1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * Functions related to sysfs handling
4 */
5 #include <linux/kernel.h>
6 #include <linux/slab.h>
7 #include <linux/module.h>
8 #include <linux/bio.h>
9 #include <linux/blkdev.h>
10 #include <linux/backing-dev.h>
11 #include <linux/blktrace_api.h>
12 #include <linux/debugfs.h>
13
14 #include "blk.h"
15 #include "blk-mq.h"
16 #include "blk-mq-debugfs.h"
17 #include "blk-mq-sched.h"
18 #include "blk-rq-qos.h"
19 #include "blk-wbt.h"
20 #include "blk-cgroup.h"
21 #include "blk-throttle.h"
22
23 struct queue_sysfs_entry {
24 struct attribute attr;
25 ssize_t (*show)(struct gendisk *disk, char *page);
26 int (*load_module)(struct gendisk *disk, const char *page, size_t count);
27 ssize_t (*store)(struct gendisk *disk, const char *page, size_t count);
28 int (*store_limit)(struct gendisk *disk, const char *page,
29 size_t count, struct queue_limits *lim);
30 };
31
32 static ssize_t
queue_var_show(unsigned long var,char * page)33 queue_var_show(unsigned long var, char *page)
34 {
35 return sprintf(page, "%lu\n", var);
36 }
37
38 static ssize_t
queue_var_store(unsigned long * var,const char * page,size_t count)39 queue_var_store(unsigned long *var, const char *page, size_t count)
40 {
41 int err;
42 unsigned long v;
43
44 err = kstrtoul(page, 10, &v);
45 if (err || v > UINT_MAX)
46 return -EINVAL;
47
48 *var = v;
49
50 return count;
51 }
52
queue_requests_show(struct gendisk * disk,char * page)53 static ssize_t queue_requests_show(struct gendisk *disk, char *page)
54 {
55 return queue_var_show(disk->queue->nr_requests, page);
56 }
57
58 static ssize_t
queue_requests_store(struct gendisk * disk,const char * page,size_t count)59 queue_requests_store(struct gendisk *disk, const char *page, size_t count)
60 {
61 unsigned long nr;
62 int ret, err;
63
64 if (!queue_is_mq(disk->queue))
65 return -EINVAL;
66
67 ret = queue_var_store(&nr, page, count);
68 if (ret < 0)
69 return ret;
70
71 if (nr < BLKDEV_MIN_RQ)
72 nr = BLKDEV_MIN_RQ;
73
74 err = blk_mq_update_nr_requests(disk->queue, nr);
75 if (err)
76 return err;
77
78 return ret;
79 }
80
queue_ra_show(struct gendisk * disk,char * page)81 static ssize_t queue_ra_show(struct gendisk *disk, char *page)
82 {
83 return queue_var_show(disk->bdi->ra_pages << (PAGE_SHIFT - 10), page);
84 }
85
86 static ssize_t
queue_ra_store(struct gendisk * disk,const char * page,size_t count)87 queue_ra_store(struct gendisk *disk, const char *page, size_t count)
88 {
89 unsigned long ra_kb;
90 ssize_t ret;
91
92 ret = queue_var_store(&ra_kb, page, count);
93 if (ret < 0)
94 return ret;
95 disk->bdi->ra_pages = ra_kb >> (PAGE_SHIFT - 10);
96 return ret;
97 }
98
99 #define QUEUE_SYSFS_LIMIT_SHOW(_field) \
100 static ssize_t queue_##_field##_show(struct gendisk *disk, char *page) \
101 { \
102 return queue_var_show(disk->queue->limits._field, page); \
103 }
104
105 QUEUE_SYSFS_LIMIT_SHOW(max_segments)
QUEUE_SYSFS_LIMIT_SHOW(max_discard_segments)106 QUEUE_SYSFS_LIMIT_SHOW(max_discard_segments)
107 QUEUE_SYSFS_LIMIT_SHOW(max_integrity_segments)
108 QUEUE_SYSFS_LIMIT_SHOW(max_segment_size)
109 QUEUE_SYSFS_LIMIT_SHOW(logical_block_size)
110 QUEUE_SYSFS_LIMIT_SHOW(physical_block_size)
111 QUEUE_SYSFS_LIMIT_SHOW(chunk_sectors)
112 QUEUE_SYSFS_LIMIT_SHOW(io_min)
113 QUEUE_SYSFS_LIMIT_SHOW(io_opt)
114 QUEUE_SYSFS_LIMIT_SHOW(discard_granularity)
115 QUEUE_SYSFS_LIMIT_SHOW(zone_write_granularity)
116 QUEUE_SYSFS_LIMIT_SHOW(virt_boundary_mask)
117 QUEUE_SYSFS_LIMIT_SHOW(dma_alignment)
118 QUEUE_SYSFS_LIMIT_SHOW(max_open_zones)
119 QUEUE_SYSFS_LIMIT_SHOW(max_active_zones)
120 QUEUE_SYSFS_LIMIT_SHOW(atomic_write_unit_min)
121 QUEUE_SYSFS_LIMIT_SHOW(atomic_write_unit_max)
122
