Christoph Hellwig | 3dcf60bc | 2019-04-30 14:42:43 -0400 | [diff] [blame] | 1 | // SPDX-License-Identifier: GPL-2.0 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3 | * Copyright (C) 1991, 1992 Linus Torvalds |
| 4 | * Copyright (C) 1994, Karl Keyte: Added support for disk statistics |
| 5 | * Elevator latency, (C) 2000 Andrea Arcangeli <andrea@suse.de> SuSE |
| 6 | * Queue request tables / lock, selectable elevator, Jens Axboe <axboe@suse.de> |
Jens Axboe | 6728cb0 | 2008-01-31 13:03:55 +0100 | [diff] [blame] | 7 | * kernel-doc documentation started by NeilBrown <neilb@cse.unsw.edu.au> |
| 8 | * - July2000 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 9 | * bio rewrite, highmem i/o, etc, Jens Axboe <axboe@suse.de> - may 2001 |
| 10 | */ |
| 11 | |
| 12 | /* |
| 13 | * This handles all read/write requests to block devices |
| 14 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 15 | #include <linux/kernel.h> |
| 16 | #include <linux/module.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 17 | #include <linux/bio.h> |
| 18 | #include <linux/blkdev.h> |
Alan Stern | 52abca6 | 2020-12-08 21:29:51 -0800 | [diff] [blame] | 19 | #include <linux/blk-pm.h> |
Christoph Hellwig | fe45e63 | 2021-09-20 14:33:27 +0200 | [diff] [blame] | 20 | #include <linux/blk-integrity.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 21 | #include <linux/highmem.h> |
| 22 | #include <linux/mm.h> |
Matthew Wilcox (Oracle) | cee9a0c | 2020-06-01 21:46:07 -0700 | [diff] [blame] | 23 | #include <linux/pagemap.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 24 | #include <linux/kernel_stat.h> |
| 25 | #include <linux/string.h> |
| 26 | #include <linux/init.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 27 | #include <linux/completion.h> |
| 28 | #include <linux/slab.h> |
| 29 | #include <linux/swap.h> |
| 30 | #include <linux/writeback.h> |
Andrew Morton | faccbd4b | 2006-12-10 02:19:35 -0800 | [diff] [blame] | 31 | #include <linux/task_io_accounting_ops.h> |
Akinobu Mita | c17bb49 | 2006-12-08 02:39:46 -0800 | [diff] [blame] | 32 | #include <linux/fault-inject.h> |
Jens Axboe | 73c1010 | 2011-03-08 13:19:51 +0100 | [diff] [blame] | 33 | #include <linux/list_sort.h> |
Tejun Heo | e3c78ca | 2011-10-19 14:32:38 +0200 | [diff] [blame] | 34 | #include <linux/delay.h> |
Tejun Heo | aaf7c68 | 2012-04-19 16:29:22 -0700 | [diff] [blame] | 35 | #include <linux/ratelimit.h> |
Lin Ming | 6c95466 | 2013-03-23 11:42:26 +0800 | [diff] [blame] | 36 | #include <linux/pm_runtime.h> |
Max Gurtovoy | 54d4e6a | 2019-09-16 18:44:29 +0300 | [diff] [blame] | 37 | #include <linux/t10-pi.h> |
Omar Sandoval | 18fbda9 | 2017-01-31 14:53:20 -0800 | [diff] [blame] | 38 | #include <linux/debugfs.h> |
Howard McLauchlan | 30abb3a | 2018-02-06 14:05:39 -0800 | [diff] [blame] | 39 | #include <linux/bpf.h> |
Christoph Hellwig | 82d981d | 2021-11-23 19:53:12 +0100 | [diff] [blame] | 40 | #include <linux/part_stat.h> |
Ming Lei | 71ac860 | 2020-05-14 16:45:09 +0800 | [diff] [blame] | 41 | #include <linux/sched/sysctl.h> |
Satya Tangirala | a892c8d | 2020-05-14 00:37:18 +0000 | [diff] [blame] | 42 | #include <linux/blk-crypto.h> |
Li Zefan | 5578213 | 2009-06-09 13:43:05 +0800 | [diff] [blame] | 43 | |
| 44 | #define CREATE_TRACE_POINTS |
| 45 | #include <trace/events/block.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 46 | |
Jens Axboe | 8324aa9 | 2008-01-29 14:51:59 +0100 | [diff] [blame] | 47 | #include "blk.h" |
Christoph Hellwig | 2aa7745 | 2021-11-23 19:53:08 +0100 | [diff] [blame] | 48 | #include "blk-mq-sched.h" |
Bart Van Assche | bca6b06 | 2018-09-26 14:01:03 -0700 | [diff] [blame] | 49 | #include "blk-pm.h" |
Ming Lei | 672fdcf | 2022-02-11 18:11:49 +0800 | [diff] [blame] | 50 | #include "blk-cgroup.h" |
Jens Axboe | a7b36ee | 2021-10-05 09:11:56 -0600 | [diff] [blame] | 51 | #include "blk-throttle.h" |
Hongyu Jin | f3c8998 | 2024-01-30 15:26:34 -0500 | [diff] [blame] | 52 | #include "blk-ioprio.h" |
Jens Axboe | 8324aa9 | 2008-01-29 14:51:59 +0100 | [diff] [blame] | 53 | |
Omar Sandoval | 18fbda9 | 2017-01-31 14:53:20 -0800 | [diff] [blame] | 54 | struct dentry *blk_debugfs_root; |
Omar Sandoval | 18fbda9 | 2017-01-31 14:53:20 -0800 | [diff] [blame] | 55 | |
Mike Snitzer | d07335e | 2010-11-16 12:52:38 +0100 | [diff] [blame] | 56 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_bio_remap); |
Jun'ichi Nomura | b0da3f0 | 2009-10-01 21:16:13 +0200 | [diff] [blame] | 57 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_rq_remap); |
Linus Torvalds | 0a82a8d | 2013-04-18 09:00:26 -0700 | [diff] [blame] | 58 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_bio_complete); |
Keith Busch | 3291fa5 | 2014-04-28 12:30:52 -0600 | [diff] [blame] | 59 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_split); |
NeilBrown | cbae8d4 | 2012-12-14 20:49:27 +0100 | [diff] [blame] | 60 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_unplug); |
Chaitanya Kulkarni | b357e4a | 2021-02-21 21:29:59 -0800 | [diff] [blame] | 61 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_rq_insert); |
Ingo Molnar | 0bfc245 | 2008-11-26 11:59:56 +0100 | [diff] [blame] | 62 | |
Christoph Hellwig | 2bd8522 | 2022-11-14 05:26:36 +0100 | [diff] [blame] | 63 | static DEFINE_IDA(blk_queue_ida); |
Tejun Heo | a73f730 | 2011-12-14 00:33:37 +0100 | [diff] [blame] | 64 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 65 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 66 | * For queue allocation |
| 67 | */ |
Christoph Hellwig | 2bd8522 | 2022-11-14 05:26:36 +0100 | [diff] [blame] | 68 | static struct kmem_cache *blk_requestq_cachep; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 69 | |
| 70 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 71 | * Controlling structure to kblockd |
| 72 | */ |
Jens Axboe | ff856ba | 2006-01-09 16:02:34 +0100 | [diff] [blame] | 73 | static struct workqueue_struct *kblockd_workqueue; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 74 | |
Bart Van Assche | 8814ce8 | 2018-03-07 17:10:04 -0800 | [diff] [blame] | 75 | /** |
| 76 | * blk_queue_flag_set - atomically set a queue flag |
| 77 | * @flag: flag to be set |
| 78 | * @q: request queue |
| 79 | */ |
| 80 | void blk_queue_flag_set(unsigned int flag, struct request_queue *q) |
| 81 | { |
Christoph Hellwig | 57d74df | 2018-11-14 17:02:07 +0100 | [diff] [blame] | 82 | set_bit(flag, &q->queue_flags); |
Bart Van Assche | 8814ce8 | 2018-03-07 17:10:04 -0800 | [diff] [blame] | 83 | } |
| 84 | EXPORT_SYMBOL(blk_queue_flag_set); |
| 85 | |
| 86 | /** |
| 87 | * blk_queue_flag_clear - atomically clear a queue flag |
| 88 | * @flag: flag to be cleared |
| 89 | * @q: request queue |
| 90 | */ |
| 91 | void blk_queue_flag_clear(unsigned int flag, struct request_queue *q) |
| 92 | { |
Christoph Hellwig | 57d74df | 2018-11-14 17:02:07 +0100 | [diff] [blame] | 93 | clear_bit(flag, &q->queue_flags); |
Bart Van Assche | 8814ce8 | 2018-03-07 17:10:04 -0800 | [diff] [blame] | 94 | } |
| 95 | EXPORT_SYMBOL(blk_queue_flag_clear); |
| 96 | |
Chaitanya Kulkarni | e47bc4e | 2019-06-20 10:59:16 -0700 | [diff] [blame] | 97 | #define REQ_OP_NAME(name) [REQ_OP_##name] = #name |
| 98 | static const char *const blk_op_name[] = { |
| 99 | REQ_OP_NAME(READ), |
| 100 | REQ_OP_NAME(WRITE), |
| 101 | REQ_OP_NAME(FLUSH), |
| 102 | REQ_OP_NAME(DISCARD), |
| 103 | REQ_OP_NAME(SECURE_ERASE), |
| 104 | REQ_OP_NAME(ZONE_RESET), |
Chaitanya Kulkarni | 6e33dbf | 2019-08-01 10:26:36 -0700 | [diff] [blame] | 105 | REQ_OP_NAME(ZONE_RESET_ALL), |
Ajay Joshi | 6c1b1da | 2019-10-27 23:05:45 +0900 | [diff] [blame] | 106 | REQ_OP_NAME(ZONE_OPEN), |
| 107 | REQ_OP_NAME(ZONE_CLOSE), |
| 108 | REQ_OP_NAME(ZONE_FINISH), |
Keith Busch | 0512a75 | 2020-05-12 17:55:47 +0900 | [diff] [blame] | 109 | REQ_OP_NAME(ZONE_APPEND), |
Chaitanya Kulkarni | e47bc4e | 2019-06-20 10:59:16 -0700 | [diff] [blame] | 110 | REQ_OP_NAME(WRITE_ZEROES), |
Chaitanya Kulkarni | e47bc4e | 2019-06-20 10:59:16 -0700 | [diff] [blame] | 111 | REQ_OP_NAME(DRV_IN), |
| 112 | REQ_OP_NAME(DRV_OUT), |
| 113 | }; |
| 114 | #undef REQ_OP_NAME |
| 115 | |
| 116 | /** |
| 117 | * blk_op_str - Return string XXX in the REQ_OP_XXX. |
| 118 | * @op: REQ_OP_XXX. |
| 119 | * |
| 120 | * Description: Centralize block layer function to convert REQ_OP_XXX into |
| 121 | * string format. Useful in the debugging and tracing bio or request. For |
| 122 | * invalid REQ_OP_XXX it returns string "UNKNOWN". |
| 123 | */ |
Bart Van Assche | 77e7ffd | 2022-07-14 11:06:28 -0700 | [diff] [blame] | 124 | inline const char *blk_op_str(enum req_op op) |
Chaitanya Kulkarni | e47bc4e | 2019-06-20 10:59:16 -0700 | [diff] [blame] | 125 | { |
| 126 | const char *op_str = "UNKNOWN"; |
| 127 | |
| 128 | if (op < ARRAY_SIZE(blk_op_name) && blk_op_name[op]) |
| 129 | op_str = blk_op_name[op]; |
| 130 | |
| 131 | return op_str; |
| 132 | } |
| 133 | EXPORT_SYMBOL_GPL(blk_op_str); |
| 134 | |
Christoph Hellwig | 2a842ac | 2017-06-03 09:38:04 +0200 | [diff] [blame] | 135 | static const struct { |
| 136 | int errno; |
| 137 | const char *name; |
| 138 | } blk_errors[] = { |
| 139 | [BLK_STS_OK] = { 0, "" }, |
| 140 | [BLK_STS_NOTSUPP] = { -EOPNOTSUPP, "operation not supported" }, |
| 141 | [BLK_STS_TIMEOUT] = { -ETIMEDOUT, "timeout" }, |
| 142 | [BLK_STS_NOSPC] = { -ENOSPC, "critical space allocation" }, |
| 143 | [BLK_STS_TRANSPORT] = { -ENOLINK, "recoverable transport" }, |
| 144 | [BLK_STS_TARGET] = { -EREMOTEIO, "critical target" }, |
Mike Christie | 7ba1508 | 2023-04-07 15:05:35 -0500 | [diff] [blame] | 145 | [BLK_STS_RESV_CONFLICT] = { -EBADE, "reservation conflict" }, |
Christoph Hellwig | 2a842ac | 2017-06-03 09:38:04 +0200 | [diff] [blame] | 146 | [BLK_STS_MEDIUM] = { -ENODATA, "critical medium" }, |
| 147 | [BLK_STS_PROTECTION] = { -EILSEQ, "protection" }, |
| 148 | [BLK_STS_RESOURCE] = { -ENOMEM, "kernel resource" }, |
Ming Lei | 86ff7c2 | 2018-01-30 22:04:57 -0500 | [diff] [blame] | 149 | [BLK_STS_DEV_RESOURCE] = { -EBUSY, "device resource" }, |
Goldwyn Rodrigues | 03a07c9 | 2017-06-20 07:05:46 -0500 | [diff] [blame] | 150 | [BLK_STS_AGAIN] = { -EAGAIN, "nonblocking retry" }, |
Song Liu | 7d32c02 | 2022-02-03 11:28:26 -0800 | [diff] [blame] | 151 | [BLK_STS_OFFLINE] = { -ENODEV, "device offline" }, |
Christoph Hellwig | 2a842ac | 2017-06-03 09:38:04 +0200 | [diff] [blame] | 152 | |
