experiments-postgresql-locks/pg_lock_hash_info.c

/* pg_lock_hash_info.c - PostgreSQL extension to expose lock and proclock hash bucket statistics */

#include "postgres.h"
#include "fmgr.h"
#include "access/htup_details.h"
#include "miscadmin.h" // Access to MaxBackends and MaxConnections
#include "access/twophase.h" // Access to max_prepared_xacts
#include "funcapi.h"
#include "storage/lock.h"
#include "storage/lwlock.h"
#include "storage/shmem.h"
#include "storage/spin.h"
#include "utils/builtins.h"
#include "utils/hsearch.h"
#include "utils/dynahash.h"

PG_MODULE_MAGIC;

HTAB *lock_hash = NULL;

void get_lock_hash();

// BEGIN Copied from dynahash.c

#define NLOCKENTS() \
        mul_size(max_locks_per_xact, add_size(MaxBackends, max_prepared_xacts))

#define DEF_SEGSIZE            256
#define DEF_SEGSIZE_SHIFT      8    /* must be log2(DEF_SEGSIZE) */
#define DEF_DIRSIZE            256
#define NUM_FREELISTS           32

typedef HASHELEMENT *HASHBUCKET;
typedef HASHBUCKET *HASHSEGMENT;

typedef struct
{
    slock_t     mutex;          /* spinlock for this freelist */
    long        nentries;       /* number of entries in associated buckets */
    HASHELEMENT *freeList;      /* chain of free elements */
} FreeListData;

struct HASHHDR
{
    FreeListData freeList[NUM_FREELISTS];
    long        dsize;                  /* directory size */
    long        nsegs;                  /* number of allocated segments (<= dsize) */
    uint32      max_bucket;             /* ID of maximum bucket in use */
    uint32      high_mask;              /* mask to modulo into entire table */
    uint32      low_mask;               /* mask to modulo into lower half of table */
    Size        keysize;                /* hash key length in bytes */
    Size        entrysize;              /* total user element size in bytes */
    long        num_partitions; /* # partitions (must be power of 2), or 0 */
    long        max_dsize;              /* 'dsize' limit if directory is fixed size */
    long        ssize;                  /* segment size --- must be power of 2 */
    int         sshift;                 /* segment shift = log2(ssize) */
    int         nelem_alloc;    /* number of entries to allocate at once */
#ifdef HASH_STATISTICS
    long        accesses;
    long        collisions;
#endif
};

struct HTAB
{
    HASHHDR    *hctl;           /* => shared control information */
    HASHSEGMENT *dir;           /* directory of segment starts */
    HashValueFunc hash;         /* hash function */
    HashCompareFunc match;      /* key comparison function */
    HashCopyFunc keycopy;       /* key copying function */
    HashAllocFunc alloc;        /* memory allocator */
    MemoryContext hcxt;         /* memory context if default allocator used */
    char       *tabname;        /* table name (for error messages) */
    bool        isshared;       /* true if table is in shared memory */
    bool        isfixed;        /* if true, don't enlarge */
    bool        frozen;         /* true = no more inserts allowed */
    Size        keysize;        /* hash key length in bytes */
    long        ssize;          /* segment size --- must be power of 2 */
    int         sshift;         /* segment shift = log2(ssize) */
};

// END Copied from dynahash.c

void
get_lock_hash()
{
    HASHCTL info;
    long init_table_size, max_table_size;
    bool found;

    max_table_size = NLOCKENTS();
    init_table_size = max_table_size / 2;

    info.keysize = sizeof(LOCKTAG);
    info.entrysize = sizeof(LOCK);
    info.num_partitions = NUM_LOCK_PARTITIONS;

    lock_hash = ShmemInitHash("LOCK hash", init_table_size, max_table_size, &info, HASH_ELEM | HASH_BLOBS | HASH_PARTITION);

    if (!lock_hash)
    {
        elog(ERROR, "Failed to find lock hash");
    }
}

int
get_free_count(FreeListData *lst)
{
    HASHELEMENT *elem;
    int count = 0;

    SpinLockAcquire(&(lst->mutex));
    elem = lst->freeList;
    while (elem != NULL)
    {
        count++;
        elem = elem->link;
    }
    SpinLockRelease(&(lst->mutex));

    return count;
}

int
get_total_free_count(HASHHDR *hctl)
{
    int i;
    int count = 0;

    for (i=0; i<NUM_FREELISTS; i++)
    {
        count += get_free_count(&(hctl->freeList[i]));
    }

    return count;
}


PG_FUNCTION_INFO_V1(pg_lock_hash_info);