123 #define QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_BYTES(_field) \
124 static ssize_t queue_##_field##_show(struct gendisk *disk, char *page) \
125 { \
126 return sprintf(page, "%llu\n", \
127 (unsigned long long)disk->queue->limits._field << \
128 SECTOR_SHIFT); \
129 }
130
131 QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_BYTES(max_discard_sectors)
132 QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_BYTES(max_hw_discard_sectors)
133 QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_BYTES(max_write_zeroes_sectors)
134 QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_BYTES(atomic_write_max_sectors)
135 QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_BYTES(atomic_write_boundary_sectors)
136
137 #define QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_KB(_field) \
138 static ssize_t queue_##_field##_show(struct gendisk *disk, char *page) \
139 { \
140 return queue_var_show(disk->queue->limits._field >> 1, page); \
141 }
142
143 QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_KB(max_sectors)
144 QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_KB(max_hw_sectors)
145
146 #define QUEUE_SYSFS_SHOW_CONST(_name, _val) \
147 static ssize_t queue_##_name##_show(struct gendisk *disk, char *page) \
148 { \
149 return sprintf(page, "%d\n", _val); \
150 }
151
152 /* deprecated fields */
153 QUEUE_SYSFS_SHOW_CONST(discard_zeroes_data, 0)
154 QUEUE_SYSFS_SHOW_CONST(write_same_max, 0)
155 QUEUE_SYSFS_SHOW_CONST(poll_delay, -1)
156
157 static int queue_max_discard_sectors_store(struct gendisk *disk,
158 const char *page, size_t count, struct queue_limits *lim)
159 {
160 unsigned long max_discard_bytes;
161 ssize_t ret;
162
163 ret = queue_var_store(&max_discard_bytes, page, count);
164 if (ret < 0)
165 return ret;
166
167 if (max_discard_bytes & (disk->queue->limits.discard_granularity - 1))
168 return -EINVAL;
169
170 if ((max_discard_bytes >> SECTOR_SHIFT) > UINT_MAX)
171 return -EINVAL;
172
173 lim->max_user_discard_sectors = max_discard_bytes >> SECTOR_SHIFT;
174 return 0;
175 }
176
177 /*
178 * For zone append queue_max_zone_append_sectors does not just return the
179 * underlying queue limits, but actually contains a calculation. Because of
180 * that we can't simply use QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_BYTES here.
181 */
queue_zone_append_max_show(struct gendisk * disk,char * page)182 static ssize_t queue_zone_append_max_show(struct gendisk *disk, char *page)
183 {
184 return sprintf(page, "%llu\n",
185 (u64)queue_max_zone_append_sectors(disk->queue) <<
186 SECTOR_SHIFT);
187 }
188
189 static int
queue_max_sectors_store(struct gendisk * disk,const char * page,size_t count,struct queue_limits * lim)190 queue_max_sectors_store(struct gendisk *disk, const char *page, size_t count,
191 struct queue_limits *lim)
192 {
193 unsigned long max_sectors_kb;
194 ssize_t ret;
195
196 ret = queue_var_store(&max_sectors_kb, page, count);
197 if (ret < 0)
198 return ret;
199
200 lim->max_user_sectors = max_sectors_kb << 1;
201 return 0;
202 }
203
queue_feature_store(struct gendisk * disk,const char * page,size_t count,struct queue_limits * lim,blk_features_t feature)204 static ssize_t queue_feature_store(struct gendisk *disk, const char *page,
205 size_t count, struct queue_limits *lim, blk_features_t feature)
206 {
207 unsigned long val;
208 ssize_t ret;
209
210 ret = queue_var_store(&val, page, count);
211 if (ret < 0)
212 return ret;
213
214 if (val)
215 lim->features |= feature;
216 else
217 lim->features &= ~feature;
218 return 0;
219 }
220
221 #define QUEUE_SYSFS_FEATURE(_name, _feature) \
222 static ssize_t queue_##_name##_show(struct gendisk *disk, char *page) \
223 { \
