Commit | Line | Data |
---|---|---|
b3b94faa DT |
1 | /* |
2 | * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. | |
091806ed | 3 | * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved. |
b3b94faa DT |
4 | * |
5 | * This copyrighted material is made available to anyone wishing to use, | |
6 | * modify, copy, or redistribute it subject to the terms and conditions | |
e9fc2aa0 | 7 | * of the GNU General Public License version 2. |
b3b94faa DT |
8 | */ |
9 | ||
10 | #include <linux/sched.h> | |
11 | #include <linux/slab.h> | |
12 | #include <linux/spinlock.h> | |
13 | #include <linux/completion.h> | |
14 | #include <linux/buffer_head.h> | |
15 | #include <linux/mm.h> | |
16 | #include <linux/pagemap.h> | |
17 | #include <linux/writeback.h> | |
18 | #include <linux/swap.h> | |
19 | #include <linux/delay.h> | |
2e565bb6 | 20 | #include <linux/bio.h> |
5c676f6d | 21 | #include <linux/gfs2_ondisk.h> |
b3b94faa DT |
22 | |
23 | #include "gfs2.h" | |
5c676f6d | 24 | #include "incore.h" |
b3b94faa DT |
25 | #include "glock.h" |
26 | #include "glops.h" | |
27 | #include "inode.h" | |
28 | #include "log.h" | |
29 | #include "lops.h" | |
30 | #include "meta_io.h" | |
31 | #include "rgrp.h" | |
32 | #include "trans.h" | |
5c676f6d | 33 | #include "util.h" |
627c10b7 | 34 | #include "trace_gfs2.h" |
b3b94faa | 35 | |
4a0f9a32 | 36 | static int gfs2_aspace_writepage(struct page *page, struct writeback_control *wbc) |
b3b94faa | 37 | { |
4a0f9a32 SW |
38 | struct buffer_head *bh, *head; |
39 | int nr_underway = 0; | |
65299a3b | 40 | int write_op = REQ_META | REQ_PRIO | |
721a9602 | 41 | (wbc->sync_mode == WB_SYNC_ALL ? WRITE_SYNC : WRITE); |
b3b94faa | 42 | |
4a0f9a32 SW |
43 | BUG_ON(!PageLocked(page)); |
44 | BUG_ON(!page_has_buffers(page)); | |
45 | ||
46 | head = page_buffers(page); | |
47 | bh = head; | |
48 | ||
49 | do { | |
50 | if (!buffer_mapped(bh)) | |
51 | continue; | |
52 | /* | |
53 | * If it's a fully non-blocking write attempt and we cannot | |
54 | * lock the buffer then redirty the page. Note that this can | |
e76e0ec9 | 55 | * potentially cause a busy-wait loop from flusher thread and kswapd |
4a0f9a32 SW |
56 | * activity, but those code paths have their own higher-level |
57 | * throttling. | |
58 | */ | |
1b430bee | 59 | if (wbc->sync_mode != WB_SYNC_NONE) { |
4a0f9a32 SW |
60 | lock_buffer(bh); |
61 | } else if (!trylock_buffer(bh)) { | |
62 | redirty_page_for_writepage(wbc, page); | |
63 | continue; | |
64 | } | |
65 | if (test_clear_buffer_dirty(bh)) { | |
66 | mark_buffer_async_write(bh); | |
67 | } else { | |
68 | unlock_buffer(bh); | |
69 | } | |
70 | } while ((bh = bh->b_this_page) != head); | |
71 | ||
72 | /* | |
73 | * The page and its buffers are protected by PageWriteback(), so we can | |
74 | * drop the bh refcounts early. | |
75 | */ | |
76 | BUG_ON(PageWriteback(page)); | |
77 | set_page_writeback(page); | |
78 | ||
79 | do { | |
80 | struct buffer_head *next = bh->b_this_page; | |
81 | if (buffer_async_write(bh)) { | |
82 | submit_bh(write_op, bh); | |
83 | nr_underway++; | |
84 | } | |
85 | bh = next; | |
86 | } while (bh != head); | |
87 | unlock_page(page); | |
88 | ||
4a0f9a32 SW |
89 | if (nr_underway == 0) |
90 | end_page_writeback(page); | |
91 | ||
eaefbf96 | 92 | return 0; |
b3b94faa DT |
93 | } |
