| 1 | |
| 2 | |
| 3 | * fs/f2fs/data.c
|
| 4 | *
|
| 5 | * Copyright (c) 2012 Samsung Electronics Co., Ltd.
|
| 6 | * http://www.samsung.com/
|
| 7 |
|
| 8 | #include <linux/fs.h>
|
| 9 | #include <linux/f2fs_fs.h>
|
| 10 | #include <linux/sched/mm.h>
|
| 11 | #include <linux/mpage.h>
|
| 12 | #include <linux/writeback.h>
|
| 13 | #include <linux/folio_batch.h>
|
| 14 | #include <linux/blkdev.h>
|
| 15 | #include <linux/bio.h>
|
| 16 | #include <linux/blk-crypto.h>
|
| 17 | #include <linux/swap.h>
|
| 18 | #include <linux/prefetch.h>
|
| 19 | #include <linux/uio.h>
|
| 20 | #include <linux/sched/signal.h>
|
| 21 | #include <linux/fiemap.h>
|
| 22 | #include <linux/iomap.h>
|
| 23 | #include <linux/fserror.h>
|
| 24 |
|
| 25 | #include "f2fs.h"
|
| 26 | #include "node.h"
|
| 27 | #include "segment.h"
|
| 28 | #include "iostat.h"
|
| 29 | #include <trace/events/f2fs.h>
|
| 30 |
|
| 31 | #define NUM_PREALLOC_POST_READ_CTXS 128
|
| 32 |
|
| 33 | static struct kmem_cache *bio_post_read_ctx_cache;
|
| 34 | static struct kmem_cache *bio_entry_slab;
|
| 35 | static struct kmem_cache *ffs_entry_slab;
|
| 36 | static mempool_t *bio_post_read_ctx_pool;
|
| 37 | static struct bio_set f2fs_bioset;
|
| 38 |
|
| 39 | struct f2fs_folio_state {
|
| 40 | spinlock_t state_lock;
|
| 41 | unsigned int read_pages_pending;
|
| 42 | };
|
| 43 |
|
| 44 | struct f2fs_bio {
|
| 45 | struct work_struct work;
|
| 46 | struct bio bio;
|
| 47 | };
|
| 48 |
|
| 49 | #define F2FS_BIO_POOL_SIZE NR_CURSEG_TYPE
|
| 50 |
|
| 51 | int __init f2fs_init_bioset(void)
|
| 52 | {
|
| 53 | return bioset_init(&f2fs_bioset, F2FS_BIO_POOL_SIZE,
|
| 54 | offsetof(struct f2fs_bio, bio), BIOSET_NEED_BVECS);
|
| 55 | }
|
| 56 |
|
| 57 | void f2fs_destroy_bioset(void)
|
| 58 | {
|
| 59 | bioset_exit(&f2fs_bioset);
|
| 60 | }
|
| 61 |
|
| 62 | bool f2fs_is_cp_guaranteed(const struct folio *folio)
|
| 63 | {
|
| 64 | struct address_space *mapping = folio->mapping;
|
| 65 | struct inode *inode;
|
| 66 | struct f2fs_sb_info *sbi;
|
| 67 |
|
| 68 | inode = mapping->host;
|
| 69 | sbi = F2FS_I_SB(inode);
|
| 70 |
|
| 71 | if (inode->i_ino == F2FS_META_INO(sbi) ||
|
| 72 | inode->i_ino == F2FS_NODE_INO(sbi) ||
|
| 73 | S_ISDIR(inode->i_mode))
|
| 74 | return true;
|
| 75 |
|
| 76 | if ((S_ISREG(inode->i_mode) && IS_NOQUOTA(inode)) ||
|
| 77 | folio_test_f2fs_gcing(folio))
|
| 78 | return true;
|
| 79 | return false;
|
| 80 | }
|
| 81 |
|
| 82 | static enum count_type __read_io_type(struct folio *folio)
|
| 83 | {
|
| 84 | struct address_space *mapping = folio->mapping;
|
| 85 |
|
| 86 | if (mapping) {
|
| 87 | struct inode *inode = mapping->host;
|
| 88 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
| 89 |
|
| 90 | if (inode->i_ino == F2FS_META_INO(sbi))
|
| 91 | return F2FS_RD_META;
|
| 92 |
|
| 93 | if (inode->i_ino == F2FS_NODE_INO(sbi))
|
| 94 | return F2FS_RD_NODE;
|
| 95 | }
|
| 96 | return F2FS_RD_DATA;
|
| 97 | }
|
| 98 |
|
| 99 |
|
| 100 | enum bio_post_read_step {
|
| 101 | #ifdef CONFIG_F2FS_FS_COMPRESSION
|
| 102 | STEP_DECOMPRESS = BIT(1),
|
| 103 | #else
|
| 104 | STEP_DECOMPRESS = 0,
|
| 105 | #endif
|
| 106 | #ifdef CONFIG_FS_VERITY
|
| 107 | STEP_VERITY = BIT(2),
|
| 108 | #else
|
| 109 | STEP_VERITY = 0,
|
| 110 | #endif
|
| 111 | };
|
| 112 |
|
| 113 | struct bio_post_read_ctx {
|
| 114 | struct bio *bio;
|
| 115 | struct f2fs_sb_info *sbi;
|
| 116 | struct fsverity_info *vi;
|
| 117 | struct work_struct work;
|
| 118 | unsigned int enabled_steps;
|
| 119 | |
| 120 | * decompression_attempted keeps track of whether
|
| 121 | * f2fs_end_read_compressed_page() has been called on the pages in the
|
| 122 | * bio that belong to a compressed cluster yet.
|
| 123 |
|
| 124 | bool decompression_attempted;
|
| 125 | block_t fs_blkaddr;
|
| 126 | };
|
| 127 |
|
| 128 | |
| 129 | * Update and unlock a bio's pages, and free the bio.
|
| 130 | *
|
| 131 | * This marks pages up-to-date only if there was no error in the bio (I/O error,
|
| 132 | * decryption error, or verity error), as indicated by bio->bi_status.
|
| 133 | *
|
| 134 | * "Compressed pages" (pagecache pages backed by a compressed cluster on-disk)
|
| 135 | * aren't marked up-to-date here, as decompression is done on a per-compression-
|
| 136 | * cluster basis rather than a per-bio basis. Instead, we only must do two
|
| 137 | * things for each compressed page here: call f2fs_end_read_compressed_page()
|
| 138 | * with failed=true if an error occurred before it would have normally gotten
|
| 139 | * called (i.e., I/O error or decryption error, but *not* verity error), and
|
| 140 | * release the bio's reference to the decompress_io_ctx of the page's cluster.
|
| 141 |
|
| 142 | static void f2fs_finish_read_bio(struct bio *bio, bool in_task)
|
| 143 | {
|
| 144 | struct folio_iter fi;
|
| 145 | struct bio_post_read_ctx *ctx = bio->bi_private;
|
| 146 | unsigned long flags;
|
| 147 |
|
| 148 | bio_for_each_folio_all(fi, bio) {
|
| 149 | struct folio *folio = fi.folio;
|
| 150 | unsigned nr_pages = fi.length >> PAGE_SHIFT;
|
| 151 | bool finished = true;
|
| 152 |
|
| 153 | if (!folio_test_large(folio) &&
|
| 154 | f2fs_is_compressed_page(folio)) {
|
| 155 | if (ctx && !ctx->decompression_attempted)
|
| 156 | f2fs_end_read_compressed_page(folio, true, 0,
|
| 157 | in_task);
|
| 158 | f2fs_put_folio_dic(folio, in_task);
|
| 159 | continue;
|
| 160 | }
|
| 161 |
|
| 162 | if (folio_test_large(folio)) {
|
| 163 | struct f2fs_folio_state *ffs = folio->private;
|
| 164 |
|
| 165 | spin_lock_irqsave(&ffs->state_lock, flags);
|
| 166 | ffs->read_pages_pending -= nr_pages;
|
| 167 | finished = !ffs->read_pages_pending;
|
| 168 | spin_unlock_irqrestore(&ffs->state_lock, flags);
|
| 169 | }
|
| 170 |
|
| 171 | while (nr_pages--)
|
| 172 | dec_page_count(F2FS_F_SB(folio), __read_io_type(folio));
|
| 173 |
|
| 174 | if (bio->bi_status == BLK_STS_OK &&
|
| 175 | F2FS_F_SB(folio)->node_inode && is_node_folio(folio) &&
|
| 176 | f2fs_sanity_check_node_footer(F2FS_F_SB(folio),
|
| 177 | folio, folio->index, NODE_TYPE_REGULAR, true))
|
| 178 | bio->bi_status = BLK_STS_IOERR;
|
| 179 |
|
| 180 | if (finished)
|
| 181 | folio_end_read(folio, bio->bi_status == BLK_STS_OK);
|
| 182 | }
|
| 183 |
|
| 184 | if (ctx)
|
| 185 | mempool_free(ctx, bio_post_read_ctx_pool);
|
| 186 | bio_put(bio);
|
| 187 | }
|
| 188 |
|
| 189 | static void f2fs_verify_bio(struct work_struct *work)
|
| 190 | {
|
| 191 | struct bio_post_read_ctx *ctx =
|
| 192 | container_of(work, struct bio_post_read_ctx, work);
|
| 193 | struct bio *bio = ctx->bio;
|
| 194 | bool may_have_compressed_pages = (ctx->enabled_steps & STEP_DECOMPRESS);
|
| 195 | struct fsverity_info *vi = ctx->vi;
|
| 196 |
|
| 197 | |
| 198 | * fsverity_verify_bio() may call readahead() again, and while verity
|
| 199 | * will be disabled for this, decryption and/or decompression may still
|
| 200 | * be needed, resulting in another bio_post_read_ctx being allocated.
|
| 201 | * So to prevent deadlocks we need to release the current ctx to the
|
| 202 | * mempool first. This assumes that verity is the last post-read step.
|
| 203 |
|
| 204 | mempool_free(ctx, bio_post_read_ctx_pool);
|
| 205 | bio->bi_private = NULL;
|
| 206 |
|
| 207 | |
| 208 | * Verify the bio's pages with fs-verity. Exclude compressed pages,
|
| 209 | * as those were handled separately by f2fs_end_read_compressed_page().
|
| 210 |
|
| 211 | if (may_have_compressed_pages) {
|
| 212 | struct folio_iter fi;
|
| 213 |
|
| 214 | bio_for_each_folio_all(fi, bio) {
|
| 215 | struct folio *folio = fi.folio;
|
| 216 |
|
| 217 | if (!f2fs_is_compressed_page(folio) &&
|
| 218 | !fsverity_verify_folio(vi, folio)) {
|
| 219 | bio->bi_status = BLK_STS_IOERR;
|
| 220 | break;
|
| 221 | }
|
| 222 | }
|
| 223 | } else {
|
| 224 | fsverity_verify_bio(vi, bio);
|
| 225 | }
|
| 226 |
|
| 227 | f2fs_finish_read_bio(bio, true);
|
| 228 | }
|
| 229 |
|
| 230 | |
| 231 | * If the bio's data needs to be verified with fs-verity, then enqueue the
|
| 232 | * verity work for the bio. Otherwise finish the bio now.
|
| 233 | *
|
| 234 | * Note that to avoid deadlocks, the verity work can't be done on the
|
| 235 | * decryption/decompression workqueue. This is because verifying the data pages
|
| 236 | * can involve reading verity metadata pages from the file, and these verity
|
| 237 | * metadata pages may be encrypted and/or compressed.
|
| 238 |
|
| 239 | static void f2fs_verify_and_finish_bio(struct bio *bio, bool in_task)
|
| 240 | {
|
| 241 | struct bio_post_read_ctx *ctx = bio->bi_private;
|
| 242 |
|
| 243 | if (ctx && (ctx->enabled_steps & STEP_VERITY)) {
|
| 244 | INIT_WORK(&ctx->work, f2fs_verify_bio);
|
| 245 | fsverity_enqueue_verify_work(&ctx->work);
|
| 246 | } else {
|
| 247 | f2fs_finish_read_bio(bio, in_task);
|
| 248 | }
|
| 249 | }
|
| 250 |
|
| 251 | |
| 252 | * Handle STEP_DECOMPRESS by decompressing any compressed clusters whose last
|
| 253 | * remaining page was read by @ctx->bio.
|
| 254 | *
|
| 255 | * Note that a bio may span clusters (even a mix of compressed and uncompressed
|
| 256 | * clusters) or be for just part of a cluster. STEP_DECOMPRESS just indicates
|
| 257 | * that the bio includes at least one compressed page. The actual decompression
|
| 258 | * is done on a per-cluster basis, not a per-bio basis.
|
| 259 |
|
| 260 | static void f2fs_handle_step_decompress(struct bio_post_read_ctx *ctx,
|
| 261 | bool in_task)
|
| 262 | {
|
| 263 | struct folio_iter fi;
|
| 264 | bool all_compressed = true;
|
| 265 | block_t blkaddr = ctx->fs_blkaddr;
|
| 266 |
|
| 267 | bio_for_each_folio_all(fi, ctx->bio) {
|
| 268 | struct folio *folio = fi.folio;
|
| 269 |
|
| 270 | if (f2fs_is_compressed_page(folio))
|
| 271 | f2fs_end_read_compressed_page(folio, false, blkaddr,
|
| 272 | in_task);
|
| 273 | else
|
| 274 | all_compressed = false;
|
| 275 |
|
| 276 | blkaddr++;
|
| 277 | }
|
| 278 |
|
| 279 | ctx->decompression_attempted = true;
|
| 280 |
|
| 281 | |
| 282 | * Optimization: if all the bio's pages are compressed, then scheduling
|
| 283 | * the per-bio verity work is unnecessary, as verity will be fully
|
| 284 | * handled at the compression cluster level.
|
| 285 |
|
| 286 | if (all_compressed)
|
| 287 | ctx->enabled_steps &= ~STEP_VERITY;
|
| 288 | }
|
| 289 |
|
| 290 | static void f2fs_post_read_work(struct work_struct *work)
|
| 291 | {
|
| 292 | struct bio_post_read_ctx *ctx =
|
| 293 | container_of(work, struct bio_post_read_ctx, work);
|
| 294 | struct bio *bio = ctx->bio;
|
| 295 |
|
| 296 | if (ctx->enabled_steps & STEP_DECOMPRESS)
|
| 297 | f2fs_handle_step_decompress(ctx, true);
|
| 298 |
|
| 299 | f2fs_verify_and_finish_bio(bio, true);
|
| 300 | }
|
| 301 |
|
| 302 | static void f2fs_read_end_io(struct bio *bio)
|
| 303 | {
|
| 304 | struct f2fs_sb_info *sbi = F2FS_F_SB(bio_first_folio_all(bio));
|
| 305 | struct bio_post_read_ctx *ctx;
|
| 306 | bool intask = in_task() && !irqs_disabled();
|
| 307 |
|
| 308 | iostat_update_and_unbind_ctx(bio);
|
| 309 | ctx = bio->bi_private;
|
| 310 |
|
| 311 | if (time_to_inject(sbi, FAULT_READ_IO))
|
| 312 | bio->bi_status = BLK_STS_IOERR;
|
| 313 |
|
| 314 | if (bio->bi_status != BLK_STS_OK) {
|
| 315 | f2fs_finish_read_bio(bio, intask);
|
| 316 | return;
|
| 317 | }
|
| 318 |
|
| 319 | if (ctx && (ctx->enabled_steps & STEP_DECOMPRESS)) {
|
| 320 | if (!f2fs_low_mem_mode(sbi)) {
|
| 321 |
|
| 322 | f2fs_handle_step_decompress(ctx, intask);
|
| 323 | } else {
|
| 324 | INIT_WORK(&ctx->work, f2fs_post_read_work);
|
| 325 | queue_work(ctx->sbi->wq, &ctx->work);
|
| 326 | return;
|
| 327 | }
|
| 328 | }
|
| 329 |
|
| 330 | f2fs_verify_and_finish_bio(bio, intask);
|
| 331 | }
|
| 332 |
|
| 333 | static void f2fs_write_end_bio(struct bio *bio)
|
| 334 | {
|
| 335 | struct f2fs_sb_info *sbi = bio->bi_private;
|
| 336 | struct folio_iter fi;
|
| 337 |
|
| 338 | if (time_to_inject(sbi, FAULT_WRITE_IO))
|
| 339 | bio->bi_status = BLK_STS_IOERR;
|
| 340 |
|
| 341 | bio_for_each_folio_all(fi, bio) {
|
| 342 | struct folio *folio = fi.folio;
|
| 343 | enum count_type type;
|
| 344 |
|
| 345 | #ifdef CONFIG_F2FS_FS_COMPRESSION
|
| 346 | if (f2fs_is_compressed_page(folio)) {
|
| 347 | f2fs_compress_write_end_io(bio, folio);
|
| 348 | continue;
|
| 349 | }
|
| 350 | #endif
|
| 351 |
|
| 352 | type = WB_DATA_TYPE(folio, false);
|
| 353 |
|
| 354 | if (unlikely(bio->bi_status != BLK_STS_OK)) {
|
| 355 | mapping_set_error(folio->mapping, -EIO);
|
| 356 | if (type == F2FS_WB_CP_DATA) {
|
| 357 | f2fs_stop_checkpoint(sbi, true,
|
| 358 | STOP_CP_REASON_WRITE_FAIL);
|
| 359 | }
|
| 360 | }
|
| 361 |
|
| 362 | if (is_node_folio(folio)) {
|
| 363 | f2fs_sanity_check_node_footer(sbi, folio,
|
| 364 | folio->index, NODE_TYPE_REGULAR, true);
|
| 365 | f2fs_bug_on(sbi, folio->index != nid_of_node(folio));
|
| 366 | }
|
| 367 | if (f2fs_in_warm_node_list(folio))
|
| 368 | f2fs_del_fsync_node_entry(sbi, folio);
|
| 369 |
|
| 370 | dec_page_count(sbi, type);
|
| 371 |
|
| 372 | |
| 373 | * we should access sbi before folio_end_writeback() to
|
| 374 | * avoid racing w/ kill_f2fs_super()
|
| 375 |
|
| 376 | if (type == F2FS_WB_CP_DATA && !get_pages(sbi, type) &&
|
| 377 | wq_has_sleeper(&sbi->cp_wait))
|
| 378 | wake_up(&sbi->cp_wait);
|
| 379 |
|
| 380 | folio_clear_f2fs_gcing(folio);
|
| 381 | folio_end_writeback(folio);
|
| 382 | }
|
| 383 |
|
| 384 | bio_put(bio);
|
| 385 | }
|
| 386 |
|
| 387 | static void f2fs_write_end_io_work(struct work_struct *work)
|
| 388 | {
|
| 389 | struct bio *bio = &container_of(work, struct f2fs_bio, work)->bio;
|
| 390 |
|
| 391 | f2fs_write_end_bio(bio);
|
| 392 | }
|
| 393 |
|
| 394 | static void f2fs_write_end_io(struct bio *bio)
|
| 395 | {
|
| 396 | struct f2fs_sb_info *sbi;
|
| 397 |
|
| 398 | iostat_update_and_unbind_ctx(bio);
|
| 399 |
|
| 400 | sbi = bio->bi_private;
|
| 401 |
|
| 402 | if (in_atomic() && bio->bi_iter.bi_size > sbi->max_atc_write_bio_size) {
|
| 403 | struct work_struct *w;
|
| 404 |
|
| 405 | w = &container_of(bio, struct f2fs_bio, bio)->work;
|
| 406 | INIT_WORK(w, f2fs_write_end_io_work);
|
| 407 | queue_work(sbi->wq, w);
|
| 408 | } else {
|
| 409 | f2fs_write_end_bio(bio);
|
| 410 | }
|
| 411 | }
|
| 412 |
|
| 413 | #ifdef CONFIG_BLK_DEV_ZONED
|
| 414 | static void f2fs_zone_write_end_io(struct bio *bio)
|
| 415 | {
|
| 416 | struct f2fs_bio_info *io = (struct f2fs_bio_info *)bio->bi_private;
|
| 417 |
|
| 418 | bio->bi_private = io->bi_private;
|
| 419 | complete(&io->zone_wait);
|
| 420 | f2fs_write_end_io(bio);
|
| 421 | }
|
| 422 | #endif
|
| 423 |
|
| 424 | struct block_device *f2fs_target_device(struct f2fs_sb_info *sbi,
|
| 425 | block_t blk_addr, sector_t *sector)
|
| 426 | {
|
| 427 | struct block_device *bdev = sbi->sb->s_bdev;
|
| 428 | int i;
|
| 429 |
|
| 430 | if (f2fs_is_multi_device(sbi)) {
|
| 431 | for (i = 0; i < sbi->s_ndevs; i++) {
|
| 432 | if (FDEV(i).start_blk <= blk_addr &&
|
| 433 | FDEV(i).end_blk >= blk_addr) {
|
| 434 | blk_addr -= FDEV(i).start_blk;
|
| 435 | bdev = FDEV(i).bdev;
|
| 436 | break;
|
| 437 | }
|
| 438 | }
|
| 439 | }
|
| 440 |
|
| 441 | if (sector)
|
| 442 | *sector = SECTOR_FROM_BLOCK(blk_addr);
|
| 443 | return bdev;
|
| 444 | }
|
| 445 |
|
| 446 | int f2fs_target_device_index(struct f2fs_sb_info *sbi, block_t blkaddr)
|
| 447 | {
|
| 448 | int i;
|
| 449 |
|
| 450 | if (!f2fs_is_multi_device(sbi))
|
| 451 | return 0;
|
| 452 |
|
| 453 | for (i = 0; i < sbi->s_ndevs; i++)
|
| 454 | if (FDEV(i).start_blk <= blkaddr && FDEV(i).end_blk >= blkaddr)
|
| 455 | return i;
|
| 456 | return 0;
|
| 457 | }
|
| 458 |
|
| 459 | static blk_opf_t f2fs_io_flags(struct f2fs_io_info *fio)
|
| 460 | {
|
| 461 | unsigned int temp_mask = GENMASK(NR_TEMP_TYPE - 1, 0);
|
| 462 | unsigned int fua_flag, meta_flag, io_flag;
|
| 463 | blk_opf_t op_flags = 0;
|
| 464 |
|
| 465 | if (fio->op != REQ_OP_WRITE)
|
| 466 | return 0;
|
| 467 | if (fio->type == DATA)
|
| 468 | io_flag = fio->sbi->data_io_flag;
|
| 469 | else if (fio->type == NODE)
|
| 470 | io_flag = fio->sbi->node_io_flag;
|
| 471 | else
|
| 472 | return 0;
|
| 473 |
|
| 474 | fua_flag = io_flag & temp_mask;
|
| 475 | meta_flag = (io_flag >> NR_TEMP_TYPE) & temp_mask;
|
| 476 |
|
| 477 | |
| 478 | * data/node io flag bits per temp:
|
| 479 | * REQ_META | REQ_FUA |
|
| 480 | * 5 | 4 | 3 | 2 | 1 | 0 |
|
| 481 | * Cold | Warm | Hot | Cold | Warm | Hot |
|
| 482 |
|
| 483 | if (BIT(fio->temp) & meta_flag)
|
| 484 | op_flags |= REQ_META;
|
| 485 | if (BIT(fio->temp) & fua_flag)
|
| 486 | op_flags |= REQ_FUA;
|
| 487 |
|
| 488 | if (fio->type == DATA &&
|
| 489 | F2FS_I(fio->folio->mapping->host)->ioprio_hint == F2FS_IOPRIO_WRITE)
|
| 490 | op_flags |= REQ_PRIO;
|
| 491 |
|
| 492 | return op_flags;
|
| 493 | }
|
| 494 |
|
| 495 | static struct bio *__bio_alloc(struct f2fs_io_info *fio, int npages)
|
| 496 | {
|
| 497 | struct f2fs_sb_info *sbi = fio->sbi;
|
| 498 | struct block_device *bdev;
|
| 499 | sector_t sector;
|
| 500 | struct bio *bio;
|
| 501 |
|
| 502 | bdev = f2fs_target_device(sbi, fio->new_blkaddr, §or);
|
| 503 | bio = bio_alloc_bioset(bdev, npages,
|
| 504 | fio->op | fio->op_flags | f2fs_io_flags(fio),
|
| 505 | GFP_NOIO, &f2fs_bioset);
|
| 506 | bio->bi_iter.bi_sector = sector;
|
| 507 | if (is_read_io(fio->op)) {
|
| 508 | bio->bi_end_io = f2fs_read_end_io;
|
| 509 | bio->bi_private = NULL;
|
| 510 | } else {
|
| 511 | bio->bi_end_io = f2fs_write_end_io;
|
| 512 | bio->bi_private = sbi;
|
| 513 | bio->bi_write_hint = f2fs_io_type_to_rw_hint(sbi,
|
| 514 | fio->type, fio->temp);
|
| 515 | bio->bi_write_stream = f2fs_io_type_to_write_stream(bdev, fio->type,
|
| 516 | fio->temp);
|
| 517 | }
|
| 518 | iostat_alloc_and_bind_ctx(sbi, bio, NULL);
|
| 519 |
|
| 520 | if (fio->io_wbc)
|
| 521 | wbc_init_bio(fio->io_wbc, bio);
|
| 522 |
|
| 523 | return bio;
|
| 524 | }
|
| 525 |
|
| 526 | static void f2fs_set_bio_crypt_ctx(struct bio *bio, const struct inode *inode,
|
| 527 | pgoff_t first_idx,
|
| 528 | const struct f2fs_io_info *fio,
|
| 529 | gfp_t gfp_mask)
|
| 530 | {
|
| 531 | |
| 532 | * The f2fs garbage collector sets ->encrypted_page when it wants to
|
| 533 | * read/write raw data without encryption.
|
| 534 |
|
| 535 | if (!fio || !fio->encrypted_page)
|
| 536 | fscrypt_set_bio_crypt_ctx(bio, inode,
|
| 537 | (loff_t)first_idx << inode->i_blkbits,
|
| 538 | gfp_mask);
|
| 539 | }
|
| 540 |
|
| 541 | static bool f2fs_crypt_mergeable_bio(struct bio *bio, const struct inode *inode,
|
| 542 | pgoff_t next_idx,
|
| 543 | const struct f2fs_io_info *fio)
|
| 544 | {
|
| 545 | |
| 546 | * The f2fs garbage collector sets ->encrypted_page when it wants to
|
| 547 | * read/write raw data without encryption.
