async.c File Reference

#include "postgres.h"
#include <limits.h>
#include <unistd.h>
#include <signal.h>
#include "access/parallel.h"
#include "access/slru.h"
#include "access/transam.h"
#include "access/xact.h"
#include "catalog/pg_database.h"
#include "commands/async.h"
#include "common/hashfn.h"
#include "funcapi.h"
#include "lib/dshash.h"
#include "libpq/libpq.h"
#include "libpq/pqformat.h"
#include "miscadmin.h"
#include "storage/dsm_registry.h"
#include "storage/ipc.h"
#include "storage/latch.h"
#include "storage/lmgr.h"
#include "storage/procsignal.h"
#include "storage/subsystems.h"
#include "tcop/tcopprot.h"
#include "utils/builtins.h"
#include "utils/dsa.h"
#include "utils/guc_hooks.h"
#include "utils/memutils.h"
#include "utils/ps_status.h"
#include "utils/snapmgr.h"
#include "utils/timestamp.h"

Include dependency graph for async.c:

Go to the source code of this file.

Data Structures
struct	AsyncQueueEntry
struct	QueuePosition
struct	QueueBackendStatus
struct	AsyncQueueControl
struct	GlobalChannelKey
struct	ListenerEntry
struct	GlobalChannelEntry
struct	ListenAction
struct	ActionList
struct	PendingListenEntry
struct	Notification
struct	NotificationList
struct	NotificationHash
struct	ChannelName

Macros
#define	NOTIFY_PAYLOAD_MAX_LENGTH   (BLCKSZ - NAMEDATALEN - 128)
#define	QUEUEALIGN(len)   INTALIGN(len)
#define	AsyncQueueEntryEmptySize   (offsetof(AsyncQueueEntry, data) + 2)
#define	QUEUE_POS_PAGE(x)   ((x).page)
#define	QUEUE_POS_OFFSET(x)   ((x).offset)
#define	SET_QUEUE_POS(x, y, z)
#define	QUEUE_POS_EQUAL(x, y)    ((x).page == (y).page && (x).offset == (y).offset)
#define	QUEUE_POS_IS_ZERO(x)    ((x).page == 0 && (x).offset == 0)
#define	QUEUE_POS_MIN(x, y)
#define	QUEUE_POS_MAX(x, y)
#define	QUEUE_POS_PRECEDES(x, y)
#define	QUEUE_CLEANUP_DELAY   4
#define	QUEUE_HEAD   (asyncQueueControl->head)
#define	QUEUE_TAIL   (asyncQueueControl->tail)
#define	QUEUE_STOP_PAGE   (asyncQueueControl->stopPage)
#define	QUEUE_FIRST_LISTENER   (asyncQueueControl->firstListener)
#define	QUEUE_BACKEND_PID(i)   (asyncQueueControl->backend[i].pid)
#define	QUEUE_BACKEND_DBOID(i)   (asyncQueueControl->backend[i].dboid)
#define	QUEUE_NEXT_LISTENER(i)   (asyncQueueControl->backend[i].nextListener)
#define	QUEUE_BACKEND_POS(i)   (asyncQueueControl->backend[i].pos)
#define	QUEUE_BACKEND_WAKEUP_PENDING(i)   (asyncQueueControl->backend[i].wakeupPending)
#define	QUEUE_BACKEND_IS_ADVANCING(i)   (asyncQueueControl->backend[i].isAdvancing)
#define	NotifyCtl   (&NotifySlruDesc)
#define	QUEUE_PAGESIZE   BLCKSZ
#define	QUEUE_FULL_WARN_INTERVAL   5000 /* warn at most once every 5s */
#define	INITIAL_LISTENERS_ARRAY_SIZE   4
#define	LocalChannelTableIsEmpty()    (localChannelTable == NULL || hash_get_num_entries(localChannelTable) == 0)
#define	MIN_HASHABLE_NOTIFIES   16 /* threshold to build hashtab */

Typedefs
typedef struct AsyncQueueEntry	AsyncQueueEntry
typedef struct QueuePosition	QueuePosition
typedef struct QueueBackendStatus	QueueBackendStatus
typedef struct AsyncQueueControl	AsyncQueueControl
typedef struct GlobalChannelKey	GlobalChannelKey
typedef struct ListenerEntry	ListenerEntry
typedef struct GlobalChannelEntry	GlobalChannelEntry
typedef struct ActionList	ActionList
typedef struct PendingListenEntry	PendingListenEntry
typedef struct Notification	Notification
typedef struct NotificationList	NotificationList
typedef struct ChannelName	ChannelName

Enumerations
enum	ListenActionKind { LISTEN_LISTEN , LISTEN_UNLISTEN , LISTEN_UNLISTEN_ALL }
enum	PendingListenAction { PENDING_LISTEN , PENDING_UNLISTEN }

Functions
static void	AsyncShmemRequest (void *arg)
static void	AsyncShmemInit (void *arg)
static bool	asyncQueuePagePrecedes (int64 p, int64 q)
static int	asyncQueueErrdetailForIoError (const void *opaque_data)
static int64	asyncQueuePageDiff (int64 p, int64 q)
static void	GlobalChannelKeyInit (GlobalChannelKey *key, Oid dboid, const char *channel)
static dshash_hash	globalChannelTableHash (const void *key, size_t size, void *arg)
static void	initGlobalChannelTable (void)
static void	initLocalChannelTable (void)
static void	queue_listen (ListenActionKind action, const char *channel)
static void	Async_UnlistenOnExit (int code, Datum arg)
static void	BecomeRegisteredListener (void)
static void	PrepareTableEntriesForListen (const char *channel)
static void	PrepareTableEntriesForUnlisten (const char *channel)
static void	PrepareTableEntriesForUnlistenAll (void)
static void	RemoveListenerFromChannel (GlobalChannelEntry **entry_ptr, ListenerEntry *listeners, int idx)
static void	ApplyPendingListenActions (bool isCommit)
static void	CleanupListenersOnExit (void)
static bool	IsListeningOn (const char *channel)
static void	asyncQueueUnregister (void)
static bool	asyncQueueIsFull (void)
static bool	asyncQueueAdvance (volatile QueuePosition *position, int entryLength)
static void	asyncQueueNotificationToEntry (Notification *n, AsyncQueueEntry *qe)
static ListCell *	asyncQueueAddEntries (ListCell *nextNotify)
static double	asyncQueueUsage (void)
static void	asyncQueueFillWarning (void)
static void	SignalBackends (void)
static void	asyncQueueReadAllNotifications (void)
static bool	asyncQueueProcessPageEntries (QueuePosition *current, QueuePosition stop, Snapshot snapshot)
static void	asyncQueueAdvanceTail (void)
static void	ProcessIncomingNotify (bool flush)
static bool	AsyncExistsPendingNotify (Notification *n)
static void	AddEventToPendingNotifies (Notification *n)
static uint32	notification_hash (const void *key, Size keysize)
static int	notification_match (const void *key1, const void *key2, Size keysize)
static void	ClearPendingActionsAndNotifies (void)
static void	initPendingListenActions (void)
Datum	pg_notify (PG_FUNCTION_ARGS)
void	Async_Notify (const char *channel, const char *payload)
void	Async_Listen (const char *channel)
void	Async_Unlisten (const char *channel)
void	Async_UnlistenAll (void)
Datum	pg_listening_channels (PG_FUNCTION_ARGS)
void	AtPrepare_Notify (void)
void	PreCommit_Notify (void)
void	AtCommit_Notify (void)
Datum	pg_notification_queue_usage (PG_FUNCTION_ARGS)
void	AtAbort_Notify (void)
void	AtSubCommit_Notify (void)
void	AtSubAbort_Notify (void)
void	HandleNotifyInterrupt (void)
void	ProcessNotifyInterrupt (bool flush)
void	AsyncNotifyFreezeXids (TransactionId newFrozenXid)
void	NotifyMyFrontEnd (const char *channel, const char *payload, int32 srcPid)
bool	check_notify_buffers (int *newval, void **extra, GucSource source)

Variables
static AsyncQueueControl *	asyncQueueControl
const ShmemCallbacks	AsyncShmemCallbacks
static SlruDesc	NotifySlruDesc
static dshash_table *	globalChannelTable = NULL
static dsa_area *	globalChannelDSA = NULL
static HTAB *	localChannelTable = NULL
static ActionList *	pendingActions = NULL
static HTAB *	pendingListenActions = NULL
static NotificationList *	pendingNotifies = NULL
volatile sig_atomic_t	notifyInterruptPending = false
static bool	unlistenExitRegistered = false
static bool	amRegisteredListener = false
static QueuePosition	queueHeadBeforeWrite
static QueuePosition	queueHeadAfterWrite
static int32 *	signalPids = NULL
static ProcNumber *	signalProcnos = NULL
static bool	tryAdvanceTail = false
bool	Trace_notify = false
int	max_notify_queue_pages = 1048576
static const dshash_parameters	globalChannelTableDSHParams

Macro Definition Documentation

AsyncQueueEntryEmptySize

#define AsyncQueueEntryEmptySize   (offsetof(AsyncQueueEntry, data) + 2)

Definition at line 227 of file async.c.

INITIAL_LISTENERS_ARRAY_SIZE

#define INITIAL_LISTENERS_ARRAY_SIZE   4

Definition at line 392 of file async.c.

LocalChannelTableIsEmpty

#define LocalChannelTableIsEmpty

(

)

(localChannelTable == NULL || hash_get_num_entries(localChannelTable) == 0)

Definition at line 425 of file async.c.

{
    LISTEN_LISTEN,
    LISTEN_UNLISTEN,
    LISTEN_UNLISTEN_ALL,
} ListenActionKind;
 
typedef struct
{
    ListenActionKind action;
    char        channel[FLEXIBLE_ARRAY_MEMBER]; /* nul-terminated string */
} ListenAction;
 
typedef struct ActionList
{
    int         nestingLevel;   /* current transaction nesting depth */
    List       *actions;        /* list of ListenAction structs */
    struct ActionList *upper;   /* details for upper transaction levels */
} ActionList;
 
static ActionList *pendingActions = NULL;
 
/*
 * Hash table recording the final listen/unlisten intent per channel for
 * the current transaction.  Key is channel name, value is PENDING_LISTEN or
 * PENDING_UNLISTEN.  This keeps critical commit/abort processing to one step
 * per channel instead of replaying every action.  This is built from the
 * pendingActions list by PreCommit_Notify, then used by AtCommit_Notify or
 * AtAbort_Notify.
 */
typedef enum
{
    PENDING_LISTEN,
    PENDING_UNLISTEN,
} PendingListenAction;
 
typedef struct PendingListenEntry
{
    char        channel[NAMEDATALEN];   /* hash key */
    PendingListenAction action; /* which action should we perform? */
} PendingListenEntry;
 
static HTAB *pendingListenActions = NULL;
 
/*
 * State for outbound notifies consists of a list of all channels+payloads
 * NOTIFYed in the current transaction.  We do not actually perform a NOTIFY
 * until and unless the transaction commits.  pendingNotifies is NULL if no
 * NOTIFYs have been done in the current (sub) transaction.
 *
 * We discard duplicate notify events issued in the same transaction.
 * Hence, in addition to the list proper (which we need to track the order
 * of the events, since we guarantee to deliver them in order), we build a
 * hash table which we can probe to detect duplicates.  Since building the
 * hash table is somewhat expensive, we do so only once we have at least
 * MIN_HASHABLE_NOTIFIES events queued in the current (sub) transaction;
 * before that we just scan the events linearly.
 *
 * The list is kept in CurTransactionContext.  In subtransactions, each
 * subtransaction has its own list in its own CurTransactionContext, but
 * successful subtransactions add their entries to their parent's list.
 * Failed subtransactions simply discard their lists.  Since these lists
 * are independent, there may be notify events in a subtransaction's list
 * that duplicate events in some ancestor (sub) transaction; we get rid of
 * the dups when merging the subtransaction's list into its parent's.
 *
 * Note: the action and notify lists do not interact within a transaction.
 * In particular, if a transaction does NOTIFY and then LISTEN on the same
 * condition name, it will get a self-notify at commit.  This is a bit odd
 * but is consistent with our historical behavior.
 */
typedef struct Notification
{
    uint16      channel_len;    /* length of channel-name string */
    uint16      payload_len;    /* length of payload string */
    /* null-terminated channel name, then null-terminated payload follow */
    char        data[FLEXIBLE_ARRAY_MEMBER];
} Notification;
 
typedef struct NotificationList
{
    int         nestingLevel;   /* current transaction nesting depth */
    List       *events;         /* list of Notification structs */
    HTAB       *hashtab;        /* hash of NotificationHash structs, or NULL */
    List       *uniqueChannelNames; /* unique channel names being notified */
    HTAB       *uniqueChannelHash;  /* hash of unique channel names, or NULL */
    struct NotificationList *upper; /* details for upper transaction levels */
} NotificationList;
 
#define MIN_HASHABLE_NOTIFIES 16    /* threshold to build hashtab */
 
struct NotificationHash
{
    Notification *event;        /* => the actual Notification struct */
};
 
static NotificationList *pendingNotifies = NULL;
 
/*
 * Hash entry in NotificationList.uniqueChannelHash or localChannelTable
 * (both just carry the channel name, with no payload).
 */
typedef struct ChannelName
{
    char        channel[NAMEDATALEN];   /* hash key */
} ChannelName;
 
/*
 * Inbound notifications are initially processed by HandleNotifyInterrupt(),
 * called from inside a signal handler. That just sets the
 * notifyInterruptPending flag and sets the process
 * latch. ProcessNotifyInterrupt() will then be called whenever it's safe to
 * actually deal with the interrupt.
 */
volatile sig_atomic_t notifyInterruptPending = false;
 
/* True if we've registered an on_shmem_exit cleanup */
static bool unlistenExitRegistered = false;
 
/* True if we're currently registered as a listener in asyncQueueControl */
static bool amRegisteredListener = false;
 
/*
 * Queue head positions for direct advancement.
 * These are captured during PreCommit_Notify while holding the heavyweight
 * lock on database 0, ensuring no other backend can insert notifications
 * between them.  SignalBackends uses these to advance idle backends.
 */
static QueuePosition queueHeadBeforeWrite;
static QueuePosition queueHeadAfterWrite;
 
/*
 * Workspace arrays for SignalBackends.  These are preallocated in
 * PreCommit_Notify to avoid needing memory allocation after committing to
 * clog.
 */
static int32 *signalPids = NULL;
static ProcNumber *signalProcnos = NULL;
 
/* have we advanced to a page that's a multiple of QUEUE_CLEANUP_DELAY? */
static bool tryAdvanceTail = false;
 
/* GUC parameters */
bool        Trace_notify = false;
 
/* For 8 KB pages this gives 8 GB of disk space */
int         max_notify_queue_pages = 1048576;
 
/* local function prototypes */
static inline int64 asyncQueuePageDiff(int64 p, int64 q);
static inline void GlobalChannelKeyInit(GlobalChannelKey *key, Oid dboid,
                                        const char *channel);
static dshash_hash globalChannelTableHash(const void *key, size_t size,
                                          void *arg);
static void initGlobalChannelTable(void);
static void initLocalChannelTable(void);
static void queue_listen(ListenActionKind action, const char *channel);
static void Async_UnlistenOnExit(int code, Datum arg);
static void BecomeRegisteredListener(void);
static void PrepareTableEntriesForListen(const char *channel);
static void PrepareTableEntriesForUnlisten(const char *channel);
static void PrepareTableEntriesForUnlistenAll(void);
static void RemoveListenerFromChannel(GlobalChannelEntry **entry_ptr,
                                      ListenerEntry *listeners,
                                      int idx);
static void ApplyPendingListenActions(bool isCommit);
static void CleanupListenersOnExit(void);
static bool IsListeningOn(const char *channel);
static void asyncQueueUnregister(void);
static bool asyncQueueIsFull(void);
static bool asyncQueueAdvance(volatile QueuePosition *position, int entryLength);
static void asyncQueueNotificationToEntry(Notification *n, AsyncQueueEntry *qe);
static ListCell *asyncQueueAddEntries(ListCell *nextNotify);
static double asyncQueueUsage(void);
static void asyncQueueFillWarning(void);
static void SignalBackends(void);
static void asyncQueueReadAllNotifications(void);
static bool asyncQueueProcessPageEntries(QueuePosition *current,
                                         QueuePosition stop,
                                         Snapshot snapshot);
static void asyncQueueAdvanceTail(void);
static void ProcessIncomingNotify(bool flush);
static bool AsyncExistsPendingNotify(Notification *n);
static void AddEventToPendingNotifies(Notification *n);
static uint32 notification_hash(const void *key, Size keysize);
static int  notification_match(const void *key1, const void *key2, Size keysize);
static void ClearPendingActionsAndNotifies(void);
 
static int
asyncQueueErrdetailForIoError(const void *opaque_data)
{
    const QueuePosition *position = opaque_data;
 
    return errdetail("Could not access async queue at page %" PRId64 ", offset %d.",
                     position->page, position->offset);
}
 
/*
 * Compute the difference between two queue page numbers.
 * Previously this function accounted for a wraparound.
 */
static inline int64
asyncQueuePageDiff(int64 p, int64 q)
{
    return p - q;
}
 
/*
 * Determines whether p precedes q.
 * Previously this function accounted for a wraparound.
 */
static inline bool
asyncQueuePagePrecedes(int64 p, int64 q)
{
    return p < q;
}
 
/*
 * GlobalChannelKeyInit
 *      Prepare a global channel table key for hashing.
 */
static inline void
GlobalChannelKeyInit(GlobalChannelKey *key, Oid dboid, const char *channel)
{
    memset(key, 0, sizeof(GlobalChannelKey));
    key->dboid = dboid;
    strlcpy(key->channel, channel, NAMEDATALEN);
}
 
/*
 * globalChannelTableHash
 *      Hash function for global channel table keys.
 */
static dshash_hash
globalChannelTableHash(const void *key, size_t size, void *arg)
{
    const GlobalChannelKey *k = (const GlobalChannelKey *) key;
    dshash_hash h;
 
    h = DatumGetUInt32(hash_uint32(k->dboid));
    h ^= DatumGetUInt32(hash_any((const unsigned char *) k->channel,
                                 strnlen(k->channel, NAMEDATALEN)));
 
    return h;
}
 
/* parameters for the global channel table */
static const dshash_parameters globalChannelTableDSHParams = {
    sizeof(GlobalChannelKey),
    sizeof(GlobalChannelEntry),
    dshash_memcmp,
    globalChannelTableHash,
    dshash_memcpy,
    LWTRANCHE_NOTIFY_CHANNEL_HASH
};
 
/*
 * initGlobalChannelTable
 *      Lazy initialization of the global channel table.
 */
static void
initGlobalChannelTable(void)
{
    MemoryContext oldcontext;
 
    /* Quick exit if we already did this */
    if (asyncQueueControl->globalChannelTableDSH != DSHASH_HANDLE_INVALID &&
        globalChannelTable != NULL)
        return;
 
    /* Otherwise, use a lock to ensure only one process creates the table */
    LWLockAcquire(NotifyQueueLock, LW_EXCLUSIVE);
 
