Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
Nathan Scott | 7b71876 | 2005-11-02 14:58:39 +1100 | [diff] [blame] | 2 | * Copyright (c) 2000-2005 Silicon Graphics, Inc. |
| 3 | * All Rights Reserved. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4 | * |
Nathan Scott | 7b71876 | 2005-11-02 14:58:39 +1100 | [diff] [blame] | 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 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 7 | * published by the Free Software Foundation. |
| 8 | * |
Nathan Scott | 7b71876 | 2005-11-02 14:58:39 +1100 | [diff] [blame] | 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. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 13 | * |
Nathan Scott | 7b71876 | 2005-11-02 14:58:39 +1100 | [diff] [blame] | 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 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 17 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 18 | #include "xfs.h" |
Nathan Scott | a844f45 | 2005-11-02 14:38:42 +1100 | [diff] [blame] | 19 | #include "xfs_fs.h" |
Dave Chinner | 6ca1c90 | 2013-08-12 20:49:26 +1000 | [diff] [blame^] | 20 | #include "xfs_format.h" |
Nathan Scott | a844f45 | 2005-11-02 14:38:42 +1100 | [diff] [blame] | 21 | #include "xfs_bit.h" |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 22 | #include "xfs_log.h" |
Nathan Scott | a844f45 | 2005-11-02 14:38:42 +1100 | [diff] [blame] | 23 | #include "xfs_inum.h" |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 24 | #include "xfs_trans.h" |
Christoph Hellwig | 211e4d4 | 2012-04-23 15:58:34 +1000 | [diff] [blame] | 25 | #include "xfs_trans_priv.h" |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 26 | #include "xfs_sb.h" |
| 27 | #include "xfs_ag.h" |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 28 | #include "xfs_dir2.h" |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 29 | #include "xfs_mount.h" |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 30 | #include "xfs_bmap_btree.h" |
Nathan Scott | a844f45 | 2005-11-02 14:38:42 +1100 | [diff] [blame] | 31 | #include "xfs_alloc_btree.h" |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 32 | #include "xfs_ialloc_btree.h" |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 33 | #include "xfs_dinode.h" |
| 34 | #include "xfs_inode.h" |
Nathan Scott | a844f45 | 2005-11-02 14:38:42 +1100 | [diff] [blame] | 35 | #include "xfs_btree.h" |
| 36 | #include "xfs_ialloc.h" |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 37 | #include "xfs_alloc.h" |
| 38 | #include "xfs_rtalloc.h" |
| 39 | #include "xfs_bmap.h" |
| 40 | #include "xfs_error.h" |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 41 | #include "xfs_quota.h" |
| 42 | #include "xfs_fsops.h" |
Christoph Hellwig | 4335509 | 2008-03-27 18:01:08 +1100 | [diff] [blame] | 43 | #include "xfs_utils.h" |
Christoph Hellwig | 0b1b213 | 2009-12-14 23:14:59 +0000 | [diff] [blame] | 44 | #include "xfs_trace.h" |
Dave Chinner | 6d8b79c | 2012-10-08 21:56:09 +1100 | [diff] [blame] | 45 | #include "xfs_icache.h" |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 46 | #include "xfs_cksum.h" |
| 47 | #include "xfs_buf_item.h" |
Christoph Hellwig | 0b1b213 | 2009-12-14 23:14:59 +0000 | [diff] [blame] | 48 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 49 | |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 50 | #ifdef HAVE_PERCPU_SB |
David Chinner | 20f4ebf | 2007-02-10 18:36:10 +1100 | [diff] [blame] | 51 | STATIC void xfs_icsb_balance_counter(xfs_mount_t *, xfs_sb_field_t, |
Christoph Hellwig | 45af6c6 | 2008-04-22 17:34:44 +1000 | [diff] [blame] | 52 | int); |
| 53 | STATIC void xfs_icsb_balance_counter_locked(xfs_mount_t *, xfs_sb_field_t, |
| 54 | int); |
David Chinner | 36fbe6e | 2008-04-10 12:19:56 +1000 | [diff] [blame] | 55 | STATIC void xfs_icsb_disable_counter(xfs_mount_t *, xfs_sb_field_t); |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 56 | #else |
| 57 | |
Christoph Hellwig | 45af6c6 | 2008-04-22 17:34:44 +1000 | [diff] [blame] | 58 | #define xfs_icsb_balance_counter(mp, a, b) do { } while (0) |
| 59 | #define xfs_icsb_balance_counter_locked(mp, a, b) do { } while (0) |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 60 | #endif |
| 61 | |
Christoph Hellwig | 1df84c9 | 2006-01-11 15:29:52 +1100 | [diff] [blame] | 62 | static const struct { |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 63 | short offset; |
| 64 | short type; /* 0 = integer |
| 65 | * 1 = binary / string (no translation) |
| 66 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 67 | } xfs_sb_info[] = { |
| 68 | { offsetof(xfs_sb_t, sb_magicnum), 0 }, |
| 69 | { offsetof(xfs_sb_t, sb_blocksize), 0 }, |
| 70 | { offsetof(xfs_sb_t, sb_dblocks), 0 }, |
| 71 | { offsetof(xfs_sb_t, sb_rblocks), 0 }, |
| 72 | { offsetof(xfs_sb_t, sb_rextents), 0 }, |
| 73 | { offsetof(xfs_sb_t, sb_uuid), 1 }, |
| 74 | { offsetof(xfs_sb_t, sb_logstart), 0 }, |
| 75 | { offsetof(xfs_sb_t, sb_rootino), 0 }, |
| 76 | { offsetof(xfs_sb_t, sb_rbmino), 0 }, |
| 77 | { offsetof(xfs_sb_t, sb_rsumino), 0 }, |
| 78 | { offsetof(xfs_sb_t, sb_rextsize), 0 }, |
| 79 | { offsetof(xfs_sb_t, sb_agblocks), 0 }, |
| 80 | { offsetof(xfs_sb_t, sb_agcount), 0 }, |
| 81 | { offsetof(xfs_sb_t, sb_rbmblocks), 0 }, |
| 82 | { offsetof(xfs_sb_t, sb_logblocks), 0 }, |
| 83 | { offsetof(xfs_sb_t, sb_versionnum), 0 }, |
| 84 | { offsetof(xfs_sb_t, sb_sectsize), 0 }, |
| 85 | { offsetof(xfs_sb_t, sb_inodesize), 0 }, |
| 86 | { offsetof(xfs_sb_t, sb_inopblock), 0 }, |
| 87 | { offsetof(xfs_sb_t, sb_fname[0]), 1 }, |
| 88 | { offsetof(xfs_sb_t, sb_blocklog), 0 }, |
| 89 | { offsetof(xfs_sb_t, sb_sectlog), 0 }, |
| 90 | { offsetof(xfs_sb_t, sb_inodelog), 0 }, |
| 91 | { offsetof(xfs_sb_t, sb_inopblog), 0 }, |
| 92 | { offsetof(xfs_sb_t, sb_agblklog), 0 }, |
| 93 | { offsetof(xfs_sb_t, sb_rextslog), 0 }, |
| 94 | { offsetof(xfs_sb_t, sb_inprogress), 0 }, |
| 95 | { offsetof(xfs_sb_t, sb_imax_pct), 0 }, |
| 96 | { offsetof(xfs_sb_t, sb_icount), 0 }, |
| 97 | { offsetof(xfs_sb_t, sb_ifree), 0 }, |
| 98 | { offsetof(xfs_sb_t, sb_fdblocks), 0 }, |
| 99 | { offsetof(xfs_sb_t, sb_frextents), 0 }, |
| 100 | { offsetof(xfs_sb_t, sb_uquotino), 0 }, |
| 101 | { offsetof(xfs_sb_t, sb_gquotino), 0 }, |
| 102 | { offsetof(xfs_sb_t, sb_qflags), 0 }, |
| 103 | { offsetof(xfs_sb_t, sb_flags), 0 }, |
| 104 | { offsetof(xfs_sb_t, sb_shared_vn), 0 }, |
| 105 | { offsetof(xfs_sb_t, sb_inoalignmt), 0 }, |
| 106 | { offsetof(xfs_sb_t, sb_unit), 0 }, |
| 107 | { offsetof(xfs_sb_t, sb_width), 0 }, |
| 108 | { offsetof(xfs_sb_t, sb_dirblklog), 0 }, |
| 109 | { offsetof(xfs_sb_t, sb_logsectlog), 0 }, |
| 110 | { offsetof(xfs_sb_t, sb_logsectsize),0 }, |
| 111 | { offsetof(xfs_sb_t, sb_logsunit), 0 }, |
| 112 | { offsetof(xfs_sb_t, sb_features2), 0 }, |
David Chinner | ee1c090 | 2008-03-06 13:45:50 +1100 | [diff] [blame] | 113 | { offsetof(xfs_sb_t, sb_bad_features2), 0 }, |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 114 | { offsetof(xfs_sb_t, sb_features_compat), 0 }, |
| 115 | { offsetof(xfs_sb_t, sb_features_ro_compat), 0 }, |
| 116 | { offsetof(xfs_sb_t, sb_features_incompat), 0 }, |
Dave Chinner | e721f50 | 2013-04-03 16:11:32 +1100 | [diff] [blame] | 117 | { offsetof(xfs_sb_t, sb_features_log_incompat), 0 }, |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 118 | { offsetof(xfs_sb_t, sb_crc), 0 }, |
Dave Chinner | e721f50 | 2013-04-03 16:11:32 +1100 | [diff] [blame] | 119 | { offsetof(xfs_sb_t, sb_pad), 0 }, |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 120 | { offsetof(xfs_sb_t, sb_pquotino), 0 }, |
| 121 | { offsetof(xfs_sb_t, sb_lsn), 0 }, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 122 | { sizeof(xfs_sb_t), 0 } |
| 123 | }; |
| 124 | |
Christoph Hellwig | 2717420 | 2009-03-30 10:21:31 +0200 | [diff] [blame] | 125 | static DEFINE_MUTEX(xfs_uuid_table_mutex); |
| 126 | static int xfs_uuid_table_size; |
| 127 | static uuid_t *xfs_uuid_table; |
| 128 | |
| 129 | /* |
| 130 | * See if the UUID is unique among mounted XFS filesystems. |
| 131 | * Mount fails if UUID is nil or a FS with the same UUID is already mounted. |
| 132 | */ |
| 133 | STATIC int |
| 134 | xfs_uuid_mount( |
| 135 | struct xfs_mount *mp) |
| 136 | { |
| 137 | uuid_t *uuid = &mp->m_sb.sb_uuid; |
| 138 | int hole, i; |
| 139 | |
| 140 | if (mp->m_flags & XFS_MOUNT_NOUUID) |
| 141 | return 0; |
| 142 | |
| 143 | if (uuid_is_nil(uuid)) { |
Dave Chinner | 0b932cc | 2011-03-07 10:08:35 +1100 | [diff] [blame] | 144 | xfs_warn(mp, "Filesystem has nil UUID - can't mount"); |
Christoph Hellwig | 2717420 | 2009-03-30 10:21:31 +0200 | [diff] [blame] | 145 | return XFS_ERROR(EINVAL); |
| 146 | } |
| 147 | |
| 148 | mutex_lock(&xfs_uuid_table_mutex); |
| 149 | for (i = 0, hole = -1; i < xfs_uuid_table_size; i++) { |
| 150 | if (uuid_is_nil(&xfs_uuid_table[i])) { |
| 151 | hole = i; |
| 152 | continue; |
| 153 | } |
| 154 | if (uuid_equal(uuid, &xfs_uuid_table[i])) |
| 155 | goto out_duplicate; |
| 156 | } |
| 157 | |
| 158 | if (hole < 0) { |
| 159 | xfs_uuid_table = kmem_realloc(xfs_uuid_table, |
| 160 | (xfs_uuid_table_size + 1) * sizeof(*xfs_uuid_table), |
| 161 | xfs_uuid_table_size * sizeof(*xfs_uuid_table), |
| 162 | KM_SLEEP); |
| 163 | hole = xfs_uuid_table_size++; |
| 164 | } |
| 165 | xfs_uuid_table[hole] = *uuid; |
| 166 | mutex_unlock(&xfs_uuid_table_mutex); |
| 167 | |
| 168 | return 0; |
| 169 | |
| 170 | out_duplicate: |
| 171 | mutex_unlock(&xfs_uuid_table_mutex); |
Mitsuo Hayasaka | 021000e | 2012-01-13 05:58:39 +0000 | [diff] [blame] | 172 | xfs_warn(mp, "Filesystem has duplicate UUID %pU - can't mount", uuid); |
Christoph Hellwig | 2717420 | 2009-03-30 10:21:31 +0200 | [diff] [blame] | 173 | return XFS_ERROR(EINVAL); |
| 174 | } |
| 175 | |
| 176 | STATIC void |
| 177 | xfs_uuid_unmount( |
| 178 | struct xfs_mount *mp) |
| 179 | { |
| 180 | uuid_t *uuid = &mp->m_sb.sb_uuid; |
| 181 | int i; |
| 182 | |
| 183 | if (mp->m_flags & XFS_MOUNT_NOUUID) |
| 184 | return; |
| 185 | |
| 186 | mutex_lock(&xfs_uuid_table_mutex); |
| 187 | for (i = 0; i < xfs_uuid_table_size; i++) { |
| 188 | if (uuid_is_nil(&xfs_uuid_table[i])) |
| 189 | continue; |
| 190 | if (!uuid_equal(uuid, &xfs_uuid_table[i])) |
| 191 | continue; |
| 192 | memset(&xfs_uuid_table[i], 0, sizeof(uuid_t)); |
| 193 | break; |
| 194 | } |
| 195 | ASSERT(i < xfs_uuid_table_size); |
| 196 | mutex_unlock(&xfs_uuid_table_mutex); |
| 197 | } |
| 198 | |
| 199 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 200 | /* |
Dave Chinner | 0fa800f | 2010-01-11 11:47:46 +0000 | [diff] [blame] | 201 | * Reference counting access wrappers to the perag structures. |
Dave Chinner | e176579 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 202 | * Because we never free per-ag structures, the only thing we |
| 203 | * have to protect against changes is the tree structure itself. |
Dave Chinner | 0fa800f | 2010-01-11 11:47:46 +0000 | [diff] [blame] | 204 | */ |
| 205 | struct xfs_perag * |
| 206 | xfs_perag_get(struct xfs_mount *mp, xfs_agnumber_t agno) |
| 207 | { |
| 208 | struct xfs_perag *pag; |
| 209 | int ref = 0; |
| 210 | |
Dave Chinner | e176579 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 211 | rcu_read_lock(); |
Dave Chinner | 0fa800f | 2010-01-11 11:47:46 +0000 | [diff] [blame] | 212 | pag = radix_tree_lookup(&mp->m_perag_tree, agno); |
| 213 | if (pag) { |
| 214 | ASSERT(atomic_read(&pag->pag_ref) >= 0); |
Dave Chinner | 0fa800f | 2010-01-11 11:47:46 +0000 | [diff] [blame] | 215 | ref = atomic_inc_return(&pag->pag_ref); |
| 216 | } |
Dave Chinner | e176579 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 217 | rcu_read_unlock(); |
Dave Chinner | 0fa800f | 2010-01-11 11:47:46 +0000 | [diff] [blame] | 218 | trace_xfs_perag_get(mp, agno, ref, _RET_IP_); |
| 219 | return pag; |
| 220 | } |
| 221 | |
Dave Chinner | 65d0f20 | 2010-09-24 18:40:15 +1000 | [diff] [blame] | 222 | /* |
| 223 | * search from @first to find the next perag with the given tag set. |
| 224 | */ |
| 225 | struct xfs_perag * |
| 226 | xfs_perag_get_tag( |
| 227 | struct xfs_mount *mp, |
| 228 | xfs_agnumber_t first, |
| 229 | int tag) |
| 230 | { |
| 231 | struct xfs_perag *pag; |
| 232 | int found; |
| 233 | int ref; |
| 234 | |
| 235 | rcu_read_lock(); |
| 236 | found = radix_tree_gang_lookup_tag(&mp->m_perag_tree, |
| 237 | (void **)&pag, first, 1, tag); |
| 238 | if (found <= 0) { |
| 239 | rcu_read_unlock(); |
| 240 | return NULL; |
| 241 | } |
| 242 | ref = atomic_inc_return(&pag->pag_ref); |
| 243 | rcu_read_unlock(); |
| 244 | trace_xfs_perag_get_tag(mp, pag->pag_agno, ref, _RET_IP_); |
| 245 | return pag; |
| 246 | } |
| 247 | |
Dave Chinner | 0fa800f | 2010-01-11 11:47:46 +0000 | [diff] [blame] | 248 | void |
| 249 | xfs_perag_put(struct xfs_perag *pag) |
| 250 | { |
| 251 | int ref; |
| 252 | |
| 253 | ASSERT(atomic_read(&pag->pag_ref) > 0); |
| 254 | ref = atomic_dec_return(&pag->pag_ref); |
| 255 | trace_xfs_perag_put(pag->pag_mount, pag->pag_agno, ref, _RET_IP_); |
| 256 | } |
| 257 | |
Dave Chinner | e176579 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 258 | STATIC void |
| 259 | __xfs_free_perag( |
| 260 | struct rcu_head *head) |
| 261 | { |
| 262 | struct xfs_perag *pag = container_of(head, struct xfs_perag, rcu_head); |
| 263 | |
| 264 | ASSERT(atomic_read(&pag->pag_ref) == 0); |
| 265 | kmem_free(pag); |
| 266 | } |
| 267 | |
Dave Chinner | 0fa800f | 2010-01-11 11:47:46 +0000 | [diff] [blame] | 268 | /* |
Dave Chinner | e176579 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 269 | * Free up the per-ag resources associated with the mount structure. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 270 | */ |
Christoph Hellwig | c962fb7 | 2008-05-20 15:10:52 +1000 | [diff] [blame] | 271 | STATIC void |
Christoph Hellwig | ff4f038 | 2008-08-13 16:50:47 +1000 | [diff] [blame] | 272 | xfs_free_perag( |
Christoph Hellwig | 745f691 | 2007-08-30 17:20:39 +1000 | [diff] [blame] | 273 | xfs_mount_t *mp) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 274 | { |
Dave Chinner | 1c1c6eb | 2010-01-11 11:47:44 +0000 | [diff] [blame] | 275 | xfs_agnumber_t agno; |
| 276 | struct xfs_perag *pag; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 277 | |
Dave Chinner | 1c1c6eb | 2010-01-11 11:47:44 +0000 | [diff] [blame] | 278 | for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) { |
| 279 | spin_lock(&mp->m_perag_lock); |
| 280 | pag = radix_tree_delete(&mp->m_perag_tree, agno); |
| 281 | spin_unlock(&mp->m_perag_lock); |
Dave Chinner | e176579 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 282 | ASSERT(pag); |
Dave Chinner | f83282a | 2010-11-08 08:55:04 +0000 | [diff] [blame] | 283 | ASSERT(atomic_read(&pag->pag_ref) == 0); |
Dave Chinner | e176579 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 284 | call_rcu(&pag->rcu_head, __xfs_free_perag); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 285 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 286 | } |
| 287 | |
Nathan Scott | 4cc929e | 2007-05-14 18:24:02 +1000 | [diff] [blame] | 288 | /* |
| 289 | * Check size of device based on the (data/realtime) block count. |
| 290 | * Note: this check is used by the growfs code as well as mount. |
| 291 | */ |
| 292 | int |
| 293 | xfs_sb_validate_fsb_count( |
| 294 | xfs_sb_t *sbp, |
| 295 | __uint64_t nblocks) |
| 296 | { |
| 297 | ASSERT(PAGE_SHIFT >= sbp->sb_blocklog); |
| 298 | ASSERT(sbp->sb_blocklog >= BBSHIFT); |
| 299 | |
| 300 | #if XFS_BIG_BLKNOS /* Limited by ULONG_MAX of page cache index */ |
| 301 | if (nblocks >> (PAGE_CACHE_SHIFT - sbp->sb_blocklog) > ULONG_MAX) |
Eric Sandeen | 657a4cf | 2010-04-30 03:42:49 +0000 | [diff] [blame] | 302 | return EFBIG; |
Nathan Scott | 4cc929e | 2007-05-14 18:24:02 +1000 | [diff] [blame] | 303 | #else /* Limited by UINT_MAX of sectors */ |
| 304 | if (nblocks << (sbp->sb_blocklog - BBSHIFT) > UINT_MAX) |
Eric Sandeen | 657a4cf | 2010-04-30 03:42:49 +0000 | [diff] [blame] | 305 | return EFBIG; |
Nathan Scott | 4cc929e | 2007-05-14 18:24:02 +1000 | [diff] [blame] | 306 | #endif |
| 307 | return 0; |
| 308 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 309 | |
| 310 | /* |
| 311 | * Check the validity of the SB found. |
| 312 | */ |
| 313 | STATIC int |
| 314 | xfs_mount_validate_sb( |
| 315 | xfs_mount_t *mp, |
Nathan Scott | 764d1f8 | 2006-03-31 13:04:17 +1000 | [diff] [blame] | 316 | xfs_sb_t *sbp, |
Dave Chinner | 3451018 | 2013-05-27 16:38:19 +1000 | [diff] [blame] | 317 | bool check_inprogress, |
| 318 | bool check_version) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 319 | { |
Dave Chinner | af34e09 | 2011-03-07 10:04:35 +1100 | [diff] [blame] | 320 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 321 | /* |
| 322 | * If the log device and data device have the |
| 323 | * same device number, the log is internal. |
| 324 | * Consequently, the sb_logstart should be non-zero. If |
| 325 | * we have a zero sb_logstart in this case, we may be trying to mount |
| 326 | * a volume filesystem in a non-volume manner. |
| 327 | */ |
| 328 | if (sbp->sb_magicnum != XFS_SB_MAGIC) { |
Dave Chinner | 9802182 | 2012-11-12 22:54:03 +1100 | [diff] [blame] | 329 | xfs_warn(mp, "bad magic number"); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 330 | return XFS_ERROR(EWRONGFS); |
| 331 | } |
| 332 | |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 333 | |
Eric Sandeen | 6211870 | 2008-03-06 13:44:28 +1100 | [diff] [blame] | 334 | if (!xfs_sb_good_version(sbp)) { |
Dave Chinner | 9802182 | 2012-11-12 22:54:03 +1100 | [diff] [blame] | 335 | xfs_warn(mp, "bad version"); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 336 | return XFS_ERROR(EWRONGFS); |
| 337 | } |
| 338 | |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 339 | /* |
Dave Chinner | e721f50 | 2013-04-03 16:11:32 +1100 | [diff] [blame] | 340 | * Version 5 superblock feature mask validation. Reject combinations the |
Dave Chinner | 3451018 | 2013-05-27 16:38:19 +1000 | [diff] [blame] | 341 | * kernel cannot support up front before checking anything else. For |
| 342 | * write validation, we don't need to check feature masks. |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 343 | */ |
Dave Chinner | 3451018 | 2013-05-27 16:38:19 +1000 | [diff] [blame] | 344 | if (check_version && XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) { |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 345 | xfs_alert(mp, |
Dave Chinner | e721f50 | 2013-04-03 16:11:32 +1100 | [diff] [blame] | 346 | "Version 5 superblock detected. This kernel has EXPERIMENTAL support enabled!\n" |
| 347 | "Use of these features in this kernel is at your own risk!"); |
| 348 | |
| 349 | if (xfs_sb_has_compat_feature(sbp, |
| 350 | XFS_SB_FEAT_COMPAT_UNKNOWN)) { |
| 351 | xfs_warn(mp, |
| 352 | "Superblock has unknown compatible features (0x%x) enabled.\n" |
| 353 | "Using a more recent kernel is recommended.", |
| 354 | (sbp->sb_features_compat & |
| 355 | XFS_SB_FEAT_COMPAT_UNKNOWN)); |
| 356 | } |
| 357 | |
| 358 | if (xfs_sb_has_ro_compat_feature(sbp, |
| 359 | XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) { |
| 360 | xfs_alert(mp, |
| 361 | "Superblock has unknown read-only compatible features (0x%x) enabled.", |
| 362 | (sbp->sb_features_ro_compat & |
| 363 | XFS_SB_FEAT_RO_COMPAT_UNKNOWN)); |
| 364 | if (!(mp->m_flags & XFS_MOUNT_RDONLY)) { |
| 365 | xfs_warn(mp, |
| 366 | "Attempted to mount read-only compatible filesystem read-write.\n" |
| 367 | "Filesystem can only be safely mounted read only."); |
| 368 | return XFS_ERROR(EINVAL); |
| 369 | } |
| 370 | } |
| 371 | if (xfs_sb_has_incompat_feature(sbp, |
