]>
Commit | Line | Data |
---|---|---|
b3b94faa DT |
1 | /* |
2 | * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. | |
3a8a9a10 | 3 | * Copyright (C) 2004-2006 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> | |
75ca61c1 | 15 | #include <linux/mempool.h> |
5c676f6d | 16 | #include <linux/gfs2_ondisk.h> |
c969f58c SW |
17 | #include <linux/bio.h> |
18 | #include <linux/fs.h> | |
b3b94faa DT |
19 | |
20 | #include "gfs2.h" | |
5c676f6d | 21 | #include "incore.h" |
2332c443 | 22 | #include "inode.h" |
b3b94faa DT |
23 | #include "glock.h" |
24 | #include "log.h" | |
25 | #include "lops.h" | |
26 | #include "meta_io.h" | |
27 | #include "recovery.h" | |
28 | #include "rgrp.h" | |
29 | #include "trans.h" | |
5c676f6d | 30 | #include "util.h" |
63997775 | 31 | #include "trace_gfs2.h" |
b3b94faa | 32 | |
9b9107a5 SW |
33 | /** |
34 | * gfs2_pin - Pin a buffer in memory | |
35 | * @sdp: The superblock | |
36 | * @bh: The buffer to be pinned | |
37 | * | |
38 | * The log lock must be held when calling this function | |
39 | */ | |
40 | static void gfs2_pin(struct gfs2_sbd *sdp, struct buffer_head *bh) | |
41 | { | |
42 | struct gfs2_bufdata *bd; | |
43 | ||
29687a2a | 44 | BUG_ON(!current->journal_info); |
9b9107a5 SW |
45 | |
46 | clear_buffer_dirty(bh); | |
47 | if (test_set_buffer_pinned(bh)) | |
48 | gfs2_assert_withdraw(sdp, 0); | |
49 | if (!buffer_uptodate(bh)) | |
50 | gfs2_io_error_bh(sdp, bh); | |
51 | bd = bh->b_private; | |
52 | /* If this buffer is in the AIL and it has already been written | |
53 | * to in-place disk block, remove it from the AIL. | |
54 | */ | |
c618e87a | 55 | spin_lock(&sdp->sd_ail_lock); |
9b9107a5 SW |
56 | if (bd->bd_ail) |
57 | list_move(&bd->bd_ail_st_list, &bd->bd_ail->ai_ail2_list); | |
c618e87a | 58 | spin_unlock(&sdp->sd_ail_lock); |
9b9107a5 | 59 | get_bh(bh); |
5e687eac | 60 | atomic_inc(&sdp->sd_log_pinned); |
63997775 | 61 | trace_gfs2_pin(bd, 1); |
9b9107a5 SW |
62 | } |
63 | ||
7c9ca621 BP |
64 | static bool buffer_is_rgrp(const struct gfs2_bufdata *bd) |
65 | { | |
66 | return bd->bd_gl->gl_name.ln_type == LM_TYPE_RGRP; | |
67 | } | |
68 | ||
69 | static void maybe_release_space(struct gfs2_bufdata *bd) | |
70 | { | |
71 | struct gfs2_glock *gl = bd->bd_gl; | |
72 | struct gfs2_sbd *sdp = gl->gl_sbd; | |
73 | struct gfs2_rgrpd *rgd = gl->gl_object; | |
74 | unsigned int index = bd->bd_bh->b_blocknr - gl->gl_name.ln_number; | |
75 | struct gfs2_bitmap *bi = rgd->rd_bits + index; | |
76 | ||
77 | if (bi->bi_clone == 0) | |
78 | return; | |
79 | if (sdp->sd_args.ar_discard) | |
66fc061b | 80 | gfs2_rgrp_send_discards(sdp, rgd->rd_data0, bd->bd_bh, bi, 1, NULL); |
7c9ca621 BP |
81 | memcpy(bi->bi_clone + bi->bi_offset, |
82 | bd->bd_bh->b_data + bi->bi_offset, bi->bi_len); | |
83 | clear_bit(GBF_FULL, &bi->bi_flags); | |
84 | rgd->rd_free_clone = rgd->rd_free; | |
85 | } | |
86 | ||
9b9107a5 SW |
87 | /** |
88 | * gfs2_unpin - Unpin a buffer | |
89 | * @sdp: the filesystem the buffer belongs to | |
90 | * @bh: The buffer to unpin | |
91 | * @ai: | |
29687a2a | 92 | * @flags: The inode dirty flags |
9b9107a5 SW |
93 | * |
94 | */ | |
95 | ||
96 | static void gfs2_unpin(struct gfs2_sbd *sdp, struct buffer_head *bh, | |
97 | struct gfs2_ail *ai) | |
98 | { | |
99 | struct gfs2_bufdata *bd = bh->b_private; | |
100 | ||
29687a2a SW |
101 | BUG_ON(!buffer_uptodate(bh)); |
102 | BUG_ON(!buffer_pinned(bh)); | |
9b9107a5 SW |
103 | |
104 | lock_buffer(bh); | |
105 | mark_buffer_dirty(bh); | |
106 | clear_buffer_pinned(bh); | |
107 | ||
7c9ca621 BP |
108 | if (buffer_is_rgrp(bd)) |
109 | maybe_release_space(bd); | |
110 | ||
d6a079e8 | 111 | spin_lock(&sdp->sd_ail_lock); |
9b9107a5 SW |
112 | if (bd->bd_ail) { |
113 | list_del(&bd->bd_ail_st_list); | |
114 | brelse(bh); | |
115 | } else { | |
116 | struct gfs2_glock *gl = bd->bd_gl; | |
117 | list_add(&bd->bd_ail_gl_list, &gl->gl_ail_list); | |
118 | atomic_inc(&gl->gl_ail_count); | |
119 | } | |
120 | bd->bd_ail = ai; | |
121 | list_add(&bd->bd_ail_st_list, &ai->ai_ail1_list); | |
d6a079e8 DC |
122 | spin_unlock(&sdp->sd_ail_lock); |
123 | ||
29687a2a | 124 | clear_bit(GLF_LFLUSH, &bd->bd_gl->gl_flags); |
63997775 | 125 | trace_gfs2_pin(bd, 0); |
9b9107a5 | 126 | unlock_buffer(bh); |
5e687eac | 127 | atomic_dec(&sdp->sd_log_pinned); |
9b9107a5 SW |
128 | } |
