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 | 100 | /** |
6802e340 | 101 | * gfs2_getbuf - Get a buffer with a given address space |
cb4c0313 | 102 | * @gl: the glock |
b3b94faa DT |
103 | * @blkno: the block number (filesystem scope) |
104 | * @create: 1 if the buffer should be created | |
105 | * | |
106 | * Returns: the buffer | |
107 | */ | |
108 | ||
6802e340 | 109 | struct buffer_head *gfs2_getbuf(struct gfs2_glock *gl, u64 blkno, int create) |
b3b94faa | 110 | { |
009d8518 | 111 | struct address_space *mapping = gfs2_glock2aspace(gl); |
cb4c0313 | 112 | struct gfs2_sbd *sdp = gl->gl_sbd; |
b3b94faa DT |
113 | struct page *page; |
114 | struct buffer_head *bh; | |
115 | unsigned int shift; | |
116 | unsigned long index; | |
117 | unsigned int bufnum; | |
118 | ||
119 | shift = PAGE_CACHE_SHIFT - sdp->sd_sb.sb_bsize_shift; | |
120 | index = blkno >> shift; /* convert block to page */ | |
121 | bufnum = blkno - (index << shift); /* block buf index within page */ | |
122 | ||
123 | if (create) { | |
124 | for (;;) { | |
cb4c0313 | 125 | page = grab_cache_page(mapping, index); |
b3b94faa DT |
126 | if (page) |
127 | break; | |
128 | yield(); | |
129 | } | |
130 | } else { | |
cb4c0313 | 131 | page = find_lock_page(mapping, index); |
b3b94faa DT |
132 | if (!page) |
133 | return NULL; | |
134 | } | |
135 | ||
136 | if (!page_has_buffers(page)) | |
137 | create_empty_buffers(page, sdp->sd_sb.sb_bsize, 0); | |
138 | ||
139 | /* Locate header for our buffer within our page */ | |
140 | for (bh = page_buffers(page); bufnum--; bh = bh->b_this_page) | |
141 | /* Do nothing */; | |
142 | get_bh(bh); | |
143 | ||
144 | if (!buffer_mapped(bh)) | |
145 | map_bh(bh, sdp->sd_vfs, blkno); | |
146 | ||
147 | unlock_page(page); | |
148 | mark_page_accessed(page); | |
149 | page_cache_release(page); | |
150 | ||
151 | return bh; | |
152 | } | |
153 | ||
154 | static void meta_prep_new(struct buffer_head *bh) | |
155 | { | |
156 | struct gfs2_meta_header *mh = (struct gfs2_meta_header *)bh->b_data; | |
157 | ||
158 | lock_buffer(bh); | |
159 | clear_buffer_dirty(bh); | |
160 | set_buffer_uptodate(bh); | |
161 | unlock_buffer(bh); | |
162 | ||
163 | mh->mh_magic = cpu_to_be32(GFS2_MAGIC); | |
164 | } | |
165 | ||
166 | /** | |
167 | * gfs2_meta_new - Get a block | |
168 | * @gl: The glock associated with this block | |
169 | * @blkno: The block number | |
170 | * | |
171 | * Returns: The buffer | |
172 | */ | |
173 | ||
cd915493 | 174 | struct buffer_head *gfs2_meta_new(struct gfs2_glock *gl, u64 blkno) |
b3b94faa DT |
175 | { |
176 | struct buffer_head *bh; | |
6802e340 | 177 | bh = gfs2_getbuf(gl, blkno, CREATE); |
b3b94faa DT |
178 | meta_prep_new(bh); |
179 | return bh; | |
180 | } | |
181 | ||
182 | /** | |
183 | * gfs2_meta_read - Read a block from disk | |
184 | * @gl: The glock covering the block | |
185 | * @blkno: The block number | |
7276b3b0 | 186 | * @flags: flags |
b3b94faa DT |
187 | * @bhp: the place where the buffer is returned (NULL on failure) |
188 | * | |
189 | * Returns: errno | |
190 | */ | |
191 | ||
cd915493 | 192 | int gfs2_meta_read(struct gfs2_glock *gl, u64 blkno, int flags, |
b3b94faa DT |
193 | struct buffer_head **bhp) |
194 | { | |
c969f58c SW |
195 | struct gfs2_sbd *sdp = gl->gl_sbd; |
196 | struct buffer_head *bh; | |
197 | ||
44b8db13 MY |
198 | if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags))) { |
199 | *bhp = NULL; | |
c969f58c | 200 | return -EIO; |
