Loading...
1/*
2 * Copyright (c) 2008-2010, 2013 Dave Chinner
3 * All Rights Reserved.
4 *
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.
8 *
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.
13 *
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
17 */
18#include "xfs.h"
19#include "xfs_fs.h"
20#include "xfs_shared.h"
21#include "xfs_format.h"
22#include "xfs_log_format.h"
23#include "xfs_trans_resv.h"
24#include "xfs_bit.h"
25#include "xfs_mount.h"
26#include "xfs_trans.h"
27#include "xfs_trans_priv.h"
28#include "xfs_error.h"
29#include "xfs_icreate_item.h"
30#include "xfs_log.h"
31
32kmem_zone_t *xfs_icreate_zone; /* inode create item zone */
33
34static inline struct xfs_icreate_item *ICR_ITEM(struct xfs_log_item *lip)
35{
36 return container_of(lip, struct xfs_icreate_item, ic_item);
37}
38
39/*
40 * This returns the number of iovecs needed to log the given inode item.
41 *
42 * We only need one iovec for the icreate log structure.
43 */
44STATIC void
45xfs_icreate_item_size(
46 struct xfs_log_item *lip,
47 int *nvecs,
48 int *nbytes)
49{
50 *nvecs += 1;
51 *nbytes += sizeof(struct xfs_icreate_log);
52}
53
54/*
55 * This is called to fill in the vector of log iovecs for the
56 * given inode create log item.
57 */
58STATIC void
59xfs_icreate_item_format(
60 struct xfs_log_item *lip,
61 struct xfs_log_vec *lv)
62{
63 struct xfs_icreate_item *icp = ICR_ITEM(lip);
64 struct xfs_log_iovec *vecp = NULL;
65
66 xlog_copy_iovec(lv, &vecp, XLOG_REG_TYPE_ICREATE,
67 &icp->ic_format,
68 sizeof(struct xfs_icreate_log));
69}
70
71
72/* Pinning has no meaning for the create item, so just return. */
73STATIC void
74xfs_icreate_item_pin(
75 struct xfs_log_item *lip)
76{
77}
78
79
80/* pinning has no meaning for the create item, so just return. */
81STATIC void
82xfs_icreate_item_unpin(
83 struct xfs_log_item *lip,
84 int remove)
85{
86}
87
88STATIC void
89xfs_icreate_item_unlock(
90 struct xfs_log_item *lip)
91{
92 struct xfs_icreate_item *icp = ICR_ITEM(lip);
93
94 if (icp->ic_item.li_flags & XFS_LI_ABORTED)
95 kmem_zone_free(xfs_icreate_zone, icp);
96 return;
97}
98
99/*
100 * Because we have ordered buffers being tracked in the AIL for the inode
101 * creation, we don't need the create item after this. Hence we can free
102 * the log item and return -1 to tell the caller we're done with the item.
103 */
104STATIC xfs_lsn_t
105xfs_icreate_item_committed(
106 struct xfs_log_item *lip,
107 xfs_lsn_t lsn)
108{
109 struct xfs_icreate_item *icp = ICR_ITEM(lip);
110
111 kmem_zone_free(xfs_icreate_zone, icp);
112 return (xfs_lsn_t)-1;
113}
114
115/* item can never get into the AIL */
116STATIC uint
117xfs_icreate_item_push(
118 struct xfs_log_item *lip,
119 struct list_head *buffer_list)
120{
121 ASSERT(0);
122 return XFS_ITEM_SUCCESS;
123}
124
125/* Ordered buffers do the dependency tracking here, so this does nothing. */
126STATIC void
127xfs_icreate_item_committing(
128 struct xfs_log_item *lip,
129 xfs_lsn_t lsn)
130{
131}
132
133/*
134 * This is the ops vector shared by all buf log items.
135 */
136static const struct xfs_item_ops xfs_icreate_item_ops = {
137 .iop_size = xfs_icreate_item_size,
138 .iop_format = xfs_icreate_item_format,
139 .iop_pin = xfs_icreate_item_pin,
140 .iop_unpin = xfs_icreate_item_unpin,
141 .iop_push = xfs_icreate_item_push,
142 .iop_unlock = xfs_icreate_item_unlock,
143 .iop_committed = xfs_icreate_item_committed,
144 .iop_committing = xfs_icreate_item_committing,
145};
146
147
148/*
149 * Initialize the inode log item for a newly allocated (in-core) inode.