    /* Be sure any local memory allocated by DSA routines is persistent */
    oldcontext = MemoryContextSwitchTo(TopMemoryContext);
 
    if (asyncQueueControl->globalChannelTableDSH == DSHASH_HANDLE_INVALID)
    {
        /* Initialize dynamic shared hash table for global channels */
        globalChannelDSA = dsa_create(LWTRANCHE_NOTIFY_CHANNEL_HASH);
        dsa_pin(globalChannelDSA);
        dsa_pin_mapping(globalChannelDSA);
        globalChannelTable = dshash_create(globalChannelDSA,
                                           &globalChannelTableDSHParams,
                                           NULL);
 
        /* Store handles in shared memory for other backends to use */
        asyncQueueControl->globalChannelTableDSA = dsa_get_handle(globalChannelDSA);
        asyncQueueControl->globalChannelTableDSH =
            dshash_get_hash_table_handle(globalChannelTable);
    }
    else if (!globalChannelTable)
    {
        /* Attach to existing dynamic shared hash table */
        globalChannelDSA = dsa_attach(asyncQueueControl->globalChannelTableDSA);
        dsa_pin_mapping(globalChannelDSA);
        globalChannelTable = dshash_attach(globalChannelDSA,
                                           &globalChannelTableDSHParams,
                                           asyncQueueControl->globalChannelTableDSH,
                                           NULL);
    }
 
    MemoryContextSwitchTo(oldcontext);
    LWLockRelease(NotifyQueueLock);
}
 
/*
 * initLocalChannelTable
 *      Lazy initialization of the local channel table.
 *      Once created, this table lasts for the life of the session.
 */
static void
initLocalChannelTable(void)
{
    HASHCTL     hash_ctl;
 
    /* Quick exit if we already did this */
    if (localChannelTable != NULL)
        return;
 
    /* Initialize local hash table for this backend's listened channels */
    hash_ctl.keysize = NAMEDATALEN;
    hash_ctl.entrysize = sizeof(ChannelName);
 
    localChannelTable =
        hash_create("Local Listen Channels",
                    64,
                    &hash_ctl,
                    HASH_ELEM | HASH_STRINGS);
}
 
/*
 * initPendingListenActions
 *      Lazy initialization of the pending listen actions hash table.
 *      This is allocated in CurTransactionContext during PreCommit_Notify,
 *      and destroyed at transaction end.
 */
static void
initPendingListenActions(void)
{
    HASHCTL     hash_ctl;
 
    if (pendingListenActions != NULL)
        return;
 
    hash_ctl.keysize = NAMEDATALEN;
    hash_ctl.entrysize = sizeof(PendingListenEntry);
    hash_ctl.hcxt = CurTransactionContext;
 
    pendingListenActions =
        hash_create("Pending Listen Actions",
                    list_length(pendingActions->actions),
                    &hash_ctl,
                    HASH_ELEM | HASH_STRINGS | HASH_CONTEXT);
}
 
/*
 * Register our shared memory needs
 */
static void
AsyncShmemRequest(void *arg)
{
    Size        size;
 
    size = mul_size(MaxBackends, sizeof(QueueBackendStatus));
    size = add_size(size, offsetof(AsyncQueueControl, backend));
 
    ShmemRequestStruct(.name = "Async Queue Control",
                       .size = size,
                       .ptr = (void **) &asyncQueueControl,
        );
 
    SimpleLruRequest(.desc = &NotifySlruDesc,
                     .name = "notify",
                     .Dir = "pg_notify",
 
    /* long segment names are used in order to avoid wraparound */
                     .long_segment_names = true,
 
                     .nslots = notify_buffers,
 
                     .sync_handler = SYNC_HANDLER_NONE,
                     .PagePrecedes = asyncQueuePagePrecedes,
                     .errdetail_for_io_error = asyncQueueErrdetailForIoError,
 
                     .buffer_tranche_id = LWTRANCHE_NOTIFY_BUFFER,
                     .bank_tranche_id = LWTRANCHE_NOTIFY_SLRU,
        );
}
 
static void
AsyncShmemInit(void *arg)
{
    SET_QUEUE_POS(QUEUE_HEAD, 0, 0);
    SET_QUEUE_POS(QUEUE_TAIL, 0, 0);
    QUEUE_STOP_PAGE = 0;
    QUEUE_FIRST_LISTENER = INVALID_PROC_NUMBER;
    asyncQueueControl->lastQueueFillWarn = 0;
    asyncQueueControl->globalChannelTableDSA = DSA_HANDLE_INVALID;
    asyncQueueControl->globalChannelTableDSH = DSHASH_HANDLE_INVALID;
    for (int i = 0; i < MaxBackends; i++)
    {
        QUEUE_BACKEND_PID(i) = InvalidPid;
        QUEUE_BACKEND_DBOID(i) = InvalidOid;
        QUEUE_NEXT_LISTENER(i) = INVALID_PROC_NUMBER;
        SET_QUEUE_POS(QUEUE_BACKEND_POS(i), 0, 0);
        QUEUE_BACKEND_WAKEUP_PENDING(i) = false;
        QUEUE_BACKEND_IS_ADVANCING(i) = false;
    }
 
    /*
     * During start or reboot, clean out the pg_notify directory.
     */
    (void) SlruScanDirectory(NotifyCtl, SlruScanDirCbDeleteAll, NULL);
}
 
 
/*
 * pg_notify -
 *    SQL function to send a notification event
 */
Datum
pg_notify(PG_FUNCTION_ARGS)
{
    const char *channel;
    const char *payload;
 
    if (PG_ARGISNULL(0))
        channel = "";
    else
        channel = text_to_cstring(PG_GETARG_TEXT_PP(0));
 
    if (PG_ARGISNULL(1))
        payload = "";
    else
        payload = text_to_cstring(PG_GETARG_TEXT_PP(1));
 
    /* For NOTIFY as a statement, this is checked in ProcessUtility */
    PreventCommandDuringRecovery("NOTIFY");
 
    Async_Notify(channel, payload);
 
    PG_RETURN_VOID();
}
 
 
/*
 * Async_Notify
 *
 *      This is executed by the SQL notify command.
 *
 *      Adds the message to the list of pending notifies.
 *      Actual notification happens during transaction commit.
 *      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 */
void
Async_Notify(const char *channel, const char *payload)
{
    int         my_level = GetCurrentTransactionNestLevel();
    size_t      channel_len;
    size_t      payload_len;
    Notification *n;
    MemoryContext oldcontext;
 
    if (IsParallelWorker())
        elog(ERROR, "cannot send notifications from a parallel worker");
 
    if (Trace_notify)
        elog(DEBUG1, "Async_Notify(%s)", channel);
 
    channel_len = channel ? strlen(channel) : 0;
    payload_len = payload ? strlen(payload) : 0;
 
    /* a channel name must be specified */
    if (channel_len == 0)
        ereport(ERROR,
                (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
                 errmsg("channel name cannot be empty")));
 
    /* enforce length limits */
    if (channel_len >= NAMEDATALEN)
        ereport(ERROR,
                (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
                 errmsg("channel name too long")));
 
    if (payload_len >= NOTIFY_PAYLOAD_MAX_LENGTH)
        ereport(ERROR,
                (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
                 errmsg("payload string too long")));
 
    /*
     * We must construct the Notification entry, even if we end up not using
     * it, in order to compare it cheaply to existing list entries.
     *
     * The notification list needs to live until end of transaction, so store
     * it in the transaction context.
     */
    oldcontext = MemoryContextSwitchTo(CurTransactionContext);
 
    n = (Notification *) palloc(offsetof(Notification, data) +
                                channel_len + payload_len + 2);
    n->channel_len = channel_len;
    n->payload_len = payload_len;
    strcpy(n->data, channel);
    if (payload)
        strcpy(n->data + channel_len + 1, payload);
    else
        n->data[channel_len + 1] = '\0';
 
    if (pendingNotifies == NULL || my_level > pendingNotifies->nestingLevel)
    {
        NotificationList *notifies;
 
        /*
         * First notify event in current (sub)xact. Note that we allocate the
         * NotificationList in TopTransactionContext; the nestingLevel might
         * get changed later by AtSubCommit_Notify.
         */
        notifies = (NotificationList *)
            MemoryContextAlloc(TopTransactionContext,
                               sizeof(NotificationList));
        notifies->nestingLevel = my_level;
        notifies->events = list_make1(n);
        /* We certainly don't need a hashtable yet */
        notifies->hashtab = NULL;
        /* We won't build uniqueChannelNames/Hash till later, either */
        notifies->uniqueChannelNames = NIL;
        notifies->uniqueChannelHash = NULL;
        notifies->upper = pendingNotifies;
        pendingNotifies = notifies;
    }
    else
    {
        /* Now check for duplicates */
        if (AsyncExistsPendingNotify(n))
        {
            /* It's a dup, so forget it */
            pfree(n);
            MemoryContextSwitchTo(oldcontext);
            return;
        }
 
        /* Append more events to existing list */
        AddEventToPendingNotifies(n);
    }
 
    MemoryContextSwitchTo(oldcontext);
}
 
/*
 * queue_listen
 *      Common code for listen, unlisten, unlisten all commands.
 *
 *      Adds the request to the list of pending actions.
 *      Actual update of localChannelTable and globalChannelTable happens during
 *      PreCommit_Notify, with staged changes committed in AtCommit_Notify.
 */
static void
queue_listen(ListenActionKind action, const char *channel)
{
    MemoryContext oldcontext;
    ListenAction *actrec;
    int         my_level = GetCurrentTransactionNestLevel();
 
    /*
     * Unlike Async_Notify, we don't try to collapse out duplicates here. We
     * keep the ordered list to preserve interactions like UNLISTEN ALL; the
     * final per-channel intent is computed during PreCommit_Notify.
     */
    oldcontext = MemoryContextSwitchTo(CurTransactionContext);
 
    /* space for terminating null is included in sizeof(ListenAction) */
    actrec = (ListenAction *) palloc(offsetof(ListenAction, channel) +
                                     strlen(channel) + 1);
    actrec->action = action;
    strcpy(actrec->channel, channel);
 
    if (pendingActions == NULL || my_level > pendingActions->nestingLevel)
    {
        ActionList *actions;
 
        /*
         * First action in current sub(xact). Note that we allocate the
         * ActionList in TopTransactionContext; the nestingLevel might get
         * changed later by AtSubCommit_Notify.
         */
        actions = (ActionList *)
            MemoryContextAlloc(TopTransactionContext, sizeof(ActionList));
        actions->nestingLevel = my_level;
        actions->actions = list_make1(actrec);
        actions->upper = pendingActions;
        pendingActions = actions;
    }
    else
        pendingActions->actions = lappend(pendingActions->actions, actrec);
 
    MemoryContextSwitchTo(oldcontext);
}
 
/*
 * Async_Listen
 *
 *      This is executed by the SQL listen command.
 */
void
Async_Listen(const char *channel)
{
    if (Trace_notify)
        elog(DEBUG1, "Async_Listen(%s,%d)", channel, MyProcPid);
 
    queue_listen(LISTEN_LISTEN, channel);
}
 
/*
 * Async_Unlisten
 *
 *      This is executed by the SQL unlisten command.
 */
void
Async_Unlisten(const char *channel)
{
    if (Trace_notify)
        elog(DEBUG1, "Async_Unlisten(%s,%d)", channel, MyProcPid);
 
    /* If we couldn't possibly be listening, no need to queue anything */
    if (pendingActions == NULL && !unlistenExitRegistered)
        return;
 
    queue_listen(LISTEN_UNLISTEN, channel);
}
 
/*
 * Async_UnlistenAll
 *
 *      This is invoked by UNLISTEN * command, and also at backend exit.
 */
void
Async_UnlistenAll(void)
{
    if (Trace_notify)
        elog(DEBUG1, "Async_UnlistenAll(%d)", MyProcPid);
 
    /* If we couldn't possibly be listening, no need to queue anything */
    if (pendingActions == NULL && !unlistenExitRegistered)
        return;
 
    queue_listen(LISTEN_UNLISTEN_ALL, "");
}
 
/*
 * SQL function: return a set of the channel names this backend is actively
 * listening to.
 *
 * Note: this coding relies on the fact that the localChannelTable cannot
 * change within a transaction.
 */
Datum
pg_listening_channels(PG_FUNCTION_ARGS)
{
    FuncCallContext *funcctx;
    HASH_SEQ_STATUS *status;
 
    /* stuff done only on the first call of the function */
    if (SRF_IS_FIRSTCALL())
    {
        /* create a function context for cross-call persistence */
        funcctx = SRF_FIRSTCALL_INIT();
 
        /* Initialize hash table iteration if we have any channels */
        if (localChannelTable != NULL)
        {
            MemoryContext oldcontext;
 
            oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx);
            status = (HASH_SEQ_STATUS *) palloc(sizeof(HASH_SEQ_STATUS));
            hash_seq_init(status, localChannelTable);
            funcctx->user_fctx = status;
            MemoryContextSwitchTo(oldcontext);
        }
        else
        {
            funcctx->user_fctx = NULL;
        }
    }
 
    /* stuff done on every call of the function */
    funcctx = SRF_PERCALL_SETUP();
    status = (HASH_SEQ_STATUS *) funcctx->user_fctx;
 
    if (status != NULL)
    {
        ChannelName *entry;
 
        entry = (ChannelName *) hash_seq_search(status);
        if (entry != NULL)
            SRF_RETURN_NEXT(funcctx, CStringGetTextDatum(entry->channel));
    }
 
    SRF_RETURN_DONE(funcctx);
}
 
/*
 * Async_UnlistenOnExit
 *
 * This is executed at backend exit if we have done any LISTENs in this
 * backend.  It might not be necessary anymore, if the user UNLISTENed
 * everything, but we don't try to detect that case.
 */
static void
Async_UnlistenOnExit(int code, Datum arg)
{
    CleanupListenersOnExit();
    asyncQueueUnregister();
}
 
/*
 * AtPrepare_Notify
 *
 *      This is called at the prepare phase of a two-phase
 *      transaction.  Save the state for possible commit later.
 */
void
AtPrepare_Notify(void)
{
    /* It's not allowed to have any pending LISTEN/UNLISTEN/NOTIFY actions */
    if (pendingActions || pendingNotifies)
        ereport(ERROR,
                (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
                 errmsg("cannot PREPARE a transaction that has executed LISTEN, UNLISTEN, or NOTIFY")));
}
 
/*
 * PreCommit_Notify
 *
 *      This is called at transaction commit, before actually committing to
 *      clog.
 *
 *      If there are pending LISTEN actions, make sure we are listed in the
 *      shared-memory listener array.  This must happen before commit to
 *      ensure we don't miss any notifies from transactions that commit
 *      just after ours.
 *
 *      If there are outbound notify requests in the pendingNotifies list,
 *      add them to the global queue.  We do that before commit so that
 *      we can still throw error if we run out of queue space.
 */
void
PreCommit_Notify(void)
{
    ListCell   *p;
 
    if (!pendingActions && !pendingNotifies)
        return;                 /* no relevant statements in this xact */
 
    if (Trace_notify)
        elog(DEBUG1, "PreCommit_Notify");
 
    /* Preflight for any pending listen/unlisten actions */
    initGlobalChannelTable();
 
    if (pendingActions != NULL)
    {
        /* Ensure we have a local channel table */
        initLocalChannelTable();
        /* Create pendingListenActions hash table for this transaction */
        initPendingListenActions();
 
        /* Stage all the actions this transaction wants to perform */
        foreach(p, pendingActions->actions)
        {
            ListenAction *actrec = (ListenAction *) lfirst(p);
 
            switch (actrec->action)
            {
                case LISTEN_LISTEN:
                    BecomeRegisteredListener();
                    PrepareTableEntriesForListen(actrec->channel);
                    break;
                case LISTEN_UNLISTEN:
                    PrepareTableEntriesForUnlisten(actrec->channel);
                    break;
                case LISTEN_UNLISTEN_ALL:
                    PrepareTableEntriesForUnlistenAll();
                    break;
            }
        }
    }
 
    /* Queue any pending notifies (must happen after the above) */
    if (pendingNotifies)
    {
        ListCell   *nextNotify;
        bool        firstIteration = true;
 
        /*
         * Build list of unique channel names being notified for use by
         * SignalBackends().
         *
         * If uniqueChannelHash is available, use it to efficiently get the
         * unique channels.  Otherwise, fall back to the O(N^2) approach.
         */
        pendingNotifies->uniqueChannelNames = NIL;
        if (pendingNotifies->uniqueChannelHash != NULL)
        {
            HASH_SEQ_STATUS status;
            ChannelName *channelEntry;
 
            hash_seq_init(&status, pendingNotifies->uniqueChannelHash);
            while ((channelEntry = (ChannelName *) hash_seq_search(&status)) != NULL)
                pendingNotifies->uniqueChannelNames =
                    lappend(pendingNotifies->uniqueChannelNames,
                            channelEntry->channel);
        }
        else
        {
            /* O(N^2) approach is better for small number of notifications */
            foreach_ptr(Notification, n, pendingNotifies->events)
            {
                char       *channel = n->data;
                bool        found = false;
 
                /* Name present in list? */
                foreach_ptr(char, oldchan, pendingNotifies->uniqueChannelNames)
                {
                    if (strcmp(oldchan, channel) == 0)
                    {
                        found = true;
                        break;
                    }
                }
                /* Add if not already in list */
                if (!found)
                    pendingNotifies->uniqueChannelNames =
                        lappend(pendingNotifies->uniqueChannelNames,
                                channel);
            }
        }
 
        /* Preallocate workspace that will be needed by SignalBackends() */
        if (signalPids == NULL)
            signalPids = MemoryContextAlloc(TopMemoryContext,
                                            MaxBackends * sizeof(int32));
 
        if (signalProcnos == NULL)
            signalProcnos = MemoryContextAlloc(TopMemoryContext,
                                               MaxBackends * sizeof(ProcNumber));
 
        /*
         * Make sure that we have an XID assigned to the current transaction.
         * GetCurrentTransactionId is cheap if we already have an XID, but not
         * so cheap if we don't, and we'd prefer not to do that work while
         * holding NotifyQueueLock.
         */
        (void) GetCurrentTransactionId();
 
        /*
         * Serialize writers by acquiring a special lock that we hold till
         * after commit.  This ensures that queue entries appear in commit
         * order, and in particular that there are never uncommitted queue
         * entries ahead of committed ones, so an uncommitted transaction
         * can't block delivery of deliverable notifications.
         *
         * We use a heavyweight lock so that it'll automatically be released
         * after either commit or abort.  This also allows deadlocks to be
         * detected, though really a deadlock shouldn't be possible here.
         *
         * The lock is on "database 0", which is pretty ugly but it doesn't
         * seem worth inventing a special locktag category just for this.
         * (Historical note: before PG 9.0, a similar lock on "database 0" was
         * used by the flatfiles mechanism.)
         */
        LockSharedObject(DatabaseRelationId, InvalidOid, 0,
                         AccessExclusiveLock);
 
