2 * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
3 * Copyright (C) 2004-2006 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 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
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>
17 #include <linux/kallsyms.h>
18 #include <linux/gfs2_ondisk.h>
29 #include "trace_gfs2.h"
31 int gfs2_trans_begin(struct gfs2_sbd *sdp, unsigned int blocks,
34 struct gfs2_trans *tr;
37 BUG_ON(current->journal_info);
38 BUG_ON(blocks == 0 && revokes == 0);
40 if (!test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags))
43 tr = kzalloc(sizeof(struct gfs2_trans), GFP_NOFS);
48 tr->tr_blocks = blocks;
49 tr->tr_revokes = revokes;
51 set_bit(TR_ALLOCED, &tr->tr_flags);
53 tr->tr_reserved += 6 + blocks;
55 tr->tr_reserved += gfs2_struct2blk(sdp, revokes,
57 INIT_LIST_HEAD(&tr->tr_databuf);
58 INIT_LIST_HEAD(&tr->tr_buf);
59 INIT_LIST_HEAD(&tr->tr_ail1_list);
60 INIT_LIST_HEAD(&tr->tr_ail2_list);
62 sb_start_intwrite(sdp->sd_vfs);
64 error = gfs2_log_reserve(sdp, tr->tr_reserved);
68 current->journal_info = tr;
73 sb_end_intwrite(sdp->sd_vfs);
79 static void gfs2_print_trans(const struct gfs2_trans *tr)
81 pr_warn("Transaction created at: %pSR\n", (void *)tr->tr_ip);
82 pr_warn("blocks=%u revokes=%u reserved=%u touched=%u\n",
83 tr->tr_blocks, tr->tr_revokes, tr->tr_reserved,
84 test_bit(TR_TOUCHED, &tr->tr_flags));
85 pr_warn("Buf %u/%u Databuf %u/%u Revoke %u/%u\n",
86 tr->tr_num_buf_new, tr->tr_num_buf_rm,
87 tr->tr_num_databuf_new, tr->tr_num_databuf_rm,
88 tr->tr_num_revoke, tr->tr_num_revoke_rm);
91 void gfs2_trans_end(struct gfs2_sbd *sdp)
93 struct gfs2_trans *tr = current->journal_info;
95 int alloced = test_bit(TR_ALLOCED, &tr->tr_flags);
98 current->journal_info = NULL;
100 if (!test_bit(TR_TOUCHED, &tr->tr_flags)) {
101 gfs2_log_release(sdp, tr->tr_reserved);
104 sb_end_intwrite(sdp->sd_vfs);
109 nbuf = tr->tr_num_buf_new + tr->tr_num_databuf_new;
110 nbuf -= tr->tr_num_buf_rm;
111 nbuf -= tr->tr_num_databuf_rm;
113 if (gfs2_assert_withdraw(sdp, (nbuf <= tr->tr_blocks) &&
114 (tr->tr_num_revoke <= tr->tr_revokes)))
115 gfs2_print_trans(tr);
117 gfs2_log_commit(sdp, tr);
118 if (alloced && !test_bit(TR_ATTACHED, &tr->tr_flags))
120 up_read(&sdp->sd_log_flush_lock);
122 if (sdp->sd_vfs->s_flags & MS_SYNCHRONOUS)
123 gfs2_log_flush(sdp, NULL, NORMAL_FLUSH);
125 sb_end_intwrite(sdp->sd_vfs);
128 static struct gfs2_bufdata *gfs2_alloc_bufdata(struct gfs2_glock *gl,
129 struct buffer_head *bh,
130 const struct gfs2_log_operations *lops)
132 struct gfs2_bufdata *bd;
134 bd = kmem_cache_zalloc(gfs2_bufdata_cachep, GFP_NOFS | __GFP_NOFAIL);
138 INIT_LIST_HEAD(&bd->bd_list);
144 * gfs2_trans_add_data - Add a databuf to the transaction.
145 * @gl: The inode glock associated with the buffer
146 * @bh: The buffer to add
148 * This is used in two distinct cases:
149 * i) In ordered write mode
150 * We put the data buffer on a list so that we can ensure that its
151 * synced to disk at the right time
152 * ii) In journaled data mode
153 * We need to journal the data block in the same way as metadata in
154 * the functions above. The difference is that here we have a tag
155 * which is two __be64's being the block number (as per meta data)
156 * and a flag which says whether the data block needs escaping or
157 * not. This means we need a new log entry for each 251 or so data
158 * blocks, which isn't an enormous overhead but twice as much as
159 * for normal metadata blocks.
