fs/gfs2/meta_io.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
   4  * Copyright (C) 2004-2008 Red Hat, Inc.  All rights reserved.
   5  */
   6
   7 #include <linux/sched.h>
   8 #include <linux/slab.h>
   9 #include <linux/spinlock.h>
  10 #include <linux/completion.h>
  11 #include <linux/buffer_head.h>
  12 #include <linux/mm.h>
  13 #include <linux/pagemap.h>
  14 #include <linux/writeback.h>
  15 #include <linux/swap.h>
  16 #include <linux/delay.h>
  17 #include <linux/bio.h>
  18 #include <linux/gfs2_ondisk.h>
  19
  20 #include "gfs2.h"
  21 #include "incore.h"
  22 #include "glock.h"
  23 #include "glops.h"
  24 #include "inode.h"
  25 #include "log.h"
  26 #include "lops.h"
  27 #include "meta_io.h"
  28 #include "rgrp.h"
  29 #include "trans.h"
  30 #include "util.h"
  31 #include "trace_gfs2.h"
  32
  33 static void gfs2_aspace_write_folio(struct folio *folio,
  34                 struct writeback_control *wbc)
  35 {
  36         struct buffer_head *bh, *head;
  37         int nr_underway = 0;
  38         blk_opf_t write_flags = REQ_META | REQ_PRIO | wbc_to_write_flags(wbc);
  39
  40         BUG_ON(!folio_test_locked(folio));
  41
  42         head = folio_buffers(folio);
  43         bh = head;
  44
  45         do {
  46                 if (!buffer_mapped(bh))
  47                         continue;
  48                 /*
  49                  * If it's a fully non-blocking write attempt and we cannot
  50                  * lock the buffer then redirty the page.  Note that this can
  51                  * potentially cause a busy-wait loop from flusher thread and kswapd
  52                  * activity, but those code paths have their own higher-level
  53                  * throttling.
  54                  */
  55                 if (wbc->sync_mode != WB_SYNC_NONE) {
  56                         lock_buffer(bh);
  57                 } else if (!trylock_buffer(bh)) {
  58                         folio_redirty_for_writepage(wbc, folio);
  59                         continue;
  60                 }
  61                 if (test_clear_buffer_dirty(bh)) {
  62                         mark_buffer_async_write(bh);
  63                 } else {
  64                         unlock_buffer(bh);
  65                 }
  66         } while ((bh = bh->b_this_page) != head);
  67
  68         /*
  69          * The folio and its buffers are protected from truncation by
  70          * the writeback flag, so we can drop the bh refcounts early.
  71          */
  72         BUG_ON(folio_test_writeback(folio));
  73         folio_start_writeback(folio);
  74
  75         do {
  76                 struct buffer_head *next = bh->b_this_page;
  77                 if (buffer_async_write(bh)) {
  78                         submit_bh(REQ_OP_WRITE | write_flags, bh);
  79                         nr_underway++;
  80                 }
  81                 bh = next;
  82         } while (bh != head);
  83         folio_unlock(folio);
  84
  85         if (nr_underway == 0)
  86                 folio_end_writeback(folio);
  87 }
  88
  89 static int gfs2_aspace_writepages(struct address_space *mapping,
  90                 struct writeback_control *wbc)
  91 {
  92         struct folio *folio = NULL;
  93         int error;
  94
  95         while ((folio = writeback_iter(mapping, wbc, folio, &error)))
  96                 gfs2_aspace_write_folio(folio, wbc);
  97
  98         return error;
  99 }
 100
 101 const struct address_space_operations gfs2_meta_aops = {
 102         .dirty_folio    = block_dirty_folio,
 103         .invalidate_folio = block_invalidate_folio,
 104         .writepages = gfs2_aspace_writepages,
 105         .release_folio = gfs2_release_folio,
 106 };
 107
 108 const struct address_space_operations gfs2_rgrp_aops = {
 109         .dirty_folio    = block_dirty_folio,
 110         .invalidate_folio = block_invalidate_folio,
