2 * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
3 * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
5 * This copyrighted material is made available to anyone wishing to use,
6 * modify, copy, or redistribute it subject to the terms and conditions
7 * of the GNU General Public License version 2.
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>
16 #include <linux/pagemap.h>
17 #include <linux/writeback.h>
18 #include <linux/swap.h>
19 #include <linux/delay.h>
20 #include <linux/bio.h>
21 #include <linux/gfs2_ondisk.h>
35 static int gfs2_aspace_writepage(struct page *page, struct writeback_control *wbc)
38 struct buffer_head *bh, *head;
40 int write_op = (1 << BIO_RW_META) | ((wbc->sync_mode == WB_SYNC_ALL ?
41 WRITE_SYNC_PLUG : WRITE));
43 BUG_ON(!PageLocked(page));
44 BUG_ON(!page_has_buffers(page));
46 head = page_buffers(page);
50 if (!buffer_mapped(bh))
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
55 * potentially cause a busy-wait loop from pdflush and kswapd
56 * activity, but those code paths have their own higher-level
59 if (wbc->sync_mode != WB_SYNC_NONE || !wbc->nonblocking) {
61 } else if (!trylock_buffer(bh)) {
62 redirty_page_for_writepage(wbc, page);
65 if (test_clear_buffer_dirty(bh)) {
66 mark_buffer_async_write(bh);
70 } while ((bh = bh->b_this_page) != head);
73 * The page and its buffers are protected by PageWriteback(), so we can
74 * drop the bh refcounts early.
76 BUG_ON(PageWriteback(page));
77 set_page_writeback(page);
80 struct buffer_head *next = bh->b_this_page;
81 if (buffer_async_write(bh)) {
82 submit_bh(write_op, bh);
91 end_page_writeback(page);
96 const struct address_space_operations gfs2_meta_aops = {
97 .writepage = gfs2_aspace_writepage,
98 .releasepage = gfs2_releasepage,
99 .sync_page = block_sync_page,
103 * gfs2_meta_sync - Sync all buffers associated with a glock
108 void gfs2_meta_sync(struct gfs2_glock *gl)
110 struct address_space *mapping = gfs2_glock2aspace(gl);
113 filemap_fdatawrite(mapping);
114 error = filemap_fdatawait(mapping);
117 gfs2_io_error(gl->gl_sbd);
121 * gfs2_getbuf - Get a buffer with a given address space
123 * @blkno: the block number (filesystem scope)
124 * @create: 1 if the buffer should be created
126 * Returns: the buffer
129 struct buffer_head *gfs2_getbuf(struct gfs2_glock *gl, u64 blkno, int create)
131 struct address_space *mapping = gfs2_glock2aspace(gl);
132 struct gfs2_sbd *sdp = gl->gl_sbd;
134 struct buffer_head *bh;
139 shift = PAGE_CACHE_SHIFT - sdp->sd_sb.sb_bsize_shift;
140 index = blkno >> shift; /* convert block to page */
141 bufnum = blkno - (index << shift); /* block buf index within page */
145 page = grab_cache_page(mapping, index);
151 page = find_lock_page(mapping, index);
156 if (!page_has_buffers(page))
157 create_empty_buffers(page, sdp->sd_sb.sb_bsize, 0);
159 /* Locate header for our buffer within our page */
160 for (bh = page_buffers(page); bufnum--; bh = bh->b_this_page)
164 if (!buffer_mapped(bh))
165 map_bh(bh, sdp->sd_vfs, blkno);
168 mark_page_accessed(page);
169 page_cache_release(page);
174 static void meta_prep_new(struct buffer_head *bh)
176 struct gfs2_meta_header *mh = (struct gfs2_meta_header *)bh->b_data;
179 clear_buffer_dirty(bh);
180 set_buffer_uptodate(bh);
183 mh->mh_magic = cpu_to_be32(GFS2_MAGIC);
187 * gfs2_meta_new - Get a block
188 * @gl: The glock associated with this block
189 * @blkno: The block number
191 * Returns: The buffer
194 struct buffer_head *gfs2_meta_new(struct gfs2_glock *gl, u64 blkno)
196 struct buffer_head *bh;
197 bh = gfs2_getbuf(gl, blkno, CREATE);
203 * gfs2_meta_read - Read a block from disk
204 * @gl: The glock covering the block
205 * @blkno: The block number
207 * @bhp: the place where the buffer is returned (NULL on failure)
212 int gfs2_meta_read(struct gfs2_glock *gl, u64 blkno, int flags,
213 struct buffer_head **bhp)
215 struct gfs2_sbd *sdp = gl->gl_sbd;
216 struct buffer_head *bh;
218 if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
221 *bhp = bh = gfs2_getbuf(gl, blkno, CREATE);
224 if (buffer_uptodate(bh)) {
228 bh->b_end_io = end_buffer_read_sync;
230 submit_bh(READ_SYNC | (1 << BIO_RW_META), bh);
