1 /* 2 * Copyright (c) 2000-2003,2005 Silicon Graphics, Inc. 3 * All Rights Reserved. 4 * 5 * This program is free software; you can redistribute it and/or 6 * modify it under the terms of the GNU General Public License as 7 * published by the Free Software Foundation. 8 * 9 * This program is distributed in the hope that it would be useful, 10 * but WITHOUT ANY WARRANTY; without even the implied warranty of 11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 12 * GNU General Public License for more details. 13 * 14 * You should have received a copy of the GNU General Public License 15 * along with this program; if not, write the Free Software Foundation, 16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA 17 */ 18 #ifndef __XFS_AG_H__ 19 #define __XFS_AG_H__ 20 21 /* 22 * Allocation group header 23 * This is divided into three structures, placed in sequential 512-byte 24 * buffers after a copy of the superblock (also in a 512-byte buffer). 25 */ 26 27 struct xfs_buf; 28 struct xfs_mount; 29 struct xfs_trans; 30 31 #define XFS_AGF_MAGIC 0x58414746 /* 'XAGF' */ 32 #define XFS_AGI_MAGIC 0x58414749 /* 'XAGI' */ 33 #define XFS_AGFL_MAGIC 0x5841464c /* 'XAFL' */ 34 #define XFS_AGF_VERSION 1 35 #define XFS_AGI_VERSION 1 36 37 #define XFS_AGF_GOOD_VERSION(v) ((v) == XFS_AGF_VERSION) 38 #define XFS_AGI_GOOD_VERSION(v) ((v) == XFS_AGI_VERSION) 39 40 /* 41 * Btree number 0 is bno, 1 is cnt. This value gives the size of the 42 * arrays below. 43 */ 44 #define XFS_BTNUM_AGF ((int)XFS_BTNUM_CNTi + 1) 45 46 /* 47 * The second word of agf_levels in the first a.g. overlaps the EFS 48 * superblock's magic number. Since the magic numbers valid for EFS 49 * are > 64k, our value cannot be confused for an EFS superblock's. 50 */ 51 52 typedef struct xfs_agf { 53 /* 54 * Common allocation group header information 55 */ 56 __be32 agf_magicnum; /* magic number == XFS_AGF_MAGIC */ 57 __be32 agf_versionnum; /* header version == XFS_AGF_VERSION */ 58 __be32 agf_seqno; /* sequence # starting from 0 */ 59 __be32 agf_length; /* size in blocks of a.g. */ 60 /* 61 * Freespace information 62 */ 63 __be32 agf_roots[XFS_BTNUM_AGF]; /* root blocks */ 64 __be32 agf_spare0; /* spare field */ 65 __be32 agf_levels[XFS_BTNUM_AGF]; /* btree levels */ 66 __be32 agf_spare1; /* spare field */ 67 68 __be32 agf_flfirst; /* first freelist block's index */ 69 __be32 agf_fllast; /* last freelist block's index */ 70 __be32 agf_flcount; /* count of blocks in freelist */ 71 __be32 agf_freeblks; /* total free blocks */ 72 73 __be32 agf_longest; /* longest free space */ 74 __be32 agf_btreeblks; /* # of blocks held in AGF btrees */ 75 uuid_t agf_uuid; /* uuid of filesystem */ 76 77 /* 78 * reserve some contiguous space for future logged fields before we add 79 * the unlogged fields. This makes the range logging via flags and 80 * structure offsets much simpler. 