mariadb/ft/ft.cc

/* -*- mode: C++; c-basic-offset: 4; indent-tabs-mode: nil -*- */
// vim: ft=cpp:expandtab:ts=8:sw=4:softtabstop=4:
#ident "$Id$"
#ident "Copyright (c) 2007-2012 Tokutek Inc.  All rights reserved."
#ident "The technology is licensed by the Massachusetts Institute of Technology, Rutgers State University of New Jersey, and the Research Foundation of State University of New York at Stony Brook under United States of America Serial No. 11/760379 and to the patents and/or patent applications resulting from it."


#include "includes.h"
#include <ft-cachetable-wrappers.h>

void
toku_ft_suppress_rollbacks(FT h, TOKUTXN txn) {
    TXNID txnid = toku_txn_get_txnid(txn);
    assert(h->txnid_that_created_or_locked_when_empty == TXNID_NONE ||
           h->txnid_that_created_or_locked_when_empty == txnid);
    h->txnid_that_created_or_locked_when_empty = txnid;
}

void 
toku_reset_root_xid_that_created(FT ft, TXNID new_root_xid_that_created) {
    // Reset the root_xid_that_created field to the given value.  
    // This redefines which xid created the dictionary.

    // hold lock around setting and clearing of dirty bit
    // (see cooperative use of dirty bit in ft_begin_checkpoint())
    toku_ft_lock (ft);
    ft->h->root_xid_that_created = new_root_xid_that_created;
    ft->h->dirty = 1;
    toku_ft_unlock (ft);
}

static void
ft_destroy(FT ft) {
    if (!ft->panic) assert(!ft->checkpoint_header);

    //header and checkpoint_header have same Blocktable pointer
    //cannot destroy since it is still in use by CURRENT
    assert(ft->h->type == FT_CURRENT);
    toku_blocktable_destroy(&ft->blocktable);
    if (ft->descriptor.dbt.data) toku_free(ft->descriptor.dbt.data);
    if (ft->cmp_descriptor.dbt.data) toku_free(ft->cmp_descriptor.dbt.data);
    toku_ft_destroy_reflock(ft);
    toku_free(ft->h);
}

// Make a copy of the header for the purpose of a checkpoint
// Not reentrant for a single FT.
// See ft_checkpoint for explanation of why
// FT lock must be held.
static void
ft_copy_for_checkpoint_unlocked(FT ft, LSN checkpoint_lsn) {
    assert(ft->h->type == FT_CURRENT);
    assert(ft->checkpoint_header == NULL);
    assert(ft->panic==0);

    FT_HEADER XMEMDUP(ch, ft->h);
    ch->type = FT_CHECKPOINT_INPROGRESS; //Different type
    //printf("checkpoint_lsn=%" PRIu64 "\n", checkpoint_lsn.lsn);
    ch->checkpoint_lsn = checkpoint_lsn;

    //ch->blocktable is SHARED between the two headers
    ft->checkpoint_header = ch;
}

void
toku_ft_free (FT ft) {
    ft_destroy(ft);
    toku_free(ft);
}

void
toku_ft_init_reflock(FT ft) {
    toku_mutex_init(&ft->ft_ref_lock, NULL);
}

void
toku_ft_destroy_reflock(FT ft) {
    toku_mutex_destroy(&ft->ft_ref_lock);
}

void
toku_ft_grab_reflock(FT ft) {
    toku_mutex_lock(&ft->ft_ref_lock);
}

void
toku_ft_release_reflock(FT ft) {
    toku_mutex_unlock(&ft->ft_ref_lock);
}

/////////////////////////////////////////////////////////////////////////
// Start of Functions that are callbacks to the cachefule
//

// maps to cf->log_fassociate_during_checkpoint
static int
ft_log_fassociate_during_checkpoint (CACHEFILE cf, void *header_v) {
    FT ft = (FT) header_v;
    char* fname_in_env = toku_cachefile_fname_in_env(cf);
    BYTESTRING bs = { .len = (uint32_t) strlen(fname_in_env), // don't include the NUL
                      .data = fname_in_env };
    TOKULOGGER logger = toku_cachefile_logger(cf);
    FILENUM filenum = toku_cachefile_filenum(cf);
    bool unlink_on_close = toku_cachefile_is_unlink_on_close(cf);
    int r = toku_log_fassociate(logger, NULL, 0, filenum, ft->h->flags, bs, unlink_on_close);
    return r;
}

// maps to cf->log_suppress_rollback_during_checkpoint
static int
ft_log_suppress_rollback_during_checkpoint (CACHEFILE cf, void *header_v) {
    int r = 0;
    FT h = (FT) header_v;
    TXNID xid = h->txnid_that_created_or_locked_when_empty;
    if (xid != TXNID_NONE) {
        //Only log if useful.
        TOKULOGGER logger = toku_cachefile_logger(cf);
        FILENUM filenum = toku_cachefile_filenum (cf);
        // We don't have access to the txn here, but the txn is
        // necessarily already marked as non-readonly.  Use NULL.
        TOKUTXN txn = NULL;
        r = toku_log_suppress_rollback(logger, NULL, 0, txn, filenum, xid);
    }
    return r;
}

