hdf5/testpar/t_cache.c

/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
 * Copyright by The HDF Group.                                               *
 * Copyright by the Board of Trustees of the University of Illinois.         *
 * All rights reserved.                                                      *
 *                                                                           *
 * This file is part of HDF5.  The full HDF5 copyright notice, including     *
 * terms governing use, modification, and redistribution, is contained in    *
 * the COPYING file, which can be found at the root of the source code       *
 * distribution tree, or in https://support.hdfgroup.org/ftp/HDF5/releases.  *
 * If you do not have access to either file, you may request a copy from     *
 * help@hdfgroup.org.                                                        *
 * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */

/*
 * Parallel metadata cache tests.
 *
 */

#include "testpar.h"

#define H5AC_FRIEND        /*suppress error about including H5ACpkg  */
#define H5C_FRIEND        /*suppress error about including H5Cpkg   */
#define H5F_FRIEND        /*suppress error about including H5Fpkg   */

#include "H5ACpkg.h"
#include "H5Cpkg.h"
#include "H5CXprivate.h"
#include "H5Fpkg.h"
#include "H5Iprivate.h"
#include "H5MFprivate.h"
#include "H5private.h"

#define BASE_ADDR               (haddr_t)1024


int         nerrors = 0;
int        failures = 0;
hbool_t        verbose = TRUE; /* used to control error messages */

#define NFILENAME 2
const char *FILENAME[NFILENAME]={"CacheTestDummy", NULL};
#ifndef PATH_MAX
#define PATH_MAX    512
#endif  /* !PATH_MAX */
char    filenames[NFILENAME][PATH_MAX];
hid_t   fapl;                           /* file access property list */
haddr_t max_addr = 0;            /* used to store the end of
                    * the address space used by
                    * the data array (see below).
                    */
hbool_t callbacks_verbose = FALSE;    /* flag used to control whether
                    * the callback functions are in
                    * verbose mode.
                    */


int        world_mpi_size = -1;
int        world_mpi_rank = -1;
int        world_server_mpi_rank = -1;
MPI_Comm    world_mpi_comm = MPI_COMM_NULL;
int        file_mpi_size = -1;
int        file_mpi_rank = -1;
MPI_Comm    file_mpi_comm = MPI_COMM_NULL;


/* the following globals are used to maintain rudementary statistics
 * to check the validity of the statistics maintained by H5C.c
 */

long datum_clears          = 0;
long datum_pinned_clears   = 0;
long datum_destroys        = 0;
long datum_flushes         = 0;
long datum_pinned_flushes  = 0;
long datum_loads           = 0;
long global_pins       = 0;
long global_dirty_pins       = 0;
long local_pins           = 0;


/* the following fields are used by the server process only */
int total_reads           = 0;
int total_writes           = 0;


/*****************************************************************************
 * struct datum
 *
 *    Instances of struct datum are used to store information on entries
 *    that may be loaded into the cache.  The individual fields are
 *    discussed below:
 *
 *    header:    Instance of H5C_cache_entry_t used by the for its data.
 *        This field is only used on the file processes, not on the
 *        server process.
 *
 *        This field MUST be the first entry in this structure.
 *
 *    base_addr: Base address of the entry.
 *
 *    len:    Length of the entry.
 *
 *    local_len:  Length of the entry according to the cache.  This
 *        value must be positive, and may not be larger than len.
 *
 *        The field exists to allow us change the sizes of entries
 *        in the cache without upsetting the server.  This value
 *        is only used locally, and is never sent to the server.
 *
 *    ver:    Version number of the entry.  This number is initialize
 *        to zero, and incremented each time the entry is modified.
 *
 *    dirty:    Boolean flag indicating whether the entry is dirty.
 *
 *        For current purposes, an entry is clean until it is
 *        modified, and dirty until written to the server (cache
 *        on process 0) or until it is marked clean (all other
 *        caches).
 *
 *    valid:    Boolean flag indicating whether the entry contains
 *        valid data.  Attempts to read an entry whose valid
 *        flag is not set should trigger an error.
 *
 *    locked:    Boolean flag that is set to true iff the entry is in
 *        the cache and locked.
 *
 *    global_pinned:    Boolean flag that is set to true iff the entry has
 *        been pinned collectively in all caches.  Since writes must
 *        be collective across all processes, only entries pinned
 *        in this fashion may be marked dirty.
 *
 *    local_pinned:    Boolean flag that is set to true iff the entry
 *        has been pinned in the local cache, but probably not all
 *        caches.  Such pins will typically not be consistant across
 *        processes, and thus cannot be marked as dirty unless they
 *        happen to overlap some collective operation.
 *
 *      cleared: Boolean flag that is set to true whenever the entry is
 *              dirty, and is cleared via a call to datum_notify with the
 *              "entry cleaned" action.
 *
 *      flushed: Boolean flag that is set to true whenever the entry is
 *              dirty, and is flushed by the metadata cache.
 *
 *    reads:  Integer field used to maintain a count of the number of
 *        times this entry has been read from the server since
 *        the last time the read and write counts were reset.
 *
 *    writes: Integer field used to maintain a count of the number of
 *        times this entry has been written to the server since
 *        the last time the read and write counts were reset.
 *
 *    index:    Index of this instance of datum in the data_index[] array
 *        discussed below.
 *
 *    aux_ptr: Pointer to the instance of H5AC_aux_t associated with the
 *        instance of the metadata cache within which this entry
 *        resides.  This field was added to allow us to pass this
 *        value to the notify callback from the serialize callback.
 *        It should be NULL when not in use.
 *
 *****************************************************************************/

struct datum
{
    H5C_cache_entry_t    header;
    haddr_t        base_addr;
    size_t        len;
    size_t        local_len;
    int            ver;
    hbool_t        dirty;
    hbool_t        valid;
    hbool_t        locked;
    hbool_t        global_pinned;
    hbool_t        local_pinned;
    hbool_t        cleared;
    hbool_t             flushed;
    int            reads;
    int            writes;
    int            index;
    struct H5AC_aux_t * aux_ptr;
};

/*****************************************************************************
 * data array
 *
 *    The data array is an array of instances of datum of size
 *    NUM_DATA_ENTRIES that is used to track the particulars of all
 *    the entries that may be loaded into the cache.
 *
 *    It exists on all processes, although the master copy is maintained
 *    by the server process.  If the cache is performing correctly, all
 *    versions should be effectively identical.  By that I mean that
 *    the data received from the server should always match that in
 *    the local version of the data array.
 *
 *****************************************************************************/

#define NUM_DATA_ENTRIES    100000

struct datum data[NUM_DATA_ENTRIES];


/* Many tests use the size of data array as the size of test loops.
 * On some machines, this results in unacceptably long test runs.
 *
 * To deal with this issue, I have introduced the virt_num_data_entries
 * global, which can be set to a lower value to throtle the length of
 * tests.
 *
 * Note that this value must always be divisible by 40, and must be an
 * even divisor of NUM_DATA_ENTRIES.  So far, all tests have been with
 * powers of 10 that meet these criteria.
 *
 * Further, this value must be consistant across all processes.
 */

#define STD_VIRT_NUM_DATA_ENTRIES    NUM_DATA_ENTRIES
#define EXPRESS_VIRT_NUM_DATA_ENTRIES    (NUM_DATA_ENTRIES / 10)
/* Use a smaller test size to avoid creating huge MPE logfiles. */
#ifdef H5_HAVE_MPE
#define MPE_VIRT_NUM_DATA_ENTIES    (NUM_DATA_ENTRIES / 100)
#endif

int virt_num_data_entries = NUM_DATA_ENTRIES;


/*****************************************************************************
 * data_index array
 *
 *    The data_index array is an array of integer used to maintain a list
 *    of instances of datum in the data array in increasing base_addr order.
 *
 *    This array is necessary, as move operations can swap the values
 *    of the base_addr fields of two instances of datum.  Without this
 *    array, we would no longer be able to use a binary search on a sorted
 *    list to find the indexes of instances of datum given the values of
 *    their base_addr fields.
 *
 *****************************************************************************/

int data_index[NUM_DATA_ENTRIES];


/*****************************************************************************
 * The following two #defines are used to control code that is in turn used
 * to force "POSIX" semantics on the server process used to simulate metadata
 * reads and writes.  Without some such mechanism, the test code contains
 * race conditions that will frequently cause spurious failures.
 *
 * When set to TRUE, DO_WRITE_REQ_ACK forces the server to send an ack after
 * each write request, and the client to wait until the ack is received
 * before proceeding.  This was my first solution to the problem, and at
 * first glance, it would seem to have a lot of unnecessary overhead.
 *
 * In an attempt to reduce the overhead, I implemented a second solution
 * in which no acks are sent after writes.  Instead, the metadata cache is
 * provided with a callback function to call after each sequence of writes.
 * This callback simply causes the client to send the server process a
 * "sync" message and and await an ack in reply.
 *
 * Strangely, at least on Phoenix, the first solution runs faster by a
 * rather large margin.  However, I can imagine this changing with
 * different OS's and MPI implementatins.
 *
 * Thus I have left code supporting the second solution in place.
 *
 * Note that while one of these two #defines must be set to TRUE, there
 * should never be any need to set both of them to TRUE (although the
 * tests will still function with this setting).
 *****************************************************************************/

#define DO_WRITE_REQ_ACK    TRUE
#define DO_SYNC_AFTER_WRITE    FALSE


/*****************************************************************************
 * struct mssg
 *
 *    The mssg structure is used as a generic container for messages to
 *    and from the server.  Not all fields are used in all cases.
 *
 *    req:    Integer field containing the type of the message.
 *
 *    src:    World communicator MPI rank of the sending process.
 *
 *    dest:    World communicator MPI rank of the destination process.
 *
 *    mssg_num:    Serial number assigned to the message by the sender.
 *
 *    base_addr:    Base address of a datum.  Not used in all mssgs.
 *
 *    len:    Length of a datum (in bytes).  Not used in all mssgs.
 *
 *    ver:    Version number of a datum.  Not used in all mssgs.
 *
 *    count:  Reported number of total/entry reads/writes.  Not used
 *        in all mssgs.
 *
 *    magic:    Magic number for error detection.  Must be set to
 *        MSSG_MAGIC.
 *
 *****************************************************************************/

#define    WRITE_REQ_CODE             0
#define    WRITE_REQ_ACK_CODE         1
#define READ_REQ_CODE             2
#define READ_REQ_REPLY_CODE         3
#define SYNC_REQ_CODE             4
#define SYNC_ACK_CODE             5
#define REQ_TTL_WRITES_CODE         6
#define REQ_TTL_WRITES_RPLY_CODE      7
#define REQ_TTL_READS_CODE         8
#define REQ_TTL_READS_RPLY_CODE      9
#define REQ_ENTRY_WRITES_CODE        10
#define REQ_ENTRY_WRITES_RPLY_CODE     11
#define REQ_ENTRY_READS_CODE        12
#define REQ_ENTRY_READS_RPLY_CODE     13
#define REQ_RW_COUNT_RESET_CODE        14
#define REQ_RW_COUNT_RESET_RPLY_CODE    15
#define DONE_REQ_CODE            16
#define MAX_REQ_CODE            16

#define MSSG_MAGIC    0x1248

struct mssg_t
{
    int        req;
    int        src;
    int        dest;
    long int    mssg_num;
    haddr_t    base_addr;
    unsigned    len;
    int        ver;
    unsigned    count;
    unsigned    magic;
};

MPI_Datatype mpi_mssg_t;    /* for MPI derived type created from mssg */


/*****************************************************************************/
/************************** function declarations ****************************/
/*****************************************************************************/

/* stats functions */

static void reset_stats(void);

/* MPI setup functions */

static hbool_t set_up_file_communicator(void);


/* data array manipulation functions */

static int addr_to_datum_index(haddr_t base_addr);
static void init_data(void);


/* test coodination related functions */

static int do_express_test(void);
static void do_sync(void);
static int get_max_nerrors(void);


/* mssg xfer related functions */

static hbool_t recv_mssg(struct mssg_t *mssg_ptr, int mssg_tag_offset);
static hbool_t send_mssg(struct mssg_t *mssg_ptr, hbool_t add_req_to_tag);
static hbool_t setup_derived_types(void);
static hbool_t takedown_derived_types(void);


/* server functions */

static hbool_t reset_server_counters(void);
static hbool_t server_main(void);
static hbool_t serve_read_request(struct mssg_t * mssg_ptr);
static hbool_t serve_sync_request(struct mssg_t * mssg_ptr);
static hbool_t serve_write_request(struct mssg_t * mssg_ptr);
static hbool_t serve_total_writes_request(struct mssg_t * mssg_ptr);
static hbool_t serve_total_reads_request(struct mssg_t * mssg_ptr);
static hbool_t serve_entry_writes_request(struct mssg_t * mssg_ptr);
static hbool_t serve_entry_reads_request(struct mssg_t * mssg_ptr);
static hbool_t serve_rw_count_reset_request(struct mssg_t * mssg_ptr);


/* call back functions & related data structures */

static herr_t datum_get_initial_load_size(void *udata_ptr,
                                  size_t *image_len_ptr);

static void * datum_deserialize(const void * image_ptr,
                                size_t len,
                                void * udata_ptr,
                                hbool_t * dirty_ptr);

static herr_t datum_image_len(const void *thing,
                              size_t *image_len_ptr);

static herr_t datum_serialize(const H5F_t *f,
                              void *image_ptr,
                              size_t len,
                              void *thing_ptr);

static herr_t datum_notify(H5C_notify_action_t action, void *thing);

static herr_t datum_free_icr(void * thing);

/* Masquerade as object header entries to the cache */
#define DATUM_ENTRY_TYPE    H5AC_OHDR_ID

#define NUMBER_OF_ENTRY_TYPES    1


/* Note the use of the H5AC__CLASS_SKIP_READS and H5AC__CLASS_SKIP_WRITES
 * flags.  As a result of these flags, the metadata cache does no file I/O
 * on metadata of the datum type.
 *
 * Instead, this test uses a server process to keep track of who has
 * written and read what, and to verify that there are no messages from
 * the past / future.
 *
 * In the callbacks for the version 2 cache, this activity was hidden in
 * the load and flush callbacks.  However, now we handle this function in
 * notify callbacks for the after load and after flush events.
 *
 *                         JRM -- 1/13/15
 */
const H5C_class_t types[NUMBER_OF_ENTRY_TYPES] =
{
  {
    /* id            */ DATUM_ENTRY_TYPE,
    /* name          */ "datum",
    /* mem_type      */ H5FD_MEM_OHDR,
    /* flags         */ H5AC__CLASS_SKIP_READS | H5AC__CLASS_SKIP_WRITES,
    /* get_initial_load_size */ datum_get_initial_load_size,
    /* get_final_load_size */ NULL,
    /* verify_chksum */ NULL,
    /* deserialize   */ datum_deserialize,
    /* image_len     */ datum_image_len,
    /* pre_serialize */ NULL,
    /* serialize     */ datum_serialize,
    /* notify        */ datum_notify,
    /* free_icr      */ datum_free_icr,
    /* fsf_size      */ NULL,
  }
};


/* test utility functions */

static void expunge_entry(H5F_t * file_ptr, int32_t idx);
static void insert_entry(H5C_t * cache_ptr, H5F_t * file_ptr,
                  int32_t idx, unsigned int flags);
static void local_pin_and_unpin_random_entries(H5F_t * file_ptr, int min_idx,
                                        int max_idx, int min_count,
                                        int max_count);
static void local_pin_random_entry(H5F_t * file_ptr, int min_idx, int max_idx);
static void local_unpin_all_entries(H5F_t * file_ptr, hbool_t via_unprotect);
static int local_unpin_next_pinned_entry(H5F_t * file_ptr, int start_idx,
                                  hbool_t via_unprotect);
static void lock_and_unlock_random_entries(H5F_t * file_ptr, int min_idx, int max_idx,
                                    int min_count, int max_count);
static void lock_and_unlock_random_entry(H5F_t * file_ptr,
                                  int min_idx, int max_idx);
static void lock_entry(H5F_t * file_ptr, int32_t idx);
static void mark_entry_dirty(int32_t idx);
static void pin_entry(H5F_t * file_ptr, int32_t idx, hbool_t global, hbool_t dirty);
static void pin_protected_entry(int32_t idx, hbool_t global);
static void move_entry(H5F_t * file_ptr, int32_t old_idx, int32_t new_idx);
static hbool_t reset_server_counts(void);
static void resize_entry(int32_t idx, size_t  new_size);
static hbool_t setup_cache_for_test(hid_t * fid_ptr,
                             H5F_t ** file_ptr_ptr,
                             H5C_t ** cache_ptr_ptr,
                             int metadata_write_strategy);
static void setup_rand(void);
static hbool_t take_down_cache(hid_t fid, H5C_t * cache_ptr);
static hbool_t verify_entry_reads(haddr_t addr, int expected_entry_reads);
static hbool_t verify_entry_writes(haddr_t addr, int expected_entry_writes);
static hbool_t verify_total_reads(int expected_total_reads);
static hbool_t verify_total_writes(unsigned expected_total_writes);
static void verify_writes(unsigned num_writes, haddr_t * written_entries_tbl);
static void unlock_entry(H5F_t * file_ptr, int32_t type, unsigned int flags);
static void unpin_entry(H5F_t * file_ptr, int32_t idx, hbool_t global,
                 hbool_t dirty, hbool_t via_unprotect);


/* test functions */

static hbool_t server_smoke_check(void);
static hbool_t smoke_check_1(int metadata_write_strategy);
static hbool_t smoke_check_2(int metadata_write_strategy);
static hbool_t smoke_check_3(int metadata_write_strategy);
static hbool_t smoke_check_4(int metadata_write_strategy);
static hbool_t smoke_check_5(int metadata_write_strategy);
static hbool_t smoke_check_6(int metadata_write_strategy);
static hbool_t trace_file_check(int metadata_write_strategy);


/*****************************************************************************/
/****************************** stats functions ******************************/
/*****************************************************************************/

#ifdef NOT_USED
/*****************************************************************************
 *
 * Function:    print_stats()
 *
 * Purpose:     Print the rudementary stats maintained by t_cache.
 *
 *              This is a debugging function, which will not normally
 *              be run as part of t_cache.
 *
 * Return:      void
 *
 * Programmer:  JRM -- 4/17/06
 *
 * Modifications:
 *
 *              None.
 *
 *****************************************************************************/

static void
print_stats(void)
{
    HDfprintf(stdout,
              "%d: datum clears / pinned clears / destroys = %ld / %ld / %ld\n",
              world_mpi_rank, datum_clears, datum_pinned_clears,
              datum_destroys );
    HDfprintf(stdout,
              "%d: datum flushes / pinned flushes / loads  = %ld / %ld / %ld\n",
              world_mpi_rank, datum_flushes, datum_pinned_flushes,
              datum_loads );
    HDfprintf(stdout,
              "%d: pins: global / global dirty / local = %ld / %ld / %ld\n",
              world_mpi_rank, global_pins, global_dirty_pins, local_pins);
    HDfflush(stdout);

    return;

} /* print_stats() */
#endif /* NOT_USED */


/*****************************************************************************
 *
 * Function:    reset_stats()
 *
 * Purpose:    Reset the rudementary stats maintained by t_cache.
 *
 * Return:    void
 *
 * Programmer:    JRM -- 4/17/06
 *
 * Modifications:
 *
 *        None.
 *
 *****************************************************************************/

static void
reset_stats(void)
{
    datum_clears          = 0;
    datum_pinned_clears   = 0;
    datum_destroys        = 0;
    datum_flushes         = 0;
    datum_pinned_flushes  = 0;
    datum_loads           = 0;
    global_pins          = 0;
    global_dirty_pins      = 0;
    local_pins          = 0;

    return;

} /* reset_stats() */


/*****************************************************************************/
/**************************** MPI setup functions ****************************/
/*****************************************************************************/

/*****************************************************************************
 *
 * Function:    set_up_file_communicator()
 *
 * Purpose:    Create the MPI communicator used to open a HDF5 file with.
 *        In passing, also initialize the file_mpi... globals.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 11/16/05
 *
 * Modifications:
 *
 *        None.
 *
 *****************************************************************************/

static hbool_t
set_up_file_communicator(void)
{
    hbool_t success = TRUE;
    int mpi_result;
    int num_excluded_ranks;
    int excluded_ranks[1];
    MPI_Group file_group;
    MPI_Group world_group;

    if ( success ) {

        mpi_result = MPI_Comm_group(world_mpi_comm, &world_group);

        if ( mpi_result != MPI_SUCCESS ) {

            nerrors++;
            success = FALSE;
        if ( verbose ) {
                HDfprintf(stdout,
                        "%d:%s: MPI_Comm_group() failed with error %d.\n",
                        world_mpi_rank, FUNC, mpi_result);
            }
        }
    }

    if ( success ) {

        num_excluded_ranks = 1;
        excluded_ranks[0] = world_server_mpi_rank;
        mpi_result = MPI_Group_excl(world_group, num_excluded_ranks,
                                    excluded_ranks, &file_group);

        if ( mpi_result != MPI_SUCCESS ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout,
                        "%d:%s: MPI_Group_excl() failed with error %d.\n",
                        world_mpi_rank, FUNC, mpi_result);
            }
        }
    }

    if ( success ) {

        mpi_result = MPI_Comm_create(world_mpi_comm, file_group,
                                     &file_mpi_comm);

        if ( mpi_result != MPI_SUCCESS ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout,
                        "%d:%s: MPI_Comm_create() failed with error %d.\n",
                        world_mpi_rank, FUNC, mpi_result);
            }

        } else {

            if ( world_mpi_rank != world_server_mpi_rank ) {

                if ( file_mpi_comm == MPI_COMM_NULL ) {

                    nerrors++;
                    success = FALSE;
                    if ( verbose ) {
                        HDfprintf(stdout,
                                "%d:%s: file_mpi_comm == MPI_COMM_NULL.\n",
                                world_mpi_rank, FUNC);
                    }
                }
            } else {

                file_mpi_size = world_mpi_size - 1; /* needed by the server */

                if ( file_mpi_comm != MPI_COMM_NULL ) {

                    nerrors++;
                    success = FALSE;
                    if ( verbose ) {
                        HDfprintf(stdout,
                                "%d:%s: file_mpi_comm != MPI_COMM_NULL.\n",
                                world_mpi_rank, FUNC);
                    }
                }
            }
        }
    }

    if ( ( success ) && ( world_mpi_rank != world_server_mpi_rank ) ) {

        mpi_result = MPI_Comm_size(file_mpi_comm, &file_mpi_size);

        if ( mpi_result != MPI_SUCCESS ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout,
                        "%d:%s: MPI_Comm_size() failed with error %d.\n",
                        world_mpi_rank, FUNC, mpi_result);
            }
        }
    }

    if ( ( success ) && ( world_mpi_rank != world_server_mpi_rank ) ) {

        mpi_result = MPI_Comm_rank(file_mpi_comm, &file_mpi_rank);

        if ( mpi_result != MPI_SUCCESS ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout,
                        "%d:%s: MPI_Comm_rank() failed with error %d.\n",
                        world_mpi_rank, FUNC, mpi_result);
            }
        }
    }

    return(success);

} /* set_up_file_communicator() */


/*****************************************************************************/
/******************** data array manipulation functions **********************/
/*****************************************************************************/

/*****************************************************************************
 *
 * Function:    addr_to_datum_index()
 *
 * Purpose:    Given the base address of a datum, find and return its index
 *        in the data array.
 *
 * Return:    Success:    index of target datum.
 *
 *        Failure:    -1.
 *
 * Programmer:    JRM -- 12/20/05
 *
 *****************************************************************************/
static int
addr_to_datum_index(haddr_t base_addr)
{
    int top = NUM_DATA_ENTRIES - 1;
    int bottom = 0;
    int middle = (NUM_DATA_ENTRIES - 1) / 2;
    int ret_value = -1;

    while ( top >= bottom )
    {
        if ( base_addr < data[data_index[middle]].base_addr ) {

            top = middle - 1;
            middle = (top + bottom) / 2;

        } else if ( base_addr > data[data_index[middle]].base_addr ) {

            bottom = middle + 1;
            middle = (top + bottom) / 2;

        } else /* ( base_addr == data[data_index[middle]].base_addr ) */ {

            ret_value = data_index[middle];
            bottom = top + 1; /* to force exit from while loop */

        }
    }

    return(ret_value);

} /* addr_to_datum_index() */


/*****************************************************************************
 *
 * Function:    init_data()
 *
 * Purpose:    Initialize the data array, from which cache entries are
 *        loaded.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 12/20/05
 *
 *****************************************************************************/
static void
init_data(void)
{
    /* The set of address offsets is chosen so as to avoid allowing the
     * base addresses to fall in a pattern of that will annoy the hash
     * table, and to give a good range of entry sizes.
     *
     * At present, I am using the first 20 entries of the Fibonacci
     * sequence multiplied by 2.  We will see how it works.
     */
    const int num_addr_offsets = 20;
    const haddr_t addr_offsets[20] = {   2,    2,    4,    6,    10,
                                        16,   26,   42,   68,   110,
                                       178,  288,  466,  754,  1220,
                                      1974, 3194, 5168, 8362, 13539};
    int i;
    int j = 0;
    haddr_t addr = BASE_ADDR;

    /* this must hold so moves don't change entry size. */
    HDassert( (NUM_DATA_ENTRIES / 2) % 20 == 0 );
    HDassert( (virt_num_data_entries / 2) % 20 == 0 );

    for ( i = 0; i < NUM_DATA_ENTRIES; i++ )
    {
        data[i].base_addr     = addr;
        data[i].len           = (size_t)(addr_offsets[j]);
        data[i].local_len     = (size_t)(addr_offsets[j]);
        data[i].ver           = 0;
        data[i].dirty         = FALSE;
        data[i].valid         = FALSE;
        data[i].locked        = FALSE;
    data[i].global_pinned = FALSE;
    data[i].local_pinned  = FALSE;
    data[i].cleared       = FALSE;
        data[i].flushed       = FALSE;
        data[i].reads         = 0;
        data[i].writes        = 0;
    data[i].index         = i;
    data[i].aux_ptr       = NULL;

        data_index[i]         = i;

        addr += addr_offsets[j];
        HDassert( addr > data[i].base_addr );

        j = (j + 1) % num_addr_offsets;
    }

    /* save the end of the address space used by the data array */
    max_addr = addr;

    return;

