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> | |
7f63257d | 19 | #include <linux/list_sort.h> |
b3b94faa | 20 | |
c1696fb8 | 21 | #include "dir.h" |
b3b94faa | 22 | #include "gfs2.h" |
5c676f6d | 23 | #include "incore.h" |
2332c443 | 24 | #include "inode.h" |
b3b94faa DT |
25 | #include "glock.h" |
26 | #include "log.h" | |
27 | #include "lops.h" | |
28 | #include "meta_io.h" | |
29 | #include "recovery.h" | |
30 | #include "rgrp.h" | |
31 | #include "trans.h" | |
5c676f6d | 32 | #include "util.h" |
63997775 | 33 | #include "trace_gfs2.h" |
b3b94faa | 34 | |
9b9107a5 SW |
35 | /** |
36 | * gfs2_pin - Pin a buffer in memory | |
37 | * @sdp: The superblock | |
38 | * @bh: The buffer to be pinned | |
39 | * | |
40 | * The log lock must be held when calling this function | |
41 | */ | |
767f433f | 42 | void gfs2_pin(struct gfs2_sbd *sdp, struct buffer_head *bh) |
9b9107a5 SW |
43 | { |
44 | struct gfs2_bufdata *bd; | |
45 | ||
29687a2a | 46 | BUG_ON(!current->journal_info); |
9b9107a5 SW |
47 | |
48 | clear_buffer_dirty(bh); | |
49 | if (test_set_buffer_pinned(bh)) | |
50 | gfs2_assert_withdraw(sdp, 0); | |
51 | if (!buffer_uptodate(bh)) | |
9e1a9ecd | 52 | gfs2_io_error_bh_wd(sdp, bh); |
9b9107a5 SW |
53 | bd = bh->b_private; |
54 | /* If this buffer is in the AIL and it has already been written | |
55 | * to in-place disk block, remove it from the AIL. | |
56 | */ | |
c618e87a | 57 | spin_lock(&sdp->sd_ail_lock); |
16ca9412 BM |
58 | if (bd->bd_tr) |
59 | list_move(&bd->bd_ail_st_list, &bd->bd_tr->tr_ail2_list); | |
c618e87a | 60 | spin_unlock(&sdp->sd_ail_lock); |
9b9107a5 | 61 | get_bh(bh); |
5e687eac | 62 | atomic_inc(&sdp->sd_log_pinned); |
63997775 | 63 | trace_gfs2_pin(bd, 1); |
9b9107a5 SW |
64 | } |
65 | ||
7c9ca621 BP |
66 | static bool buffer_is_rgrp(const struct gfs2_bufdata *bd) |
67 | { | |
68 | return bd->bd_gl->gl_name.ln_type == LM_TYPE_RGRP; | |
69 | } | |
70 | ||
71 | static void maybe_release_space(struct gfs2_bufdata *bd) | |
72 | { | |
73 | struct gfs2_glock *gl = bd->bd_gl; | |
15562c43 | 74 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
6f6597ba | 75 | struct gfs2_rgrpd *rgd = gfs2_glock2rgrp(gl); |
7c9ca621 BP |
76 | unsigned int index = bd->bd_bh->b_blocknr - gl->gl_name.ln_number; |
77 | struct gfs2_bitmap *bi = rgd->rd_bits + index; | |
78 | ||
5a7c6690 | 79 | if (bi->bi_clone == NULL) |
7c9ca621 BP |
80 | return; |
81 | if (sdp->sd_args.ar_discard) | |
66fc061b | 82 | gfs2_rgrp_send_discards(sdp, rgd->rd_data0, bd->bd_bh, bi, 1, NULL); |
7c9ca621 | 83 | memcpy(bi->bi_clone + bi->bi_offset, |
281b4952 | 84 | bd->bd_bh->b_data + bi->bi_offset, bi->bi_bytes); |
7c9ca621 BP |
85 | clear_bit(GBF_FULL, &bi->bi_flags); |
86 | rgd->rd_free_clone = rgd->rd_free; | |
5ea5050c | 87 | rgd->rd_extfail_pt = rgd->rd_free; |
7c9ca621 BP |
88 | } |
89 | ||
9b9107a5 SW |
90 | /** |
91 | * gfs2_unpin - Unpin a buffer | |
92 | * @sdp: the filesystem the buffer belongs to | |
93 | * @bh: The buffer to unpin | |
94 | * @ai: | |
29687a2a | 95 | * @flags: The inode dirty flags |
9b9107a5 SW |
96 | * |
97 | */ | |
98 | ||
99 | static void gfs2_unpin(struct gfs2_sbd *sdp, struct buffer_head *bh, | |
16ca9412 | 100 | struct gfs2_trans *tr) |
9b9107a5 SW |
101 | { |
102 | struct gfs2_bufdata *bd = bh->b_private; | |
103 | ||
29687a2a SW |
104 | BUG_ON(!buffer_uptodate(bh)); |
105 | BUG_ON(!buffer_pinned(bh)); | |
9b9107a5 SW |
106 | |
107 | lock_buffer(bh); | |
108 | mark_buffer_dirty(bh); | |
109 | clear_buffer_pinned(bh); | |
110 | ||
7c9ca621 BP |
111 | if (buffer_is_rgrp(bd)) |
112 | maybe_release_space(bd); | |
113 | ||
d6a079e8 | 114 | spin_lock(&sdp->sd_ail_lock); |
16ca9412 | 115 | if (bd->bd_tr) { |
9b9107a5 SW |
116 | list_del(&bd->bd_ail_st_list); |
117 | brelse(bh); | |
118 | } else { | |
119 | struct gfs2_glock *gl = bd->bd_gl; | |
120 | list_add(&bd->bd_ail_gl_list, &gl->gl_ail_list); | |
121 | atomic_inc(&gl->gl_ail_count); | |
122 | } | |
16ca9412 BM |
123 | bd->bd_tr = tr; |
124 | list_add(&bd->bd_ail_st_list, &tr->tr_ail1_list); | |
d6a079e8 DC |
125 | spin_unlock(&sdp->sd_ail_lock); |
126 | ||