44b8db13 | 201 | } |
c969f58c SW |
202 | |
203 | *bhp = bh = gfs2_getbuf(gl, blkno, CREATE); | |
204 | ||
205 | lock_buffer(bh); | |
206 | if (buffer_uptodate(bh)) { | |
207 | unlock_buffer(bh); | |
208 | return 0; | |
209 | } | |
210 | bh->b_end_io = end_buffer_read_sync; | |
211 | get_bh(bh); | |
65299a3b | 212 | submit_bh(READ_SYNC | REQ_META | REQ_PRIO, bh); |
c969f58c SW |
213 | if (!(flags & DIO_WAIT)) |
214 | return 0; | |
215 | ||
216 | wait_on_buffer(bh); | |
217 | if (unlikely(!buffer_uptodate(bh))) { | |
218 | struct gfs2_trans *tr = current->journal_info; | |
219 | if (tr && tr->tr_touched) | |
220 | gfs2_io_error_bh(sdp, bh); | |
221 | brelse(bh); | |
44b8db13 | 222 | *bhp = NULL; |
c969f58c | 223 | return -EIO; |
7276b3b0 | 224 | } |
b3b94faa | 225 | |
7276b3b0 | 226 | return 0; |
b3b94faa DT |
227 | } |
228 | ||
229 | /** | |
7276b3b0 | 230 | * gfs2_meta_wait - Reread a block from disk |
b3b94faa | 231 | * @sdp: the filesystem |
7276b3b0 | 232 | * @bh: The block to wait for |
b3b94faa DT |
233 | * |
234 | * Returns: errno | |
235 | */ | |
236 | ||
7276b3b0 | 237 | int gfs2_meta_wait(struct gfs2_sbd *sdp, struct buffer_head *bh) |
b3b94faa DT |
238 | { |
239 | if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags))) | |
240 | return -EIO; | |
241 | ||
7276b3b0 | 242 | wait_on_buffer(bh); |
b3b94faa | 243 | |
7276b3b0 SW |
244 | if (!buffer_uptodate(bh)) { |
245 | struct gfs2_trans *tr = current->journal_info; | |
246 | if (tr && tr->tr_touched) | |
247 | gfs2_io_error_bh(sdp, bh); | |
248 | return -EIO; | |
b3b94faa | 249 | } |
7276b3b0 SW |
250 | if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags))) |
251 | return -EIO; | |
b3b94faa DT |
252 | |
253 | return 0; | |
254 | } | |
255 | ||
16615be1 SW |
256 | void gfs2_remove_from_journal(struct buffer_head *bh, struct gfs2_trans *tr, int meta) |
257 | { | |
009d8518 SW |
258 | struct address_space *mapping = bh->b_page->mapping; |
259 | struct gfs2_sbd *sdp = gfs2_mapping2sbd(mapping); | |
16615be1 | 260 | struct gfs2_bufdata *bd = bh->b_private; |
009d8518 | 261 | |
16615be1 | 262 | if (test_clear_buffer_pinned(bh)) { |
627c10b7 | 263 | trace_gfs2_pin(bd, 0); |
5e687eac | 264 | atomic_dec(&sdp->sd_log_pinned); |
c0752aa7 | 265 | list_del_init(&bd->bd_list); |
16615be1 SW |
266 | if (meta) { |
267 | gfs2_assert_warn(sdp, sdp->sd_log_num_buf); | |
268 | sdp->sd_log_num_buf--; | |
269 | tr->tr_num_buf_rm++; | |
270 | } else { | |
271 | gfs2_assert_warn(sdp, sdp->sd_log_num_databuf); | |
272 | sdp->sd_log_num_databuf--; | |
273 | tr->tr_num_databuf_rm++; | |
274 | } | |
275 | tr->tr_touched = 1; | |
276 | brelse(bh); | |
277 | } | |
278 | if (bd) { | |
c618e87a | 279 | spin_lock(&sdp->sd_ail_lock); |
16ca9412 | 280 | if (bd->bd_tr) { |
16615be1 SW |
281 | gfs2_trans_add_revoke(sdp, bd); |
282 | } | |
c618e87a | 283 | spin_unlock(&sdp->sd_ail_lock); |
16615be1 SW |
284 | } |
285 | clear_buffer_dirty(bh); | |
286 | clear_buffer_uptodate(bh); | |
287 | } | |
288 | ||
b3b94faa DT |
289 | /** |
290 | * gfs2_meta_wipe - make inode's buffers so they aren't dirty/pinned anymore | |
291 | * @ip: the inode who owns the buffers | |
292 | * @bstart: the first buffer in the run | |
293 | * @blen: the number of buffers in the run | |
294 | * | |
295 | */ | |
296 | ||
cd915493 | 297 | void gfs2_meta_wipe(struct gfs2_inode *ip, u64 bstart, u32 blen) |
b3b94faa | 298 | { |
feaa7bba | 299 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa DT |
