linux 2.6.16.38 w/ vs2.0.3-rc1
[linux-2.6.git] / fs / xfs / xfs_mount.c
index 9dfae18..62188ea 100644 (file)
 #include "xfs_trans.h"
 #include "xfs_sb.h"
 #include "xfs_ag.h"
+#include "xfs_dir.h"
 #include "xfs_dir2.h"
 #include "xfs_dmapi.h"
 #include "xfs_mount.h"
 #include "xfs_bmap_btree.h"
 #include "xfs_alloc_btree.h"
 #include "xfs_ialloc_btree.h"
+#include "xfs_dir_sf.h"
 #include "xfs_dir2_sf.h"
 #include "xfs_attr_sf.h"
 #include "xfs_dinode.h"
@@ -49,32 +51,11 @@ STATIC int  xfs_uuid_mount(xfs_mount_t *);
 STATIC void    xfs_uuid_unmount(xfs_mount_t *mp);
 STATIC void    xfs_unmountfs_wait(xfs_mount_t *);
 
-
-#ifdef HAVE_PERCPU_SB
-STATIC void    xfs_icsb_destroy_counters(xfs_mount_t *);
-STATIC void    xfs_icsb_balance_counter(xfs_mount_t *, xfs_sb_field_t, int);
-STATIC void    xfs_icsb_sync_counters(xfs_mount_t *);
-STATIC int     xfs_icsb_modify_counters(xfs_mount_t *, xfs_sb_field_t,
-                                               int, int);
-STATIC int     xfs_icsb_modify_counters_locked(xfs_mount_t *, xfs_sb_field_t,
-                                               int, int);
-STATIC int     xfs_icsb_disable_counter(xfs_mount_t *, xfs_sb_field_t);
-
-#else
-
-#define xfs_icsb_destroy_counters(mp)                  do { } while (0)
-#define xfs_icsb_balance_counter(mp, a, b)             do { } while (0)
-#define xfs_icsb_sync_counters(mp)                     do { } while (0)
-#define xfs_icsb_modify_counters(mp, a, b, c)          do { } while (0)
-#define xfs_icsb_modify_counters_locked(mp, a, b, c)   do { } while (0)
-
-#endif
-
 static const struct {
-       short offset;
-       short type;     /* 0 = integer
-                        * 1 = binary / string (no translation)
-                        */
+    short offset;
+    short type;     /* 0 = integer
+               * 1 = binary / string (no translation)
+               */
 } xfs_sb_info[] = {
     { offsetof(xfs_sb_t, sb_magicnum),   0 },
     { offsetof(xfs_sb_t, sb_blocksize),  0 },
@@ -132,11 +113,7 @@ xfs_mount_init(void)
 {
        xfs_mount_t *mp;
 
-       mp = kmem_zalloc(sizeof(xfs_mount_t), KM_SLEEP);
-
-       if (xfs_icsb_init_counters(mp)) {
-               mp->m_flags |= XFS_MOUNT_NO_PERCPU_SB;
-       }
+       mp = kmem_zalloc(sizeof(*mp), KM_SLEEP);
 
        AIL_LOCKINIT(&mp->m_ail_lock, "xfs_ail");
        spinlock_init(&mp->m_sb_lock, "xfs_sb");
@@ -159,8 +136,8 @@ xfs_mount_init(void)
  */
 void
 xfs_mount_free(
-       xfs_mount_t     *mp,
-       int             remove_bhv)
+       xfs_mount_t *mp,
+       int         remove_bhv)
 {
        if (mp->m_ihash)
                xfs_ihash_free(mp);
@@ -194,13 +171,12 @@ xfs_mount_free(
                kmem_free(mp->m_logname, strlen(mp->m_logname) + 1);
 
        if (remove_bhv) {
-               struct bhv_vfs  *vfsp = XFS_MTOVFS(mp);
+               struct vfs      *vfsp = XFS_MTOVFS(mp);
 
                bhv_remove_all_vfsops(vfsp, 0);
                VFS_REMOVEBHV(vfsp, &mp->m_bhv);
        }
 
-       xfs_icsb_destroy_counters(mp);
        kmem_free(mp, sizeof(xfs_mount_t));
 }
 
@@ -211,8 +187,7 @@ xfs_mount_free(
 STATIC int
 xfs_mount_validate_sb(
        xfs_mount_t     *mp,
-       xfs_sb_t        *sbp,
-       int             flags)
+       xfs_sb_t        *sbp)
 {
        /*
         * If the log device and data device have the
@@ -222,29 +197,33 @@ xfs_mount_validate_sb(
         * a volume filesystem in a non-volume manner.
         */
        if (sbp->sb_magicnum != XFS_SB_MAGIC) {
-               xfs_fs_mount_cmn_err(flags, "bad magic number");
+               cmn_err(CE_WARN, "XFS: bad magic number");
                return XFS_ERROR(EWRONGFS);
        }
 
        if (!XFS_SB_GOOD_VERSION(sbp)) {
-               xfs_fs_mount_cmn_err(flags, "bad version");
+               cmn_err(CE_WARN, "XFS: bad version");
                return XFS_ERROR(EWRONGFS);
        }
 