        /*
         * For the direct advancement optimization in SignalBackends(), we
         * need to ensure that no other backend can insert queue entries
         * between queueHeadBeforeWrite and queueHeadAfterWrite.  The
         * heavyweight lock above provides this guarantee, since it serializes
         * all writers.
         *
         * Note: if the heavyweight lock were ever removed for scalability
         * reasons, we could achieve the same guarantee by holding
         * NotifyQueueLock in EXCLUSIVE mode across all our insertions, rather
         * than releasing and reacquiring it for each page as we do below.
         */
 
        /* Initialize values to a safe default in case list is empty */
        SET_QUEUE_POS(queueHeadBeforeWrite, 0, 0);
        SET_QUEUE_POS(queueHeadAfterWrite, 0, 0);
 
        /* Now push the notifications into the queue */
        nextNotify = list_head(pendingNotifies->events);
        while (nextNotify != NULL)
        {
            /*
             * Add the pending notifications to the queue.  We acquire and
             * release NotifyQueueLock once per page, which might be overkill
             * but it does allow readers to get in while we're doing this.
             *
             * A full queue is very uncommon and should really not happen,
             * given that we have so much space available in the SLRU pages.
             * Nevertheless we need to deal with this possibility. Note that
             * when we get here we are in the process of committing our
             * transaction, but we have not yet committed to clog, so at this
             * point in time we can still roll the transaction back.
             */
            LWLockAcquire(NotifyQueueLock, LW_EXCLUSIVE);
            if (firstIteration)
            {
                queueHeadBeforeWrite = QUEUE_HEAD;
                firstIteration = false;
            }
            asyncQueueFillWarning();
            if (asyncQueueIsFull())
                ereport(ERROR,
                        (errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
                         errmsg("too many notifications in the NOTIFY queue")));
            nextNotify = asyncQueueAddEntries(nextNotify);
            queueHeadAfterWrite = QUEUE_HEAD;
            LWLockRelease(NotifyQueueLock);
        }
 
        /* Note that we don't clear pendingNotifies; AtCommit_Notify will. */
    }
}
 
/*
 * AtCommit_Notify
 *
 *      This is called at transaction commit, after committing to clog.
 *
 *      Apply pending listen/unlisten changes and clear transaction-local state.
 *
 *      If we issued any notifications in the transaction, send signals to
 *      listening backends (possibly including ourselves) to process them.
 *      Also, if we filled enough queue pages with new notifies, try to
 *      advance the queue tail pointer.
 */
void
AtCommit_Notify(void)
{
    /*
     * Allow transactions that have not executed LISTEN/UNLISTEN/NOTIFY to
     * return as soon as possible
     */
    if (!pendingActions && !pendingNotifies)
        return;
 
    if (Trace_notify)
        elog(DEBUG1, "AtCommit_Notify");
 
    /* Apply staged listen/unlisten changes */
    ApplyPendingListenActions(true);
 
    /* If no longer listening to anything, get out of listener array */
    if (amRegisteredListener && LocalChannelTableIsEmpty())
        asyncQueueUnregister();
 
    /*
     * Send signals to listening backends.  We need do this only if there are
     * pending notifies, which were previously added to the shared queue by
     * PreCommit_Notify().
     */
    if (pendingNotifies != NULL)
        SignalBackends();
 
    /*
     * If it's time to try to advance the global tail pointer, do that.
     *
     * (It might seem odd to do this in the sender, when more than likely the
     * listeners won't yet have read the messages we just sent.  However,
     * there's less contention if only the sender does it, and there is little
     * need for urgency in advancing the global tail.  So this typically will
     * be clearing out messages that were sent some time ago.)
     */
    if (tryAdvanceTail)
    {
        tryAdvanceTail = false;
        asyncQueueAdvanceTail();
    }
 
    /* And clean up */
    ClearPendingActionsAndNotifies();
}
 
/*
 * BecomeRegisteredListener --- subroutine for PreCommit_Notify
 *
 * This function must make sure we are ready to catch any incoming messages.
 */
static void
BecomeRegisteredListener(void)
{
    QueuePosition head;
    QueuePosition max;
    ProcNumber  prevListener;
 
    /*
     * Nothing to do if we are already listening to something, nor if we
     * already ran this routine in this transaction.
     */
    if (amRegisteredListener)
        return;
 
    if (Trace_notify)
        elog(DEBUG1, "BecomeRegisteredListener(%d)", MyProcPid);
 
    /*
     * Before registering, make sure we will unlisten before dying. (Note:
     * this action does not get undone if we abort later.)
     */
    if (!unlistenExitRegistered)
    {
        before_shmem_exit(Async_UnlistenOnExit, 0);
        unlistenExitRegistered = true;
    }
 
    /*
     * This is our first LISTEN, so establish our pointer.
     *
     * We set our pointer to the global tail pointer and then move it forward
     * over already-committed notifications.  This ensures we cannot miss any
     * not-yet-committed notifications.  We might get a few more but that
     * doesn't hurt.
     *
     * In some scenarios there might be a lot of committed notifications that
     * have not yet been pruned away (because some backend is being lazy about
     * reading them).  To reduce our startup time, we can look at other
     * backends and adopt the maximum "pos" pointer of any backend that's in
     * our database; any notifications it's already advanced over are surely
     * committed and need not be re-examined by us.  (We must consider only
     * backends connected to our DB, because others will not have bothered to
     * check committed-ness of notifications in our DB.)
     *
     * We need exclusive lock here so we can look at other backends' entries
     * and manipulate the list links.
     */
    LWLockAcquire(NotifyQueueLock, LW_EXCLUSIVE);
    head = QUEUE_HEAD;
    max = QUEUE_TAIL;
    prevListener = INVALID_PROC_NUMBER;
    for (ProcNumber i = QUEUE_FIRST_LISTENER; i != INVALID_PROC_NUMBER; i = QUEUE_NEXT_LISTENER(i))
    {
        if (QUEUE_BACKEND_DBOID(i) == MyDatabaseId)
            max = QUEUE_POS_MAX(max, QUEUE_BACKEND_POS(i));
        /* Also find last listening backend before this one */
        if (i < MyProcNumber)
            prevListener = i;
    }
    QUEUE_BACKEND_POS(MyProcNumber) = max;
    QUEUE_BACKEND_PID(MyProcNumber) = MyProcPid;
    QUEUE_BACKEND_DBOID(MyProcNumber) = MyDatabaseId;
    QUEUE_BACKEND_WAKEUP_PENDING(MyProcNumber) = false;
    QUEUE_BACKEND_IS_ADVANCING(MyProcNumber) = false;
    /* Insert backend into list of listeners at correct position */
    if (prevListener != INVALID_PROC_NUMBER)
    {
        QUEUE_NEXT_LISTENER(MyProcNumber) = QUEUE_NEXT_LISTENER(prevListener);
        QUEUE_NEXT_LISTENER(prevListener) = MyProcNumber;
    }
    else
    {
        QUEUE_NEXT_LISTENER(MyProcNumber) = QUEUE_FIRST_LISTENER;
        QUEUE_FIRST_LISTENER = MyProcNumber;
    }
    LWLockRelease(NotifyQueueLock);
 
    /* Now we are listed in the global array, so remember we're listening */
    amRegisteredListener = true;
 
    /*
     * Try to move our pointer forward as far as possible.  This will skip
     * over already-committed notifications, which we want to do because they
     * might be quite stale.  Note that we are not yet listening on anything,
     * so we won't deliver such notifications to our frontend.  Also, although
     * our transaction might have executed NOTIFY, those message(s) aren't
     * queued yet so we won't skip them here.
     */
    if (!QUEUE_POS_EQUAL(max, head))
        asyncQueueReadAllNotifications();
}
 
/*
 * PrepareTableEntriesForListen --- subroutine for PreCommit_Notify
 *
 * Prepare a LISTEN by recording it in pendingListenActions, pre-allocating
 * an entry in localChannelTable, and pre-allocating an entry in the shared
 * globalChannelTable with removeOnAbort set.  AtCommit_Notify will clear
 * removeOnAbort; abort processing will remove entries still marked so.
 */
static void
PrepareTableEntriesForListen(const char *channel)
{
    GlobalChannelKey key;
    GlobalChannelEntry *entry;
    bool        found;
    ListenerEntry *listeners;
    PendingListenEntry *pending;
 
    /*
     * Record in local pending hash that we want to LISTEN, overwriting any
     * earlier attempt to UNLISTEN.
     */
    pending = (PendingListenEntry *)
        hash_search(pendingListenActions, channel, HASH_ENTER, NULL);
    pending->action = PENDING_LISTEN;
 
    /*
     * Ensure that there is an entry for the channel in localChannelTable.
     * (Should this fail, we can just roll back.)  If the transaction fails
     * after this point, we will remove the entry if appropriate during
     * ApplyPendingListenActions.  Note that this entry allows IsListeningOn()
     * to return TRUE; we assume nothing is going to consult that before
     * AtCommit_Notify/AtAbort_Notify.  However, if later actions attempt to
     * UNLISTEN this channel or UNLISTEN *, we need to have the local entry
     * present to ensure they do the right things; see
     * PrepareTableEntriesForUnlisten and PrepareTableEntriesForUnlistenAll.
     */
    (void) hash_search(localChannelTable, channel, HASH_ENTER, NULL);
 
    /* Pre-allocate entry in shared globalChannelTable */
    GlobalChannelKeyInit(&key, MyDatabaseId, channel);
    entry = dshash_find_or_insert(globalChannelTable, &key, &found);
 
    if (!found)
    {
        /* New channel entry, so initialize it to a safe state */
        entry->listenersArray = InvalidDsaPointer;
        entry->numListeners = 0;
        entry->allocatedListeners = 0;
    }
 
    /*
     * Create listenersArray if entry doesn't have one.  It's tempting to fold
     * this into the !found case, but this coding allows us to cope in case
     * dsa_allocate() failed in an earlier attempt.
     */
    if (!DsaPointerIsValid(entry->listenersArray))
    {
        entry->listenersArray = dsa_allocate(globalChannelDSA,
                                             sizeof(ListenerEntry) * INITIAL_LISTENERS_ARRAY_SIZE);
        entry->allocatedListeners = INITIAL_LISTENERS_ARRAY_SIZE;
    }
 
    listeners = (ListenerEntry *)
        dsa_get_address(globalChannelDSA, entry->listenersArray);
 
    /*
     * Check if we already have a ListenerEntry (possibly from earlier in this
     * transaction)
     */
    for (int i = 0; i < entry->numListeners; i++)
    {
        if (listeners[i].procNo == MyProcNumber)
        {
            /* Already have an entry; leave removeOnAbort as-is */
            dshash_release_lock(globalChannelTable, entry);
            return;
        }
    }
 
    /* Need to add a new entry; grow array if necessary */
    if (entry->numListeners >= entry->allocatedListeners)
    {
        int         new_size = entry->allocatedListeners * 2;
        dsa_pointer old_array = entry->listenersArray;
        dsa_pointer new_array = dsa_allocate(globalChannelDSA,
                                             sizeof(ListenerEntry) * new_size);
        ListenerEntry *new_listeners = (ListenerEntry *) dsa_get_address(globalChannelDSA, new_array);
 
        memcpy(new_listeners, listeners, sizeof(ListenerEntry) * entry->numListeners);
        entry->listenersArray = new_array;
        entry->allocatedListeners = new_size;
        dsa_free(globalChannelDSA, old_array);
        listeners = new_listeners;
    }
 
    listeners[entry->numListeners].procNo = MyProcNumber;
    listeners[entry->numListeners].removeOnAbort = true;
    entry->numListeners++;
 
    dshash_release_lock(globalChannelTable, entry);
}
 
/*
 * PrepareTableEntriesForUnlisten --- subroutine for PreCommit_Notify
 *
 * Prepare an UNLISTEN by recording it in pendingListenActions, but only if
 * we're currently listening (committed or staged).  We don't touch
 * globalChannelTable yet - the listener keeps receiving signals until
 * commit, when the entry is removed.
 */
static void
PrepareTableEntriesForUnlisten(const char *channel)
{
    PendingListenEntry *pending;
 
    /*
     * If the channel name is not in localChannelTable, then we are neither
     * listening on it nor preparing to listen on it, so we don't need to
     * record an UNLISTEN action.
     */
    Assert(localChannelTable != NULL);
    if (hash_search(localChannelTable, channel, HASH_FIND, NULL) == NULL)
        return;
 
    /*
     * Record in local pending hash that we want to UNLISTEN, overwriting any
     * earlier attempt to LISTEN.  Don't touch localChannelTable or
     * globalChannelTable yet - we keep receiving signals until commit.
     */
    pending = (PendingListenEntry *)
        hash_search(pendingListenActions, channel, HASH_ENTER, NULL);
    pending->action = PENDING_UNLISTEN;
}
 
/*
 * PrepareTableEntriesForUnlistenAll --- subroutine for PreCommit_Notify
 *
 * Prepare UNLISTEN * by recording an UNLISTEN for all listened or
 * about-to-be-listened channels in pendingListenActions.
 */
static void
PrepareTableEntriesForUnlistenAll(void)
{
    HASH_SEQ_STATUS seq;
    ChannelName *channelEntry;
    PendingListenEntry *pending;
 
    /*
     * Scan localChannelTable, which will have the names of all channels that
     * we are listening on or have prepared to listen on.  Record an UNLISTEN
     * action for each one, overwriting any earlier attempt to LISTEN.
     */
    hash_seq_init(&seq, localChannelTable);
    while ((channelEntry = (ChannelName *) hash_seq_search(&seq)) != NULL)
    {
        pending = (PendingListenEntry *)
            hash_search(pendingListenActions, channelEntry->channel, HASH_ENTER, NULL);
        pending->action = PENDING_UNLISTEN;
    }
}
 
/*
 * RemoveListenerFromChannel --- remove idx'th listener from global channel entry
 *
 * Decrements numListeners, compacts the array, and frees the entry if empty.
 * Sets *entry_ptr to NULL if the entry was deleted.
 *
 * We could get the listeners pointer from the entry, but all callers
 * already have it at hand.
 */
static void
RemoveListenerFromChannel(GlobalChannelEntry **entry_ptr,
                          ListenerEntry *listeners,
                          int idx)
{
    GlobalChannelEntry *entry = *entry_ptr;
 
    entry->numListeners--;
    if (idx < entry->numListeners)
        memmove(&listeners[idx], &listeners[idx + 1],
                sizeof(ListenerEntry) * (entry->numListeners - idx));
 
    if (entry->numListeners == 0)
    {
        dsa_free(globalChannelDSA, entry->listenersArray);
        dshash_delete_entry(globalChannelTable, entry);
        /* tells caller not to release the entry's lock: */
        *entry_ptr = NULL;
    }
}
 
/*
 * ApplyPendingListenActions
 *
 * Apply, or revert, staged listen/unlisten changes to the local and global
 * hash tables.
 */
static void
ApplyPendingListenActions(bool isCommit)
{
    HASH_SEQ_STATUS seq;
    PendingListenEntry *pending;
 
    /* Quick exit if nothing to do */
    if (pendingListenActions == NULL)
        return;
 
    /* We made a globalChannelTable before building pendingListenActions */
    if (globalChannelTable == NULL)
        elog(PANIC, "global channel table missing post-commit/abort");
 
    /* For each staged action ... */
    hash_seq_init(&seq, pendingListenActions);
    while ((pending = (PendingListenEntry *) hash_seq_search(&seq)) != NULL)
    {
        GlobalChannelKey key;
        GlobalChannelEntry *entry;
        bool        removeLocal = true;
        bool        foundListener = false;
 
        /*
         * Find the global entry for this channel.  If isCommit, it had better
         * exist (it was created in PreCommit).  In an abort, it might not
         * exist, in which case we are not listening and should discard any
         * local entry that PreCommit may have managed to create.
         */
        GlobalChannelKeyInit(&key, MyDatabaseId, pending->channel);
        entry = dshash_find(globalChannelTable, &key, true);
        if (entry != NULL)
        {
            /* Scan entry to find the ListenerEntry for this backend */
            ListenerEntry *listeners;
 
            listeners = (ListenerEntry *)
                dsa_get_address(globalChannelDSA, entry->listenersArray);
 
            for (int i = 0; i < entry->numListeners; i++)
            {
                if (listeners[i].procNo != MyProcNumber)
                    continue;
                foundListener = true;
                if (isCommit)
                {
                    if (pending->action == PENDING_LISTEN)
                    {
                        /*
                         * LISTEN being committed: entry is now permanent.
                         * localChannelTable entry was created during
                         * PreCommit and should be kept.
                         */
                        listeners[i].removeOnAbort = false;
                        removeLocal = false;
                    }
                    else
                    {
                        /*
                         * UNLISTEN being committed: remove pre-allocated
                         * entries from both tables.
                         */
                        RemoveListenerFromChannel(&entry, listeners, i);
                    }
                }
                else
                {
                    /*
                     * Note: this part is reachable only if the transaction
                     * aborts after PreCommit_Notify() has made some
                     * pendingListenActions entries, so it's pretty hard to
                     * test.
                     */
                    if (listeners[i].removeOnAbort)
                    {
                        /*
                         * Staged LISTEN (or LISTEN+UNLISTEN) being aborted,
                         * so remove pre-allocated entries from both tables.
                         */
                        RemoveListenerFromChannel(&entry, listeners, i);
                    }
                    else
                    {
                        /*
                         * Entry predates this transaction, so keep the
                         * localChannelTable entry.
                         */
                        removeLocal = false;
                    }
                }
                break;          /* there shouldn't be another match */
            }
 
            /* We might have already released the entry by removing it */
            if (entry != NULL)
                dshash_release_lock(globalChannelTable, entry);
        }
 
        /*
         * If we're committing a LISTEN action, we should have found a
         * matching ListenerEntry, but otherwise it's okay if we didn't.
         */
        if (isCommit && pending->action == PENDING_LISTEN && !foundListener)
            elog(PANIC, "could not find listener entry for channel \"%s\" backend %d",
                 pending->channel, MyProcNumber);
 
        /*
         * If we did not find a globalChannelTable entry for our backend, or
         * if we are unlistening, remove any localChannelTable entry that may
         * exist.  (Note in particular that this cleans up if we created a
         * localChannelTable entry and then failed while trying to create a
         * globalChannelTable entry.)
         */
        if (removeLocal && localChannelTable != NULL)
            (void) hash_search(localChannelTable, pending->channel,
                               HASH_REMOVE, NULL);
    }
}
 
/*
 * CleanupListenersOnExit --- called from Async_UnlistenOnExit
 *
 *      Remove this backend from all channels in the shared global table.
 */
static void
CleanupListenersOnExit(void)
{
    dshash_seq_status status;
    GlobalChannelEntry *entry;
 
    if (Trace_notify)
        elog(DEBUG1, "CleanupListenersOnExit(%d)", MyProcPid);
 
    /* Clear our local cache (not really necessary, but be consistent) */
    if (localChannelTable != NULL)
    {
        hash_destroy(localChannelTable);
        localChannelTable = NULL;
    }
 