} /* init_data() */


/*****************************************************************************/
/******************** test coodination related functions *********************/
/*****************************************************************************/

/*****************************************************************************
 *
 * Function:    do_express_test()
 *
 * Purpose:    Do an MPI_Allreduce to obtain the maximum value returned
 *         by GetTestExpress() across all processes.  Return this
 *         value.
 *
 *         Envirmoment variables can be different across different
 *         processes.  This function ensures that all processes agree
 *         on whether to do an express test.
 *
 * Return:    Success:    Maximum of the values returned by
 *                 GetTestExpress() across    all processes.
 *
 *        Failure:    -1
 *
 * Programmer:    JRM -- 4/25/06
 *
 *****************************************************************************/
static int
do_express_test(void)
{
    int express_test;
    int max_express_test;
    int result;

    express_test = GetTestExpress();

    result = MPI_Allreduce((void *)&express_test,
                           (void *)&max_express_test,
                           1,
                           MPI_INT,
                           MPI_MAX,
                           world_mpi_comm);

    if ( result != MPI_SUCCESS ) {

        nerrors++;
        max_express_test = -1;
        if ( verbose ) {
            HDfprintf(stdout, "%d:%s: MPI_Allreduce() failed.\n",
                      world_mpi_rank, FUNC );
        }
    }

    return(max_express_test);

} /* do_express_test() */


/*****************************************************************************
 *
 * Function:    do_sync()
 *
 * Purpose:    Ensure that all messages sent by this process have been
 *         processed before proceeding.
 *
 *         Do this by exchanging sync req / sync ack messages with
 *         the server.
 *
 *         Do nothing if nerrors is greater than zero.
 *
 * Return:    void
 *
 * Programmer:    JRM -- 5/10/06
 *
 *****************************************************************************/
static void
do_sync(void)
{

    struct mssg_t mssg;

    if ( nerrors <= 0 ) {

        /* compose the message */
    mssg.req       = SYNC_REQ_CODE;
        mssg.src       = world_mpi_rank;
        mssg.dest      = world_server_mpi_rank;
        mssg.mssg_num  = -1; /* set by send function */
        mssg.base_addr = 0;
        mssg.len       = 0;
        mssg.ver       = 0;
        mssg.count     = 0;
        mssg.magic     = MSSG_MAGIC;

    if ( ! send_mssg(&mssg, FALSE) ) {

        nerrors++;
        if ( verbose ) {
                HDfprintf(stdout, "%d:%s: send_mssg() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }

    if ( nerrors <= 0 ) {

    if ( ! recv_mssg(&mssg, SYNC_ACK_CODE) ) {

            nerrors++;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: recv_mssg() failed.\n",
                          world_mpi_rank, FUNC);
            }
    } else if ( ( mssg.req != SYNC_ACK_CODE ) ||
                    ( mssg.src != world_server_mpi_rank ) ||
                    ( mssg.dest != world_mpi_rank ) ||
            ( mssg.magic != MSSG_MAGIC ) ) {

            nerrors++;
        if ( verbose ) {
                HDfprintf(stdout, "%d:%s: Bad data in sync ack.\n",
                          world_mpi_rank, FUNC);
            }
    }
    }

    return;

} /* do_sync() */


/*****************************************************************************
 *
 * Function:    get_max_nerrors()
 *
 * Purpose:    Do an MPI_Allreduce to obtain the maximum value of nerrors
 *        across all processes.  Return this value.
 *
 * Return:    Success:    Maximum of the nerrors global variables across
 *                all processes.
 *
 *        Failure:    -1
 *
 * Programmer:    JRM -- 1/3/06
 *
 *****************************************************************************/
static int
get_max_nerrors(void)
{
    int max_nerrors;
    int result;

    result = MPI_Allreduce((void *)&nerrors,
                           (void *)&max_nerrors,
                           1,
                           MPI_INT,
                           MPI_MAX,
                           world_mpi_comm);

    if ( result != MPI_SUCCESS ) {

        nerrors++;
        max_nerrors = -1;
        if ( verbose ) {
            HDfprintf(stdout, "%d:%s: MPI_Allreduce() failed.\n",
                      world_mpi_rank, FUNC );
        }
    }

    return(max_nerrors);

} /* get_max_nerrors() */


/*****************************************************************************/
/************************ mssg xfer related functions ************************/
/*****************************************************************************/

/*****************************************************************************
 *
 * Function:    recv_mssg()
 *
 * Purpose:    Receive a message from any process in the provided instance
 *        of struct mssg.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 12/22/05
 *
 * Modifications:
 *
 *         JRM -- 5/10/06
 *        Added mssg_tag_offset parameter and supporting code.
 *
 *****************************************************************************/

#define CACHE_TEST_TAG    99 /* different from any used by the library */

static hbool_t
recv_mssg(struct mssg_t *mssg_ptr,
    int mssg_tag_offset)
{
    hbool_t success = TRUE;
    int mssg_tag = CACHE_TEST_TAG;
    int result;
    MPI_Status status;

    if ( ( mssg_ptr == NULL ) ||
         ( mssg_tag_offset < 0 ) ||
         ( mssg_tag_offset> MAX_REQ_CODE ) ) {

        nerrors++;
        success = FALSE;
        if ( verbose ) {
            HDfprintf(stdout, "%d:%s: bad param(s) on entry.\n",
                      world_mpi_rank, FUNC);
        }
    } else {

        mssg_tag += mssg_tag_offset;
    }

    if ( success ) {

        result = MPI_Recv((void *)mssg_ptr, 1, mpi_mssg_t, MPI_ANY_SOURCE,
                          mssg_tag, world_mpi_comm, &status);

        if ( result != MPI_SUCCESS ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: MPI_Recv() failed.\n",
                          world_mpi_rank, FUNC );
            }
        } else if ( mssg_ptr->magic != MSSG_MAGIC ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: invalid magic.\n", world_mpi_rank,
                          FUNC);
            }
        } else if ( mssg_ptr->src != status.MPI_SOURCE ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout,
                          "%d:%s: mssg_ptr->src != status.MPI_SOURCE.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }

    return(success);

} /* recv_mssg() */


/*****************************************************************************
 *
 * Function:    send_mssg()
 *
 * Purpose:    Send the provided instance of mssg to the indicated target.
 *
 *        Note that all source and destination ranks are in the
 *        global communicator.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 12/22/05
 *
 * Modifications:
 *
 *         JRM -- 5/10/06
 *        Added the add_req_to_tag parameter and supporting code.
 *
 *****************************************************************************/
static hbool_t
send_mssg(struct mssg_t *mssg_ptr,
    hbool_t add_req_to_tag)
{
    hbool_t success = TRUE;
    int mssg_tag = CACHE_TEST_TAG;
    int result;
    static long mssg_num = 0;

    if ( ( mssg_ptr == NULL ) ||
         ( mssg_ptr->src != world_mpi_rank ) ||
         ( mssg_ptr->dest < 0 ) ||
         ( mssg_ptr->dest == mssg_ptr->src ) ||
         ( mssg_ptr->dest >= world_mpi_size ) ||
         ( mssg_ptr->req < 0 ) ||
         ( mssg_ptr->req > MAX_REQ_CODE ) ||
         ( mssg_ptr->magic != MSSG_MAGIC ) ) {

        nerrors++;
        success = FALSE;
        if ( verbose ) {
            HDfprintf(stdout, "%d:%s: Invalid mssg on entry.\n",
                      world_mpi_rank, FUNC);
        }
    }

    if ( success ) {

        mssg_ptr->mssg_num = mssg_num++;

    if ( add_req_to_tag ) {

        mssg_tag += mssg_ptr->req;
    }

        result = MPI_Send((void *)mssg_ptr, 1, mpi_mssg_t,
                          mssg_ptr->dest, mssg_tag, world_mpi_comm);

        if ( result != MPI_SUCCESS ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: MPI_Send() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }

    return(success);

} /* send_mssg() */


/*****************************************************************************
 *
 * Function:    setup_derived_types()
 *
 * Purpose:    Set up the derived types used by the test bed.  At present,
 *        only the mpi_mssg derived type is needed.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 12/22/05
 *
 *****************************************************************************/
static hbool_t
setup_derived_types(void)
{
    hbool_t success = TRUE;
    int i;
    int result;
    MPI_Datatype mpi_types[9] = {MPI_INT, MPI_INT, MPI_INT, MPI_LONG,
                                 HADDR_AS_MPI_TYPE, MPI_INT, MPI_INT,
                                 MPI_UNSIGNED, MPI_UNSIGNED};
    int block_len[9] = {1, 1, 1, 1, 1, 1, 1, 1, 1};
    MPI_Aint displs[9];
    struct mssg_t sample; /* used to compute displacements */

    /* setup the displacements array */
    if ( ( MPI_SUCCESS != MPI_Get_address(&sample.req, &displs[0]) ) ||
         ( MPI_SUCCESS != MPI_Get_address(&sample.src, &displs[1]) ) ||
         ( MPI_SUCCESS != MPI_Get_address(&sample.dest, &displs[2]) ) ||
         ( MPI_SUCCESS != MPI_Get_address(&sample.mssg_num, &displs[3]) ) ||
         ( MPI_SUCCESS != MPI_Get_address(&sample.base_addr, &displs[4]) ) ||
         ( MPI_SUCCESS != MPI_Get_address(&sample.len, &displs[5]) ) ||
         ( MPI_SUCCESS != MPI_Get_address(&sample.ver, &displs[6]) ) ||
         ( MPI_SUCCESS != MPI_Get_address(&sample.count, &displs[7]) ) ||
         ( MPI_SUCCESS != MPI_Get_address(&sample.magic, &displs[8]) ) ) {

        nerrors++;
        success = FALSE;
        if ( verbose ) {
            HDfprintf(stdout, "%d:%s: MPI_Get_address() call failed.\n",
                      world_mpi_rank, FUNC);
        }

    } else {

        /* Now calculate the actual displacements */
        for ( i = 8; i >= 0; --i)
        {
            displs[i] -= displs[0];
        }
    }

    if ( success ) {

        result = MPI_Type_create_struct(9, block_len, displs, mpi_types, &mpi_mssg_t);

        if ( result != MPI_SUCCESS ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: MPI_Type_create_struct() call failed.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }

    if ( success ) {

        result = MPI_Type_commit(&mpi_mssg_t);

        if ( result != MPI_SUCCESS) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: MPI_Type_commit() call failed.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }

    return(success);

} /* setup_derived_types */


/*****************************************************************************
 *
 * Function:    takedown_derived_types()
 *
 * Purpose:    take down the derived types used by the test bed.  At present,
 *        only the mpi_mssg derived type is needed.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 12/22/05
 *
 *****************************************************************************/
static hbool_t
takedown_derived_types(void)
{
    hbool_t success = TRUE;
    int result;

    result = MPI_Type_free(&mpi_mssg_t);

    if ( result != MPI_SUCCESS ) {

        nerrors++;
        success = FALSE;
        if ( verbose ) {
            HDfprintf(stdout, "%d:%s: MPI_Type_free() call failed.\n",
            world_mpi_rank, FUNC);
        }
    }

    return(success);

} /* takedown_derived_types() */


/*****************************************************************************/
/***************************** server functions ******************************/
/*****************************************************************************/

/*****************************************************************************
 *
 * Function:    reset_server_counters()
 *
 * Purpose:    Reset the counters maintained by the server, doing a
 *        sanity check in passing.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 5/5/10
 *
 *****************************************************************************/
static hbool_t
reset_server_counters(void)
{
    hbool_t success = TRUE;
    int i;
    long actual_total_reads = 0;
    long actual_total_writes = 0;

    for ( i = 0; i < NUM_DATA_ENTRIES; i++ )
    {
        if ( data[i].reads > 0 ) {

            actual_total_reads += data[i].reads;
            data[i].reads = 0;
        }

        if ( data[i].writes > 0 ) {

            actual_total_writes += data[i].writes;
            data[i].writes = 0;
        }
    }

    if ( actual_total_reads != total_reads ) {

        success = FALSE;
        nerrors++;
        if ( verbose ) {
            HDfprintf(stdout, "%d:%s: actual/total reads mismatch (%ld/%ld).\n",
                      world_mpi_rank, FUNC,
                      actual_total_reads, total_reads);
        }
    }

    if ( actual_total_writes != total_writes ) {

        success = FALSE;
        nerrors++;
        if ( verbose ) {
            HDfprintf(stdout, "%d:%s: actual/total writes mismatch (%ld/%ld).\n",
                      world_mpi_rank, FUNC,
                      actual_total_writes, total_writes);
        }
    }

    total_reads = 0;
    total_writes = 0;

    return(success);

} /* reset_server_counters() */


/*****************************************************************************
 *
 * Function:    server_main()
 *
 * Purpose:    Main function for the server process.  This process exists
 *        to provide an independant view of the data array.
 *
 *        The function handles request from the other processes in
 *        the test until the count of done messages received equals
 *        the number of client processes.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 12/22/05
 *
 * Modifications:
 *
 *         JRM -- 5/10/06
 *        Updated for sync message.
 *
 *****************************************************************************/
static hbool_t
server_main(void)
{
    hbool_t done = FALSE;
    hbool_t success = TRUE;
    int done_count = 0;
    struct mssg_t mssg;

    if ( world_mpi_rank != world_server_mpi_rank ) {

        nerrors++;
        success = FALSE;
        if ( verbose ) {
            HDfprintf(stdout, "%d:%s: This isn't the server process?!?!?\n",
                      world_mpi_rank, FUNC);
        }
    }


    while ( ( success ) && ( ! done ) )
    {
        success = recv_mssg(&mssg, 0);

        if ( success ) {

            switch ( mssg.req )
            {
        case WRITE_REQ_CODE:
            success = serve_write_request(&mssg);
            break;

        case WRITE_REQ_ACK_CODE:
                    success = FALSE;
                    if(verbose)
                        HDfprintf(stdout, "%s: Received write ack?!?.\n", FUNC);
            break;

        case READ_REQ_CODE:
                    success = serve_read_request(&mssg);
            break;

        case READ_REQ_REPLY_CODE:
                    success = FALSE;
                    if(verbose)
                        HDfprintf(stdout, "%s: Received read req reply?!?.\n", FUNC);
            break;

        case SYNC_REQ_CODE:
                    success = serve_sync_request(&mssg);
            break;

        case SYNC_ACK_CODE:
                    success = FALSE;
                    if(verbose)
                        HDfprintf(stdout, "%s: Received sync ack?!?.\n", FUNC);
            break;

        case REQ_TTL_WRITES_CODE:
            success = serve_total_writes_request(&mssg);
            break;

        case REQ_TTL_WRITES_RPLY_CODE:
                    success = FALSE;
                    if(verbose)
                        HDfprintf(stdout, "%s: Received total writes reply?!?.\n", FUNC);
            break;

        case REQ_TTL_READS_CODE:
            success = serve_total_reads_request(&mssg);
            break;

        case REQ_TTL_READS_RPLY_CODE:
                    success = FALSE;
                    if(verbose)
                        HDfprintf(stdout, "%s: Received total reads reply?!?.\n", FUNC);
            break;

        case REQ_ENTRY_WRITES_CODE:
            success = serve_entry_writes_request(&mssg);
            break;

        case REQ_ENTRY_WRITES_RPLY_CODE:
                    success = FALSE;
                    if(verbose)
                        HDfprintf(stdout, "%s: Received entry writes reply?!?.\n", FUNC);
            break;

        case REQ_ENTRY_READS_CODE:
            success = serve_entry_reads_request(&mssg);
            break;

        case REQ_ENTRY_READS_RPLY_CODE:
                    success = FALSE;
                    if(verbose)
                        HDfprintf(stdout, "%s: Received entry reads reply?!?.\n", FUNC);
            break;

        case REQ_RW_COUNT_RESET_CODE:
            success = serve_rw_count_reset_request(&mssg);
            break;

        case REQ_RW_COUNT_RESET_RPLY_CODE:
                    success = FALSE;
                    if(verbose)
                        HDfprintf(stdout, "%s: Received RW count reset reply?!?.\n", FUNC);
            break;

        case DONE_REQ_CODE:
            done_count++;
            if(done_count >= file_mpi_size)
            done = TRUE;
            break;

        default:
                    nerrors++;
                    success = FALSE;
                    if(verbose)
                HDfprintf(stdout, "%d:%s: Unknown request code.\n", world_mpi_rank, FUNC);
            break;
            }
        }
    }

    return(success);

} /* server_main() */


/*****************************************************************************
 *
 * Function:    serve_read_request()
 *
 * Purpose:    Serve a read request.
 *
 *        The function accepts a pointer to an instance of struct
 *        mssg_t as input.  If all sanity checks pass, it sends
 *        a copy of the indicated datum from the data array to
 *        the requesting process.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 12/22/05
 *
 *****************************************************************************/
static hbool_t
serve_read_request(struct mssg_t * mssg_ptr)
{
    hbool_t report_mssg = FALSE;
    hbool_t success = TRUE;
    int target_index;
    haddr_t target_addr;
    struct mssg_t reply;

    if ( ( mssg_ptr == NULL ) ||
         ( mssg_ptr->req != READ_REQ_CODE ) ||
         ( mssg_ptr->magic != MSSG_MAGIC ) ) {

        nerrors++;
        success = FALSE;
        if ( verbose ) {
            HDfprintf(stdout, "%d:%s: Bad mssg on entry.\n",
                      world_mpi_rank, FUNC);
        }
    }

    if ( success ) {

        target_addr = mssg_ptr->base_addr;
        target_index = addr_to_datum_index(target_addr);

        if ( target_index < 0 ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: addr lookup failed for %a.\n",
                          world_mpi_rank, FUNC, target_addr);
            }
        } else if ( data[target_index].len != mssg_ptr->len ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout,
                          "%d:%s: data[i].len = %Zu != mssg->len = %d.\n",
                           world_mpi_rank, FUNC,
                           data[target_index].len, mssg_ptr->len);
            }
        } else if ( ! (data[target_index].valid) ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout,
                  "%d:%s: proc %d read invalid entry. idx/base_addr = %d/%a.\n",
                         world_mpi_rank, FUNC,
                         mssg_ptr->src,
            target_index,
            data[target_index].base_addr);
            }
        } else {

            /* compose the reply message */
            reply.req       = READ_REQ_REPLY_CODE;
            reply.src       = world_mpi_rank;
            reply.dest      = mssg_ptr->src;
            reply.mssg_num  = -1; /* set by send function */
            reply.base_addr = data[target_index].base_addr;
            H5_CHECKED_ASSIGN(reply.len, unsigned, data[target_index].len, size_t);
            reply.ver       = data[target_index].ver;
            reply.count     = 0;
            reply.magic     = MSSG_MAGIC;

        /* and update the counters */
        total_reads++;
            (data[target_index].reads)++;
        }
    }

    if ( success ) {

        success = send_mssg(&reply, TRUE);
    }

    if ( report_mssg ) {

        if ( success ) {

            HDfprintf(stdout, "%d read 0x%llx. len = %d. ver = %d.\n",
                      (int)(mssg_ptr->src),
                      (long long)(data[target_index].base_addr),
                      (int)(data[target_index].len),
                      (int)(data[target_index].ver));

        } else {

            HDfprintf(stdout, "%d read 0x%llx FAILED. len = %d. ver = %d.\n",
                      (int)(mssg_ptr->src),
                      (long long)(data[target_index].base_addr),
                      (int)(data[target_index].len),
                      (int)(data[target_index].ver));

        }
    }

    return(success);

} /* serve_read_request() */


/*****************************************************************************
 *
 * Function:    serve_sync_request()
 *
 * Purpose:    Serve a sync request.
 *
 *        The function accepts a pointer to an instance of struct
 *        mssg_t as input.  If all sanity checks pass, it sends a
 *        sync ack to the requesting process.
 *
 *        This service exist to allow the sending process to ensure
 *        that all previous messages have been processed before
 *        proceeding.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 5/10/06
 *
 *****************************************************************************/
static hbool_t
serve_sync_request(struct mssg_t * mssg_ptr)
{
    hbool_t report_mssg = FALSE;
    hbool_t success = TRUE;
    struct mssg_t reply;

    if ( ( mssg_ptr == NULL ) ||
         ( mssg_ptr->req != SYNC_REQ_CODE ) ||
         ( mssg_ptr->magic != MSSG_MAGIC ) ) {

        nerrors++;
        success = FALSE;
        if ( verbose ) {
            HDfprintf(stdout, "%d:%s: Bad mssg on entry.\n",
                      world_mpi_rank, FUNC);
        }
    }

    if ( success ) {

        /* compose the reply message */
        reply.req       = SYNC_ACK_CODE;
        reply.src       = world_mpi_rank;
        reply.dest      = mssg_ptr->src;
        reply.mssg_num  = -1; /* set by send function */
        reply.base_addr = 0;
        reply.len       = 0;
        reply.ver       = 0;
    reply.count     = 0;
        reply.magic     = MSSG_MAGIC;
    }

    if ( success ) {

        success = send_mssg(&reply, TRUE);
    }

    if ( report_mssg ) {

        if ( success ) {

            HDfprintf(stdout, "%d sync.\n", (int)(mssg_ptr->src));

        } else {

            HDfprintf(stdout, "%d sync FAILED.\n", (int)(mssg_ptr->src));

        }
    }

    return(success);

} /* serve_sync_request() */


/*****************************************************************************
 *
 * Function:    serve_write_request()
 *
 * Purpose:    Serve a write request.
 *
 *        The function accepts a pointer to an instance of struct
 *        mssg_t as input.  If all sanity checks pass, it updates
 *        the version number of the target data array entry as
 *        specified in the message.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 12/21/05
 *
 *****************************************************************************/
static hbool_t
serve_write_request(struct mssg_t * mssg_ptr)
{
    hbool_t report_mssg = FALSE;
    hbool_t success = TRUE;
    int target_index;
    int new_ver_num = 0;
    haddr_t target_addr;
#if DO_WRITE_REQ_ACK
    struct mssg_t reply;
#endif /* DO_WRITE_REQ_ACK */

    if ( ( mssg_ptr == NULL ) ||
         ( mssg_ptr->req != WRITE_REQ_CODE ) ||
         ( mssg_ptr->magic != MSSG_MAGIC ) ) {

        nerrors++;
        success = FALSE;
        if ( verbose ) {
            HDfprintf(stdout, "%d:%s: Bad mssg on entry.\n",
                      world_mpi_rank, FUNC);
        }
    }

    if ( success ) {

        target_addr = mssg_ptr->base_addr;
        target_index = addr_to_datum_index(target_addr);

        if ( target_index < 0 ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: addr lookup failed for %a.\n",
                          world_mpi_rank, FUNC, target_addr);
            }
        } else if ( data[target_index].len != mssg_ptr->len ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout,
                          "%d:%s: data[i].len = %Zu != mssg->len = %d.\n",
                          world_mpi_rank, FUNC,
                          data[target_index].len, mssg_ptr->len);
            }
        }
    }

    if ( success ) {

        new_ver_num = mssg_ptr->ver;

        /* this check should catch duplicate writes */
    if ( new_ver_num <= data[target_index].ver ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: new ver = %d <= old ver = %d.\n",
                          world_mpi_rank, FUNC,
                          new_ver_num, data[target_index].ver);
            }
        }
    }

    if ( success ) {

    /* process the write */
        data[target_index].ver = new_ver_num;
        data[target_index].valid = TRUE;

        /* and update the counters */
    total_writes++;
        (data[target_index].writes)++;

#if DO_WRITE_REQ_ACK

        /* compose the reply message */
        reply.req       = WRITE_REQ_ACK_CODE;
        reply.src       = world_mpi_rank;
        reply.dest      = mssg_ptr->src;
        reply.mssg_num  = -1; /* set by send function */
        reply.base_addr = data[target_index].base_addr;
        H5_CHECKED_ASSIGN(reply.len, unsigned, data[target_index].len, size_t);
        reply.ver       = data[target_index].ver;
        reply.count     = 0;
        reply.magic     = MSSG_MAGIC;

    /* and send it */
        success = send_mssg(&reply, TRUE);

#endif /* DO_WRITE_REQ_ACK */

    }

    if ( report_mssg ) {

        if ( success ) {

            HDfprintf(stdout, "%d write 0x%llx. len = %d. ver = %d.\n",
                      (int)(mssg_ptr->src),
                      (long long)(data[target_index].base_addr),
                      (int)(data[target_index].len),
                      (int)(data[target_index].ver));

        } else {

            HDfprintf(stdout, "%d write 0x%llx FAILED. len = %d. ver = %d.\n",
                      (int)(mssg_ptr->src),
                      (long long)(data[target_index].base_addr),
                      (int)(data[target_index].len),
                      (int)(data[target_index].ver));

        }
    }

    return(success);

} /* serve_write_request() */


/*****************************************************************************
 *
 * Function:    serve_total_writes_request()
 *
 * Purpose:    Serve a request for the total number of writes recorded since
 *        the last reset.
 *
 *        The function accepts a pointer to an instance of struct
 *        mssg_t as input.  If all sanity checks pass, it sends
 *        the current value of the total_writes global variable to
 *        the requesting process.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 5/5/10
 *
 *****************************************************************************/
static hbool_t
serve_total_writes_request(struct mssg_t * mssg_ptr)
{
    hbool_t report_mssg = FALSE;
    hbool_t success = TRUE;
    struct mssg_t reply;

    if ( ( mssg_ptr == NULL ) ||
         ( mssg_ptr->req != REQ_TTL_WRITES_CODE ) ||
         ( mssg_ptr->magic != MSSG_MAGIC ) ) {

        nerrors++;
        success = FALSE;
        if ( verbose ) {
            HDfprintf(stdout, "%d:%s: Bad mssg on entry.\n",
                      world_mpi_rank, FUNC);
        }
    }

    if ( success ) {

        /* compose the reply message */
        reply.req       = REQ_TTL_WRITES_RPLY_CODE;
        reply.src       = world_mpi_rank;
        reply.dest      = mssg_ptr->src;
        reply.mssg_num  = -1; /* set by send function */
        reply.base_addr = 0;
        reply.len       = 0;
        reply.ver       = 0;
        reply.count     = (unsigned)total_writes;
        reply.magic     = MSSG_MAGIC;
    }

    if ( success ) {

        success = send_mssg(&reply, TRUE);
    }

    if ( report_mssg ) {

        if ( success ) {

            HDfprintf(stdout, "%d request total writes %ld.\n",
                      (int)(mssg_ptr->src),
                      total_writes);

