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 | ||
d77d1b58 JP |
10 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
11 | ||
b3b94faa DT |
12 | #include <linux/sched.h> |
13 | #include <linux/slab.h> | |
14 | #include <linux/spinlock.h> | |
15 | #include <linux/completion.h> | |
16 | #include <linux/buffer_head.h> | |
d0dc80db | 17 | #include <linux/kallsyms.h> |
f057f6cd | 18 | #include <linux/gfs2_ondisk.h> |
b3b94faa DT |
19 | |
20 | #include "gfs2.h" | |
5c676f6d | 21 | #include "incore.h" |
b3b94faa | 22 | #include "glock.h" |
767f433f | 23 | #include "inode.h" |
b3b94faa DT |
24 | #include "log.h" |
25 | #include "lops.h" | |
26 | #include "meta_io.h" | |
27 | #include "trans.h" | |
5c676f6d | 28 | #include "util.h" |
5e687eac | 29 | #include "trace_gfs2.h" |
b3b94faa | 30 | |
d0dc80db SW |
31 | int gfs2_trans_begin(struct gfs2_sbd *sdp, unsigned int blocks, |
32 | unsigned int revokes) | |
b3b94faa DT |
33 | { |
34 | struct gfs2_trans *tr; | |
35 | int error; | |
36 | ||
d0dc80db SW |
37 | BUG_ON(current->journal_info); |
38 | BUG_ON(blocks == 0 && revokes == 0); | |
b3b94faa | 39 | |
a1c0643f SW |
40 | if (!test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags)) |
41 | return -EROFS; | |
42 | ||
f55ab26a | 43 | tr = kzalloc(sizeof(struct gfs2_trans), GFP_NOFS); |
b3b94faa DT |
44 | if (!tr) |
45 | return -ENOMEM; | |
46 | ||
d0dc80db | 47 | tr->tr_ip = (unsigned long)__builtin_return_address(0); |
b3b94faa DT |
48 | tr->tr_blocks = blocks; |
49 | tr->tr_revokes = revokes; | |
50 | tr->tr_reserved = 1; | |
51 | if (blocks) | |
f4154ea0 | 52 | tr->tr_reserved += 6 + blocks; |
b3b94faa DT |
53 | if (revokes) |
54 | tr->tr_reserved += gfs2_struct2blk(sdp, revokes, | |
cd915493 | 55 | sizeof(u64)); |
d69a3c65 SW |
56 | INIT_LIST_HEAD(&tr->tr_databuf); |
57 | INIT_LIST_HEAD(&tr->tr_buf); | |
58 | ||
39263d5e | 59 | sb_start_intwrite(sdp->sd_vfs); |
579b78a4 | 60 | gfs2_holder_init(sdp->sd_trans_gl, LM_ST_SHARED, 0, &tr->tr_t_gh); |
b3b94faa | 61 | |
e317ffcb | 62 | error = gfs2_glock_nq(&tr->tr_t_gh); |
b3b94faa | 63 | if (error) |
e317ffcb | 64 | goto fail_holder_uninit; |
b3b94faa | 65 | |
b3b94faa DT |
66 | error = gfs2_log_reserve(sdp, tr->tr_reserved); |
67 | if (error) | |
68 | goto fail_gunlock; | |
69 | ||
5c676f6d | 70 | current->journal_info = tr; |
b3b94faa DT |
71 | |
72 | return 0; | |
73 | ||
484adff8 | 74 | fail_gunlock: |
e317ffcb | 75 | gfs2_glock_dq(&tr->tr_t_gh); |
b3b94faa | 76 | |
484adff8 | 77 | fail_holder_uninit: |
39263d5e | 78 | sb_end_intwrite(sdp->sd_vfs); |
e317ffcb | 79 | gfs2_holder_uninit(&tr->tr_t_gh); |
b3b94faa DT |
80 | kfree(tr); |
81 | ||
82 | return error; | |
83 | } | |
84 | ||
5e687eac BM |
85 | /** |
86 | * gfs2_log_release - Release a given number of log blocks | |
87 | * @sdp: The GFS2 superblock | |
88 | * @blks: The number of blocks | |
89 | * | |