 111         .writepages = gfs2_aspace_writepages,
 112         .release_folio = gfs2_release_folio,
 113 };
 114
 115 /**
 116  * gfs2_getbuf - Get a buffer with a given address space
 117  * @gl: the glock
 118  * @blkno: the block number (filesystem scope)
 119  * @create: 1 if the buffer should be created
 120  *
 121  * Returns: the buffer
 122  */
 123
 124 struct buffer_head *gfs2_getbuf(struct gfs2_glock *gl, u64 blkno, int create)
 125 {
 126         struct address_space *mapping = gfs2_glock2aspace(gl);
 127         struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
 128         struct folio *folio;
 129         struct buffer_head *bh;
 130         unsigned int shift;
 131         unsigned long index;
 132         unsigned int bufnum;
 133
 134         if (mapping == NULL)
 135                 mapping = &sdp->sd_aspace;
 136
 137         shift = PAGE_SHIFT - sdp->sd_sb.sb_bsize_shift;
 138         index = blkno >> shift;             /* convert block to page */
 139         bufnum = blkno - (index << shift);  /* block buf index within page */
 140
 141         if (create) {
 142                 folio = __filemap_get_folio(mapping, index,
 143                                 FGP_LOCK | FGP_ACCESSED | FGP_CREAT,
 144                                 mapping_gfp_mask(mapping) | __GFP_NOFAIL);
 145                 bh = folio_buffers(folio);
 146                 if (!bh)
 147                         bh = create_empty_buffers(folio,
 148                                 sdp->sd_sb.sb_bsize, 0);
 149         } else {
 150                 folio = __filemap_get_folio(mapping, index,
 151                                 FGP_LOCK | FGP_ACCESSED, 0);
 152                 if (IS_ERR(folio))
 153                         return NULL;
 154                 bh = folio_buffers(folio);
 155         }
 156
 157         if (!bh)
 158                 goto out_unlock;
 159
 160         bh = get_nth_bh(bh, bufnum);
 161         if (!buffer_mapped(bh))
 162                 map_bh(bh, sdp->sd_vfs, blkno);
 163
 164 out_unlock:
 165         folio_unlock(folio);
 166         folio_put(folio);
 167
 168         return bh;
 169 }
 170
 171 static void meta_prep_new(struct buffer_head *bh)
 172 {
 173         struct gfs2_meta_header *mh = (struct gfs2_meta_header *)bh->b_data;
 174
 175         lock_buffer(bh);
 176         clear_buffer_dirty(bh);
 177         set_buffer_uptodate(bh);
 178         unlock_buffer(bh);
 179
 180         mh->mh_magic = cpu_to_be32(GFS2_MAGIC);
 181 }
 182
 183 /**
 184  * gfs2_meta_new - Get a block
 185  * @gl: The glock associated with this block
 186  * @blkno: The block number
 187  *
 188  * Returns: The buffer
 189  */
 190
 191 struct buffer_head *gfs2_meta_new(struct gfs2_glock *gl, u64 blkno)
 192 {
 193         struct buffer_head *bh;
 194         bh = gfs2_getbuf(gl, blkno, CREATE);
 195         meta_prep_new(bh);
 196         return bh;
 197 }
 198
 199 static void gfs2_meta_read_endio(struct bio *bio)
 200 {
 201         struct bio_vec *bvec;
 202         struct bvec_iter_all iter_all;
 203
 204         bio_for_each_segment_all(bvec, bio, iter_all) {
 205                 struct page *page = bvec->bv_page;
 206                 struct buffer_head *bh = page_buffers(page);
 207                 unsigned int len = bvec->bv_len;
 208
 209                 while (bh_offset(bh) < bvec->bv_offset)
 210                         bh = bh->b_this_page;
 211                 do {
 212                         struct buffer_head *next = bh->b_this_page;
 213                         len -= bh->b_size;
 214                         bh->b_end_io(bh, !bio->bi_status);
 215                         bh = next;
 216                 } while (bh && len);
 217         }
 218         bio_put(bio);
 219 }
 220
 221 /*
 222  * Submit several consecutive buffer head I/O requests as a single bio I/O
 223  * request.  (See submit_bh_wbc.)