|
| 548 |
|
| 549 | if (fio && fio->encrypted_page)
|
| 550 | return !bio_has_crypt_ctx(bio);
|
| 551 |
|
| 552 | return fscrypt_mergeable_bio(bio, inode,
|
| 553 | (loff_t)next_idx << inode->i_blkbits);
|
| 554 | }
|
| 555 |
|
| 556 | void f2fs_submit_read_bio(struct f2fs_sb_info *sbi, struct bio *bio,
|
| 557 | enum page_type type)
|
| 558 | {
|
| 559 | if (!bio)
|
| 560 | return;
|
| 561 |
|
| 562 | WARN_ON_ONCE(!is_read_io(bio_op(bio)));
|
| 563 | trace_f2fs_submit_read_bio(sbi->sb, type, bio);
|
| 564 |
|
| 565 | iostat_update_submit_ctx(bio, type);
|
| 566 | blk_crypto_submit_bio(bio);
|
| 567 | }
|
| 568 |
|
| 569 | static void f2fs_submit_write_bio(struct f2fs_sb_info *sbi, struct bio *bio,
|
| 570 | enum page_type type)
|
| 571 | {
|
| 572 | WARN_ON_ONCE(is_read_io(bio_op(bio)));
|
| 573 | trace_f2fs_submit_write_bio(sbi->sb, type, bio);
|
| 574 | iostat_update_submit_ctx(bio, type);
|
| 575 | blk_crypto_submit_bio(bio);
|
| 576 | }
|
| 577 |
|
| 578 | static void __submit_merged_bio(struct f2fs_bio_info *io)
|
| 579 | {
|
| 580 | struct f2fs_io_info *fio = &io->fio;
|
| 581 |
|
| 582 | if (!io->bio)
|
| 583 | return;
|
| 584 |
|
| 585 | if (is_read_io(fio->op)) {
|
| 586 | trace_f2fs_prepare_read_bio(io->sbi->sb, fio->type, io->bio);
|
| 587 | f2fs_submit_read_bio(io->sbi, io->bio, fio->type);
|
| 588 | } else {
|
| 589 | trace_f2fs_prepare_write_bio(io->sbi->sb, fio->type, io->bio);
|
| 590 | f2fs_submit_write_bio(io->sbi, io->bio, fio->type);
|
| 591 | }
|
| 592 | io->bio = NULL;
|
| 593 | }
|
| 594 |
|
| 595 | static bool __has_merged_page(struct bio *bio, struct inode *inode,
|
| 596 | struct folio *folio, nid_t ino)
|
| 597 | {
|
| 598 | struct folio_iter fi;
|
| 599 |
|
| 600 | if (!bio)
|
| 601 | return false;
|
| 602 |
|
| 603 | if (!inode && !folio && !ino)
|
| 604 | return true;
|
| 605 |
|
| 606 | bio_for_each_folio_all(fi, bio) {
|
| 607 | struct folio *target = fi.folio;
|
| 608 |
|
| 609 | if (f2fs_is_compressed_page(target)) {
|
| 610 | target = f2fs_compress_control_folio(target);
|
| 611 | if (IS_ERR(target))
|
| 612 | continue;
|
| 613 | }
|
| 614 |
|
| 615 | if (inode && inode == target->mapping->host)
|
| 616 | return true;
|
| 617 | if (folio && folio == target)
|
| 618 | return true;
|
| 619 | if (ino && ino == ino_of_node(target))
|
| 620 | return true;
|
| 621 | }
|
| 622 |
|
| 623 | return false;
|
| 624 | }
|
| 625 |
|
| 626 | int f2fs_init_write_merge_io(struct f2fs_sb_info *sbi)
|
| 627 | {
|
| 628 | int i;
|
| 629 |
|
| 630 | for (i = 0; i < NR_PAGE_TYPE; i++) {
|
| 631 | int n = (i == META) ? 1 : NR_TEMP_TYPE;
|
| 632 | int j;
|
| 633 |
|
| 634 | sbi->write_io[i] = f2fs_kmalloc(sbi,
|
| 635 | array_size(n, sizeof(struct f2fs_bio_info)),
|
| 636 | GFP_KERNEL);
|
| 637 | if (!sbi->write_io[i])
|
| 638 | return -ENOMEM;
|
| 639 |
|
| 640 | for (j = HOT; j < n; j++) {
|
| 641 | struct f2fs_bio_info *io = &sbi->write_io[i][j];
|
| 642 |
|
| 643 | init_f2fs_rwsem_trace(&io->io_rwsem, sbi,
|
| 644 | LOCK_NAME_IO_RWSEM);
|
| 645 | io->sbi = sbi;
|
| 646 | io->bio = NULL;
|
| 647 | io->last_block_in_bio = 0;
|
| 648 | spin_lock_init(&io->io_lock);
|
| 649 | INIT_LIST_HEAD(&io->io_list);
|
| 650 | INIT_LIST_HEAD(&io->bio_list);
|
| 651 | init_f2fs_rwsem(&io->bio_list_lock);
|
| 652 | #ifdef CONFIG_BLK_DEV_ZONED
|
| 653 | init_completion(&io->zone_wait);
|
| 654 | io->zone_pending_bio = NULL;
|
| 655 | io->bi_private = NULL;
|
| 656 | #endif
|
| 657 | }
|
| 658 | }
|
| 659 |
|
| 660 | return 0;
|
| 661 | }
|
| 662 |
|
| 663 | static void __f2fs_submit_merged_write(struct f2fs_sb_info *sbi,
|
| 664 | enum page_type type, enum temp_type temp)
|
| 665 | {
|
| 666 | enum page_type btype = PAGE_TYPE_OF_BIO(type);
|
| 667 | struct f2fs_bio_info *io = sbi->write_io[btype] + temp;
|
| 668 | struct f2fs_lock_context lc;
|
| 669 |
|
| 670 | f2fs_down_write_trace(&io->io_rwsem, &lc);
|
| 671 |
|
| 672 | if (!io->bio)
|
| 673 | goto unlock_out;
|
| 674 |
|
| 675 |
|
| 676 | if (type >= META_FLUSH) {
|
| 677 | io->fio.type = META_FLUSH;
|
| 678 | io->bio->bi_opf |= REQ_META | REQ_PRIO | REQ_SYNC;
|
| 679 | if (!test_opt(sbi, NOBARRIER))
|
| 680 | io->bio->bi_opf |= REQ_PREFLUSH | REQ_FUA;
|
| 681 | }
|
| 682 | __submit_merged_bio(io);
|
| 683 | unlock_out:
|
| 684 | f2fs_up_write_trace(&io->io_rwsem, &lc);
|
| 685 | }
|
| 686 |
|
| 687 | static void __submit_merged_write_cond(struct f2fs_sb_info *sbi,
|
| 688 | struct inode *inode, struct folio *folio,
|
| 689 | nid_t ino, enum page_type type, bool writeback)
|
| 690 | {
|
| 691 | enum temp_type temp;
|
| 692 | bool ret = true;
|
| 693 | bool force = !inode && !folio && !ino;
|
| 694 |
|
| 695 | for (temp = HOT; temp < NR_TEMP_TYPE; temp++) {
|
| 696 | if (!force) {
|
| 697 | enum page_type btype = PAGE_TYPE_OF_BIO(type);
|
| 698 | struct f2fs_bio_info *io = sbi->write_io[btype] + temp;
|
| 699 | struct f2fs_lock_context lc;
|
| 700 |
|
| 701 | f2fs_down_read_trace(&io->io_rwsem, &lc);
|
| 702 | ret = __has_merged_page(io->bio, inode, folio, ino);
|
| 703 | f2fs_up_read_trace(&io->io_rwsem, &lc);
|
| 704 | }
|
| 705 | if (ret) {
|
| 706 | __f2fs_submit_merged_write(sbi, type, temp);
|
| 707 | |
| 708 | * For waitting writebck case, if the bio owned by the
|
| 709 | * folio is already submitted, we do not need to submit
|
| 710 | * other types of bios.
|
| 711 |
|
| 712 | if (writeback)
|
| 713 | break;
|
| 714 | }
|
| 715 |
|
| 716 |
|
| 717 | if (type >= META)
|
| 718 | break;
|
| 719 | }
|
| 720 | }
|
| 721 |
|
| 722 | void f2fs_submit_merged_write(struct f2fs_sb_info *sbi, enum page_type type)
|
| 723 | {
|
| 724 | __submit_merged_write_cond(sbi, NULL, NULL, 0, type, false);
|
| 725 | }
|
| 726 |
|
| 727 | void f2fs_submit_merged_write_cond(struct f2fs_sb_info *sbi,
|
| 728 | struct inode *inode, struct folio *folio,
|
| 729 | nid_t ino, enum page_type type)
|
| 730 | {
|
| 731 | __submit_merged_write_cond(sbi, inode, folio, ino, type, false);
|
| 732 | }
|
| 733 |
|
| 734 | void f2fs_submit_merged_write_folio(struct f2fs_sb_info *sbi,
|
| 735 | struct folio *folio, enum page_type type)
|
| 736 | {
|
| 737 | __submit_merged_write_cond(sbi, NULL, folio, 0, type, true);
|
| 738 | }
|
| 739 |
|
| 740 | void f2fs_flush_merged_writes(struct f2fs_sb_info *sbi)
|
| 741 | {
|
| 742 | f2fs_submit_merged_write(sbi, DATA);
|
| 743 | f2fs_submit_merged_write(sbi, NODE);
|
| 744 | f2fs_submit_merged_write(sbi, META);
|
| 745 | }
|
| 746 |
|
| 747 | |
| 748 | * Fill the locked page with data located in the block address.
|
| 749 | * A caller needs to unlock the page on failure.
|
| 750 |
|
| 751 | int f2fs_submit_page_bio(struct f2fs_io_info *fio)
|
| 752 | {
|
| 753 | struct bio *bio;
|
| 754 | struct folio *fio_folio = fio->folio;
|
| 755 | struct folio *data_folio = fio->encrypted_page ?
|
| 756 | page_folio(fio->encrypted_page) : fio_folio;
|
| 757 |
|
| 758 | if (!f2fs_is_valid_blkaddr(fio->sbi, fio->new_blkaddr,
|
| 759 | fio->is_por ? META_POR : (__is_meta_io(fio) ?
|
| 760 | META_GENERIC : DATA_GENERIC_ENHANCE)))
|
| 761 | return -EFSCORRUPTED;
|
| 762 |
|
| 763 | trace_f2fs_submit_folio_bio(data_folio, fio);
|
| 764 |
|
| 765 |
|
| 766 | bio = __bio_alloc(fio, 1);
|
| 767 |
|
| 768 | f2fs_set_bio_crypt_ctx(bio, fio_folio->mapping->host,
|
| 769 | fio_folio->index, fio, GFP_NOIO);
|
| 770 | bio_add_folio_nofail(bio, data_folio, folio_size(data_folio), 0);
|
| 771 |
|
| 772 | if (fio->io_wbc && !is_read_io(fio->op))
|
| 773 | wbc_account_cgroup_owner(fio->io_wbc, fio_folio, PAGE_SIZE);
|
| 774 |
|
| 775 | inc_page_count(fio->sbi, is_read_io(fio->op) ?
|
| 776 | __read_io_type(data_folio) : WB_DATA_TYPE(fio->folio, false));
|
| 777 |
|
| 778 | if (is_read_io(bio_op(bio)))
|
| 779 | f2fs_submit_read_bio(fio->sbi, bio, fio->type);
|
| 780 | else
|
| 781 | f2fs_submit_write_bio(fio->sbi, bio, fio->type);
|
| 782 | return 0;
|
| 783 | }
|
| 784 |
|
| 785 | static bool page_is_mergeable(struct f2fs_sb_info *sbi, struct bio *bio,
|
| 786 | block_t last_blkaddr, block_t cur_blkaddr)
|
| 787 | {
|
| 788 | if (unlikely(sbi->max_io_bytes &&
|
| 789 | bio->bi_iter.bi_size >= sbi->max_io_bytes))
|
| 790 | return false;
|
| 791 | if (last_blkaddr + 1 != cur_blkaddr)
|
| 792 | return false;
|
| 793 | return bio->bi_bdev == f2fs_target_device(sbi, cur_blkaddr, NULL);
|
| 794 | }
|
| 795 |
|
| 796 | static bool io_type_is_mergeable(struct f2fs_bio_info *io,
|
| 797 | struct f2fs_io_info *fio)
|
| 798 | {
|
| 799 | blk_opf_t mask = ~(REQ_PREFLUSH | REQ_FUA);
|
| 800 |
|
| 801 | if (io->fio.op != fio->op)
|
| 802 | return false;
|
| 803 | return (io->fio.op_flags & mask) == (fio->op_flags & mask);
|
| 804 | }
|
| 805 |
|
| 806 | static bool io_is_mergeable(struct f2fs_sb_info *sbi, struct bio *bio,
|
| 807 | struct f2fs_bio_info *io,
|
| 808 | struct f2fs_io_info *fio,
|
| 809 | block_t last_blkaddr,
|
| 810 | block_t cur_blkaddr)
|
| 811 | {
|
| 812 | if (!page_is_mergeable(sbi, bio, last_blkaddr, cur_blkaddr))
|
| 813 | return false;
|
| 814 | return io_type_is_mergeable(io, fio);
|
| 815 | }
|
| 816 |
|
| 817 | static void add_bio_entry(struct f2fs_sb_info *sbi, struct bio *bio,
|
| 818 | struct folio *folio, enum temp_type temp)
|
| 819 | {
|
| 820 | struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
|
| 821 | struct bio_entry *be;
|
| 822 |
|
| 823 | be = f2fs_kmem_cache_alloc(bio_entry_slab, GFP_NOFS, true, NULL);
|
| 824 | be->bio = bio;
|
| 825 | bio_get(bio);
|
| 826 |
|
| 827 | bio_add_folio_nofail(bio, folio, folio_size(folio), 0);
|
| 828 |
|
| 829 | f2fs_down_write(&io->bio_list_lock);
|
| 830 | list_add_tail(&be->list, &io->bio_list);
|
| 831 | f2fs_up_write(&io->bio_list_lock);
|
| 832 | }
|
| 833 |
|
| 834 | static void del_bio_entry(struct bio_entry *be)
|
| 835 | {
|
| 836 | list_del(&be->list);
|
| 837 | kmem_cache_free(bio_entry_slab, be);
|
| 838 | }
|
| 839 |
|
| 840 | static int add_ipu_page(struct f2fs_io_info *fio, struct bio **bio,
|
| 841 | struct folio *folio)
|
| 842 | {
|
| 843 | struct folio *fio_folio = fio->folio;
|
| 844 | struct f2fs_sb_info *sbi = fio->sbi;
|
| 845 | enum temp_type temp;
|
| 846 | bool found = false;
|
| 847 | int ret = -EAGAIN;
|
| 848 |
|
| 849 | for (temp = HOT; temp < NR_TEMP_TYPE && !found; temp++) {
|
| 850 | struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
|
| 851 | struct list_head *head = &io->bio_list;
|
| 852 | struct bio_entry *be;
|
| 853 |
|
| 854 | f2fs_down_write(&io->bio_list_lock);
|
| 855 | list_for_each_entry(be, head, list) {
|
| 856 | if (be->bio != *bio)
|
| 857 | continue;
|
| 858 |
|
| 859 | found = true;
|
| 860 |
|
| 861 | f2fs_bug_on(sbi, !page_is_mergeable(sbi, *bio,
|
| 862 | *fio->last_block,
|
| 863 | fio->new_blkaddr));
|
| 864 | if (f2fs_crypt_mergeable_bio(*bio,
|
| 865 | fio_folio->mapping->host,
|
| 866 | fio_folio->index, fio) &&
|
| 867 | bio_add_folio(*bio, folio, folio_size(folio), 0)) {
|
| 868 | ret = 0;
|
| 869 | break;
|
| 870 | }
|
| 871 |
|
| 872 |
|
| 873 | del_bio_entry(be);
|
| 874 | f2fs_submit_write_bio(sbi, *bio, DATA);
|
| 875 | break;
|
| 876 | }
|
| 877 | f2fs_up_write(&io->bio_list_lock);
|
| 878 | }
|
| 879 |
|
| 880 | if (ret) {
|
| 881 | bio_put(*bio);
|
| 882 | *bio = NULL;
|
| 883 | }
|
| 884 |
|
| 885 | return ret;
|
| 886 | }
|
| 887 |
|
| 888 | void f2fs_submit_merged_ipu_write(struct f2fs_sb_info *sbi,
|
| 889 | struct bio **bio, struct folio *folio)
|
| 890 | {
|
| 891 | enum temp_type temp;
|
| 892 | bool found = false;
|
| 893 | struct bio *target = bio ? *bio : NULL;
|
| 894 |
|
| 895 | f2fs_bug_on(sbi, !target && !folio);
|
| 896 |
|
| 897 | for (temp = HOT; temp < NR_TEMP_TYPE && !found; temp++) {
|
| 898 | struct f2fs_bio_info *io = sbi->write_io[DATA] + temp;
|
| 899 | struct list_head *head = &io->bio_list;
|
| 900 | struct bio_entry *be;
|
| 901 |
|
| 902 | if (list_empty(head))
|
| 903 | continue;
|
| 904 |
|
| 905 | f2fs_down_read(&io->bio_list_lock);
|
| 906 | list_for_each_entry(be, head, list) {
|
| 907 | if (target)
|
| 908 | found = (target == be->bio);
|
| 909 | else
|
| 910 | found = __has_merged_page(be->bio, NULL,
|
| 911 | folio, 0);
|
| 912 | if (found)
|
| 913 | break;
|
| 914 | }
|
| 915 | f2fs_up_read(&io->bio_list_lock);
|
| 916 |
|
| 917 | if (!found)
|
| 918 | continue;
|
| 919 |
|
| 920 | found = false;
|
| 921 |
|
| 922 | f2fs_down_write(&io->bio_list_lock);
|
| 923 | list_for_each_entry(be, head, list) {
|
| 924 | if (target)
|
| 925 | found = (target == be->bio);
|
| 926 | else
|
| 927 | found = __has_merged_page(be->bio, NULL,
|
| 928 | folio, 0);
|
| 929 | if (found) {
|
| 930 | target = be->bio;
|
| 931 | del_bio_entry(be);
|
| 932 | break;
|
| 933 | }
|
| 934 | }
|
| 935 | f2fs_up_write(&io->bio_list_lock);
|
| 936 | }
|
| 937 |
|
| 938 | if (found)
|
| 939 | f2fs_submit_write_bio(sbi, target, DATA);
|
| 940 | if (bio && *bio) {
|
| 941 | bio_put(*bio);
|
| 942 | *bio = NULL;
|
| 943 | }
|
| 944 | }
|
| 945 |
|
| 946 | void f2fs_submit_all_merged_ipu_writes(struct f2fs_sb_info *sbi)
|
| 947 | {
|
| 948 | struct bio_entry *be, *tmp;
|
| 949 | struct f2fs_bio_info *io;
|
| 950 | enum temp_type temp;
|
| 951 |
|
| 952 | for (temp = HOT; temp < NR_TEMP_TYPE; temp++) {
|
| 953 | LIST_HEAD(list);
|
| 954 |
|
| 955 | io = sbi->write_io[DATA] + temp;
|
| 956 |
|
| 957 | |
| 958 | * other kworkers that we miss will be submitted by those
|
| 959 | * kworkers accordingly.
|
| 960 |
|
| 961 | if (list_empty(&io->bio_list))
|
| 962 | continue;
|
| 963 |
|
| 964 | f2fs_down_write(&io->bio_list_lock);
|
| 965 | list_splice_init(&io->bio_list, &list);
|
| 966 | f2fs_up_write(&io->bio_list_lock);
|
| 967 |
|
| 968 | list_for_each_entry_safe(be, tmp, &list, list) {
|
| 969 | f2fs_submit_write_bio(sbi, be->bio, DATA);
|
| 970 | del_bio_entry(be);
|
| 971 | }
|
| 972 | }
|
| 973 | }
|
| 974 |
|
| 975 | int f2fs_merge_page_bio(struct f2fs_io_info *fio)
|
| 976 | {
|
| 977 | struct bio *bio = *fio->bio;
|
| 978 | struct folio *data_folio = fio->encrypted_page ?
|
| 979 | page_folio(fio->encrypted_page) : fio->folio;
|
| 980 | struct folio *folio = fio->folio;
|
| 981 |
|
| 982 | if (!f2fs_is_valid_blkaddr(fio->sbi, fio->new_blkaddr,
|
| 983 | __is_meta_io(fio) ? META_GENERIC : DATA_GENERIC))
|
| 984 | return -EFSCORRUPTED;
|
| 985 |
|
| 986 | trace_f2fs_submit_folio_bio(data_folio, fio);
|
| 987 |
|
| 988 | if (bio && !page_is_mergeable(fio->sbi, bio, *fio->last_block,
|
| 989 | fio->new_blkaddr))
|
| 990 | f2fs_submit_merged_ipu_write(fio->sbi, &bio, NULL);
|
| 991 | alloc_new:
|
| 992 | if (!bio) {
|
| 993 | bio = __bio_alloc(fio, BIO_MAX_VECS);
|
| 994 | f2fs_set_bio_crypt_ctx(bio, folio->mapping->host,
|
| 995 | folio->index, fio, GFP_NOIO);
|
| 996 |
|
| 997 | add_bio_entry(fio->sbi, bio, data_folio, fio->temp);
|
| 998 | } else {
|
| 999 | if (add_ipu_page(fio, &bio, data_folio))
|
| 1000 | goto alloc_new;
|
| 1001 | }
|
| 1002 |
|
| 1003 | if (fio->io_wbc)
|
| 1004 | wbc_account_cgroup_owner(fio->io_wbc, folio, folio_size(folio));
|
| 1005 |
|
| 1006 | inc_page_count(fio->sbi, WB_DATA_TYPE(folio, false));
|
| 1007 |
|
| 1008 | *fio->last_block = fio->new_blkaddr;
|
| 1009 | *fio->bio = bio;
|
| 1010 |
|
| 1011 | return 0;
|
| 1012 | }
|
| 1013 |
|
| 1014 | #ifdef CONFIG_BLK_DEV_ZONED
|
| 1015 | static bool is_end_zone_blkaddr(struct f2fs_sb_info *sbi, block_t blkaddr)
|
| 1016 | {
|
| 1017 | struct block_device *bdev = sbi->sb->s_bdev;
|
| 1018 | int devi = 0;
|
| 1019 |
|
| 1020 | if (f2fs_is_multi_device(sbi)) {
|
| 1021 | devi = f2fs_target_device_index(sbi, blkaddr);
|
| 1022 | if (blkaddr < FDEV(devi).start_blk ||
|
| 1023 | blkaddr > FDEV(devi).end_blk) {
|
| 1024 | f2fs_err(sbi, "Invalid block %x", blkaddr);
|
| 1025 | return false;
|
| 1026 | }
|
| 1027 | blkaddr -= FDEV(devi).start_blk;
|
| 1028 | bdev = FDEV(devi).bdev;
|
| 1029 | }
|
| 1030 | return bdev_is_zoned(bdev) &&
|
| 1031 | f2fs_blkz_is_seq(sbi, devi, blkaddr) &&
|
| 1032 | (blkaddr % sbi->blocks_per_blkz == sbi->blocks_per_blkz - 1);
|
| 1033 | }
|
| 1034 | #endif
|
| 1035 |
|
| 1036 | void f2fs_submit_page_write(struct f2fs_io_info *fio)
|
| 1037 | {
|
| 1038 | struct f2fs_sb_info *sbi = fio->sbi;
|
| 1039 | enum page_type btype = PAGE_TYPE_OF_BIO(fio->type);
|
| 1040 | struct f2fs_bio_info *io = sbi->write_io[btype] + fio->temp;
|
| 1041 | struct folio *bio_folio;
|
| 1042 | struct f2fs_lock_context lc;
|
| 1043 | enum count_type type;
|
| 1044 |
|
| 1045 | f2fs_bug_on(sbi, is_read_io(fio->op));
|
| 1046 |
|
| 1047 | f2fs_down_write_trace(&io->io_rwsem, &lc);
|
| 1048 | next:
|
| 1049 | #ifdef CONFIG_BLK_DEV_ZONED
|
| 1050 | if (f2fs_sb_has_blkzoned(sbi) && btype < META && io->zone_pending_bio) {
|
| 1051 | wait_for_completion_io(&io->zone_wait);
|
| 1052 | bio_put(io->zone_pending_bio);
|
| 1053 | io->zone_pending_bio = NULL;
|
| 1054 | io->bi_private = NULL;
|
| 1055 | }
|
| 1056 | #endif
|
| 1057 |
|
| 1058 | if (fio->in_list) {
|
| 1059 | spin_lock(&io->io_lock);
|
| 1060 | if (list_empty(&io->io_list)) {
|
| 1061 | spin_unlock(&io->io_lock);
|
| 1062 | goto out;
|
| 1063 | }
|
| 1064 | fio = list_first_entry(&io->io_list,
|
| 1065 | struct f2fs_io_info, list);
|
| 1066 | list_del(&fio->list);
|
| 1067 | spin_unlock(&io->io_lock);
|
| 1068 | }
|
| 1069 |
|
| 1070 | verify_fio_blkaddr(fio);
|
| 1071 |
|
| 1072 | if (fio->encrypted_page)
|
| 1073 | bio_folio = page_folio(fio->encrypted_page);
|
| 1074 | else if (fio->compressed_page)
|
| 1075 | bio_folio = page_folio(fio->compressed_page);
|
| 1076 | else
|
| 1077 | bio_folio = fio->folio;
|
| 1078 |
|
| 1079 |
|
| 1080 | fio->submitted = 1;
|
| 1081 |
|
| 1082 | type = WB_DATA_TYPE(bio_folio, fio->compressed_page);
|
| 1083 | inc_page_count(sbi, type);
|
| 1084 |
|
| 1085 | if (io->bio &&
|
| 1086 | (!io_is_mergeable(sbi, io->bio, io, fio, io->last_block_in_bio,
|
| 1087 | fio->new_blkaddr) ||
|
| 1088 | !f2fs_crypt_mergeable_bio(io->bio, fio_inode(fio),
|
| 1089 | bio_folio->index, fio)))
|
| 1090 | __submit_merged_bio(io);
|
| 1091 | alloc_new:
|
| 1092 | if (io->bio == NULL) {
|
| 1093 | io->bio = __bio_alloc(fio, BIO_MAX_VECS);
|
| 1094 | f2fs_set_bio_crypt_ctx(io->bio, fio_inode(fio),
|
| 1095 | bio_folio->index, fio, GFP_NOIO);
|
| 1096 | io->fio = *fio;
|
| 1097 | }
|
| 1098 |
|
| 1099 | if (!bio_add_folio(io->bio, bio_folio, folio_size(bio_folio), 0)) {
|
| 1100 | __submit_merged_bio(io);
|
| 1101 | goto alloc_new;
|
| 1102 | }
|
| 1103 |
|
| 1104 | if (fio->io_wbc)
|
| 1105 | wbc_account_cgroup_owner(fio->io_wbc, fio->folio,
|
| 1106 | folio_size(fio->folio));
|
| 1107 |
|
| 1108 | io->last_block_in_bio = fio->new_blkaddr;
|
| 1109 |
|
| 1110 | trace_f2fs_submit_folio_write(fio->folio, fio);
|
| 1111 | #ifdef CONFIG_BLK_DEV_ZONED
|
| 1112 | if (f2fs_sb_has_blkzoned(sbi) && btype < META &&
|
| 1113 | is_end_zone_blkaddr(sbi, fio->new_blkaddr)) {
|
| 1114 | bio_get(io->bio);
|
| 1115 | reinit_completion(&io->zone_wait);
|
| 1116 | io->bi_private = io->bio->bi_private;
|
| 1117 | io->bio->bi_private = io;
|
| 1118 | io->bio->bi_end_io = f2fs_zone_write_end_io;
|
| 1119 | io->zone_pending_bio = io->bio;
|
| 1120 | __submit_merged_bio(io);
|
| 1121 | }
|
| 1122 | #endif
|
| 1123 | if (fio->in_list)
|
| 1124 | goto next;
|
| 1125 | out:
|
| 1126 | if (is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN) ||
|
| 1127 | !f2fs_is_checkpoint_ready(sbi))
|
| 1128 | __submit_merged_bio(io);
|
| 1129 | f2fs_up_write_trace(&io->io_rwsem, &lc);
|
| 1130 | }
|
| 1131 |
|
| 1132 | static struct bio *f2fs_grab_read_bio(struct inode *inode,
|
| 1133 | struct fsverity_info *vi, block_t blkaddr,
|
| 1134 | unsigned nr_pages, blk_opf_t op_flag,
|
| 1135 | pgoff_t first_idx, bool for_write)
|
| 1136 | {
|
| 1137 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
| 1138 | struct bio *bio;
|
| 1139 | struct bio_post_read_ctx *ctx = NULL;
|
| 1140 | unsigned int post_read_steps = 0;
|
| 1141 | sector_t sector;
|
| 1142 | struct block_device *bdev = f2fs_target_device(sbi, blkaddr, §or);
|
| 1143 |
|
| 1144 | bio = bio_alloc_bioset(bdev, bio_max_segs(nr_pages),
|
| 1145 | REQ_OP_READ | op_flag,
|
| 1146 | for_write ? GFP_NOIO : GFP_KERNEL, &f2fs_bioset);
|
| 1147 | bio->bi_iter.bi_sector = sector;
|
| 1148 | f2fs_set_bio_crypt_ctx(bio, inode, first_idx, NULL, GFP_NOFS);
|
| 1149 | bio->bi_end_io = f2fs_read_end_io;
|
| 1150 |
|
| 1151 | if (vi)
|
| 1152 | post_read_steps |= STEP_VERITY;
|
| 1153 |
|
| 1154 | |
| 1155 | * STEP_DECOMPRESS is handled specially, since a compressed file might
|
| 1156 | * contain both compressed and uncompressed clusters. We'll allocate a
|
| 1157 | * bio_post_read_ctx if the file is compressed, but the caller is
|
| 1158 | * responsible for enabling STEP_DECOMPRESS if it's actually needed.