231 if (!(flags & DIO_WAIT))
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);
247 * gfs2_meta_wait - Reread a block from disk
248 * @sdp: the filesystem
249 * @bh: The block to wait for
254 int gfs2_meta_wait(struct gfs2_sbd *sdp, struct buffer_head *bh)
256 if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
261 if (!buffer_uptodate(bh)) {
262 struct gfs2_trans *tr = current->journal_info;
263 if (tr && tr->tr_touched)
264 gfs2_io_error_bh(sdp, bh);
267 if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
274 * gfs2_attach_bufdata - attach a struct gfs2_bufdata structure to a buffer
275 * @gl: the glock the buffer belongs to
276 * @bh: The buffer to be attached to
277 * @meta: Flag to indicate whether its metadata or not
280 void gfs2_attach_bufdata(struct gfs2_glock *gl, struct buffer_head *bh,
283 struct gfs2_bufdata *bd;
286 lock_page(bh->b_page);
290 unlock_page(bh->b_page);
294 bd = kmem_cache_zalloc(gfs2_bufdata_cachep, GFP_NOFS | __GFP_NOFAIL);
298 INIT_LIST_HEAD(&bd->bd_list_tr);
300 lops_init_le(&bd->bd_le, &gfs2_buf_lops);
302 lops_init_le(&bd->bd_le, &gfs2_databuf_lops);
306 unlock_page(bh->b_page);
309 void gfs2_remove_from_journal(struct buffer_head *bh, struct gfs2_trans *tr, int meta)
311 struct address_space *mapping = bh->b_page->mapping;
312 struct gfs2_sbd *sdp = gfs2_mapping2sbd(mapping);
313 struct gfs2_bufdata *bd = bh->b_private;
315 if (test_clear_buffer_pinned(bh)) {
316 list_del_init(&bd->bd_le.le_list);
318 gfs2_assert_warn(sdp, sdp->sd_log_num_buf);
319 sdp->sd_log_num_buf--;
322 gfs2_assert_warn(sdp, sdp->sd_log_num_databuf);
323 sdp->sd_log_num_databuf--;
324 tr->tr_num_databuf_rm++;
331 gfs2_remove_from_ail(bd);
332 bh->b_private = NULL;
334 bd->bd_blkno = bh->b_blocknr;
335 gfs2_trans_add_revoke(sdp, bd);
338 clear_buffer_dirty(bh);
339 clear_buffer_uptodate(bh);
343 * gfs2_meta_wipe - make inode's buffers so they aren't dirty/pinned anymore
344 * @ip: the inode who owns the buffers
345 * @bstart: the first buffer in the run
346 * @blen: the number of buffers in the run
350 void gfs2_meta_wipe(struct gfs2_inode *ip, u64 bstart, u32 blen)
352 struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
353 struct buffer_head *bh;
356 bh = gfs2_getbuf(ip->i_gl, bstart, NO_CREATE);
360 gfs2_remove_from_journal(bh, current->journal_info, 1);
361 gfs2_log_unlock(sdp);
372 * gfs2_meta_indirect_buffer - Get a metadata buffer
373 * @ip: The GFS2 inode
374 * @height: The level of this buf in the metadata (indir addr) tree (if any)
375 * @num: The block number (device relative) of the buffer
376 * @new: Non-zero if we may create a new buffer
377 * @bhp: the buffer is returned here
382 int gfs2_meta_indirect_buffer(struct gfs2_inode *ip, int height, u64 num,
383 int new, struct buffer_head **bhp)
385 struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
386 struct gfs2_glock *gl = ip->i_gl;
387 struct buffer_head *bh;
392 bh = gfs2_meta_new(gl, num);
393 gfs2_trans_add_bh(ip->i_gl, bh, 1);
394 gfs2_metatype_set(bh, GFS2_METATYPE_IN, GFS2_FORMAT_IN);
395 gfs2_buffer_clear_tail(bh, sizeof(struct gfs2_meta_header));
397 u32 mtype = height ? GFS2_METATYPE_IN : GFS2_METATYPE_DI;
398 ret = gfs2_meta_read(gl, num, DIO_WAIT, &bh);
399 if (ret == 0 && gfs2_metatype_check(sdp, bh, mtype)) {
409 * gfs2_meta_ra - start readahead on an extent of a file
410 * @gl: the glock the blocks belong to
411 * @dblock: the starting disk block
412 * @extlen: the number of blocks in the extent
414 * returns: the first buffer in the extent
417 struct buffer_head *gfs2_meta_ra(struct gfs2_glock *gl, u64 dblock, u32 extlen)
419 struct gfs2_sbd *sdp = gl->gl_sbd;
420 struct buffer_head *first_bh, *bh;
421 u32 max_ra = gfs2_tune_get(sdp, gt_max_readahead) >>
422 sdp->sd_sb.sb_bsize_shift;
431 first_bh = gfs2_getbuf(gl, dblock, CREATE);
433 if (buffer_uptodate(first_bh))
435 if (!buffer_locked(first_bh))
436 ll_rw_block(READ_SYNC | (1 << BIO_RW_META), 1, &first_bh);
442 bh = gfs2_getbuf(gl, dblock, CREATE);
444 if (!buffer_uptodate(bh) && !buffer_locked(bh))
445 ll_rw_block(READA, 1, &bh);
449 if (!buffer_locked(first_bh) && buffer_uptodate(first_bh))
453 wait_on_buffer(first_bh);