    /* Now remove our entries from the shared globalChannelTable */
    if (globalChannelTable == NULL)
        return;
 
    dshash_seq_init(&status, globalChannelTable, true);
    while ((entry = dshash_seq_next(&status)) != NULL)
    {
        ListenerEntry *listeners;
 
        if (entry->key.dboid != MyDatabaseId)
            continue;           /* not relevant */
 
        listeners = (ListenerEntry *)
            dsa_get_address(globalChannelDSA, entry->listenersArray);
 
        for (int i = 0; i < entry->numListeners; i++)
        {
            if (listeners[i].procNo == MyProcNumber)
            {
                entry->numListeners--;
                if (i < entry->numListeners)
                    memmove(&listeners[i], &listeners[i + 1],
                            sizeof(ListenerEntry) * (entry->numListeners - i));
 
                if (entry->numListeners == 0)
                {
                    dsa_free(globalChannelDSA, entry->listenersArray);
                    dshash_delete_current(&status);
                }
                break;
            }
        }
    }
    dshash_seq_term(&status);
}
 
/*
 * Test whether we are actively listening on the given channel name.
 *
 * Note: this function is executed for every notification found in the queue.
 */
static bool
IsListeningOn(const char *channel)
{
    if (localChannelTable == NULL)
        return false;
 
    return (hash_search(localChannelTable, channel, HASH_FIND, NULL) != NULL);
}
 
/*
 * Remove our entry from the listeners array when we are no longer listening
 * on any channel.  NB: must not fail if we're already not listening.
 */
static void
asyncQueueUnregister(void)
{
    Assert(LocalChannelTableIsEmpty()); /* else caller error */
 
    if (!amRegisteredListener)  /* nothing to do */
        return;
 
    /*
     * Need exclusive lock here to manipulate list links.
     */
    LWLockAcquire(NotifyQueueLock, LW_EXCLUSIVE);
    /* Mark our entry as invalid */
    QUEUE_BACKEND_PID(MyProcNumber) = InvalidPid;
    QUEUE_BACKEND_DBOID(MyProcNumber) = InvalidOid;
    QUEUE_BACKEND_WAKEUP_PENDING(MyProcNumber) = false;
    QUEUE_BACKEND_IS_ADVANCING(MyProcNumber) = false;
    /* and remove it from the list */
    if (QUEUE_FIRST_LISTENER == MyProcNumber)
        QUEUE_FIRST_LISTENER = QUEUE_NEXT_LISTENER(MyProcNumber);
    else
    {
        for (ProcNumber i = QUEUE_FIRST_LISTENER; i != INVALID_PROC_NUMBER; i = QUEUE_NEXT_LISTENER(i))
        {
            if (QUEUE_NEXT_LISTENER(i) == MyProcNumber)
            {
                QUEUE_NEXT_LISTENER(i) = QUEUE_NEXT_LISTENER(MyProcNumber);
                break;
            }
        }
    }
    QUEUE_NEXT_LISTENER(MyProcNumber) = INVALID_PROC_NUMBER;
    LWLockRelease(NotifyQueueLock);
 
    /* mark ourselves as no longer listed in the global array */
    amRegisteredListener = false;
}
 
/*
 * Test whether there is room to insert more notification messages.
 *
 * Caller must hold at least shared NotifyQueueLock.
 */
static bool
asyncQueueIsFull(void)
{
    int64       headPage = QUEUE_POS_PAGE(QUEUE_HEAD);
    int64       tailPage = QUEUE_POS_PAGE(QUEUE_TAIL);
    int64       occupied = headPage - tailPage;
 
    return occupied >= max_notify_queue_pages;
}
 
/*
 * Advance the QueuePosition to the next entry, assuming that the current
 * entry is of length entryLength.  If we jump to a new page the function
 * returns true, else false.
 */
static bool
asyncQueueAdvance(volatile QueuePosition *position, int entryLength)
{
    int64       pageno = QUEUE_POS_PAGE(*position);
    int         offset = QUEUE_POS_OFFSET(*position);
    bool        pageJump = false;
 
    /*
     * Move to the next writing position: First jump over what we have just
     * written or read.
     */
    offset += entryLength;
    Assert(offset <= QUEUE_PAGESIZE);
 
    /*
     * In a second step check if another entry can possibly be written to the
     * page. If so, stay here, we have reached the next position. If not, then
     * we need to move on to the next page.
     */
    if (offset + QUEUEALIGN(AsyncQueueEntryEmptySize) > QUEUE_PAGESIZE)
    {
        pageno++;
        offset = 0;
        pageJump = true;
    }
 
    SET_QUEUE_POS(*position, pageno, offset);
    return pageJump;
}
 
/*
 * Fill the AsyncQueueEntry at *qe with an outbound notification message.
 */
static void
asyncQueueNotificationToEntry(Notification *n, AsyncQueueEntry *qe)
{
    size_t      channellen = n->channel_len;
    size_t      payloadlen = n->payload_len;
    int         entryLength;
 
    Assert(channellen < NAMEDATALEN);
    Assert(payloadlen < NOTIFY_PAYLOAD_MAX_LENGTH);
 
    /* The terminators are already included in AsyncQueueEntryEmptySize */
    entryLength = AsyncQueueEntryEmptySize + payloadlen + channellen;
    entryLength = QUEUEALIGN(entryLength);
    qe->length = entryLength;
    qe->dboid = MyDatabaseId;
    qe->xid = GetCurrentTransactionId();
    qe->srcPid = MyProcPid;
    memcpy(qe->data, n->data, channellen + payloadlen + 2);
}
 
/*
 * Add pending notifications to the queue.
 *
 * We go page by page here, i.e. we stop once we have to go to a new page but
 * we will be called again and then fill that next page. If an entry does not
 * fit into the current page, we write a dummy entry with an InvalidOid as the
 * database OID in order to fill the page. So every page is always used up to
 * the last byte which simplifies reading the page later.
 *
 * We are passed the list cell (in pendingNotifies->events) containing the next
 * notification to write and return the first still-unwritten cell back.
 * Eventually we will return NULL indicating all is done.
 *
 * We are holding NotifyQueueLock already from the caller and grab
 * page specific SLRU bank lock locally in this function.
 */
static ListCell *
asyncQueueAddEntries(ListCell *nextNotify)
{
    AsyncQueueEntry qe;
    QueuePosition queue_head;
    int64       pageno;
    int         offset;
    int         slotno;
    LWLock     *prevlock;
 
    /*
     * We work with a local copy of QUEUE_HEAD, which we write back to shared
     * memory upon exiting.  The reason for this is that if we have to advance
     * to a new page, SimpleLruZeroPage might fail (out of disk space, for
     * instance), and we must not advance QUEUE_HEAD if it does.  (Otherwise,
     * subsequent insertions would try to put entries into a page that slru.c
     * thinks doesn't exist yet.)  So, use a local position variable.  Note
     * that if we do fail, any already-inserted queue entries are forgotten;
     * this is okay, since they'd be useless anyway after our transaction
     * rolls back.
     */
    queue_head = QUEUE_HEAD;
 
    /*
     * If this is the first write since the postmaster started, we need to
     * initialize the first page of the async SLRU.  Otherwise, the current
     * page should be initialized already, so just fetch it.
     */
    pageno = QUEUE_POS_PAGE(queue_head);
    prevlock = SimpleLruGetBankLock(NotifyCtl, pageno);
 
    /* We hold both NotifyQueueLock and SLRU bank lock during this operation */
    LWLockAcquire(prevlock, LW_EXCLUSIVE);
 
    if (QUEUE_POS_IS_ZERO(queue_head))
        slotno = SimpleLruZeroPage(NotifyCtl, pageno);
    else
        slotno = SimpleLruReadPage(NotifyCtl, pageno, true, &queue_head);
 
    /* Note we mark the page dirty before writing in it */
    NotifyCtl->shared->page_dirty[slotno] = true;
 
    while (nextNotify != NULL)
    {
        Notification *n = (Notification *) lfirst(nextNotify);
 
        /* Construct a valid queue entry in local variable qe */
        asyncQueueNotificationToEntry(n, &qe);
 
        offset = QUEUE_POS_OFFSET(queue_head);
 
        /* Check whether the entry really fits on the current page */
        if (offset + qe.length <= QUEUE_PAGESIZE)
        {
            /* OK, so advance nextNotify past this item */
            nextNotify = lnext(pendingNotifies->events, nextNotify);
        }
        else
        {
            /*
             * Write a dummy entry to fill up the page. Actually readers will
             * only check dboid and since it won't match any reader's database
             * OID, they will ignore this entry and move on.
             */
            qe.length = QUEUE_PAGESIZE - offset;
            qe.dboid = InvalidOid;
            qe.xid = InvalidTransactionId;
            qe.data[0] = '\0';  /* empty channel */
            qe.data[1] = '\0';  /* empty payload */
        }
 
        /* Now copy qe into the shared buffer page */
        memcpy(NotifyCtl->shared->page_buffer[slotno] + offset,
               &qe,
               qe.length);
 
        /* Advance queue_head appropriately, and detect if page is full */
        if (asyncQueueAdvance(&(queue_head), qe.length))
        {
            LWLock     *lock;
 
            pageno = QUEUE_POS_PAGE(queue_head);
            lock = SimpleLruGetBankLock(NotifyCtl, pageno);
            if (lock != prevlock)
            {
                LWLockRelease(prevlock);
                LWLockAcquire(lock, LW_EXCLUSIVE);
                prevlock = lock;
            }
 
            /*
             * Page is full, so we're done here, but first fill the next page
             * with zeroes.  The reason to do this is to ensure that slru.c's
             * idea of the head page is always the same as ours, which avoids
             * boundary problems in SimpleLruTruncate.  The test in
             * asyncQueueIsFull() ensured that there is room to create this
             * page without overrunning the queue.
             */
            slotno = SimpleLruZeroPage(NotifyCtl, QUEUE_POS_PAGE(queue_head));
 
            /*
             * If the new page address is a multiple of QUEUE_CLEANUP_DELAY,
             * set flag to remember that we should try to advance the tail
             * pointer (we don't want to actually do that right here).
             */
            if (QUEUE_POS_PAGE(queue_head) % QUEUE_CLEANUP_DELAY == 0)
                tryAdvanceTail = true;
 
            /* And exit the loop */
            break;
        }
    }
 
    /* Success, so update the global QUEUE_HEAD */
    QUEUE_HEAD = queue_head;
 
    LWLockRelease(prevlock);
 
    return nextNotify;
}
 
/*
 * SQL function to return the fraction of the notification queue currently
 * occupied.
 */
Datum
pg_notification_queue_usage(PG_FUNCTION_ARGS)
{
    double      usage;
 
    /* Advance the queue tail so we don't report a too-large result */
    asyncQueueAdvanceTail();
 
    LWLockAcquire(NotifyQueueLock, LW_SHARED);
    usage = asyncQueueUsage();
    LWLockRelease(NotifyQueueLock);
 
    PG_RETURN_FLOAT8(usage);
}
 
/*
 * Return the fraction of the queue that is currently occupied.
 *
 * The caller must hold NotifyQueueLock in (at least) shared mode.
 *
 * Note: we measure the distance to the logical tail page, not the physical
 * tail page.  In some sense that's wrong, but the relative position of the
 * physical tail is affected by details such as SLRU segment boundaries,
 * so that a result based on that is unpleasantly unstable.
 */
static double
asyncQueueUsage(void)
{
    int64       headPage = QUEUE_POS_PAGE(QUEUE_HEAD);
    int64       tailPage = QUEUE_POS_PAGE(QUEUE_TAIL);
    int64       occupied = headPage - tailPage;
 
    if (occupied == 0)
        return (double) 0;      /* fast exit for common case */
 
    return (double) occupied / (double) max_notify_queue_pages;
}
 
/*
 * Check whether the queue is at least half full, and emit a warning if so.
 *
 * This is unlikely given the size of the queue, but possible.
 * The warnings show up at most once every QUEUE_FULL_WARN_INTERVAL.
 *
 * Caller must hold exclusive NotifyQueueLock.
 */
static void
asyncQueueFillWarning(void)
{
    double      fillDegree;
    TimestampTz t;
 
    fillDegree = asyncQueueUsage();
    if (fillDegree < 0.5)
        return;
 
    t = GetCurrentTimestamp();
 
    if (TimestampDifferenceExceeds(asyncQueueControl->lastQueueFillWarn,
                                   t, QUEUE_FULL_WARN_INTERVAL))
    {
        QueuePosition min = QUEUE_HEAD;
        int32       minPid = InvalidPid;
 
        for (ProcNumber i = QUEUE_FIRST_LISTENER; i != INVALID_PROC_NUMBER; i = QUEUE_NEXT_LISTENER(i))
        {
            Assert(QUEUE_BACKEND_PID(i) != InvalidPid);
            min = QUEUE_POS_MIN(min, QUEUE_BACKEND_POS(i));
            if (QUEUE_POS_EQUAL(min, QUEUE_BACKEND_POS(i)))
                minPid = QUEUE_BACKEND_PID(i);
        }
 
        ereport(WARNING,
                (errmsg("NOTIFY queue is %.0f%% full", fillDegree * 100),
                 (minPid != InvalidPid ?
                  errdetail("The server process with PID %d is among those with the oldest transactions.", minPid)
                  : 0),
                 (minPid != InvalidPid ?
                  errhint("The NOTIFY queue cannot be emptied until that process ends its current transaction.")
                  : 0)));
 
        asyncQueueControl->lastQueueFillWarn = t;
    }
}
 
/*
 * Send signals to listening backends.
 *
 * Normally we signal only backends that are interested in the notifies that
 * we just sent.  However, that will leave idle listeners falling further and
 * further behind.  Waken them anyway if they're far enough behind, so they'll
 * advance their queue position pointers, allowing the global tail to advance.
 *
 * Since we know the ProcNumber and the Pid the signaling is quite cheap.
 *
 * This is called during CommitTransaction(), so it's important for it
 * to have very low probability of failure.
 */
static void
SignalBackends(void)
{
    int         count;
 
    /* Can't get here without PreCommit_Notify having made the global table */
    Assert(globalChannelTable != NULL);
 
    /* It should have set up these arrays, too */
    Assert(signalPids != NULL && signalProcnos != NULL);
 
    /*
     * Identify backends that we need to signal.  We don't want to send
     * signals while holding the NotifyQueueLock, so this part just builds a
     * list of target PIDs in signalPids[] and signalProcnos[].
     */
    count = 0;
 
    LWLockAcquire(NotifyQueueLock, LW_EXCLUSIVE);
 
    /* Scan each channel name that we notified in this transaction */
    foreach_ptr(char, channel, pendingNotifies->uniqueChannelNames)
    {
        GlobalChannelKey key;
        GlobalChannelEntry *entry;
        ListenerEntry *listeners;
 
        GlobalChannelKeyInit(&key, MyDatabaseId, channel);
        entry = dshash_find(globalChannelTable, &key, false);
        if (entry == NULL)
            continue;           /* nobody is listening */
 
        listeners = (ListenerEntry *) dsa_get_address(globalChannelDSA,
                                                      entry->listenersArray);
 
        /*
         * Identify listeners that now need waking, add them to arrays.
         *
         * Note that we signal listeners regardless of the state of their
         * removeOnAbort flags.  Hence a new listener that reached PreCommit,
         * but then failed before AtCommit_Notify, can receive a signal even
         * though it was never really listening.  This is okay because it will
         * not do anything in response to that signal.  If we did not do it
         * like this then a new listener might miss some messages due to the
         * direct-advance logic below.
         */
        for (int j = 0; j < entry->numListeners; j++)
        {
            ProcNumber  i = listeners[j].procNo;
            int32       pid;
            QueuePosition pos;
 
            if (QUEUE_BACKEND_WAKEUP_PENDING(i))
                continue;       /* already signaled, no need to repeat */
 
            pid = QUEUE_BACKEND_PID(i);
            pos = QUEUE_BACKEND_POS(i);
 
            if (QUEUE_POS_EQUAL(pos, QUEUE_HEAD))
                continue;       /* it's fully caught up already */
 
            Assert(pid != InvalidPid);
 
            QUEUE_BACKEND_WAKEUP_PENDING(i) = true;
            signalPids[count] = pid;
            signalProcnos[count] = i;
            count++;
        }
 
        dshash_release_lock(globalChannelTable, entry);
    }
 
    /*
     * Scan all listeners.  Any that are not already pending wakeup must not
     * be interested in our notifications (else we'd have set their wakeup
     * flags above).  Check to see if we can directly advance their queue
     * pointers to save a wakeup.  Otherwise, if they are far behind, wake
     * them anyway so they will catch up.
     */
    for (ProcNumber i = QUEUE_FIRST_LISTENER; i != INVALID_PROC_NUMBER; i = QUEUE_NEXT_LISTENER(i))
    {
        int32       pid;
        QueuePosition pos;
 
        if (QUEUE_BACKEND_WAKEUP_PENDING(i))
            continue;
 
        /* If it's currently advancing, we should not touch it */
        if (QUEUE_BACKEND_IS_ADVANCING(i))
            continue;
 
        pid = QUEUE_BACKEND_PID(i);
        pos = QUEUE_BACKEND_POS(i);
 
        /*
         * We can directly advance the other backend's queue pointer if it's
         * not currently advancing (else there are race conditions), and its
         * current pointer is not behind queueHeadBeforeWrite (else we'd make
         * it miss some older messages), and we'd not be moving the pointer
         * backward.
         */
        if (!QUEUE_POS_PRECEDES(pos, queueHeadBeforeWrite) &&
            QUEUE_POS_PRECEDES(pos, queueHeadAfterWrite))
        {
            /* We can directly advance its pointer past what we wrote */
            QUEUE_BACKEND_POS(i) = queueHeadAfterWrite;
        }
        else if (asyncQueuePageDiff(QUEUE_POS_PAGE(QUEUE_HEAD),
                                    QUEUE_POS_PAGE(pos)) >= QUEUE_CLEANUP_DELAY)
        {
            /* It's idle and far behind, so wake it up */
            Assert(pid != InvalidPid);
 
            QUEUE_BACKEND_WAKEUP_PENDING(i) = true;
            signalPids[count] = pid;
            signalProcnos[count] = i;
            count++;
        }
    }
 
    LWLockRelease(NotifyQueueLock);
 
    /* Now send signals */
    for (int i = 0; i < count; i++)
    {
        int32       pid = signalPids[i];
 
        /*
         * If we are signaling our own process, no need to involve the kernel;
         * just set the flag directly.
         */
        if (pid == MyProcPid)
        {
            notifyInterruptPending = true;
            continue;
        }
 
        /*
         * Note: assuming things aren't broken, a signal failure here could
         * only occur if the target backend exited since we released
         * NotifyQueueLock; which is unlikely but certainly possible. So we
         * just log a low-level debug message if it happens.
         */
        if (SendProcSignal(pid, PROCSIG_NOTIFY_INTERRUPT, signalProcnos[i]) < 0)
            elog(DEBUG3, "could not signal backend with PID %d: %m", pid);
    }
}
 