|
| 1159 |
|
| 1160 |
|
| 1161 | if (post_read_steps || f2fs_compressed_file(inode)) {
|
| 1162 |
|
| 1163 | ctx = mempool_alloc(bio_post_read_ctx_pool, GFP_NOFS);
|
| 1164 | ctx->bio = bio;
|
| 1165 | ctx->sbi = sbi;
|
| 1166 | ctx->vi = vi;
|
| 1167 | ctx->enabled_steps = post_read_steps;
|
| 1168 | ctx->fs_blkaddr = blkaddr;
|
| 1169 | ctx->decompression_attempted = false;
|
| 1170 | bio->bi_private = ctx;
|
| 1171 | }
|
| 1172 | iostat_alloc_and_bind_ctx(sbi, bio, ctx);
|
| 1173 |
|
| 1174 | return bio;
|
| 1175 | }
|
| 1176 |
|
| 1177 |
|
| 1178 | static void f2fs_submit_page_read(struct inode *inode, struct fsverity_info *vi,
|
| 1179 | struct folio *folio, block_t blkaddr,
|
| 1180 | blk_opf_t op_flags, bool for_write)
|
| 1181 | {
|
| 1182 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
| 1183 | struct bio *bio;
|
| 1184 |
|
| 1185 | bio = f2fs_grab_read_bio(inode, vi, blkaddr, 1, op_flags, folio->index,
|
| 1186 | for_write);
|
| 1187 |
|
| 1188 |
|
| 1189 | f2fs_wait_on_block_writeback(inode, blkaddr);
|
| 1190 |
|
| 1191 | if (!bio_add_folio(bio, folio, PAGE_SIZE, 0))
|
| 1192 | f2fs_bug_on(sbi, 1);
|
| 1193 |
|
| 1194 | inc_page_count(sbi, F2FS_RD_DATA);
|
| 1195 | f2fs_update_iostat(sbi, NULL, FS_DATA_READ_IO, F2FS_BLKSIZE);
|
| 1196 | f2fs_submit_read_bio(sbi, bio, DATA);
|
| 1197 | }
|
| 1198 |
|
| 1199 | static void __set_data_blkaddr(struct dnode_of_data *dn, block_t blkaddr)
|
| 1200 | {
|
| 1201 | __le32 *addr = get_dnode_addr(dn->inode, dn->node_folio);
|
| 1202 |
|
| 1203 | dn->data_blkaddr = blkaddr;
|
| 1204 | addr[dn->ofs_in_node] = cpu_to_le32(dn->data_blkaddr);
|
| 1205 | }
|
| 1206 |
|
| 1207 | |
| 1208 | * Lock ordering for the change of data block address:
|
| 1209 | * ->data_page
|
| 1210 | * ->node_folio
|
| 1211 | * update block addresses in the node page
|
| 1212 |
|
| 1213 | void f2fs_set_data_blkaddr(struct dnode_of_data *dn, block_t blkaddr)
|
| 1214 | {
|
| 1215 | f2fs_folio_wait_writeback(dn->node_folio, NODE, true, true);
|
| 1216 | __set_data_blkaddr(dn, blkaddr);
|
| 1217 | if (folio_mark_dirty(dn->node_folio))
|
| 1218 | dn->node_changed = true;
|
| 1219 | }
|
| 1220 |
|
| 1221 | void f2fs_update_data_blkaddr(struct dnode_of_data *dn, block_t blkaddr)
|
| 1222 | {
|
| 1223 | f2fs_set_data_blkaddr(dn, blkaddr);
|
| 1224 | f2fs_update_read_extent_cache(dn);
|
| 1225 | }
|
| 1226 |
|
| 1227 |
|
| 1228 | int f2fs_reserve_new_blocks(struct dnode_of_data *dn, blkcnt_t count)
|
| 1229 | {
|
| 1230 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
|
| 1231 | int err;
|
| 1232 |
|
| 1233 | if (!count)
|
| 1234 | return 0;
|
| 1235 |
|
| 1236 | if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
|
| 1237 | return -EPERM;
|
| 1238 | err = inc_valid_block_count(sbi, dn->inode, &count, true, false);
|
| 1239 | if (unlikely(err))
|
| 1240 | return err;
|
| 1241 |
|
| 1242 | trace_f2fs_reserve_new_blocks(dn->inode, dn->nid,
|
| 1243 | dn->ofs_in_node, count);
|
| 1244 |
|
| 1245 | f2fs_folio_wait_writeback(dn->node_folio, NODE, true, true);
|
| 1246 |
|
| 1247 | for (; count > 0; dn->ofs_in_node++) {
|
| 1248 | block_t blkaddr = f2fs_data_blkaddr(dn);
|
| 1249 |
|
| 1250 | if (blkaddr == NULL_ADDR) {
|
| 1251 | __set_data_blkaddr(dn, NEW_ADDR);
|
| 1252 | count--;
|
| 1253 | }
|
| 1254 | }
|
| 1255 |
|
| 1256 | if (folio_mark_dirty(dn->node_folio))
|
| 1257 | dn->node_changed = true;
|
| 1258 | return 0;
|
| 1259 | }
|
| 1260 |
|
| 1261 |
|
| 1262 | int f2fs_reserve_new_block(struct dnode_of_data *dn)
|
| 1263 | {
|
| 1264 | unsigned int ofs_in_node = dn->ofs_in_node;
|
| 1265 | int ret;
|
| 1266 |
|
| 1267 | ret = f2fs_reserve_new_blocks(dn, 1);
|
| 1268 | dn->ofs_in_node = ofs_in_node;
|
| 1269 | return ret;
|
| 1270 | }
|
| 1271 |
|
| 1272 | int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index)
|
| 1273 | {
|
| 1274 | bool need_put = dn->inode_folio ? false : true;
|
| 1275 | int err;
|
| 1276 |
|
| 1277 | err = f2fs_get_dnode_of_data(dn, index, ALLOC_NODE);
|
| 1278 | if (err)
|
| 1279 | return err;
|
| 1280 |
|
| 1281 | if (dn->data_blkaddr == NULL_ADDR)
|
| 1282 | err = f2fs_reserve_new_block(dn);
|
| 1283 | if (err || need_put)
|
| 1284 | f2fs_put_dnode(dn);
|
| 1285 | return err;
|
| 1286 | }
|
| 1287 |
|
| 1288 | static inline struct fsverity_info *f2fs_need_verity(const struct inode *inode,
|
| 1289 | pgoff_t idx)
|
| 1290 | {
|
| 1291 | if (idx < DIV_ROUND_UP(inode->i_size, PAGE_SIZE))
|
| 1292 | return fsverity_get_info(inode);
|
| 1293 | return NULL;
|
| 1294 | }
|
| 1295 |
|
| 1296 | struct folio *f2fs_get_read_data_folio(struct inode *inode, pgoff_t index,
|
| 1297 | blk_opf_t op_flags, bool for_write, pgoff_t *next_pgofs)
|
| 1298 | {
|
| 1299 | struct address_space *mapping = inode->i_mapping;
|
| 1300 | struct dnode_of_data dn;
|
| 1301 | struct folio *folio;
|
| 1302 | int err;
|
| 1303 | retry:
|
| 1304 | folio = f2fs_grab_cache_folio(mapping, index, for_write);
|
| 1305 | if (IS_ERR(folio))
|
| 1306 | return folio;
|
| 1307 |
|
| 1308 | if (folio_test_large(folio)) {
|
| 1309 | pgoff_t folio_index = mapping_align_index(mapping, index);
|
| 1310 | unsigned long nr_pages = folio_nr_pages(folio);
|
| 1311 |
|
| 1312 | f2fs_folio_put(folio, true);
|
| 1313 | invalidate_inode_pages2_range(mapping, folio_index,
|
| 1314 | folio_index + nr_pages - 1);
|
| 1315 | f2fs_schedule_timeout(DEFAULT_SCHEDULE_TIMEOUT);
|
| 1316 | goto retry;
|
| 1317 | }
|
| 1318 |
|
| 1319 | if (f2fs_lookup_read_extent_cache_block(inode, index,
|
| 1320 | &dn.data_blkaddr)) {
|
| 1321 | if (!f2fs_is_valid_blkaddr(F2FS_I_SB(inode), dn.data_blkaddr,
|
| 1322 | DATA_GENERIC_ENHANCE_READ)) {
|
| 1323 | err = -EFSCORRUPTED;
|
| 1324 | goto put_err;
|
| 1325 | }
|
| 1326 | goto got_it;
|
| 1327 | }
|
| 1328 |
|
| 1329 | set_new_dnode(&dn, inode, NULL, NULL, 0);
|
| 1330 | err = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
|
| 1331 | if (err) {
|
| 1332 | if (err == -ENOENT && next_pgofs)
|
| 1333 | *next_pgofs = f2fs_get_next_page_offset(&dn, index);
|
| 1334 | goto put_err;
|
| 1335 | }
|
| 1336 | f2fs_put_dnode(&dn);
|
| 1337 |
|
| 1338 | if (unlikely(dn.data_blkaddr == NULL_ADDR)) {
|
| 1339 | err = -ENOENT;
|
| 1340 | if (next_pgofs)
|
| 1341 | *next_pgofs = index + 1;
|
| 1342 | goto put_err;
|
| 1343 | }
|
| 1344 | if (dn.data_blkaddr != NEW_ADDR &&
|
| 1345 | !f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
|
| 1346 | dn.data_blkaddr,
|
| 1347 | DATA_GENERIC_ENHANCE)) {
|
| 1348 | err = -EFSCORRUPTED;
|
| 1349 | goto put_err;
|
| 1350 | }
|
| 1351 | got_it:
|
| 1352 | if (folio_test_uptodate(folio)) {
|
| 1353 | folio_unlock(folio);
|
| 1354 | return folio;
|
| 1355 | }
|
| 1356 |
|
| 1357 | |
| 1358 | * A new dentry page is allocated but not able to be written, since its
|
| 1359 | * new inode page couldn't be allocated due to -ENOSPC.
|
| 1360 | * In such the case, its blkaddr can be remained as NEW_ADDR.
|
| 1361 | * see, f2fs_add_link -> f2fs_get_new_data_folio ->
|
| 1362 | * f2fs_init_inode_metadata.
|
| 1363 |
|
| 1364 | if (dn.data_blkaddr == NEW_ADDR) {
|
| 1365 | folio_zero_segment(folio, 0, folio_size(folio));
|
| 1366 | if (!folio_test_uptodate(folio))
|
| 1367 | folio_mark_uptodate(folio);
|
| 1368 | folio_unlock(folio);
|
| 1369 | return folio;
|
| 1370 | }
|
| 1371 |
|
| 1372 | f2fs_submit_page_read(inode, f2fs_need_verity(inode, folio->index),
|
| 1373 | folio, dn.data_blkaddr, op_flags, for_write);
|
| 1374 | return folio;
|
| 1375 |
|
| 1376 | put_err:
|
| 1377 | f2fs_folio_put(folio, true);
|
| 1378 | return ERR_PTR(err);
|
| 1379 | }
|
| 1380 |
|
| 1381 | struct folio *f2fs_find_data_folio(struct inode *inode, pgoff_t index,
|
| 1382 | pgoff_t *next_pgofs)
|
| 1383 | {
|
| 1384 | struct address_space *mapping = inode->i_mapping;
|
| 1385 | struct folio *folio;
|
| 1386 |
|
| 1387 | folio = f2fs_filemap_get_folio(mapping, index, FGP_ACCESSED, 0);
|
| 1388 | if (IS_ERR(folio))
|
| 1389 | goto read;
|
| 1390 | if (folio_test_uptodate(folio))
|
| 1391 | return folio;
|
| 1392 | f2fs_folio_put(folio, false);
|
| 1393 |
|
| 1394 | read:
|
| 1395 | folio = f2fs_get_read_data_folio(inode, index, 0, false, next_pgofs);
|
| 1396 | if (IS_ERR(folio))
|
| 1397 | return folio;
|
| 1398 |
|
| 1399 | if (folio_test_uptodate(folio))
|
| 1400 | return folio;
|
| 1401 |
|
| 1402 | folio_wait_locked(folio);
|
| 1403 | if (unlikely(!folio_test_uptodate(folio))) {
|
| 1404 | f2fs_folio_put(folio, false);
|
| 1405 | return ERR_PTR(-EIO);
|
| 1406 | }
|
| 1407 | return folio;
|
| 1408 | }
|
| 1409 |
|
| 1410 | |
| 1411 | * If it tries to access a hole, return an error.
|
| 1412 | * Because, the callers, functions in dir.c and GC, should be able to know
|
| 1413 | * whether this page exists or not.
|
| 1414 |
|
| 1415 | struct folio *f2fs_get_lock_data_folio(struct inode *inode, pgoff_t index,
|
| 1416 | bool for_write)
|
| 1417 | {
|
| 1418 | struct address_space *mapping = inode->i_mapping;
|
| 1419 | struct folio *folio;
|
| 1420 |
|
| 1421 | folio = f2fs_get_read_data_folio(inode, index, 0, for_write, NULL);
|
| 1422 | if (IS_ERR(folio))
|
| 1423 | return folio;
|
| 1424 |
|
| 1425 |
|
| 1426 | folio_lock(folio);
|
| 1427 | if (unlikely(folio->mapping != mapping || !folio_test_uptodate(folio))) {
|
| 1428 | f2fs_folio_put(folio, true);
|
| 1429 | return ERR_PTR(-EIO);
|
| 1430 | }
|
| 1431 | return folio;
|
| 1432 | }
|
| 1433 |
|
| 1434 | |
| 1435 | * Caller ensures that this data page is never allocated.
|
| 1436 | * A new zero-filled data page is allocated in the page cache.
|
| 1437 | *
|
| 1438 | * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and
|
| 1439 | * f2fs_unlock_op().
|
| 1440 | * Note that, ifolio is set only by make_empty_dir, and if any error occur,
|
| 1441 | * ifolio should be released by this function.
|
| 1442 |
|
| 1443 | struct folio *f2fs_get_new_data_folio(struct inode *inode,
|
| 1444 | struct folio *ifolio, pgoff_t index, bool new_i_size)
|
| 1445 | {
|
| 1446 | struct address_space *mapping = inode->i_mapping;
|
| 1447 | struct folio *folio;
|
| 1448 | struct dnode_of_data dn;
|
| 1449 | int err;
|
| 1450 |
|
| 1451 | folio = f2fs_grab_cache_folio(mapping, index, true);
|
| 1452 | if (IS_ERR(folio)) {
|
| 1453 | |
| 1454 | * before exiting, we should make sure ifolio will be released
|
| 1455 | * if any error occur.
|
| 1456 |
|
| 1457 | f2fs_folio_put(ifolio, true);
|
| 1458 | return ERR_PTR(-ENOMEM);
|
| 1459 | }
|
| 1460 |
|
| 1461 | set_new_dnode(&dn, inode, ifolio, NULL, 0);
|
| 1462 | err = f2fs_reserve_block(&dn, index);
|
| 1463 | if (err) {
|
| 1464 | f2fs_folio_put(folio, true);
|
| 1465 | return ERR_PTR(err);
|
| 1466 | }
|
| 1467 | if (!ifolio)
|
| 1468 | f2fs_put_dnode(&dn);
|
| 1469 |
|
| 1470 | if (folio_test_uptodate(folio))
|
| 1471 | goto got_it;
|
| 1472 |
|
| 1473 | if (dn.data_blkaddr == NEW_ADDR) {
|
| 1474 | folio_zero_segment(folio, 0, folio_size(folio));
|
| 1475 | if (!folio_test_uptodate(folio))
|
| 1476 | folio_mark_uptodate(folio);
|
| 1477 | } else {
|
| 1478 | f2fs_folio_put(folio, true);
|
| 1479 |
|
| 1480 |
|
| 1481 | f2fs_bug_on(F2FS_I_SB(inode), ifolio);
|
| 1482 | folio = f2fs_get_lock_data_folio(inode, index, true);
|
| 1483 | if (IS_ERR(folio))
|
| 1484 | return folio;
|
| 1485 | }
|
| 1486 | got_it:
|
| 1487 | if (new_i_size && i_size_read(inode) <
|
| 1488 | ((loff_t)(index + 1) << PAGE_SHIFT))
|
| 1489 | f2fs_i_size_write(inode, ((loff_t)(index + 1) << PAGE_SHIFT));
|
| 1490 | return folio;
|
| 1491 | }
|
| 1492 |
|
| 1493 | static int __allocate_data_block(struct dnode_of_data *dn, int seg_type)
|
| 1494 | {
|
| 1495 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
|
| 1496 | struct f2fs_summary sum;
|
| 1497 | struct node_info ni;
|
| 1498 | block_t old_blkaddr;
|
| 1499 | blkcnt_t count = 1;
|
| 1500 | int err;
|
| 1501 |
|
| 1502 | if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
|
| 1503 | return -EPERM;
|
| 1504 |
|
| 1505 | err = f2fs_get_node_info(sbi, dn->nid, &ni, false);
|
| 1506 | if (err)
|
| 1507 | return err;
|
| 1508 |
|
| 1509 | dn->data_blkaddr = f2fs_data_blkaddr(dn);
|
| 1510 | if (dn->data_blkaddr == NULL_ADDR) {
|
| 1511 | err = inc_valid_block_count(sbi, dn->inode, &count, true, false);
|
| 1512 | if (unlikely(err))
|
| 1513 | return err;
|
| 1514 | }
|
| 1515 |
|
| 1516 | set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);
|
| 1517 | old_blkaddr = dn->data_blkaddr;
|
| 1518 | err = f2fs_allocate_data_block(sbi, NULL, old_blkaddr,
|
| 1519 | &dn->data_blkaddr, &sum, seg_type, NULL);
|
| 1520 | if (err) {
|
| 1521 | if (old_blkaddr == NULL_ADDR)
|
| 1522 | dec_valid_block_count(sbi, dn->inode, count);
|
| 1523 | return err;
|
| 1524 | }
|
| 1525 |
|
| 1526 | if (GET_SEGNO(sbi, old_blkaddr) != NULL_SEGNO)
|
| 1527 | f2fs_invalidate_internal_cache(sbi, old_blkaddr, 1);
|
| 1528 |
|
| 1529 | f2fs_update_data_blkaddr(dn, dn->data_blkaddr);
|
| 1530 | return 0;
|
| 1531 | }
|
| 1532 |
|
| 1533 | static void f2fs_map_lock(struct f2fs_sb_info *sbi,
|
| 1534 | struct f2fs_lock_context *lc,
|
| 1535 | int flag)
|
| 1536 | {
|
| 1537 | if (flag == F2FS_GET_BLOCK_PRE_AIO)
|
| 1538 | f2fs_down_read_trace(&sbi->node_change, lc);
|
| 1539 | else
|
| 1540 | f2fs_lock_op(sbi, lc);
|
| 1541 | }
|
| 1542 |
|
| 1543 | static void f2fs_map_unlock(struct f2fs_sb_info *sbi,
|
| 1544 | struct f2fs_lock_context *lc,
|
| 1545 | int flag)
|
| 1546 | {
|
| 1547 | if (flag == F2FS_GET_BLOCK_PRE_AIO)
|
| 1548 | f2fs_up_read_trace(&sbi->node_change, lc);
|
| 1549 | else
|
| 1550 | f2fs_unlock_op(sbi, lc);
|
| 1551 | }
|
| 1552 |
|
| 1553 | int f2fs_get_block_locked(struct dnode_of_data *dn, pgoff_t index)
|
| 1554 | {
|
| 1555 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
|
| 1556 | struct f2fs_lock_context lc;
|
| 1557 | int err = 0;
|
| 1558 |
|
| 1559 | f2fs_map_lock(sbi, &lc, F2FS_GET_BLOCK_PRE_AIO);
|
| 1560 | if (!f2fs_lookup_read_extent_cache_block(dn->inode, index,
|
| 1561 | &dn->data_blkaddr))
|
| 1562 | err = f2fs_reserve_block(dn, index);
|
| 1563 | f2fs_map_unlock(sbi, &lc, F2FS_GET_BLOCK_PRE_AIO);
|
| 1564 |
|
| 1565 | return err;
|
| 1566 | }
|
| 1567 |
|
| 1568 | static int f2fs_map_no_dnode(struct inode *inode,
|
| 1569 | struct f2fs_map_blocks *map, struct dnode_of_data *dn,
|
| 1570 | pgoff_t pgoff)
|
| 1571 | {
|
| 1572 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
| 1573 |
|
| 1574 | |
| 1575 | * There is one exceptional case that read_node_page() may return
|
| 1576 | * -ENOENT due to filesystem has been shutdown or cp_error, return
|
| 1577 | * -EIO in that case.
|
| 1578 |
|
| 1579 | if (map->m_may_create &&
|
| 1580 | (is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN) || f2fs_cp_error(sbi)))
|
| 1581 | return -EIO;
|
| 1582 |
|
| 1583 | if (map->m_next_pgofs)
|
| 1584 | *map->m_next_pgofs = f2fs_get_next_page_offset(dn, pgoff);
|
| 1585 | if (map->m_next_extent)
|
| 1586 | *map->m_next_extent = f2fs_get_next_page_offset(dn, pgoff);
|
| 1587 | return 0;
|
| 1588 | }
|
| 1589 |
|
| 1590 | static bool f2fs_map_blocks_cached(struct inode *inode,
|
| 1591 | struct f2fs_map_blocks *map, int flag)
|
| 1592 | {
|
| 1593 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
| 1594 | unsigned int maxblocks = map->m_len;
|
| 1595 | pgoff_t pgoff = (pgoff_t)map->m_lblk;
|
| 1596 | struct extent_info ei = {};
|
| 1597 |
|
| 1598 | if (!f2fs_lookup_read_extent_cache(inode, pgoff, &ei))
|
| 1599 | return false;
|
| 1600 |
|
| 1601 | map->m_pblk = ei.blk + pgoff - ei.fofs;
|
| 1602 | map->m_len = min((pgoff_t)maxblocks, ei.fofs + ei.len - pgoff);
|
| 1603 | map->m_flags = F2FS_MAP_MAPPED;
|
| 1604 | if (map->m_next_extent)
|
| 1605 | *map->m_next_extent = pgoff + map->m_len;
|
| 1606 |
|
| 1607 |
|
| 1608 | if (flag == F2FS_GET_BLOCK_DIO)
|
| 1609 | f2fs_wait_on_block_writeback_range(inode,
|
| 1610 | map->m_pblk, map->m_len);
|
| 1611 |
|
| 1612 | map->m_multidev_dio = f2fs_allow_multi_device_dio(sbi, flag);
|
| 1613 | if (map->m_multidev_dio) {
|
| 1614 | int bidx = f2fs_target_device_index(sbi, map->m_pblk);
|
| 1615 | struct f2fs_dev_info *dev = &sbi->devs[bidx];
|
| 1616 |
|
| 1617 | map->m_bdev = dev->bdev;
|
| 1618 | map->m_len = min(map->m_len, dev->end_blk + 1 - map->m_pblk);
|
| 1619 | map->m_pblk -= dev->start_blk;
|
| 1620 | } else {
|
| 1621 | map->m_bdev = inode->i_sb->s_bdev;
|
| 1622 | }
|
| 1623 | return true;
|
| 1624 | }
|
| 1625 |
|
| 1626 | static bool map_is_mergeable(struct f2fs_sb_info *sbi,
|
| 1627 | struct f2fs_map_blocks *map,
|
| 1628 | block_t blkaddr, int flag, int bidx,
|
| 1629 | int ofs)
|
| 1630 | {
|
| 1631 | if (map->m_multidev_dio && map->m_bdev != FDEV(bidx).bdev)
|
| 1632 | return false;
|
| 1633 | if (map->m_pblk != NEW_ADDR && blkaddr == (map->m_pblk + ofs))
|
| 1634 | return true;
|
| 1635 | if (map->m_pblk == NEW_ADDR && blkaddr == NEW_ADDR)
|
| 1636 | return true;
|
| 1637 | if (flag == F2FS_GET_BLOCK_PRE_DIO)
|
| 1638 | return true;
|
| 1639 | if (flag == F2FS_GET_BLOCK_DIO &&
|
| 1640 | map->m_pblk == NULL_ADDR && blkaddr == NULL_ADDR)
|
| 1641 | return true;
|
| 1642 | return false;
|
| 1643 | }
|
| 1644 |
|
| 1645 | |
| 1646 | * f2fs_map_blocks() tries to find or build mapping relationship which
|
| 1647 | * maps continuous logical blocks to physical blocks, and return such
|
| 1648 | * info via f2fs_map_blocks structure.
|
| 1649 |
|
| 1650 | int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map, int flag)
|
| 1651 | {
|
| 1652 | unsigned int maxblocks = map->m_len;
|
| 1653 | struct dnode_of_data dn;
|
| 1654 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
| 1655 | struct f2fs_lock_context lc;
|
| 1656 | int mode = map->m_may_create ? ALLOC_NODE : LOOKUP_NODE;
|
| 1657 | pgoff_t pgofs, end_offset, end;
|
| 1658 | int err = 0, ofs = 1;
|
| 1659 | unsigned int ofs_in_node, last_ofs_in_node;
|
| 1660 | blkcnt_t prealloc;
|
| 1661 | block_t blkaddr;
|
| 1662 | unsigned int start_pgofs;
|
| 1663 | int bidx = 0;
|
| 1664 | bool is_hole;
|
| 1665 | bool lfs_dio_write;
|
| 1666 |
|
| 1667 | if (!maxblocks)
|
| 1668 | return 0;
|
| 1669 |
|
| 1670 | lfs_dio_write = (flag == F2FS_GET_BLOCK_DIO && f2fs_lfs_mode(sbi) &&
|
| 1671 | map->m_may_create);
|
| 1672 |
|
| 1673 | if (!map->m_may_create && f2fs_map_blocks_cached(inode, map, flag)) {
|
| 1674 | struct extent_info ei;
|
| 1675 |
|
| 1676 | |
| 1677 | * 1. If map->m_multidev_dio is true, map->m_pblk cannot be
|
| 1678 | * waitted by f2fs_wait_on_block_writeback_range() and are not
|
| 1679 | * mergeable.
|
| 1680 | * 2. If pgofs hits the read extent cache, it means the mapping
|
| 1681 | * is already cached in the extent cache, but it is not
|
| 1682 | * mergeable, and there is no need to query the mapping again
|
| 1683 | * via f2fs_get_dnode_of_data().