224 return sprintf(page, "%u\n", \
225 !!(disk->queue->limits.features & _feature)); \
226 } \
227 static int queue_##_name##_store(struct gendisk *disk, \
228 const char *page, size_t count, struct queue_limits *lim) \
229 { \
230 return queue_feature_store(disk, page, count, lim, _feature); \
231 }
232
233 QUEUE_SYSFS_FEATURE(rotational, BLK_FEAT_ROTATIONAL)
234 QUEUE_SYSFS_FEATURE(add_random, BLK_FEAT_ADD_RANDOM)
235 QUEUE_SYSFS_FEATURE(iostats, BLK_FEAT_IO_STAT)
236 QUEUE_SYSFS_FEATURE(stable_writes, BLK_FEAT_STABLE_WRITES);
237
238 #define QUEUE_SYSFS_FEATURE_SHOW(_name, _feature) \
239 static ssize_t queue_##_name##_show(struct gendisk *disk, char *page) \
240 { \
241 return sprintf(page, "%u\n", \
242 !!(disk->queue->limits.features & _feature)); \
243 }
244
245 QUEUE_SYSFS_FEATURE_SHOW(fua, BLK_FEAT_FUA);
246 QUEUE_SYSFS_FEATURE_SHOW(dax, BLK_FEAT_DAX);
247
queue_poll_show(struct gendisk * disk,char * page)248 static ssize_t queue_poll_show(struct gendisk *disk, char *page)
249 {
250 if (queue_is_mq(disk->queue))
251 return sysfs_emit(page, "%u\n", blk_mq_can_poll(disk->queue));
252 return sysfs_emit(page, "%u\n",
253 !!(disk->queue->limits.features & BLK_FEAT_POLL));
254 }
255
queue_zoned_show(struct gendisk * disk,char * page)256 static ssize_t queue_zoned_show(struct gendisk *disk, char *page)
257 {
258 if (blk_queue_is_zoned(disk->queue))
259 return sprintf(page, "host-managed\n");
260 return sprintf(page, "none\n");
261 }
262
queue_nr_zones_show(struct gendisk * disk,char * page)263 static ssize_t queue_nr_zones_show(struct gendisk *disk, char *page)
264 {
265 return queue_var_show(disk_nr_zones(disk), page);
266 }
267
queue_nomerges_show(struct gendisk * disk,char * page)268 static ssize_t queue_nomerges_show(struct gendisk *disk, char *page)
269 {
270 return queue_var_show((blk_queue_nomerges(disk->queue) << 1) |
271 blk_queue_noxmerges(disk->queue), page);
272 }
273
queue_nomerges_store(struct gendisk * disk,const char * page,size_t count)274 static ssize_t queue_nomerges_store(struct gendisk *disk, const char *page,
275 size_t count)
276 {
277 unsigned long nm;
278 ssize_t ret = queue_var_store(&nm, page, count);
279
280 if (ret < 0)
281 return ret;
282
283 blk_queue_flag_clear(QUEUE_FLAG_NOMERGES, disk->queue);
284 blk_queue_flag_clear(QUEUE_FLAG_NOXMERGES, disk->queue);
285 if (nm == 2)
286 blk_queue_flag_set(QUEUE_FLAG_NOMERGES, disk->queue);
287 else if (nm)
288 blk_queue_flag_set(QUEUE_FLAG_NOXMERGES, disk->queue);
289
290 return ret;
291 }
292
queue_rq_affinity_show(struct gendisk * disk,char * page)293 static ssize_t queue_rq_affinity_show(struct gendisk *disk, char *page)
294 {
295 bool set = test_bit(QUEUE_FLAG_SAME_COMP, &disk->queue->queue_flags);
296 bool force = test_bit(QUEUE_FLAG_SAME_FORCE, &disk->queue->queue_flags);
297
298 return queue_var_show(set << force, page);
299 }
300
301 static ssize_t
queue_rq_affinity_store(struct gendisk * disk,const char * page,size_t count)302 queue_rq_affinity_store(struct gendisk *disk, const char *page, size_t count)
303 {
304 ssize_t ret = -EINVAL;
305 #ifdef CONFIG_SMP
306 struct request_queue *q = disk->queue;
307 unsigned long val;
308
309 ret = queue_var_store(&val, page, count);
310 if (ret < 0)
311 return ret;
312
313 if (val == 2) {
314 blk_queue_flag_set(QUEUE_FLAG_SAME_COMP, q);
315 blk_queue_flag_set(QUEUE_FLAG_SAME_FORCE, q);
316 } else if (val == 1) {
317 blk_queue_flag_set(QUEUE_FLAG_SAME_COMP, q);
318 blk_queue_flag_clear(QUEUE_FLAG_SAME_FORCE, q);
319 } else if (val == 0) {
320 blk_queue_flag_clear(QUEUE_FLAG_SAME_COMP, q);
321 blk_queue_flag_clear(QUEUE_FLAG_SAME_FORCE, q);
322 }
323 #endif
324 return ret;
325 }
326