81 */ 82 __be64 agf_spare64[16]; 83 84 /* unlogged fields, written during buffer writeback. */ 85 __be64 agf_lsn; /* last write sequence */ 86 __be32 agf_crc; /* crc of agf sector */ 87 __be32 agf_spare2; 88 89 /* structure must be padded to 64 bit alignment */ 90 } xfs_agf_t; 91 92 #define XFS_AGF_MAGICNUM 0x00000001 93 #define XFS_AGF_VERSIONNUM 0x00000002 94 #define XFS_AGF_SEQNO 0x00000004 95 #define XFS_AGF_LENGTH 0x00000008 96 #define XFS_AGF_ROOTS 0x00000010 97 #define XFS_AGF_LEVELS 0x00000020 98 #define XFS_AGF_FLFIRST 0x00000040 99 #define XFS_AGF_FLLAST 0x00000080 100 #define XFS_AGF_FLCOUNT 0x00000100 101 #define XFS_AGF_FREEBLKS 0x00000200 102 #define XFS_AGF_LONGEST 0x00000400 103 #define XFS_AGF_BTREEBLKS 0x00000800 104 #define XFS_AGF_UUID 0x00001000 105 #define XFS_AGF_NUM_BITS 13 106 #define XFS_AGF_ALL_BITS ((1 << XFS_AGF_NUM_BITS) - 1) 107 108 #define XFS_AGF_FLAGS \ 109 { XFS_AGF_MAGICNUM, "MAGICNUM" }, \ 110 { XFS_AGF_VERSIONNUM, "VERSIONNUM" }, \ 111 { XFS_AGF_SEQNO, "SEQNO" }, \ 112 { XFS_AGF_LENGTH, "LENGTH" }, \ 113 { XFS_AGF_ROOTS, "ROOTS" }, \ 114 { XFS_AGF_LEVELS, "LEVELS" }, \ 115 { XFS_AGF_FLFIRST, "FLFIRST" }, \ 116 { XFS_AGF_FLLAST, "FLLAST" }, \ 117 { XFS_AGF_FLCOUNT, "FLCOUNT" }, \ 118 { XFS_AGF_FREEBLKS, "FREEBLKS" }, \ 119 { XFS_AGF_LONGEST, "LONGEST" }, \ 120 { XFS_AGF_BTREEBLKS, "BTREEBLKS" }, \ 121 { XFS_AGF_UUID, "UUID" } 122 123 /* disk block (xfs_daddr_t) in the AG */ 124 #define XFS_AGF_DADDR(mp) ((xfs_daddr_t)(1 << (mp)->m_sectbb_log)) 125 #define XFS_AGF_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_AGF_DADDR(mp)) 126 #define XFS_BUF_TO_AGF(bp) ((xfs_agf_t *)((bp)->b_addr)) 127 128 extern int xfs_read_agf(struct xfs_mount *mp, struct xfs_trans *tp, 129 xfs_agnumber_t agno, int flags, struct xfs_buf **bpp); 130 131 extern const struct xfs_buf_ops xfs_agf_buf_ops; 132 133 /* 134 * Size of the unlinked inode hash table in the agi. 135 */ 136 #define XFS_AGI_UNLINKED_BUCKETS 64 137 138 typedef struct xfs_agi { 139 /* 140 * Common allocation group header information 141 */ 142 __be32 agi_magicnum; /* magic number == XFS_AGI_MAGIC */ 143 __be32 agi_versionnum; /* header version == XFS_AGI_VERSION */ 144 __be32 agi_seqno; /* sequence # starting from 0 */ 145 __be32 agi_length; /* size in blocks of a.g. */ 146 /* 147 * Inode information 148 * Inodes are mapped by interpreting the inode number, so no 149 * mapping data is needed here. 150 */ 151 __be32 agi_count; /* count of allocated inodes */ 152 __be32 agi_root; /* root of inode btree */ 153 __be32 agi_level; /* levels in inode btree */ 154 __be32 agi_freecount; /* number of free inodes */ 155 156 __be32 agi_newino; /* new inode just allocated */ 157 __be32 agi_dirino; /* last directory inode chunk */ 158 /* 159 * Hash table of inodes which have been unlinked but are 160 * still being referenced. 