fs/xfs/xfs_mount.h

   1 /*
   2  * Copyright (c) 2000-2005 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_MOUNT_H__
  19 #define __XFS_MOUNT_H__
  20
  21 #ifdef __KERNEL__
  22
  23 struct xlog;
  24 struct xfs_inode;
  25 struct xfs_mru_cache;
  26 struct xfs_nameops;
  27 struct xfs_ail;
  28 struct xfs_quotainfo;
  29 struct xfs_dir_ops;
  30 struct xfs_da_geometry;
  31
  32 #ifdef HAVE_PERCPU_SB
  33
  34 /*
  35  * Valid per-cpu incore superblock counters. Note that if you add new counters,
  36  * you may need to define new counter disabled bit field descriptors as there
  37  * are more possible fields in the superblock that can fit in a bitfield on a
  38  * 32 bit platform. The XFS_SBS_* values for the current current counters just
  39  * fit.
  40  */
  41 typedef struct xfs_icsb_cnts {
  42         uint64_t        icsb_fdblocks;
  43         uint64_t        icsb_ifree;
  44         unsigned long   icsb_flags;
  45 } xfs_icsb_cnts_t;
  46
  47 #define XFS_ICSB_FLAG_LOCK      (1 << 0)        /* counter lock bit */
  48
  49 #define XFS_ICSB_LAZY_COUNT     (1 << 1)        /* accuracy not needed */
  50
  51 extern int      xfs_icsb_init_counters(struct xfs_mount *);
  52 extern void     xfs_icsb_reinit_counters(struct xfs_mount *);
  53 extern void     xfs_icsb_destroy_counters(struct xfs_mount *);
  54 extern void     xfs_icsb_sync_counters(struct xfs_mount *, int);
  55 extern void     xfs_icsb_sync_counters_locked(struct xfs_mount *, int);
  56 extern int      xfs_icsb_modify_counters(struct xfs_mount *, xfs_sb_field_t,
  57                                                 int64_t, int);
  58
  59 #else
  60 #define xfs_icsb_init_counters(mp)              (0)
  61 #define xfs_icsb_destroy_counters(mp)           do { } while (0)
  62 #define xfs_icsb_reinit_counters(mp)            do { } while (0)
  63 #define xfs_icsb_sync_counters(mp, flags)       do { } while (0)
  64 #define xfs_icsb_sync_counters_locked(mp, flags) do { } while (0)
  65 #define xfs_icsb_modify_counters(mp, field, delta, rsvd) \
  66         xfs_mod_incore_sb(mp, field, delta, rsvd)
  67 #endif
  68
  69 /* dynamic preallocation free space thresholds, 5% down to 1% */
  70 enum {
  71         XFS_LOWSP_1_PCNT = 0,
  72         XFS_LOWSP_2_PCNT,
  73         XFS_LOWSP_3_PCNT,
  74         XFS_LOWSP_4_PCNT,
  75         XFS_LOWSP_5_PCNT,
  76         XFS_LOWSP_MAX,
  77 };
  78
  79 typedef struct xfs_mount {
  80         struct super_block      *m_super;
  81         xfs_tid_t               m_tid;          /* next unused tid for fs */
  82         struct xfs_ail          *m_ail;         /* fs active log item list */
  83
  84         struct xfs_sb           m_sb;           /* copy of fs superblock */
  85         spinlock_t              m_sb_lock;      /* sb counter lock */
  86         struct percpu_counter   m_icount;       /* allocated inodes counter */
  87         struct percpu_counter   m_ifree;        /* free inodes counter */
  88         struct percpu_counter   m_fdblocks;     /* free block counter */
  89
  90         struct xfs_buf          *m_sb_bp;       /* buffer for superblock */
  91         char                    *m_fsname;      /* filesystem name */
  92         int                     m_fsname_len;   /* strlen of fs name */
  93         char                    *m_rtname;      /* realtime device name */
  94         char                    *m_logname;     /* external log device name */
  95         int                     m_bsize;        /* fs logical block size */
  96         xfs_agnumber_t          m_agfrotor;     /* last ag where space found */
  97         xfs_agnumber_t          m_agirotor;     /* last ag dir inode alloced */
  98         spinlock_t              m_agirotor_lock;/* .. and lock protecting it */
  99         xfs_agnumber_t          m_maxagi;       /* highest inode alloc group */
 100         uint                    m_readio_log;   /* min read size log bytes */
 101         uint                    m_readio_blocks; /* min read size blocks */
 102         uint                    m_writeio_log;  /* min write size log bytes */
 103         uint                    m_writeio_blocks; /* min write size blocks */
 104         struct xfs_da_geometry  *m_dir_geo;     /* directory block geometry */
 105         struct xfs_da_geometry  *m_attr_geo;    /* attribute block geometry */
 106         struct xlog             *m_log;         /* log specific stuff */
 107         int                     m_logbufs;      /* number of log buffers */
