1 // SPDX-License-Identifier: GPL-2.0-only
3 * Copyright (c) 2018 Red Hat, Inc.
5 * This is a test "dust" device, which fails reads on specified
6 * sectors, emulating the behavior of a hard disk drive sending
7 * a "Read Medium Error" sense.
11 #include <linux/device-mapper.h>
12 #include <linux/module.h>
13 #include <linux/rbtree.h>
15 #define DM_MSG_PREFIX "dust"
20 unsigned char wr_fail_cnt;
25 struct rb_root badblocklist;
26 unsigned long long badblock_count;
29 int sect_per_block_shift;
30 unsigned int sect_per_block;
32 bool fail_read_on_bb:1;
36 static struct badblock *dust_rb_search(struct rb_root *root, sector_t blk)
38 struct rb_node *node = root->rb_node;
41 struct badblock *bblk = rb_entry(node, struct badblock, node);
45 else if (bblk->bb < blk)
46 node = node->rb_right;
54 static bool dust_rb_insert(struct rb_root *root, struct badblock *new)
56 struct badblock *bblk;
57 struct rb_node **link = &root->rb_node, *parent = NULL;
58 sector_t value = new->bb;
62 bblk = rb_entry(parent, struct badblock, node);
65 link = &(*link)->rb_left;
66 else if (bblk->bb < value)
67 link = &(*link)->rb_right;
72 rb_link_node(&new->node, parent, link);
73 rb_insert_color(&new->node, root);
78 static int dust_remove_block(struct dust_device *dd, unsigned long long block)
80 struct badblock *bblock;
83 spin_lock_irqsave(&dd->dust_lock, flags);
84 bblock = dust_rb_search(&dd->badblocklist, block);
87 if (!dd->quiet_mode) {
88 DMERR("%s: block %llu not found in badblocklist",
91 spin_unlock_irqrestore(&dd->dust_lock, flags);
95 rb_erase(&bblock->node, &dd->badblocklist);
98 DMINFO("%s: badblock removed at block %llu", __func__, block);
100 spin_unlock_irqrestore(&dd->dust_lock, flags);
105 static int dust_add_block(struct dust_device *dd, unsigned long long block,
106 unsigned char wr_fail_cnt)
108 struct badblock *bblock;
111 bblock = kmalloc(sizeof(*bblock), GFP_KERNEL);
112 if (bblock == NULL) {
114 DMERR("%s: badblock allocation failed", __func__);
118 spin_lock_irqsave(&dd->dust_lock, flags);
120 bblock->wr_fail_cnt = wr_fail_cnt;
121 if (!dust_rb_insert(&dd->badblocklist, bblock)) {
122 if (!dd->quiet_mode) {
123 DMERR("%s: block %llu already in badblocklist",
126 spin_unlock_irqrestore(&dd->dust_lock, flags);
131 dd->badblock_count++;
132 if (!dd->quiet_mode) {
133 DMINFO("%s: badblock added at block %llu with write fail count %u",
134 __func__, block, wr_fail_cnt);
136 spin_unlock_irqrestore(&dd->dust_lock, flags);
141 static int dust_query_block(struct dust_device *dd, unsigned long long block, char *result,
142 unsigned int maxlen, unsigned int *sz_ptr)
144 struct badblock *bblock;
146 unsigned int sz = *sz_ptr;
148 spin_lock_irqsave(&dd->dust_lock, flags);
149 bblock = dust_rb_search(&dd->badblocklist, block);
151 DMEMIT("%s: block %llu found in badblocklist", __func__, block);
153 DMEMIT("%s: block %llu not found in badblocklist", __func__, block);
154 spin_unlock_irqrestore(&dd->dust_lock, flags);
159 static int __dust_map_read(struct dust_device *dd, sector_t thisblock)
161 struct badblock *bblk = dust_rb_search(&dd->badblocklist, thisblock);
164 return DM_MAPIO_KILL;
166 return DM_MAPIO_REMAPPED;
169 static int dust_map_read(struct dust_device *dd, sector_t thisblock,
170 bool fail_read_on_bb)
173 int r = DM_MAPIO_REMAPPED;
175 if (fail_read_on_bb) {
176 thisblock >>= dd->sect_per_block_shift;
177 spin_lock_irqsave(&dd->dust_lock, flags);
178 r = __dust_map_read(dd, thisblock);
179 spin_unlock_irqrestore(&dd->dust_lock, flags);
185 static int __dust_map_write(struct dust_device *dd, sector_t thisblock)
187 struct badblock *bblk = dust_rb_search(&dd->badblocklist, thisblock);
189 if (bblk && bblk->wr_fail_cnt > 0) {
191 return DM_MAPIO_KILL;
195 rb_erase(&bblk->node, &dd->badblocklist);
196 dd->badblock_count--;
198 if (!dd->quiet_mode) {
199 sector_div(thisblock, dd->sect_per_block);