90 | */ | |
91 | ||
92 | static void gfs2_log_release(struct gfs2_sbd *sdp, unsigned int blks) | |
93 | { | |
94 | ||
95 | atomic_add(blks, &sdp->sd_log_blks_free); | |
96 | trace_gfs2_log_blocks(sdp, blks); | |
97 | gfs2_assert_withdraw(sdp, atomic_read(&sdp->sd_log_blks_free) <= | |
98 | sdp->sd_jdesc->jd_blocks); | |
99 | up_read(&sdp->sd_log_flush_lock); | |
100 | } | |
101 | ||
c50b91c4 SW |
102 | static void gfs2_print_trans(const struct gfs2_trans *tr) |
103 | { | |
d77d1b58 JP |
104 | pr_warn("Transaction created at: %pSR\n", (void *)tr->tr_ip); |
105 | pr_warn("blocks=%u revokes=%u reserved=%u touched=%u\n", | |
106 | tr->tr_blocks, tr->tr_revokes, tr->tr_reserved, tr->tr_touched); | |
107 | pr_warn("Buf %u/%u Databuf %u/%u Revoke %u/%u\n", | |
108 | tr->tr_num_buf_new, tr->tr_num_buf_rm, | |
109 | tr->tr_num_databuf_new, tr->tr_num_databuf_rm, | |
110 | tr->tr_num_revoke, tr->tr_num_revoke_rm); | |
c50b91c4 SW |
111 | } |
112 | ||
b3b94faa DT |
113 | void gfs2_trans_end(struct gfs2_sbd *sdp) |
114 | { | |
f4154ea0 | 115 | struct gfs2_trans *tr = current->journal_info; |
c50b91c4 | 116 | s64 nbuf; |
f4154ea0 | 117 | BUG_ON(!tr); |
5c676f6d | 118 | current->journal_info = NULL; |
b3b94faa | 119 | |
b3b94faa DT |
120 | if (!tr->tr_touched) { |
121 | gfs2_log_release(sdp, tr->tr_reserved); | |
d8348de0 SW |
122 | if (tr->tr_t_gh.gh_gl) { |
123 | gfs2_glock_dq(&tr->tr_t_gh); | |
124 | gfs2_holder_uninit(&tr->tr_t_gh); | |
125 | kfree(tr); | |
126 | } | |
39263d5e | 127 | sb_end_intwrite(sdp->sd_vfs); |
b3b94faa DT |
128 | return; |
129 | } | |
130 | ||
c50b91c4 SW |
131 | nbuf = tr->tr_num_buf_new + tr->tr_num_databuf_new; |
132 | nbuf -= tr->tr_num_buf_rm; | |
133 | nbuf -= tr->tr_num_databuf_rm; | |
134 | ||
135 | if (gfs2_assert_withdraw(sdp, (nbuf <= tr->tr_blocks) && | |
136 | (tr->tr_num_revoke <= tr->tr_revokes))) | |
137 | gfs2_print_trans(tr); | |
b3b94faa DT |
138 | |
139 | gfs2_log_commit(sdp, tr); | |
d8348de0 SW |
140 | if (tr->tr_t_gh.gh_gl) { |
141 | gfs2_glock_dq(&tr->tr_t_gh); | |
142 | gfs2_holder_uninit(&tr->tr_t_gh); | |
16ca9412 BM |
143 | if (!tr->tr_attached) |
144 | kfree(tr); | |
d8348de0 | 145 | } |
16ca9412 | 146 | up_read(&sdp->sd_log_flush_lock); |
b3b94faa DT |
147 | |
148 | if (sdp->sd_vfs->s_flags & MS_SYNCHRONOUS) | |
b09e593d | 149 | gfs2_log_flush(sdp, NULL); |
39263d5e | 150 | sb_end_intwrite(sdp->sd_vfs); |
b3b94faa DT |
151 | } |
152 | ||
c76c4d96 SW |
153 | static struct gfs2_bufdata *gfs2_alloc_bufdata(struct gfs2_glock *gl, |
154 | struct buffer_head *bh, | |
155 | const struct gfs2_log_operations *lops) | |
156 | { | |
157 | struct gfs2_bufdata *bd; | |
158 | ||
159 | bd = kmem_cache_zalloc(gfs2_bufdata_cachep, GFP_NOFS | __GFP_NOFAIL); | |
160 | bd->bd_bh = bh; | |
161 | bd->bd_gl = gl; | |
162 | bd->bd_ops = lops; | |
163 | INIT_LIST_HEAD(&bd->bd_list); | |