| 372 | XFS_SB_FEAT_INCOMPAT_UNKNOWN)) { |
| 373 | xfs_warn(mp, |
| 374 | "Superblock has unknown incompatible features (0x%x) enabled.\n" |
| 375 | "Filesystem can not be safely mounted by this kernel.", |
| 376 | (sbp->sb_features_incompat & |
| 377 | XFS_SB_FEAT_INCOMPAT_UNKNOWN)); |
| 378 | return XFS_ERROR(EINVAL); |
| 379 | } |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 380 | } |
| 381 | |
Chandra Seetharaman | d892d58 | 2013-07-19 17:36:02 -0500 | [diff] [blame] | 382 | if (xfs_sb_version_has_pquotino(sbp)) { |
| 383 | if (sbp->sb_qflags & (XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD)) { |
| 384 | xfs_notice(mp, |
| 385 | "Version 5 of Super block has XFS_OQUOTA bits.\n"); |
| 386 | return XFS_ERROR(EFSCORRUPTED); |
| 387 | } |
| 388 | } else if (sbp->sb_qflags & (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD | |
| 389 | XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) { |
| 390 | xfs_notice(mp, |
| 391 | "Superblock earlier than Version 5 has XFS_[PQ]UOTA_{ENFD|CHKD} bits.\n"); |
| 392 | return XFS_ERROR(EFSCORRUPTED); |
| 393 | } |
| 394 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 395 | if (unlikely( |
| 396 | sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) { |
Dave Chinner | 9802182 | 2012-11-12 22:54:03 +1100 | [diff] [blame] | 397 | xfs_warn(mp, |
Dave Chinner | af34e09 | 2011-03-07 10:04:35 +1100 | [diff] [blame] | 398 | "filesystem is marked as having an external log; " |
| 399 | "specify logdev on the mount command line."); |
Nathan Scott | 764d1f8 | 2006-03-31 13:04:17 +1000 | [diff] [blame] | 400 | return XFS_ERROR(EINVAL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 401 | } |
| 402 | |
| 403 | if (unlikely( |
| 404 | sbp->sb_logstart != 0 && mp->m_logdev_targp != mp->m_ddev_targp)) { |
Dave Chinner | 9802182 | 2012-11-12 22:54:03 +1100 | [diff] [blame] | 405 | xfs_warn(mp, |
Dave Chinner | af34e09 | 2011-03-07 10:04:35 +1100 | [diff] [blame] | 406 | "filesystem is marked as having an internal log; " |
| 407 | "do not specify logdev on the mount command line."); |
Nathan Scott | 764d1f8 | 2006-03-31 13:04:17 +1000 | [diff] [blame] | 408 | return XFS_ERROR(EINVAL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 409 | } |
| 410 | |
| 411 | /* |
Eric Sandeen | c0e090c | 2011-05-20 21:52:17 +0000 | [diff] [blame] | 412 | * More sanity checking. Most of these were stolen directly from |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 413 | * xfs_repair. |
| 414 | */ |
| 415 | if (unlikely( |
| 416 | sbp->sb_agcount <= 0 || |
| 417 | sbp->sb_sectsize < XFS_MIN_SECTORSIZE || |
| 418 | sbp->sb_sectsize > XFS_MAX_SECTORSIZE || |
| 419 | sbp->sb_sectlog < XFS_MIN_SECTORSIZE_LOG || |
| 420 | sbp->sb_sectlog > XFS_MAX_SECTORSIZE_LOG || |
Olaf Weber | 2ac00af | 2009-04-17 16:12:45 -0500 | [diff] [blame] | 421 | sbp->sb_sectsize != (1 << sbp->sb_sectlog) || |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 422 | sbp->sb_blocksize < XFS_MIN_BLOCKSIZE || |
| 423 | sbp->sb_blocksize > XFS_MAX_BLOCKSIZE || |
| 424 | sbp->sb_blocklog < XFS_MIN_BLOCKSIZE_LOG || |
| 425 | sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG || |
Olaf Weber | 2ac00af | 2009-04-17 16:12:45 -0500 | [diff] [blame] | 426 | sbp->sb_blocksize != (1 << sbp->sb_blocklog) || |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 427 | sbp->sb_inodesize < XFS_DINODE_MIN_SIZE || |
| 428 | sbp->sb_inodesize > XFS_DINODE_MAX_SIZE || |
Nathan Scott | 9f989c9 | 2006-03-14 13:29:32 +1100 | [diff] [blame] | 429 | sbp->sb_inodelog < XFS_DINODE_MIN_LOG || |
| 430 | sbp->sb_inodelog > XFS_DINODE_MAX_LOG || |
Olaf Weber | 2ac00af | 2009-04-17 16:12:45 -0500 | [diff] [blame] | 431 | sbp->sb_inodesize != (1 << sbp->sb_inodelog) || |
Nathan Scott | 9f989c9 | 2006-03-14 13:29:32 +1100 | [diff] [blame] | 432 | (sbp->sb_blocklog - sbp->sb_inodelog != sbp->sb_inopblog) || |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 433 | (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE) || |
| 434 | (sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE) || |
Eric Sandeen | c0e090c | 2011-05-20 21:52:17 +0000 | [diff] [blame] | 435 | (sbp->sb_imax_pct > 100 /* zero sb_imax_pct is valid */) || |
| 436 | sbp->sb_dblocks == 0 || |
| 437 | sbp->sb_dblocks > XFS_MAX_DBLOCKS(sbp) || |
| 438 | sbp->sb_dblocks < XFS_MIN_DBLOCKS(sbp))) { |
Dave Chinner | 9802182 | 2012-11-12 22:54:03 +1100 | [diff] [blame] | 439 | XFS_CORRUPTION_ERROR("SB sanity check failed", |
Eric Sandeen | c0e090c | 2011-05-20 21:52:17 +0000 | [diff] [blame] | 440 | XFS_ERRLEVEL_LOW, mp, sbp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 441 | return XFS_ERROR(EFSCORRUPTED); |
| 442 | } |
| 443 | |
Lachlan McIlroy | 2edbddd | 2008-06-27 13:34:34 +1000 | [diff] [blame] | 444 | /* |
| 445 | * Until this is fixed only page-sized or smaller data blocks work. |
| 446 | */ |
| 447 | if (unlikely(sbp->sb_blocksize > PAGE_SIZE)) { |
Dave Chinner | 9802182 | 2012-11-12 22:54:03 +1100 | [diff] [blame] | 448 | xfs_warn(mp, |
Dave Chinner | af34e09 | 2011-03-07 10:04:35 +1100 | [diff] [blame] | 449 | "File system with blocksize %d bytes. " |
| 450 | "Only pagesize (%ld) or less will currently work.", |
| 451 | sbp->sb_blocksize, PAGE_SIZE); |
Lachlan McIlroy | 2edbddd | 2008-06-27 13:34:34 +1000 | [diff] [blame] | 452 | return XFS_ERROR(ENOSYS); |
| 453 | } |
| 454 | |
Christoph Hellwig | 1a5902c | 2009-03-29 19:26:46 +0200 | [diff] [blame] | 455 | /* |
| 456 | * Currently only very few inode sizes are supported. |
| 457 | */ |
| 458 | switch (sbp->sb_inodesize) { |
| 459 | case 256: |
| 460 | case 512: |
| 461 | case 1024: |
| 462 | case 2048: |
| 463 | break; |
| 464 | default: |
Dave Chinner | 9802182 | 2012-11-12 22:54:03 +1100 | [diff] [blame] | 465 | xfs_warn(mp, "inode size of %d bytes not supported", |
Dave Chinner | af34e09 | 2011-03-07 10:04:35 +1100 | [diff] [blame] | 466 | sbp->sb_inodesize); |
Christoph Hellwig | 1a5902c | 2009-03-29 19:26:46 +0200 | [diff] [blame] | 467 | return XFS_ERROR(ENOSYS); |
| 468 | } |
| 469 | |
Nathan Scott | 4cc929e | 2007-05-14 18:24:02 +1000 | [diff] [blame] | 470 | if (xfs_sb_validate_fsb_count(sbp, sbp->sb_dblocks) || |
| 471 | xfs_sb_validate_fsb_count(sbp, sbp->sb_rblocks)) { |
Dave Chinner | 9802182 | 2012-11-12 22:54:03 +1100 | [diff] [blame] | 472 | xfs_warn(mp, |
Dave Chinner | af34e09 | 2011-03-07 10:04:35 +1100 | [diff] [blame] | 473 | "file system too large to be mounted on this system."); |
Eric Sandeen | 657a4cf | 2010-04-30 03:42:49 +0000 | [diff] [blame] | 474 | return XFS_ERROR(EFBIG); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 475 | } |
| 476 | |
Dave Chinner | 9802182 | 2012-11-12 22:54:03 +1100 | [diff] [blame] | 477 | if (check_inprogress && sbp->sb_inprogress) { |
| 478 | xfs_warn(mp, "Offline file system operation in progress!"); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 479 | return XFS_ERROR(EFSCORRUPTED); |
| 480 | } |
| 481 | |
| 482 | /* |
Nathan Scott | de20614 | 2005-05-05 13:24:13 -0700 | [diff] [blame] | 483 | * Version 1 directory format has never worked on Linux. |
| 484 | */ |
Eric Sandeen | 6211870 | 2008-03-06 13:44:28 +1100 | [diff] [blame] | 485 | if (unlikely(!xfs_sb_version_hasdirv2(sbp))) { |
Dave Chinner | 9802182 | 2012-11-12 22:54:03 +1100 | [diff] [blame] | 486 | xfs_warn(mp, "file system using version 1 directory format"); |
Nathan Scott | de20614 | 2005-05-05 13:24:13 -0700 | [diff] [blame] | 487 | return XFS_ERROR(ENOSYS); |
| 488 | } |
| 489 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 490 | return 0; |
| 491 | } |
| 492 | |
Dave Chinner | 1c1c6eb | 2010-01-11 11:47:44 +0000 | [diff] [blame] | 493 | int |
Nathan Scott | c11e2c3 | 2005-11-02 15:11:45 +1100 | [diff] [blame] | 494 | xfs_initialize_perag( |
Nathan Scott | c11e2c3 | 2005-11-02 15:11:45 +1100 | [diff] [blame] | 495 | xfs_mount_t *mp, |
Dave Chinner | 1c1c6eb | 2010-01-11 11:47:44 +0000 | [diff] [blame] | 496 | xfs_agnumber_t agcount, |
| 497 | xfs_agnumber_t *maxagi) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 498 | { |
Carlos Maiolino | 2d2194f | 2012-09-20 10:32:38 -0300 | [diff] [blame] | 499 | xfs_agnumber_t index; |
Dave Chinner | 8b26c58 | 2010-01-11 11:47:48 +0000 | [diff] [blame] | 500 | xfs_agnumber_t first_initialised = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 501 | xfs_perag_t *pag; |
| 502 | xfs_agino_t agino; |
| 503 | xfs_ino_t ino; |
| 504 | xfs_sb_t *sbp = &mp->m_sb; |
Dave Chinner | 8b26c58 | 2010-01-11 11:47:48 +0000 | [diff] [blame] | 505 | int error = -ENOMEM; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 506 | |
Dave Chinner | 1c1c6eb | 2010-01-11 11:47:44 +0000 | [diff] [blame] | 507 | /* |
| 508 | * Walk the current per-ag tree so we don't try to initialise AGs |
| 509 | * that already exist (growfs case). Allocate and insert all the |
| 510 | * AGs we don't find ready for initialisation. |
| 511 | */ |
| 512 | for (index = 0; index < agcount; index++) { |
| 513 | pag = xfs_perag_get(mp, index); |
| 514 | if (pag) { |
| 515 | xfs_perag_put(pag); |
| 516 | continue; |
| 517 | } |
Dave Chinner | 8b26c58 | 2010-01-11 11:47:48 +0000 | [diff] [blame] | 518 | if (!first_initialised) |
| 519 | first_initialised = index; |
Christoph Hellwig | fb3b504 | 2010-05-28 19:03:10 +0000 | [diff] [blame] | 520 | |
Dave Chinner | 1c1c6eb | 2010-01-11 11:47:44 +0000 | [diff] [blame] | 521 | pag = kmem_zalloc(sizeof(*pag), KM_MAYFAIL); |
| 522 | if (!pag) |
Dave Chinner | 8b26c58 | 2010-01-11 11:47:48 +0000 | [diff] [blame] | 523 | goto out_unwind; |
Christoph Hellwig | fb3b504 | 2010-05-28 19:03:10 +0000 | [diff] [blame] | 524 | pag->pag_agno = index; |
| 525 | pag->pag_mount = mp; |
Dave Chinner | 1a427ab | 2010-12-16 17:08:41 +1100 | [diff] [blame] | 526 | spin_lock_init(&pag->pag_ici_lock); |
Dave Chinner | 69b491c | 2010-09-27 11:09:51 +1000 | [diff] [blame] | 527 | mutex_init(&pag->pag_ici_reclaim_lock); |
Christoph Hellwig | fb3b504 | 2010-05-28 19:03:10 +0000 | [diff] [blame] | 528 | INIT_RADIX_TREE(&pag->pag_ici_root, GFP_ATOMIC); |
Dave Chinner | 74f75a0 | 2010-09-24 19:59:04 +1000 | [diff] [blame] | 529 | spin_lock_init(&pag->pag_buf_lock); |
| 530 | pag->pag_buf_tree = RB_ROOT; |
Christoph Hellwig | fb3b504 | 2010-05-28 19:03:10 +0000 | [diff] [blame] | 531 | |
Dave Chinner | 1c1c6eb | 2010-01-11 11:47:44 +0000 | [diff] [blame] | 532 | if (radix_tree_preload(GFP_NOFS)) |
Dave Chinner | 8b26c58 | 2010-01-11 11:47:48 +0000 | [diff] [blame] | 533 | goto out_unwind; |
Christoph Hellwig | fb3b504 | 2010-05-28 19:03:10 +0000 | [diff] [blame] | 534 | |
Dave Chinner | 1c1c6eb | 2010-01-11 11:47:44 +0000 | [diff] [blame] | 535 | spin_lock(&mp->m_perag_lock); |
| 536 | if (radix_tree_insert(&mp->m_perag_tree, index, pag)) { |
| 537 | BUG(); |
| 538 | spin_unlock(&mp->m_perag_lock); |
Dave Chinner | 8b26c58 | 2010-01-11 11:47:48 +0000 | [diff] [blame] | 539 | radix_tree_preload_end(); |
| 540 | error = -EEXIST; |
| 541 | goto out_unwind; |
Dave Chinner | 1c1c6eb | 2010-01-11 11:47:44 +0000 | [diff] [blame] | 542 | } |
| 543 | spin_unlock(&mp->m_perag_lock); |
| 544 | radix_tree_preload_end(); |
| 545 | } |
| 546 | |
Christoph Hellwig | fb3b504 | 2010-05-28 19:03:10 +0000 | [diff] [blame] | 547 | /* |
| 548 | * If we mount with the inode64 option, or no inode overflows |
| 549 | * the legacy 32-bit address space clear the inode32 option. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 550 | */ |
Christoph Hellwig | fb3b504 | 2010-05-28 19:03:10 +0000 | [diff] [blame] | 551 | agino = XFS_OFFBNO_TO_AGINO(mp, sbp->sb_agblocks - 1, 0); |
| 552 | ino = XFS_AGINO_TO_INO(mp, agcount - 1, agino); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 553 | |
Christoph Hellwig | fb3b504 | 2010-05-28 19:03:10 +0000 | [diff] [blame] | 554 | if ((mp->m_flags & XFS_MOUNT_SMALL_INUMS) && ino > XFS_MAXINUMBER_32) |
| 555 | mp->m_flags |= XFS_MOUNT_32BITINODES; |
| 556 | else |
| 557 | mp->m_flags &= ~XFS_MOUNT_32BITINODES; |
| 558 | |
Carlos Maiolino | 2d2194f | 2012-09-20 10:32:38 -0300 | [diff] [blame] | 559 | if (mp->m_flags & XFS_MOUNT_32BITINODES) |
| 560 | index = xfs_set_inode32(mp); |
| 561 | else |
| 562 | index = xfs_set_inode64(mp); |
Christoph Hellwig | fb3b504 | 2010-05-28 19:03:10 +0000 | [diff] [blame] | 563 | |
Dave Chinner | 1c1c6eb | 2010-01-11 11:47:44 +0000 | [diff] [blame] | 564 | if (maxagi) |
| 565 | *maxagi = index; |
| 566 | return 0; |
Dave Chinner | 8b26c58 | 2010-01-11 11:47:48 +0000 | [diff] [blame] | 567 | |
| 568 | out_unwind: |
| 569 | kmem_free(pag); |
| 570 | for (; index > first_initialised; index--) { |
| 571 | pag = radix_tree_delete(&mp->m_perag_tree, index); |
| 572 | kmem_free(pag); |
| 573 | } |
| 574 | return error; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 575 | } |
| 576 | |
Chandra Seetharaman | 83e782e | 2013-06-27 17:25:10 -0500 | [diff] [blame] | 577 | static void |
| 578 | xfs_sb_quota_from_disk(struct xfs_sb *sbp) |
| 579 | { |
Chandra Seetharaman | 0102629 | 2013-07-19 17:32:55 -0500 | [diff] [blame] | 580 | /* |
| 581 | * older mkfs doesn't initialize quota inodes to NULLFSINO. This |
| 582 | * leads to in-core values having two different values for a quota |
| 583 | * inode to be invalid: 0 and NULLFSINO. Change it to a single value |
| 584 | * NULLFSINO. |
| 585 | * |
| 586 | * Note that this change affect only the in-core values. These |
| 587 | * values are not written back to disk unless any quota information |
| 588 | * is written to the disk. Even in that case, sb_pquotino field is |
| 589 | * not written to disk unless the superblock supports pquotino. |
| 590 | */ |
| 591 | if (sbp->sb_uquotino == 0) |
| 592 | sbp->sb_uquotino = NULLFSINO; |
| 593 | if (sbp->sb_gquotino == 0) |
| 594 | sbp->sb_gquotino = NULLFSINO; |
| 595 | if (sbp->sb_pquotino == 0) |
| 596 | sbp->sb_pquotino = NULLFSINO; |
| 597 | |
Chandra Seetharaman | d892d58 | 2013-07-19 17:36:02 -0500 | [diff] [blame] | 598 | /* |
| 599 | * We need to do these manipilations only if we are working |
| 600 | * with an older version of on-disk superblock. |
| 601 | */ |
| 602 | if (xfs_sb_version_has_pquotino(sbp)) |
| 603 | return; |
| 604 | |
Chandra Seetharaman | 83e782e | 2013-06-27 17:25:10 -0500 | [diff] [blame] | 605 | if (sbp->sb_qflags & XFS_OQUOTA_ENFD) |
| 606 | sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ? |
| 607 | XFS_PQUOTA_ENFD : XFS_GQUOTA_ENFD; |
| 608 | if (sbp->sb_qflags & XFS_OQUOTA_CHKD) |
| 609 | sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ? |
| 610 | XFS_PQUOTA_CHKD : XFS_GQUOTA_CHKD; |
| 611 | sbp->sb_qflags &= ~(XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD); |
Chandra Seetharaman | d892d58 | 2013-07-19 17:36:02 -0500 | [diff] [blame] | 612 | |
| 613 | if (sbp->sb_qflags & XFS_PQUOTA_ACCT) { |
| 614 | /* |
| 615 | * In older version of superblock, on-disk superblock only |
| 616 | * has sb_gquotino, and in-core superblock has both sb_gquotino |
| 617 | * and sb_pquotino. But, only one of them is supported at any |
| 618 | * point of time. So, if PQUOTA is set in disk superblock, |
| 619 | * copy over sb_gquotino to sb_pquotino. |
| 620 | */ |
| 621 | sbp->sb_pquotino = sbp->sb_gquotino; |
| 622 | sbp->sb_gquotino = NULLFSINO; |
| 623 | } |
Chandra Seetharaman | 83e782e | 2013-06-27 17:25:10 -0500 | [diff] [blame] | 624 | } |
| 625 | |
Christoph Hellwig | 2bdf7cd | 2007-08-28 13:58:06 +1000 | [diff] [blame] | 626 | void |
| 627 | xfs_sb_from_disk( |
Dave Chinner | 9802182 | 2012-11-12 22:54:03 +1100 | [diff] [blame] | 628 | struct xfs_sb *to, |
Christoph Hellwig | 2bdf7cd | 2007-08-28 13:58:06 +1000 | [diff] [blame] | 629 | xfs_dsb_t *from) |
| 630 | { |
| 631 | to->sb_magicnum = be32_to_cpu(from->sb_magicnum); |
| 632 | to->sb_blocksize = be32_to_cpu(from->sb_blocksize); |
| 633 | to->sb_dblocks = be64_to_cpu(from->sb_dblocks); |
| 634 | to->sb_rblocks = be64_to_cpu(from->sb_rblocks); |
| 635 | to->sb_rextents = be64_to_cpu(from->sb_rextents); |
| 636 | memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid)); |
| 637 | to->sb_logstart = be64_to_cpu(from->sb_logstart); |
| 638 | to->sb_rootino = be64_to_cpu(from->sb_rootino); |
| 639 | to->sb_rbmino = be64_to_cpu(from->sb_rbmino); |
| 640 | to->sb_rsumino = be64_to_cpu(from->sb_rsumino); |
| 641 | to->sb_rextsize = be32_to_cpu(from->sb_rextsize); |
| 642 | to->sb_agblocks = be32_to_cpu(from->sb_agblocks); |
| 643 | to->sb_agcount = be32_to_cpu(from->sb_agcount); |
| 644 | to->sb_rbmblocks = be32_to_cpu(from->sb_rbmblocks); |
| 645 | to->sb_logblocks = be32_to_cpu(from->sb_logblocks); |
| 646 | to->sb_versionnum = be16_to_cpu(from->sb_versionnum); |
| 647 | to->sb_sectsize = be16_to_cpu(from->sb_sectsize); |
| 648 | to->sb_inodesize = be16_to_cpu(from->sb_inodesize); |
| 649 | to->sb_inopblock = be16_to_cpu(from->sb_inopblock); |
| 650 | memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname)); |
| 651 | to->sb_blocklog = from->sb_blocklog; |
| 652 | to->sb_sectlog = from->sb_sectlog; |
| 653 | to->sb_inodelog = from->sb_inodelog; |
| 654 | to->sb_inopblog = from->sb_inopblog; |
| 655 | to->sb_agblklog = from->sb_agblklog; |
| 656 | to->sb_rextslog = from->sb_rextslog; |
| 657 | to->sb_inprogress = from->sb_inprogress; |
| 658 | to->sb_imax_pct = from->sb_imax_pct; |
| 659 | to->sb_icount = be64_to_cpu(from->sb_icount); |
| 660 | to->sb_ifree = be64_to_cpu(from->sb_ifree); |
| 661 | to->sb_fdblocks = be64_to_cpu(from->sb_fdblocks); |
| 662 | to->sb_frextents = be64_to_cpu(from->sb_frextents); |
| 663 | to->sb_uquotino = be64_to_cpu(from->sb_uquotino); |
| 664 | to->sb_gquotino = be64_to_cpu(from->sb_gquotino); |
| 665 | to->sb_qflags = be16_to_cpu(from->sb_qflags); |
| 666 | to->sb_flags = from->sb_flags; |
| 667 | to->sb_shared_vn = from->sb_shared_vn; |
| 668 | to->sb_inoalignmt = be32_to_cpu(from->sb_inoalignmt); |
| 669 | to->sb_unit = be32_to_cpu(from->sb_unit); |
| 670 | to->sb_width = be32_to_cpu(from->sb_width); |
| 671 | to->sb_dirblklog = from->sb_dirblklog; |
| 672 | to->sb_logsectlog = from->sb_logsectlog; |
| 673 | to->sb_logsectsize = be16_to_cpu(from->sb_logsectsize); |
| 674 | to->sb_logsunit = be32_to_cpu(from->sb_logsunit); |
| 675 | to->sb_features2 = be32_to_cpu(from->sb_features2); |
David Chinner | ee1c090 | 2008-03-06 13:45:50 +1100 | [diff] [blame] | 676 | to->sb_bad_features2 = be32_to_cpu(from->sb_bad_features2); |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 677 | to->sb_features_compat = be32_to_cpu(from->sb_features_compat); |
| 678 | to->sb_features_ro_compat = be32_to_cpu(from->sb_features_ro_compat); |
| 679 | to->sb_features_incompat = be32_to_cpu(from->sb_features_incompat); |
Dave Chinner | e721f50 | 2013-04-03 16:11:32 +1100 | [diff] [blame] | 680 | to->sb_features_log_incompat = |
| 681 | be32_to_cpu(from->sb_features_log_incompat); |
| 682 | to->sb_pad = 0; |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 683 | to->sb_pquotino = be64_to_cpu(from->sb_pquotino); |
| 684 | to->sb_lsn = be64_to_cpu(from->sb_lsn); |
Christoph Hellwig | 2bdf7cd | 2007-08-28 13:58:06 +1000 | [diff] [blame] | 685 | } |
| 686 | |
Chandra Seetharaman | 83e782e | 2013-06-27 17:25:10 -0500 | [diff] [blame] | 687 | static inline void |
| 688 | xfs_sb_quota_to_disk( |
| 689 | xfs_dsb_t *to, |
| 690 | xfs_sb_t *from, |
| 691 | __int64_t *fields) |
| 692 | { |
| 693 | __uint16_t qflags = from->sb_qflags; |
| 694 | |
Chandra Seetharaman | d892d58 | 2013-07-19 17:36:02 -0500 | [diff] [blame] | 695 | /* |
| 696 | * We need to do these manipilations only if we are working |
| 697 | * with an older version of on-disk superblock. |
| 698 | */ |