queue_poll_delay_store(struct gendisk * disk,const char * page,size_t count)327 static ssize_t queue_poll_delay_store(struct gendisk *disk, const char *page,
328 size_t count)
329 {
330 return count;
331 }
332
queue_poll_store(struct gendisk * disk,const char * page,size_t count)333 static ssize_t queue_poll_store(struct gendisk *disk, const char *page,
334 size_t count)
335 {
336 if (!(disk->queue->limits.features & BLK_FEAT_POLL))
337 return -EINVAL;
338 pr_info_ratelimited("writes to the poll attribute are ignored.\n");
339 pr_info_ratelimited("please use driver specific parameters instead.\n");
340 return count;
341 }
342
queue_io_timeout_show(struct gendisk * disk,char * page)343 static ssize_t queue_io_timeout_show(struct gendisk *disk, char *page)
344 {
345 return sprintf(page, "%u\n", jiffies_to_msecs(disk->queue->rq_timeout));
346 }
347
queue_io_timeout_store(struct gendisk * disk,const char * page,size_t count)348 static ssize_t queue_io_timeout_store(struct gendisk *disk, const char *page,
349 size_t count)
350 {
351 unsigned int val;
352 int err;
353
354 err = kstrtou32(page, 10, &val);
355 if (err || val == 0)
356 return -EINVAL;
357
358 blk_queue_rq_timeout(disk->queue, msecs_to_jiffies(val));
359
360 return count;
361 }
362
queue_wc_show(struct gendisk * disk,char * page)363 static ssize_t queue_wc_show(struct gendisk *disk, char *page)
364 {
365 if (blk_queue_write_cache(disk->queue))
366 return sprintf(page, "write back\n");
367 return sprintf(page, "write through\n");
368 }
369
queue_wc_store(struct gendisk * disk,const char * page,size_t count,struct queue_limits * lim)370 static int queue_wc_store(struct gendisk *disk, const char *page,
371 size_t count, struct queue_limits *lim)
372 {
373 bool disable;
374
375 if (!strncmp(page, "write back", 10)) {
376 disable = false;
377 } else if (!strncmp(page, "write through", 13) ||
378 !strncmp(page, "none", 4)) {
379 disable = true;
380 } else {
381 return -EINVAL;
382 }
383
384 if (disable)
385 lim->flags |= BLK_FLAG_WRITE_CACHE_DISABLED;
386 else
387 lim->flags &= ~BLK_FLAG_WRITE_CACHE_DISABLED;
388 return 0;
389 }
390
391 #define QUEUE_RO_ENTRY(_prefix, _name) \
392 static struct queue_sysfs_entry _prefix##_entry = { \
393 .attr = { .name = _name, .mode = 0444 }, \
394 .show = _prefix##_show, \
395 };
396
397 #define QUEUE_RW_ENTRY(_prefix, _name) \
398 static struct queue_sysfs_entry _prefix##_entry = { \
399 .attr = { .name = _name, .mode = 0644 }, \
400 .show = _prefix##_show, \
401 .store = _prefix##_store, \
402 };
403
404 #define QUEUE_LIM_RW_ENTRY(_prefix, _name) \
405 static struct queue_sysfs_entry _prefix##_entry = { \
406 .attr = { .name = _name, .mode = 0644 }, \
407 .show = _prefix##_show, \
408 .store_limit = _prefix##_store, \
409 }
410
411 #define QUEUE_RW_LOAD_MODULE_ENTRY(_prefix, _name) \
412 static struct queue_sysfs_entry _prefix##_entry = { \
413 .attr = { .name = _name, .mode = 0644 }, \
414 .show = _prefix##_show, \
415 .load_module = _prefix##_load_module, \
416 .store = _prefix##_store, \
417 }
418
419 QUEUE_RW_ENTRY(queue_requests, "nr_requests");
420 QUEUE_RW_ENTRY(queue_ra, "read_ahead_kb");
421 QUEUE_LIM_RW_ENTRY(queue_max_sectors, "max_sectors_kb");
422 QUEUE_RO_ENTRY(queue_max_hw_sectors, "max_hw_sectors_kb");
423 QUEUE_RO_ENTRY(queue_max_segments, "max_segments");
424 QUEUE_RO_ENTRY(queue_max_integrity_segments, "max_integrity_segments");
425 QUEUE_RO_ENTRY(queue_max_segment_size, "max_segment_size");
426 QUEUE_RW_LOAD_MODULE_ENTRY(elv_iosched, "scheduler");
427
428 QUEUE_RO_ENTRY(queue_logical_block_size, "logical_block_size");
429 QUEUE_RO_ENTRY(queue_physical_block_size, "physical_block_size");
430 QUEUE_RO_ENTRY(queue_chunk_sectors, "chunk_sectors");
431 QUEUE_RO_ENTRY(queue_io_min, "minimum_io_size");
