]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/gfs2/meta_io.c
convert gfs2
[net-next-2.6.git] / fs / gfs2 / meta_io.c
CommitLineData
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"
b3b94faa 34
4a0f9a32 35static int gfs2_aspace_writepage(struct page *page, struct writeback_control *wbc)
b3b94faa 36{
4a0f9a32
SW
37 struct buffer_head *bh, *head;
38 int nr_underway = 0;
7b6d91da
CH
39 int write_op = REQ_META |
40 (wbc->sync_mode == WB_SYNC_ALL ? WRITE_SYNC_PLUG : WRITE);
b3b94faa 41
4a0f9a32
SW
42 BUG_ON(!PageLocked(page));
43 BUG_ON(!page_has_buffers(page));
44
45 head = page_buffers(page);
46 bh = head;
47
48 do {
49 if (!buffer_mapped(bh))
50 continue;
51 /*
52 * If it's a fully non-blocking write attempt and we cannot
53 * lock the buffer then redirty the page. Note that this can
54 * potentially cause a busy-wait loop from pdflush and kswapd
55 * activity, but those code paths have their own higher-level
56 * throttling.
57 */
1b430bee 58 if (wbc->sync_mode != WB_SYNC_NONE) {
4a0f9a32
SW
59 lock_buffer(bh);
60 } else if (!trylock_buffer(bh)) {
61 redirty_page_for_writepage(wbc, page);
62 continue;
63 }
64 if (test_clear_buffer_dirty(bh)) {
65 mark_buffer_async_write(bh);
66 } else {
67 unlock_buffer(bh);
68 }
69 } while ((bh = bh->b_this_page) != head);
70
71 /*
72 * The page and its buffers are protected by PageWriteback(), so we can
73 * drop the bh refcounts early.
74 */
75 BUG_ON(PageWriteback(page));
76 set_page_writeback(page);
77
78 do {
79 struct buffer_head *next = bh->b_this_page;
80 if (buffer_async_write(bh)) {
81 submit_bh(write_op, bh);
82 nr_underway++;
83 }
84 bh = next;
85 } while (bh != head);
86 unlock_page(page);
87
4a0f9a32
SW
88 if (nr_underway == 0)
89 end_page_writeback(page);
90
eaefbf96 91 return 0;
b3b94faa
DT
92}
93
009d8518 94const struct address_space_operations gfs2_meta_aops = {
b3b94faa 95 .writepage = gfs2_aspace_writepage,
4340fe62 96 .releasepage = gfs2_releasepage,
52d4c74b 97 .sync_page = block_sync_page,
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 106void 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 127struct 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
172static 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 192struct 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 210int 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
216 if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
217 return -EIO;
218
219 *bhp = bh = gfs2_getbuf(gl, blkno, CREATE);
220
221 lock_buffer(bh);
222 if (buffer_uptodate(bh)) {
223 unlock_buffer(bh);
224 return 0;
225 }
226 bh->b_end_io = end_buffer_read_sync;
227 get_bh(bh);
7b6d91da 228 submit_bh(READ_SYNC | REQ_META, bh);
c969f58c
SW
229 if (!(flags & DIO_WAIT))
230 return 0;
231
232 wait_on_buffer(bh);
233 if (unlikely(!buffer_uptodate(bh))) {
234 struct gfs2_trans *tr = current->journal_info;
235 if (tr && tr->tr_touched)
236 gfs2_io_error_bh(sdp, bh);
237 brelse(bh);
238 return -EIO;
7276b3b0 239 }
b3b94faa 240
7276b3b0 241 return 0;
b3b94faa
DT
242}
243
244/**
7276b3b0 245 * gfs2_meta_wait - Reread a block from disk
b3b94faa 246 * @sdp: the filesystem