| 699 | if (xfs_sb_version_has_pquotino(from)) |
| 700 | return; |
| 701 | |
Chandra Seetharaman | 83e782e | 2013-06-27 17:25:10 -0500 | [diff] [blame] | 702 | if (*fields & XFS_SB_QFLAGS) { |
| 703 | /* |
| 704 | * The in-core version of sb_qflags do not have |
| 705 | * XFS_OQUOTA_* flags, whereas the on-disk version |
| 706 | * does. So, convert incore XFS_{PG}QUOTA_* flags |
| 707 | * to on-disk XFS_OQUOTA_* flags. |
| 708 | */ |
| 709 | qflags &= ~(XFS_PQUOTA_ENFD | XFS_PQUOTA_CHKD | |
| 710 | XFS_GQUOTA_ENFD | XFS_GQUOTA_CHKD); |
| 711 | |
| 712 | if (from->sb_qflags & |
| 713 | (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD)) |
| 714 | qflags |= XFS_OQUOTA_ENFD; |
| 715 | if (from->sb_qflags & |
| 716 | (XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) |
| 717 | qflags |= XFS_OQUOTA_CHKD; |
| 718 | to->sb_qflags = cpu_to_be16(qflags); |
| 719 | *fields &= ~XFS_SB_QFLAGS; |
| 720 | } |
Chandra Seetharaman | d892d58 | 2013-07-19 17:36:02 -0500 | [diff] [blame] | 721 | |
| 722 | /* |
| 723 | * GQUOTINO and PQUOTINO cannot be used together in versions |
| 724 | * of superblock that do not have pquotino. from->sb_flags |
| 725 | * tells us which quota is active and should be copied to |
| 726 | * disk. |
| 727 | */ |
| 728 | if ((*fields & XFS_SB_GQUOTINO) && |
| 729 | (from->sb_qflags & XFS_GQUOTA_ACCT)) |
| 730 | to->sb_gquotino = cpu_to_be64(from->sb_gquotino); |
| 731 | else if ((*fields & XFS_SB_PQUOTINO) && |
| 732 | (from->sb_qflags & XFS_PQUOTA_ACCT)) |
| 733 | to->sb_gquotino = cpu_to_be64(from->sb_pquotino); |
| 734 | |
| 735 | *fields &= ~(XFS_SB_PQUOTINO | XFS_SB_GQUOTINO); |
Chandra Seetharaman | 83e782e | 2013-06-27 17:25:10 -0500 | [diff] [blame] | 736 | } |
| 737 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 738 | /* |
Christoph Hellwig | 2bdf7cd | 2007-08-28 13:58:06 +1000 | [diff] [blame] | 739 | * Copy in core superblock to ondisk one. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 740 | * |
Christoph Hellwig | 2bdf7cd | 2007-08-28 13:58:06 +1000 | [diff] [blame] | 741 | * The fields argument is mask of superblock fields to copy. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 742 | */ |
| 743 | void |
Christoph Hellwig | 2bdf7cd | 2007-08-28 13:58:06 +1000 | [diff] [blame] | 744 | xfs_sb_to_disk( |
| 745 | xfs_dsb_t *to, |
| 746 | xfs_sb_t *from, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 747 | __int64_t fields) |
| 748 | { |
Christoph Hellwig | 2bdf7cd | 2007-08-28 13:58:06 +1000 | [diff] [blame] | 749 | xfs_caddr_t to_ptr = (xfs_caddr_t)to; |
| 750 | xfs_caddr_t from_ptr = (xfs_caddr_t)from; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 751 | xfs_sb_field_t f; |
| 752 | int first; |
| 753 | int size; |
| 754 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 755 | ASSERT(fields); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 756 | if (!fields) |
| 757 | return; |
| 758 | |
Chandra Seetharaman | 83e782e | 2013-06-27 17:25:10 -0500 | [diff] [blame] | 759 | xfs_sb_quota_to_disk(to, from, &fields); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 760 | while (fields) { |
| 761 | f = (xfs_sb_field_t)xfs_lowbit64((__uint64_t)fields); |
| 762 | first = xfs_sb_info[f].offset; |
| 763 | size = xfs_sb_info[f + 1].offset - first; |
| 764 | |
| 765 | ASSERT(xfs_sb_info[f].type == 0 || xfs_sb_info[f].type == 1); |
| 766 | |
| 767 | if (size == 1 || xfs_sb_info[f].type == 1) { |
Christoph Hellwig | 2bdf7cd | 2007-08-28 13:58:06 +1000 | [diff] [blame] | 768 | memcpy(to_ptr + first, from_ptr + first, size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 769 | } else { |
| 770 | switch (size) { |
| 771 | case 2: |
Christoph Hellwig | 2bdf7cd | 2007-08-28 13:58:06 +1000 | [diff] [blame] | 772 | *(__be16 *)(to_ptr + first) = |
| 773 | cpu_to_be16(*(__u16 *)(from_ptr + first)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 774 | break; |
| 775 | case 4: |
Christoph Hellwig | 2bdf7cd | 2007-08-28 13:58:06 +1000 | [diff] [blame] | 776 | *(__be32 *)(to_ptr + first) = |
| 777 | cpu_to_be32(*(__u32 *)(from_ptr + first)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 778 | break; |
| 779 | case 8: |
Christoph Hellwig | 2bdf7cd | 2007-08-28 13:58:06 +1000 | [diff] [blame] | 780 | *(__be64 *)(to_ptr + first) = |
| 781 | cpu_to_be64(*(__u64 *)(from_ptr + first)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 782 | break; |
| 783 | default: |
| 784 | ASSERT(0); |
| 785 | } |
| 786 | } |
| 787 | |
| 788 | fields &= ~(1LL << f); |
| 789 | } |
| 790 | } |
| 791 | |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 792 | static int |
Dave Chinner | 612cfbf | 2012-11-14 17:52:32 +1100 | [diff] [blame] | 793 | xfs_sb_verify( |
Dave Chinner | 3451018 | 2013-05-27 16:38:19 +1000 | [diff] [blame] | 794 | struct xfs_buf *bp, |
| 795 | bool check_version) |
Dave Chinner | 9802182 | 2012-11-12 22:54:03 +1100 | [diff] [blame] | 796 | { |
| 797 | struct xfs_mount *mp = bp->b_target->bt_mount; |
| 798 | struct xfs_sb sb; |
Dave Chinner | 9802182 | 2012-11-12 22:54:03 +1100 | [diff] [blame] | 799 | |
| 800 | xfs_sb_from_disk(&sb, XFS_BUF_TO_SBP(bp)); |
| 801 | |
| 802 | /* |
| 803 | * Only check the in progress field for the primary superblock as |
| 804 | * mkfs.xfs doesn't clear it from secondary superblocks. |
| 805 | */ |
Dave Chinner | 3451018 | 2013-05-27 16:38:19 +1000 | [diff] [blame] | 806 | return xfs_mount_validate_sb(mp, &sb, bp->b_bn == XFS_SB_DADDR, |
| 807 | check_version); |
Dave Chinner | 612cfbf | 2012-11-14 17:52:32 +1100 | [diff] [blame] | 808 | } |
| 809 | |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 810 | /* |
| 811 | * If the superblock has the CRC feature bit set or the CRC field is non-null, |
| 812 | * check that the CRC is valid. We check the CRC field is non-null because a |
| 813 | * single bit error could clear the feature bit and unused parts of the |
| 814 | * superblock are supposed to be zero. Hence a non-null crc field indicates that |
| 815 | * we've potentially lost a feature bit and we should check it anyway. |
| 816 | */ |
Dave Chinner | 1813dd6 | 2012-11-14 17:54:40 +1100 | [diff] [blame] | 817 | static void |
Dave Chinner | 612cfbf | 2012-11-14 17:52:32 +1100 | [diff] [blame] | 818 | xfs_sb_read_verify( |
| 819 | struct xfs_buf *bp) |
| 820 | { |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 821 | struct xfs_mount *mp = bp->b_target->bt_mount; |
| 822 | struct xfs_dsb *dsb = XFS_BUF_TO_SBP(bp); |
| 823 | int error; |
| 824 | |
| 825 | /* |
| 826 | * open code the version check to avoid needing to convert the entire |
| 827 | * superblock from disk order just to check the version number |
| 828 | */ |
| 829 | if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC) && |
| 830 | (((be16_to_cpu(dsb->sb_versionnum) & XFS_SB_VERSION_NUMBITS) == |
| 831 | XFS_SB_VERSION_5) || |
| 832 | dsb->sb_crc != 0)) { |
| 833 | |
| 834 | if (!xfs_verify_cksum(bp->b_addr, be16_to_cpu(dsb->sb_sectsize), |
| 835 | offsetof(struct xfs_sb, sb_crc))) { |
| 836 | error = EFSCORRUPTED; |
| 837 | goto out_error; |
| 838 | } |
| 839 | } |
Dave Chinner | 3451018 | 2013-05-27 16:38:19 +1000 | [diff] [blame] | 840 | error = xfs_sb_verify(bp, true); |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 841 | |
| 842 | out_error: |
| 843 | if (error) { |
| 844 | XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, bp->b_addr); |
| 845 | xfs_buf_ioerror(bp, error); |
| 846 | } |
Dave Chinner | 9802182 | 2012-11-12 22:54:03 +1100 | [diff] [blame] | 847 | } |
| 848 | |
| 849 | /* |
| 850 | * We may be probed for a filesystem match, so we may not want to emit |
| 851 | * messages when the superblock buffer is not actually an XFS superblock. |
| 852 | * If we find an XFS superblock, the run a normal, noisy mount because we are |
| 853 | * really going to mount it and want to know about errors. |
| 854 | */ |
Dave Chinner | 1813dd6 | 2012-11-14 17:54:40 +1100 | [diff] [blame] | 855 | static void |
Dave Chinner | 9802182 | 2012-11-12 22:54:03 +1100 | [diff] [blame] | 856 | xfs_sb_quiet_read_verify( |
| 857 | struct xfs_buf *bp) |
| 858 | { |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 859 | struct xfs_dsb *dsb = XFS_BUF_TO_SBP(bp); |
Dave Chinner | 9802182 | 2012-11-12 22:54:03 +1100 | [diff] [blame] | 860 | |
Dave Chinner | 9802182 | 2012-11-12 22:54:03 +1100 | [diff] [blame] | 861 | |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 862 | if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC)) { |
Dave Chinner | 9802182 | 2012-11-12 22:54:03 +1100 | [diff] [blame] | 863 | /* XFS filesystem, verify noisily! */ |
| 864 | xfs_sb_read_verify(bp); |
| 865 | return; |
| 866 | } |
| 867 | /* quietly fail */ |
Eric Sandeen | aeb4f20 | 2013-01-16 17:33:53 -0600 | [diff] [blame] | 868 | xfs_buf_ioerror(bp, EWRONGFS); |
Dave Chinner | 9802182 | 2012-11-12 22:54:03 +1100 | [diff] [blame] | 869 | } |
| 870 | |
Dave Chinner | 1813dd6 | 2012-11-14 17:54:40 +1100 | [diff] [blame] | 871 | static void |
| 872 | xfs_sb_write_verify( |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 873 | struct xfs_buf *bp) |
Dave Chinner | 1813dd6 | 2012-11-14 17:54:40 +1100 | [diff] [blame] | 874 | { |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 875 | struct xfs_mount *mp = bp->b_target->bt_mount; |
| 876 | struct xfs_buf_log_item *bip = bp->b_fspriv; |
| 877 | int error; |
| 878 | |
Dave Chinner | 3451018 | 2013-05-27 16:38:19 +1000 | [diff] [blame] | 879 | error = xfs_sb_verify(bp, false); |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 880 | if (error) { |
| 881 | XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, bp->b_addr); |
| 882 | xfs_buf_ioerror(bp, error); |
| 883 | return; |
| 884 | } |
| 885 | |
| 886 | if (!xfs_sb_version_hascrc(&mp->m_sb)) |
| 887 | return; |
| 888 | |
| 889 | if (bip) |
| 890 | XFS_BUF_TO_SBP(bp)->sb_lsn = cpu_to_be64(bip->bli_item.li_lsn); |
| 891 | |
| 892 | xfs_update_cksum(bp->b_addr, BBTOB(bp->b_length), |
| 893 | offsetof(struct xfs_sb, sb_crc)); |
Dave Chinner | 1813dd6 | 2012-11-14 17:54:40 +1100 | [diff] [blame] | 894 | } |
| 895 | |
| 896 | const struct xfs_buf_ops xfs_sb_buf_ops = { |
| 897 | .verify_read = xfs_sb_read_verify, |
| 898 | .verify_write = xfs_sb_write_verify, |
| 899 | }; |
| 900 | |
| 901 | static const struct xfs_buf_ops xfs_sb_quiet_buf_ops = { |
| 902 | .verify_read = xfs_sb_quiet_read_verify, |
| 903 | .verify_write = xfs_sb_write_verify, |
| 904 | }; |
| 905 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 906 | /* |
| 907 | * xfs_readsb |
| 908 | * |
| 909 | * Does the initial read of the superblock. |
| 910 | */ |
| 911 | int |
Nathan Scott | 764d1f8 | 2006-03-31 13:04:17 +1000 | [diff] [blame] | 912 | xfs_readsb(xfs_mount_t *mp, int flags) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 913 | { |
| 914 | unsigned int sector_size; |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 915 | struct xfs_buf *bp; |
| 916 | struct xfs_sb *sbp = &mp->m_sb; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 917 | int error; |
Dave Chinner | af34e09 | 2011-03-07 10:04:35 +1100 | [diff] [blame] | 918 | int loud = !(flags & XFS_MFSI_QUIET); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 919 | |
| 920 | ASSERT(mp->m_sb_bp == NULL); |
| 921 | ASSERT(mp->m_ddev_targp != NULL); |
| 922 | |
| 923 | /* |
| 924 | * Allocate a (locked) buffer to hold the superblock. |
| 925 | * This will be kept around at all times to optimize |
| 926 | * access to the superblock. |
| 927 | */ |
| 928 | sector_size = xfs_getsize_buftarg(mp->m_ddev_targp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 929 | |
Dave Chinner | 26af655 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 930 | reread: |
Dave Chinner | e70b73f | 2012-04-23 15:58:49 +1000 | [diff] [blame] | 931 | bp = xfs_buf_read_uncached(mp->m_ddev_targp, XFS_SB_DADDR, |
Dave Chinner | 9802182 | 2012-11-12 22:54:03 +1100 | [diff] [blame] | 932 | BTOBB(sector_size), 0, |
Dave Chinner | 1813dd6 | 2012-11-14 17:54:40 +1100 | [diff] [blame] | 933 | loud ? &xfs_sb_buf_ops |
| 934 | : &xfs_sb_quiet_buf_ops); |
Dave Chinner | 26af655 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 935 | if (!bp) { |
Dave Chinner | af34e09 | 2011-03-07 10:04:35 +1100 | [diff] [blame] | 936 | if (loud) |
| 937 | xfs_warn(mp, "SB buffer read failed"); |
Dave Chinner | 26af655 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 938 | return EIO; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 939 | } |
Dave Chinner | eab4e63 | 2012-11-12 22:54:02 +1100 | [diff] [blame] | 940 | if (bp->b_error) { |
| 941 | error = bp->b_error; |
| 942 | if (loud) |
Dave Chinner | e721f50 | 2013-04-03 16:11:32 +1100 | [diff] [blame] | 943 | xfs_warn(mp, "SB validate failed with error %d.", error); |
Dave Chinner | eab4e63 | 2012-11-12 22:54:02 +1100 | [diff] [blame] | 944 | goto release_buf; |
| 945 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 946 | |
| 947 | /* |
| 948 | * Initialize the mount structure from the superblock. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 949 | */ |
Dave Chinner | 9802182 | 2012-11-12 22:54:03 +1100 | [diff] [blame] | 950 | xfs_sb_from_disk(&mp->m_sb, XFS_BUF_TO_SBP(bp)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 951 | |
Chandra Seetharaman | 83e782e | 2013-06-27 17:25:10 -0500 | [diff] [blame] | 952 | xfs_sb_quota_from_disk(&mp->m_sb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 953 | /* |
| 954 | * We must be able to do sector-sized and sector-aligned IO. |
| 955 | */ |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 956 | if (sector_size > sbp->sb_sectsize) { |
Dave Chinner | af34e09 | 2011-03-07 10:04:35 +1100 | [diff] [blame] | 957 | if (loud) |
| 958 | xfs_warn(mp, "device supports %u byte sectors (not %u)", |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 959 | sector_size, sbp->sb_sectsize); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 960 | error = ENOSYS; |
Dave Chinner | 26af655 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 961 | goto release_buf; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 962 | } |
| 963 | |
| 964 | /* |
| 965 | * If device sector size is smaller than the superblock size, |
| 966 | * re-read the superblock so the buffer is correctly sized. |
| 967 | */ |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 968 | if (sector_size < sbp->sb_sectsize) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 969 | xfs_buf_relse(bp); |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 970 | sector_size = sbp->sb_sectsize; |
Dave Chinner | 26af655 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 971 | goto reread; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 972 | } |
| 973 | |
Lachlan McIlroy | 5478eea | 2007-02-10 18:36:29 +1100 | [diff] [blame] | 974 | /* Initialize per-cpu counters */ |
| 975 | xfs_icsb_reinit_counters(mp); |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 976 | |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 977 | /* no need to be quiet anymore, so reset the buf ops */ |
| 978 | bp->b_ops = &xfs_sb_buf_ops; |
| 979 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 980 | mp->m_sb_bp = bp; |
Dave Chinner | 26af655 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 981 | xfs_buf_unlock(bp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 982 | return 0; |
| 983 | |
Dave Chinner | 26af655 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 984 | release_buf: |
| 985 | xfs_buf_relse(bp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 986 | return error; |
| 987 | } |
| 988 | |
| 989 | |
| 990 | /* |
| 991 | * xfs_mount_common |
| 992 | * |
| 993 | * Mount initialization code establishing various mount |
| 994 | * fields from the superblock associated with the given |
| 995 | * mount structure |
| 996 | */ |
Christoph Hellwig | ba0f32d | 2005-06-21 15:36:52 +1000 | [diff] [blame] | 997 | STATIC void |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 998 | xfs_mount_common(xfs_mount_t *mp, xfs_sb_t *sbp) |
| 999 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1000 | mp->m_agfrotor = mp->m_agirotor = 0; |
Eric Sandeen | 007c61c | 2007-10-11 17:43:56 +1000 | [diff] [blame] | 1001 | spin_lock_init(&mp->m_agirotor_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1002 | mp->m_maxagi = mp->m_sb.sb_agcount; |
| 1003 | mp->m_blkbit_log = sbp->sb_blocklog + XFS_NBBYLOG; |
| 1004 | mp->m_blkbb_log = sbp->sb_blocklog - BBSHIFT; |
| 1005 | mp->m_sectbb_log = sbp->sb_sectlog - BBSHIFT; |
| 1006 | mp->m_agno_log = xfs_highbit32(sbp->sb_agcount - 1) + 1; |
| 1007 | mp->m_agino_log = sbp->sb_inopblog + sbp->sb_agblklog; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1008 | mp->m_blockmask = sbp->sb_blocksize - 1; |
| 1009 | mp->m_blockwsize = sbp->sb_blocksize >> XFS_WORDLOG; |
| 1010 | mp->m_blockwmask = mp->m_blockwsize - 1; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1011 | |
Christoph Hellwig | 60197e8 | 2008-10-30 17:11:19 +1100 | [diff] [blame] | 1012 | mp->m_alloc_mxr[0] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 1); |
| 1013 | mp->m_alloc_mxr[1] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 0); |
| 1014 | mp->m_alloc_mnr[0] = mp->m_alloc_mxr[0] / 2; |
| 1015 | mp->m_alloc_mnr[1] = mp->m_alloc_mxr[1] / 2; |
| 1016 | |
| 1017 | mp->m_inobt_mxr[0] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 1); |
| 1018 | mp->m_inobt_mxr[1] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 0); |
| 1019 | mp->m_inobt_mnr[0] = mp->m_inobt_mxr[0] / 2; |
| 1020 | mp->m_inobt_mnr[1] = mp->m_inobt_mxr[1] / 2; |
| 1021 | |
| 1022 | mp->m_bmap_dmxr[0] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 1); |
| 1023 | mp->m_bmap_dmxr[1] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 0); |
| 1024 | mp->m_bmap_dmnr[0] = mp->m_bmap_dmxr[0] / 2; |
| 1025 | mp->m_bmap_dmnr[1] = mp->m_bmap_dmxr[1] / 2; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1026 | |
| 1027 | mp->m_bsize = XFS_FSB_TO_BB(mp, 1); |
| 1028 | mp->m_ialloc_inos = (int)MAX((__uint16_t)XFS_INODES_PER_CHUNK, |
| 1029 | sbp->sb_inopblock); |
| 1030 | mp->m_ialloc_blks = mp->m_ialloc_inos >> sbp->sb_inopblog; |
| 1031 | } |
David Chinner | 92821e2 | 2007-05-24 15:26:31 +1000 | [diff] [blame] | 1032 | |
| 1033 | /* |
| 1034 | * xfs_initialize_perag_data |
| 1035 | * |
| 1036 | * Read in each per-ag structure so we can count up the number of |
| 1037 | * allocated inodes, free inodes and used filesystem blocks as this |
| 1038 | * information is no longer persistent in the superblock. Once we have |
| 1039 | * this information, write it into the in-core superblock structure. |
| 1040 | */ |
| 1041 | STATIC int |
| 1042 | xfs_initialize_perag_data(xfs_mount_t *mp, xfs_agnumber_t agcount) |
| 1043 | { |
| 1044 | xfs_agnumber_t index; |
| 1045 | xfs_perag_t *pag; |
| 1046 | xfs_sb_t *sbp = &mp->m_sb; |
| 1047 | uint64_t ifree = 0; |
| 1048 | uint64_t ialloc = 0; |
| 1049 | uint64_t bfree = 0; |
| 1050 | uint64_t bfreelst = 0; |
| 1051 | uint64_t btree = 0; |
| 1052 | int error; |
David Chinner | 92821e2 | 2007-05-24 15:26:31 +1000 | [diff] [blame] | 1053 | |
| 1054 | for (index = 0; index < agcount; index++) { |
| 1055 | /* |
| 1056 | * read the agf, then the agi. This gets us |
Malcolm Parsons | 9da096f | 2009-03-29 09:55:42 +0200 | [diff] [blame] | 1057 | * all the information we need and populates the |
David Chinner | 92821e2 | 2007-05-24 15:26:31 +1000 | [diff] [blame] | 1058 | * per-ag structures for us. |
| 1059 | */ |
| 1060 | error = xfs_alloc_pagf_init(mp, NULL, index, 0); |
| 1061 | if (error) |
| 1062 | return error; |
| 1063 | |