164 | bh->b_private = bd; | |
165 | return bd; | |
166 | } | |
167 | ||
b3b94faa | 168 | /** |
45138990 SW |
169 | * gfs2_trans_add_data - Add a databuf to the transaction. |
170 | * @gl: The inode glock associated with the buffer | |
171 | * @bh: The buffer to add | |
b3b94faa | 172 | * |
767f433f SW |
173 | * This is used in two distinct cases: |
174 | * i) In ordered write mode | |
175 | * We put the data buffer on a list so that we can ensure that its | |
176 | * synced to disk at the right time | |
177 | * ii) In journaled data mode | |
178 | * We need to journal the data block in the same way as metadata in | |
179 | * the functions above. The difference is that here we have a tag | |
180 | * which is two __be64's being the block number (as per meta data) | |
181 | * and a flag which says whether the data block needs escaping or | |
182 | * not. This means we need a new log entry for each 251 or so data | |
183 | * blocks, which isn't an enormous overhead but twice as much as | |
184 | * for normal metadata blocks. | |
b3b94faa | 185 | */ |
45138990 | 186 | void gfs2_trans_add_data(struct gfs2_glock *gl, struct buffer_head *bh) |
767f433f SW |
187 | { |
188 | struct gfs2_trans *tr = current->journal_info; | |
45138990 SW |
189 | struct gfs2_sbd *sdp = gl->gl_sbd; |
190 | struct address_space *mapping = bh->b_page->mapping; | |
767f433f | 191 | struct gfs2_inode *ip = GFS2_I(mapping->host); |
45138990 | 192 | struct gfs2_bufdata *bd; |
767f433f | 193 | |
45138990 SW |
194 | if (!gfs2_is_jdata(ip)) { |
195 | gfs2_ordered_add_inode(ip); | |
767f433f | 196 | return; |
767f433f | 197 | } |
b3b94faa | 198 | |
96e5d1d3 BM |
199 | lock_buffer(bh); |
200 | gfs2_log_lock(sdp); | |
5c676f6d | 201 | bd = bh->b_private; |
c76c4d96 | 202 | if (bd == NULL) { |
96e5d1d3 BM |
203 | gfs2_log_unlock(sdp); |
204 | unlock_buffer(bh); | |
c76c4d96 SW |
205 | if (bh->b_private == NULL) |
206 | bd = gfs2_alloc_bufdata(gl, bh, &gfs2_databuf_lops); | |
96e5d1d3 BM |
207 | lock_buffer(bh); |
208 | gfs2_log_lock(sdp); | |
b3b94faa | 209 | } |
c76c4d96 | 210 | gfs2_assert(sdp, bd->bd_gl == gl); |
45138990 SW |
211 | tr->tr_touched = 1; |
212 | if (list_empty(&bd->bd_list)) { | |
213 | set_bit(GLF_LFLUSH, &bd->bd_gl->gl_flags); | |
214 | set_bit(GLF_DIRTY, &bd->bd_gl->gl_flags); | |
215 | gfs2_pin(sdp, bd->bd_bh); | |
216 | tr->tr_num_databuf_new++; | |
d69a3c65 | 217 | list_add_tail(&bd->bd_list, &tr->tr_databuf); |
45138990 | 218 | } |
96e5d1d3 BM |
219 | gfs2_log_unlock(sdp); |
220 | unlock_buffer(bh); | |
b3b94faa DT |
221 | } |
222 | ||
767f433f | 223 | static void meta_lo_add(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd) |
350a9b0a | 224 | { |
767f433f SW |
225 | struct gfs2_meta_header *mh; |
226 | struct gfs2_trans *tr; | |
227 | ||
228 | tr = current->journal_info; | |
229 | tr->tr_touched = 1; | |
230 | if (!list_empty(&bd->bd_list)) | |