|
| 1684 |
|
| 1685 | pgofs = (pgoff_t)map->m_lblk + map->m_len;
|
| 1686 | if (map->m_len == maxblocks ||
|
| 1687 | map->m_multidev_dio ||
|
| 1688 | f2fs_lookup_read_extent_cache(inode, pgofs, &ei))
|
| 1689 | goto out;
|
| 1690 | ofs = map->m_len;
|
| 1691 | goto map_more;
|
| 1692 | }
|
| 1693 |
|
| 1694 | map->m_bdev = inode->i_sb->s_bdev;
|
| 1695 | map->m_multidev_dio =
|
| 1696 | f2fs_allow_multi_device_dio(F2FS_I_SB(inode), flag);
|
| 1697 |
|
| 1698 | map->m_len = 0;
|
| 1699 | map->m_flags = 0;
|
| 1700 |
|
| 1701 |
|
| 1702 | pgofs = (pgoff_t)map->m_lblk;
|
| 1703 | map_more:
|
| 1704 | end = (pgoff_t)map->m_lblk + maxblocks;
|
| 1705 |
|
| 1706 | if (flag == F2FS_GET_BLOCK_PRECACHE)
|
| 1707 | mode = LOOKUP_NODE_RA;
|
| 1708 |
|
| 1709 | next_dnode:
|
| 1710 | if (map->m_may_create) {
|
| 1711 | if (f2fs_lfs_mode(sbi))
|
| 1712 | f2fs_balance_fs(sbi, true);
|
| 1713 | f2fs_map_lock(sbi, &lc, flag);
|
| 1714 | }
|
| 1715 |
|
| 1716 |
|
| 1717 | set_new_dnode(&dn, inode, NULL, NULL, 0);
|
| 1718 | err = f2fs_get_dnode_of_data(&dn, pgofs, mode);
|
| 1719 | if (err) {
|
| 1720 | if (flag == F2FS_GET_BLOCK_BMAP)
|
| 1721 | map->m_pblk = 0;
|
| 1722 | if (err == -ENOENT)
|
| 1723 | err = f2fs_map_no_dnode(inode, map, &dn, pgofs);
|
| 1724 | goto unlock_out;
|
| 1725 | }
|
| 1726 |
|
| 1727 | start_pgofs = pgofs;
|
| 1728 | prealloc = 0;
|
| 1729 | last_ofs_in_node = ofs_in_node = dn.ofs_in_node;
|
| 1730 | end_offset = ADDRS_PER_PAGE(dn.node_folio, inode);
|
| 1731 |
|
| 1732 | next_block:
|
| 1733 | blkaddr = f2fs_data_blkaddr(&dn);
|
| 1734 | is_hole = !__is_valid_data_blkaddr(blkaddr);
|
| 1735 | if (!is_hole &&
|
| 1736 | !f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC_ENHANCE)) {
|
| 1737 | err = -EFSCORRUPTED;
|
| 1738 | goto sync_out;
|
| 1739 | }
|
| 1740 |
|
| 1741 |
|
| 1742 | if (map->m_may_create && (is_hole ||
|
| 1743 | (flag == F2FS_GET_BLOCK_DIO && f2fs_lfs_mode(sbi) &&
|
| 1744 | !f2fs_is_pinned_file(inode) && map->m_last_pblk != blkaddr))) {
|
| 1745 | if (unlikely(f2fs_cp_error(sbi))) {
|
| 1746 | err = -EIO;
|
| 1747 | goto sync_out;
|
| 1748 | }
|
| 1749 |
|
| 1750 | switch (flag) {
|
| 1751 | case F2FS_GET_BLOCK_PRE_AIO:
|
| 1752 | if (blkaddr == NULL_ADDR) {
|
| 1753 | prealloc++;
|
| 1754 | last_ofs_in_node = dn.ofs_in_node;
|
| 1755 | }
|
| 1756 | break;
|
| 1757 | case F2FS_GET_BLOCK_PRE_DIO:
|
| 1758 | case F2FS_GET_BLOCK_DIO:
|
| 1759 | err = __allocate_data_block(&dn, map->m_seg_type);
|
| 1760 | if (err)
|
| 1761 | goto sync_out;
|
| 1762 | if (flag == F2FS_GET_BLOCK_PRE_DIO)
|
| 1763 | file_need_truncate(inode);
|
| 1764 | set_inode_flag(inode, FI_APPEND_WRITE);
|
| 1765 | break;
|
| 1766 | default:
|
| 1767 | WARN_ON_ONCE(1);
|
| 1768 | err = -EIO;
|
| 1769 | goto sync_out;
|
| 1770 | }
|
| 1771 |
|
| 1772 | blkaddr = dn.data_blkaddr;
|
| 1773 | if (is_hole)
|
| 1774 | map->m_flags |= F2FS_MAP_NEW;
|
| 1775 | } else if (is_hole) {
|
| 1776 | if (f2fs_compressed_file(inode) &&
|
| 1777 | f2fs_sanity_check_cluster(&dn)) {
|
| 1778 | err = -EFSCORRUPTED;
|
| 1779 | f2fs_handle_error(sbi,
|
| 1780 | ERROR_CORRUPTED_CLUSTER);
|
| 1781 | fserror_report_file_metadata(inode, err, GFP_NOFS);
|
| 1782 | goto sync_out;
|
| 1783 | }
|
| 1784 |
|
| 1785 | switch (flag) {
|
| 1786 | case F2FS_GET_BLOCK_PRECACHE:
|
| 1787 | goto sync_out;
|
| 1788 | case F2FS_GET_BLOCK_BMAP:
|
| 1789 | map->m_pblk = 0;
|
| 1790 | goto sync_out;
|
| 1791 | case F2FS_GET_BLOCK_FIEMAP:
|
| 1792 | if (blkaddr == NULL_ADDR) {
|
| 1793 | if (map->m_next_pgofs)
|
| 1794 | *map->m_next_pgofs = pgofs + 1;
|
| 1795 | goto sync_out;
|
| 1796 | }
|
| 1797 | break;
|
| 1798 | case F2FS_GET_BLOCK_DIO:
|
| 1799 | if (map->m_next_pgofs)
|
| 1800 | *map->m_next_pgofs = pgofs + 1;
|
| 1801 | break;
|
| 1802 | default:
|
| 1803 |
|
| 1804 | if (map->m_next_pgofs)
|
| 1805 | *map->m_next_pgofs = pgofs + 1;
|
| 1806 | goto sync_out;
|
| 1807 | }
|
| 1808 | }
|
| 1809 |
|
| 1810 | if (flag == F2FS_GET_BLOCK_PRE_AIO)
|
| 1811 | goto skip;
|
| 1812 |
|
| 1813 | if (map->m_multidev_dio)
|
| 1814 | bidx = f2fs_target_device_index(sbi, blkaddr);
|
| 1815 |
|
| 1816 | if (map->m_len == 0) {
|
| 1817 |
|
| 1818 | if (blkaddr == NEW_ADDR)
|
| 1819 | map->m_flags |= F2FS_MAP_DELALLOC;
|
| 1820 |
|
| 1821 | if (!(flag == F2FS_GET_BLOCK_DIO && is_hole && !map->m_may_create))
|
| 1822 | map->m_flags |= F2FS_MAP_MAPPED;
|
| 1823 |
|
| 1824 | map->m_pblk = blkaddr;
|
| 1825 | map->m_len = 1;
|
| 1826 |
|
| 1827 | if (map->m_multidev_dio)
|
| 1828 | map->m_bdev = FDEV(bidx).bdev;
|
| 1829 |
|
| 1830 | if (lfs_dio_write)
|
| 1831 | map->m_last_pblk = NULL_ADDR;
|
| 1832 | } else if (map_is_mergeable(sbi, map, blkaddr, flag, bidx, ofs)) {
|
| 1833 | ofs++;
|
| 1834 | map->m_len++;
|
| 1835 | } else {
|
| 1836 | if (lfs_dio_write && !f2fs_is_pinned_file(inode))
|
| 1837 | map->m_last_pblk = blkaddr;
|
| 1838 | goto sync_out;
|
| 1839 | }
|
| 1840 |
|
| 1841 | skip:
|
| 1842 | dn.ofs_in_node++;
|
| 1843 | pgofs++;
|
| 1844 |
|
| 1845 |
|
| 1846 | if (flag == F2FS_GET_BLOCK_PRE_AIO &&
|
| 1847 | (pgofs == end || dn.ofs_in_node == end_offset)) {
|
| 1848 |
|
| 1849 | dn.ofs_in_node = ofs_in_node;
|
| 1850 | err = f2fs_reserve_new_blocks(&dn, prealloc);
|
| 1851 | if (err)
|
| 1852 | goto sync_out;
|
| 1853 |
|
| 1854 | map->m_len += dn.ofs_in_node - ofs_in_node;
|
| 1855 | if (prealloc && dn.ofs_in_node != last_ofs_in_node + 1) {
|
| 1856 | err = -ENOSPC;
|
| 1857 | goto sync_out;
|
| 1858 | }
|
| 1859 | dn.ofs_in_node = end_offset;
|
| 1860 | }
|
| 1861 |
|
| 1862 | if (pgofs >= end)
|
| 1863 | goto sync_out;
|
| 1864 | else if (dn.ofs_in_node < end_offset)
|
| 1865 | goto next_block;
|
| 1866 |
|
| 1867 | if (flag == F2FS_GET_BLOCK_PRECACHE) {
|
| 1868 | if (map->m_flags & F2FS_MAP_MAPPED) {
|
| 1869 | unsigned int ofs = start_pgofs - map->m_lblk;
|
| 1870 |
|
| 1871 | f2fs_update_read_extent_cache_range(&dn,
|
| 1872 | start_pgofs, map->m_pblk + ofs,
|
| 1873 | map->m_len - ofs);
|
| 1874 | }
|
| 1875 | }
|
| 1876 |
|
| 1877 | f2fs_put_dnode(&dn);
|
| 1878 |
|
| 1879 | if (map->m_may_create) {
|
| 1880 | f2fs_map_unlock(sbi, &lc, flag);
|
| 1881 | f2fs_balance_fs(sbi, dn.node_changed);
|
| 1882 | }
|
| 1883 | goto next_dnode;
|
| 1884 |
|
| 1885 | sync_out:
|
| 1886 |
|
| 1887 | if (flag == F2FS_GET_BLOCK_DIO && map->m_flags & F2FS_MAP_MAPPED) {
|
| 1888 | |
| 1889 | * for hardware encryption, but to avoid potential issue
|
| 1890 | * in future
|
| 1891 |
|
| 1892 | f2fs_wait_on_block_writeback_range(inode,
|
| 1893 | map->m_pblk, map->m_len);
|
| 1894 |
|
| 1895 | if (map->m_multidev_dio) {
|
| 1896 | block_t blk_addr = map->m_pblk;
|
| 1897 |
|
| 1898 | bidx = f2fs_target_device_index(sbi, map->m_pblk);
|
| 1899 |
|
| 1900 | map->m_bdev = FDEV(bidx).bdev;
|
| 1901 | map->m_pblk -= FDEV(bidx).start_blk;
|
| 1902 |
|
| 1903 | if (map->m_may_create)
|
| 1904 | f2fs_update_device_state(sbi, inode->i_ino,
|
| 1905 | blk_addr, map->m_len);
|
| 1906 |
|
| 1907 | f2fs_bug_on(sbi, blk_addr + map->m_len >
|
| 1908 | FDEV(bidx).end_blk + 1);
|
| 1909 | }
|
| 1910 | }
|
| 1911 |
|
| 1912 | if (flag == F2FS_GET_BLOCK_PRECACHE) {
|
| 1913 | if (map->m_flags & F2FS_MAP_MAPPED) {
|
| 1914 | unsigned int ofs = start_pgofs - map->m_lblk;
|
| 1915 |
|
| 1916 | if (map->m_len > ofs)
|
| 1917 | f2fs_update_read_extent_cache_range(&dn,
|
| 1918 | start_pgofs, map->m_pblk + ofs,
|
| 1919 | map->m_len - ofs);
|
| 1920 | }
|
| 1921 | if (map->m_next_extent)
|
| 1922 | *map->m_next_extent = is_hole ? pgofs + 1 : pgofs;
|
| 1923 | }
|
| 1924 | f2fs_put_dnode(&dn);
|
| 1925 | unlock_out:
|
| 1926 | if (map->m_may_create) {
|
| 1927 | f2fs_map_unlock(sbi, &lc, flag);
|
| 1928 | f2fs_balance_fs(sbi, dn.node_changed);
|
| 1929 | }
|
| 1930 | out:
|
| 1931 | trace_f2fs_map_blocks(inode, map, flag, err);
|
| 1932 | return err;
|
| 1933 | }
|
| 1934 |
|
| 1935 | static bool __f2fs_overwrite_io(struct inode *inode, loff_t pos, size_t len,
|
| 1936 | bool check_first)
|
| 1937 | {
|
| 1938 | struct f2fs_map_blocks map;
|
| 1939 | block_t last_lblk;
|
| 1940 | int err;
|
| 1941 |
|
| 1942 | if (pos + len > i_size_read(inode))
|
| 1943 | return false;
|
| 1944 |
|
| 1945 | map.m_lblk = F2FS_BYTES_TO_BLK(pos);
|
| 1946 | map.m_next_pgofs = NULL;
|
| 1947 | map.m_next_extent = NULL;
|
| 1948 | map.m_seg_type = NO_CHECK_TYPE;
|
| 1949 | map.m_may_create = false;
|
| 1950 | last_lblk = F2FS_BLK_ALIGN(pos + len);
|
| 1951 |
|
| 1952 | while (map.m_lblk < last_lblk) {
|
| 1953 | map.m_len = last_lblk - map.m_lblk;
|
| 1954 | err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT);
|
| 1955 | if (err || map.m_len == 0)
|
| 1956 | return false;
|
| 1957 | map.m_lblk += map.m_len;
|
| 1958 | if (check_first)
|
| 1959 | break;
|
| 1960 | }
|
| 1961 | return true;
|
| 1962 | }
|
| 1963 |
|
| 1964 | bool f2fs_overwrite_io(struct inode *inode, loff_t pos, size_t len)
|
| 1965 | {
|
| 1966 | return __f2fs_overwrite_io(inode, pos, len, false);
|
| 1967 | }
|
| 1968 |
|
| 1969 | static int f2fs_xattr_fiemap(struct inode *inode,
|
| 1970 | struct fiemap_extent_info *fieinfo)
|
| 1971 | {
|
| 1972 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
| 1973 | struct node_info ni;
|
| 1974 | __u64 phys = 0, len;
|
| 1975 | __u32 flags;
|
| 1976 | nid_t xnid = F2FS_I(inode)->i_xattr_nid;
|
| 1977 | int err = 0;
|
| 1978 |
|
| 1979 | if (f2fs_has_inline_xattr(inode)) {
|
| 1980 | int offset;
|
| 1981 | struct folio *folio = f2fs_grab_cache_folio(NODE_MAPPING(sbi),
|
| 1982 | inode->i_ino, false);
|
| 1983 |
|
| 1984 | if (IS_ERR(folio))
|
| 1985 | return PTR_ERR(folio);
|
| 1986 |
|
| 1987 | err = f2fs_get_node_info(sbi, inode->i_ino, &ni, false);
|
| 1988 | if (err) {
|
| 1989 | f2fs_folio_put(folio, true);
|
| 1990 | return err;
|
| 1991 | }
|
| 1992 |
|
| 1993 | phys = F2FS_BLK_TO_BYTES(ni.blk_addr);
|
| 1994 | offset = offsetof(struct f2fs_inode, i_addr) +
|
| 1995 | sizeof(__le32) * (DEF_ADDRS_PER_INODE -
|
| 1996 | get_inline_xattr_addrs(inode));
|
| 1997 |
|
| 1998 | phys += offset;
|
| 1999 | len = inline_xattr_size(inode);
|
| 2000 |
|
| 2001 | f2fs_folio_put(folio, true);
|
| 2002 |
|
| 2003 | flags = FIEMAP_EXTENT_DATA_INLINE | FIEMAP_EXTENT_NOT_ALIGNED;
|
| 2004 |
|
| 2005 | if (!xnid)
|
| 2006 | flags |= FIEMAP_EXTENT_LAST;
|
| 2007 |
|
| 2008 | err = fiemap_fill_next_extent(fieinfo, 0, phys, len, flags);
|
| 2009 | trace_f2fs_fiemap(inode, 0, phys, len, flags, err);
|
| 2010 | if (err)
|
| 2011 | return err;
|
| 2012 | }
|
| 2013 |
|
| 2014 | if (xnid) {
|
| 2015 | struct folio *folio = f2fs_grab_cache_folio(NODE_MAPPING(sbi),
|
| 2016 | xnid, false);
|
| 2017 |
|
| 2018 | if (IS_ERR(folio))
|
| 2019 | return PTR_ERR(folio);
|
| 2020 |
|
| 2021 | err = f2fs_get_node_info(sbi, xnid, &ni, false);
|
| 2022 | if (err) {
|
| 2023 | f2fs_folio_put(folio, true);
|
| 2024 | return err;
|
| 2025 | }
|
| 2026 |
|
| 2027 | phys = F2FS_BLK_TO_BYTES(ni.blk_addr);
|
| 2028 | len = inode->i_sb->s_blocksize;
|
| 2029 |
|
| 2030 | f2fs_folio_put(folio, true);
|
| 2031 |
|
| 2032 | flags = FIEMAP_EXTENT_LAST;
|
| 2033 | }
|
| 2034 |
|
| 2035 | if (phys) {
|
| 2036 | err = fiemap_fill_next_extent(fieinfo, 0, phys, len, flags);
|
| 2037 | trace_f2fs_fiemap(inode, 0, phys, len, flags, err);
|
| 2038 | }
|
| 2039 |
|
| 2040 | return (err < 0 ? err : 0);
|
| 2041 | }
|
| 2042 |
|
| 2043 | int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
|
| 2044 | u64 start, u64 len)
|
| 2045 | {
|
| 2046 | struct f2fs_map_blocks map;
|
| 2047 | sector_t start_blk, last_blk, blk_len, max_len;
|
| 2048 | pgoff_t next_pgofs;
|
| 2049 | u64 logical = 0, phys = 0, size = 0;
|
| 2050 | u32 flags = 0;
|
| 2051 | int ret = 0;
|
| 2052 | bool compr_cluster = false, compr_appended;
|
| 2053 | unsigned int cluster_size = F2FS_I(inode)->i_cluster_size;
|
| 2054 | unsigned int count_in_cluster = 0;
|
| 2055 | loff_t maxbytes;
|
| 2056 |
|
| 2057 | if (fieinfo->fi_flags & FIEMAP_FLAG_CACHE) {
|
| 2058 | ret = f2fs_precache_extents(inode);
|
| 2059 | if (ret)
|
| 2060 | return ret;
|
| 2061 | }
|
| 2062 |
|
| 2063 | ret = fiemap_prep(inode, fieinfo, start, &len, FIEMAP_FLAG_XATTR);
|
| 2064 | if (ret)
|
| 2065 | return ret;
|
| 2066 |
|
| 2067 | inode_lock_shared(inode);
|
| 2068 |
|
| 2069 | maxbytes = F2FS_BLK_TO_BYTES(max_file_blocks(inode));
|
| 2070 | if (start > maxbytes) {
|
| 2071 | ret = -EFBIG;
|
| 2072 | goto out;
|
| 2073 | }
|
| 2074 |
|
| 2075 | if (len > maxbytes || (maxbytes - len) < start)
|
| 2076 | len = maxbytes - start;
|
| 2077 |
|
| 2078 | if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR) {
|
| 2079 | ret = f2fs_xattr_fiemap(inode, fieinfo);
|
| 2080 | goto out;
|
| 2081 | }
|
| 2082 |
|
| 2083 | if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
|
| 2084 | ret = f2fs_inline_data_fiemap(inode, fieinfo, start, len);
|
| 2085 | if (ret != -EAGAIN)
|
| 2086 | goto out;
|
| 2087 | }
|
| 2088 |
|
| 2089 | start_blk = F2FS_BYTES_TO_BLK(start);
|
| 2090 | last_blk = F2FS_BYTES_TO_BLK(start + len - 1);
|
| 2091 | blk_len = last_blk - start_blk + 1;
|
| 2092 | max_len = F2FS_BYTES_TO_BLK(maxbytes) - start_blk;
|
| 2093 |
|
| 2094 | next:
|
| 2095 | memset(&map, 0, sizeof(map));
|
| 2096 | map.m_lblk = start_blk;
|
| 2097 | map.m_len = blk_len;
|
| 2098 | map.m_next_pgofs = &next_pgofs;
|
| 2099 | map.m_seg_type = NO_CHECK_TYPE;
|
| 2100 |
|
| 2101 | if (compr_cluster) {
|
| 2102 | map.m_lblk += 1;
|
| 2103 | map.m_len = cluster_size - count_in_cluster;
|
| 2104 | }
|
| 2105 |
|
| 2106 | ret = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_FIEMAP);
|
| 2107 | if (ret)
|
| 2108 | goto out;
|
| 2109 |
|
| 2110 |
|
| 2111 | if (!compr_cluster && !(map.m_flags & F2FS_MAP_FLAGS)) {
|
| 2112 | start_blk = next_pgofs;
|
| 2113 |
|
| 2114 | if (F2FS_BLK_TO_BYTES(start_blk) < maxbytes)
|
| 2115 | goto prep_next;
|
| 2116 |
|
| 2117 | flags |= FIEMAP_EXTENT_LAST;
|
| 2118 | }
|
| 2119 |
|
| 2120 | |
| 2121 | * current extent may cross boundary of inquiry, increase len to
|
| 2122 | * requery.
|
| 2123 |
|
| 2124 | if (!compr_cluster && (map.m_flags & F2FS_MAP_MAPPED) &&
|
| 2125 | map.m_lblk + map.m_len - 1 == last_blk &&
|
| 2126 | blk_len != max_len) {
|
| 2127 | blk_len = max_len;
|
| 2128 | goto next;
|
| 2129 | }
|
| 2130 |
|
| 2131 | compr_appended = false;
|
| 2132 |
|
| 2133 | if (compr_cluster && ((map.m_flags & F2FS_MAP_DELALLOC) ||
|
| 2134 | !(map.m_flags & F2FS_MAP_FLAGS))) {
|
| 2135 | compr_appended = true;
|
| 2136 | goto skip_fill;
|
| 2137 | }
|
| 2138 |
|
| 2139 | if (size) {
|
| 2140 | flags |= FIEMAP_EXTENT_MERGED;
|
| 2141 | if (IS_ENCRYPTED(inode))
|
| 2142 | flags |= FIEMAP_EXTENT_DATA_ENCRYPTED;
|
| 2143 |
|
| 2144 | ret = fiemap_fill_next_extent(fieinfo, logical,
|
| 2145 | phys, size, flags);
|
| 2146 | trace_f2fs_fiemap(inode, logical, phys, size, flags, ret);
|
| 2147 | if (ret)
|
| 2148 | goto out;
|
| 2149 | size = 0;
|
| 2150 | }
|
| 2151 |
|
| 2152 | if (start_blk > last_blk)
|
| 2153 | goto out;
|
| 2154 |
|
| 2155 | skip_fill:
|
| 2156 | if (map.m_pblk == COMPRESS_ADDR) {
|
| 2157 | compr_cluster = true;
|
| 2158 | count_in_cluster = 1;
|
| 2159 | } else if (compr_appended) {
|
| 2160 | unsigned int appended_blks = cluster_size -
|
| 2161 | count_in_cluster + 1;
|
| 2162 | size += F2FS_BLK_TO_BYTES(appended_blks);
|
| 2163 | start_blk += appended_blks;
|
| 2164 | compr_cluster = false;
|
| 2165 | } else {
|
| 2166 | logical = F2FS_BLK_TO_BYTES(start_blk);
|
| 2167 | phys = __is_valid_data_blkaddr(map.m_pblk) ?
|
| 2168 | F2FS_BLK_TO_BYTES(map.m_pblk) : 0;
|
| 2169 | size = F2FS_BLK_TO_BYTES(map.m_len);
|
| 2170 | flags = 0;
|
| 2171 |
|
| 2172 | if (compr_cluster) {
|
| 2173 | flags = FIEMAP_EXTENT_ENCODED;
|
| 2174 | count_in_cluster += map.m_len;
|
| 2175 | if (count_in_cluster == cluster_size) {
|
| 2176 | compr_cluster = false;
|
| 2177 | size += F2FS_BLKSIZE;
|
| 2178 | }
|
| 2179 | } else if (map.m_flags & F2FS_MAP_DELALLOC) {
|
| 2180 | flags = FIEMAP_EXTENT_UNWRITTEN;
|
| 2181 | }
|
| 2182 |
|
| 2183 | start_blk += F2FS_BYTES_TO_BLK(size);
|
| 2184 | }
|
| 2185 |
|
| 2186 | prep_next:
|
| 2187 | cond_resched();
|
| 2188 | if (fatal_signal_pending(current))
|
| 2189 | ret = -EINTR;
|
| 2190 | else
|
| 2191 | goto next;
|
| 2192 | out:
|
| 2193 | if (ret == 1)
|
| 2194 | ret = 0;
|
| 2195 |
|
| 2196 | inode_unlock_shared(inode);
|
| 2197 | return ret;
|
| 2198 | }
|
| 2199 |
|
| 2200 | static inline loff_t f2fs_readpage_limit(struct inode *inode)
|
| 2201 | {
|
| 2202 | if (IS_ENABLED(CONFIG_FS_VERITY) && IS_VERITY(inode))
|
| 2203 | return F2FS_BLK_TO_BYTES(max_file_blocks(inode));
|
| 2204 |
|
| 2205 | return i_size_read(inode);
|
| 2206 | }
|
| 2207 |
|
| 2208 | static inline blk_opf_t f2fs_ra_op_flags(struct readahead_control *rac)
|
| 2209 | {
|
| 2210 | return rac ? REQ_RAHEAD : 0;
|
| 2211 | }
|
| 2212 |
|
| 2213 | static int f2fs_read_single_page(struct inode *inode, struct fsverity_info *vi,
|
| 2214 | struct folio *folio, unsigned int nr_pages,
|
| 2215 | struct f2fs_map_blocks *map,
|
| 2216 | struct bio **bio_ret,
|
| 2217 | sector_t *last_block_in_bio,
|
| 2218 | struct readahead_control *rac)
|
| 2219 | {
|
| 2220 | struct bio *bio = *bio_ret;
|
| 2221 | const unsigned int blocksize = F2FS_BLKSIZE;
|
| 2222 | sector_t block_in_file;
|
| 2223 | sector_t last_block;
|
| 2224 | sector_t last_block_in_file;
|
| 2225 | sector_t block_nr;
|
| 2226 | pgoff_t index = folio->index;
|
| 2227 | int ret = 0;
|
| 2228 |
|
| 2229 | block_in_file = (sector_t)index;
|
| 2230 | last_block = block_in_file + nr_pages;
|
| 2231 | last_block_in_file = F2FS_BYTES_TO_BLK(f2fs_readpage_limit(inode) +
|
| 2232 | blocksize - 1);
|
| 2233 | if (last_block > last_block_in_file)
|
| 2234 | last_block = last_block_in_file;
|
| 2235 |
|
| 2236 |
|
| 2237 | if (block_in_file >= last_block)
|
| 2238 | goto zero_out;
|
| 2239 | |
| 2240 | * Map blocks using the previous result first.
|
| 2241 |
|
| 2242 | if (map->m_flags & F2FS_MAP_MAPPED) {
|
| 2243 | if (block_in_file > map->m_lblk &&
|
| 2244 | block_in_file < (map->m_lblk + map->m_len))
|
| 2245 | goto got_it;
|
| 2246 | } else if (block_in_file < *map->m_next_pgofs) {
|
| 2247 | goto got_it;
|
| 2248 | }
|
| 2249 |
|
| 2250 | |
| 2251 | * Then do more f2fs_map_blocks() calls until we are
|
| 2252 | * done with this page.
|
| 2253 |
|
| 2254 | map->m_lblk = block_in_file;
|
| 2255 | map->m_len = last_block - block_in_file;
|
| 2256 |
|
| 2257 | ret = f2fs_map_blocks(inode, map, F2FS_GET_BLOCK_DEFAULT);
|
| 2258 | if (ret)
|
| 2259 | goto out;
|
| 2260 | got_it:
|
| 2261 | if ((map->m_flags & F2FS_MAP_MAPPED)) {
|
| 2262 | block_nr = map->m_pblk + block_in_file - map->m_lblk;
|
| 2263 | folio_set_mappedtodisk(folio);
|
| 2264 |
|
| 2265 | if (!f2fs_is_valid_blkaddr(F2FS_I_SB(inode), block_nr,
|
| 2266 | DATA_GENERIC_ENHANCE_READ)) {
|
| 2267 | ret = -EFSCORRUPTED;
|
| 2268 | goto out;
|
| 2269 | }
|
| 2270 | } else {
|
| 2271 | zero_out:
|
| 2272 | folio_zero_segment(folio, 0, folio_size(folio));
|
| 2273 | if (vi && !fsverity_verify_folio(vi, folio)) {
|
| 2274 | ret = -EIO;
|
| 2275 | goto out;
|
| 2276 | }
|
| 2277 | if (!folio_test_uptodate(folio))
|
| 2278 | folio_mark_uptodate(folio);
|
| 2279 | folio_unlock(folio);
|
| 2280 | goto out;
|
| 2281 | }
|
| 2282 |
|
| 2283 | |
| 2284 | * This page will go to BIO. Do we need to send this
|
| 2285 | * BIO off first?