150 *
151 * Inode extents can only reside within an AG. Hence specify the starting
152 * block for the inode chunk by offset within an AG as well as the
153 * length of the allocated extent.
154 *
155 * This joins the item to the transaction and marks it dirty so
156 * that we don't need a separate call to do this, nor does the
157 * caller need to know anything about the icreate item.
158 */
159void
160xfs_icreate_log(
161 struct xfs_trans *tp,
162 xfs_agnumber_t agno,
163 xfs_agblock_t agbno,
164 unsigned int count,
165 unsigned int inode_size,
166 xfs_agblock_t length,
167 unsigned int generation)
168{
169 struct xfs_icreate_item *icp;
170
171 icp = kmem_zone_zalloc(xfs_icreate_zone, KM_SLEEP);
172
173 xfs_log_item_init(tp->t_mountp, &icp->ic_item, XFS_LI_ICREATE,
174 &xfs_icreate_item_ops);
175
176 icp->ic_format.icl_type = XFS_LI_ICREATE;
177 icp->ic_format.icl_size = 1; /* single vector */
178 icp->ic_format.icl_ag = cpu_to_be32(agno);
179 icp->ic_format.icl_agbno = cpu_to_be32(agbno);
180 icp->ic_format.icl_count = cpu_to_be32(count);
181 icp->ic_format.icl_isize = cpu_to_be32(inode_size);
182 icp->ic_format.icl_length = cpu_to_be32(length);
183 icp->ic_format.icl_gen = cpu_to_be32(generation);
184
185 xfs_trans_add_item(tp, &icp->ic_item);
186 tp->t_flags |= XFS_TRANS_DIRTY;
187 icp->ic_item.li_desc->lid_flags |= XFS_LID_DIRTY;
188}
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * Copyright (c) 2008-2010, 2013 Dave Chinner
4 * All Rights Reserved.
5 */
6#include "xfs.h"
7#include "xfs_fs.h"
8#include "xfs_shared.h"
9#include "xfs_format.h"
10#include "xfs_log_format.h"
11#include "xfs_trans_resv.h"
12#include "xfs_mount.h"
13#include "xfs_inode.h"
14#include "xfs_trans.h"
15#include "xfs_trans_priv.h"
16#include "xfs_icreate_item.h"
17#include "xfs_log.h"
18#include "xfs_log_priv.h"
19#include "xfs_log_recover.h"
20#include "xfs_ialloc.h"
21#include "xfs_trace.h"
22
23struct kmem_cache *xfs_icreate_cache; /* inode create item */
24
25static inline struct xfs_icreate_item *ICR_ITEM(struct xfs_log_item *lip)
26{
27 return container_of(lip, struct xfs_icreate_item, ic_item);
28}
29
30/*
31 * This returns the number of iovecs needed to log the given inode item.
32 *
33 * We only need one iovec for the icreate log structure.
34 */
35STATIC void
36xfs_icreate_item_size(
37 struct xfs_log_item *lip,
38 int *nvecs,
39 int *nbytes)
40{
41 *nvecs += 1;
42 *nbytes += sizeof(struct xfs_icreate_log);
43}
44
45/*
46 * This is called to fill in the vector of log iovecs for the
47 * given inode create log item.
48 */
49STATIC void
50xfs_icreate_item_format(
51 struct xfs_log_item *lip,
52 struct xfs_log_vec *lv)
53{
54 struct xfs_icreate_item *icp = ICR_ITEM(lip);
55 struct xfs_log_iovec *vecp = NULL;
56
57 xlog_copy_iovec(lv, &vecp, XLOG_REG_TYPE_ICREATE,
58 &icp->ic_format,
59 sizeof(struct xfs_icreate_log));
60}
61
62STATIC void
63xfs_icreate_item_release(
64 struct xfs_log_item *lip)
65{
66 kvfree(ICR_ITEM(lip)->ic_item.li_lv_shadow);
67 kmem_cache_free(xfs_icreate_cache, ICR_ITEM(lip));
68}
69
70static const struct xfs_item_ops xfs_icreate_item_ops = {
71 .flags = XFS_ITEM_RELEASE_WHEN_COMMITTED,
72 .iop_size = xfs_icreate_item_size,
73 .iop_format = xfs_icreate_item_format,
74 .iop_release = xfs_icreate_item_release,
75};
76
77
78/*
79 * Initialize the inode log item for a newly allocated (in-core) inode.