// Maps to cf->begin_checkpoint_userdata
// Create checkpoint-in-progress versions of header and translation (btt) (and fifo for now...).
// Has access to fd (it is protected).
//
// Not reentrant for a single FT (see ft_checkpoint)
static int
ft_begin_checkpoint (LSN checkpoint_lsn, void *header_v) {
    FT ft = (FT) header_v;
    int r = ft->panic;
    if (r==0) {
        // hold lock around copying and clearing of dirty bit
        toku_ft_lock (ft);
        assert(ft->h->type == FT_CURRENT);
        assert(ft->checkpoint_header == NULL);
        ft_copy_for_checkpoint_unlocked(ft, checkpoint_lsn);
        ft->h->dirty = 0;             // this is only place this bit is cleared        (in currentheader)
        toku_block_translation_note_start_checkpoint_unlocked(ft->blocktable);
        toku_ft_unlock (ft);
    }
    return r;
}

// #4922: Hack to remove data corruption race condition.
// Reading (and upgrading) a node up to version 19 causes this.
// We COULD skip this if we know that no nodes remained (as of last checkpoint)
// that are below version 19.
// If there are no nodes < version 19 this is harmless (field is unused).
// If there are, this will make certain the value is at least as low as necessary,
// and not much lower.  (Too low is good, too high can cause data corruption).
// TODO(yoni): If we ever stop supporting upgrades of nodes < version 19 we can delete this.
// TODO(yoni): If we know no nodes are left to upgrade, we can skip this. (Probably not worth doing).
static void
ft_hack_highest_unused_msn_for_upgrade_for_checkpoint(FT ft) {
    if (ft->h->layout_version_original < FT_LAYOUT_VERSION_19) {
        ft->checkpoint_header->highest_unused_msn_for_upgrade = ft->h->highest_unused_msn_for_upgrade;
    }
}

// maps to cf->checkpoint_userdata
// Write checkpoint-in-progress versions of header and translation to disk (really to OS internal buffer).
// Copy current header's version of checkpoint_staging stat64info to checkpoint header.
// Must have access to fd (protected).
// Requires: all pending bits are clear.  This implies that no thread will modify the checkpoint_staging
// version of the stat64info.
//
// No locks are taken for checkpoint_count/lsn because this is single threaded.  Can be called by:
//  - ft_close
//  - end_checkpoint
// checkpoints hold references to FTs and so they cannot be closed during a checkpoint.
// ft_close is not reentrant for a single FT
// end_checkpoint is not reentrant period
static int
ft_checkpoint (CACHEFILE cf, int fd, void *header_v) {
    FT ft = (FT) header_v;
    FT_HEADER ch = ft->checkpoint_header;
    int r = 0;
    if (ft->panic!=0) goto handle_error;
    //printf("%s:%d allocated_limit=%lu writing queue to %lu\n", __FILE__, __LINE__,
    //             block_allocator_allocated_limit(h->block_allocator), h->unused_blocks.b*h->nodesize);
    assert(ch);
    assert(ch->type == FT_CHECKPOINT_INPROGRESS);
    if (ch->dirty) {            // this is only place this bit is tested (in checkpoint_header)
        TOKULOGGER logger = toku_cachefile_logger(cf);
        if (logger) {
            r = toku_logger_fsync_if_lsn_not_fsynced(logger, ch->checkpoint_lsn);
            if (r!=0) goto handle_error;
        }
        uint64_t now = (uint64_t) time(NULL);
        ft->h->time_of_last_modification = now;
        ch->time_of_last_modification = now;
        ch->checkpoint_count++;
        ft_hack_highest_unused_msn_for_upgrade_for_checkpoint(ft);
                                                             
        // write translation and header to disk (or at least to OS internal buffer)
        r = toku_serialize_ft_to(fd, ch, ft->blocktable, ft->cf);
        if (r!=0) goto handle_error;
        ch->dirty = 0;                      // this is only place this bit is cleared (in checkpoint_header)
        
        // fsync the cachefile
        r = toku_cachefile_fsync(cf);
        if (r!=0) {
            goto handle_error;
        }
        ft->h->checkpoint_count++;        // checkpoint succeeded, next checkpoint will save to alternate header location
        ft->h->checkpoint_lsn = ch->checkpoint_lsn;  //Header updated.
    } 
    else {
        toku_block_translation_note_skipped_checkpoint(ft->blocktable);
    }
    if (0) {
handle_error:
        if (ft->panic) r = ft->panic;
        else toku_block_translation_note_failed_checkpoint(ft->blocktable);
    }
    return r;

}

// maps to cf->end_checkpoint_userdata
// free unused disk space 
// (i.e. tell BlockAllocator to liberate blocks used by previous checkpoint).
// Must have access to fd (protected)
static int
ft_end_checkpoint (CACHEFILE UU(cachefile), int fd, void *header_v) {
    FT ft = (FT) header_v;
    int r = ft->panic;
    if (r==0) {
        assert(ft->h->type == FT_CURRENT);
        toku_block_translation_note_end_checkpoint(ft->blocktable, fd);
    }
    if (ft->checkpoint_header) {         // could be NULL only if panic was true at begin_checkpoint
        toku_free(ft->checkpoint_header);
        ft->checkpoint_header = NULL;
    }
    return r;
}

