mm: fix exec activate_mm vs TLB shootdown and lazy tlb switching race
[linux/fpc-iii.git] / block / partition-generic.c
blobdb57cced9b987371e6c8a3c72ff6721b9d540bda
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * Code extracted from drivers/block/genhd.c
4 * Copyright (C) 1991-1998 Linus Torvalds
5 * Re-organised Feb 1998 Russell King
7 * We now have independent partition support from the
8 * block drivers, which allows all the partition code to
9 * be grouped in one location, and it to be mostly self
10 * contained.
13 #include <linux/init.h>
14 #include <linux/module.h>
15 #include <linux/fs.h>
16 #include <linux/slab.h>
17 #include <linux/kmod.h>
18 #include <linux/ctype.h>
19 #include <linux/genhd.h>
20 #include <linux/blktrace_api.h>
22 #include "partitions/check.h"
24 #ifdef CONFIG_BLK_DEV_MD
25 extern void md_autodetect_dev(dev_t dev);
26 #endif
29 * disk_name() is used by partition check code and the genhd driver.
30 * It formats the devicename of the indicated disk into
31 * the supplied buffer (of size at least 32), and returns
32 * a pointer to that same buffer (for convenience).
35 char *disk_name(struct gendisk *hd, int partno, char *buf)
37 if (!partno)
38 snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name);
39 else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1]))
40 snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno);
41 else
42 snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno);
44 return buf;
47 const char *bdevname(struct block_device *bdev, char *buf)
49 return disk_name(bdev->bd_disk, bdev->bd_part->partno, buf);
52 EXPORT_SYMBOL(bdevname);
54 const char *bio_devname(struct bio *bio, char *buf)
56 return disk_name(bio->bi_disk, bio->bi_partno, buf);
58 EXPORT_SYMBOL(bio_devname);
61 * There's very little reason to use this, you should really
62 * have a struct block_device just about everywhere and use
63 * bdevname() instead.
65 const char *__bdevname(dev_t dev, char *buffer)
67 scnprintf(buffer, BDEVNAME_SIZE, "unknown-block(%u,%u)",
68 MAJOR(dev), MINOR(dev));
69 return buffer;
72 EXPORT_SYMBOL(__bdevname);
74 static ssize_t part_partition_show(struct device *dev,
75 struct device_attribute *attr, char *buf)
77 struct hd_struct *p = dev_to_part(dev);
79 return sprintf(buf, "%d\n", p->partno);
82 static ssize_t part_start_show(struct device *dev,
83 struct device_attribute *attr, char *buf)
85 struct hd_struct *p = dev_to_part(dev);
87 return sprintf(buf, "%llu\n",(unsigned long long)p->start_sect);
90 ssize_t part_size_show(struct device *dev,
91 struct device_attribute *attr, char *buf)
93 struct hd_struct *p = dev_to_part(dev);
94 return sprintf(buf, "%llu\n",(unsigned long long)part_nr_sects_read(p));
97 static ssize_t part_ro_show(struct device *dev,
98 struct device_attribute *attr, char *buf)
100 struct hd_struct *p = dev_to_part(dev);
101 return sprintf(buf, "%d\n", p->policy ? 1 : 0);
104 static ssize_t part_alignment_offset_show(struct device *dev,
105 struct device_attribute *attr, char *buf)
107 struct hd_struct *p = dev_to_part(dev);
108 return sprintf(buf, "%llu\n", (unsigned long long)p->alignment_offset);
111 static ssize_t part_discard_alignment_show(struct device *dev,
112 struct device_attribute *attr, char *buf)
114 struct hd_struct *p = dev_to_part(dev);
115 return sprintf(buf, "%u\n", p->discard_alignment);
118 ssize_t part_stat_show(struct device *dev,
119 struct device_attribute *attr, char *buf)
121 struct hd_struct *p = dev_to_part(dev);
122 struct request_queue *q = part_to_disk(p)->queue;
123 unsigned int inflight[2];
124 int cpu;
126 cpu = part_stat_lock();
127 part_round_stats(q, cpu, p);
128 part_stat_unlock();
129 part_in_flight(q, p, inflight);
130 return sprintf(buf,
131 "%8lu %8lu %8llu %8u "
132 "%8lu %8lu %8llu %8u "
133 "%8u %8u %8u"
