hammer2 - Change freemap segment delineation, other stuff

[dragonfly.git] / sys / vfs / hammer2 / hammer2_disk.h

/*
 * Copyright (c) 2011-2012 The DragonFly Project.  All rights reserved.
 *
 * This code is derived from software contributed to The DragonFly Project
 * by Matthew Dillon <dillon@dragonflybsd.org>
 * by Venkatesh Srinivas <vsrinivas@dragonflybsd.org>
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 *
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in
 *    the documentation and/or other materials provided with the
 *    distribution.
 * 3. Neither the name of The DragonFly Project nor the names of its
 *    contributors may be used to endorse or promote products derived
 *    from this software without specific, prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 */
#ifndef VFS_HAMMER2_DISK_H_
#define VFS_HAMMER2_DISK_H_

#ifndef _SYS_UUID_H_
#include <sys/uuid.h>
#endif
#ifndef _SYS_DMSG_H_
#include <sys/dmsg.h>
#endif

/*
 * The structures below represent the on-disk media structures for the HAMMER2
 * filesystem.  Note that all fields for on-disk structures are naturally
 * aligned.  The host endian format is typically used - compatibility is
 * possible if the implementation detects reversed endian and adjusts accesses
 * accordingly.
 *
 * HAMMER2 primarily revolves around the directory topology:  inodes,
 * directory entries, and block tables.  Block device buffer cache buffers
 * are always 64KB.  Logical file buffers are typically 16KB.  All data
 * references utilize 64-bit byte offsets.
 *
 * Free block management is handled independently using blocks reserved by
 * the media topology.
 */

/*
 * The data at the end of a file or directory may be a fragment in order
 * to optimize storage efficiency.  The minimum fragment size is 1KB.
 * Since allocations are in powers of 2 fragments must also be sized in
 * powers of 2 (1024, 2048, ... 65536).
 *
 * For the moment the maximum allocation size is HAMMER2_PBUFSIZE (64K),
 * which is 2^16.  Larger extents may be supported in the future.  Smaller
 * fragments might be supported in the future (down to 64 bytes is possible),
 * but probably will not be.
 *
 * A full indirect block use supports 1024 x 64-byte blockrefs in a 64KB
 * buffer.  Indirect blocks down to 1KB are supported to keep small
 * directories small.
 *
 * A maximally sized file (2^64-1 bytes) requires 5 indirect block levels.
 * The hammer2_blockset in the volume header or file inode has another 8
 * entries, giving us 66+3 = 69 bits of address space.  However, some bits
 * are taken up by (potentially) requests for redundant copies.  HAMMER2
 * currently supports up to 8 copies, which brings the address space down
 * to 66 bits and gives us 2 bits of leeway.
 */
#define HAMMER2_MIN_ALLOC       1024    /* minimum allocation size */
#define HAMMER2_MIN_RADIX       10      /* minimum allocation size 2^N */
#define HAMMER2_MAX_ALLOC       65536   /* maximum allocation size */
#define HAMMER2_MAX_RADIX       16      /* maximum allocation size 2^N */
#define HAMMER2_KEY_RADIX       64      /* number of bits in key */

/*
 * MINALLOCSIZE         - The minimum allocation size.  This can be smaller
 *                        or larger than the minimum physical IO size.
 *
 *                        NOTE: Should not be larger than 1K since inodes
 *                              are 1K.
 *
 * MINIOSIZE            - The minimum IO size.  This must be less than
 *                        or equal to HAMMER2_LBUFSIZE.
 *
 * HAMMER2_LBUFSIZE     - Nominal buffer size for I/O rollups.
 *
 * HAMMER2_PBUFSIZE     - Topological block size used by files for all
 *                        blocks except the block straddling EOF.
 *
 * HAMMER2_SEGSIZE      - Allocation map segment size, typically 2MB
 *                        (space represented by a level0 bitmap).
 */

#define HAMMER2_SEGSIZE         (1 << HAMMER2_FREEMAP_LEVEL0_RADIX)
#define HAMMER2_SEGRADIX        HAMMER2_FREEMAP_LEVEL0_RADIX

#define HAMMER2_PBUFRADIX       16      /* physical buf (1<<16) bytes */
#define HAMMER2_PBUFSIZE        65536
#define HAMMER2_LBUFRADIX       14      /* logical buf (1<<14) bytes */
#define HAMMER2_LBUFSIZE        16384

/*
 * Generally speaking we want to use 16K and 64K I/Os
 */
#define HAMMER2_MINIORADIX      HAMMER2_LBUFRADIX
#define HAMMER2_MINIOSIZE       HAMMER2_LBUFSIZE

#define HAMMER2_IND_BYTES_MIN   HAMMER2_LBUFSIZE
#define HAMMER2_IND_BYTES_MAX   HAMMER2_PBUFSIZE
#define HAMMER2_IND_COUNT_MIN   (HAMMER2_IND_BYTES_MIN / \
                                 sizeof(hammer2_blockref_t))
#define HAMMER2_IND_COUNT_MAX   (HAMMER2_IND_BYTES_MAX / \
                                 sizeof(hammer2_blockref_t))

/*
 * In HAMMER2, arrays of blockrefs are fully set-associative, meaning that
 * any element can occur at any index and holes can be anywhere.  As a
 * future optimization we will be able to flag that such arrays are sorted
 * and thus optimize lookups, but for now we don't.
 *
 * Inodes embed either 512 bytes of direct data or an array of 8 blockrefs,
 * resulting in highly efficient storage for files <= 512 bytes and for files
 * <= 512KB.  Up to 8 directory entries can be referenced from a directory
 * without requiring an indirect block.
 *
 * Indirect blocks are typically either 4KB (64 blockrefs / ~4MB represented),
 * or 64KB (1024 blockrefs / ~64MB represented).
 */
#define HAMMER2_SET_COUNT               8       /* direct entries */
#define HAMMER2_SET_RADIX               3
#define HAMMER2_EMBEDDED_BYTES          512     /* inode blockset/dd size */
#define HAMMER2_EMBEDDED_RADIX          9