// maps to cf->close_userdata
// Has access to fd (it is protected).
static int
ft_close (CACHEFILE cachefile, int fd, void *header_v, char **malloced_error_string, bool oplsn_valid, LSN oplsn) {
    FT ft = (FT) header_v;
    assert(ft->h->type == FT_CURRENT);
    // We already have exclusive access to this field already, so skip the locking.
    // This should already never fail.
    invariant(!toku_ft_needed_unlocked(ft));
    int r = 0;
    if (ft->panic) {
        r = ft->panic;
    } else {
        assert(ft->cf == cachefile);
        TOKULOGGER logger = toku_cachefile_logger(cachefile);
        LSN lsn = ZERO_LSN;
        //Get LSN
        if (oplsn_valid) {
            //Use recovery-specified lsn
            lsn = oplsn;
            //Recovery cannot reduce lsn of a header.
            if (lsn.lsn < ft->h->checkpoint_lsn.lsn)
                lsn = ft->h->checkpoint_lsn;
        }
        else {
            //Get LSN from logger
            lsn = ZERO_LSN; // if there is no logger, we use zero for the lsn
            if (logger) {
                char* fname_in_env = toku_cachefile_fname_in_env(cachefile);
                assert(fname_in_env);
                BYTESTRING bs = {.len=(uint32_t) strlen(fname_in_env), .data=fname_in_env};
                r = toku_log_fclose(logger, &lsn, ft->h->dirty, bs, toku_cachefile_filenum(cachefile)); // flush the log on close (if new header is being written), otherwise it might not make it out.
                if (r!=0) return r;
            }
        }
        if (ft->h->dirty) {               // this is the only place this bit is tested (in currentheader)
            if (logger) { //Rollback cachefile MUST NOT BE CLOSED DIRTY
                          //It can be checkpointed only via 'checkpoint'
                assert(logger->rollback_cachefile != cachefile);
            }
            int r2;
            //assert(lsn.lsn!=0);
            r2 = ft_begin_checkpoint(lsn, header_v);
            if (r==0) r = r2;
            r2 = ft_checkpoint(cachefile, fd, ft);
            if (r==0) r = r2;
            r2 = ft_end_checkpoint(cachefile, fd, header_v);
            if (r==0) r = r2;
            if (!ft->panic) assert(!ft->h->dirty);             // dirty bit should be cleared by begin_checkpoint and never set again (because we're closing the dictionary)
        }
    }
    if (malloced_error_string) *malloced_error_string = ft->panic_string;
    if (r == 0) {
        r = ft->panic;
    }
    toku_ft_free(ft);
    return r;
}

// maps to cf->note_pin_by_checkpoint
//Must be protected by ydb lock.
//Is only called by checkpoint begin, which holds it
static int
ft_note_pin_by_checkpoint (CACHEFILE UU(cachefile), void *header_v)
{
    //Set arbitrary brt (for given header) as pinned by checkpoint.
    //Only one can be pinned (only one checkpoint at a time), but not worth verifying.
    FT ft = (FT) header_v;

    // Note: open_close lock is held by checkpoint begin
    toku_ft_grab_reflock(ft);
    assert(!ft->pinned_by_checkpoint);
    assert(toku_ft_needed_unlocked(ft));
    ft->pinned_by_checkpoint = true;
    toku_ft_release_reflock(ft);
    return 0;
}

static void
unpin_by_checkpoint_callback(FT ft, void *extra) 
// Requires: the reflock is held.
{
    invariant(extra == NULL);
    invariant(ft->pinned_by_checkpoint);
    ft->pinned_by_checkpoint = false; //Unpin
}

// maps to cf->note_unpin_by_checkpoint
//Must be protected by ydb lock.
//Called by end_checkpoint, which grabs ydb lock around note_unpin
static int
ft_note_unpin_by_checkpoint (CACHEFILE UU(cachefile), void *header_v)
{
    FT ft = (FT) header_v;
    toku_ft_remove_reference(ft, false, ZERO_LSN, unpin_by_checkpoint_callback, NULL);
    return 0;
}

//
// End of Functions that are callbacks to the cachefile
/////////////////////////////////////////////////////////////////////////


void toku_node_save_ct_pair(void *value_data, PAIR p) {
    FTNODE CAST_FROM_VOIDP(node, value_data);
    node->ct_pair = p;
}

static int setup_initial_ft_root_node (FT ft, BLOCKNUM blocknum) {
    FTNODE XMALLOC(node);
    toku_initialize_empty_ftnode(node, blocknum, 0, 1, ft->h->layout_version, ft->h->flags);
    BP_STATE(node,0) = PT_AVAIL;

    uint32_t fullhash = toku_cachetable_hash(ft->cf, blocknum);
    node->fullhash = fullhash;
    int r = toku_cachetable_put(ft->cf, blocknum, fullhash,
                                node, make_ftnode_pair_attr(node),
                                get_write_callbacks_for_node(ft),
                                toku_node_save_ct_pair);
    if (r != 0)
        toku_free(node);
    else
        toku_unpin_ftnode(ft, node);
    return r;
}