200 DMINFO("block %llu removed from badblocklist by write",
201 (unsigned long long)thisblock);
205 return DM_MAPIO_REMAPPED;
208 static int dust_map_write(struct dust_device *dd, sector_t thisblock,
209 bool fail_read_on_bb)
212 int r = DM_MAPIO_REMAPPED;
214 if (fail_read_on_bb) {
215 thisblock >>= dd->sect_per_block_shift;
216 spin_lock_irqsave(&dd->dust_lock, flags);
217 r = __dust_map_write(dd, thisblock);
218 spin_unlock_irqrestore(&dd->dust_lock, flags);
224 static int dust_map(struct dm_target *ti, struct bio *bio)
226 struct dust_device *dd = ti->private;
229 bio_set_dev(bio, dd->dev->bdev);
230 bio->bi_iter.bi_sector = dd->start + dm_target_offset(ti, bio->bi_iter.bi_sector);
232 if (bio_data_dir(bio) == READ)
233 r = dust_map_read(dd, bio->bi_iter.bi_sector, dd->fail_read_on_bb);
235 r = dust_map_write(dd, bio->bi_iter.bi_sector, dd->fail_read_on_bb);
240 static bool __dust_clear_badblocks(struct rb_root *tree,
241 unsigned long long count)
243 struct rb_node *node = NULL, *nnode = NULL;
245 nnode = rb_first(tree);
253 nnode = rb_next(node);
254 rb_erase(node, tree);
259 BUG_ON(tree->rb_node != NULL);
264 static int dust_clear_badblocks(struct dust_device *dd, char *result, unsigned int maxlen,
265 unsigned int *sz_ptr)
268 struct rb_root badblocklist;
269 unsigned long long badblock_count;
270 unsigned int sz = *sz_ptr;
272 spin_lock_irqsave(&dd->dust_lock, flags);
273 badblocklist = dd->badblocklist;
274 badblock_count = dd->badblock_count;
275 dd->badblocklist = RB_ROOT;
276 dd->badblock_count = 0;
277 spin_unlock_irqrestore(&dd->dust_lock, flags);
279 if (!__dust_clear_badblocks(&badblocklist, badblock_count))
280 DMEMIT("%s: no badblocks found", __func__);
282 DMEMIT("%s: badblocks cleared", __func__);
287 static int dust_list_badblocks(struct dust_device *dd, char *result, unsigned int maxlen,
288 unsigned int *sz_ptr)
291 struct rb_root badblocklist;
292 struct rb_node *node;
293 struct badblock *bblk;
294 unsigned int sz = *sz_ptr;
295 unsigned long long num = 0;
297 spin_lock_irqsave(&dd->dust_lock, flags);
298 badblocklist = dd->badblocklist;
299 for (node = rb_first(&badblocklist); node; node = rb_next(node)) {
300 bblk = rb_entry(node, struct badblock, node);
301 DMEMIT("%llu\n", bblk->bb);
305 spin_unlock_irqrestore(&dd->dust_lock, flags);
307 DMEMIT("No blocks in badblocklist");
315 * <device_path> <offset> <blksz>
317 * device_path: path to the block device
318 * offset: offset to data area from start of device_path
319 * blksz: block size (minimum 512, maximum 1073741824, must be a power of 2)
321 static int dust_ctr(struct dm_target *ti, unsigned int argc, char **argv)
323 struct dust_device *dd;
324 unsigned long long tmp;
327 unsigned int sect_per_block;
328 sector_t DUST_MAX_BLKSZ_SECTORS = 2097152;
329 sector_t max_block_sectors = min(ti->len, DUST_MAX_BLKSZ_SECTORS);
332 ti->error = "Invalid argument count";
336 if (kstrtouint(argv[2], 10, &blksz) || !blksz) {
337 ti->error = "Invalid block size parameter";
342 ti->error = "Block size must be at least 512";
346 if (!is_power_of_2(blksz)) {
347 ti->error = "Block size must be a power of 2";
351 if (to_sector(blksz) > max_block_sectors) {
352 ti->error = "Block size is too large";
356 sect_per_block = (blksz >> SECTOR_SHIFT);
358 if (sscanf(argv[1], "%llu%c", &tmp, &dummy) != 1 || tmp != (sector_t)tmp) {
359 ti->error = "Invalid device offset sector";
363 dd = kzalloc(sizeof(struct dust_device), GFP_KERNEL);
365 ti->error = "Cannot allocate context";
369 if (dm_get_device(ti, argv[0], dm_table_get_mode(ti->table), &dd->dev)) {
370 ti->error = "Device lookup failed";
375 dd->sect_per_block = sect_per_block;
379 dd->sect_per_block_shift = __ffs(sect_per_block);
382 * Whether to fail a read on a "bad" block.
383 * Defaults to false; enabled later by message.
385 dd->fail_read_on_bb = false;
388 * Initialize bad block list rbtree.