 108         int                     m_logbsize;     /* size of each log buffer */
 109         uint                    m_rsumlevels;   /* rt summary levels */
 110         uint                    m_rsumsize;     /* size of rt summary, bytes */
 111         struct xfs_inode        *m_rbmip;       /* pointer to bitmap inode */
 112         struct xfs_inode        *m_rsumip;      /* pointer to summary inode */
 113         struct xfs_inode        *m_rootip;      /* pointer to root directory */
 114         struct xfs_quotainfo    *m_quotainfo;   /* disk quota information */
 115         xfs_buftarg_t           *m_ddev_targp;  /* saves taking the address */
 116         xfs_buftarg_t           *m_logdev_targp;/* ptr to log device */
 117         xfs_buftarg_t           *m_rtdev_targp; /* ptr to rt device */
 118         __uint8_t               m_blkbit_log;   /* blocklog + NBBY */
 119         __uint8_t               m_blkbb_log;    /* blocklog - BBSHIFT */
 120         __uint8_t               m_agno_log;     /* log #ag's */
 121         __uint8_t               m_agino_log;    /* #bits for agino in inum */
 122         uint                    m_inode_cluster_size;/* min inode buf size */
 123         uint                    m_blockmask;    /* sb_blocksize-1 */
 124         uint                    m_blockwsize;   /* sb_blocksize in words */
 125         uint                    m_blockwmask;   /* blockwsize-1 */
 126         uint                    m_alloc_mxr[2]; /* max alloc btree records */
 127         uint                    m_alloc_mnr[2]; /* min alloc btree records */
 128         uint                    m_bmap_dmxr[2]; /* max bmap btree records */
 129         uint                    m_bmap_dmnr[2]; /* min bmap btree records */
 130         uint                    m_inobt_mxr[2]; /* max inobt btree records */
 131         uint                    m_inobt_mnr[2]; /* min inobt btree records */
 132         uint                    m_ag_maxlevels; /* XFS_AG_MAXLEVELS */
 133         uint                    m_bm_maxlevels[2]; /* XFS_BM_MAXLEVELS */
 134         uint                    m_in_maxlevels; /* max inobt btree levels. */
 135         struct radix_tree_root  m_perag_tree;   /* per-ag accounting info */
 136         spinlock_t              m_perag_lock;   /* lock for m_perag_tree */
 137         struct mutex            m_growlock;     /* growfs mutex */
 138         int                     m_fixedfsid[2]; /* unchanged for life of FS */
 139         uint                    m_dmevmask;     /* DMI events for this FS */
 140         __uint64_t              m_flags;        /* global mount flags */
 141         int                     m_ialloc_inos;  /* inodes in inode allocation */
 142         int                     m_ialloc_blks;  /* blocks in inode allocation */
 143         int                     m_inoalign_mask;/* mask sb_inoalignmt if used */
 144         uint                    m_qflags;       /* quota status flags */
 145         struct xfs_trans_resv   m_resv;         /* precomputed res values */
 146         __uint64_t              m_maxicount;    /* maximum inode count */
 147         __uint64_t              m_resblks;      /* total reserved blocks */
 148         __uint64_t              m_resblks_avail;/* available reserved blocks */
 149         __uint64_t              m_resblks_save; /* reserved blks @ remount,ro */
 150         int                     m_dalign;       /* stripe unit */
 151         int                     m_swidth;       /* stripe width */
 152         int                     m_sinoalign;    /* stripe unit inode alignment */
 153         __uint8_t               m_sectbb_log;   /* sectlog - BBSHIFT */
 154         const struct xfs_nameops *m_dirnameops; /* vector of dir name ops */
 155         const struct xfs_dir_ops *m_dir_inode_ops; /* vector of dir inode ops */
 156         const struct xfs_dir_ops *m_nondir_inode_ops; /* !dir inode ops */
 157         uint                    m_chsize;       /* size of next field */
 158         atomic_t                m_active_trans; /* number trans frozen */
 159 #ifdef HAVE_PERCPU_SB
 160         xfs_icsb_cnts_t __percpu *m_sb_cnts;    /* per-cpu superblock counters */
 161         unsigned long           m_icsb_counters; /* disabled per-cpu counters */
 162         struct notifier_block   m_icsb_notifier; /* hotplug cpu notifier */
 163         struct mutex            m_icsb_mutex;   /* balancer sync lock */