/*
 * AtAbort_Notify
 *
 *  This is called at transaction abort.
 *
 *  Revert any staged listen/unlisten changes and clean up transaction state.
 *  This only does anything if we abort after PreCommit_Notify has staged
 *  some entries.
 */
void
AtAbort_Notify(void)
{
    /* Revert staged listen/unlisten changes */
    ApplyPendingListenActions(false);
 
    /* If we're no longer listening on anything, unregister */
    if (amRegisteredListener && LocalChannelTableIsEmpty())
        asyncQueueUnregister();
 
    /* And clean up */
    ClearPendingActionsAndNotifies();
}
 
/*
 * AtSubCommit_Notify() --- Take care of subtransaction commit.
 *
 * Reassign all items in the pending lists to the parent transaction.
 */
void
AtSubCommit_Notify(void)
{
    int         my_level = GetCurrentTransactionNestLevel();
 
    /* If there are actions at our nesting level, we must reparent them. */
    if (pendingActions != NULL &&
        pendingActions->nestingLevel >= my_level)
    {
        if (pendingActions->upper == NULL ||
            pendingActions->upper->nestingLevel < my_level - 1)
        {
            /* nothing to merge; give the whole thing to the parent */
            --pendingActions->nestingLevel;
        }
        else
        {
            ActionList *childPendingActions = pendingActions;
 
            pendingActions = pendingActions->upper;
 
            /*
             * Mustn't try to eliminate duplicates here --- see queue_listen()
             */
            pendingActions->actions =
                list_concat(pendingActions->actions,
                            childPendingActions->actions);
            pfree(childPendingActions);
        }
    }
 
    /* If there are notifies at our nesting level, we must reparent them. */
    if (pendingNotifies != NULL &&
        pendingNotifies->nestingLevel >= my_level)
    {
        Assert(pendingNotifies->nestingLevel == my_level);
 
        if (pendingNotifies->upper == NULL ||
            pendingNotifies->upper->nestingLevel < my_level - 1)
        {
            /* nothing to merge; give the whole thing to the parent */
            --pendingNotifies->nestingLevel;
        }
        else
        {
            /*
             * Formerly, we didn't bother to eliminate duplicates here, but
             * now we must, else we fall foul of "Assert(!found)", either here
             * or during a later attempt to build the parent-level hashtable.
             */
            NotificationList *childPendingNotifies = pendingNotifies;
            ListCell   *l;
 
            pendingNotifies = pendingNotifies->upper;
            /* Insert all the subxact's events into parent, except for dups */
            foreach(l, childPendingNotifies->events)
            {
                Notification *childn = (Notification *) lfirst(l);
 
                if (!AsyncExistsPendingNotify(childn))
                    AddEventToPendingNotifies(childn);
            }
            pfree(childPendingNotifies);
        }
    }
}
 
/*
 * AtSubAbort_Notify() --- Take care of subtransaction abort.
 */
void
AtSubAbort_Notify(void)
{
    int         my_level = GetCurrentTransactionNestLevel();
 
    /*
     * All we have to do is pop the stack --- the actions/notifies made in
     * this subxact are no longer interesting, and the space will be freed
     * when CurTransactionContext is recycled. We still have to free the
     * ActionList and NotificationList objects themselves, though, because
     * those are allocated in TopTransactionContext.
     *
     * Note that there might be no entries at all, or no entries for the
     * current subtransaction level, either because none were ever created, or
     * because we reentered this routine due to trouble during subxact abort.
     */
    while (pendingActions != NULL &&
           pendingActions->nestingLevel >= my_level)
    {
        ActionList *childPendingActions = pendingActions;
 
        pendingActions = pendingActions->upper;
        pfree(childPendingActions);
    }
 
    while (pendingNotifies != NULL &&
           pendingNotifies->nestingLevel >= my_level)
    {
        NotificationList *childPendingNotifies = pendingNotifies;
 
        pendingNotifies = pendingNotifies->upper;
        pfree(childPendingNotifies);
    }
}
 
/*
 * HandleNotifyInterrupt
 *
 *      Signal handler portion of interrupt handling. Let the backend know
 *      that there's a pending notify interrupt. If we're currently reading
 *      from the client, this will interrupt the read and
 *      ProcessClientReadInterrupt() will call ProcessNotifyInterrupt().
 */
void
HandleNotifyInterrupt(void)
{
    /*
     * Note: this is called by a SIGNAL HANDLER. You must be very wary what
     * you do here.
     */
 
    /* signal that work needs to be done */
    notifyInterruptPending = true;
 
    /* latch will be set by procsignal_sigusr1_handler */
}
 
/*
 * ProcessNotifyInterrupt
 *
 *      This is called if we see notifyInterruptPending set, just before
 *      transmitting ReadyForQuery at the end of a frontend command, and
 *      also if a notify signal occurs while reading from the frontend.
 *      HandleNotifyInterrupt() will cause the read to be interrupted
 *      via the process's latch, and this routine will get called.
 *      If we are truly idle (ie, *not* inside a transaction block),
 *      process the incoming notifies.
 *
 *      If "flush" is true, force any frontend messages out immediately.
 *      This can be false when being called at the end of a frontend command,
 *      since we'll flush after sending ReadyForQuery.
 */
void
ProcessNotifyInterrupt(bool flush)
{
    if (IsTransactionOrTransactionBlock())
        return;                 /* not really idle */
 
    /* Loop in case another signal arrives while sending messages */
    while (notifyInterruptPending)
        ProcessIncomingNotify(flush);
}
 
 
/*
 * Read all pending notifications from the queue, and deliver appropriate
 * ones to my frontend.  Stop when we reach queue head or an uncommitted
 * notification.
 */
static void
asyncQueueReadAllNotifications(void)
{
    QueuePosition pos;
    QueuePosition head;
    Snapshot    snapshot;
 
    /*
     * Fetch current state, indicate to others that we have woken up, and that
     * we are in process of advancing our position.
     */
    LWLockAcquire(NotifyQueueLock, LW_SHARED);
    /* Assert checks that we have a valid state entry */
    Assert(MyProcPid == QUEUE_BACKEND_PID(MyProcNumber));
    QUEUE_BACKEND_WAKEUP_PENDING(MyProcNumber) = false;
    pos = QUEUE_BACKEND_POS(MyProcNumber);
    head = QUEUE_HEAD;
 
    if (QUEUE_POS_EQUAL(pos, head))
    {
        /* Nothing to do, we have read all notifications already. */
        LWLockRelease(NotifyQueueLock);
        return;
    }
 
    QUEUE_BACKEND_IS_ADVANCING(MyProcNumber) = true;
    LWLockRelease(NotifyQueueLock);
 
    /*----------
     * Get snapshot we'll use to decide which xacts are still in progress.
     * This is trickier than it might seem, because of race conditions.
     * Consider the following example:
     *
     * Backend 1:                    Backend 2:
     *
     * transaction starts
     * UPDATE foo SET ...;
     * NOTIFY foo;
     * commit starts
     * queue the notify message
     *                               transaction starts
     *                               LISTEN foo;  -- first LISTEN in session
     *                               SELECT * FROM foo WHERE ...;
     * commit to clog
     *                               commit starts
     *                               add backend 2 to array of listeners
     *                               advance to queue head (this code)
     *                               commit to clog
     *
     * Transaction 2's SELECT has not seen the UPDATE's effects, since that
     * wasn't committed yet.  Ideally we'd ensure that client 2 would
     * eventually get transaction 1's notify message, but there's no way
     * to do that; until we're in the listener array, there's no guarantee
     * that the notify message doesn't get removed from the queue.
     *
     * Therefore the coding technique transaction 2 is using is unsafe:
     * applications must commit a LISTEN before inspecting database state,
     * if they want to ensure they will see notifications about subsequent
     * changes to that state.
     *
     * What we do guarantee is that we'll see all notifications from
     * transactions committing after the snapshot we take here.
     * BecomeRegisteredListener has already added us to the listener array,
     * so no not-yet-committed messages can be removed from the queue
     * before we see them.
     *----------
     */
    snapshot = RegisterSnapshot(GetLatestSnapshot());
 
    /*
     * It is possible that we fail while trying to send a message to our
     * frontend (for example, because of encoding conversion failure).  If
     * that happens it is critical that we not try to send the same message
     * over and over again.  Therefore, we set ExitOnAnyError to upgrade any
     * ERRORs to FATAL, causing the client connection to be closed on error.
     *
     * We used to only skip over the offending message and try to soldier on,
     * but it was somewhat questionable to lose a notification and give the
     * client an ERROR instead.  A client application is not be prepared for
     * that and can't tell that a notification was missed.  It was also not
     * very useful in practice because notifications are often processed while
     * a connection is idle and reading a message from the client, and in that
     * state, any error is upgraded to FATAL anyway.  Closing the connection
     * is a clear signal to the application that it might have missed
     * notifications.
     */
    {
        bool        save_ExitOnAnyError = ExitOnAnyError;
        bool        reachedStop;
 
        ExitOnAnyError = true;
 
        do
        {
            /*
             * Process messages up to the stop position, end of page, or an
             * uncommitted message.
             *
             * Our stop position is what we found to be the head's position
             * when we entered this function. It might have changed already.
             * But if it has, we will receive (or have already received and
             * queued) another signal and come here again.
             *
             * We are not holding NotifyQueueLock here! The queue can only
             * extend beyond the head pointer (see above) and we leave our
             * backend's pointer where it is so nobody will truncate or
             * rewrite pages under us. Especially we don't want to hold a lock
             * while sending the notifications to the frontend.
             */
            reachedStop = asyncQueueProcessPageEntries(&pos, head, snapshot);
        } while (!reachedStop);
 
        /* Update shared state */
        LWLockAcquire(NotifyQueueLock, LW_SHARED);
        QUEUE_BACKEND_POS(MyProcNumber) = pos;
        QUEUE_BACKEND_IS_ADVANCING(MyProcNumber) = false;
        LWLockRelease(NotifyQueueLock);
 
        ExitOnAnyError = save_ExitOnAnyError;
    }
 
    /* Done with snapshot */
    UnregisterSnapshot(snapshot);
}
 
/*
 * Fetch notifications from the shared queue, beginning at position current,
 * and deliver relevant ones to my frontend.
 *
 * The function returns true once we have reached the stop position or an
 * uncommitted notification, and false if we have finished with the page.
 * In other words: once it returns true there is no need to look further.
 * The QueuePosition *current is advanced past all processed messages.
 */
static bool
asyncQueueProcessPageEntries(QueuePosition *current,
                             QueuePosition stop,
                             Snapshot snapshot)
{
    int64       curpage = QUEUE_POS_PAGE(*current);
    int         slotno;
    char       *page_buffer;
    bool        reachedStop = false;
    bool        reachedEndOfPage;
 
    /*
     * We copy the entries into a local buffer to avoid holding the SLRU lock
     * while we transmit them to our frontend.  The local buffer must be
     * adequately aligned.
     */
    alignas(AsyncQueueEntry) char local_buf[QUEUE_PAGESIZE];
    char       *local_buf_end = local_buf;
 
    slotno = SimpleLruReadPage_ReadOnly(NotifyCtl, curpage, current);
    page_buffer = NotifyCtl->shared->page_buffer[slotno];
 
    do
    {
        QueuePosition thisentry = *current;
        AsyncQueueEntry *qe;
 
        if (QUEUE_POS_EQUAL(thisentry, stop))
            break;
 
        qe = (AsyncQueueEntry *) (page_buffer + QUEUE_POS_OFFSET(thisentry));
 
        /*
         * Advance *current over this message, possibly to the next page. As
         * noted in the comments for asyncQueueReadAllNotifications, we must
         * do this before possibly failing while processing the message.
         */
        reachedEndOfPage = asyncQueueAdvance(current, qe->length);
 
        /* Ignore messages destined for other databases */
        if (qe->dboid == MyDatabaseId)
        {
            if (XidInMVCCSnapshot(qe->xid, snapshot))
            {
                /*
                 * The source transaction is still in progress, so we can't
                 * process this message yet.  Break out of the loop, but first
                 * back up *current so we will reprocess the message next
                 * time.  (Note: it is unlikely but not impossible for
                 * TransactionIdDidCommit to fail, so we can't really avoid
                 * this advance-then-back-up behavior when dealing with an
                 * uncommitted message.)
                 *
                 * Note that we must test XidInMVCCSnapshot before we test
                 * TransactionIdDidCommit, else we might return a message from
                 * a transaction that is not yet visible to snapshots; compare
                 * the comments at the head of heapam_visibility.c.
                 *
                 * Also, while our own xact won't be listed in the snapshot,
                 * we need not check for TransactionIdIsCurrentTransactionId
                 * because our transaction cannot (yet) have queued any
                 * messages.
                 */
                *current = thisentry;
                reachedStop = true;
                break;
            }
 
            /*
             * Quick check for the case that we're not listening on any
             * channels, before calling TransactionIdDidCommit().  This makes
             * that case a little faster, but more importantly, it ensures
             * that if there's a bad entry in the queue for which
             * TransactionIdDidCommit() fails for some reason, we can skip
             * over it on the first LISTEN in a session, and not get stuck on
             * it indefinitely.  (This is a little trickier than it looks: it
             * works because BecomeRegisteredListener runs this code before we
             * have made the first entry in localChannelTable.)
             */
            if (LocalChannelTableIsEmpty())
                continue;
 
            if (TransactionIdDidCommit(qe->xid))
            {
                memcpy(local_buf_end, qe, qe->length);
                local_buf_end += qe->length;
            }
            else
            {
                /*
                 * The source transaction aborted or crashed, so we just
                 * ignore its notifications.
                 */
            }
        }
 
        /* Loop back if we're not at end of page */
    } while (!reachedEndOfPage);
 
    /* Release lock that we got from SimpleLruReadPage_ReadOnly() */
    LWLockRelease(SimpleLruGetBankLock(NotifyCtl, curpage));
 
    /*
     * Now that we have let go of the SLRU bank lock, send the notifications
     * to our backend
     */
    Assert(local_buf_end - local_buf <= BLCKSZ);
    for (char *p = local_buf; p < local_buf_end;)
    {
        AsyncQueueEntry *qe = (AsyncQueueEntry *) p;
 
        /* qe->data is the null-terminated channel name */
        char       *channel = qe->data;
 
        if (IsListeningOn(channel))
        {
            /* payload follows channel name */
            char       *payload = qe->data + strlen(channel) + 1;
 
            NotifyMyFrontEnd(channel, payload, qe->srcPid);
        }
 
        p += qe->length;
    }
 
    if (QUEUE_POS_EQUAL(*current, stop))
        reachedStop = true;
 
    return reachedStop;
}
 
/*
 * Advance the shared queue tail variable to the minimum of all the
 * per-backend tail pointers.  Truncate pg_notify space if possible.
 *
 * This is (usually) called during CommitTransaction(), so it's important for
 * it to have very low probability of failure.
 */
static void
asyncQueueAdvanceTail(void)
{
    QueuePosition min;
    int64       oldtailpage;
    int64       newtailpage;
    int64       boundary;
 
    /* Restrict task to one backend per cluster; see SimpleLruTruncate(). */
    LWLockAcquire(NotifyQueueTailLock, LW_EXCLUSIVE);
 
    /*
     * Compute the new tail.  Pre-v13, it's essential that QUEUE_TAIL be exact
     * (ie, exactly match at least one backend's queue position), so it must
     * be updated atomically with the actual computation.  Since v13, we could
     * get away with not doing it like that, but it seems prudent to keep it
     * so.
     *
     * Also, because incoming backends will scan forward from QUEUE_TAIL, that
     * must be advanced before we can truncate any data.  Thus, QUEUE_TAIL is
     * the logical tail, while QUEUE_STOP_PAGE is the physical tail, or oldest
     * un-truncated page.  When QUEUE_STOP_PAGE != QUEUE_POS_PAGE(QUEUE_TAIL),
     * there are pages we can truncate but haven't yet finished doing so.
     *
     * For concurrency's sake, we don't want to hold NotifyQueueLock while
     * performing SimpleLruTruncate.  This is OK because no backend will try
     * to access the pages we are in the midst of truncating.
     */
    LWLockAcquire(NotifyQueueLock, LW_EXCLUSIVE);
    min = QUEUE_HEAD;
    for (ProcNumber i = QUEUE_FIRST_LISTENER; i != INVALID_PROC_NUMBER; i = QUEUE_NEXT_LISTENER(i))
    {
        Assert(QUEUE_BACKEND_PID(i) != InvalidPid);
        min = QUEUE_POS_MIN(min, QUEUE_BACKEND_POS(i));
    }
    QUEUE_TAIL = min;
    oldtailpage = QUEUE_STOP_PAGE;
    LWLockRelease(NotifyQueueLock);
 
    /*
     * We can truncate something if the global tail advanced across an SLRU
     * segment boundary.
     *
     * XXX it might be better to truncate only once every several segments, to
     * reduce the number of directory scans.
     */
    newtailpage = QUEUE_POS_PAGE(min);
    boundary = newtailpage - (newtailpage % SLRU_PAGES_PER_SEGMENT);
    if (asyncQueuePagePrecedes(oldtailpage, boundary))
    {
        /*
         * SimpleLruTruncate() will ask for SLRU bank locks but will also
         * release the lock again.
         */
        SimpleLruTruncate(NotifyCtl, newtailpage);
 
        LWLockAcquire(NotifyQueueLock, LW_EXCLUSIVE);
        QUEUE_STOP_PAGE = newtailpage;
        LWLockRelease(NotifyQueueLock);
    }
 
    LWLockRelease(NotifyQueueTailLock);
}
 
/*
 * AsyncNotifyFreezeXids
 *
 * Prepare the async notification queue for CLOG truncation by freezing
 * transaction IDs that are about to become inaccessible.
 *
 * This function is called by VACUUM before advancing datfrozenxid. It scans
 * the notification queue and replaces XIDs that would become inaccessible
 * after CLOG truncation with special markers:
 * - Committed transactions are set to FrozenTransactionId
 * - Aborted/crashed transactions are set to InvalidTransactionId
 *
 * Only XIDs < newFrozenXid are processed, as those are the ones whose CLOG
 * pages will be truncated. If XID < newFrozenXid, it cannot still be running
 * (or it would have held back newFrozenXid through ProcArray).
 * Therefore, if TransactionIdDidCommit returns false, we know the transaction
 * either aborted explicitly or crashed, and we can safely mark it invalid.
 */
void
AsyncNotifyFreezeXids(TransactionId newFrozenXid)
{
    QueuePosition pos;
    QueuePosition head;
    int64       curpage = -1;
    int         slotno = -1;
    char       *page_buffer = NULL;
    bool        page_dirty = false;
 
    /*
     * Acquire locks in the correct order to avoid deadlocks. As per the
     * locking protocol: NotifyQueueTailLock, then NotifyQueueLock, then SLRU
     * bank locks.
     *
     * We only need SHARED mode since we're just reading the head/tail
     * positions, not modifying them.
     */
    LWLockAcquire(NotifyQueueTailLock, LW_SHARED);
    LWLockAcquire(NotifyQueueLock, LW_SHARED);
 
    pos = QUEUE_TAIL;
    head = QUEUE_HEAD;
 
