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>
26 LIST_HEAD(nvdimm_bus_list
);
27 DEFINE_MUTEX(nvdimm_bus_list_mutex
);
28 static DEFINE_IDA(nd_ida
);
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
);
60 u64
nd_fletcher64(void *addr
, size_t len
, bool le
)
67 for (i
= 0; i
< len
/ sizeof(u32
); i
++) {
68 lo32
+= le
? le32_to_cpu((__le32
) buf
[i
]) : buf
[i
];
72 return hi32
<< 32 | lo32
;
74 EXPORT_SYMBOL_GPL(nd_fletcher64
);
76 static void nvdimm_bus_release(struct device
*dev
)
78 struct nvdimm_bus
*nvdimm_bus
;
80 nvdimm_bus
= container_of(dev
, struct nvdimm_bus
, dev
);
81 ida_simple_remove(&nd_ida
, nvdimm_bus
->id
);
85 struct nvdimm_bus
*to_nvdimm_bus(struct device
*dev
)
87 struct nvdimm_bus
*nvdimm_bus
;
89 nvdimm_bus
= container_of(dev
, struct nvdimm_bus
, dev
);
90 WARN_ON(nvdimm_bus
->dev
.release
!= nvdimm_bus_release
);
93 EXPORT_SYMBOL_GPL(to_nvdimm_bus
);
95 struct nvdimm_bus_descriptor
*to_nd_desc(struct nvdimm_bus
*nvdimm_bus
)
97 /* struct nvdimm_bus definition is private to libnvdimm */
98 return nvdimm_bus
->nd_desc
;
100 EXPORT_SYMBOL_GPL(to_nd_desc
);
102 struct nvdimm_bus
*walk_to_nvdimm_bus(struct device
*nd_dev
)
106 for (dev
= nd_dev
; dev
; dev
= dev
->parent
)
107 if (dev
->release
== nvdimm_bus_release
)
109 dev_WARN_ONCE(nd_dev
, !dev
, "invalid dev, not on nd bus\n");
111 return to_nvdimm_bus(dev
);
115 static bool is_uuid_sep(char sep
)
117 if (sep
== '\n' || sep
== '-' || sep
== ':' || sep
== '\0')
122 static int nd_uuid_parse(struct device
*dev
, u8
*uuid_out
, const char *buf
,
125 const char *str
= buf
;
129 for (i
= 0; i
< 16; i
++) {
130 if (!isxdigit(str
[0]) || !isxdigit(str
[1])) {
131 dev_dbg(dev
, "%s: pos: %d buf[%zd]: %c buf[%zd]: %c\n",
132 __func__
, i
, str
- buf
, str
[0],
133 str
+ 1 - buf
, str
[1]);
137 uuid
[i
] = (hex_to_bin(str
[0]) << 4) | hex_to_bin(str
[1]);
139 if (is_uuid_sep(*str
))
143 memcpy(uuid_out
, uuid
, sizeof(uuid
));
148 * nd_uuid_store: common implementation for writing 'uuid' sysfs attributes
149 * @dev: container device for the uuid property
150 * @uuid_out: uuid buffer to replace
151 * @buf: raw sysfs buffer to parse
153 * Enforce that uuids can only be changed while the device is disabled
155 * LOCKING: expects device_lock() is held on entry
157 int nd_uuid_store(struct device
*dev
, u8
**uuid_out
, const char *buf
,
166 rc
= nd_uuid_parse(dev
, uuid
, buf
, len
);
171 *uuid_out
= kmemdup(uuid
, sizeof(uuid
), GFP_KERNEL
);
178 ssize_t
nd_sector_size_show(unsigned long current_lbasize
,
179 const unsigned long *supported
, char *buf
)
184 for (i
= 0; supported
[i
]; i
++)
185 if (current_lbasize
== supported
[i
])
186 len
+= sprintf(buf
+ len
, "[%ld] ", supported
[i
]);
188 len
+= sprintf(buf
+ len
, "%ld ", supported
[i
]);
189 len
+= sprintf(buf
+ len
, "\n");
193 ssize_t
nd_sector_size_store(struct device
*dev
, const char *buf
,
194 unsigned long *current_lbasize
, const unsigned long *supported
)
196 unsigned long lbasize
;
202 rc
= kstrtoul(buf
, 0, &lbasize
);
206 for (i
= 0; supported
[i
]; i
++)
207 if (lbasize
== supported
[i
])
211 *current_lbasize
= lbasize
;
218 void __nd_iostat_start(struct bio
*bio
, unsigned long *start
)
220 struct gendisk
*disk
= bio
->bi_bdev
->bd_disk
;
221 const int rw
= bio_data_dir(bio
);
222 int cpu
= part_stat_lock();
225 part_round_stats(cpu
, &disk
->part0
);
226 part_stat_inc(cpu
, &disk
->part0
, ios
[rw
]);
227 part_stat_add(cpu
, &disk
->part0
, sectors
[rw
], bio_sectors(bio
));
228 part_inc_in_flight(&disk
->part0
, rw
);
231 EXPORT_SYMBOL(__nd_iostat_start
);
233 void nd_iostat_end(struct bio
*bio
, unsigned long start
)
235 struct gendisk