#define HAMMER2_PBUFMASK        (HAMMER2_PBUFSIZE - 1)
#define HAMMER2_LBUFMASK        (HAMMER2_LBUFSIZE - 1)
#define HAMMER2_SEGMASK         (HAMMER2_SEGSIZE - 1)

#define HAMMER2_LBUFMASK64      ((hammer2_off_t)HAMMER2_LBUFMASK)
#define HAMMER2_PBUFSIZE64      ((hammer2_off_t)HAMMER2_PBUFSIZE)
#define HAMMER2_PBUFMASK64      ((hammer2_off_t)HAMMER2_PBUFMASK)
#define HAMMER2_SEGSIZE64       ((hammer2_off_t)HAMMER2_SEGSIZE)
#define HAMMER2_SEGMASK64       ((hammer2_off_t)HAMMER2_SEGMASK)

#define HAMMER2_UUID_STRING     "5cbb9ad1-862d-11dc-a94d-01301bb8a9f5"

/*
 * A HAMMER2 filesystem is always sized in multiples of 8MB.
 *
 * A 4MB segment is reserved at the beginning of each 2GB zone.  This segment
 * contains the volume header (or backup volume header), the free block
 * table, and possibly other information in the future.
 *
 * 4MB = 64 x 64K blocks.  Each 4MB segment is broken down as follows:
 *
 *      +-----------------------+
 *      |       Volume Hdr      | block 0       volume header & alternates
 *      +-----------------------+               (first four zones only)
 *      |   FreeBlk Section A   | block 1-4
 *      +-----------------------+
 *      |   FreeBlk Section B   | block 5-8
 *      +-----------------------+
 *      |   FreeBlk Section C   | block 9-12
 *      +-----------------------+
 *      |   FreeBlk Section D   | block 13-16
 *      +-----------------------+
 *      |                       | block 17...63
 *      |       reserved        |
 *      |                       |
 *      +-----------------------+
 *
 * The first few 2GB zones contain volume headers and volume header backups.
 * After that the volume header block# is reserved.
 *
 *                      Freemap (see the FREEMAP document)
 *
 * The freemap utilizes blocks #1-16 for now, see the FREEMAP document.
 * The filesystems rotations through the sections to avoid disturbing the
 * 'previous' version of the freemap during a flush.
 *
 * Each freemap section is 4 x 64K blocks and represents 2GB, 2TB, 2PB,
 * and 2EB indirect map, plus the volume header has a set of 8 blockrefs
 * for another 3 bits for a total of 64 bits of address space.  The Level 0
 * 64KB block representing 2GB of storage is a hammer2_bmap_data[1024].
 * Each element contains a 128x2 bit bitmap representing 16KB per chunk for
 * 2MB of storage (x1024 elements = 2GB).  2 bits per chunk:
 *
 *      00      Free
 *      01      Possibly fragmented
 *      10      Possibly free
 *      11      Allocated
 *
 * One important thing to note here is that the freemap resolution is 16KB,
 * but the minimuim storage allocation size is 1KB.  The hammer2 vfs keeps
 * track of sub-allocations in memory (on umount or reboot obvious the whole
 * 16KB will be considered allocated even if only 1KB is allocated).  It is
 * possible for fragmentation to build up over time.
 *
 * The Second thing to note is that due to the way snapshots and inode
 * replication works, deleting a file cannot immediately free the related
 * space.  Instead, the freemap elements transition from 11->10.  The bulk
 * freeing code which does a complete scan is then responsible for
 * transitioning the elements to 00 or back to 11 or to 01 for that matter.
 *
 * WARNING!  ZONE_SEG and VOLUME_ALIGN must be a multiple of 1<<LEVEL0_RADIX
 *           (i.e. a multiple of 2MB).  VOLUME_ALIGN must be >= ZONE_SEG.
 */
#define HAMMER2_VOLUME_ALIGN            (8 * 1024 * 1024)
#define HAMMER2_VOLUME_ALIGN64          ((hammer2_off_t)HAMMER2_VOLUME_ALIGN)
#define HAMMER2_VOLUME_ALIGNMASK        (HAMMER2_VOLUME_ALIGN - 1)
#define HAMMER2_VOLUME_ALIGNMASK64     ((hammer2_off_t)HAMMER2_VOLUME_ALIGNMASK)

#define HAMMER2_NEWFS_ALIGN             (HAMMER2_VOLUME_ALIGN)
#define HAMMER2_NEWFS_ALIGN64           ((hammer2_off_t)HAMMER2_VOLUME_ALIGN)
#define HAMMER2_NEWFS_ALIGNMASK         (HAMMER2_VOLUME_ALIGN - 1)
#define HAMMER2_NEWFS_ALIGNMASK64       ((hammer2_off_t)HAMMER2_NEWFS_ALIGNMASK)

#define HAMMER2_ZONE_BYTES64            (2LLU * 1024 * 1024 * 1024)
#define HAMMER2_ZONE_MASK64             (HAMMER2_ZONE_BYTES64 - 1)
#define HAMMER2_ZONE_SEG                (4 * 1024 * 1024)
#define HAMMER2_ZONE_SEG64              ((hammer2_off_t)HAMMER2_ZONE_SEG)
#define HAMMER2_ZONE_BLOCKS_SEG         (HAMMER2_ZONE_SEG / HAMMER2_PBUFSIZE)