Datum
pg_lock_hash_info(PG_FUNCTION_ARGS)
{
    TupleDesc tupdesc;
    Datum values[14];
    bool nulls[14];
    HeapTuple tuple;

    // Get a pointer to the lock hash
    get_lock_hash();

    if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE)
        ereport(ERROR,
                (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
                 errmsg("function returning record called in context that cannot accept type record")));

    values[0] = Int32GetDatum(lock_hash->hctl->dsize);
    //values[0] = Int32GetDatum(MaxBackends);
    nulls[0] = false;

    values[1] = Int32GetDatum(lock_hash->hctl->nsegs);
    //values[1] = Int32GetDatum(MaxConnections);
    nulls[1] = false;

    values[2] = UInt32GetDatum(lock_hash->hctl->max_bucket);
    nulls[2] = false;

    values[3] = UInt32GetDatum(lock_hash->hctl->keysize);
    nulls[3] = false;

    values[4] = UInt32GetDatum(lock_hash->hctl->entrysize);
    nulls[4] = false;

    values[5] = Int32GetDatum(lock_hash->hctl->num_partitions);
    nulls[5] = false;

    values[6] = Int32GetDatum(lock_hash->hctl->max_dsize);
    nulls[6] = false;

    values[7] = Int32GetDatum(lock_hash->hctl->ssize);
    nulls[7] = false;

    values[8] = Int32GetDatum(lock_hash->hctl->sshift);
    nulls[8] = false;

    values[9] = Int32GetDatum(lock_hash->hctl->nelem_alloc);
    nulls[9] = false;

    values[10] = Int32GetDatum(get_total_free_count(lock_hash->hctl));
    nulls[10] = false;

    values[11] = Int32GetDatum(NLOCKENTS());
    nulls[11] = false;

    values[12] = Int32GetDatum(lock_hash->hctl->low_mask);
    nulls[12] = false;

    values[13] = Int32GetDatum(lock_hash->hctl->high_mask);
    nulls[13] = false;

    tuple = heap_form_tuple(tupdesc, values, nulls);
    PG_RETURN_DATUM(HeapTupleGetDatum(tuple));
}



PG_FUNCTION_INFO_V1(pg_lock_hash_freelist_info);

Datum
pg_lock_hash_freelist_info(PG_FUNCTION_ARGS)
{
    FuncCallContext *funcctx;
    int *list_id;

    if (SRF_IS_FIRSTCALL())
    {
        TupleDesc tupdesc;
        MemoryContext oldcontext;

        funcctx = SRF_FIRSTCALL_INIT();
        oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx);

        // Get a pointer to the lock hash
        get_lock_hash();

        tupdesc = CreateTemplateTupleDesc(3);
        TupleDescInitEntry(tupdesc, (AttrNumber) 1, "freelist", INT4OID, -1, 0);
        TupleDescInitEntry(tupdesc, (AttrNumber) 2, "nentries", INT4OID, -1, 0);
        TupleDescInitEntry(tupdesc, (AttrNumber) 3, "count", INT4OID, -1, 0);
        funcctx->tuple_desc = BlessTupleDesc(tupdesc);

        /* Allocate and initialize the position */
        funcctx->user_fctx = palloc(sizeof(int));
        list_id = (int *) funcctx->user_fctx;
        *list_id = 0;

        MemoryContextSwitchTo(oldcontext);
    }

    funcctx = SRF_PERCALL_SETUP();
    list_id = (int *) funcctx->user_fctx;

    if (*list_id >= NUM_FREELISTS)
    {
        SRF_RETURN_DONE(funcctx);
    }
    else
    {
        Datum values[4];
        bool nulls[4];
        HeapTuple tuple;
        FreeListData *freelist = &(lock_hash->hctl->freeList[*list_id]);

        values[0] = Int32GetDatum(*list_id);
        nulls[0] = false;

        values[1] = Int32GetDatum(freelist->nentries);
        nulls[1] = false;

        values[2] = Int32GetDatum(get_free_count(freelist));
        nulls[2] = false;

        tuple = heap_form_tuple(funcctx->tuple_desc, values, nulls);