    /* Release NotifyQueueLock early, we only needed to read the positions */
    LWLockRelease(NotifyQueueLock);
 
    /*
     * Scan the queue from tail to head, freezing XIDs as needed. We hold
     * NotifyQueueTailLock throughout to ensure the tail doesn't move while
     * we're working.
     */
    while (!QUEUE_POS_EQUAL(pos, head))
    {
        AsyncQueueEntry *qe;
        TransactionId xid;
        int64       pageno = QUEUE_POS_PAGE(pos);
        int         offset = QUEUE_POS_OFFSET(pos);
 
        /* If we need a different page, release old lock and get new one */
        if (pageno != curpage)
        {
            LWLock     *lock;
 
            /* Release previous page if any */
            if (slotno >= 0)
            {
                if (page_dirty)
                {
                    NotifyCtl->shared->page_dirty[slotno] = true;
                    page_dirty = false;
                }
                LWLockRelease(SimpleLruGetBankLock(NotifyCtl, curpage));
            }
 
            lock = SimpleLruGetBankLock(NotifyCtl, pageno);
            LWLockAcquire(lock, LW_EXCLUSIVE);
            slotno = SimpleLruReadPage(NotifyCtl, pageno, true, &pos);
            page_buffer = NotifyCtl->shared->page_buffer[slotno];
            curpage = pageno;
        }
 
        qe = (AsyncQueueEntry *) (page_buffer + offset);
        xid = qe->xid;
 
        if (TransactionIdIsNormal(xid) &&
            TransactionIdPrecedes(xid, newFrozenXid))
        {
            if (TransactionIdDidCommit(xid))
            {
                qe->xid = FrozenTransactionId;
                page_dirty = true;
            }
            else
            {
                qe->xid = InvalidTransactionId;
                page_dirty = true;
            }
        }
 
        /* Advance to next entry */
        asyncQueueAdvance(&pos, qe->length);
    }
 
    /* Release final page lock if we acquired one */
    if (slotno >= 0)
    {
        if (page_dirty)
            NotifyCtl->shared->page_dirty[slotno] = true;
        LWLockRelease(SimpleLruGetBankLock(NotifyCtl, curpage));
    }
 
    LWLockRelease(NotifyQueueTailLock);
}
 
/*
 * ProcessIncomingNotify
 *
 *      Scan the queue for arriving notifications and report them to the front
 *      end.  The notifications might be from other sessions, or our own;
 *      there's no need to distinguish here.
 *
 *      If "flush" is true, force any frontend messages out immediately.
 *
 *      NOTE: since we are outside any transaction, we must create our own.
 */
static void
ProcessIncomingNotify(bool flush)
{
    /* We *must* reset the flag */
    notifyInterruptPending = false;
 
    /* Do nothing else if we aren't actively listening */
    if (LocalChannelTableIsEmpty())
        return;
 
    if (Trace_notify)
        elog(DEBUG1, "ProcessIncomingNotify");
 
    set_ps_display("notify interrupt");
 
    /*
     * We must run asyncQueueReadAllNotifications inside a transaction, else
     * bad things happen if it gets an error.
     */
    StartTransactionCommand();
 
    asyncQueueReadAllNotifications();
 
    CommitTransactionCommand();
 
    /*
     * If this isn't an end-of-command case, we must flush the notify messages
     * to ensure frontend gets them promptly.
     */
    if (flush)
        pq_flush();
 
    set_ps_display("idle");
 
    if (Trace_notify)
        elog(DEBUG1, "ProcessIncomingNotify: done");
}
 
/*
 * Send NOTIFY message to my front end.
 */
void
NotifyMyFrontEnd(const char *channel, const char *payload, int32 srcPid)
{
    if (whereToSendOutput == DestRemote)
    {
        StringInfoData buf;
 
        pq_beginmessage(&buf, PqMsg_NotificationResponse);
        pq_sendint32(&buf, srcPid);
        pq_sendstring(&buf, channel);
        pq_sendstring(&buf, payload);
        pq_endmessage(&buf);
 
        /*
         * NOTE: we do not do pq_flush() here.  Some level of caller will
         * handle it later, allowing this message to be combined into a packet
         * with other ones.
         */
    }
    else
        elog(INFO, "NOTIFY for \"%s\" payload \"%s\"", channel, payload);
}
 
/* Does pendingNotifies include a match for the given event? */
static bool
AsyncExistsPendingNotify(Notification *n)
{
    if (pendingNotifies == NULL)
        return false;
 
    if (pendingNotifies->hashtab != NULL)
    {
        /* Use the hash table to probe for a match */
        if (hash_search(pendingNotifies->hashtab,
                        &n,
                        HASH_FIND,
                        NULL))
            return true;
    }
    else
    {
        /* Must scan the event list */
        ListCell   *l;
 
        foreach(l, pendingNotifies->events)
        {
            Notification *oldn = (Notification *) lfirst(l);
 
            if (n->channel_len == oldn->channel_len &&
                n->payload_len == oldn->payload_len &&
                memcmp(n->data, oldn->data,
                       n->channel_len + n->payload_len + 2) == 0)
                return true;
        }
    }
 
    return false;
}
 
/*
 * Add a notification event to a pre-existing pendingNotifies list.
 *
 * Because pendingNotifies->events is already nonempty, this works
 * correctly no matter what CurrentMemoryContext is.
 */
static void
AddEventToPendingNotifies(Notification *n)
{
    Assert(pendingNotifies->events != NIL);
 
    /* Create the hash tables if it's time to */
    if (list_length(pendingNotifies->events) >= MIN_HASHABLE_NOTIFIES &&
        pendingNotifies->hashtab == NULL)
    {
        HASHCTL     hash_ctl;
        ListCell   *l;
 
        /* Create the hash table */
        hash_ctl.keysize = sizeof(Notification *);
        hash_ctl.entrysize = sizeof(struct NotificationHash);
        hash_ctl.hash = notification_hash;
        hash_ctl.match = notification_match;
        hash_ctl.hcxt = CurTransactionContext;
        pendingNotifies->hashtab =
            hash_create("Pending Notifies",
                        256L,
                        &hash_ctl,
                        HASH_ELEM | HASH_FUNCTION | HASH_COMPARE | HASH_CONTEXT);
 
        /* Create the unique channel name table */
        Assert(pendingNotifies->uniqueChannelHash == NULL);
        hash_ctl.keysize = NAMEDATALEN;
        hash_ctl.entrysize = sizeof(ChannelName);
        hash_ctl.hcxt = CurTransactionContext;
        pendingNotifies->uniqueChannelHash =
            hash_create("Pending Notify Channel Names",
                        64L,
                        &hash_ctl,
                        HASH_ELEM | HASH_STRINGS | HASH_CONTEXT);
 
        /* Insert all the already-existing events */
        foreach(l, pendingNotifies->events)
        {
            Notification *oldn = (Notification *) lfirst(l);
            char       *channel = oldn->data;
            bool        found;
 
            (void) hash_search(pendingNotifies->hashtab,
                               &oldn,
                               HASH_ENTER,
                               &found);
            Assert(!found);
 
            /* Add channel name to uniqueChannelHash; might be there already */
            (void) hash_search(pendingNotifies->uniqueChannelHash,
                               channel,
                               HASH_ENTER,
                               NULL);
        }
    }
 
    /* Add new event to the list, in order */
    pendingNotifies->events = lappend(pendingNotifies->events, n);
 
    /* Add event to the hash tables if needed */
    if (pendingNotifies->hashtab != NULL)
    {
        char       *channel = n->data;
        bool        found;
 
        (void) hash_search(pendingNotifies->hashtab,
                           &n,
                           HASH_ENTER,
                           &found);
        Assert(!found);
 
        /* Add channel name to uniqueChannelHash; might be there already */
        (void) hash_search(pendingNotifies->uniqueChannelHash,
                           channel,
                           HASH_ENTER,
                           NULL);
    }
}
 
/*
 * notification_hash: hash function for notification hash table
 *
 * The hash "keys" are pointers to Notification structs.
 */
static uint32
notification_hash(const void *key, Size keysize)
{
    const Notification *k = *(const Notification *const *) key;
 
    Assert(keysize == sizeof(Notification *));
    /* We don't bother to include the payload's trailing null in the hash */
    return DatumGetUInt32(hash_any((const unsigned char *) k->data,
                                   k->channel_len + k->payload_len + 1));
}
 
/*
 * notification_match: match function to use with notification_hash
 */
static int
notification_match(const void *key1, const void *key2, Size keysize)
{
    const Notification *k1 = *(const Notification *const *) key1;
    const Notification *k2 = *(const Notification *const *) key2;
 
    Assert(keysize == sizeof(Notification *));
    if (k1->channel_len == k2->channel_len &&
        k1->payload_len == k2->payload_len &&
        memcmp(k1->data, k2->data,
               k1->channel_len + k1->payload_len + 2) == 0)
        return 0;               /* equal */
    return 1;                   /* not equal */
}
 
/* Clear the pendingActions and pendingNotifies lists. */
static void
ClearPendingActionsAndNotifies(void)
{
    /*
     * Everything's allocated in either TopTransactionContext or the context
     * for the subtransaction to which it corresponds.  So, there's nothing to
     * do here except reset the pointers; the space will be reclaimed when the
     * contexts are deleted.
     */
    pendingActions = NULL;
    pendingNotifies = NULL;
    /* Also clear pendingListenActions, which is derived from pendingActions */
    pendingListenActions = NULL;
}
 
/*
 * GUC check_hook for notify_buffers
 */
bool
check_notify_buffers(int *newval, void **extra, GucSource source)
{
    return check_slru_buffers("notify_buffers", newval);
}

MIN_HASHABLE_NOTIFIES

#define MIN_HASHABLE_NOTIFIES   16 /* threshold to build hashtab */

Definition at line 527 of file async.c.

NOTIFY_PAYLOAD_MAX_LENGTH

#define NOTIFY_PAYLOAD_MAX_LENGTH   (BLCKSZ - NAMEDATALEN - 128)

Definition at line 201 of file async.c.

NotifyCtl

#define NotifyCtl   (&NotifySlruDesc)

Definition at line 378 of file async.c.

QUEUE_BACKEND_DBOID

#define QUEUE_BACKEND_DBOID

(

i

)

(asyncQueueControl->backend[i].dboid)

Definition at line 363 of file async.c.

QUEUE_BACKEND_IS_ADVANCING

#define QUEUE_BACKEND_IS_ADVANCING

(

i

)

(asyncQueueControl->backend[i].isAdvancing)

Definition at line 367 of file async.c.

QUEUE_BACKEND_PID

#define QUEUE_BACKEND_PID

(

i

)

(asyncQueueControl->backend[i].pid)

Definition at line 362 of file async.c.

QUEUE_BACKEND_POS

#define QUEUE_BACKEND_POS

(

i

)

(asyncQueueControl->backend[i].pos)

Definition at line 365 of file async.c.

QUEUE_BACKEND_WAKEUP_PENDING

#define QUEUE_BACKEND_WAKEUP_PENDING

(

i

)

(asyncQueueControl->backend[i].wakeupPending)

Definition at line 366 of file async.c.

QUEUE_CLEANUP_DELAY

#define QUEUE_CLEANUP_DELAY   4

Definition at line 282 of file async.c.

QUEUE_FIRST_LISTENER

#define QUEUE_FIRST_LISTENER   (asyncQueueControl->firstListener)

Definition at line 361 of file async.c.

QUEUE_FULL_WARN_INTERVAL

#define QUEUE_FULL_WARN_INTERVAL   5000 /* warn at most once every 5s */

Definition at line 381 of file async.c.

QUEUE_HEAD

#define QUEUE_HEAD   (asyncQueueControl->head)

Definition at line 358 of file async.c.

QUEUE_NEXT_LISTENER

#define QUEUE_NEXT_LISTENER

(

i

)

(asyncQueueControl->backend[i].nextListener)

Definition at line 364 of file async.c.

QUEUE_PAGESIZE

#define QUEUE_PAGESIZE   BLCKSZ

Definition at line 379 of file async.c.

QUEUE_POS_EQUAL

#define QUEUE_POS_EQUAL	(		x,
			y
	)		((x).page == (y).page && (x).offset == (y).offset)

Definition at line 247 of file async.c.

QUEUE_POS_IS_ZERO

#define QUEUE_POS_IS_ZERO

(

x

)

((x).page == 0 && (x).offset == 0)

Definition at line 250 of file async.c.

QUEUE_POS_MAX

#define QUEUE_POS_MAX	(		x,
			y
	)

Value:

    (asyncQueuePagePrecedes((x).page, (y).page) ? (y) : \
     (x).page != (y).page ? (x) : \
     (x).offset > (y).offset ? (x) : (y))

Definition at line 260 of file async.c.

                                                      : \
     (x).page != (y).page ? (x) : \
     (x).offset > (y).offset ? (x) : (y))

QUEUE_POS_MIN

#define QUEUE_POS_MIN	(		x,
			y
	)

Value:

    (asyncQueuePagePrecedes((x).page, (y).page) ? (x) : \
     (x).page != (y).page ? (y) : \
     (x).offset < (y).offset ? (x) : (y))

Definition at line 254 of file async.c.

                                                      : \
     (x).page != (y).page ? (y) : \
     (x).offset < (y).offset ? (x) : (y))

QUEUE_POS_OFFSET

#define QUEUE_POS_OFFSET

(

x

)

((x).offset)

Definition at line 239 of file async.c.

QUEUE_POS_PAGE

#define QUEUE_POS_PAGE

(

x

)

((x).page)

Definition at line 238 of file async.c.

QUEUE_POS_PRECEDES

#define QUEUE_POS_PRECEDES	(		x,
			y
	)

Value:

    (asyncQueuePagePrecedes((x).page, (y).page) || \
     ((x).page == (y).page && (x).offset < (y).offset))

Definition at line 266 of file async.c.

QUEUE_STOP_PAGE

#define QUEUE_STOP_PAGE   (asyncQueueControl->stopPage)

Definition at line 360 of file async.c.

QUEUE_TAIL

#define QUEUE_TAIL   (asyncQueueControl->tail)

Definition at line 359 of file async.c.

QUEUEALIGN

#define QUEUEALIGN

(

len

)

INTALIGN(len)

Definition at line 225 of file async.c.

SET_QUEUE_POS

#define SET_QUEUE_POS	(		x,
			y,
			z
	)

Value:

    do { \
        (x).page = (y); \
        (x).offset = (z); \
    } while (0)

Definition at line 241 of file async.c.

       { \
        (x).page = (y); \
        (x).offset = (z); \
    } while (0)

Typedef Documentation

ActionList

typedef struct ActionList ActionList

AsyncQueueControl

typedef struct AsyncQueueControl AsyncQueueControl

AsyncQueueEntry

typedef struct AsyncQueueEntry AsyncQueueEntry

ChannelName

typedef struct ChannelName ChannelName

GlobalChannelEntry

typedef struct GlobalChannelEntry GlobalChannelEntry

GlobalChannelKey

typedef struct GlobalChannelKey GlobalChannelKey

ListenerEntry

typedef struct ListenerEntry ListenerEntry

Notification

typedef struct Notification Notification

NotificationList

typedef struct NotificationList NotificationList

PendingListenEntry

typedef struct PendingListenEntry PendingListenEntry

QueueBackendStatus

typedef struct QueueBackendStatus QueueBackendStatus

QueuePosition

typedef struct QueuePosition QueuePosition

Enumeration Type Documentation

ListenActionKind

enum ListenActionKind

Enumerator
LISTEN_LISTEN
LISTEN_UNLISTEN
LISTEN_UNLISTEN_ALL

Definition at line 438 of file async.c.

{
    LISTEN_LISTEN,
    LISTEN_UNLISTEN,
    LISTEN_UNLISTEN_ALL,
} ListenActionKind;

PendingListenAction

enum PendingListenAction

Enumerator
PENDING_LISTEN
PENDING_UNLISTEN

Definition at line 468 of file async.c.

{
    PENDING_LISTEN,
    PENDING_UNLISTEN,
} PendingListenAction;

Function Documentation

AddEventToPendingNotifies()

static void AddEventToPendingNotifies ( Notification *  n )

static

Definition at line 3164 of file async.c.

{
    Assert(pendingNotifies->events != NIL);
 
    /* Create the hash tables if it's time to */
    if (list_length(pendingNotifies->events) >= MIN_HASHABLE_NOTIFIES &&
        pendingNotifies->hashtab == NULL)
    {
        HASHCTL     hash_ctl;
        ListCell   *l;
 
        /* Create the hash table */
        hash_ctl.keysize = sizeof(Notification *);
        hash_ctl.entrysize = sizeof(struct NotificationHash);
        hash_ctl.hash = notification_hash;
        hash_ctl.match = notification_match;
        hash_ctl.hcxt = CurTransactionContext;
        pendingNotifies->hashtab =
            hash_create("Pending Notifies",
                        256L,
                        &hash_ctl,
                        HASH_ELEM | HASH_FUNCTION | HASH_COMPARE | HASH_CONTEXT);
 
        /* Create the unique channel name table */
        Assert(pendingNotifies->uniqueChannelHash == NULL);
        hash_ctl.keysize = NAMEDATALEN;
        hash_ctl.entrysize = sizeof(ChannelName);
        hash_ctl.hcxt = CurTransactionContext;
        pendingNotifies->uniqueChannelHash =
            hash_create("Pending Notify Channel Names",
                        64L,
                        &hash_ctl,
                        HASH_ELEM | HASH_STRINGS | HASH_CONTEXT);
 
        /* Insert all the already-existing events */
        foreach(l, pendingNotifies->events)
        {
            Notification *oldn = (Notification *) lfirst(l);
            char       *channel = oldn->data;
            bool        found;
 
            (void) hash_search(pendingNotifies->hashtab,
                               &oldn,
                               HASH_ENTER,
                               &found);
            Assert(!found);
 
            /* Add channel name to uniqueChannelHash; might be there already */
            (void) hash_search(pendingNotifies->uniqueChannelHash,
                               channel,
                               HASH_ENTER,
                               NULL);
        }
    }
 
    /* Add new event to the list, in order */
    pendingNotifies->events = lappend(pendingNotifies->events, n);
 
    /* Add event to the hash tables if needed */
    if (pendingNotifies->hashtab != NULL)
    {
        char       *channel = n->data;
        bool        found;
 
        (void) hash_search(pendingNotifies->hashtab,
                           &n,
                           HASH_ENTER,
                           &found);
        Assert(!found);
 
        /* Add channel name to uniqueChannelHash; might be there already */
        (void) hash_search(pendingNotifies->uniqueChannelHash,
                           channel,
                           HASH_ENTER,
                           NULL);
    }
}

References Assert, CurTransactionContext, Notification::data, NotificationList::events, fb(), HASH_COMPARE, HASH_CONTEXT, hash_create(), HASH_ELEM, HASH_ENTER, HASH_FUNCTION, hash_search(), HASH_STRINGS, NotificationList::hashtab, lappend(), lfirst, list_length(), MIN_HASHABLE_NOTIFIES, NAMEDATALEN, NIL, notification_hash(), notification_match(), pendingNotifies, and NotificationList::uniqueChannelHash.