 224  */
 225 static void gfs2_submit_bhs(blk_opf_t opf, struct buffer_head *bhs[], int num)
 226 {
 227         while (num > 0) {
 228                 struct buffer_head *bh = *bhs;
 229                 struct bio *bio;
 230
 231                 bio = bio_alloc(bh->b_bdev, num, opf, GFP_NOIO);
 232                 bio->bi_iter.bi_sector = bh->b_blocknr * (bh->b_size >> 9);
 233                 while (num > 0) {
 234                         bh = *bhs;
 235                         if (!bio_add_page(bio, bh->b_page, bh->b_size, bh_offset(bh))) {
 236                                 BUG_ON(bio->bi_iter.bi_size == 0);
 237                                 break;
 238                         }
 239                         bhs++;
 240                         num--;
 241                 }
 242                 bio->bi_end_io = gfs2_meta_read_endio;
 243                 submit_bio(bio);
 244         }
 245 }
 246
 247 /**
 248  * gfs2_meta_read - Read a block from disk
 249  * @gl: The glock covering the block
 250  * @blkno: The block number
 251  * @flags: flags
 252  * @rahead: Do read-ahead
 253  * @bhp: the place where the buffer is returned (NULL on failure)
 254  *
 255  * Returns: errno
 256  */
 257
 258 int gfs2_meta_read(struct gfs2_glock *gl, u64 blkno, int flags,
 259                    int rahead, struct buffer_head **bhp)
 260 {
 261         struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
 262         struct buffer_head *bh, *bhs[2];
 263         int num = 0;
 264
 265         if (gfs2_withdrawing_or_withdrawn(sdp) &&
 266             !gfs2_withdraw_in_prog(sdp)) {
 267                 *bhp = NULL;
 268                 return -EIO;
 269         }
 270
 271         *bhp = bh = gfs2_getbuf(gl, blkno, CREATE);
 272
 273         lock_buffer(bh);
 274         if (buffer_uptodate(bh)) {
 275                 unlock_buffer(bh);
 276                 flags &= ~DIO_WAIT;
 277         } else {
 278                 bh->b_end_io = end_buffer_read_sync;
 279                 get_bh(bh);
 280                 bhs[num++] = bh;
 281         }
 282
 283         if (rahead) {
 284                 bh = gfs2_getbuf(gl, blkno + 1, CREATE);
 285
 286                 lock_buffer(bh);
 287                 if (buffer_uptodate(bh)) {
 288                         unlock_buffer(bh);
 289                         brelse(bh);
 290                 } else {
 291                         bh->b_end_io = end_buffer_read_sync;
 292                         bhs[num++] = bh;
 293                 }
 294         }
 295
 296         gfs2_submit_bhs(REQ_OP_READ | REQ_META | REQ_PRIO, bhs, num);
 297         if (!(flags & DIO_WAIT))
 298                 return 0;
 299
 300         bh = *bhp;
 301         wait_on_buffer(bh);
 302         if (unlikely(!buffer_uptodate(bh))) {
 303                 struct gfs2_trans *tr = current->journal_info;
 304                 if (tr && test_bit(TR_TOUCHED, &tr->tr_flags))
 305                         gfs2_io_error_bh_wd(sdp, bh);
 306                 brelse(bh);
 307                 *bhp = NULL;
 308                 return -EIO;
 309         }
 310
 311         return 0;
 312 }
 313
 314 /**
 315  * gfs2_meta_wait - Reread a block from disk
 316  * @sdp: the filesystem
 317  * @bh: The block to wait for
 318  *
 319  * Returns: errno
 320  */
 321
 322 int gfs2_meta_wait(struct gfs2_sbd *sdp, struct buffer_head *bh)
 323 {
 324         if (gfs2_withdrawing_or_withdrawn(sdp) &&
 325             !gfs2_withdraw_in_prog(sdp))
 326                 return -EIO;
 327
 328         wait_on_buffer(bh);
 329
 330         if (!buffer_uptodate(bh)) {
 331                 struct gfs2_trans *tr = current->journal_info;
 332                 if (tr && test_bit(TR_TOUCHED, &tr->tr_flags))
 333                         gfs2_io_error_bh_wd(sdp, bh);
 334                 return -EIO;
 335         }
 336         if (gfs2_withdrawing_or_withdrawn(sdp) &&
 337             !gfs2_withdraw_in_prog(sdp))
 338                 return -EIO;
 339
 340         return 0;
