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 | ||
b3b94faa DT |
100 | /** |
101 | * gfs2_meta_sync - Sync all buffers associated with a glock | |
102 | * @gl: The glock | |
b3b94faa DT |
103 | * |
104 | */ | |
105 | ||
7276b3b0 | 106 | void gfs2_meta_sync(struct gfs2_glock *gl) |
b3b94faa | 107 | { |
009d8518 | 108 | struct address_space *mapping = gfs2_glock2aspace(gl); |
7276b3b0 | 109 | int error; |
b3b94faa | 110 | |
7276b3b0 SW |
111 | filemap_fdatawrite(mapping); |
112 | error = filemap_fdatawait(mapping); | |
b3b94faa DT |
113 | |
114 | if (error) | |
115 | gfs2_io_error(gl->gl_sbd); | |
116 | } | |
117 | ||
118 | /** | |
6802e340 | 119 | * gfs2_getbuf - Get a buffer with a given address space |
cb4c0313 | 120 | * @gl: the glock |
b3b94faa DT |
121 | * @blkno: the block number (filesystem scope) |
122 | * @create: 1 if the buffer should be created | |
123 | * | |
124 | * Returns: the buffer | |
125 | */ | |
126 | ||
6802e340 | 127 | struct buffer_head *gfs2_getbuf(struct gfs2_glock *gl, u64 blkno, int create) |
b3b94faa | 128 | { |
009d8518 | 129 | struct address_space *mapping = gfs2_glock2aspace(gl); |
cb4c0313 | 130 | struct gfs2_sbd *sdp = gl->gl_sbd; |
b3b94faa DT |
131 | struct page *page; |
132 | struct buffer_head *bh; | |
133 | unsigned int shift; | |
134 | unsigned long index; | |
135 | unsigned int bufnum; | |
136 | ||
137 | shift = PAGE_CACHE_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 | for (;;) { | |
cb4c0313 | 143 | page = grab_cache_page(mapping, index); |
b3b94faa DT |
144 | if (page) |
145 | break; | |
146 | yield(); | |
147 | } | |
148 | } else { | |
cb4c0313 | 149 | page = find_lock_page(mapping, index); |
b3b94faa DT |
150 | if (!page) |
151 | return NULL; | |
152 | } | |
153 | ||
154 | if (!page_has_buffers(page)) | |
155 | create_empty_buffers(page, sdp->sd_sb.sb_bsize, 0); | |
156 | ||
157 | /* Locate header for our buffer within our page */ | |
158 | for (bh = page_buffers(page); bufnum--; bh = bh->b_this_page) | |
159 | /* Do nothing */; | |
160 | get_bh(bh); | |
161 | ||
162 | if (!buffer_mapped(bh)) | |
163 | map_bh(bh, sdp->sd_vfs, blkno); | |
164 | ||
165 | unlock_page(page); | |
166 | mark_page_accessed(page); | |
167 | page_cache_release(page); | |
168 | ||
169 | return bh; | |
170 | } | |
171 | ||
172 | static void meta_prep_new(struct buffer_head *bh) | |
173 | { | |
174 | struct gfs2_meta_header *mh = (struct gfs2_meta_header *)bh->b_data; | |
175 | ||
176 | lock_buffer(bh); | |
177 | clear_buffer_dirty(bh); | |
178 | set_buffer_uptodate(bh); | |
179 | unlock_buffer(bh); | |
180 | ||
181 | mh->mh_magic = cpu_to_be32(GFS2_MAGIC); | |
182 | } | |
183 | ||
184 | /** | |
185 | * gfs2_meta_new - Get a block | |
186 | * @gl: The glock associated with this block | |
187 | * @blkno: The block number | |
188 | * | |
189 | * Returns: The buffer | |
190 | */ | |
191 | ||
cd915493 | 192 | struct buffer_head *gfs2_meta_new(struct gfs2_glock *gl, u64 blkno) |
b3b94faa DT |
193 | { |
194 | struct buffer_head *bh; | |
6802e340 | 195 | bh = gfs2_getbuf(gl, blkno, CREATE); |
b3b94faa DT |
196 | meta_prep_new(bh); |
197 | return bh; | |
198 | } | |
199 | ||
200 | /** | |
201 | * gfs2_meta_read - Read a block from disk | |
202 | * @gl: The glock covering the block | |
203 | * @blkno: The block number | |
7276b3b0 | 204 | * @flags: flags |
b3b94faa DT |
205 | * @bhp: the place where the buffer is returned (NULL on failure) |
206 | * | |
207 | * Returns: errno | |
208 | */ | |
209 | ||
cd915493 | 210 | int gfs2_meta_read(struct gfs2_glock *gl, u64 blkno, int flags, |
b3b94faa DT |
211 | struct buffer_head **bhp) |
212 | { | |