29687a2a | 127 | clear_bit(GLF_LFLUSH, &bd->bd_gl->gl_flags); |
63997775 | 128 | trace_gfs2_pin(bd, 0); |
9b9107a5 | 129 | unlock_buffer(bh); |
5e687eac | 130 | atomic_dec(&sdp->sd_log_pinned); |
9b9107a5 SW |
131 | } |
132 | ||
e8c92ed7 | 133 | static void gfs2_log_incr_head(struct gfs2_sbd *sdp) |
16615be1 | 134 | { |
e8c92ed7 SW |
135 | BUG_ON((sdp->sd_log_flush_head == sdp->sd_log_tail) && |
136 | (sdp->sd_log_flush_head != sdp->sd_log_head)); | |
137 | ||
722f6f62 | 138 | if (++sdp->sd_log_flush_head == sdp->sd_jdesc->jd_blocks) |
e8c92ed7 | 139 | sdp->sd_log_flush_head = 0; |
16615be1 SW |
140 | } |
141 | ||
c1696fb8 | 142 | u64 gfs2_log_bmap(struct gfs2_sbd *sdp) |
16615be1 | 143 | { |
e8c92ed7 SW |
144 | unsigned int lbn = sdp->sd_log_flush_head; |
145 | struct gfs2_journal_extent *je; | |
146 | u64 block; | |
147 | ||
b50f227b SW |
148 | list_for_each_entry(je, &sdp->sd_jdesc->extent_list, list) { |
149 | if ((lbn >= je->lblock) && (lbn < (je->lblock + je->blocks))) { | |
e8c92ed7 SW |
150 | block = je->dblock + lbn - je->lblock; |
151 | gfs2_log_incr_head(sdp); | |
152 | return block; | |
153 | } | |
154 | } | |
155 | ||
156 | return -1; | |
16615be1 SW |
157 | } |
158 | ||
e8c92ed7 SW |
159 | /** |
160 | * gfs2_end_log_write_bh - end log write of pagecache data with buffers | |
161 | * @sdp: The superblock | |
162 | * @bvec: The bio_vec | |
163 | * @error: The i/o status | |
164 | * | |
4519eaad | 165 | * This finds the relevant buffers and unlocks them and sets the |
e8c92ed7 SW |
166 | * error flag according to the status of the i/o request. This is |
167 | * used when the log is writing data which has an in-place version | |
168 | * that is pinned in the pagecache. | |
169 | */ | |
170 | ||
6dc4f100 ML |
171 | static void gfs2_end_log_write_bh(struct gfs2_sbd *sdp, |
172 | struct bio_vec *bvec, | |
4e4cbee9 | 173 | blk_status_t error) |
16615be1 | 174 | { |
e8c92ed7 SW |
175 | struct buffer_head *bh, *next; |
176 | struct page *page = bvec->bv_page; | |
177 | unsigned size; | |
178 | ||
179 | bh = page_buffers(page); | |
180 | size = bvec->bv_len; | |
181 | while (bh_offset(bh) < bvec->bv_offset) | |
182 | bh = bh->b_this_page; | |
183 | do { | |
184 | if (error) | |
87354e5d | 185 | mark_buffer_write_io_error(bh); |
e8c92ed7 SW |
186 | unlock_buffer(bh); |
187 | next = bh->b_this_page; | |
188 | size -= bh->b_size; | |
189 | brelse(bh); | |
190 | bh = next; | |
191 | } while(bh && size); | |
16615be1 SW |
192 | } |
193 | ||
47ac5537 | 194 | /** |
e8c92ed7 SW |
195 | * gfs2_end_log_write - end of i/o to the log |
196 | * @bio: The bio | |
23e93c9b | 197 | * @error: Status of i/o request |
e8c92ed7 SW |
198 | * |
199 | * Each bio_vec contains either data from the pagecache or data | |
200 | * relating to the log itself. Here we iterate over the bio_vec | |
201 | * array, processing both kinds of data. | |
47ac5537 SW |
202 | * |
203 | */ | |
204 | ||
4246a0b6 | 205 | static void gfs2_end_log_write(struct bio *bio) |
47ac5537 | 206 | { |
e8c92ed7 SW |
207 | struct gfs2_sbd *sdp = bio->bi_private; |
208 | struct bio_vec *bvec; | |
209 | struct page *page; | |
210 | int i; | |
6dc4f100 | 211 | struct bvec_iter_all iter_all; |
e8c92ed7 | 212 | |
942b0cdd BP |
213 | if (bio->bi_status) { |
214 | fs_err(sdp, "Error %d writing to journal, jid=%u\n", | |
215 | bio->bi_status, sdp->sd_jdesc->jd_jid); | |
216 | wake_up(&sdp->sd_logd_waitq); | |
217 | } | |
e8c92ed7 | 218 | |
6dc4f100 | 219 | bio_for_each_segment_all(bvec, bio, i, iter_all) { |
e8c92ed7 SW |
220 | page = bvec->bv_page; |
221 | if (page_has_buffers(page)) | |
4e4cbee9 | 222 | gfs2_end_log_write_bh(sdp, bvec, bio->bi_status); |
e8c92ed7 SW |
223 | else |
224 | mempool_free(page, gfs2_page_pool); | |
225 | } | |
47ac5537 | 226 | |
e8c92ed7 | 227 | bio_put(bio); |
47ac5537 SW |
228 | if (atomic_dec_and_test(&sdp->sd_log_in_flight)) |
229 | wake_up(&sdp->sd_log_flush_wait); | |
230 | } | |
231 | ||
232 | /** | |
5b846095 AD |
233 | * gfs2_log_submit_bio - Submit any pending log bio |
234 | * @biop: Address of the bio pointer | |
23e93c9b BP |
235 | * @op: REQ_OP |
236 | * @op_flags: req_flag_bits | |
