimm_bus_descriptor
*to_nd_desc(struct nvdimm_bus
*nvdimm_bus
)
199 /* struct nvdimm_bus definition is private to libnvdimm */
200 return nvdimm_bus
->nd_desc
;
202 EXPORT_SYMBOL_GPL(to_nd_desc
);
204 struct device
*to_nvdimm_bus_dev(struct nvdimm_bus
*nvdimm_bus
)
206 /* struct nvdimm_bus definition is private to libnvdimm */
207 return &nvdimm_bus
->dev
;
209 EXPORT_SYMBOL_GPL(to_nvdimm_bus_dev
);
211 static bool is_uuid_sep(char sep
)
213 if (sep
== '\n' || sep
== '-' || sep
== ':' || sep
== '\0')
218 static int nd_uuid_parse(struct device
*dev
, u8
*uuid_out
, const char *buf
,
221 const char *str
= buf
;
225 for (i
= 0; i
< 16; i
++) {
226 if (!isxdigit(str
[0]) || !isxdigit(str
[1])) {
227 dev_dbg(dev
, "%s: pos: %d buf[%zd]: %c buf[%zd]: %c\n",
228 __func__
, i
, str
- buf
, str
[0],
229 str
+ 1 - buf
, str
[1]);
233 uuid
[i
] = (hex_to_bin(str
[0]) << 4) | hex_to_bin(str
[1]);
235 if (is_uuid_sep(*str
))
239 memcpy(uuid_out
, uuid
, sizeof(uuid
));
244 * nd_uuid_store: common implementation for writing 'uuid' sysfs attributes
245 * @dev: container device for the uuid property
246 * @uuid_out: uuid buffer to replace
247 * @buf: raw sysfs buffer to parse
249 * Enforce that uuids can only be changed while the device is disabled
251 * LOCKING: expects device_lock() is held on entry
253 int nd_uuid_store(struct device
*dev
, u8
**uuid_out
, const char *buf
,
262 rc
= nd_uuid_parse(dev
, uuid
, buf
, len
);
267 *uuid_out
= kmemdup(uuid
, sizeof(uuid
), GFP_KERNEL
);
274 ssize_t
nd_sector_size_show(unsigned long current_lbasize
,
275 const unsigned long *supported
, char *buf
)
280 for (i
= 0; supported
[i
]; i
++)
281 if (current_lbasize
== supported
[i
])
282 len
+= sprintf(buf
+ len
, "[%ld] ", supported
[i
]);
284 len
+= sprintf(buf
+ len
, "%ld ", supported
[i
]);
285 len
+= sprintf(buf
+ len
, "\n");
289 ssize_t
nd_sector_size_store(struct device
*dev
, const char *buf
,
290 unsigned long *current_lbasize
, const unsigned long *supported
)
292 unsigned long lbasize
;
298 rc
= kstrtoul(buf
, 0, &lbasize
);
302 for (i
= 0; supported
[i
]; i
++)
303 if (lbasize
== supported
[i
])
307 *current_lbasize
= lbasize
;
314 void __nd_iostat_start(struct bio
*bio
, unsigned long *start
)
316 struct gendisk
*disk
= bio
->bi_bdev
->bd_disk
;
317 const int rw
= bio_data_dir(bio
);
318 int cpu
= part_stat_lock();
321 part_round_stats(cpu
, &disk
->part0
);
322 part_stat_inc(cpu
, &disk
->part0
, ios
[rw
]);
323 part_stat_add(cpu
, &disk
->part0
, sectors
[rw
], bio_sectors(bio
));
324 part_inc_in_flight(&disk
->part0
, rw
);
327 EXPORT_SYMBOL(__nd_iostat_start
);
329 void nd_iostat_end(struct bio
*bio
, unsigned long start
)
331 struct gendisk
*disk
= bio
->bi_bdev
->bd_disk
;
332 unsigned long duration
= jiffies
- start
;
333 const int rw
= bio_data_dir(bio
);
334 int cpu
= part_stat_lock();
336 part_stat_add(cpu
, &disk
->part0
, ticks
[rw
], duration
);
337 part_round_stats(cpu
, &disk
->part0
);
338 part_dec_in_flight(&disk
->part0
, rw
);
341 EXPORT_SYMBOL(nd_iostat_end
);
343 static ssize_t
commands_show(struct device
*dev
,
344 struct device_attribute
*attr
, char *buf
)
347 struct nvdimm_bus
*nvdimm_bus
= to_nvdimm_bus(dev
);
348 struct nvdimm_bus_descriptor
*nd_desc
= nvdimm_bus
->nd_desc
;
350 for_each_set_bit(cmd
, &nd_desc
->cmd_mask
, BITS_PER_LONG
)
351 len
+= sprintf(buf
+ len
, "%s ", nvdimm_bus_cmd_name(cmd
));
352 len
+= sprintf(buf
+ len
, "\n");
355 static DEVICE_ATTR_RO(commands
);
357 static const char *nvdimm_bus_provider(struct nvdimm_bus