/*
 * 64 x 64KB blocks are reserved at the base of each 2GB zone.  These blocks
 * are used to store the volume header or volume header backups, allocation
 * tree, and other information in the future.
 *
 * All specified blocks are not necessarily used in all 2GB zones.  However,
 * dead areas are reserved for future use and MUST NOT BE USED for other
 * purposes.
 *
 * The freemap is arranged into four groups.  Modifications rotate through
 * the groups on a block by block basis (so all the blocks are not necessarily
 * synchronized to the same group).  Because the freemap is flushed
 * independent of the main filesystem, the freemap only really needs two
 * groups to operate efficiently.
 *
 *
 *
 */
#define HAMMER2_ZONE_VOLHDR             0       /* volume header or backup */
#define HAMMER2_ZONE_FREEMAP_A          1       /* freemap layer group A */
#define HAMMER2_ZONE_FREEMAP_B          5       /* freemap layer group B */
#define HAMMER2_ZONE_FREEMAP_C          9       /* freemap layer group C */
#define HAMMER2_ZONE_FREEMAP_D          13      /* freemap layer group D */

                                                /* relative to FREEMAP_x */
#define HAMMER2_ZONEFM_LEVEL1           0       /* 2GB leafmap */
#define HAMMER2_ZONEFM_LEVEL2           1       /* 2TB indmap */
#define HAMMER2_ZONEFM_LEVEL3           2       /* 2PB indmap */
#define HAMMER2_ZONEFM_LEVEL4           3       /* 2EB indmap */
/* LEVEL5 is a set of 8 blockrefs in the volume header 16EB */


/*
 * Freemap radii.  Please note that LEVEL 1 blockref array entries
 * point to 256-byte sections of the bitmap representing 2MB of storage.
 * Even though the chain structures represent only 256 bytes, they are
 * mapped using larger 16K or 64K buffer cache buffers.
 */
#define HAMMER2_FREEMAP_LEVEL5_RADIX    64      /* 16EB */
#define HAMMER2_FREEMAP_LEVEL4_RADIX    61      /* 2EB */
#define HAMMER2_FREEMAP_LEVEL3_RADIX    51      /* 2PB */
#define HAMMER2_FREEMAP_LEVEL2_RADIX    41      /* 2TB */
#define HAMMER2_FREEMAP_LEVEL1_RADIX    31      /* 2GB */
#define HAMMER2_FREEMAP_LEVEL0_RADIX    21      /* 2MB (entry in l-1 leaf) */

#define HAMMER2_FREEMAP_LEVELN_PSIZE    65536   /* physical bytes */

#define HAMMER2_FREEMAP_COUNT           (int)(HAMMER2_FREEMAP_LEVELN_PSIZE / \
                                         sizeof(hammer2_bmap_data_t))
#define HAMMER2_FREEMAP_BLOCK_RADIX     14
#define HAMMER2_FREEMAP_BLOCK_SIZE      (1 << HAMMER2_FREEMAP_BLOCK_RADIX)
#define HAMMER2_FREEMAP_BLOCK_MASK      (HAMMER2_FREEMAP_BLOCK_SIZE - 1)

/*
 * Two linear areas can be reserved after the initial 2MB segment in the base
 * zone (the one starting at offset 0).  These areas are NOT managed by the
 * block allocator and do not fall under HAMMER2 crc checking rules based
 * at the volume header (but can be self-CRCd internally, depending).
 */
#define HAMMER2_BOOT_MIN_BYTES          HAMMER2_VOLUME_ALIGN
#define HAMMER2_BOOT_NOM_BYTES          (64*1024*1024)
#define HAMMER2_BOOT_MAX_BYTES          (256*1024*1024)

#define HAMMER2_REDO_MIN_BYTES          HAMMER2_VOLUME_ALIGN
#define HAMMER2_REDO_NOM_BYTES          (256*1024*1024)
#define HAMMER2_REDO_MAX_BYTES          (1024*1024*1024)

/*
 * Most HAMMER2 types are implemented as unsigned 64-bit integers.
 * Transaction ids are monotonic.
 *
 * We utilize 32-bit iSCSI CRCs.
 */
typedef uint64_t hammer2_tid_t;
typedef uint64_t hammer2_off_t;
typedef uint64_t hammer2_key_t;
typedef uint32_t hammer2_crc32_t;

/*
 * Miscellanious ranges (all are unsigned).
 */
#define HAMMER2_MIN_TID         1ULL
#define HAMMER2_MAX_TID         0xFFFFFFFFFFFFFFFFULL
#define HAMMER2_MIN_KEY         0ULL
#define HAMMER2_MAX_KEY         0xFFFFFFFFFFFFFFFFULL
#define HAMMER2_MIN_OFFSET      0ULL
#define HAMMER2_MAX_OFFSET      0xFFFFFFFFFFFFFFFFULL

/*
 * HAMMER2 data offset special cases and masking.
 *
 * All HAMMER2 data offsets have to be broken down into a 64K buffer base
 * offset (HAMMER2_OFF_MASK_HI) and a 64K buffer index (HAMMER2_OFF_MASK_LO).
 *
 * Indexes into physical buffers are always 64-byte aligned.  The low 6 bits
 * of the data offset field specifies how large the data chunk being pointed
 * to as a power of 2.  The theoretical minimum radix is thus 6 (The space
 * needed in the low bits of the data offset field).  However, the practical
 * minimum allocation chunk size is 1KB (a radix of 10), so HAMMER2 sets
 * HAMMER2_MIN_RADIX to 10.  The maximum radix is currently 16 (64KB), but
 * we fully intend to support larger extents in the future.
 */
#define HAMMER2_OFF_BAD         ((hammer2_off_t)-1)
#define HAMMER2_OFF_MASK        0xFFFFFFFFFFFFFFC0ULL
#define HAMMER2_OFF_MASK_LO     (HAMMER2_OFF_MASK & HAMMER2_PBUFMASK64)
#define HAMMER2_OFF_MASK_HI     (~HAMMER2_PBUFMASK64)
#define HAMMER2_OFF_MASK_RADIX  0x000000000000003FULL
#define HAMMER2_MAX_COPIES      6