| 1064 | error = xfs_ialloc_pagi_init(mp, NULL, index); |
| 1065 | if (error) |
| 1066 | return error; |
Dave Chinner | 44b56e0 | 2010-01-11 11:47:43 +0000 | [diff] [blame] | 1067 | pag = xfs_perag_get(mp, index); |
David Chinner | 92821e2 | 2007-05-24 15:26:31 +1000 | [diff] [blame] | 1068 | ifree += pag->pagi_freecount; |
| 1069 | ialloc += pag->pagi_count; |
| 1070 | bfree += pag->pagf_freeblks; |
| 1071 | bfreelst += pag->pagf_flcount; |
| 1072 | btree += pag->pagf_btreeblks; |
Dave Chinner | 44b56e0 | 2010-01-11 11:47:43 +0000 | [diff] [blame] | 1073 | xfs_perag_put(pag); |
David Chinner | 92821e2 | 2007-05-24 15:26:31 +1000 | [diff] [blame] | 1074 | } |
| 1075 | /* |
| 1076 | * Overwrite incore superblock counters with just-read data |
| 1077 | */ |
Eric Sandeen | 3685c2a | 2007-10-11 17:42:32 +1000 | [diff] [blame] | 1078 | spin_lock(&mp->m_sb_lock); |
David Chinner | 92821e2 | 2007-05-24 15:26:31 +1000 | [diff] [blame] | 1079 | sbp->sb_ifree = ifree; |
| 1080 | sbp->sb_icount = ialloc; |
| 1081 | sbp->sb_fdblocks = bfree + bfreelst + btree; |
Eric Sandeen | 3685c2a | 2007-10-11 17:42:32 +1000 | [diff] [blame] | 1082 | spin_unlock(&mp->m_sb_lock); |
David Chinner | 92821e2 | 2007-05-24 15:26:31 +1000 | [diff] [blame] | 1083 | |
| 1084 | /* Fixup the per-cpu counters as well. */ |
| 1085 | xfs_icsb_reinit_counters(mp); |
| 1086 | |
| 1087 | return 0; |
| 1088 | } |
| 1089 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1090 | /* |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1091 | * Update alignment values based on mount options and sb values |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1092 | */ |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1093 | STATIC int |
Christoph Hellwig | 7884bc8 | 2009-01-19 02:04:07 +0100 | [diff] [blame] | 1094 | xfs_update_alignment(xfs_mount_t *mp) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1095 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1096 | xfs_sb_t *sbp = &(mp->m_sb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1097 | |
Christoph Hellwig | 4249023 | 2008-08-13 16:49:32 +1000 | [diff] [blame] | 1098 | if (mp->m_dalign) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1099 | /* |
| 1100 | * If stripe unit and stripe width are not multiples |
| 1101 | * of the fs blocksize turn off alignment. |
| 1102 | */ |
| 1103 | if ((BBTOB(mp->m_dalign) & mp->m_blockmask) || |
| 1104 | (BBTOB(mp->m_swidth) & mp->m_blockmask)) { |
Jie Liu | 39a45d8 | 2013-05-02 19:27:47 +0800 | [diff] [blame] | 1105 | xfs_warn(mp, |
| 1106 | "alignment check failed: sunit/swidth vs. blocksize(%d)", |
| 1107 | sbp->sb_blocksize); |
| 1108 | return XFS_ERROR(EINVAL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1109 | } else { |
| 1110 | /* |
| 1111 | * Convert the stripe unit and width to FSBs. |
| 1112 | */ |
| 1113 | mp->m_dalign = XFS_BB_TO_FSBT(mp, mp->m_dalign); |
| 1114 | if (mp->m_dalign && (sbp->sb_agblocks % mp->m_dalign)) { |
Dave Chinner | 5348778 | 2011-03-07 10:05:35 +1100 | [diff] [blame] | 1115 | xfs_warn(mp, |
Jie Liu | 39a45d8 | 2013-05-02 19:27:47 +0800 | [diff] [blame] | 1116 | "alignment check failed: sunit/swidth vs. agsize(%d)", |
| 1117 | sbp->sb_agblocks); |
| 1118 | return XFS_ERROR(EINVAL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1119 | } else if (mp->m_dalign) { |
| 1120 | mp->m_swidth = XFS_BB_TO_FSBT(mp, mp->m_swidth); |
| 1121 | } else { |
Jie Liu | 39a45d8 | 2013-05-02 19:27:47 +0800 | [diff] [blame] | 1122 | xfs_warn(mp, |
| 1123 | "alignment check failed: sunit(%d) less than bsize(%d)", |
| 1124 | mp->m_dalign, sbp->sb_blocksize); |
| 1125 | return XFS_ERROR(EINVAL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1126 | } |
| 1127 | } |
| 1128 | |
| 1129 | /* |
| 1130 | * Update superblock with new values |
| 1131 | * and log changes |
| 1132 | */ |
Eric Sandeen | 6211870 | 2008-03-06 13:44:28 +1100 | [diff] [blame] | 1133 | if (xfs_sb_version_hasdalign(sbp)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1134 | if (sbp->sb_unit != mp->m_dalign) { |
| 1135 | sbp->sb_unit = mp->m_dalign; |
Christoph Hellwig | 7884bc8 | 2009-01-19 02:04:07 +0100 | [diff] [blame] | 1136 | mp->m_update_flags |= XFS_SB_UNIT; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1137 | } |
| 1138 | if (sbp->sb_width != mp->m_swidth) { |
| 1139 | sbp->sb_width = mp->m_swidth; |
Christoph Hellwig | 7884bc8 | 2009-01-19 02:04:07 +0100 | [diff] [blame] | 1140 | mp->m_update_flags |= XFS_SB_WIDTH; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1141 | } |
Jie Liu | 34d7f60 | 2013-05-02 19:27:53 +0800 | [diff] [blame] | 1142 | } else { |
| 1143 | xfs_warn(mp, |
| 1144 | "cannot change alignment: superblock does not support data alignment"); |
| 1145 | return XFS_ERROR(EINVAL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1146 | } |
| 1147 | } else if ((mp->m_flags & XFS_MOUNT_NOALIGN) != XFS_MOUNT_NOALIGN && |
Eric Sandeen | 6211870 | 2008-03-06 13:44:28 +1100 | [diff] [blame] | 1148 | xfs_sb_version_hasdalign(&mp->m_sb)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1149 | mp->m_dalign = sbp->sb_unit; |
| 1150 | mp->m_swidth = sbp->sb_width; |
| 1151 | } |
| 1152 | |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1153 | return 0; |
| 1154 | } |
| 1155 | |
| 1156 | /* |
| 1157 | * Set the maximum inode count for this filesystem |
| 1158 | */ |
| 1159 | STATIC void |
| 1160 | xfs_set_maxicount(xfs_mount_t *mp) |
| 1161 | { |
| 1162 | xfs_sb_t *sbp = &(mp->m_sb); |
| 1163 | __uint64_t icount; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1164 | |
| 1165 | if (sbp->sb_imax_pct) { |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1166 | /* |
| 1167 | * Make sure the maximum inode count is a multiple |
| 1168 | * of the units we allocate inodes in. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1169 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1170 | icount = sbp->sb_dblocks * sbp->sb_imax_pct; |
| 1171 | do_div(icount, 100); |
| 1172 | do_div(icount, mp->m_ialloc_blks); |
| 1173 | mp->m_maxicount = (icount * mp->m_ialloc_blks) << |
| 1174 | sbp->sb_inopblog; |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1175 | } else { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1176 | mp->m_maxicount = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1177 | } |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1178 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1179 | |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1180 | /* |
| 1181 | * Set the default minimum read and write sizes unless |
| 1182 | * already specified in a mount option. |
| 1183 | * We use smaller I/O sizes when the file system |
| 1184 | * is being used for NFS service (wsync mount option). |
| 1185 | */ |
| 1186 | STATIC void |
| 1187 | xfs_set_rw_sizes(xfs_mount_t *mp) |
| 1188 | { |
| 1189 | xfs_sb_t *sbp = &(mp->m_sb); |
| 1190 | int readio_log, writeio_log; |
| 1191 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1192 | if (!(mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)) { |
| 1193 | if (mp->m_flags & XFS_MOUNT_WSYNC) { |
| 1194 | readio_log = XFS_WSYNC_READIO_LOG; |
| 1195 | writeio_log = XFS_WSYNC_WRITEIO_LOG; |
| 1196 | } else { |
| 1197 | readio_log = XFS_READIO_LOG_LARGE; |
| 1198 | writeio_log = XFS_WRITEIO_LOG_LARGE; |
| 1199 | } |
| 1200 | } else { |
| 1201 | readio_log = mp->m_readio_log; |
| 1202 | writeio_log = mp->m_writeio_log; |
| 1203 | } |
| 1204 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1205 | if (sbp->sb_blocklog > readio_log) { |
| 1206 | mp->m_readio_log = sbp->sb_blocklog; |
| 1207 | } else { |
| 1208 | mp->m_readio_log = readio_log; |
| 1209 | } |
| 1210 | mp->m_readio_blocks = 1 << (mp->m_readio_log - sbp->sb_blocklog); |
| 1211 | if (sbp->sb_blocklog > writeio_log) { |
| 1212 | mp->m_writeio_log = sbp->sb_blocklog; |
| 1213 | } else { |
| 1214 | mp->m_writeio_log = writeio_log; |
| 1215 | } |
| 1216 | mp->m_writeio_blocks = 1 << (mp->m_writeio_log - sbp->sb_blocklog); |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1217 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1218 | |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1219 | /* |
Dave Chinner | 055388a | 2011-01-04 11:35:03 +1100 | [diff] [blame] | 1220 | * precalculate the low space thresholds for dynamic speculative preallocation. |
| 1221 | */ |
| 1222 | void |
| 1223 | xfs_set_low_space_thresholds( |
| 1224 | struct xfs_mount *mp) |
| 1225 | { |
| 1226 | int i; |
| 1227 | |
| 1228 | for (i = 0; i < XFS_LOWSP_MAX; i++) { |
| 1229 | __uint64_t space = mp->m_sb.sb_dblocks; |
| 1230 | |
| 1231 | do_div(space, 100); |
| 1232 | mp->m_low_space[i] = space * (i + 1); |
| 1233 | } |
| 1234 | } |
| 1235 | |
| 1236 | |
| 1237 | /* |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1238 | * Set whether we're using inode alignment. |
| 1239 | */ |
| 1240 | STATIC void |
| 1241 | xfs_set_inoalignment(xfs_mount_t *mp) |
| 1242 | { |
Eric Sandeen | 6211870 | 2008-03-06 13:44:28 +1100 | [diff] [blame] | 1243 | if (xfs_sb_version_hasalign(&mp->m_sb) && |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1244 | mp->m_sb.sb_inoalignmt >= |
| 1245 | XFS_B_TO_FSBT(mp, mp->m_inode_cluster_size)) |
| 1246 | mp->m_inoalign_mask = mp->m_sb.sb_inoalignmt - 1; |
| 1247 | else |
| 1248 | mp->m_inoalign_mask = 0; |
| 1249 | /* |
| 1250 | * If we are using stripe alignment, check whether |
| 1251 | * the stripe unit is a multiple of the inode alignment |
| 1252 | */ |
| 1253 | if (mp->m_dalign && mp->m_inoalign_mask && |
| 1254 | !(mp->m_dalign & mp->m_inoalign_mask)) |
| 1255 | mp->m_sinoalign = mp->m_dalign; |
| 1256 | else |
| 1257 | mp->m_sinoalign = 0; |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1258 | } |
| 1259 | |
| 1260 | /* |
| 1261 | * Check that the data (and log if separate) are an ok size. |
| 1262 | */ |
| 1263 | STATIC int |
Christoph Hellwig | 4249023 | 2008-08-13 16:49:32 +1000 | [diff] [blame] | 1264 | xfs_check_sizes(xfs_mount_t *mp) |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1265 | { |
| 1266 | xfs_buf_t *bp; |
| 1267 | xfs_daddr_t d; |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1268 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1269 | d = (xfs_daddr_t)XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks); |
| 1270 | if (XFS_BB_TO_FSB(mp, d) != mp->m_sb.sb_dblocks) { |
Dave Chinner | 0b932cc | 2011-03-07 10:08:35 +1100 | [diff] [blame] | 1271 | xfs_warn(mp, "filesystem size mismatch detected"); |
Eric Sandeen | 657a4cf | 2010-04-30 03:42:49 +0000 | [diff] [blame] | 1272 | return XFS_ERROR(EFBIG); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1273 | } |
Dave Chinner | e70b73f | 2012-04-23 15:58:49 +1000 | [diff] [blame] | 1274 | bp = xfs_buf_read_uncached(mp->m_ddev_targp, |
Dave Chinner | 1922c94 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 1275 | d - XFS_FSS_TO_BB(mp, 1), |
Dave Chinner | c3f8fc7 | 2012-11-12 22:54:01 +1100 | [diff] [blame] | 1276 | XFS_FSS_TO_BB(mp, 1), 0, NULL); |
Dave Chinner | 1922c94 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 1277 | if (!bp) { |
Dave Chinner | 0b932cc | 2011-03-07 10:08:35 +1100 | [diff] [blame] | 1278 | xfs_warn(mp, "last sector read failed"); |
Dave Chinner | 1922c94 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 1279 | return EIO; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1280 | } |
Dave Chinner | 1922c94 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 1281 | xfs_buf_relse(bp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1282 | |
Christoph Hellwig | 4249023 | 2008-08-13 16:49:32 +1000 | [diff] [blame] | 1283 | if (mp->m_logdev_targp != mp->m_ddev_targp) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1284 | d = (xfs_daddr_t)XFS_FSB_TO_BB(mp, mp->m_sb.sb_logblocks); |
| 1285 | if (XFS_BB_TO_FSB(mp, d) != mp->m_sb.sb_logblocks) { |
Dave Chinner | 0b932cc | 2011-03-07 10:08:35 +1100 | [diff] [blame] | 1286 | xfs_warn(mp, "log size mismatch detected"); |
Eric Sandeen | 657a4cf | 2010-04-30 03:42:49 +0000 | [diff] [blame] | 1287 | return XFS_ERROR(EFBIG); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1288 | } |
Dave Chinner | e70b73f | 2012-04-23 15:58:49 +1000 | [diff] [blame] | 1289 | bp = xfs_buf_read_uncached(mp->m_logdev_targp, |
Dave Chinner | 1922c94 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 1290 | d - XFS_FSB_TO_BB(mp, 1), |
Dave Chinner | c3f8fc7 | 2012-11-12 22:54:01 +1100 | [diff] [blame] | 1291 | XFS_FSB_TO_BB(mp, 1), 0, NULL); |
Dave Chinner | 1922c94 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 1292 | if (!bp) { |
Dave Chinner | 0b932cc | 2011-03-07 10:08:35 +1100 | [diff] [blame] | 1293 | xfs_warn(mp, "log device read failed"); |
Dave Chinner | 1922c94 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 1294 | return EIO; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1295 | } |
Dave Chinner | 1922c94 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 1296 | xfs_buf_relse(bp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1297 | } |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1298 | return 0; |
| 1299 | } |
| 1300 | |
| 1301 | /* |
Christoph Hellwig | 7d09525 | 2009-06-08 15:33:32 +0200 | [diff] [blame] | 1302 | * Clear the quotaflags in memory and in the superblock. |
| 1303 | */ |
| 1304 | int |
| 1305 | xfs_mount_reset_sbqflags( |
| 1306 | struct xfs_mount *mp) |
| 1307 | { |
| 1308 | int error; |
| 1309 | struct xfs_trans *tp; |
| 1310 | |
| 1311 | mp->m_qflags = 0; |
| 1312 | |
| 1313 | /* |
| 1314 | * It is OK to look at sb_qflags here in mount path, |
| 1315 | * without m_sb_lock. |
| 1316 | */ |
| 1317 | if (mp->m_sb.sb_qflags == 0) |
| 1318 | return 0; |
| 1319 | spin_lock(&mp->m_sb_lock); |
| 1320 | mp->m_sb.sb_qflags = 0; |
| 1321 | spin_unlock(&mp->m_sb_lock); |
| 1322 | |
| 1323 | /* |
| 1324 | * If the fs is readonly, let the incore superblock run |
| 1325 | * with quotas off but don't flush the update out to disk |
| 1326 | */ |
| 1327 | if (mp->m_flags & XFS_MOUNT_RDONLY) |
| 1328 | return 0; |
| 1329 | |
Christoph Hellwig | 7d09525 | 2009-06-08 15:33:32 +0200 | [diff] [blame] | 1330 | tp = xfs_trans_alloc(mp, XFS_TRANS_QM_SBCHANGE); |
Jeff Liu | b0c10b98 | 2013-01-28 21:26:16 +0800 | [diff] [blame] | 1331 | error = xfs_trans_reserve(tp, 0, XFS_QM_SBCHANGE_LOG_RES(mp), |
| 1332 | 0, 0, XFS_DEFAULT_LOG_COUNT); |
Christoph Hellwig | 7d09525 | 2009-06-08 15:33:32 +0200 | [diff] [blame] | 1333 | if (error) { |
| 1334 | xfs_trans_cancel(tp, 0); |
Dave Chinner | 5348778 | 2011-03-07 10:05:35 +1100 | [diff] [blame] | 1335 | xfs_alert(mp, "%s: Superblock update failed!", __func__); |
Christoph Hellwig | 7d09525 | 2009-06-08 15:33:32 +0200 | [diff] [blame] | 1336 | return error; |
| 1337 | } |
| 1338 | |
| 1339 | xfs_mod_sb(tp, XFS_SB_QFLAGS); |
| 1340 | return xfs_trans_commit(tp, 0); |
| 1341 | } |
| 1342 | |
Eric Sandeen | d5db0f9 | 2010-02-05 22:59:53 +0000 | [diff] [blame] | 1343 | __uint64_t |
| 1344 | xfs_default_resblks(xfs_mount_t *mp) |
| 1345 | { |
| 1346 | __uint64_t resblks; |
| 1347 | |
| 1348 | /* |
Dave Chinner | 8babd8a | 2010-03-04 01:46:25 +0000 | [diff] [blame] | 1349 | * We default to 5% or 8192 fsbs of space reserved, whichever is |
| 1350 | * smaller. This is intended to cover concurrent allocation |
| 1351 | * transactions when we initially hit enospc. These each require a 4 |
| 1352 | * block reservation. Hence by default we cover roughly 2000 concurrent |
| 1353 | * allocation reservations. |
Eric Sandeen | d5db0f9 | 2010-02-05 22:59:53 +0000 | [diff] [blame] | 1354 | */ |
| 1355 | resblks = mp->m_sb.sb_dblocks; |
| 1356 | do_div(resblks, 20); |
Dave Chinner | 8babd8a | 2010-03-04 01:46:25 +0000 | [diff] [blame] | 1357 | resblks = min_t(__uint64_t, resblks, 8192); |
Eric Sandeen | d5db0f9 | 2010-02-05 22:59:53 +0000 | [diff] [blame] | 1358 | return resblks; |
| 1359 | } |
| 1360 | |
Christoph Hellwig | 7d09525 | 2009-06-08 15:33:32 +0200 | [diff] [blame] | 1361 | /* |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1362 | * This function does the following on an initial mount of a file system: |
| 1363 | * - reads the superblock from disk and init the mount struct |
| 1364 | * - if we're a 32-bit kernel, do a size check on the superblock |
| 1365 | * so we don't mount terabyte filesystems |
| 1366 | * - init mount struct realtime fields |
| 1367 | * - allocate inode hash table for fs |
| 1368 | * - init directory manager |
| 1369 | * - perform recovery and init the log manager |
| 1370 | */ |
| 1371 | int |
| 1372 | xfs_mountfs( |
Christoph Hellwig | 4249023 | 2008-08-13 16:49:32 +1000 | [diff] [blame] | 1373 | xfs_mount_t *mp) |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1374 | { |
| 1375 | xfs_sb_t *sbp = &(mp->m_sb); |
| 1376 | xfs_inode_t *rip; |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1377 | __uint64_t resblks; |
Christoph Hellwig | 7d09525 | 2009-06-08 15:33:32 +0200 | [diff] [blame] | 1378 | uint quotamount = 0; |
| 1379 | uint quotaflags = 0; |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1380 | int error = 0; |
| 1381 | |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1382 | xfs_mount_common(mp, sbp); |
| 1383 | |
| 1384 | /* |
Eric Sandeen | e6957ea | 2008-04-10 12:19:34 +1000 | [diff] [blame] | 1385 | * Check for a mismatched features2 values. Older kernels |
| 1386 | * read & wrote into the wrong sb offset for sb_features2 |
| 1387 | * on some platforms due to xfs_sb_t not being 64bit size aligned |
| 1388 | * when sb_features2 was added, which made older superblock |
| 1389 | * reading/writing routines swap it as a 64-bit value. |
David Chinner | ee1c090 | 2008-03-06 13:45:50 +1100 | [diff] [blame] | 1390 | * |
Eric Sandeen | e6957ea | 2008-04-10 12:19:34 +1000 | [diff] [blame] | 1391 | * For backwards compatibility, we make both slots equal. |
| 1392 | * |
| 1393 | * If we detect a mismatched field, we OR the set bits into the |
| 1394 | * existing features2 field in case it has already been modified; we |
| 1395 | * don't want to lose any features. We then update the bad location |
| 1396 | * with the ORed value so that older kernels will see any features2 |
| 1397 | * flags, and mark the two fields as needing updates once the |
| 1398 | * transaction subsystem is online. |
David Chinner | ee1c090 | 2008-03-06 13:45:50 +1100 | [diff] [blame] | 1399 | */ |
Eric Sandeen | e6957ea | 2008-04-10 12:19:34 +1000 | [diff] [blame] | 1400 | if (xfs_sb_has_mismatched_features2(sbp)) { |
Dave Chinner | 0b932cc | 2011-03-07 10:08:35 +1100 | [diff] [blame] | 1401 | xfs_warn(mp, "correcting sb_features alignment problem"); |
David Chinner | ee1c090 | 2008-03-06 13:45:50 +1100 | [diff] [blame] | 1402 | sbp->sb_features2 |= sbp->sb_bad_features2; |
Eric Sandeen | e6957ea | 2008-04-10 12:19:34 +1000 | [diff] [blame] | 1403 | sbp->sb_bad_features2 = sbp->sb_features2; |
Christoph Hellwig | 7884bc8 | 2009-01-19 02:04:07 +0100 | [diff] [blame] | 1404 | mp->m_update_flags |= XFS_SB_FEATURES2 | XFS_SB_BAD_FEATURES2; |
Eric Sandeen | e6957ea | 2008-04-10 12:19:34 +1000 | [diff] [blame] | 1405 | |
| 1406 | /* |
| 1407 | * Re-check for ATTR2 in case it was found in bad_features2 |
| 1408 | * slot. |
| 1409 | */ |
Tim Shimmin | 7c12f29 | 2008-04-30 18:15:28 +1000 | [diff] [blame] | 1410 | if (xfs_sb_version_hasattr2(&mp->m_sb) && |
| 1411 | !(mp->m_flags & XFS_MOUNT_NOATTR2)) |
Eric Sandeen | e6957ea | 2008-04-10 12:19:34 +1000 | [diff] [blame] | 1412 | mp->m_flags |= XFS_MOUNT_ATTR2; |
Tim Shimmin | 7c12f29 | 2008-04-30 18:15:28 +1000 | [diff] [blame] | 1413 | } |
Eric Sandeen | e6957ea | 2008-04-10 12:19:34 +1000 | [diff] [blame] | 1414 | |