129 | ||
e8c92ed7 | 130 | static void gfs2_log_incr_head(struct gfs2_sbd *sdp) |
16615be1 | 131 | { |
e8c92ed7 SW |
132 | BUG_ON((sdp->sd_log_flush_head == sdp->sd_log_tail) && |
133 | (sdp->sd_log_flush_head != sdp->sd_log_head)); | |
134 | ||
135 | if (++sdp->sd_log_flush_head == sdp->sd_jdesc->jd_blocks) { | |
136 | sdp->sd_log_flush_head = 0; | |
137 | sdp->sd_log_flush_wrapped = 1; | |
138 | } | |
16615be1 SW |
139 | } |
140 | ||
e8c92ed7 | 141 | static u64 gfs2_log_bmap(struct gfs2_sbd *sdp) |
16615be1 | 142 | { |
e8c92ed7 SW |
143 | unsigned int lbn = sdp->sd_log_flush_head; |
144 | struct gfs2_journal_extent *je; | |
145 | u64 block; | |
146 | ||
147 | list_for_each_entry(je, &sdp->sd_jdesc->extent_list, extent_list) { | |
148 | if (lbn >= je->lblock && lbn < je->lblock + je->blocks) { | |
149 | block = je->dblock + lbn - je->lblock; | |
150 | gfs2_log_incr_head(sdp); | |
151 | return block; | |
152 | } | |
153 | } | |
154 | ||
155 | return -1; | |
16615be1 SW |
156 | } |
157 | ||
e8c92ed7 SW |
158 | /** |
159 | * gfs2_end_log_write_bh - end log write of pagecache data with buffers | |
160 | * @sdp: The superblock | |
161 | * @bvec: The bio_vec | |
162 | * @error: The i/o status | |
163 | * | |
164 | * This finds the relavent buffers and unlocks then and sets the | |
165 | * error flag according to the status of the i/o request. This is | |
166 | * used when the log is writing data which has an in-place version | |
167 | * that is pinned in the pagecache. | |
168 | */ | |
169 | ||
170 | static void gfs2_end_log_write_bh(struct gfs2_sbd *sdp, struct bio_vec *bvec, | |
171 | int error) | |
16615be1 | 172 | { |
e8c92ed7 SW |
173 | struct buffer_head *bh, *next; |
174 | struct page *page = bvec->bv_page; | |
175 | unsigned size; | |
176 | ||
177 | bh = page_buffers(page); | |
178 | size = bvec->bv_len; | |
179 | while (bh_offset(bh) < bvec->bv_offset) | |
180 | bh = bh->b_this_page; | |
181 | do { | |
182 | if (error) | |
183 | set_buffer_write_io_error(bh); | |
184 | unlock_buffer(bh); | |
185 | next = bh->b_this_page; | |
186 | size -= bh->b_size; | |
187 | brelse(bh); | |
188 | bh = next; | |
189 | } while(bh && size); | |
16615be1 SW |
190 | } |
191 | ||
47ac5537 | 192 | /** |
e8c92ed7 SW |
193 | * gfs2_end_log_write - end of i/o to the log |
194 | * @bio: The bio | |
195 | * @error: Status of i/o request | |
196 | * | |
197 | * Each bio_vec contains either data from the pagecache or data | |
198 | * relating to the log itself. Here we iterate over the bio_vec | |
199 | * array, processing both kinds of data. | |
47ac5537 SW |
200 | * |
201 | */ | |
202 | ||
e8c92ed7 | 203 | static void gfs2_end_log_write(struct bio *bio, int error) |
47ac5537 | 204 | { |
e8c92ed7 SW |
205 | struct gfs2_sbd *sdp = bio->bi_private; |
206 | struct bio_vec *bvec; | |
207 | struct page *page; | |
208 | int i; | |
209 | ||
210 | if (error) { | |
211 | sdp->sd_log_error = error; | |
212 | fs_err(sdp, "Error %d writing to log\n", error); | |
213 | } | |
214 | ||
215 | bio_for_each_segment(bvec, bio, i) { | |
216 | page = bvec->bv_page; | |
217 | if (page_has_buffers(page)) | |
218 | gfs2_end_log_write_bh(sdp, bvec, error); | |
219 | else | |
220 | mempool_free(page, gfs2_page_pool); | |
221 | } | |
47ac5537 | 222 | |
e8c92ed7 | 223 | bio_put(bio); |
47ac5537 SW |
224 | if (atomic_dec_and_test(&sdp->sd_log_in_flight)) |
225 | wake_up(&sdp->sd_log_flush_wait); | |
226 | } | |
227 | ||
228 | /** | |
e8c92ed7 SW |
229 | * gfs2_log_flush_bio - Submit any pending log bio |
230 | * @sdp: The superblock | |
231 | * @rw: The rw flags | |
47ac5537 | 232 | * |
e8c92ed7 SW |
233 | * Submit any pending part-built or full bio to the block device. If |
234 | * there is no pending bio, then this is a no-op. | |
47ac5537 SW |
235 | */ |
236 | ||
e8c92ed7 | 237 | void gfs2_log_flush_bio(struct gfs2_sbd *sdp, int rw) |
47ac5537 | 238 | { |
e8c92ed7 SW |
239 | if (sdp->sd_log_bio) { |
240 | atomic_inc(&sdp->sd_log_in_flight); | |
241 | submit_bio(rw, sdp->sd_log_bio); | |
242 | sdp->sd_log_bio = NULL; | |
243 | } | |
244 | } | |
47ac5537 | 245 | |
e8c92ed7 SW |
246 | /** |
247 | * gfs2_log_alloc_bio - Allocate a new bio for log writing | |
248 | * @sdp: The superblock | |
249 | * @blkno: The next device block number we want to write to | |
250 | * | |