        if (unlikely(
            sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) {
-               xfs_fs_mount_cmn_err(flags,
-                       "filesystem is marked as having an external log; "
-                       "specify logdev on the\nmount command line.");
-               return XFS_ERROR(EINVAL);
+               cmn_err(CE_WARN,
+       "XFS: filesystem is marked as having an external log; "
+       "specify logdev on the\nmount command line.");
+               XFS_CORRUPTION_ERROR("xfs_mount_validate_sb(1)",
+                                    XFS_ERRLEVEL_HIGH, mp, sbp);
+               return XFS_ERROR(EFSCORRUPTED);
        }
 
        if (unlikely(
            sbp->sb_logstart != 0 && mp->m_logdev_targp != mp->m_ddev_targp)) {
-               xfs_fs_mount_cmn_err(flags,
-                       "filesystem is marked as having an internal log; "
-                       "do not specify logdev on\nthe mount command line.");
-               return XFS_ERROR(EINVAL);
+               cmn_err(CE_WARN,
+       "XFS: filesystem is marked as having an internal log; "
+       "don't specify logdev on\nthe mount command line.");
+               XFS_CORRUPTION_ERROR("xfs_mount_validate_sb(2)",
+                                    XFS_ERRLEVEL_HIGH, mp, sbp);
+               return XFS_ERROR(EFSCORRUPTED);
        }
 
        /*
@@ -263,13 +242,12 @@ xfs_mount_validate_sb(
            sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG                    ||
            sbp->sb_inodesize < XFS_DINODE_MIN_SIZE                     ||
            sbp->sb_inodesize > XFS_DINODE_MAX_SIZE                     ||
-           sbp->sb_inodelog < XFS_DINODE_MIN_LOG                       ||
-           sbp->sb_inodelog > XFS_DINODE_MAX_LOG                       ||
-           (sbp->sb_blocklog - sbp->sb_inodelog != sbp->sb_inopblog)   ||
            (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE)  ||
            (sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE)  ||
-           (sbp->sb_imax_pct > 100 /* zero sb_imax_pct is valid */))) {
-               xfs_fs_mount_cmn_err(flags, "SB sanity check 1 failed");
+           sbp->sb_imax_pct > 100)) {
+               cmn_err(CE_WARN, "XFS: SB sanity check 1 failed");
+               XFS_CORRUPTION_ERROR("xfs_mount_validate_sb(3)",
+                                    XFS_ERRLEVEL_LOW, mp, sbp);
                return XFS_ERROR(EFSCORRUPTED);
        }
 
@@ -282,7 +260,9 @@ xfs_mount_validate_sb(
             (xfs_drfsbno_t)sbp->sb_agcount * sbp->sb_agblocks ||
            sbp->sb_dblocks < (xfs_drfsbno_t)(sbp->sb_agcount - 1) *
                              sbp->sb_agblocks + XFS_MIN_AG_BLOCKS)) {
-               xfs_fs_mount_cmn_err(flags, "SB sanity check 2 failed");
+               cmn_err(CE_WARN, "XFS: SB sanity check 2 failed");
+               XFS_ERROR_REPORT("xfs_mount_validate_sb(4)",
+                                XFS_ERRLEVEL_LOW, mp);
                return XFS_ERROR(EFSCORRUPTED);
        }
 
@@ -298,13 +278,15 @@ xfs_mount_validate_sb(
            (sbp->sb_dblocks << (sbp->sb_blocklog - BBSHIFT)) > UINT_MAX ||
            (sbp->sb_rblocks << (sbp->sb_blocklog - BBSHIFT)) > UINT_MAX)) {
 #endif
-               xfs_fs_mount_cmn_err(flags,
-                       "file system too large to be mounted on this system.");
+               cmn_err(CE_WARN,
+       "XFS: File system is too large to be mounted on this system.");
                return XFS_ERROR(E2BIG);
        }
 
        if (unlikely(sbp->sb_inprogress)) {
-               xfs_fs_mount_cmn_err(flags, "file system busy");
+               cmn_err(CE_WARN, "XFS: file system busy");
+               XFS_ERROR_REPORT("xfs_mount_validate_sb(5)",
+                                XFS_ERRLEVEL_LOW, mp);
                return XFS_ERROR(EFSCORRUPTED);
        }
 
@@ -312,8 +294,8 @@ xfs_mount_validate_sb(
         * Version 1 directory format has never worked on Linux.
         */
        if (unlikely(!XFS_SB_VERSION_HASDIRV2(sbp))) {
-               xfs_fs_mount_cmn_err(flags,
-                       "file system using version 1 directory format");
+               cmn_err(CE_WARN,
+       "XFS: Attempted to mount file system using version 1 directory format");
                return XFS_ERROR(ENOSYS);
        }
 
@@ -321,11 +303,11 @@ xfs_mount_validate_sb(
         * Until this is fixed only page-sized or smaller data blocks work.
         */
        if (unlikely(sbp->sb_blocksize > PAGE_SIZE)) {
-               xfs_fs_mount_cmn_err(flags,
-                       "file system with blocksize %d bytes",
+               cmn_err(CE_WARN,
+               "XFS: Attempted to mount file system with blocksize %d bytes",
                        sbp->sb_blocksize);
-               xfs_fs_mount_cmn_err(flags,
-                       "only pagesize (%ld) or less will currently work.",
+               cmn_err(CE_WARN,
+               "XFS: Only page-sized (%ld) or less blocksizes currently work.",
                        PAGE_SIZE);
                return XFS_ERROR(ENOSYS);
        }
@@ -335,7 +317,7 @@ xfs_mount_validate_sb(
 
 xfs_agnumber_t
 xfs_initialize_perag(
-       bhv_vfs_t       *vfs,
+       struct vfs      *vfs,
        xfs_mount_t     *mp,
        xfs_agnumber_t  agcount)
 {
@@ -382,7 +364,7 @@ xfs_initialize_perag(
                                break;
                        }
 