432 QUEUE_RO_ENTRY(queue_io_opt, "optimal_io_size");
433
434 QUEUE_RO_ENTRY(queue_max_discard_segments, "max_discard_segments");
435 QUEUE_RO_ENTRY(queue_discard_granularity, "discard_granularity");
436 QUEUE_RO_ENTRY(queue_max_hw_discard_sectors, "discard_max_hw_bytes");
437 QUEUE_LIM_RW_ENTRY(queue_max_discard_sectors, "discard_max_bytes");
438 QUEUE_RO_ENTRY(queue_discard_zeroes_data, "discard_zeroes_data");
439
440 QUEUE_RO_ENTRY(queue_atomic_write_max_sectors, "atomic_write_max_bytes");
441 QUEUE_RO_ENTRY(queue_atomic_write_boundary_sectors,
442 "atomic_write_boundary_bytes");
443 QUEUE_RO_ENTRY(queue_atomic_write_unit_max, "atomic_write_unit_max_bytes");
444 QUEUE_RO_ENTRY(queue_atomic_write_unit_min, "atomic_write_unit_min_bytes");
445
446 QUEUE_RO_ENTRY(queue_write_same_max, "write_same_max_bytes");
447 QUEUE_RO_ENTRY(queue_max_write_zeroes_sectors, "write_zeroes_max_bytes");
448 QUEUE_RO_ENTRY(queue_zone_append_max, "zone_append_max_bytes");
449 QUEUE_RO_ENTRY(queue_zone_write_granularity, "zone_write_granularity");
450
451 QUEUE_RO_ENTRY(queue_zoned, "zoned");
452 QUEUE_RO_ENTRY(queue_nr_zones, "nr_zones");
453 QUEUE_RO_ENTRY(queue_max_open_zones, "max_open_zones");
454 QUEUE_RO_ENTRY(queue_max_active_zones, "max_active_zones");
455
456 QUEUE_RW_ENTRY(queue_nomerges, "nomerges");
457 QUEUE_RW_ENTRY(queue_rq_affinity, "rq_affinity");
458 QUEUE_RW_ENTRY(queue_poll, "io_poll");
459 QUEUE_RW_ENTRY(queue_poll_delay, "io_poll_delay");
460 QUEUE_LIM_RW_ENTRY(queue_wc, "write_cache");
461 QUEUE_RO_ENTRY(queue_fua, "fua");
462 QUEUE_RO_ENTRY(queue_dax, "dax");
463 QUEUE_RW_ENTRY(queue_io_timeout, "io_timeout");
464 QUEUE_RO_ENTRY(queue_virt_boundary_mask, "virt_boundary_mask");
465 QUEUE_RO_ENTRY(queue_dma_alignment, "dma_alignment");
466
467 /* legacy alias for logical_block_size: */
468 static struct queue_sysfs_entry queue_hw_sector_size_entry = {
469 .attr = {.name = "hw_sector_size", .mode = 0444 },
470 .show = queue_logical_block_size_show,
471 };
472
473 QUEUE_LIM_RW_ENTRY(queue_rotational, "rotational");
474 QUEUE_LIM_RW_ENTRY(queue_iostats, "iostats");
475 QUEUE_LIM_RW_ENTRY(queue_add_random, "add_random");
476 QUEUE_LIM_RW_ENTRY(queue_stable_writes, "stable_writes");
477
478 #ifdef CONFIG_BLK_WBT
queue_var_store64(s64 * var,const char * page)479 static ssize_t queue_var_store64(s64 *var, const char *page)
480 {
481 int err;
482 s64 v;
483
484 err = kstrtos64(page, 10, &v);
485 if (err < 0)
486 return err;
487
488 *var = v;
489 return 0;
490 }
491
queue_wb_lat_show(struct gendisk * disk,char * page)492 static ssize_t queue_wb_lat_show(struct gendisk *disk, char *page)
493 {
494 if (!wbt_rq_qos(disk->queue))
495 return -EINVAL;
496
497 if (wbt_disabled(disk->queue))
498 return sprintf(page, "0\n");
499
500 return sprintf(page, "%llu\n",
501 div_u64(wbt_get_min_lat(disk->queue), 1000));
502 }
503
queue_wb_lat_store(struct gendisk * disk,const char * page,size_t count)504 static ssize_t queue_wb_lat_store(struct gendisk *disk, const char *page,
505 size_t count)
506 {
507 struct request_queue *q = disk->queue;
508 struct rq_qos *rqos;
509 ssize_t ret;
510 s64 val;
511
512 ret = queue_var_store64(&val, page);
513 if (ret < 0)
514 return ret;
515 if (val < -1)
516 return -EINVAL;
517
518 rqos = wbt_rq_qos(q);
519 if (!rqos) {
520 ret = wbt_init(disk);
521 if (ret)
522 return ret;
523 }
524
525 if (val == -1)
526 val = wbt_default_latency_nsec(q);
527 else if (val >= 0)
528 val *= 1000ULL;
529
530 if (wbt_get_min_lat(q) == val)
531 return count;
532
533 /*
534 * Ensure that the queue is idled, in case the latency update
535 * ends up either enabling or disabling wbt completely. We can't
536 * have IO inflight if that happens.