Christoph Hellwig | 4e4cbee | 2017-06-03 09:38:06 +0200 | [diff] [blame] | 153 | /* device mapper special case, should not leak out: */ |
| 154 | [BLK_STS_DM_REQUEUE] = { -EREMCHG, "dm internal retry" }, |
| 155 | |
Keith Busch | 3b481d9 | 2020-09-24 13:53:28 -0700 | [diff] [blame] | 156 | /* zone device specific errors */ |
| 157 | [BLK_STS_ZONE_OPEN_RESOURCE] = { -ETOOMANYREFS, "open zones exceeded" }, |
| 158 | [BLK_STS_ZONE_ACTIVE_RESOURCE] = { -EOVERFLOW, "active zones exceeded" }, |
| 159 | |
Damien Le Moal | dffc480 | 2023-05-11 03:13:36 +0200 | [diff] [blame] | 160 | /* Command duration limit device-side timeout */ |
| 161 | [BLK_STS_DURATION_LIMIT] = { -ETIME, "duration limit exceeded" }, |
| 162 | |
John Garry | 9da3d1e | 2024-06-20 12:53:54 +0000 | [diff] [blame] | 163 | [BLK_STS_INVAL] = { -EINVAL, "invalid" }, |
| 164 | |
Christoph Hellwig | 2a842ac | 2017-06-03 09:38:04 +0200 | [diff] [blame] | 165 | /* everything else not covered above: */ |
| 166 | [BLK_STS_IOERR] = { -EIO, "I/O" }, |
| 167 | }; |
| 168 | |
| 169 | blk_status_t errno_to_blk_status(int errno) |
| 170 | { |
| 171 | int i; |
| 172 | |
| 173 | for (i = 0; i < ARRAY_SIZE(blk_errors); i++) { |
| 174 | if (blk_errors[i].errno == errno) |
| 175 | return (__force blk_status_t)i; |
| 176 | } |
| 177 | |
| 178 | return BLK_STS_IOERR; |
| 179 | } |
| 180 | EXPORT_SYMBOL_GPL(errno_to_blk_status); |
| 181 | |
| 182 | int blk_status_to_errno(blk_status_t status) |
| 183 | { |
| 184 | int idx = (__force int)status; |
| 185 | |
Bart Van Assche | 34bd9c1 | 2017-06-21 10:55:46 -0700 | [diff] [blame] | 186 | if (WARN_ON_ONCE(idx >= ARRAY_SIZE(blk_errors))) |
Christoph Hellwig | 2a842ac | 2017-06-03 09:38:04 +0200 | [diff] [blame] | 187 | return -EIO; |
| 188 | return blk_errors[idx].errno; |
| 189 | } |
| 190 | EXPORT_SYMBOL_GPL(blk_status_to_errno); |
| 191 | |
Christoph Hellwig | 0d7a29a | 2021-11-17 07:14:03 +0100 | [diff] [blame] | 192 | const char *blk_status_to_str(blk_status_t status) |
Christoph Hellwig | 2a842ac | 2017-06-03 09:38:04 +0200 | [diff] [blame] | 193 | { |
| 194 | int idx = (__force int)status; |
| 195 | |
Bart Van Assche | 34bd9c1 | 2017-06-21 10:55:46 -0700 | [diff] [blame] | 196 | if (WARN_ON_ONCE(idx >= ARRAY_SIZE(blk_errors))) |
Christoph Hellwig | 0d7a29a | 2021-11-17 07:14:03 +0100 | [diff] [blame] | 197 | return "<null>"; |
| 198 | return blk_errors[idx].name; |
Christoph Hellwig | 2a842ac | 2017-06-03 09:38:04 +0200 | [diff] [blame] | 199 | } |
Kent Overstreet | 7ba3792 | 2023-08-13 14:26:34 -0400 | [diff] [blame] | 200 | EXPORT_SYMBOL_GPL(blk_status_to_str); |
Christoph Hellwig | 2a842ac | 2017-06-03 09:38:04 +0200 | [diff] [blame] | 201 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 202 | /** |
| 203 | * blk_sync_queue - cancel any pending callbacks on a queue |
| 204 | * @q: the queue |
| 205 | * |
| 206 | * Description: |
| 207 | * The block layer may perform asynchronous callback activity |
| 208 | * on a queue, such as calling the unplug function after a timeout. |
| 209 | * A block device may call blk_sync_queue to ensure that any |
| 210 | * such activity is cancelled, thus allowing it to release resources |
Michael Opdenacker | 59c5159 | 2007-05-09 08:57:56 +0200 | [diff] [blame] | 211 | * that the callbacks might use. The caller must already have made sure |
Christoph Hellwig | c62b37d | 2020-07-01 10:59:43 +0200 | [diff] [blame] | 212 | * that its ->submit_bio will not re-add plugging prior to calling |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 213 | * this function. |
| 214 | * |
Vivek Goyal | da52777 | 2011-03-02 19:05:33 -0500 | [diff] [blame] | 215 | * This function does not cancel any asynchronous activity arising |
Masanari Iida | da3dae5 | 2014-09-09 01:27:23 +0900 | [diff] [blame] | 216 | * out of elevator or throttling code. That would require elevator_exit() |
Tejun Heo | 5efd611 | 2012-03-05 13:15:12 -0800 | [diff] [blame] | 217 | * and blkcg_exit_queue() to be called with queue lock initialized. |
Vivek Goyal | da52777 | 2011-03-02 19:05:33 -0500 | [diff] [blame] | 218 | * |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 219 | */ |
| 220 | void blk_sync_queue(struct request_queue *q) |
| 221 | { |
Jens Axboe | 70ed28b | 2008-11-19 14:38:39 +0100 | [diff] [blame] | 222 | del_timer_sync(&q->timeout); |
Bart Van Assche | 4e9b6f2 | 2017-10-19 10:00:48 -0700 | [diff] [blame] | 223 | cancel_work_sync(&q->timeout_work); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 224 | } |
| 225 | EXPORT_SYMBOL(blk_sync_queue); |
| 226 | |
| 227 | /** |
Bart Van Assche | cd84a62 | 2018-09-26 14:01:04 -0700 | [diff] [blame] | 228 | * blk_set_pm_only - increment pm_only counter |
Bart Van Assche | c9254f2 | 2017-11-09 10:49:57 -0800 | [diff] [blame] | 229 | * @q: request queue pointer |
Bart Van Assche | c9254f2 | 2017-11-09 10:49:57 -0800 | [diff] [blame] | 230 | */ |
Bart Van Assche | cd84a62 | 2018-09-26 14:01:04 -0700 | [diff] [blame] | 231 | void blk_set_pm_only(struct request_queue *q) |
Bart Van Assche | c9254f2 | 2017-11-09 10:49:57 -0800 | [diff] [blame] | 232 | { |
Bart Van Assche | cd84a62 | 2018-09-26 14:01:04 -0700 | [diff] [blame] | 233 | atomic_inc(&q->pm_only); |
Bart Van Assche | c9254f2 | 2017-11-09 10:49:57 -0800 | [diff] [blame] | 234 | } |
Bart Van Assche | cd84a62 | 2018-09-26 14:01:04 -0700 | [diff] [blame] | 235 | EXPORT_SYMBOL_GPL(blk_set_pm_only); |
Bart Van Assche | c9254f2 | 2017-11-09 10:49:57 -0800 | [diff] [blame] | 236 | |
Bart Van Assche | cd84a62 | 2018-09-26 14:01:04 -0700 | [diff] [blame] | 237 | void blk_clear_pm_only(struct request_queue *q) |
Bart Van Assche | c9254f2 | 2017-11-09 10:49:57 -0800 | [diff] [blame] | 238 | { |
Bart Van Assche | cd84a62 | 2018-09-26 14:01:04 -0700 | [diff] [blame] | 239 | int pm_only; |
| 240 | |
| 241 | pm_only = atomic_dec_return(&q->pm_only); |
| 242 | WARN_ON_ONCE(pm_only < 0); |
| 243 | if (pm_only == 0) |
| 244 | wake_up_all(&q->mq_freeze_wq); |
Bart Van Assche | c9254f2 | 2017-11-09 10:49:57 -0800 | [diff] [blame] | 245 | } |
Bart Van Assche | cd84a62 | 2018-09-26 14:01:04 -0700 | [diff] [blame] | 246 | EXPORT_SYMBOL_GPL(blk_clear_pm_only); |
Bart Van Assche | c9254f2 | 2017-11-09 10:49:57 -0800 | [diff] [blame] | 247 | |
Christoph Hellwig | 2bd8522 | 2022-11-14 05:26:36 +0100 | [diff] [blame] | 248 | static void blk_free_queue_rcu(struct rcu_head *rcu_head) |
| 249 | { |
Ming Lei | d36a9ea | 2022-12-15 10:16:29 +0800 | [diff] [blame] | 250 | struct request_queue *q = container_of(rcu_head, |
| 251 | struct request_queue, rcu_head); |
| 252 | |
| 253 | percpu_ref_exit(&q->q_usage_counter); |
| 254 | kmem_cache_free(blk_requestq_cachep, q); |
Christoph Hellwig | 2bd8522 | 2022-11-14 05:26:36 +0100 | [diff] [blame] | 255 | } |
| 256 | |
| 257 | static void blk_free_queue(struct request_queue *q) |
| 258 | { |
Christoph Hellwig | 2bd8522 | 2022-11-14 05:26:36 +0100 | [diff] [blame] | 259 | blk_free_queue_stats(q->stats); |
Christoph Hellwig | 2bd8522 | 2022-11-14 05:26:36 +0100 | [diff] [blame] | 260 | if (queue_is_mq(q)) |
| 261 | blk_mq_release(q); |
| 262 | |
| 263 | ida_free(&blk_queue_ida, q->id); |
| 264 | call_rcu(&q->rcu_head, blk_free_queue_rcu); |
| 265 | } |
| 266 | |
Luis Chamberlain | b5bd357 | 2020-06-19 20:47:23 +0000 | [diff] [blame] | 267 | /** |
| 268 | * blk_put_queue - decrement the request_queue refcount |
| 269 | * @q: the request_queue structure to decrement the refcount for |
| 270 | * |
Christoph Hellwig | 2bd8522 | 2022-11-14 05:26:36 +0100 | [diff] [blame] | 271 | * Decrements the refcount of the request_queue and free it when the refcount |
| 272 | * reaches 0. |
Luis Chamberlain | b5bd357 | 2020-06-19 20:47:23 +0000 | [diff] [blame] | 273 | */ |
Jens Axboe | 165125e | 2007-07-24 09:28:11 +0200 | [diff] [blame] | 274 | void blk_put_queue(struct request_queue *q) |
Al Viro | 483f4af | 2006-03-18 18:34:37 -0500 | [diff] [blame] | 275 | { |
Christoph Hellwig | 2bd8522 | 2022-11-14 05:26:36 +0100 | [diff] [blame] | 276 | if (refcount_dec_and_test(&q->refs)) |
| 277 | blk_free_queue(q); |
Al Viro | 483f4af | 2006-03-18 18:34:37 -0500 | [diff] [blame] | 278 | } |
Jens Axboe | d86e0e8 | 2011-05-27 07:44:43 +0200 | [diff] [blame] | 279 | EXPORT_SYMBOL(blk_put_queue); |
Al Viro | 483f4af | 2006-03-18 18:34:37 -0500 | [diff] [blame] | 280 | |
Christoph Hellwig | 8e141f9 | 2021-09-29 09:12:40 +0200 | [diff] [blame] | 281 | void blk_queue_start_drain(struct request_queue *q) |
Jens Axboe | aed3ea9 | 2014-12-22 14:04:42 -0700 | [diff] [blame] | 282 | { |
Ming Lei | d3cfb2a | 2017-03-27 20:06:58 +0800 | [diff] [blame] | 283 | /* |
| 284 | * When queue DYING flag is set, we need to block new req |
| 285 | * entering queue, so we call blk_freeze_queue_start() to |
| 286 | * prevent I/O from crossing blk_queue_enter(). |
| 287 | */ |
| 288 | blk_freeze_queue_start(q); |
Jens Axboe | 344e9ff | 2018-11-15 12:22:51 -0700 | [diff] [blame] | 289 | if (queue_is_mq(q)) |
Jens Axboe | aed3ea9 | 2014-12-22 14:04:42 -0700 | [diff] [blame] | 290 | blk_mq_wake_waiters(q); |
Ming Lei | 055f6e1 | 2017-11-09 10:49:53 -0800 | [diff] [blame] | 291 | /* Make blk_queue_enter() reexamine the DYING flag. */ |
| 292 | wake_up_all(&q->mq_freeze_wq); |
Jens Axboe | aed3ea9 | 2014-12-22 14:04:42 -0700 | [diff] [blame] | 293 | } |
Christoph Hellwig | 8e141f9 | 2021-09-29 09:12:40 +0200 | [diff] [blame] | 294 | |
Tejun Heo | d732580 | 2012-03-05 13:14:58 -0800 | [diff] [blame] | 295 | /** |
Bart Van Assche | 3a0a529 | 2017-11-09 10:49:58 -0800 | [diff] [blame] | 296 | * blk_queue_enter() - try to increase q->q_usage_counter |
| 297 | * @q: request queue pointer |
Bart Van Assche | a4d34da | 2020-12-08 21:29:50 -0800 | [diff] [blame] | 298 | * @flags: BLK_MQ_REQ_NOWAIT and/or BLK_MQ_REQ_PM |
Bart Van Assche | 3a0a529 | 2017-11-09 10:49:58 -0800 | [diff] [blame] | 299 | */ |
Bart Van Assche | 9a95e4e | 2017-11-09 10:49:59 -0800 | [diff] [blame] | 300 | int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags) |
Dan Williams | 3ef28e8 | 2015-10-21 13:20:12 -0400 | [diff] [blame] | 301 | { |
Bart Van Assche | a4d34da | 2020-12-08 21:29:50 -0800 | [diff] [blame] | 302 | const bool pm = flags & BLK_MQ_REQ_PM; |
Bart Van Assche | 3a0a529 | 2017-11-09 10:49:58 -0800 | [diff] [blame] | 303 | |
Christoph Hellwig | 1f14a09 | 2021-09-29 09:12:38 +0200 | [diff] [blame] | 304 | while (!blk_try_enter_queue(q, pm)) { |