/*
 * HAMMER2 directory support and pre-defined keys
 */
#define HAMMER2_DIRHASH_VISIBLE 0x8000000000000000ULL
#define HAMMER2_DIRHASH_USERMSK 0x7FFFFFFFFFFFFFFFULL
#define HAMMER2_DIRHASH_LOMASK  0x0000000000007FFFULL
#define HAMMER2_DIRHASH_HIMASK  0xFFFFFFFFFFFF0000ULL
#define HAMMER2_DIRHASH_FORCED  0x0000000000008000ULL   /* bit forced on */

#define HAMMER2_SROOT_KEY       0x0000000000000000ULL   /* volume to sroot */

/*
 * The media block reference structure.  This forms the core of the HAMMER2
 * media topology recursion.  This 64-byte data structure is embedded in the
 * volume header, in inodes (which are also directory entries), and in
 * indirect blocks.
 *
 * A blockref references a single media item, which typically can be a
 * directory entry (aka inode), indirect block, or data block.
 *
 * The primary feature a blockref represents is the ability to validate
 * the entire tree underneath it via its check code.  Any modification to
 * anything propagates up the blockref tree all the way to the root, replacing
 * the related blocks.  Propagations can shortcut to the volume root to
 * implement the 'fast syncing' feature but this only delays the eventual
 * propagation.
 *
 * The check code can be a simple 32-bit iscsi code, a 64-bit crc,
 * or as complex as a 192 bit cryptographic hash.  192 bits is the maximum
 * supported check code size, which is not sufficient for unverified dedup
 * UNLESS one doesn't mind once-in-a-blue-moon data corruption (such as when
 * farming web data).  HAMMER2 has an unverified dedup feature for just this
 * purpose.
 *
 * --
 *
 * NOTE: The range of keys represented by the blockref is (key) to
 *       ((key) + (1LL << keybits) - 1).  HAMMER2 usually populates
 *       blocks bottom-up, inserting a new root when radix expansion
 *       is required.
 */
struct hammer2_blockref {               /* MUST BE EXACTLY 64 BYTES */
        uint8_t         type;           /* type of underlying item */
        uint8_t         methods;        /* check method & compression method */
        uint8_t         copyid;         /* specify which copy this is */
        uint8_t         keybits;        /* #of keybits masked off 0=leaf */
        uint8_t         vradix;         /* virtual data/meta-data size */
        uint8_t         flags;          /* blockref flags */
        uint8_t         reserved06;
        uint8_t         reserved07;
        hammer2_key_t   key;            /* key specification */
        hammer2_tid_t   mirror_tid;     /* propagate for mirror scan */
        hammer2_tid_t   modify_tid;     /* modifications sans propagation */
        hammer2_off_t   data_off;       /* low 6 bits is phys size (radix)*/
        union {                         /* check info */
                char    buf[24];
                struct {
                        uint32_t value;
                        uint32_t unused[5];
                } iscsi32;
                struct {
                        uint64_t value;
                        uint64_t unused[2];
                } crc64;
                struct {
                        char data[24];
                } sha192;

                /*
                 * Freemap hints are embedded in addition to the icrc32.
                 *
                 * bigmask - Radixes available for allocation (0-31).
                 *           Heuristical (may be permissive but not
                 *           restrictive).  Typically only radix values
                 *           10-16 are used (i.e. (1<<10) through (1<<16)).
                 *
                 * avail   - Total available space remaining, in bytes
                 */
                struct {
                        uint32_t icrc32;
                        uint32_t bigmask;       /* available radixes */
                        uint64_t avail;         /* total available bytes */
                        uint64_t unused;        /* unused must be 0 */
                } freemap;
        } check;
};

typedef struct hammer2_blockref hammer2_blockref_t;

#if 0
#define HAMMER2_BREF_SYNC1              0x01    /* modification synchronized */
#define HAMMER2_BREF_SYNC2              0x02    /* modification committed */
#define HAMMER2_BREF_DESYNCCHLD         0x04    /* desynchronize children */
#define HAMMER2_BREF_DELETED            0x80    /* indicates a deletion */
#endif

#define HAMMER2_BLOCKREF_BYTES          64      /* blockref struct in bytes */

/*
 * On-media and off-media blockref types.
 */
#define HAMMER2_BREF_TYPE_EMPTY         0
#define HAMMER2_BREF_TYPE_INODE         1
#define HAMMER2_BREF_TYPE_INDIRECT      2
#define HAMMER2_BREF_TYPE_DATA          3
#define HAMMER2_BREF_TYPE_UNUSED04      4
#define HAMMER2_BREF_TYPE_FREEMAP_NODE  5
#define HAMMER2_BREF_TYPE_FREEMAP_LEAF  6
#define HAMMER2_BREF_TYPE_FREEMAP       254     /* pseudo-type */
#define HAMMER2_BREF_TYPE_VOLUME        255     /* pseudo-type */

#define HAMMER2_ENC_CHECK(n)            ((n) << 4)
#define HAMMER2_DEC_CHECK(n)            (((n) >> 4) & 15)

#define HAMMER2_CHECK_NONE              0
#define HAMMER2_CHECK_ISCSI32           1
#define HAMMER2_CHECK_CRC64             2
#define HAMMER2_CHECK_SHA192            3
#define HAMMER2_CHECK_FREEMAP           4

#define HAMMER2_ENC_COMP(n)             (n)
#define HAMMER2_DEC_COMP(n)             ((n) & 15)

#define HAMMER2_COMP_NONE               0
#define HAMMER2_COMP_AUTOZERO           1