251 | * This should never be called when there is a cached bio in the | |
252 | * super block. When it returns, there will be a cached bio in the | |
253 | * super block which will have as many bio_vecs as the device is | |
254 | * happy to handle. | |
255 | * | |
256 | * Returns: Newly allocated bio | |
257 | */ | |
258 | ||
259 | static struct bio *gfs2_log_alloc_bio(struct gfs2_sbd *sdp, u64 blkno) | |
260 | { | |
261 | struct super_block *sb = sdp->sd_vfs; | |
262 | unsigned nrvecs = bio_get_nr_vecs(sb->s_bdev); | |
263 | struct bio *bio; | |
264 | ||
265 | BUG_ON(sdp->sd_log_bio); | |
266 | ||
267 | while (1) { | |
268 | bio = bio_alloc(GFP_NOIO, nrvecs); | |
269 | if (likely(bio)) | |
270 | break; | |
271 | nrvecs = max(nrvecs/2, 1U); | |
272 | } | |
47ac5537 | 273 | |
e8c92ed7 SW |
274 | bio->bi_sector = blkno * (sb->s_blocksize >> 9); |
275 | bio->bi_bdev = sb->s_bdev; | |
276 | bio->bi_end_io = gfs2_end_log_write; | |
277 | bio->bi_private = sdp; | |
278 | ||
279 | sdp->sd_log_bio = bio; | |
280 | ||
281 | return bio; | |
47ac5537 SW |
282 | } |
283 | ||
284 | /** | |
e8c92ed7 SW |
285 | * gfs2_log_get_bio - Get cached log bio, or allocate a new one |
286 | * @sdp: The superblock | |
287 | * @blkno: The device block number we want to write to | |
288 | * | |
289 | * If there is a cached bio, then if the next block number is sequential | |
290 | * with the previous one, return it, otherwise flush the bio to the | |
291 | * device. If there is not a cached bio, or we just flushed it, then | |
292 | * allocate a new one. | |
47ac5537 | 293 | * |
e8c92ed7 | 294 | * Returns: The bio to use for log writes |
47ac5537 SW |
295 | */ |
296 | ||
e8c92ed7 | 297 | static struct bio *gfs2_log_get_bio(struct gfs2_sbd *sdp, u64 blkno) |
47ac5537 | 298 | { |
e8c92ed7 SW |
299 | struct bio *bio = sdp->sd_log_bio; |
300 | u64 nblk; | |
301 | ||
302 | if (bio) { | |
303 | nblk = bio->bi_sector + bio_sectors(bio); | |
304 | nblk >>= sdp->sd_fsb2bb_shift; | |
305 | if (blkno == nblk) | |
306 | return bio; | |
307 | gfs2_log_flush_bio(sdp, WRITE); | |
308 | } | |
309 | ||
310 | return gfs2_log_alloc_bio(sdp, blkno); | |
47ac5537 SW |
311 | } |
312 | ||
e8c92ed7 | 313 | |
47ac5537 | 314 | /** |
e8c92ed7 | 315 | * gfs2_log_write - write to log |
47ac5537 | 316 | * @sdp: the filesystem |
e8c92ed7 SW |
317 | * @page: the page to write |
318 | * @size: the size of the data to write | |
319 | * @offset: the offset within the page | |
47ac5537 | 320 | * |
e8c92ed7 SW |
321 | * Try and add the page segment to the current bio. If that fails, |
322 | * submit the current bio to the device and create a new one, and | |
323 | * then add the page segment to that. | |
47ac5537 SW |
324 | */ |
325 | ||
e8c92ed7 SW |
326 | static void gfs2_log_write(struct gfs2_sbd *sdp, struct page *page, |
327 | unsigned size, unsigned offset) | |
47ac5537 | 328 | { |
e8c92ed7 SW |
329 | u64 blkno = gfs2_log_bmap(sdp); |
330 | struct bio *bio; | |
331 | int ret; | |
332 | ||
333 | bio = gfs2_log_get_bio(sdp, blkno); | |
334 | ret = bio_add_page(bio, page, size, offset); | |
335 | if (ret == 0) { | |
336 | gfs2_log_flush_bio(sdp, WRITE); | |
337 | bio = gfs2_log_alloc_bio(sdp, blkno); | |
338 | ret = bio_add_page(bio, page, size, offset); | |
339 | WARN_ON(ret == 0); | |
340 | } | |
341 | } | |
47ac5537 | 342 | |
e8c92ed7 SW |
343 | /** |
344 | * gfs2_log_write_bh - write a buffer's content to the log | |
345 | * @sdp: The super block | |
346 | * @bh: The buffer pointing to the in-place location | |
347 | * | |
348 | * This writes the content of the buffer to the next available location | |
349 | * in the log. The buffer will be unlocked once the i/o to the log has | |
350 | * completed. | |
351 | */ | |
352 | ||
353 | static void gfs2_log_write_bh(struct gfs2_sbd *sdp, struct buffer_head *bh) | |
354 | { | |
355 | gfs2_log_write(sdp, bh->b_page, bh->b_size, bh_offset(bh)); | |
356 | } | |
47ac5537 | 357 | |
e8c92ed7 SW |
358 | /** |
359 | * gfs2_log_write_page - write one block stored in a page, into the log | |
360 | * @sdp: The superblock | |
361 | * @page: The struct page | |
362 | * | |
363 | * This writes the first block-sized part of the page into the log. Note | |
364 | * that the page must have been allocated from the gfs2_page_pool mempool | |
365 | * and that after this has been called, ownership has been transferred and | |