161 */ 162 __be32 agi_unlinked[XFS_AGI_UNLINKED_BUCKETS]; 163 164 uuid_t agi_uuid; /* uuid of filesystem */ 165 __be32 agi_crc; /* crc of agi sector */ 166 __be32 agi_pad32; 167 __be64 agi_lsn; /* last write sequence */ 168 169 /* structure must be padded to 64 bit alignment */ 170 } xfs_agi_t; 171 172 #define XFS_AGI_MAGICNUM 0x00000001 173 #define XFS_AGI_VERSIONNUM 0x00000002 174 #define XFS_AGI_SEQNO 0x00000004 175 #define XFS_AGI_LENGTH 0x00000008 176 #define XFS_AGI_COUNT 0x00000010 177 #define XFS_AGI_ROOT 0x00000020 178 #define XFS_AGI_LEVEL 0x00000040 179 #define XFS_AGI_FREECOUNT 0x00000080 180 #define XFS_AGI_NEWINO 0x00000100 181 #define XFS_AGI_DIRINO 0x00000200 182 #define XFS_AGI_UNLINKED 0x00000400 183 #define XFS_AGI_NUM_BITS 11 184 #define XFS_AGI_ALL_BITS ((1 << XFS_AGI_NUM_BITS) - 1) 185 186 /* disk block (xfs_daddr_t) in the AG */ 187 #define XFS_AGI_DADDR(mp) ((xfs_daddr_t)(2 << (mp)->m_sectbb_log)) 188 #define XFS_AGI_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_AGI_DADDR(mp)) 189 #define XFS_BUF_TO_AGI(bp) ((xfs_agi_t *)((bp)->b_addr)) 190 191 extern int xfs_read_agi(struct xfs_mount *mp, struct xfs_trans *tp, 192 xfs_agnumber_t agno, struct xfs_buf **bpp); 193 194 extern const struct xfs_buf_ops xfs_agi_buf_ops; 195 196 /* 197 * The third a.g. block contains the a.g. freelist, an array 198 * of block pointers to blocks owned by the allocation btree code. 199 */ 200 #define XFS_AGFL_DADDR(mp) ((xfs_daddr_t)(3 << (mp)->m_sectbb_log)) 201 #define XFS_AGFL_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_AGFL_DADDR(mp)) 202 #define XFS_BUF_TO_AGFL(bp) ((xfs_agfl_t *)((bp)->b_addr)) 203 204 #define XFS_BUF_TO_AGFL_BNO(mp, bp) \ 205 (xfs_sb_version_hascrc(&((mp)->m_sb)) ? \ 206 &(XFS_BUF_TO_AGFL(bp)->agfl_bno[0]) : \ 207 (__be32 *)(bp)->b_addr) 208 209 /* 210 * Size of the AGFL. For CRC-enabled filesystes we steal a couple of 211 * slots in the beginning of the block for a proper header with the 212 * location information and CRC. 213 */ 214 #define XFS_AGFL_SIZE(mp) \ 215 (((mp)->m_sb.sb_sectsize - \ 216 (xfs_sb_version_hascrc(&((mp)->m_sb)) ? \ 217 sizeof(struct xfs_agfl) : 0)) / \ 218 sizeof(xfs_agblock_t)) 219 220 typedef struct xfs_agfl { 221 __be32 agfl_magicnum; 222 __be32 agfl_seqno; 223 uuid_t agfl_uuid; 224 __be64 agfl_lsn; 225 __be32 agfl_crc; 226 __be32 agfl_bno[]; /* actually XFS_AGFL_SIZE(mp) */ 227 } xfs_agfl_t; 228 229 /* 230 * Per-ag incore structure, copies of information in agf and agi, 231 * to improve the performance of allocation group selection. 232 */ 233 #define XFS_PAGB_NUM_SLOTS 128 234 235 typedef struct xfs_perag { 236 struct xfs_mount *pag_mount; /* owner filesystem */ 237 xfs_agnumber_t pag_agno; /* AG this structure belongs to */ 238 atomic_t pag_ref; /* perag reference count */ 239 char pagf_init; /* this agf's entry is initialized */ 240 char pagi_init; /* this agi's entry is initialized */ 241 char pagf_metadata; /* the agf is preferred to be metadata */ 242 char pagi_inodeok; /* The agi is ok for inodes */ 243 __uint8_t pagf_levels[XFS_BTNUM_AGF]; 244 /* # of levels in bno & cnt btree */ 245 __uint32_t pagf_flcount; /* count of blocks in freelist */ 246 xfs_extlen_t pagf_freeblks; /* total free blocks */ 247 xfs_extlen_t pagf_longest; /* longest free space */ 248 __uint32_t pagf_btreeblks; /* # of blocks held in AGF btrees */ 249 xfs_agino_t pagi_freecount; /* number of free inodes */ 250 xfs_agino_t pagi_count; /* number of allocated inodes */ 251 252 /* 253 * Inode allocation search lookup optimisation. 