/*
 * HAMMER2 block references are collected into sets of 8 blockrefs.  These
 * sets are fully associative, meaning the elements making up a set are
 * not sorted in any way and may contain duplicate entries, holes, or
 * entries which shortcut multiple levels of indirection.  Sets are used
 * in various ways:
 *
 * (1) When redundancy is desired a set may contain several duplicate
 *     entries pointing to different copies of the same data.  Up to 8 copies
 *     are supported but the set structure becomes a bit inefficient once
 *     you go over 4.
 *
 * (2) The blockrefs in a set can shortcut multiple levels of indirections
 *     within the bounds imposed by the parent of set.
 *
 * When a set fills up another level of indirection is inserted, moving
 * some or all of the set's contents into indirect blocks placed under the
 * set.  This is a top-down approach in that indirect blocks are not created
 * until the set actually becomes full (that is, the entries in the set can
 * shortcut the indirect blocks when the set is not full).  Depending on how
 * things are filled multiple indirect blocks will eventually be created.
 *
 * Indirect blocks are typically 4KB (64 entres) or 64KB (1024 entries) and
 * are also treated as fully set-associative.
 */
struct hammer2_blockset {
        hammer2_blockref_t      blockref[HAMMER2_SET_COUNT];
};

typedef struct hammer2_blockset hammer2_blockset_t;

/*
 * Catch programmer snafus
 */
#if (1 << HAMMER2_SET_RADIX) != HAMMER2_SET_COUNT
#error "hammer2 direct radix is incorrect"
#endif
#if (1 << HAMMER2_PBUFRADIX) != HAMMER2_PBUFSIZE
#error "HAMMER2_PBUFRADIX and HAMMER2_PBUFSIZE are inconsistent"
#endif
#if (1 << HAMMER2_MIN_RADIX) != HAMMER2_MIN_ALLOC
#error "HAMMER2_MIN_RADIX and HAMMER2_MIN_ALLOC are inconsistent"
#endif

/*
 * hammer2_bmap_data - A freemap entry in the LEVEL1 block.
 *
 * Each 64-byte entry contains the bitmap and meta-data required to manage
 * a LEVEL0 (2MB) block of storage.  The storage is managed in 128 x 16KB
 * chunks.  Smaller allocation granularity is supported via a linear iterator
 * and/or must otherwise be tracked in ram.
 *
 * (data structure must be 64 bytes exactly)
 *
 * linear  - A BYTE linear allocation offset used for sub-16KB allocations
 *           only.  May contain values between 0 and 2MB.  Must be ignored
 *           if 16KB-aligned (i.e. force bitmap scan), otherwise may be
 *           used to sub-allocate within the 16KB block (which is already
 *           marked as allocated in the bitmap).
 *
 *           Sub-allocations need only be 1KB-aligned and do not have to be
 *           size-aligned, and 16KB or larger allocations do not update this
 *           field, resulting in pretty good packing.
 *
 *           Please note that file data granularity may be limited by
 *           other issues such as buffer cache direct-mapping and the
 *           desire to support sector sizes up to 16KB (so H2 only issues
 *           I/O's in multiples of 16KB anyway).
 *
 * class   - Clustering class.  Cleared to 0 only if the entire leaf becomes
 *           free.  Used to cluster device buffers so all elements must have
 *           the same device block size, but may mix logical sizes.
 *
 *           Typically integrated with the blockref type in the upper 8 bits
 *           to localize inodes and indrect blocks, improving bulk free scans
 *           and directory scans.
 *
 * bitmap  - Two bits per 16KB allocation block arranged in arrays of
 *           32-bit elements, 128x2 bits representing ~2MB worth of media
 *           storage.  Bit patterns are as follows:
 *
 *           00 Unallocated
 *           01 Armed for bulk free scan
 *           10 Possibly free
 *           11 Allocated
 */
struct hammer2_bmap_data {
        int32_t linear;         /* 00 linear sub-granular allocation offset */
        uint16_t class;         /* 04-05 clustering class ((type<<8)|radix) */
        uint8_t reserved06;     /* 06 */
        uint8_t reserved07;     /* 07 */
        uint32_t reserved08;    /* 08 */
        uint32_t reserved0C;    /* 0C */
        uint32_t reserved10;    /* 10 */
        uint32_t reserved14;    /* 14 */
        uint32_t reserved18;    /* 18 */
        uint32_t avail;         /* 1C */
        uint32_t bitmap[8];     /* 20-3F 256 bits manages 2MB/16KB/2-bits */
};

typedef struct hammer2_bmap_data hammer2_bmap_data_t;

/*
 * In HAMMER2 inodes ARE directory entries, with a special exception for
 * hardlinks.  The inode number is stored in the inode rather than being
 * based on the location of the inode (since the location moves every time
 * the inode or anything underneath the inode is modified).
 *
 * The inode is 1024 bytes, made up of 256 bytes of meta-data, 256 bytes
 * for the filename, and 512 bytes worth of direct file data OR an embedded
 * blockset.
 *
 * Directories represent one inode per blockref.  Inodes are not laid out
 * as a file but instead are represented by the related blockrefs.  The
 * blockrefs, in turn, are indexed by the 64-bit directory hash key.  Remember
 * that blocksets are fully associative, so a certain degree efficiency is
 * achieved just from that.
 *
 * Up to 512 bytes of direct data can be embedded in an inode, and since
 * inodes are essentially directory entries this also means that small data
 * files end up simply being laid out linearly in the directory, resulting
 * in fewer seeks and highly optimal access.
 *
 * The compression mode can be changed at any time in the inode and is
 * recorded on a blockref-by-blockref basis.
 *
 * Hardlinks are supported via the inode map.  Essentially the way a hardlink
 * works is that all individual directory entries representing the same file
 * are special cased and specify the same inode number.  The actual file
 * is placed in the nearest parent directory that is parent to all instances
 * of the hardlink.  If all hardlinks to a file are in the same directory
 * the actual file will also be placed in that directory.  This file uses
 * the inode number as the directory entry key and is invisible to normal
 * directory scans.  Real directory entry keys are differentiated from the
 * inode number key via bit 63.  Access to the hardlink silently looks up
 * the real file and forwards all operations to that file.  Removal of the
 * last hardlink also removes the real file.
 *
 * (attr_tid) is only updated when the inode's specific attributes or regular
 * file size has changed, and affects path lookups and stat.  (attr_tid)
 * represents a special cache coherency lock under the inode.  The inode
 * blockref's modify_tid will always cover it.
 *
 * (dirent_tid) is only updated when an entry under a directory inode has
 * been created, deleted, renamed, or had its attributes change, and affects
 * directory lookups and scans.  (dirent_tid) represents another special cache
 * coherency lock under the inode.  The inode blockref's modify_tid will
 * always cover it.
 */
#define HAMMER2_INODE_BYTES             1024    /* (asserted by code) */
#define HAMMER2_INODE_MAXNAME           256     /* maximum name in bytes */
#define HAMMER2_INODE_VERSION_ONE       1