        } else {

            HDfprintf(stdout, "%d request total writes %ld -- FAILED.\n",
                      (int)(mssg_ptr->src),
                      total_writes);

        }
    }

    return(success);

} /* serve_total_writes_request() */


/*****************************************************************************
 *
 * Function:    serve_total_reads_request()
 *
 * Purpose:    Serve a request for the total number of reads recorded since
 *        the last reset.
 *
 *        The function accepts a pointer to an instance of struct
 *        mssg_t as input.  If all sanity checks pass, it sends
 *        the current value of the total_reads global variable to
 *        the requesting process.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 5/5/10
 *
 *****************************************************************************/
static hbool_t
serve_total_reads_request(struct mssg_t * mssg_ptr)
{
    hbool_t report_mssg = FALSE;
    hbool_t success = TRUE;
    struct mssg_t reply;

    if ( ( mssg_ptr == NULL ) ||
         ( mssg_ptr->req != REQ_TTL_READS_CODE ) ||
         ( mssg_ptr->magic != MSSG_MAGIC ) ) {

        nerrors++;
        success = FALSE;
        if ( verbose ) {
            HDfprintf(stdout, "%d:%s: Bad mssg on entry.\n",
                      world_mpi_rank, FUNC);
        }
    }

    if ( success ) {

        /* compose the reply message */
        reply.req       = REQ_TTL_READS_RPLY_CODE;
        reply.src       = world_mpi_rank;
        reply.dest      = mssg_ptr->src;
        reply.mssg_num  = -1; /* set by send function */
        reply.base_addr = 0;
        reply.len       = 0;
        reply.ver       = 0;
        reply.count     = (unsigned)total_reads;
        reply.magic     = MSSG_MAGIC;
    }

    if ( success ) {

        success = send_mssg(&reply, TRUE);
    }

    if ( report_mssg ) {

        if ( success ) {

            HDfprintf(stdout, "%d request total reads %ld.\n",
                      (int)(mssg_ptr->src),
                      total_reads);

        } else {

            HDfprintf(stdout, "%d request total reads %ld -- FAILED.\n",
                      (int)(mssg_ptr->src),
                      total_reads);

        }
    }

    return(success);

} /* serve_total_reads_request() */


/*****************************************************************************
 *
 * Function:    serve_entry_writes_request()
 *
 * Purpose:    Serve an entry writes request.
 *
 *        The function accepts a pointer to an instance of struct
 *        mssg_t as input.  If all sanity checks pass, it sends
 *        the number of times that the indicated datum has been
 *        written since the last counter reset to the requesting
 *        process.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 5/5/10
 *
 *****************************************************************************/
static hbool_t
serve_entry_writes_request(struct mssg_t * mssg_ptr)
{
    hbool_t report_mssg = FALSE;
    hbool_t success = TRUE;
    int target_index;
    haddr_t target_addr;
    struct mssg_t reply;

    if ( ( mssg_ptr == NULL ) ||
         ( mssg_ptr->req != REQ_ENTRY_WRITES_CODE ) ||
         ( mssg_ptr->magic != MSSG_MAGIC ) ) {

        nerrors++;
        success = FALSE;
        if ( verbose ) {
            HDfprintf(stdout, "%d:%s: Bad mssg on entry.\n",
                      world_mpi_rank, FUNC);
        }
    }

    if ( success ) {

        target_addr = mssg_ptr->base_addr;
        target_index = addr_to_datum_index(target_addr);

        if ( target_index < 0 ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: addr lookup failed for %a.\n",
                          world_mpi_rank, FUNC, target_addr);
            }
        } else {

            /* compose the reply message */
            reply.req       = REQ_ENTRY_WRITES_RPLY_CODE;
            reply.src       = world_mpi_rank;
            reply.dest      = mssg_ptr->src;
            reply.mssg_num  = -1; /* set by send function */
            reply.base_addr = target_addr;
            reply.len       = 0;
            reply.ver       = 0;
            reply.count     = (unsigned)data[target_index].writes;
            reply.magic     = MSSG_MAGIC;
        }
    }

    if ( success ) {

        success = send_mssg(&reply, TRUE);
    }

    if ( report_mssg ) {

        if ( success ) {

            HDfprintf(stdout, "%d request entry 0x%llx writes = %ld.\n",
                      (int)(mssg_ptr->src),
                      (long long)(data[target_index].base_addr),
                      (long)(data[target_index].writes));

        } else {

            HDfprintf(stdout, "%d request entry 0x%llx writes = %ld FAILED.\n",
                      (int)(mssg_ptr->src),
                      (long long)(data[target_index].base_addr),
                      (long)(data[target_index].writes));

        }
    }

    return(success);

} /* serve_entry_writes_request() */


/*****************************************************************************
 *
 * Function:    serve_entry_reads_request()
 *
 * Purpose:    Serve an entry reads request.
 *
 *        The function accepts a pointer to an instance of struct
 *        mssg_t as input.  If all sanity checks pass, it sends
 *        the number of times that the indicated datum has been
 *        read since the last counter reset to the requesting
 *        process.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 5/5/10
 *
 *****************************************************************************/
static hbool_t
serve_entry_reads_request(struct mssg_t * mssg_ptr)
{
    hbool_t report_mssg = FALSE;
    hbool_t success = TRUE;
    int target_index;
    haddr_t target_addr;
    struct mssg_t reply;

    if ( ( mssg_ptr == NULL ) ||
         ( mssg_ptr->req != REQ_ENTRY_READS_CODE ) ||
         ( mssg_ptr->magic != MSSG_MAGIC ) ) {

        nerrors++;
        success = FALSE;
        if ( verbose ) {
            HDfprintf(stdout, "%d:%s: Bad mssg on entry.\n",
                      world_mpi_rank, FUNC);
        }
    }

    if ( success ) {

        target_addr = mssg_ptr->base_addr;
        target_index = addr_to_datum_index(target_addr);

        if ( target_index < 0 ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: addr lookup failed for %a.\n",
                          world_mpi_rank, FUNC, target_addr);
            }
        } else {

            /* compose the reply message */
            reply.req       = REQ_ENTRY_READS_RPLY_CODE;
            reply.src       = world_mpi_rank;
            reply.dest      = mssg_ptr->src;
            reply.mssg_num  = -1; /* set by send function */
            reply.base_addr = target_addr;
            reply.len       = 0;
            reply.ver       = 0;
            reply.count     = (unsigned)(data[target_index].reads);
            reply.magic     = MSSG_MAGIC;
        }
    }

    if ( success ) {

        success = send_mssg(&reply, TRUE);
    }

    if ( report_mssg ) {

        if ( success ) {

            HDfprintf(stdout, "%d request entry 0x%llx reads = %ld.\n",
                      (int)(mssg_ptr->src),
                      (long long)(data[target_index].base_addr),
                      (long)(data[target_index].reads));

        } else {

            HDfprintf(stdout, "%d request entry 0x%llx reads = %ld FAILED.\n",
                      (int)(mssg_ptr->src),
                      (long long)(data[target_index].base_addr),
                      (long)(data[target_index].reads));

        }
    }

    return(success);

} /* serve_entry_reads_request() */


/*****************************************************************************
 *
 * Function:    serve_rw_count_reset_request()
 *
 * Purpose:    Serve read/write count reset request.
 *
 *        The function accepts a pointer to an instance of struct
 *        mssg_t as input.  If all sanity checks pass, it resets the
 *        read/write counters, and sends a confirmation message to
 *        the calling process.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 5/5/10
 *
 *****************************************************************************/
static hbool_t
serve_rw_count_reset_request(struct mssg_t * mssg_ptr)
{
    hbool_t report_mssg = FALSE;
    hbool_t success = TRUE;
    struct mssg_t reply;

    if ( ( mssg_ptr == NULL ) ||
         ( mssg_ptr->req != REQ_RW_COUNT_RESET_CODE ) ||
         ( mssg_ptr->magic != MSSG_MAGIC ) ) {

        nerrors++;
        success = FALSE;
        if ( verbose ) {
            HDfprintf(stdout, "%d:%s: Bad mssg on entry.\n",
                      world_mpi_rank, FUNC);
        }
    }

    if ( success ) {

        success = reset_server_counters();
    }

    if ( success ) {

        /* compose the reply message */
        reply.req       = REQ_RW_COUNT_RESET_RPLY_CODE;
        reply.src       = world_mpi_rank;
        reply.dest      = mssg_ptr->src;
        reply.mssg_num  = -1; /* set by send function */
        reply.base_addr = 0;
        reply.len       = 0;
        reply.ver       = 0;
        reply.count     = 0;
        reply.magic     = MSSG_MAGIC;
    }

    if ( success ) {

        success = send_mssg(&reply, TRUE);
    }

    if ( report_mssg ) {

        if ( success ) {

            HDfprintf(stdout, "%d request R/W counter reset.\n",
                      (int)(mssg_ptr->src));

        } else {

            HDfprintf(stdout, "%d request R/w counter reset FAILED.\n",
                      (int)(mssg_ptr->src));

        }
    }

    return(success);

} /* serve_rw_count_reset_request() */


/*****************************************************************************/
/**************************** Call back functions ****************************/
/*****************************************************************************/


/*-------------------------------------------------------------------------
 * Function:    datum_get_initial_load_size
 *
 * Purpose:    Query the image size for an entry before deserializing it
 *
 * Return:    SUCCEED
 *
 * Programmer:    Quincey Koziol
 *              5/18/10
 *
 *-------------------------------------------------------------------------
 */
static herr_t
datum_get_initial_load_size(void *udata_ptr, size_t *image_len_ptr)
{
    haddr_t addr = *(haddr_t *)udata_ptr;
    int idx;
    struct datum * entry_ptr;

    HDassert( udata_ptr );
    HDassert( image_len_ptr );

    idx = addr_to_datum_index(addr);

    HDassert( idx >= 0 );
    HDassert( idx < NUM_DATA_ENTRIES );
    HDassert( idx < virt_num_data_entries );

    entry_ptr = &(data[idx]);

    HDassert( addr == entry_ptr->base_addr );
    HDassert( ! entry_ptr->global_pinned );
    HDassert( ! entry_ptr->local_pinned );

    if ( callbacks_verbose ) {

        HDfprintf(stdout,
    "%d: get_initial_load_size() idx = %d, addr = %ld, len = %d.\n",
              world_mpi_rank, idx, (long)addr, (int)entry_ptr->local_len);
    fflush(stdout);
    }

    /* Set image length size */
    *image_len_ptr = entry_ptr->local_len;

    return(SUCCEED);
} /* get_initial_load_size() */


/*-------------------------------------------------------------------------
 * Function:    datum_deserialize
 *
 * Purpose:    deserialize the entry.
 *
 * Return:    void * (pointer to the in core representation of the entry)
 *
 * Programmer:    John Mainzer
 *              9/20/07
 *
 *-------------------------------------------------------------------------
 */
static void *
datum_deserialize(const void H5_ATTR_NDEBUG_UNUSED *image_ptr,
                  H5_ATTR_UNUSED size_t len,
                  void * udata_ptr,
                  hbool_t * dirty_ptr)
{
    haddr_t addr = *(haddr_t *)udata_ptr;
    hbool_t success = TRUE;
    int idx;
    struct datum * entry_ptr = NULL;

    HDassert( image_ptr != NULL );

    idx = addr_to_datum_index(addr);

    HDassert( idx >= 0 );
    HDassert( idx < NUM_DATA_ENTRIES );
    HDassert( idx < virt_num_data_entries );

    entry_ptr = &(data[idx]);

    HDassert( addr == entry_ptr->base_addr );
    HDassert( ! entry_ptr->global_pinned );
    HDassert( ! entry_ptr->local_pinned );

    HDassert( dirty_ptr );

    if ( callbacks_verbose ) {

        HDfprintf(stdout,
    "%d: deserialize() idx = %d, addr = %ld, len = %d, is_dirty = %d.\n",
    world_mpi_rank, idx, (long)addr, (int)len,
    (int)(entry_ptr->header.is_dirty));
    fflush(stdout);
    }

    *dirty_ptr = FALSE;

    if ( ! success ) {

        entry_ptr = NULL;

    }

    return(entry_ptr);

} /* deserialize() */


/*-------------------------------------------------------------------------
 * Function:    datum_image_len
 *
 * Purpose:    Return the real (and possibly reduced) length of the image.
 *         The helper functions verify that the correct version of
 *         deserialize is being called, and then call deserialize
 *         proper.
 *
 * Return:    SUCCEED
 *
 * Programmer:    John Mainzer
 *              9/19/07
 *
 *-------------------------------------------------------------------------
 */
static herr_t
datum_image_len(const void *thing, size_t *image_len)
{
    int idx;
    struct datum * entry_ptr;

    HDassert( thing );
    HDassert( image_len );

    entry_ptr = (struct datum *)thing;

    idx = addr_to_datum_index(entry_ptr->base_addr);

    HDassert( idx >= 0 );
    HDassert( idx < NUM_DATA_ENTRIES );
    HDassert( idx < virt_num_data_entries );
    HDassert( &(data[idx]) == entry_ptr );
    HDassert( entry_ptr->local_len > 0 );
    HDassert( entry_ptr->local_len <= entry_ptr->len );

    if(callbacks_verbose) {
        HDfprintf(stdout,
        "%d: image_len() idx = %d, addr = %ld, len = %d.\n",
        world_mpi_rank, idx, (long)(entry_ptr->base_addr),
        (int)(entry_ptr->local_len));
    fflush(stdout);
    }

    HDassert( entry_ptr->header.addr == entry_ptr->base_addr );

    *image_len = entry_ptr->local_len;

    return(SUCCEED);
} /* datum_image_len() */


/*-------------------------------------------------------------------------
 * Function:    datum_serialize
 *
 * Purpose:    Serialize the supplied entry.
 *
 * Return:    SUCCEED if successful, FAIL otherwise.
 *
 * Programmer:    John Mainzer
 *              10/30/07
 *
 *-------------------------------------------------------------------------
 */
static herr_t
datum_serialize(const H5F_t *f,
                void H5_ATTR_NDEBUG_UNUSED *image_ptr,
                size_t len,
                void *thing_ptr)
{
    herr_t ret_value = SUCCEED;
    int idx;
    struct datum * entry_ptr;
    struct H5AC_aux_t * aux_ptr;

    HDassert( thing_ptr );
    HDassert( image_ptr );

    entry_ptr = (struct datum *)thing_ptr;

    HDassert( f );
    HDassert( f->shared );
    HDassert( f->shared->cache );
    HDassert( f->shared->cache->magic == H5C__H5C_T_MAGIC );
    HDassert( f->shared->cache->aux_ptr );

    aux_ptr = (H5AC_aux_t *)(f->shared->cache->aux_ptr);

    HDassert( aux_ptr );
    HDassert( aux_ptr->magic == H5AC__H5AC_AUX_T_MAGIC );

    entry_ptr->aux_ptr = aux_ptr;

    idx = addr_to_datum_index(entry_ptr->base_addr);

    HDassert( idx >= 0 );
    HDassert( idx < NUM_DATA_ENTRIES );
    HDassert( idx < virt_num_data_entries );
    HDassert( &(data[idx]) == entry_ptr );

    if ( callbacks_verbose ) {

        HDfprintf(stdout,
        "%d: serialize() idx = %d, addr = %ld, len = %d.\n",
        world_mpi_rank, idx, (long)entry_ptr->header.addr, (int)len);
    fflush(stdout);
    }

    HDassert( entry_ptr->header.addr == entry_ptr->base_addr );
    HDassert( ( entry_ptr->header.size == entry_ptr->len ) ||
              ( entry_ptr->header.size == entry_ptr->local_len ) );

    HDassert( entry_ptr->header.is_dirty == entry_ptr->dirty );

    datum_flushes++;

    if ( entry_ptr->header.is_pinned ) {

        datum_pinned_flushes++;
        HDassert( entry_ptr->global_pinned || entry_ptr->local_pinned );
    }

    return(ret_value);

} /* datum_serialize() */


/*-------------------------------------------------------------------------
 * Function:    datum_notify
 *
 * Purpose:    Do the communication with the server we used to do in the
 *        flush and load callbacks in the version 2 cache.
 *
 * Return:    SUCCEED
 *
 * Programmer:    John Mainzer
 *              1/12/15
 *
 *-------------------------------------------------------------------------
 */
static herr_t
datum_notify(H5C_notify_action_t action, void *thing)
{
    hbool_t was_dirty = FALSE;
    herr_t ret_value = SUCCEED;
    struct datum * entry_ptr;
    struct H5AC_aux_t * aux_ptr;
    struct mssg_t mssg;
    int idx;

    HDassert( thing );

    entry_ptr = (struct datum *)thing;

    idx = addr_to_datum_index(entry_ptr->base_addr);

    HDassert( idx >= 0 );
    HDassert( idx < NUM_DATA_ENTRIES );
    HDassert( idx < virt_num_data_entries );
    HDassert( &(data[idx]) == entry_ptr );

    if ( callbacks_verbose ) {

        HDfprintf(stdout,
                  "%d: notify() action = %d, idx = %d, addr = %ld.\n",
                  world_mpi_rank, (int) action, idx,
                  (long)entry_ptr->header.addr);
        fflush(stdout);
    }

    HDassert( entry_ptr->header.addr == entry_ptr->base_addr );
    /* Skip this check when the entry is being dirtied, since the resize
     * operation sends the message before the len/local_len is updated
     * (after the resize operation completes successfully) (QAK - 2016/10/19)
     */
    if(H5AC_NOTIFY_ACTION_ENTRY_DIRTIED != action)
        HDassert( ( entry_ptr->header.size == entry_ptr->len ) ||
                  ( entry_ptr->header.size == entry_ptr->local_len ) );

    switch ( action )
    {
        case H5AC_NOTIFY_ACTION_AFTER_INSERT:
            if ( callbacks_verbose ) {

                HDfprintf(stdout,
                      "%d: notify() action = insert, idx = %d, addr = %ld.\n",
                      world_mpi_rank, idx, (long)entry_ptr->header.addr);
                fflush(stdout);
            }
            /* do nothing */
            break;

        case H5AC_NOTIFY_ACTION_AFTER_LOAD:
            if ( callbacks_verbose ) {

                HDfprintf(stdout,
                      "%d: notify() action = load, idx = %d, addr = %ld.\n",
                      world_mpi_rank, idx, (long)entry_ptr->header.addr);
                fflush(stdout);
            }

            /* compose the read message */
            mssg.req       = READ_REQ_CODE;
            mssg.src       = world_mpi_rank;
            mssg.dest      = world_server_mpi_rank;
            mssg.mssg_num  = -1; /* set by send function */
            mssg.base_addr = entry_ptr->base_addr;
            H5_CHECKED_ASSIGN(mssg.len, unsigned, entry_ptr->len, size_t);
            mssg.ver       = 0; /* bogus -- should be corrected by server */
            mssg.count     = 0; /* not used */
            mssg.magic     = MSSG_MAGIC;

            if ( ! send_mssg(&mssg, FALSE) ) {

                nerrors++;
                ret_value = FAIL;
                if ( verbose ) {
                    HDfprintf(stdout, "%d:%s: send_mssg() failed.\n",
                              world_mpi_rank, FUNC);
                }
            }

            if ( ret_value == SUCCEED ) {

                if ( ! recv_mssg(&mssg, READ_REQ_REPLY_CODE) ) {

                    nerrors++;
                    ret_value = FAIL;
                    if ( verbose ) {
                        HDfprintf(stdout, "%d:%s: recv_mssg() failed.\n",
                                  world_mpi_rank, FUNC);
                    }
                }
            }

            if ( ret_value == SUCCEED ) {

                if ( ( mssg.req != READ_REQ_REPLY_CODE ) ||
                     ( mssg.src != world_server_mpi_rank ) ||
                     ( mssg.dest != world_mpi_rank ) ||
                     ( mssg.base_addr != entry_ptr->base_addr ) ||
                     ( mssg.len != entry_ptr->len ) ||
                     ( mssg.ver < entry_ptr->ver ) ||
                     ( mssg.magic != MSSG_MAGIC ) ) {

                    nerrors++;
                    ret_value = FAIL;
                    if ( verbose ) {
                        HDfprintf(stdout,
                                  "%d:%s: Bad data in read req reply.\n",
                                  world_mpi_rank, FUNC);
                    }

#if 0 /* This has been useful debugging code -- keep it for now. */
                if ( mssg.req != READ_REQ_REPLY_CODE ) {

                HDfprintf(stdout,
                                  "%d:%s: mssg.req != READ_REQ_REPLY_CODE.\n",
                    world_mpi_rank, FUNC);
                HDfprintf(stdout, "%d:%s: mssg.req = %d.\n",
                    world_mpi_rank, FUNC, (int)(mssg.req));
                }

                if ( mssg.src != world_server_mpi_rank ) {

                HDfprintf(stdout,
                                 "%d:%s: mssg.src != world_server_mpi_rank.\n",
                    world_mpi_rank, FUNC);
                }

                if ( mssg.dest != world_mpi_rank ) {

                HDfprintf(stdout,
                                  "%d:%s: mssg.dest != world_mpi_rank.\n",
                    world_mpi_rank, FUNC);
                    }

                if ( mssg.base_addr != entry_ptr->base_addr ) {

                HDfprintf(stdout,
                "%d:%s: mssg.base_addr != entry_ptr->base_addr.\n",
                world_mpi_rank, FUNC);
                HDfprintf(stdout, "%d:%s: mssg.base_addr = %a.\n",
                    world_mpi_rank, FUNC, mssg.base_addr);
                HDfprintf(stdout,
                                  "%d:%s: entry_ptr->base_addr = %a.\n",
                    world_mpi_rank, FUNC,
                                   entry_ptr->base_addr);
                    }

                if ( mssg.len != entry_ptr->len ) {

                HDfprintf(stdout,
                                  "%d:%s: mssg.len != entry_ptr->len.\n",
                    world_mpi_rank, FUNC);
                HDfprintf(stdout, "%d:%s: mssg.len = %a.\n",
                    world_mpi_rank, FUNC, mssg.len);
                    }

                if ( mssg.ver < entry_ptr->ver ) {

                HDfprintf(stdout,
                                  "%d:%s: mssg.ver < entry_ptr->ver.\n",
                    world_mpi_rank, FUNC);
                    }

                if ( mssg.magic != MSSG_MAGIC ) {

                HDfprintf(stdout, "%d:%s: mssg.magic != MSSG_MAGIC.\n",
                    world_mpi_rank, FUNC);
                     }
#endif /* JRM */

                } else {

                    entry_ptr->ver = mssg.ver;
                    entry_ptr->dirty = FALSE;
                    datum_loads++;
                }
            }
            break;

    case H5C_NOTIFY_ACTION_AFTER_FLUSH:
            if ( callbacks_verbose ) {

                HDfprintf(stdout,
                      "%d: notify() action = flush, idx = %d, addr = %ld.\n",
                      world_mpi_rank, idx, (long)entry_ptr->header.addr);
                fflush(stdout);
            }

            HDassert( entry_ptr->aux_ptr );
            HDassert( entry_ptr->aux_ptr->magic == H5AC__H5AC_AUX_T_MAGIC );
            aux_ptr = entry_ptr->aux_ptr;
            entry_ptr->aux_ptr = NULL;

        HDassert(entry_ptr->header.is_dirty); /* JRM */

            if ( ( file_mpi_rank != 0 ) &&
                 ( entry_ptr->dirty ) &&
                 ( aux_ptr->metadata_write_strategy ==
                   H5AC_METADATA_WRITE_STRATEGY__PROCESS_0_ONLY ) ) {

                ret_value = FAIL;
                HDfprintf(stdout,
                     "%d:%s: Flushed dirty entry from non-zero file process.",
                     world_mpi_rank, FUNC);
            }

            if ( ret_value == SUCCEED ) {

                if ( entry_ptr->header.is_dirty ) {

                was_dirty = TRUE; /* so we will receive the ack
                                       * if requested
                                       */

                    /* compose the message */
                    mssg.req       = WRITE_REQ_CODE;
                    mssg.src       = world_mpi_rank;
                    mssg.dest      = world_server_mpi_rank;
                    mssg.mssg_num  = -1; /* set by send function */
                    mssg.base_addr = entry_ptr->base_addr;
                    H5_CHECKED_ASSIGN(mssg.len, unsigned, entry_ptr->len, size_t);
                    mssg.ver       = entry_ptr->ver;
                    mssg.count     = 0;
                    mssg.magic     = MSSG_MAGIC;

                    if ( ! send_mssg(&mssg, FALSE) ) {

                        nerrors++;
                        ret_value = FAIL;
                        if ( verbose ) {
                            HDfprintf(stdout, "%d:%s: send_mssg() failed.\n",
                                      world_mpi_rank, FUNC);
                        }
                    }
                    else
                    {
                        entry_ptr->dirty = FALSE;
            entry_ptr->flushed = TRUE;
                    }
                }
            }

#if DO_WRITE_REQ_ACK

            if ( ( ret_value == SUCCEED ) && ( was_dirty ) ) {

                if ( ! recv_mssg(&mssg, WRITE_REQ_ACK_CODE) ) {

                    nerrors++;
                    ret_value = FAIL;
                    if ( verbose ) {
                        HDfprintf(stdout, "%d:%s: recv_mssg() failed.\n",
                                  world_mpi_rank, FUNC);
                    }
                } else if ( ( mssg.req != WRITE_REQ_ACK_CODE ) ||
                            ( mssg.src != world_server_mpi_rank ) ||
                            ( mssg.dest != world_mpi_rank ) ||
                            ( mssg.base_addr != entry_ptr->base_addr ) ||
                            ( mssg.len != entry_ptr->len ) ||
                            ( mssg.ver != entry_ptr->ver ) ||
                            ( mssg.magic != MSSG_MAGIC ) ) {

                    nerrors++;
                    ret_value = FAIL;
                    if ( verbose ) {
                        HDfprintf(stdout,
                                  "%d:%s: Bad data in write req ack.\n",
                                  world_mpi_rank, FUNC);
                    }
                }
            }

#endif /* DO_WRITE_REQ_ACK */

            datum_flushes++;

            if ( entry_ptr->header.is_pinned ) {

                datum_pinned_flushes++;
                HDassert(entry_ptr->global_pinned || entry_ptr->local_pinned);
            }
        break;

        case H5AC_NOTIFY_ACTION_BEFORE_EVICT:
            if ( callbacks_verbose ) {

                HDfprintf(stdout,
                      "%d: notify() action = evict, idx = %d, addr = %ld.\n",
                      world_mpi_rank, idx, (long)entry_ptr->header.addr);
                fflush(stdout);
            }

