1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (c) 2008-2010, 2013 Dave Chinner
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"
18 #include "xfs_log_priv.h"
19 #include "xfs_log_recover.h"
20 #include "xfs_ialloc.h"
21 #include "xfs_trace.h"
23 kmem_zone_t *xfs_icreate_zone; /* inode create item zone */
25 static inline struct xfs_icreate_item *ICR_ITEM(struct xfs_log_item *lip)
27 return container_of(lip, struct xfs_icreate_item, ic_item);
31 * This returns the number of iovecs needed to log the given inode item.
33 * We only need one iovec for the icreate log structure.
36 xfs_icreate_item_size(
37 struct xfs_log_item *lip,
42 *nbytes += sizeof(struct xfs_icreate_log);
46 * This is called to fill in the vector of log iovecs for the
47 * given inode create log item.
50 xfs_icreate_item_format(
51 struct xfs_log_item *lip,
52 struct xfs_log_vec *lv)
54 struct xfs_icreate_item *icp = ICR_ITEM(lip);
55 struct xfs_log_iovec *vecp = NULL;
57 xlog_copy_iovec(lv, &vecp, XLOG_REG_TYPE_ICREATE,
59 sizeof(struct xfs_icreate_log));
63 xfs_icreate_item_release(
64 struct xfs_log_item *lip)
66 kmem_cache_free(xfs_icreate_zone, ICR_ITEM(lip));
69 static const struct xfs_item_ops xfs_icreate_item_ops = {
70 .flags = XFS_ITEM_RELEASE_WHEN_COMMITTED,
71 .iop_size = xfs_icreate_item_size,
72 .iop_format = xfs_icreate_item_format,
73 .iop_release = xfs_icreate_item_release,
78 * Initialize the inode log item for a newly allocated (in-core) inode.
80 * Inode extents can only reside within an AG. Hence specify the starting
81 * block for the inode chunk by offset within an AG as well as the
82 * length of the allocated extent.
84 * This joins the item to the transaction and marks it dirty so
85 * that we don't need a separate call to do this, nor does the
86 * caller need to know anything about the icreate item.
94 unsigned int inode_size,
96 unsigned int generation)
98 struct xfs_icreate_item *icp;
100 icp = kmem_cache_zalloc(xfs_icreate_zone, GFP_KERNEL | __GFP_NOFAIL);
102 xfs_log_item_init(tp->t_mountp, &icp->ic_item, XFS_LI_ICREATE,
103 &xfs_icreate_item_ops);
105 icp->ic_format.icl_type = XFS_LI_ICREATE;
106 icp->ic_format.icl_size = 1; /* single vector */
107 icp->ic_format.icl_ag = cpu_to_be32(agno);
108 icp->ic_format.icl_agbno = cpu_to_be32(agbno);
109 icp->ic_format.icl_count = cpu_to_be32(count);
110 icp->ic_format.icl_isize = cpu_to_be32(inode_size);
111 icp->ic_format.icl_length = cpu_to_be32(length);
112 icp->ic_format.icl_gen = cpu_to_be32(generation);
114 xfs_trans_add_item(tp, &icp->ic_item);
115 tp->t_flags |= XFS_TRANS_DIRTY;
116 set_bit(XFS_LI_DIRTY, &icp->ic_item.li_flags);
119 static enum xlog_recover_reorder
120 xlog_recover_icreate_reorder(
121 struct xlog_recover_item *item)
124 * Inode allocation buffers must be replayed before subsequent inode
125 * items try to modify those buffers. ICREATE items are the logical
126 * equivalent of logging a newly initialized inode buffer, so recover
127 * these at the same time that we recover logged buffers.
129 return XLOG_REORDER_BUFFER_LIST;
133 * This routine is called when an inode create format structure is found in a
134 * committed transaction in the log. It's purpose is to initialise the inodes
135 * being allocated on disk. This requires us to get inode cluster buffers that
136 * match the range to be initialised, stamped with inode templates and written
137 * by delayed write so that subsequent modifications will hit the cached buffer
138 * and only need writing out at the end of recovery.