c969f58c SW |
213 | struct gfs2_sbd *sdp = gl->gl_sbd; |
214 | struct buffer_head *bh; | |
215 | ||
44b8db13 MY |
216 | if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags))) { |
217 | *bhp = NULL; | |
c969f58c | 218 | return -EIO; |
44b8db13 | 219 | } |
c969f58c SW |
220 | |
221 | *bhp = bh = gfs2_getbuf(gl, blkno, CREATE); | |
222 | ||
223 | lock_buffer(bh); | |
224 | if (buffer_uptodate(bh)) { | |
225 | unlock_buffer(bh); | |
226 | return 0; | |
227 | } | |
228 | bh->b_end_io = end_buffer_read_sync; | |
229 | get_bh(bh); | |
65299a3b | 230 | submit_bh(READ_SYNC | REQ_META | REQ_PRIO, bh); |
c969f58c SW |
231 | if (!(flags & DIO_WAIT)) |
232 | return 0; | |
233 | ||
234 | wait_on_buffer(bh); | |
235 | if (unlikely(!buffer_uptodate(bh))) { | |
236 | struct gfs2_trans *tr = current->journal_info; | |
237 | if (tr && tr->tr_touched) | |
238 | gfs2_io_error_bh(sdp, bh); | |
239 | brelse(bh); | |
44b8db13 | 240 | *bhp = NULL; |
c969f58c | 241 | return -EIO; |
7276b3b0 | 242 | } |
b3b94faa | 243 | |
7276b3b0 | 244 | return 0; |
b3b94faa DT |
245 | } |
246 | ||
247 | /** | |
7276b3b0 | 248 | * gfs2_meta_wait - Reread a block from disk |
b3b94faa | 249 | * @sdp: the filesystem |
7276b3b0 | 250 | * @bh: The block to wait for |
b3b94faa DT |
251 | * |
252 | * Returns: errno | |
253 | */ | |
254 | ||
7276b3b0 | 255 | int gfs2_meta_wait(struct gfs2_sbd *sdp, struct buffer_head *bh) |
b3b94faa DT |
256 | { |
257 | if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags))) | |
258 | return -EIO; | |
259 | ||
7276b3b0 | 260 | wait_on_buffer(bh); |
b3b94faa | 261 | |
7276b3b0 SW |
262 | if (!buffer_uptodate(bh)) { |
263 | struct gfs2_trans *tr = current->journal_info; | |
264 | if (tr && tr->tr_touched) | |
265 | gfs2_io_error_bh(sdp, bh); | |
266 | return -EIO; | |
b3b94faa | 267 | } |
7276b3b0 SW |
268 | if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags))) |
269 | return -EIO; | |
b3b94faa DT |
270 | |
271 | return 0; | |
272 | } | |
273 | ||
16615be1 SW |
274 | void gfs2_remove_from_journal(struct buffer_head *bh, struct gfs2_trans *tr, int meta) |
275 | { | |
009d8518 SW |
276 | struct address_space *mapping = bh->b_page->mapping; |
277 | struct gfs2_sbd *sdp = gfs2_mapping2sbd(mapping); | |
16615be1 | 278 | struct gfs2_bufdata *bd = bh->b_private; |
009d8518 | 279 | |
16615be1 | 280 | if (test_clear_buffer_pinned(bh)) { |
627c10b7 | 281 | trace_gfs2_pin(bd, 0); |
5e687eac | 282 | atomic_dec(&sdp->sd_log_pinned); |
c0752aa7 | 283 | list_del_init(&bd->bd_list); |
16615be1 SW |
284 | if (meta) { |
285 | gfs2_assert_warn(sdp, sdp->sd_log_num_buf); | |
286 | sdp->sd_log_num_buf--; | |
287 | tr->tr_num_buf_rm++; | |
288 | } else { | |
289 | gfs2_assert_warn(sdp, sdp->sd_log_num_databuf); | |
290 | sdp->sd_log_num_databuf--; | |
291 | tr->tr_num_databuf_rm++; | |
292 | } | |
293 | tr->tr_touched = 1; | |
294 | brelse(bh); | |
295 | } | |
296 | if (bd) { | |
c618e87a | 297 | spin_lock(&sdp->sd_ail_lock); |
16ca9412 | 298 | if (bd->bd_tr) { |
f91a0d3e | 299 | gfs2_remove_from_ail(bd); |
16615be1 SW |
300 | bh->b_private = NULL; |
301 | bd->bd_bh = NULL; | |
302 | bd->bd_blkno = bh->b_blocknr; | |
303 | gfs2_trans_add_revoke(sdp, bd); | |
304 | } | |
c618e87a | 305 | spin_unlock(&sdp->sd_ail_lock); |
16615be1 SW |
306 | } |
307 | clear_buffer_dirty(bh); | |
308 | clear_buffer_uptodate(bh); | |
309 | } | |
310 | ||
b3b94faa DT |
311 | /** |
312 | * gfs2_meta_wipe - make inode's buffers so they aren't dirty/pinned anymore | |
313 | * @ip: the inode who owns the buffers | |
314 | * @bstart: the first buffer in the run | |
315 | * @blen: the number of buffers in the run | |