47ac5537 | 237 | * |
e8c92ed7 SW |
238 | * Submit any pending part-built or full bio to the block device. If |
239 | * there is no pending bio, then this is a no-op. | |
47ac5537 SW |
240 | */ |
241 | ||
23e93c9b | 242 | void gfs2_log_submit_bio(struct bio **biop, int op, int op_flags) |
47ac5537 | 243 | { |
5b846095 AD |
244 | struct bio *bio = *biop; |
245 | if (bio) { | |
246 | struct gfs2_sbd *sdp = bio->bi_private; | |
e8c92ed7 | 247 | atomic_inc(&sdp->sd_log_in_flight); |
23e93c9b | 248 | bio_set_op_attrs(bio, op, op_flags); |
5b846095 AD |
249 | submit_bio(bio); |
250 | *biop = NULL; | |
e8c92ed7 SW |
251 | } |
252 | } | |
47ac5537 | 253 | |
e8c92ed7 | 254 | /** |
5b846095 AD |
255 | * gfs2_log_alloc_bio - Allocate a bio |
256 | * @sdp: The super block | |
257 | * @blkno: The device block number we want to write to | |
258 | * @end_io: The bi_end_io callback | |
e8c92ed7 | 259 | * |
5b846095 | 260 | * Allocate a new bio, initialize it with the given parameters and return it. |
e8c92ed7 | 261 | * |
5b846095 | 262 | * Returns: The newly allocated bio |
e8c92ed7 SW |
263 | */ |
264 | ||
5b846095 AD |
265 | static struct bio *gfs2_log_alloc_bio(struct gfs2_sbd *sdp, u64 blkno, |
266 | bio_end_io_t *end_io) | |
e8c92ed7 SW |
267 | { |
268 | struct super_block *sb = sdp->sd_vfs; | |
5b846095 | 269 | struct bio *bio = bio_alloc(GFP_NOIO, BIO_MAX_PAGES); |
e8c92ed7 | 270 | |
4f024f37 | 271 | bio->bi_iter.bi_sector = blkno * (sb->s_blocksize >> 9); |
74d46992 | 272 | bio_set_dev(bio, sb->s_bdev); |
5b846095 | 273 | bio->bi_end_io = end_io; |
e8c92ed7 SW |
274 | bio->bi_private = sdp; |
275 | ||
e8c92ed7 | 276 | return bio; |
47ac5537 SW |
277 | } |
278 | ||
279 | /** | |
e8c92ed7 | 280 | * gfs2_log_get_bio - Get cached log bio, or allocate a new one |
5b846095 | 281 | * @sdp: The super block |
e8c92ed7 | 282 | * @blkno: The device block number we want to write to |
5b846095 AD |
283 | * @bio: The bio to get or allocate |
284 | * @op: REQ_OP | |
285 | * @end_io: The bi_end_io callback | |
286 | * @flush: Always flush the current bio and allocate a new one? | |
e8c92ed7 SW |
287 | * |
288 | * If there is a cached bio, then if the next block number is sequential | |
289 | * with the previous one, return it, otherwise flush the bio to the | |
5b846095 | 290 | * device. If there is no cached bio, or we just flushed it, then |
e8c92ed7 | 291 | * allocate a new one. |
47ac5537 | 292 | * |
e8c92ed7 | 293 | * Returns: The bio to use for log writes |
47ac5537 SW |
294 | */ |
295 | ||
5b846095 AD |
296 | static struct bio *gfs2_log_get_bio(struct gfs2_sbd *sdp, u64 blkno, |
297 | struct bio **biop, int op, | |
298 | bio_end_io_t *end_io, bool flush) | |
47ac5537 | 299 | { |
5b846095 | 300 | struct bio *bio = *biop; |
e8c92ed7 SW |
301 | |
302 | if (bio) { | |
5b846095 AD |
303 | u64 nblk; |
304 | ||
f73a1c7d | 305 | nblk = bio_end_sector(bio); |
e8c92ed7 | 306 | nblk >>= sdp->sd_fsb2bb_shift; |
5b846095 | 307 | if (blkno == nblk && !flush) |
e8c92ed7 | 308 | return bio; |
23e93c9b | 309 | gfs2_log_submit_bio(biop, op, 0); |
e8c92ed7 SW |
310 | } |
311 | ||
5b846095 AD |
312 | *biop = gfs2_log_alloc_bio(sdp, blkno, end_io); |
313 | return *biop; | |
47ac5537 SW |
314 | } |
315 | ||
316 | /** | |
e8c92ed7 | 317 | * gfs2_log_write - write to log |
47ac5537 | 318 | * @sdp: the filesystem |
e8c92ed7 SW |
319 | * @page: the page to write |
320 | * @size: the size of the data to write | |
321 | * @offset: the offset within the page | |
c1696fb8 | 322 | * @blkno: block number of the log entry |
47ac5537 | 323 | * |
e8c92ed7 SW |
324 | * Try and add the page segment to the current bio. If that fails, |
325 | * submit the current bio to the device and create a new one, and | |
326 | * then add the page segment to that. | |
47ac5537 SW |
327 | */ |
328 | ||
c1696fb8 BP |
329 | void gfs2_log_write(struct gfs2_sbd *sdp, struct page *page, |
330 | unsigned size, unsigned offset, u64 blkno) | |
47ac5537 | 331 | { |
e8c92ed7 SW |
332 | struct bio *bio; |
333 | int ret; | |
334 | ||
5b846095 AD |
335 | bio = gfs2_log_get_bio(sdp, blkno, &sdp->sd_log_bio, REQ_OP_WRITE, |