 164 #endif
 165         struct xfs_mru_cache    *m_filestream;  /* per-mount filestream data */
 166         struct delayed_work     m_reclaim_work; /* background inode reclaim */
 167         struct delayed_work     m_eofblocks_work; /* background eof blocks
 168                                                      trimming */
 169         bool                    m_update_sb;    /* sb needs update in mount */
 170         int64_t                 m_low_space[XFS_LOWSP_MAX];
 171                                                 /* low free space thresholds */
 172         struct xfs_kobj         m_kobj;
 173
 174         struct workqueue_struct *m_buf_workqueue;
 175         struct workqueue_struct *m_data_workqueue;
 176         struct workqueue_struct *m_unwritten_workqueue;
 177         struct workqueue_struct *m_cil_workqueue;
 178         struct workqueue_struct *m_reclaim_workqueue;
 179         struct workqueue_struct *m_log_workqueue;
 180         struct workqueue_struct *m_eofblocks_workqueue;
 181
 182         /*
 183          * Generation of the filesysyem layout.  This is incremented by each
 184          * growfs, and used by the pNFS server to ensure the client updates
 185          * its view of the block device once it gets a layout that might
 186          * reference the newly added blocks.  Does not need to be persistent
 187          * as long as we only allow file system size increments, but if we
 188          * ever support shrinks it would have to be persisted in addition
 189          * to various other kinds of pain inflicted on the pNFS server.
 190          */
 191         __uint32_t              m_generation;
 192 } xfs_mount_t;
 193
 194 /*
 195  * Flags for m_flags.
 196  */
 197 #define XFS_MOUNT_WSYNC         (1ULL << 0)     /* for nfs - all metadata ops
 198                                                    must be synchronous except
 199                                                    for space allocations */
 200 #define XFS_MOUNT_WAS_CLEAN     (1ULL << 3)
 201 #define XFS_MOUNT_FS_SHUTDOWN   (1ULL << 4)     /* atomic stop of all filesystem
 202                                                    operations, typically for
 203                                                    disk errors in metadata */
 204 #define XFS_MOUNT_DISCARD       (1ULL << 5)     /* discard unused blocks */
 205 #define XFS_MOUNT_NOALIGN       (1ULL << 7)     /* turn off stripe alignment
 206                                                    allocations */
 207 #define XFS_MOUNT_ATTR2         (1ULL << 8)     /* allow use of attr2 format */
 208 #define XFS_MOUNT_GRPID         (1ULL << 9)     /* group-ID assigned from directory */
 209 #define XFS_MOUNT_NORECOVERY    (1ULL << 10)    /* no recovery - dirty fs */
 210 #define XFS_MOUNT_DFLT_IOSIZE   (1ULL << 12)    /* set default i/o size */
 211 #define XFS_MOUNT_32BITINODES   (1ULL << 14)    /* do not create inodes above
 212                                                  * 32 bits in size */
 213 #define XFS_MOUNT_SMALL_INUMS   (1ULL << 15)    /* users wants 32bit inodes */
 214 #define XFS_MOUNT_NOUUID        (1ULL << 16)    /* ignore uuid during mount */
 215 #define XFS_MOUNT_BARRIER       (1ULL << 17)
 216 #define XFS_MOUNT_IKEEP         (1ULL << 18)    /* keep empty inode clusters*/
 217 #define XFS_MOUNT_SWALLOC       (1ULL << 19)    /* turn on stripe width
 218                                                  * allocation */
 219 #define XFS_MOUNT_RDONLY        (1ULL << 20)    /* read-only fs */
 220 #define XFS_MOUNT_DIRSYNC       (1ULL << 21)    /* synchronous directory ops */
 221 #define XFS_MOUNT_COMPAT_IOSIZE (1ULL << 22)    /* don't report large preferred
 222                                                  * I/O size in stat() */
 223 #define XFS_MOUNT_FILESTREAMS   (1ULL << 24)    /* enable the filestreams
 224                                                    allocator */
 225 #define XFS_MOUNT_NOATTR2       (1ULL << 25)    /* disable use of attr2 format */
 226
 227
 228 /*
 229  * Default minimum read and write sizes.
 230  */
 231 #define XFS_READIO_LOG_LARGE    16
 232 #define XFS_WRITEIO_LOG_LARGE   16
 233
 234 /*
 235  * Max and min values for mount-option defined I/O
 236  * preallocation sizes.
 237  */
 238 #define XFS_MAX_IO_LOG          30      /* 1G */
 239 #define XFS_MIN_IO_LOG          PAGE_SHIFT
 240
 241 /*
 242  * Synchronous read and write sizes.  This should be
 243  * better for NFSv2 wsync filesystems.