80 *
81 * Inode extents can only reside within an AG. Hence specify the starting
82 * block for the inode chunk by offset within an AG as well as the
83 * length of the allocated extent.
84 *
85 * This joins the item to the transaction and marks it dirty so
86 * that we don't need a separate call to do this, nor does the
87 * caller need to know anything about the icreate item.
88 */
89void
90xfs_icreate_log(
91 struct xfs_trans *tp,
92 xfs_agnumber_t agno,
93 xfs_agblock_t agbno,
94 unsigned int count,
95 unsigned int inode_size,
96 xfs_agblock_t length,
97 unsigned int generation)
98{
99 struct xfs_icreate_item *icp;
100
101 icp = kmem_cache_zalloc(xfs_icreate_cache, GFP_KERNEL | __GFP_NOFAIL);
102
103 xfs_log_item_init(tp->t_mountp, &icp->ic_item, XFS_LI_ICREATE,
104 &xfs_icreate_item_ops);
105
106 icp->ic_format.icl_type = XFS_LI_ICREATE;
107 icp->ic_format.icl_size = 1; /* single vector */
108 icp->ic_format.icl_ag = cpu_to_be32(agno);
109 icp->ic_format.icl_agbno = cpu_to_be32(agbno);
110 icp->ic_format.icl_count = cpu_to_be32(count);
111 icp->ic_format.icl_isize = cpu_to_be32(inode_size);
112 icp->ic_format.icl_length = cpu_to_be32(length);
113 icp->ic_format.icl_gen = cpu_to_be32(generation);
114
115 xfs_trans_add_item(tp, &icp->ic_item);
116 tp->t_flags |= XFS_TRANS_DIRTY;
117 set_bit(XFS_LI_DIRTY, &icp->ic_item.li_flags);
118}
119
120static enum xlog_recover_reorder
121xlog_recover_icreate_reorder(
122 struct xlog_recover_item *item)
123{
124 /*
125 * Inode allocation buffers must be replayed before subsequent inode
126 * items try to modify those buffers. ICREATE items are the logical
127 * equivalent of logging a newly initialized inode buffer, so recover
128 * these at the same time that we recover logged buffers.
129 */
130 return XLOG_REORDER_BUFFER_LIST;
131}
132
133/*
134 * This routine is called when an inode create format structure is found in a
135 * committed transaction in the log. It's purpose is to initialise the inodes
136 * being allocated on disk. This requires us to get inode cluster buffers that
137 * match the range to be initialised, stamped with inode templates and written
138 * by delayed write so that subsequent modifications will hit the cached buffer
139 * and only need writing out at the end of recovery.