141 xlog_recover_icreate_commit_pass2(
143 struct list_head *buffer_list,
144 struct xlog_recover_item *item,
147 struct xfs_mount *mp = log->l_mp;
148 struct xfs_icreate_log *icl;
149 struct xfs_ino_geometry *igeo = M_IGEO(mp);
154 xfs_agblock_t length;
160 icl = (struct xfs_icreate_log *)item->ri_buf[0].i_addr;
161 if (icl->icl_type != XFS_LI_ICREATE) {
162 xfs_warn(log->l_mp, "xlog_recover_do_icreate_trans: bad type");
166 if (icl->icl_size != 1) {
167 xfs_warn(log->l_mp, "xlog_recover_do_icreate_trans: bad icl size");
171 agno = be32_to_cpu(icl->icl_ag);
172 if (agno >= mp->m_sb.sb_agcount) {
173 xfs_warn(log->l_mp, "xlog_recover_do_icreate_trans: bad agno");
176 agbno = be32_to_cpu(icl->icl_agbno);
177 if (!agbno || agbno == NULLAGBLOCK || agbno >= mp->m_sb.sb_agblocks) {
178 xfs_warn(log->l_mp, "xlog_recover_do_icreate_trans: bad agbno");
181 isize = be32_to_cpu(icl->icl_isize);
182 if (isize != mp->m_sb.sb_inodesize) {
183 xfs_warn(log->l_mp, "xlog_recover_do_icreate_trans: bad isize");
186 count = be32_to_cpu(icl->icl_count);
188 xfs_warn(log->l_mp, "xlog_recover_do_icreate_trans: bad count");
191 length = be32_to_cpu(icl->icl_length);
192 if (!length || length >= mp->m_sb.sb_agblocks) {
193 xfs_warn(log->l_mp, "xlog_recover_do_icreate_trans: bad length");
198 * The inode chunk is either full or sparse and we only support
199 * m_ino_geo.ialloc_min_blks sized sparse allocations at this time.
201 if (length != igeo->ialloc_blks &&
202 length != igeo->ialloc_min_blks) {
204 "%s: unsupported chunk length", __func__);
208 /* verify inode count is consistent with extent length */
209 if ((count >> mp->m_sb.sb_inopblog) != length) {
211 "%s: inconsistent inode count and chunk length",
217 * The icreate transaction can cover multiple cluster buffers and these
218 * buffers could have been freed and reused. Check the individual
219 * buffers for cancellation so we don't overwrite anything written after
222 bb_per_cluster = XFS_FSB_TO_BB(mp, igeo->blocks_per_cluster);
223 nbufs = length / igeo->blocks_per_cluster;
224 for (i = 0, cancel_count = 0; i < nbufs; i++) {
227 daddr = XFS_AGB_TO_DADDR(mp, agno,
228 agbno + i * igeo->blocks_per_cluster);
229 if (xlog_is_buffer_cancelled(log, daddr, bb_per_cluster))
234 * We currently only use icreate for a single allocation at a time. This
235 * means we should expect either all or none of the buffers to be
236 * cancelled. Be conservative and skip replay if at least one buffer is
237 * cancelled, but warn the user that something is awry if the buffers
238 * are not consistent.
240 * XXX: This must be refined to only skip cancelled clusters once we use
241 * icreate for multiple chunk allocations.
243 ASSERT(!cancel_count || cancel_count == nbufs);
245 if (cancel_count != nbufs)
247 "WARNING: partial inode chunk cancellation, skipped icreate.");
248 trace_xfs_log_recover_icreate_cancel(log, icl);
252 trace_xfs_log_recover_icreate_recover(log, icl);
253 return xfs_ialloc_inode_init(mp, NULL, buffer_list, count, agno, agbno,
254 length, be32_to_cpu(icl->icl_gen));
257 const struct xlog_recover_item_ops xlog_icreate_item_ops = {
258 .item_type = XFS_LI_ICREATE,
259 .reorder = xlog_recover_icreate_reorder,
260 .commit_pass2 = xlog_recover_icreate_commit_pass2,