Tim Shimmin | 7c12f29 | 2008-04-30 18:15:28 +1000 | [diff] [blame] | 1415 | if (xfs_sb_version_hasattr2(&mp->m_sb) && |
| 1416 | (mp->m_flags & XFS_MOUNT_NOATTR2)) { |
| 1417 | xfs_sb_version_removeattr2(&mp->m_sb); |
Christoph Hellwig | 7884bc8 | 2009-01-19 02:04:07 +0100 | [diff] [blame] | 1418 | mp->m_update_flags |= XFS_SB_FEATURES2; |
Tim Shimmin | 7c12f29 | 2008-04-30 18:15:28 +1000 | [diff] [blame] | 1419 | |
| 1420 | /* update sb_versionnum for the clearing of the morebits */ |
| 1421 | if (!sbp->sb_features2) |
Christoph Hellwig | 7884bc8 | 2009-01-19 02:04:07 +0100 | [diff] [blame] | 1422 | mp->m_update_flags |= XFS_SB_VERSIONNUM; |
David Chinner | ee1c090 | 2008-03-06 13:45:50 +1100 | [diff] [blame] | 1423 | } |
| 1424 | |
| 1425 | /* |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1426 | * Check if sb_agblocks is aligned at stripe boundary |
| 1427 | * If sb_agblocks is NOT aligned turn off m_dalign since |
| 1428 | * allocator alignment is within an ag, therefore ag has |
| 1429 | * to be aligned at stripe boundary. |
| 1430 | */ |
Christoph Hellwig | 7884bc8 | 2009-01-19 02:04:07 +0100 | [diff] [blame] | 1431 | error = xfs_update_alignment(mp); |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1432 | if (error) |
Christoph Hellwig | f9057e3 | 2009-02-04 09:31:52 +0100 | [diff] [blame] | 1433 | goto out; |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1434 | |
| 1435 | xfs_alloc_compute_maxlevels(mp); |
| 1436 | xfs_bmap_compute_maxlevels(mp, XFS_DATA_FORK); |
| 1437 | xfs_bmap_compute_maxlevels(mp, XFS_ATTR_FORK); |
| 1438 | xfs_ialloc_compute_maxlevels(mp); |
| 1439 | |
| 1440 | xfs_set_maxicount(mp); |
| 1441 | |
Christoph Hellwig | 2717420 | 2009-03-30 10:21:31 +0200 | [diff] [blame] | 1442 | error = xfs_uuid_mount(mp); |
| 1443 | if (error) |
| 1444 | goto out; |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1445 | |
| 1446 | /* |
| 1447 | * Set the minimum read and write sizes |
| 1448 | */ |
| 1449 | xfs_set_rw_sizes(mp); |
| 1450 | |
Dave Chinner | 055388a | 2011-01-04 11:35:03 +1100 | [diff] [blame] | 1451 | /* set the low space thresholds for dynamic preallocation */ |
| 1452 | xfs_set_low_space_thresholds(mp); |
| 1453 | |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1454 | /* |
| 1455 | * Set the inode cluster size. |
| 1456 | * This may still be overridden by the file system |
| 1457 | * block size if it is larger than the chosen cluster size. |
| 1458 | */ |
| 1459 | mp->m_inode_cluster_size = XFS_INODE_BIG_CLUSTER_SIZE; |
| 1460 | |
| 1461 | /* |
| 1462 | * Set inode alignment fields |
| 1463 | */ |
| 1464 | xfs_set_inoalignment(mp); |
| 1465 | |
| 1466 | /* |
| 1467 | * Check that the data (and log if separate) are an ok size. |
| 1468 | */ |
Christoph Hellwig | 4249023 | 2008-08-13 16:49:32 +1000 | [diff] [blame] | 1469 | error = xfs_check_sizes(mp); |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1470 | if (error) |
Christoph Hellwig | f9057e3 | 2009-02-04 09:31:52 +0100 | [diff] [blame] | 1471 | goto out_remove_uuid; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1472 | |
| 1473 | /* |
| 1474 | * Initialize realtime fields in the mount structure |
| 1475 | */ |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1476 | error = xfs_rtmount_init(mp); |
| 1477 | if (error) { |
Dave Chinner | 0b932cc | 2011-03-07 10:08:35 +1100 | [diff] [blame] | 1478 | xfs_warn(mp, "RT mount failed"); |
Christoph Hellwig | f9057e3 | 2009-02-04 09:31:52 +0100 | [diff] [blame] | 1479 | goto out_remove_uuid; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1480 | } |
| 1481 | |
| 1482 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1483 | * Copies the low order bits of the timestamp and the randomly |
| 1484 | * set "sequence" number out of a UUID. |
| 1485 | */ |
| 1486 | uuid_getnodeuniq(&sbp->sb_uuid, mp->m_fixedfsid); |
| 1487 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1488 | mp->m_dmevmask = 0; /* not persistent; set after each mount */ |
| 1489 | |
Nathan Scott | f6c2d1f | 2006-06-20 13:04:51 +1000 | [diff] [blame] | 1490 | xfs_dir_mount(mp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1491 | |
| 1492 | /* |
| 1493 | * Initialize the attribute manager's entries. |
| 1494 | */ |
| 1495 | mp->m_attr_magicpct = (mp->m_sb.sb_blocksize * 37) / 100; |
| 1496 | |
| 1497 | /* |
| 1498 | * Initialize the precomputed transaction reservations values. |
| 1499 | */ |
| 1500 | xfs_trans_init(mp); |
| 1501 | |
| 1502 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1503 | * Allocate and initialize the per-ag data. |
| 1504 | */ |
Dave Chinner | 1c1c6eb | 2010-01-11 11:47:44 +0000 | [diff] [blame] | 1505 | spin_lock_init(&mp->m_perag_lock); |
Dave Chinner | 9b98b6f | 2010-05-27 01:58:13 +0000 | [diff] [blame] | 1506 | INIT_RADIX_TREE(&mp->m_perag_tree, GFP_ATOMIC); |
Dave Chinner | 1c1c6eb | 2010-01-11 11:47:44 +0000 | [diff] [blame] | 1507 | error = xfs_initialize_perag(mp, sbp->sb_agcount, &mp->m_maxagi); |
| 1508 | if (error) { |
Dave Chinner | 0b932cc | 2011-03-07 10:08:35 +1100 | [diff] [blame] | 1509 | xfs_warn(mp, "Failed per-ag init: %d", error); |
Christoph Hellwig | f9057e3 | 2009-02-04 09:31:52 +0100 | [diff] [blame] | 1510 | goto out_remove_uuid; |
Dave Chinner | 1c1c6eb | 2010-01-11 11:47:44 +0000 | [diff] [blame] | 1511 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1512 | |
Christoph Hellwig | f9057e3 | 2009-02-04 09:31:52 +0100 | [diff] [blame] | 1513 | if (!sbp->sb_logblocks) { |
Dave Chinner | 0b932cc | 2011-03-07 10:08:35 +1100 | [diff] [blame] | 1514 | xfs_warn(mp, "no log defined"); |
Christoph Hellwig | f9057e3 | 2009-02-04 09:31:52 +0100 | [diff] [blame] | 1515 | XFS_ERROR_REPORT("xfs_mountfs", XFS_ERRLEVEL_LOW, mp); |
| 1516 | error = XFS_ERROR(EFSCORRUPTED); |
| 1517 | goto out_free_perag; |
| 1518 | } |
| 1519 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1520 | /* |
| 1521 | * log's mount-time initialization. Perform 1st part recovery if needed |
| 1522 | */ |
Christoph Hellwig | f9057e3 | 2009-02-04 09:31:52 +0100 | [diff] [blame] | 1523 | error = xfs_log_mount(mp, mp->m_logdev_targp, |
| 1524 | XFS_FSB_TO_DADDR(mp, sbp->sb_logstart), |
| 1525 | XFS_FSB_TO_BB(mp, sbp->sb_logblocks)); |
| 1526 | if (error) { |
Dave Chinner | 0b932cc | 2011-03-07 10:08:35 +1100 | [diff] [blame] | 1527 | xfs_warn(mp, "log mount failed"); |
Dave Chinner | d4f3512 | 2012-04-23 15:59:06 +1000 | [diff] [blame] | 1528 | goto out_fail_wait; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1529 | } |
| 1530 | |
| 1531 | /* |
David Chinner | 92821e2 | 2007-05-24 15:26:31 +1000 | [diff] [blame] | 1532 | * Now the log is mounted, we know if it was an unclean shutdown or |
| 1533 | * not. If it was, with the first phase of recovery has completed, we |
| 1534 | * have consistent AG blocks on disk. We have not recovered EFIs yet, |
| 1535 | * but they are recovered transactionally in the second recovery phase |
| 1536 | * later. |
| 1537 | * |
| 1538 | * Hence we can safely re-initialise incore superblock counters from |
| 1539 | * the per-ag data. These may not be correct if the filesystem was not |
| 1540 | * cleanly unmounted, so we need to wait for recovery to finish before |
| 1541 | * doing this. |
| 1542 | * |
| 1543 | * If the filesystem was cleanly unmounted, then we can trust the |
| 1544 | * values in the superblock to be correct and we don't need to do |
| 1545 | * anything here. |
| 1546 | * |
| 1547 | * If we are currently making the filesystem, the initialisation will |
| 1548 | * fail as the perag data is in an undefined state. |
| 1549 | */ |
David Chinner | 92821e2 | 2007-05-24 15:26:31 +1000 | [diff] [blame] | 1550 | if (xfs_sb_version_haslazysbcount(&mp->m_sb) && |
| 1551 | !XFS_LAST_UNMOUNT_WAS_CLEAN(mp) && |
| 1552 | !mp->m_sb.sb_inprogress) { |
| 1553 | error = xfs_initialize_perag_data(mp, sbp->sb_agcount); |
Christoph Hellwig | f9057e3 | 2009-02-04 09:31:52 +0100 | [diff] [blame] | 1554 | if (error) |
Dave Chinner | d4f3512 | 2012-04-23 15:59:06 +1000 | [diff] [blame] | 1555 | goto out_fail_wait; |
David Chinner | 92821e2 | 2007-05-24 15:26:31 +1000 | [diff] [blame] | 1556 | } |
Christoph Hellwig | f9057e3 | 2009-02-04 09:31:52 +0100 | [diff] [blame] | 1557 | |
David Chinner | 92821e2 | 2007-05-24 15:26:31 +1000 | [diff] [blame] | 1558 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1559 | * Get and sanity-check the root inode. |
| 1560 | * Save the pointer to it in the mount structure. |
| 1561 | */ |
Dave Chinner | 7b6259e | 2010-06-24 11:35:17 +1000 | [diff] [blame] | 1562 | error = xfs_iget(mp, NULL, sbp->sb_rootino, 0, XFS_ILOCK_EXCL, &rip); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1563 | if (error) { |
Dave Chinner | 0b932cc | 2011-03-07 10:08:35 +1100 | [diff] [blame] | 1564 | xfs_warn(mp, "failed to read root inode"); |
Christoph Hellwig | f9057e3 | 2009-02-04 09:31:52 +0100 | [diff] [blame] | 1565 | goto out_log_dealloc; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1566 | } |
| 1567 | |
| 1568 | ASSERT(rip != NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1569 | |
Al Viro | abbede1 | 2011-07-26 02:31:30 -0400 | [diff] [blame] | 1570 | if (unlikely(!S_ISDIR(rip->i_d.di_mode))) { |
Dave Chinner | 0b932cc | 2011-03-07 10:08:35 +1100 | [diff] [blame] | 1571 | xfs_warn(mp, "corrupted root inode %llu: not a directory", |
Nathan Scott | b657452 | 2006-06-09 15:29:40 +1000 | [diff] [blame] | 1572 | (unsigned long long)rip->i_ino); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1573 | xfs_iunlock(rip, XFS_ILOCK_EXCL); |
| 1574 | XFS_ERROR_REPORT("xfs_mountfs_int(2)", XFS_ERRLEVEL_LOW, |
| 1575 | mp); |
| 1576 | error = XFS_ERROR(EFSCORRUPTED); |
Christoph Hellwig | f9057e3 | 2009-02-04 09:31:52 +0100 | [diff] [blame] | 1577 | goto out_rele_rip; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1578 | } |
| 1579 | mp->m_rootip = rip; /* save it */ |
| 1580 | |
| 1581 | xfs_iunlock(rip, XFS_ILOCK_EXCL); |
| 1582 | |
| 1583 | /* |
| 1584 | * Initialize realtime inode pointers in the mount structure |
| 1585 | */ |
Eric Sandeen | 0771fb4 | 2007-10-12 11:03:40 +1000 | [diff] [blame] | 1586 | error = xfs_rtmount_inodes(mp); |
| 1587 | if (error) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1588 | /* |
| 1589 | * Free up the root inode. |
| 1590 | */ |
Dave Chinner | 0b932cc | 2011-03-07 10:08:35 +1100 | [diff] [blame] | 1591 | xfs_warn(mp, "failed to read RT inodes"); |
Christoph Hellwig | f9057e3 | 2009-02-04 09:31:52 +0100 | [diff] [blame] | 1592 | goto out_rele_rip; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1593 | } |
| 1594 | |
| 1595 | /* |
Christoph Hellwig | 7884bc8 | 2009-01-19 02:04:07 +0100 | [diff] [blame] | 1596 | * If this is a read-only mount defer the superblock updates until |
| 1597 | * the next remount into writeable mode. Otherwise we would never |
| 1598 | * perform the update e.g. for the root filesystem. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1599 | */ |
Christoph Hellwig | 7884bc8 | 2009-01-19 02:04:07 +0100 | [diff] [blame] | 1600 | if (mp->m_update_flags && !(mp->m_flags & XFS_MOUNT_RDONLY)) { |
| 1601 | error = xfs_mount_log_sb(mp, mp->m_update_flags); |
David Chinner | e5720ee | 2008-04-10 12:21:18 +1000 | [diff] [blame] | 1602 | if (error) { |
Dave Chinner | 0b932cc | 2011-03-07 10:08:35 +1100 | [diff] [blame] | 1603 | xfs_warn(mp, "failed to write sb changes"); |
Christoph Hellwig | b93b6e4 | 2009-02-04 09:33:58 +0100 | [diff] [blame] | 1604 | goto out_rtunmount; |
David Chinner | e5720ee | 2008-04-10 12:21:18 +1000 | [diff] [blame] | 1605 | } |
| 1606 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1607 | |
| 1608 | /* |
| 1609 | * Initialise the XFS quota management subsystem for this mount |
| 1610 | */ |
Christoph Hellwig | 7d09525 | 2009-06-08 15:33:32 +0200 | [diff] [blame] | 1611 | if (XFS_IS_QUOTA_RUNNING(mp)) { |
| 1612 | error = xfs_qm_newmount(mp, "amount, "aflags); |
| 1613 | if (error) |
| 1614 | goto out_rtunmount; |
| 1615 | } else { |
| 1616 | ASSERT(!XFS_IS_QUOTA_ON(mp)); |
| 1617 | |
| 1618 | /* |
| 1619 | * If a file system had quotas running earlier, but decided to |
| 1620 | * mount without -o uquota/pquota/gquota options, revoke the |
| 1621 | * quotachecked license. |
| 1622 | */ |
| 1623 | if (mp->m_sb.sb_qflags & XFS_ALL_QUOTA_ACCT) { |
Dave Chinner | 0b932cc | 2011-03-07 10:08:35 +1100 | [diff] [blame] | 1624 | xfs_notice(mp, "resetting quota flags"); |
Christoph Hellwig | 7d09525 | 2009-06-08 15:33:32 +0200 | [diff] [blame] | 1625 | error = xfs_mount_reset_sbqflags(mp); |
| 1626 | if (error) |
| 1627 | return error; |
| 1628 | } |
| 1629 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1630 | |
| 1631 | /* |
| 1632 | * Finish recovering the file system. This part needed to be |
| 1633 | * delayed until after the root and real-time bitmap inodes |
| 1634 | * were consistently read in. |
| 1635 | */ |
Christoph Hellwig | 4249023 | 2008-08-13 16:49:32 +1000 | [diff] [blame] | 1636 | error = xfs_log_mount_finish(mp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1637 | if (error) { |
Dave Chinner | 0b932cc | 2011-03-07 10:08:35 +1100 | [diff] [blame] | 1638 | xfs_warn(mp, "log mount finish failed"); |
Christoph Hellwig | b93b6e4 | 2009-02-04 09:33:58 +0100 | [diff] [blame] | 1639 | goto out_rtunmount; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1640 | } |
| 1641 | |
| 1642 | /* |
| 1643 | * Complete the quota initialisation, post-log-replay component. |
| 1644 | */ |
Christoph Hellwig | 7d09525 | 2009-06-08 15:33:32 +0200 | [diff] [blame] | 1645 | if (quotamount) { |
| 1646 | ASSERT(mp->m_qflags == 0); |
| 1647 | mp->m_qflags = quotaflags; |
| 1648 | |
| 1649 | xfs_qm_mount_quotas(mp); |
| 1650 | } |
| 1651 | |
David Chinner | 84e1e99 | 2007-06-18 16:50:27 +1000 | [diff] [blame] | 1652 | /* |
| 1653 | * Now we are mounted, reserve a small amount of unused space for |
| 1654 | * privileged transactions. This is needed so that transaction |
| 1655 | * space required for critical operations can dip into this pool |
| 1656 | * when at ENOSPC. This is needed for operations like create with |
| 1657 | * attr, unwritten extent conversion at ENOSPC, etc. Data allocations |
| 1658 | * are not allowed to use this reserved space. |
Dave Chinner | 8babd8a | 2010-03-04 01:46:25 +0000 | [diff] [blame] | 1659 | * |
| 1660 | * This may drive us straight to ENOSPC on mount, but that implies |
| 1661 | * we were already there on the last unmount. Warn if this occurs. |
David Chinner | 84e1e99 | 2007-06-18 16:50:27 +1000 | [diff] [blame] | 1662 | */ |
Eric Sandeen | d5db0f9 | 2010-02-05 22:59:53 +0000 | [diff] [blame] | 1663 | if (!(mp->m_flags & XFS_MOUNT_RDONLY)) { |
| 1664 | resblks = xfs_default_resblks(mp); |
| 1665 | error = xfs_reserve_blocks(mp, &resblks, NULL); |
| 1666 | if (error) |
Dave Chinner | 0b932cc | 2011-03-07 10:08:35 +1100 | [diff] [blame] | 1667 | xfs_warn(mp, |
| 1668 | "Unable to allocate reserve blocks. Continuing without reserve pool."); |
Eric Sandeen | d5db0f9 | 2010-02-05 22:59:53 +0000 | [diff] [blame] | 1669 | } |
David Chinner | 84e1e99 | 2007-06-18 16:50:27 +1000 | [diff] [blame] | 1670 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1671 | return 0; |
| 1672 | |
Christoph Hellwig | b93b6e4 | 2009-02-04 09:33:58 +0100 | [diff] [blame] | 1673 | out_rtunmount: |
| 1674 | xfs_rtunmount_inodes(mp); |
Christoph Hellwig | f9057e3 | 2009-02-04 09:31:52 +0100 | [diff] [blame] | 1675 | out_rele_rip: |
Christoph Hellwig | 4335509 | 2008-03-27 18:01:08 +1100 | [diff] [blame] | 1676 | IRELE(rip); |
Christoph Hellwig | f9057e3 | 2009-02-04 09:31:52 +0100 | [diff] [blame] | 1677 | out_log_dealloc: |
Christoph Hellwig | 21b699c | 2009-03-16 08:19:29 +0100 | [diff] [blame] | 1678 | xfs_log_unmount(mp); |
Dave Chinner | d4f3512 | 2012-04-23 15:59:06 +1000 | [diff] [blame] | 1679 | out_fail_wait: |
| 1680 | if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) |
| 1681 | xfs_wait_buftarg(mp->m_logdev_targp); |
| 1682 | xfs_wait_buftarg(mp->m_ddev_targp); |
Christoph Hellwig | f9057e3 | 2009-02-04 09:31:52 +0100 | [diff] [blame] | 1683 | out_free_perag: |
Christoph Hellwig | ff4f038 | 2008-08-13 16:50:47 +1000 | [diff] [blame] | 1684 | xfs_free_perag(mp); |
Christoph Hellwig | f9057e3 | 2009-02-04 09:31:52 +0100 | [diff] [blame] | 1685 | out_remove_uuid: |
Christoph Hellwig | 2717420 | 2009-03-30 10:21:31 +0200 | [diff] [blame] | 1686 | xfs_uuid_unmount(mp); |
Christoph Hellwig | f9057e3 | 2009-02-04 09:31:52 +0100 | [diff] [blame] | 1687 | out: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1688 | return error; |
| 1689 | } |
| 1690 | |
| 1691 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1692 | * This flushes out the inodes,dquots and the superblock, unmounts the |
| 1693 | * log and makes sure that incore structures are freed. |
| 1694 | */ |
Christoph Hellwig | 41b5c2e | 2008-08-13 16:49:57 +1000 | [diff] [blame] | 1695 | void |
| 1696 | xfs_unmountfs( |
| 1697 | struct xfs_mount *mp) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1698 | { |
Christoph Hellwig | 41b5c2e | 2008-08-13 16:49:57 +1000 | [diff] [blame] | 1699 | __uint64_t resblks; |
| 1700 | int error; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1701 | |
Brian Foster | 579b62f | 2012-11-06 09:50:47 -0500 | [diff] [blame] | 1702 | cancel_delayed_work_sync(&mp->m_eofblocks_work); |
| 1703 | |
Christoph Hellwig | 7d09525 | 2009-06-08 15:33:32 +0200 | [diff] [blame] | 1704 | xfs_qm_unmount_quotas(mp); |
Christoph Hellwig | b93b6e4 | 2009-02-04 09:33:58 +0100 | [diff] [blame] | 1705 | xfs_rtunmount_inodes(mp); |
Christoph Hellwig | 77508ec | 2008-08-13 16:49:04 +1000 | [diff] [blame] | 1706 | IRELE(mp->m_rootip); |
| 1707 | |
David Chinner | 641c56f | 2007-06-18 16:50:17 +1000 | [diff] [blame] | 1708 | /* |
| 1709 | * We can potentially deadlock here if we have an inode cluster |
Malcolm Parsons | 9da096f | 2009-03-29 09:55:42 +0200 | [diff] [blame] | 1710 | * that has been freed has its buffer still pinned in memory because |
David Chinner | 641c56f | 2007-06-18 16:50:17 +1000 | [diff] [blame] | 1711 | * the transaction is still sitting in a iclog. The stale inodes |
| 1712 | * on that buffer will have their flush locks held until the |
| 1713 | * transaction hits the disk and the callbacks run. the inode |
| 1714 | * flush takes the flush lock unconditionally and with nothing to |
| 1715 | * push out the iclog we will never get that unlocked. hence we |
| 1716 | * need to force the log first. |
| 1717 | */ |
Christoph Hellwig | a14a348 | 2010-01-19 09:56:46 +0000 | [diff] [blame] | 1718 | xfs_log_force(mp, XFS_LOG_SYNC); |
Dave Chinner | c854363 | 2010-02-06 12:39:36 +1100 | [diff] [blame] | 1719 | |
| 1720 | /* |
Christoph Hellwig | 211e4d4 | 2012-04-23 15:58:34 +1000 | [diff] [blame] | 1721 | * Flush all pending changes from the AIL. |
Dave Chinner | c854363 | 2010-02-06 12:39:36 +1100 | [diff] [blame] | 1722 | */ |
Christoph Hellwig | 211e4d4 | 2012-04-23 15:58:34 +1000 | [diff] [blame] | 1723 | xfs_ail_push_all_sync(mp->m_ail); |
| 1724 | |
| 1725 | /* |
| 1726 | * And reclaim all inodes. At this point there should be no dirty |
Dave Chinner | 7e18530 | 2012-10-08 21:56:00 +1100 | [diff] [blame] | 1727 | * inodes and none should be pinned or locked, but use synchronous |
| 1728 | * reclaim just to be sure. We can stop background inode reclaim |
| 1729 | * here as well if it is still running. |
Christoph Hellwig | 211e4d4 | 2012-04-23 15:58:34 +1000 | [diff] [blame] | 1730 | */ |
Dave Chinner | 7e18530 | 2012-10-08 21:56:00 +1100 | [diff] [blame] | 1731 | cancel_delayed_work_sync(&mp->m_reclaim_work); |