366 | * the page may be freed at any time. | |
367 | */ | |
47ac5537 | 368 | |
e8c92ed7 SW |
369 | void gfs2_log_write_page(struct gfs2_sbd *sdp, struct page *page) |
370 | { | |
371 | struct super_block *sb = sdp->sd_vfs; | |
372 | gfs2_log_write(sdp, page, sb->s_blocksize, 0); | |
47ac5537 | 373 | } |
16615be1 | 374 | |
dad30e90 SW |
375 | static struct page *gfs2_get_log_desc(struct gfs2_sbd *sdp, u32 ld_type, |
376 | u32 ld_length, u32 ld_data1) | |
16615be1 | 377 | { |
144a4c2f | 378 | struct page *page = mempool_alloc(gfs2_page_pool, GFP_NOIO); |
e8c92ed7 SW |
379 | struct gfs2_log_descriptor *ld = page_address(page); |
380 | clear_page(ld); | |
16615be1 SW |
381 | ld->ld_header.mh_magic = cpu_to_be32(GFS2_MAGIC); |
382 | ld->ld_header.mh_type = cpu_to_be32(GFS2_METATYPE_LD); | |
383 | ld->ld_header.mh_format = cpu_to_be32(GFS2_FORMAT_LD); | |
384 | ld->ld_type = cpu_to_be32(ld_type); | |
dad30e90 SW |
385 | ld->ld_length = cpu_to_be32(ld_length); |
386 | ld->ld_data1 = cpu_to_be32(ld_data1); | |
16615be1 | 387 | ld->ld_data2 = 0; |
e8c92ed7 | 388 | return page; |
16615be1 SW |
389 | } |
390 | ||
c0752aa7 | 391 | static void buf_lo_add(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd) |
b3b94faa | 392 | { |
0ab7d13f | 393 | struct gfs2_meta_header *mh; |
b3b94faa DT |
394 | struct gfs2_trans *tr; |
395 | ||
9b9107a5 | 396 | lock_buffer(bd->bd_bh); |
8bd95727 | 397 | gfs2_log_lock(sdp); |
5c676f6d | 398 | tr = current->journal_info; |
b3b94faa | 399 | tr->tr_touched = 1; |
c0752aa7 | 400 | if (!list_empty(&bd->bd_list)) |
9b9107a5 | 401 | goto out; |
2bcd610d SW |
402 | set_bit(GLF_LFLUSH, &bd->bd_gl->gl_flags); |
403 | set_bit(GLF_DIRTY, &bd->bd_gl->gl_flags); | |
0ab7d13f | 404 | mh = (struct gfs2_meta_header *)bd->bd_bh->b_data; |
1b8ba31a SW |
405 | if (unlikely(mh->mh_magic != cpu_to_be32(GFS2_MAGIC))) { |
406 | printk(KERN_ERR | |
407 | "Attempting to add uninitialised block to journal (inplace block=%lld)\n", | |
408 | (unsigned long long)bd->bd_bh->b_blocknr); | |
409 | BUG(); | |
410 | } | |
411 | gfs2_pin(sdp, bd->bd_bh); | |
0ab7d13f SW |
412 | mh->__pad0 = cpu_to_be64(0); |
413 | mh->mh_jid = cpu_to_be32(sdp->sd_jdesc->jd_jid); | |
b3b94faa | 414 | sdp->sd_log_num_buf++; |
c0752aa7 | 415 | list_add(&bd->bd_list, &sdp->sd_log_le_buf); |
b3b94faa | 416 | tr->tr_num_buf_new++; |
9b9107a5 SW |
417 | out: |
418 | gfs2_log_unlock(sdp); | |
419 | unlock_buffer(bd->bd_bh); | |
b3b94faa DT |
420 | } |
421 | ||
dad30e90 SW |
422 | static void gfs2_check_magic(struct buffer_head *bh) |
423 | { | |
424 | void *kaddr; | |
425 | __be32 *ptr; | |
426 | ||
427 | clear_buffer_escaped(bh); | |
428 | kaddr = kmap_atomic(bh->b_page); | |
429 | ptr = kaddr + bh_offset(bh); | |
430 | if (*ptr == cpu_to_be32(GFS2_MAGIC)) | |
431 | set_buffer_escaped(bh); | |
432 | kunmap_atomic(kaddr); | |
433 | } | |
434 | ||
435 | static void gfs2_before_commit(struct gfs2_sbd *sdp, unsigned int limit, | |
436 | unsigned int total, struct list_head *blist, | |
437 | bool is_databuf) | |
b3b94faa | 438 | { |
b3b94faa DT |
439 | struct gfs2_log_descriptor *ld; |
440 | struct gfs2_bufdata *bd1 = NULL, *bd2; | |
e8c92ed7 | 441 | struct page *page; |
b3b94faa DT |
442 | unsigned int num; |
443 | unsigned n; | |
444 | __be64 *ptr; | |
445 | ||
905d2aef | 446 | gfs2_log_lock(sdp); |
c0752aa7 | 447 | bd1 = bd2 = list_prepare_entry(bd1, blist, bd_list); |
b3b94faa DT |
448 | while(total) { |
449 | num = total; | |
450 | if (total > limit) | |
451 | num = limit; | |
905d2aef | 452 | gfs2_log_unlock(sdp); |
dad30e90 | 453 | page = gfs2_get_log_desc(sdp, GFS2_LOG_DESC_METADATA, num + 1, num); |
e8c92ed7 | 454 | ld = page_address(page); |
905d2aef | 455 | gfs2_log_lock(sdp); |
e8c92ed7 | 456 | ptr = (__be64 *)(ld + 1); |
b3b94faa DT |
457 | |
458 | n = 0; | |
c0752aa7 | 459 | list_for_each_entry_continue(bd1, blist, bd_list) { |
b3b94faa | 460 | *ptr++ = cpu_to_be64(bd1->bd_bh->b_blocknr); |
dad30e90 SW |
461 | if (is_databuf) { |
462 | gfs2_check_magic(bd1->bd_bh); | |
463 | *ptr++ = cpu_to_be64(buffer_escaped(bd1->bd_bh) ? 1 : 0); | |
464 | } | |
b3b94faa DT |
465 | if (++n >= num) |
466 | break; | |
467 | } | |
468 | ||
905d2aef | 469 | gfs2_log_unlock(sdp); |
e8c92ed7 | 470 | gfs2_log_write_page(sdp, page); |
905d2aef | 471 | gfs2_log_lock(sdp); |
b3b94faa DT |
472 | |