|
| 2286 |
|
| 2287 | if (bio && (!page_is_mergeable(F2FS_I_SB(inode), bio,
|
| 2288 | *last_block_in_bio, block_nr) ||
|
| 2289 | !f2fs_crypt_mergeable_bio(bio, inode, index, NULL))) {
|
| 2290 | submit_and_realloc:
|
| 2291 | f2fs_submit_read_bio(F2FS_I_SB(inode), bio, DATA);
|
| 2292 | bio = NULL;
|
| 2293 | }
|
| 2294 | if (bio == NULL)
|
| 2295 | bio = f2fs_grab_read_bio(inode, vi, block_nr, nr_pages,
|
| 2296 | f2fs_ra_op_flags(rac), index, false);
|
| 2297 |
|
| 2298 | |
| 2299 | * If the page is under writeback, we need to wait for
|
| 2300 | * its completion to see the correct decrypted data.
|
| 2301 |
|
| 2302 | f2fs_wait_on_block_writeback(inode, block_nr);
|
| 2303 |
|
| 2304 | if (!bio_add_folio(bio, folio, blocksize, 0))
|
| 2305 | goto submit_and_realloc;
|
| 2306 |
|
| 2307 | inc_page_count(F2FS_I_SB(inode), F2FS_RD_DATA);
|
| 2308 | f2fs_update_iostat(F2FS_I_SB(inode), NULL, FS_DATA_READ_IO,
|
| 2309 | F2FS_BLKSIZE);
|
| 2310 | *last_block_in_bio = block_nr;
|
| 2311 | out:
|
| 2312 | *bio_ret = bio;
|
| 2313 | return ret;
|
| 2314 | }
|
| 2315 |
|
| 2316 | #ifdef CONFIG_F2FS_FS_COMPRESSION
|
| 2317 | int f2fs_read_multi_pages(struct compress_ctx *cc, struct bio **bio_ret,
|
| 2318 | unsigned nr_pages, sector_t *last_block_in_bio,
|
| 2319 | struct readahead_control *rac, bool for_write)
|
| 2320 | {
|
| 2321 | struct dnode_of_data dn;
|
| 2322 | struct inode *inode = cc->inode;
|
| 2323 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
| 2324 | struct bio *bio = *bio_ret;
|
| 2325 | unsigned int start_idx = cc->cluster_idx << cc->log_cluster_size;
|
| 2326 | sector_t last_block_in_file;
|
| 2327 | const unsigned int blocksize = F2FS_BLKSIZE;
|
| 2328 | struct decompress_io_ctx *dic = NULL;
|
| 2329 | struct extent_info ei = {};
|
| 2330 | bool from_dnode = true;
|
| 2331 | int i;
|
| 2332 | int ret = 0;
|
| 2333 |
|
| 2334 | if (unlikely(f2fs_cp_error(sbi))) {
|
| 2335 | ret = -EIO;
|
| 2336 | from_dnode = false;
|
| 2337 | goto out_put_dnode;
|
| 2338 | }
|
| 2339 |
|
| 2340 | f2fs_bug_on(sbi, f2fs_cluster_is_empty(cc));
|
| 2341 |
|
| 2342 | last_block_in_file = F2FS_BYTES_TO_BLK(f2fs_readpage_limit(inode) +
|
| 2343 | blocksize - 1);
|
| 2344 |
|
| 2345 |
|
| 2346 | for (i = 0; i < cc->cluster_size; i++) {
|
| 2347 | struct page *page = cc->rpages[i];
|
| 2348 | struct folio *folio;
|
| 2349 |
|
| 2350 | if (!page)
|
| 2351 | continue;
|
| 2352 |
|
| 2353 | folio = page_folio(page);
|
| 2354 | if ((sector_t)folio->index >= last_block_in_file) {
|
| 2355 | folio_zero_segment(folio, 0, folio_size(folio));
|
| 2356 | if (!folio_test_uptodate(folio))
|
| 2357 | folio_mark_uptodate(folio);
|
| 2358 | } else if (!folio_test_uptodate(folio)) {
|
| 2359 | continue;
|
| 2360 | }
|
| 2361 | folio_unlock(folio);
|
| 2362 | if (for_write)
|
| 2363 | folio_put(folio);
|
| 2364 | cc->rpages[i] = NULL;
|
| 2365 | cc->nr_rpages--;
|
| 2366 | }
|
| 2367 |
|
| 2368 |
|
| 2369 | if (f2fs_cluster_is_empty(cc))
|
| 2370 | goto out;
|
| 2371 |
|
| 2372 | if (f2fs_lookup_read_extent_cache(inode, start_idx, &ei))
|
| 2373 | from_dnode = false;
|
| 2374 |
|
| 2375 | if (!from_dnode)
|
| 2376 | goto skip_reading_dnode;
|
| 2377 |
|
| 2378 | set_new_dnode(&dn, inode, NULL, NULL, 0);
|
| 2379 | ret = f2fs_get_dnode_of_data(&dn, start_idx, LOOKUP_NODE);
|
| 2380 | if (ret)
|
| 2381 | goto out;
|
| 2382 |
|
| 2383 | f2fs_bug_on(sbi, dn.data_blkaddr != COMPRESS_ADDR);
|
| 2384 |
|
| 2385 | skip_reading_dnode:
|
| 2386 | for (i = 1; i < cc->cluster_size; i++) {
|
| 2387 | block_t blkaddr;
|
| 2388 |
|
| 2389 | blkaddr = from_dnode ? data_blkaddr(dn.inode, dn.node_folio,
|
| 2390 | dn.ofs_in_node + i) :
|
| 2391 | ei.blk + i - 1;
|
| 2392 |
|
| 2393 | if (!__is_valid_data_blkaddr(blkaddr))
|
| 2394 | break;
|
| 2395 |
|
| 2396 | if (!f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC)) {
|
| 2397 | ret = -EFAULT;
|
| 2398 | goto out_put_dnode;
|
| 2399 | }
|
| 2400 | cc->nr_cpages++;
|
| 2401 |
|
| 2402 | if (!from_dnode && i >= ei.c_len)
|
| 2403 | break;
|
| 2404 | }
|
| 2405 |
|
| 2406 |
|
| 2407 | if (cc->nr_cpages == 0) {
|
| 2408 | ret = 0;
|
| 2409 | goto out_put_dnode;
|
| 2410 | }
|
| 2411 |
|
| 2412 | dic = f2fs_alloc_dic(cc);
|
| 2413 | if (IS_ERR(dic)) {
|
| 2414 | ret = PTR_ERR(dic);
|
| 2415 | goto out_put_dnode;
|
| 2416 | }
|
| 2417 |
|
| 2418 | for (i = 0; i < cc->nr_cpages; i++) {
|
| 2419 | struct folio *folio = page_folio(dic->cpages[i]);
|
| 2420 | block_t blkaddr;
|
| 2421 | struct bio_post_read_ctx *ctx;
|
| 2422 |
|
| 2423 | blkaddr = from_dnode ? data_blkaddr(dn.inode, dn.node_folio,
|
| 2424 | dn.ofs_in_node + i + 1) :
|
| 2425 | ei.blk + i;
|
| 2426 |
|
| 2427 | f2fs_wait_on_block_writeback(inode, blkaddr);
|
| 2428 |
|
| 2429 | if (f2fs_load_compressed_folio(sbi, folio, blkaddr)) {
|
| 2430 | if (atomic_dec_and_test(&dic->remaining_pages)) {
|
| 2431 | f2fs_decompress_cluster(dic, true);
|
| 2432 | break;
|
| 2433 | }
|
| 2434 | continue;
|
| 2435 | }
|
| 2436 |
|
| 2437 | if (bio && (!page_is_mergeable(sbi, bio,
|
| 2438 | *last_block_in_bio, blkaddr) ||
|
| 2439 | !f2fs_crypt_mergeable_bio(bio, inode, folio->index, NULL))) {
|
| 2440 | submit_and_realloc:
|
| 2441 | f2fs_submit_read_bio(sbi, bio, DATA);
|
| 2442 | bio = NULL;
|
| 2443 | }
|
| 2444 |
|
| 2445 | if (!bio)
|
| 2446 | bio = f2fs_grab_read_bio(inode, cc->vi, blkaddr,
|
| 2447 | nr_pages - i,
|
| 2448 | f2fs_ra_op_flags(rac),
|
| 2449 | folio->index, for_write);
|
| 2450 |
|
| 2451 | if (!bio_add_folio(bio, folio, blocksize, 0))
|
| 2452 | goto submit_and_realloc;
|
| 2453 |
|
| 2454 | ctx = get_post_read_ctx(bio);
|
| 2455 | ctx->enabled_steps |= STEP_DECOMPRESS;
|
| 2456 | refcount_inc(&dic->refcnt);
|
| 2457 |
|
| 2458 | inc_page_count(sbi, F2FS_RD_DATA);
|
| 2459 | f2fs_update_iostat(sbi, inode, FS_DATA_READ_IO, F2FS_BLKSIZE);
|
| 2460 | *last_block_in_bio = blkaddr;
|
| 2461 | }
|
| 2462 |
|
| 2463 | if (from_dnode)
|
| 2464 | f2fs_put_dnode(&dn);
|
| 2465 |
|
| 2466 | *bio_ret = bio;
|
| 2467 | return 0;
|
| 2468 |
|
| 2469 | out_put_dnode:
|
| 2470 | if (from_dnode)
|
| 2471 | f2fs_put_dnode(&dn);
|
| 2472 | out:
|
| 2473 | for (i = 0; i < cc->cluster_size; i++) {
|
| 2474 | if (cc->rpages[i]) {
|
| 2475 | ClearPageUptodate(cc->rpages[i]);
|
| 2476 | unlock_page(cc->rpages[i]);
|
| 2477 | }
|
| 2478 | }
|
| 2479 | *bio_ret = bio;
|
| 2480 | return ret;
|
| 2481 | }
|
| 2482 | #endif
|
| 2483 |
|
| 2484 | static struct f2fs_folio_state *ffs_find_or_alloc(struct folio *folio)
|
| 2485 | {
|
| 2486 | struct f2fs_folio_state *ffs = folio->private;
|
| 2487 |
|
| 2488 | if (ffs)
|
| 2489 | return ffs;
|
| 2490 |
|
| 2491 | ffs = f2fs_kmem_cache_alloc(ffs_entry_slab,
|
| 2492 | GFP_NOIO | __GFP_ZERO, true, NULL);
|
| 2493 |
|
| 2494 | spin_lock_init(&ffs->state_lock);
|
| 2495 | folio_attach_private(folio, ffs);
|
| 2496 | return ffs;
|
| 2497 | }
|
| 2498 |
|
| 2499 | static void ffs_detach_free(struct folio *folio)
|
| 2500 | {
|
| 2501 | struct f2fs_folio_state *ffs;
|
| 2502 |
|
| 2503 | if (!folio_test_large(folio)) {
|
| 2504 | folio_detach_private(folio);
|
| 2505 | return;
|
| 2506 | }
|
| 2507 |
|
| 2508 | ffs = folio_detach_private(folio);
|
| 2509 | if (!ffs)
|
| 2510 | return;
|
| 2511 |
|
| 2512 | WARN_ON_ONCE(ffs->read_pages_pending != 0);
|
| 2513 | kmem_cache_free(ffs_entry_slab, ffs);
|
| 2514 | }
|
| 2515 |
|
| 2516 | static int f2fs_read_data_large_folio(struct inode *inode,
|
| 2517 | struct fsverity_info *vi,
|
| 2518 | struct readahead_control *rac, struct folio *folio)
|
| 2519 | {
|
| 2520 | struct bio *bio = NULL;
|
| 2521 | sector_t last_block_in_bio = 0;
|
| 2522 | struct f2fs_map_blocks map = {0, };
|
| 2523 | pgoff_t index, offset, next_pgofs = 0;
|
| 2524 | unsigned max_nr_pages = rac ? readahead_count(rac) :
|
| 2525 | folio_nr_pages(folio);
|
| 2526 | unsigned nrpages;
|
| 2527 | struct f2fs_folio_state *ffs;
|
| 2528 | int ret = 0;
|
| 2529 | bool folio_in_bio = false;
|
| 2530 |
|
| 2531 | if (!IS_IMMUTABLE(inode) || f2fs_compressed_file(inode)) {
|
| 2532 | if (folio)
|
| 2533 | folio_unlock(folio);
|
| 2534 | return -EOPNOTSUPP;
|
| 2535 | }
|
| 2536 |
|
| 2537 | map.m_seg_type = NO_CHECK_TYPE;
|
| 2538 |
|
| 2539 | if (rac)
|
| 2540 | folio = readahead_folio(rac);
|
| 2541 | next_folio:
|
| 2542 | if (!folio)
|
| 2543 | goto out;
|
| 2544 |
|
| 2545 | f2fs_update_read_folio_count(F2FS_I_SB(inode), folio);
|
| 2546 |
|
| 2547 | folio_in_bio = false;
|
| 2548 | index = folio->index;
|
| 2549 | offset = 0;
|
| 2550 | ffs = NULL;
|
| 2551 | nrpages = folio_nr_pages(folio);
|
| 2552 |
|
| 2553 | for (; nrpages; nrpages--, max_nr_pages--, index++, offset++) {
|
| 2554 | sector_t block_nr;
|
| 2555 | |
| 2556 | * Map blocks using the previous result first.
|
| 2557 |
|
| 2558 | if (map.m_flags & F2FS_MAP_MAPPED) {
|
| 2559 | if (index > map.m_lblk &&
|
| 2560 | index < (map.m_lblk + map.m_len))
|
| 2561 | goto got_it;
|
| 2562 | } else if (index < next_pgofs) {
|
| 2563 |
|
| 2564 | goto got_it;
|
| 2565 | }
|
| 2566 |
|
| 2567 | |
| 2568 | * Then do more f2fs_map_blocks() calls until we are
|
| 2569 | * done with this page.
|
| 2570 |
|
| 2571 | memset(&map, 0, sizeof(map));
|
| 2572 | map.m_next_pgofs = &next_pgofs;
|
| 2573 | map.m_seg_type = NO_CHECK_TYPE;
|
| 2574 | map.m_lblk = index;
|
| 2575 | map.m_len = max_nr_pages;
|
| 2576 |
|
| 2577 | ret = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DEFAULT);
|
| 2578 | if (ret)
|
| 2579 | goto err_out;
|
| 2580 | got_it:
|
| 2581 | if ((map.m_flags & F2FS_MAP_MAPPED)) {
|
| 2582 | block_nr = map.m_pblk + index - map.m_lblk;
|
| 2583 | if (!f2fs_is_valid_blkaddr(F2FS_I_SB(inode), block_nr,
|
| 2584 | DATA_GENERIC_ENHANCE_READ)) {
|
| 2585 | ret = -EFSCORRUPTED;
|
| 2586 | goto err_out;
|
| 2587 | }
|
| 2588 | } else {
|
| 2589 | size_t page_offset = offset << PAGE_SHIFT;
|
| 2590 | folio_zero_range(folio, page_offset, PAGE_SIZE);
|
| 2591 | if (vi && !fsverity_verify_blocks(vi, folio, PAGE_SIZE, page_offset)) {
|
| 2592 | ret = -EIO;
|
| 2593 | goto err_out;
|
| 2594 | }
|
| 2595 | continue;
|
| 2596 | }
|
| 2597 |
|
| 2598 | |
| 2599 | * to prevent from premature folio_end_read() call on folio
|
| 2600 |
|
| 2601 | if (folio_test_large(folio)) {
|
| 2602 | ffs = ffs_find_or_alloc(folio);
|
| 2603 |
|
| 2604 |
|
| 2605 | spin_lock_irq(&ffs->state_lock);
|
| 2606 | ffs->read_pages_pending++;
|
| 2607 | spin_unlock_irq(&ffs->state_lock);
|
| 2608 | }
|
| 2609 |
|
| 2610 | |
| 2611 | * This page will go to BIO. Do we need to send this
|
| 2612 | * BIO off first?
|
| 2613 |
|
| 2614 | if (bio && (!page_is_mergeable(F2FS_I_SB(inode), bio,
|
| 2615 | last_block_in_bio, block_nr) ||
|
| 2616 | !f2fs_crypt_mergeable_bio(bio, inode, index, NULL))) {
|
| 2617 | submit_and_realloc:
|
| 2618 | f2fs_submit_read_bio(F2FS_I_SB(inode), bio, DATA);
|
| 2619 | bio = NULL;
|
| 2620 | }
|
| 2621 | if (bio == NULL)
|
| 2622 | bio = f2fs_grab_read_bio(inode, vi,
|
| 2623 | block_nr, max_nr_pages,
|
| 2624 | f2fs_ra_op_flags(rac),
|
| 2625 | index, false);
|
| 2626 |
|
| 2627 | |
| 2628 | * If the page is under writeback, we need to wait for
|
| 2629 | * its completion to see the correct decrypted data.
|
| 2630 |
|
| 2631 | f2fs_wait_on_block_writeback(inode, block_nr);
|
| 2632 |
|
| 2633 | if (!bio_add_folio(bio, folio, F2FS_BLKSIZE,
|
| 2634 | offset << PAGE_SHIFT))
|
| 2635 | goto submit_and_realloc;
|
| 2636 |
|
| 2637 | folio_in_bio = true;
|
| 2638 | inc_page_count(F2FS_I_SB(inode), F2FS_RD_DATA);
|
| 2639 | f2fs_update_iostat(F2FS_I_SB(inode), NULL, FS_DATA_READ_IO,
|
| 2640 | F2FS_BLKSIZE);
|
| 2641 | last_block_in_bio = block_nr;
|
| 2642 | }
|
| 2643 | trace_f2fs_read_folio(folio, DATA);
|
| 2644 | err_out:
|
| 2645 | if (!folio_in_bio)
|
| 2646 | folio_end_read(folio, !ret);
|
| 2647 | if (ret)
|
| 2648 | goto out;
|
| 2649 | if (rac) {
|
| 2650 | folio = readahead_folio(rac);
|
| 2651 | goto next_folio;
|
| 2652 | }
|
| 2653 | out:
|
| 2654 | f2fs_submit_read_bio(F2FS_I_SB(inode), bio, DATA);
|
| 2655 | if (ret && folio_in_bio) {
|
| 2656 |
|
| 2657 | folio_lock(folio);
|
| 2658 | folio_clear_uptodate(folio);
|
| 2659 | folio_unlock(folio);
|
| 2660 | }
|
| 2661 | return ret;
|
| 2662 | }
|
| 2663 |
|
| 2664 | |
| 2665 | * This function was originally taken from fs/mpage.c, and customized for f2fs.
|
| 2666 | * Major change was from block_size == page_size in f2fs by default.
|
| 2667 |
|
| 2668 | static int f2fs_mpage_readpages(struct inode *inode, struct fsverity_info *vi,
|
| 2669 | struct readahead_control *rac, struct folio *folio)
|
| 2670 | {
|
| 2671 | struct bio *bio = NULL;
|
| 2672 | sector_t last_block_in_bio = 0;
|
| 2673 | struct f2fs_map_blocks map;
|
| 2674 | #ifdef CONFIG_F2FS_FS_COMPRESSION
|
| 2675 | struct compress_ctx cc = {
|
| 2676 | .inode = inode,
|
| 2677 | .log_cluster_size = F2FS_I(inode)->i_log_cluster_size,
|
| 2678 | .cluster_size = F2FS_I(inode)->i_cluster_size,
|
| 2679 | .cluster_idx = NULL_CLUSTER,
|
| 2680 | .rpages = NULL,
|
| 2681 | .cpages = NULL,
|
| 2682 | .nr_rpages = 0,
|
| 2683 | .nr_cpages = 0,
|
| 2684 | };
|
| 2685 | pgoff_t nc_cluster_idx = NULL_CLUSTER;
|
| 2686 | pgoff_t index;
|
| 2687 | #endif
|
| 2688 | pgoff_t next_pgofs = 0;
|
| 2689 | unsigned nr_pages = rac ? readahead_count(rac) : 1;
|
| 2690 | struct address_space *mapping = rac ? rac->mapping : folio->mapping;
|
| 2691 | unsigned max_nr_pages = nr_pages;
|
| 2692 | int ret = 0;
|
| 2693 |
|
| 2694 | if (mapping_large_folio_support(mapping))
|
| 2695 | return f2fs_read_data_large_folio(inode, vi, rac, folio);
|
| 2696 |
|
| 2697 | #ifdef CONFIG_F2FS_FS_COMPRESSION
|
| 2698 | if (f2fs_compressed_file(inode)) {
|
| 2699 | index = rac ? readahead_index(rac) : folio->index;
|
| 2700 | max_nr_pages = round_up(index + nr_pages, cc.cluster_size) -
|
| 2701 | round_down(index, cc.cluster_size);
|
| 2702 | }
|
| 2703 | #endif
|
| 2704 |
|
| 2705 | map.m_pblk = 0;
|
| 2706 | map.m_lblk = 0;
|
| 2707 | map.m_len = 0;
|
| 2708 | map.m_flags = 0;
|
| 2709 | map.m_next_pgofs = &next_pgofs;
|
| 2710 | map.m_next_extent = NULL;
|
| 2711 | map.m_seg_type = NO_CHECK_TYPE;
|
| 2712 | map.m_may_create = false;
|
| 2713 |
|
| 2714 | for (; nr_pages; nr_pages--) {
|
| 2715 | if (rac) {
|
| 2716 | folio = readahead_folio(rac);
|
| 2717 | prefetchw(&folio->flags);
|
| 2718 | }
|
| 2719 |
|
| 2720 | f2fs_update_read_folio_count(F2FS_I_SB(inode), folio);
|
| 2721 |
|
| 2722 | #ifdef CONFIG_F2FS_FS_COMPRESSION
|
| 2723 | index = folio->index;
|
| 2724 |
|
| 2725 | if (!f2fs_compressed_file(inode))
|
| 2726 | goto read_single_page;
|
| 2727 |
|
| 2728 |
|
| 2729 | if (!f2fs_cluster_can_merge_page(&cc, index)) {
|
| 2730 | cc.vi = vi;
|
| 2731 | ret = f2fs_read_multi_pages(&cc, &bio,
|
| 2732 | max_nr_pages,
|
| 2733 | &last_block_in_bio,
|
| 2734 | rac, false);
|
| 2735 | f2fs_destroy_compress_ctx(&cc, false);
|
| 2736 | if (ret)
|
| 2737 | goto set_error_page;
|
| 2738 | }
|
| 2739 | if (cc.cluster_idx == NULL_CLUSTER) {
|
| 2740 | if (nc_cluster_idx == index >> cc.log_cluster_size)
|
| 2741 | goto read_single_page;
|
| 2742 |
|
| 2743 | ret = f2fs_is_compressed_cluster(inode, index);
|
| 2744 | if (ret < 0)
|
| 2745 | goto set_error_page;
|
| 2746 | else if (!ret) {
|
| 2747 | nc_cluster_idx =
|
| 2748 | index >> cc.log_cluster_size;
|
| 2749 | goto read_single_page;
|
| 2750 | }
|
| 2751 |
|
| 2752 | nc_cluster_idx = NULL_CLUSTER;
|
| 2753 | }
|
| 2754 | ret = f2fs_init_compress_ctx(&cc);
|
| 2755 | if (ret)
|
| 2756 | goto set_error_page;
|
| 2757 |
|
| 2758 | f2fs_compress_ctx_add_page(&cc, folio);
|
| 2759 |
|
| 2760 | goto next_page;
|
| 2761 | read_single_page:
|
| 2762 | #endif
|
| 2763 |
|
| 2764 | ret = f2fs_read_single_page(inode, vi, folio, max_nr_pages,
|
| 2765 | &map, &bio, &last_block_in_bio,
|
| 2766 | rac);
|
| 2767 | if (ret) {
|
| 2768 | #ifdef CONFIG_F2FS_FS_COMPRESSION
|
| 2769 | set_error_page:
|
| 2770 | #endif
|
| 2771 | folio_zero_segment(folio, 0, folio_size(folio));
|
| 2772 | folio_unlock(folio);
|
| 2773 | }
|
| 2774 | #ifdef CONFIG_F2FS_FS_COMPRESSION
|
| 2775 | next_page:
|
| 2776 | #endif
|
| 2777 |
|
| 2778 | #ifdef CONFIG_F2FS_FS_COMPRESSION
|
| 2779 | if (f2fs_compressed_file(inode)) {
|
| 2780 |
|
| 2781 | if (nr_pages == 1 && !f2fs_cluster_is_empty(&cc)) {
|
| 2782 | cc.vi = vi;
|
| 2783 | ret = f2fs_read_multi_pages(&cc, &bio,
|
| 2784 | max_nr_pages,
|
| 2785 | &last_block_in_bio,
|
| 2786 | rac, false);
|
| 2787 | f2fs_destroy_compress_ctx(&cc, false);
|
| 2788 | }
|
| 2789 | }
|
| 2790 | #endif
|
| 2791 | }
|
| 2792 | f2fs_submit_read_bio(F2FS_I_SB(inode), bio, DATA);
|
| 2793 | return ret;
|
| 2794 | }
|
| 2795 |
|
| 2796 | static int f2fs_read_data_folio(struct file *file, struct folio *folio)
|
| 2797 | {
|
| 2798 | struct inode *inode = folio->mapping->host;
|
| 2799 | struct fsverity_info *vi = NULL;
|
| 2800 | int ret;
|
| 2801 |
|
| 2802 | trace_f2fs_readpage(folio, DATA);
|
| 2803 |
|
| 2804 | if (!f2fs_is_compress_backend_ready(inode)) {
|
| 2805 | folio_unlock(folio);
|
| 2806 | return -EOPNOTSUPP;
|
| 2807 | }
|
| 2808 |
|
| 2809 |
|
| 2810 | if (f2fs_has_inline_data(inode)) {
|
| 2811 | ret = f2fs_read_inline_data(inode, folio);
|
| 2812 | if (ret != -EAGAIN)
|
| 2813 | return ret;
|
| 2814 | }
|
| 2815 |
|
| 2816 | vi = f2fs_need_verity(inode, folio->index);
|
| 2817 | if (vi)
|
| 2818 | fsverity_readahead(vi, folio->index, folio_nr_pages(folio));
|
| 2819 | return f2fs_mpage_readpages(inode, vi, NULL, folio);
|
| 2820 | }
|
| 2821 |
|
| 2822 | static void f2fs_readahead(struct readahead_control *rac)
|
| 2823 | {
|
| 2824 | struct inode *inode = rac->mapping->host;
|
| 2825 | struct fsverity_info *vi = NULL;
|
| 2826 |
|
| 2827 | trace_f2fs_readpages(inode, readahead_index(rac), readahead_count(rac));
|
| 2828 |
|
| 2829 | if (!f2fs_is_compress_backend_ready(inode))
|
| 2830 | return;
|
| 2831 |
|
| 2832 |
|
| 2833 | if (f2fs_has_inline_data(inode))
|
| 2834 | return;
|
| 2835 |
|
| 2836 | vi = f2fs_need_verity(inode, readahead_index(rac));
|
| 2837 | if (vi)
|
| 2838 | fsverity_readahead(vi, readahead_index(rac),
|
| 2839 | readahead_count(rac));
|
| 2840 | f2fs_mpage_readpages(inode, vi, rac, NULL);
|
| 2841 | }
|
| 2842 |
|
| 2843 | static inline bool check_inplace_update_policy(struct inode *inode,
|
| 2844 | struct f2fs_io_info *fio)
|
| 2845 | {
|
| 2846 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
| 2847 |
|
| 2848 | if (IS_F2FS_IPU_HONOR_OPU_WRITE(sbi) &&
|
| 2849 | is_inode_flag_set(inode, FI_OPU_WRITE))
|
| 2850 | return false;
|
| 2851 | if (IS_F2FS_IPU_FORCE(sbi))
|
| 2852 | return true;
|
| 2853 | if (IS_F2FS_IPU_SSR(sbi) && f2fs_need_SSR(sbi))
|
| 2854 | return true;
|
| 2855 | if (IS_F2FS_IPU_UTIL(sbi) && utilization(sbi) > SM_I(sbi)->min_ipu_util)
|
| 2856 | return true;
|
| 2857 | if (IS_F2FS_IPU_SSR_UTIL(sbi) && f2fs_need_SSR(sbi) &&
|
| 2858 | utilization(sbi) > SM_I(sbi)->min_ipu_util)
|
| 2859 | return true;
|
| 2860 |
|
| 2861 | |
| 2862 | * IPU for rewrite async pages
|
| 2863 |
|
| 2864 | if (IS_F2FS_IPU_ASYNC(sbi) && fio && fio->op == REQ_OP_WRITE &&
|
| 2865 | !(fio->op_flags & REQ_SYNC) && !IS_ENCRYPTED(inode))
|
| 2866 | return true;
|
| 2867 |
|
| 2868 |
|
| 2869 | if (IS_F2FS_IPU_FSYNC(sbi) && is_inode_flag_set(inode, FI_NEED_IPU))
|
| 2870 | return true;
|
| 2871 |
|
| 2872 | if (unlikely(fio && is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
|
| 2873 | !f2fs_is_checkpointed_data(sbi, fio->old_blkaddr)))
|
| 2874 | return true;
|
| 2875 |
|
| 2876 | return false;
|
| 2877 | }
|
| 2878 |
|
| 2879 | bool f2fs_should_update_inplace(struct inode *inode, struct f2fs_io_info *fio)
|
| 2880 | {
|
| 2881 |
|
| 2882 | if (is_inode_flag_set(inode, FI_ALIGNED_WRITE))
|
| 2883 | return false;
|
| 2884 |
|
| 2885 | if (f2fs_is_pinned_file(inode))
|
| 2886 | return true;
|
| 2887 |
|
| 2888 |
|
| 2889 | if (file_is_cold(inode) && !is_inode_flag_set(inode, FI_OPU_WRITE))
|
| 2890 | return true;
|
| 2891 |
|
| 2892 | return check_inplace_update_policy(inode, fio);
|
| 2893 | }
|
| 2894 |
|
| 2895 | bool f2fs_should_update_outplace(struct inode *inode, struct f2fs_io_info *fio)
|
| 2896 | {
|
| 2897 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
| 2898 |
|
| 2899 |
|
| 2900 | if (f2fs_is_pinned_file(inode))
|
| 2901 | return false;
|
| 2902 | if (fio && is_sbi_flag_set(sbi, SBI_NEED_FSCK))
|
| 2903 | return true;
|
| 2904 | if (f2fs_lfs_mode(sbi))
|
| 2905 | return true;
|
| 2906 | if (S_ISDIR(inode->i_mode))
|
| 2907 | return true;
|
| 2908 | if (IS_NOQUOTA(inode))
|
| 2909 | return true;
|
| 2910 | if (f2fs_used_in_atomic_write(inode))
|
| 2911 | return true;
|
| 2912 |
|
| 2913 | if (f2fs_compressed_file(inode) &&
|
| 2914 | F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER &&
|
| 2915 | is_inode_flag_set(inode, FI_ENABLE_COMPRESS))
|
| 2916 | return true;
|
| 2917 |
|
| 2918 |
|
| 2919 | if (is_inode_flag_set(inode, FI_ALIGNED_WRITE))
|
| 2920 | return true;
|
| 2921 |
|
| 2922 | if (is_inode_flag_set(inode, FI_OPU_WRITE))
|
| 2923 | return true;
|
| 2924 |
|
| 2925 | if (fio) {
|
| 2926 | if (page_private_gcing(fio->page))
|
| 2927 | return true;
|
| 2928 | if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
|
| 2929 | f2fs_is_checkpointed_data(sbi, fio->old_blkaddr)))
|
| 2930 | return true;
|
| 2931 | }
|
| 2932 | return false;
|
| 2933 | }
|
| 2934 |
|
| 2935 | static inline bool need_inplace_update(struct f2fs_io_info *fio)
|
| 2936 | {
|
| 2937 | struct inode *inode = fio_inode(fio);
|
| 2938 |
|
| 2939 | if (f2fs_should_update_outplace(inode, fio))
|
| 2940 | return false;
|
| 2941 |
|
| 2942 | return f2fs_should_update_inplace(inode, fio);
|
| 2943 | }
|
| 2944 |
|
| 2945 | int f2fs_do_write_data_page(struct f2fs_io_info *fio)
|
| 2946 | {
|
| 2947 | struct folio *folio = fio->folio;
|
| 2948 | struct inode *inode = folio->mapping->host;
|
| 2949 | struct dnode_of_data dn;
|
| 2950 | struct node_info ni;
|
| 2951 | struct f2fs_lock_context lc;
|
| 2952 | bool ipu_force = false;
|
| 2953 | bool atomic_commit;
|
| 2954 | int err = 0;
|
| 2955 |
|
| 2956 |
|
| 2957 | atomic_commit = f2fs_is_atomic_file(inode) &&
|
| 2958 | folio_test_f2fs_atomic(folio);
|
| 2959 | if (atomic_commit)
|
| 2960 | set_new_dnode(&dn, F2FS_I(inode)->cow_inode, NULL, NULL, 0);
|
| 2961 | else
|
| 2962 | set_new_dnode(&dn, inode, NULL, NULL, 0);
|
| 2963 |
|
| 2964 | if (need_inplace_update(fio) &&
|
| 2965 | f2fs_lookup_read_extent_cache_block(inode, folio->index,
|
| 2966 | &fio->old_blkaddr)) {
|
| 2967 | if (!f2fs_is_valid_blkaddr(fio->sbi, fio->old_blkaddr,
|
| 2968 | DATA_GENERIC_ENHANCE))
|
| 2969 | return -EFSCORRUPTED;
|
| 2970 |
|
| 2971 | ipu_force = true;
|
| 2972 | fio->need_lock = LOCK_DONE;
|
| 2973 | goto got_it;
|
| 2974 | }
|
| 2975 |
|
| 2976 | if (is_sbi_flag_set(fio->sbi, SBI_ENABLE_CHECKPOINT) &&
|
| 2977 | time_to_inject(fio->sbi, FAULT_SKIP_WRITE))
|
| 2978 | return -EINVAL;
|
| 2979 |
|
| 2980 |
|
| 2981 | if (fio->need_lock == LOCK_REQ && !f2fs_trylock_op(fio->sbi, &lc))
|
| 2982 | return -EAGAIN;
|
| 2983 |
|
| 2984 | err = f2fs_get_dnode_of_data(&dn, folio->index, LOOKUP_NODE);
|
| 2985 | if (err)
|
| 2986 | goto out;
|
| 2987 |
|
| 2988 | fio->old_blkaddr = dn.data_blkaddr;
|
| 2989 |
|
| 2990 |
|
| 2991 | if (fio->old_blkaddr == NULL_ADDR) {
|
| 2992 | folio_clear_uptodate(folio);
|
| 2993 | folio_clear_f2fs_gcing(folio);
|
| 2994 | goto out_writepage;
|
| 2995 | }
|
| 2996 | got_it:
|
| 2997 | if (__is_valid_data_blkaddr(fio->old_blkaddr) &&
|
| 2998 | !f2fs_is_valid_blkaddr(fio->sbi, fio->old_blkaddr,
|
| 2999 | DATA_GENERIC_ENHANCE)) {
|
| 3000 | err = -EFSCORRUPTED;
|
| 3001 | goto out_writepage;
|
| 3002 | }
|
| 3003 |
|
| 3004 |
|
| 3005 | if (fio->meta_gc)
|
| 3006 | f2fs_wait_on_block_writeback(inode, fio->old_blkaddr);
|
| 3007 |
|
| 3008 | |
| 3009 | * If current allocation needs SSR,
|
| 3010 | * it had better in-place writes for updated data.