Dave Chinner | c854363 | 2010-02-06 12:39:36 +1100 | [diff] [blame] | 1732 | xfs_reclaim_inodes(mp, SYNC_WAIT); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1733 | |
Christoph Hellwig | 7d09525 | 2009-06-08 15:33:32 +0200 | [diff] [blame] | 1734 | xfs_qm_unmount(mp); |
Lachlan McIlroy | a357a12 | 2008-10-30 16:53:25 +1100 | [diff] [blame] | 1735 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1736 | /* |
David Chinner | 84e1e99 | 2007-06-18 16:50:27 +1000 | [diff] [blame] | 1737 | * Unreserve any blocks we have so that when we unmount we don't account |
| 1738 | * the reserved free space as used. This is really only necessary for |
| 1739 | * lazy superblock counting because it trusts the incore superblock |
Malcolm Parsons | 9da096f | 2009-03-29 09:55:42 +0200 | [diff] [blame] | 1740 | * counters to be absolutely correct on clean unmount. |
David Chinner | 84e1e99 | 2007-06-18 16:50:27 +1000 | [diff] [blame] | 1741 | * |
| 1742 | * We don't bother correcting this elsewhere for lazy superblock |
| 1743 | * counting because on mount of an unclean filesystem we reconstruct the |
| 1744 | * correct counter value and this is irrelevant. |
| 1745 | * |
| 1746 | * For non-lazy counter filesystems, this doesn't matter at all because |
| 1747 | * we only every apply deltas to the superblock and hence the incore |
| 1748 | * value does not matter.... |
| 1749 | */ |
| 1750 | resblks = 0; |
David Chinner | 714082b | 2008-04-10 12:20:03 +1000 | [diff] [blame] | 1751 | error = xfs_reserve_blocks(mp, &resblks, NULL); |
| 1752 | if (error) |
Dave Chinner | 0b932cc | 2011-03-07 10:08:35 +1100 | [diff] [blame] | 1753 | xfs_warn(mp, "Unable to free reserved block pool. " |
David Chinner | 714082b | 2008-04-10 12:20:03 +1000 | [diff] [blame] | 1754 | "Freespace may not be correct on next mount."); |
| 1755 | |
Chandra Seetharaman | adab0f6 | 2011-06-29 22:10:14 +0000 | [diff] [blame] | 1756 | error = xfs_log_sbcount(mp); |
David Chinner | e5720ee | 2008-04-10 12:21:18 +1000 | [diff] [blame] | 1757 | if (error) |
Dave Chinner | 0b932cc | 2011-03-07 10:08:35 +1100 | [diff] [blame] | 1758 | xfs_warn(mp, "Unable to update superblock counters. " |
David Chinner | e5720ee | 2008-04-10 12:21:18 +1000 | [diff] [blame] | 1759 | "Freespace may not be correct on next mount."); |
Christoph Hellwig | 87c7bec | 2011-09-14 14:08:26 +0000 | [diff] [blame] | 1760 | |
Christoph Hellwig | 21b699c | 2009-03-16 08:19:29 +0100 | [diff] [blame] | 1761 | xfs_log_unmount(mp); |
Christoph Hellwig | 2717420 | 2009-03-30 10:21:31 +0200 | [diff] [blame] | 1762 | xfs_uuid_unmount(mp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1763 | |
Christoph Hellwig | 1550d0b | 2008-08-13 16:17:37 +1000 | [diff] [blame] | 1764 | #if defined(DEBUG) |
Christoph Hellwig | 0ce4cfd | 2007-08-30 17:20:53 +1000 | [diff] [blame] | 1765 | xfs_errortag_clearall(mp, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1766 | #endif |
Christoph Hellwig | ff4f038 | 2008-08-13 16:50:47 +1000 | [diff] [blame] | 1767 | xfs_free_perag(mp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1768 | } |
| 1769 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1770 | int |
David Chinner | 92821e2 | 2007-05-24 15:26:31 +1000 | [diff] [blame] | 1771 | xfs_fs_writable(xfs_mount_t *mp) |
| 1772 | { |
Jan Kara | d9457dc | 2012-06-12 16:20:39 +0200 | [diff] [blame] | 1773 | return !(mp->m_super->s_writers.frozen || XFS_FORCED_SHUTDOWN(mp) || |
Christoph Hellwig | bd186aa | 2007-08-30 17:21:12 +1000 | [diff] [blame] | 1774 | (mp->m_flags & XFS_MOUNT_RDONLY)); |
David Chinner | 92821e2 | 2007-05-24 15:26:31 +1000 | [diff] [blame] | 1775 | } |
| 1776 | |
| 1777 | /* |
Alex Elder | b2ce397 | 2011-07-11 09:51:44 -0500 | [diff] [blame] | 1778 | * xfs_log_sbcount |
| 1779 | * |
Chandra Seetharaman | adab0f6 | 2011-06-29 22:10:14 +0000 | [diff] [blame] | 1780 | * Sync the superblock counters to disk. |
Alex Elder | b2ce397 | 2011-07-11 09:51:44 -0500 | [diff] [blame] | 1781 | * |
| 1782 | * Note this code can be called during the process of freezing, so |
Chandra Seetharaman | adab0f6 | 2011-06-29 22:10:14 +0000 | [diff] [blame] | 1783 | * we may need to use the transaction allocator which does not |
Alex Elder | b2ce397 | 2011-07-11 09:51:44 -0500 | [diff] [blame] | 1784 | * block when the transaction subsystem is in its frozen state. |
David Chinner | 92821e2 | 2007-05-24 15:26:31 +1000 | [diff] [blame] | 1785 | */ |
| 1786 | int |
Chandra Seetharaman | adab0f6 | 2011-06-29 22:10:14 +0000 | [diff] [blame] | 1787 | xfs_log_sbcount(xfs_mount_t *mp) |
David Chinner | 92821e2 | 2007-05-24 15:26:31 +1000 | [diff] [blame] | 1788 | { |
| 1789 | xfs_trans_t *tp; |
| 1790 | int error; |
| 1791 | |
| 1792 | if (!xfs_fs_writable(mp)) |
| 1793 | return 0; |
| 1794 | |
Christoph Hellwig | d4d90b5 | 2008-04-22 17:34:37 +1000 | [diff] [blame] | 1795 | xfs_icsb_sync_counters(mp, 0); |
David Chinner | 92821e2 | 2007-05-24 15:26:31 +1000 | [diff] [blame] | 1796 | |
| 1797 | /* |
| 1798 | * we don't need to do this if we are updating the superblock |
| 1799 | * counters on every modification. |
| 1800 | */ |
| 1801 | if (!xfs_sb_version_haslazysbcount(&mp->m_sb)) |
| 1802 | return 0; |
| 1803 | |
Alex Elder | b2ce397 | 2011-07-11 09:51:44 -0500 | [diff] [blame] | 1804 | tp = _xfs_trans_alloc(mp, XFS_TRANS_SB_COUNT, KM_SLEEP); |
Jeff Liu | e457274 | 2013-01-28 21:27:31 +0800 | [diff] [blame] | 1805 | error = xfs_trans_reserve(tp, 0, XFS_SB_LOG_RES(mp), 0, 0, |
| 1806 | XFS_DEFAULT_LOG_COUNT); |
David Chinner | 92821e2 | 2007-05-24 15:26:31 +1000 | [diff] [blame] | 1807 | if (error) { |
| 1808 | xfs_trans_cancel(tp, 0); |
| 1809 | return error; |
| 1810 | } |
| 1811 | |
| 1812 | xfs_mod_sb(tp, XFS_SB_IFREE | XFS_SB_ICOUNT | XFS_SB_FDBLOCKS); |
Chandra Seetharaman | adab0f6 | 2011-06-29 22:10:14 +0000 | [diff] [blame] | 1813 | xfs_trans_set_sync(tp); |
David Chinner | e5720ee | 2008-04-10 12:21:18 +1000 | [diff] [blame] | 1814 | error = xfs_trans_commit(tp, 0); |
| 1815 | return error; |
David Chinner | 92821e2 | 2007-05-24 15:26:31 +1000 | [diff] [blame] | 1816 | } |
| 1817 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1818 | /* |
| 1819 | * xfs_mod_sb() can be used to copy arbitrary changes to the |
| 1820 | * in-core superblock into the superblock buffer to be logged. |
| 1821 | * It does not provide the higher level of locking that is |
| 1822 | * needed to protect the in-core superblock from concurrent |
| 1823 | * access. |
| 1824 | */ |
| 1825 | void |
| 1826 | xfs_mod_sb(xfs_trans_t *tp, __int64_t fields) |
| 1827 | { |
| 1828 | xfs_buf_t *bp; |
| 1829 | int first; |
| 1830 | int last; |
| 1831 | xfs_mount_t *mp; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1832 | xfs_sb_field_t f; |
| 1833 | |
| 1834 | ASSERT(fields); |
| 1835 | if (!fields) |
| 1836 | return; |
| 1837 | mp = tp->t_mountp; |
| 1838 | bp = xfs_trans_getsb(tp, mp, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1839 | first = sizeof(xfs_sb_t); |
| 1840 | last = 0; |
| 1841 | |
| 1842 | /* translate/copy */ |
| 1843 | |
Christoph Hellwig | 2bdf7cd | 2007-08-28 13:58:06 +1000 | [diff] [blame] | 1844 | xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, fields); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1845 | |
| 1846 | /* find modified range */ |
Dave Chinner | 587aa0f | 2010-01-20 12:04:53 +1100 | [diff] [blame] | 1847 | f = (xfs_sb_field_t)xfs_highbit64((__uint64_t)fields); |
| 1848 | ASSERT((1LL << f) & XFS_SB_MOD_BITS); |
| 1849 | last = xfs_sb_info[f + 1].offset - 1; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1850 | |
| 1851 | f = (xfs_sb_field_t)xfs_lowbit64((__uint64_t)fields); |
| 1852 | ASSERT((1LL << f) & XFS_SB_MOD_BITS); |
| 1853 | first = xfs_sb_info[f].offset; |
| 1854 | |
Dave Chinner | 04a1e6c | 2013-04-03 16:11:31 +1100 | [diff] [blame] | 1855 | xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SB_BUF); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1856 | xfs_trans_log_buf(tp, bp, first, last); |
| 1857 | } |
Yingping Lu | d210a28 | 2006-06-09 14:55:18 +1000 | [diff] [blame] | 1858 | |
Yingping Lu | d210a28 | 2006-06-09 14:55:18 +1000 | [diff] [blame] | 1859 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1860 | /* |
| 1861 | * xfs_mod_incore_sb_unlocked() is a utility routine common used to apply |
| 1862 | * a delta to a specified field in the in-core superblock. Simply |
| 1863 | * switch on the field indicated and apply the delta to that field. |
| 1864 | * Fields are not allowed to dip below zero, so if the delta would |
| 1865 | * do this do not apply it and return EINVAL. |
| 1866 | * |
Eric Sandeen | 3685c2a | 2007-10-11 17:42:32 +1000 | [diff] [blame] | 1867 | * The m_sb_lock must be held when this routine is called. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1868 | */ |
Eric Sandeen | d96f8f8 | 2009-07-02 00:09:33 -0500 | [diff] [blame] | 1869 | STATIC int |
David Chinner | 20f4ebf | 2007-02-10 18:36:10 +1100 | [diff] [blame] | 1870 | xfs_mod_incore_sb_unlocked( |
| 1871 | xfs_mount_t *mp, |
| 1872 | xfs_sb_field_t field, |
| 1873 | int64_t delta, |
| 1874 | int rsvd) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1875 | { |
| 1876 | int scounter; /* short counter for 32 bit fields */ |
| 1877 | long long lcounter; /* long counter for 64 bit fields */ |
| 1878 | long long res_used, rem; |
| 1879 | |
| 1880 | /* |
| 1881 | * With the in-core superblock spin lock held, switch |
| 1882 | * on the indicated field. Apply the delta to the |
| 1883 | * proper field. If the fields value would dip below |
| 1884 | * 0, then do not apply the delta and return EINVAL. |
| 1885 | */ |
| 1886 | switch (field) { |
| 1887 | case XFS_SBS_ICOUNT: |
| 1888 | lcounter = (long long)mp->m_sb.sb_icount; |
| 1889 | lcounter += delta; |
| 1890 | if (lcounter < 0) { |
| 1891 | ASSERT(0); |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 1892 | return XFS_ERROR(EINVAL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1893 | } |
| 1894 | mp->m_sb.sb_icount = lcounter; |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 1895 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1896 | case XFS_SBS_IFREE: |
| 1897 | lcounter = (long long)mp->m_sb.sb_ifree; |
| 1898 | lcounter += delta; |
| 1899 | if (lcounter < 0) { |
| 1900 | ASSERT(0); |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 1901 | return XFS_ERROR(EINVAL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1902 | } |
| 1903 | mp->m_sb.sb_ifree = lcounter; |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 1904 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1905 | case XFS_SBS_FDBLOCKS: |
David Chinner | 4be536d | 2006-09-07 14:26:50 +1000 | [diff] [blame] | 1906 | lcounter = (long long) |
| 1907 | mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1908 | res_used = (long long)(mp->m_resblks - mp->m_resblks_avail); |
| 1909 | |
| 1910 | if (delta > 0) { /* Putting blocks back */ |
| 1911 | if (res_used > delta) { |
| 1912 | mp->m_resblks_avail += delta; |
| 1913 | } else { |
| 1914 | rem = delta - res_used; |
| 1915 | mp->m_resblks_avail = mp->m_resblks; |
| 1916 | lcounter += rem; |
| 1917 | } |
| 1918 | } else { /* Taking blocks away */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1919 | lcounter += delta; |
Dave Chinner | 8babd8a | 2010-03-04 01:46:25 +0000 | [diff] [blame] | 1920 | if (lcounter >= 0) { |
| 1921 | mp->m_sb.sb_fdblocks = lcounter + |
| 1922 | XFS_ALLOC_SET_ASIDE(mp); |
| 1923 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1924 | } |
Dave Chinner | 8babd8a | 2010-03-04 01:46:25 +0000 | [diff] [blame] | 1925 | |
| 1926 | /* |
| 1927 | * We are out of blocks, use any available reserved |
| 1928 | * blocks if were allowed to. |
| 1929 | */ |
| 1930 | if (!rsvd) |
| 1931 | return XFS_ERROR(ENOSPC); |
| 1932 | |
| 1933 | lcounter = (long long)mp->m_resblks_avail + delta; |
| 1934 | if (lcounter >= 0) { |
| 1935 | mp->m_resblks_avail = lcounter; |
| 1936 | return 0; |
| 1937 | } |
| 1938 | printk_once(KERN_WARNING |
| 1939 | "Filesystem \"%s\": reserve blocks depleted! " |
| 1940 | "Consider increasing reserve pool size.", |
| 1941 | mp->m_fsname); |
| 1942 | return XFS_ERROR(ENOSPC); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1943 | } |
| 1944 | |
David Chinner | 4be536d | 2006-09-07 14:26:50 +1000 | [diff] [blame] | 1945 | mp->m_sb.sb_fdblocks = lcounter + XFS_ALLOC_SET_ASIDE(mp); |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 1946 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1947 | case XFS_SBS_FREXTENTS: |
| 1948 | lcounter = (long long)mp->m_sb.sb_frextents; |
| 1949 | lcounter += delta; |
| 1950 | if (lcounter < 0) { |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 1951 | return XFS_ERROR(ENOSPC); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1952 | } |
| 1953 | mp->m_sb.sb_frextents = lcounter; |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 1954 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1955 | case XFS_SBS_DBLOCKS: |
| 1956 | lcounter = (long long)mp->m_sb.sb_dblocks; |
| 1957 | lcounter += delta; |
| 1958 | if (lcounter < 0) { |
| 1959 | ASSERT(0); |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 1960 | return XFS_ERROR(EINVAL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1961 | } |
| 1962 | mp->m_sb.sb_dblocks = lcounter; |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 1963 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1964 | case XFS_SBS_AGCOUNT: |
| 1965 | scounter = mp->m_sb.sb_agcount; |
| 1966 | scounter += delta; |
| 1967 | if (scounter < 0) { |
| 1968 | ASSERT(0); |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 1969 | return XFS_ERROR(EINVAL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1970 | } |
| 1971 | mp->m_sb.sb_agcount = scounter; |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 1972 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1973 | case XFS_SBS_IMAX_PCT: |
| 1974 | scounter = mp->m_sb.sb_imax_pct; |
| 1975 | scounter += delta; |
| 1976 | if (scounter < 0) { |
| 1977 | ASSERT(0); |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 1978 | return XFS_ERROR(EINVAL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1979 | } |
| 1980 | mp->m_sb.sb_imax_pct = scounter; |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 1981 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1982 | case XFS_SBS_REXTSIZE: |
| 1983 | scounter = mp->m_sb.sb_rextsize; |
| 1984 | scounter += delta; |
| 1985 | if (scounter < 0) { |
| 1986 | ASSERT(0); |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 1987 | return XFS_ERROR(EINVAL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1988 | } |
| 1989 | mp->m_sb.sb_rextsize = scounter; |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 1990 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1991 | case XFS_SBS_RBMBLOCKS: |
| 1992 | scounter = mp->m_sb.sb_rbmblocks; |
| 1993 | scounter += delta; |
| 1994 | if (scounter < 0) { |
| 1995 | ASSERT(0); |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 1996 | return XFS_ERROR(EINVAL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1997 | } |
| 1998 | mp->m_sb.sb_rbmblocks = scounter; |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 1999 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2000 | case XFS_SBS_RBLOCKS: |
| 2001 | lcounter = (long long)mp->m_sb.sb_rblocks; |
| 2002 | lcounter += delta; |
| 2003 | if (lcounter < 0) { |
| 2004 | ASSERT(0); |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 2005 | return XFS_ERROR(EINVAL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2006 | } |
| 2007 | mp->m_sb.sb_rblocks = lcounter; |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 2008 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2009 | case XFS_SBS_REXTENTS: |
| 2010 | lcounter = (long long)mp->m_sb.sb_rextents; |
| 2011 | lcounter += delta; |
| 2012 | if (lcounter < 0) { |
| 2013 | ASSERT(0); |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 2014 | return XFS_ERROR(EINVAL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2015 | } |
| 2016 | mp->m_sb.sb_rextents = lcounter; |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 2017 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2018 | case XFS_SBS_REXTSLOG: |
| 2019 | scounter = mp->m_sb.sb_rextslog; |
| 2020 | scounter += delta; |
| 2021 | if (scounter < 0) { |
| 2022 | ASSERT(0); |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 2023 | return XFS_ERROR(EINVAL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2024 | } |
| 2025 | mp->m_sb.sb_rextslog = scounter; |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 2026 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2027 | default: |
| 2028 | ASSERT(0); |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 2029 | return XFS_ERROR(EINVAL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2030 | } |
| 2031 | } |
| 2032 | |
| 2033 | /* |
| 2034 | * xfs_mod_incore_sb() is used to change a field in the in-core |
| 2035 | * superblock structure by the specified delta. This modification |
Eric Sandeen | 3685c2a | 2007-10-11 17:42:32 +1000 | [diff] [blame] | 2036 | * is protected by the m_sb_lock. Just use the xfs_mod_incore_sb_unlocked() |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2037 | * routine to do the work. |
| 2038 | */ |
| 2039 | int |
David Chinner | 20f4ebf | 2007-02-10 18:36:10 +1100 | [diff] [blame] | 2040 | xfs_mod_incore_sb( |
Christoph Hellwig | 96540c7 | 2010-09-30 02:25:55 +0000 | [diff] [blame] | 2041 | struct xfs_mount *mp, |
| 2042 | xfs_sb_field_t field, |
| 2043 | int64_t delta, |
| 2044 | int rsvd) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2045 | { |
Christoph Hellwig | 96540c7 | 2010-09-30 02:25:55 +0000 | [diff] [blame] | 2046 | int status; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2047 | |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2048 | #ifdef HAVE_PERCPU_SB |
Christoph Hellwig | 96540c7 | 2010-09-30 02:25:55 +0000 | [diff] [blame] | 2049 | ASSERT(field < XFS_SBS_ICOUNT || field > XFS_SBS_FDBLOCKS); |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2050 | #endif |
Christoph Hellwig | 96540c7 | 2010-09-30 02:25:55 +0000 | [diff] [blame] | 2051 | spin_lock(&mp->m_sb_lock); |
| 2052 | status = xfs_mod_incore_sb_unlocked(mp, field, delta, rsvd); |
| 2053 | spin_unlock(&mp->m_sb_lock); |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2054 | |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 2055 | return status; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2056 | } |
| 2057 | |
| 2058 | /* |
Christoph Hellwig | 1b04071 | 2010-09-30 02:25:56 +0000 | [diff] [blame] | 2059 | * Change more than one field in the in-core superblock structure at a time. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2060 | * |
Christoph Hellwig | 1b04071 | 2010-09-30 02:25:56 +0000 | [diff] [blame] | 2061 | * The fields and changes to those fields are specified in the array of |
| 2062 | * xfs_mod_sb structures passed in. Either all of the specified deltas |
| 2063 | * will be applied or none of them will. If any modified field dips below 0, |
| 2064 | * then all modifications will be backed out and EINVAL will be returned. |
| 2065 | * |
| 2066 | * Note that this function may not be used for the superblock values that |
| 2067 | * are tracked with the in-memory per-cpu counters - a direct call to |
| 2068 | * xfs_icsb_modify_counters is required for these. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2069 | */ |
| 2070 | int |
Christoph Hellwig | 1b04071 | 2010-09-30 02:25:56 +0000 | [diff] [blame] | 2071 | xfs_mod_incore_sb_batch( |