struct hammer2_inode_data {
        uint16_t        version;        /* 0000 inode data version */
        uint16_t        reserved02;     /* 0002 */

        /*
         * core inode attributes, inode type, misc flags
         */
        uint32_t        uflags;         /* 0004 chflags */
        uint32_t        rmajor;         /* 0008 available for device nodes */
        uint32_t        rminor;         /* 000C available for device nodes */
        uint64_t        ctime;          /* 0010 inode change time */
        uint64_t        mtime;          /* 0018 modified time */
        uint64_t        atime;          /* 0020 access time (unsupported) */
        uint64_t        btime;          /* 0028 birth time */
        uuid_t          uid;            /* 0030 uid / degenerate unix uid */
        uuid_t          gid;            /* 0040 gid / degenerate unix gid */

        uint8_t         type;           /* 0050 object type */
        uint8_t         op_flags;       /* 0051 operational flags */
        uint16_t        cap_flags;      /* 0052 capability flags */
        uint32_t        mode;           /* 0054 unix modes (typ low 16 bits) */

        /*
         * inode size, identification, localized recursive configuration
         * for compression and backup copies.
         */
        hammer2_tid_t   inum;           /* 0058 inode number */
        hammer2_off_t   size;           /* 0060 size of file */
        uint64_t        nlinks;         /* 0068 hard links (typ only dirs) */
        hammer2_tid_t   iparent;        /* 0070 parent inum (recovery only) */
        hammer2_key_t   name_key;       /* 0078 full filename key */
        uint16_t        name_len;       /* 0080 filename length */
        uint8_t         ncopies;        /* 0082 ncopies to local media */
        uint8_t         comp_algo;      /* 0083 compression request & algo */

        /*
         * These fields are currently only applicable to PFSROOTs.
         *
         * NOTE: We can't use {volume_data->fsid, pfs_clid} to uniquely
         *       identify an instance of a PFS in the cluster because
         *       a mount may contain more than one copy of the PFS as
         *       a separate node.  {pfs_clid, pfs_fsid} must be used for
         *       registration in the cluster.
         */
        uint8_t         target_type;    /* 0084 hardlink target type */
        uint8_t         reserved85;     /* 0085 */
        uint8_t         reserved86;     /* 0086 */
        uint8_t         pfs_type;       /* 0087 (if PFSROOT) node type */
        uint64_t        pfs_inum;       /* 0088 (if PFSROOT) inum allocator */
        uuid_t          pfs_clid;       /* 0090 (if PFSROOT) cluster uuid */
        uuid_t          pfs_fsid;       /* 00A0 (if PFSROOT) unique uuid */

        /*
         * Quotas and cumulative sub-tree counters.
         */
        hammer2_key_t   data_quota;     /* 00B0 subtree quota in bytes */
        hammer2_key_t   data_count;     /* 00B8 subtree byte count */
        hammer2_key_t   inode_quota;    /* 00C0 subtree quota inode count */
        hammer2_key_t   inode_count;    /* 00C8 subtree inode count */
        hammer2_tid_t   attr_tid;       /* 00D0 attributes changed */
        hammer2_tid_t   dirent_tid;     /* 00D8 directory/attr changed */

        /*
         * Tracks (possibly degenerate) free areas covering all sub-tree
         * allocations under inode, not counting the inode itself.
         * 0/0 indicates empty entry.  fully set-associative.
         */
        hammer2_off_t   freezones[4];   /* 00E0/E8/F0/F8 base|radix */

        unsigned char   filename[HAMMER2_INODE_MAXNAME];
                                        /* 0100-01FF (256 char, unterminated) */
        union {                         /* 0200-03FF (64x8 = 512 bytes) */
                struct hammer2_blockset blockset;
                char data[HAMMER2_EMBEDDED_BYTES];
        } u;
};

typedef struct hammer2_inode_data hammer2_inode_data_t;

#define HAMMER2_OPFLAG_DIRECTDATA       0x01
#define HAMMER2_OPFLAG_PFSROOT          0x02
#define HAMMER2_OPFLAG_COPYIDS          0x04    /* copyids override parent */

#define HAMMER2_OBJTYPE_UNKNOWN         0
#define HAMMER2_OBJTYPE_DIRECTORY       1
#define HAMMER2_OBJTYPE_REGFILE         2
#define HAMMER2_OBJTYPE_FIFO            4
#define HAMMER2_OBJTYPE_CDEV            5
#define HAMMER2_OBJTYPE_BDEV            6
#define HAMMER2_OBJTYPE_SOFTLINK        7
#define HAMMER2_OBJTYPE_HARDLINK        8       /* dummy entry for hardlink */
#define HAMMER2_OBJTYPE_SOCKET          9
#define HAMMER2_OBJTYPE_WHITEOUT        10

#define HAMMER2_COPYID_NONE             0
#define HAMMER2_COPYID_LOCAL            ((uint8_t)-1)