Bart Van Assche | 3a0a529 | 2017-11-09 10:49:58 -0800 | [diff] [blame] | 305 | if (flags & BLK_MQ_REQ_NOWAIT) |
Stefan Roesch | 56f99b8 | 2022-09-12 09:53:25 -0700 | [diff] [blame] | 306 | return -EAGAIN; |
Dan Williams | 3ef28e8 | 2015-10-21 13:20:12 -0400 | [diff] [blame] | 307 | |
Ming Lei | 5ed61d3 | 2017-03-27 20:06:56 +0800 | [diff] [blame] | 308 | /* |
Christoph Hellwig | 1f14a09 | 2021-09-29 09:12:38 +0200 | [diff] [blame] | 309 | * read pair of barrier in blk_freeze_queue_start(), we need to |
| 310 | * order reading __PERCPU_REF_DEAD flag of .q_usage_counter and |
| 311 | * reading .mq_freeze_depth or queue dying flag, otherwise the |
| 312 | * following wait may never return if the two reads are |
| 313 | * reordered. |
Ming Lei | 5ed61d3 | 2017-03-27 20:06:56 +0800 | [diff] [blame] | 314 | */ |
| 315 | smp_rmb(); |
Alan Jenkins | 1dc3039 | 2018-04-12 19:11:58 +0100 | [diff] [blame] | 316 | wait_event(q->mq_freeze_wq, |
Bob Liu | 7996a8b | 2019-05-21 11:25:55 +0800 | [diff] [blame] | 317 | (!q->mq_freeze_depth && |
Alan Stern | 52abca6 | 2020-12-08 21:29:51 -0800 | [diff] [blame] | 318 | blk_pm_resume_queue(pm, q)) || |
Alan Jenkins | 1dc3039 | 2018-04-12 19:11:58 +0100 | [diff] [blame] | 319 | blk_queue_dying(q)); |
Dan Williams | 3ef28e8 | 2015-10-21 13:20:12 -0400 | [diff] [blame] | 320 | if (blk_queue_dying(q)) |
| 321 | return -ENODEV; |
Dan Williams | 3ef28e8 | 2015-10-21 13:20:12 -0400 | [diff] [blame] | 322 | } |
Christoph Hellwig | 1f14a09 | 2021-09-29 09:12:38 +0200 | [diff] [blame] | 323 | |
| 324 | return 0; |
Dan Williams | 3ef28e8 | 2015-10-21 13:20:12 -0400 | [diff] [blame] | 325 | } |
| 326 | |
Jens Axboe | c98cb5b | 2021-11-04 12:45:51 -0600 | [diff] [blame] | 327 | int __bio_queue_enter(struct request_queue *q, struct bio *bio) |
Christoph Hellwig | accea32 | 2020-04-28 13:27:56 +0200 | [diff] [blame] | 328 | { |
Christoph Hellwig | a674153 | 2021-09-29 09:12:39 +0200 | [diff] [blame] | 329 | while (!blk_try_enter_queue(q, false)) { |
Pavel Begunkov | eab4e02 | 2021-10-14 15:03:29 +0100 | [diff] [blame] | 330 | struct gendisk *disk = bio->bi_bdev->bd_disk; |
| 331 | |
Christoph Hellwig | a674153 | 2021-09-29 09:12:39 +0200 | [diff] [blame] | 332 | if (bio->bi_opf & REQ_NOWAIT) { |
Christoph Hellwig | 8e141f9 | 2021-09-29 09:12:40 +0200 | [diff] [blame] | 333 | if (test_bit(GD_DEAD, &disk->state)) |
Christoph Hellwig | a674153 | 2021-09-29 09:12:39 +0200 | [diff] [blame] | 334 | goto dead; |
Christoph Hellwig | accea32 | 2020-04-28 13:27:56 +0200 | [diff] [blame] | 335 | bio_wouldblock_error(bio); |
Stefan Roesch | 56f99b8 | 2022-09-12 09:53:25 -0700 | [diff] [blame] | 336 | return -EAGAIN; |
Christoph Hellwig | a674153 | 2021-09-29 09:12:39 +0200 | [diff] [blame] | 337 | } |
| 338 | |
| 339 | /* |
| 340 | * read pair of barrier in blk_freeze_queue_start(), we need to |
| 341 | * order reading __PERCPU_REF_DEAD flag of .q_usage_counter and |
| 342 | * reading .mq_freeze_depth or queue dying flag, otherwise the |
| 343 | * following wait may never return if the two reads are |
| 344 | * reordered. |
| 345 | */ |
| 346 | smp_rmb(); |
| 347 | wait_event(q->mq_freeze_wq, |
| 348 | (!q->mq_freeze_depth && |
| 349 | blk_pm_resume_queue(false, q)) || |
Christoph Hellwig | 8e141f9 | 2021-09-29 09:12:40 +0200 | [diff] [blame] | 350 | test_bit(GD_DEAD, &disk->state)); |
| 351 | if (test_bit(GD_DEAD, &disk->state)) |
Christoph Hellwig | a674153 | 2021-09-29 09:12:39 +0200 | [diff] [blame] | 352 | goto dead; |
Christoph Hellwig | accea32 | 2020-04-28 13:27:56 +0200 | [diff] [blame] | 353 | } |
| 354 | |
Christoph Hellwig | a674153 | 2021-09-29 09:12:39 +0200 | [diff] [blame] | 355 | return 0; |
| 356 | dead: |
| 357 | bio_io_error(bio); |
| 358 | return -ENODEV; |
Christoph Hellwig | accea32 | 2020-04-28 13:27:56 +0200 | [diff] [blame] | 359 | } |
| 360 | |
Dan Williams | 3ef28e8 | 2015-10-21 13:20:12 -0400 | [diff] [blame] | 361 | void blk_queue_exit(struct request_queue *q) |
| 362 | { |
| 363 | percpu_ref_put(&q->q_usage_counter); |
| 364 | } |
| 365 | |
| 366 | static void blk_queue_usage_counter_release(struct percpu_ref *ref) |
| 367 | { |
| 368 | struct request_queue *q = |
| 369 | container_of(ref, struct request_queue, q_usage_counter); |
| 370 | |
| 371 | wake_up_all(&q->mq_freeze_wq); |
| 372 | } |
| 373 | |
Kees Cook | bca237a | 2017-08-28 15:03:41 -0700 | [diff] [blame] | 374 | static void blk_rq_timed_out_timer(struct timer_list *t) |
Christoph Hellwig | 287922e | 2015-10-30 20:57:30 +0800 | [diff] [blame] | 375 | { |
Kees Cook | bca237a | 2017-08-28 15:03:41 -0700 | [diff] [blame] | 376 | struct request_queue *q = from_timer(q, t, timeout); |
Christoph Hellwig | 287922e | 2015-10-30 20:57:30 +0800 | [diff] [blame] | 377 | |
| 378 | kblockd_schedule_work(&q->timeout_work); |
| 379 | } |
| 380 | |
Tetsuo Handa | 2e3c18d | 2019-01-30 22:21:45 +0900 | [diff] [blame] | 381 | static void blk_timeout_work(struct work_struct *work) |
| 382 | { |
| 383 | } |
| 384 | |
Christoph Hellwig | ad751ba | 2024-02-13 08:34:18 +0100 | [diff] [blame] | 385 | struct request_queue *blk_alloc_queue(struct queue_limits *lim, int node_id) |
Christoph Lameter | 1946089 | 2005-06-23 00:08:19 -0700 | [diff] [blame] | 386 | { |
Jens Axboe | 165125e | 2007-07-24 09:28:11 +0200 | [diff] [blame] | 387 | struct request_queue *q; |
Christoph Hellwig | ad751ba | 2024-02-13 08:34:18 +0100 | [diff] [blame] | 388 | int error; |
Christoph Lameter | 1946089 | 2005-06-23 00:08:19 -0700 | [diff] [blame] | 389 | |
Christoph Hellwig | 80bd4a7 | 2022-11-01 16:00:47 +0100 | [diff] [blame] | 390 | q = kmem_cache_alloc_node(blk_requestq_cachep, GFP_KERNEL | __GFP_ZERO, |
| 391 | node_id); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 392 | if (!q) |
Christoph Hellwig | ad751ba | 2024-02-13 08:34:18 +0100 | [diff] [blame] | 393 | return ERR_PTR(-ENOMEM); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 394 | |
Christoph Hellwig | cbf62af | 2018-05-31 19:11:36 +0200 | [diff] [blame] | 395 | q->last_merge = NULL; |
Christoph Hellwig | cbf62af | 2018-05-31 19:11:36 +0200 | [diff] [blame] | 396 | |
Bo Liu | 798f2a6 | 2022-06-15 04:18:16 -0400 | [diff] [blame] | 397 | q->id = ida_alloc(&blk_queue_ida, GFP_KERNEL); |
Christoph Hellwig | ad751ba | 2024-02-13 08:34:18 +0100 | [diff] [blame] | 398 | if (q->id < 0) { |
| 399 | error = q->id; |
Christoph Hellwig | 80bd4a7 | 2022-11-01 16:00:47 +0100 | [diff] [blame] | 400 | goto fail_q; |
Christoph Hellwig | ad751ba | 2024-02-13 08:34:18 +0100 | [diff] [blame] | 401 | } |
Tejun Heo | a73f730 | 2011-12-14 00:33:37 +0100 | [diff] [blame] | 402 | |
Jens Axboe | a83b576 | 2017-03-21 17:20:01 -0600 | [diff] [blame] | 403 | q->stats = blk_alloc_queue_stats(); |
Christoph Hellwig | ad751ba | 2024-02-13 08:34:18 +0100 | [diff] [blame] | 404 | if (!q->stats) { |
| 405 | error = -ENOMEM; |
Christoph Hellwig | 46754bd | 2022-07-27 12:22:57 -0400 | [diff] [blame] | 406 | goto fail_id; |
Christoph Hellwig | ad751ba | 2024-02-13 08:34:18 +0100 | [diff] [blame] | 407 | } |
| 408 | |
| 409 | error = blk_set_default_limits(lim); |
| 410 | if (error) |
| 411 | goto fail_stats; |
| 412 | q->limits = *lim; |
Jens Axboe | a83b576 | 2017-03-21 17:20:01 -0600 | [diff] [blame] | 413 | |
Mike Snitzer | 5151412 | 2011-11-23 10:59:13 +0100 | [diff] [blame] | 414 | q->node = node_id; |
Jens Axboe | 0989a02 | 2009-06-12 14:42:56 +0200 | [diff] [blame] | 415 | |
John Garry | 079a2e3 | 2021-10-05 18:23:39 +0800 | [diff] [blame] | 416 | atomic_set(&q->nr_active_requests_shared_tags, 0); |
John Garry | bccf5e2 | 2020-08-19 23:20:26 +0800 | [diff] [blame] | 417 | |
Kees Cook | bca237a | 2017-08-28 15:03:41 -0700 | [diff] [blame] | 418 | timer_setup(&q->timeout, blk_rq_timed_out_timer, 0); |
Tetsuo Handa | 2e3c18d | 2019-01-30 22:21:45 +0900 | [diff] [blame] | 419 | INIT_WORK(&q->timeout_work, blk_timeout_work); |
Tejun Heo | a612fdd | 2011-12-14 00:33:41 +0100 | [diff] [blame] | 420 | INIT_LIST_HEAD(&q->icq_list); |
Al Viro | 483f4af | 2006-03-18 18:34:37 -0500 | [diff] [blame] | 421 | |
Christoph Hellwig | 2bd8522 | 2022-11-14 05:26:36 +0100 | [diff] [blame] | 422 | refcount_set(&q->refs, 1); |
Luis Chamberlain | 85e0cbb | 2020-06-19 20:47:30 +0000 | [diff] [blame] | 423 | mutex_init(&q->debugfs_mutex); |
Al Viro | 483f4af | 2006-03-18 18:34:37 -0500 | [diff] [blame] | 424 | mutex_init(&q->sysfs_lock); |
Ming Lei | cecf5d8 | 2019-08-27 19:01:48 +0800 | [diff] [blame] | 425 | mutex_init(&q->sysfs_dir_lock); |
Christoph Hellwig | d690cb8a | 2024-02-13 08:34:14 +0100 | [diff] [blame] | 426 | mutex_init(&q->limits_lock); |
Yu Kuai | a13bd91 | 2023-04-14 16:40:08 +0800 | [diff] [blame] | 427 | mutex_init(&q->rq_qos_mutex); |
Christoph Hellwig | 0d945c1 | 2018-11-15 12:17:28 -0700 | [diff] [blame] | 428 | spin_lock_init(&q->queue_lock); |
Vivek Goyal | c94a96a | 2011-03-02 19:04:42 -0500 | [diff] [blame] | 429 | |
Jens Axboe | 320ae51 | 2013-10-24 09:20:05 +0100 | [diff] [blame] | 430 | init_waitqueue_head(&q->mq_freeze_wq); |
Bob Liu | 7996a8b | 2019-05-21 11:25:55 +0800 | [diff] [blame] | 431 | mutex_init(&q->mq_freeze_lock); |
Jens Axboe | 320ae51 | 2013-10-24 09:20:05 +0100 | [diff] [blame] | 432 | |
Ming Lei | 8b8ace0 | 2024-04-07 20:59:10 +0800 | [diff] [blame] | 433 | blkg_init_queue(q); |
| 434 | |
Dan Williams | 3ef28e8 | 2015-10-21 13:20:12 -0400 | [diff] [blame] | 435 | /* |
| 436 | * Init percpu_ref in atomic mode so that it's faster to shutdown. |
| 437 | * See blk_register_queue() for details. |
| 438 | */ |
Christoph Hellwig | ad751ba | 2024-02-13 08:34:18 +0100 | [diff] [blame] | 439 | error = percpu_ref_init(&q->q_usage_counter, |
Dan Williams | 3ef28e8 | 2015-10-21 13:20:12 -0400 | [diff] [blame] | 440 | blk_queue_usage_counter_release, |
Christoph Hellwig | ad751ba | 2024-02-13 08:34:18 +0100 | [diff] [blame] | 441 | PERCPU_REF_INIT_ATOMIC, GFP_KERNEL); |
| 442 | if (error) |
Christoph Hellwig | edb0872 | 2021-08-09 16:17:43 +0200 | [diff] [blame] | 443 | goto fail_stats; |
Tejun Heo | f51b802 | 2012-03-05 13:15:05 -0800 | [diff] [blame] | 444 | |
John Garry | d2a2796 | 2021-10-05 18:23:27 +0800 | [diff] [blame] | 445 | q->nr_requests = BLKDEV_DEFAULT_RQ; |
Christoph Hellwig | 3d745ea | 2020-03-27 09:30:11 +0100 | [diff] [blame] | 446 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 447 | return q; |
Tejun Heo | a73f730 | 2011-12-14 00:33:37 +0100 | [diff] [blame] | 448 | |
Jens Axboe | a83b576 | 2017-03-21 17:20:01 -0600 | [diff] [blame] | 449 | fail_stats: |