94 | ||
009d8518 | 95 | const struct address_space_operations gfs2_meta_aops = { |
b3b94faa | 96 | .writepage = gfs2_aspace_writepage, |
4340fe62 | 97 | .releasepage = gfs2_releasepage, |
b3b94faa DT |
98 | }; |
99 | ||
1b2ad412 SW |
100 | const struct address_space_operations gfs2_rgrp_aops = { |
101 | .writepage = gfs2_aspace_writepage, | |
102 | .releasepage = gfs2_releasepage, | |
103 | }; | |
104 | ||
b3b94faa | 105 | /** |
6802e340 | 106 | * gfs2_getbuf - Get a buffer with a given address space |
cb4c0313 | 107 | * @gl: the glock |
b3b94faa DT |
108 | * @blkno: the block number (filesystem scope) |
109 | * @create: 1 if the buffer should be created | |
110 | * | |
111 | * Returns: the buffer | |
112 | */ | |
113 | ||
6802e340 | 114 | struct buffer_head *gfs2_getbuf(struct gfs2_glock *gl, u64 blkno, int create) |
b3b94faa | 115 | { |
009d8518 | 116 | struct address_space *mapping = gfs2_glock2aspace(gl); |
15562c43 | 117 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
b3b94faa DT |
118 | struct page *page; |
119 | struct buffer_head *bh; | |
120 | unsigned int shift; | |
121 | unsigned long index; | |
122 | unsigned int bufnum; | |
123 | ||
70d4ee94 SW |
124 | if (mapping == NULL) |
125 | mapping = &sdp->sd_aspace; | |
126 | ||
09cbfeaf | 127 | shift = PAGE_SHIFT - sdp->sd_sb.sb_bsize_shift; |
b3b94faa DT |
128 | index = blkno >> shift; /* convert block to page */ |
129 | bufnum = blkno - (index << shift); /* block buf index within page */ | |
130 | ||
131 | if (create) { | |
132 | for (;;) { | |
cb4c0313 | 133 | page = grab_cache_page(mapping, index); |
b3b94faa DT |
134 | if (page) |
135 | break; | |
136 | yield(); | |
137 | } | |
138 | } else { | |
2457aec6 MG |
139 | page = find_get_page_flags(mapping, index, |
140 | FGP_LOCK|FGP_ACCESSED); | |
b3b94faa DT |
141 | if (!page) |
142 | return NULL; | |
143 | } | |
144 | ||
145 | if (!page_has_buffers(page)) | |
146 | create_empty_buffers(page, sdp->sd_sb.sb_bsize, 0); | |
147 | ||
148 | /* Locate header for our buffer within our page */ | |
149 | for (bh = page_buffers(page); bufnum--; bh = bh->b_this_page) | |
150 | /* Do nothing */; | |
151 | get_bh(bh); | |
152 | ||
153 | if (!buffer_mapped(bh)) | |
154 | map_bh(bh, sdp->sd_vfs, blkno); | |
155 | ||
156 | unlock_page(page); | |
09cbfeaf | 157 | put_page(page); |
b3b94faa DT |
158 | |
159 | return bh; | |
160 | } | |
161 | ||
162 | static void meta_prep_new(struct buffer_head *bh) | |
163 | { | |
164 | struct gfs2_meta_header *mh = (struct gfs2_meta_header *)bh->b_data; | |
165 | ||
166 | lock_buffer(bh); | |
167 | clear_buffer_dirty(bh); | |
168 | set_buffer_uptodate(bh); | |
169 | unlock_buffer(bh); | |
170 | ||
171 | mh->mh_magic = cpu_to_be32(GFS2_MAGIC); | |
172 | } | |
173 | ||
174 | /** | |
175 | * gfs2_meta_new - Get a block | |
176 | * @gl: The glock associated with this block | |
177 | * @blkno: The block number | |
178 | * | |
179 | * Returns: The buffer | |
180 | */ | |
181 | ||
cd915493 | 182 | struct buffer_head *gfs2_meta_new(struct gfs2_glock *gl, u64 blkno) |
b3b94faa DT |
183 | { |
184 | struct buffer_head *bh; | |
6802e340 | 185 | bh = gfs2_getbuf(gl, blkno, CREATE); |
b3b94faa DT |
186 | meta_prep_new(bh); |
187 | return bh; | |
188 | } | |
189 | ||
39b0555f | 190 | static void gfs2_meta_read_endio(struct bio *bio) |