/*
 * PEER types identify connections and help cluster controller filter
 * out unwanted SPANs.
 */
#define HAMMER2_PEER_NONE               DMSG_PEER_NONE
#define HAMMER2_PEER_CLUSTER            DMSG_PEER_CLUSTER
#define HAMMER2_PEER_BLOCK              DMSG_PEER_BLOCK
#define HAMMER2_PEER_HAMMER2            DMSG_PEER_HAMMER2

#define HAMMER2_COPYID_COUNT            DMSG_COPYID_COUNT

/*
 * PFS types identify a PFS on media and in LNK_SPAN messages.
 */
#define HAMMER2_PFSTYPE_NONE            DMSG_PFSTYPE_NONE
#define HAMMER2_PFSTYPE_ADMIN           DMSG_PFSTYPE_ADMIN
#define HAMMER2_PFSTYPE_CLIENT          DMSG_PFSTYPE_CLIENT
#define HAMMER2_PFSTYPE_CACHE           DMSG_PFSTYPE_CACHE
#define HAMMER2_PFSTYPE_COPY            DMSG_PFSTYPE_COPY
#define HAMMER2_PFSTYPE_SLAVE           DMSG_PFSTYPE_SLAVE
#define HAMMER2_PFSTYPE_SOFT_SLAVE      DMSG_PFSTYPE_SOFT_SLAVE
#define HAMMER2_PFSTYPE_SOFT_MASTER     DMSG_PFSTYPE_SOFT_MASTER
#define HAMMER2_PFSTYPE_MASTER          DMSG_PFSTYPE_MASTER
#define HAMMER2_PFSTYPE_SNAPSHOT        DMSG_PFSTYPE_SNAPSHOT
#define HAMMER2_PFSTYPE_MAX             DMSG_PFSTYPE_MAX

/*
 *                              Allocation Table
 *
 */


/*
 * Flags (8 bits) - blockref, for freemap only
 *
 * Note that the minimum chunk size is 1KB so we could theoretically have
 * 10 bits here, but we might have some future extension that allows a
 * chunk size down to 256 bytes and if so we will need bits 8 and 9.
 */
#define HAMMER2_AVF_SELMASK             0x03    /* select group */
#define HAMMER2_AVF_ALL_ALLOC           0x04    /* indicate all allocated */
#define HAMMER2_AVF_ALL_FREE            0x08    /* indicate all free */
#define HAMMER2_AVF_RESERVED10          0x10
#define HAMMER2_AVF_RESERVED20          0x20
#define HAMMER2_AVF_RESERVED40          0x40
#define HAMMER2_AVF_RESERVED80          0x80
#define HAMMER2_AVF_AVMASK32            ((uint32_t)0xFFFFFF00LU)
#define HAMMER2_AVF_AVMASK64            ((uint64_t)0xFFFFFFFFFFFFFF00LLU)

#define HAMMER2_AV_SELECT_A             0x00
#define HAMMER2_AV_SELECT_B             0x01
#define HAMMER2_AV_SELECT_C             0x02
#define HAMMER2_AV_SELECT_D             0x03

/*
 * The volume header eats a 64K block.  There is currently an issue where
 * we want to try to fit all nominal filesystem updates in a 512-byte section
 * but it may be a lost cause due to the need for a blockset.
 *
 * All information is stored in host byte order.  The volume header's magic
 * number may be checked to determine the byte order.  If you wish to mount
 * between machines w/ different endian modes you'll need filesystem code
 * which acts on the media data consistently (either all one way or all the
 * other).  Our code currently does not do that.
 *
 * A read-write mount may have to recover missing allocations by doing an
 * incremental mirror scan looking for modifications made after alloc_tid.
 * If alloc_tid == last_tid then no recovery operation is needed.  Recovery
 * operations are usually very, very fast.
 *
 * Read-only mounts do not need to do any recovery, access to the filesystem
 * topology is always consistent after a crash (is always consistent, period).
 * However, there may be shortcutted blockref updates present from deep in
 * the tree which are stored in the volumeh eader and must be tracked on
 * the fly.
 *
 * NOTE: The copyinfo[] array contains the configuration for both the
 *       cluster connections and any local media copies.  The volume
 *       header will be replicated for each local media copy.
 *
 *       The mount command may specify multiple medias or just one and
 *       allow HAMMER2 to pick up the others when it checks the copyinfo[]
 *       array on mount.
 *
 * NOTE: root_blockref points to the super-root directory, not the root
 *       directory.  The root directory will be a subdirectory under the
 *       super-root.
 *
 *       The super-root directory contains all root directories and all
 *       snapshots (readonly or writable).  It is possible to do a
 *       null-mount of the super-root using special path constructions
 *       relative to your mounted root.
 *
 * NOTE: HAMMER2 allows any subdirectory tree to be managed as if it were
 *       a PFS, including mirroring and storage quota operations, and this is
 *       prefered over creating discrete PFSs in the super-root.  Instead
 *       the super-root is most typically used to create writable snapshots,
 *       alternative roots, and so forth.  The super-root is also used by
 *       the automatic snapshotting mechanism.
 */
#define HAMMER2_VOLUME_ID_HBO   0x48414d3205172011LLU
#define HAMMER2_VOLUME_ID_ABO   0x11201705324d4148LLU

struct hammer2_volume_data {
        /*
         * sector #0 - 512 bytes
         */
        uint64_t        magic;                  /* 0000 Signature */
        hammer2_off_t   boot_beg;               /* 0008 Boot area (future) */
        hammer2_off_t   boot_end;               /* 0010 (size = end - beg) */
        hammer2_off_t   aux_beg;                /* 0018 Aux area (future) */
        hammer2_off_t   aux_end;                /* 0020 (size = end - beg) */
        hammer2_off_t   volu_size;              /* 0028 Volume size, bytes */

        uint32_t        version;                /* 0030 */
        uint32_t        flags;                  /* 0034 */
        uint8_t         copyid;                 /* 0038 copyid of phys vol */
        uint8_t         freemap_version;        /* 0039 freemap algorithm */
        uint8_t         peer_type;              /* 003A HAMMER2_PEER_xxx */
        uint8_t         reserved003B;           /* 003B */
        uint32_t        reserved003C;           /* 003C */

        uuid_t          fsid;                   /* 0040 */
        uuid_t          fstype;                 /* 0050 */