Christoph Hellwig | edb0872 | 2021-08-09 16:17:43 +0200 | [diff] [blame] | 450 | blk_free_queue_stats(q->stats); |
Tejun Heo | a73f730 | 2011-12-14 00:33:37 +0100 | [diff] [blame] | 451 | fail_id: |
Bo Liu | 798f2a6 | 2022-06-15 04:18:16 -0400 | [diff] [blame] | 452 | ida_free(&blk_queue_ida, q->id); |
Tejun Heo | a73f730 | 2011-12-14 00:33:37 +0100 | [diff] [blame] | 453 | fail_q: |
Christoph Hellwig | 80bd4a7 | 2022-11-01 16:00:47 +0100 | [diff] [blame] | 454 | kmem_cache_free(blk_requestq_cachep, q); |
Christoph Hellwig | ad751ba | 2024-02-13 08:34:18 +0100 | [diff] [blame] | 455 | return ERR_PTR(error); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 456 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 457 | |
Luis Chamberlain | b5bd357 | 2020-06-19 20:47:23 +0000 | [diff] [blame] | 458 | /** |
| 459 | * blk_get_queue - increment the request_queue refcount |
| 460 | * @q: the request_queue structure to increment the refcount for |
| 461 | * |
| 462 | * Increment the refcount of the request_queue kobject. |
Luis Chamberlain | 763b589 | 2020-06-19 20:47:24 +0000 | [diff] [blame] | 463 | * |
| 464 | * Context: Any context. |
Luis Chamberlain | b5bd357 | 2020-06-19 20:47:23 +0000 | [diff] [blame] | 465 | */ |
Tejun Heo | 09ac46c | 2011-12-14 00:33:38 +0100 | [diff] [blame] | 466 | bool blk_get_queue(struct request_queue *q) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 467 | { |
Christoph Hellwig | 828b5f0 | 2022-07-21 08:34:32 +0200 | [diff] [blame] | 468 | if (unlikely(blk_queue_dying(q))) |
| 469 | return false; |
Christoph Hellwig | 2bd8522 | 2022-11-14 05:26:36 +0100 | [diff] [blame] | 470 | refcount_inc(&q->refs); |
Christoph Hellwig | 828b5f0 | 2022-07-21 08:34:32 +0200 | [diff] [blame] | 471 | return true; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 472 | } |
Jens Axboe | d86e0e8 | 2011-05-27 07:44:43 +0200 | [diff] [blame] | 473 | EXPORT_SYMBOL(blk_get_queue); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 474 | |
Akinobu Mita | c17bb49 | 2006-12-08 02:39:46 -0800 | [diff] [blame] | 475 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
| 476 | |
| 477 | static DECLARE_FAULT_ATTR(fail_make_request); |
| 478 | |
| 479 | static int __init setup_fail_make_request(char *str) |
| 480 | { |
| 481 | return setup_fault_attr(&fail_make_request, str); |
| 482 | } |
| 483 | __setup("fail_make_request=", setup_fail_make_request); |
| 484 | |
Christoph Hellwig | 06c8c69 | 2021-11-17 07:13:58 +0100 | [diff] [blame] | 485 | bool should_fail_request(struct block_device *part, unsigned int bytes) |
Akinobu Mita | c17bb49 | 2006-12-08 02:39:46 -0800 | [diff] [blame] | 486 | { |
Al Viro | 811ba89 | 2024-04-28 00:15:07 -0400 | [diff] [blame] | 487 | return bdev_test_flag(part, BD_MAKE_IT_FAIL) && |
| 488 | should_fail(&fail_make_request, bytes); |
Akinobu Mita | c17bb49 | 2006-12-08 02:39:46 -0800 | [diff] [blame] | 489 | } |
| 490 | |
| 491 | static int __init fail_make_request_debugfs(void) |
| 492 | { |
Akinobu Mita | dd48c08 | 2011-08-03 16:21:01 -0700 | [diff] [blame] | 493 | struct dentry *dir = fault_create_debugfs_attr("fail_make_request", |
| 494 | NULL, &fail_make_request); |
| 495 | |
Duan Jiong | 21f9fcd | 2014-04-11 15:58:56 +0800 | [diff] [blame] | 496 | return PTR_ERR_OR_ZERO(dir); |
Akinobu Mita | c17bb49 | 2006-12-08 02:39:46 -0800 | [diff] [blame] | 497 | } |
| 498 | |
| 499 | late_initcall(fail_make_request_debugfs); |
Akinobu Mita | c17bb49 | 2006-12-08 02:39:46 -0800 | [diff] [blame] | 500 | #endif /* CONFIG_FAIL_MAKE_REQUEST */ |
| 501 | |
Miaohe Lin | bdb7d42 | 2022-09-05 18:27:54 +0800 | [diff] [blame] | 502 | static inline void bio_check_ro(struct bio *bio) |
Ilya Dryomov | 721c7fc | 2018-01-11 14:09:11 +0100 | [diff] [blame] | 503 | { |
Christoph Hellwig | 2f9f622 | 2021-01-24 11:02:35 +0100 | [diff] [blame] | 504 | if (op_is_write(bio_op(bio)) && bdev_read_only(bio->bi_bdev)) { |
Mikulas Patocka | 8b2ded1 | 2018-09-05 16:14:36 -0600 | [diff] [blame] | 505 | if (op_is_flush(bio->bi_opf) && !bio_sectors(bio)) |
Miaohe Lin | bdb7d42 | 2022-09-05 18:27:54 +0800 | [diff] [blame] | 506 | return; |
Yu Kuai | 67d995e | 2023-11-28 20:30:27 +0800 | [diff] [blame] | 507 | |
Al Viro | 49a43da | 2024-04-12 01:24:27 -0400 | [diff] [blame] | 508 | if (bdev_test_flag(bio->bi_bdev, BD_RO_WARNED)) |
Yu Kuai | 67d995e | 2023-11-28 20:30:27 +0800 | [diff] [blame] | 509 | return; |
| 510 | |
Al Viro | 49a43da | 2024-04-12 01:24:27 -0400 | [diff] [blame] | 511 | bdev_set_flag(bio->bi_bdev, BD_RO_WARNED); |
| 512 | |
Yu Kuai | 67d995e | 2023-11-28 20:30:27 +0800 | [diff] [blame] | 513 | /* |
| 514 | * Use ioctl to set underlying disk of raid/dm to read-only |
| 515 | * will trigger this. |
| 516 | */ |
| 517 | pr_warn("Trying to write to read-only block-device %pg\n", |
| 518 | bio->bi_bdev); |
Ilya Dryomov | 721c7fc | 2018-01-11 14:09:11 +0100 | [diff] [blame] | 519 | } |
Ilya Dryomov | 721c7fc | 2018-01-11 14:09:11 +0100 | [diff] [blame] | 520 | } |
| 521 | |
Howard McLauchlan | 30abb3a | 2018-02-06 14:05:39 -0800 | [diff] [blame] | 522 | static noinline int should_fail_bio(struct bio *bio) |
| 523 | { |
Christoph Hellwig | 309dca30 | 2021-01-24 11:02:34 +0100 | [diff] [blame] | 524 | if (should_fail_request(bdev_whole(bio->bi_bdev), bio->bi_iter.bi_size)) |
Howard McLauchlan | 30abb3a | 2018-02-06 14:05:39 -0800 | [diff] [blame] | 525 | return -EIO; |
| 526 | return 0; |
| 527 | } |
| 528 | ALLOW_ERROR_INJECTION(should_fail_bio, ERRNO); |
| 529 | |
Jens Axboe | c07e2b4 | 2007-07-18 13:27:58 +0200 | [diff] [blame] | 530 | /* |
Christoph Hellwig | 52c5e62 | 2018-03-14 16:56:53 +0100 | [diff] [blame] | 531 | * Check whether this bio extends beyond the end of the device or partition. |
| 532 | * This may well happen - the kernel calls bread() without checking the size of |
| 533 | * the device, e.g., when mounting a file system. |
| 534 | */ |
Christoph Hellwig | 2f9f622 | 2021-01-24 11:02:35 +0100 | [diff] [blame] | 535 | static inline int bio_check_eod(struct bio *bio) |
Christoph Hellwig | 52c5e62 | 2018-03-14 16:56:53 +0100 | [diff] [blame] | 536 | { |
Christoph Hellwig | 2f9f622 | 2021-01-24 11:02:35 +0100 | [diff] [blame] | 537 | sector_t maxsector = bdev_nr_sectors(bio->bi_bdev); |
Christoph Hellwig | 52c5e62 | 2018-03-14 16:56:53 +0100 | [diff] [blame] | 538 | unsigned int nr_sectors = bio_sectors(bio); |
| 539 | |
Christoph Hellwig | 3eb9694 | 2023-05-24 08:05:38 +0200 | [diff] [blame] | 540 | if (nr_sectors && |
Christoph Hellwig | 52c5e62 | 2018-03-14 16:56:53 +0100 | [diff] [blame] | 541 | (nr_sectors > maxsector || |
| 542 | bio->bi_iter.bi_sector > maxsector - nr_sectors)) { |
Christoph Hellwig | ad74078 | 2022-03-04 19:00:57 +0100 | [diff] [blame] | 543 | pr_info_ratelimited("%s: attempt to access beyond end of device\n" |
Christoph Hellwig | 069adba | 2022-05-04 07:33:55 -0700 | [diff] [blame] | 544 | "%pg: rw=%d, sector=%llu, nr_sectors = %u limit=%llu\n", |
| 545 | current->comm, bio->bi_bdev, bio->bi_opf, |
| 546 | bio->bi_iter.bi_sector, nr_sectors, maxsector); |
Christoph Hellwig | 52c5e62 | 2018-03-14 16:56:53 +0100 | [diff] [blame] | 547 | return -EIO; |
| 548 | } |
| 549 | return 0; |
| 550 | } |
| 551 | |
| 552 | /* |
Christoph Hellwig | 74d4699 | 2017-08-23 19:10:32 +0200 | [diff] [blame] | 553 | * Remap block n of partition p to block n+start(p) of the disk. |
| 554 | */ |
Christoph Hellwig | 2f9f622 | 2021-01-24 11:02:35 +0100 | [diff] [blame] | 555 | static int blk_partition_remap(struct bio *bio) |
Christoph Hellwig | 74d4699 | 2017-08-23 19:10:32 +0200 | [diff] [blame] | 556 | { |
Christoph Hellwig | 309dca30 | 2021-01-24 11:02:34 +0100 | [diff] [blame] | 557 | struct block_device *p = bio->bi_bdev; |
Christoph Hellwig | 74d4699 | 2017-08-23 19:10:32 +0200 | [diff] [blame] | 558 | |
Christoph Hellwig | 52c5e62 | 2018-03-14 16:56:53 +0100 | [diff] [blame] | 559 | if (unlikely(should_fail_request(p, bio->bi_iter.bi_size))) |
Christoph Hellwig | 2f9f622 | 2021-01-24 11:02:35 +0100 | [diff] [blame] | 560 | return -EIO; |
Damien Le Moal | 5eac3eb | 2019-11-11 11:39:25 +0900 | [diff] [blame] | 561 | if (bio_sectors(bio)) { |
Christoph Hellwig | 8446fe9 | 2020-11-24 09:36:54 +0100 | [diff] [blame] | 562 | bio->bi_iter.bi_sector += p->bd_start_sect; |
Christoph Hellwig | 1c02fca | 2020-12-03 17:21:38 +0100 | [diff] [blame] | 563 | trace_block_bio_remap(bio, p->bd_dev, |
Christoph Hellwig | 29ff57c | 2020-11-24 09:34:24 +0100 | [diff] [blame] | 564 | bio->bi_iter.bi_sector - |
Christoph Hellwig | 8446fe9 | 2020-11-24 09:36:54 +0100 | [diff] [blame] | 565 | p->bd_start_sect); |
Christoph Hellwig | 52c5e62 | 2018-03-14 16:56:53 +0100 | [diff] [blame] | 566 | } |
Christoph Hellwig | 30c5d34 | 2021-01-24 11:02:36 +0100 | [diff] [blame] | 567 | bio_set_flag(bio, BIO_REMAPPED); |
Christoph Hellwig | 2f9f622 | 2021-01-24 11:02:35 +0100 | [diff] [blame] | 568 | return 0; |
Christoph Hellwig | 74d4699 | 2017-08-23 19:10:32 +0200 | [diff] [blame] | 569 | } |
| 570 | |
Keith Busch | 0512a75 | 2020-05-12 17:55:47 +0900 | [diff] [blame] | 571 | /* |
| 572 | * Check write append to a zoned block device. |
| 573 | */ |
| 574 | static inline blk_status_t blk_check_zone_append(struct request_queue *q, |
| 575 | struct bio *bio) |
| 576 | { |
Keith Busch | 0512a75 | 2020-05-12 17:55:47 +0900 | [diff] [blame] | 577 | int nr_sectors = bio_sectors(bio); |
| 578 | |
| 579 | /* Only applicable to zoned block devices */ |
Christoph Hellwig | edd1dbc | 2022-07-06 09:03:37 +0200 | [diff] [blame] | 580 | if (!bdev_is_zoned(bio->bi_bdev)) |
Keith Busch | 0512a75 | 2020-05-12 17:55:47 +0900 | [diff] [blame] | 581 | return BLK_STS_NOTSUPP; |
| 582 | |
| 583 | /* The bio sector must point to the start of a sequential zone */ |
Damien Le Moal | bca150f | 2024-04-08 10:41:23 +0900 | [diff] [blame] | 584 | if (!bdev_is_zone_start(bio->bi_bdev, bio->bi_iter.bi_sector)) |
Keith Busch | 0512a75 | 2020-05-12 17:55:47 +0900 | [diff] [blame] | 585 | return BLK_STS_IOERR; |
| 586 | |
| 587 | /* |
| 588 | * Not allowed to cross zone boundaries. Otherwise, the BIO will be |
| 589 | * split and could result in non-contiguous sectors being written in |
| 590 | * different zones. |
| 591 | */ |
| 592 | if (nr_sectors > q->limits.chunk_sectors) |
| 593 | return BLK_STS_IOERR; |
| 594 | |
| 595 | /* Make sure the BIO is small enough and will not get split */ |