537 */
538 blk_mq_quiesce_queue(q);
539
540 wbt_set_min_lat(q, val);
541
542 blk_mq_unquiesce_queue(q);
543
544 return count;
545 }
546
547 QUEUE_RW_ENTRY(queue_wb_lat, "wbt_lat_usec");
548 #endif
549
550 /* Common attributes for bio-based and request-based queues. */
551 static struct attribute *queue_attrs[] = {
552 &queue_ra_entry.attr,
553 &queue_max_hw_sectors_entry.attr,
554 &queue_max_sectors_entry.attr,
555 &queue_max_segments_entry.attr,
556 &queue_max_discard_segments_entry.attr,
557 &queue_max_integrity_segments_entry.attr,
558 &queue_max_segment_size_entry.attr,
559 &queue_hw_sector_size_entry.attr,
560 &queue_logical_block_size_entry.attr,
561 &queue_physical_block_size_entry.attr,
562 &queue_chunk_sectors_entry.attr,
563 &queue_io_min_entry.attr,
564 &queue_io_opt_entry.attr,
565 &queue_discard_granularity_entry.attr,
566 &queue_max_discard_sectors_entry.attr,
567 &queue_max_hw_discard_sectors_entry.attr,
568 &queue_discard_zeroes_data_entry.attr,
569 &queue_atomic_write_max_sectors_entry.attr,
570 &queue_atomic_write_boundary_sectors_entry.attr,
571 &queue_atomic_write_unit_min_entry.attr,
572 &queue_atomic_write_unit_max_entry.attr,
573 &queue_write_same_max_entry.attr,
574 &queue_max_write_zeroes_sectors_entry.attr,
575 &queue_zone_append_max_entry.attr,
576 &queue_zone_write_granularity_entry.attr,
577 &queue_rotational_entry.attr,
578 &queue_zoned_entry.attr,
579 &queue_nr_zones_entry.attr,
580 &queue_max_open_zones_entry.attr,
581 &queue_max_active_zones_entry.attr,
582 &queue_nomerges_entry.attr,
583 &queue_iostats_entry.attr,
584 &queue_stable_writes_entry.attr,
585 &queue_add_random_entry.attr,
586 &queue_poll_entry.attr,
587 &queue_wc_entry.attr,
588 &queue_fua_entry.attr,
589 &queue_dax_entry.attr,
590 &queue_poll_delay_entry.attr,
591 &queue_virt_boundary_mask_entry.attr,
592 &queue_dma_alignment_entry.attr,
593 NULL,
594 };
595
596 /* Request-based queue attributes that are not relevant for bio-based queues. */
597 static struct attribute *blk_mq_queue_attrs[] = {
598 &queue_requests_entry.attr,
599 &elv_iosched_entry.attr,
600 &queue_rq_affinity_entry.attr,
601 &queue_io_timeout_entry.attr,
602 #ifdef CONFIG_BLK_WBT
603 &queue_wb_lat_entry.attr,
604 #endif
605 NULL,
606 };
607
queue_attr_visible(struct kobject * kobj,struct attribute * attr,int n)608 static umode_t queue_attr_visible(struct kobject *kobj, struct attribute *attr,
609 int n)
610 {
611 struct gendisk *disk = container_of(kobj, struct gendisk, queue_kobj);
612 struct request_queue *q = disk->queue;
613
614 if ((attr == &queue_max_open_zones_entry.attr ||
615 attr == &queue_max_active_zones_entry.attr) &&
616 !blk_queue_is_zoned(q))
617 return 0;
618
619 return attr->mode;
620 }
621
blk_mq_queue_attr_visible(struct kobject * kobj,struct attribute * attr,int n)622 static umode_t blk_mq_queue_attr_visible(struct kobject *kobj,
623 struct attribute *attr, int n)
624 {
625 struct gendisk *disk = container_of(kobj, struct gendisk, queue_kobj);
626 struct request_queue *q = disk->queue;
627
628 if (!queue_is_mq(q))
629 return 0;
630
631 if (attr == &queue_io_timeout_entry.attr && !q->mq_ops->timeout)
632 return 0;
633
634 return attr->mode;
635 }
636
637 static struct attribute_group queue_attr_group = {
638 .attrs = queue_attrs,
639 .is_visible = queue_attr_visible,