            /* do nothing */
            break;

        case H5AC_NOTIFY_ACTION_ENTRY_DIRTIED:
            if ( callbacks_verbose ) {

                HDfprintf(stdout,
                      "%d: notify() action = entry dirty, idx = %d, addr = %ld.\n",
                      world_mpi_rank, idx, (long)entry_ptr->header.addr);
                fflush(stdout);
            }

            /* do nothing */
            break;

        case H5AC_NOTIFY_ACTION_ENTRY_CLEANED:
            if ( callbacks_verbose ) {

                HDfprintf(stdout,
                      "%d: notify() action = entry clean, idx = %d, addr = %ld.\n",
                      world_mpi_rank, idx, (long)entry_ptr->header.addr);
                fflush(stdout);
            }

            entry_ptr->cleared = TRUE;
            entry_ptr->dirty = FALSE;

            datum_clears++;

            if(entry_ptr->header.is_pinned) {
                datum_pinned_clears++;
                HDassert( entry_ptr->global_pinned || entry_ptr->local_pinned );
            } /* end if */

            break;

        case H5AC_NOTIFY_ACTION_CHILD_DIRTIED:
            if ( callbacks_verbose ) {

                HDfprintf(stdout,
                      "%d: notify() action = child entry dirty, idx = %d, addr = %ld.\n",
                      world_mpi_rank, idx, (long)entry_ptr->header.addr);
                fflush(stdout);
            }

            /* do nothing */
            break;

        case H5AC_NOTIFY_ACTION_CHILD_CLEANED:
            if ( callbacks_verbose ) {

                HDfprintf(stdout,
                      "%d: notify() action = child entry clean, idx = %d, addr = %ld.\n",
                      world_mpi_rank, idx, (long)entry_ptr->header.addr);
                fflush(stdout);
            }

            /* do nothing */
            break;

        case H5AC_NOTIFY_ACTION_CHILD_UNSERIALIZED:
            if ( callbacks_verbose ) {

                HDfprintf(stdout,
                      "%d: notify() action = child entry unserialized, idx = %d, addr = %ld.\n",
                      world_mpi_rank, idx, (long)entry_ptr->header.addr);
                fflush(stdout);
            }

            /* do nothing */
            break;

        case H5AC_NOTIFY_ACTION_CHILD_SERIALIZED:
            if ( callbacks_verbose ) {

                HDfprintf(stdout,
                      "%d: notify() action = child entry serialized, idx = %d, addr = %ld.\n",
                      world_mpi_rank, idx, (long)entry_ptr->header.addr);
                fflush(stdout);
            }

            /* do nothing */
            break;

    default:
            nerrors++;
            ret_value = FAIL;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: Unknown notify action.\n",
                          world_mpi_rank, FUNC);
            }
        break;
    }

    return(ret_value);

} /* datum_notify() */


/*-------------------------------------------------------------------------
 * Function:    datum_free_icr
 *
 * Purpose:    Nominally, this callback is supposed to free the
 *         in core representation of the entry.
 *
 *         In the context of this test bed, we use it to do
 *         do all the processing we used to do on a destroy.
 *
 * Return:    SUCCEED
 *
 * Programmer:    John Mainzer
 *              9/19/07
 *
 *-------------------------------------------------------------------------
 */
static herr_t
datum_free_icr(void * thing)
{
    int idx;
    struct datum * entry_ptr;

    HDassert( thing );

    entry_ptr = (struct datum *)thing;

    idx = addr_to_datum_index(entry_ptr->base_addr);

    HDassert( idx >= 0 );
    HDassert( idx < NUM_DATA_ENTRIES );
    HDassert( idx < virt_num_data_entries );
    HDassert( &(data[idx]) == entry_ptr );

    if ( callbacks_verbose ) {

        HDfprintf(stdout,
    "%d: free_icr() idx = %d, dirty = %d.\n",
        world_mpi_rank, idx, (int)(entry_ptr->dirty));
    fflush(stdout);
    }

    HDassert( entry_ptr->header.addr == entry_ptr->base_addr );
    HDassert( ( entry_ptr->header.size == entry_ptr->len ) ||
              ( entry_ptr->header.size == entry_ptr->local_len ) );

    HDassert( !(entry_ptr->header.is_dirty) );
    HDassert( !(entry_ptr->global_pinned) );
    HDassert( !(entry_ptr->local_pinned) );
    HDassert( !(entry_ptr->header.is_pinned) );

    datum_destroys++;

    return(SUCCEED);
} /* datum_free_icr() */


/*****************************************************************************/
/************************** test utility functions ***************************/
/*****************************************************************************/

/*****************************************************************************
 * Function:    expunge_entry()
 *
 * Purpose:     Expunge the entry indicated by the type and index, mark it
 *        as clean, and don't increment its version number.
 *
 *        Do nothing if nerrors is non-zero on entry.
 *
 * Return:      void
 *
 * Programmer:  John Mainzer
 *              07/11/06
 *
 *****************************************************************************/
static void
expunge_entry(H5F_t * file_ptr,
              int32_t idx)
{
    hbool_t in_cache;
    herr_t result;
    struct datum * entry_ptr;

    HDassert( file_ptr );
    HDassert( ( 0 <= idx ) && ( idx < NUM_DATA_ENTRIES ) );
    HDassert( idx < virt_num_data_entries );

    entry_ptr = &(data[idx]);

    HDassert( !(entry_ptr->locked) );
    HDassert( !(entry_ptr->global_pinned) );
    HDassert( !(entry_ptr->local_pinned) );

    entry_ptr->dirty = FALSE;

    if ( nerrors == 0 ) {

        result = H5AC_expunge_entry(file_ptr, &(types[0]), entry_ptr->header.addr, H5AC__NO_FLAGS_SET);

        if ( result < 0 ) {

            nerrors++;
            if ( verbose ) {
            HDfprintf(stdout, "%d:%s: Error in H5AC_expunge_entry().\n",
                    world_mpi_rank, FUNC);
            }
        }

        HDassert( ((entry_ptr->header).type)->id == DATUM_ENTRY_TYPE );
    HDassert( ! ((entry_ptr->header).is_dirty) );

    result = H5C_get_entry_status(file_ptr, entry_ptr->base_addr,
                    NULL, &in_cache, NULL, NULL, NULL, NULL, NULL, NULL, NULL);

    if ( result < 0 ) {

            nerrors++;
            if ( verbose ) {
            HDfprintf(stdout, "%d:%s: Error in H5C_get_entry_status().\n",
                    world_mpi_rank, FUNC);
            }
        } else if ( in_cache ) {

            nerrors++;
            if ( verbose ) {
            HDfprintf(stdout, "%d:%s: Expunged entry still in cache?!?\n",
                    world_mpi_rank, FUNC);
            }
        }
    }

    return;

} /* expunge_entry() */


/*****************************************************************************
 * Function:    insert_entry()
 *
 * Purpose:     Insert the entry indicated by the type and index, mark it
 *        as dirty, and increment its version number.
 *
 *        Do nothing if nerrors is non-zero on entry.
 *
 * Return:      void
 *
 * Programmer:  John Mainzer
 *              01/04/06
 *
 * Modifications:
 *
 *              JRM -- 8/11/06
 *              Updated code to reflect the fact that entries can now be
 *              inserted pinned.  Note that since all inserts are dirty,
 *              any pins must be global pins.
 *
 *****************************************************************************/
static void
insert_entry(H5C_t * cache_ptr,
             H5F_t * file_ptr,
             int32_t idx,
             unsigned int flags)
{
    hbool_t insert_pinned;
    herr_t result;
    struct datum * entry_ptr;

    HDassert( cache_ptr );
    HDassert( file_ptr );
    HDassert( ( 0 <= idx ) && ( idx < NUM_DATA_ENTRIES ) );
    HDassert( idx < virt_num_data_entries );

    entry_ptr = &(data[idx]);

    HDassert( !(entry_ptr->locked) );

    insert_pinned = ((flags & H5C__PIN_ENTRY_FLAG) != 0 );

    if ( nerrors == 0 ) {

        (entry_ptr->ver)++;
        entry_ptr->dirty = TRUE;

        result = H5AC_insert_entry(file_ptr, &(types[0]),
                entry_ptr->base_addr, (void *)(&(entry_ptr->header)), flags);

        if ( ( result < 0 ) ||
             ( entry_ptr->header.type != &(types[0]) ) ||
             ( entry_ptr->len != entry_ptr->header.size ) ||
             ( entry_ptr->base_addr != entry_ptr->header.addr ) ) {

            nerrors++;
            if ( verbose ) {
            HDfprintf(stdout, "%d:%s: Error in H5AC_insert_entry().\n",
                    world_mpi_rank, FUNC);
            }
        }

        if ( ! (entry_ptr->header.is_dirty) ) {

        /* it is possible that we just exceeded the dirty bytes
        * threshold, triggering a write of the newly inserted
        * entry.  Test for this, and only flag an error if this
        * is not the case.
        */

        struct H5AC_aux_t * aux_ptr;

        aux_ptr = ((H5AC_aux_t *)(cache_ptr->aux_ptr));

        if ( ! ( ( aux_ptr != NULL ) &&
            ( aux_ptr->magic == H5AC__H5AC_AUX_T_MAGIC ) &&
            ( aux_ptr->dirty_bytes == 0 ) ) ) {

                nerrors++;
                if ( verbose ) {
                HDfprintf(stdout, "%d:%s: data[%d].header.is_dirty = %d.\n",
                        world_mpi_rank, FUNC, idx,
                              (int)(data[idx].header.is_dirty));
        }
            }
        }

        if ( insert_pinned ) {

            HDassert( entry_ptr->header.is_pinned );
            entry_ptr->global_pinned = TRUE;
        global_pins++;

        } else {

            HDassert( ! ( entry_ptr->header.is_pinned ) );
            entry_ptr->global_pinned = FALSE;

        }

        /* HDassert( entry_ptr->header.is_dirty ); */
        HDassert( ((entry_ptr->header).type)->id == DATUM_ENTRY_TYPE );
    }

    return;

} /* insert_entry() */


/*****************************************************************************
 * Function:    local_pin_and_unpin_random_entries()
 *
 * Purpose:     Pin a random number of randomly selected entries in cache, and
 *              then unpin a random number of entries.
 *
 *              Do nothing if nerrors is non-zero on entry.
 *
 * Return:      void
 *
 * Programmer:  John Mainzer
 *              4/12/06
 *
 *****************************************************************************/
static void
local_pin_and_unpin_random_entries(H5F_t * file_ptr,
                                   int min_idx,
                                   int max_idx,
                int min_count,
                int max_count)
{

    if ( nerrors == 0 ) {

        hbool_t via_unprotect;
        int count;
        int i;
        int idx;

        HDassert( file_ptr );
        HDassert( 0 <= min_idx );
        HDassert( min_idx < max_idx );
        HDassert( max_idx < NUM_DATA_ENTRIES );
        HDassert( max_idx < virt_num_data_entries );
    HDassert( 0 <= min_count );
    HDassert( min_count < max_count );

    count = (HDrand() % (max_count - min_count)) + min_count;

    HDassert( min_count <= count );
    HDassert( count <= max_count );

    for ( i = 0; i < count; i++ )
    {
            local_pin_random_entry(file_ptr, min_idx, max_idx);
    }

    count = (HDrand() % (max_count - min_count)) + min_count;

    HDassert( min_count <= count );
    HDassert( count <= max_count );

        i = 0;
    idx = 0;

    while ( ( i < count ) && ( idx >= 0 ) )
    {
        via_unprotect = ( (((unsigned)i) & 0x0001) == 0 );
        idx = local_unpin_next_pinned_entry(file_ptr, idx, via_unprotect);
        i++;
    }
    }

    return;

} /* local_pin_and_unpin_random_entries() */


/*****************************************************************************
 * Function:    local_pin_random_entry()
 *
 * Purpose:     Pin a randomly selected entry in cache, and mark the entry
 *              as being locally pinned.  Since this entry will not in
 *              general be pinned in any other cache, we can't mark it
 *              dirty.
 *
 *              Do nothing if nerrors is non-zero on entry.
 *
 * Return:      void
 *
 * Programmer:  John Mainzer
 *              4/12/06
 *
 *****************************************************************************/
static void
local_pin_random_entry(H5F_t * file_ptr,
                       int min_idx,
                       int max_idx)
{
    int idx;

    if ( nerrors == 0 ) {

        HDassert( file_ptr );
        HDassert( 0 <= min_idx );
        HDassert( min_idx < max_idx );
        HDassert( max_idx < NUM_DATA_ENTRIES );
        HDassert( max_idx < virt_num_data_entries );

    do
    {
            idx = (HDrand() % (max_idx - min_idx)) + min_idx;
            HDassert( min_idx <= idx );
            HDassert( idx <= max_idx );
    }
    while ( data[idx].global_pinned || data[idx].local_pinned );

        pin_entry(file_ptr, idx, FALSE, FALSE);
    }

    return;

} /* local_pin_random_entry() */


/*****************************************************************************
 * Function:    local_unpin_all_entries()
 *
 * Purpose:     Unpin all local pinned entries.
 *
 *              Do nothing if nerrors is non-zero on entry.
 *
 * Return:      void
 *
 * Programmer:  John Mainzer
 *              4/12/06
 *
 *****************************************************************************/
static void
local_unpin_all_entries(H5F_t * file_ptr,
            hbool_t via_unprotect)
{

    if ( nerrors == 0 ) {

        int idx;

        HDassert( file_ptr );

    idx = 0;

    while ( idx >= 0 )
    {
        idx = local_unpin_next_pinned_entry(file_ptr,
                                    idx, via_unprotect);
    }
    }

    return;

} /* local_unpin_all_entries() */


/*****************************************************************************
 * Function:    local_unpin_next_pinned_entry()
 *
 * Purpose:     Find the next locally pinned entry after the specified
 *         starting point, and unpin it.
 *
 *              Do nothing if nerrors is non-zero on entry.
 *
 * Return:      Index of the unpinned entry if there is one, or -1 if
 *              nerrors is non-zero on entry, or if there is no locally
 *              pinned entry.
 *
 * Programmer:  John Mainzer
 *              4/12/06
 *
 *****************************************************************************/
static int
local_unpin_next_pinned_entry(H5F_t * file_ptr,
                              int start_idx,
                hbool_t via_unprotect)
{
    int i = 0;
    int idx = -1;

    if ( nerrors == 0 ) {

        HDassert( file_ptr );
        HDassert( 0 <= start_idx );
        HDassert( start_idx < NUM_DATA_ENTRIES );
        HDassert( start_idx < virt_num_data_entries );

    idx = start_idx;

    while ( ( i < virt_num_data_entries ) &&
        ( ! ( data[idx].local_pinned ) ) )
    {
        i++;
        idx++;
        if ( idx >= virt_num_data_entries ) {
        idx = 0;
        }
    }

    if ( data[idx].local_pinned ) {

        unpin_entry(file_ptr, idx, FALSE, FALSE, via_unprotect);

    } else {

        idx = -1;
    }
    }

    return(idx);

} /* local_unpin_next_pinned_entry() */


/*****************************************************************************
 * Function:    lock_and_unlock_random_entries()
 *
 * Purpose:     Obtain a random number in the closed interval [min_count,
 *        max_count].  Then protect and unprotect that number of
 *        random entries.
 *
 *              Do nothing if nerrors is non-zero on entry.
 *
 * Return:      void
 *
 * Programmer:  John Mainzer
 *              1/12/06
 *
 *****************************************************************************/
static void
lock_and_unlock_random_entries(H5F_t * file_ptr,
                               int min_idx,
                               int max_idx,
                               int min_count,
                               int max_count)
{
    int count;
    int i;

    if ( nerrors == 0 ) {

        HDassert( file_ptr );
        HDassert( 0 <= min_count );
        HDassert( min_count < max_count );

        count = (HDrand() % (max_count - min_count)) + min_count;

        HDassert( min_count <= count );
        HDassert( count <= max_count );

        for ( i = 0; i < count; i++ )
        {
            lock_and_unlock_random_entry(file_ptr, min_idx, max_idx);
        }
    }

    return;

} /* lock_and_unlock_random_entries() */


/*****************************************************************************
 * Function:    lock_and_unlock_random_entry()
 *
 * Purpose:     Protect and then unprotect a random entry with index in
 *        the data[] array in the close interval [min_idx, max_idx].
 *
 *              Do nothing if nerrors is non-zero on entry.
 *
 * Return:      void
 *
 * Programmer:  John Mainzer
 *              1/4/06
 *
 *****************************************************************************/
static void
lock_and_unlock_random_entry(H5F_t * file_ptr,
                             int min_idx,
                             int max_idx)
{
    int idx;

    if ( nerrors == 0 ) {

        HDassert( file_ptr );
        HDassert( 0 <= min_idx );
        HDassert( min_idx < max_idx );
        HDassert( max_idx < NUM_DATA_ENTRIES );
        HDassert( max_idx < virt_num_data_entries );

        idx = (HDrand() % (max_idx - min_idx)) + min_idx;

        HDassert( min_idx <= idx );
        HDassert( idx <= max_idx );

    lock_entry(file_ptr, idx);
    unlock_entry(file_ptr, idx, H5AC__NO_FLAGS_SET);
    }

    return;

} /* lock_and_unlock_random_entry() */


/*****************************************************************************
 * Function:    lock_entry()
 *
 * Purpose:     Protect the entry indicated by the index.
 *
 *              Do nothing if nerrors is non-zero on entry.
 *
 * Return:      void
 *
 * Programmer:  John Mainzer
 *              1/4/06
 *
 * Modifications:
 *
 *         JRM -- 7/11/06
 *         Modified asserts to handle the new local_len field in
 *         datum.
 *
 *****************************************************************************/
static void
lock_entry(H5F_t * file_ptr,
           int32_t idx)
{
    struct datum * entry_ptr;
    H5C_cache_entry_t * cache_entry_ptr;

    if ( nerrors == 0 ) {

        HDassert( ( 0 <= idx ) && ( idx < NUM_DATA_ENTRIES ) );
        HDassert( idx < virt_num_data_entries );

        entry_ptr = &(data[idx]);

    HDassert( ! (entry_ptr->locked) );

        cache_entry_ptr = (H5C_cache_entry_t *)H5AC_protect(file_ptr,
                                        &(types[0]), entry_ptr->base_addr,
                                        &entry_ptr->base_addr,
                                        H5AC__NO_FLAGS_SET);

        if ( ( cache_entry_ptr != (void *)(&(entry_ptr->header)) ) ||
             ( entry_ptr->header.type != &(types[0]) ) ||
             ( ( entry_ptr->len != entry_ptr->header.size ) &&
        ( entry_ptr->local_len != entry_ptr->header.size ) ) ||
             ( entry_ptr->base_addr != entry_ptr->header.addr ) ) {

            nerrors++;
            if ( verbose ) {
        HDfprintf(stdout, "%d:%s: error in H5AC_protect().\n",
                    world_mpi_rank, FUNC);
            }
        } else {

        entry_ptr->locked = TRUE;

    }

        HDassert( ((entry_ptr->header).type)->id == DATUM_ENTRY_TYPE );
    }

    return;

} /* lock_entry() */


/*****************************************************************************
 * Function:    mark_entry_dirty()
 *
 * Purpose:     Mark dirty the entry indicated by the index,
 *
 *              Do nothing if nerrors is non-zero on entry.
 *
 * Return:      void
 *
 * Programmer:  John Mainzer
 *              4/14/06
 *
 *****************************************************************************/
static void
mark_entry_dirty(int32_t idx)
{
    herr_t result;
    struct datum * entry_ptr;

    if ( nerrors == 0 ) {

        HDassert( ( 0 <= idx ) && ( idx < NUM_DATA_ENTRIES ) );
        HDassert( idx < virt_num_data_entries );

        entry_ptr = &(data[idx]);

        HDassert ( entry_ptr->locked || entry_ptr->global_pinned );
    HDassert ( ! (entry_ptr->local_pinned) );

        (entry_ptr->ver)++;
        entry_ptr->dirty = TRUE;

    result = H5AC_mark_entry_dirty( (void *)entry_ptr);

        if ( result < 0 ) {

            nerrors++;
            if ( verbose ) {
            HDfprintf(stdout,
                          "%d:%s: error in H5AC_mark_entry_dirty().\n",
                          world_mpi_rank, FUNC);
            }
        }
    else if ( ! ( entry_ptr->locked ) )
    {
        global_dirty_pins++;
    }
    }

    return;

} /* mark_entry_dirty() */


/*****************************************************************************
 * Function:    pin_entry()
 *
 * Purpose:     Pin the entry indicated by the index.
 *
 *              Do nothing if nerrors is non-zero on entry.
 *
 * Return:      void
 *
 * Programmer:  John Mainzer
 *              4/11/06
 *
 *****************************************************************************/
static void
pin_entry(H5F_t * file_ptr,
          int32_t idx,
    hbool_t global,
    hbool_t dirty)
{
    unsigned int flags = H5AC__PIN_ENTRY_FLAG;
    struct datum * entry_ptr;

    if ( nerrors == 0 ) {

        HDassert( file_ptr );
        HDassert( ( 0 <= idx ) && ( idx < NUM_DATA_ENTRIES ) );
        HDassert( idx < virt_num_data_entries );

        entry_ptr = &(data[idx]);

    HDassert ( ! (entry_ptr->global_pinned) );
    HDassert ( ! (entry_ptr->local_pinned) );
    HDassert ( ! ( dirty && ( ! global ) ) );

    lock_entry(file_ptr, idx);

    if ( dirty ) {

        flags |= H5AC__DIRTIED_FLAG;
    }

    unlock_entry(file_ptr, idx, flags);

        HDassert( (entry_ptr->header).is_pinned );
    HDassert( ( ! dirty ) || ( (entry_ptr->header).is_dirty ) );

    if ( global ) {

        entry_ptr->global_pinned = TRUE;

        global_pins++;

    } else {

        entry_ptr->local_pinned = TRUE;

        local_pins++;

    }
    }

    return;

} /* pin_entry() */


/*****************************************************************************
 * Function:    pin_protected_entry()
 *
 * Purpose:     Insert the entry indicated by the type and index, mark it
 *        as dirty, and increment its version number.
 *
 *        Do nothing if nerrors is non-zero on entry.
 *
 * Return:      void
 *
 * Programmer:  John Mainzer
 *              01/04/06
 *
 *****************************************************************************/
static void
pin_protected_entry(int32_t idx,
            hbool_t global)
{
    herr_t result;
    struct datum * entry_ptr;

    HDassert( ( 0 <= idx ) && ( idx < NUM_DATA_ENTRIES ) );
    HDassert( idx < virt_num_data_entries );

    entry_ptr = &(data[idx]);

    HDassert( entry_ptr->locked );

    if ( nerrors == 0 ) {

    result = H5AC_pin_protected_entry((void *)entry_ptr);

        if ( ( result < 0 ) ||
             ( entry_ptr->header.type != &(types[0]) ) ||
             ( ( entry_ptr->len != entry_ptr->header.size ) &&
        ( entry_ptr->local_len != entry_ptr->header.size ) )||
             ( entry_ptr->base_addr != entry_ptr->header.addr ) ||
        ( ! ( (entry_ptr->header).is_pinned ) ) ) {

            nerrors++;
            if ( verbose ) {
            HDfprintf(stdout,
            "%d:%s: Error in H5AC_pin_protected entry().\n",
                    world_mpi_rank, FUNC);
            }
        }

        if ( global ) {

        entry_ptr->global_pinned = TRUE;

        global_pins++;

    } else {

        entry_ptr->local_pinned = TRUE;

        local_pins++;

    }

        HDassert( ((entry_ptr->header).type)->id == DATUM_ENTRY_TYPE );
    }

    return;

} /* pin_protected_entry() */


/*****************************************************************************
 * Function:    move_entry()
 *
 * Purpose:     Move the entry indicated old_idx to the entry indicated
 *        by new_idex.  Touch up the data array so that flush will
 *        not choke.
 *
 *        Do nothing if nerrors isn't zero, or if old_idx equals
 *        new_idx.
 *
 * Return:      void
 *
 * Programmer:  John Mainzer
 *              1/10/06
 *
 *****************************************************************************/
static void
move_entry(H5F_t * file_ptr,
           int32_t old_idx,
           int32_t new_idx)
{
    herr_t         result;
    int           tmp;
    size_t       tmp_len;
    haddr_t        old_addr = HADDR_UNDEF;
    haddr_t        new_addr = HADDR_UNDEF;
    struct datum * old_entry_ptr;
    struct datum * new_entry_ptr;

    if ( ( nerrors == 0 ) && ( old_idx != new_idx ) ) {

        HDassert( file_ptr );
        HDassert( ( 0 <= old_idx ) && ( old_idx < NUM_DATA_ENTRIES ) );
        HDassert( old_idx < virt_num_data_entries );
        HDassert( ( 0 <= new_idx ) && ( new_idx < NUM_DATA_ENTRIES ) );
        HDassert( new_idx < virt_num_data_entries );

        old_entry_ptr = &(data[old_idx]);
        new_entry_ptr = &(data[new_idx]);

        HDassert( ((old_entry_ptr->header).type)->id == DATUM_ENTRY_TYPE );
        HDassert( !(old_entry_ptr->header.is_protected) );
        HDassert( !(old_entry_ptr->locked) );
        HDassert( old_entry_ptr->len == new_entry_ptr->len );

        old_addr = old_entry_ptr->base_addr;
        new_addr = new_entry_ptr->base_addr;

        /* Moving will mark the entry dirty if it is not already */
        old_entry_ptr->dirty = TRUE;