Damien Le Moal | ccdbf0a | 2024-04-08 10:41:09 +0900 | [diff] [blame] | 596 | if (nr_sectors > queue_max_zone_append_sectors(q)) |
Keith Busch | 0512a75 | 2020-05-12 17:55:47 +0900 | [diff] [blame] | 597 | return BLK_STS_IOERR; |
| 598 | |
| 599 | bio->bi_opf |= REQ_NOMERGE; |
| 600 | |
| 601 | return BLK_STS_OK; |
| 602 | } |
| 603 | |
Christoph Hellwig | 3e08773 | 2021-10-12 13:12:24 +0200 | [diff] [blame] | 604 | static void __submit_bio(struct bio *bio) |
Christoph Hellwig | ac7c567 | 2020-05-16 20:28:01 +0200 | [diff] [blame] | 605 | { |
Yu Kuai | 9a42891 | 2024-05-22 04:03:08 +0800 | [diff] [blame] | 606 | /* If plug is not used, add new plug here to cache nsecs time. */ |
| 607 | struct blk_plug plug; |
| 608 | |
Ming Lei | 7f36b7d | 2022-02-16 12:45:08 +0800 | [diff] [blame] | 609 | if (unlikely(!blk_crypto_bio_prep(&bio))) |
| 610 | return; |
| 611 | |
Yu Kuai | 9a42891 | 2024-05-22 04:03:08 +0800 | [diff] [blame] | 612 | blk_start_plug(&plug); |
| 613 | |
Al Viro | ac2b6f9 | 2024-04-12 01:21:45 -0400 | [diff] [blame] | 614 | if (!bdev_test_flag(bio->bi_bdev, BD_HAS_SUBMIT_BIO)) { |
Christoph Hellwig | 3e08773 | 2021-10-12 13:12:24 +0200 | [diff] [blame] | 615 | blk_mq_submit_bio(bio); |
Ming Lei | 7f36b7d | 2022-02-16 12:45:08 +0800 | [diff] [blame] | 616 | } else if (likely(bio_queue_enter(bio) == 0)) { |
Jens Axboe | 9f4107b | 2023-04-14 07:32:02 -0600 | [diff] [blame] | 617 | struct gendisk *disk = bio->bi_bdev->bd_disk; |
| 618 | |
Ming Lei | 7f36b7d | 2022-02-16 12:45:08 +0800 | [diff] [blame] | 619 | disk->fops->submit_bio(bio); |
| 620 | blk_queue_exit(disk->queue); |
| 621 | } |
Yu Kuai | 9a42891 | 2024-05-22 04:03:08 +0800 | [diff] [blame] | 622 | |
| 623 | blk_finish_plug(&plug); |
Christoph Hellwig | ac7c567 | 2020-05-16 20:28:01 +0200 | [diff] [blame] | 624 | } |
| 625 | |
Christoph Hellwig | 566acf2 | 2020-07-01 10:59:45 +0200 | [diff] [blame] | 626 | /* |
| 627 | * The loop in this function may be a bit non-obvious, and so deserves some |
| 628 | * explanation: |
| 629 | * |
| 630 | * - Before entering the loop, bio->bi_next is NULL (as all callers ensure |
| 631 | * that), so we have a list with a single bio. |
| 632 | * - We pretend that we have just taken it off a longer list, so we assign |
| 633 | * bio_list to a pointer to the bio_list_on_stack, thus initialising the |
| 634 | * bio_list of new bios to be added. ->submit_bio() may indeed add some more |
| 635 | * bios through a recursive call to submit_bio_noacct. If it did, we find a |
| 636 | * non-NULL value in bio_list and re-enter the loop from the top. |
| 637 | * - In this case we really did just take the bio of the top of the list (no |
| 638 | * pretending) and so remove it from bio_list, and call into ->submit_bio() |
| 639 | * again. |
| 640 | * |
| 641 | * bio_list_on_stack[0] contains bios submitted by the current ->submit_bio. |
| 642 | * bio_list_on_stack[1] contains bios that were submitted before the current |
Ming Lei | 69fe0f2 | 2022-03-04 21:08:03 -0500 | [diff] [blame] | 643 | * ->submit_bio, but that haven't been processed yet. |
Christoph Hellwig | 566acf2 | 2020-07-01 10:59:45 +0200 | [diff] [blame] | 644 | */ |
Christoph Hellwig | 3e08773 | 2021-10-12 13:12:24 +0200 | [diff] [blame] | 645 | static void __submit_bio_noacct(struct bio *bio) |
Christoph Hellwig | 566acf2 | 2020-07-01 10:59:45 +0200 | [diff] [blame] | 646 | { |
| 647 | struct bio_list bio_list_on_stack[2]; |
Christoph Hellwig | 566acf2 | 2020-07-01 10:59:45 +0200 | [diff] [blame] | 648 | |
| 649 | BUG_ON(bio->bi_next); |
| 650 | |
| 651 | bio_list_init(&bio_list_on_stack[0]); |
| 652 | current->bio_list = bio_list_on_stack; |
| 653 | |
| 654 | do { |
Pavel Begunkov | eab4e02 | 2021-10-14 15:03:29 +0100 | [diff] [blame] | 655 | struct request_queue *q = bdev_get_queue(bio->bi_bdev); |
Christoph Hellwig | 566acf2 | 2020-07-01 10:59:45 +0200 | [diff] [blame] | 656 | struct bio_list lower, same; |
| 657 | |
Christoph Hellwig | 566acf2 | 2020-07-01 10:59:45 +0200 | [diff] [blame] | 658 | /* |
| 659 | * Create a fresh bio_list for all subordinate requests. |
| 660 | */ |
| 661 | bio_list_on_stack[1] = bio_list_on_stack[0]; |
| 662 | bio_list_init(&bio_list_on_stack[0]); |
| 663 | |
Christoph Hellwig | 3e08773 | 2021-10-12 13:12:24 +0200 | [diff] [blame] | 664 | __submit_bio(bio); |
Christoph Hellwig | 566acf2 | 2020-07-01 10:59:45 +0200 | [diff] [blame] | 665 | |
| 666 | /* |
| 667 | * Sort new bios into those for a lower level and those for the |
| 668 | * same level. |
| 669 | */ |
| 670 | bio_list_init(&lower); |
| 671 | bio_list_init(&same); |
| 672 | while ((bio = bio_list_pop(&bio_list_on_stack[0])) != NULL) |
Pavel Begunkov | eab4e02 | 2021-10-14 15:03:29 +0100 | [diff] [blame] | 673 | if (q == bdev_get_queue(bio->bi_bdev)) |
Christoph Hellwig | 566acf2 | 2020-07-01 10:59:45 +0200 | [diff] [blame] | 674 | bio_list_add(&same, bio); |
| 675 | else |
| 676 | bio_list_add(&lower, bio); |
| 677 | |
| 678 | /* |
| 679 | * Now assemble so we handle the lowest level first. |
| 680 | */ |
| 681 | bio_list_merge(&bio_list_on_stack[0], &lower); |
| 682 | bio_list_merge(&bio_list_on_stack[0], &same); |
| 683 | bio_list_merge(&bio_list_on_stack[0], &bio_list_on_stack[1]); |
| 684 | } while ((bio = bio_list_pop(&bio_list_on_stack[0]))); |
| 685 | |
| 686 | current->bio_list = NULL; |
Christoph Hellwig | 566acf2 | 2020-07-01 10:59:45 +0200 | [diff] [blame] | 687 | } |
| 688 | |
Christoph Hellwig | 3e08773 | 2021-10-12 13:12:24 +0200 | [diff] [blame] | 689 | static void __submit_bio_noacct_mq(struct bio *bio) |
Christoph Hellwig | ff93ea0 | 2020-07-01 10:59:46 +0200 | [diff] [blame] | 690 | { |
Christoph Hellwig | 7c792f3 | 2020-07-02 21:21:25 +0200 | [diff] [blame] | 691 | struct bio_list bio_list[2] = { }; |
Christoph Hellwig | ff93ea0 | 2020-07-01 10:59:46 +0200 | [diff] [blame] | 692 | |
Christoph Hellwig | 7c792f3 | 2020-07-02 21:21:25 +0200 | [diff] [blame] | 693 | current->bio_list = bio_list; |
Christoph Hellwig | ff93ea0 | 2020-07-01 10:59:46 +0200 | [diff] [blame] | 694 | |
| 695 | do { |
Christoph Hellwig | 3e08773 | 2021-10-12 13:12:24 +0200 | [diff] [blame] | 696 | __submit_bio(bio); |
Christoph Hellwig | 7c792f3 | 2020-07-02 21:21:25 +0200 | [diff] [blame] | 697 | } while ((bio = bio_list_pop(&bio_list[0]))); |
Christoph Hellwig | ff93ea0 | 2020-07-01 10:59:46 +0200 | [diff] [blame] | 698 | |
| 699 | current->bio_list = NULL; |
Christoph Hellwig | ff93ea0 | 2020-07-01 10:59:46 +0200 | [diff] [blame] | 700 | } |
| 701 | |
Ming Lei | 3f98c75 | 2022-02-16 12:45:10 +0800 | [diff] [blame] | 702 | void submit_bio_noacct_nocheck(struct bio *bio) |
| 703 | { |
Jinke Han | 0f7c8f0 | 2023-02-16 11:22:50 +0800 | [diff] [blame] | 704 | blk_cgroup_bio_start(bio); |
| 705 | blkcg_bio_issue_init(bio); |
| 706 | |
| 707 | if (!bio_flagged(bio, BIO_TRACE_COMPLETION)) { |
| 708 | trace_block_bio_queue(bio); |
| 709 | /* |
| 710 | * Now that enqueuing has been traced, we need to trace |
| 711 | * completion as well. |
| 712 | */ |
| 713 | bio_set_flag(bio, BIO_TRACE_COMPLETION); |
| 714 | } |
| 715 | |
Ming Lei | 3f98c75 | 2022-02-16 12:45:10 +0800 | [diff] [blame] | 716 | /* |
| 717 | * We only want one ->submit_bio to be active at a time, else stack |
| 718 | * usage with stacked devices could be a problem. Use current->bio_list |
| 719 | * to collect a list of requests submited by a ->submit_bio method while |
| 720 | * it is active, and then process them after it returned. |
| 721 | */ |
| 722 | if (current->bio_list) |
| 723 | bio_list_add(¤t->bio_list[0], bio); |
Al Viro | ac2b6f9 | 2024-04-12 01:21:45 -0400 | [diff] [blame] | 724 | else if (!bdev_test_flag(bio->bi_bdev, BD_HAS_SUBMIT_BIO)) |
Ming Lei | 3f98c75 | 2022-02-16 12:45:10 +0800 | [diff] [blame] | 725 | __submit_bio_noacct_mq(bio); |
| 726 | else |
| 727 | __submit_bio_noacct(bio); |
| 728 | } |
| 729 | |
John Garry | 9da3d1e | 2024-06-20 12:53:54 +0000 | [diff] [blame] | 730 | static blk_status_t blk_validate_atomic_write_op_size(struct request_queue *q, |
| 731 | struct bio *bio) |
| 732 | { |
| 733 | if (bio->bi_iter.bi_size > queue_atomic_write_unit_max_bytes(q)) |
| 734 | return BLK_STS_INVAL; |
| 735 | |
| 736 | if (bio->bi_iter.bi_size % queue_atomic_write_unit_min_bytes(q)) |
| 737 | return BLK_STS_INVAL; |
| 738 | |
| 739 | return BLK_STS_OK; |
| 740 | } |
| 741 | |
Christoph Hellwig | 27a84d5 | 2011-09-15 14:01:40 +0200 | [diff] [blame] | 742 | /** |
Christoph Hellwig | ed00aab | 2020-07-01 10:59:44 +0200 | [diff] [blame] | 743 | * submit_bio_noacct - re-submit a bio to the block device layer for I/O |
Christoph Hellwig | 27a84d5 | 2011-09-15 14:01:40 +0200 | [diff] [blame] | 744 | * @bio: The bio describing the location in memory and on the device. |
| 745 | * |
Christoph Hellwig | 3fdd408 | 2020-04-28 13:27:53 +0200 | [diff] [blame] | 746 | * This is a version of submit_bio() that shall only be used for I/O that is |
| 747 | * resubmitted to lower level drivers by stacking block drivers. All file |
| 748 | * systems and other upper level users of the block layer should use |
| 749 | * submit_bio() instead. |
Neil Brown | d89d879 | 2007-05-01 09:53:42 +0200 | [diff] [blame] | 750 | */ |
Christoph Hellwig | 3e08773 | 2021-10-12 13:12:24 +0200 | [diff] [blame] | 751 | void submit_bio_noacct(struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 752 | { |
| 753 | struct block_device *bdev = bio->bi_bdev; |
| 754 | struct request_queue *q = bdev_get_queue(bdev); |
| 755 | blk_status_t status = BLK_STS_IOERR; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 756 | |
| 757 | might_sleep(); |
| 758 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 759 | /* |
| 760 | * For a REQ_NOWAIT based request, return -EOPNOTSUPP |
| 761 | * if queue does not support NOWAIT. |
NeilBrown | 5ddfe96 | 2006-10-30 22:07:21 -0800 | [diff] [blame] | 762 | */ |
Christoph Hellwig | 568ec93 | 2022-09-27 09:58:15 +0200 | [diff] [blame] | 763 | if ((bio->bi_opf & REQ_NOWAIT) && !bdev_nowait(bdev)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 764 | goto not_supported; |
| 765 | |
| 766 | if (should_fail_bio(bio)) |
| 767 | goto end_io; |
Miaohe Lin | bdb7d42 | 2022-09-05 18:27:54 +0800 | [diff] [blame] | 768 | bio_check_ro(bio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 769 | if (!bio_flagged(bio, BIO_REMAPPED)) { |
| 770 | if (unlikely(bio_check_eod(bio))) |