473 | n = 0; | |
c0752aa7 | 474 | list_for_each_entry_continue(bd2, blist, bd_list) { |
16615be1 | 475 | get_bh(bd2->bd_bh); |
905d2aef | 476 | gfs2_log_unlock(sdp); |
16615be1 | 477 | lock_buffer(bd2->bd_bh); |
dad30e90 SW |
478 | |
479 | if (buffer_escaped(bd2->bd_bh)) { | |
480 | void *kaddr; | |
481 | page = mempool_alloc(gfs2_page_pool, GFP_NOIO); | |
482 | ptr = page_address(page); | |
483 | kaddr = kmap_atomic(bd2->bd_bh->b_page); | |
484 | memcpy(ptr, kaddr + bh_offset(bd2->bd_bh), | |
485 | bd2->bd_bh->b_size); | |
486 | kunmap_atomic(kaddr); | |
487 | *(__be32 *)ptr = 0; | |
488 | clear_buffer_escaped(bd2->bd_bh); | |
489 | unlock_buffer(bd2->bd_bh); | |
490 | brelse(bd2->bd_bh); | |
491 | gfs2_log_write_page(sdp, page); | |
492 | } else { | |
493 | gfs2_log_write_bh(sdp, bd2->bd_bh); | |
494 | } | |
905d2aef | 495 | gfs2_log_lock(sdp); |
b3b94faa DT |
496 | if (++n >= num) |
497 | break; | |
498 | } | |
499 | ||
905d2aef | 500 | BUG_ON(total < num); |
b3b94faa DT |
501 | total -= num; |
502 | } | |
905d2aef | 503 | gfs2_log_unlock(sdp); |
b3b94faa DT |
504 | } |
505 | ||
dad30e90 SW |
506 | static void buf_lo_before_commit(struct gfs2_sbd *sdp) |
507 | { | |
508 | unsigned int limit = buf_limit(sdp); /* 503 for 4k blocks */ | |
509 | ||
510 | gfs2_before_commit(sdp, limit, sdp->sd_log_num_buf, | |
511 | &sdp->sd_log_le_buf, 0); | |
512 | } | |
513 | ||
b3b94faa DT |
514 | static void buf_lo_after_commit(struct gfs2_sbd *sdp, struct gfs2_ail *ai) |
515 | { | |
516 | struct list_head *head = &sdp->sd_log_le_buf; | |
517 | struct gfs2_bufdata *bd; | |
518 | ||
519 | while (!list_empty(head)) { | |
c0752aa7 BP |
520 | bd = list_entry(head->next, struct gfs2_bufdata, bd_list); |
521 | list_del_init(&bd->bd_list); | |
b3b94faa DT |
522 | sdp->sd_log_num_buf--; |
523 | ||
a98ab220 | 524 | gfs2_unpin(sdp, bd->bd_bh, ai); |
b3b94faa DT |
525 | } |
526 | gfs2_assert_warn(sdp, !sdp->sd_log_num_buf); | |
527 | } | |
528 | ||
529 | static void buf_lo_before_scan(struct gfs2_jdesc *jd, | |
55167622 | 530 | struct gfs2_log_header_host *head, int pass) |
b3b94faa | 531 | { |
feaa7bba | 532 | struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode); |
b3b94faa DT |
533 | |
534 | if (pass != 0) | |
535 | return; | |
536 | ||
537 | sdp->sd_found_blocks = 0; | |
538 | sdp->sd_replayed_blocks = 0; | |
539 | } | |
540 | ||
541 | static int buf_lo_scan_elements(struct gfs2_jdesc *jd, unsigned int start, | |
542 | struct gfs2_log_descriptor *ld, __be64 *ptr, | |
543 | int pass) | |
544 | { | |
feaa7bba SW |
545 | struct gfs2_inode *ip = GFS2_I(jd->jd_inode); |
546 | struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode); | |
5c676f6d | 547 | struct gfs2_glock *gl = ip->i_gl; |
b3b94faa DT |
548 | unsigned int blks = be32_to_cpu(ld->ld_data1); |
549 | struct buffer_head *bh_log, *bh_ip; | |
cd915493 | 550 | u64 blkno; |
b3b94faa DT |
551 | int error = 0; |
552 | ||
553 | if (pass != 1 || be32_to_cpu(ld->ld_type) != GFS2_LOG_DESC_METADATA) | |
554 | return 0; | |
555 | ||
556 | gfs2_replay_incr_blk(sdp, &start); | |
557 | ||
558 | for (; blks; gfs2_replay_incr_blk(sdp, &start), blks--) { | |
559 | blkno = be64_to_cpu(*ptr++); | |
560 | ||
561 | sdp->sd_found_blocks++; | |
562 | ||
563 | if (gfs2_revoke_check(sdp, blkno, start)) | |
564 | continue; | |
565 | ||
566 | error = gfs2_replay_read_block(jd, start, &bh_log); | |
82ffa516 SW |
567 | if (error) |
568 | return error; | |
b3b94faa DT |
569 | |
570 | bh_ip = gfs2_meta_new(gl, blkno); | |
571 | memcpy(bh_ip->b_data, bh_log->b_data, bh_log->b_size); | |
572 | ||
573 | if (gfs2_meta_check(sdp, bh_ip)) | |
574 | error = -EIO; | |
575 | else | |
576 | mark_buffer_dirty(bh_ip); | |
577 | ||
578 | brelse(bh_log); | |
579 | brelse(bh_ip); | |
580 | ||
581 | if (error) | |
582 | break; | |
583 | ||
584 | sdp->sd_replayed_blocks++; | |
585 | } | |
586 | ||
587 | return error; | |
588 | } | |
589 | ||
590 | static void buf_lo_after_scan(struct gfs2_jdesc *jd, int error, int pass) | |
591 | { | |
feaa7bba SW |
592 | struct gfs2_inode *ip = GFS2_I(jd->jd_inode); |
593 | struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode); | |
b3b94faa DT |
594 | |
595 | if (error) { | |
7276b3b0 | 596 | gfs2_meta_sync(ip->i_gl); |
b3b94faa DT |
597 | return; |
598 | } | |
599 | if (pass != 1) | |
600 | return; | |