316 | * | |
317 | */ | |
318 | ||
cd915493 | 319 | void gfs2_meta_wipe(struct gfs2_inode *ip, u64 bstart, u32 blen) |
b3b94faa | 320 | { |
feaa7bba | 321 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa DT |
322 | struct buffer_head *bh; |
323 | ||
324 | while (blen) { | |
6802e340 | 325 | bh = gfs2_getbuf(ip->i_gl, bstart, NO_CREATE); |
b3b94faa | 326 | if (bh) { |
1ad38c43 SW |
327 | lock_buffer(bh); |
328 | gfs2_log_lock(sdp); | |
16615be1 | 329 | gfs2_remove_from_journal(bh, current->journal_info, 1); |
1ad38c43 | 330 | gfs2_log_unlock(sdp); |
b3b94faa | 331 | unlock_buffer(bh); |
b3b94faa DT |
332 | brelse(bh); |
333 | } | |
334 | ||
335 | bstart++; | |
336 | blen--; | |
337 | } | |
338 | } | |
339 | ||
b3b94faa DT |
340 | /** |
341 | * gfs2_meta_indirect_buffer - Get a metadata buffer | |
342 | * @ip: The GFS2 inode | |
343 | * @height: The level of this buf in the metadata (indir addr) tree (if any) | |
344 | * @num: The block number (device relative) of the buffer | |
b3b94faa DT |
345 | * @bhp: the buffer is returned here |
346 | * | |
b3b94faa DT |
347 | * Returns: errno |
348 | */ | |
349 | ||
cd915493 | 350 | int gfs2_meta_indirect_buffer(struct gfs2_inode *ip, int height, u64 num, |
f2f9c812 | 351 | struct buffer_head **bhp) |
b3b94faa | 352 | { |
7276b3b0 SW |
353 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
354 | struct gfs2_glock *gl = ip->i_gl; | |
f91a0d3e SW |
355 | struct buffer_head *bh; |
356 | int ret = 0; | |
f2f9c812 | 357 | u32 mtype = height ? GFS2_METATYPE_IN : GFS2_METATYPE_DI; |
b3b94faa | 358 | |
f2f9c812 BP |
359 | ret = gfs2_meta_read(gl, num, DIO_WAIT, &bh); |
360 | if (ret == 0 && gfs2_metatype_check(sdp, bh, mtype)) { | |
361 | brelse(bh); | |
362 | ret = -EIO; | |
b3b94faa | 363 | } |
b3b94faa | 364 | *bhp = bh; |
f91a0d3e | 365 | return ret; |
b3b94faa DT |
366 | } |
367 | ||
368 | /** | |
369 | * gfs2_meta_ra - start readahead on an extent of a file | |
370 | * @gl: the glock the blocks belong to | |
371 | * @dblock: the starting disk block | |
372 | * @extlen: the number of blocks in the extent | |
373 | * | |
7276b3b0 | 374 | * returns: the first buffer in the extent |
b3b94faa DT |
375 | */ |
376 | ||
7276b3b0 | 377 | struct buffer_head *gfs2_meta_ra(struct gfs2_glock *gl, u64 dblock, u32 extlen) |
b3b94faa DT |
378 | { |
379 | struct gfs2_sbd *sdp = gl->gl_sbd; | |
b3b94faa | 380 | struct buffer_head *first_bh, *bh; |
cd915493 | 381 | u32 max_ra = gfs2_tune_get(sdp, gt_max_readahead) >> |
568f4c96 | 382 | sdp->sd_sb.sb_bsize_shift; |
b3b94faa | 383 | |
7276b3b0 SW |
384 | BUG_ON(!extlen); |
385 | ||
386 | if (max_ra < 1) | |
387 | max_ra = 1; | |
b3b94faa DT |
388 | if (extlen > max_ra) |
389 | extlen = max_ra; | |
390 | ||
6802e340 | 391 | first_bh = gfs2_getbuf(gl, dblock, CREATE); |
b3b94faa DT |
392 | |
393 | if (buffer_uptodate(first_bh)) | |
394 | goto out; | |
7276b3b0 | 395 | if (!buffer_locked(first_bh)) |
20ed0535 | 396 | ll_rw_block(READ_SYNC | REQ_META, 1, &first_bh); |
b3b94faa DT |
397 | |
398 | dblock++; | |
399 | extlen--; | |
400 | ||
401 | while (extlen) { | |
6802e340 | 402 | bh = gfs2_getbuf(gl, dblock, CREATE); |
b3b94faa | 403 | |
7276b3b0 | 404 | if (!buffer_uptodate(bh) && !buffer_locked(bh)) |
20ed0535 | 405 | ll_rw_block(READA | REQ_META, 1, &bh); |
7276b3b0 | 406 | brelse(bh); |
b3b94faa DT |
407 | dblock++; |
408 | extlen--; | |
7276b3b0 SW |
409 | if (!buffer_locked(first_bh) && buffer_uptodate(first_bh)) |
410 | goto out; | |
b3b94faa DT |
411 | } |
412 | ||
7276b3b0 | 413 | wait_on_buffer(first_bh); |
a91ea69f | 414 | out: |
7276b3b0 | 415 | return first_bh; |
b3b94faa DT |
416 | } |
417 |