static int
ft_init(FT ft, FT_OPTIONS options, CACHEFILE cf) {
    ft->checkpoint_header = NULL;
    ft->layout_version_read_from_disk = FT_LAYOUT_VERSION;             // fake, prevent unnecessary upgrade logic

    toku_list_init(&ft->live_ft_handles);

    ft->compare_fun = options->compare_fun;
    ft->update_fun = options->update_fun;

    if (ft->cf != NULL) {
        assert(ft->cf == cf);
    }
    ft->cf = cf;
    ft->in_memory_stats = ZEROSTATS;

    int r;
    r = setup_initial_ft_root_node(ft, ft->h->root_blocknum);
    if (r != 0) {
        goto exit;
    }
    //printf("%s:%d putting %p (%d)\n", __FILE__, __LINE__, ft, 0);
    toku_cachefile_set_userdata(ft->cf,
                                ft,
                                ft_log_fassociate_during_checkpoint,
                                ft_log_suppress_rollback_during_checkpoint,
                                ft_close,
                                ft_checkpoint,
                                ft_begin_checkpoint,
                                ft_end_checkpoint,
                                ft_note_pin_by_checkpoint,
                                ft_note_unpin_by_checkpoint);

    toku_block_verify_no_free_blocknums(ft->blocktable);
    r = 0;
exit:
    return r;
}


static FT_HEADER
ft_header_new(FT_OPTIONS options, BLOCKNUM root_blocknum, TXNID root_xid_that_created)
{
    uint64_t now = (uint64_t) time(NULL);
    struct ft_header h = {
        .type = FT_CURRENT,
        .dirty = 0,
        .checkpoint_count = 0,
        .checkpoint_lsn = ZERO_LSN,
        .layout_version = FT_LAYOUT_VERSION,
        .layout_version_original = FT_LAYOUT_VERSION,
        .build_id = BUILD_ID,
        .build_id_original = BUILD_ID,
        .time_of_creation = now,
        .root_xid_that_created = root_xid_that_created,
        .time_of_last_modification = now,
        .time_of_last_verification = 0,
        .root_blocknum = root_blocknum,
        .flags = options->flags,
        .nodesize = options->nodesize,
        .basementnodesize = options->basementnodesize,
        .compression_method = options->compression_method,
        .highest_unused_msn_for_upgrade = { .msn = (MIN_MSN.msn - 1) },
        .time_of_last_optimize_begin = 0,
        .time_of_last_optimize_end = 0,
        .count_of_optimize_in_progress = 0,
        .count_of_optimize_in_progress_read_from_disk = 0,
        .msn_at_start_of_last_completed_optimize = ZERO_MSN,
        .on_disk_stats = ZEROSTATS
    };
    return (FT_HEADER) toku_xmemdup(&h, sizeof h);
}

// allocate and initialize a fractal tree.
// t->ft->cf is not set to anything. TODO(leif): I don't think that's true
int
toku_create_new_ft(FT *ftp, FT_OPTIONS options, CACHEFILE cf, TOKUTXN txn) {
    int r;
    invariant(ftp);

    FT XCALLOC(ft);

    memset(&ft->descriptor, 0, sizeof(ft->descriptor));
    memset(&ft->cmp_descriptor, 0, sizeof(ft->cmp_descriptor));

    ft->h = ft_header_new(options, make_blocknum(0), (txn ? txn->ancestor_txnid64 : TXNID_NONE));

    toku_ft_init_reflock(ft);
    toku_blocktable_create_new(&ft->blocktable);
    //Assign blocknum for root block, also dirty the header
    toku_allocate_blocknum(ft->blocktable, &ft->h->root_blocknum, ft);

    r = ft_init(ft, options, cf);
    if (r != 0) {
        goto exit;
    }

    *ftp = ft;
    r = 0;
exit:
    if (r != 0) {
        if (ft) {
            toku_free(ft);
            ft = NULL;
        }
        return r;
    }
    return r;
}

// TODO: (Zardosht) get rid of brt parameter
int toku_read_ft_and_store_in_cachefile (FT_HANDLE brt, CACHEFILE cf, LSN max_acceptable_lsn, FT *header, bool* was_open)
// If the cachefile already has the header, then just get it.
// If the cachefile has not been initialized, then don't modify anything.
// max_acceptable_lsn is the latest acceptable checkpointed version of the file.
{
    {
        FT h;
        if ((h = (FT) toku_cachefile_get_userdata(cf))!=0) {
            *header = h;
            *was_open = true;
            assert(brt->options.update_fun == h->update_fun);
            assert(brt->options.compare_fun == h->compare_fun);
            return 0;
        }
    }
    *was_open = false;
    FT h;
    int r;
    {
        int fd = toku_cachefile_get_fd(cf);
        r = toku_deserialize_ft_from(fd, max_acceptable_lsn, &h);
        if (r == TOKUDB_BAD_CHECKSUM) {
            fprintf(stderr, "Checksum failure while reading header in file %s.\n", toku_cachefile_fname_in_env(cf));
            assert(false);  // make absolutely sure we crash before doing anything else
        }
    }
    if (r!=0) return r;
    h->cf = cf;
    h->compare_fun = brt->options.compare_fun;
    h->update_fun = brt->options.update_fun;
    toku_cachefile_set_userdata(cf,
                                (void*)h,
                                ft_log_fassociate_during_checkpoint,
                                ft_log_suppress_rollback_during_checkpoint,
                                ft_close,
                                ft_checkpoint,
                                ft_begin_checkpoint,
                                ft_end_checkpoint,
                                ft_note_pin_by_checkpoint,
                                ft_note_unpin_by_checkpoint);
    *header = h;
    return 0;
}