601 | ||
7276b3b0 | 602 | gfs2_meta_sync(ip->i_gl); |
b3b94faa DT |
603 | |
604 | fs_info(sdp, "jid=%u: Replayed %u of %u blocks\n", | |
605 | jd->jd_jid, sdp->sd_replayed_blocks, sdp->sd_found_blocks); | |
606 | } | |
607 | ||
c0752aa7 | 608 | static void revoke_lo_add(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd) |
b3b94faa | 609 | { |
f42ab085 | 610 | struct gfs2_glock *gl = bd->bd_gl; |
b3b94faa DT |
611 | struct gfs2_trans *tr; |
612 | ||
5c676f6d | 613 | tr = current->journal_info; |
b3b94faa DT |
614 | tr->tr_touched = 1; |
615 | tr->tr_num_revoke++; | |
b3b94faa | 616 | sdp->sd_log_num_revoke++; |
f42ab085 SW |
617 | atomic_inc(&gl->gl_revokes); |
618 | set_bit(GLF_LFLUSH, &gl->gl_flags); | |
c0752aa7 | 619 | list_add(&bd->bd_list, &sdp->sd_log_le_revoke); |
b3b94faa DT |
620 | } |
621 | ||
622 | static void revoke_lo_before_commit(struct gfs2_sbd *sdp) | |
623 | { | |
624 | struct gfs2_log_descriptor *ld; | |
625 | struct gfs2_meta_header *mh; | |
b3b94faa DT |
626 | unsigned int offset; |
627 | struct list_head *head = &sdp->sd_log_le_revoke; | |
82e86087 | 628 | struct gfs2_bufdata *bd; |
e8c92ed7 | 629 | struct page *page; |
dad30e90 | 630 | unsigned int length; |
b3b94faa DT |
631 | |
632 | if (!sdp->sd_log_num_revoke) | |
633 | return; | |
634 | ||
dad30e90 SW |
635 | length = gfs2_struct2blk(sdp, sdp->sd_log_num_revoke, sizeof(u64)); |
636 | page = gfs2_get_log_desc(sdp, GFS2_LOG_DESC_REVOKE, length, sdp->sd_log_num_revoke); | |
e8c92ed7 | 637 | ld = page_address(page); |
b3b94faa DT |
638 | offset = sizeof(struct gfs2_log_descriptor); |
639 | ||
c0752aa7 | 640 | list_for_each_entry(bd, head, bd_list) { |
b3b94faa DT |
641 | sdp->sd_log_num_revoke--; |
642 | ||
cd915493 | 643 | if (offset + sizeof(u64) > sdp->sd_sb.sb_bsize) { |
b3b94faa | 644 | |
e8c92ed7 SW |
645 | gfs2_log_write_page(sdp, page); |
646 | page = mempool_alloc(gfs2_page_pool, GFP_NOIO); | |
647 | mh = page_address(page); | |
648 | clear_page(mh); | |
b3b94faa | 649 | mh->mh_magic = cpu_to_be32(GFS2_MAGIC); |
e3167ded SW |
650 | mh->mh_type = cpu_to_be32(GFS2_METATYPE_LB); |
651 | mh->mh_format = cpu_to_be32(GFS2_FORMAT_LB); | |
b3b94faa DT |
652 | offset = sizeof(struct gfs2_meta_header); |
653 | } | |
654 | ||
e8c92ed7 | 655 | *(__be64 *)(page_address(page) + offset) = cpu_to_be64(bd->bd_blkno); |
cd915493 | 656 | offset += sizeof(u64); |
b3b94faa DT |
657 | } |
658 | gfs2_assert_withdraw(sdp, !sdp->sd_log_num_revoke); | |
659 | ||
e8c92ed7 | 660 | gfs2_log_write_page(sdp, page); |
b3b94faa DT |
661 | } |
662 | ||
f42ab085 SW |
663 | static void revoke_lo_after_commit(struct gfs2_sbd *sdp, struct gfs2_ail *ai) |
664 | { | |
665 | struct list_head *head = &sdp->sd_log_le_revoke; | |
666 | struct gfs2_bufdata *bd; | |
667 | struct gfs2_glock *gl; | |
668 | ||
669 | while (!list_empty(head)) { | |
c0752aa7 BP |
670 | bd = list_entry(head->next, struct gfs2_bufdata, bd_list); |
671 | list_del_init(&bd->bd_list); | |
f42ab085 SW |
672 | gl = bd->bd_gl; |
673 | atomic_dec(&gl->gl_revokes); | |
674 | clear_bit(GLF_LFLUSH, &gl->gl_flags); | |
675 | kmem_cache_free(gfs2_bufdata_cachep, bd); | |
676 | } | |
677 | } | |
678 | ||
b3b94faa | 679 | static void revoke_lo_before_scan(struct gfs2_jdesc *jd, |
55167622 | 680 | struct gfs2_log_header_host *head, int pass) |
b3b94faa | 681 | { |
feaa7bba | 682 | struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode); |
b3b94faa DT |
683 | |
684 | if (pass != 0) | |
685 | return; | |
686 | ||
687 | sdp->sd_found_revokes = 0; | |
688 | sdp->sd_replay_tail = head->lh_tail; | |
689 | } | |
690 | ||
691 | static int revoke_lo_scan_elements(struct gfs2_jdesc *jd, unsigned int start, | |
692 | struct gfs2_log_descriptor *ld, __be64 *ptr, | |
693 | int pass) | |
694 | { | |
feaa7bba | 695 | struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode); |
b3b94faa DT |
696 | unsigned int blks = be32_to_cpu(ld->ld_length); |
697 | unsigned int revokes = be32_to_cpu(ld->ld_data1); | |
698 | struct buffer_head *bh; | |
699 | unsigned int offset; | |
cd915493 | 700 | u64 blkno; |
b3b94faa DT |
701 | int first = 1; |
702 | int error; | |
703 | ||
704 | if (pass != 0 || be32_to_cpu(ld->ld_type) != GFS2_LOG_DESC_REVOKE) | |
705 | return 0; | |
706 | ||
707 | offset = sizeof(struct gfs2_log_descriptor); | |
708 | ||