140 */
141STATIC int
142xlog_recover_icreate_commit_pass2(
143 struct xlog *log,
144 struct list_head *buffer_list,
145 struct xlog_recover_item *item,
146 xfs_lsn_t lsn)
147{
148 struct xfs_mount *mp = log->l_mp;
149 struct xfs_icreate_log *icl;
150 struct xfs_ino_geometry *igeo = M_IGEO(mp);
151 xfs_agnumber_t agno;
152 xfs_agblock_t agbno;
153 unsigned int count;
154 unsigned int isize;
155 xfs_agblock_t length;
156 int bb_per_cluster;
157 int cancel_count;
158 int nbufs;
159 int i;
160
161 icl = (struct xfs_icreate_log *)item->ri_buf[0].i_addr;
162 if (icl->icl_type != XFS_LI_ICREATE) {
163 xfs_warn(log->l_mp, "xlog_recover_do_icreate_trans: bad type");
164 return -EINVAL;
165 }
166
167 if (icl->icl_size != 1) {
168 xfs_warn(log->l_mp, "xlog_recover_do_icreate_trans: bad icl size");
169 return -EINVAL;
170 }
171
172 agno = be32_to_cpu(icl->icl_ag);
173 if (agno >= mp->m_sb.sb_agcount) {
174 xfs_warn(log->l_mp, "xlog_recover_do_icreate_trans: bad agno");
175 return -EINVAL;
176 }
177 agbno = be32_to_cpu(icl->icl_agbno);
178 if (!agbno || agbno == NULLAGBLOCK || agbno >= mp->m_sb.sb_agblocks) {
179 xfs_warn(log->l_mp, "xlog_recover_do_icreate_trans: bad agbno");
180 return -EINVAL;
181 }
182 isize = be32_to_cpu(icl->icl_isize);
183 if (isize != mp->m_sb.sb_inodesize) {
184 xfs_warn(log->l_mp, "xlog_recover_do_icreate_trans: bad isize");
185 return -EINVAL;
186 }
187 count = be32_to_cpu(icl->icl_count);
188 if (!count) {
189 xfs_warn(log->l_mp, "xlog_recover_do_icreate_trans: bad count");
190 return -EINVAL;
191 }
192 length = be32_to_cpu(icl->icl_length);
193 if (!length || length >= mp->m_sb.sb_agblocks) {
194 xfs_warn(log->l_mp, "xlog_recover_do_icreate_trans: bad length");
195 return -EINVAL;
196 }
197
198 /*
199 * The inode chunk is either full or sparse and we only support
200 * m_ino_geo.ialloc_min_blks sized sparse allocations at this time.
201 */
202 if (length != igeo->ialloc_blks &&
203 length != igeo->ialloc_min_blks) {
204 xfs_warn(log->l_mp,
205 "%s: unsupported chunk length", __func__);
206 return -EINVAL;
207 }
208
209 /* verify inode count is consistent with extent length */
210 if ((count >> mp->m_sb.sb_inopblog) != length) {
211 xfs_warn(log->l_mp,
212 "%s: inconsistent inode count and chunk length",
213 __func__);
214 return -EINVAL;
215 }
216
217 /*
218 * The icreate transaction can cover multiple cluster buffers and these
219 * buffers could have been freed and reused. Check the individual
220 * buffers for cancellation so we don't overwrite anything written after
221 * a cancellation.
222 */
223 bb_per_cluster = XFS_FSB_TO_BB(mp, igeo->blocks_per_cluster);
224 nbufs = length / igeo->blocks_per_cluster;
225 for (i = 0, cancel_count = 0; i < nbufs; i++) {
226 xfs_daddr_t daddr;
227
228 daddr = XFS_AGB_TO_DADDR(mp, agno,
229 agbno + i * igeo->blocks_per_cluster);
230 if (xlog_is_buffer_cancelled(log, daddr, bb_per_cluster))
231 cancel_count++;
232 }
233
234 /*
235 * We currently only use icreate for a single allocation at a time. This
236 * means we should expect either all or none of the buffers to be
237 * cancelled. Be conservative and skip replay if at least one buffer is
238 * cancelled, but warn the user that something is awry if the buffers
239 * are not consistent.
240 *
241 * XXX: This must be refined to only skip cancelled clusters once we use
242 * icreate for multiple chunk allocations.
243 */
244 ASSERT(!cancel_count || cancel_count == nbufs);
245 if (cancel_count) {
246 if (cancel_count != nbufs)
247 xfs_warn(mp,
248 "WARNING: partial inode chunk cancellation, skipped icreate.");
249 trace_xfs_log_recover_icreate_cancel(log, icl);
250 return 0;
251 }
252
253 trace_xfs_log_recover_icreate_recover(log, icl);
254 return xfs_ialloc_inode_init(mp, NULL, buffer_list, count, agno, agbno,
255 length, be32_to_cpu(icl->icl_gen));
256}
257
258const struct xlog_recover_item_ops xlog_icreate_item_ops = {
259 .item_type = XFS_LI_ICREATE,
260 .reorder = xlog_recover_icreate_reorder,
261 .commit_pass2 = xlog_recover_icreate_commit_pass2,
262};