        /* touch up versions, base_addrs, and data_index.  Do this
         * now as it is possible that the rename will trigger a
         * sync point.
         */
        if(old_entry_ptr->ver < new_entry_ptr->ver)
        old_entry_ptr->ver = new_entry_ptr->ver;
        else
            (old_entry_ptr->ver)++;

        old_entry_ptr->base_addr = new_addr;
        new_entry_ptr->base_addr = old_addr;

        data_index[old_entry_ptr->index] = new_idx;
        data_index[new_entry_ptr->index] = old_idx;

        tmp                  = old_entry_ptr->index;
        old_entry_ptr->index = new_entry_ptr->index;
        new_entry_ptr->index = tmp;

    if(old_entry_ptr->local_len != new_entry_ptr->local_len) {
        tmp_len                  = old_entry_ptr->local_len;
        old_entry_ptr->local_len = new_entry_ptr->local_len;
        new_entry_ptr->local_len = tmp_len;
    } /* end if */

        result = H5AC_move_entry(file_ptr, &(types[0]), old_addr, new_addr);

        if ( ( result < 0 ) || ( old_entry_ptr->header.addr != new_addr ) ) {

            nerrors++;
            if ( verbose ) {
        HDfprintf(stdout, "%d:%s: H5AC_move_entry() failed.\n",
                    world_mpi_rank, FUNC);
            }

        } else {

            HDassert( ((old_entry_ptr->header).type)->id == DATUM_ENTRY_TYPE );

            if ( ! (old_entry_ptr->header.is_dirty) ) {

            /* it is possible that we just exceeded the dirty bytes
            * threshold, triggering a write of the newly inserted
            * entry.  Test for this, and only flag an error if this
            * is not the case.
            */

            struct H5AC_aux_t * aux_ptr;

            aux_ptr = ((H5AC_aux_t *)(file_ptr->shared->cache->aux_ptr));

            if ( ! ( ( aux_ptr != NULL ) &&
                ( aux_ptr->magic == H5AC__H5AC_AUX_T_MAGIC ) &&
                ( aux_ptr->dirty_bytes == 0 ) ) ) {

                    nerrors++;
                    if ( verbose ) {
                    HDfprintf(stdout,
                                  "%d:%s: data[%d].header.is_dirty = %d.\n",
                            world_mpi_rank, FUNC, new_idx,
                                  (int)(data[new_idx].header.is_dirty));
            }
                }
            } else {

                HDassert( old_entry_ptr->header.is_dirty );
            }
        }
    }

} /* move_entry() */


/*****************************************************************************
 *
 * Function:    reset_server_counts()
 *
 * Purpose:    Send a message to the server process requesting it to reset
 *        its counters.  Await confirmation message.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 5/6/10
 *
 *****************************************************************************/
static hbool_t
reset_server_counts(void)
{
    hbool_t success = TRUE; /* will set to FALSE if appropriate. */
    struct mssg_t mssg;

    if ( success ) {

        /* compose the message */
        mssg.req       = REQ_RW_COUNT_RESET_CODE;
        mssg.src       = world_mpi_rank;
        mssg.dest      = world_server_mpi_rank;
        mssg.mssg_num  = -1; /* set by send function */
        mssg.base_addr = 0;
        mssg.len       = 0;
        mssg.ver       = 0;
        mssg.count     = 0;
        mssg.magic     = MSSG_MAGIC;

        if ( ! send_mssg(&mssg, FALSE) ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: send_mssg() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }

    if ( success ) {

        if ( ! recv_mssg(&mssg, REQ_RW_COUNT_RESET_RPLY_CODE) ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: recv_mssg() failed.\n",
                          world_mpi_rank, FUNC);
            }
        } else if ( ( mssg.req != REQ_RW_COUNT_RESET_RPLY_CODE ) ||
                    ( mssg.src != world_server_mpi_rank ) ||
                    ( mssg.dest != world_mpi_rank ) ||
                    ( mssg.base_addr != 0 ) ||
                    ( mssg.len != 0 ) ||
                    ( mssg.ver != 0 ) ||
                    ( mssg.count != 0 ) ||
                    ( mssg.magic != MSSG_MAGIC ) ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout,
                          "%d:%s: Bad data in req r/w counter reset reply.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }

    return(success);

} /* reset_server_counts() */


/*****************************************************************************
 * Function:    resize_entry()
 *
 * Purpose:     Resize the pinned entry indicated by idx to the new_size.
 *         Note that new_size must be greater than 0, and must be
 *         less than or equal to the original size of the entry.
 *
 *        Do nothing if nerrors isn't zero.
 *
 * Return:      void
 *
 * Programmer:  John Mainzer
 *              7/11/06
 *
 *****************************************************************************/
static void
resize_entry(int32_t idx,
        size_t  new_size)
{
    herr_t         result;
    struct datum * entry_ptr;

    if ( nerrors == 0 ) {

        HDassert( ( 0 <= idx ) && ( idx < NUM_DATA_ENTRIES ) );
        HDassert( idx < virt_num_data_entries );

        entry_ptr = &(data[idx]);

        HDassert( ((entry_ptr->header).type)->id == DATUM_ENTRY_TYPE );
        HDassert( !(entry_ptr->locked) );
    HDassert( ( entry_ptr->global_pinned ) &&
        ( ! entry_ptr->local_pinned ) );
    HDassert( ( entry_ptr->header.size == entry_ptr->len ) ||
            ( entry_ptr->header.size == entry_ptr->local_len ) );
    HDassert( new_size > 0 );
    HDassert( new_size <= entry_ptr->len );

    result = H5AC_resize_entry((void *)entry_ptr, new_size);

        if ( result < 0 ) {

            nerrors++;
            if ( verbose ) {
        HDfprintf(stdout, "%d:%s: H5AC_resize_entry() failed.\n",
                    world_mpi_rank, FUNC);
            }

        } else {

            HDassert( ((entry_ptr->header).type)->id == DATUM_ENTRY_TYPE );
            HDassert( entry_ptr->header.is_dirty );
            HDassert( entry_ptr->header.size == new_size );

            entry_ptr->dirty = TRUE;
        entry_ptr->local_len = new_size;

            /* touch up version. */

            (entry_ptr->ver)++;
        }
    }

    return;

} /* resize_entry() */


/*****************************************************************************
 *
 * Function:    setup_cache_for_test()
 *
 * Purpose:    Setup the parallel cache for a test, and return the file id
 *        and a pointer to the cache's internal data structures.
 *
 *        To do this, we must create a file, flush it (so that we
 *        don't have to worry about entries in the metadata cache),
 *        look up the address of the metadata cache, and then instruct
 *        the cache to omit sanity checks on dxpl IDs.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 1/4/06
 *
 *****************************************************************************/
static hbool_t
setup_cache_for_test(hid_t * fid_ptr,
                     H5F_t ** file_ptr_ptr,
                     H5C_t ** cache_ptr_ptr,
                     int metadata_write_strategy)
{
    hbool_t success = FALSE; /* will set to TRUE if appropriate. */
    hbool_t enable_rpt_fcn = FALSE;
    hid_t fid = -1;
    H5AC_cache_config_t config;
    H5AC_cache_config_t test_config;
    H5F_t * file_ptr = NULL;
    H5C_t * cache_ptr = NULL;
    haddr_t actual_base_addr;

    HDassert ( fid_ptr != NULL );
    HDassert ( file_ptr_ptr != NULL );
    HDassert ( cache_ptr_ptr != NULL );

    fid = H5Fcreate(filenames[0], H5F_ACC_TRUNC, H5P_DEFAULT, fapl);

    /* Push API context */
    H5CX_push();

    if ( fid < 0 ) {
        nerrors++;
        if ( verbose ) {
        HDfprintf(stdout, "%d:%s: H5Fcreate() failed.\n",
                      world_mpi_rank, FUNC);
        }
    } else if ( H5Fflush(fid, H5F_SCOPE_GLOBAL) < 0 ) {
        nerrors++;
        if ( verbose ) {
        HDfprintf(stdout, "%d:%s: H5Fflush() failed.\n",
                      world_mpi_rank, FUNC);
        }
    } else {
        file_ptr = (H5F_t *)H5VL_object_verify(fid, H5I_FILE);
    }

    if ( file_ptr == NULL ) {
        nerrors++;
        if ( verbose ) {
        HDfprintf(stdout, "%d:%s: Can't get file_ptr.\n",
                      world_mpi_rank, FUNC);
        }
    } else {
        cache_ptr = file_ptr->shared->cache;
    }

    if ( cache_ptr == NULL ) {
        nerrors++;
        if ( verbose ) {
        HDfprintf(stdout, "%d:%s: Can't get cache_ptr.\n",
                      world_mpi_rank, FUNC);
        }
    } else if ( cache_ptr->magic != H5C__H5C_T_MAGIC ) {
        nerrors++;
        if ( verbose ) {
        HDfprintf(stdout, "%d:%s: Bad cache_ptr magic.\n",
                      world_mpi_rank, FUNC);
        }
    } else {
        cache_ptr->ignore_tags = TRUE;
        *fid_ptr = fid;
        *file_ptr_ptr = file_ptr;
        *cache_ptr_ptr = cache_ptr;
        H5C_stats__reset(cache_ptr);
        success = TRUE;
    }

    if ( success ) {

        config.version = H5AC__CURR_CACHE_CONFIG_VERSION;

        if ( H5AC_get_cache_auto_resize_config(cache_ptr, &config)
             != SUCCEED ) {

        HDfprintf(stdout,
                      "%d:%s: H5AC_get_cache_auto_resize_config(1) failed.\n",
                      world_mpi_rank, FUNC);

        } else {

            config.rpt_fcn_enabled         = enable_rpt_fcn;
            config.metadata_write_strategy = metadata_write_strategy;

            if ( H5AC_set_cache_auto_resize_config(cache_ptr, &config)
        != SUCCEED ) {

            HDfprintf(stdout,
                         "%d:%s: H5AC_set_cache_auto_resize_config() failed.\n",
                          world_mpi_rank, FUNC);

            } else if ( enable_rpt_fcn ) {

                HDfprintf(stdout, "%d:%s: rpt_fcn enabled.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }

    /* verify that the metadata write strategy is set as expected.  Must
     * do this here, as this field is only set in the parallel case.  Hence
     * we can't do our usual checks in the serial case.
     */

    if ( success ) /* verify that the metadata write strategy is as expected */
    {
        if ( cache_ptr->aux_ptr == NULL ) {

            nerrors++;
            if ( verbose ) {
            HDfprintf(stdout, "%d:%s: cache_ptr->aux_ptr == NULL.\n",
                          world_mpi_rank, FUNC);
            }
        } else if ( ((H5AC_aux_t *)(cache_ptr->aux_ptr))->magic !=
                    H5AC__H5AC_AUX_T_MAGIC ) {

            nerrors++;
            if ( verbose ) {
            HDfprintf(stdout,
                "%d:%s: cache_ptr->aux_ptr->magic != H5AC__H5AC_AUX_T_MAGIC.\n",
                world_mpi_rank, FUNC);
            }
        } else if( ((H5AC_aux_t *)(cache_ptr->aux_ptr))->metadata_write_strategy
                   != metadata_write_strategy ) {

            nerrors++;
            if ( verbose ) {
            HDfprintf(stdout,
                    "%d:%s: bad cache_ptr->aux_ptr->metadata_write_strategy\n",
                    world_mpi_rank, FUNC);
            }
        }
    }

    /* also verify that the expected metadata write strategy is reported
     * when we get the current configuration.
     */

    if ( success ) {

        test_config.version = H5AC__CURR_CACHE_CONFIG_VERSION;

        if ( H5AC_get_cache_auto_resize_config(cache_ptr, &test_config)
             != SUCCEED ) {

        HDfprintf(stdout,
                      "%d:%s: H5AC_get_cache_auto_resize_config(2) failed.\n",
                      world_mpi_rank, FUNC);

        } else if ( test_config.metadata_write_strategy !=
                    metadata_write_strategy ) {

            nerrors++;

            if ( verbose ) {

                HDfprintf(stdout,
                          "%d:%s: unexpected metadata_write_strategy.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }

    /* allocate space for test entries -- do this before we set the
     * sync point done callback as it will dirty the superblock, requiring
     * another flush.  If the sync point done callback is set, this will
     * cause a spurious failure.
     */
    if ( success ) { /* allocate space for test entries */

        actual_base_addr = H5MF_alloc(file_ptr, H5FD_MEM_DEFAULT,
                                      (hsize_t)(max_addr + BASE_ADDR));

        if ( actual_base_addr == HADDR_UNDEF ) {

            success = FALSE;
        nerrors++;

            if ( verbose ) {
            HDfprintf(stdout, "%d:%s: H5MF_alloc() failed.\n",
                          world_mpi_rank, FUNC);
            }

        } else if ( actual_base_addr > BASE_ADDR ) {

            /* If this happens, must increase BASE_ADDR so that the
             * actual_base_addr is <= BASE_ADDR.  This should only happen
             * if the size of the superblock is increase.
             */
            success = FALSE;
        nerrors++;

            if ( verbose ) {
            HDfprintf(stdout, "%d:%s: actual_base_addr > BASE_ADDR.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }


    /* flush the file again -- space allocation dirtied superblock */
    if ( success ) {

        if ( H5Fflush(fid, H5F_SCOPE_GLOBAL) < 0 ) {
            nerrors++;
            if ( verbose ) {
            HDfprintf(stdout, "%d:%s: second H5Fflush() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }

#if DO_SYNC_AFTER_WRITE

    if ( success ) {

    if ( H5AC__set_write_done_callback(cache_ptr, do_sync) != SUCCEED ) {

            nerrors++;
            if ( verbose ) {
            HDfprintf(stdout,
            "%d:%s: H5C_set_write_done_callback failed.\n",
                          world_mpi_rank, FUNC);
            }
    }
    }

#endif /* DO_SYNC_AFTER_WRITE */

    if ( success ) {

    if ( H5AC__set_sync_point_done_callback(cache_ptr, verify_writes) != SUCCEED ) {

            nerrors++;
            if ( verbose ) {
            HDfprintf(stdout,
            "%d:%s: H5AC__set_sync_point_done_callback failed.\n",
                          world_mpi_rank, FUNC);
            }
    }
    }

    return(success);

} /* setup_cache_for_test() */


/*****************************************************************************
 *
 * Function:    verify_writes()
 *
 * Purpose:    Verify that the indicated entries have been written exactly
 *        once each, and that the indicated total number of writes
 *        has been processed by the server process.  Flag an error if
 *        discrepency is noted.  Finally reset the counters maintained
 *        by the server process.
 *
 *        This function should only be called by the metadata cache
 *        as the "sync point done" function, as it must do some
 *        synchronization to avoid false positives.
 *
 *        Note that at present, this function does not allow for the
 *        case in which one or more of the indicated entries should
 *        have been written more than once since the last time the
 *        server process's counters were reset.  That is fine for now,
 *        as with the current metadata write strategies, no entry
 *        should be written more than once per sync point.  If this
 *        changes this limitation will have to be revisited.
 *
 * Return:    void.
 *
 * Programmer:    JRM -- 5/9/10
 *
 *****************************************************************************/
static void
verify_writes(unsigned num_writes, haddr_t *written_entries_tbl)
{
    const hbool_t report = FALSE;
    hbool_t proceed = TRUE;
    unsigned u = 0;

    HDassert( world_mpi_rank != world_server_mpi_rank );
    HDassert( ( num_writes == 0 ) ||
              ( written_entries_tbl != NULL ) );

    /* barrier to ensure that all other processes are ready to leave
     * the sync point as well.
     */
    if ( proceed ) {

        if ( MPI_SUCCESS != MPI_Barrier(file_mpi_comm) ) {

            proceed = FALSE;
            nerrors++;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: barrier 1 failed.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }

    if(proceed)
        proceed = verify_total_writes(num_writes);

    while(proceed && u < num_writes) {
        proceed = verify_entry_writes(written_entries_tbl[u], 1);
        u++;
    }

    /* barrier to ensure that all other processes have finished verifying
     * the number of writes before we reset the counters.
     */
    if ( proceed ) {

        if ( MPI_SUCCESS != MPI_Barrier(file_mpi_comm) ) {

            proceed = FALSE;
            nerrors++;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: barrier 2 failed.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }

    if ( proceed ) {

        proceed = reset_server_counts();
    }

    /* if requested, display status of check to stdout */
    if ( ( report ) && ( file_mpi_rank == 0 ) ) {

        if ( proceed ) {

            HDfprintf(stdout, "%d:%s: verified %u writes.\n",
                      world_mpi_rank, FUNC, num_writes);

        } else {

            HDfprintf(stdout, "%d:%s: FAILED to verify %u writes.\n",
                      world_mpi_rank, FUNC, num_writes);

        }
    }

    /* final barrier to ensure that all processes think that the server
     * counters have been reset before we leave the sync point.  This
     * barrier is probaby not necessary at this point in time (5/9/10),
     * but I can think of at least one likely change to the metadata write
     * strategies that will require it -- hence its insertion now.
     */
    if ( proceed ) {

        if ( MPI_SUCCESS != MPI_Barrier(file_mpi_comm) ) {

            proceed = FALSE;
            nerrors++;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: barrier 3 failed.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }

    return;

} /* verify_writes() */


/*****************************************************************************
 *
 * Function:    setup_rand()
 *
 * Purpose:    Use gettimeofday() to obtain a seed for rand(), print the
 *        seed to stdout, and then pass it to srand().
 *
 *        Increment nerrors if any errors are detected.
 *
 * Return:    void.
 *
 * Programmer:    JRM -- 1/12/06
 *
 * Modifications:
 *
 *        JRM -- 5/9/06
 *        Modified function to facilitate setting predefined seeds.
 *
 *****************************************************************************/
static void
setup_rand(void)
{
    hbool_t use_predefined_seeds = FALSE;
    int num_predefined_seeds = 3;
    unsigned predefined_seeds[3] = {18669, 89925, 12577};
    unsigned seed;
    struct timeval tv;

    if ( ( use_predefined_seeds ) &&
         ( world_mpi_size == num_predefined_seeds ) ) {

        HDassert( world_mpi_rank >= 0 );
        HDassert( world_mpi_rank < world_mpi_size );

            seed = predefined_seeds[world_mpi_rank];
        HDfprintf(stdout, "%d:%s: predefined_seed = %d.\n",
                      world_mpi_rank, FUNC, seed);
        fflush(stdout);
            HDsrand(seed);

    } else {

        if ( HDgettimeofday(&tv, NULL) != 0 ) {

            nerrors++;
            if ( verbose ) {
            HDfprintf(stdout, "%d:%s: gettimeofday() failed.\n",
                          world_mpi_rank, FUNC);
            }
        } else {
            seed = (unsigned)tv.tv_usec;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: seed = %d.\n",
                          world_mpi_rank, FUNC, seed);
                fflush(stdout);
            }
            HDsrand(seed);
        }
    }

    return;

} /* setup_rand() */


/*****************************************************************************
 *
 * Function:    take_down_cache()
 *
 * Purpose:    Take down the parallel cache after a test.
 *
 *        To do this, we must close the file, and delete if if
 *        possible.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 1/4/06
 *
 *****************************************************************************/
static hbool_t
take_down_cache(hid_t fid, H5C_t * cache_ptr)
{
    hbool_t success = TRUE; /* will set to FALSE if appropriate. */

    /* flush the file -- this should write out any remaining test
     * entries in the cache.
     */
    if ( ( success ) && ( H5Fflush(fid, H5F_SCOPE_GLOBAL) < 0 ) ) {

        success = FALSE;
        nerrors++;
        if ( verbose ) {
            HDfprintf(stdout, "%d:%s: H5Fflush() failed.\n",
                      world_mpi_rank, FUNC);
        }
    }

    /* Now reset the sync point done callback.  Must do this as with
     * the SWMR mods, the cache will do additional I/O on file close
     * un-related to the test entries, and thereby corrupt our counts
     * of entry writes.
     */
    if ( success ) {

        if ( H5AC__set_sync_point_done_callback(cache_ptr, NULL) != SUCCEED ) {

            success = FALSE;
            nerrors++;
            if ( verbose ) {
                HDfprintf(stdout,
                          "%d:%s: H5AC__set_sync_point_done_callback failed.\n",
                          world_mpi_rank, FUNC);
            }
        }


    }

    /* close the file */
    if ( ( success ) && ( H5Fclose(fid) < 0 ) ) {

        success = FALSE;
        nerrors++;
        if ( verbose ) {
            HDfprintf(stdout, "%d:%s: H5Fclose() failed.\n",
                      world_mpi_rank, FUNC);
        }

    }

    /* Pop API context */
    H5CX_pop();

    if ( success ) {

        if ( world_mpi_rank == world_server_mpi_rank ) {

            if ( HDremove(filenames[0]) < 0 ) {

                success = FALSE;
                nerrors++;
                if ( verbose ) {
                    HDfprintf(stdout, "%d:%s: HDremove() failed.\n",
                              world_mpi_rank, FUNC);
                }
            }
        } else {

        /* verify that there have been no further writes of test
             * entries during the close
             */
            success = verify_total_writes(0);

        }
    }

    return(success);

} /* take_down_cache() */


/*****************************************************************************
 * Function:    verify_entry_reads
 *
 * Purpose:     Query the server to determine the number of times the
 *        indicated entry has been read since the last time the
 *        server counters were reset.
 *
 *        Return TRUE if successful, and if the supplied expected
 *        number of reads matches the number of reads reported by
 *        the server process.
 *
 *        Return FALSE and flag an error otherwise.
 *
 * Return:      TRUE if successful, FALSE otherwise.
 *
 * Programmer:  John Mainzer
 *              5/6/10
 *
 *-------------------------------------------------------------------------
 */
static hbool_t
verify_entry_reads(haddr_t addr,
                   int expected_entry_reads)
{
    hbool_t success = TRUE;
    int reported_entry_reads = 0;
    struct mssg_t mssg;

    if ( success ) {

        /* compose the message */
        mssg.req       = REQ_ENTRY_READS_CODE;
        mssg.src       = world_mpi_rank;
        mssg.dest      = world_server_mpi_rank;
        mssg.mssg_num  = -1; /* set by send function */
        mssg.base_addr = addr;
        mssg.len       = 0; /* not used */
        mssg.ver       = 0; /* not used */
        mssg.count     = 0; /* not used */
        mssg.magic     = MSSG_MAGIC;

        if ( ! send_mssg(&mssg, FALSE) ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: send_mssg() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }

    if ( success ) {

        if ( ! recv_mssg(&mssg, REQ_ENTRY_READS_RPLY_CODE) ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: recv_mssg() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }

    if ( success ) {

        if ( ( mssg.req != REQ_ENTRY_READS_RPLY_CODE ) ||
             ( mssg.src != world_server_mpi_rank ) ||
             ( mssg.dest != world_mpi_rank ) ||
             ( mssg.base_addr != addr ) ||
             ( mssg.len != 0 ) ||
             ( mssg.ver != 0 ) ||
             ( mssg.magic != MSSG_MAGIC ) ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: Bad data in req entry reads reply.\n",
                          world_mpi_rank, FUNC);
            }
        } else {

            H5_CHECKED_ASSIGN(reported_entry_reads, int, mssg.count, unsigned);
        }
    }

    if ( ! success ) {

        if ( reported_entry_reads != expected_entry_reads ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout,
                    "%d:%s: rep/exp entry 0x%llx reads mismatch (%ld/%ld).\n",
                    world_mpi_rank, FUNC, (long long)addr,
                    reported_entry_reads, expected_entry_reads);
            }
        }
    }

    return(success);

} /* verify_entry_reads() */


/*****************************************************************************
 * Function:    verify_entry_writes
 *
 * Purpose:     Query the server to determine the number of times the
 *        indicated entry has been written since the last time the
 *        server counters were reset.
 *
 *        Return TRUE if successful, and if the supplied expected
 *        number of reads matches the number of reads reported by
 *        the server process.
 *
 *        Return FALSE and flag an error otherwise.
 *
 * Return:      TRUE if successful, FALSE otherwise.
 *
 * Programmer:  John Mainzer
 *              5/6/10
 *
 *-------------------------------------------------------------------------
 */
static hbool_t
verify_entry_writes(haddr_t addr,
                    int expected_entry_writes)
{
    hbool_t success = TRUE;
    int reported_entry_writes = 0;
    struct mssg_t mssg;

    if ( success ) {

        /* compose the message */
        mssg.req       = REQ_ENTRY_WRITES_CODE;
        mssg.src       = world_mpi_rank;
        mssg.dest      = world_server_mpi_rank;
        mssg.mssg_num  = -1; /* set by send function */
        mssg.base_addr = addr;
        mssg.len       = 0; /* not used */
        mssg.ver       = 0; /* not used */
        mssg.count     = 0; /* not used */
        mssg.magic     = MSSG_MAGIC;

        if ( ! send_mssg(&mssg, FALSE) ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: send_mssg() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }

    if ( success ) {

        if ( ! recv_mssg(&mssg, REQ_ENTRY_WRITES_RPLY_CODE) ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: recv_mssg() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }

    if ( success ) {

        if ( ( mssg.req != REQ_ENTRY_WRITES_RPLY_CODE ) ||
             ( mssg.src != world_server_mpi_rank ) ||
             ( mssg.dest != world_mpi_rank ) ||
             ( mssg.base_addr != addr ) ||
             ( mssg.len != 0 ) ||
             ( mssg.ver != 0 ) ||
             ( mssg.magic != MSSG_MAGIC ) ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: Bad data in req entry writes reply.\n",
                          world_mpi_rank, FUNC);
            }
        } else {

            H5_CHECKED_ASSIGN(reported_entry_writes, int, mssg.count, unsigned);
        }
    }

    if ( ! success ) {

        if ( reported_entry_writes != expected_entry_writes ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout,
                    "%d:%s: rep/exp entry 0x%llx writes mismatch (%ld/%ld).\n",
                    world_mpi_rank, FUNC, (long long)addr,
                    reported_entry_writes, expected_entry_writes);
            }
        }
    }

    return(success);