134 "\n",
135 part_stat_read(p, ios[READ]),
136 part_stat_read(p, merges[READ]),
137 (unsigned long long)part_stat_read(p, sectors[READ]),
138 jiffies_to_msecs(part_stat_read(p, ticks[READ])),
139 part_stat_read(p, ios[WRITE]),
140 part_stat_read(p, merges[WRITE]),
141 (unsigned long long)part_stat_read(p, sectors[WRITE]),
142 jiffies_to_msecs(part_stat_read(p, ticks[WRITE])),
143 inflight[0],
144 jiffies_to_msecs(part_stat_read(p, io_ticks)),
145 jiffies_to_msecs(part_stat_read(p, time_in_queue)));
148 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr,
149 char *buf)
151 struct hd_struct *p = dev_to_part(dev);
152 struct request_queue *q = part_to_disk(p)->queue;
153 unsigned int inflight[2];
155 part_in_flight_rw(q, p, inflight);
156 return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
159 #ifdef CONFIG_FAIL_MAKE_REQUEST
160 ssize_t part_fail_show(struct device *dev,
161 struct device_attribute *attr, char *buf)
163 struct hd_struct *p = dev_to_part(dev);
165 return sprintf(buf, "%d\n", p->make_it_fail);
168 ssize_t part_fail_store(struct device *dev,
169 struct device_attribute *attr,
170 const char *buf, size_t count)
172 struct hd_struct *p = dev_to_part(dev);
173 int i;
175 if (count > 0 && sscanf(buf, "%d", &i) > 0)
176 p->make_it_fail = (i == 0) ? 0 : 1;
178 return count;
180 #endif
182 static DEVICE_ATTR(partition, S_IRUGO, part_partition_show, NULL);
183 static DEVICE_ATTR(start, S_IRUGO, part_start_show, NULL);
184 static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
185 static DEVICE_ATTR(ro, S_IRUGO, part_ro_show, NULL);
186 static DEVICE_ATTR(alignment_offset, S_IRUGO, part_alignment_offset_show, NULL);
187 static DEVICE_ATTR(discard_alignment, S_IRUGO, part_discard_alignment_show,
188 NULL);
189 static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
190 static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
191 #ifdef CONFIG_FAIL_MAKE_REQUEST
192 static struct device_attribute dev_attr_fail =
193 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
194 #endif
196 static struct attribute *part_attrs[] = {
197 &dev_attr_partition.attr,
198 &dev_attr_start.attr,
199 &dev_attr_size.attr,
200 &dev_attr_ro.attr,
201 &dev_attr_alignment_offset.attr,
202 &dev_attr_discard_alignment.attr,
203 &dev_attr_stat.attr,
204 &dev_attr_inflight.attr,
205 #ifdef CONFIG_FAIL_MAKE_REQUEST
206 &dev_attr_fail.attr,
207 #endif
208 NULL
211 static struct attribute_group part_attr_group = {
212 .attrs = part_attrs,
215 static const struct attribute_group *part_attr_groups[] = {
216 &part_attr_group,
217 #ifdef CONFIG_BLK_DEV_IO_TRACE
218 &blk_trace_attr_group,
219 #endif
220 NULL
223 static void part_release(struct device *dev)
225 struct hd_struct *p = dev_to_part(dev);
226 blk_free_devt(dev->devt);
227 hd_free_part(p);
228 kfree(p);
231 static int part_uevent(struct device *dev, struct kobj_uevent_env *env)
233 struct hd_struct *part = dev_to_part(dev);
235 add_uevent_var(env, "PARTN=%u", part->partno);
236 if (part->info && part->info->volname[0])
237 add_uevent_var(env, "PARTNAME=%s", part->info->volname);
238 return 0;
241 struct device_type part_type = {
242 .name = "partition",
243 .groups = part_attr_groups,
244 .release = part_release,
245 .uevent = part_uevent,
248 static void delete_partition_rcu_cb(struct rcu_head *head)
250 struct hd_struct *part = container_of(head, struct hd_struct, rcu_head);
252 part->start_sect = 0;
253 part->nr_sects = 0;
254 part_stat_set_all(part, 0);
255 put_device(part_to_dev(part));
258 void __delete_partition(struct percpu_ref *ref)
260 struct hd_struct *part = container_of(ref, struct hd_struct, ref);
261 call_rcu(&part->rcu_head, delete_partition_rcu_cb);
265 * Must be called either with bd_mutex held, before a disk can be opened or
266 * after all disk users are gone.