231 | return; | |
232 | set_bit(GLF_LFLUSH, &bd->bd_gl->gl_flags); | |
233 | set_bit(GLF_DIRTY, &bd->bd_gl->gl_flags); | |
234 | mh = (struct gfs2_meta_header *)bd->bd_bh->b_data; | |
235 | if (unlikely(mh->mh_magic != cpu_to_be32(GFS2_MAGIC))) { | |
d77d1b58 | 236 | pr_err("Attempting to add uninitialised block to journal (inplace block=%lld)\n", |
767f433f SW |
237 | (unsigned long long)bd->bd_bh->b_blocknr); |
238 | BUG(); | |
239 | } | |
240 | gfs2_pin(sdp, bd->bd_bh); | |
241 | mh->__pad0 = cpu_to_be64(0); | |
242 | mh->mh_jid = cpu_to_be32(sdp->sd_jdesc->jd_jid); | |
d69a3c65 | 243 | list_add(&bd->bd_list, &tr->tr_buf); |
767f433f | 244 | tr->tr_num_buf_new++; |
350a9b0a SW |
245 | } |
246 | ||
247 | void gfs2_trans_add_meta(struct gfs2_glock *gl, struct buffer_head *bh) | |
248 | { | |
767f433f SW |
249 | |
250 | struct gfs2_sbd *sdp = gl->gl_sbd; | |
251 | struct gfs2_bufdata *bd; | |
252 | ||
253 | lock_buffer(bh); | |
254 | gfs2_log_lock(sdp); | |
255 | bd = bh->b_private; | |
c76c4d96 | 256 | if (bd == NULL) { |
767f433f SW |
257 | gfs2_log_unlock(sdp); |
258 | unlock_buffer(bh); | |
c76c4d96 SW |
259 | lock_page(bh->b_page); |
260 | if (bh->b_private == NULL) | |
261 | bd = gfs2_alloc_bufdata(gl, bh, &gfs2_buf_lops); | |
262 | unlock_page(bh->b_page); | |
767f433f SW |
263 | lock_buffer(bh); |
264 | gfs2_log_lock(sdp); | |
265 | } | |
c76c4d96 | 266 | gfs2_assert(sdp, bd->bd_gl == gl); |
767f433f SW |
267 | meta_lo_add(sdp, bd); |
268 | gfs2_log_unlock(sdp); | |
269 | unlock_buffer(bh); | |
350a9b0a SW |
270 | } |
271 | ||
1ad38c43 | 272 | void gfs2_trans_add_revoke(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd) |
b3b94faa | 273 | { |
75f2b879 SW |
274 | struct gfs2_trans *tr = current->journal_info; |
275 | ||
c0752aa7 | 276 | BUG_ON(!list_empty(&bd->bd_list)); |
5d054964 | 277 | gfs2_add_revoke(sdp, bd); |
75f2b879 SW |
278 | tr->tr_touched = 1; |
279 | tr->tr_num_revoke++; | |
b3b94faa DT |
280 | } |
281 | ||
5731be53 | 282 | void gfs2_trans_add_unrevoke(struct gfs2_sbd *sdp, u64 blkno, unsigned int len) |
b3b94faa | 283 | { |
5731be53 SW |
284 | struct gfs2_bufdata *bd, *tmp; |
285 | struct gfs2_trans *tr = current->journal_info; | |
286 | unsigned int n = len; | |
b3b94faa DT |
287 | |
288 | gfs2_log_lock(sdp); | |
c0752aa7 | 289 | list_for_each_entry_safe(bd, tmp, &sdp->sd_log_le_revoke, bd_list) { |
5731be53 | 290 | if ((bd->bd_blkno >= blkno) && (bd->bd_blkno < (blkno + len))) { |
c0752aa7 | 291 | list_del_init(&bd->bd_list); |
b3b94faa DT |
292 | gfs2_assert_withdraw(sdp, sdp->sd_log_num_revoke); |
293 | sdp->sd_log_num_revoke--; | |
5731be53 SW |
294 | kmem_cache_free(gfs2_bufdata_cachep, bd); |
295 | tr->tr_num_revoke_rm++; | |
296 | if (--n == 0) | |
297 | break; | |
b3b94faa DT |
298 | } |
299 | } | |
b3b94faa | 300 | gfs2_log_unlock(sdp); |
b3b94faa DT |
301 | } |
302 |