        /*
         * allocator_size is precalculated at newfs time and does not include
         * reserved blocks, boot, or redo areas.
         *
         * Initial non-reserved-area allocations do not use the freemap
         * but instead adjust alloc_iterator.  Dynamic allocations take
         * over starting at (allocator_beg).  This makes newfs_hammer2's
         * job a lot easier and can also serve as a testing jig.
         */
        hammer2_off_t   allocator_size;         /* 0060 Total data space */
        hammer2_off_t   allocator_free;         /* 0068 Free space */
        hammer2_off_t   allocator_beg;          /* 0070 Initial allocations */
        hammer2_tid_t   mirror_tid;             /* 0078 best committed tid */
        hammer2_tid_t   alloc_tid;              /* 0080 Alloctable modify tid */
        hammer2_blockref_t reserved0088;        /* 0088-00C7 */

        /*
         * Copyids are allocated dynamically from the copyexists bitmap.
         * An id from the active copies set (up to 8, see copyinfo later on)
         * may still exist after the copy set has been removed from the
         * volume header and its bit will remain active in the bitmap and
         * cannot be reused until it is 100% removed from the hierarchy.
         */
        uint32_t        copyexists[8];          /* 00C8-00E7 copy exists bmap */
        char            reserved0140[248];      /* 00E8-01DF */

        /*
         * 32 bit CRC array at the end of the first 512 byte sector.
         *
         * icrc_sects[7] - First 512-4 bytes of volume header (including all
         *                 the other icrc's except this one).
         *
         * icrc_sects[6] - Sector 1 (512 bytes) of volume header, which is
         *                 the blockset for the root.
         *
         * icrc_sects[5] - Sector 2
         * icrc_sects[4] - Sector 3
         * icrc_sects[3] - Sector 4 (the freemap blockset)
         */
        hammer2_crc32_t icrc_sects[8];          /* 01E0-01FF */

        /*
         * sector #1 - 512 bytes
         *
         * The entire sector is used by a blockset.
         */
        hammer2_blockset_t sroot_blockset;      /* 0200-03FF Superroot dir */

        /*
         * sector #2-7
         */
        char    sector2[512];                   /* 0400-05FF reserved */
        char    sector3[512];                   /* 0600-07FF reserved */
        hammer2_blockset_t freemap_blockset;    /* 0800-09FF freemap  */
        char    sector5[512];                   /* 0A00-0BFF reserved */
        char    sector6[512];                   /* 0C00-0DFF reserved */
        char    sector7[512];                   /* 0E00-0FFF reserved */

        /*
         * sector #8-71 - 32768 bytes
         *
         * Contains the configuration for up to 256 copyinfo targets.  These
         * specify local and remote copies operating as masters or slaves.
         * copyid's 0 and 255 are reserved (0 indicates an empty slot and 255
         * indicates the local media).
         *
         * Each inode contains a set of up to 8 copyids, either inherited
         * from its parent or explicitly specified in the inode, which
         * indexes into this array.
         */
                                                /* 1000-8FFF copyinfo config */
        dmsg_vol_data_t copyinfo[HAMMER2_COPYID_COUNT];

        /*
         * Remaining sections are reserved for future use.
         */
        char            reserved0400[0x6FFC];   /* 9000-FFFB reserved */

        /*
         * icrc on entire volume header
         */
        hammer2_crc32_t icrc_volheader;         /* FFFC-FFFF full volume icrc*/
};

typedef struct hammer2_volume_data hammer2_volume_data_t;

/*
 * Various parts of the volume header have their own iCRCs.
 *
 * The first 512 bytes has its own iCRC stored at the end of the 512 bytes
 * and not included the icrc calculation.
 *
 * The second 512 bytes also has its own iCRC but it is stored in the first
 * 512 bytes so it covers the entire second 512 bytes.
 *
 * The whole volume block (64KB) has an iCRC covering all but the last 4 bytes,
 * which is where the iCRC for the whole volume is stored.  This is currently
 * a catch-all for anything not individually iCRCd.
 */
#define HAMMER2_VOL_ICRC_SECT0          7
#define HAMMER2_VOL_ICRC_SECT1          6

#define HAMMER2_VOLUME_BYTES            65536

#define HAMMER2_VOLUME_ICRC0_OFF        0
#define HAMMER2_VOLUME_ICRC1_OFF        512
#define HAMMER2_VOLUME_ICRCVH_OFF       0

#define HAMMER2_VOLUME_ICRC0_SIZE       (512 - 4)
#define HAMMER2_VOLUME_ICRC1_SIZE       (512)
#define HAMMER2_VOLUME_ICRCVH_SIZE      (65536 - 4)

#define HAMMER2_VOL_VERSION_MIN         1
#define HAMMER2_VOL_VERSION_DEFAULT     1
#define HAMMER2_VOL_VERSION_WIP         2

#define HAMMER2_NUM_VOLHDRS             4

union hammer2_media_data {
        hammer2_volume_data_t   voldata;
        hammer2_inode_data_t    ipdata;
        hammer2_blockref_t      npdata[HAMMER2_IND_COUNT_MAX];
        hammer2_bmap_data_t     bmdata[HAMMER2_FREEMAP_COUNT];
        char                    buf[HAMMER2_PBUFSIZE];
};

typedef union hammer2_media_data hammer2_media_data_t;

#endif