640 };
641
642 static struct attribute_group blk_mq_queue_attr_group = {
643 .attrs = blk_mq_queue_attrs,
644 .is_visible = blk_mq_queue_attr_visible,
645 };
646
647 #define to_queue(atr) container_of((atr), struct queue_sysfs_entry, attr)
648
649 static ssize_t
queue_attr_show(struct kobject * kobj,struct attribute * attr,char * page)650 queue_attr_show(struct kobject *kobj, struct attribute *attr, char *page)
651 {
652 struct queue_sysfs_entry *entry = to_queue(attr);
653 struct gendisk *disk = container_of(kobj, struct gendisk, queue_kobj);
654 ssize_t res;
655
656 if (!entry->show)
657 return -EIO;
658 mutex_lock(&disk->queue->sysfs_lock);
659 res = entry->show(disk, page);
660 mutex_unlock(&disk->queue->sysfs_lock);
661 return res;
662 }
663
664 static ssize_t
queue_attr_store(struct kobject * kobj,struct attribute * attr,const char * page,size_t length)665 queue_attr_store(struct kobject *kobj, struct attribute *attr,
666 const char *page, size_t length)
667 {
668 struct queue_sysfs_entry *entry = to_queue(attr);
669 struct gendisk *disk = container_of(kobj, struct gendisk, queue_kobj);
670 struct request_queue *q = disk->queue;
671 ssize_t res;
672
673 if (!entry->store_limit && !entry->store)
674 return -EIO;
675
676 /*
677 * If the attribute needs to load a module, do it before freezing the
678 * queue to ensure that the module file can be read when the request
679 * queue is the one for the device storing the module file.
680 */
681 if (entry->load_module) {
682 res = entry->load_module(disk, page, length);
683 if (res)
684 return res;
685 }
686
687 if (entry->store_limit) {
688 struct queue_limits lim = queue_limits_start_update(q);
689
690 res = entry->store_limit(disk, page, length, &lim);
691 if (res < 0) {
692 queue_limits_cancel_update(q);
693 return res;
694 }
695
696 res = queue_limits_commit_update_frozen(q, &lim);
697 if (res)
698 return res;
699 return length;
700 }
701
702 mutex_lock(&q->sysfs_lock);
703 blk_mq_freeze_queue(q);
704 res = entry->store(disk, page, length);
705 blk_mq_unfreeze_queue(q);
706 mutex_unlock(&q->sysfs_lock);
707 return res;
708 }
709
710 static const struct sysfs_ops queue_sysfs_ops = {
711 .show = queue_attr_show,
712 .store = queue_attr_store,
713 };
714
715 static const struct attribute_group *blk_queue_attr_groups[] = {
716 &queue_attr_group,
717 &blk_mq_queue_attr_group,
718 NULL
719 };
720
blk_queue_release(struct kobject * kobj)721 static void blk_queue_release(struct kobject *kobj)
722 {
723 /* nothing to do here, all data is associated with the parent gendisk */
724 }
725
726 static const struct kobj_type blk_queue_ktype = {
727 .default_groups = blk_queue_attr_groups,
728 .sysfs_ops = &queue_sysfs_ops,
729 .release = blk_queue_release,
730 };
731
blk_debugfs_remove(struct gendisk * disk)732 static void blk_debugfs_remove(struct gendisk *disk)
733 {
734 struct request_queue *q = disk->queue;
735
736 mutex_lock(&q->debugfs_mutex);
737 blk_trace_shutdown(q);
738 debugfs_remove_recursive(q->debugfs_dir);
739 q->debugfs_dir = NULL;
740 q->sched_debugfs_dir = NULL;
741 q->rqos_debugfs_dir = NULL;
742 mutex_unlock(&q->debugfs_mutex);
743 }
744
745 /**
746 * blk_register_queue - register a block layer queue with sysfs
747 * @disk: Disk of which the request queue should be registered with sysfs.