|
| 3011 |
|
| 3012 | if (ipu_force ||
|
| 3013 | (__is_valid_data_blkaddr(fio->old_blkaddr) &&
|
| 3014 | need_inplace_update(fio))) {
|
| 3015 | folio_start_writeback(folio);
|
| 3016 | f2fs_put_dnode(&dn);
|
| 3017 | if (fio->need_lock == LOCK_REQ)
|
| 3018 | f2fs_unlock_op(fio->sbi, &lc);
|
| 3019 | err = f2fs_inplace_write_data(fio);
|
| 3020 | if (err)
|
| 3021 | folio_end_writeback(folio);
|
| 3022 | else
|
| 3023 | set_inode_flag(inode, FI_UPDATE_WRITE);
|
| 3024 | trace_f2fs_do_write_data_page(folio, IPU);
|
| 3025 | return err;
|
| 3026 | }
|
| 3027 |
|
| 3028 | if (fio->need_lock == LOCK_RETRY) {
|
| 3029 | if (!f2fs_trylock_op(fio->sbi, &lc)) {
|
| 3030 | err = -EAGAIN;
|
| 3031 | goto out_writepage;
|
| 3032 | }
|
| 3033 | fio->need_lock = LOCK_REQ;
|
| 3034 | }
|
| 3035 |
|
| 3036 | err = f2fs_get_node_info(fio->sbi, dn.nid, &ni, false);
|
| 3037 | if (err)
|
| 3038 | goto out_writepage;
|
| 3039 |
|
| 3040 | fio->version = ni.version;
|
| 3041 |
|
| 3042 | folio_start_writeback(folio);
|
| 3043 |
|
| 3044 | if (fio->compr_blocks && fio->old_blkaddr == COMPRESS_ADDR)
|
| 3045 | f2fs_i_compr_blocks_update(inode, fio->compr_blocks - 1, false);
|
| 3046 |
|
| 3047 |
|
| 3048 | f2fs_outplace_write_data(&dn, fio);
|
| 3049 | trace_f2fs_do_write_data_page(folio, OPU);
|
| 3050 | set_inode_flag(inode, FI_APPEND_WRITE);
|
| 3051 | if (atomic_commit)
|
| 3052 | folio_clear_f2fs_atomic(folio);
|
| 3053 | out_writepage:
|
| 3054 | f2fs_put_dnode(&dn);
|
| 3055 | out:
|
| 3056 | if (fio->need_lock == LOCK_REQ)
|
| 3057 | f2fs_unlock_op(fio->sbi, &lc);
|
| 3058 | return err;
|
| 3059 | }
|
| 3060 |
|
| 3061 | int f2fs_write_single_data_page(struct folio *folio, int *submitted,
|
| 3062 | struct bio **bio,
|
| 3063 | sector_t *last_block,
|
| 3064 | struct writeback_control *wbc,
|
| 3065 | enum iostat_type io_type,
|
| 3066 | int compr_blocks,
|
| 3067 | bool allow_balance)
|
| 3068 | {
|
| 3069 | struct inode *inode = folio->mapping->host;
|
| 3070 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
| 3071 | loff_t i_size = i_size_read(inode);
|
| 3072 | const pgoff_t end_index = ((unsigned long long)i_size)
|
| 3073 | >> PAGE_SHIFT;
|
| 3074 | loff_t psize = (loff_t)(folio->index + 1) << PAGE_SHIFT;
|
| 3075 | unsigned offset = 0;
|
| 3076 | bool need_balance_fs = false;
|
| 3077 | bool quota_inode = IS_NOQUOTA(inode);
|
| 3078 | int err = 0;
|
| 3079 | struct f2fs_io_info fio = {
|
| 3080 | .sbi = sbi,
|
| 3081 | .ino = inode->i_ino,
|
| 3082 | .type = DATA,
|
| 3083 | .op = REQ_OP_WRITE,
|
| 3084 | .op_flags = wbc_to_write_flags(wbc),
|
| 3085 | .old_blkaddr = NULL_ADDR,
|
| 3086 | .folio = folio,
|
| 3087 | .encrypted_page = NULL,
|
| 3088 | .submitted = 0,
|
| 3089 | .compr_blocks = compr_blocks,
|
| 3090 | .need_lock = compr_blocks ? LOCK_DONE : LOCK_RETRY,
|
| 3091 | .meta_gc = f2fs_meta_inode_gc_required(inode) ? 1 : 0,
|
| 3092 | .io_type = io_type,
|
| 3093 | .io_wbc = wbc,
|
| 3094 | .bio = bio,
|
| 3095 | .last_block = last_block,
|
| 3096 | };
|
| 3097 |
|
| 3098 | trace_f2fs_writepage(folio, DATA);
|
| 3099 |
|
| 3100 |
|
| 3101 | if (unlikely(f2fs_cp_error(sbi))) {
|
| 3102 | mapping_set_error(folio->mapping, -EIO);
|
| 3103 | |
| 3104 | * don't drop any dirty dentry pages for keeping lastest
|
| 3105 | * directory structure.
|
| 3106 |
|
| 3107 | if (S_ISDIR(inode->i_mode) &&
|
| 3108 | !is_sbi_flag_set(sbi, SBI_IS_CLOSE))
|
| 3109 | goto redirty_out;
|
| 3110 |
|
| 3111 |
|
| 3112 | if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_READONLY)
|
| 3113 | goto redirty_out;
|
| 3114 | goto out;
|
| 3115 | }
|
| 3116 |
|
| 3117 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
|
| 3118 | goto redirty_out;
|
| 3119 |
|
| 3120 | if (folio->index < end_index ||
|
| 3121 | f2fs_verity_in_progress(inode) ||
|
| 3122 | compr_blocks)
|
| 3123 | goto write;
|
| 3124 |
|
| 3125 | |
| 3126 | * If the offset is out-of-range of file size,
|
| 3127 | * this page does not have to be written to disk.
|
| 3128 |
|
| 3129 | offset = i_size & (PAGE_SIZE - 1);
|
| 3130 | if ((folio->index >= end_index + 1) || !offset)
|
| 3131 | goto out;
|
| 3132 |
|
| 3133 | folio_zero_segment(folio, offset, folio_size(folio));
|
| 3134 | write:
|
| 3135 |
|
| 3136 | if (S_ISDIR(inode->i_mode) || quota_inode) {
|
| 3137 | struct f2fs_lock_context lc;
|
| 3138 |
|
| 3139 | |
| 3140 | * We need to wait for node_write to avoid block allocation during
|
| 3141 | * checkpoint. This can only happen to quota writes which can cause
|
| 3142 | * the below discard race condition.
|
| 3143 |
|
| 3144 | if (quota_inode)
|
| 3145 | f2fs_down_read_trace(&sbi->node_write, &lc);
|
| 3146 |
|
| 3147 | fio.need_lock = LOCK_DONE;
|
| 3148 | err = f2fs_do_write_data_page(&fio);
|
| 3149 |
|
| 3150 | if (quota_inode)
|
| 3151 | f2fs_up_read_trace(&sbi->node_write, &lc);
|
| 3152 |
|
| 3153 | goto done;
|
| 3154 | }
|
| 3155 |
|
| 3156 | need_balance_fs = true;
|
| 3157 | err = -EAGAIN;
|
| 3158 | if (f2fs_has_inline_data(inode)) {
|
| 3159 | err = f2fs_write_inline_data(inode, folio);
|
| 3160 | if (!err)
|
| 3161 | goto out;
|
| 3162 | }
|
| 3163 |
|
| 3164 | if (err == -EAGAIN) {
|
| 3165 | err = f2fs_do_write_data_page(&fio);
|
| 3166 | if (err == -EAGAIN) {
|
| 3167 | f2fs_bug_on(sbi, compr_blocks);
|
| 3168 | fio.need_lock = LOCK_REQ;
|
| 3169 | err = f2fs_do_write_data_page(&fio);
|
| 3170 | }
|
| 3171 | }
|
| 3172 |
|
| 3173 | if (err) {
|
| 3174 | file_set_keep_isize(inode);
|
| 3175 | } else {
|
| 3176 | spin_lock(&F2FS_I(inode)->i_size_lock);
|
| 3177 | if (F2FS_I(inode)->last_disk_size < psize)
|
| 3178 | F2FS_I(inode)->last_disk_size = psize;
|
| 3179 | spin_unlock(&F2FS_I(inode)->i_size_lock);
|
| 3180 | }
|
| 3181 |
|
| 3182 | done:
|
| 3183 | if (err && err != -ENOENT)
|
| 3184 | goto redirty_out;
|
| 3185 |
|
| 3186 | out:
|
| 3187 | inode_dec_dirty_pages(inode);
|
| 3188 | if (err) {
|
| 3189 | folio_clear_uptodate(folio);
|
| 3190 | folio_clear_f2fs_gcing(folio);
|
| 3191 | }
|
| 3192 | folio_unlock(folio);
|
| 3193 | if (!S_ISDIR(inode->i_mode) && !IS_NOQUOTA(inode) &&
|
| 3194 | !F2FS_I(inode)->wb_task && allow_balance)
|
| 3195 | f2fs_balance_fs(sbi, need_balance_fs);
|
| 3196 |
|
| 3197 | if (unlikely(f2fs_cp_error(sbi))) {
|
| 3198 | f2fs_submit_merged_write(sbi, DATA);
|
| 3199 | if (bio && *bio)
|
| 3200 | f2fs_submit_merged_ipu_write(sbi, bio, NULL);
|
| 3201 | submitted = NULL;
|
| 3202 | }
|
| 3203 |
|
| 3204 | if (submitted)
|
| 3205 | *submitted = fio.submitted;
|
| 3206 |
|
| 3207 | return 0;
|
| 3208 |
|
| 3209 | redirty_out:
|
| 3210 | folio_redirty_for_writepage(wbc, folio);
|
| 3211 | |
| 3212 | * pageout() in MM translates EAGAIN, so calls handle_write_error()
|
| 3213 | * -> mapping_set_error() -> set_bit(AS_EIO, ...).
|
| 3214 | * file_write_and_wait_range() will see EIO error, which is critical
|
| 3215 | * to return value of fsync() followed by atomic_write failure to user.
|
| 3216 |
|
| 3217 | folio_unlock(folio);
|
| 3218 | if (!err)
|
| 3219 | return 1;
|
| 3220 | return err;
|
| 3221 | }
|
| 3222 |
|
| 3223 | |
| 3224 | * This function was copied from write_cache_pages from mm/page-writeback.c.
|
| 3225 | * The major change is making write step of cold data page separately from
|
| 3226 | * warm/hot data page.
|
| 3227 |
|
| 3228 | static int f2fs_write_cache_pages(struct address_space *mapping,
|
| 3229 | struct writeback_control *wbc,
|
| 3230 | enum iostat_type io_type)
|
| 3231 | {
|
| 3232 | int ret = 0;
|
| 3233 | int done = 0, retry = 0;
|
| 3234 | struct page *pages_local[F2FS_ONSTACK_PAGES];
|
| 3235 | struct page **pages = pages_local;
|
| 3236 | struct folio_batch fbatch;
|
| 3237 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping);
|
| 3238 | struct bio *bio = NULL;
|
| 3239 | sector_t last_block;
|
| 3240 | #ifdef CONFIG_F2FS_FS_COMPRESSION
|
| 3241 | struct inode *inode = mapping->host;
|
| 3242 | struct compress_ctx cc = {
|
| 3243 | .inode = inode,
|
| 3244 | .log_cluster_size = F2FS_I(inode)->i_log_cluster_size,
|
| 3245 | .cluster_size = F2FS_I(inode)->i_cluster_size,
|
| 3246 | .cluster_idx = NULL_CLUSTER,
|
| 3247 | .rpages = NULL,
|
| 3248 | .nr_rpages = 0,
|
| 3249 | .cpages = NULL,
|
| 3250 | .valid_nr_cpages = 0,
|
| 3251 | .rbuf = NULL,
|
| 3252 | .cbuf = NULL,
|
| 3253 | .rlen = PAGE_SIZE * F2FS_I(inode)->i_cluster_size,
|
| 3254 | .private = NULL,
|
| 3255 | };
|
| 3256 | #endif
|
| 3257 | int nr_folios, p, idx;
|
| 3258 | int nr_pages;
|
| 3259 | unsigned int max_pages = F2FS_ONSTACK_PAGES;
|
| 3260 | pgoff_t index;
|
| 3261 | pgoff_t end;
|
| 3262 | pgoff_t done_index;
|
| 3263 | int range_whole = 0;
|
| 3264 | xa_mark_t tag;
|
| 3265 | int nwritten = 0;
|
| 3266 | int submitted = 0;
|
| 3267 | int i;
|
| 3268 |
|
| 3269 | #ifdef CONFIG_F2FS_FS_COMPRESSION
|
| 3270 | if (f2fs_compressed_file(inode) &&
|
| 3271 | 1 << cc.log_cluster_size > F2FS_ONSTACK_PAGES) {
|
| 3272 | pages = f2fs_kzalloc(sbi, sizeof(struct page *) <<
|
| 3273 | cc.log_cluster_size, GFP_NOFS | __GFP_NOFAIL);
|
| 3274 | max_pages = 1 << cc.log_cluster_size;
|
| 3275 | }
|
| 3276 | #endif
|
| 3277 |
|
| 3278 | folio_batch_init(&fbatch);
|
| 3279 |
|
| 3280 | if (get_dirty_pages(mapping->host) <=
|
| 3281 | SM_I(F2FS_M_SB(mapping))->min_hot_blocks)
|
| 3282 | set_inode_flag(mapping->host, FI_HOT_DATA);
|
| 3283 | else
|
| 3284 | clear_inode_flag(mapping->host, FI_HOT_DATA);
|
| 3285 |
|
| 3286 | if (wbc->range_cyclic) {
|
| 3287 | index = mapping->writeback_index;
|
| 3288 | end = -1;
|
| 3289 | } else {
|
| 3290 | index = wbc->range_start >> PAGE_SHIFT;
|
| 3291 | end = wbc->range_end >> PAGE_SHIFT;
|
| 3292 | if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
|
| 3293 | range_whole = 1;
|
| 3294 | }
|
| 3295 | tag = wbc_to_tag(wbc);
|
| 3296 | retry:
|
| 3297 | retry = 0;
|
| 3298 | if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
|
| 3299 | tag_pages_for_writeback(mapping, index, end);
|
| 3300 | done_index = index;
|
| 3301 | while (!done && !retry && (index <= end)) {
|
| 3302 | nr_pages = 0;
|
| 3303 | again:
|
| 3304 | nr_folios = filemap_get_folios_tag(mapping, &index, end,
|
| 3305 | tag, &fbatch);
|
| 3306 | if (nr_folios == 0) {
|
| 3307 | if (nr_pages)
|
| 3308 | goto write;
|
| 3309 | break;
|
| 3310 | }
|
| 3311 |
|
| 3312 | for (i = 0; i < nr_folios; i++) {
|
| 3313 | struct folio *folio = fbatch.folios[i];
|
| 3314 |
|
| 3315 | idx = 0;
|
| 3316 | p = folio_nr_pages(folio);
|
| 3317 | add_more:
|
| 3318 | pages[nr_pages] = folio_page(folio, idx);
|
| 3319 | folio_get(folio);
|
| 3320 | if (++nr_pages == max_pages) {
|
| 3321 | index = folio->index + idx + 1;
|
| 3322 | folio_batch_release(&fbatch);
|
| 3323 | goto write;
|
| 3324 | }
|
| 3325 | if (++idx < p)
|
| 3326 | goto add_more;
|
| 3327 | }
|
| 3328 | folio_batch_release(&fbatch);
|
| 3329 | goto again;
|
| 3330 | write:
|
| 3331 | for (i = 0; i < nr_pages; i++) {
|
| 3332 | struct page *page = pages[i];
|
| 3333 | struct folio *folio = page_folio(page);
|
| 3334 | bool need_readd;
|
| 3335 | readd:
|
| 3336 | need_readd = false;
|
| 3337 | #ifdef CONFIG_F2FS_FS_COMPRESSION
|
| 3338 | if (f2fs_compressed_file(inode)) {
|
| 3339 | void *fsdata = NULL;
|
| 3340 | struct page *pagep;
|
| 3341 | int ret2;
|
| 3342 |
|
| 3343 | ret = f2fs_init_compress_ctx(&cc);
|
| 3344 | if (ret) {
|
| 3345 | done = 1;
|
| 3346 | break;
|
| 3347 | }
|
| 3348 |
|
| 3349 | if (!f2fs_cluster_can_merge_page(&cc,
|
| 3350 | folio->index)) {
|
| 3351 | ret = f2fs_write_multi_pages(&cc,
|
| 3352 | &submitted, wbc, io_type);
|
| 3353 | if (!ret)
|
| 3354 | need_readd = true;
|
| 3355 | goto result;
|
| 3356 | }
|
| 3357 |
|
| 3358 | if (unlikely(f2fs_cp_error(sbi)))
|
| 3359 | goto lock_folio;
|
| 3360 |
|
| 3361 | if (!f2fs_cluster_is_empty(&cc))
|
| 3362 | goto lock_folio;
|
| 3363 |
|
| 3364 | if (f2fs_all_cluster_page_ready(&cc,
|
| 3365 | pages, i, nr_pages, true))
|
| 3366 | goto lock_folio;
|
| 3367 |
|
| 3368 | ret2 = f2fs_prepare_compress_overwrite(
|
| 3369 | inode, &pagep,
|
| 3370 | folio->index, &fsdata);
|
| 3371 | if (ret2 < 0) {
|
| 3372 | ret = ret2;
|
| 3373 | done = 1;
|
| 3374 | break;
|
| 3375 | } else if (ret2 &&
|
| 3376 | (!f2fs_compress_write_end(inode,
|
| 3377 | fsdata, folio->index, 1) ||
|
| 3378 | !f2fs_all_cluster_page_ready(&cc,
|
| 3379 | pages, i, nr_pages,
|
| 3380 | false))) {
|
| 3381 | retry = 1;
|
| 3382 | break;
|
| 3383 | }
|
| 3384 | }
|
| 3385 | #endif
|
| 3386 |
|
| 3387 | if (atomic_read(&sbi->wb_sync_req[DATA]) &&
|
| 3388 | wbc->sync_mode == WB_SYNC_NONE) {
|
| 3389 | done = 1;
|
| 3390 | break;
|
| 3391 | }
|
| 3392 | #ifdef CONFIG_F2FS_FS_COMPRESSION
|
| 3393 | lock_folio:
|
| 3394 | #endif
|
| 3395 | done_index = folio->index;
|
| 3396 | retry_write:
|
| 3397 | folio_lock(folio);
|
| 3398 |
|
| 3399 | if (unlikely(folio->mapping != mapping)) {
|
| 3400 | continue_unlock:
|
| 3401 | folio_unlock(folio);
|
| 3402 | continue;
|
| 3403 | }
|
| 3404 |
|
| 3405 | if (!folio_test_dirty(folio)) {
|
| 3406 |
|
| 3407 | goto continue_unlock;
|
| 3408 | }
|
| 3409 |
|
| 3410 | if (folio_test_writeback(folio)) {
|
| 3411 | if (wbc->sync_mode == WB_SYNC_NONE)
|
| 3412 | goto continue_unlock;
|
| 3413 | f2fs_folio_wait_writeback(folio, DATA, true, true);
|
| 3414 | }
|
| 3415 |
|
| 3416 | if (!folio_clear_dirty_for_io(folio))
|
| 3417 | goto continue_unlock;
|
| 3418 |
|
| 3419 | #ifdef CONFIG_F2FS_FS_COMPRESSION
|
| 3420 | if (f2fs_compressed_file(inode)) {
|
| 3421 | folio_get(folio);
|
| 3422 | f2fs_compress_ctx_add_page(&cc, folio);
|
| 3423 | continue;
|
| 3424 | }
|
| 3425 | #endif
|
| 3426 | submitted = 0;
|
| 3427 | ret = f2fs_write_single_data_page(folio,
|
| 3428 | &submitted, &bio, &last_block,
|
| 3429 | wbc, io_type, 0, true);
|
| 3430 | #ifdef CONFIG_F2FS_FS_COMPRESSION
|
| 3431 | result:
|
| 3432 | #endif
|
| 3433 | nwritten += submitted;
|
| 3434 | wbc->nr_to_write -= submitted;
|
| 3435 |
|
| 3436 | if (unlikely(ret)) {
|
| 3437 | |
| 3438 | * keep nr_to_write, since vfs uses this to
|
| 3439 | * get # of written pages.