268 void delete_partition(struct gendisk *disk, int partno)
270 struct disk_part_tbl *ptbl =
271 rcu_dereference_protected(disk->part_tbl, 1);
272 struct hd_struct *part;
274 if (partno >= ptbl->len)
275 return;
277 part = rcu_dereference_protected(ptbl->part[partno], 1);
278 if (!part)
279 return;
281 rcu_assign_pointer(ptbl->part[partno], NULL);
282 rcu_assign_pointer(ptbl->last_lookup, NULL);
283 kobject_put(part->holder_dir);
284 device_del(part_to_dev(part));
286 hd_struct_kill(part);
289 static ssize_t whole_disk_show(struct device *dev,
290 struct device_attribute *attr, char *buf)
292 return 0;
294 static DEVICE_ATTR(whole_disk, S_IRUSR | S_IRGRP | S_IROTH,
295 whole_disk_show, NULL);
298 * Must be called either with bd_mutex held, before a disk can be opened or
299 * after all disk users are gone.
301 struct hd_struct *add_partition(struct gendisk *disk, int partno,
302 sector_t start, sector_t len, int flags,
303 struct partition_meta_info *info)
305 struct hd_struct *p;
306 dev_t devt = MKDEV(0, 0);
307 struct device *ddev = disk_to_dev(disk);
308 struct device *pdev;
309 struct disk_part_tbl *ptbl;
310 const char *dname;
311 int err;
313 err = disk_expand_part_tbl(disk, partno);
314 if (err)
315 return ERR_PTR(err);
316 ptbl = rcu_dereference_protected(disk->part_tbl, 1);
318 if (ptbl->part[partno])
319 return ERR_PTR(-EBUSY);
321 p = kzalloc(sizeof(*p), GFP_KERNEL);
322 if (!p)
323 return ERR_PTR(-EBUSY);
325 if (!init_part_stats(p)) {
326 err = -ENOMEM;
327 goto out_free;
330 seqcount_init(&p->nr_sects_seq);
331 pdev = part_to_dev(p);
333 p->start_sect = start;
334 p->alignment_offset =
335 queue_limit_alignment_offset(&disk->queue->limits, start);
336 p->discard_alignment =
337 queue_limit_discard_alignment(&disk->queue->limits, start);
338 p->nr_sects = len;
339 p->partno = partno;
340 p->policy = get_disk_ro(disk);
342 if (info) {
343 struct partition_meta_info *pinfo = alloc_part_info(disk);
344 if (!pinfo) {
345 err = -ENOMEM;
346 goto out_free_stats;
348 memcpy(pinfo, info, sizeof(*info));
349 p->info = pinfo;
352 dname = dev_name(ddev);
353 if (isdigit(dname[strlen(dname) - 1]))
354 dev_set_name(pdev, "%sp%d", dname, partno);
355 else
356 dev_set_name(pdev, "%s%d", dname, partno);
358 device_initialize(pdev);
359 pdev->class = &block_class;
360 pdev->type = &part_type;
361 pdev->parent = ddev;
363 err = blk_alloc_devt(p, &devt);
364 if (err)
365 goto out_free_info;
366 pdev->devt = devt;
368 /* delay uevent until 'holders' subdir is created */
369 dev_set_uevent_suppress(pdev, 1);
370 err = device_add(pdev);
371 if (err)
372 goto out_put;
374 err = -ENOMEM;
375 p->holder_dir = kobject_create_and_add("holders", &pdev->kobj);
376 if (!p->holder_dir)
377 goto out_del;
379 dev_set_uevent_suppress(pdev, 0);
380 if (flags & ADDPART_FLAG_WHOLEDISK) {
381 err = device_create_file(pdev, &dev_attr_whole_disk);
382 if (err)
383 goto out_del;
386 err = hd_ref_init(p);
387 if (err) {
388 if (flags & ADDPART_FLAG_WHOLEDISK)
389 goto out_remove_file;
390 goto out_del;
393 /* everything is up and running, commence */
394 rcu_assign_pointer(ptbl->part[partno], p);
396 /* suppress uevent if the disk suppresses it */
397 if (!dev_get_uevent_suppress(ddev))
398 kobject_uevent(&pdev->kobj, KOBJ_ADD);
399 return p;
401 out_free_info:
402 free_part_info(p);
403 out_free_stats:
404 free_part_stats(p);