c8d57703 | 191 | { |
39b0555f AG |
192 | struct bio_vec *bvec; |
193 | int i; | |
194 | ||
195 | bio_for_each_segment_all(bvec, bio, i) { | |
196 | struct page *page = bvec->bv_page; | |
197 | struct buffer_head *bh = page_buffers(page); | |
198 | unsigned int len = bvec->bv_len; | |
199 | ||
200 | while (bh_offset(bh) < bvec->bv_offset) | |
201 | bh = bh->b_this_page; | |
202 | do { | |
203 | struct buffer_head *next = bh->b_this_page; | |
204 | len -= bh->b_size; | |
205 | bh->b_end_io(bh, !bio->bi_error); | |
206 | bh = next; | |
207 | } while (bh && len); | |
208 | } | |
209 | bio_put(bio); | |
210 | } | |
c8d57703 | 211 | |
39b0555f AG |
212 | /* |
213 | * Submit several consecutive buffer head I/O requests as a single bio I/O | |
214 | * request. (See submit_bh_wbc.) | |
215 | */ | |
216 | static void gfs2_submit_bhs(int rw, struct buffer_head *bhs[], int num) | |
217 | { | |
218 | struct buffer_head *bh = bhs[0]; | |
219 | struct bio *bio; | |
220 | int i; | |
221 | ||
222 | if (!num) | |
c8d57703 | 223 | return; |
39b0555f AG |
224 | |
225 | bio = bio_alloc(GFP_NOIO, num); | |
226 | bio->bi_iter.bi_sector = bh->b_blocknr * (bh->b_size >> 9); | |
227 | bio->bi_bdev = bh->b_bdev; | |
228 | for (i = 0; i < num; i++) { | |
229 | bh = bhs[i]; | |
230 | bio_add_page(bio, bh->b_page, bh->b_size, bh_offset(bh)); | |
c8d57703 | 231 | } |
39b0555f AG |
232 | bio->bi_end_io = gfs2_meta_read_endio; |
233 | submit_bio(rw, bio); | |
c8d57703 AG |
234 | } |
235 | ||
b3b94faa DT |
236 | /** |
237 | * gfs2_meta_read - Read a block from disk | |
238 | * @gl: The glock covering the block | |
239 | * @blkno: The block number | |
7276b3b0 | 240 | * @flags: flags |
b3b94faa DT |
241 | * @bhp: the place where the buffer is returned (NULL on failure) |
242 | * | |
243 | * Returns: errno | |
244 | */ | |
245 | ||
cd915493 | 246 | int gfs2_meta_read(struct gfs2_glock *gl, u64 blkno, int flags, |
c8d57703 | 247 | int rahead, struct buffer_head **bhp) |
b3b94faa | 248 | { |
15562c43 | 249 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
39b0555f AG |
250 | struct buffer_head *bh, *bhs[2]; |
251 | int num = 0; | |
c969f58c | 252 | |
44b8db13 MY |
253 | if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags))) { |
254 | *bhp = NULL; | |
c969f58c | 255 | return -EIO; |
44b8db13 | 256 | } |
c969f58c SW |
257 | |
258 | *bhp = bh = gfs2_getbuf(gl, blkno, CREATE); | |
259 | ||
260 | lock_buffer(bh); | |
261 | if (buffer_uptodate(bh)) { | |
262 | unlock_buffer(bh); | |
39b0555f AG |
263 | flags &= ~DIO_WAIT; |
264 | } else { | |
265 | bh->b_end_io = end_buffer_read_sync; | |
266 | get_bh(bh); | |
267 | bhs[num++] = bh; | |
c969f58c | 268 | } |
39b0555f AG |
269 | |
270 | if (rahead) { | |
271 | bh = gfs2_getbuf(gl, blkno + 1, CREATE); | |
272 | ||
273 | lock_buffer(bh); | |
274 | if (buffer_uptodate(bh)) { | |
275 | unlock_buffer(bh); | |
276 | brelse(bh); | |
277 | } else { | |
278 | bh->b_end_io = end_buffer_read_sync; | |
279 | bhs[num++] = bh; | |
280 | } | |
281 | } | |
282 | ||
283 | gfs2_submit_bhs(READ_SYNC | REQ_META | REQ_PRIO, bhs, num); | |
c969f58c SW |
284 | if (!(flags & DIO_WAIT)) |
285 | return 0; | |
286 | ||
39b0555f | 287 | bh = *bhp; |
c969f58c SW |
288 | wait_on_buffer(bh); |
289 | if (unlikely(!buffer_uptodate(bh))) { | |