Referenced by Async_Notify(), and AtSubCommit_Notify().

ApplyPendingListenActions()

static void ApplyPendingListenActions ( bool  isCommit )

static

Definition at line 1719 of file async.c.

{
    HASH_SEQ_STATUS seq;
    PendingListenEntry *pending;
 
    /* Quick exit if nothing to do */
    if (pendingListenActions == NULL)
        return;
 
    /* We made a globalChannelTable before building pendingListenActions */
    if (globalChannelTable == NULL)
        elog(PANIC, "global channel table missing post-commit/abort");
 
    /* For each staged action ... */
    hash_seq_init(&seq, pendingListenActions);
    while ((pending = (PendingListenEntry *) hash_seq_search(&seq)) != NULL)
    {
        GlobalChannelKey key;
        GlobalChannelEntry *entry;
        bool        removeLocal = true;
        bool        foundListener = false;
 
        /*
         * Find the global entry for this channel.  If isCommit, it had better
         * exist (it was created in PreCommit).  In an abort, it might not
         * exist, in which case we are not listening and should discard any
         * local entry that PreCommit may have managed to create.
         */
        GlobalChannelKeyInit(&key, MyDatabaseId, pending->channel);
        entry = dshash_find(globalChannelTable, &key, true);
        if (entry != NULL)
        {
            /* Scan entry to find the ListenerEntry for this backend */
            ListenerEntry *listeners;
 
            listeners = (ListenerEntry *)
                dsa_get_address(globalChannelDSA, entry->listenersArray);
 
            for (int i = 0; i < entry->numListeners; i++)
            {
                if (listeners[i].procNo != MyProcNumber)
                    continue;
                foundListener = true;
                if (isCommit)
                {
                    if (pending->action == PENDING_LISTEN)
                    {
                        /*
                         * LISTEN being committed: entry is now permanent.
                         * localChannelTable entry was created during
                         * PreCommit and should be kept.
                         */
                        listeners[i].removeOnAbort = false;
                        removeLocal = false;
                    }
                    else
                    {
                        /*
                         * UNLISTEN being committed: remove pre-allocated
                         * entries from both tables.
                         */
                        RemoveListenerFromChannel(&entry, listeners, i);
                    }
                }
                else
                {
                    /*
                     * Note: this part is reachable only if the transaction
                     * aborts after PreCommit_Notify() has made some
                     * pendingListenActions entries, so it's pretty hard to
                     * test.
                     */
                    if (listeners[i].removeOnAbort)
                    {
                        /*
                         * Staged LISTEN (or LISTEN+UNLISTEN) being aborted,
                         * so remove pre-allocated entries from both tables.
                         */
                        RemoveListenerFromChannel(&entry, listeners, i);
                    }
                    else
                    {
                        /*
                         * Entry predates this transaction, so keep the
                         * localChannelTable entry.
                         */
                        removeLocal = false;
                    }
                }
                break;          /* there shouldn't be another match */
            }
 
            /* We might have already released the entry by removing it */
            if (entry != NULL)
                dshash_release_lock(globalChannelTable, entry);
        }
 
        /*
         * If we're committing a LISTEN action, we should have found a
         * matching ListenerEntry, but otherwise it's okay if we didn't.
         */
        if (isCommit && pending->action == PENDING_LISTEN && !foundListener)
            elog(PANIC, "could not find listener entry for channel \"%s\" backend %d",
                 pending->channel, MyProcNumber);
 
        /*
         * If we did not find a globalChannelTable entry for our backend, or
         * if we are unlistening, remove any localChannelTable entry that may
         * exist.  (Note in particular that this cleans up if we created a
         * localChannelTable entry and then failed while trying to create a
         * globalChannelTable entry.)
         */
        if (removeLocal && localChannelTable != NULL)
            (void) hash_search(localChannelTable, pending->channel,
                               HASH_REMOVE, NULL);
    }
}

References PendingListenEntry::action, PendingListenEntry::channel, dsa_get_address(), dshash_find(), dshash_release_lock(), elog, fb(), globalChannelDSA, GlobalChannelKeyInit(), globalChannelTable, HASH_REMOVE, hash_search(), hash_seq_init(), hash_seq_search(), i, GlobalChannelEntry::listenersArray, localChannelTable, MyDatabaseId, MyProcNumber, GlobalChannelEntry::numListeners, PANIC, PENDING_LISTEN, pendingListenActions, RemoveListenerFromChannel(), and ListenerEntry::removeOnAbort.

Referenced by AtAbort_Notify(), and AtCommit_Notify().

Async_Listen()

void Async_Listen

(

const char *

channel

)

Definition at line 1043 of file async.c.

{
    if (Trace_notify)
        elog(DEBUG1, "Async_Listen(%s,%d)", channel, MyProcPid);
 
    queue_listen(LISTEN_LISTEN, channel);
}

References DEBUG1, elog, LISTEN_LISTEN, MyProcPid, queue_listen(), and Trace_notify.

Referenced by standard_ProcessUtility().

Async_Notify()

void Async_Notify	(	const char *	channel,
		const char *	payload
	)

Definition at line 894 of file async.c.

{
    int         my_level = GetCurrentTransactionNestLevel();
    size_t      channel_len;
    size_t      payload_len;
    Notification *n;
    MemoryContext oldcontext;
 
    if (IsParallelWorker())
        elog(ERROR, "cannot send notifications from a parallel worker");
 
    if (Trace_notify)
        elog(DEBUG1, "Async_Notify(%s)", channel);
 
    channel_len = channel ? strlen(channel) : 0;
    payload_len = payload ? strlen(payload) : 0;
 
    /* a channel name must be specified */
    if (channel_len == 0)
        ereport(ERROR,
                (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
                 errmsg("channel name cannot be empty")));
 
    /* enforce length limits */
    if (channel_len >= NAMEDATALEN)
        ereport(ERROR,
                (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
                 errmsg("channel name too long")));
 
    if (payload_len >= NOTIFY_PAYLOAD_MAX_LENGTH)
        ereport(ERROR,
                (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
                 errmsg("payload string too long")));
 
    /*
     * We must construct the Notification entry, even if we end up not using
     * it, in order to compare it cheaply to existing list entries.
     *
     * The notification list needs to live until end of transaction, so store
     * it in the transaction context.
     */
    oldcontext = MemoryContextSwitchTo(CurTransactionContext);
 
    n = (Notification *) palloc(offsetof(Notification, data) +
                                channel_len + payload_len + 2);
    n->channel_len = channel_len;
    n->payload_len = payload_len;
    strcpy(n->data, channel);
    if (payload)
        strcpy(n->data + channel_len + 1, payload);
    else
        n->data[channel_len + 1] = '\0';
 
    if (pendingNotifies == NULL || my_level > pendingNotifies->nestingLevel)
    {
        NotificationList *notifies;
 
        /*
         * First notify event in current (sub)xact. Note that we allocate the
         * NotificationList in TopTransactionContext; the nestingLevel might
         * get changed later by AtSubCommit_Notify.
         */
        notifies = (NotificationList *)
            MemoryContextAlloc(TopTransactionContext,
                               sizeof(NotificationList));
        notifies->nestingLevel = my_level;
        notifies->events = list_make1(n);
        /* We certainly don't need a hashtable yet */
        notifies->hashtab = NULL;
        /* We won't build uniqueChannelNames/Hash till later, either */
        notifies->uniqueChannelNames = NIL;
        notifies->uniqueChannelHash = NULL;
        notifies->upper = pendingNotifies;
        pendingNotifies = notifies;
    }
    else
    {
        /* Now check for duplicates */
        if (AsyncExistsPendingNotify(n))
        {
            /* It's a dup, so forget it */
            pfree(n);
            MemoryContextSwitchTo(oldcontext);
            return;
        }
 
        /* Append more events to existing list */
        AddEventToPendingNotifies(n);
    }
 
    MemoryContextSwitchTo(oldcontext);
}

References AddEventToPendingNotifies(), AsyncExistsPendingNotify(), Notification::channel_len, CurTransactionContext, Notification::data, data, DEBUG1, elog, ereport, errcode(), errmsg, ERROR, fb(), GetCurrentTransactionNestLevel(), IsParallelWorker, list_make1, MemoryContextAlloc(), MemoryContextSwitchTo(), NAMEDATALEN, NotificationList::nestingLevel, NIL, NOTIFY_PAYLOAD_MAX_LENGTH, palloc(), Notification::payload_len, pendingNotifies, pfree(), TopTransactionContext, and Trace_notify.

Referenced by pg_notify(), standard_ProcessUtility(), and triggered_change_notification().

Async_Unlisten()

void Async_Unlisten

(

const char *

channel

)

Definition at line 1057 of file async.c.

{
    if (Trace_notify)
        elog(DEBUG1, "Async_Unlisten(%s,%d)", channel, MyProcPid);
 
    /* If we couldn't possibly be listening, no need to queue anything */
    if (pendingActions == NULL && !unlistenExitRegistered)
        return;
 
    queue_listen(LISTEN_UNLISTEN, channel);
}

References DEBUG1, elog, fb(), LISTEN_UNLISTEN, MyProcPid, pendingActions, queue_listen(), Trace_notify, and unlistenExitRegistered.

Referenced by standard_ProcessUtility().

Async_UnlistenAll()

void Async_UnlistenAll

(

void

)

Definition at line 1075 of file async.c.

{
    if (Trace_notify)
        elog(DEBUG1, "Async_UnlistenAll(%d)", MyProcPid);
 
    /* If we couldn't possibly be listening, no need to queue anything */
    if (pendingActions == NULL && !unlistenExitRegistered)
        return;
 
    queue_listen(LISTEN_UNLISTEN_ALL, "");
}

References DEBUG1, elog, fb(), LISTEN_UNLISTEN_ALL, MyProcPid, pendingActions, queue_listen(), Trace_notify, and unlistenExitRegistered.

Referenced by DiscardAll(), and standard_ProcessUtility().

Async_UnlistenOnExit()

static void Async_UnlistenOnExit ( int  code, Datum  arg  )

static

Definition at line 1147 of file async.c.

{
    CleanupListenersOnExit();
    asyncQueueUnregister();
}

References asyncQueueUnregister(), and CleanupListenersOnExit().

Referenced by BecomeRegisteredListener().

AsyncExistsPendingNotify()

static bool AsyncExistsPendingNotify ( Notification *  n )

static

Definition at line 3123 of file async.c.

{
    if (pendingNotifies == NULL)
        return false;
 
    if (pendingNotifies->hashtab != NULL)
    {
        /* Use the hash table to probe for a match */
        if (hash_search(pendingNotifies->hashtab,
                        &n,
                        HASH_FIND,
                        NULL))
            return true;
    }
    else
    {
        /* Must scan the event list */
        ListCell   *l;
 
        foreach(l, pendingNotifies->events)
        {
            Notification *oldn = (Notification *) lfirst(l);
 
            if (n->channel_len == oldn->channel_len &&
                n->payload_len == oldn->payload_len &&
                memcmp(n->data, oldn->data,
                       n->channel_len + n->payload_len + 2) == 0)
                return true;
        }
    }
 
    return false;
}

References Notification::channel_len, Notification::data, NotificationList::events, fb(), HASH_FIND, hash_search(), NotificationList::hashtab, lfirst, Notification::payload_len, and pendingNotifies.

Referenced by Async_Notify(), and AtSubCommit_Notify().

AsyncNotifyFreezeXids()

void AsyncNotifyFreezeXids

(

TransactionId

newFrozenXid

)

Definition at line 2952 of file async.c.

{
    QueuePosition pos;
    QueuePosition head;
    int64       curpage = -1;
    int         slotno = -1;
    char       *page_buffer = NULL;
    bool        page_dirty = false;
 
    /*
     * Acquire locks in the correct order to avoid deadlocks. As per the
     * locking protocol: NotifyQueueTailLock, then NotifyQueueLock, then SLRU
     * bank locks.
     *
     * We only need SHARED mode since we're just reading the head/tail
     * positions, not modifying them.
     */
    LWLockAcquire(NotifyQueueTailLock, LW_SHARED);
    LWLockAcquire(NotifyQueueLock, LW_SHARED);
 
    pos = QUEUE_TAIL;
    head = QUEUE_HEAD;
 
    /* Release NotifyQueueLock early, we only needed to read the positions */
    LWLockRelease(NotifyQueueLock);
 
    /*
     * Scan the queue from tail to head, freezing XIDs as needed. We hold
     * NotifyQueueTailLock throughout to ensure the tail doesn't move while
     * we're working.
     */
    while (!QUEUE_POS_EQUAL(pos, head))
    {
        AsyncQueueEntry *qe;
        TransactionId xid;
        int64       pageno = QUEUE_POS_PAGE(pos);
        int         offset = QUEUE_POS_OFFSET(pos);
 
        /* If we need a different page, release old lock and get new one */
        if (pageno != curpage)
        {
            LWLock     *lock;
 
            /* Release previous page if any */
            if (slotno >= 0)
            {
                if (page_dirty)
                {
                    NotifyCtl->shared->page_dirty[slotno] = true;
                    page_dirty = false;
                }
                LWLockRelease(SimpleLruGetBankLock(NotifyCtl, curpage));
            }
 
            lock = SimpleLruGetBankLock(NotifyCtl, pageno);
            LWLockAcquire(lock, LW_EXCLUSIVE);
            slotno = SimpleLruReadPage(NotifyCtl, pageno, true, &pos);
            page_buffer = NotifyCtl->shared->page_buffer[slotno];
            curpage = pageno;
        }
 
        qe = (AsyncQueueEntry *) (page_buffer + offset);
        xid = qe->xid;
 
        if (TransactionIdIsNormal(xid) &&
            TransactionIdPrecedes(xid, newFrozenXid))
        {
            if (TransactionIdDidCommit(xid))
            {
                qe->xid = FrozenTransactionId;
                page_dirty = true;
            }
            else
            {
                qe->xid = InvalidTransactionId;
                page_dirty = true;
            }
        }
 
        /* Advance to next entry */
        asyncQueueAdvance(&pos, qe->length);
    }
 
    /* Release final page lock if we acquired one */
    if (slotno >= 0)
    {
        if (page_dirty)
            NotifyCtl->shared->page_dirty[slotno] = true;
        LWLockRelease(SimpleLruGetBankLock(NotifyCtl, curpage));
    }
 
    LWLockRelease(NotifyQueueTailLock);
}

References asyncQueueAdvance(), fb(), FrozenTransactionId, InvalidTransactionId, LW_EXCLUSIVE, LW_SHARED, LWLockAcquire(), LWLockRelease(), NotifyCtl, QUEUE_HEAD, QUEUE_POS_EQUAL, QUEUE_POS_OFFSET, QUEUE_POS_PAGE, QUEUE_TAIL, SimpleLruGetBankLock(), SimpleLruReadPage(), TransactionIdDidCommit(), TransactionIdIsNormal, and TransactionIdPrecedes().

Referenced by vac_truncate_clog().

asyncQueueAddEntries()

static ListCell * asyncQueueAddEntries ( ListCell *  nextNotify )

static

Definition at line 2040 of file async.c.

{
    AsyncQueueEntry qe;
    QueuePosition queue_head;
    int64       pageno;
    int         offset;
    int         slotno;
    LWLock     *prevlock;
 
    /*
     * We work with a local copy of QUEUE_HEAD, which we write back to shared
     * memory upon exiting.  The reason for this is that if we have to advance
     * to a new page, SimpleLruZeroPage might fail (out of disk space, for
     * instance), and we must not advance QUEUE_HEAD if it does.  (Otherwise,
     * subsequent insertions would try to put entries into a page that slru.c
     * thinks doesn't exist yet.)  So, use a local position variable.  Note
     * that if we do fail, any already-inserted queue entries are forgotten;
     * this is okay, since they'd be useless anyway after our transaction
     * rolls back.
     */
    queue_head = QUEUE_HEAD;
 
    /*
     * If this is the first write since the postmaster started, we need to
     * initialize the first page of the async SLRU.  Otherwise, the current
     * page should be initialized already, so just fetch it.
     */
    pageno = QUEUE_POS_PAGE(queue_head);
    prevlock = SimpleLruGetBankLock(NotifyCtl, pageno);
 
    /* We hold both NotifyQueueLock and SLRU bank lock during this operation */
    LWLockAcquire(prevlock, LW_EXCLUSIVE);
 
    if (QUEUE_POS_IS_ZERO(queue_head))
        slotno = SimpleLruZeroPage(NotifyCtl, pageno);
    else
        slotno = SimpleLruReadPage(NotifyCtl, pageno, true, &queue_head);
 
    /* Note we mark the page dirty before writing in it */
    NotifyCtl->shared->page_dirty[slotno] = true;
 
    while (nextNotify != NULL)
    {
        Notification *n = (Notification *) lfirst(nextNotify);
 
        /* Construct a valid queue entry in local variable qe */
        asyncQueueNotificationToEntry(n, &qe);
 
        offset = QUEUE_POS_OFFSET(queue_head);
 
        /* Check whether the entry really fits on the current page */
        if (offset + qe.length <= QUEUE_PAGESIZE)
        {
            /* OK, so advance nextNotify past this item */
            nextNotify = lnext(pendingNotifies->events, nextNotify);
        }
        else
        {
            /*
             * Write a dummy entry to fill up the page. Actually readers will
             * only check dboid and since it won't match any reader's database
             * OID, they will ignore this entry and move on.
             */
            qe.length = QUEUE_PAGESIZE - offset;
            qe.dboid = InvalidOid;
            qe.xid = InvalidTransactionId;
            qe.data[0] = '\0';  /* empty channel */
            qe.data[1] = '\0';  /* empty payload */
        }
 
        /* Now copy qe into the shared buffer page */
        memcpy(NotifyCtl->shared->page_buffer[slotno] + offset,
               &qe,
               qe.length);
 
        /* Advance queue_head appropriately, and detect if page is full */
        if (asyncQueueAdvance(&(queue_head), qe.length))
        {
            LWLock     *lock;
 
            pageno = QUEUE_POS_PAGE(queue_head);
            lock = SimpleLruGetBankLock(NotifyCtl, pageno);
            if (lock != prevlock)
            {
                LWLockRelease(prevlock);
                LWLockAcquire(lock, LW_EXCLUSIVE);
                prevlock = lock;
            }
 
            /*
             * Page is full, so we're done here, but first fill the next page
             * with zeroes.  The reason to do this is to ensure that slru.c's
             * idea of the head page is always the same as ours, which avoids
             * boundary problems in SimpleLruTruncate.  The test in
             * asyncQueueIsFull() ensured that there is room to create this
             * page without overrunning the queue.
             */
            slotno = SimpleLruZeroPage(NotifyCtl, QUEUE_POS_PAGE(queue_head));
 
            /*
             * If the new page address is a multiple of QUEUE_CLEANUP_DELAY,
             * set flag to remember that we should try to advance the tail
             * pointer (we don't want to actually do that right here).
             */
            if (QUEUE_POS_PAGE(queue_head) % QUEUE_CLEANUP_DELAY == 0)
                tryAdvanceTail = true;
 
            /* And exit the loop */
            break;
        }
    }
 
    /* Success, so update the global QUEUE_HEAD */
    QUEUE_HEAD = queue_head;
 
    LWLockRelease(prevlock);
 
    return nextNotify;
}

References asyncQueueAdvance(), asyncQueueNotificationToEntry(), NotificationList::events, fb(), InvalidOid, InvalidTransactionId, lfirst, lnext(), LW_EXCLUSIVE, LWLockAcquire(), LWLockRelease(), memcpy(), NotifyCtl, pendingNotifies, QUEUE_CLEANUP_DELAY, QUEUE_HEAD, QUEUE_PAGESIZE, QUEUE_POS_IS_ZERO, QUEUE_POS_OFFSET, QUEUE_POS_PAGE, SimpleLruGetBankLock(), SimpleLruReadPage(), SimpleLruZeroPage(), and tryAdvanceTail.