|
| 3440 |
|
| 3441 | if (ret == 1) {
|
| 3442 | ret = 0;
|
| 3443 | goto next;
|
| 3444 | } else if (ret == -EAGAIN) {
|
| 3445 | ret = 0;
|
| 3446 | if (wbc->sync_mode == WB_SYNC_ALL) {
|
| 3447 | f2fs_schedule_timeout(
|
| 3448 | DEFAULT_SCHEDULE_TIMEOUT);
|
| 3449 | goto retry_write;
|
| 3450 | }
|
| 3451 | goto next;
|
| 3452 | }
|
| 3453 | done_index = folio_next_index(folio);
|
| 3454 | done = 1;
|
| 3455 | break;
|
| 3456 | }
|
| 3457 |
|
| 3458 | if (wbc->nr_to_write <= 0 &&
|
| 3459 | wbc->sync_mode == WB_SYNC_NONE) {
|
| 3460 | done = 1;
|
| 3461 | break;
|
| 3462 | }
|
| 3463 | next:
|
| 3464 | if (need_readd)
|
| 3465 | goto readd;
|
| 3466 | }
|
| 3467 | release_pages(pages, nr_pages);
|
| 3468 | cond_resched();
|
| 3469 | }
|
| 3470 | #ifdef CONFIG_F2FS_FS_COMPRESSION
|
| 3471 |
|
| 3472 | if (f2fs_compressed_file(inode) && !f2fs_cluster_is_empty(&cc)) {
|
| 3473 | ret = f2fs_write_multi_pages(&cc, &submitted, wbc, io_type);
|
| 3474 | nwritten += submitted;
|
| 3475 | wbc->nr_to_write -= submitted;
|
| 3476 | if (ret) {
|
| 3477 | done = 1;
|
| 3478 | retry = 0;
|
| 3479 | }
|
| 3480 | }
|
| 3481 | if (f2fs_compressed_file(inode))
|
| 3482 | f2fs_destroy_compress_ctx(&cc, false);
|
| 3483 | #endif
|
| 3484 | if (retry) {
|
| 3485 | index = 0;
|
| 3486 | end = -1;
|
| 3487 | goto retry;
|
| 3488 | }
|
| 3489 | if (wbc->range_cyclic && !done)
|
| 3490 | done_index = 0;
|
| 3491 | if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
|
| 3492 | mapping->writeback_index = done_index;
|
| 3493 |
|
| 3494 | if (nwritten)
|
| 3495 | f2fs_submit_merged_write_cond(F2FS_M_SB(mapping), mapping->host,
|
| 3496 | NULL, 0, DATA);
|
| 3497 |
|
| 3498 | if (bio)
|
| 3499 | f2fs_submit_merged_ipu_write(sbi, &bio, NULL);
|
| 3500 |
|
| 3501 | #ifdef CONFIG_F2FS_FS_COMPRESSION
|
| 3502 | if (pages != pages_local)
|
| 3503 | kfree(pages);
|
| 3504 | #endif
|
| 3505 |
|
| 3506 | return ret;
|
| 3507 | }
|
| 3508 |
|
| 3509 | static inline bool __should_serialize_io(struct inode *inode,
|
| 3510 | struct writeback_control *wbc)
|
| 3511 | {
|
| 3512 |
|
| 3513 | if (F2FS_I(inode)->wb_task)
|
| 3514 | return false;
|
| 3515 |
|
| 3516 | if (!S_ISREG(inode->i_mode))
|
| 3517 | return false;
|
| 3518 | if (IS_NOQUOTA(inode))
|
| 3519 | return false;
|
| 3520 |
|
| 3521 | if (f2fs_is_pinned_file(inode))
|
| 3522 | return false;
|
| 3523 | if (f2fs_need_compress_data(inode))
|
| 3524 | return true;
|
| 3525 | if (wbc->sync_mode != WB_SYNC_ALL)
|
| 3526 | return true;
|
| 3527 | if (get_dirty_pages(inode) >= SM_I(F2FS_I_SB(inode))->min_seq_blocks)
|
| 3528 | return true;
|
| 3529 | return false;
|
| 3530 | }
|
| 3531 |
|
| 3532 | static inline void account_writeback(struct inode *inode, bool inc)
|
| 3533 | {
|
| 3534 | if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
|
| 3535 | return;
|
| 3536 |
|
| 3537 | f2fs_down_read(&F2FS_I(inode)->i_sem);
|
| 3538 | if (inc)
|
| 3539 | atomic_inc(&F2FS_I(inode)->writeback);
|
| 3540 | else
|
| 3541 | atomic_dec(&F2FS_I(inode)->writeback);
|
| 3542 | f2fs_up_read(&F2FS_I(inode)->i_sem);
|
| 3543 | }
|
| 3544 |
|
| 3545 | static inline void update_skipped_write(struct f2fs_sb_info *sbi,
|
| 3546 | struct writeback_control *wbc)
|
| 3547 | {
|
| 3548 | long skipped = wbc->pages_skipped;
|
| 3549 |
|
| 3550 | if (is_sbi_flag_set(sbi, SBI_ENABLE_CHECKPOINT) && skipped &&
|
| 3551 | wbc->sync_mode == WB_SYNC_ALL)
|
| 3552 | atomic_add(skipped, &sbi->nr_pages[F2FS_SKIPPED_WRITE]);
|
| 3553 | }
|
| 3554 |
|
| 3555 | static int __f2fs_write_data_pages(struct address_space *mapping,
|
| 3556 | struct writeback_control *wbc,
|
| 3557 | enum iostat_type io_type)
|
| 3558 | {
|
| 3559 | struct inode *inode = mapping->host;
|
| 3560 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
| 3561 | struct blk_plug plug;
|
| 3562 | int ret;
|
| 3563 | bool locked = false;
|
| 3564 |
|
| 3565 |
|
| 3566 | if (!get_dirty_pages(inode) && wbc->sync_mode == WB_SYNC_NONE)
|
| 3567 | return 0;
|
| 3568 |
|
| 3569 |
|
| 3570 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
|
| 3571 | goto skip_write;
|
| 3572 |
|
| 3573 | if ((S_ISDIR(inode->i_mode) || IS_NOQUOTA(inode)) &&
|
| 3574 | wbc->sync_mode == WB_SYNC_NONE &&
|
| 3575 | get_dirty_pages(inode) < nr_pages_to_skip(sbi, DATA) &&
|
| 3576 | f2fs_available_free_memory(sbi, DIRTY_DENTS))
|
| 3577 | goto skip_write;
|
| 3578 |
|
| 3579 |
|
| 3580 | if (is_inode_flag_set(inode, FI_SKIP_WRITES))
|
| 3581 | goto skip_write;
|
| 3582 |
|
| 3583 | trace_f2fs_writepages(mapping->host, wbc, DATA);
|
| 3584 |
|
| 3585 |
|
| 3586 | if (wbc->sync_mode == WB_SYNC_ALL)
|
| 3587 | atomic_inc(&sbi->wb_sync_req[DATA]);
|
| 3588 | else if (atomic_read(&sbi->wb_sync_req[DATA])) {
|
| 3589 |
|
| 3590 | if (current->plug)
|
| 3591 | blk_finish_plug(current->plug);
|
| 3592 | goto skip_write;
|
| 3593 | }
|
| 3594 |
|
| 3595 | if (__should_serialize_io(inode, wbc)) {
|
| 3596 | mutex_lock(&sbi->writepages);
|
| 3597 | locked = true;
|
| 3598 | }
|
| 3599 |
|
| 3600 | account_writeback(inode, true);
|
| 3601 |
|
| 3602 | blk_start_plug(&plug);
|
| 3603 | ret = f2fs_write_cache_pages(mapping, wbc, io_type);
|
| 3604 | blk_finish_plug(&plug);
|
| 3605 |
|
| 3606 | account_writeback(inode, false);
|
| 3607 |
|
| 3608 | if (locked)
|
| 3609 | mutex_unlock(&sbi->writepages);
|
| 3610 |
|
| 3611 | if (wbc->sync_mode == WB_SYNC_ALL)
|
| 3612 | atomic_dec(&sbi->wb_sync_req[DATA]);
|
| 3613 | |
| 3614 | * if some pages were truncated, we cannot guarantee its mapping->host
|
| 3615 | * to detect pending bios.
|
| 3616 |
|
| 3617 |
|
| 3618 | f2fs_remove_dirty_inode(inode);
|
| 3619 |
|
| 3620 | |
| 3621 | * f2fs_write_cache_pages() has retry logic for EAGAIN case which is
|
| 3622 | * common when racing w/ checkpoint, so only update skipped write
|
| 3623 | * when ret is non-zero.
|
| 3624 |
|
| 3625 | if (ret)
|
| 3626 | update_skipped_write(sbi, wbc);
|
| 3627 | return ret;
|
| 3628 |
|
| 3629 | skip_write:
|
| 3630 | wbc->pages_skipped += get_dirty_pages(inode);
|
| 3631 | update_skipped_write(sbi, wbc);
|
| 3632 | trace_f2fs_writepages(mapping->host, wbc, DATA);
|
| 3633 | return 0;
|
| 3634 | }
|
| 3635 |
|
| 3636 | static int f2fs_write_data_pages(struct address_space *mapping,
|
| 3637 | struct writeback_control *wbc)
|
| 3638 | {
|
| 3639 | struct inode *inode = mapping->host;
|
| 3640 |
|
| 3641 | return __f2fs_write_data_pages(mapping, wbc,
|
| 3642 | F2FS_I(inode)->cp_task == current ?
|
| 3643 | FS_CP_DATA_IO : FS_DATA_IO);
|
| 3644 | }
|
| 3645 |
|
| 3646 | void f2fs_write_failed(struct inode *inode, loff_t to)
|
| 3647 | {
|
| 3648 | loff_t i_size = i_size_read(inode);
|
| 3649 |
|
| 3650 | if (IS_NOQUOTA(inode))
|
| 3651 | return;
|
| 3652 |
|
| 3653 |
|
| 3654 | if (to > i_size && !f2fs_verity_in_progress(inode)) {
|
| 3655 | f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
|
| 3656 | filemap_invalidate_lock(inode->i_mapping);
|
| 3657 |
|
| 3658 | truncate_pagecache(inode, i_size);
|
| 3659 | f2fs_truncate_blocks(inode, i_size, true);
|
| 3660 |
|
| 3661 | filemap_invalidate_unlock(inode->i_mapping);
|
| 3662 | f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
|
| 3663 | }
|
| 3664 | }
|
| 3665 |
|
| 3666 | static int prepare_write_begin(struct f2fs_sb_info *sbi,
|
| 3667 | struct folio *folio, loff_t pos, unsigned int len,
|
| 3668 | block_t *blk_addr, bool *node_changed)
|
| 3669 | {
|
| 3670 | struct inode *inode = folio->mapping->host;
|
| 3671 | pgoff_t index = folio->index;
|
| 3672 | struct dnode_of_data dn;
|
| 3673 | struct f2fs_lock_context lc;
|
| 3674 | struct folio *ifolio;
|
| 3675 | bool locked = false;
|
| 3676 | int flag = F2FS_GET_BLOCK_PRE_AIO;
|
| 3677 | int err = 0;
|
| 3678 |
|
| 3679 | if (!f2fs_has_inline_data(inode) && !f2fs_compressed_file(inode) &&
|
| 3680 | (pos & PAGE_MASK) < i_size_read(inode) &&
|
| 3681 | f2fs_lookup_read_extent_cache_block(inode, index, blk_addr))
|
| 3682 | return 0;
|
| 3683 |
|
| 3684 | |
| 3685 | * If a whole page is being written and we already preallocated all the
|
| 3686 | * blocks, then there is no need to get a block address now.
|
| 3687 |
|
| 3688 | if (len == PAGE_SIZE && is_inode_flag_set(inode, FI_PREALLOCATED_ALL))
|
| 3689 | return 0;
|
| 3690 |
|
| 3691 |
|
| 3692 | if (f2fs_has_inline_data(inode)) {
|
| 3693 | if (pos + len > MAX_INLINE_DATA(inode))
|
| 3694 | flag = F2FS_GET_BLOCK_DEFAULT;
|
| 3695 | f2fs_map_lock(sbi, &lc, flag);
|
| 3696 | locked = true;
|
| 3697 | } else if ((pos & PAGE_MASK) >= i_size_read(inode)) {
|
| 3698 | f2fs_map_lock(sbi, &lc, flag);
|
| 3699 | locked = true;
|
| 3700 | }
|
| 3701 |
|
| 3702 | restart:
|
| 3703 |
|
| 3704 | ifolio = f2fs_get_inode_folio(sbi, inode->i_ino);
|
| 3705 | if (IS_ERR(ifolio)) {
|
| 3706 | err = PTR_ERR(ifolio);
|
| 3707 | goto unlock_out;
|
| 3708 | }
|
| 3709 |
|
| 3710 | set_new_dnode(&dn, inode, ifolio, ifolio, 0);
|
| 3711 |
|
| 3712 | if (f2fs_has_inline_data(inode)) {
|
| 3713 | if (pos + len <= MAX_INLINE_DATA(inode)) {
|
| 3714 | f2fs_do_read_inline_data(folio, ifolio);
|
| 3715 | set_inode_flag(inode, FI_DATA_EXIST);
|
| 3716 | if (inode->i_nlink)
|
| 3717 | folio_set_f2fs_inline(ifolio);
|
| 3718 | goto out;
|
| 3719 | }
|
| 3720 | err = f2fs_convert_inline_folio(&dn, folio);
|
| 3721 | if (err || dn.data_blkaddr != NULL_ADDR)
|
| 3722 | goto out;
|
| 3723 | }
|
| 3724 |
|
| 3725 | if (!f2fs_lookup_read_extent_cache_block(inode, index,
|
| 3726 | &dn.data_blkaddr)) {
|
| 3727 | if (IS_DEVICE_ALIASING(inode)) {
|
| 3728 | err = -ENODATA;
|
| 3729 | goto out;
|
| 3730 | }
|
| 3731 |
|
| 3732 | if (locked) {
|
| 3733 | err = f2fs_reserve_block(&dn, index);
|
| 3734 | goto out;
|
| 3735 | }
|
| 3736 |
|
| 3737 |
|
| 3738 | err = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
|
| 3739 | if (!err && dn.data_blkaddr != NULL_ADDR)
|
| 3740 | goto out;
|
| 3741 | f2fs_put_dnode(&dn);
|
| 3742 | f2fs_map_lock(sbi, &lc, F2FS_GET_BLOCK_PRE_AIO);
|
| 3743 | WARN_ON(flag != F2FS_GET_BLOCK_PRE_AIO);
|
| 3744 | locked = true;
|
| 3745 | goto restart;
|
| 3746 | }
|
| 3747 | out:
|
| 3748 | if (!err) {
|
| 3749 |
|
| 3750 | *blk_addr = dn.data_blkaddr;
|
| 3751 | *node_changed = dn.node_changed;
|
| 3752 | }
|
| 3753 | f2fs_put_dnode(&dn);
|
| 3754 | unlock_out:
|
| 3755 | if (locked)
|
| 3756 | f2fs_map_unlock(sbi, &lc, flag);
|
| 3757 | return err;
|
| 3758 | }
|
| 3759 |
|
| 3760 | static int __find_data_block(struct inode *inode, pgoff_t index,
|
| 3761 | block_t *blk_addr)
|
| 3762 | {
|
| 3763 | struct dnode_of_data dn;
|
| 3764 | struct folio *ifolio;
|
| 3765 | int err = 0;
|
| 3766 |
|
| 3767 | ifolio = f2fs_get_inode_folio(F2FS_I_SB(inode), inode->i_ino);
|
| 3768 | if (IS_ERR(ifolio))
|
| 3769 | return PTR_ERR(ifolio);
|
| 3770 |
|
| 3771 | set_new_dnode(&dn, inode, ifolio, ifolio, 0);
|
| 3772 |
|
| 3773 | if (!f2fs_lookup_read_extent_cache_block(inode, index,
|
| 3774 | &dn.data_blkaddr)) {
|
| 3775 |
|
| 3776 | err = f2fs_get_dnode_of_data(&dn, index, LOOKUP_NODE);
|
| 3777 | if (err) {
|
| 3778 | dn.data_blkaddr = NULL_ADDR;
|
| 3779 | err = 0;
|
| 3780 | }
|
| 3781 | }
|
| 3782 | *blk_addr = dn.data_blkaddr;
|
| 3783 | f2fs_put_dnode(&dn);
|
| 3784 | return err;
|
| 3785 | }
|
| 3786 |
|
| 3787 | static int __reserve_data_block(struct inode *inode, pgoff_t index,
|
| 3788 | block_t *blk_addr, bool *node_changed)
|
| 3789 | {
|
| 3790 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
| 3791 | struct dnode_of_data dn;
|
| 3792 | struct f2fs_lock_context lc;
|
| 3793 | struct folio *ifolio;
|
| 3794 | int err = 0;
|
| 3795 |
|
| 3796 | f2fs_map_lock(sbi, &lc, F2FS_GET_BLOCK_PRE_AIO);
|
| 3797 |
|
| 3798 | ifolio = f2fs_get_inode_folio(sbi, inode->i_ino);
|
| 3799 | if (IS_ERR(ifolio)) {
|
| 3800 | err = PTR_ERR(ifolio);
|
| 3801 | goto unlock_out;
|
| 3802 | }
|
| 3803 | set_new_dnode(&dn, inode, ifolio, ifolio, 0);
|
| 3804 |
|
| 3805 | if (!f2fs_lookup_read_extent_cache_block(dn.inode, index,
|
| 3806 | &dn.data_blkaddr))
|
| 3807 | err = f2fs_reserve_block(&dn, index);
|
| 3808 |
|
| 3809 | *blk_addr = dn.data_blkaddr;
|
| 3810 | *node_changed = dn.node_changed;
|
| 3811 | f2fs_put_dnode(&dn);
|
| 3812 |
|
| 3813 | unlock_out:
|
| 3814 | f2fs_map_unlock(sbi, &lc, F2FS_GET_BLOCK_PRE_AIO);
|
| 3815 | return err;
|
| 3816 | }
|
| 3817 |
|
| 3818 | static int prepare_atomic_write_begin(struct f2fs_sb_info *sbi,
|
| 3819 | struct folio *folio, loff_t pos, unsigned int len,
|
| 3820 | block_t *blk_addr, bool *node_changed)
|
| 3821 | {
|
| 3822 | struct inode *inode = folio->mapping->host;
|
| 3823 | struct inode *cow_inode = F2FS_I(inode)->cow_inode;
|
| 3824 | pgoff_t index = folio->index;
|
| 3825 | int err = 0;
|
| 3826 | block_t ori_blk_addr = NULL_ADDR;
|
| 3827 | bool cow_has_reserved_block = false;
|
| 3828 |
|
| 3829 |
|
| 3830 | if ((pos & PAGE_MASK) >= i_size_read(inode))
|
| 3831 | goto reserve_block;
|
| 3832 |
|
| 3833 |
|
| 3834 | err = __find_data_block(cow_inode, index, blk_addr);
|
| 3835 | if (err)
|
| 3836 | return err;
|
| 3837 |
|
| 3838 | if (__is_valid_data_blkaddr(*blk_addr))
|
| 3839 | return 0;
|
| 3840 |
|
| 3841 | if (*blk_addr == NEW_ADDR)
|
| 3842 | cow_has_reserved_block = true;
|
| 3843 |
|
| 3844 | if (is_inode_flag_set(inode, FI_ATOMIC_REPLACE))
|
| 3845 | goto reserve_block;
|
| 3846 |
|
| 3847 |
|
| 3848 | err = __find_data_block(inode, index, &ori_blk_addr);
|
| 3849 | if (err)
|
| 3850 | return err;
|
| 3851 |
|
| 3852 | reserve_block:
|
| 3853 |
|
| 3854 | if (!cow_has_reserved_block) {
|
| 3855 | err = __reserve_data_block(cow_inode, index, blk_addr,
|
| 3856 | node_changed);
|
| 3857 | if (err)
|
| 3858 | return err;
|
| 3859 | inc_atomic_write_cnt(inode);
|
| 3860 | }
|
| 3861 |
|
| 3862 | if (ori_blk_addr != NULL_ADDR)
|
| 3863 | *blk_addr = ori_blk_addr;
|
| 3864 | return 0;
|
| 3865 | }
|
| 3866 |
|
| 3867 | static int f2fs_write_begin(const struct kiocb *iocb,
|
| 3868 | struct address_space *mapping,
|
| 3869 | loff_t pos, unsigned len, struct folio **foliop,
|
| 3870 | void **fsdata)
|
| 3871 | {
|
| 3872 | struct inode *inode = mapping->host;
|
| 3873 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
| 3874 | struct folio *folio;
|
| 3875 | pgoff_t index = pos >> PAGE_SHIFT;
|
| 3876 | bool need_balance = false;
|
| 3877 | block_t blkaddr = NULL_ADDR;
|
| 3878 | int err = 0;
|
| 3879 |
|
| 3880 | trace_f2fs_write_begin(inode, pos, len);
|
| 3881 |
|
| 3882 | if (!f2fs_is_checkpoint_ready(sbi)) {
|
| 3883 | err = -ENOSPC;
|
| 3884 | goto fail;
|
| 3885 | }
|
| 3886 |
|
| 3887 | |
| 3888 | * We should check this at this moment to avoid deadlock on inode page
|
| 3889 | * and #0 page. The locking rule for inline_data conversion should be:
|
| 3890 | * folio_lock(folio #0) -> folio_lock(inode_page)
|
| 3891 |
|
| 3892 | if (index != 0) {
|
| 3893 | err = f2fs_convert_inline_inode(inode);
|
| 3894 | if (err)
|
| 3895 | goto fail;
|
| 3896 | }
|
| 3897 |
|
| 3898 | #ifdef CONFIG_F2FS_FS_COMPRESSION
|
| 3899 | if (f2fs_compressed_file(inode)) {
|
| 3900 | int ret;
|
| 3901 | struct page *page;
|
| 3902 |
|
| 3903 | *fsdata = NULL;
|
| 3904 |
|
| 3905 | if (len == PAGE_SIZE && !(f2fs_is_atomic_file(inode)))
|
| 3906 | goto repeat;
|
| 3907 |
|
| 3908 | ret = f2fs_prepare_compress_overwrite(inode, &page,
|
| 3909 | index, fsdata);
|
| 3910 | if (ret < 0) {
|
| 3911 | err = ret;
|
| 3912 | goto fail;
|
| 3913 | } else if (ret) {
|
| 3914 | *foliop = page_folio(page);
|
| 3915 | return 0;
|
| 3916 | }
|
| 3917 | }
|
| 3918 | #endif
|
| 3919 |
|
| 3920 | repeat:
|
| 3921 | |
| 3922 | * Do not use FGP_STABLE to avoid deadlock.
|
| 3923 | * Will wait that below with our IO control.
|
| 3924 |
|
| 3925 | folio = f2fs_filemap_get_folio(mapping, index,
|
| 3926 | FGP_LOCK | FGP_WRITE | FGP_CREAT,
|
| 3927 | mapping_gfp_mask(mapping));
|
| 3928 | if (IS_ERR(folio)) {
|
| 3929 | err = PTR_ERR(folio);
|
| 3930 | goto fail;
|
| 3931 | }
|
| 3932 |
|
| 3933 |
|
| 3934 |
|
| 3935 | *foliop = folio;
|
| 3936 |
|
| 3937 | if (f2fs_is_atomic_file(inode))
|
| 3938 | err = prepare_atomic_write_begin(sbi, folio, pos, len,
|
| 3939 | &blkaddr, &need_balance);
|
| 3940 | else
|
| 3941 | err = prepare_write_begin(sbi, folio, pos, len,
|
| 3942 | &blkaddr, &need_balance);
|
| 3943 | if (err)
|
| 3944 | goto put_folio;
|
| 3945 |
|
| 3946 | if (need_balance && !IS_NOQUOTA(inode) &&
|
| 3947 | has_not_enough_free_secs(sbi, 0, 0)) {
|
| 3948 | folio_unlock(folio);
|
| 3949 | f2fs_balance_fs(sbi, true);
|
| 3950 | folio_lock(folio);
|
| 3951 | if (folio->mapping != mapping) {
|
| 3952 |
|
| 3953 | folio_unlock(folio);
|
| 3954 | folio_put(folio);
|
| 3955 | goto repeat;
|
| 3956 | }
|
| 3957 | }
|
| 3958 |
|
| 3959 | f2fs_folio_wait_writeback(folio, DATA, false, true);
|
| 3960 |
|
| 3961 | if (len == folio_size(folio) || folio_test_uptodate(folio))
|
| 3962 | return 0;
|
| 3963 |
|
| 3964 | if (!(pos & (PAGE_SIZE - 1)) && (pos + len) >= i_size_read(inode) &&
|
| 3965 | !f2fs_verity_in_progress(inode)) {
|
| 3966 | folio_zero_segment(folio, len, folio_size(folio));
|
| 3967 | return 0;
|
| 3968 | }
|
| 3969 |
|
| 3970 | if (blkaddr == NEW_ADDR) {
|
| 3971 | folio_zero_segment(folio, 0, folio_size(folio));
|
| 3972 | folio_mark_uptodate(folio);
|
| 3973 | } else {
|
| 3974 | if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
|
| 3975 | DATA_GENERIC_ENHANCE_READ)) {
|
| 3976 | err = -EFSCORRUPTED;
|
| 3977 | goto put_folio;
|
| 3978 | }
|
| 3979 | |
| 3980 | * Although the block may be stored in the COW inode, the folio
|
| 3981 | * belongs to @inode and its data was encrypted (or not) using
|
| 3982 | * @inode's context (see f2fs_set_bio_crypt_ctx()). Read with
|
| 3983 | * @inode so the post-read decryption decision matches the
|
| 3984 | * folio's owner; otherwise an unencrypted @inode whose COW inode
|
| 3985 | * is encrypted hits a NULL ->i_crypt_info on decryption.
|
| 3986 |
|
| 3987 | f2fs_submit_page_read(inode,
|
| 3988 | NULL,
|
| 3989 | folio, blkaddr, 0, true);
|
| 3990 |
|
| 3991 | folio_lock(folio);
|
| 3992 | if (unlikely(folio->mapping != mapping)) {
|
| 3993 | folio_unlock(folio);
|
| 3994 | folio_put(folio);
|
| 3995 | goto repeat;
|
| 3996 | }
|
| 3997 | if (unlikely(!folio_test_uptodate(folio))) {
|
| 3998 | err = -EIO;
|
| 3999 | goto put_folio;
|
| 4000 | }
|
| 4001 | }
|
| 4002 | return 0;
|
| 4003 |
|
| 4004 | put_folio:
|
| 4005 | f2fs_folio_put(folio, true);
|
| 4006 | fail:
|
| 4007 | f2fs_write_failed(inode, pos + len);
|
| 4008 | return err;
|
| 4009 | }
|
| 4010 |
|
| 4011 | static int f2fs_write_end(const struct kiocb *iocb,
|
| 4012 | struct address_space *mapping,
|
| 4013 | loff_t pos, unsigned len, unsigned copied,
|
| 4014 | struct folio *folio, void *fsdata)
|
| 4015 | {
|
| 4016 | struct inode *inode = folio->mapping->host;
|
| 4017 |
|
| 4018 | trace_f2fs_write_end(inode, pos, len, copied);
|
| 4019 |
|
| 4020 | |
| 4021 | * This should be come from len == PAGE_SIZE, and we expect copied
|
| 4022 | * should be PAGE_SIZE. Otherwise, we treat it with zero copied and
|
| 4023 | * let generic_perform_write() try to copy data again through copied=0.