        // Increment the position, handling segment wraparound
        (*list_id)++;

        SRF_RETURN_NEXT(funcctx, HeapTupleGetDatum(tuple));
    }
}


int
count_list(HASHBUCKET root)
{
    int count = 0;
    HASHELEMENT *elem;

    if (!root)
    {
        return 0;
    }

    elem = root->link;

    while (elem != NULL) {
      count++;
      elem = elem->link;
    }

    return count;
}


PG_FUNCTION_INFO_V1(pg_lock_hash_bucket_info);

typedef struct {
    int segment_id;
    int bucket_id;
} LockHashBucketInfoPosition;

Datum
pg_lock_hash_bucket_info(PG_FUNCTION_ARGS)
{
    FuncCallContext *funcctx;
    LockHashBucketInfoPosition *pos;
    LWLock *part_lock;

    if (SRF_IS_FIRSTCALL())
    {
        TupleDesc tupdesc;
        MemoryContext oldcontext;

        funcctx = SRF_FIRSTCALL_INIT();
        oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx);

        // Get a pointer to the lock hash
        get_lock_hash();

        tupdesc = CreateTemplateTupleDesc(6);
        TupleDescInitEntry(tupdesc, (AttrNumber) 1, "segment", INT4OID, -1, 0);
        TupleDescInitEntry(tupdesc, (AttrNumber) 2, "bucket", INT4OID, -1, 0);
        TupleDescInitEntry(tupdesc, (AttrNumber) 3, "bucket_hash", INT4OID, -1, 0);
        TupleDescInitEntry(tupdesc, (AttrNumber) 4, "bucket_hash_low", INT4OID, -1, 0);
        TupleDescInitEntry(tupdesc, (AttrNumber) 5, "bucket_hash_high", INT4OID, -1, 0);
        TupleDescInitEntry(tupdesc, (AttrNumber) 6, "entries", INT4OID, -1, 0);
        funcctx->tuple_desc = BlessTupleDesc(tupdesc);

        /* Allocate and initialize the position */
        funcctx->user_fctx = palloc(sizeof(LockHashBucketInfoPosition));
        pos = (LockHashBucketInfoPosition *) funcctx->user_fctx;
        pos->segment_id = 0;
        pos->bucket_id = 0;

        MemoryContextSwitchTo(oldcontext);
    }

    funcctx = SRF_PERCALL_SETUP();
    pos = (LockHashBucketInfoPosition *) funcctx->user_fctx;

    if (pos->segment_id >= lock_hash->hctl->nsegs)
    {
        SRF_RETURN_DONE(funcctx);
    }
    else
    {
        Datum values[6];
        bool nulls[6];
        HeapTuple tuple;
        HASHBUCKET bucketp = lock_hash->dir[pos->segment_id][pos->bucket_id];

        values[0] = Int32GetDatum(pos->segment_id);
        nulls[0] = false;

        values[1] = Int32GetDatum(pos->bucket_id);
        nulls[1] = false;

        if (bucketp)
        {
            values[2] = Int32GetDatum(bucketp->hashvalue);
            nulls[2] = false;

            values[3] = Int32GetDatum(bucketp->hashvalue & lock_hash->hctl->low_mask);
            nulls[3] = false;

            values[4] = Int32GetDatum(bucketp->hashvalue & lock_hash->hctl->high_mask);
            nulls[4] = false;

            part_lock = LockHashPartitionLock(bucketp->hashvalue);
            LWLockAcquire(part_lock, LW_SHARED);
            values[5] = UInt32GetDatum(count_list(bucketp));
            nulls[5] = false;
            LWLockRelease(part_lock);
        }
        else
        {
            values[2] = Int32GetDatum(0);
            nulls[2] = true;

            values[3] = Int32GetDatum(0);
            nulls[3] = true;

            values[4] = Int32GetDatum(0);
            nulls[4] = true;

            values[5] = Int32GetDatum(0);
            nulls[5] = false;
        }

        tuple = heap_form_tuple(funcctx->tuple_desc, values, nulls);

        // Increment the position, handling segment wraparound
        pos->bucket_id++;
        while (pos->bucket_id >= lock_hash->hctl->ssize)
        {
            pos->bucket_id -= lock_hash->hctl->ssize;
            pos->segment_id++;
        }

        SRF_RETURN_NEXT(funcctx, HeapTupleGetDatum(tuple));
    }
}