 244  */
 245 #define XFS_WSYNC_READIO_LOG    15      /* 32k */
 246 #define XFS_WSYNC_WRITEIO_LOG   14      /* 16k */
 247
 248 /*
 249  * Allow large block sizes to be reported to userspace programs if the
 250  * "largeio" mount option is used.
 251  *
 252  * If compatibility mode is specified, simply return the basic unit of caching
 253  * so that we don't get inefficient read/modify/write I/O from user apps.
 254  * Otherwise....
 255  *
 256  * If the underlying volume is a stripe, then return the stripe width in bytes
 257  * as the recommended I/O size. It is not a stripe and we've set a default
 258  * buffered I/O size, return that, otherwise return the compat default.
 259  */
 260 static inline unsigned long
 261 xfs_preferred_iosize(xfs_mount_t *mp)
 262 {
 263         if (mp->m_flags & XFS_MOUNT_COMPAT_IOSIZE)
 264                 return PAGE_CACHE_SIZE;
 265         return (mp->m_swidth ?
 266                 (mp->m_swidth << mp->m_sb.sb_blocklog) :
 267                 ((mp->m_flags & XFS_MOUNT_DFLT_IOSIZE) ?
 268                         (1 << (int)MAX(mp->m_readio_log, mp->m_writeio_log)) :
 269                         PAGE_CACHE_SIZE));
 270 }
 271
 272 #define XFS_LAST_UNMOUNT_WAS_CLEAN(mp)  \
 273                                 ((mp)->m_flags & XFS_MOUNT_WAS_CLEAN)
 274 #define XFS_FORCED_SHUTDOWN(mp) ((mp)->m_flags & XFS_MOUNT_FS_SHUTDOWN)
 275 void xfs_do_force_shutdown(struct xfs_mount *mp, int flags, char *fname,
 276                 int lnnum);
 277 #define xfs_force_shutdown(m,f) \
 278         xfs_do_force_shutdown(m, f, __FILE__, __LINE__)
 279
 280 #define SHUTDOWN_META_IO_ERROR  0x0001  /* write attempt to metadata failed */
 281 #define SHUTDOWN_LOG_IO_ERROR   0x0002  /* write attempt to the log failed */
 282 #define SHUTDOWN_FORCE_UMOUNT   0x0004  /* shutdown from a forced unmount */
 283 #define SHUTDOWN_CORRUPT_INCORE 0x0008  /* corrupt in-memory data structures */
 284 #define SHUTDOWN_REMOTE_REQ     0x0010  /* shutdown came from remote cell */
 285 #define SHUTDOWN_DEVICE_REQ     0x0020  /* failed all paths to the device */
 286
 287 /*
 288  * Flags for xfs_mountfs
 289  */
 290 #define XFS_MFSI_QUIET          0x40    /* Be silent if mount errors found */
 291
 292 static inline xfs_agnumber_t
 293 xfs_daddr_to_agno(struct xfs_mount *mp, xfs_daddr_t d)
 294 {
 295         xfs_daddr_t ld = XFS_BB_TO_FSBT(mp, d);
 296         do_div(ld, mp->m_sb.sb_agblocks);
 297         return (xfs_agnumber_t) ld;
 298 }
 299
 300 static inline xfs_agblock_t
 301 xfs_daddr_to_agbno(struct xfs_mount *mp, xfs_daddr_t d)
 302 {
 303         xfs_daddr_t ld = XFS_BB_TO_FSBT(mp, d);
 304         return (xfs_agblock_t) do_div(ld, mp->m_sb.sb_agblocks);
 305 }
 306
 307 /*
 308  * Per-cpu superblock locking functions
 309  */
 310 #ifdef HAVE_PERCPU_SB
 311 static inline void
 312 xfs_icsb_lock(xfs_mount_t *mp)
 313 {
 314         mutex_lock(&mp->m_icsb_mutex);
 315 }
 316
 317 static inline void
 318 xfs_icsb_unlock(xfs_mount_t *mp)
 319 {
 320         mutex_unlock(&mp->m_icsb_mutex);
 321 }
 322 #else
 323 #define xfs_icsb_lock(mp)
 324 #define xfs_icsb_unlock(mp)
 325 #endif
 326
 327 /*
 328  * This structure is for use by the xfs_mod_incore_sb_batch() routine.
 329  * xfs_growfs can specify a few fields which are more than int limit
 330  */
 331 typedef struct xfs_mod_sb {
 332         xfs_sb_field_t  msb_field;      /* Field to modify, see below */
 333         int64_t         msb_delta;      /* Change to make to specified field */
 334 } xfs_mod_sb_t;
 335
 336 /*
 337  * Per-ag incore structure, copies of information in agf and agi, to improve the
 338  * performance of allocation group selection.