 341 }
 342
 343 void gfs2_remove_from_journal(struct buffer_head *bh, int meta)
 344 {
 345         struct address_space *mapping = bh->b_folio->mapping;
 346         struct gfs2_sbd *sdp = gfs2_mapping2sbd(mapping);
 347         struct gfs2_bufdata *bd = bh->b_private;
 348         struct gfs2_trans *tr = current->journal_info;
 349         int was_pinned = 0;
 350
 351         if (test_clear_buffer_pinned(bh)) {
 352                 trace_gfs2_pin(bd, 0);
 353                 atomic_dec(&sdp->sd_log_pinned);
 354                 list_del_init(&bd->bd_list);
 355                 if (meta == REMOVE_META)
 356                         tr->tr_num_buf_rm++;
 357                 else
 358                         tr->tr_num_databuf_rm++;
 359                 set_bit(TR_TOUCHED, &tr->tr_flags);
 360                 was_pinned = 1;
 361                 brelse(bh);
 362         }
 363         if (bd) {
 364                 if (bd->bd_tr) {
 365                         gfs2_trans_add_revoke(sdp, bd);
 366                 } else if (was_pinned) {
 367                         bh->b_private = NULL;
 368                         kmem_cache_free(gfs2_bufdata_cachep, bd);
 369                 } else if (!list_empty(&bd->bd_ail_st_list) &&
 370                                         !list_empty(&bd->bd_ail_gl_list)) {
 371                         gfs2_remove_from_ail(bd);
 372                 }
 373         }
 374         clear_buffer_dirty(bh);
 375         clear_buffer_uptodate(bh);
 376 }
 377
 378 /**
 379  * gfs2_ail1_wipe - remove deleted/freed buffers from the ail1 list
 380  * @sdp: superblock
 381  * @bstart: starting block address of buffers to remove
 382  * @blen: length of buffers to be removed
 383  *
 384  * This function is called from gfs2_journal wipe, whose job is to remove
 385  * buffers, corresponding to deleted blocks, from the journal. If we find any
 386  * bufdata elements on the system ail1 list, they haven't been written to
 387  * the journal yet. So we remove them.
 388  */
 389 static void gfs2_ail1_wipe(struct gfs2_sbd *sdp, u64 bstart, u32 blen)
 390 {
 391         struct gfs2_trans *tr, *s;
 392         struct gfs2_bufdata *bd, *bs;
 393         struct buffer_head *bh;
 394         u64 end = bstart + blen;
 395
 396         gfs2_log_lock(sdp);
 397         spin_lock(&sdp->sd_ail_lock);
 398         list_for_each_entry_safe(tr, s, &sdp->sd_ail1_list, tr_list) {
 399                 list_for_each_entry_safe(bd, bs, &tr->tr_ail1_list,
 400                                          bd_ail_st_list) {
 401                         bh = bd->bd_bh;
 402                         if (bh->b_blocknr < bstart || bh->b_blocknr >= end)
 403                                 continue;
 404
 405                         gfs2_remove_from_journal(bh, REMOVE_JDATA);
 406                 }
 407         }
 408         spin_unlock(&sdp->sd_ail_lock);
 409         gfs2_log_unlock(sdp);
 410 }
 411
 412 static struct buffer_head *gfs2_getjdatabuf(struct gfs2_inode *ip, u64 blkno)
 413 {
 414         struct address_space *mapping = ip->i_inode.i_mapping;
 415         struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 416         struct folio *folio;
 417         struct buffer_head *bh;
 418         unsigned int shift = PAGE_SHIFT - sdp->sd_sb.sb_bsize_shift;
 419         unsigned long index = blkno >> shift; /* convert block to page */
 420         unsigned int bufnum = blkno - (index << shift);
 421
 422         folio = __filemap_get_folio(mapping, index, FGP_LOCK | FGP_ACCESSED, 0);
 423         if (IS_ERR(folio))
 424                 return NULL;
 425         bh = folio_buffers(folio);
 426         if (bh)
 427                 bh = get_nth_bh(bh, bufnum);
 428         folio_unlock(folio);
 429         folio_put(folio);
 430         return bh;
 431 }
 432
 433 /**
 434  * gfs2_journal_wipe - make inode's buffers so they aren't dirty/pinned anymore
 435  * @ip: the inode who owns the buffers
 436  * @bstart: the first buffer in the run