void
toku_ft_note_ft_handle_open(FT ft, FT_HANDLE live) {
    toku_ft_grab_reflock(ft);
    live->ft = ft;
    toku_list_push(&ft->live_ft_handles, &live->live_ft_handle_link);
    toku_ft_release_reflock(ft);
}

// the reference count for a ft is the number of txn's that
// touched it plus the number of open handles plus one if
// pinned by a checkpoint.
static int
ft_get_reference_count(FT ft) {
    uint32_t pinned_by_checkpoint = ft->pinned_by_checkpoint ? 1 : 0;
    int num_handles = toku_list_num_elements_est(&ft->live_ft_handles);
    return pinned_by_checkpoint + ft->num_txns + num_handles;
}

// a ft is needed in memory iff its reference count is non-zero
bool
toku_ft_needed_unlocked(FT ft) {
    return ft_get_reference_count(ft) != 0;
}

// get the reference count and return true if it was 1
bool
toku_ft_has_one_reference_unlocked(FT ft) {
    return ft_get_reference_count(ft) == 1;
}

// evict a ft from memory by closing its cachefile. any future work
// will have to read in the ft in a new cachefile and new FT object.
int toku_ft_evict_from_memory(FT ft, char **error_string, bool oplsn_valid, LSN oplsn) {
    int r = 0;
    assert(ft->cf);
    if (error_string) {
        assert(*error_string == 0);
    }
    r = toku_cachefile_close(&ft->cf, error_string, oplsn_valid, oplsn);
    if (r == 0 && error_string) {
        assert(*error_string == 0);
    }
    return r;
}

// Verifies there exists exactly one ft handle and returns it.
FT_HANDLE toku_ft_get_only_existing_ft_handle(FT h) {
    FT_HANDLE ft_handle_ret = NULL;
    toku_ft_grab_reflock(h);
    assert(toku_list_num_elements_est(&h->live_ft_handles) == 1);
    ft_handle_ret = toku_list_struct(toku_list_head(&h->live_ft_handles), struct ft_handle, live_ft_handle_link);
    toku_ft_release_reflock(h);
    return ft_handle_ret;
}

// Purpose: set fields in brt_header to capture accountability info for start of HOT optimize.
// Note: HOT accountability variables in header are modified only while holding header lock.
//       (Header lock is really needed for touching the dirty bit, but it's useful and 
//       convenient here for keeping the HOT variables threadsafe.)
void
toku_ft_note_hot_begin(FT_HANDLE brt) {
    FT ft = brt->ft;
    time_t now = time(NULL);

    // hold lock around setting and clearing of dirty bit
    // (see cooperative use of dirty bit in ft_begin_checkpoint())
    toku_ft_lock(ft);
    ft->h->time_of_last_optimize_begin = now;
    ft->h->count_of_optimize_in_progress++;
    ft->h->dirty = 1;
    toku_ft_unlock(ft);
}


// Purpose: set fields in brt_header to capture accountability info for end of HOT optimize.
// Note: See note for toku_ft_note_hot_begin().
void
toku_ft_note_hot_complete(FT_HANDLE brt, bool success, MSN msn_at_start_of_hot) {
    FT ft = brt->ft;
    time_t now = time(NULL);

    toku_ft_lock(ft);
    ft->h->count_of_optimize_in_progress--;
    if (success) {
        ft->h->time_of_last_optimize_end = now;
        ft->h->msn_at_start_of_last_completed_optimize = msn_at_start_of_hot;
        // If we just successfully completed an optimization and no other thread is performing
        // an optimization, then the number of optimizations in progress is zero.
        // If there was a crash during a HOT optimization, this is how count_of_optimize_in_progress
        // would be reset to zero on the disk after recovery from that crash.  
        if (ft->h->count_of_optimize_in_progress == ft->h->count_of_optimize_in_progress_read_from_disk)
            ft->h->count_of_optimize_in_progress = 0;
    }
    ft->h->dirty = 1;
    toku_ft_unlock(ft);
}


void
toku_ft_init(FT ft,
             BLOCKNUM root_blocknum_on_disk,
             LSN checkpoint_lsn,
             TXNID root_xid_that_created,
             uint32_t target_nodesize,
             uint32_t target_basementnodesize,
             enum toku_compression_method compression_method)
{
    memset(ft, 0, sizeof *ft);
    struct ft_options options = {
        .nodesize = target_nodesize,
        .basementnodesize = target_basementnodesize,
        .compression_method = compression_method,
        .flags = 0,
        .compare_fun = NULL,
        .update_fun = NULL
    };
    ft->h = ft_header_new(&options, root_blocknum_on_disk, root_xid_that_created);
    ft->h->checkpoint_count = 1;
    ft->h->checkpoint_lsn   = checkpoint_lsn;
}