} /* verify_entry_writes() */


/*****************************************************************************
 *
 * Function:    verify_total_reads()
 *
 * Purpose:    Query the server to obtain the total reads since the last
 *        server counter reset, and compare this value with the supplied
 *        expected value.
 *
 *        If the values match, return TRUE.
 *
 *        If the values don't match, flag an error and return FALSE.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 5/6/10
 *
 *****************************************************************************/
static hbool_t
verify_total_reads(int expected_total_reads)
{
    hbool_t success = TRUE; /* will set to FALSE if appropriate. */
    long reported_total_reads;
    struct mssg_t mssg;

    if ( success ) {

        /* compose the message */
        mssg.req       = REQ_TTL_READS_CODE;
        mssg.src       = world_mpi_rank;
        mssg.dest      = world_server_mpi_rank;
        mssg.mssg_num  = -1; /* set by send function */
        mssg.base_addr = 0;
        mssg.len       = 0;
        mssg.ver       = 0;
        mssg.count     = 0;
        mssg.magic     = MSSG_MAGIC;

        if ( ! send_mssg(&mssg, FALSE) ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: send_mssg() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }

    if ( success ) {

        if ( ! recv_mssg(&mssg, REQ_TTL_READS_RPLY_CODE) ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: recv_mssg() failed.\n",
                          world_mpi_rank, FUNC);
            }
        } else if ( ( mssg.req != REQ_TTL_READS_RPLY_CODE ) ||
                    ( mssg.src != world_server_mpi_rank ) ||
                    ( mssg.dest != world_mpi_rank ) ||
                    ( mssg.base_addr != 0 ) ||
                    ( mssg.len != 0 ) ||
                    ( mssg.ver != 0 ) ||
                    ( mssg.magic != MSSG_MAGIC ) ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: Bad data in req total reads reply.\n",
                          world_mpi_rank, FUNC);
            }
        } else {

            reported_total_reads = mssg.count;
        }
    }

    if ( success ) {

        if ( reported_total_reads != expected_total_reads ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout,
                    "%d:%s: reported/expected total reads mismatch (%ld/%ld).\n",
                    world_mpi_rank, FUNC,
                    reported_total_reads, expected_total_reads);

            }
        }
    }

    return(success);

} /* verify_total_reads() */


/*****************************************************************************
 *
 * Function:    verify_total_writes()
 *
 * Purpose:    Query the server to obtain the total writes since the last
 *        server counter reset, and compare this value with the supplied
 *        expected value.
 *
 *        If the values match, return TRUE.
 *
 *        If the values don't match, flag an error and return FALSE.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 5/6/10
 *
 *****************************************************************************/
static hbool_t
verify_total_writes(unsigned expected_total_writes)
{
    hbool_t success = TRUE; /* will set to FALSE if appropriate. */
    unsigned reported_total_writes;
    struct mssg_t mssg;

    if ( success ) {

        /* compose the message */
        mssg.req       = REQ_TTL_WRITES_CODE;
        mssg.src       = world_mpi_rank;
        mssg.dest      = world_server_mpi_rank;
        mssg.mssg_num  = -1; /* set by send function */
        mssg.base_addr = 0;
        mssg.len       = 0;
        mssg.ver       = 0;
        mssg.count     = 0;
        mssg.magic     = MSSG_MAGIC;

        if ( ! send_mssg(&mssg, FALSE) ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: send_mssg() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }

    if ( success ) {

        if ( ! recv_mssg(&mssg, REQ_TTL_WRITES_RPLY_CODE) ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: recv_mssg() failed.\n",
                          world_mpi_rank, FUNC);
            }
        } else if ( ( mssg.req != REQ_TTL_WRITES_RPLY_CODE ) ||
                    ( mssg.src != world_server_mpi_rank ) ||
                    ( mssg.dest != world_mpi_rank ) ||
                    ( mssg.base_addr != 0 ) ||
                    ( mssg.len != 0 ) ||
                    ( mssg.ver != 0 ) ||
                    ( mssg.magic != MSSG_MAGIC ) ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: Bad data in req total reads reply.\n",
                          world_mpi_rank, FUNC);
            }
        } else {

            reported_total_writes = mssg.count;
        }
    }

    if ( success ) {

        if ( reported_total_writes != expected_total_writes ) {

            nerrors++;
            success = FALSE;
            if ( verbose ) {
                HDfprintf(stdout,
                   "%d:%s: reported/expected total writes mismatch (%u/%u).\n",
                   world_mpi_rank, FUNC,
                   reported_total_writes, expected_total_writes);
            }
        }
    }

    return(success);

} /* verify_total_writes() */


/*****************************************************************************
 * Function:    unlock_entry()
 *
 * Purpose:     Unprotect the entry indicated by the index.
 *
 *              Do nothing if nerrors is non-zero on entry.
 *
 * Return:      void
 *
 * Programmer:  John Mainzer
 *              1/4/06
 *
 * Modifications:
 *
 *         7/11/06
 *         Updated for the new local_len field in datum.
 *
 *****************************************************************************/
static void
unlock_entry(H5F_t * file_ptr,
             int32_t idx,
             unsigned int flags)
{
    herr_t dirtied;
    herr_t result;
    struct datum * entry_ptr;

    if ( nerrors == 0 ) {

        HDassert( file_ptr );
        HDassert( ( 0 <= idx ) && ( idx < NUM_DATA_ENTRIES ) );
        HDassert( idx < virt_num_data_entries );

        entry_ptr = &(data[idx]);

    HDassert( entry_ptr->locked );

        dirtied = ((flags & H5AC__DIRTIED_FLAG) == H5AC__DIRTIED_FLAG );

        if ( dirtied ) {

            (entry_ptr->ver)++;
            entry_ptr->dirty = TRUE;
        }

        result = H5AC_unprotect(file_ptr, &(types[0]),
                entry_ptr->base_addr, (void *)(&(entry_ptr->header)), flags);

        if ( ( result < 0 ) ||
             ( entry_ptr->header.type != &(types[0]) ) ||
             ( ( entry_ptr->len != entry_ptr->header.size ) &&
        ( entry_ptr->local_len != entry_ptr->header.size ) ) ||
             ( entry_ptr->base_addr != entry_ptr->header.addr ) ) {

            nerrors++;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: error in H5AC_unprotect().\n",
                          world_mpi_rank, FUNC);
            }
        } else {

            entry_ptr->locked = FALSE;

    }

        HDassert( ((entry_ptr->header).type)->id == DATUM_ENTRY_TYPE );

        if ( ( (flags & H5AC__DIRTIED_FLAG) != 0 ) &&
             ( (flags & H5C__DELETED_FLAG) == 0 ) &&
             ( ! ( ( ( world_mpi_rank == 0 ) && ( entry_ptr->flushed ) )
                   ||
                   ( ( world_mpi_rank != 0 ) && ( entry_ptr->cleared ) )
                 )
             )
           ) {
            HDassert( entry_ptr->header.is_dirty );
            HDassert( entry_ptr->dirty );
        }
    }

    return;

} /* unlock_entry() */


/*****************************************************************************
 * Function:    unpin_entry()
 *
 * Purpose:     Unpin the entry indicated by the index.
 *
 *              Do nothing if nerrors is non-zero on entry.
 *
 * Return:      void
 *
 * Programmer:  John Mainzer
 *              4/12/06
 *
 * Modifications:
 *
 *         JRM -- 8/15/06
 *         Added assertion that entry is pinned on entry.
 *
 *****************************************************************************/
static void
unpin_entry(H5F_t * file_ptr,
            int32_t idx,
            hbool_t global,
            hbool_t dirty,
            hbool_t via_unprotect)
{
    herr_t result;
    unsigned int flags = H5AC__UNPIN_ENTRY_FLAG;
    struct datum * entry_ptr;

    if ( nerrors == 0 ) {

        HDassert( file_ptr );
        HDassert( ( 0 <= idx ) && ( idx < NUM_DATA_ENTRIES ) );
        HDassert( idx < virt_num_data_entries );

        entry_ptr = &(data[idx]);

    HDassert( (entry_ptr->header).is_pinned );
    HDassert ( ! ( entry_ptr->global_pinned && entry_ptr->local_pinned) );
    HDassert ( ( global && entry_ptr->global_pinned ) ||
        ( ! global && entry_ptr->local_pinned ) );
    HDassert ( ! ( dirty && ( ! global ) ) );

    if ( via_unprotect ) {

        lock_entry(file_ptr, idx);

        if ( dirty ) {

            flags |= H5AC__DIRTIED_FLAG;
        }

        unlock_entry(file_ptr, idx, flags);

    } else {

        if ( dirty ) {

        mark_entry_dirty(idx);

        }

        result = H5AC_unpin_entry(entry_ptr);

        if ( result < 0 ) {

                nerrors++;
                if ( verbose ) {
                    HDfprintf(stdout, "%d:%s: error in H5AC_unpin_entry().\n",
                        world_mpi_rank, FUNC);
                }
        }
    }

        HDassert( ! ((entry_ptr->header).is_pinned) );

    if ( global ) {

        entry_ptr->global_pinned = FALSE;

    } else {

        entry_ptr->local_pinned = FALSE;

    }
    }

    return;

} /* unpin_entry() */


/*****************************************************************************/
/****************************** test functions *******************************/
/*****************************************************************************/


/*****************************************************************************
 *
 * Function:    server_smoke_check()
 *
 * Purpose:    Quick smoke check for the server process.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 12/21/05
 *
 *****************************************************************************/
static hbool_t
server_smoke_check(void)
{
    hbool_t success = TRUE;
    int max_nerrors;
    struct mssg_t mssg;

    if ( world_mpi_rank == 0 ) {

        TESTING("server smoke check");
    }

    nerrors = 0;
    init_data();
    reset_stats();

    if ( world_mpi_rank == world_server_mpi_rank ) {

    if ( ! server_main() ) {

            /* some error occured in the server -- report failure */
            nerrors++;
            if ( verbose ) {
        HDfprintf(stdout, "%d:%s: server_main() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }
    else /* run the clients */
    {
        /* compose the write message */
        mssg.req       = WRITE_REQ_CODE;
        mssg.src       = world_mpi_rank;
        mssg.dest      = world_server_mpi_rank;
        mssg.mssg_num  = -1; /* set by send function */
        mssg.base_addr = data[world_mpi_rank].base_addr;
        H5_CHECKED_ASSIGN(mssg.len, unsigned, data[world_mpi_rank].len, size_t);
        mssg.ver       = ++(data[world_mpi_rank].ver);
        mssg.count     = 0;
        mssg.magic     = MSSG_MAGIC;

        if ( ! ( success = send_mssg(&mssg, FALSE) ) ) {

            nerrors++;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: send_mssg() failed on write.\n",
                          world_mpi_rank, FUNC);
            }
        }

#if DO_WRITE_REQ_ACK

        /* try to receive the write ack from the server */
        if ( success ) {

            success = recv_mssg(&mssg, WRITE_REQ_ACK_CODE);

            if ( ! success ) {

                nerrors++;
                if ( verbose ) {
                    HDfprintf(stdout, "%d:%s: recv_mssg() failed.\n",
                              world_mpi_rank, FUNC);
                }
            }
        }

        /* verify that we received the expected ack message */
        if ( success ) {

            if ( ( mssg.req != WRITE_REQ_ACK_CODE ) ||
                 ( mssg.src != world_server_mpi_rank ) ||
                 ( mssg.dest != world_mpi_rank ) ||
                 ( mssg.base_addr != data[world_mpi_rank].base_addr ) ||
                 ( mssg.len != data[world_mpi_rank].len ) ||
                 ( mssg.ver != data[world_mpi_rank].ver ) ||
                 ( mssg.magic != MSSG_MAGIC ) ) {

                success = FALSE;
                nerrors++;
                if ( verbose ) {
                    HDfprintf(stdout, "%d:%s: Bad data in write req ack.\n",
                              world_mpi_rank, FUNC);
                }
            }
        }

#endif /* DO_WRITE_REQ_ACK */

    do_sync();

    /* barrier to allow all writes to complete */
        if ( MPI_SUCCESS != MPI_Barrier(file_mpi_comm) ) {

            success = FALSE;
            nerrors++;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: barrier 1 failed.\n",
                          world_mpi_rank, FUNC);
            }
        }

        /* verify that the expected entries have been written, the total */
        if ( success ) {

            success = verify_entry_writes(data[world_mpi_rank].base_addr, 1);
        }

        if ( success ) {

            success = verify_entry_reads(data[world_mpi_rank].base_addr, 0);
        }

        if ( success ) {

            success = verify_total_writes((unsigned)(world_mpi_size - 1));
        }

        if ( success ) {

            success = verify_total_reads(0);
        }

    /* barrier to allow all writes to complete */
        if ( MPI_SUCCESS != MPI_Barrier(file_mpi_comm) ) {

            success = FALSE;
            nerrors++;
            if ( verbose ) {

                HDfprintf(stdout, "%d:%s: barrier 2 failed.\n",
                          world_mpi_rank, FUNC);
            }
        }

        /* compose the read message */
        mssg.req       = READ_REQ_CODE;
        mssg.src       = world_mpi_rank;
        mssg.dest      = world_server_mpi_rank;
        mssg.mssg_num  = -1; /* set by send function */
        mssg.base_addr = data[world_mpi_rank].base_addr;
        H5_CHECKED_ASSIGN(mssg.len, unsigned, data[world_mpi_rank].len, size_t);
        mssg.ver       = 0; /* bogus -- should be corrected by server */
        mssg.count     = 0;
        mssg.magic     = MSSG_MAGIC;

        if ( success ) {

            success = send_mssg(&mssg, FALSE);

            if ( ! success ) {

                nerrors++;
                if ( verbose ) {
                    HDfprintf(stdout, "%d:%s: send_mssg() failed on write.\n",
                              world_mpi_rank, FUNC);
                }
            }
        }

        /* try to receive the reply from the server */
        if ( success ) {

            success = recv_mssg(&mssg, READ_REQ_REPLY_CODE);

            if ( ! success ) {

                nerrors++;
                if ( verbose ) {
                    HDfprintf(stdout, "%d:%s: recv_mssg() failed.\n",
                              world_mpi_rank, FUNC);
                }
            }
        }

        /* verify that we got the expected result */
        if ( success ) {

            if ( ( mssg.req != READ_REQ_REPLY_CODE ) ||
                 ( mssg.src != world_server_mpi_rank ) ||
                 ( mssg.dest != world_mpi_rank ) ||
                 ( mssg.base_addr != data[world_mpi_rank].base_addr ) ||
                 ( mssg.len != data[world_mpi_rank].len ) ||
                 ( mssg.ver != data[world_mpi_rank].ver ) ||
                 ( mssg.magic != MSSG_MAGIC ) ) {

                success = FALSE;
                nerrors++;
                if ( verbose ) {
                    HDfprintf(stdout, "%d:%s: Bad data in read req reply.\n",
                              world_mpi_rank, FUNC);
                }
            }
        }

    /* barrier to allow all writes to complete */
        if ( MPI_SUCCESS != MPI_Barrier(file_mpi_comm) ) {

            success = FALSE;
            nerrors++;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: barrier 3 failed.\n",
                          world_mpi_rank, FUNC);
            }
        }

        /* verify that the expected entries have been read, and the total */
        if ( success ) {

            success = verify_entry_writes(data[world_mpi_rank].base_addr, 1);
        }

        if ( success ) {

            success = verify_entry_reads(data[world_mpi_rank].base_addr, 1);
        }

        if ( success ) {

            success = verify_total_writes((unsigned)(world_mpi_size - 1));
        }

        if ( success ) {

            success = verify_total_reads(world_mpi_size - 1);
        }

        if ( MPI_SUCCESS != MPI_Barrier(file_mpi_comm) ) {

            success = FALSE;
            nerrors++;
            if ( verbose ) {

                HDfprintf(stdout, "%d:%s: barrier 4 failed.\n",
                          world_mpi_rank, FUNC);
            }
        }

        /* reset the counters */
        if ( success ) {

            success = reset_server_counts();
        }

        if ( MPI_SUCCESS != MPI_Barrier(file_mpi_comm) ) {

            success = FALSE;
            nerrors++;
            if ( verbose ) {

                HDfprintf(stdout, "%d:%s: barrier 5 failed.\n",
                          world_mpi_rank, FUNC);
            }
        }

        /* verify that the counters have been reset */
        if ( success ) {

            success = verify_entry_writes(data[world_mpi_rank].base_addr, 0);
        }

        if ( success ) {

            success = verify_entry_reads(data[world_mpi_rank].base_addr, 0);
        }

        if ( success ) {

            success = verify_total_writes(0);
        }

        if ( success ) {

            success = verify_total_reads(0);
        }

        if ( MPI_SUCCESS != MPI_Barrier(file_mpi_comm) ) {

            success = FALSE;
            nerrors++;
            if ( verbose ) {

                HDfprintf(stdout, "%d:%s: barrier 6 failed.\n",
                          world_mpi_rank, FUNC);
            }
        }

        /* compose the done message */
        mssg.req       = DONE_REQ_CODE;
        mssg.src       = world_mpi_rank;
        mssg.dest      = world_server_mpi_rank;
        mssg.mssg_num  = -1; /* set by send function */
        mssg.base_addr = 0; /* not used */
        mssg.len       = 0; /* not used */
        mssg.ver       = 0; /* not used */
        mssg.count     = 0;
        mssg.magic     = MSSG_MAGIC;

        if ( success ) {

            success = send_mssg(&mssg, FALSE);

            if ( ! success ) {

                nerrors++;
                if ( verbose ) {
                    HDfprintf(stdout, "%d:%s: send_mssg() failed on done.\n",
                              world_mpi_rank, FUNC);
                }
            }
        }
    }

    max_nerrors = get_max_nerrors();

    if ( world_mpi_rank == 0 ) {

    if ( max_nerrors == 0 ) {

        PASSED();

        } else {

            failures++;
            H5_FAILED();
        }
    }

    success = ( ( success ) && ( max_nerrors == 0 ) );

    return(success);

} /* server_smoke_check() */


/*****************************************************************************
 *
 * Function:    smoke_check_1()
 *
 * Purpose:    First smoke check for the parallel cache.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 1/4/06
 *
 *****************************************************************************/
static hbool_t
smoke_check_1(int metadata_write_strategy)
{
    hbool_t success = TRUE;
    int i;
    int max_nerrors;
    hid_t fid = -1;
    H5F_t * file_ptr = NULL;
    H5C_t * cache_ptr = NULL;
    struct mssg_t mssg;

    switch ( metadata_write_strategy ) {

    case H5AC_METADATA_WRITE_STRATEGY__PROCESS_0_ONLY:
            if ( world_mpi_rank == 0 ) {
            TESTING("smoke check #1 -- process 0 only md write strategy");
            }
        break;

    case H5AC_METADATA_WRITE_STRATEGY__DISTRIBUTED:
            if ( world_mpi_rank == 0 ) {
            TESTING("smoke check #1 -- distributed md write strategy");
            }
        break;

        default:
            if ( world_mpi_rank == 0 ) {
            TESTING("smoke check #1 -- unknown md write strategy");
            }
        break;
    }

    nerrors = 0;
    init_data();
    reset_stats();

    if ( world_mpi_rank == world_server_mpi_rank ) {

    if ( ! server_main() ) {

            /* some error occured in the server -- report failure */
            nerrors++;
            if ( verbose ) {
        HDfprintf(stdout, "%d:%s: server_main() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }
    else /* run the clients */
    {
        if ( ! setup_cache_for_test(&fid, &file_ptr, &cache_ptr,
                                    metadata_write_strategy) ) {

            nerrors++;
            fid = -1;
            cache_ptr = NULL;
            if ( verbose ) {
        HDfprintf(stdout, "%d:%s: setup_cache_for_test() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }

        for ( i = 0; i < (virt_num_data_entries / 2); i++ )
        {
            insert_entry(cache_ptr, file_ptr, i, H5AC__NO_FLAGS_SET);
        }

        for ( i = (virt_num_data_entries / 2) - 1; i >= 0; i-- )
        {
        lock_entry(file_ptr, i);
        unlock_entry(file_ptr, i, H5AC__NO_FLAGS_SET);
        }

        /* Move the first half of the entries... */
        for ( i = 0; i < (virt_num_data_entries / 2); i++ )
        {
        lock_entry(file_ptr, i);
        unlock_entry(file_ptr, i, H5AC__NO_FLAGS_SET);
        move_entry(file_ptr, i, (i + (virt_num_data_entries / 2)));
        }

        /* ...and then move them back. */
        for ( i = (virt_num_data_entries / 2) - 1; i >= 0; i-- )
        {
        lock_entry(file_ptr, i);
        unlock_entry(file_ptr, i, H5AC__NO_FLAGS_SET);
        move_entry(file_ptr, i, (i + (virt_num_data_entries / 2)));
        }

        if ( fid >= 0 ) {

            if ( ! take_down_cache(fid, cache_ptr) ) {

                nerrors++;
                if ( verbose ) {
            HDfprintf(stdout, "%d:%s: take_down_cache() failed.\n",
                              world_mpi_rank, FUNC);
                }
            }
        }

        /* verify that all instance of datum are back where the started
         * and are clean.
         */

        for ( i = 0; i < NUM_DATA_ENTRIES; i++ )
        {
            HDassert( data_index[i] == i );
            HDassert( ! (data[i].dirty) );
        }

        /* compose the done message */
        mssg.req       = DONE_REQ_CODE;
        mssg.src       = world_mpi_rank;
        mssg.dest      = world_server_mpi_rank;
        mssg.mssg_num  = -1; /* set by send function */
        mssg.base_addr = 0; /* not used */
        mssg.len       = 0; /* not used */
        mssg.ver       = 0; /* not used */
        mssg.count     = 0; /* not used */
        mssg.magic     = MSSG_MAGIC;

        if ( success ) {

            success = send_mssg(&mssg, FALSE);

            if ( ! success ) {

                nerrors++;
                if ( verbose ) {
                    HDfprintf(stdout, "%d:%s: send_mssg() failed on done.\n",
                              world_mpi_rank, FUNC);
                }
            }
        }
    }

    max_nerrors = get_max_nerrors();

    if ( world_mpi_rank == 0 ) {

    if ( max_nerrors == 0 ) {

        PASSED();

        } else {

            failures++;
            H5_FAILED();
        }
    }

    success = ( ( success ) && ( max_nerrors == 0 ) );

    return(success);

} /* smoke_check_1() */


/*****************************************************************************
 *
 * Function:    smoke_check_2()
 *
 * Purpose:    Second smoke check for the parallel cache.
 *
 *        Introduce random reads, but keep all processes with roughly
 *        the same work load.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 1/12/06
 *
 *****************************************************************************/
static hbool_t
smoke_check_2(int metadata_write_strategy)
{
    hbool_t success = TRUE;
    int i;
    int max_nerrors;
    hid_t fid = -1;
    H5F_t * file_ptr = NULL;
    H5C_t * cache_ptr = NULL;
    struct mssg_t mssg;

    switch ( metadata_write_strategy ) {

    case H5AC_METADATA_WRITE_STRATEGY__PROCESS_0_ONLY:
            if ( world_mpi_rank == 0 ) {
            TESTING("smoke check #2 -- process 0 only md write strategy");
            }
        break;

    case H5AC_METADATA_WRITE_STRATEGY__DISTRIBUTED:
            if ( world_mpi_rank == 0 ) {
            TESTING("smoke check #2 -- distributed md write strategy");
            }
        break;

        default:
            if ( world_mpi_rank == 0 ) {
            TESTING("smoke check #2 -- unknown md write strategy");
            }
        break;
    }

    nerrors = 0;
    init_data();
    reset_stats();

    if ( world_mpi_rank == world_server_mpi_rank ) {

    if ( ! server_main() ) {

            /* some error occured in the server -- report failure */
            nerrors++;
            if ( verbose ) {
        HDfprintf(stdout, "%d:%s: server_main() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }
    else /* run the clients */
    {
        if ( ! setup_cache_for_test(&fid, &file_ptr, &cache_ptr,
                                    metadata_write_strategy) ) {

            nerrors++;
            fid = -1;
            cache_ptr = NULL;
            if ( verbose ) {
        HDfprintf(stdout, "%d:%s: setup_cache_for_test() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }

        for ( i = 0; i < (virt_num_data_entries / 2); i++ )
        {
            insert_entry(cache_ptr, file_ptr, i, H5AC__NO_FLAGS_SET);

            if ( i > 100 ) {

        lock_and_unlock_random_entries(file_ptr, (i - 100), i, 0, 10);
            }
        }

    for ( i = 0; i < (virt_num_data_entries / 2); i+=61 )
    {
        /* Make sure we don't step on any locally pinned entries */
        if ( data[i].local_pinned ) {
        unpin_entry(file_ptr, i, FALSE, FALSE, FALSE);
        }

        pin_entry(file_ptr, i, TRUE, FALSE);
    }

        for ( i = (virt_num_data_entries / 2) - 1; i >= 0; i-=2 )
        {
        lock_entry(file_ptr, i);
        unlock_entry(file_ptr, i, H5AC__NO_FLAGS_SET);
        lock_and_unlock_random_entries(file_ptr, 0,
                            (virt_num_data_entries / 20),
                    0, 100);
        local_pin_and_unpin_random_entries(file_ptr, 0,
                                (virt_num_data_entries / 4),
                        0, 3);
        }

        for ( i = 0; i < (virt_num_data_entries / 2); i+=2 )
        {
        lock_entry(file_ptr, i);
        unlock_entry(file_ptr, i, H5AC__DIRTIED_FLAG);
        lock_and_unlock_random_entries(file_ptr, 0,
                            (virt_num_data_entries / 10),
                    0, 100);
        }

    /* we can't move pinned entries, so release any local pins now. */
    local_unpin_all_entries(file_ptr, FALSE);

        /* Move the first half of the entries... */
        for ( i = 0; i < (virt_num_data_entries / 2); i++ )
        {
        lock_entry(file_ptr, i);
        unlock_entry(file_ptr, i, H5AC__NO_FLAGS_SET);
        move_entry(file_ptr, i, (i + (virt_num_data_entries / 2)));
        lock_and_unlock_random_entries(file_ptr, 0,
                            ((virt_num_data_entries / 50) - 1),
                                           0, 100);
        }

        /* ...and then move them back. */
        for ( i = (virt_num_data_entries / 2) - 1; i >= 0; i-- )
        {
        lock_entry(file_ptr, i);
        unlock_entry(file_ptr, i, H5AC__DIRTIED_FLAG);
        move_entry(file_ptr, i, (i + (virt_num_data_entries / 2)));
        lock_and_unlock_random_entries(file_ptr, 0,
                            (virt_num_data_entries / 100),
                    0, 100);
        }

    for ( i = 0; i < (virt_num_data_entries / 2); i+=61 )
    {
        hbool_t via_unprotect = ( (((unsigned)i) & 0x01) == 0 );
        hbool_t dirty = ( (((unsigned)i) & 0x02) == 0 );

        unpin_entry(file_ptr, i, TRUE, dirty, via_unprotect);
    }

        if ( fid >= 0 ) {

            if ( ! take_down_cache(fid, cache_ptr) ) {

                nerrors++;
                if ( verbose ) {
            HDfprintf(stdout, "%d:%s: take_down_cache() failed.\n",
                              world_mpi_rank, FUNC);
                }
            }
        }