| 2072 | struct xfs_mount *mp, |
| 2073 | xfs_mod_sb_t *msb, |
| 2074 | uint nmsb, |
| 2075 | int rsvd) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2076 | { |
David Sterba | 45c51b9 | 2011-04-13 22:03:28 +0000 | [diff] [blame] | 2077 | xfs_mod_sb_t *msbp; |
Christoph Hellwig | 1b04071 | 2010-09-30 02:25:56 +0000 | [diff] [blame] | 2078 | int error = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2079 | |
| 2080 | /* |
Christoph Hellwig | 1b04071 | 2010-09-30 02:25:56 +0000 | [diff] [blame] | 2081 | * Loop through the array of mod structures and apply each individually. |
| 2082 | * If any fail, then back out all those which have already been applied. |
| 2083 | * Do all of this within the scope of the m_sb_lock so that all of the |
| 2084 | * changes will be atomic. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2085 | */ |
Eric Sandeen | 3685c2a | 2007-10-11 17:42:32 +1000 | [diff] [blame] | 2086 | spin_lock(&mp->m_sb_lock); |
David Sterba | 45c51b9 | 2011-04-13 22:03:28 +0000 | [diff] [blame] | 2087 | for (msbp = msb; msbp < (msb + nmsb); msbp++) { |
Christoph Hellwig | 1b04071 | 2010-09-30 02:25:56 +0000 | [diff] [blame] | 2088 | ASSERT(msbp->msb_field < XFS_SBS_ICOUNT || |
| 2089 | msbp->msb_field > XFS_SBS_FDBLOCKS); |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2090 | |
Christoph Hellwig | 1b04071 | 2010-09-30 02:25:56 +0000 | [diff] [blame] | 2091 | error = xfs_mod_incore_sb_unlocked(mp, msbp->msb_field, |
| 2092 | msbp->msb_delta, rsvd); |
| 2093 | if (error) |
| 2094 | goto unwind; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2095 | } |
Eric Sandeen | 3685c2a | 2007-10-11 17:42:32 +1000 | [diff] [blame] | 2096 | spin_unlock(&mp->m_sb_lock); |
Christoph Hellwig | 1b04071 | 2010-09-30 02:25:56 +0000 | [diff] [blame] | 2097 | return 0; |
| 2098 | |
| 2099 | unwind: |
| 2100 | while (--msbp >= msb) { |
| 2101 | error = xfs_mod_incore_sb_unlocked(mp, msbp->msb_field, |
| 2102 | -msbp->msb_delta, rsvd); |
| 2103 | ASSERT(error == 0); |
| 2104 | } |
| 2105 | spin_unlock(&mp->m_sb_lock); |
| 2106 | return error; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2107 | } |
| 2108 | |
| 2109 | /* |
| 2110 | * xfs_getsb() is called to obtain the buffer for the superblock. |
| 2111 | * The buffer is returned locked and read in from disk. |
| 2112 | * The buffer should be released with a call to xfs_brelse(). |
| 2113 | * |
| 2114 | * If the flags parameter is BUF_TRYLOCK, then we'll only return |
| 2115 | * the superblock buffer if it can be locked without sleeping. |
| 2116 | * If it can't then we'll return NULL. |
| 2117 | */ |
Christoph Hellwig | 0c842ad | 2011-07-08 14:36:19 +0200 | [diff] [blame] | 2118 | struct xfs_buf * |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2119 | xfs_getsb( |
Christoph Hellwig | 0c842ad | 2011-07-08 14:36:19 +0200 | [diff] [blame] | 2120 | struct xfs_mount *mp, |
| 2121 | int flags) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2122 | { |
Christoph Hellwig | 0c842ad | 2011-07-08 14:36:19 +0200 | [diff] [blame] | 2123 | struct xfs_buf *bp = mp->m_sb_bp; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2124 | |
Christoph Hellwig | 0c842ad | 2011-07-08 14:36:19 +0200 | [diff] [blame] | 2125 | if (!xfs_buf_trylock(bp)) { |
| 2126 | if (flags & XBF_TRYLOCK) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2127 | return NULL; |
Christoph Hellwig | 0c842ad | 2011-07-08 14:36:19 +0200 | [diff] [blame] | 2128 | xfs_buf_lock(bp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2129 | } |
Christoph Hellwig | 0c842ad | 2011-07-08 14:36:19 +0200 | [diff] [blame] | 2130 | |
Chandra Seetharaman | 72790aa | 2011-07-22 23:40:04 +0000 | [diff] [blame] | 2131 | xfs_buf_hold(bp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2132 | ASSERT(XFS_BUF_ISDONE(bp)); |
Jesper Juhl | 014c254 | 2006-01-15 02:37:08 +0100 | [diff] [blame] | 2133 | return bp; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2134 | } |
| 2135 | |
| 2136 | /* |
| 2137 | * Used to free the superblock along various error paths. |
| 2138 | */ |
| 2139 | void |
| 2140 | xfs_freesb( |
Dave Chinner | 26af655 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 2141 | struct xfs_mount *mp) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2142 | { |
Dave Chinner | 26af655 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 2143 | struct xfs_buf *bp = mp->m_sb_bp; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2144 | |
Dave Chinner | 26af655 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 2145 | xfs_buf_lock(bp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2146 | mp->m_sb_bp = NULL; |
Dave Chinner | 26af655 | 2010-09-22 10:47:20 +1000 | [diff] [blame] | 2147 | xfs_buf_relse(bp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2148 | } |
| 2149 | |
| 2150 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2151 | * Used to log changes to the superblock unit and width fields which could |
Eric Sandeen | e6957ea | 2008-04-10 12:19:34 +1000 | [diff] [blame] | 2152 | * be altered by the mount options, as well as any potential sb_features2 |
| 2153 | * fixup. Only the first superblock is updated. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2154 | */ |
Christoph Hellwig | 7884bc8 | 2009-01-19 02:04:07 +0100 | [diff] [blame] | 2155 | int |
David Chinner | ee1c090 | 2008-03-06 13:45:50 +1100 | [diff] [blame] | 2156 | xfs_mount_log_sb( |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2157 | xfs_mount_t *mp, |
| 2158 | __int64_t fields) |
| 2159 | { |
| 2160 | xfs_trans_t *tp; |
David Chinner | e5720ee | 2008-04-10 12:21:18 +1000 | [diff] [blame] | 2161 | int error; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2162 | |
David Chinner | ee1c090 | 2008-03-06 13:45:50 +1100 | [diff] [blame] | 2163 | ASSERT(fields & (XFS_SB_UNIT | XFS_SB_WIDTH | XFS_SB_UUID | |
David Chinner | 4b166de | 2008-05-20 11:30:27 +1000 | [diff] [blame] | 2164 | XFS_SB_FEATURES2 | XFS_SB_BAD_FEATURES2 | |
| 2165 | XFS_SB_VERSIONNUM)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2166 | |
| 2167 | tp = xfs_trans_alloc(mp, XFS_TRANS_SB_UNIT); |
Jeff Liu | 5166ab0 | 2013-01-28 21:27:39 +0800 | [diff] [blame] | 2168 | error = xfs_trans_reserve(tp, 0, XFS_SB_LOG_RES(mp), 0, 0, |
| 2169 | XFS_DEFAULT_LOG_COUNT); |
David Chinner | e5720ee | 2008-04-10 12:21:18 +1000 | [diff] [blame] | 2170 | if (error) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2171 | xfs_trans_cancel(tp, 0); |
David Chinner | e5720ee | 2008-04-10 12:21:18 +1000 | [diff] [blame] | 2172 | return error; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2173 | } |
| 2174 | xfs_mod_sb(tp, fields); |
David Chinner | e5720ee | 2008-04-10 12:21:18 +1000 | [diff] [blame] | 2175 | error = xfs_trans_commit(tp, 0); |
| 2176 | return error; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2177 | } |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2178 | |
Christoph Hellwig | dda35b8 | 2010-02-15 09:44:46 +0000 | [diff] [blame] | 2179 | /* |
| 2180 | * If the underlying (data/log/rt) device is readonly, there are some |
| 2181 | * operations that cannot proceed. |
| 2182 | */ |
| 2183 | int |
| 2184 | xfs_dev_is_read_only( |
| 2185 | struct xfs_mount *mp, |
| 2186 | char *message) |
| 2187 | { |
| 2188 | if (xfs_readonly_buftarg(mp->m_ddev_targp) || |
| 2189 | xfs_readonly_buftarg(mp->m_logdev_targp) || |
| 2190 | (mp->m_rtdev_targp && xfs_readonly_buftarg(mp->m_rtdev_targp))) { |
Dave Chinner | 0b932cc | 2011-03-07 10:08:35 +1100 | [diff] [blame] | 2191 | xfs_notice(mp, "%s required on read-only device.", message); |
| 2192 | xfs_notice(mp, "write access unavailable, cannot proceed."); |
Christoph Hellwig | dda35b8 | 2010-02-15 09:44:46 +0000 | [diff] [blame] | 2193 | return EROFS; |
| 2194 | } |
| 2195 | return 0; |
| 2196 | } |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2197 | |
| 2198 | #ifdef HAVE_PERCPU_SB |
| 2199 | /* |
| 2200 | * Per-cpu incore superblock counters |
| 2201 | * |
| 2202 | * Simple concept, difficult implementation |
| 2203 | * |
| 2204 | * Basically, replace the incore superblock counters with a distributed per cpu |
| 2205 | * counter for contended fields (e.g. free block count). |
| 2206 | * |
| 2207 | * Difficulties arise in that the incore sb is used for ENOSPC checking, and |
| 2208 | * hence needs to be accurately read when we are running low on space. Hence |
| 2209 | * there is a method to enable and disable the per-cpu counters based on how |
| 2210 | * much "stuff" is available in them. |
| 2211 | * |
| 2212 | * Basically, a counter is enabled if there is enough free resource to justify |
| 2213 | * running a per-cpu fast-path. If the per-cpu counter runs out (i.e. a local |
| 2214 | * ENOSPC), then we disable the counters to synchronise all callers and |
| 2215 | * re-distribute the available resources. |
| 2216 | * |
| 2217 | * If, once we redistributed the available resources, we still get a failure, |
| 2218 | * we disable the per-cpu counter and go through the slow path. |
| 2219 | * |
| 2220 | * The slow path is the current xfs_mod_incore_sb() function. This means that |
Malcolm Parsons | 9da096f | 2009-03-29 09:55:42 +0200 | [diff] [blame] | 2221 | * when we disable a per-cpu counter, we need to drain its resources back to |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2222 | * the global superblock. We do this after disabling the counter to prevent |
| 2223 | * more threads from queueing up on the counter. |
| 2224 | * |
| 2225 | * Essentially, this means that we still need a lock in the fast path to enable |
| 2226 | * synchronisation between the global counters and the per-cpu counters. This |
| 2227 | * is not a problem because the lock will be local to a CPU almost all the time |
| 2228 | * and have little contention except when we get to ENOSPC conditions. |
| 2229 | * |
| 2230 | * Basically, this lock becomes a barrier that enables us to lock out the fast |
| 2231 | * path while we do things like enabling and disabling counters and |
| 2232 | * synchronising the counters. |
| 2233 | * |
| 2234 | * Locking rules: |
| 2235 | * |
Eric Sandeen | 3685c2a | 2007-10-11 17:42:32 +1000 | [diff] [blame] | 2236 | * 1. m_sb_lock before picking up per-cpu locks |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2237 | * 2. per-cpu locks always picked up via for_each_online_cpu() order |
Eric Sandeen | 3685c2a | 2007-10-11 17:42:32 +1000 | [diff] [blame] | 2238 | * 3. accurate counter sync requires m_sb_lock + per cpu locks |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2239 | * 4. modifying per-cpu counters requires holding per-cpu lock |
Eric Sandeen | 3685c2a | 2007-10-11 17:42:32 +1000 | [diff] [blame] | 2240 | * 5. modifying global counters requires holding m_sb_lock |
| 2241 | * 6. enabling or disabling a counter requires holding the m_sb_lock |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2242 | * and _none_ of the per-cpu locks. |
| 2243 | * |
| 2244 | * Disabled counters are only ever re-enabled by a balance operation |
| 2245 | * that results in more free resources per CPU than a given threshold. |
| 2246 | * To ensure counters don't remain disabled, they are rebalanced when |
| 2247 | * the global resource goes above a higher threshold (i.e. some hysteresis |
| 2248 | * is present to prevent thrashing). |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2249 | */ |
David Chinner | e8234a6 | 2006-03-14 13:23:52 +1100 | [diff] [blame] | 2250 | |
Chandra Seetharaman | 5a67e4c | 2006-06-27 02:54:11 -0700 | [diff] [blame] | 2251 | #ifdef CONFIG_HOTPLUG_CPU |
David Chinner | e8234a6 | 2006-03-14 13:23:52 +1100 | [diff] [blame] | 2252 | /* |
| 2253 | * hot-plug CPU notifier support. |
| 2254 | * |
Chandra Seetharaman | 5a67e4c | 2006-06-27 02:54:11 -0700 | [diff] [blame] | 2255 | * We need a notifier per filesystem as we need to be able to identify |
| 2256 | * the filesystem to balance the counters out. This is achieved by |
| 2257 | * having a notifier block embedded in the xfs_mount_t and doing pointer |
| 2258 | * magic to get the mount pointer from the notifier block address. |
David Chinner | e8234a6 | 2006-03-14 13:23:52 +1100 | [diff] [blame] | 2259 | */ |
| 2260 | STATIC int |
| 2261 | xfs_icsb_cpu_notify( |
| 2262 | struct notifier_block *nfb, |
| 2263 | unsigned long action, |
| 2264 | void *hcpu) |
| 2265 | { |
| 2266 | xfs_icsb_cnts_t *cntp; |
| 2267 | xfs_mount_t *mp; |
David Chinner | e8234a6 | 2006-03-14 13:23:52 +1100 | [diff] [blame] | 2268 | |
| 2269 | mp = (xfs_mount_t *)container_of(nfb, xfs_mount_t, m_icsb_notifier); |
| 2270 | cntp = (xfs_icsb_cnts_t *) |
| 2271 | per_cpu_ptr(mp->m_sb_cnts, (unsigned long)hcpu); |
| 2272 | switch (action) { |
| 2273 | case CPU_UP_PREPARE: |
Rafael J. Wysocki | 8bb7844 | 2007-05-09 02:35:10 -0700 | [diff] [blame] | 2274 | case CPU_UP_PREPARE_FROZEN: |
David Chinner | e8234a6 | 2006-03-14 13:23:52 +1100 | [diff] [blame] | 2275 | /* Easy Case - initialize the area and locks, and |
| 2276 | * then rebalance when online does everything else for us. */ |
David Chinner | 01e1b69 | 2006-03-14 13:29:16 +1100 | [diff] [blame] | 2277 | memset(cntp, 0, sizeof(xfs_icsb_cnts_t)); |
David Chinner | e8234a6 | 2006-03-14 13:23:52 +1100 | [diff] [blame] | 2278 | break; |
| 2279 | case CPU_ONLINE: |
Rafael J. Wysocki | 8bb7844 | 2007-05-09 02:35:10 -0700 | [diff] [blame] | 2280 | case CPU_ONLINE_FROZEN: |
David Chinner | 03135cf | 2007-02-10 18:35:15 +1100 | [diff] [blame] | 2281 | xfs_icsb_lock(mp); |
Christoph Hellwig | 45af6c6 | 2008-04-22 17:34:44 +1000 | [diff] [blame] | 2282 | xfs_icsb_balance_counter(mp, XFS_SBS_ICOUNT, 0); |
| 2283 | xfs_icsb_balance_counter(mp, XFS_SBS_IFREE, 0); |
| 2284 | xfs_icsb_balance_counter(mp, XFS_SBS_FDBLOCKS, 0); |
David Chinner | 03135cf | 2007-02-10 18:35:15 +1100 | [diff] [blame] | 2285 | xfs_icsb_unlock(mp); |
David Chinner | e8234a6 | 2006-03-14 13:23:52 +1100 | [diff] [blame] | 2286 | break; |
| 2287 | case CPU_DEAD: |
Rafael J. Wysocki | 8bb7844 | 2007-05-09 02:35:10 -0700 | [diff] [blame] | 2288 | case CPU_DEAD_FROZEN: |
David Chinner | e8234a6 | 2006-03-14 13:23:52 +1100 | [diff] [blame] | 2289 | /* Disable all the counters, then fold the dead cpu's |
| 2290 | * count into the total on the global superblock and |
| 2291 | * re-enable the counters. */ |
David Chinner | 03135cf | 2007-02-10 18:35:15 +1100 | [diff] [blame] | 2292 | xfs_icsb_lock(mp); |
Eric Sandeen | 3685c2a | 2007-10-11 17:42:32 +1000 | [diff] [blame] | 2293 | spin_lock(&mp->m_sb_lock); |
David Chinner | e8234a6 | 2006-03-14 13:23:52 +1100 | [diff] [blame] | 2294 | xfs_icsb_disable_counter(mp, XFS_SBS_ICOUNT); |
| 2295 | xfs_icsb_disable_counter(mp, XFS_SBS_IFREE); |
| 2296 | xfs_icsb_disable_counter(mp, XFS_SBS_FDBLOCKS); |
| 2297 | |
| 2298 | mp->m_sb.sb_icount += cntp->icsb_icount; |
| 2299 | mp->m_sb.sb_ifree += cntp->icsb_ifree; |
| 2300 | mp->m_sb.sb_fdblocks += cntp->icsb_fdblocks; |
| 2301 | |
David Chinner | 01e1b69 | 2006-03-14 13:29:16 +1100 | [diff] [blame] | 2302 | memset(cntp, 0, sizeof(xfs_icsb_cnts_t)); |
David Chinner | e8234a6 | 2006-03-14 13:23:52 +1100 | [diff] [blame] | 2303 | |
Christoph Hellwig | 45af6c6 | 2008-04-22 17:34:44 +1000 | [diff] [blame] | 2304 | xfs_icsb_balance_counter_locked(mp, XFS_SBS_ICOUNT, 0); |
| 2305 | xfs_icsb_balance_counter_locked(mp, XFS_SBS_IFREE, 0); |
| 2306 | xfs_icsb_balance_counter_locked(mp, XFS_SBS_FDBLOCKS, 0); |
Eric Sandeen | 3685c2a | 2007-10-11 17:42:32 +1000 | [diff] [blame] | 2307 | spin_unlock(&mp->m_sb_lock); |
David Chinner | 03135cf | 2007-02-10 18:35:15 +1100 | [diff] [blame] | 2308 | xfs_icsb_unlock(mp); |
David Chinner | e8234a6 | 2006-03-14 13:23:52 +1100 | [diff] [blame] | 2309 | break; |
| 2310 | } |
| 2311 | |
| 2312 | return NOTIFY_OK; |
| 2313 | } |
Chandra Seetharaman | 5a67e4c | 2006-06-27 02:54:11 -0700 | [diff] [blame] | 2314 | #endif /* CONFIG_HOTPLUG_CPU */ |
David Chinner | e8234a6 | 2006-03-14 13:23:52 +1100 | [diff] [blame] | 2315 | |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2316 | int |
| 2317 | xfs_icsb_init_counters( |
| 2318 | xfs_mount_t *mp) |
| 2319 | { |
| 2320 | xfs_icsb_cnts_t *cntp; |
| 2321 | int i; |
| 2322 | |
| 2323 | mp->m_sb_cnts = alloc_percpu(xfs_icsb_cnts_t); |
| 2324 | if (mp->m_sb_cnts == NULL) |
| 2325 | return -ENOMEM; |
| 2326 | |
Chandra Seetharaman | 5a67e4c | 2006-06-27 02:54:11 -0700 | [diff] [blame] | 2327 | #ifdef CONFIG_HOTPLUG_CPU |
David Chinner | e8234a6 | 2006-03-14 13:23:52 +1100 | [diff] [blame] | 2328 | mp->m_icsb_notifier.notifier_call = xfs_icsb_cpu_notify; |
| 2329 | mp->m_icsb_notifier.priority = 0; |
Chandra Seetharaman | 5a67e4c | 2006-06-27 02:54:11 -0700 | [diff] [blame] | 2330 | register_hotcpu_notifier(&mp->m_icsb_notifier); |
| 2331 | #endif /* CONFIG_HOTPLUG_CPU */ |
David Chinner | e8234a6 | 2006-03-14 13:23:52 +1100 | [diff] [blame] | 2332 | |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2333 | for_each_online_cpu(i) { |
| 2334 | cntp = (xfs_icsb_cnts_t *)per_cpu_ptr(mp->m_sb_cnts, i); |
David Chinner | 01e1b69 | 2006-03-14 13:29:16 +1100 | [diff] [blame] | 2335 | memset(cntp, 0, sizeof(xfs_icsb_cnts_t)); |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2336 | } |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2337 | |
| 2338 | mutex_init(&mp->m_icsb_mutex); |
| 2339 | |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2340 | /* |
| 2341 | * start with all counters disabled so that the |
| 2342 | * initial balance kicks us off correctly |
| 2343 | */ |
| 2344 | mp->m_icsb_counters = -1; |
| 2345 | return 0; |
| 2346 | } |
| 2347 | |
Lachlan McIlroy | 5478eea | 2007-02-10 18:36:29 +1100 | [diff] [blame] | 2348 | void |
| 2349 | xfs_icsb_reinit_counters( |
| 2350 | xfs_mount_t *mp) |
| 2351 | { |
| 2352 | xfs_icsb_lock(mp); |
| 2353 | /* |
| 2354 | * start with all counters disabled so that the |
| 2355 | * initial balance kicks us off correctly |
| 2356 | */ |
| 2357 | mp->m_icsb_counters = -1; |
Christoph Hellwig | 45af6c6 | 2008-04-22 17:34:44 +1000 | [diff] [blame] | 2358 | xfs_icsb_balance_counter(mp, XFS_SBS_ICOUNT, 0); |
| 2359 | xfs_icsb_balance_counter(mp, XFS_SBS_IFREE, 0); |
| 2360 | xfs_icsb_balance_counter(mp, XFS_SBS_FDBLOCKS, 0); |
Lachlan McIlroy | 5478eea | 2007-02-10 18:36:29 +1100 | [diff] [blame] | 2361 | xfs_icsb_unlock(mp); |
| 2362 | } |
| 2363 | |
Christoph Hellwig | c962fb7 | 2008-05-20 15:10:52 +1000 | [diff] [blame] | 2364 | void |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2365 | xfs_icsb_destroy_counters( |
| 2366 | xfs_mount_t *mp) |
| 2367 | { |
David Chinner | e8234a6 | 2006-03-14 13:23:52 +1100 | [diff] [blame] | 2368 | if (mp->m_sb_cnts) { |
Chandra Seetharaman | 5a67e4c | 2006-06-27 02:54:11 -0700 | [diff] [blame] | 2369 | unregister_hotcpu_notifier(&mp->m_icsb_notifier); |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2370 | free_percpu(mp->m_sb_cnts); |
David Chinner | e8234a6 | 2006-03-14 13:23:52 +1100 | [diff] [blame] | 2371 | } |
David Chinner | 03135cf | 2007-02-10 18:35:15 +1100 | [diff] [blame] | 2372 | mutex_destroy(&mp->m_icsb_mutex); |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2373 | } |
| 2374 | |
Christoph Hellwig | b8f82a4 | 2009-11-14 16:17:22 +0000 | [diff] [blame] | 2375 | STATIC void |
David Chinner | 01e1b69 | 2006-03-14 13:29:16 +1100 | [diff] [blame] | 2376 | xfs_icsb_lock_cntr( |
| 2377 | xfs_icsb_cnts_t *icsbp) |
| 2378 | { |
| 2379 | while (test_and_set_bit(XFS_ICSB_FLAG_LOCK, &icsbp->icsb_flags)) { |
| 2380 | ndelay(1000); |
| 2381 | } |
| 2382 | } |
| 2383 | |