290 | struct gfs2_trans *tr = current->journal_info; | |
291 | if (tr && tr->tr_touched) | |
292 | gfs2_io_error_bh(sdp, bh); | |
293 | brelse(bh); | |
44b8db13 | 294 | *bhp = NULL; |
c969f58c | 295 | return -EIO; |
7276b3b0 | 296 | } |
b3b94faa | 297 | |
7276b3b0 | 298 | return 0; |
b3b94faa DT |
299 | } |
300 | ||
301 | /** | |
7276b3b0 | 302 | * gfs2_meta_wait - Reread a block from disk |
b3b94faa | 303 | * @sdp: the filesystem |
7276b3b0 | 304 | * @bh: The block to wait for |
b3b94faa DT |
305 | * |
306 | * Returns: errno | |
307 | */ | |
308 | ||
7276b3b0 | 309 | int gfs2_meta_wait(struct gfs2_sbd *sdp, struct buffer_head *bh) |
b3b94faa DT |
310 | { |
311 | if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags))) | |
312 | return -EIO; | |
313 | ||
7276b3b0 | 314 | wait_on_buffer(bh); |
b3b94faa | 315 | |
7276b3b0 SW |
316 | if (!buffer_uptodate(bh)) { |
317 | struct gfs2_trans *tr = current->journal_info; | |
318 | if (tr && tr->tr_touched) | |
319 | gfs2_io_error_bh(sdp, bh); | |
320 | return -EIO; | |
b3b94faa | 321 | } |
7276b3b0 SW |
322 | if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags))) |
323 | return -EIO; | |
b3b94faa DT |
324 | |
325 | return 0; | |
326 | } | |
327 | ||
16615be1 SW |
328 | void gfs2_remove_from_journal(struct buffer_head *bh, struct gfs2_trans *tr, int meta) |
329 | { | |
009d8518 SW |
330 | struct address_space *mapping = bh->b_page->mapping; |
331 | struct gfs2_sbd *sdp = gfs2_mapping2sbd(mapping); | |
16615be1 | 332 | struct gfs2_bufdata *bd = bh->b_private; |
502be2a3 | 333 | int was_pinned = 0; |
009d8518 | 334 | |
16615be1 | 335 | if (test_clear_buffer_pinned(bh)) { |
627c10b7 | 336 | trace_gfs2_pin(bd, 0); |
5e687eac | 337 | atomic_dec(&sdp->sd_log_pinned); |
c0752aa7 | 338 | list_del_init(&bd->bd_list); |
022ef4fe | 339 | if (meta) |
16615be1 | 340 | tr->tr_num_buf_rm++; |
022ef4fe | 341 | else |
16615be1 | 342 | tr->tr_num_databuf_rm++; |
16615be1 | 343 | tr->tr_touched = 1; |
502be2a3 | 344 | was_pinned = 1; |
16615be1 SW |
345 | brelse(bh); |
346 | } | |
347 | if (bd) { | |
c618e87a | 348 | spin_lock(&sdp->sd_ail_lock); |
16ca9412 | 349 | if (bd->bd_tr) { |
16615be1 | 350 | gfs2_trans_add_revoke(sdp, bd); |
502be2a3 BP |
351 | } else if (was_pinned) { |
352 | bh->b_private = NULL; | |
353 | kmem_cache_free(gfs2_bufdata_cachep, bd); | |
16615be1 | 354 | } |
c618e87a | 355 | spin_unlock(&sdp->sd_ail_lock); |
16615be1 SW |
356 | } |
357 | clear_buffer_dirty(bh); | |
358 | clear_buffer_uptodate(bh); | |
359 | } | |
360 | ||
b3b94faa DT |
361 | /** |
362 | * gfs2_meta_wipe - make inode's buffers so they aren't dirty/pinned anymore | |
363 | * @ip: the inode who owns the buffers | |
364 | * @bstart: the first buffer in the run | |
365 | * @blen: the number of buffers in the run | |
366 | * | |
367 | */ | |
368 | ||
cd915493 | 369 | void gfs2_meta_wipe(struct gfs2_inode *ip, u64 bstart, u32 blen) |
b3b94faa | 370 | { |
feaa7bba | 371 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa DT |
372 | struct buffer_head *bh; |
373 | ||
374 | while (blen) { | |
6802e340 | 375 | bh = gfs2_getbuf(ip->i_gl, bstart, NO_CREATE); |
b3b94faa | 376 | if (bh) { |
1ad38c43 SW |
377 | lock_buffer(bh); |
378 | gfs2_log_lock(sdp); | |
16615be1 | 379 | gfs2_remove_from_journal(bh, current->journal_info, 1); |