336 | gfs2_end_log_write, false); | |
e8c92ed7 SW |
337 | ret = bio_add_page(bio, page, size, offset); |
338 | if (ret == 0) { | |
5b846095 AD |
339 | bio = gfs2_log_get_bio(sdp, blkno, &sdp->sd_log_bio, |
340 | REQ_OP_WRITE, gfs2_end_log_write, true); | |
e8c92ed7 SW |
341 | ret = bio_add_page(bio, page, size, offset); |
342 | WARN_ON(ret == 0); | |
343 | } | |
344 | } | |
47ac5537 | 345 | |
e8c92ed7 SW |
346 | /** |
347 | * gfs2_log_write_bh - write a buffer's content to the log | |
348 | * @sdp: The super block | |
349 | * @bh: The buffer pointing to the in-place location | |
350 | * | |
351 | * This writes the content of the buffer to the next available location | |
352 | * in the log. The buffer will be unlocked once the i/o to the log has | |
353 | * completed. | |
354 | */ | |
355 | ||
356 | static void gfs2_log_write_bh(struct gfs2_sbd *sdp, struct buffer_head *bh) | |
357 | { | |
c1696fb8 BP |
358 | gfs2_log_write(sdp, bh->b_page, bh->b_size, bh_offset(bh), |
359 | gfs2_log_bmap(sdp)); | |
e8c92ed7 | 360 | } |
47ac5537 | 361 | |
e8c92ed7 SW |
362 | /** |
363 | * gfs2_log_write_page - write one block stored in a page, into the log | |
364 | * @sdp: The superblock | |
365 | * @page: The struct page | |
366 | * | |
367 | * This writes the first block-sized part of the page into the log. Note | |
368 | * that the page must have been allocated from the gfs2_page_pool mempool | |
369 | * and that after this has been called, ownership has been transferred and | |
370 | * the page may be freed at any time. | |
371 | */ | |
47ac5537 | 372 | |
e8c92ed7 SW |
373 | void gfs2_log_write_page(struct gfs2_sbd *sdp, struct page *page) |
374 | { | |
375 | struct super_block *sb = sdp->sd_vfs; | |
c1696fb8 BP |
376 | gfs2_log_write(sdp, page, sb->s_blocksize, 0, |
377 | gfs2_log_bmap(sdp)); | |
47ac5537 | 378 | } |
16615be1 | 379 | |
dad30e90 SW |
380 | static struct page *gfs2_get_log_desc(struct gfs2_sbd *sdp, u32 ld_type, |
381 | u32 ld_length, u32 ld_data1) | |
16615be1 | 382 | { |
144a4c2f | 383 | struct page *page = mempool_alloc(gfs2_page_pool, GFP_NOIO); |
e8c92ed7 SW |
384 | struct gfs2_log_descriptor *ld = page_address(page); |
385 | clear_page(ld); | |
16615be1 SW |
386 | ld->ld_header.mh_magic = cpu_to_be32(GFS2_MAGIC); |
387 | ld->ld_header.mh_type = cpu_to_be32(GFS2_METATYPE_LD); | |
388 | ld->ld_header.mh_format = cpu_to_be32(GFS2_FORMAT_LD); | |
389 | ld->ld_type = cpu_to_be32(ld_type); | |
dad30e90 SW |
390 | ld->ld_length = cpu_to_be32(ld_length); |
391 | ld->ld_data1 = cpu_to_be32(ld_data1); | |
16615be1 | 392 | ld->ld_data2 = 0; |
e8c92ed7 | 393 | return page; |
16615be1 SW |
394 | } |
395 | ||
dad30e90 SW |
396 | static void gfs2_check_magic(struct buffer_head *bh) |
397 | { | |
398 | void *kaddr; | |
399 | __be32 *ptr; | |
400 | ||
401 | clear_buffer_escaped(bh); | |
402 | kaddr = kmap_atomic(bh->b_page); | |
403 | ptr = kaddr + bh_offset(bh); | |
404 | if (*ptr == cpu_to_be32(GFS2_MAGIC)) | |
405 | set_buffer_escaped(bh); | |
406 | kunmap_atomic(kaddr); | |
407 | } | |
408 | ||
7f63257d BM |
409 | static int blocknr_cmp(void *priv, struct list_head *a, struct list_head *b) |
410 | { | |
411 | struct gfs2_bufdata *bda, *bdb; | |
412 | ||
413 | bda = list_entry(a, struct gfs2_bufdata, bd_list); | |
414 | bdb = list_entry(b, struct gfs2_bufdata, bd_list); | |
415 | ||
416 | if (bda->bd_bh->b_blocknr < bdb->bd_bh->b_blocknr) | |
417 | return -1; | |
418 | if (bda->bd_bh->b_blocknr > bdb->bd_bh->b_blocknr) | |
419 | return 1; | |
420 | return 0; | |
421 | } | |
422 | ||
dad30e90 SW |
423 | static void gfs2_before_commit(struct gfs2_sbd *sdp, unsigned int limit, |
424 | unsigned int total, struct list_head *blist, | |
425 | bool is_databuf) | |
b3b94faa | 426 | { |
b3b94faa DT |
427 | struct gfs2_log_descriptor *ld; |
428 | struct gfs2_bufdata *bd1 = NULL, *bd2; | |
e8c92ed7 | 429 | struct page *page; |
b3b94faa DT |
430 | unsigned int num; |
431 | unsigned n; | |
432 | __be64 *ptr; | |
433 | ||
905d2aef | 434 | gfs2_log_lock(sdp); |
7f63257d | 435 | list_sort(NULL, blist, blocknr_cmp); |