Christoph Hellwig | b8f82a4 | 2009-11-14 16:17:22 +0000 | [diff] [blame] | 2384 | STATIC void |
David Chinner | 01e1b69 | 2006-03-14 13:29:16 +1100 | [diff] [blame] | 2385 | xfs_icsb_unlock_cntr( |
| 2386 | xfs_icsb_cnts_t *icsbp) |
| 2387 | { |
| 2388 | clear_bit(XFS_ICSB_FLAG_LOCK, &icsbp->icsb_flags); |
| 2389 | } |
| 2390 | |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2391 | |
Christoph Hellwig | b8f82a4 | 2009-11-14 16:17:22 +0000 | [diff] [blame] | 2392 | STATIC void |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2393 | xfs_icsb_lock_all_counters( |
| 2394 | xfs_mount_t *mp) |
| 2395 | { |
| 2396 | xfs_icsb_cnts_t *cntp; |
| 2397 | int i; |
| 2398 | |
| 2399 | for_each_online_cpu(i) { |
| 2400 | cntp = (xfs_icsb_cnts_t *)per_cpu_ptr(mp->m_sb_cnts, i); |
David Chinner | 01e1b69 | 2006-03-14 13:29:16 +1100 | [diff] [blame] | 2401 | xfs_icsb_lock_cntr(cntp); |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2402 | } |
| 2403 | } |
| 2404 | |
Christoph Hellwig | b8f82a4 | 2009-11-14 16:17:22 +0000 | [diff] [blame] | 2405 | STATIC void |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2406 | xfs_icsb_unlock_all_counters( |
| 2407 | xfs_mount_t *mp) |
| 2408 | { |
| 2409 | xfs_icsb_cnts_t *cntp; |
| 2410 | int i; |
| 2411 | |
| 2412 | for_each_online_cpu(i) { |
| 2413 | cntp = (xfs_icsb_cnts_t *)per_cpu_ptr(mp->m_sb_cnts, i); |
David Chinner | 01e1b69 | 2006-03-14 13:29:16 +1100 | [diff] [blame] | 2414 | xfs_icsb_unlock_cntr(cntp); |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2415 | } |
| 2416 | } |
| 2417 | |
| 2418 | STATIC void |
| 2419 | xfs_icsb_count( |
| 2420 | xfs_mount_t *mp, |
| 2421 | xfs_icsb_cnts_t *cnt, |
| 2422 | int flags) |
| 2423 | { |
| 2424 | xfs_icsb_cnts_t *cntp; |
| 2425 | int i; |
| 2426 | |
| 2427 | memset(cnt, 0, sizeof(xfs_icsb_cnts_t)); |
| 2428 | |
| 2429 | if (!(flags & XFS_ICSB_LAZY_COUNT)) |
| 2430 | xfs_icsb_lock_all_counters(mp); |
| 2431 | |
| 2432 | for_each_online_cpu(i) { |
| 2433 | cntp = (xfs_icsb_cnts_t *)per_cpu_ptr(mp->m_sb_cnts, i); |
| 2434 | cnt->icsb_icount += cntp->icsb_icount; |
| 2435 | cnt->icsb_ifree += cntp->icsb_ifree; |
| 2436 | cnt->icsb_fdblocks += cntp->icsb_fdblocks; |
| 2437 | } |
| 2438 | |
| 2439 | if (!(flags & XFS_ICSB_LAZY_COUNT)) |
| 2440 | xfs_icsb_unlock_all_counters(mp); |
| 2441 | } |
| 2442 | |
| 2443 | STATIC int |
| 2444 | xfs_icsb_counter_disabled( |
| 2445 | xfs_mount_t *mp, |
| 2446 | xfs_sb_field_t field) |
| 2447 | { |
| 2448 | ASSERT((field >= XFS_SBS_ICOUNT) && (field <= XFS_SBS_FDBLOCKS)); |
| 2449 | return test_bit(field, &mp->m_icsb_counters); |
| 2450 | } |
| 2451 | |
David Chinner | 36fbe6e | 2008-04-10 12:19:56 +1000 | [diff] [blame] | 2452 | STATIC void |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2453 | xfs_icsb_disable_counter( |
| 2454 | xfs_mount_t *mp, |
| 2455 | xfs_sb_field_t field) |
| 2456 | { |
| 2457 | xfs_icsb_cnts_t cnt; |
| 2458 | |
| 2459 | ASSERT((field >= XFS_SBS_ICOUNT) && (field <= XFS_SBS_FDBLOCKS)); |
| 2460 | |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2461 | /* |
| 2462 | * If we are already disabled, then there is nothing to do |
| 2463 | * here. We check before locking all the counters to avoid |
| 2464 | * the expensive lock operation when being called in the |
| 2465 | * slow path and the counter is already disabled. This is |
| 2466 | * safe because the only time we set or clear this state is under |
| 2467 | * the m_icsb_mutex. |
| 2468 | */ |
| 2469 | if (xfs_icsb_counter_disabled(mp, field)) |
David Chinner | 36fbe6e | 2008-04-10 12:19:56 +1000 | [diff] [blame] | 2470 | return; |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2471 | |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2472 | xfs_icsb_lock_all_counters(mp); |
| 2473 | if (!test_and_set_bit(field, &mp->m_icsb_counters)) { |
| 2474 | /* drain back to superblock */ |
| 2475 | |
Christoph Hellwig | ce46193 | 2008-04-22 17:34:50 +1000 | [diff] [blame] | 2476 | xfs_icsb_count(mp, &cnt, XFS_ICSB_LAZY_COUNT); |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2477 | switch(field) { |
| 2478 | case XFS_SBS_ICOUNT: |
| 2479 | mp->m_sb.sb_icount = cnt.icsb_icount; |
| 2480 | break; |
| 2481 | case XFS_SBS_IFREE: |
| 2482 | mp->m_sb.sb_ifree = cnt.icsb_ifree; |
| 2483 | break; |
| 2484 | case XFS_SBS_FDBLOCKS: |
| 2485 | mp->m_sb.sb_fdblocks = cnt.icsb_fdblocks; |
| 2486 | break; |
| 2487 | default: |
| 2488 | BUG(); |
| 2489 | } |
| 2490 | } |
| 2491 | |
| 2492 | xfs_icsb_unlock_all_counters(mp); |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2493 | } |
| 2494 | |
| 2495 | STATIC void |
| 2496 | xfs_icsb_enable_counter( |
| 2497 | xfs_mount_t *mp, |
| 2498 | xfs_sb_field_t field, |
| 2499 | uint64_t count, |
| 2500 | uint64_t resid) |
| 2501 | { |
| 2502 | xfs_icsb_cnts_t *cntp; |
| 2503 | int i; |
| 2504 | |
| 2505 | ASSERT((field >= XFS_SBS_ICOUNT) && (field <= XFS_SBS_FDBLOCKS)); |
| 2506 | |
| 2507 | xfs_icsb_lock_all_counters(mp); |
| 2508 | for_each_online_cpu(i) { |
| 2509 | cntp = per_cpu_ptr(mp->m_sb_cnts, i); |
| 2510 | switch (field) { |
| 2511 | case XFS_SBS_ICOUNT: |
| 2512 | cntp->icsb_icount = count + resid; |
| 2513 | break; |
| 2514 | case XFS_SBS_IFREE: |
| 2515 | cntp->icsb_ifree = count + resid; |
| 2516 | break; |
| 2517 | case XFS_SBS_FDBLOCKS: |
| 2518 | cntp->icsb_fdblocks = count + resid; |
| 2519 | break; |
| 2520 | default: |
| 2521 | BUG(); |
| 2522 | break; |
| 2523 | } |
| 2524 | resid = 0; |
| 2525 | } |
| 2526 | clear_bit(field, &mp->m_icsb_counters); |
| 2527 | xfs_icsb_unlock_all_counters(mp); |
| 2528 | } |
| 2529 | |
David Chinner | dbcabad | 2007-02-10 18:36:17 +1100 | [diff] [blame] | 2530 | void |
Christoph Hellwig | d4d90b5 | 2008-04-22 17:34:37 +1000 | [diff] [blame] | 2531 | xfs_icsb_sync_counters_locked( |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2532 | xfs_mount_t *mp, |
| 2533 | int flags) |
| 2534 | { |
| 2535 | xfs_icsb_cnts_t cnt; |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2536 | |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2537 | xfs_icsb_count(mp, &cnt, flags); |
| 2538 | |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2539 | if (!xfs_icsb_counter_disabled(mp, XFS_SBS_ICOUNT)) |
| 2540 | mp->m_sb.sb_icount = cnt.icsb_icount; |
| 2541 | if (!xfs_icsb_counter_disabled(mp, XFS_SBS_IFREE)) |
| 2542 | mp->m_sb.sb_ifree = cnt.icsb_ifree; |
| 2543 | if (!xfs_icsb_counter_disabled(mp, XFS_SBS_FDBLOCKS)) |
| 2544 | mp->m_sb.sb_fdblocks = cnt.icsb_fdblocks; |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2545 | } |
| 2546 | |
| 2547 | /* |
| 2548 | * Accurate update of per-cpu counters to incore superblock |
| 2549 | */ |
Christoph Hellwig | d4d90b5 | 2008-04-22 17:34:37 +1000 | [diff] [blame] | 2550 | void |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2551 | xfs_icsb_sync_counters( |
Christoph Hellwig | d4d90b5 | 2008-04-22 17:34:37 +1000 | [diff] [blame] | 2552 | xfs_mount_t *mp, |
| 2553 | int flags) |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2554 | { |
Christoph Hellwig | d4d90b5 | 2008-04-22 17:34:37 +1000 | [diff] [blame] | 2555 | spin_lock(&mp->m_sb_lock); |
| 2556 | xfs_icsb_sync_counters_locked(mp, flags); |
| 2557 | spin_unlock(&mp->m_sb_lock); |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2558 | } |
| 2559 | |
| 2560 | /* |
| 2561 | * Balance and enable/disable counters as necessary. |
| 2562 | * |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2563 | * Thresholds for re-enabling counters are somewhat magic. inode counts are |
| 2564 | * chosen to be the same number as single on disk allocation chunk per CPU, and |
| 2565 | * free blocks is something far enough zero that we aren't going thrash when we |
| 2566 | * get near ENOSPC. We also need to supply a minimum we require per cpu to |
| 2567 | * prevent looping endlessly when xfs_alloc_space asks for more than will |
| 2568 | * be distributed to a single CPU but each CPU has enough blocks to be |
| 2569 | * reenabled. |
| 2570 | * |
| 2571 | * Note that we can be called when counters are already disabled. |
| 2572 | * xfs_icsb_disable_counter() optimises the counter locking in this case to |
| 2573 | * prevent locking every per-cpu counter needlessly. |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2574 | */ |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2575 | |
| 2576 | #define XFS_ICSB_INO_CNTR_REENABLE (uint64_t)64 |
David Chinner | 4be536d | 2006-09-07 14:26:50 +1000 | [diff] [blame] | 2577 | #define XFS_ICSB_FDBLK_CNTR_REENABLE(mp) \ |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2578 | (uint64_t)(512 + XFS_ALLOC_SET_ASIDE(mp)) |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2579 | STATIC void |
Christoph Hellwig | 45af6c6 | 2008-04-22 17:34:44 +1000 | [diff] [blame] | 2580 | xfs_icsb_balance_counter_locked( |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2581 | xfs_mount_t *mp, |
| 2582 | xfs_sb_field_t field, |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2583 | int min_per_cpu) |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2584 | { |
Nathan Scott | 6fdf8cc | 2006-06-28 10:13:52 +1000 | [diff] [blame] | 2585 | uint64_t count, resid; |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2586 | int weight = num_online_cpus(); |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2587 | uint64_t min = (uint64_t)min_per_cpu; |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2588 | |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2589 | /* disable counter and sync counter */ |
| 2590 | xfs_icsb_disable_counter(mp, field); |
| 2591 | |
| 2592 | /* update counters - first CPU gets residual*/ |
| 2593 | switch (field) { |
| 2594 | case XFS_SBS_ICOUNT: |
| 2595 | count = mp->m_sb.sb_icount; |
| 2596 | resid = do_div(count, weight); |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2597 | if (count < max(min, XFS_ICSB_INO_CNTR_REENABLE)) |
Christoph Hellwig | 45af6c6 | 2008-04-22 17:34:44 +1000 | [diff] [blame] | 2598 | return; |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2599 | break; |
| 2600 | case XFS_SBS_IFREE: |
| 2601 | count = mp->m_sb.sb_ifree; |
| 2602 | resid = do_div(count, weight); |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2603 | if (count < max(min, XFS_ICSB_INO_CNTR_REENABLE)) |
Christoph Hellwig | 45af6c6 | 2008-04-22 17:34:44 +1000 | [diff] [blame] | 2604 | return; |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2605 | break; |
| 2606 | case XFS_SBS_FDBLOCKS: |
| 2607 | count = mp->m_sb.sb_fdblocks; |
| 2608 | resid = do_div(count, weight); |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2609 | if (count < max(min, XFS_ICSB_FDBLK_CNTR_REENABLE(mp))) |
Christoph Hellwig | 45af6c6 | 2008-04-22 17:34:44 +1000 | [diff] [blame] | 2610 | return; |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2611 | break; |
| 2612 | default: |
| 2613 | BUG(); |
Nathan Scott | 6fdf8cc | 2006-06-28 10:13:52 +1000 | [diff] [blame] | 2614 | count = resid = 0; /* quiet, gcc */ |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2615 | break; |
| 2616 | } |
| 2617 | |
| 2618 | xfs_icsb_enable_counter(mp, field, count, resid); |
Christoph Hellwig | 45af6c6 | 2008-04-22 17:34:44 +1000 | [diff] [blame] | 2619 | } |
| 2620 | |
| 2621 | STATIC void |
| 2622 | xfs_icsb_balance_counter( |
| 2623 | xfs_mount_t *mp, |
| 2624 | xfs_sb_field_t fields, |
| 2625 | int min_per_cpu) |
| 2626 | { |
| 2627 | spin_lock(&mp->m_sb_lock); |
| 2628 | xfs_icsb_balance_counter_locked(mp, fields, min_per_cpu); |
| 2629 | spin_unlock(&mp->m_sb_lock); |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2630 | } |
| 2631 | |
Christoph Hellwig | 1b04071 | 2010-09-30 02:25:56 +0000 | [diff] [blame] | 2632 | int |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2633 | xfs_icsb_modify_counters( |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2634 | xfs_mount_t *mp, |
| 2635 | xfs_sb_field_t field, |
David Chinner | 20f4ebf | 2007-02-10 18:36:10 +1100 | [diff] [blame] | 2636 | int64_t delta, |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2637 | int rsvd) |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2638 | { |
| 2639 | xfs_icsb_cnts_t *icsbp; |
| 2640 | long long lcounter; /* long counter for 64 bit fields */ |
Christoph Lameter | 7a9e02d | 2009-10-03 19:48:23 +0900 | [diff] [blame] | 2641 | int ret = 0; |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2642 | |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2643 | might_sleep(); |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2644 | again: |
Christoph Lameter | 7a9e02d | 2009-10-03 19:48:23 +0900 | [diff] [blame] | 2645 | preempt_disable(); |
| 2646 | icsbp = this_cpu_ptr(mp->m_sb_cnts); |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2647 | |
| 2648 | /* |
| 2649 | * if the counter is disabled, go to slow path |
| 2650 | */ |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2651 | if (unlikely(xfs_icsb_counter_disabled(mp, field))) |
| 2652 | goto slow_path; |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2653 | xfs_icsb_lock_cntr(icsbp); |
| 2654 | if (unlikely(xfs_icsb_counter_disabled(mp, field))) { |
| 2655 | xfs_icsb_unlock_cntr(icsbp); |
| 2656 | goto slow_path; |
| 2657 | } |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2658 | |
| 2659 | switch (field) { |
| 2660 | case XFS_SBS_ICOUNT: |
| 2661 | lcounter = icsbp->icsb_icount; |
| 2662 | lcounter += delta; |
| 2663 | if (unlikely(lcounter < 0)) |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2664 | goto balance_counter; |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2665 | icsbp->icsb_icount = lcounter; |
| 2666 | break; |
| 2667 | |
| 2668 | case XFS_SBS_IFREE: |
| 2669 | lcounter = icsbp->icsb_ifree; |
| 2670 | lcounter += delta; |
| 2671 | if (unlikely(lcounter < 0)) |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2672 | goto balance_counter; |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2673 | icsbp->icsb_ifree = lcounter; |
| 2674 | break; |
| 2675 | |
| 2676 | case XFS_SBS_FDBLOCKS: |
| 2677 | BUG_ON((mp->m_resblks - mp->m_resblks_avail) != 0); |
| 2678 | |
David Chinner | 4be536d | 2006-09-07 14:26:50 +1000 | [diff] [blame] | 2679 | lcounter = icsbp->icsb_fdblocks - XFS_ALLOC_SET_ASIDE(mp); |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2680 | lcounter += delta; |
| 2681 | if (unlikely(lcounter < 0)) |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2682 | goto balance_counter; |
David Chinner | 4be536d | 2006-09-07 14:26:50 +1000 | [diff] [blame] | 2683 | icsbp->icsb_fdblocks = lcounter + XFS_ALLOC_SET_ASIDE(mp); |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2684 | break; |
| 2685 | default: |
| 2686 | BUG(); |
| 2687 | break; |
| 2688 | } |
David Chinner | 01e1b69 | 2006-03-14 13:29:16 +1100 | [diff] [blame] | 2689 | xfs_icsb_unlock_cntr(icsbp); |
Christoph Lameter | 7a9e02d | 2009-10-03 19:48:23 +0900 | [diff] [blame] | 2690 | preempt_enable(); |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2691 | return 0; |
| 2692 | |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2693 | slow_path: |
Christoph Lameter | 7a9e02d | 2009-10-03 19:48:23 +0900 | [diff] [blame] | 2694 | preempt_enable(); |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2695 | |
| 2696 | /* |
| 2697 | * serialise with a mutex so we don't burn lots of cpu on |
| 2698 | * the superblock lock. We still need to hold the superblock |
| 2699 | * lock, however, when we modify the global structures. |
| 2700 | */ |
David Chinner | 03135cf | 2007-02-10 18:35:15 +1100 | [diff] [blame] | 2701 | xfs_icsb_lock(mp); |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2702 | |
| 2703 | /* |
| 2704 | * Now running atomically. |
| 2705 | * |
| 2706 | * If the counter is enabled, someone has beaten us to rebalancing. |
| 2707 | * Drop the lock and try again in the fast path.... |
| 2708 | */ |
| 2709 | if (!(xfs_icsb_counter_disabled(mp, field))) { |
David Chinner | 03135cf | 2007-02-10 18:35:15 +1100 | [diff] [blame] | 2710 | xfs_icsb_unlock(mp); |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2711 | goto again; |
| 2712 | } |
| 2713 | |
| 2714 | /* |
| 2715 | * The counter is currently disabled. Because we are |
| 2716 | * running atomically here, we know a rebalance cannot |
| 2717 | * be in progress. Hence we can go straight to operating |
| 2718 | * on the global superblock. We do not call xfs_mod_incore_sb() |
Eric Sandeen | 3685c2a | 2007-10-11 17:42:32 +1000 | [diff] [blame] | 2719 | * here even though we need to get the m_sb_lock. Doing so |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2720 | * will cause us to re-enter this function and deadlock. |
Eric Sandeen | 3685c2a | 2007-10-11 17:42:32 +1000 | [diff] [blame] | 2721 | * Hence we get the m_sb_lock ourselves and then call |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2722 | * xfs_mod_incore_sb_unlocked() as the unlocked path operates |
| 2723 | * directly on the global counters. |
| 2724 | */ |
Eric Sandeen | 3685c2a | 2007-10-11 17:42:32 +1000 | [diff] [blame] | 2725 | spin_lock(&mp->m_sb_lock); |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2726 | ret = xfs_mod_incore_sb_unlocked(mp, field, delta, rsvd); |
Eric Sandeen | 3685c2a | 2007-10-11 17:42:32 +1000 | [diff] [blame] | 2727 | spin_unlock(&mp->m_sb_lock); |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2728 | |
| 2729 | /* |
| 2730 | * Now that we've modified the global superblock, we |
| 2731 | * may be able to re-enable the distributed counters |
| 2732 | * (e.g. lots of space just got freed). After that |
| 2733 | * we are done. |
| 2734 | */ |
| 2735 | if (ret != ENOSPC) |
Christoph Hellwig | 45af6c6 | 2008-04-22 17:34:44 +1000 | [diff] [blame] | 2736 | xfs_icsb_balance_counter(mp, field, 0); |
David Chinner | 03135cf | 2007-02-10 18:35:15 +1100 | [diff] [blame] | 2737 | xfs_icsb_unlock(mp); |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2738 | return ret; |
| 2739 | |
| 2740 | balance_counter: |
David Chinner | 01e1b69 | 2006-03-14 13:29:16 +1100 | [diff] [blame] | 2741 | xfs_icsb_unlock_cntr(icsbp); |
Christoph Lameter | 7a9e02d | 2009-10-03 19:48:23 +0900 | [diff] [blame] | 2742 | preempt_enable(); |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2743 | |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2744 | /* |
| 2745 | * We may have multiple threads here if multiple per-cpu |
| 2746 | * counters run dry at the same time. This will mean we can |
| 2747 | * do more balances than strictly necessary but it is not |
| 2748 | * the common slowpath case. |
| 2749 | */ |
David Chinner | 03135cf | 2007-02-10 18:35:15 +1100 | [diff] [blame] | 2750 | xfs_icsb_lock(mp); |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2751 | |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2752 | /* |
| 2753 | * running atomically. |
| 2754 | * |
| 2755 | * This will leave the counter in the correct state for future |
| 2756 | * accesses. After the rebalance, we simply try again and our retry |
| 2757 | * will either succeed through the fast path or slow path without |
| 2758 | * another balance operation being required. |
| 2759 | */ |
Christoph Hellwig | 45af6c6 | 2008-04-22 17:34:44 +1000 | [diff] [blame] | 2760 | xfs_icsb_balance_counter(mp, field, delta); |
David Chinner | 03135cf | 2007-02-10 18:35:15 +1100 | [diff] [blame] | 2761 | xfs_icsb_unlock(mp); |
David Chinner | 20b6428 | 2007-02-10 18:35:09 +1100 | [diff] [blame] | 2762 | goto again; |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2763 | } |
| 2764 | |
David Chinner | 8d280b9 | 2006-03-14 13:13:09 +1100 | [diff] [blame] | 2765 | #endif |