| 771 | goto end_io; |
Al Viro | 3f9b8fb | 2024-04-12 00:54:19 -0400 | [diff] [blame] | 772 | if (bdev_is_partition(bdev) && |
| 773 | unlikely(blk_partition_remap(bio))) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 774 | goto end_io; |
| 775 | } |
| 776 | |
| 777 | /* |
| 778 | * Filter flush bio's early so that bio based drivers without flush |
| 779 | * support don't have to worry about them. |
| 780 | */ |
Christoph Hellwig | b4a6bb3 | 2022-11-02 00:09:03 -0700 | [diff] [blame] | 781 | if (op_is_flush(bio->bi_opf)) { |
| 782 | if (WARN_ON_ONCE(bio_op(bio) != REQ_OP_WRITE && |
| 783 | bio_op(bio) != REQ_OP_ZONE_APPEND)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 784 | goto end_io; |
Christoph Hellwig | 1122c0c | 2024-06-17 08:04:40 +0200 | [diff] [blame] | 785 | if (!bdev_write_cache(bdev)) { |
Christoph Hellwig | b4a6bb3 | 2022-11-02 00:09:03 -0700 | [diff] [blame] | 786 | bio->bi_opf &= ~(REQ_PREFLUSH | REQ_FUA); |
| 787 | if (!bio_sectors(bio)) { |
| 788 | status = BLK_STS_OK; |
| 789 | goto end_io; |
| 790 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 791 | } |
| 792 | } |
| 793 | |
hexue | 73e59d3 | 2024-07-18 15:08:17 +0800 | [diff] [blame] | 794 | if (!(q->limits.features & BLK_FEAT_POLL) && |
| 795 | (bio->bi_opf & REQ_POLLED)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 796 | bio_clear_polled(bio); |
hexue | 73e59d3 | 2024-07-18 15:08:17 +0800 | [diff] [blame] | 797 | goto not_supported; |
| 798 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 799 | |
| 800 | switch (bio_op(bio)) { |
Christoph Hellwig | 1c042f8 | 2023-12-21 08:05:38 +0100 | [diff] [blame] | 801 | case REQ_OP_READ: |
| 802 | case REQ_OP_WRITE: |
John Garry | 9da3d1e | 2024-06-20 12:53:54 +0000 | [diff] [blame] | 803 | if (bio->bi_opf & REQ_ATOMIC) { |
| 804 | status = blk_validate_atomic_write_op_size(q, bio); |
| 805 | if (status != BLK_STS_OK) |
| 806 | goto end_io; |
| 807 | } |
Christoph Hellwig | 1c042f8 | 2023-12-21 08:05:38 +0100 | [diff] [blame] | 808 | break; |
| 809 | case REQ_OP_FLUSH: |
| 810 | /* |
| 811 | * REQ_OP_FLUSH can't be submitted through bios, it is only |
| 812 | * synthetized in struct request by the flush state machine. |
| 813 | */ |
| 814 | goto not_supported; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 815 | case REQ_OP_DISCARD: |
Christoph Hellwig | 7020057 | 2022-04-15 06:52:55 +0200 | [diff] [blame] | 816 | if (!bdev_max_discard_sectors(bdev)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 817 | goto not_supported; |
NeilBrown | 5ddfe96 | 2006-10-30 22:07:21 -0800 | [diff] [blame] | 818 | break; |
Jens Axboe | 2056a78 | 2006-03-23 20:00:26 +0100 | [diff] [blame] | 819 | case REQ_OP_SECURE_ERASE: |
Christoph Hellwig | 44abff2c | 2022-04-15 06:52:57 +0200 | [diff] [blame] | 820 | if (!bdev_max_secure_erase_sectors(bdev)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 821 | goto not_supported; |
| 822 | break; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 823 | case REQ_OP_ZONE_APPEND: |
| 824 | status = blk_check_zone_append(q, bio); |
| 825 | if (status != BLK_STS_OK) |
| 826 | goto end_io; |
| 827 | break; |
Christoph Hellwig | 1c042f8 | 2023-12-21 08:05:38 +0100 | [diff] [blame] | 828 | case REQ_OP_WRITE_ZEROES: |
| 829 | if (!q->limits.max_write_zeroes_sectors) |
| 830 | goto not_supported; |
| 831 | break; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 832 | case REQ_OP_ZONE_RESET: |
| 833 | case REQ_OP_ZONE_OPEN: |
| 834 | case REQ_OP_ZONE_CLOSE: |
| 835 | case REQ_OP_ZONE_FINISH: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 836 | case REQ_OP_ZONE_RESET_ALL: |
Damien Le Moal | f2a7bea | 2024-07-04 14:28:15 +0900 | [diff] [blame] | 837 | if (!bdev_is_zoned(bio->bi_bdev)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 838 | goto not_supported; |
| 839 | break; |
Christoph Hellwig | 1c042f8 | 2023-12-21 08:05:38 +0100 | [diff] [blame] | 840 | case REQ_OP_DRV_IN: |
| 841 | case REQ_OP_DRV_OUT: |
| 842 | /* |
| 843 | * Driver private operations are only used with passthrough |
| 844 | * requests. |
| 845 | */ |
| 846 | fallthrough; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 847 | default: |
Christoph Hellwig | 1c042f8 | 2023-12-21 08:05:38 +0100 | [diff] [blame] | 848 | goto not_supported; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 849 | } |
| 850 | |
| 851 | if (blk_throtl_bio(bio)) |
Ming Lei | a650628 | 2022-02-16 12:45:07 +0800 | [diff] [blame] | 852 | return; |
Ming Lei | 3f98c75 | 2022-02-16 12:45:10 +0800 | [diff] [blame] | 853 | submit_bio_noacct_nocheck(bio); |
Ming Lei | d24c670 | 2022-02-16 12:45:11 +0800 | [diff] [blame] | 854 | return; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 855 | |
Neil Brown | d89d879 | 2007-05-01 09:53:42 +0200 | [diff] [blame] | 856 | not_supported: |
| 857 | status = BLK_STS_NOTSUPP; |
| 858 | end_io: |
| 859 | bio->bi_status = status; |
| 860 | bio_endio(bio); |
Neil Brown | d89d879 | 2007-05-01 09:53:42 +0200 | [diff] [blame] | 861 | } |
Christoph Hellwig | ed00aab | 2020-07-01 10:59:44 +0200 | [diff] [blame] | 862 | EXPORT_SYMBOL(submit_bio_noacct); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 863 | |
Hongyu Jin | f3c8998 | 2024-01-30 15:26:34 -0500 | [diff] [blame] | 864 | static void bio_set_ioprio(struct bio *bio) |
| 865 | { |
| 866 | /* Nobody set ioprio so far? Initialize it based on task's nice value */ |
| 867 | if (IOPRIO_PRIO_CLASS(bio->bi_ioprio) == IOPRIO_CLASS_NONE) |
| 868 | bio->bi_ioprio = get_current_ioprio(); |
| 869 | blkcg_set_ioprio(bio); |
| 870 | } |
| 871 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 872 | /** |
Randy Dunlap | 710027a | 2008-08-19 20:13:11 +0200 | [diff] [blame] | 873 | * submit_bio - submit a bio to the block device layer for I/O |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 874 | * @bio: The &struct bio which describes the I/O |
| 875 | * |
Christoph Hellwig | 3fdd408 | 2020-04-28 13:27:53 +0200 | [diff] [blame] | 876 | * submit_bio() is used to submit I/O requests to block devices. It is passed a |
| 877 | * fully set up &struct bio that describes the I/O that needs to be done. The |
Christoph Hellwig | 309dca30 | 2021-01-24 11:02:34 +0100 | [diff] [blame] | 878 | * bio will be send to the device described by the bi_bdev field. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 879 | * |
Christoph Hellwig | 3fdd408 | 2020-04-28 13:27:53 +0200 | [diff] [blame] | 880 | * The success/failure status of the request, along with notification of |
| 881 | * completion, is delivered asynchronously through the ->bi_end_io() callback |
Ping-Xiang Chen | e884808 | 2022-09-14 00:42:37 -0700 | [diff] [blame] | 882 | * in @bio. The bio must NOT be touched by the caller until ->bi_end_io() has |
Christoph Hellwig | 3fdd408 | 2020-04-28 13:27:53 +0200 | [diff] [blame] | 883 | * been called. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 884 | */ |
Christoph Hellwig | 3e08773 | 2021-10-12 13:12:24 +0200 | [diff] [blame] | 885 | void submit_bio(struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 886 | { |
Christoph Hellwig | a3e7689 | 2022-05-16 08:36:54 +0200 | [diff] [blame] | 887 | if (bio_op(bio) == REQ_OP_READ) { |
| 888 | task_io_account_read(bio->bi_iter.bi_size); |
| 889 | count_vm_events(PGPGIN, bio_sectors(bio)); |
| 890 | } else if (bio_op(bio) == REQ_OP_WRITE) { |
| 891 | count_vm_events(PGPGOUT, bio_sectors(bio)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 892 | } |
| 893 | |
Hongyu Jin | f3c8998 | 2024-01-30 15:26:34 -0500 | [diff] [blame] | 894 | bio_set_ioprio(bio); |
Christoph Hellwig | 3e08773 | 2021-10-12 13:12:24 +0200 | [diff] [blame] | 895 | submit_bio_noacct(bio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 896 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 897 | EXPORT_SYMBOL(submit_bio); |
| 898 | |
Jens Axboe | 1052b8a | 2018-11-26 08:21:49 -0700 | [diff] [blame] | 899 | /** |
Christoph Hellwig | 3e08773 | 2021-10-12 13:12:24 +0200 | [diff] [blame] | 900 | * bio_poll - poll for BIO completions |
| 901 | * @bio: bio to poll for |
Yang Guang | e30028a | 2021-11-26 00:20:55 +0800 | [diff] [blame] | 902 | * @iob: batches of IO |
Christoph Hellwig | 3e08773 | 2021-10-12 13:12:24 +0200 | [diff] [blame] | 903 | * @flags: BLK_POLL_* flags that control the behavior |
| 904 | * |
| 905 | * Poll for completions on queue associated with the bio. Returns number of |
| 906 | * completed entries found. |
| 907 | * |
| 908 | * Note: the caller must either be the context that submitted @bio, or |
| 909 | * be in a RCU critical section to prevent freeing of @bio. |
| 910 | */ |
Jens Axboe | 5a72e89 | 2021-10-12 09:24:29 -0600 | [diff] [blame] | 911 | int bio_poll(struct bio *bio, struct io_comp_batch *iob, unsigned int flags) |
Christoph Hellwig | 3e08773 | 2021-10-12 13:12:24 +0200 | [diff] [blame] | 912 | { |
Christoph Hellwig | 3e08773 | 2021-10-12 13:12:24 +0200 | [diff] [blame] | 913 | blk_qc_t cookie = READ_ONCE(bio->bi_cookie); |
Jens Axboe | 310726c | 2023-02-24 10:01:19 -0700 | [diff] [blame] | 914 | struct block_device *bdev; |
| 915 | struct request_queue *q; |
Ming Lei | 69fe0f2 | 2022-03-04 21:08:03 -0500 | [diff] [blame] | 916 | int ret = 0; |
Christoph Hellwig | 3e08773 | 2021-10-12 13:12:24 +0200 | [diff] [blame] | 917 | |
Jens Axboe | 310726c | 2023-02-24 10:01:19 -0700 | [diff] [blame] | 918 | bdev = READ_ONCE(bio->bi_bdev); |
| 919 | if (!bdev) |
| 920 | return 0; |
| 921 | |
| 922 | q = bdev_get_queue(bdev); |
Christoph Hellwig | 8023e14 | 2024-06-17 08:04:48 +0200 | [diff] [blame] | 923 | if (cookie == BLK_QC_T_NONE || !(q->limits.features & BLK_FEAT_POLL)) |
Christoph Hellwig | 3e08773 | 2021-10-12 13:12:24 +0200 | [diff] [blame] | 924 | return 0; |
| 925 | |
Christoph Hellwig | aa8dccc | 2022-01-27 08:05:49 +0100 | [diff] [blame] | 926 | blk_flush_plug(current->plug, false); |
Christoph Hellwig | 3e08773 | 2021-10-12 13:12:24 +0200 | [diff] [blame] | 927 | |
Jens Axboe | 33391ee | 2023-01-20 07:51:07 -0700 | [diff] [blame] | 928 | /* |
| 929 | * We need to be able to enter a frozen queue, similar to how |
| 930 | * timeouts also need to do that. If that is blocked, then we can |
| 931 | * have pending IO when a queue freeze is started, and then the |
| 932 | * wait for the freeze to finish will wait for polled requests to |
| 933 | * timeout as the poller is preventer from entering the queue and |