*disk
= bio
->bi_bdev
->bd_disk
;
236 unsigned long duration
= jiffies
- start
;
237 const int rw
= bio_data_dir(bio
);
238 int cpu
= part_stat_lock();
240 part_stat_add(cpu
, &disk
->part0
, ticks
[rw
], duration
);
241 part_round_stats(cpu
, &disk
->part0
);
242 part_dec_in_flight(&disk
->part0
, rw
);
245 EXPORT_SYMBOL(nd_iostat_end
);
247 static ssize_t
commands_show(struct device
*dev
,
248 struct device_attribute
*attr
, char *buf
)
251 struct nvdimm_bus
*nvdimm_bus
= to_nvdimm_bus(dev
);
252 struct nvdimm_bus_descriptor
*nd_desc
= nvdimm_bus
->nd_desc
;
254 for_each_set_bit(cmd
, &nd_desc
->cmd_mask
, BITS_PER_LONG
)
255 len
+= sprintf(buf
+ len
, "%s ", nvdimm_bus_cmd_name(cmd
));
256 len
+= sprintf(buf
+ len
, "\n");
259 static DEVICE_ATTR_RO(commands
);
261 static const char *nvdimm_bus_provider(struct nvdimm_bus
*nvdimm_bus
)
263 struct nvdimm_bus_descriptor
*nd_desc
= nvdimm_bus
->nd_desc
;
264 struct device
*parent
= nvdimm_bus
->dev
.parent
;
266 if (nd_desc
->provider_name
)
267 return nd_desc
->provider_name
;
269 return dev_name(parent
);
274 static ssize_t
provider_show(struct device
*dev
,
275 struct device_attribute
*attr
, char *buf
)
277 struct nvdimm_bus
*nvdimm_bus
= to_nvdimm_bus(dev
);
279 return sprintf(buf
, "%s\n", nvdimm_bus_provider(nvdimm_bus
));
281 static DEVICE_ATTR_RO(provider
);
283 static int flush_namespaces(struct device
*dev
, void *data
)
290 static int flush_regions_dimms(struct device
*dev
, void *data
)
294 device_for_each_child(dev
, NULL
, flush_namespaces
);
298 static ssize_t
wait_probe_show(struct device
*dev
,
299 struct device_attribute
*attr
, char *buf
)
301 struct nvdimm_bus
*nvdimm_bus
= to_nvdimm_bus(dev
);
302 struct nvdimm_bus_descriptor
*nd_desc
= nvdimm_bus
->nd_desc
;
305 if (nd_desc
->flush_probe
) {
306 rc
= nd_desc
->flush_probe(nd_desc
);
311 device_for_each_child(dev
, NULL
, flush_regions_dimms
);
312 return sprintf(buf
, "1\n");
314 static DEVICE_ATTR_RO(wait_probe
);
316 static struct attribute
*nvdimm_bus_attributes
[] = {
317 &dev_attr_commands
.attr
,
318 &dev_attr_wait_probe
.attr
,
319 &dev_attr_provider
.attr
,
323 struct attribute_group nvdimm_bus_attribute_group
= {
324 .attrs
= nvdimm_bus_attributes
,
326 EXPORT_SYMBOL_GPL(nvdimm_bus_attribute_group
);
328 struct nvdimm_bus
*__nvdimm_bus_register(struct device
*parent
,
329 struct nvdimm_bus_descriptor
*nd_desc
, struct module
*module
)
331 struct nvdimm_bus
*nvdimm_bus
;
334 nvdimm_bus
= kzalloc(sizeof(*nvdimm_bus
), GFP_KERNEL
);
337 INIT_LIST_HEAD(&nvdimm_bus
->list
);
338 INIT_LIST_HEAD(&nvdimm_bus
->poison_list
);
339 init_waitqueue_head(&nvdimm_bus
->probe_wait
);
340 nvdimm_bus
->id
= ida_simple_get(&nd_ida
, 0, 0, GFP_KERNEL
);
341 mutex_init(&nvdimm_bus
->reconfig_mutex
);
342 if (nvdimm_bus
->id
< 0) {
346 nvdimm_bus
->nd_desc
= nd_desc
;
347 nvdimm_bus
->module
= module
;
348 nvdimm_bus
->dev
.parent
= parent
;
349 nvdimm_bus
->dev
.release
= nvdimm_bus_release
;
350 nvdimm_bus
->dev
.groups
= nd_desc
->attr_groups
;
351 dev_set_name(&nvdimm_bus
->dev
, "ndbus%d", nvdimm_bus
->id
);
352 rc
= device_register(&nvdimm_bus
->dev
);
354 dev_dbg(&nvdimm_bus
->dev
, "registration failed: %d\n", rc
);
358 rc
= nvdimm_bus_create_ndctl(nvdimm_bus
);
362 mutex_lock(&nvdimm_bus_list_mutex
);
363 list_add_tail(&nvdimm_bus
->list
, &nvdimm_bus_list
);
364 mutex_unlock(&nvdimm_bus_list_mutex
);
368 put_device(&nvdimm_bus
->dev
);
371 EXPORT_SYMBOL_GPL(__nvdimm_bus_register
);
373 static void set_badblock(struct badblocks
*bb
, sector_t s
, int num
)
375 dev_dbg(bb
->dev
, "Found a poison range (0x%llx, 0x%llx)\n",
376 (u64
) s
* 512, (u64
) num
* 512);