-                       /* This ag is preferred for inodes */
+                       /* This ag is prefered for inodes */
                        pag = &mp->m_perag[index];
                        pag->pagi_inodeok = 1;
                        if (index < max_metadata)
@@ -473,7 +455,7 @@ xfs_xlatesb(
  * Does the initial read of the superblock.
  */
 int
-xfs_readsb(xfs_mount_t *mp, int flags)
+xfs_readsb(xfs_mount_t *mp)
 {
        unsigned int    sector_size;
        unsigned int    extra_flags;
@@ -495,7 +477,7 @@ xfs_readsb(xfs_mount_t *mp, int flags)
        bp = xfs_buf_read_flags(mp->m_ddev_targp, XFS_SB_DADDR,
                                BTOBB(sector_size), extra_flags);
        if (!bp || XFS_BUF_ISERROR(bp)) {
-               xfs_fs_mount_cmn_err(flags, "SB read failed");
+               cmn_err(CE_WARN, "XFS: SB read failed");
                error = bp ? XFS_BUF_GETERROR(bp) : ENOMEM;
                goto fail;
        }
@@ -509,9 +491,9 @@ xfs_readsb(xfs_mount_t *mp, int flags)
        sbp = XFS_BUF_TO_SBP(bp);
        xfs_xlatesb(XFS_BUF_PTR(bp), &(mp->m_sb), 1, XFS_SB_ALL_BITS);
 
-       error = xfs_mount_validate_sb(mp, &(mp->m_sb), flags);
+       error = xfs_mount_validate_sb(mp, &(mp->m_sb));
        if (error) {
-               xfs_fs_mount_cmn_err(flags, "SB validate failed");
+               cmn_err(CE_WARN, "XFS: SB validate failed");
                goto fail;
        }
 
@@ -519,8 +501,8 @@ xfs_readsb(xfs_mount_t *mp, int flags)
         * We must be able to do sector-sized and sector-aligned IO.
         */
        if (sector_size > mp->m_sb.sb_sectsize) {
-               xfs_fs_mount_cmn_err(flags,
-                       "device supports only %u byte sectors (not %u)",
+               cmn_err(CE_WARN,
+                       "XFS: device supports only %u byte sectors (not %u)",
                        sector_size, mp->m_sb.sb_sectsize);
                error = ENOSYS;
                goto fail;
@@ -537,7 +519,7 @@ xfs_readsb(xfs_mount_t *mp, int flags)
                bp = xfs_buf_read_flags(mp->m_ddev_targp, XFS_SB_DADDR,
                                        BTOBB(sector_size), extra_flags);
                if (!bp || XFS_BUF_ISERROR(bp)) {
-                       xfs_fs_mount_cmn_err(flags, "SB re-read failed");
+                       cmn_err(CE_WARN, "XFS: SB re-read failed");
                        error = bp ? XFS_BUF_GETERROR(bp) : ENOMEM;
                        goto fail;
                }
@@ -545,10 +527,6 @@ xfs_readsb(xfs_mount_t *mp, int flags)
                ASSERT(XFS_BUF_VALUSEMA(bp) <= 0);
        }
 
-       xfs_icsb_balance_counter(mp, XFS_SBS_ICOUNT, 0);
-       xfs_icsb_balance_counter(mp, XFS_SBS_IFREE, 0);
-       xfs_icsb_balance_counter(mp, XFS_SBS_FDBLOCKS, 0);
-
        mp->m_sb_bp = bp;
        xfs_buf_relse(bp);
        ASSERT(XFS_BUF_VALUSEMA(bp) > 0);
@@ -649,14 +627,14 @@ xfs_mount_common(xfs_mount_t *mp, xfs_sb_t *sbp)
  */
 int
 xfs_mountfs(
-       bhv_vfs_t       *vfsp,
+       vfs_t           *vfsp,
        xfs_mount_t     *mp,
        int             mfsi_flags)
 {
        xfs_buf_t       *bp;
        xfs_sb_t        *sbp = &(mp->m_sb);
        xfs_inode_t     *rip;
-       bhv_vnode_t     *rvp = NULL;
+       vnode_t         *rvp = NULL;
        int             readio_log, writeio_log;
        xfs_daddr_t     d;
        __uint64_t      ret64;
@@ -667,7 +645,7 @@ xfs_mountfs(
        int             error = 0;
 
        if (mp->m_sb_bp == NULL) {
-               if ((error = xfs_readsb(mp, mfsi_flags))) {
+               if ((error = xfs_readsb(mp))) {
                        return error;
                }
        }
@@ -932,7 +910,18 @@ xfs_mountfs(
        vfsp->vfs_altfsid = (xfs_fsid_t *)mp->m_fixedfsid;
        mp->m_dmevmask = 0;     /* not persistent; set after each mount */
 
-       xfs_dir_mount(mp);
+       /*
+        * Select the right directory manager.
+        */
+       mp->m_dirops =
+               XFS_SB_VERSION_HASDIRV2(&mp->m_sb) ?
+                       xfsv2_dirops :
+                       xfsv1_dirops;
+
+       /*
+        * Initialize directory manager's entries.
+        */
+       XFS_DIR_MOUNT(mp);
 