748 */
blk_register_queue(struct gendisk * disk)749 int blk_register_queue(struct gendisk *disk)
750 {
751 struct request_queue *q = disk->queue;
752 int ret;
753
754 mutex_lock(&q->sysfs_dir_lock);
755 kobject_init(&disk->queue_kobj, &blk_queue_ktype);
756 ret = kobject_add(&disk->queue_kobj, &disk_to_dev(disk)->kobj, "queue");
757 if (ret < 0)
758 goto out_put_queue_kobj;
759
760 if (queue_is_mq(q)) {
761 ret = blk_mq_sysfs_register(disk);
762 if (ret)
763 goto out_put_queue_kobj;
764 }
765 mutex_lock(&q->sysfs_lock);
766
767 mutex_lock(&q->debugfs_mutex);
768 q->debugfs_dir = debugfs_create_dir(disk->disk_name, blk_debugfs_root);
769 if (queue_is_mq(q))
770 blk_mq_debugfs_register(q);
771 mutex_unlock(&q->debugfs_mutex);
772
773 ret = disk_register_independent_access_ranges(disk);
774 if (ret)
775 goto out_debugfs_remove;
776
777 if (q->elevator) {
778 ret = elv_register_queue(q, false);
779 if (ret)
780 goto out_unregister_ia_ranges;
781 }
782
783 ret = blk_crypto_sysfs_register(disk);
784 if (ret)
785 goto out_elv_unregister;
786
787 blk_queue_flag_set(QUEUE_FLAG_REGISTERED, q);
788 wbt_enable_default(disk);
789
790 /* Now everything is ready and send out KOBJ_ADD uevent */
791 kobject_uevent(&disk->queue_kobj, KOBJ_ADD);
792 if (q->elevator)
793 kobject_uevent(&q->elevator->kobj, KOBJ_ADD);
794 mutex_unlock(&q->sysfs_lock);
795 mutex_unlock(&q->sysfs_dir_lock);
796
797 /*
798 * SCSI probing may synchronously create and destroy a lot of
799 * request_queues for non-existent devices. Shutting down a fully
800 * functional queue takes measureable wallclock time as RCU grace
801 * periods are involved. To avoid excessive latency in these
802 * cases, a request_queue starts out in a degraded mode which is
803 * faster to shut down and is made fully functional here as
804 * request_queues for non-existent devices never get registered.
805 */
806 blk_queue_flag_set(QUEUE_FLAG_INIT_DONE, q);
807 percpu_ref_switch_to_percpu(&q->q_usage_counter);
808
809 return ret;
810
811 out_elv_unregister:
812 elv_unregister_queue(q);
813 out_unregister_ia_ranges:
814 disk_unregister_independent_access_ranges(disk);
815 out_debugfs_remove:
816 blk_debugfs_remove(disk);
817 mutex_unlock(&q->sysfs_lock);
818 if (queue_is_mq(q))
819 blk_mq_sysfs_unregister(disk);
820 out_put_queue_kobj:
821 kobject_put(&disk->queue_kobj);
822 mutex_unlock(&q->sysfs_dir_lock);
823 return ret;
824 }
825
826 /**
827 * blk_unregister_queue - counterpart of blk_register_queue()
828 * @disk: Disk of which the request queue should be unregistered from sysfs.
829 *
830 * Note: the caller is responsible for guaranteeing that this function is called
831 * after blk_register_queue() has finished.
832 */
blk_unregister_queue(struct gendisk * disk)833 void blk_unregister_queue(struct gendisk *disk)
834 {
835 struct request_queue *q = disk->queue;
836
837 if (WARN_ON(!q))
838 return;
839
840 /* Return early if disk->queue was never registered. */
841 if (!blk_queue_registered(q))
842 return;
843
844 /*
845 * Since sysfs_remove_dir() prevents adding new directory entries
846 * before removal of existing entries starts, protect against
847 * concurrent elv_iosched_store() calls.
848 */
849 mutex_lock(&q->sysfs_lock);
850 blk_queue_flag_clear(QUEUE_FLAG_REGISTERED, q);
851 mutex_unlock(&q->sysfs_lock);
852
853 mutex_lock(&q->sysfs_dir_lock);
854 /*
855 * Remove the sysfs attributes before unregistering the queue data
856 * structures that can be modified through sysfs.
857 */
858 if (queue_is_mq(q))
859 blk_mq_sysfs_unregister(disk);
860 blk_crypto_sysfs_unregister(disk);
861
862 mutex_lock(&q->sysfs_lock);
863 elv_unregister_queue(q);
864 disk_unregister_independent_access_ranges(disk);
865 mutex_unlock(&q->sysfs_lock);
866
867 /* Now that we've deleted all child objects, we can delete the queue. */
868 kobject_uevent(&disk->queue_kobj, KOBJ_REMOVE);
869 kobject_del(&disk->queue_kobj);
870 mutex_unlock(&q->sysfs_dir_lock);
871
872 blk_debugfs_remove(disk);
873 kobject_put(&disk->queue_kobj);
874 }
875