| 934 | * completing them. As long as we prevent new IO from being queued, |
| 935 | * that should be all that matters. |
| 936 | */ |
| 937 | if (!percpu_ref_tryget(&q->q_usage_counter)) |
Christoph Hellwig | 3e08773 | 2021-10-12 13:12:24 +0200 | [diff] [blame] | 938 | return 0; |
Ming Lei | 69fe0f2 | 2022-03-04 21:08:03 -0500 | [diff] [blame] | 939 | if (queue_is_mq(q)) { |
Jens Axboe | 5a72e89 | 2021-10-12 09:24:29 -0600 | [diff] [blame] | 940 | ret = blk_mq_poll(q, cookie, iob, flags); |
Ming Lei | 69fe0f2 | 2022-03-04 21:08:03 -0500 | [diff] [blame] | 941 | } else { |
| 942 | struct gendisk *disk = q->disk; |
| 943 | |
| 944 | if (disk && disk->fops->poll_bio) |
| 945 | ret = disk->fops->poll_bio(bio, iob, flags); |
| 946 | } |
Christoph Hellwig | 3e08773 | 2021-10-12 13:12:24 +0200 | [diff] [blame] | 947 | blk_queue_exit(q); |
| 948 | return ret; |
| 949 | } |
| 950 | EXPORT_SYMBOL_GPL(bio_poll); |
| 951 | |
| 952 | /* |
| 953 | * Helper to implement file_operations.iopoll. Requires the bio to be stored |
| 954 | * in iocb->private, and cleared before freeing the bio. |
| 955 | */ |
Jens Axboe | 5a72e89 | 2021-10-12 09:24:29 -0600 | [diff] [blame] | 956 | int iocb_bio_iopoll(struct kiocb *kiocb, struct io_comp_batch *iob, |
| 957 | unsigned int flags) |
Christoph Hellwig | 3e08773 | 2021-10-12 13:12:24 +0200 | [diff] [blame] | 958 | { |
| 959 | struct bio *bio; |
| 960 | int ret = 0; |
| 961 | |
| 962 | /* |
| 963 | * Note: the bio cache only uses SLAB_TYPESAFE_BY_RCU, so bio can |
| 964 | * point to a freshly allocated bio at this point. If that happens |
| 965 | * we have a few cases to consider: |
| 966 | * |
| 967 | * 1) the bio is beeing initialized and bi_bdev is NULL. We can just |
| 968 | * simply nothing in this case |
| 969 | * 2) the bio points to a not poll enabled device. bio_poll will catch |
| 970 | * this and return 0 |
| 971 | * 3) the bio points to a poll capable device, including but not |
| 972 | * limited to the one that the original bio pointed to. In this |
| 973 | * case we will call into the actual poll method and poll for I/O, |
| 974 | * even if we don't need to, but it won't cause harm either. |
| 975 | * |
| 976 | * For cases 2) and 3) above the RCU grace period ensures that bi_bdev |
| 977 | * is still allocated. Because partitions hold a reference to the whole |
| 978 | * device bdev and thus disk, the disk is also still valid. Grabbing |
| 979 | * a reference to the queue in bio_poll() ensures the hctxs and requests |
| 980 | * are still valid as well. |
| 981 | */ |
| 982 | rcu_read_lock(); |
| 983 | bio = READ_ONCE(kiocb->private); |
Jens Axboe | 310726c | 2023-02-24 10:01:19 -0700 | [diff] [blame] | 984 | if (bio) |
Jens Axboe | 5a72e89 | 2021-10-12 09:24:29 -0600 | [diff] [blame] | 985 | ret = bio_poll(bio, iob, flags); |
Christoph Hellwig | 3e08773 | 2021-10-12 13:12:24 +0200 | [diff] [blame] | 986 | rcu_read_unlock(); |
| 987 | |
| 988 | return ret; |
| 989 | } |
| 990 | EXPORT_SYMBOL_GPL(iocb_bio_iopoll); |
| 991 | |
Christoph Hellwig | 450b787 | 2021-11-17 07:14:01 +0100 | [diff] [blame] | 992 | void update_io_ticks(struct block_device *part, unsigned long now, bool end) |
Christoph Hellwig | 9123bf6 | 2020-05-27 07:24:13 +0200 | [diff] [blame] | 993 | { |
| 994 | unsigned long stamp; |
| 995 | again: |
Christoph Hellwig | 8446fe9 | 2020-11-24 09:36:54 +0100 | [diff] [blame] | 996 | stamp = READ_ONCE(part->bd_stamp); |
Yu Kuai | 99dc422 | 2024-05-09 20:37:16 +0800 | [diff] [blame] | 997 | if (unlikely(time_after(now, stamp)) && |
| 998 | likely(try_cmpxchg(&part->bd_stamp, &stamp, now)) && |
| 999 | (end || part_in_flight(part))) |
| 1000 | __part_stat_add(part, io_ticks, now - stamp); |
| 1001 | |
Al Viro | 3f9b8fb | 2024-04-12 00:54:19 -0400 | [diff] [blame] | 1002 | if (bdev_is_partition(part)) { |
Christoph Hellwig | 8446fe9 | 2020-11-24 09:36:54 +0100 | [diff] [blame] | 1003 | part = bdev_whole(part); |
Christoph Hellwig | 9123bf6 | 2020-05-27 07:24:13 +0200 | [diff] [blame] | 1004 | goto again; |
| 1005 | } |
| 1006 | } |
| 1007 | |
Yu Kuai | 5f27571 | 2023-02-23 17:12:26 +0800 | [diff] [blame] | 1008 | unsigned long bdev_start_io_acct(struct block_device *bdev, enum req_op op, |
Ming Lei | 5f0614a | 2022-04-17 22:27:13 -0400 | [diff] [blame] | 1009 | unsigned long start_time) |
Christoph Hellwig | 956d510 | 2020-05-27 07:24:04 +0200 | [diff] [blame] | 1010 | { |
Christoph Hellwig | 956d510 | 2020-05-27 07:24:04 +0200 | [diff] [blame] | 1011 | part_stat_lock(); |
Ming Lei | 5f0614a | 2022-04-17 22:27:13 -0400 | [diff] [blame] | 1012 | update_io_ticks(bdev, start_time, false); |
Ming Lei | 5f0614a | 2022-04-17 22:27:13 -0400 | [diff] [blame] | 1013 | part_stat_local_inc(bdev, in_flight[op_is_write(op)]); |
Christoph Hellwig | 956d510 | 2020-05-27 07:24:04 +0200 | [diff] [blame] | 1014 | part_stat_unlock(); |
| 1015 | |
Mike Snitzer | e45c47d | 2022-01-28 10:58:39 -0500 | [diff] [blame] | 1016 | return start_time; |
Christoph Hellwig | 956d510 | 2020-05-27 07:24:04 +0200 | [diff] [blame] | 1017 | } |
Ming Lei | 5f0614a | 2022-04-17 22:27:13 -0400 | [diff] [blame] | 1018 | EXPORT_SYMBOL(bdev_start_io_acct); |
Song Liu | 7b26410 | 2020-08-31 15:27:23 -0700 | [diff] [blame] | 1019 | |
Christoph Hellwig | 99dfc43 | 2021-01-24 11:02:37 +0100 | [diff] [blame] | 1020 | /** |
| 1021 | * bio_start_io_acct - start I/O accounting for bio based drivers |
| 1022 | * @bio: bio to start account for |
| 1023 | * |
| 1024 | * Returns the start time that should be passed back to bio_end_io_acct(). |
| 1025 | */ |
| 1026 | unsigned long bio_start_io_acct(struct bio *bio) |
Song Liu | 7b26410 | 2020-08-31 15:27:23 -0700 | [diff] [blame] | 1027 | { |
Yu Kuai | 5f27571 | 2023-02-23 17:12:26 +0800 | [diff] [blame] | 1028 | return bdev_start_io_acct(bio->bi_bdev, bio_op(bio), jiffies); |
Song Liu | 7b26410 | 2020-08-31 15:27:23 -0700 | [diff] [blame] | 1029 | } |
Christoph Hellwig | 99dfc43 | 2021-01-24 11:02:37 +0100 | [diff] [blame] | 1030 | EXPORT_SYMBOL_GPL(bio_start_io_acct); |
Song Liu | 7b26410 | 2020-08-31 15:27:23 -0700 | [diff] [blame] | 1031 | |
Bart Van Assche | 77e7ffd | 2022-07-14 11:06:28 -0700 | [diff] [blame] | 1032 | void bdev_end_io_acct(struct block_device *bdev, enum req_op op, |
Yu Kuai | 5f27571 | 2023-02-23 17:12:26 +0800 | [diff] [blame] | 1033 | unsigned int sectors, unsigned long start_time) |
Christoph Hellwig | 956d510 | 2020-05-27 07:24:04 +0200 | [diff] [blame] | 1034 | { |
Christoph Hellwig | 956d510 | 2020-05-27 07:24:04 +0200 | [diff] [blame] | 1035 | const int sgrp = op_stat_group(op); |
| 1036 | unsigned long now = READ_ONCE(jiffies); |
| 1037 | unsigned long duration = now - start_time; |
| 1038 | |
| 1039 | part_stat_lock(); |
Ming Lei | 5f0614a | 2022-04-17 22:27:13 -0400 | [diff] [blame] | 1040 | update_io_ticks(bdev, now, true); |
Yu Kuai | 5f27571 | 2023-02-23 17:12:26 +0800 | [diff] [blame] | 1041 | part_stat_inc(bdev, ios[sgrp]); |
| 1042 | part_stat_add(bdev, sectors[sgrp], sectors); |
Ming Lei | 5f0614a | 2022-04-17 22:27:13 -0400 | [diff] [blame] | 1043 | part_stat_add(bdev, nsecs[sgrp], jiffies_to_nsecs(duration)); |
| 1044 | part_stat_local_dec(bdev, in_flight[op_is_write(op)]); |
Christoph Hellwig | 956d510 | 2020-05-27 07:24:04 +0200 | [diff] [blame] | 1045 | part_stat_unlock(); |
| 1046 | } |
Ming Lei | 5f0614a | 2022-04-17 22:27:13 -0400 | [diff] [blame] | 1047 | EXPORT_SYMBOL(bdev_end_io_acct); |
Song Liu | 7b26410 | 2020-08-31 15:27:23 -0700 | [diff] [blame] | 1048 | |
Christoph Hellwig | 99dfc43 | 2021-01-24 11:02:37 +0100 | [diff] [blame] | 1049 | void bio_end_io_acct_remapped(struct bio *bio, unsigned long start_time, |
Ming Lei | 5f0614a | 2022-04-17 22:27:13 -0400 | [diff] [blame] | 1050 | struct block_device *orig_bdev) |
Song Liu | 7b26410 | 2020-08-31 15:27:23 -0700 | [diff] [blame] | 1051 | { |
Yu Kuai | 5f27571 | 2023-02-23 17:12:26 +0800 | [diff] [blame] | 1052 | bdev_end_io_acct(orig_bdev, bio_op(bio), bio_sectors(bio), start_time); |
Song Liu | 7b26410 | 2020-08-31 15:27:23 -0700 | [diff] [blame] | 1053 | } |
Christoph Hellwig | 99dfc43 | 2021-01-24 11:02:37 +0100 | [diff] [blame] | 1054 | EXPORT_SYMBOL_GPL(bio_end_io_acct_remapped); |
Song Liu | 7b26410 | 2020-08-31 15:27:23 -0700 | [diff] [blame] | 1055 | |
Kiyoshi Ueda | ef9e3fa | 2008-10-01 16:12:15 +0200 | [diff] [blame] | 1056 | /** |
| 1057 | * blk_lld_busy - Check if underlying low-level drivers of a device are busy |
| 1058 | * @q : the queue of the device being checked |
| 1059 | * |
| 1060 | * Description: |
| 1061 | * Check if underlying low-level drivers of a device are busy. |
| 1062 | * If the drivers want to export their busy state, they must set own |
| 1063 | * exporting function using blk_queue_lld_busy() first. |
| 1064 | * |
| 1065 | * Basically, this function is used only by request stacking drivers |
| 1066 | * to stop dispatching requests to underlying devices when underlying |
| 1067 | * devices are busy. This behavior helps more I/O merging on the queue |
| 1068 | * of the request stacking driver and prevents I/O throughput regression |
| 1069 | * on burst I/O load. |
| 1070 | * |
| 1071 | * Return: |
| 1072 | * 0 - Not busy (The request stacking driver should dispatch request) |
| 1073 | * 1 - Busy (The request stacking driver should stop dispatching request) |
| 1074 | */ |
| 1075 | int blk_lld_busy(struct request_queue *q) |
| 1076 | { |
Jens Axboe | 344e9ff | 2018-11-15 12:22:51 -0700 | [diff] [blame] | 1077 | if (queue_is_mq(q) && q->mq_ops->busy) |
Jens Axboe | 9ba2052 | 2018-10-29 10:15:10 -0600 | [diff] [blame] | 1078 | return q->mq_ops->busy(q); |
Kiyoshi Ueda | ef9e3fa | 2008-10-01 16:12:15 +0200 | [diff] [blame] | 1079 | |
| 1080 | return 0; |
| 1081 | } |
| 1082 | EXPORT_SYMBOL_GPL(blk_lld_busy); |
| 1083 | |
Jens Axboe | 59c3d45 | 2014-04-08 09:15:35 -0600 | [diff] [blame] | 1084 | int kblockd_schedule_work(struct work_struct *work) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1085 | { |
| 1086 | return queue_work(kblockd_workqueue, work); |
| 1087 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1088 | EXPORT_SYMBOL(kblockd_schedule_work); |
| 1089 | |
Jens Axboe | 818cd1c | 2017-04-10 09:54:55 -0600 | [diff] [blame] | 1090 | int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, |
| 1091 | unsigned long delay) |
| 1092 | { |
| 1093 | return mod_delayed_work_on(cpu, kblockd_workqueue, dwork, delay); |
| 1094 | } |