*nvdimm_bus
)
359 struct nvdimm_bus_descriptor
*nd_desc
= nvdimm_bus
->nd_desc
;
360 struct device
*parent
= nvdimm_bus
->dev
.parent
;
362 if (nd_desc
->provider_name
)
363 return nd_desc
->provider_name
;
365 return dev_name(parent
);
370 static ssize_t
provider_show(struct device
*dev
,
371 struct device_attribute
*attr
, char *buf
)
373 struct nvdimm_bus
*nvdimm_bus
= to_nvdimm_bus(dev
);
375 return sprintf(buf
, "%s\n", nvdimm_bus_provider(nvdimm_bus
));
377 static DEVICE_ATTR_RO(provider
);
379 static int flush_namespaces(struct device
*dev
, void *data
)
386 static int flush_regions_dimms(struct device
*dev
, void *data
)
390 device_for_each_child(dev
, NULL
, flush_namespaces
);
394 static ssize_t
wait_probe_show(struct device
*dev
,
395 struct device_attribute
*attr
, char *buf
)
397 struct nvdimm_bus
*nvdimm_bus
= to_nvdimm_bus(dev
);
398 struct nvdimm_bus_descriptor
*nd_desc
= nvdimm_bus
->nd_desc
;
401 if (nd_desc
->flush_probe
) {
402 rc
= nd_desc
->flush_probe(nd_desc
);
407 device_for_each_child(dev
, NULL
, flush_regions_dimms
);
408 return sprintf(buf
, "1\n");
410 static DEVICE_ATTR_RO(wait_probe
);
412 static struct attribute
*nvdimm_bus_attributes
[] = {
413 &dev_attr_commands
.attr
,
414 &dev_attr_wait_probe
.attr
,
415 &dev_attr_provider
.attr
,
419 struct attribute_group nvdimm_bus_attribute_group
= {
420 .attrs
= nvdimm_bus_attributes
,
422 EXPORT_SYMBOL_GPL(nvdimm_bus_attribute_group
);
424 static void set_badblock(struct badblocks
*bb
, sector_t s
, int num
)
426 dev_dbg(bb
->dev
, "Found a poison range (0x%llx, 0x%llx)\n",
427 (u64
) s
* 512, (u64
) num
* 512);
428 /* this isn't an error as the hardware will still throw an exception */
429 if (badblocks_set(bb
, s
, num
, 1))
430 dev_info_once(bb
->dev
, "%s: failed for sector %llx\n",
435 * __add_badblock_range() - Convert a physical address range to bad sectors
436 * @bb: badblocks instance to populate
437 * @ns_offset: namespace offset where the error range begins (in bytes)
438 * @len: number of bytes of poison to be added
440 * This assumes that the range provided with (ns_offset, len) is within
441 * the bounds of physical addresses for this namespace, i.e. lies in the
442 * interval [ns_start, ns_start + ns_size)
444 static void __add_badblock_range(struct badblocks
*bb
, u64 ns_offset
, u64 len
)
446 const unsigned int sector_size
= 512;
447 sector_t start_sector
;
451 start_sector
= div_u64(ns_offset
, sector_size
);
452 num_sectors
= div_u64_rem(len
, sector_size
, &rem
);
456 if (unlikely(num_sectors
> (u64
)INT_MAX
)) {
457 u64 remaining
= num_sectors
;
458 sector_t s
= start_sector
;
461 int done
= min_t(u64
, remaining
, INT_MAX
);
463 set_badblock(bb
, s
, done
);
468 set_badblock(bb
, start_sector
, num_sectors
);
471 static void badblocks_populate(struct list_head
*poison_list
,
472 struct badblocks
*bb
, const struct resource
*res
)
474 struct nd_poison
*pl
;
476 if (list_empty(poison_list
))
479 list_for_each_entry(pl
, poison_list
, list
) {
480 u64 pl_end
= pl
->start
+ pl
->length
- 1;
482 /* Discard intervals with no intersection */
483 if (pl_end
< res
->start
)
485 if (pl
->start
> res
->end
)
487 /* Deal with any overlap after start of the namespace */
488 if (pl
->start
>= res
->start
) {
489 u64 start
= pl
->start
;
492 if (pl_end
<= res
->end
)
495 len
= res
->start
+ resource_size(res
)
497 __add_badblock_range(bb
, start
- res
->start
, len
);
500 /* Deal with overlap for poison starting before the namespace */