709 | for (; blks; gfs2_replay_incr_blk(sdp, &start), blks--) { | |
710 | error = gfs2_replay_read_block(jd, start, &bh); | |
711 | if (error) | |
712 | return error; | |
713 | ||
714 | if (!first) | |
715 | gfs2_metatype_check(sdp, bh, GFS2_METATYPE_LB); | |
716 | ||
cd915493 | 717 | while (offset + sizeof(u64) <= sdp->sd_sb.sb_bsize) { |
b3b94faa DT |
718 | blkno = be64_to_cpu(*(__be64 *)(bh->b_data + offset)); |
719 | ||
720 | error = gfs2_revoke_add(sdp, blkno, start); | |
3ad62e87 BP |
721 | if (error < 0) { |
722 | brelse(bh); | |
b3b94faa | 723 | return error; |
3ad62e87 | 724 | } |
b3b94faa DT |
725 | else if (error) |
726 | sdp->sd_found_revokes++; | |
727 | ||
728 | if (!--revokes) | |
729 | break; | |
cd915493 | 730 | offset += sizeof(u64); |
b3b94faa DT |
731 | } |
732 | ||
733 | brelse(bh); | |
734 | offset = sizeof(struct gfs2_meta_header); | |
735 | first = 0; | |
736 | } | |
737 | ||
738 | return 0; | |
739 | } | |
740 | ||
741 | static void revoke_lo_after_scan(struct gfs2_jdesc *jd, int error, int pass) | |
742 | { | |
feaa7bba | 743 | struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode); |
b3b94faa DT |
744 | |
745 | if (error) { | |
746 | gfs2_revoke_clean(sdp); | |
747 | return; | |
748 | } | |
749 | if (pass != 1) | |
750 | return; | |
751 | ||
752 | fs_info(sdp, "jid=%u: Found %u revoke tags\n", | |
753 | jd->jd_jid, sdp->sd_found_revokes); | |
754 | ||
755 | gfs2_revoke_clean(sdp); | |
756 | } | |
757 | ||
18ec7d5c SW |
758 | /** |
759 | * databuf_lo_add - Add a databuf to the transaction. | |
760 | * | |
761 | * This is used in two distinct cases: | |
762 | * i) In ordered write mode | |
763 | * We put the data buffer on a list so that we can ensure that its | |
764 | * synced to disk at the right time | |
765 | * ii) In journaled data mode | |
766 | * We need to journal the data block in the same way as metadata in | |
767 | * the functions above. The difference is that here we have a tag | |
768 | * which is two __be64's being the block number (as per meta data) | |
769 | * and a flag which says whether the data block needs escaping or | |
770 | * not. This means we need a new log entry for each 251 or so data | |
771 | * blocks, which isn't an enormous overhead but twice as much as | |
772 | * for normal metadata blocks. | |
773 | */ | |
c0752aa7 | 774 | static void databuf_lo_add(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd) |
b3b94faa | 775 | { |
5c676f6d | 776 | struct gfs2_trans *tr = current->journal_info; |
18ec7d5c | 777 | struct address_space *mapping = bd->bd_bh->b_page->mapping; |
feaa7bba | 778 | struct gfs2_inode *ip = GFS2_I(mapping->host); |
b3b94faa | 779 | |
9b9107a5 | 780 | lock_buffer(bd->bd_bh); |
8bd95727 | 781 | gfs2_log_lock(sdp); |
c50b91c4 | 782 | if (tr) |
9ff8ec32 | 783 | tr->tr_touched = 1; |
c0752aa7 | 784 | if (!list_empty(&bd->bd_list)) |
9b9107a5 | 785 | goto out; |
2bcd610d SW |
786 | set_bit(GLF_LFLUSH, &bd->bd_gl->gl_flags); |
787 | set_bit(GLF_DIRTY, &bd->bd_gl->gl_flags); | |
2332c443 | 788 | if (gfs2_is_jdata(ip)) { |
2332c443 RP |
789 | gfs2_pin(sdp, bd->bd_bh); |
790 | tr->tr_num_databuf_new++; | |
d7b616e2 | 791 | sdp->sd_log_num_databuf++; |
c0752aa7 | 792 | list_add_tail(&bd->bd_list, &sdp->sd_log_le_databuf); |
d7b616e2 | 793 | } else { |
c0752aa7 | 794 | list_add_tail(&bd->bd_list, &sdp->sd_log_le_ordered); |
9b9107a5 | 795 | } |
9b9107a5 | 796 | out: |
b3b94faa | 797 | gfs2_log_unlock(sdp); |
9b9107a5 | 798 | unlock_buffer(bd->bd_bh); |
b3b94faa DT |
799 | } |
800 | ||
16615be1 SW |
801 | /** |
802 | * databuf_lo_before_commit - Scan the data buffers, writing as we go | |
803 | * | |
804 | */ | |
805 | ||
806 | static void databuf_lo_before_commit(struct gfs2_sbd *sdp) | |
807 | { | |
dad30e90 | 808 | unsigned int limit = buf_limit(sdp) / 2; |
16615be1 | 809 | |
dad30e90 SW |
810 | gfs2_before_commit(sdp, limit, sdp->sd_log_num_databuf, |
811 | &sdp->sd_log_le_databuf, 1); | |
18ec7d5c SW |
812 | } |
813 | ||
814 | static int databuf_lo_scan_elements(struct gfs2_jdesc *jd, unsigned int start, | |
815 | struct gfs2_log_descriptor *ld, | |
816 | __be64 *ptr, int pass) | |
817 | { | |
feaa7bba SW |
818 | struct gfs2_inode *ip = GFS2_I(jd->jd_inode); |
819 | struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode); | |
5c676f6d | 820 | struct gfs2_glock *gl = ip->i_gl; |
18ec7d5c SW |