c0752aa7 | 436 | bd1 = bd2 = list_prepare_entry(bd1, blist, bd_list); |
b3b94faa DT |
437 | while(total) { |
438 | num = total; | |
439 | if (total > limit) | |
440 | num = limit; | |
905d2aef | 441 | gfs2_log_unlock(sdp); |
4a586812 BP |
442 | page = gfs2_get_log_desc(sdp, |
443 | is_databuf ? GFS2_LOG_DESC_JDATA : | |
444 | GFS2_LOG_DESC_METADATA, num + 1, num); | |
e8c92ed7 | 445 | ld = page_address(page); |
905d2aef | 446 | gfs2_log_lock(sdp); |
e8c92ed7 | 447 | ptr = (__be64 *)(ld + 1); |
b3b94faa DT |
448 | |
449 | n = 0; | |
c0752aa7 | 450 | list_for_each_entry_continue(bd1, blist, bd_list) { |
b3b94faa | 451 | *ptr++ = cpu_to_be64(bd1->bd_bh->b_blocknr); |
dad30e90 SW |
452 | if (is_databuf) { |
453 | gfs2_check_magic(bd1->bd_bh); | |
454 | *ptr++ = cpu_to_be64(buffer_escaped(bd1->bd_bh) ? 1 : 0); | |
455 | } | |
b3b94faa DT |
456 | if (++n >= num) |
457 | break; | |
458 | } | |
459 | ||
905d2aef | 460 | gfs2_log_unlock(sdp); |
e8c92ed7 | 461 | gfs2_log_write_page(sdp, page); |
905d2aef | 462 | gfs2_log_lock(sdp); |
b3b94faa DT |
463 | |
464 | n = 0; | |
c0752aa7 | 465 | list_for_each_entry_continue(bd2, blist, bd_list) { |
16615be1 | 466 | get_bh(bd2->bd_bh); |
905d2aef | 467 | gfs2_log_unlock(sdp); |
16615be1 | 468 | lock_buffer(bd2->bd_bh); |
dad30e90 SW |
469 | |
470 | if (buffer_escaped(bd2->bd_bh)) { | |
471 | void *kaddr; | |
472 | page = mempool_alloc(gfs2_page_pool, GFP_NOIO); | |
473 | ptr = page_address(page); | |
474 | kaddr = kmap_atomic(bd2->bd_bh->b_page); | |
475 | memcpy(ptr, kaddr + bh_offset(bd2->bd_bh), | |
476 | bd2->bd_bh->b_size); | |
477 | kunmap_atomic(kaddr); | |
478 | *(__be32 *)ptr = 0; | |
479 | clear_buffer_escaped(bd2->bd_bh); | |
480 | unlock_buffer(bd2->bd_bh); | |
481 | brelse(bd2->bd_bh); | |
482 | gfs2_log_write_page(sdp, page); | |
483 | } else { | |
484 | gfs2_log_write_bh(sdp, bd2->bd_bh); | |
485 | } | |
905d2aef | 486 | gfs2_log_lock(sdp); |
b3b94faa DT |
487 | if (++n >= num) |
488 | break; | |
489 | } | |
490 | ||
905d2aef | 491 | BUG_ON(total < num); |
b3b94faa DT |
492 | total -= num; |
493 | } | |
905d2aef | 494 | gfs2_log_unlock(sdp); |
b3b94faa DT |
495 | } |
496 | ||
d69a3c65 | 497 | static void buf_lo_before_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr) |
dad30e90 SW |
498 | { |
499 | unsigned int limit = buf_limit(sdp); /* 503 for 4k blocks */ | |
022ef4fe | 500 | unsigned int nbuf; |
d69a3c65 SW |
501 | if (tr == NULL) |
502 | return; | |
022ef4fe SW |
503 | nbuf = tr->tr_num_buf_new - tr->tr_num_buf_rm; |
504 | gfs2_before_commit(sdp, limit, nbuf, &tr->tr_buf, 0); | |
dad30e90 SW |
505 | } |
506 | ||
16ca9412 | 507 | static void buf_lo_after_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr) |
b3b94faa | 508 | { |
d69a3c65 | 509 | struct list_head *head; |
b3b94faa DT |
510 | struct gfs2_bufdata *bd; |
511 | ||
d69a3c65 | 512 | if (tr == NULL) |
16ca9412 | 513 | return; |
16ca9412 | 514 | |
d69a3c65 | 515 | head = &tr->tr_buf; |
b3b94faa | 516 | while (!list_empty(head)) { |
c0752aa7 BP |
517 | bd = list_entry(head->next, struct gfs2_bufdata, bd_list); |
518 | list_del_init(&bd->bd_list); | |
16ca9412 | 519 | gfs2_unpin(sdp, bd->bd_bh, tr); |
b3b94faa | 520 | } |
b3b94faa DT |
521 | } |
522 | ||
523 | static void buf_lo_before_scan(struct gfs2_jdesc *jd, | |
55167622 | 524 | struct gfs2_log_header_host *head, int pass) |
b3b94faa | 525 | { |
b3b94faa DT |
526 | if (pass != 0) |
527 | return; | |
528 | ||
a17d758b BP |
529 | jd->jd_found_blocks = 0; |
530 | jd->jd_replayed_blocks = 0; | |
b3b94faa DT |
531 | } |
532 | ||
533 | static int buf_lo_scan_elements(struct gfs2_jdesc *jd, unsigned int start, | |
534 | struct gfs2_log_descriptor *ld, __be64 *ptr, | |
535 | int pass) | |
536 | { | |
feaa7bba SW |
537 | struct gfs2_inode *ip = GFS2_I(jd->jd_inode); |
538 | struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode); | |
5c676f6d | 539 | struct gfs2_glock *gl = ip->i_gl; |
b3b94faa DT |
540 | unsigned int blks = be32_to_cpu(ld->ld_data1); |
541 | struct buffer_head *bh_log, *bh_ip; | |
cd915493 | 542 | u64 blkno; |
b3b94faa DT |
543 | int error = 0; |
544 | ||