405 out_free:
406 kfree(p);
407 return ERR_PTR(err);
408 out_remove_file:
409 device_remove_file(pdev, &dev_attr_whole_disk);
410 out_del:
411 kobject_put(p->holder_dir);
412 device_del(pdev);
413 out_put:
414 put_device(pdev);
415 return ERR_PTR(err);
418 static bool disk_unlock_native_capacity(struct gendisk *disk)
420 const struct block_device_operations *bdops = disk->fops;
422 if (bdops->unlock_native_capacity &&
423 !(disk->flags & GENHD_FL_NATIVE_CAPACITY)) {
424 printk(KERN_CONT "enabling native capacity\n");
425 bdops->unlock_native_capacity(disk);
426 disk->flags |= GENHD_FL_NATIVE_CAPACITY;
427 return true;
428 } else {
429 printk(KERN_CONT "truncated\n");
430 return false;
434 static int drop_partitions(struct gendisk *disk, struct block_device *bdev)
436 struct disk_part_iter piter;
437 struct hd_struct *part;
438 int res;
440 if (bdev->bd_part_count || bdev->bd_super)
441 return -EBUSY;
442 res = invalidate_partition(disk, 0);
443 if (res)
444 return res;
446 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
447 while ((part = disk_part_iter_next(&piter)))
448 delete_partition(disk, part->partno);
449 disk_part_iter_exit(&piter);
451 return 0;
454 static bool part_zone_aligned(struct gendisk *disk,
455 struct block_device *bdev,
456 sector_t from, sector_t size)
458 unsigned int zone_sectors = bdev_zone_sectors(bdev);
461 * If this function is called, then the disk is a zoned block device
462 * (host-aware or host-managed). This can be detected even if the
463 * zoned block device support is disabled (CONFIG_BLK_DEV_ZONED not
464 * set). In this case, however, only host-aware devices will be seen
465 * as a block device is not created for host-managed devices. Without
466 * zoned block device support, host-aware drives can still be used as
467 * regular block devices (no zone operation) and their zone size will
468 * be reported as 0. Allow this case.
470 if (!zone_sectors)
471 return true;
474 * Check partition start and size alignement. If the drive has a
475 * smaller last runt zone, ignore it and allow the partition to
476 * use it. Check the zone size too: it should be a power of 2 number
477 * of sectors.
479 if (WARN_ON_ONCE(!is_power_of_2(zone_sectors))) {
480 u32 rem;
482 div_u64_rem(from, zone_sectors, &rem);
483 if (rem)
484 return false;
485 if ((from + size) < get_capacity(disk)) {
486 div_u64_rem(size, zone_sectors, &rem);
487 if (rem)
488 return false;
491 } else {
493 if (from & (zone_sectors - 1))
494 return false;
495 if ((from + size) < get_capacity(disk) &&
496 (size & (zone_sectors - 1)))
497 return false;
501 return true;
504 int rescan_partitions(struct gendisk *disk, struct block_device *bdev)
506 struct parsed_partitions *state = NULL;
507 struct hd_struct *part;
508 int p, highest, res;
509 rescan:
510 if (state && !IS_ERR(state)) {
511 free_partitions(state);
512 state = NULL;
515 res = drop_partitions(disk, bdev);
516 if (res)
517 return res;
519 if (disk->fops->revalidate_disk)
520 disk->fops->revalidate_disk(disk);
521 check_disk_size_change(disk, bdev);
522 bdev->bd_invalidated = 0;
523 if (!get_capacity(disk) || !(state = check_partition(disk, bdev)))
524 return 0;
525 if (IS_ERR(state)) {
527 * I/O error reading the partition table. If any
528 * partition code tried to read beyond EOD, retry
529 * after unlocking native capacity.