7276b3b0 247 * @bh: The block to wait for
b3b94faa
DT
248 *
249 * Returns: errno
250 */
251
7276b3b0 252int gfs2_meta_wait(struct gfs2_sbd *sdp, struct buffer_head *bh)
b3b94faa
DT
253{
254 if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
255 return -EIO;
256
7276b3b0 257 wait_on_buffer(bh);
b3b94faa 258
7276b3b0
SW
259 if (!buffer_uptodate(bh)) {
260 struct gfs2_trans *tr = current->journal_info;
261 if (tr && tr->tr_touched)
262 gfs2_io_error_bh(sdp, bh);
263 return -EIO;
b3b94faa 264 }
7276b3b0
SW
265 if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
266 return -EIO;
b3b94faa
DT
267
268 return 0;
269}
270
271/**
586dfdaa 272 * gfs2_attach_bufdata - attach a struct gfs2_bufdata structure to a buffer
b3b94faa
DT
273 * @gl: the glock the buffer belongs to
274 * @bh: The buffer to be attached to
586dfdaa 275 * @meta: Flag to indicate whether its metadata or not
b3b94faa
DT
276 */
277
568f4c96
SW
278void gfs2_attach_bufdata(struct gfs2_glock *gl, struct buffer_head *bh,
279 int meta)
b3b94faa
DT
280{
281 struct gfs2_bufdata *bd;
282
18ec7d5c
SW
283 if (meta)
284 lock_page(bh->b_page);
b3b94faa 285
5c676f6d 286 if (bh->b_private) {
18ec7d5c
SW
287 if (meta)
288 unlock_page(bh->b_page);
b3b94faa
DT
289 return;
290 }
291
3e5cd087 292 bd = kmem_cache_zalloc(gfs2_bufdata_cachep, GFP_NOFS | __GFP_NOFAIL);
b3b94faa
DT
293 bd->bd_bh = bh;
294 bd->bd_gl = gl;
295
296 INIT_LIST_HEAD(&bd->bd_list_tr);
82ffa516 297 if (meta)
586dfdaa 298 lops_init_le(&bd->bd_le, &gfs2_buf_lops);
82ffa516 299 else
586dfdaa 300 lops_init_le(&bd->bd_le, &gfs2_databuf_lops);
5c676f6d 301 bh->b_private = bd;
b3b94faa 302
18ec7d5c
SW
303 if (meta)
304 unlock_page(bh->b_page);
b3b94faa
DT
305}
306
16615be1
SW
307void gfs2_remove_from_journal(struct buffer_head *bh, struct gfs2_trans *tr, int meta)
308{
009d8518
SW
309 struct address_space *mapping = bh->b_page->mapping;
310 struct gfs2_sbd *sdp = gfs2_mapping2sbd(mapping);
16615be1 311 struct gfs2_bufdata *bd = bh->b_private;
009d8518 312
16615be1 313 if (test_clear_buffer_pinned(bh)) {
5e687eac 314 atomic_dec(&sdp->sd_log_pinned);
16615be1
SW
315 list_del_init(&bd->bd_le.le_list);
316 if (meta) {
317 gfs2_assert_warn(sdp, sdp->sd_log_num_buf);
318 sdp->sd_log_num_buf--;
319 tr->tr_num_buf_rm++;
320 } else {
321 gfs2_assert_warn(sdp, sdp->sd_log_num_databuf);
322 sdp->sd_log_num_databuf--;
323 tr->tr_num_databuf_rm++;
324 }
325 tr->tr_touched = 1;
326 brelse(bh);
327 }
328 if (bd) {
329 if (bd->bd_ail) {
f91a0d3e 330 gfs2_remove_from_ail(bd);
16615be1
SW
331 bh->b_private = NULL;
332 bd->bd_bh = NULL;
333 bd->bd_blkno = bh->b_blocknr;
334 gfs2_trans_add_revoke(sdp, bd);
335 }
336 }
337 clear_buffer_dirty(bh);
338 clear_buffer_uptodate(bh);
339}
340
b3b94faa
DT
341/**
342 * gfs2_meta_wipe - make inode's buffers so they aren't dirty/pinned anymore
343 * @ip: the inode who owns the buffers
344 * @bstart: the first buffer in the run
345 * @blen: the number of buffers in the run
346 *
347 */
348
cd915493 349void gfs2_meta_wipe(struct gfs2_inode *ip, u64 bstart, u32 blen)
b3b94faa 350{
feaa7bba 351 struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
b3b94faa
DT
352 struct buffer_head *bh;
353