|
| 4024 |
|
| 4025 | if (!folio_test_uptodate(folio)) {
|
| 4026 | if (unlikely(copied != len))
|
| 4027 | copied = 0;
|
| 4028 | else
|
| 4029 | folio_mark_uptodate(folio);
|
| 4030 | }
|
| 4031 |
|
| 4032 | #ifdef CONFIG_F2FS_FS_COMPRESSION
|
| 4033 |
|
| 4034 | if (f2fs_compressed_file(inode) && fsdata) {
|
| 4035 | f2fs_compress_write_end(inode, fsdata, folio->index, copied);
|
| 4036 | f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
|
| 4037 |
|
| 4038 | if (pos + copied > i_size_read(inode) &&
|
| 4039 | !f2fs_verity_in_progress(inode))
|
| 4040 | f2fs_i_size_write(inode, pos + copied);
|
| 4041 | return copied;
|
| 4042 | }
|
| 4043 | #endif
|
| 4044 |
|
| 4045 | if (!copied)
|
| 4046 | goto unlock_out;
|
| 4047 |
|
| 4048 | folio_mark_dirty(folio);
|
| 4049 |
|
| 4050 | if (f2fs_is_atomic_file(inode))
|
| 4051 | folio_set_f2fs_atomic(folio);
|
| 4052 |
|
| 4053 | if (pos + copied > i_size_read(inode) &&
|
| 4054 | !f2fs_verity_in_progress(inode)) {
|
| 4055 | f2fs_i_size_write(inode, pos + copied);
|
| 4056 | if (f2fs_is_atomic_file(inode))
|
| 4057 | f2fs_i_size_write(F2FS_I(inode)->cow_inode,
|
| 4058 | pos + copied);
|
| 4059 | }
|
| 4060 | unlock_out:
|
| 4061 | f2fs_folio_put(folio, true);
|
| 4062 | f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
|
| 4063 | return copied;
|
| 4064 | }
|
| 4065 |
|
| 4066 | void f2fs_invalidate_folio(struct folio *folio, size_t offset, size_t length)
|
| 4067 | {
|
| 4068 | struct inode *inode = folio->mapping->host;
|
| 4069 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
| 4070 |
|
| 4071 | if (inode->i_ino >= F2FS_ROOT_INO(sbi) &&
|
| 4072 | (offset || length != folio_size(folio)))
|
| 4073 | return;
|
| 4074 |
|
| 4075 | if (folio_test_dirty(folio)) {
|
| 4076 | if (inode->i_ino == F2FS_META_INO(sbi)) {
|
| 4077 | dec_page_count(sbi, F2FS_DIRTY_META);
|
| 4078 | } else if (inode->i_ino == F2FS_NODE_INO(sbi)) {
|
| 4079 | dec_page_count(sbi, F2FS_DIRTY_NODES);
|
| 4080 | } else {
|
| 4081 | inode_dec_dirty_pages(inode);
|
| 4082 | f2fs_remove_dirty_inode(inode);
|
| 4083 | }
|
| 4084 | }
|
| 4085 |
|
| 4086 | if (offset || length != folio_size(folio))
|
| 4087 | return;
|
| 4088 |
|
| 4089 | folio_cancel_dirty(folio);
|
| 4090 | ffs_detach_free(folio);
|
| 4091 | }
|
| 4092 |
|
| 4093 | bool f2fs_release_folio(struct folio *folio, gfp_t wait)
|
| 4094 | {
|
| 4095 |
|
| 4096 | if (folio_test_dirty(folio))
|
| 4097 | return false;
|
| 4098 |
|
| 4099 | ffs_detach_free(folio);
|
| 4100 | return true;
|
| 4101 | }
|
| 4102 |
|
| 4103 | static bool f2fs_dirty_data_folio(struct address_space *mapping,
|
| 4104 | struct folio *folio)
|
| 4105 | {
|
| 4106 | struct inode *inode = mapping->host;
|
| 4107 |
|
| 4108 | trace_f2fs_set_page_dirty(folio, DATA);
|
| 4109 |
|
| 4110 | if (!folio_test_uptodate(folio))
|
| 4111 | folio_mark_uptodate(folio);
|
| 4112 | BUG_ON(folio_test_swapcache(folio));
|
| 4113 |
|
| 4114 | if (filemap_dirty_folio(mapping, folio)) {
|
| 4115 | f2fs_update_dirty_folio(inode, folio);
|
| 4116 | return true;
|
| 4117 | }
|
| 4118 | return false;
|
| 4119 | }
|
| 4120 |
|
| 4121 |
|
| 4122 | static sector_t f2fs_bmap_compress(struct inode *inode, sector_t block)
|
| 4123 | {
|
| 4124 | #ifdef CONFIG_F2FS_FS_COMPRESSION
|
| 4125 | struct dnode_of_data dn;
|
| 4126 | sector_t start_idx, blknr = 0;
|
| 4127 | int ret;
|
| 4128 |
|
| 4129 | start_idx = round_down(block, F2FS_I(inode)->i_cluster_size);
|
| 4130 |
|
| 4131 | set_new_dnode(&dn, inode, NULL, NULL, 0);
|
| 4132 | ret = f2fs_get_dnode_of_data(&dn, start_idx, LOOKUP_NODE);
|
| 4133 | if (ret)
|
| 4134 | return 0;
|
| 4135 |
|
| 4136 | if (dn.data_blkaddr != COMPRESS_ADDR) {
|
| 4137 | dn.ofs_in_node += block - start_idx;
|
| 4138 | blknr = f2fs_data_blkaddr(&dn);
|
| 4139 | if (!__is_valid_data_blkaddr(blknr))
|
| 4140 | blknr = 0;
|
| 4141 | }
|
| 4142 |
|
| 4143 | f2fs_put_dnode(&dn);
|
| 4144 | return blknr;
|
| 4145 | #else
|
| 4146 | return 0;
|
| 4147 | #endif
|
| 4148 | }
|
| 4149 |
|
| 4150 |
|
| 4151 | static sector_t f2fs_bmap(struct address_space *mapping, sector_t block)
|
| 4152 | {
|
| 4153 | struct inode *inode = mapping->host;
|
| 4154 | sector_t blknr = 0;
|
| 4155 |
|
| 4156 | if (f2fs_has_inline_data(inode))
|
| 4157 | goto out;
|
| 4158 |
|
| 4159 |
|
| 4160 | if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
|
| 4161 | filemap_write_and_wait(mapping);
|
| 4162 |
|
| 4163 |
|
| 4164 | if (unlikely(block >= max_file_blocks(inode)))
|
| 4165 | goto out;
|
| 4166 |
|
| 4167 | if (f2fs_compressed_file(inode)) {
|
| 4168 | blknr = f2fs_bmap_compress(inode, block);
|
| 4169 | } else {
|
| 4170 | struct f2fs_map_blocks map;
|
| 4171 |
|
| 4172 | memset(&map, 0, sizeof(map));
|
| 4173 | map.m_lblk = block;
|
| 4174 | map.m_len = 1;
|
| 4175 | map.m_next_pgofs = NULL;
|
| 4176 | map.m_seg_type = NO_CHECK_TYPE;
|
| 4177 |
|
| 4178 | if (!f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_BMAP))
|
| 4179 | blknr = map.m_pblk;
|
| 4180 | }
|
| 4181 | out:
|
| 4182 | trace_f2fs_bmap(inode, block, blknr);
|
| 4183 | return blknr;
|
| 4184 | }
|
| 4185 |
|
| 4186 | #ifdef CONFIG_SWAP
|
| 4187 | static int f2fs_migrate_blocks(struct inode *inode, block_t start_blk,
|
| 4188 | unsigned int blkcnt)
|
| 4189 | {
|
| 4190 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
| 4191 | unsigned int blkofs;
|
| 4192 | unsigned int blk_per_sec = BLKS_PER_SEC(sbi);
|
| 4193 | unsigned int end_blk = start_blk + blkcnt - 1;
|
| 4194 | unsigned int secidx = start_blk / blk_per_sec;
|
| 4195 | unsigned int end_sec;
|
| 4196 | int ret = 0;
|
| 4197 |
|
| 4198 | if (!blkcnt)
|
| 4199 | return 0;
|
| 4200 | end_sec = end_blk / blk_per_sec;
|
| 4201 |
|
| 4202 | f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
|
| 4203 | filemap_invalidate_lock(inode->i_mapping);
|
| 4204 |
|
| 4205 | set_inode_flag(inode, FI_ALIGNED_WRITE);
|
| 4206 | set_inode_flag(inode, FI_OPU_WRITE);
|
| 4207 |
|
| 4208 | for (; secidx <= end_sec; secidx++) {
|
| 4209 | unsigned int blkofs_end = secidx == end_sec ?
|
| 4210 | end_blk % blk_per_sec : blk_per_sec - 1;
|
| 4211 |
|
| 4212 | f2fs_down_write(&sbi->pin_sem);
|
| 4213 |
|
| 4214 | ret = f2fs_allocate_pinning_section(sbi);
|
| 4215 | if (ret) {
|
| 4216 | f2fs_up_write(&sbi->pin_sem);
|
| 4217 | break;
|
| 4218 | }
|
| 4219 |
|
| 4220 | set_inode_flag(inode, FI_SKIP_WRITES);
|
| 4221 |
|
| 4222 | for (blkofs = 0; blkofs <= blkofs_end; blkofs++) {
|
| 4223 | struct folio *folio;
|
| 4224 | unsigned int blkidx = secidx * blk_per_sec + blkofs;
|
| 4225 |
|
| 4226 | folio = f2fs_get_lock_data_folio(inode, blkidx, true);
|
| 4227 | if (IS_ERR(folio)) {
|
| 4228 | f2fs_up_write(&sbi->pin_sem);
|
| 4229 | ret = PTR_ERR(folio);
|
| 4230 | goto done;
|
| 4231 | }
|
| 4232 |
|
| 4233 | folio_mark_dirty(folio);
|
| 4234 | f2fs_folio_put(folio, true);
|
| 4235 | }
|
| 4236 |
|
| 4237 | clear_inode_flag(inode, FI_SKIP_WRITES);
|
| 4238 |
|
| 4239 | ret = filemap_fdatawrite(inode->i_mapping);
|
| 4240 |
|
| 4241 | f2fs_up_write(&sbi->pin_sem);
|
| 4242 |
|
| 4243 | if (ret)
|
| 4244 | break;
|
| 4245 | }
|
| 4246 |
|
| 4247 | done:
|
| 4248 | clear_inode_flag(inode, FI_SKIP_WRITES);
|
| 4249 | clear_inode_flag(inode, FI_OPU_WRITE);
|
| 4250 | clear_inode_flag(inode, FI_ALIGNED_WRITE);
|
| 4251 |
|
| 4252 | filemap_invalidate_unlock(inode->i_mapping);
|
| 4253 | f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
|
| 4254 |
|
| 4255 | return ret;
|
| 4256 | }
|
| 4257 |
|
| 4258 | static int check_swap_activate(struct swap_info_struct *sis,
|
| 4259 | struct file *swap_file, sector_t *span)
|
| 4260 | {
|
| 4261 | struct address_space *mapping = swap_file->f_mapping;
|
| 4262 | struct inode *inode = mapping->host;
|
| 4263 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
| 4264 | block_t cur_lblock;
|
| 4265 | block_t last_lblock;
|
| 4266 | block_t pblock;
|
| 4267 | block_t lowest_pblock = -1;
|
| 4268 | block_t highest_pblock = 0;
|
| 4269 | int nr_extents = 0;
|
| 4270 | unsigned int nr_pblocks;
|
| 4271 | unsigned int blks_per_sec = BLKS_PER_SEC(sbi);
|
| 4272 | unsigned int not_aligned = 0;
|
| 4273 | int ret = 0;
|
| 4274 |
|
| 4275 | |
| 4276 | * Map all the blocks into the extent list. This code doesn't try
|
| 4277 | * to be very smart.
|
| 4278 |
|
| 4279 | cur_lblock = 0;
|
| 4280 | last_lblock = F2FS_BYTES_TO_BLK(i_size_read(inode));
|
| 4281 |
|
| 4282 | while (cur_lblock < last_lblock && cur_lblock < sis->max) {
|
| 4283 | struct f2fs_map_blocks map;
|
| 4284 | bool last_extent = false;
|
| 4285 | retry:
|
| 4286 | cond_resched();
|
| 4287 |
|
| 4288 | memset(&map, 0, sizeof(map));
|
| 4289 | map.m_lblk = cur_lblock;
|
| 4290 | map.m_len = last_lblock - cur_lblock;
|
| 4291 | map.m_next_pgofs = NULL;
|
| 4292 | map.m_next_extent = NULL;
|
| 4293 | map.m_seg_type = NO_CHECK_TYPE;
|
| 4294 | map.m_may_create = false;
|
| 4295 |
|
| 4296 | ret = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_FIEMAP);
|
| 4297 | if (ret)
|
| 4298 | goto out;
|
| 4299 |
|
| 4300 |
|
| 4301 | if (!(map.m_flags & F2FS_MAP_FLAGS)) {
|
| 4302 | f2fs_err(sbi, "Swapfile has holes");
|
| 4303 | ret = -EINVAL;
|
| 4304 | goto out;
|
| 4305 | }
|
| 4306 |
|
| 4307 | pblock = map.m_pblk;
|
| 4308 | nr_pblocks = map.m_len;
|
| 4309 |
|
| 4310 | if (!last_extent &&
|
| 4311 | ((pblock - SM_I(sbi)->main_blkaddr) % blks_per_sec ||
|
| 4312 | nr_pblocks % blks_per_sec ||
|
| 4313 | f2fs_is_sequential_zone_area(sbi, pblock))) {
|
| 4314 | not_aligned++;
|
| 4315 |
|
| 4316 | nr_pblocks = roundup(nr_pblocks, blks_per_sec);
|
| 4317 | if (cur_lblock + nr_pblocks > sis->max)
|
| 4318 | nr_pblocks -= blks_per_sec;
|
| 4319 |
|
| 4320 |
|
| 4321 | if (!nr_pblocks) {
|
| 4322 | nr_pblocks = last_lblock - cur_lblock;
|
| 4323 | last_extent = true;
|
| 4324 | }
|
| 4325 |
|
| 4326 | ret = f2fs_migrate_blocks(inode, cur_lblock,
|
| 4327 | nr_pblocks);
|
| 4328 | if (ret) {
|
| 4329 | if (ret == -ENOENT)
|
| 4330 | ret = -EINVAL;
|
| 4331 | goto out;
|
| 4332 | }
|
| 4333 |
|
| 4334 |
|
| 4335 | goto retry;
|
| 4336 | }
|
| 4337 |
|
| 4338 | if (cur_lblock + nr_pblocks >= sis->max)
|
| 4339 | nr_pblocks = sis->max - cur_lblock;
|
| 4340 |
|
| 4341 | if (cur_lblock) {
|
| 4342 | if (pblock < lowest_pblock)
|
| 4343 | lowest_pblock = pblock;
|
| 4344 | if (pblock + nr_pblocks - 1 > highest_pblock)
|
| 4345 | highest_pblock = pblock + nr_pblocks - 1;
|
| 4346 | }
|
| 4347 |
|
| 4348 | |
| 4349 | * We found a PAGE_SIZE-length, PAGE_SIZE-aligned run of blocks
|
| 4350 |
|
| 4351 | ret = add_swap_extent(sis, cur_lblock, nr_pblocks, pblock);
|
| 4352 | if (ret < 0)
|
| 4353 | goto out;
|
| 4354 | nr_extents += ret;
|
| 4355 | cur_lblock += nr_pblocks;
|
| 4356 | }
|
| 4357 | ret = nr_extents;
|
| 4358 | *span = 1 + highest_pblock - lowest_pblock;
|
| 4359 | if (cur_lblock == 0)
|
| 4360 | cur_lblock = 1;
|
| 4361 | sis->max = cur_lblock;
|
| 4362 | sis->pages = cur_lblock - 1;
|
| 4363 | out:
|
| 4364 | if (not_aligned)
|
| 4365 | f2fs_warn(sbi, "Swapfile (%u) is not align to section: 1) creat(), 2) ioctl(F2FS_IOC_SET_PIN_FILE), 3) fallocate(%lu * N)",
|
| 4366 | not_aligned, blks_per_sec * F2FS_BLKSIZE);
|
| 4367 | return ret;
|
| 4368 | }
|
| 4369 |
|
| 4370 | static int f2fs_swap_activate(struct swap_info_struct *sis, struct file *file,
|
| 4371 | sector_t *span)
|
| 4372 | {
|
| 4373 | struct inode *inode = file_inode(file);
|
| 4374 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
| 4375 | int ret;
|
| 4376 |
|
| 4377 | if (!S_ISREG(inode->i_mode))
|
| 4378 | return -EINVAL;
|
| 4379 |
|
| 4380 | if (f2fs_readonly(sbi->sb))
|
| 4381 | return -EROFS;
|
| 4382 |
|
| 4383 | if (f2fs_lfs_mode(sbi) && !f2fs_sb_has_blkzoned(sbi)) {
|
| 4384 | f2fs_err(sbi, "Swapfile not supported in LFS mode");
|
| 4385 | return -EINVAL;
|
| 4386 | }
|
| 4387 |
|
| 4388 | ret = f2fs_convert_inline_inode(inode);
|
| 4389 | if (ret)
|
| 4390 | return ret;
|
| 4391 |
|
| 4392 | if (!f2fs_disable_compressed_file(inode))
|
| 4393 | return -EINVAL;
|
| 4394 |
|
| 4395 | ret = filemap_fdatawrite(inode->i_mapping);
|
| 4396 | if (ret < 0)
|
| 4397 | return ret;
|
| 4398 |
|
| 4399 | f2fs_precache_extents(inode);
|
| 4400 |
|
| 4401 | ret = check_swap_activate(sis, file, span);
|
| 4402 | if (ret < 0)
|
| 4403 | return ret;
|
| 4404 |
|
| 4405 | stat_inc_swapfile_inode(inode);
|
| 4406 | set_inode_flag(inode, FI_PIN_FILE);
|
| 4407 | f2fs_update_time(sbi, REQ_TIME);
|
| 4408 | return ret;
|
| 4409 | }
|
| 4410 |
|
| 4411 | static void f2fs_swap_deactivate(struct file *file)
|
| 4412 | {
|
| 4413 | struct inode *inode = file_inode(file);
|
| 4414 |
|
| 4415 | stat_dec_swapfile_inode(inode);
|
| 4416 | clear_inode_flag(inode, FI_PIN_FILE);
|
| 4417 | }
|
| 4418 | #else
|
| 4419 | static int f2fs_swap_activate(struct swap_info_struct *sis, struct file *file,
|
| 4420 | sector_t *span)
|
| 4421 | {
|
| 4422 | return -EOPNOTSUPP;
|
| 4423 | }
|
| 4424 |
|
| 4425 | static void f2fs_swap_deactivate(struct file *file)
|
| 4426 | {
|
| 4427 | }
|
| 4428 | #endif
|
| 4429 |
|
| 4430 | const struct address_space_operations f2fs_dblock_aops = {
|
| 4431 | .read_folio = f2fs_read_data_folio,
|
| 4432 | .readahead = f2fs_readahead,
|
| 4433 | .writepages = f2fs_write_data_pages,
|
| 4434 | .write_begin = f2fs_write_begin,
|
| 4435 | .write_end = f2fs_write_end,
|
| 4436 | .dirty_folio = f2fs_dirty_data_folio,
|
| 4437 | .migrate_folio = filemap_migrate_folio,
|
| 4438 | .invalidate_folio = f2fs_invalidate_folio,
|
| 4439 | .release_folio = f2fs_release_folio,
|
| 4440 | .bmap = f2fs_bmap,
|
| 4441 | .swap_activate = f2fs_swap_activate,
|
| 4442 | .swap_deactivate = f2fs_swap_deactivate,
|
| 4443 | };
|
| 4444 |
|
| 4445 | void f2fs_clear_page_cache_dirty_tag(struct folio *folio)
|
| 4446 | {
|
| 4447 | struct address_space *mapping = folio->mapping;
|
| 4448 | unsigned long flags;
|
| 4449 |
|
| 4450 | xa_lock_irqsave(&mapping->i_pages, flags);
|
| 4451 | __xa_clear_mark(&mapping->i_pages, folio->index,
|
| 4452 | PAGECACHE_TAG_DIRTY);
|
| 4453 | xa_unlock_irqrestore(&mapping->i_pages, flags);
|
| 4454 | }
|
| 4455 |
|
| 4456 | int __init f2fs_init_post_read_processing(void)
|
| 4457 | {
|
| 4458 | bio_post_read_ctx_cache =
|
| 4459 | kmem_cache_create("f2fs_bio_post_read_ctx",
|
| 4460 | sizeof(struct bio_post_read_ctx), 0, 0, NULL);
|
| 4461 | if (!bio_post_read_ctx_cache)
|
| 4462 | goto fail;
|
| 4463 | bio_post_read_ctx_pool =
|
| 4464 | mempool_create_slab_pool(NUM_PREALLOC_POST_READ_CTXS,
|
| 4465 | bio_post_read_ctx_cache);
|
| 4466 | if (!bio_post_read_ctx_pool)
|
| 4467 | goto fail_free_cache;
|
| 4468 | return 0;
|
| 4469 |
|
| 4470 | fail_free_cache:
|
| 4471 | kmem_cache_destroy(bio_post_read_ctx_cache);
|
| 4472 | fail:
|
| 4473 | return -ENOMEM;
|
| 4474 | }
|
| 4475 |
|
| 4476 | void f2fs_destroy_post_read_processing(void)
|
| 4477 | {
|
| 4478 | mempool_destroy(bio_post_read_ctx_pool);
|
| 4479 | kmem_cache_destroy(bio_post_read_ctx_cache);
|
| 4480 | }
|
| 4481 |
|
| 4482 | int f2fs_init_wq(struct f2fs_sb_info *sbi)
|
| 4483 | {
|
| 4484 | sbi->wq = alloc_workqueue("f2fs_wq", WQ_UNBOUND | WQ_HIGHPRI,
|
| 4485 | num_online_cpus());
|
| 4486 | if (!sbi->wq)
|
| 4487 | return -ENOMEM;
|
| 4488 |
|
| 4489 | sbi->evict_wq = alloc_workqueue("f2fs_evict_wq",
|
| 4490 | WQ_UNBOUND | WQ_HIGHPRI, num_online_cpus());
|
| 4491 | if (!sbi->evict_wq) {
|
| 4492 | destroy_workqueue(sbi->wq);
|
| 4493 | return -ENOMEM;
|
| 4494 | }
|
| 4495 | return 0;
|
| 4496 | }
|
| 4497 |
|
| 4498 | void f2fs_destroy_wq(struct f2fs_sb_info *sbi)
|
| 4499 | {
|
| 4500 | if (sbi->wq)
|
| 4501 | destroy_workqueue(sbi->wq);
|
| 4502 | if (sbi->evict_wq)
|
| 4503 | destroy_workqueue(sbi->evict_wq);
|
| 4504 | }
|
| 4505 |
|
| 4506 | int __init f2fs_init_bio_entry_cache(void)
|
| 4507 | {
|
| 4508 | bio_entry_slab = f2fs_kmem_cache_create("f2fs_bio_entry_slab",
|
| 4509 | sizeof(struct bio_entry));
|
| 4510 |
|
| 4511 | if (!bio_entry_slab)
|
| 4512 | return -ENOMEM;
|
| 4513 |
|
| 4514 | ffs_entry_slab = f2fs_kmem_cache_create("f2fs_ffs_slab",
|
| 4515 | sizeof(struct f2fs_folio_state));
|
| 4516 |
|
| 4517 | if (!ffs_entry_slab) {
|
| 4518 | kmem_cache_destroy(bio_entry_slab);
|
| 4519 | return -ENOMEM;
|
| 4520 | }
|
| 4521 |
|
| 4522 | return 0;
|
| 4523 | }
|
| 4524 |
|
| 4525 | void f2fs_destroy_bio_entry_cache(void)
|
| 4526 | {
|
| 4527 | kmem_cache_destroy(bio_entry_slab);
|
| 4528 | kmem_cache_destroy(ffs_entry_slab);
|
| 4529 | }
|
| 4530 |
|
| 4531 | static int f2fs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
|
| 4532 | unsigned int flags, struct iomap *iomap,
|
| 4533 | struct iomap *srcmap)
|
| 4534 | {
|
| 4535 | struct f2fs_map_blocks map = { NULL, };
|
| 4536 | pgoff_t next_pgofs = 0;
|
| 4537 | int err;
|
| 4538 |
|
| 4539 | map.m_lblk = F2FS_BYTES_TO_BLK(offset);
|
| 4540 | map.m_len = F2FS_BYTES_TO_BLK(offset + length - 1) - map.m_lblk + 1;
|
| 4541 | map.m_next_pgofs = &next_pgofs;
|
| 4542 | map.m_seg_type = f2fs_rw_hint_to_seg_type(F2FS_I_SB(inode),
|
| 4543 | inode->i_write_hint);
|
| 4544 | if (flags & IOMAP_WRITE && iomap->private) {
|
| 4545 | map.m_last_pblk = (unsigned long)iomap->private;
|
| 4546 | iomap->private = NULL;
|
| 4547 | }
|
| 4548 |
|
| 4549 | |
| 4550 | * If the blocks being overwritten are already allocated,
|
| 4551 | * f2fs_map_lock and f2fs_balance_fs are not necessary.
|
| 4552 |
|
| 4553 | if ((flags & IOMAP_WRITE) &&
|
| 4554 | !__f2fs_overwrite_io(inode, offset, length, true))
|
| 4555 | map.m_may_create = true;
|
| 4556 |
|
| 4557 | err = f2fs_map_blocks(inode, &map, F2FS_GET_BLOCK_DIO);
|
| 4558 | if (err)
|
| 4559 | return err;
|
| 4560 |
|
| 4561 | iomap->offset = F2FS_BLK_TO_BYTES(map.m_lblk);
|
| 4562 |
|
| 4563 | |
| 4564 | * Sometimes I/O to an encrypted file has to be broken up to guarantee
|
| 4565 | * DUN contiguity. Handle this by limiting the length of the mapping
|
| 4566 | * returned.
|
| 4567 |
|
| 4568 | map.m_len = fscrypt_limit_io_blocks(inode, map.m_lblk, map.m_len);
|
| 4569 |
|
| 4570 | |
| 4571 | * We should never see delalloc or compressed extents here based on
|
| 4572 | * prior flushing and checks.
|
| 4573 |
|
| 4574 | if (WARN_ON_ONCE(map.m_pblk == COMPRESS_ADDR))
|
| 4575 | return -EINVAL;
|
| 4576 |
|
| 4577 | if (map.m_flags & F2FS_MAP_MAPPED) {
|
| 4578 | if (WARN_ON_ONCE(map.m_pblk == NEW_ADDR))
|
| 4579 | return -EINVAL;
|
| 4580 |
|
| 4581 | iomap->length = F2FS_BLK_TO_BYTES(map.m_len);
|
| 4582 | iomap->type = IOMAP_MAPPED;
|
| 4583 | iomap->flags |= IOMAP_F_MERGED;
|
| 4584 | iomap->bdev = map.m_bdev;
|
| 4585 | iomap->addr = F2FS_BLK_TO_BYTES(map.m_pblk);
|
| 4586 |
|
| 4587 | if (flags & IOMAP_WRITE && map.m_last_pblk)
|
| 4588 | iomap->private = (void *)map.m_last_pblk;
|
| 4589 | } else {
|
| 4590 | if (flags & IOMAP_WRITE)
|
| 4591 | return -ENOTBLK;
|
| 4592 |
|
| 4593 | if (map.m_pblk == NULL_ADDR) {
|
| 4594 | iomap->length = F2FS_BLK_TO_BYTES(next_pgofs) -
|
| 4595 | iomap->offset;
|
| 4596 | iomap->type = IOMAP_HOLE;
|
| 4597 | } else if (map.m_pblk == NEW_ADDR) {
|
| 4598 | iomap->length = F2FS_BLK_TO_BYTES(map.m_len);
|
| 4599 | iomap->type = IOMAP_UNWRITTEN;
|
| 4600 | } else {
|
| 4601 | f2fs_bug_on(F2FS_I_SB(inode), 1);
|
| 4602 | }
|
| 4603 | iomap->addr = IOMAP_NULL_ADDR;
|
| 4604 | }
|
| 4605 |
|
| 4606 | if (map.m_flags & F2FS_MAP_NEW)
|
| 4607 | iomap->flags |= IOMAP_F_NEW;
|
| 4608 | if ((inode_state_read_once(inode) & I_DIRTY_DATASYNC) ||
|
| 4609 | offset + length > i_size_read(inode))
|
| 4610 | iomap->flags |= IOMAP_F_DIRTY;
|
| 4611 |
|
| 4612 | return 0;
|
| 4613 | }
|
| 4614 |
|
| 4615 | static DEFINE_IOMAP_ITER_NEXT(f2fs_iomap_next, f2fs_iomap_begin);
|
| 4616 |
|
| 4617 | const struct iomap_ops f2fs_iomap_ops = {
|
| 4618 | .iomap_next = f2fs_iomap_next,
|
| 4619 | };
|