        /* verify that all instance of datum are back where the started
         * and are clean.
         */

        for ( i = 0; i < NUM_DATA_ENTRIES; i++ )
        {
            HDassert( data_index[i] == i );
            HDassert( ! (data[i].dirty) );
        }

        /* compose the done message */
        mssg.req       = DONE_REQ_CODE;
        mssg.src       = world_mpi_rank;
        mssg.dest      = world_server_mpi_rank;
        mssg.mssg_num  = -1; /* set by send function */
        mssg.base_addr = 0; /* not used */
        mssg.len       = 0; /* not used */
        mssg.ver       = 0; /* not used */
        mssg.count     = 0; /* not used */
        mssg.magic     = MSSG_MAGIC;

        if ( success ) {

            success = send_mssg(&mssg, FALSE);

            if ( ! success ) {

                nerrors++;
                if ( verbose ) {
                    HDfprintf(stdout, "%d:%s: send_mssg() failed on done.\n",
                              world_mpi_rank, FUNC);
                }
            }
        }
    }

    max_nerrors = get_max_nerrors();

    if ( world_mpi_rank == 0 ) {

    if ( max_nerrors == 0 ) {

        PASSED();

        } else {

            failures++;
            H5_FAILED();
        }
    }

    success = ( ( success ) && ( max_nerrors == 0 ) );

    return(success);

} /* smoke_check_2() */


/*****************************************************************************
 *
 * Function:    smoke_check_3()
 *
 * Purpose:    Third smoke check for the parallel cache.
 *
 *        Use random reads to vary the loads on the diffferent
 *        processors.  Also force different cache size adjustments.
 *
 *        In this test, load process 0 heavily, and the other
 *        processes lightly.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 1/13/06
 *
 *****************************************************************************/
static hbool_t
smoke_check_3(int metadata_write_strategy)
{
    hbool_t success = TRUE;
    int i;
    int max_nerrors;
    int min_count;
    int max_count;
    int min_idx;
    int max_idx;
    hid_t fid = -1;
    H5F_t * file_ptr = NULL;
    H5C_t * cache_ptr = NULL;
    struct mssg_t mssg;

    switch ( metadata_write_strategy ) {

    case H5AC_METADATA_WRITE_STRATEGY__PROCESS_0_ONLY:
            if ( world_mpi_rank == 0 ) {
            TESTING("smoke check #3 -- process 0 only md write strategy");
            }
        break;

    case H5AC_METADATA_WRITE_STRATEGY__DISTRIBUTED:
            if ( world_mpi_rank == 0 ) {
            TESTING("smoke check #3 -- distributed md write strategy");
            }
        break;

        default:
            if ( world_mpi_rank == 0 ) {
            TESTING("smoke check #3 -- unknown md write strategy");
            }
        break;
    }

    nerrors = 0;
    init_data();
    reset_stats();

    if ( world_mpi_rank == world_server_mpi_rank ) {

    if ( ! server_main() ) {

            /* some error occured in the server -- report failure */
            nerrors++;
            if ( verbose ) {
        HDfprintf(stdout, "%d:%s: server_main() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }
    else /* run the clients */
    {
        if ( ! setup_cache_for_test(&fid, &file_ptr, &cache_ptr,
                                    metadata_write_strategy) ) {

            nerrors++;
            fid = -1;
            cache_ptr = NULL;
            if ( verbose ) {
        HDfprintf(stdout, "%d:%s: setup_cache_for_test() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }

        min_count = 100 / ((file_mpi_rank + 1) * (file_mpi_rank + 1));
        max_count = min_count + 50;

        for ( i = 0; i < (virt_num_data_entries / 4); i++ )
        {
            insert_entry(cache_ptr, file_ptr, i, H5AC__NO_FLAGS_SET);

            if ( i > 100 ) {

        lock_and_unlock_random_entries(file_ptr, (i - 100), i,
                                               min_count, max_count);
            }
        }


        min_count = 100 / ((file_mpi_rank + 2) * (file_mpi_rank + 2));
        max_count = min_count + 50;

        for ( i = (virt_num_data_entries / 4);
        i < (virt_num_data_entries / 2);
        i++ )
        {

            insert_entry(cache_ptr, file_ptr, i, H5AC__NO_FLAGS_SET);

        if ( i % 59 == 0 ) {

        hbool_t dirty = ( (i % 2) == 0);

        if ( data[i].local_pinned ) {
            unpin_entry(file_ptr, i, FALSE, FALSE, FALSE);
        }

        pin_entry(file_ptr, i, TRUE, dirty);

            HDassert( !dirty || data[i].header.is_dirty );
            HDassert( data[i].header.is_pinned );
            HDassert( data[i].global_pinned );
            HDassert( ! data[i].local_pinned );
        }

            if ( i > 100 ) {

        lock_and_unlock_random_entries(file_ptr, (i - 100), i,
                                               min_count, max_count);
            }

        local_pin_and_unpin_random_entries(file_ptr, 0,
                                               virt_num_data_entries / 4,
                        0, (file_mpi_rank + 2));

    }


    /* flush the file to be sure that we have no problems flushing
    * pinned entries
    */
        if ( H5Fflush(fid, H5F_SCOPE_GLOBAL) < 0 ) {
            nerrors++;
            if ( verbose ) {
            HDfprintf(stdout, "%d:%s: H5Fflush() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }


        min_idx = 0;
        max_idx = ((virt_num_data_entries / 10) /
                   ((file_mpi_rank + 1) * (file_mpi_rank + 1))) - 1;
        if ( max_idx <= min_idx ) {

            max_idx = min_idx + 10;
        }

        for ( i = (virt_num_data_entries / 2) - 1; i >= 0; i-- )
        {
        if ( ( i >= (virt_num_data_entries / 4) ) && ( i % 59 == 0 ) ) {

                hbool_t via_unprotect = ( (((unsigned)i) & 0x02) == 0 );
            hbool_t dirty = ( (((unsigned)i) & 0x04) == 0 );

        HDassert( data[i].global_pinned );
        HDassert( ! data[i].local_pinned );

        unpin_entry(file_ptr, i, TRUE, dirty,
                via_unprotect);
        }
        if ( i % 2 == 0 ) {

            lock_entry(file_ptr, i);
            unlock_entry(file_ptr, i, H5AC__NO_FLAGS_SET);
            local_pin_and_unpin_random_entries(file_ptr, 0,
                                virt_num_data_entries / 2,
                        0, 2);
            lock_and_unlock_random_entries(file_ptr,
                                               min_idx, max_idx, 0, 100);
        }
        }

        min_idx = 0;
        max_idx = ((virt_num_data_entries / 10) /
                   ((file_mpi_rank + 3) * (file_mpi_rank + 3))) - 1;
        if ( max_idx <= min_idx ) {

            max_idx = min_idx + 10;
        }

        for ( i = 0; i < (virt_num_data_entries / 2); i+=2 )
        {
        lock_entry(file_ptr, i);
        unlock_entry(file_ptr, i, H5AC__DIRTIED_FLAG);
        lock_and_unlock_random_entries(file_ptr,
                                           min_idx, max_idx, 0, 100);
        }

        /* we can't move pinned entries, so release any local pins now. */
        local_unpin_all_entries(file_ptr, FALSE);

        min_count = 10 / (file_mpi_rank + 1);
        max_count = min_count + 100;

        /* move the first half of the entries... */
        for ( i = 0; i < (virt_num_data_entries / 2); i++ )
        {
        lock_entry(file_ptr, i);
        unlock_entry(file_ptr, i, H5AC__NO_FLAGS_SET);
        move_entry(file_ptr, i, (i + (virt_num_data_entries / 2)));
        lock_and_unlock_random_entries(file_ptr, 0,
                            (virt_num_data_entries / 20),
                                           min_count, max_count);
        }

        /* ...and then move them back. */
        for ( i = (virt_num_data_entries / 2) - 1; i >= 0; i-- )
        {
        lock_entry(file_ptr, i);
        unlock_entry(file_ptr, i, H5AC__DIRTIED_FLAG);
        move_entry(file_ptr, i, (i + (virt_num_data_entries / 2)));
        lock_and_unlock_random_entries(file_ptr, 0,
                            (virt_num_data_entries / 40),
                                           min_count, max_count);
        }

        /* finally, do some dirty lock/unlocks while we give the cache
         * a chance t reduce its size.
         */
        min_count = 200 / ((file_mpi_rank + 1) * (file_mpi_rank + 1));
        max_count = min_count + 100;

        for ( i = 0; i < (virt_num_data_entries / 2); i+=2 )
        {
        local_pin_and_unpin_random_entries(file_ptr, 0,
                                (virt_num_data_entries / 2),
                        0, 5);

        lock_entry(file_ptr, i);
        unlock_entry(file_ptr, i, H5AC__DIRTIED_FLAG);

            if ( i > 100 ) {

        lock_and_unlock_random_entries(file_ptr, (i - 100), i,
                                               min_count, max_count);
            }
        }

        /* release any local pins before we take down the cache. */
        local_unpin_all_entries(file_ptr, FALSE);

        if ( fid >= 0 ) {

            if ( ! take_down_cache(fid, cache_ptr) ) {

                nerrors++;
                if ( verbose ) {
            HDfprintf(stdout, "%d:%s: take_down_cache() failed.\n",
                              world_mpi_rank, FUNC);
                }
            }
        }

        /* verify that all instances of datum are back where the started
         * and are clean.
         */

        for ( i = 0; i < NUM_DATA_ENTRIES; i++ )
        {
            HDassert( data_index[i] == i );
            HDassert( ! (data[i].dirty) );
        }

        /* compose the done message */
        mssg.req       = DONE_REQ_CODE;
        mssg.src       = world_mpi_rank;
        mssg.dest      = world_server_mpi_rank;
        mssg.mssg_num  = -1; /* set by send function */
        mssg.base_addr = 0; /* not used */
        mssg.len       = 0; /* not used */
        mssg.ver       = 0; /* not used */
        mssg.count     = 0; /* not used */
        mssg.magic     = MSSG_MAGIC;

        if ( success ) {


            success = send_mssg(&mssg, FALSE);

            if ( ! success ) {

                nerrors++;
                if ( verbose ) {
                    HDfprintf(stdout, "%d:%s: send_mssg() failed on done.\n",
                              world_mpi_rank, FUNC);
                }
            }
        }
    }

    max_nerrors = get_max_nerrors();

    if ( world_mpi_rank == 0 ) {

    if ( max_nerrors == 0 ) {

        PASSED();

        } else {

            failures++;
            H5_FAILED();
        }
    }

    success = ( ( success ) && ( max_nerrors == 0 ) );

    return(success);

} /* smoke_check_3() */


/*****************************************************************************
 *
 * Function:    smoke_check_4()
 *
 * Purpose:    Fourth smoke check for the parallel cache.
 *
 *        Use random reads to vary the loads on the diffferent
 *        processors.  Also force different cache size adjustments.
 *
 *        In this test, load process 0 lightly, and the other
 *        processes heavily.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 1/13/06
 *
 *****************************************************************************/
static hbool_t
smoke_check_4(int metadata_write_strategy)
{
    hbool_t success = TRUE;
    int i;
    int max_nerrors;
    int min_count;
    int max_count;
    int min_idx;
    int max_idx;
    hid_t fid = -1;
    H5F_t * file_ptr = NULL;
    H5C_t * cache_ptr = NULL;
    struct mssg_t mssg;

    switch ( metadata_write_strategy ) {

    case H5AC_METADATA_WRITE_STRATEGY__PROCESS_0_ONLY:
            if ( world_mpi_rank == 0 ) {
            TESTING("smoke check #4 -- process 0 only md write strategy");
            }
        break;

    case H5AC_METADATA_WRITE_STRATEGY__DISTRIBUTED:
            if ( world_mpi_rank == 0 ) {
            TESTING("smoke check #4 -- distributed md write strategy");
            }
        break;

        default:
            if ( world_mpi_rank == 0 ) {
            TESTING("smoke check #4 -- unknown md write strategy");
            }
        break;
    }

    nerrors = 0;
    init_data();
    reset_stats();

    if ( world_mpi_rank == world_server_mpi_rank ) {

    if ( ! server_main() ) {

            /* some error occured in the server -- report failure */
            nerrors++;
            if ( verbose ) {
        HDfprintf(stdout, "%d:%s: server_main() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }
    else /* run the clients */
    {
        if ( ! setup_cache_for_test(&fid, &file_ptr, &cache_ptr,
                                    metadata_write_strategy) ) {

            nerrors++;
            fid = -1;
            cache_ptr = NULL;
            if ( verbose ) {
        HDfprintf(stdout, "%d:%s: setup_cache_for_test() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }


        min_count = 100 * (file_mpi_rank % 4);
        max_count = min_count + 50;

        for ( i = 0; i < (virt_num_data_entries / 4); i++ )
        {
            insert_entry(cache_ptr, file_ptr, i, H5AC__NO_FLAGS_SET);

            if ( i > 100 ) {

        lock_and_unlock_random_entries(file_ptr, (i - 100), i,
                                               min_count, max_count);
            }
        }

        min_count = 10 * (file_mpi_rank % 4);
        max_count = min_count + 100;

        for ( i = (virt_num_data_entries / 4);
        i < (virt_num_data_entries / 2);
        i++ )
        {
        if ( i % 2 == 0 ) {

                insert_entry(cache_ptr, file_ptr, i, H5AC__NO_FLAGS_SET);

        } else {

        /* Insert some entries pinned, and then unpin them
        * immediately.  We have tested pinned entries elsewhere,
        * so it should be sufficient to verify that the
        * entries are in fact pinned (which unpin_entry() should do).
        */
                insert_entry(cache_ptr, file_ptr, i, H5C__PIN_ENTRY_FLAG);
                unpin_entry(file_ptr, i, TRUE, FALSE, FALSE);
        }

            if ( i % 59 == 0 ) {

                hbool_t dirty = ( (i % 2) == 0);

                if ( data[i].local_pinned ) {
                    unpin_entry(file_ptr, i, FALSE, FALSE, FALSE);
                }

                pin_entry(file_ptr, i, TRUE, dirty);

                HDassert( !dirty || data[i].header.is_dirty );
                HDassert( data[i].header.is_pinned );
                HDassert( data[i].global_pinned );
                HDassert( ! data[i].local_pinned );
            }

            if ( i > 100 ) {

        lock_and_unlock_random_entries(file_ptr, (i - 100), i,
                                               min_count, max_count);
            }

            local_pin_and_unpin_random_entries(file_ptr, 0,
                                (virt_num_data_entries / 4),
                                               0, (file_mpi_rank + 2));
        }


        /* flush the file to be sure that we have no problems flushing
    * pinned entries
    */
        if ( H5Fflush(fid, H5F_SCOPE_GLOBAL) < 0 ) {
            nerrors++;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: H5Fflush() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }


        min_idx = 0;
        max_idx = (((virt_num_data_entries / 10) / 4) *
                   ((file_mpi_rank % 4) + 1)) - 1;

        for ( i = (virt_num_data_entries / 2) - 1; i >= 0; i-- )
        {
            if ( ( i >= (virt_num_data_entries / 4) ) && ( i % 59 == 0 ) ) {

                hbool_t via_unprotect = ( (((unsigned)i) & 0x02) == 0 );
                hbool_t dirty = ( (((unsigned)i) & 0x04) == 0 );

                HDassert( data[i].global_pinned );
                HDassert( ! data[i].local_pinned );

                unpin_entry(file_ptr, i, TRUE, dirty, via_unprotect);
            }

        if ( i % 2 == 0 ) {

            lock_entry(file_ptr, i);
            unlock_entry(file_ptr, i, H5AC__NO_FLAGS_SET);
            lock_and_unlock_random_entries(file_ptr,
                                               min_idx, max_idx, 0, 100);
        }
        }

        min_idx = 0;
        max_idx = (((virt_num_data_entries / 10) / 8) *
                   ((file_mpi_rank % 4) + 1)) - 1;

        for ( i = 0; i < (virt_num_data_entries / 2); i+=2 )
        {
        lock_entry(file_ptr, i);
        unlock_entry(file_ptr, i, H5AC__DIRTIED_FLAG);
        lock_and_unlock_random_entries(file_ptr,
                                           min_idx, max_idx, 0, 100);
        }

    /* we can't move pinned entries, so release any local pins now. */
    local_unpin_all_entries(file_ptr, FALSE);

        min_count = 10 * (file_mpi_rank % 4);
        max_count = min_count + 100;

        /* move the first half of the entries... */
        for ( i = 0; i < (virt_num_data_entries / 2); i++ )
        {
        lock_entry(file_ptr, i);
        unlock_entry(file_ptr, i, H5AC__NO_FLAGS_SET);
        move_entry(file_ptr, i, (i + (virt_num_data_entries / 2)));
        lock_and_unlock_random_entries(file_ptr, 0,
                            (virt_num_data_entries / 20),
                                           min_count, max_count);
        }

        /* ...and then move them back. */
        for ( i = (virt_num_data_entries / 2) - 1; i >= 0; i-- )
        {
        lock_entry(file_ptr, i);
        unlock_entry(file_ptr, i, H5AC__DIRTIED_FLAG);
        move_entry(file_ptr, i, (i + (virt_num_data_entries / 2)));
        lock_and_unlock_random_entries(file_ptr, 0,
                            (virt_num_data_entries / 40),
                                           min_count, max_count);
        }

        /* finally, do some dirty lock/unlocks while we give the cache
         * a chance t reduce its size.
         */
        min_count = 100 * (file_mpi_rank % 4);
        max_count = min_count + 100;

        for ( i = 0; i < (virt_num_data_entries / 2); i+=2 )
        {
        lock_entry(file_ptr, i);
        unlock_entry(file_ptr, i, H5AC__DIRTIED_FLAG);

            if ( i > 100 ) {

        lock_and_unlock_random_entries(file_ptr, (i - 100), i,
                                               min_count, max_count);
            }
        }

        if ( fid >= 0 ) {

            if ( ! take_down_cache(fid, cache_ptr) ) {

                nerrors++;
                if ( verbose ) {
            HDfprintf(stdout, "%d:%s: take_down_cache() failed.\n",
                              world_mpi_rank, FUNC);
                }
            }
        }

        /* verify that all instance of datum are back where the started
         * and are clean.
         */

        for ( i = 0; i < NUM_DATA_ENTRIES; i++ )
        {
            HDassert( data_index[i] == i );
            HDassert( ! (data[i].dirty) );
        }

        /* compose the done message */
        mssg.req       = DONE_REQ_CODE;
        mssg.src       = world_mpi_rank;
        mssg.dest      = world_server_mpi_rank;
        mssg.mssg_num  = -1; /* set by send function */
        mssg.base_addr = 0; /* not used */
        mssg.len       = 0; /* not used */
        mssg.ver       = 0; /* not used */
        mssg.count     = 0; /* not used */
        mssg.magic     = MSSG_MAGIC;

        if ( success ) {


            success = send_mssg(&mssg, FALSE);

            if ( ! success ) {

                nerrors++;
                if ( verbose ) {
                    HDfprintf(stdout, "%d:%s: send_mssg() failed on done.\n",
                              world_mpi_rank, FUNC);
                }
            }
        }
    }

    max_nerrors = get_max_nerrors();

    if ( world_mpi_rank == 0 ) {

    if ( max_nerrors == 0 ) {

        PASSED();

        } else {

            failures++;
            H5_FAILED();
        }
    }

    success = ( ( success ) && ( max_nerrors == 0 ) );

    return(success);

} /* smoke_check_4() */


/*****************************************************************************
 *
 * Function:    smoke_check_5()
 *
 * Purpose:    Similar to smoke check 1, but modified to verify that
 *         H5AC_mark_entry_dirty() works in the parallel case.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 5/18/06
 *
 *****************************************************************************/
static hbool_t
smoke_check_5(int metadata_write_strategy)
{
    hbool_t success = TRUE;
    int i;
    int max_nerrors;
    hid_t fid = -1;
    H5F_t * file_ptr = NULL;
    H5C_t * cache_ptr = NULL;
    struct mssg_t mssg;

    switch ( metadata_write_strategy ) {

    case H5AC_METADATA_WRITE_STRATEGY__PROCESS_0_ONLY:
            if ( world_mpi_rank == 0 ) {
            TESTING("smoke check #5 -- process 0 only md write strategy");
            }
        break;

    case H5AC_METADATA_WRITE_STRATEGY__DISTRIBUTED:
            if ( world_mpi_rank == 0 ) {
            TESTING("smoke check #5 -- distributed md write strategy");
            }
        break;

        default:
            if ( world_mpi_rank == 0 ) {
            TESTING("smoke check #5 -- unknown md write strategy");
            }
        break;
    }


    nerrors = 0;
    init_data();
    reset_stats();

    if ( world_mpi_rank == world_server_mpi_rank ) {

    if ( ! server_main() ) {

            /* some error occured in the server -- report failure */
            nerrors++;
            if ( verbose ) {
        HDfprintf(stdout, "%d:%s: server_main() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }
    else /* run the clients */
    {

        if ( ! setup_cache_for_test(&fid, &file_ptr, &cache_ptr,
                                    metadata_write_strategy) ) {

            nerrors++;
            fid = -1;
            cache_ptr = NULL;
            if ( verbose ) {
        HDfprintf(stdout, "%d:%s: setup_cache_for_test() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }

        for ( i = 0; i < (virt_num_data_entries / 2); i++ )
        {
            insert_entry(cache_ptr, file_ptr, i, H5AC__NO_FLAGS_SET);
        }

    /* flush the file so we can lock known clean entries. */
        if ( H5Fflush(fid, H5F_SCOPE_GLOBAL) < 0 ) {
            nerrors++;
            if ( verbose ) {
            HDfprintf(stdout, "%d:%s: H5Fflush() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }

        for ( i = 0; i < (virt_num_data_entries / 4); i++ )
        {
        lock_entry(file_ptr, i);

        if ( i % 2 == 0 )
        {
        mark_entry_dirty(i);
        }

        unlock_entry(file_ptr, i, H5AC__NO_FLAGS_SET);

        if ( i % 2 == 1 )
        {
        if ( i % 4 == 1 ) {

                lock_entry(file_ptr, i);
                unlock_entry(file_ptr, i, H5AC__DIRTIED_FLAG);
        }

            expunge_entry(file_ptr, i);
        }
        }

        for ( i = (virt_num_data_entries / 2) - 1;
              i >= (virt_num_data_entries / 4);
        i-- )
        {
        pin_entry(file_ptr, i, TRUE, FALSE);

        if ( i % 2 == 0 )
        {
        if ( i % 8 <= 4 ) {

            resize_entry(i, data[i].len / 2);
        }

                mark_entry_dirty(i);

        if ( i % 8 <= 4 ) {

            resize_entry(i, data[i].len);
        }
        }

        unpin_entry(file_ptr, i, TRUE, FALSE, FALSE);
        }

        if ( fid >= 0 ) {

            if ( ! take_down_cache(fid, cache_ptr) ) {

                nerrors++;
                if ( verbose ) {
            HDfprintf(stdout, "%d:%s: take_down_cache() failed.\n",
                              world_mpi_rank, FUNC);
                }
            }
        }

        /* verify that all instance of datum are back where the started
         * and are clean.
         */

        for ( i = 0; i < NUM_DATA_ENTRIES; i++ )
        {
            HDassert( data_index[i] == i );
            HDassert( ! (data[i].dirty) );
        }

        /* compose the done message */
        mssg.req       = DONE_REQ_CODE;
        mssg.src       = world_mpi_rank;
        mssg.dest      = world_server_mpi_rank;
        mssg.mssg_num  = -1; /* set by send function */
        mssg.base_addr = 0; /* not used */
        mssg.len       = 0; /* not used */
        mssg.ver       = 0; /* not used */
        mssg.count     = 0; /* not used */
        mssg.magic     = MSSG_MAGIC;

        if ( success ) {

            success = send_mssg(&mssg, FALSE);

            if ( ! success ) {

                nerrors++;
                if ( verbose ) {
                    HDfprintf(stdout, "%d:%s: send_mssg() failed on done.\n",
                              world_mpi_rank, FUNC);
                }
            }
        }
    }

    max_nerrors = get_max_nerrors();

    if ( world_mpi_rank == 0 ) {

    if ( max_nerrors == 0 ) {

        PASSED();

        } else {

            failures++;
            H5_FAILED();
        }
    }

    success = ( ( success ) && ( max_nerrors == 0 ) );

    return(success);