// Open a brt for use by redirect.  The new brt must have the same dict_id as the old_ft passed in.  (FILENUM is assigned by the ft_handle_open() function.)
static int
ft_handle_open_for_redirect(FT_HANDLE *new_ftp, const char *fname_in_env, TOKUTXN txn, FT old_h) {
    int r;
    FT_HANDLE t;
    assert(old_h->dict_id.dictid != DICTIONARY_ID_NONE.dictid);
    r = toku_ft_handle_create(&t);
    assert_zero(r);
    r = toku_ft_set_bt_compare(t, old_h->compare_fun);
    assert_zero(r);
    r = toku_ft_set_update(t, old_h->update_fun);
    assert_zero(r);
    toku_ft_handle_set_nodesize(t, old_h->h->nodesize);
    toku_ft_handle_set_basementnodesize(t, old_h->h->basementnodesize);
    toku_ft_handle_set_compression_method(t, old_h->h->compression_method);
    CACHETABLE ct = toku_cachefile_get_cachetable(old_h->cf);
    r = toku_ft_handle_open_with_dict_id(t, fname_in_env, 0, 0, ct, txn, old_h->dict_id);
    assert_zero(r);
    assert(t->ft->dict_id.dictid == old_h->dict_id.dictid);

    *new_ftp = t;
    return r;
}

// This function performs most of the work to redirect a dictionary to different file.
// It is called for redirect and to abort a redirect.  (This function is almost its own inverse.)
static int
dictionary_redirect_internal(const char *dst_fname_in_env, FT src_h, TOKUTXN txn, FT *dst_hp) {
    int r;

    FILENUM src_filenum = toku_cachefile_filenum(src_h->cf);
    FILENUM dst_filenum = FILENUM_NONE;

    FT dst_h = NULL;
    struct toku_list *list;
    // open a dummy brt based off of 
    // dst_fname_in_env to get the header
    // then we will change all the brt's to have
    // their headers point to dst_h instead of src_h
    FT_HANDLE tmp_dst_ft = NULL;
    r = ft_handle_open_for_redirect(&tmp_dst_ft, dst_fname_in_env, txn, src_h);
    assert_zero(r);
    dst_h = tmp_dst_ft->ft;

    // some sanity checks on dst_filenum
    dst_filenum = toku_cachefile_filenum(dst_h->cf);
    assert(dst_filenum.fileid!=FILENUM_NONE.fileid);
    assert(dst_filenum.fileid!=src_filenum.fileid); //Cannot be same file.

    // for each live brt, brt->ft is currently src_h
    // we want to change it to dummy_dst
    toku_ft_grab_reflock(src_h);
    while (!toku_list_empty(&src_h->live_ft_handles)) {
        list = src_h->live_ft_handles.next;
        FT_HANDLE src_handle = NULL;
        src_handle = toku_list_struct(list, struct ft_handle, live_ft_handle_link);

        toku_list_remove(&src_handle->live_ft_handle_link);
        
        toku_ft_note_ft_handle_open(dst_h, src_handle);
        if (src_handle->redirect_callback) {
            src_handle->redirect_callback(src_handle, src_handle->redirect_callback_extra);
        }
    }
    assert(dst_h);
    // making sure that we are not leaking src_h
    assert(toku_ft_needed_unlocked(src_h));
    toku_ft_release_reflock(src_h);

    toku_ft_handle_close(tmp_dst_ft);

    *dst_hp = dst_h;
    return r;
}



//This is the 'abort redirect' function.  The redirect of old_h to new_h was done
//and now must be undone, so here we redirect new_h back to old_h.
int
toku_dictionary_redirect_abort(FT old_h, FT new_h, TOKUTXN txn) {
    char *old_fname_in_env = toku_cachefile_fname_in_env(old_h->cf);
    int r;
    {
        FILENUM old_filenum = toku_cachefile_filenum(old_h->cf);
        FILENUM new_filenum = toku_cachefile_filenum(new_h->cf);
        assert(old_filenum.fileid!=new_filenum.fileid); //Cannot be same file.

        //No living brts in old header.
        toku_ft_grab_reflock(old_h);
        assert(toku_list_empty(&old_h->live_ft_handles));
        toku_ft_release_reflock(old_h);
    }

    FT dst_h;
    // redirect back from new_h to old_h
    r = dictionary_redirect_internal(old_fname_in_env, new_h, txn, &dst_h);
    assert_zero(r);
    assert(dst_h == old_h);
    return r;
}