        /*
         * Initialize the attribute manager's entries.
@@ -993,9 +982,8 @@ xfs_mountfs(
 
        if (unlikely((rip->i_d.di_mode & S_IFMT) != S_IFDIR)) {
                cmn_err(CE_WARN, "XFS: corrupted root inode");
-               cmn_err(CE_WARN, "Device %s - root %llu is not a directory",
-                       XFS_BUFTARG_NAME(mp->m_ddev_targp),
-                       (unsigned long long)rip->i_ino);
+               prdev("Root inode %llu is not a directory",
+                     mp->m_ddev_targp, (unsigned long long)rip->i_ino);
                xfs_iunlock(rip, XFS_ILOCK_EXCL);
                XFS_ERROR_REPORT("xfs_mountfs_int(2)", XFS_ERRLEVEL_LOW,
                                 mp);
@@ -1082,7 +1070,7 @@ xfs_mountfs(
 int
 xfs_unmountfs(xfs_mount_t *mp, struct cred *cr)
 {
-       struct bhv_vfs  *vfsp = XFS_MTOVFS(mp);
+       struct vfs      *vfsp = XFS_MTOVFS(mp);
 #if defined(DEBUG) || defined(INDUCE_IO_ERROR)
        int64_t         fsid;
 #endif
@@ -1166,9 +1154,6 @@ xfs_unmountfs_writesb(xfs_mount_t *mp)
        sbp = xfs_getsb(mp, 0);
        if (!(XFS_MTOVFS(mp)->vfs_flag & VFS_RDONLY ||
                XFS_FORCED_SHUTDOWN(mp))) {
-
-               xfs_icsb_sync_counters(mp);
-
                /*
                 * mark shared-readonly if desired
                 */
@@ -1243,7 +1228,6 @@ xfs_mod_sb(xfs_trans_t *tp, __int64_t fields)
        xfs_trans_log_buf(tp, bp, first, last);
 }
 
-
 /*
  * xfs_mod_incore_sb_unlocked() is a utility routine common used to apply
  * a delta to a specified field in the in-core superblock.  Simply
@@ -1253,7 +1237,7 @@ xfs_mod_sb(xfs_trans_t *tp, __int64_t fields)
  *
  * The SB_LOCK must be held when this routine is called.
  */
-int
+STATIC int
 xfs_mod_incore_sb_unlocked(xfs_mount_t *mp, xfs_sb_field_t field,
                        int delta, int rsvd)
 {
@@ -1288,8 +1272,7 @@ xfs_mod_incore_sb_unlocked(xfs_mount_t *mp, xfs_sb_field_t field,
                return 0;
        case XFS_SBS_FDBLOCKS:
 
-               lcounter = (long long)
-                       mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
+               lcounter = (long long)mp->m_sb.sb_fdblocks;
                res_used = (long long)(mp->m_resblks - mp->m_resblks_avail);
 
                if (delta > 0) {                /* Putting blocks back */
@@ -1323,7 +1306,7 @@ xfs_mod_incore_sb_unlocked(xfs_mount_t *mp, xfs_sb_field_t field,
                        }
                }
 
-               mp->m_sb.sb_fdblocks = lcounter + XFS_ALLOC_SET_ASIDE(mp);
+               mp->m_sb.sb_fdblocks = lcounter;
                return 0;
        case XFS_SBS_FREXTENTS:
                lcounter = (long long)mp->m_sb.sb_frextents;
@@ -1423,26 +1406,9 @@ xfs_mod_incore_sb(xfs_mount_t *mp, xfs_sb_field_t field, int delta, int rsvd)
        unsigned long   s;
        int     status;
 
-       /* check for per-cpu counters */
-       switch (field) {
-#ifdef HAVE_PERCPU_SB
-       case XFS_SBS_ICOUNT:
-       case XFS_SBS_IFREE:
-       case XFS_SBS_FDBLOCKS:
-               if (!(mp->m_flags & XFS_MOUNT_NO_PERCPU_SB)) {
-                       status = xfs_icsb_modify_counters(mp, field,
-                                                       delta, rsvd);
-                       break;
-               }
-               /* FALLTHROUGH */
-#endif
-       default:
-               s = XFS_SB_LOCK(mp);
-               status = xfs_mod_incore_sb_unlocked(mp, field, delta, rsvd);
-               XFS_SB_UNLOCK(mp, s);
-               break;
-       }
-
+       s = XFS_SB_LOCK(mp);
+       status = xfs_mod_incore_sb_unlocked(mp, field, delta, rsvd);
+       XFS_SB_UNLOCK(mp, s);
        return status;
 }
 