300 | struct buffer_head *bh; |
301 | ||
302 | while (blen) { | |
6802e340 | 303 | bh = gfs2_getbuf(ip->i_gl, bstart, NO_CREATE); |
b3b94faa | 304 | if (bh) { |
1ad38c43 SW |
305 | lock_buffer(bh); |
306 | gfs2_log_lock(sdp); | |
16615be1 | 307 | gfs2_remove_from_journal(bh, current->journal_info, 1); |
1ad38c43 | 308 | gfs2_log_unlock(sdp); |
b3b94faa | 309 | unlock_buffer(bh); |
b3b94faa DT |
310 | brelse(bh); |
311 | } | |
312 | ||
313 | bstart++; | |
314 | blen--; | |
315 | } | |
316 | } | |
317 | ||
b3b94faa DT |
318 | /** |
319 | * gfs2_meta_indirect_buffer - Get a metadata buffer | |
320 | * @ip: The GFS2 inode | |
321 | * @height: The level of this buf in the metadata (indir addr) tree (if any) | |
322 | * @num: The block number (device relative) of the buffer | |
b3b94faa DT |
323 | * @bhp: the buffer is returned here |
324 | * | |
b3b94faa DT |
325 | * Returns: errno |
326 | */ | |
327 | ||
cd915493 | 328 | int gfs2_meta_indirect_buffer(struct gfs2_inode *ip, int height, u64 num, |
f2f9c812 | 329 | struct buffer_head **bhp) |
b3b94faa | 330 | { |
7276b3b0 SW |
331 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
332 | struct gfs2_glock *gl = ip->i_gl; | |
f91a0d3e SW |
333 | struct buffer_head *bh; |
334 | int ret = 0; | |
f2f9c812 | 335 | u32 mtype = height ? GFS2_METATYPE_IN : GFS2_METATYPE_DI; |
b3b94faa | 336 | |
f2f9c812 BP |
337 | ret = gfs2_meta_read(gl, num, DIO_WAIT, &bh); |
338 | if (ret == 0 && gfs2_metatype_check(sdp, bh, mtype)) { | |
339 | brelse(bh); | |
340 | ret = -EIO; | |
b3b94faa | 341 | } |
b3b94faa | 342 | *bhp = bh; |
f91a0d3e | 343 | return ret; |
b3b94faa DT |
344 | } |
345 | ||
346 | /** | |
347 | * gfs2_meta_ra - start readahead on an extent of a file | |
348 | * @gl: the glock the blocks belong to | |
349 | * @dblock: the starting disk block | |
350 | * @extlen: the number of blocks in the extent | |
351 | * | |
7276b3b0 | 352 | * returns: the first buffer in the extent |
b3b94faa DT |
353 | */ |
354 | ||
7276b3b0 | 355 | struct buffer_head *gfs2_meta_ra(struct gfs2_glock *gl, u64 dblock, u32 extlen) |
b3b94faa DT |
356 | { |
357 | struct gfs2_sbd *sdp = gl->gl_sbd; | |
b3b94faa | 358 | struct buffer_head *first_bh, *bh; |
cd915493 | 359 | u32 max_ra = gfs2_tune_get(sdp, gt_max_readahead) >> |
568f4c96 | 360 | sdp->sd_sb.sb_bsize_shift; |
b3b94faa | 361 | |
7276b3b0 SW |
362 | BUG_ON(!extlen); |
363 | ||
364 | if (max_ra < 1) | |
365 | max_ra = 1; | |
b3b94faa DT |
366 | if (extlen > max_ra) |
367 | extlen = max_ra; | |
368 | ||
6802e340 | 369 | first_bh = gfs2_getbuf(gl, dblock, CREATE); |
b3b94faa DT |
370 | |
371 | if (buffer_uptodate(first_bh)) | |
372 | goto out; | |
7276b3b0 | 373 | if (!buffer_locked(first_bh)) |
20ed0535 | 374 | ll_rw_block(READ_SYNC | REQ_META, 1, &first_bh); |
b3b94faa DT |
375 | |
376 | dblock++; | |
377 | extlen--; | |
378 | ||
379 | while (extlen) { | |
6802e340 | 380 | bh = gfs2_getbuf(gl, dblock, CREATE); |
b3b94faa | 381 | |
7276b3b0 | 382 | if (!buffer_uptodate(bh) && !buffer_locked(bh)) |
20ed0535 | 383 | ll_rw_block(READA | REQ_META, 1, &bh); |
7276b3b0 | 384 | brelse(bh); |
b3b94faa DT |
385 | dblock++; |
386 | extlen--; | |
7276b3b0 SW |
387 | if (!buffer_locked(first_bh) && buffer_uptodate(first_bh)) |
388 | goto out; | |
b3b94faa DT |
389 | } |
390 | ||
7276b3b0 | 391 | wait_on_buffer(first_bh); |
a91ea69f | 392 | out: |
7276b3b0 | 393 | return first_bh; |
b3b94faa DT |
394 | } |
395 |