/****
 * on redirect or abort:
 *  if redirect txn_note_doing_work(txn)
 *  if redirect connect src brt to txn (txn modified this brt)
 *  for each src brt
 *    open brt to dst file (create new brt struct)
 *    if redirect connect dst brt to txn 
 *    redirect db to new brt
 *    redirect cursors to new brt
 *  close all src brts
 *  if redirect make rollback log entry
 * 
 * on commit:
 *   nothing to do
 *
 *****/

int 
toku_dictionary_redirect (const char *dst_fname_in_env, FT_HANDLE old_ft_h, TOKUTXN txn) {
// Input args:
//   new file name for dictionary (relative to env)
//   old_ft_h is a live brt of open handle ({DB, BRT} pair) that currently refers to old dictionary file.
//   (old_ft_h may be one of many handles to the dictionary.)
//   txn that created the loader
// Requires: 
//   multi operation lock is held.
//   The brt is open.  (which implies there can be no zombies.)
//   The new file must be a valid dictionary.
//   The block size and flags in the new file must match the existing BRT.
//   The new file must already have its descriptor in it (and it must match the existing descriptor).
// Effect:   
//   Open new FTs (and related header and cachefile) to the new dictionary file with a new FILENUM.
//   Redirect all DBs that point to brts that point to the old file to point to brts that point to the new file.
//   Copy the dictionary id (dict_id) from the header of the original file to the header of the new file.
//   Create a rollback log entry.
//   The original BRT, header, cachefile and file remain unchanged.  They will be cleaned up on commmit.
//   If the txn aborts, then this operation will be undone
    int r;

    FT old_ft = old_ft_h->ft;

    // dst file should not be open.  (implies that dst and src are different because src must be open.)
    {
        CACHETABLE ct = toku_cachefile_get_cachetable(old_ft->cf);
        CACHEFILE cf;
        r = toku_cachefile_of_iname_in_env(ct, dst_fname_in_env, &cf);
        if (r==0) {
            r = EINVAL;
            goto cleanup;
        }
        assert(r==ENOENT);
        r = 0;
    }

    if (txn) {
        toku_txn_maybe_note_ft(txn, old_ft);  // mark old ft as touched by this txn
    }

    FT new_ft;
    r = dictionary_redirect_internal(dst_fname_in_env, old_ft, txn, &new_ft);
    assert_zero(r);

    // make rollback log entry
    if (txn) {
        toku_txn_maybe_note_ft(txn, new_ft); // mark new ft as touched by this txn

        FILENUM old_filenum = toku_cachefile_filenum(old_ft->cf);
        FILENUM new_filenum = toku_cachefile_filenum(new_ft->cf);
        r = toku_logger_save_rollback_dictionary_redirect(txn, old_filenum, new_filenum);
        assert_zero(r);

        TXNID xid = toku_txn_get_txnid(txn);
        toku_ft_suppress_rollbacks(new_ft, txn);
        r = toku_log_suppress_rollback(txn->logger, NULL, 0, txn, new_filenum, xid);
        assert_zero(r);
    }
    
cleanup:
    return r;
}

// Insert reference to transaction into ft
void
toku_ft_add_txn_ref(FT ft) {
    toku_ft_grab_reflock(ft);
    ++ft->num_txns;
    toku_ft_release_reflock(ft);
}

static void
remove_txn_ref_callback(FT ft, void *UU(context)) {
    invariant(ft->num_txns > 0);
    --ft->num_txns;
}

void
toku_ft_remove_txn_ref(FT ft) {
    toku_ft_remove_reference(ft, false, ZERO_LSN, remove_txn_ref_callback, NULL);
}

void toku_calculate_root_offset_pointer (
    FT ft, 
    CACHEKEY* root_key, 
    uint32_t *roothash
    ) 
{
    *roothash = toku_cachetable_hash(ft->cf, ft->h->root_blocknum);
    *root_key = ft->h->root_blocknum;
}

void toku_ft_set_new_root_blocknum(
    FT ft, 
    CACHEKEY new_root_key
    ) 
{
    ft->h->root_blocknum = new_root_key;
}

LSN toku_ft_checkpoint_lsn(FT ft) {
    return ft->h->checkpoint_lsn;
}

int toku_ft_set_panic(FT ft, int panic, const char *panic_string) {
    if (ft->panic == 0) {
        ft->panic = panic;
        if (ft->panic_string) {
            toku_free(ft->panic_string);
        }
        ft->panic_string = toku_strdup(panic_string);
    }
    return 0;
}

void 
toku_ft_stat64 (FT ft, struct ftstat64_s *s) {
    s->fsize = toku_cachefile_size(ft->cf);
    // just use the in memory stats from the header
    // prevent appearance of negative numbers for numrows, numbytes
    int64_t n = ft->in_memory_stats.numrows;
    if (n < 0) {
        n = 0;
    }
    s->nkeys = s->ndata = n;
    n = ft->in_memory_stats.numbytes;
    if (n < 0) {
        n = 0;
    }
    s->dsize = n; 
    s->create_time_sec = ft->h->time_of_creation;
    s->modify_time_sec = ft->h->time_of_last_modification;
    s->verify_time_sec = ft->h->time_of_last_verification;    
}

void 
toku_ft_update_descriptor(FT ft, DESCRIPTOR d) 
// Effect: Changes the descriptor in a tree (log the change, make sure it makes it to disk eventually).
// requires: the ft is fully user-opened with a valid cachefile.
//           descriptor updates cannot happen in parallel for an FT 
//           (ydb layer uses a row lock to enforce this)
{
    assert(ft->cf);
    int fd = toku_cachefile_get_fd(ft->cf);
    toku_ft_update_descriptor_with_fd(ft, d, fd);
}