@@ -1479,26 +1445,8 @@ xfs_mod_incore_sb_batch(xfs_mount_t *mp, xfs_mod_sb_t *msb, uint nmsb, int rsvd)
                 * from the loop so we'll fall into the undo loop
                 * below.
                 */
-               switch (msbp->msb_field) {
-#ifdef HAVE_PERCPU_SB
-               case XFS_SBS_ICOUNT:
-               case XFS_SBS_IFREE:
-               case XFS_SBS_FDBLOCKS:
-                       if (!(mp->m_flags & XFS_MOUNT_NO_PERCPU_SB)) {
-                               status = xfs_icsb_modify_counters_locked(mp,
-                                                       msbp->msb_field,
-                                                       msbp->msb_delta, rsvd);
-                               break;
-                       }
-                       /* FALLTHROUGH */
-#endif
-               default:
-                       status = xfs_mod_incore_sb_unlocked(mp,
-                                               msbp->msb_field,
-                                               msbp->msb_delta, rsvd);
-                       break;
-               }
-
+               status = xfs_mod_incore_sb_unlocked(mp, msbp->msb_field,
+                                                   msbp->msb_delta, rsvd);
                if (status != 0) {
                        break;
                }
@@ -1515,28 +1463,8 @@ xfs_mod_incore_sb_batch(xfs_mount_t *mp, xfs_mod_sb_t *msb, uint nmsb, int rsvd)
        if (status != 0) {
                msbp--;
                while (msbp >= msb) {
-                       switch (msbp->msb_field) {
-#ifdef HAVE_PERCPU_SB
-                       case XFS_SBS_ICOUNT:
-                       case XFS_SBS_IFREE:
-                       case XFS_SBS_FDBLOCKS:
-                               if (!(mp->m_flags & XFS_MOUNT_NO_PERCPU_SB)) {
-                                       status =
-                                           xfs_icsb_modify_counters_locked(mp,
-                                                       msbp->msb_field,
-                                                       -(msbp->msb_delta),
-                                                       rsvd);
-                                       break;
-                               }
-                               /* FALLTHROUGH */
-#endif
-                       default:
-                               status = xfs_mod_incore_sb_unlocked(mp,
-                                                       msbp->msb_field,
-                                                       -(msbp->msb_delta),
-                                                       rsvd);
-                               break;
-                       }
+                       status = xfs_mod_incore_sb_unlocked(mp,
+                                   msbp->msb_field, -(msbp->msb_delta), rsvd);
                        ASSERT(status == 0);
                        msbp--;
                }
@@ -1649,529 +1577,3 @@ xfs_mount_log_sbunit(
        xfs_mod_sb(tp, fields);
        xfs_trans_commit(tp, 0, NULL);
 }
-
-
-#ifdef HAVE_PERCPU_SB
-/*
- * Per-cpu incore superblock counters
- *
- * Simple concept, difficult implementation
- *
- * Basically, replace the incore superblock counters with a distributed per cpu
- * counter for contended fields (e.g.  free block count).
- *
- * Difficulties arise in that the incore sb is used for ENOSPC checking, and
- * hence needs to be accurately read when we are running low on space. Hence
- * there is a method to enable and disable the per-cpu counters based on how
- * much "stuff" is available in them.
- *
- * Basically, a counter is enabled if there is enough free resource to justify
- * running a per-cpu fast-path. If the per-cpu counter runs out (i.e. a local
- * ENOSPC), then we disable the counters to synchronise all callers and
- * re-distribute the available resources.
- *
- * If, once we redistributed the available resources, we still get a failure,
- * we disable the per-cpu counter and go through the slow path.
- *
- * The slow path is the current xfs_mod_incore_sb() function.  This means that
- * when we disable a per-cpu counter, we need to drain it's resources back to
- * the global superblock. We do this after disabling the counter to prevent
- * more threads from queueing up on the counter.
- *
- * Essentially, this means that we still need a lock in the fast path to enable
- * synchronisation between the global counters and the per-cpu counters. This
- * is not a problem because the lock will be local to a CPU almost all the time
- * and have little contention except when we get to ENOSPC conditions.
- *
- * Basically, this lock becomes a barrier that enables us to lock out the fast
- * path while we do things like enabling and disabling counters and
- * synchronising the counters.
- *
- * Locking rules:
- *
- *     1. XFS_SB_LOCK() before picking up per-cpu locks
- *     2. per-cpu locks always picked up via for_each_online_cpu() order
- *     3. accurate counter sync requires XFS_SB_LOCK + per cpu locks
