Btrfs: deal with DIO bios that span more than one ordered extent
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / xfs / xfs_utils.c
blobb7d5769d2df04a6d5599fec813ce59693275c701
1 /*
2 * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
18 #include "xfs.h"
19 #include "xfs_fs.h"
20 #include "xfs_types.h"
21 #include "xfs_bit.h"
22 #include "xfs_log.h"
23 #include "xfs_inum.h"
24 #include "xfs_trans.h"
25 #include "xfs_sb.h"
26 #include "xfs_ag.h"
27 #include "xfs_dir2.h"
28 #include "xfs_mount.h"
29 #include "xfs_bmap_btree.h"
30 #include "xfs_dinode.h"
31 #include "xfs_inode.h"
32 #include "xfs_inode_item.h"
33 #include "xfs_bmap.h"
34 #include "xfs_error.h"
35 #include "xfs_quota.h"
36 #include "xfs_itable.h"
37 #include "xfs_utils.h"
41 * Allocates a new inode from disk and return a pointer to the
42 * incore copy. This routine will internally commit the current
43 * transaction and allocate a new one if the Space Manager needed
44 * to do an allocation to replenish the inode free-list.
46 * This routine is designed to be called from xfs_create and
47 * xfs_create_dir.
50 int
51 xfs_dir_ialloc(
52 xfs_trans_t **tpp, /* input: current transaction;
53 output: may be a new transaction. */
54 xfs_inode_t *dp, /* directory within whose allocate
55 the inode. */
56 mode_t mode,
57 xfs_nlink_t nlink,
58 xfs_dev_t rdev,
59 cred_t *credp,
60 prid_t prid, /* project id */
61 int okalloc, /* ok to allocate new space */
62 xfs_inode_t **ipp, /* pointer to inode; it will be
63 locked. */
64 int *committed)
67 xfs_trans_t *tp;
68 xfs_trans_t *ntp;
69 xfs_inode_t *ip;
70 xfs_buf_t *ialloc_context = NULL;
71 boolean_t call_again = B_FALSE;
72 int code;
73 uint log_res;
74 uint log_count;
75 void *dqinfo;
76 uint tflags;
78 tp = *tpp;
79 ASSERT(tp->t_flags & XFS_TRANS_PERM_LOG_RES);
82 * xfs_ialloc will return a pointer to an incore inode if
83 * the Space Manager has an available inode on the free
84 * list. Otherwise, it will do an allocation and replenish
85 * the freelist. Since we can only do one allocation per
86 * transaction without deadlocks, we will need to commit the
87 * current transaction and start a new one. We will then
88 * need to call xfs_ialloc again to get the inode.
90 * If xfs_ialloc did an allocation to replenish the freelist,
91 * it returns the bp containing the head of the freelist as
92 * ialloc_context. We will hold a lock on it across the
93 * transaction commit so that no other process can steal
94 * the inode(s) that we've just allocated.
96 code = xfs_ialloc(tp, dp, mode, nlink, rdev, credp, prid, okalloc,
97 &ialloc_context, &call_again, &ip);
100 * Return an error if we were unable to allocate a new inode.
101 * This should only happen if we run out of space on disk or
102 * encounter a disk error.
104 if (code) {
105 *ipp = NULL;
106 return code;
108 if (!call_again && (ip == NULL)) {
109 *ipp = NULL;
110 return XFS_ERROR(ENOSPC);
114 * If call_again is set, then we were unable to get an
115 * inode in one operation. We need to commit the current
116 * transaction and call xfs_ialloc() again. It is guaranteed
117 * to succeed the second time.
119 if (call_again) {
122 * Normally, xfs_trans_commit releases all the locks.
123 * We call bhold to hang on to the ialloc_context across
124 * the commit. Holding this buffer prevents any other
125 * processes from doing any allocations in this
126 * allocation group.
128 xfs_trans_bhold(tp, ialloc_context);
130 * Save the log reservation so we can use
131 * them in the next transaction.
133 log_res = xfs_trans_get_log_res(tp);
134 log_count = xfs_trans_get_log_count(tp);
137 * We want the quota changes to be associated with the next
138 * transaction, NOT this one. So, detach the dqinfo from this
139 * and attach it to the next transaction.
141 dqinfo = NULL;
142 tflags = 0;
143 if (tp->t_dqinfo) {
144 dqinfo = (void *)tp->t_dqinfo;
145 tp->t_dqinfo = NULL;
146 tflags = tp->t_flags & XFS_TRANS_DQ_DIRTY;
147 tp->t_flags &= ~(XFS_TRANS_DQ_DIRTY);
150 ntp = xfs_trans_dup(tp);
151 code = xfs_trans_commit(tp, 0);
152 tp = ntp;
153 if (committed != NULL) {
154 *committed = 1;
157 * If we get an error during the commit processing,
158 * release the buffer that is still held and return
159 * to the caller.