377 /* this isn't an error as the hardware will still throw an exception */
378 if (badblocks_set(bb
, s
, num
, 1))
379 dev_info_once(bb
->dev
, "%s: failed for sector %llx\n",
384 * __add_badblock_range() - Convert a physical address range to bad sectors
385 * @bb: badblocks instance to populate
386 * @ns_offset: namespace offset where the error range begins (in bytes)
387 * @len: number of bytes of poison to be added
389 * This assumes that the range provided with (ns_offset, len) is within
390 * the bounds of physical addresses for this namespace, i.e. lies in the
391 * interval [ns_start, ns_start + ns_size)
393 static void __add_badblock_range(struct badblocks
*bb
, u64 ns_offset
, u64 len
)
395 const unsigned int sector_size
= 512;
396 sector_t start_sector
;
400 start_sector
= div_u64(ns_offset
, sector_size
);
401 num_sectors
= div_u64_rem(len
, sector_size
, &rem
);
405 if (unlikely(num_sectors
> (u64
)INT_MAX
)) {
406 u64 remaining
= num_sectors
;
407 sector_t s
= start_sector
;
410 int done
= min_t(u64
, remaining
, INT_MAX
);
412 set_badblock(bb
, s
, done
);
417 set_badblock(bb
, start_sector
, num_sectors
);
420 static void badblocks_populate(struct list_head
*poison_list
,
421 struct badblocks
*bb
, const struct resource
*res
)
423 struct nd_poison
*pl
;
425 if (list_empty(poison_list
))
428 list_for_each_entry(pl
, poison_list
, list
) {
429 u64 pl_end
= pl
->start
+ pl
->length
- 1;
431 /* Discard intervals with no intersection */
432 if (pl_end
< res
->start
)
434 if (pl
->start
> res
->end
)
436 /* Deal with any overlap after start of the namespace */
437 if (pl
->start
>= res
->start
) {
438 u64 start
= pl
->start
;
441 if (pl_end
<= res
->end
)
444 len
= res
->start
+ resource_size(res
)
446 __add_badblock_range(bb
, start
- res
->start
, len
);
449 /* Deal with overlap for poison starting before the namespace */
450 if (pl
->start
< res
->start
) {
453 if (pl_end
< res
->end
)
454 len
= pl
->start
+ pl
->length
- res
->start
;
456 len
= resource_size(res
);
457 __add_badblock_range(bb
, 0, len
);
463 * nvdimm_badblocks_populate() - Convert a list of poison ranges to badblocks
464 * @region: parent region of the range to interrogate
465 * @bb: badblocks instance to populate
466 * @res: resource range to consider
468 * The poison list generated during bus initialization may contain
469 * multiple, possibly overlapping physical address ranges. Compare each
470 * of these ranges to the resource range currently being initialized,
471 * and add badblocks entries for all matching sub-ranges
473 void nvdimm_badblocks_populate(struct nd_region
*nd_region
,
474 struct badblocks
*bb
, const struct resource
*res
)
476 struct nvdimm_bus
*nvdimm_bus
;
477 struct list_head
*poison_list
;
479 if (!is_nd_pmem(&nd_region
->dev
)) {
480 dev_WARN_ONCE(&nd_region
->dev
, 1,
481 "%s only valid for pmem regions\n", __func__
);
484 nvdimm_bus
= walk_to_nvdimm_bus(&nd_region
->dev
);
485 poison_list
= &nvdimm_bus
->poison_list
;
487 nvdimm_bus_lock(&nvdimm_bus
->dev
);
488 badblocks_populate(poison_list
, bb
, res
);
489 nvdimm_bus_unlock(&nvdimm_bus
->dev
);
491 EXPORT_SYMBOL_GPL(nvdimm_badblocks_populate
);
493 static int add_poison(struct nvdimm_bus
*nvdimm_bus
, u64 addr
, u64 length
)
495 struct nd_poison
*pl
;
497 pl
= kzalloc(sizeof(*pl
), GFP_KERNEL
);
503 list_add_tail(&pl
->list
, &nvdimm_bus
->poison_list
);
508 static int bus_add_poison(struct nvdimm_bus
*nvdimm_bus
, u64 addr
, u64 length
)
510 struct nd_poison
*pl
;
512 if (list_empty(&nvdimm_bus
->poison_list
))
513 return add_poison(nvdimm_bus
, addr
, length
);
516 * There is a chance this is a duplicate, check for those first.