- *     4. modifying per-cpu counters requires holding per-cpu lock
- *     5. modifying global counters requires holding XFS_SB_LOCK
- *     6. enabling or disabling a counter requires holding the XFS_SB_LOCK
- *        and _none_ of the per-cpu locks.
- *
- * Disabled counters are only ever re-enabled by a balance operation
- * that results in more free resources per CPU than a given threshold.
- * To ensure counters don't remain disabled, they are rebalanced when
- * the global resource goes above a higher threshold (i.e. some hysteresis
- * is present to prevent thrashing).
- */
-
-#ifdef CONFIG_HOTPLUG_CPU
-/*
- * hot-plug CPU notifier support.
- *
- * We need a notifier per filesystem as we need to be able to identify
- * the filesystem to balance the counters out. This is achieved by
- * having a notifier block embedded in the xfs_mount_t and doing pointer
- * magic to get the mount pointer from the notifier block address.
- */
-STATIC int
-xfs_icsb_cpu_notify(
-       struct notifier_block *nfb,
-       unsigned long action,
-       void *hcpu)
-{
-       xfs_icsb_cnts_t *cntp;
-       xfs_mount_t     *mp;
-       int             s;
-
-       mp = (xfs_mount_t *)container_of(nfb, xfs_mount_t, m_icsb_notifier);
-       cntp = (xfs_icsb_cnts_t *)
-                       per_cpu_ptr(mp->m_sb_cnts, (unsigned long)hcpu);
-       switch (action) {
-       case CPU_UP_PREPARE:
-               /* Easy Case - initialize the area and locks, and
-                * then rebalance when online does everything else for us. */
-               memset(cntp, 0, sizeof(xfs_icsb_cnts_t));
-               break;
-       case CPU_ONLINE:
-               xfs_icsb_balance_counter(mp, XFS_SBS_ICOUNT, 0);
-               xfs_icsb_balance_counter(mp, XFS_SBS_IFREE, 0);
-               xfs_icsb_balance_counter(mp, XFS_SBS_FDBLOCKS, 0);
-               break;
-       case CPU_DEAD:
-               /* Disable all the counters, then fold the dead cpu's
-                * count into the total on the global superblock and
-                * re-enable the counters. */
-               s = XFS_SB_LOCK(mp);
-               xfs_icsb_disable_counter(mp, XFS_SBS_ICOUNT);
-               xfs_icsb_disable_counter(mp, XFS_SBS_IFREE);
-               xfs_icsb_disable_counter(mp, XFS_SBS_FDBLOCKS);
-
-               mp->m_sb.sb_icount += cntp->icsb_icount;
-               mp->m_sb.sb_ifree += cntp->icsb_ifree;
-               mp->m_sb.sb_fdblocks += cntp->icsb_fdblocks;
-
-               memset(cntp, 0, sizeof(xfs_icsb_cnts_t));
-
-               xfs_icsb_balance_counter(mp, XFS_SBS_ICOUNT, XFS_ICSB_SB_LOCKED);
-               xfs_icsb_balance_counter(mp, XFS_SBS_IFREE, XFS_ICSB_SB_LOCKED);
-               xfs_icsb_balance_counter(mp, XFS_SBS_FDBLOCKS, XFS_ICSB_SB_LOCKED);
-               XFS_SB_UNLOCK(mp, s);
-               break;
-       }
-
-       return NOTIFY_OK;
-}
-#endif /* CONFIG_HOTPLUG_CPU */
-
-int
-xfs_icsb_init_counters(
-       xfs_mount_t     *mp)
-{
-       xfs_icsb_cnts_t *cntp;
-       int             i;
-
-       mp->m_sb_cnts = alloc_percpu(xfs_icsb_cnts_t);
-       if (mp->m_sb_cnts == NULL)
-               return -ENOMEM;
-
-#ifdef CONFIG_HOTPLUG_CPU
-       mp->m_icsb_notifier.notifier_call = xfs_icsb_cpu_notify;
-       mp->m_icsb_notifier.priority = 0;
-       register_hotcpu_notifier(&mp->m_icsb_notifier);
-#endif /* CONFIG_HOTPLUG_CPU */
-
-       for_each_online_cpu(i) {
-               cntp = (xfs_icsb_cnts_t *)per_cpu_ptr(mp->m_sb_cnts, i);
-               memset(cntp, 0, sizeof(xfs_icsb_cnts_t));
-       }
-       /*
-        * start with all counters disabled so that the
-        * initial balance kicks us off correctly
-        */
-       mp->m_icsb_counters = -1;
-       return 0;
-}
-
-STATIC void
-xfs_icsb_destroy_counters(
-       xfs_mount_t     *mp)
-{
-       if (mp->m_sb_cnts) {
-               unregister_hotcpu_notifier(&mp->m_icsb_notifier);
-               free_percpu(mp->m_sb_cnts);
-       }
-}
-
-STATIC inline void
-xfs_icsb_lock_cntr(
-       xfs_icsb_cnts_t *icsbp)
-{
-       while (test_and_set_bit(XFS_ICSB_FLAG_LOCK, &icsbp->icsb_flags)) {
-               ndelay(1000);
-       }
-}
-
-STATIC inline void
-xfs_icsb_unlock_cntr(
-       xfs_icsb_cnts_t *icsbp)
-{
-       clear_bit(XFS_ICSB_FLAG_LOCK, &icsbp->icsb_flags);
-}
-
-
-STATIC inline void
-xfs_icsb_lock_all_counters(
-       xfs_mount_t     *mp)
-{
-       xfs_icsb_cnts_t *cntp;
-       int             i;
-
-       for_each_online_cpu(i) {
-               cntp = (xfs_icsb_cnts_t *)per_cpu_ptr(mp->m_sb_cnts, i);
-               xfs_icsb_lock_cntr(cntp);
-       }
-}
-
-STATIC inline void
-xfs_icsb_unlock_all_counters(
-       xfs_mount_t     *mp)
-{
-       xfs_icsb_cnts_t *cntp;
-       int             i;
-
-       for_each_online_cpu(i) {
-               cntp = (xfs_icsb_cnts_t *)per_cpu_ptr(mp->m_sb_cnts, i);
-               xfs_icsb_unlock_cntr(cntp);
-       }
-}
-
-STATIC void
-xfs_icsb_count(