// upadate the descriptor for an ft and serialize it using
// the given descriptor instead of reading the descriptor
// from the ft's cachefile. we do this so serialize code can
// update a descriptor before the ft is fully opened and has
// a valid cachefile.
void
toku_ft_update_descriptor_with_fd(FT ft, DESCRIPTOR d, int fd) {
    // the checksum is four bytes, so that's where the magic number comes from
    // make space for the new descriptor and write it out to disk
    DISKOFF offset, size;
    size = toku_serialize_descriptor_size(d) + 4;
    toku_realloc_descriptor_on_disk(ft->blocktable, size, &offset, ft, fd);
    toku_serialize_descriptor_contents_to_fd(fd, d, offset);

    // cleanup the old descriptor and set the in-memory descriptor to the new one
    if (ft->descriptor.dbt.data) {
        toku_free(ft->descriptor.dbt.data);
    }
    ft->descriptor.dbt.size = d->dbt.size;
    ft->descriptor.dbt.data = toku_memdup(d->dbt.data, d->dbt.size);
}

void 
toku_ft_update_cmp_descriptor(FT ft) {
    if (ft->cmp_descriptor.dbt.data != NULL) {
        toku_free(ft->cmp_descriptor.dbt.data);
    }
    ft->cmp_descriptor.dbt.size = ft->descriptor.dbt.size;
    ft->cmp_descriptor.dbt.data = toku_xmemdup(
        ft->descriptor.dbt.data, 
        ft->descriptor.dbt.size
        );
}

DESCRIPTOR
toku_ft_get_descriptor(FT_HANDLE ft_handle) {
    return &ft_handle->ft->descriptor;
}

DESCRIPTOR
toku_ft_get_cmp_descriptor(FT_HANDLE ft_handle) {
    return &ft_handle->ft->cmp_descriptor;
}

void
toku_ft_update_stats(STAT64INFO headerstats, STAT64INFO_S delta) {
    (void) __sync_fetch_and_add(&(headerstats->numrows),  delta.numrows);
    (void) __sync_fetch_and_add(&(headerstats->numbytes), delta.numbytes);
}

void
toku_ft_decrease_stats(STAT64INFO headerstats, STAT64INFO_S delta) {
    (void) __sync_fetch_and_sub(&(headerstats->numrows),  delta.numrows);
    (void) __sync_fetch_and_sub(&(headerstats->numbytes), delta.numbytes);
}

void
toku_ft_remove_reference(FT ft, bool oplsn_valid, LSN oplsn, remove_ft_ref_callback remove_ref, void *extra) {
    toku_ft_grab_reflock(ft);
    if (toku_ft_has_one_reference_unlocked(ft)) {
        toku_ft_release_reflock(ft);

        toku_ft_open_close_lock();
        toku_ft_grab_reflock(ft);

        remove_ref(ft, extra);
        bool needed = toku_ft_needed_unlocked(ft);
        toku_ft_release_reflock(ft);

        // if we're running during recovery, we must close the underlying ft.
        // we know we're running in recovery if we were passed a valid lsn.
        if (oplsn_valid) {
            assert(!needed);
        }
        if (!needed) {
            // close header
            char *error_string = NULL;
            int r;
            r = toku_ft_evict_from_memory(ft, &error_string, oplsn_valid, oplsn);
            assert_zero(r);
            assert(error_string == NULL);
        }
    
        toku_ft_open_close_unlock();
    }
    else {
        remove_ref(ft, extra);
        toku_ft_release_reflock(ft);
    }
}

void toku_ft_set_nodesize(FT ft, unsigned int nodesize) {
    toku_ft_lock(ft);
    ft->h->nodesize = nodesize;
    ft->h->dirty = 1;
    toku_ft_unlock(ft);
}

void toku_ft_get_nodesize(FT ft, unsigned int *nodesize) {
    toku_ft_lock(ft);
    *nodesize = ft->h->nodesize;
    toku_ft_unlock(ft);
}

void toku_ft_set_basementnodesize(FT ft, unsigned int basementnodesize) {
    toku_ft_lock(ft);
    ft->h->basementnodesize = basementnodesize;
    ft->h->dirty = 1;
    toku_ft_unlock(ft);
}

void toku_ft_get_basementnodesize(FT ft, unsigned int *basementnodesize) {
    toku_ft_lock(ft);
    *basementnodesize = ft->h->basementnodesize;
    toku_ft_unlock(ft);
}

void toku_ft_set_compression_method(FT ft, enum toku_compression_method method) {
    toku_ft_lock(ft);
    ft->h->compression_method = method;
    ft->h->dirty = 1;
    toku_ft_unlock(ft);
}

void toku_ft_get_compression_method(FT ft, enum toku_compression_method *methodp) {
    toku_ft_lock(ft);
    *methodp = ft->h->compression_method;
    toku_ft_unlock(ft);
}

// mark the ft as a blackhole. any message injections will be a no op.
void toku_ft_set_blackhole(FT_HANDLE ft_handle) {
    ft_handle->ft->blackhole = true;
}