517 * This will be the common case as ARS_STATUS returns all known
518 * errors in the SPA space, and we can't query it per region
520 list_for_each_entry(pl
, &nvdimm_bus
->poison_list
, list
)
521 if (pl
->start
== addr
) {
522 /* If length has changed, update this list entry */
523 if (pl
->length
!= length
)
529 * If not a duplicate or a simple length update, add the entry as is,
530 * as any overlapping ranges will get resolved when the list is consumed
531 * and converted to badblocks
533 return add_poison(nvdimm_bus
, addr
, length
);
536 int nvdimm_bus_add_poison(struct nvdimm_bus
*nvdimm_bus
, u64 addr
, u64 length
)
540 nvdimm_bus_lock(&nvdimm_bus
->dev
);
541 rc
= bus_add_poison(nvdimm_bus
, addr
, length
);
542 nvdimm_bus_unlock(&nvdimm_bus
->dev
);
546 EXPORT_SYMBOL_GPL(nvdimm_bus_add_poison
);
548 static void free_poison_list(struct list_head
*poison_list
)
550 struct nd_poison
*pl
, *next
;
552 list_for_each_entry_safe(pl
, next
, poison_list
, list
) {
556 list_del_init(poison_list
);
559 static int child_unregister(struct device
*dev
, void *data
)
562 * the singular ndctl class device per bus needs to be
563 * "device_destroy"ed, so skip it here
565 * i.e. remove classless children
570 nd_device_unregister(dev
, ND_SYNC
);
574 void nvdimm_bus_unregister(struct nvdimm_bus
*nvdimm_bus
)
579 mutex_lock(&nvdimm_bus_list_mutex
);
580 list_del_init(&nvdimm_bus
->list
);
581 mutex_unlock(&nvdimm_bus_list_mutex
);
584 device_for_each_child(&nvdimm_bus
->dev
, NULL
, child_unregister
);
586 nvdimm_bus_lock(&nvdimm_bus
->dev
);
587 free_poison_list(&nvdimm_bus
->poison_list
);
588 nvdimm_bus_unlock(&nvdimm_bus
->dev
);
590 nvdimm_bus_destroy_ndctl(nvdimm_bus
);
592 device_unregister(&nvdimm_bus
->dev
);
594 EXPORT_SYMBOL_GPL(nvdimm_bus_unregister
);
596 #ifdef CONFIG_BLK_DEV_INTEGRITY
597 int nd_integrity_init(struct gendisk
*disk
, unsigned long meta_size
)
599 struct blk_integrity bi
;
605 bi
.tuple_size
= meta_size
;
606 bi
.tag_size
= meta_size
;
608 blk_integrity_register(disk
, &bi
);
609 blk_queue_max_integrity_segments(disk
->queue
, 1);
613 EXPORT_SYMBOL(nd_integrity_init
);
615 #else /* CONFIG_BLK_DEV_INTEGRITY */
616 int nd_integrity_init(struct gendisk
*disk
, unsigned long meta_size
)
620 EXPORT_SYMBOL(nd_integrity_init
);
624 static __init
int libnvdimm_init(void)
628 rc
= nvdimm_bus_init();
634 rc
= nd_region_init();
645 static __exit
void libnvdimm_exit(void)
647 WARN_ON(!list_empty(&nvdimm_bus_list
));
651 nd_region_devs_exit();
653 ida_destroy(&nd_ida
);
656 MODULE_LICENSE("GPL v2");
657 MODULE_AUTHOR("Intel Corporation");
658 subsys_initcall(libnvdimm_init
);
659 module_exit(libnvdimm_exit
);