390 dd->badblocklist = RB_ROOT;
391 dd->badblock_count = 0;
392 spin_lock_init(&dd->dust_lock);
394 dd->quiet_mode = false;
396 BUG_ON(dm_set_target_max_io_len(ti, dd->sect_per_block) != 0);
398 ti->num_discard_bios = 1;
399 ti->num_flush_bios = 1;
405 static void dust_dtr(struct dm_target *ti)
407 struct dust_device *dd = ti->private;
409 __dust_clear_badblocks(&dd->badblocklist, dd->badblock_count);
410 dm_put_device(ti, dd->dev);
414 static int dust_message(struct dm_target *ti, unsigned int argc, char **argv,
415 char *result, unsigned int maxlen)
417 struct dust_device *dd = ti->private;
418 sector_t size = bdev_nr_sectors(dd->dev->bdev);
419 bool invalid_msg = false;
421 unsigned long long tmp, block;
422 unsigned char wr_fail_cnt;
429 if (!strcasecmp(argv[0], "addbadblock") ||
430 !strcasecmp(argv[0], "removebadblock") ||
431 !strcasecmp(argv[0], "queryblock")) {
432 DMERR("%s requires an additional argument", argv[0]);
433 } else if (!strcasecmp(argv[0], "disable")) {
434 DMINFO("disabling read failures on bad sectors");
435 dd->fail_read_on_bb = false;
437 } else if (!strcasecmp(argv[0], "enable")) {
438 DMINFO("enabling read failures on bad sectors");
439 dd->fail_read_on_bb = true;
441 } else if (!strcasecmp(argv[0], "countbadblocks")) {
442 spin_lock_irqsave(&dd->dust_lock, flags);
443 DMEMIT("countbadblocks: %llu badblock(s) found",
445 spin_unlock_irqrestore(&dd->dust_lock, flags);
447 } else if (!strcasecmp(argv[0], "clearbadblocks")) {
448 r = dust_clear_badblocks(dd, result, maxlen, &sz);
449 } else if (!strcasecmp(argv[0], "quiet")) {
451 dd->quiet_mode = true;
453 dd->quiet_mode = false;
455 } else if (!strcasecmp(argv[0], "listbadblocks")) {
456 r = dust_list_badblocks(dd, result, maxlen, &sz);
460 } else if (argc == 2) {
461 if (sscanf(argv[1], "%llu%c", &tmp, &dummy) != 1)
465 sector_div(size, dd->sect_per_block);
467 DMERR("selected block value out of range");
471 if (!strcasecmp(argv[0], "addbadblock"))
472 r = dust_add_block(dd, block, 0);
473 else if (!strcasecmp(argv[0], "removebadblock"))
474 r = dust_remove_block(dd, block);
475 else if (!strcasecmp(argv[0], "queryblock"))
476 r = dust_query_block(dd, block, result, maxlen, &sz);
480 } else if (argc == 3) {
481 if (sscanf(argv[1], "%llu%c", &tmp, &dummy) != 1)
484 if (sscanf(argv[2], "%u%c", &tmp_ui, &dummy) != 1)
489 DMERR("selected write fail count out of range");
492 wr_fail_cnt = tmp_ui;
493 sector_div(size, dd->sect_per_block);
495 DMERR("selected block value out of range");
499 if (!strcasecmp(argv[0], "addbadblock"))
500 r = dust_add_block(dd, block, wr_fail_cnt);
505 DMERR("invalid number of arguments '%d'", argc);
508 DMERR("unrecognized message '%s' received", argv[0]);
513 static void dust_status(struct dm_target *ti, status_type_t type,
514 unsigned int status_flags, char *result, unsigned int maxlen)
516 struct dust_device *dd = ti->private;
520 case STATUSTYPE_INFO:
521 DMEMIT("%s %s %s", dd->dev->name,
522 dd->fail_read_on_bb ? "fail_read_on_bad_block" : "bypass",
523 dd->quiet_mode ? "quiet" : "verbose");
526 case STATUSTYPE_TABLE:
527 DMEMIT("%s %llu %u", dd->dev->name,
528 (unsigned long long)dd->start, dd->blksz);
537 static int dust_prepare_ioctl(struct dm_target *ti, struct block_device **bdev)
539 struct dust_device *dd = ti->private;
540 struct dm_dev *dev = dd->dev;
545 * Only pass ioctls through if the device sizes match exactly.
547 if (dd->start || ti->len != bdev_nr_sectors(dev->bdev))
553 static int dust_iterate_devices(struct dm_target *ti, iterate_devices_callout_fn fn,
556 struct dust_device *dd = ti->private;
558 return fn(ti, dd->dev, dd->start, ti->len, data);
561 static struct target_type dust_target = {
563 .version = {1, 0, 0},
564 .module = THIS_MODULE,
567 .iterate_devices = dust_iterate_devices,
569 .message = dust_message,
570 .status = dust_status,
571 .prepare_ioctl = dust_prepare_ioctl,
575 MODULE_DESCRIPTION(DM_NAME " dust test target");
576 MODULE_AUTHOR("Bryan Gurney <dm-devel@redhat.com>");
577 MODULE_LICENSE("GPL");