161 if (code) {
162 xfs_buf_relse(ialloc_context);
163 if (dqinfo) {
164 tp->t_dqinfo = dqinfo;
165 xfs_trans_free_dqinfo(tp);
167 *tpp = ntp;
168 *ipp = NULL;
169 return code;
173 * transaction commit worked ok so we can drop the extra ticket
174 * reference that we gained in xfs_trans_dup()
176 xfs_log_ticket_put(tp->t_ticket);
177 code = xfs_trans_reserve(tp, 0, log_res, 0,
178 XFS_TRANS_PERM_LOG_RES, log_count);
180 * Re-attach the quota info that we detached from prev trx.
182 if (dqinfo) {
183 tp->t_dqinfo = dqinfo;
184 tp->t_flags |= tflags;
187 if (code) {
188 xfs_buf_relse(ialloc_context);
189 *tpp = ntp;
190 *ipp = NULL;
191 return code;
193 xfs_trans_bjoin(tp, ialloc_context);
196 * Call ialloc again. Since we've locked out all
197 * other allocations in this allocation group,
198 * this call should always succeed.
200 code = xfs_ialloc(tp, dp, mode, nlink, rdev, credp, prid,
201 okalloc, &ialloc_context, &call_again, &ip);
204 * If we get an error at this point, return to the caller
205 * so that the current transaction can be aborted.
207 if (code) {
208 *tpp = tp;
209 *ipp = NULL;
210 return code;
212 ASSERT ((!call_again) && (ip != NULL));
214 } else {
215 if (committed != NULL) {
216 *committed = 0;
220 *ipp = ip;
221 *tpp = tp;
223 return 0;
227 * Decrement the link count on an inode & log the change.
228 * If this causes the link count to go to zero, initiate the
229 * logging activity required to truncate a file.
231 int /* error */
232 xfs_droplink(
233 xfs_trans_t *tp,
234 xfs_inode_t *ip)
236 int error;
238 xfs_ichgtime(ip, XFS_ICHGTIME_CHG);
240 ASSERT (ip->i_d.di_nlink > 0);
241 ip->i_d.di_nlink--;
242 drop_nlink(VFS_I(ip));
243 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
245 error = 0;
246 if (ip->i_d.di_nlink == 0) {
248 * We're dropping the last link to this file.
249 * Move the on-disk inode to the AGI unlinked list.
250 * From xfs_inactive() we will pull the inode from
251 * the list and free it.
253 error = xfs_iunlink(tp, ip);
255 return error;
259 * This gets called when the inode's version needs to be changed from 1 to 2.
260 * Currently this happens when the nlink field overflows the old 16-bit value
261 * or when chproj is called to change the project for the first time.
262 * As a side effect the superblock version will also get rev'd
263 * to contain the NLINK bit.
265 void
266 xfs_bump_ino_vers2(
267 xfs_trans_t *tp,
268 xfs_inode_t *ip)
270 xfs_mount_t *mp;
272 ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
273 ASSERT(ip->i_d.di_version == 1);
275 ip->i_d.di_version = 2;
276 ip->i_d.di_onlink = 0;
277 memset(&(ip->i_d.di_pad[0]), 0, sizeof(ip->i_d.di_pad));
278 mp = tp->t_mountp;
279 if (!xfs_sb_version_hasnlink(&mp->m_sb)) {
280 spin_lock(&mp->m_sb_lock);
281 if (!xfs_sb_version_hasnlink(&mp->m_sb)) {
282 xfs_sb_version_addnlink(&mp->m_sb);
283 spin_unlock(&mp->m_sb_lock);
284 xfs_mod_sb(tp, XFS_SB_VERSIONNUM);
285 } else {
286 spin_unlock(&mp->m_sb_lock);
289 /* Caller must log the inode */
293 * Increment the link count on an inode & log the change.
296 xfs_bumplink(
297 xfs_trans_t *tp,
298 xfs_inode_t *ip)
300 if (ip->i_d.di_nlink >= XFS_MAXLINK)
301 return XFS_ERROR(EMLINK);
302 xfs_ichgtime(ip, XFS_ICHGTIME_CHG);
304 ASSERT(ip->i_d.di_nlink > 0);
305 ip->i_d.di_nlink++;
306 inc_nlink(VFS_I(ip));
307 if ((ip->i_d.di_version == 1) &&
308 (ip->i_d.di_nlink > XFS_MAXLINK_1)) {
310 * The inode has increased its number of links beyond
311 * what can fit in an old format inode. It now needs
312 * to be converted to a version 2 inode with a 32 bit
313 * link count. If this is the first inode in the file
314 * system to do this, then we need to bump the superblock
315 * version number as well.
317 xfs_bump_ino_vers2(tp, ip);
320 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
321 return 0;