-       xfs_mount_t     *mp,
-       xfs_icsb_cnts_t *cnt,
-       int             flags)
-{
-       xfs_icsb_cnts_t *cntp;
-       int             i;
-
-       memset(cnt, 0, sizeof(xfs_icsb_cnts_t));
-
-       if (!(flags & XFS_ICSB_LAZY_COUNT))
-               xfs_icsb_lock_all_counters(mp);
-
-       for_each_online_cpu(i) {
-               cntp = (xfs_icsb_cnts_t *)per_cpu_ptr(mp->m_sb_cnts, i);
-               cnt->icsb_icount += cntp->icsb_icount;
-               cnt->icsb_ifree += cntp->icsb_ifree;
-               cnt->icsb_fdblocks += cntp->icsb_fdblocks;
-       }
-
-       if (!(flags & XFS_ICSB_LAZY_COUNT))
-               xfs_icsb_unlock_all_counters(mp);
-}
-
-STATIC int
-xfs_icsb_counter_disabled(
-       xfs_mount_t     *mp,
-       xfs_sb_field_t  field)
-{
-       ASSERT((field >= XFS_SBS_ICOUNT) && (field <= XFS_SBS_FDBLOCKS));
-       return test_bit(field, &mp->m_icsb_counters);
-}
-
-STATIC int
-xfs_icsb_disable_counter(
-       xfs_mount_t     *mp,
-       xfs_sb_field_t  field)
-{
-       xfs_icsb_cnts_t cnt;
-
-       ASSERT((field >= XFS_SBS_ICOUNT) && (field <= XFS_SBS_FDBLOCKS));
-
-       xfs_icsb_lock_all_counters(mp);
-       if (!test_and_set_bit(field, &mp->m_icsb_counters)) {
-               /* drain back to superblock */
-
-               xfs_icsb_count(mp, &cnt, XFS_ICSB_SB_LOCKED|XFS_ICSB_LAZY_COUNT);
-               switch(field) {
-               case XFS_SBS_ICOUNT:
-                       mp->m_sb.sb_icount = cnt.icsb_icount;
-                       break;
-               case XFS_SBS_IFREE:
-                       mp->m_sb.sb_ifree = cnt.icsb_ifree;
-                       break;
-               case XFS_SBS_FDBLOCKS:
-                       mp->m_sb.sb_fdblocks = cnt.icsb_fdblocks;
-                       break;
-               default:
-                       BUG();
-               }
-       }
-
-       xfs_icsb_unlock_all_counters(mp);
-
-       return 0;
-}
-
-STATIC void
-xfs_icsb_enable_counter(
-       xfs_mount_t     *mp,
-       xfs_sb_field_t  field,
-       uint64_t        count,
-       uint64_t        resid)
-{
-       xfs_icsb_cnts_t *cntp;
-       int             i;
-
-       ASSERT((field >= XFS_SBS_ICOUNT) && (field <= XFS_SBS_FDBLOCKS));
-
-       xfs_icsb_lock_all_counters(mp);
-       for_each_online_cpu(i) {
-               cntp = per_cpu_ptr(mp->m_sb_cnts, i);
-               switch (field) {
-               case XFS_SBS_ICOUNT:
-                       cntp->icsb_icount = count + resid;
-                       break;
-               case XFS_SBS_IFREE:
-                       cntp->icsb_ifree = count + resid;
-                       break;
-               case XFS_SBS_FDBLOCKS:
-                       cntp->icsb_fdblocks = count + resid;
-                       break;
-               default:
-                       BUG();
-                       break;
-               }
-               resid = 0;
-       }
-       clear_bit(field, &mp->m_icsb_counters);
-       xfs_icsb_unlock_all_counters(mp);
-}
-
-STATIC void
-xfs_icsb_sync_counters_int(
-       xfs_mount_t     *mp,
-       int             flags)
-{
-       xfs_icsb_cnts_t cnt;
-       int             s;
-
-       /* Pass 1: lock all counters */
-       if ((flags & XFS_ICSB_SB_LOCKED) == 0)
-               s = XFS_SB_LOCK(mp);
-
-       xfs_icsb_count(mp, &cnt, flags);
-
-       /* Step 3: update mp->m_sb fields */
-       if (!xfs_icsb_counter_disabled(mp, XFS_SBS_ICOUNT))
-               mp->m_sb.sb_icount = cnt.icsb_icount;
-       if (!xfs_icsb_counter_disabled(mp, XFS_SBS_IFREE))
-               mp->m_sb.sb_ifree = cnt.icsb_ifree;
-       if (!xfs_icsb_counter_disabled(mp, XFS_SBS_FDBLOCKS))
-               mp->m_sb.sb_fdblocks = cnt.icsb_fdblocks;
-
-       if ((flags & XFS_ICSB_SB_LOCKED) == 0)
-               XFS_SB_UNLOCK(mp, s);
-}
-
-/*
- * Accurate update of per-cpu counters to incore superblock
- */
-STATIC void
-xfs_icsb_sync_counters(
-       xfs_mount_t     *mp)
-{
-       xfs_icsb_sync_counters_int(mp, 0);
-}
-
-/*
- * lazy addition used for things like df, background sb syncs, etc
- */
-void
-xfs_icsb_sync_counters_lazy(
-       xfs_mount_t     *mp)
-{
-       xfs_icsb_sync_counters_int(mp, XFS_ICSB_LAZY_COUNT);
-}
-
-/*
- * Balance and enable/disable counters as necessary.
- *
- * Thresholds for re-enabling counters are somewhat magic.
- * inode counts are chosen to be the same number as single
- * on disk allocation chunk per CPU, and free blocks is
- * something far enough zero that we aren't going thrash
- * when we get near ENOSPC.
- */
-#define XFS_ICSB_INO_CNTR_REENABLE     64
-#define XFS_ICSB_FDBLK_CNTR_REENABLE(mp) \
-               (512 + XFS_ALLOC_SET_ASIDE(mp))
-STATIC void
-xfs_icsb_balance_counter(
-       xfs_mount_t     *mp,
-       xfs_sb_field_t  field,
-       int             flags)
-{
-       uint64_t        count, resid;
-       int             weight = num_online_cpus();
-       int             s;
-
-       if (!(flags & XFS_ICSB_SB_LOCKED))