} /* smoke_check_5() */


/*****************************************************************************
 *
 * Function:    trace_file_check()
 *
 * Purpose:     A basic test of the trace file capability.  In essence,
 *              we invoke all operations that generate trace file output,
 *              and then verify that the expected output was generated.
 *
 *              Note that the trace file is currently implemented at the
 *              H5AC level, so all calls have to go through H5AC.  Thus it
 *              is more convenient to test trace file capabilities in the
 *              parallel cache test which works at the H5AC level, instead
 *              of in the serial test code which does everything at the
 *              H5C level.
 *
 *              The function must test trace file output in the following
 *              functions:
 *
 *                    - H5AC_flush()
 *                    - H5AC_insert_entry()
 *                    - H5AC_mark_entry_dirty()
 *                    - H5AC_move_entry()
 *                    - H5AC_pin_protected_entry()
 *                    - H5AC_protect()
 *                    - H5AC_unpin_entry()
 *                    - H5AC_unprotect()
 *                    - H5AC_set_cache_auto_resize_config()
 *                    - H5AC_expunge_entry()
 *                    - H5AC_resize_entry()
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 6/13/06
 *
 *****************************************************************************/
static hbool_t
trace_file_check(int metadata_write_strategy)
{
    hbool_t success = TRUE;

    const char *((* expected_output)[]) = NULL;
    const char * expected_output_0[] =
    {
      "### HDF5 metadata cache trace file version 1 ###\n",
      "H5AC_set_cache_auto_resize_config",
      "H5AC_insert_entry",
      "H5AC_insert_entry",
      "H5AC_insert_entry",
      "H5AC_insert_entry",
      "H5AC_protect",
      "H5AC_mark_entry_dirty",
      "H5AC_unprotect",
      "H5AC_protect",
      "H5AC_pin_protected_entry",
      "H5AC_unprotect",
      "H5AC_unpin_entry",
      "H5AC_expunge_entry",
      "H5AC_protect",
      "H5AC_pin_protected_entry",
      "H5AC_unprotect",
      "H5AC_mark_entry_dirty",
      "H5AC_resize_entry",
      "H5AC_resize_entry",
      "H5AC_unpin_entry",
      "H5AC_move_entry",
      "H5AC_move_entry",
      "H5AC_flush",
      "H5AC_flush",
      NULL
    };
    const char * expected_output_1[] =
    {
      "### HDF5 metadata cache trace file version 1 ###\n",
      "H5AC_set_cache_auto_resize_config",
      "H5AC_insert_entry",
      "H5AC_insert_entry",
      "H5AC_insert_entry",
      "H5AC_insert_entry",
      "H5AC_protect",
      "H5AC_mark_entry_dirty",
      "H5AC_unprotect",
      "H5AC_protect",
      "H5AC_pin_protected_entry",
      "H5AC_unprotect",
      "H5AC_unpin_entry",
      "H5AC_expunge_entry",
      "H5AC_protect",
      "H5AC_pin_protected_entry",
      "H5AC_unprotect",
      "H5AC_mark_entry_dirty",
      "H5AC_resize_entry",
      "H5AC_resize_entry",
      "H5AC_unpin_entry",
      "H5AC_move_entry",
      "H5AC_move_entry",
      "H5AC_flush",
      "H5AC_flush",
      NULL
    };
    char buffer[256];
    char trace_file_name[64];
    hbool_t done = FALSE;
    int i;
    int max_nerrors;
    size_t expected_line_len;
    size_t actual_line_len;
    hid_t fid = -1;
    H5F_t * file_ptr = NULL;
    H5C_t * cache_ptr = NULL;
    FILE * trace_file_ptr = NULL;
    H5AC_cache_config_t config;
    struct mssg_t mssg;


    switch(metadata_write_strategy) {

        case H5AC_METADATA_WRITE_STRATEGY__PROCESS_0_ONLY:

            expected_output = &expected_output_0;

            if(world_mpi_rank == 0)
                TESTING("trace file collection -- process 0 only md write strategy");
            break;

        case H5AC_METADATA_WRITE_STRATEGY__DISTRIBUTED:

            expected_output = &expected_output_1;

            if(world_mpi_rank == 0)
                TESTING("trace file collection -- distributed md write strategy");
            break;

        default:

            /* this will almost certainly cause a failure, but it keeps us
             * from de-referenceing a NULL pointer.
             */
            expected_output = &expected_output_0;

            if(world_mpi_rank == 0)
                TESTING("trace file collection -- unknown md write strategy");
            break;
    } /* end switch */


    nerrors = 0;
    init_data();
    reset_stats();

    if(world_mpi_rank == world_server_mpi_rank) {

        if(!server_main()) {

            /* some error occured in the server -- report failure */
            nerrors++;
            if ( verbose )
                HDfprintf(stdout, "%d:%s: server_main() failed.\n", world_mpi_rank, FUNC);
        }
    }
    else {
        /* run the clients */

        if(!setup_cache_for_test(&fid, &file_ptr, &cache_ptr, metadata_write_strategy) ) {

            nerrors++;
            fid = -1;
            cache_ptr = NULL;
            if(verbose)
                HDfprintf(stdout, "%d:%s: setup_cache_for_test() failed.\n", world_mpi_rank, FUNC);
        }

        if(nerrors == 0) {

            config.version = H5AC__CURR_CACHE_CONFIG_VERSION;

            if(H5AC_get_cache_auto_resize_config(cache_ptr, &config) != SUCCEED) {
                nerrors++;
                HDfprintf(stdout, "%d:%s: H5AC_get_cache_auto_resize_config() failed.\n", world_mpi_rank, FUNC);
            }
            else {
                config.open_trace_file = TRUE;
                strcpy(config.trace_file_name, "t_cache_trace.txt");

                if(H5AC_set_cache_auto_resize_config(cache_ptr, &config) != SUCCEED) {
                    nerrors++;
                    HDfprintf(stdout, "%d:%s: H5AC_set_cache_auto_resize_config() failed.\n", world_mpi_rank, FUNC);
                }
            }
        } /* end if */

        insert_entry(cache_ptr, file_ptr, 0, H5AC__NO_FLAGS_SET);
        insert_entry(cache_ptr, file_ptr, 1, H5AC__NO_FLAGS_SET);
        insert_entry(cache_ptr, file_ptr, 2, H5AC__NO_FLAGS_SET);
        insert_entry(cache_ptr, file_ptr, 3, H5AC__NO_FLAGS_SET);

        lock_entry(file_ptr, 0);
        mark_entry_dirty(0);
        unlock_entry(file_ptr, 0, H5AC__NO_FLAGS_SET);

        lock_entry(file_ptr, 1);
        pin_protected_entry(1, TRUE);
        unlock_entry(file_ptr, 1, H5AC__NO_FLAGS_SET);
        unpin_entry(file_ptr, 1, TRUE, FALSE, FALSE);

        expunge_entry(file_ptr, 1);

        lock_entry(file_ptr, 2);
        pin_protected_entry(2, TRUE);
        unlock_entry(file_ptr, 2, H5AC__NO_FLAGS_SET);
        mark_entry_dirty(2);
        resize_entry(2, data[2].len / 2);
        resize_entry(2, data[2].len);
        unpin_entry(file_ptr, 2, TRUE, FALSE, FALSE);

        move_entry(file_ptr, 0, 20);
        move_entry(file_ptr, 0, 20);

        if(H5Fflush(fid, H5F_SCOPE_GLOBAL) < 0) {
            nerrors++;
            if(verbose)
                HDfprintf(stdout, "%d:%s: H5Fflush() failed.\n", world_mpi_rank, FUNC);
        }

        if(nerrors == 0) {
            config.version = H5AC__CURR_CACHE_CONFIG_VERSION;

            if(H5AC_get_cache_auto_resize_config(cache_ptr, &config) != SUCCEED) {
                nerrors++;
                HDfprintf(stdout, "%d:%s: H5AC_get_cache_auto_resize_config() failed.\n", world_mpi_rank, FUNC);
            }
            else {
                config.open_trace_file = FALSE;
                config.close_trace_file = TRUE;
                config.trace_file_name[0] = '\0';

                if(H5AC_set_cache_auto_resize_config(cache_ptr, &config) != SUCCEED) {
                    nerrors++;
                    HDfprintf(stdout, "%d:%s: H5AC_set_cache_auto_resize_config() failed.\n", world_mpi_rank, FUNC);
                }
            }
        } /* end if */

        if(fid >= 0) {
            if(!take_down_cache(fid, cache_ptr)) {

                nerrors++;
                if(verbose)
                    HDfprintf(stdout, "%d:%s: take_down_cache() failed.\n", world_mpi_rank, FUNC);
            }
        } /* end if */

        /* verify that all instance of datum are back where the started
         * and are clean.
         */

        for(i = 0; i < NUM_DATA_ENTRIES; i++) {
            HDassert(data_index[i] == i);
            HDassert(!(data[i].dirty));
        }

        /* compose the done message */
        mssg.req       = DONE_REQ_CODE;
        mssg.src       = world_mpi_rank;
        mssg.dest      = world_server_mpi_rank;
        mssg.mssg_num  = -1; /* set by send function */
        mssg.base_addr = 0; /* not used */
        mssg.len       = 0; /* not used */
        mssg.ver       = 0; /* not used */
        mssg.count     = 0; /* not used */
        mssg.magic     = MSSG_MAGIC;

        if(success) {
            success = send_mssg(&mssg, FALSE);

            if(!success) {
                nerrors++;
                if(verbose)
                    HDfprintf(stdout, "%d:%s: send_mssg() failed on done.\n", world_mpi_rank, FUNC);
            }
        } /* end if */

        if(nerrors == 0) {
            HDsprintf(trace_file_name, "t_cache_trace.txt.%d", (int)file_mpi_rank);

            if((trace_file_ptr = HDfopen(trace_file_name, "r")) == NULL ) {

                nerrors++;
                if(verbose)
                    HDfprintf(stdout, "%d:%s: HDfopen failed.\n", world_mpi_rank, FUNC);
            }
        } /* end if */


        i = 0;
        while((nerrors == 0) && (!done)) {
            /* Get lines of actual and expected data */
            if((*expected_output)[i] == NULL)
                expected_line_len = (size_t)0;
            else
                expected_line_len = HDstrlen((*expected_output)[i]);

            if(HDfgets(buffer, 255, trace_file_ptr) != NULL)
                actual_line_len = HDstrlen(buffer);
            else
                actual_line_len = (size_t)0;

            /* Compare the lines */
            /* Handle running out of data */
            if((actual_line_len == 0) || (expected_line_len == 0)) {
                if((actual_line_len == 0) && (expected_line_len == 0)) {
                    /* Both ran out at the same time - we're done */
                    done = TRUE;
                }
                else {
                    /* One ran out before the other - BADNESS */
                    nerrors++;
                    if(verbose) {
                        HDfprintf(stdout, "%d:%s: Unexpected data in trace file line %d.\n", world_mpi_rank, FUNC, i);
                        if(expected_line_len == 0) {
                            HDfprintf(stdout, "%d:%s: expected = \"%s\" %d\n", world_mpi_rank, FUNC, "<EMPTY>", expected_line_len);
                            HDfprintf(stdout, "%d:%s: actual   = \"%s\" %d\n", world_mpi_rank, FUNC, buffer, actual_line_len);
                        }
                        if(actual_line_len == 0) {
                            HDfprintf(stdout, "%d:%s: expected = \"%s\" %d\n", world_mpi_rank, FUNC, (*expected_output)[i], expected_line_len);
                            HDfprintf(stdout, "%d:%s: actual   = \"%s\" %d\n", world_mpi_rank, FUNC, "<EMPTY>", actual_line_len);
                        }
                    }
                    HDfprintf(stdout, "BADNESS BADNESS BADNESS\n");
                }
            }
            /* We directly compare the header line (line 0) */
            else if(0 == i) {
                if((actual_line_len != expected_line_len) || (HDstrcmp(buffer, (*expected_output)[i]) != 0 )) {

                    nerrors++;
                    if(verbose) {
                        HDfprintf(stdout, "%d:%s: Unexpected data in trace file line %d.\n", world_mpi_rank, FUNC, i);
                        HDfprintf(stdout, "%d:%s: expected = \"%s\" %d\n", world_mpi_rank, FUNC, (*expected_output)[i], expected_line_len);
                        HDfprintf(stdout, "%d:%s: actual   = \"%s\" %d\n", world_mpi_rank, FUNC, buffer, actual_line_len);
                    }
                }
            }
            /* All other lines we tokenize and just compare the function name. This
             * keeps the test from being too fragile.
             */
            else {
                char *tok = NULL;     /* token for actual line */

                tok = HDstrtok(buffer, " ");

                if(HDstrcmp(tok, (*expected_output)[i]) != 0 ) {

                    nerrors++;
                    if(verbose) {
                        HDfprintf(stdout, "%d:%s: Unexpected data in trace file line %d.\n", world_mpi_rank, FUNC, i);
                        HDfprintf(stdout, "%d:%s: expected = \"%s\"\n", world_mpi_rank, FUNC, (*expected_output)[i]);
                        HDfprintf(stdout, "%d:%s: actual   = \"%s\"\n", world_mpi_rank, FUNC, tok);
                    }
                }
            } /* end else */

            i++;
        } /* end while */

        /* Clean up the trace file */
        if(trace_file_ptr != NULL) {
            HDfclose(trace_file_ptr);
            trace_file_ptr = NULL;
            HDremove(trace_file_name);
        }
    } /* end giant else that runs clients */

    max_nerrors = get_max_nerrors();

    if(world_mpi_rank == 0) {

        if(max_nerrors == 0) {
            PASSED();
        }
        else {
            failures++;
            H5_FAILED();
        }
    }

    success = ((success) && (max_nerrors == 0));

    return(success);

} /* trace_file_check() */


/*****************************************************************************
 *
 * Function:    smoke_check_6()
 *
 * Purpose:    Sixth smoke check for the parallel cache.
 *
 * Return:    Success:    TRUE
 *
 *        Failure:    FALSE
 *
 * Programmer:    JRM -- 1/13/06
 *
 *****************************************************************************/
static hbool_t
smoke_check_6(int metadata_write_strategy)
{
    hbool_t success = TRUE;
    int i;
    int max_nerrors;
    hid_t fid = -1;
    H5F_t * file_ptr = NULL;
    H5C_t * cache_ptr = NULL;
    struct mssg_t mssg;

    switch ( metadata_write_strategy ) {

    case H5AC_METADATA_WRITE_STRATEGY__PROCESS_0_ONLY:
            if ( world_mpi_rank == 0 ) {
            TESTING("smoke check #6 -- process 0 only md write strategy");
            }
        break;

    case H5AC_METADATA_WRITE_STRATEGY__DISTRIBUTED:
            if ( world_mpi_rank == 0 ) {
            TESTING("smoke check #6 -- distributed md write strategy");
            }
        break;

        default:
            if ( world_mpi_rank == 0 ) {
            TESTING("smoke check #6 -- unknown md write strategy");
            }
        break;
    }

    nerrors = 0;
    init_data();
    reset_stats();

    if ( world_mpi_rank == world_server_mpi_rank ) {

    if ( ! server_main() ) {

            /* some error occured in the server -- report failure */
            nerrors++;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: server_main() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }
    else /* run the clients */
    {
        int temp;

        if ( ! setup_cache_for_test(&fid, &file_ptr, &cache_ptr,
                                    metadata_write_strategy) ) {

            nerrors++;
            fid = -1;
            cache_ptr = NULL;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: setup_cache_for_test() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }

        temp = virt_num_data_entries;
        virt_num_data_entries = NUM_DATA_ENTRIES;

        /* insert the first half collectively */
        H5CX_set_coll_metadata_read(TRUE);
        for ( i = 0; i < virt_num_data_entries/2; i++ )
        {
            struct datum * entry_ptr;
            entry_ptr = &(data[i]);

            insert_entry(cache_ptr, file_ptr, i, H5AC__NO_FLAGS_SET);

            if(TRUE != entry_ptr->header.coll_access) {
                nerrors++;
                if ( verbose ) {
                    HDfprintf(stdout, "%d:%s: Entry inserted not marked as collective.\n",
                              world_mpi_rank, FUNC);
                }
            }

            /* Make sure coll entries do not cross the 80% threshold */
            H5_CHECK_OVERFLOW(cache_ptr->max_cache_size, size_t, double);
            HDassert((double)cache_ptr->max_cache_size*0.8 > cache_ptr->coll_list_size);
        }

        /* insert the other half independently */
        H5CX_set_coll_metadata_read(FALSE);
        for ( i = virt_num_data_entries/2; i < virt_num_data_entries; i++ )
        {
            struct datum * entry_ptr;
            entry_ptr = &(data[i]);

            insert_entry(cache_ptr, file_ptr, i, H5AC__NO_FLAGS_SET);

            if(FALSE != entry_ptr->header.coll_access) {
                nerrors++;
                if ( verbose ) {
                    HDfprintf(stdout, "%d:%s: Entry inserted indepedently marked as collective.\n",
                              world_mpi_rank, FUNC);
                }
            }

            /* Make sure coll entries do not cross the 80% threshold */
            HDassert((double)cache_ptr->max_cache_size*0.8 > cache_ptr->coll_list_size);
        }

        /* flush the file */
        if ( H5Fflush(fid, H5F_SCOPE_GLOBAL) < 0 ) {
            nerrors++;
            if ( verbose ) {
            HDfprintf(stdout, "%d:%s: H5Fflush() failed.\n",
                          world_mpi_rank, FUNC);
            }
        }

        /* Protect the first half of the entries collectively */
        H5CX_set_coll_metadata_read(TRUE);
        for ( i = 0; i < (virt_num_data_entries / 2); i++ )
        {
            struct datum * entry_ptr;
            entry_ptr = &(data[i]);

        lock_entry(file_ptr, i);

            if(TRUE != entry_ptr->header.coll_access) {
                nerrors++;
                if ( verbose ) {
                    HDfprintf(stdout, "%d:%s: Entry protected not marked as collective.\n",
                              world_mpi_rank, FUNC);
                }
            }

            /* Make sure coll entries do not cross the 80% threshold */
            HDassert((double)cache_ptr->max_cache_size*0.8 > cache_ptr->coll_list_size);
        }

        /* protect the other half independently */
        H5CX_set_coll_metadata_read(FALSE);
        for ( i = virt_num_data_entries/2; i < virt_num_data_entries; i++ )
        {
            struct datum * entry_ptr;
            entry_ptr = &(data[i]);

            lock_entry(file_ptr, i);

            if(FALSE != entry_ptr->header.coll_access) {
                nerrors++;
                if ( verbose ) {
                    HDfprintf(stdout, "%d:%s: Entry inserted indepedently marked as collective.\n",
                              world_mpi_rank, FUNC);
                }
            }

            /* Make sure coll entries do not cross the 80% threshold */
            HDassert((double)cache_ptr->max_cache_size*0.8 > cache_ptr->coll_list_size);
        }

        for ( i = 0; i < (virt_num_data_entries); i++ )
        {
            unlock_entry(file_ptr, i, H5AC__NO_FLAGS_SET);
        }

        if ( fid >= 0 ) {

            if ( ! take_down_cache(fid, cache_ptr) ) {

                nerrors++;
                if ( verbose ) {
            HDfprintf(stdout, "%d:%s: take_down_cache() failed.\n",
                              world_mpi_rank, FUNC);
                }
            }
        }

        /* verify that all instances of datum are back where the started
         * and are clean.
         */

        for ( i = 0; i < NUM_DATA_ENTRIES; i++ )
        {
            HDassert( data_index[i] == i );
            HDassert( ! (data[i].dirty) );
        }

        /* compose the done message */
        mssg.req       = DONE_REQ_CODE;
        mssg.src       = world_mpi_rank;
        mssg.dest      = world_server_mpi_rank;
        mssg.mssg_num  = -1; /* set by send function */
        mssg.base_addr = 0; /* not used */
        mssg.len       = 0; /* not used */
        mssg.ver       = 0; /* not used */
        mssg.count     = 0; /* not used */
        mssg.magic     = MSSG_MAGIC;

        if ( success ) {


            success = send_mssg(&mssg, FALSE);

            if ( ! success ) {

                nerrors++;
                if ( verbose ) {
                    HDfprintf(stdout, "%d:%s: send_mssg() failed on done.\n",
                              world_mpi_rank, FUNC);
                }
            }
        }
        virt_num_data_entries = temp;
    }

    max_nerrors = get_max_nerrors();

    if ( world_mpi_rank == 0 ) {

    if ( max_nerrors == 0 ) {

        PASSED();

        } else {

            failures++;
            H5_FAILED();
        }
    }

    success = ( ( success ) && ( max_nerrors == 0 ) );

    return(success);

} /* smoke_check_6() */


/*****************************************************************************
 *
 * Function:    main()
 *
 * Purpose:    Main function for the parallel cache test.
 *
 * Return:    Success:    0
 *
 *        Failure:    1
 *
 * Programmer:    JRM -- 12/23/05
 *
 *****************************************************************************/
int
main(int argc, char **argv)
{
    int express_test;
    unsigned u;
    int mpi_size;
    int mpi_rank;
    int max_nerrors;

    MPI_Init(&argc, &argv);
    MPI_Comm_size(MPI_COMM_WORLD, &mpi_size);
    MPI_Comm_rank(MPI_COMM_WORLD, &mpi_rank);

    world_mpi_size = mpi_size;
    world_mpi_rank = mpi_rank;
    world_server_mpi_rank = mpi_size - 1;
    world_mpi_comm = MPI_COMM_WORLD;

    /* Attempt to turn off atexit post processing so that in case errors
     * happen during the test and the process is aborted, it will not get
     * hang in the atexit post processing in which it may try to make MPI
     * calls.  By then, MPI calls may not work.
     */
    if (H5dont_atexit() < 0){
        HDprintf("%d:Failed to turn off atexit processing. Continue.\n",
               mpi_rank);
    };
    H5open();

    express_test = do_express_test();
#if 0 /* JRM */
    express_test = 0;
#endif /* JRM */
    if ( express_test ) {

    virt_num_data_entries = EXPRESS_VIRT_NUM_DATA_ENTRIES;

    } else {

    virt_num_data_entries = STD_VIRT_NUM_DATA_ENTRIES;
    }

#ifdef H5_HAVE_MPE
    if ( MAINPROCESS ) { HDprintf("    Tests compiled for MPE.\n"); }
    virt_num_data_entries = MPE_VIRT_NUM_DATA_ENTIES;
#endif /* H5_HAVE_MPE */


    if (MAINPROCESS){
    HDprintf("===================================\n");
    HDprintf("Parallel metadata cache tests\n");
    HDprintf("    mpi_size     = %d\n", mpi_size);
    HDprintf("    express_test = %d\n", express_test);
    HDprintf("===================================\n");
    }

    if ( mpi_size < 3 ) {

        if ( MAINPROCESS ) {

            HDprintf("    Need at least 3 processes.  Exiting.\n");
        }
        goto finish;
    }

    set_up_file_communicator();

    setup_derived_types();

    /* h5_fixname() will hang some processes don't participate.
     *
     * Thus we set up the fapl global with the world communicator,
     * make our calls to h5_fixname(), discard the fapl, and then
     * create it again with the file communicator.
     */

    /* setup file access property list with the world communicator */
    if ( FAIL == (fapl = H5Pcreate(H5P_FILE_ACCESS)) ) {
        nerrors++;
    if ( verbose ) {
        HDfprintf(stdout, "%d:%s: H5Pcreate() failed 1.\n",
                      world_mpi_rank, FUNC);
        }
    }

    if ( H5Pset_fapl_mpio(fapl, world_mpi_comm, MPI_INFO_NULL) < 0 ) {

        nerrors++;
        if ( verbose ) {
            HDfprintf(stdout, "%d:%s: H5Pset_fapl_mpio() failed 1.\n",
                      world_mpi_rank, FUNC);
        }
    }

    /* fix the file names */
    for ( u = 0; u < sizeof(FILENAME) / sizeof(FILENAME[0]) - 1; ++u )
    {
        if ( h5_fixname(FILENAME[u], fapl, filenames[u],
                        sizeof(filenames[u])) == NULL ) {

            nerrors++;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: h5_fixname() failed.\n",
                          world_mpi_rank, FUNC);
            }
            break;
        }
    }

    /* close the fapl before we set it up again */
    if ( H5Pclose(fapl) < 0 ) {
        nerrors++;
    if ( verbose ) {
        HDfprintf(stdout, "%d:%s: H5Pclose() failed.\n",
                      world_mpi_rank, FUNC);
        }
    }

    /* now create the fapl again, excluding the server process. */
    if ( world_mpi_rank != world_server_mpi_rank ) {

        /* setup file access property list */
        if ( FAIL == (fapl = H5Pcreate(H5P_FILE_ACCESS)) ) {
        nerrors++;
        if ( verbose ) {
            HDfprintf(stdout, "%d:%s: H5Pcreate() failed 2.\n",
                          world_mpi_rank, FUNC);
            }
        }

        if ( H5Pset_fapl_mpio(fapl, file_mpi_comm, MPI_INFO_NULL) < 0 ) {

            nerrors++;
        if ( verbose ) {
            HDfprintf(stdout, "%d:%s: H5Pset_fapl_mpio() failed 2.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }

    setup_rand();

    max_nerrors = get_max_nerrors();

    if ( max_nerrors != 0 ) {

        /* errors in setup -- no point in continuing */

        if ( world_mpi_rank == 0 ) {

            HDfprintf(stdout, "Errors in test initialization.  Exiting.\n");
        }
    goto finish;
    }

    /* run the tests */
#if 1
    server_smoke_check();
#endif
#if 1
    smoke_check_1(H5AC_METADATA_WRITE_STRATEGY__PROCESS_0_ONLY);
    smoke_check_1(H5AC_METADATA_WRITE_STRATEGY__DISTRIBUTED);
#endif
#if 1
    smoke_check_2(H5AC_METADATA_WRITE_STRATEGY__PROCESS_0_ONLY);
    smoke_check_2(H5AC_METADATA_WRITE_STRATEGY__DISTRIBUTED);
#endif
#if 1
    smoke_check_3(H5AC_METADATA_WRITE_STRATEGY__PROCESS_0_ONLY);
    smoke_check_3(H5AC_METADATA_WRITE_STRATEGY__DISTRIBUTED);
#endif
#if 1
    smoke_check_4(H5AC_METADATA_WRITE_STRATEGY__PROCESS_0_ONLY);
    smoke_check_4(H5AC_METADATA_WRITE_STRATEGY__DISTRIBUTED);
#endif
#if 1
    smoke_check_5(H5AC_METADATA_WRITE_STRATEGY__PROCESS_0_ONLY);
    smoke_check_5(H5AC_METADATA_WRITE_STRATEGY__DISTRIBUTED);
#endif
    /* enable the collective metadata read property */
    if ( world_mpi_rank != world_server_mpi_rank ) {
        if ( H5Pset_all_coll_metadata_ops(fapl, TRUE) < 0 ) {

            nerrors++;
            if ( verbose ) {
                HDfprintf(stdout, "%d:%s: H5Pset_all_coll_metadata_ops() failed 1.\n",
                          world_mpi_rank, FUNC);
            }
        }
    }
#if 1
    smoke_check_6(H5AC_METADATA_WRITE_STRATEGY__PROCESS_0_ONLY);
    smoke_check_6(H5AC_METADATA_WRITE_STRATEGY__DISTRIBUTED);
#endif

#if 1
    trace_file_check(H5AC_METADATA_WRITE_STRATEGY__PROCESS_0_ONLY);
    trace_file_check(H5AC_METADATA_WRITE_STRATEGY__DISTRIBUTED);
#endif

finish:
    /* make sure all processes are finished before final report, cleanup
     * and exit.
     */
    MPI_Barrier(MPI_COMM_WORLD);
    if (MAINPROCESS){        /* only process 0 reports */
    HDprintf("===================================\n");
    if (failures){
        HDprintf("***metadata cache tests detected %d failures***\n",
                   failures);
    }
    else{
        HDprintf("metadata cache tests finished with no failures\n");
    }
    HDprintf("===================================\n");
    }

    takedown_derived_types();

    /* close HDF5 library */
    H5close();

    /* MPI_Finalize must be called AFTER H5close which may use MPI calls */
    MPI_Finalize();

    /* cannot just return (failures) because exit code is limited to 1byte */
    return(failures != 0);
}