501 if (pl
->start
< res
->start
) {
504 if (pl_end
< res
->end
)
505 len
= pl
->start
+ pl
->length
- res
->start
;
507 len
= resource_size(res
);
508 __add_badblock_range(bb
, 0, len
);
514 * nvdimm_badblocks_populate() - Convert a list of poison ranges to badblocks
515 * @region: parent region of the range to interrogate
516 * @bb: badblocks instance to populate
517 * @res: resource range to consider
519 * The poison list generated during bus initialization may contain
520 * multiple, possibly overlapping physical address ranges. Compare each
521 * of these ranges to the resource range currently being initialized,
522 * and add badblocks entries for all matching sub-ranges
524 void nvdimm_badblocks_populate(struct nd_region
*nd_region
,
525 struct badblocks
*bb
, const struct resource
*res
)
527 struct nvdimm_bus
*nvdimm_bus
;
528 struct list_head
*poison_list
;
530 if (!is_nd_pmem(&nd_region
->dev
)) {
531 dev_WARN_ONCE(&nd_region
->dev
, 1,
532 "%s only valid for pmem regions\n", __func__
);
535 nvdimm_bus
= walk_to_nvdimm_bus(&nd_region
->dev
);
536 poison_list
= &nvdimm_bus
->poison_list
;
538 nvdimm_bus_lock(&nvdimm_bus
->dev
);
539 badblocks_populate(poison_list
, bb
, res
);
540 nvdimm_bus_unlock(&nvdimm_bus
->dev
);
542 EXPORT_SYMBOL_GPL(nvdimm_badblocks_populate
);
544 static int add_poison(struct nvdimm_bus
*nvdimm_bus
, u64 addr
, u64 length
)
546 struct nd_poison
*pl
;
548 pl
= kzalloc(sizeof(*pl
), GFP_KERNEL
);
554 list_add_tail(&pl
->list
, &nvdimm_bus
->poison_list
);
559 static int bus_add_poison(struct nvdimm_bus
*nvdimm_bus
, u64 addr
, u64 length
)
561 struct nd_poison
*pl
;
563 if (list_empty(&nvdimm_bus
->poison_list
))
564 return add_poison(nvdimm_bus
, addr
, length
);
567 * There is a chance this is a duplicate, check for those first.
568 * This will be the common case as ARS_STATUS returns all known
569 * errors in the SPA space, and we can't query it per region
571 list_for_each_entry(pl
, &nvdimm_bus
->poison_list
, list
)
572 if (pl
->start
== addr
) {
573 /* If length has changed, update this list entry */
574 if (pl
->length
!= length
)
580 * If not a duplicate or a simple length update, add the entry as is,
581 * as any overlapping ranges will get resolved when the list is consumed
582 * and converted to badblocks
584 return add_poison(nvdimm_bus
, addr
, length
);
587 int nvdimm_bus_add_poison(struct nvdimm_bus
*nvdimm_bus
, u64 addr
, u64 length
)
591 nvdimm_bus_lock(&nvdimm_bus
->dev
);
592 rc
= bus_add_poison(nvdimm_bus
, addr
, length
);
593 nvdimm_bus_unlock(&nvdimm_bus
->dev
);
597 EXPORT_SYMBOL_GPL(nvdimm_bus_add_poison
);
599 #ifdef CONFIG_BLK_DEV_INTEGRITY
600 int nd_integrity_init(struct gendisk
*disk
, unsigned long meta_size
)
602 struct blk_integrity bi
;
607 memset(&bi
, 0, sizeof(bi
));
609 bi
.tuple_size
= meta_size
;
610 bi
.tag_size
= meta_size
;
612 blk_integrity_register(disk
, &bi
);
613 blk_queue_max_integrity_segments(disk
->queue
, 1);
617 EXPORT_SYMBOL(nd_integrity_init
);
619 #else /* CONFIG_BLK_DEV_INTEGRITY */
620 int nd_integrity_init(struct gendisk
*disk
, unsigned long meta_size
)
624 EXPORT_SYMBOL(nd_integrity_init
);
628 static __init
int libnvdimm_init(void)
632 rc
= nvdimm_bus_init();
638 rc
= nd_region_init();
649 static __exit
void libnvdimm_exit(void)
651 WARN_ON(!list_empty(&nvdimm_bus_list
));
655 nd_region_devs_exit();
659 MODULE_LICENSE("GPL v2");
660 MODULE_AUTHOR("Intel Corporation");
661 subsys_initcall(libnvdimm_init
);
662 module_exit(libnvdimm_exit
);