161 void gfs2_trans_add_data(struct gfs2_glock *gl, struct buffer_head *bh)
163 struct gfs2_trans *tr = current->journal_info;
164 struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
165 struct address_space *mapping = bh->b_page->mapping;
166 struct gfs2_inode *ip = GFS2_I(mapping->host);
167 struct gfs2_bufdata *bd;
169 if (!gfs2_is_jdata(ip)) {
170 gfs2_ordered_add_inode(ip);
175 if (buffer_pinned(bh)) {
176 set_bit(TR_TOUCHED, &tr->tr_flags);
182 gfs2_log_unlock(sdp);
184 if (bh->b_private == NULL)
185 bd = gfs2_alloc_bufdata(gl, bh, &gfs2_databuf_lops);
191 gfs2_assert(sdp, bd->bd_gl == gl);
192 set_bit(TR_TOUCHED, &tr->tr_flags);
193 if (list_empty(&bd->bd_list)) {
194 set_bit(GLF_LFLUSH, &bd->bd_gl->gl_flags);
195 set_bit(GLF_DIRTY, &bd->bd_gl->gl_flags);
196 gfs2_pin(sdp, bd->bd_bh);
197 tr->tr_num_databuf_new++;
198 list_add_tail(&bd->bd_list, &tr->tr_databuf);
200 gfs2_log_unlock(sdp);
205 void gfs2_trans_add_meta(struct gfs2_glock *gl, struct buffer_head *bh)
208 struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
209 struct gfs2_bufdata *bd;
210 struct gfs2_meta_header *mh;
211 struct gfs2_trans *tr = current->journal_info;
212 enum gfs2_freeze_state state = atomic_read(&sdp->sd_freeze_state);
215 if (buffer_pinned(bh)) {
216 set_bit(TR_TOUCHED, &tr->tr_flags);
222 gfs2_log_unlock(sdp);
224 lock_page(bh->b_page);
225 if (bh->b_private == NULL)
226 bd = gfs2_alloc_bufdata(gl, bh, &gfs2_buf_lops);
229 unlock_page(bh->b_page);
233 gfs2_assert(sdp, bd->bd_gl == gl);
234 set_bit(TR_TOUCHED, &tr->tr_flags);
235 if (!list_empty(&bd->bd_list))
237 set_bit(GLF_LFLUSH, &bd->bd_gl->gl_flags);
238 set_bit(GLF_DIRTY, &bd->bd_gl->gl_flags);
239 mh = (struct gfs2_meta_header *)bd->bd_bh->b_data;
240 if (unlikely(mh->mh_magic != cpu_to_be32(GFS2_MAGIC))) {
241 pr_err("Attempting to add uninitialised block to journal (inplace block=%lld)\n",
242 (unsigned long long)bd->bd_bh->b_blocknr);
245 if (unlikely(state == SFS_FROZEN)) {
246 printk(KERN_INFO "GFS2:adding buf while frozen\n");
247 gfs2_assert_withdraw(sdp, 0);
249 gfs2_pin(sdp, bd->bd_bh);
250 mh->__pad0 = cpu_to_be64(0);
251 mh->mh_jid = cpu_to_be32(sdp->sd_jdesc->jd_jid);
252 list_add(&bd->bd_list, &tr->tr_buf);
253 tr->tr_num_buf_new++;
255 gfs2_log_unlock(sdp);
260 void gfs2_trans_add_revoke(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd)
262 struct gfs2_trans *tr = current->journal_info;
264 BUG_ON(!list_empty(&bd->bd_list));
265 gfs2_add_revoke(sdp, bd);
266 set_bit(TR_TOUCHED, &tr->tr_flags);
270 void gfs2_trans_add_unrevoke(struct gfs2_sbd *sdp, u64 blkno, unsigned int len)
272 struct gfs2_bufdata *bd, *tmp;
273 struct gfs2_trans *tr = current->journal_info;
274 unsigned int n = len;
277 list_for_each_entry_safe(bd, tmp, &sdp->sd_log_le_revoke, bd_list) {
278 if ((bd->bd_blkno >= blkno) && (bd->bd_blkno < (blkno + len))) {
279 list_del_init(&bd->bd_list);
280 gfs2_assert_withdraw(sdp, sdp->sd_log_num_revoke);
281 sdp->sd_log_num_revoke--;
283 gfs2_glock_remove_revoke(bd->bd_gl);
284 kmem_cache_free(gfs2_bufdata_cachep, bd);
285 tr->tr_num_revoke_rm++;
290 gfs2_log_unlock(sdp);