545 | if (pass != 1 || be32_to_cpu(ld->ld_type) != GFS2_LOG_DESC_METADATA) | |
546 | return 0; | |
547 | ||
e1cb6be9 | 548 | gfs2_replay_incr_blk(jd, &start); |
b3b94faa | 549 | |
e1cb6be9 | 550 | for (; blks; gfs2_replay_incr_blk(jd, &start), blks--) { |
b3b94faa DT |
551 | blkno = be64_to_cpu(*ptr++); |
552 | ||
a17d758b | 553 | jd->jd_found_blocks++; |
b3b94faa | 554 | |
a17d758b | 555 | if (gfs2_revoke_check(jd, blkno, start)) |
b3b94faa DT |
556 | continue; |
557 | ||
558 | error = gfs2_replay_read_block(jd, start, &bh_log); | |
82ffa516 SW |
559 | if (error) |
560 | return error; | |
b3b94faa DT |
561 | |
562 | bh_ip = gfs2_meta_new(gl, blkno); | |
563 | memcpy(bh_ip->b_data, bh_log->b_data, bh_log->b_size); | |
564 | ||
565 | if (gfs2_meta_check(sdp, bh_ip)) | |
566 | error = -EIO; | |
567 | else | |
568 | mark_buffer_dirty(bh_ip); | |
569 | ||
570 | brelse(bh_log); | |
571 | brelse(bh_ip); | |
572 | ||
573 | if (error) | |
574 | break; | |
575 | ||
a17d758b | 576 | jd->jd_replayed_blocks++; |
b3b94faa DT |
577 | } |
578 | ||
579 | return error; | |
580 | } | |
581 | ||
7c0ef28a SW |
582 | /** |
583 | * gfs2_meta_sync - Sync all buffers associated with a glock | |
584 | * @gl: The glock | |
585 | * | |
586 | */ | |
587 | ||
588 | static void gfs2_meta_sync(struct gfs2_glock *gl) | |
589 | { | |
590 | struct address_space *mapping = gfs2_glock2aspace(gl); | |
15562c43 | 591 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
7c0ef28a SW |
592 | int error; |
593 | ||
70d4ee94 SW |
594 | if (mapping == NULL) |
595 | mapping = &sdp->sd_aspace; | |
596 | ||
7c0ef28a SW |
597 | filemap_fdatawrite(mapping); |
598 | error = filemap_fdatawait(mapping); | |
599 | ||
600 | if (error) | |
15562c43 | 601 | gfs2_io_error(gl->gl_name.ln_sbd); |
7c0ef28a SW |
602 | } |
603 | ||
b3b94faa DT |
604 | static void buf_lo_after_scan(struct gfs2_jdesc *jd, int error, int pass) |
605 | { | |
feaa7bba SW |
606 | struct gfs2_inode *ip = GFS2_I(jd->jd_inode); |
607 | struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode); | |
b3b94faa DT |
608 | |
609 | if (error) { | |
7276b3b0 | 610 | gfs2_meta_sync(ip->i_gl); |
b3b94faa DT |
611 | return; |
612 | } | |
613 | if (pass != 1) | |
614 | return; | |
615 | ||
7276b3b0 | 616 | gfs2_meta_sync(ip->i_gl); |
b3b94faa DT |
617 | |
618 | fs_info(sdp, "jid=%u: Replayed %u of %u blocks\n", | |
a17d758b | 619 | jd->jd_jid, jd->jd_replayed_blocks, jd->jd_found_blocks); |
b3b94faa DT |
620 | } |
621 | ||
d69a3c65 | 622 | static void revoke_lo_before_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr) |
b3b94faa | 623 | { |
b3b94faa | 624 | struct gfs2_meta_header *mh; |
b3b94faa DT |
625 | unsigned int offset; |
626 | struct list_head *head = &sdp->sd_log_le_revoke; | |
82e86087 | 627 | struct gfs2_bufdata *bd; |
e8c92ed7 | 628 | struct page *page; |
dad30e90 | 629 | unsigned int length; |
b3b94faa | 630 | |
5d054964 | 631 | gfs2_write_revokes(sdp); |
b3b94faa DT |
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); | |
b3b94faa DT |
637 | offset = sizeof(struct gfs2_log_descriptor); |
638 | ||
c0752aa7 | 639 | list_for_each_entry(bd, head, bd_list) { |
b3b94faa DT |
640 | sdp->sd_log_num_revoke--; |
641 | ||
cd915493 | 642 | if (offset + sizeof(u64) > sdp->sd_sb.sb_bsize) { |
b3b94faa | 643 | |
e8c92ed7 SW |
644 | gfs2_log_write_page(sdp, page); |
645 | page = mempool_alloc(gfs2_page_pool, GFP_NOIO); | |
646 | mh = page_address(page); | |
647 | clear_page(mh); | |
b3b94faa | 648 | mh->mh_magic = cpu_to_be32(GFS2_MAGIC); |
e3167ded SW |
649 | mh->mh_type = cpu_to_be32(GFS2_METATYPE_LB); |
650 | mh->mh_format = cpu_to_be32(GFS2_FORMAT_LB); | |
b3b94faa DT |
651 | offset = sizeof(struct gfs2_meta_header); |
652 | } | |
653 | ||
e8c92ed7 | 654 | *(__be64 *)(page_address(page) + offset) = cpu_to_be64(bd->bd_blkno); |
cd915493 | 655 | offset += sizeof(u64); |
b3b94faa DT |
656 | } |
657 | gfs2_assert_withdraw(sdp, !sdp->sd_log_num_revoke); | |
658 | ||
e8c92ed7 | 659 | gfs2_log_write_page(sdp, page); |
b3b94faa DT |
660 | } |
661 | ||
16ca9412 | 662 | static void revoke_lo_after_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr) |
f42ab085 SW |