354 while (blen) {
6802e340 355 bh = gfs2_getbuf(ip->i_gl, bstart, NO_CREATE);
b3b94faa 356 if (bh) {
1ad38c43
SW
357 lock_buffer(bh);
358 gfs2_log_lock(sdp);
16615be1 359 gfs2_remove_from_journal(bh, current->journal_info, 1);
1ad38c43 360 gfs2_log_unlock(sdp);
b3b94faa 361 unlock_buffer(bh);
b3b94faa
DT
362 brelse(bh);
363 }
364
365 bstart++;
366 blen--;
367 }
368}
369
b3b94faa
DT
370/**
371 * gfs2_meta_indirect_buffer - Get a metadata buffer
372 * @ip: The GFS2 inode
373 * @height: The level of this buf in the metadata (indir addr) tree (if any)
374 * @num: The block number (device relative) of the buffer
375 * @new: Non-zero if we may create a new buffer
376 * @bhp: the buffer is returned here
377 *
b3b94faa
DT
378 * Returns: errno
379 */
380
cd915493 381int gfs2_meta_indirect_buffer(struct gfs2_inode *ip, int height, u64 num,
b3b94faa
DT
382 int new, struct buffer_head **bhp)
383{
7276b3b0
SW
384 struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
385 struct gfs2_glock *gl = ip->i_gl;
f91a0d3e
SW
386 struct buffer_head *bh;
387 int ret = 0;
b3b94faa
DT
388
389 if (new) {
f91a0d3e
SW
390 BUG_ON(height == 0);
391 bh = gfs2_meta_new(gl, num);
d4e9c4c3 392 gfs2_trans_add_bh(ip->i_gl, bh, 1);
b3b94faa
DT
393 gfs2_metatype_set(bh, GFS2_METATYPE_IN, GFS2_FORMAT_IN);
394 gfs2_buffer_clear_tail(bh, sizeof(struct gfs2_meta_header));
7276b3b0
SW
395 } else {
396 u32 mtype = height ? GFS2_METATYPE_IN : GFS2_METATYPE_DI;
f91a0d3e
SW
397 ret = gfs2_meta_read(gl, num, DIO_WAIT, &bh);
398 if (ret == 0 && gfs2_metatype_check(sdp, bh, mtype)) {
399 brelse(bh);
400 ret = -EIO;
7276b3b0 401 }
b3b94faa 402 }
b3b94faa 403 *bhp = bh;
f91a0d3e 404 return ret;
b3b94faa
DT
405}
406
407/**
408 * gfs2_meta_ra - start readahead on an extent of a file
409 * @gl: the glock the blocks belong to
410 * @dblock: the starting disk block
411 * @extlen: the number of blocks in the extent
412 *
7276b3b0 413 * returns: the first buffer in the extent
b3b94faa
DT
414 */
415
7276b3b0 416struct buffer_head *gfs2_meta_ra(struct gfs2_glock *gl, u64 dblock, u32 extlen)
b3b94faa
DT
417{
418 struct gfs2_sbd *sdp = gl->gl_sbd;
b3b94faa 419 struct buffer_head *first_bh, *bh;
cd915493 420 u32 max_ra = gfs2_tune_get(sdp, gt_max_readahead) >>
568f4c96 421 sdp->sd_sb.sb_bsize_shift;
b3b94faa 422
7276b3b0
SW
423 BUG_ON(!extlen);
424
425 if (max_ra < 1)
426 max_ra = 1;
b3b94faa
DT
427 if (extlen > max_ra)
428 extlen = max_ra;
429
6802e340 430 first_bh = gfs2_getbuf(gl, dblock, CREATE);
b3b94faa
DT
431
432 if (buffer_uptodate(first_bh))
433 goto out;
7276b3b0 434 if (!buffer_locked(first_bh))
7b6d91da 435 ll_rw_block(READ_SYNC | REQ_META, 1, &first_bh);
b3b94faa
DT
436
437 dblock++;
438 extlen--;
439
440 while (extlen) {
6802e340 441 bh = gfs2_getbuf(gl, dblock, CREATE);
b3b94faa 442
7276b3b0
SW
443 if (!buffer_uptodate(bh) && !buffer_locked(bh))
444 ll_rw_block(READA, 1, &bh);
445 brelse(bh);
b3b94faa
DT
446 dblock++;
447 extlen--;
7276b3b0
SW
448 if (!buffer_locked(first_bh) && buffer_uptodate(first_bh))
449 goto out;
b3b94faa
DT
450 }
451
7276b3b0 452 wait_on_buffer(first_bh);
a91ea69f 453out:
7276b3b0 454 return first_bh;
b3b94faa
DT
455}
456