1ad38c43 | 380 | gfs2_log_unlock(sdp); |
b3b94faa | 381 | unlock_buffer(bh); |
b3b94faa DT |
382 | brelse(bh); |
383 | } | |
384 | ||
385 | bstart++; | |
386 | blen--; | |
387 | } | |
388 | } | |
389 | ||
b3b94faa DT |
390 | /** |
391 | * gfs2_meta_indirect_buffer - Get a metadata buffer | |
392 | * @ip: The GFS2 inode | |
393 | * @height: The level of this buf in the metadata (indir addr) tree (if any) | |
394 | * @num: The block number (device relative) of the buffer | |
b3b94faa DT |
395 | * @bhp: the buffer is returned here |
396 | * | |
b3b94faa DT |
397 | * Returns: errno |
398 | */ | |
399 | ||
cd915493 | 400 | int gfs2_meta_indirect_buffer(struct gfs2_inode *ip, int height, u64 num, |
f2f9c812 | 401 | struct buffer_head **bhp) |
b3b94faa | 402 | { |
7276b3b0 SW |
403 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
404 | struct gfs2_glock *gl = ip->i_gl; | |
f91a0d3e SW |
405 | struct buffer_head *bh; |
406 | int ret = 0; | |
f2f9c812 | 407 | u32 mtype = height ? GFS2_METATYPE_IN : GFS2_METATYPE_DI; |
c8d57703 AG |
408 | int rahead = 0; |
409 | ||
410 | if (num == ip->i_no_addr) | |
411 | rahead = ip->i_rahead; | |
b3b94faa | 412 | |
c8d57703 | 413 | ret = gfs2_meta_read(gl, num, DIO_WAIT, rahead, &bh); |
f2f9c812 BP |
414 | if (ret == 0 && gfs2_metatype_check(sdp, bh, mtype)) { |
415 | brelse(bh); | |
416 | ret = -EIO; | |
b3b94faa | 417 | } |
b3b94faa | 418 | *bhp = bh; |
f91a0d3e | 419 | return ret; |
b3b94faa DT |
420 | } |
421 | ||
422 | /** | |
423 | * gfs2_meta_ra - start readahead on an extent of a file | |
424 | * @gl: the glock the blocks belong to | |
425 | * @dblock: the starting disk block | |
426 | * @extlen: the number of blocks in the extent | |
427 | * | |
7276b3b0 | 428 | * returns: the first buffer in the extent |
b3b94faa DT |
429 | */ |
430 | ||
7276b3b0 | 431 | struct buffer_head *gfs2_meta_ra(struct gfs2_glock *gl, u64 dblock, u32 extlen) |
b3b94faa | 432 | { |
15562c43 | 433 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
b3b94faa | 434 | struct buffer_head *first_bh, *bh; |
cd915493 | 435 | u32 max_ra = gfs2_tune_get(sdp, gt_max_readahead) >> |
568f4c96 | 436 | sdp->sd_sb.sb_bsize_shift; |
b3b94faa | 437 | |
7276b3b0 SW |
438 | BUG_ON(!extlen); |
439 | ||
440 | if (max_ra < 1) | |
441 | max_ra = 1; | |
b3b94faa DT |
442 | if (extlen > max_ra) |
443 | extlen = max_ra; | |
444 | ||
6802e340 | 445 | first_bh = gfs2_getbuf(gl, dblock, CREATE); |
b3b94faa DT |
446 | |
447 | if (buffer_uptodate(first_bh)) | |
448 | goto out; | |
7276b3b0 | 449 | if (!buffer_locked(first_bh)) |
20ed0535 | 450 | ll_rw_block(READ_SYNC | REQ_META, 1, &first_bh); |
b3b94faa DT |
451 | |
452 | dblock++; | |
453 | extlen--; | |
454 | ||
455 | while (extlen) { | |
6802e340 | 456 | bh = gfs2_getbuf(gl, dblock, CREATE); |
b3b94faa | 457 | |
7276b3b0 | 458 | if (!buffer_uptodate(bh) && !buffer_locked(bh)) |
20ed0535 | 459 | ll_rw_block(READA | REQ_META, 1, &bh); |
7276b3b0 | 460 | brelse(bh); |
b3b94faa DT |
461 | dblock++; |
462 | extlen--; | |
7276b3b0 SW |
463 | if (!buffer_locked(first_bh) && buffer_uptodate(first_bh)) |
464 | goto out; | |
b3b94faa DT |
465 | } |
466 | ||
7276b3b0 | 467 | wait_on_buffer(first_bh); |
a91ea69f | 468 | out: |
7276b3b0 | 469 | return first_bh; |
b3b94faa DT |
470 | } |
471 |