 437  * @blen: the number of buffers in the run
 438  *
 439  */
 440
 441 void gfs2_journal_wipe(struct gfs2_inode *ip, u64 bstart, u32 blen)
 442 {
 443         struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 444         struct buffer_head *bh;
 445         int ty;
 446
 447         if (!ip->i_gl) {
 448                 /* This can only happen during incomplete inode creation. */
 449                 BUG_ON(!test_bit(GIF_ALLOC_FAILED, &ip->i_flags));
 450                 return;
 451         }
 452
 453         gfs2_ail1_wipe(sdp, bstart, blen);
 454         while (blen) {
 455                 ty = REMOVE_META;
 456                 bh = gfs2_getbuf(ip->i_gl, bstart, NO_CREATE);
 457                 if (!bh && gfs2_is_jdata(ip)) {
 458                         bh = gfs2_getjdatabuf(ip, bstart);
 459                         ty = REMOVE_JDATA;
 460                 }
 461                 if (bh) {
 462                         lock_buffer(bh);
 463                         gfs2_log_lock(sdp);
 464                         spin_lock(&sdp->sd_ail_lock);
 465                         gfs2_remove_from_journal(bh, ty);
 466                         spin_unlock(&sdp->sd_ail_lock);
 467                         gfs2_log_unlock(sdp);
 468                         unlock_buffer(bh);
 469                         brelse(bh);
 470                 }
 471
 472                 bstart++;
 473                 blen--;
 474         }
 475 }
 476
 477 /**
 478  * gfs2_meta_buffer - Get a metadata buffer
 479  * @ip: The GFS2 inode
 480  * @mtype: The block type (GFS2_METATYPE_*)
 481  * @num: The block number (device relative) of the buffer
 482  * @bhp: the buffer is returned here
 483  *
 484  * Returns: errno
 485  */
 486
 487 int gfs2_meta_buffer(struct gfs2_inode *ip, u32 mtype, u64 num,
 488                      struct buffer_head **bhp)
 489 {
 490         struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 491         struct gfs2_glock *gl = ip->i_gl;
 492         struct buffer_head *bh;
 493         int ret = 0;
 494         int rahead = 0;
 495
 496         if (num == ip->i_no_addr)
 497                 rahead = ip->i_rahead;
 498
 499         ret = gfs2_meta_read(gl, num, DIO_WAIT, rahead, &bh);
 500         if (ret == 0 && gfs2_metatype_check(sdp, bh, mtype)) {
 501                 brelse(bh);
 502                 ret = -EIO;
 503         } else {
 504                 *bhp = bh;
 505         }
 506         return ret;
 507 }
 508
 509 /**
 510  * gfs2_meta_ra - start readahead on an extent of a file
 511  * @gl: the glock the blocks belong to
 512  * @dblock: the starting disk block
 513  * @extlen: the number of blocks in the extent
 514  *
 515  * returns: the first buffer in the extent
 516  */
 517
 518 struct buffer_head *gfs2_meta_ra(struct gfs2_glock *gl, u64 dblock, u32 extlen)
 519 {
 520         struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
 521         struct buffer_head *first_bh, *bh;
 522         u32 max_ra = gfs2_tune_get(sdp, gt_max_readahead) >>
 523                           sdp->sd_sb.sb_bsize_shift;
 524
 525         BUG_ON(!extlen);
 526
 527         if (max_ra < 1)
 528                 max_ra = 1;
 529         if (extlen > max_ra)
 530                 extlen = max_ra;
 531
 532         first_bh = gfs2_getbuf(gl, dblock, CREATE);
 533
 534         if (buffer_uptodate(first_bh))
 535                 goto out;
 536         bh_read_nowait(first_bh, REQ_META | REQ_PRIO);
 537
 538         dblock++;
 539         extlen--;
 540
 541         while (extlen) {
 542                 bh = gfs2_getbuf(gl, dblock, CREATE);
 543
 544                 bh_readahead(bh, REQ_RAHEAD | REQ_META | REQ_PRIO);
 545                 brelse(bh);
 546                 dblock++;
 547                 extlen--;
 548                 if (!buffer_locked(first_bh) && buffer_uptodate(first_bh))
 549                         goto out;
 550         }
 551
 552         wait_on_buffer(first_bh);
 553 out:
 554         return first_bh;
 555 }
 556