| 1095 | EXPORT_SYMBOL(kblockd_mod_delayed_work_on); |
| 1096 | |
Jens Axboe | 47c122e | 2021-10-06 06:34:11 -0600 | [diff] [blame] | 1097 | void blk_start_plug_nr_ios(struct blk_plug *plug, unsigned short nr_ios) |
| 1098 | { |
| 1099 | struct task_struct *tsk = current; |
| 1100 | |
| 1101 | /* |
| 1102 | * If this is a nested plug, don't actually assign it. |
| 1103 | */ |
| 1104 | if (tsk->plug) |
| 1105 | return; |
| 1106 | |
Jens Axboe | da4c8c3 | 2024-01-15 14:46:03 -0700 | [diff] [blame] | 1107 | plug->cur_ktime = 0; |
Jens Axboe | bc490f8 | 2021-10-18 10:12:12 -0600 | [diff] [blame] | 1108 | plug->mq_list = NULL; |
Jens Axboe | 47c122e | 2021-10-06 06:34:11 -0600 | [diff] [blame] | 1109 | plug->cached_rq = NULL; |
| 1110 | plug->nr_ios = min_t(unsigned short, nr_ios, BLK_MAX_REQUEST_COUNT); |
| 1111 | plug->rq_count = 0; |
| 1112 | plug->multiple_queues = false; |
Jens Axboe | dc5fc361 | 2021-10-19 06:02:30 -0600 | [diff] [blame] | 1113 | plug->has_elevator = false; |
Jens Axboe | 47c122e | 2021-10-06 06:34:11 -0600 | [diff] [blame] | 1114 | INIT_LIST_HEAD(&plug->cb_list); |
| 1115 | |
| 1116 | /* |
| 1117 | * Store ordering should not be needed here, since a potential |
| 1118 | * preempt will imply a full memory barrier |
| 1119 | */ |
| 1120 | tsk->plug = plug; |
| 1121 | } |
| 1122 | |
Suresh Jayaraman | 75df713 | 2011-09-21 10:00:16 +0200 | [diff] [blame] | 1123 | /** |
| 1124 | * blk_start_plug - initialize blk_plug and track it inside the task_struct |
| 1125 | * @plug: The &struct blk_plug that needs to be initialized |
| 1126 | * |
| 1127 | * Description: |
Jeff Moyer | 4040585 | 2019-01-08 16:57:34 -0500 | [diff] [blame] | 1128 | * blk_start_plug() indicates to the block layer an intent by the caller |
| 1129 | * to submit multiple I/O requests in a batch. The block layer may use |
| 1130 | * this hint to defer submitting I/Os from the caller until blk_finish_plug() |
| 1131 | * is called. However, the block layer may choose to submit requests |
| 1132 | * before a call to blk_finish_plug() if the number of queued I/Os |
| 1133 | * exceeds %BLK_MAX_REQUEST_COUNT, or if the size of the I/O is larger than |
| 1134 | * %BLK_PLUG_FLUSH_SIZE. The queued I/Os may also be submitted early if |
| 1135 | * the task schedules (see below). |
| 1136 | * |
Suresh Jayaraman | 75df713 | 2011-09-21 10:00:16 +0200 | [diff] [blame] | 1137 | * Tracking blk_plug inside the task_struct will help with auto-flushing the |
| 1138 | * pending I/O should the task end up blocking between blk_start_plug() and |
| 1139 | * blk_finish_plug(). This is important from a performance perspective, but |
| 1140 | * also ensures that we don't deadlock. For instance, if the task is blocking |
| 1141 | * for a memory allocation, memory reclaim could end up wanting to free a |
| 1142 | * page belonging to that request that is currently residing in our private |
| 1143 | * plug. By flushing the pending I/O when the process goes to sleep, we avoid |
| 1144 | * this kind of deadlock. |
| 1145 | */ |
Jens Axboe | 73c1010 | 2011-03-08 13:19:51 +0100 | [diff] [blame] | 1146 | void blk_start_plug(struct blk_plug *plug) |
| 1147 | { |
Jens Axboe | 47c122e | 2021-10-06 06:34:11 -0600 | [diff] [blame] | 1148 | blk_start_plug_nr_ios(plug, 1); |
Jens Axboe | 73c1010 | 2011-03-08 13:19:51 +0100 | [diff] [blame] | 1149 | } |
| 1150 | EXPORT_SYMBOL(blk_start_plug); |
| 1151 | |
NeilBrown | 74018dc | 2012-07-31 09:08:15 +0200 | [diff] [blame] | 1152 | static void flush_plug_callbacks(struct blk_plug *plug, bool from_schedule) |
NeilBrown | 048c937 | 2011-04-18 09:52:22 +0200 | [diff] [blame] | 1153 | { |
| 1154 | LIST_HEAD(callbacks); |
| 1155 | |
Shaohua Li | 2a7d555 | 2012-07-31 09:08:15 +0200 | [diff] [blame] | 1156 | while (!list_empty(&plug->cb_list)) { |
| 1157 | list_splice_init(&plug->cb_list, &callbacks); |
NeilBrown | 048c937 | 2011-04-18 09:52:22 +0200 | [diff] [blame] | 1158 | |
Shaohua Li | 2a7d555 | 2012-07-31 09:08:15 +0200 | [diff] [blame] | 1159 | while (!list_empty(&callbacks)) { |
| 1160 | struct blk_plug_cb *cb = list_first_entry(&callbacks, |
NeilBrown | 048c937 | 2011-04-18 09:52:22 +0200 | [diff] [blame] | 1161 | struct blk_plug_cb, |
| 1162 | list); |
Shaohua Li | 2a7d555 | 2012-07-31 09:08:15 +0200 | [diff] [blame] | 1163 | list_del(&cb->list); |
NeilBrown | 74018dc | 2012-07-31 09:08:15 +0200 | [diff] [blame] | 1164 | cb->callback(cb, from_schedule); |
Shaohua Li | 2a7d555 | 2012-07-31 09:08:15 +0200 | [diff] [blame] | 1165 | } |
NeilBrown | 048c937 | 2011-04-18 09:52:22 +0200 | [diff] [blame] | 1166 | } |
| 1167 | } |
| 1168 | |
NeilBrown | 9cbb175 | 2012-07-31 09:08:14 +0200 | [diff] [blame] | 1169 | struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug, void *data, |
| 1170 | int size) |
| 1171 | { |
| 1172 | struct blk_plug *plug = current->plug; |
| 1173 | struct blk_plug_cb *cb; |
| 1174 | |
| 1175 | if (!plug) |
| 1176 | return NULL; |
| 1177 | |
| 1178 | list_for_each_entry(cb, &plug->cb_list, list) |
| 1179 | if (cb->callback == unplug && cb->data == data) |
| 1180 | return cb; |
| 1181 | |
| 1182 | /* Not currently on the callback list */ |
| 1183 | BUG_ON(size < sizeof(*cb)); |
| 1184 | cb = kzalloc(size, GFP_ATOMIC); |
| 1185 | if (cb) { |
| 1186 | cb->data = data; |
| 1187 | cb->callback = unplug; |
| 1188 | list_add(&cb->list, &plug->cb_list); |
| 1189 | } |
| 1190 | return cb; |
| 1191 | } |
| 1192 | EXPORT_SYMBOL(blk_check_plugged); |
| 1193 | |
Christoph Hellwig | aa8dccc | 2022-01-27 08:05:49 +0100 | [diff] [blame] | 1194 | void __blk_flush_plug(struct blk_plug *plug, bool from_schedule) |
Jens Axboe | 73c1010 | 2011-03-08 13:19:51 +0100 | [diff] [blame] | 1195 | { |
Pavel Begunkov | b600455 | 2021-10-20 16:41:18 +0200 | [diff] [blame] | 1196 | if (!list_empty(&plug->cb_list)) |
| 1197 | flush_plug_callbacks(plug, from_schedule); |
Ross Lagerwall | 7090426 | 2023-07-14 11:11:06 +0100 | [diff] [blame] | 1198 | blk_mq_flush_plug_list(plug, from_schedule); |
Jens Axboe | c5fc7b9 | 2021-11-03 05:49:07 -0600 | [diff] [blame] | 1199 | /* |
| 1200 | * Unconditionally flush out cached requests, even if the unplug |
| 1201 | * event came from schedule. Since we know hold references to the |
| 1202 | * queue for cached requests, we don't want a blocked task holding |
| 1203 | * up a queue freeze/quiesce event. |
| 1204 | */ |
| 1205 | if (unlikely(!rq_list_empty(plug->cached_rq))) |
Jens Axboe | 47c122e | 2021-10-06 06:34:11 -0600 | [diff] [blame] | 1206 | blk_mq_free_plug_rqs(plug); |
Jens Axboe | 06b23f92 | 2024-01-16 09:18:39 -0700 | [diff] [blame] | 1207 | |
Yu Kuai | 3ec4848 | 2024-04-11 11:23:48 +0800 | [diff] [blame] | 1208 | plug->cur_ktime = 0; |
Jens Axboe | 06b23f92 | 2024-01-16 09:18:39 -0700 | [diff] [blame] | 1209 | current->flags &= ~PF_BLOCK_TS; |
Jens Axboe | 73c1010 | 2011-03-08 13:19:51 +0100 | [diff] [blame] | 1210 | } |
Jens Axboe | 73c1010 | 2011-03-08 13:19:51 +0100 | [diff] [blame] | 1211 | |
Jeff Moyer | 4040585 | 2019-01-08 16:57:34 -0500 | [diff] [blame] | 1212 | /** |
| 1213 | * blk_finish_plug - mark the end of a batch of submitted I/O |
| 1214 | * @plug: The &struct blk_plug passed to blk_start_plug() |
| 1215 | * |
| 1216 | * Description: |
| 1217 | * Indicate that a batch of I/O submissions is complete. This function |
| 1218 | * must be paired with an initial call to blk_start_plug(). The intent |
| 1219 | * is to allow the block layer to optimize I/O submission. See the |
| 1220 | * documentation for blk_start_plug() for more information. |
| 1221 | */ |
Jens Axboe | 73c1010 | 2011-03-08 13:19:51 +0100 | [diff] [blame] | 1222 | void blk_finish_plug(struct blk_plug *plug) |
| 1223 | { |
Christoph Hellwig | 008f75a | 2021-10-20 16:41:19 +0200 | [diff] [blame] | 1224 | if (plug == current->plug) { |
Christoph Hellwig | aa8dccc | 2022-01-27 08:05:49 +0100 | [diff] [blame] | 1225 | __blk_flush_plug(plug, false); |
Christoph Hellwig | 008f75a | 2021-10-20 16:41:19 +0200 | [diff] [blame] | 1226 | current->plug = NULL; |
| 1227 | } |
Jens Axboe | 73c1010 | 2011-03-08 13:19:51 +0100 | [diff] [blame] | 1228 | } |
| 1229 | EXPORT_SYMBOL(blk_finish_plug); |
| 1230 | |
Ming Lei | 71ac860 | 2020-05-14 16:45:09 +0800 | [diff] [blame] | 1231 | void blk_io_schedule(void) |
| 1232 | { |
| 1233 | /* Prevent hang_check timer from firing at us during very long I/O */ |
| 1234 | unsigned long timeout = sysctl_hung_task_timeout_secs * HZ / 2; |
| 1235 | |
| 1236 | if (timeout) |
| 1237 | io_schedule_timeout(timeout); |
| 1238 | else |
| 1239 | io_schedule(); |
| 1240 | } |
| 1241 | EXPORT_SYMBOL_GPL(blk_io_schedule); |
| 1242 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1243 | int __init blk_dev_init(void) |
| 1244 | { |
Bart Van Assche | 16458cf | 2022-07-14 11:06:32 -0700 | [diff] [blame] | 1245 | BUILD_BUG_ON((__force u32)REQ_OP_LAST >= (1 << REQ_OP_BITS)); |
Christoph Hellwig | ef295ec | 2016-10-28 08:48:16 -0600 | [diff] [blame] | 1246 | BUILD_BUG_ON(REQ_OP_BITS + REQ_FLAG_BITS > 8 * |
Pankaj Bharadiya | c593642 | 2019-12-09 10:31:43 -0800 | [diff] [blame] | 1247 | sizeof_field(struct request, cmd_flags)); |
Christoph Hellwig | ef295ec | 2016-10-28 08:48:16 -0600 | [diff] [blame] | 1248 | BUILD_BUG_ON(REQ_OP_BITS + REQ_FLAG_BITS > 8 * |
Pankaj Bharadiya | c593642 | 2019-12-09 10:31:43 -0800 | [diff] [blame] | 1249 | sizeof_field(struct bio, bi_opf)); |
Nikanth Karthikesan | 9eb55b0 | 2009-04-27 14:53:54 +0200 | [diff] [blame] | 1250 | |
Tejun Heo | 89b90be | 2011-01-03 15:01:47 +0100 | [diff] [blame] | 1251 | /* used for unplugging and affects IO latency/throughput - HIGHPRI */ |
| 1252 | kblockd_workqueue = alloc_workqueue("kblockd", |
Matias Bjørling | 28747fc | 2014-06-11 23:43:54 +0200 | [diff] [blame] | 1253 | WQ_MEM_RECLAIM | WQ_HIGHPRI, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1254 | if (!kblockd_workqueue) |
| 1255 | panic("Failed to create kblockd\n"); |
| 1256 | |
Kunwu Chan | 48ff13a | 2024-01-31 17:43:23 +0800 | [diff] [blame] | 1257 | blk_requestq_cachep = KMEM_CACHE(request_queue, SLAB_PANIC); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1258 | |
Omar Sandoval | 18fbda9 | 2017-01-31 14:53:20 -0800 | [diff] [blame] | 1259 | blk_debugfs_root = debugfs_create_dir("block", NULL); |
Omar Sandoval | 18fbda9 | 2017-01-31 14:53:20 -0800 | [diff] [blame] | 1260 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1261 | return 0; |
| 1262 | } |