821 | unsigned int blks = be32_to_cpu(ld->ld_data1); |
822 | struct buffer_head *bh_log, *bh_ip; | |
cd915493 SW |
823 | u64 blkno; |
824 | u64 esc; | |
18ec7d5c SW |
825 | int error = 0; |
826 | ||
827 | if (pass != 1 || be32_to_cpu(ld->ld_type) != GFS2_LOG_DESC_JDATA) | |
828 | return 0; | |
829 | ||
830 | gfs2_replay_incr_blk(sdp, &start); | |
831 | for (; blks; gfs2_replay_incr_blk(sdp, &start), blks--) { | |
832 | blkno = be64_to_cpu(*ptr++); | |
833 | esc = be64_to_cpu(*ptr++); | |
834 | ||
835 | sdp->sd_found_blocks++; | |
836 | ||
837 | if (gfs2_revoke_check(sdp, blkno, start)) | |
838 | continue; | |
839 | ||
840 | error = gfs2_replay_read_block(jd, start, &bh_log); | |
841 | if (error) | |
842 | return error; | |
843 | ||
844 | bh_ip = gfs2_meta_new(gl, blkno); | |
845 | memcpy(bh_ip->b_data, bh_log->b_data, bh_log->b_size); | |
846 | ||
847 | /* Unescape */ | |
848 | if (esc) { | |
849 | __be32 *eptr = (__be32 *)bh_ip->b_data; | |
850 | *eptr = cpu_to_be32(GFS2_MAGIC); | |
851 | } | |
852 | mark_buffer_dirty(bh_ip); | |
853 | ||
854 | brelse(bh_log); | |
855 | brelse(bh_ip); | |
18ec7d5c SW |
856 | |
857 | sdp->sd_replayed_blocks++; | |
858 | } | |
859 | ||
860 | return error; | |
861 | } | |
862 | ||
863 | /* FIXME: sort out accounting for log blocks etc. */ | |
864 | ||
865 | static void databuf_lo_after_scan(struct gfs2_jdesc *jd, int error, int pass) | |
866 | { | |
feaa7bba SW |
867 | struct gfs2_inode *ip = GFS2_I(jd->jd_inode); |
868 | struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode); | |
18ec7d5c SW |
869 | |
870 | if (error) { | |
7276b3b0 | 871 | gfs2_meta_sync(ip->i_gl); |
18ec7d5c SW |
872 | return; |
873 | } | |
874 | if (pass != 1) | |
875 | return; | |
876 | ||
877 | /* data sync? */ | |
7276b3b0 | 878 | gfs2_meta_sync(ip->i_gl); |
18ec7d5c SW |
879 | |
880 | fs_info(sdp, "jid=%u: Replayed %u of %u data blocks\n", | |
881 | jd->jd_jid, sdp->sd_replayed_blocks, sdp->sd_found_blocks); | |
882 | } | |
883 | ||
884 | static void databuf_lo_after_commit(struct gfs2_sbd *sdp, struct gfs2_ail *ai) | |
885 | { | |
886 | struct list_head *head = &sdp->sd_log_le_databuf; | |
887 | struct gfs2_bufdata *bd; | |
888 | ||
889 | while (!list_empty(head)) { | |
c0752aa7 BP |
890 | bd = list_entry(head->next, struct gfs2_bufdata, bd_list); |
891 | list_del_init(&bd->bd_list); | |
18ec7d5c | 892 | sdp->sd_log_num_databuf--; |
18ec7d5c | 893 | gfs2_unpin(sdp, bd->bd_bh, ai); |
18ec7d5c | 894 | } |
b3b94faa DT |
895 | gfs2_assert_warn(sdp, !sdp->sd_log_num_databuf); |
896 | } | |
897 | ||
18ec7d5c | 898 | |
b09e593d | 899 | const struct gfs2_log_operations gfs2_buf_lops = { |
b3b94faa | 900 | .lo_add = buf_lo_add, |
b3b94faa DT |
901 | .lo_before_commit = buf_lo_before_commit, |
902 | .lo_after_commit = buf_lo_after_commit, | |
903 | .lo_before_scan = buf_lo_before_scan, | |
904 | .lo_scan_elements = buf_lo_scan_elements, | |
905 | .lo_after_scan = buf_lo_after_scan, | |
ea67eedb | 906 | .lo_name = "buf", |
b3b94faa DT |
907 | }; |
908 | ||
b09e593d | 909 | const struct gfs2_log_operations gfs2_revoke_lops = { |
b3b94faa DT |
910 | .lo_add = revoke_lo_add, |
911 | .lo_before_commit = revoke_lo_before_commit, | |
f42ab085 | 912 | .lo_after_commit = revoke_lo_after_commit, |
b3b94faa DT |
913 | .lo_before_scan = revoke_lo_before_scan, |
914 | .lo_scan_elements = revoke_lo_scan_elements, | |
915 | .lo_after_scan = revoke_lo_after_scan, | |
ea67eedb | 916 | .lo_name = "revoke", |
b3b94faa DT |
917 | }; |
918 | ||
b09e593d | 919 | const struct gfs2_log_operations gfs2_rg_lops = { |
ea67eedb | 920 | .lo_name = "rg", |
b3b94faa DT |
921 | }; |
922 | ||
b09e593d | 923 | const struct gfs2_log_operations gfs2_databuf_lops = { |
b3b94faa DT |
924 | .lo_add = databuf_lo_add, |
925 | .lo_before_commit = databuf_lo_before_commit, | |
18ec7d5c SW |
926 | .lo_after_commit = databuf_lo_after_commit, |
927 | .lo_scan_elements = databuf_lo_scan_elements, | |
928 | .lo_after_scan = databuf_lo_after_scan, | |
ea67eedb | 929 | .lo_name = "databuf", |
b3b94faa DT |
930 | }; |
931 | ||
b09e593d | 932 | const struct gfs2_log_operations *gfs2_log_ops[] = { |
16615be1 | 933 | &gfs2_databuf_lops, |
b3b94faa | 934 | &gfs2_buf_lops, |
b3b94faa | 935 | &gfs2_rg_lops, |
16615be1 | 936 | &gfs2_revoke_lops, |
ea67eedb | 937 | NULL, |
b3b94faa DT |
938 | }; |
939 |