254 * If the pagino matches, the search for new inodes 255 * doesn't need to search the near ones again straight away 256 */ 257 xfs_agino_t pagl_pagino; 258 xfs_agino_t pagl_leftrec; 259 xfs_agino_t pagl_rightrec; 260 #ifdef __KERNEL__ 261 spinlock_t pagb_lock; /* lock for pagb_tree */ 262 struct rb_root pagb_tree; /* ordered tree of busy extents */ 263 264 atomic_t pagf_fstrms; /* # of filestreams active in this AG */ 265 266 spinlock_t pag_ici_lock; /* incore inode cache lock */ 267 struct radix_tree_root pag_ici_root; /* incore inode cache root */ 268 int pag_ici_reclaimable; /* reclaimable inodes */ 269 struct mutex pag_ici_reclaim_lock; /* serialisation point */ 270 unsigned long pag_ici_reclaim_cursor; /* reclaim restart point */ 271 272 /* buffer cache index */ 273 spinlock_t pag_buf_lock; /* lock for pag_buf_tree */ 274 struct rb_root pag_buf_tree; /* ordered tree of active buffers */ 275 276 /* for rcu-safe freeing */ 277 struct rcu_head rcu_head; 278 #endif 279 int pagb_count; /* pagb slots in use */ 280 } xfs_perag_t; 281 282 /* 283 * tags for inode radix tree 284 */ 285 #define XFS_ICI_NO_TAG (-1) /* special flag for an untagged lookup 286 in xfs_inode_ag_iterator */ 287 #define XFS_ICI_RECLAIM_TAG 0 /* inode is to be reclaimed */ 288 #define XFS_ICI_EOFBLOCKS_TAG 1 /* inode has blocks beyond EOF */ 289 290 #define XFS_AG_MAXLEVELS(mp) ((mp)->m_ag_maxlevels) 291 #define XFS_MIN_FREELIST_RAW(bl,cl,mp) \ 292 (MIN(bl + 1, XFS_AG_MAXLEVELS(mp)) + MIN(cl + 1, XFS_AG_MAXLEVELS(mp))) 293 #define XFS_MIN_FREELIST(a,mp) \ 294 (XFS_MIN_FREELIST_RAW( \ 295 be32_to_cpu((a)->agf_levels[XFS_BTNUM_BNOi]), \ 296 be32_to_cpu((a)->agf_levels[XFS_BTNUM_CNTi]), mp)) 297 #define XFS_MIN_FREELIST_PAG(pag,mp) \ 298 (XFS_MIN_FREELIST_RAW( \ 299 (unsigned int)(pag)->pagf_levels[XFS_BTNUM_BNOi], \ 300 (unsigned int)(pag)->pagf_levels[XFS_BTNUM_CNTi], mp)) 301 302 #define XFS_AGB_TO_FSB(mp,agno,agbno) \ 303 (((xfs_fsblock_t)(agno) << (mp)->m_sb.sb_agblklog) | (agbno)) 304 #define XFS_FSB_TO_AGNO(mp,fsbno) \ 305 ((xfs_agnumber_t)((fsbno) >> (mp)->m_sb.sb_agblklog)) 306 #define XFS_FSB_TO_AGBNO(mp,fsbno) \ 307 ((xfs_agblock_t)((fsbno) & xfs_mask32lo((mp)->m_sb.sb_agblklog))) 308 #define XFS_AGB_TO_DADDR(mp,agno,agbno) \ 309 ((xfs_daddr_t)XFS_FSB_TO_BB(mp, \ 310 (xfs_fsblock_t)(agno) * (mp)->m_sb.sb_agblocks + (agbno))) 311 #define XFS_AG_DADDR(mp,agno,d) (XFS_AGB_TO_DADDR(mp, agno, 0) + (d)) 312 313 /* 314 * For checking for bad ranges of xfs_daddr_t's, covering multiple 315 * allocation groups or a single xfs_daddr_t that's a superblock copy. 316 */ 317 #define XFS_AG_CHECK_DADDR(mp,d,len) \ 318 ((len) == 1 ? \ 319 ASSERT((d) == XFS_SB_DADDR || \ 320 xfs_daddr_to_agbno(mp, d) != XFS_SB_DADDR) : \ 321 ASSERT(xfs_daddr_to_agno(mp, d) == \ 322 xfs_daddr_to_agno(mp, (d) + (len) - 1))) 323 324 #endif /* __XFS_AG_H__ */ 325