531 if (PTR_ERR(state) == -ENOSPC) {
532 printk(KERN_WARNING "%s: partition table beyond EOD, ",
533 disk->disk_name);
534 if (disk_unlock_native_capacity(disk))
535 goto rescan;
537 return -EIO;
540 * If any partition code tried to read beyond EOD, try
541 * unlocking native capacity even if partition table is
542 * successfully read as we could be missing some partitions.
544 if (state->access_beyond_eod) {
545 printk(KERN_WARNING
546 "%s: partition table partially beyond EOD, ",
547 disk->disk_name);
548 if (disk_unlock_native_capacity(disk))
549 goto rescan;
552 /* tell userspace that the media / partition table may have changed */
553 kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE);
555 /* Detect the highest partition number and preallocate
556 * disk->part_tbl. This is an optimization and not strictly
557 * necessary.
559 for (p = 1, highest = 0; p < state->limit; p++)
560 if (state->parts[p].size)
561 highest = p;
563 disk_expand_part_tbl(disk, highest);
565 /* add partitions */
566 for (p = 1; p < state->limit; p++) {
567 sector_t size, from;
569 size = state->parts[p].size;
570 if (!size)
571 continue;
573 from = state->parts[p].from;
574 if (from >= get_capacity(disk)) {
575 printk(KERN_WARNING
576 "%s: p%d start %llu is beyond EOD, ",
577 disk->disk_name, p, (unsigned long long) from);
578 if (disk_unlock_native_capacity(disk))
579 goto rescan;
580 continue;
583 if (from + size > get_capacity(disk)) {
584 printk(KERN_WARNING
585 "%s: p%d size %llu extends beyond EOD, ",
586 disk->disk_name, p, (unsigned long long) size);
588 if (disk_unlock_native_capacity(disk)) {
589 /* free state and restart */
590 goto rescan;
591 } else {
593 * we can not ignore partitions of broken tables
594 * created by for example camera firmware, but
595 * we limit them to the end of the disk to avoid
596 * creating invalid block devices
598 size = get_capacity(disk) - from;
603 * On a zoned block device, partitions should be aligned on the
604 * device zone size (i.e. zone boundary crossing not allowed).
605 * Otherwise, resetting the write pointer of the last zone of
606 * one partition may impact the following partition.
608 if (bdev_is_zoned(bdev) &&
609 !part_zone_aligned(disk, bdev, from, size)) {
610 printk(KERN_WARNING
611 "%s: p%d start %llu+%llu is not zone aligned\n",
612 disk->disk_name, p, (unsigned long long) from,
613 (unsigned long long) size);
614 continue;
617 part = add_partition(disk, p, from, size,
618 state->parts[p].flags,
619 &state->parts[p].info);
620 if (IS_ERR(part)) {
621 printk(KERN_ERR " %s: p%d could not be added: %ld\n",
622 disk->disk_name, p, -PTR_ERR(part));
623 continue;
625 #ifdef CONFIG_BLK_DEV_MD
626 if (state->parts[p].flags & ADDPART_FLAG_RAID)
627 md_autodetect_dev(part_to_dev(part)->devt);
628 #endif
630 free_partitions(state);
631 return 0;
634 int invalidate_partitions(struct gendisk *disk, struct block_device *bdev)
636 int res;
638 if (!bdev->bd_invalidated)
639 return 0;
641 res = drop_partitions(disk, bdev);
642 if (res)
643 return res;
645 set_capacity(disk, 0);
646 check_disk_size_change(disk, bdev);
647 bdev->bd_invalidated = 0;
648 /* tell userspace that the media / partition table may have changed */
649 kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE);
651 return 0;
654 unsigned char *read_dev_sector(struct block_device *bdev, sector_t n, Sector *p)
656 struct address_space *mapping = bdev->bd_inode->i_mapping;
657 struct page *page;
659 page = read_mapping_page(mapping, (pgoff_t)(n >> (PAGE_SHIFT-9)), NULL);
660 if (!IS_ERR(page)) {
661 if (PageError(page))
662 goto fail;
663 p->v = page;
664 return (unsigned char *)page_address(page) + ((n & ((1 << (PAGE_SHIFT - 9)) - 1)) << 9);
665 fail:
666 put_page(page);
668 p->v = NULL;
669 return NULL;
672 EXPORT_SYMBOL(read_dev_sector);