 339  */
 340 typedef struct xfs_perag {
 341         struct xfs_mount *pag_mount;    /* owner filesystem */
 342         xfs_agnumber_t  pag_agno;       /* AG this structure belongs to */
 343         atomic_t        pag_ref;        /* perag reference count */
 344         char            pagf_init;      /* this agf's entry is initialized */
 345         char            pagi_init;      /* this agi's entry is initialized */
 346         char            pagf_metadata;  /* the agf is preferred to be metadata */
 347         char            pagi_inodeok;   /* The agi is ok for inodes */
 348         __uint8_t       pagf_levels[XFS_BTNUM_AGF];
 349                                         /* # of levels in bno & cnt btree */
 350         __uint32_t      pagf_flcount;   /* count of blocks in freelist */
 351         xfs_extlen_t    pagf_freeblks;  /* total free blocks */
 352         xfs_extlen_t    pagf_longest;   /* longest free space */
 353         __uint32_t      pagf_btreeblks; /* # of blocks held in AGF btrees */
 354         xfs_agino_t     pagi_freecount; /* number of free inodes */
 355         xfs_agino_t     pagi_count;     /* number of allocated inodes */
 356
 357         /*
 358          * Inode allocation search lookup optimisation.
 359          * If the pagino matches, the search for new inodes
 360          * doesn't need to search the near ones again straight away
 361          */
 362         xfs_agino_t     pagl_pagino;
 363         xfs_agino_t     pagl_leftrec;
 364         xfs_agino_t     pagl_rightrec;
 365         spinlock_t      pagb_lock;      /* lock for pagb_tree */
 366         struct rb_root  pagb_tree;      /* ordered tree of busy extents */
 367
 368         atomic_t        pagf_fstrms;    /* # of filestreams active in this AG */
 369
 370         spinlock_t      pag_ici_lock;   /* incore inode cache lock */
 371         struct radix_tree_root pag_ici_root;    /* incore inode cache root */
 372         int             pag_ici_reclaimable;    /* reclaimable inodes */
 373         struct mutex    pag_ici_reclaim_lock;   /* serialisation point */
 374         unsigned long   pag_ici_reclaim_cursor; /* reclaim restart point */
 375
 376         /* buffer cache index */
 377         spinlock_t      pag_buf_lock;   /* lock for pag_buf_tree */
 378         struct rb_root  pag_buf_tree;   /* ordered tree of active buffers */
 379
 380         /* for rcu-safe freeing */
 381         struct rcu_head rcu_head;
 382         int             pagb_count;     /* pagb slots in use */
 383 } xfs_perag_t;
 384
 385 extern int      xfs_log_sbcount(xfs_mount_t *);
 386 extern __uint64_t xfs_default_resblks(xfs_mount_t *mp);
 387 extern int      xfs_mountfs(xfs_mount_t *mp);
 388 extern int      xfs_initialize_perag(xfs_mount_t *mp, xfs_agnumber_t agcount,
 389                                      xfs_agnumber_t *maxagi);
 390
 391 extern void     xfs_unmountfs(xfs_mount_t *);
 392 extern int      xfs_mod_incore_sb(xfs_mount_t *, xfs_sb_field_t, int64_t, int);
 393 extern int      xfs_mod_incore_sb_batch(xfs_mount_t *, xfs_mod_sb_t *,
 394                         uint, int);
 395 extern int      xfs_mod_icount(struct xfs_mount *mp, int64_t delta);
 396 extern int      xfs_mod_ifree(struct xfs_mount *mp, int64_t delta);
 397 extern int      xfs_mod_fdblocks(struct xfs_mount *mp, int64_t delta,
 398                                  bool reserved);
 399 extern int      xfs_mount_log_sb(xfs_mount_t *);
 400 extern struct xfs_buf *xfs_getsb(xfs_mount_t *, int);
 401 extern int      xfs_readsb(xfs_mount_t *, int);
 402 extern void     xfs_freesb(xfs_mount_t *);
 403 extern bool     xfs_fs_writable(struct xfs_mount *mp, int level);
 404 extern int      xfs_sb_validate_fsb_count(struct xfs_sb *, __uint64_t);
 405
 406 extern int      xfs_dev_is_read_only(struct xfs_mount *, char *);
 407
 408 extern void     xfs_set_low_space_thresholds(struct xfs_mount *);
 409
 410 #endif  /* __KERNEL__ */
 411
 412 #endif  /* __XFS_MOUNT_H__ */