663 | { |
664 | struct list_head *head = &sdp->sd_log_le_revoke; | |
665 | struct gfs2_bufdata *bd; | |
666 | struct gfs2_glock *gl; | |
667 | ||
668 | while (!list_empty(head)) { | |
c0752aa7 BP |
669 | bd = list_entry(head->next, struct gfs2_bufdata, bd_list); |
670 | list_del_init(&bd->bd_list); | |
f42ab085 SW |
671 | gl = bd->bd_gl; |
672 | atomic_dec(&gl->gl_revokes); | |
673 | clear_bit(GLF_LFLUSH, &gl->gl_flags); | |
674 | kmem_cache_free(gfs2_bufdata_cachep, bd); | |
675 | } | |
676 | } | |
677 | ||
b3b94faa | 678 | static void revoke_lo_before_scan(struct gfs2_jdesc *jd, |
55167622 | 679 | struct gfs2_log_header_host *head, int pass) |
b3b94faa | 680 | { |
b3b94faa DT |
681 | if (pass != 0) |
682 | return; | |
683 | ||
a17d758b BP |
684 | jd->jd_found_revokes = 0; |
685 | jd->jd_replay_tail = head->lh_tail; | |
b3b94faa DT |
686 | } |
687 | ||
688 | static int revoke_lo_scan_elements(struct gfs2_jdesc *jd, unsigned int start, | |
689 | struct gfs2_log_descriptor *ld, __be64 *ptr, | |
690 | int pass) | |
691 | { | |
feaa7bba | 692 | struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode); |
b3b94faa DT |
693 | unsigned int blks = be32_to_cpu(ld->ld_length); |
694 | unsigned int revokes = be32_to_cpu(ld->ld_data1); | |
695 | struct buffer_head *bh; | |
696 | unsigned int offset; | |
cd915493 | 697 | u64 blkno; |
b3b94faa DT |
698 | int first = 1; |
699 | int error; | |
700 | ||
701 | if (pass != 0 || be32_to_cpu(ld->ld_type) != GFS2_LOG_DESC_REVOKE) | |
702 | return 0; | |
703 | ||
704 | offset = sizeof(struct gfs2_log_descriptor); | |
705 | ||
e1cb6be9 | 706 | for (; blks; gfs2_replay_incr_blk(jd, &start), blks--) { |
b3b94faa DT |
707 | error = gfs2_replay_read_block(jd, start, &bh); |
708 | if (error) | |
709 | return error; | |
710 | ||
711 | if (!first) | |
712 | gfs2_metatype_check(sdp, bh, GFS2_METATYPE_LB); | |
713 | ||
cd915493 | 714 | while (offset + sizeof(u64) <= sdp->sd_sb.sb_bsize) { |
b3b94faa DT |
715 | blkno = be64_to_cpu(*(__be64 *)(bh->b_data + offset)); |
716 | ||
a17d758b | 717 | error = gfs2_revoke_add(jd, blkno, start); |
3ad62e87 BP |
718 | if (error < 0) { |
719 | brelse(bh); | |
b3b94faa | 720 | return error; |
3ad62e87 | 721 | } |
b3b94faa | 722 | else if (error) |
a17d758b | 723 | jd->jd_found_revokes++; |
b3b94faa DT |
724 | |
725 | if (!--revokes) | |
726 | break; | |
cd915493 | 727 | offset += sizeof(u64); |
b3b94faa DT |
728 | } |
729 | ||
730 | brelse(bh); | |
731 | offset = sizeof(struct gfs2_meta_header); | |
732 | first = 0; | |
733 | } | |
734 | ||
735 | return 0; | |
736 | } | |
737 | ||
738 | static void revoke_lo_after_scan(struct gfs2_jdesc *jd, int error, int pass) | |
739 | { | |
feaa7bba | 740 | struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode); |
b3b94faa DT |
741 | |
742 | if (error) { | |
a17d758b | 743 | gfs2_revoke_clean(jd); |
b3b94faa DT |
744 | return; |
745 | } | |
746 | if (pass != 1) | |
747 | return; | |
748 | ||
749 | fs_info(sdp, "jid=%u: Found %u revoke tags\n", | |
a17d758b | 750 | jd->jd_jid, jd->jd_found_revokes); |
b3b94faa | 751 | |
a17d758b | 752 | gfs2_revoke_clean(jd); |
b3b94faa DT |
753 | } |
754 | ||
16615be1 SW |
755 | /** |
756 | * databuf_lo_before_commit - Scan the data buffers, writing as we go | |
757 | * | |
758 | */ | |
759 | ||
d69a3c65 | 760 | static void databuf_lo_before_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr) |
16615be1 | 761 | { |
022ef4fe SW |
762 | unsigned int limit = databuf_limit(sdp); |
763 | unsigned int nbuf; | |
d69a3c65 SW |
764 | if (tr == NULL) |
765 | return; | |
022ef4fe SW |
766 | nbuf = tr->tr_num_databuf_new - tr->tr_num_databuf_rm; |
767 | gfs2_before_commit(sdp, limit, nbuf, &tr->tr_databuf, 1); | |
18ec7d5c SW |
768 | } |
769 | ||
770 | static int databuf_lo_scan_elements(struct gfs2_jdesc *jd, unsigned int start, | |
771 | struct gfs2_log_descriptor *ld, | |
772 | __be64 *ptr, int pass) | |
773 | { | |
feaa7bba | 774 | struct gfs2_inode *ip = GFS2_I(jd->jd_inode); |
5c676f6d | 775 | struct gfs2_glock *gl = ip->i_gl; |
18ec7d5c SW |
776 | unsigned int blks = be32_to_cpu(ld->ld_data1); |