-               s = XFS_SB_LOCK(mp);
-
-       /* disable counter and sync counter */
-       xfs_icsb_disable_counter(mp, field);
-
-       /* update counters  - first CPU gets residual*/
-       switch (field) {
-       case XFS_SBS_ICOUNT:
-               count = mp->m_sb.sb_icount;
-               resid = do_div(count, weight);
-               if (count < XFS_ICSB_INO_CNTR_REENABLE)
-                       goto out;
-               break;
-       case XFS_SBS_IFREE:
-               count = mp->m_sb.sb_ifree;
-               resid = do_div(count, weight);
-               if (count < XFS_ICSB_INO_CNTR_REENABLE)
-                       goto out;
-               break;
-       case XFS_SBS_FDBLOCKS:
-               count = mp->m_sb.sb_fdblocks;
-               resid = do_div(count, weight);
-               if (count < XFS_ICSB_FDBLK_CNTR_REENABLE(mp))
-                       goto out;
-               break;
-       default:
-               BUG();
-               count = resid = 0;      /* quiet, gcc */
-               break;
-       }
-
-       xfs_icsb_enable_counter(mp, field, count, resid);
-out:
-       if (!(flags & XFS_ICSB_SB_LOCKED))
-               XFS_SB_UNLOCK(mp, s);
-}
-
-STATIC int
-xfs_icsb_modify_counters_int(
-       xfs_mount_t     *mp,
-       xfs_sb_field_t  field,
-       int             delta,
-       int             rsvd,
-       int             flags)
-{
-       xfs_icsb_cnts_t *icsbp;
-       long long       lcounter;       /* long counter for 64 bit fields */
-       int             cpu, s, locked = 0;
-       int             ret = 0, balance_done = 0;
-
-again:
-       cpu = get_cpu();
-       icsbp = (xfs_icsb_cnts_t *)per_cpu_ptr(mp->m_sb_cnts, cpu),
-       xfs_icsb_lock_cntr(icsbp);
-       if (unlikely(xfs_icsb_counter_disabled(mp, field)))
-               goto slow_path;
-
-       switch (field) {
-       case XFS_SBS_ICOUNT:
-               lcounter = icsbp->icsb_icount;
-               lcounter += delta;
-               if (unlikely(lcounter < 0))
-                       goto slow_path;
-               icsbp->icsb_icount = lcounter;
-               break;
-
-       case XFS_SBS_IFREE:
-               lcounter = icsbp->icsb_ifree;
-               lcounter += delta;
-               if (unlikely(lcounter < 0))
-                       goto slow_path;
-               icsbp->icsb_ifree = lcounter;
-               break;
-
-       case XFS_SBS_FDBLOCKS:
-               BUG_ON((mp->m_resblks - mp->m_resblks_avail) != 0);
-
-               lcounter = icsbp->icsb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
-               lcounter += delta;
-               if (unlikely(lcounter < 0))
-                       goto slow_path;
-               icsbp->icsb_fdblocks = lcounter + XFS_ALLOC_SET_ASIDE(mp);
-               break;
-       default:
-               BUG();
-               break;
-       }
-       xfs_icsb_unlock_cntr(icsbp);
-       put_cpu();
-       if (locked)
-               XFS_SB_UNLOCK(mp, s);
-       return 0;
-
-       /*
-        * The slow path needs to be run with the SBLOCK
-        * held so that we prevent other threads from
-        * attempting to run this path at the same time.
-        * this provides exclusion for the balancing code,
-        * and exclusive fallback if the balance does not
-        * provide enough resources to continue in an unlocked
-        * manner.
-        */
-slow_path:
-       xfs_icsb_unlock_cntr(icsbp);
-       put_cpu();
-
-       /* need to hold superblock incase we need
-        * to disable a counter */
-       if (!(flags & XFS_ICSB_SB_LOCKED)) {
-               s = XFS_SB_LOCK(mp);
-               locked = 1;
-               flags |= XFS_ICSB_SB_LOCKED;
-       }
-       if (!balance_done) {
-               xfs_icsb_balance_counter(mp, field, flags);
-               balance_done = 1;
-               goto again;
-       } else {
-               /*
-                * we might not have enough on this local
-                * cpu to allocate for a bulk request.
-                * We need to drain this field from all CPUs
-                * and disable the counter fastpath
-                */
-               xfs_icsb_disable_counter(mp, field);
-       }
-
-       ret = xfs_mod_incore_sb_unlocked(mp, field, delta, rsvd);
-
-       if (locked)
-               XFS_SB_UNLOCK(mp, s);
-       return ret;
-}
-
-STATIC int
-xfs_icsb_modify_counters(
-       xfs_mount_t     *mp,
-       xfs_sb_field_t  field,
-       int             delta,
-       int             rsvd)
-{
-       return xfs_icsb_modify_counters_int(mp, field, delta, rsvd, 0);
-}
-
-/*
- * Called when superblock is already locked
- */
-STATIC int
-xfs_icsb_modify_counters_locked(
-       xfs_mount_t     *mp,
-       xfs_sb_field_t  field,
-       int             delta,
-       int             rsvd)
-{
-       return xfs_icsb_modify_counters_int(mp, field, delta,
-                                               rsvd, XFS_ICSB_SB_LOCKED);
-}
-#endif