777 | struct buffer_head *bh_log, *bh_ip; | |
cd915493 SW |
778 | u64 blkno; |
779 | u64 esc; | |
18ec7d5c SW |
780 | int error = 0; |
781 | ||
782 | if (pass != 1 || be32_to_cpu(ld->ld_type) != GFS2_LOG_DESC_JDATA) | |
783 | return 0; | |
784 | ||
e1cb6be9 BP |
785 | gfs2_replay_incr_blk(jd, &start); |
786 | for (; blks; gfs2_replay_incr_blk(jd, &start), blks--) { | |
18ec7d5c SW |
787 | blkno = be64_to_cpu(*ptr++); |
788 | esc = be64_to_cpu(*ptr++); | |
789 | ||
a17d758b | 790 | jd->jd_found_blocks++; |
18ec7d5c | 791 | |
a17d758b | 792 | if (gfs2_revoke_check(jd, blkno, start)) |
18ec7d5c SW |
793 | continue; |
794 | ||
795 | error = gfs2_replay_read_block(jd, start, &bh_log); | |
796 | if (error) | |
797 | return error; | |
798 | ||
799 | bh_ip = gfs2_meta_new(gl, blkno); | |
800 | memcpy(bh_ip->b_data, bh_log->b_data, bh_log->b_size); | |
801 | ||
802 | /* Unescape */ | |
803 | if (esc) { | |
804 | __be32 *eptr = (__be32 *)bh_ip->b_data; | |
805 | *eptr = cpu_to_be32(GFS2_MAGIC); | |
806 | } | |
807 | mark_buffer_dirty(bh_ip); | |
808 | ||
809 | brelse(bh_log); | |
810 | brelse(bh_ip); | |
18ec7d5c | 811 | |
a17d758b | 812 | jd->jd_replayed_blocks++; |
18ec7d5c SW |
813 | } |
814 | ||
815 | return error; | |
816 | } | |
817 | ||
818 | /* FIXME: sort out accounting for log blocks etc. */ | |
819 | ||
820 | static void databuf_lo_after_scan(struct gfs2_jdesc *jd, int error, int pass) | |
821 | { | |
feaa7bba SW |
822 | struct gfs2_inode *ip = GFS2_I(jd->jd_inode); |
823 | struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode); | |
18ec7d5c SW |
824 | |
825 | if (error) { | |
7276b3b0 | 826 | gfs2_meta_sync(ip->i_gl); |
18ec7d5c SW |
827 | return; |
828 | } | |
829 | if (pass != 1) | |
830 | return; | |
831 | ||
832 | /* data sync? */ | |
7276b3b0 | 833 | gfs2_meta_sync(ip->i_gl); |
18ec7d5c SW |
834 | |
835 | fs_info(sdp, "jid=%u: Replayed %u of %u data blocks\n", | |
a17d758b | 836 | jd->jd_jid, jd->jd_replayed_blocks, jd->jd_found_blocks); |
18ec7d5c SW |
837 | } |
838 | ||
16ca9412 | 839 | static void databuf_lo_after_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr) |
18ec7d5c | 840 | { |
d69a3c65 | 841 | struct list_head *head; |
18ec7d5c SW |
842 | struct gfs2_bufdata *bd; |
843 | ||
d69a3c65 | 844 | if (tr == NULL) |
16ca9412 | 845 | return; |
16ca9412 | 846 | |
d69a3c65 | 847 | head = &tr->tr_databuf; |
18ec7d5c | 848 | while (!list_empty(head)) { |
c0752aa7 BP |
849 | bd = list_entry(head->next, struct gfs2_bufdata, bd_list); |
850 | list_del_init(&bd->bd_list); | |
16ca9412 | 851 | gfs2_unpin(sdp, bd->bd_bh, tr); |
18ec7d5c | 852 | } |
b3b94faa DT |
853 | } |
854 | ||
18ec7d5c | 855 | |
b09e593d | 856 | const struct gfs2_log_operations gfs2_buf_lops = { |
b3b94faa DT |
857 | .lo_before_commit = buf_lo_before_commit, |
858 | .lo_after_commit = buf_lo_after_commit, | |
859 | .lo_before_scan = buf_lo_before_scan, | |
860 | .lo_scan_elements = buf_lo_scan_elements, | |
861 | .lo_after_scan = buf_lo_after_scan, | |
ea67eedb | 862 | .lo_name = "buf", |
b3b94faa DT |
863 | }; |
864 | ||
b09e593d | 865 | const struct gfs2_log_operations gfs2_revoke_lops = { |
b3b94faa | 866 | .lo_before_commit = revoke_lo_before_commit, |
f42ab085 | 867 | .lo_after_commit = revoke_lo_after_commit, |
b3b94faa DT |
868 | .lo_before_scan = revoke_lo_before_scan, |
869 | .lo_scan_elements = revoke_lo_scan_elements, | |
870 | .lo_after_scan = revoke_lo_after_scan, | |
ea67eedb | 871 | .lo_name = "revoke", |
b3b94faa DT |
872 | }; |
873 | ||
b09e593d | 874 | const struct gfs2_log_operations gfs2_databuf_lops = { |
b3b94faa | 875 | .lo_before_commit = databuf_lo_before_commit, |
18ec7d5c SW |
876 | .lo_after_commit = databuf_lo_after_commit, |
877 | .lo_scan_elements = databuf_lo_scan_elements, | |
878 | .lo_after_scan = databuf_lo_after_scan, | |
ea67eedb | 879 | .lo_name = "databuf", |
b3b94faa DT |
880 | }; |
881 | ||
b09e593d | 882 | const struct gfs2_log_operations *gfs2_log_ops[] = { |
16615be1 | 883 | &gfs2_databuf_lops, |
b3b94faa | 884 | &gfs2_buf_lops, |
16615be1 | 885 | &gfs2_revoke_lops, |
ea67eedb | 886 | NULL, |
b3b94faa DT |
887 | }; |
888 |