Referenced by PreCommit_Notify().

asyncQueueAdvance()

static bool asyncQueueAdvance ( volatile QueuePosition *  position, int  entryLength  )

static

Definition at line 1971 of file async.c.

{
    int64       pageno = QUEUE_POS_PAGE(*position);
    int         offset = QUEUE_POS_OFFSET(*position);
    bool        pageJump = false;
 
    /*
     * Move to the next writing position: First jump over what we have just
     * written or read.
     */
    offset += entryLength;
    Assert(offset <= QUEUE_PAGESIZE);
 
    /*
     * In a second step check if another entry can possibly be written to the
     * page. If so, stay here, we have reached the next position. If not, then
     * we need to move on to the next page.
     */
    if (offset + QUEUEALIGN(AsyncQueueEntryEmptySize) > QUEUE_PAGESIZE)
    {
        pageno++;
        offset = 0;
        pageJump = true;
    }
 
    SET_QUEUE_POS(*position, pageno, offset);
    return pageJump;
}

References Assert, AsyncQueueEntryEmptySize, fb(), QUEUE_PAGESIZE, QUEUE_POS_OFFSET, QUEUE_POS_PAGE, QUEUEALIGN, and SET_QUEUE_POS.

Referenced by AsyncNotifyFreezeXids(), asyncQueueAddEntries(), and asyncQueueProcessPageEntries().

asyncQueueAdvanceTail()

static void asyncQueueAdvanceTail ( void  )

static

Definition at line 2870 of file async.c.

{
    QueuePosition min;
    int64       oldtailpage;
    int64       newtailpage;
    int64       boundary;
 
    /* Restrict task to one backend per cluster; see SimpleLruTruncate(). */
    LWLockAcquire(NotifyQueueTailLock, LW_EXCLUSIVE);
 
    /*
     * Compute the new tail.  Pre-v13, it's essential that QUEUE_TAIL be exact
     * (ie, exactly match at least one backend's queue position), so it must
     * be updated atomically with the actual computation.  Since v13, we could
     * get away with not doing it like that, but it seems prudent to keep it
     * so.
     *
     * Also, because incoming backends will scan forward from QUEUE_TAIL, that
     * must be advanced before we can truncate any data.  Thus, QUEUE_TAIL is
     * the logical tail, while QUEUE_STOP_PAGE is the physical tail, or oldest
     * un-truncated page.  When QUEUE_STOP_PAGE != QUEUE_POS_PAGE(QUEUE_TAIL),
     * there are pages we can truncate but haven't yet finished doing so.
     *
     * For concurrency's sake, we don't want to hold NotifyQueueLock while
     * performing SimpleLruTruncate.  This is OK because no backend will try
     * to access the pages we are in the midst of truncating.
     */
    LWLockAcquire(NotifyQueueLock, LW_EXCLUSIVE);
    min = QUEUE_HEAD;
    for (ProcNumber i = QUEUE_FIRST_LISTENER; i != INVALID_PROC_NUMBER; i = QUEUE_NEXT_LISTENER(i))
    {
        Assert(QUEUE_BACKEND_PID(i) != InvalidPid);
        min = QUEUE_POS_MIN(min, QUEUE_BACKEND_POS(i));
    }
    QUEUE_TAIL = min;
    oldtailpage = QUEUE_STOP_PAGE;
    LWLockRelease(NotifyQueueLock);
 
    /*
     * We can truncate something if the global tail advanced across an SLRU
     * segment boundary.
     *
     * XXX it might be better to truncate only once every several segments, to
     * reduce the number of directory scans.
     */
    newtailpage = QUEUE_POS_PAGE(min);
    boundary = newtailpage - (newtailpage % SLRU_PAGES_PER_SEGMENT);
    if (asyncQueuePagePrecedes(oldtailpage, boundary))
    {
        /*
         * SimpleLruTruncate() will ask for SLRU bank locks but will also
         * release the lock again.
         */
        SimpleLruTruncate(NotifyCtl, newtailpage);
 
        LWLockAcquire(NotifyQueueLock, LW_EXCLUSIVE);
        QUEUE_STOP_PAGE = newtailpage;
        LWLockRelease(NotifyQueueLock);
    }
 
    LWLockRelease(NotifyQueueTailLock);
}

References Assert, asyncQueuePagePrecedes(), fb(), i, INVALID_PROC_NUMBER, InvalidPid, LW_EXCLUSIVE, LWLockAcquire(), LWLockRelease(), NotifyCtl, QUEUE_BACKEND_PID, QUEUE_BACKEND_POS, QUEUE_FIRST_LISTENER, QUEUE_HEAD, QUEUE_NEXT_LISTENER, QUEUE_POS_MIN, QUEUE_POS_PAGE, QUEUE_STOP_PAGE, QUEUE_TAIL, SimpleLruTruncate(), and SLRU_PAGES_PER_SEGMENT.

Referenced by AtCommit_Notify(), and pg_notification_queue_usage().

asyncQueueErrdetailForIoError()

static int asyncQueueErrdetailForIoError ( const void *  opaque_data )

static

Definition at line 627 of file async.c.

{
    const QueuePosition *position = opaque_data;
 
    return errdetail("Could not access async queue at page %" PRId64 ", offset %d.",
                     position->page, position->offset);
}

References errdetail(), fb(), QueuePosition::offset, and QueuePosition::page.

Referenced by AsyncShmemRequest().

asyncQueueFillWarning()

static void asyncQueueFillWarning ( void  )

static

Definition at line 2211 of file async.c.

{
    double      fillDegree;
    TimestampTz t;
 
    fillDegree = asyncQueueUsage();
    if (fillDegree < 0.5)
        return;
 
    t = GetCurrentTimestamp();
 
    if (TimestampDifferenceExceeds(asyncQueueControl->lastQueueFillWarn,
                                   t, QUEUE_FULL_WARN_INTERVAL))
    {
        QueuePosition min = QUEUE_HEAD;
        int32       minPid = InvalidPid;
 
        for (ProcNumber i = QUEUE_FIRST_LISTENER; i != INVALID_PROC_NUMBER; i = QUEUE_NEXT_LISTENER(i))
        {
            Assert(QUEUE_BACKEND_PID(i) != InvalidPid);
            min = QUEUE_POS_MIN(min, QUEUE_BACKEND_POS(i));
            if (QUEUE_POS_EQUAL(min, QUEUE_BACKEND_POS(i)))
                minPid = QUEUE_BACKEND_PID(i);
        }
 
        ereport(WARNING,
                (errmsg("NOTIFY queue is %.0f%% full", fillDegree * 100),
                 (minPid != InvalidPid ?
                  errdetail("The server process with PID %d is among those with the oldest transactions.", minPid)
                  : 0),
                 (minPid != InvalidPid ?
                  errhint("The NOTIFY queue cannot be emptied until that process ends its current transaction.")
                  : 0)));
 
        asyncQueueControl->lastQueueFillWarn = t;
    }
}

References Assert, asyncQueueControl, asyncQueueUsage(), ereport, errdetail(), errhint(), errmsg, fb(), GetCurrentTimestamp(), i, INVALID_PROC_NUMBER, InvalidPid, AsyncQueueControl::lastQueueFillWarn, QUEUE_BACKEND_PID, QUEUE_BACKEND_POS, QUEUE_FIRST_LISTENER, QUEUE_FULL_WARN_INTERVAL, QUEUE_HEAD, QUEUE_NEXT_LISTENER, QUEUE_POS_EQUAL, QUEUE_POS_MIN, TimestampDifferenceExceeds(), and WARNING.

Referenced by PreCommit_Notify().

asyncQueueIsFull()

static bool asyncQueueIsFull ( void  )

static

Definition at line 1956 of file async.c.

{
    int64       headPage = QUEUE_POS_PAGE(QUEUE_HEAD);
    int64       tailPage = QUEUE_POS_PAGE(QUEUE_TAIL);
    int64       occupied = headPage - tailPage;
 
    return occupied >= max_notify_queue_pages;
}

References fb(), max_notify_queue_pages, QUEUE_HEAD, QUEUE_POS_PAGE, and QUEUE_TAIL.

Referenced by PreCommit_Notify().

asyncQueueNotificationToEntry()

static void asyncQueueNotificationToEntry ( Notification *  n, AsyncQueueEntry *  qe  )

static

Definition at line 2004 of file async.c.

{
    size_t      channellen = n->channel_len;
    size_t      payloadlen = n->payload_len;
    int         entryLength;
 
    Assert(channellen < NAMEDATALEN);
    Assert(payloadlen < NOTIFY_PAYLOAD_MAX_LENGTH);
 
    /* The terminators are already included in AsyncQueueEntryEmptySize */
    entryLength = AsyncQueueEntryEmptySize + payloadlen + channellen;
    entryLength = QUEUEALIGN(entryLength);
    qe->length = entryLength;
    qe->dboid = MyDatabaseId;
    qe->xid = GetCurrentTransactionId();
    qe->srcPid = MyProcPid;
    memcpy(qe->data, n->data, channellen + payloadlen + 2);
}

References Assert, AsyncQueueEntryEmptySize, Notification::channel_len, Notification::data, fb(), GetCurrentTransactionId(), memcpy(), MyDatabaseId, MyProcPid, NAMEDATALEN, NOTIFY_PAYLOAD_MAX_LENGTH, Notification::payload_len, and QUEUEALIGN.

Referenced by asyncQueueAddEntries().

asyncQueuePageDiff()

static int64 asyncQueuePageDiff ( int64  p, int64  q  )

inlinestatic

Definition at line 640 of file async.c.

{
    return p - q;
}

Referenced by SignalBackends().

asyncQueuePagePrecedes()

static bool asyncQueuePagePrecedes ( int64  p, int64  q  )

inlinestatic

Definition at line 650 of file async.c.

{
    return p < q;
}

Referenced by asyncQueueAdvanceTail(), and AsyncShmemRequest().

asyncQueueProcessPageEntries()

static bool asyncQueueProcessPageEntries ( QueuePosition *  current, QueuePosition  stop, Snapshot  snapshot  )

static

Definition at line 2732 of file async.c.

{
    int64       curpage = QUEUE_POS_PAGE(*current);
    int         slotno;
    char       *page_buffer;
    bool        reachedStop = false;
    bool        reachedEndOfPage;
 
    /*
     * We copy the entries into a local buffer to avoid holding the SLRU lock
     * while we transmit them to our frontend.  The local buffer must be
     * adequately aligned.
     */
    alignas(AsyncQueueEntry) char local_buf[QUEUE_PAGESIZE];
    char       *local_buf_end = local_buf;
 
    slotno = SimpleLruReadPage_ReadOnly(NotifyCtl, curpage, current);
    page_buffer = NotifyCtl->shared->page_buffer[slotno];
 
    do
    {
        QueuePosition thisentry = *current;
        AsyncQueueEntry *qe;
 
        if (QUEUE_POS_EQUAL(thisentry, stop))
            break;
 
        qe = (AsyncQueueEntry *) (page_buffer + QUEUE_POS_OFFSET(thisentry));
 
        /*
         * Advance *current over this message, possibly to the next page. As
         * noted in the comments for asyncQueueReadAllNotifications, we must
         * do this before possibly failing while processing the message.
         */
        reachedEndOfPage = asyncQueueAdvance(current, qe->length);
 
        /* Ignore messages destined for other databases */
        if (qe->dboid == MyDatabaseId)
        {
            if (XidInMVCCSnapshot(qe->xid, snapshot))
            {
                /*
                 * The source transaction is still in progress, so we can't
                 * process this message yet.  Break out of the loop, but first
                 * back up *current so we will reprocess the message next
                 * time.  (Note: it is unlikely but not impossible for
                 * TransactionIdDidCommit to fail, so we can't really avoid
                 * this advance-then-back-up behavior when dealing with an
                 * uncommitted message.)
                 *
                 * Note that we must test XidInMVCCSnapshot before we test
                 * TransactionIdDidCommit, else we might return a message from
                 * a transaction that is not yet visible to snapshots; compare
                 * the comments at the head of heapam_visibility.c.
                 *
                 * Also, while our own xact won't be listed in the snapshot,
                 * we need not check for TransactionIdIsCurrentTransactionId
                 * because our transaction cannot (yet) have queued any
                 * messages.
                 */
                *current = thisentry;
                reachedStop = true;
                break;
            }
 
            /*
             * Quick check for the case that we're not listening on any
             * channels, before calling TransactionIdDidCommit().  This makes
             * that case a little faster, but more importantly, it ensures
             * that if there's a bad entry in the queue for which
             * TransactionIdDidCommit() fails for some reason, we can skip
             * over it on the first LISTEN in a session, and not get stuck on
             * it indefinitely.  (This is a little trickier than it looks: it
             * works because BecomeRegisteredListener runs this code before we
             * have made the first entry in localChannelTable.)
             */
            if (LocalChannelTableIsEmpty())
                continue;
 
            if (TransactionIdDidCommit(qe->xid))
            {
                memcpy(local_buf_end, qe, qe->length);
                local_buf_end += qe->length;
            }
            else
            {
                /*
                 * The source transaction aborted or crashed, so we just
                 * ignore its notifications.
                 */
            }
        }
 
        /* Loop back if we're not at end of page */
    } while (!reachedEndOfPage);
 
    /* Release lock that we got from SimpleLruReadPage_ReadOnly() */
    LWLockRelease(SimpleLruGetBankLock(NotifyCtl, curpage));
 
    /*
     * Now that we have let go of the SLRU bank lock, send the notifications
     * to our backend
     */
    Assert(local_buf_end - local_buf <= BLCKSZ);
    for (char *p = local_buf; p < local_buf_end;)
    {
        AsyncQueueEntry *qe = (AsyncQueueEntry *) p;
 
        /* qe->data is the null-terminated channel name */
        char       *channel = qe->data;
 
        if (IsListeningOn(channel))
        {
            /* payload follows channel name */
            char       *payload = qe->data + strlen(channel) + 1;
 
            NotifyMyFrontEnd(channel, payload, qe->srcPid);
        }
 
        p += qe->length;
    }
 
    if (QUEUE_POS_EQUAL(*current, stop))
        reachedStop = true;
 
    return reachedStop;
}

References Assert, asyncQueueAdvance(), AsyncQueueEntry::data, fb(), IsListeningOn(), LocalChannelTableIsEmpty, LWLockRelease(), memcpy(), MyDatabaseId, NotifyCtl, NotifyMyFrontEnd(), QUEUE_PAGESIZE, QUEUE_POS_EQUAL, QUEUE_POS_OFFSET, QUEUE_POS_PAGE, SimpleLruGetBankLock(), SimpleLruReadPage_ReadOnly(), TransactionIdDidCommit(), and XidInMVCCSnapshot().

Referenced by asyncQueueReadAllNotifications().

asyncQueueReadAllNotifications()

static void asyncQueueReadAllNotifications ( void  )

static

Definition at line 2598 of file async.c.

{
    QueuePosition pos;
    QueuePosition head;
    Snapshot    snapshot;
 
    /*
     * Fetch current state, indicate to others that we have woken up, and that
     * we are in process of advancing our position.
     */
    LWLockAcquire(NotifyQueueLock, LW_SHARED);
    /* Assert checks that we have a valid state entry */
    Assert(MyProcPid == QUEUE_BACKEND_PID(MyProcNumber));
    QUEUE_BACKEND_WAKEUP_PENDING(MyProcNumber) = false;
    pos = QUEUE_BACKEND_POS(MyProcNumber);
    head = QUEUE_HEAD;
 
    if (QUEUE_POS_EQUAL(pos, head))
    {
        /* Nothing to do, we have read all notifications already. */
        LWLockRelease(NotifyQueueLock);
        return;
    }
 
    QUEUE_BACKEND_IS_ADVANCING(MyProcNumber) = true;
    LWLockRelease(NotifyQueueLock);
 
    /*----------
     * Get snapshot we'll use to decide which xacts are still in progress.
     * This is trickier than it might seem, because of race conditions.
     * Consider the following example:
     *
     * Backend 1:                    Backend 2:
     *
     * transaction starts
     * UPDATE foo SET ...;
     * NOTIFY foo;
     * commit starts
     * queue the notify message
     *                               transaction starts
     *                               LISTEN foo;  -- first LISTEN in session
     *                               SELECT * FROM foo WHERE ...;
     * commit to clog
     *                               commit starts
     *                               add backend 2 to array of listeners
     *                               advance to queue head (this code)
     *                               commit to clog
     *
     * Transaction 2's SELECT has not seen the UPDATE's effects, since that
     * wasn't committed yet.  Ideally we'd ensure that client 2 would
     * eventually get transaction 1's notify message, but there's no way
     * to do that; until we're in the listener array, there's no guarantee
     * that the notify message doesn't get removed from the queue.
     *
     * Therefore the coding technique transaction 2 is using is unsafe:
     * applications must commit a LISTEN before inspecting database state,
     * if they want to ensure they will see notifications about subsequent
     * changes to that state.
     *
     * What we do guarantee is that we'll see all notifications from
     * transactions committing after the snapshot we take here.
     * BecomeRegisteredListener has already added us to the listener array,
     * so no not-yet-committed messages can be removed from the queue
     * before we see them.
     *----------
     */
    snapshot = RegisterSnapshot(GetLatestSnapshot());
 
    /*
     * It is possible that we fail while trying to send a message to our
     * frontend (for example, because of encoding conversion failure).  If
     * that happens it is critical that we not try to send the same message
     * over and over again.  Therefore, we set ExitOnAnyError to upgrade any
     * ERRORs to FATAL, causing the client connection to be closed on error.
     *
     * We used to only skip over the offending message and try to soldier on,
     * but it was somewhat questionable to lose a notification and give the
     * client an ERROR instead.  A client application is not be prepared for
     * that and can't tell that a notification was missed.  It was also not
     * very useful in practice because notifications are often processed while
     * a connection is idle and reading a message from the client, and in that
     * state, any error is upgraded to FATAL anyway.  Closing the connection
     * is a clear signal to the application that it might have missed
     * notifications.
     */
    {
        bool        save_ExitOnAnyError = ExitOnAnyError;
        bool        reachedStop;
 
        ExitOnAnyError = true;
 
        do
        {
            /*
             * Process messages up to the stop position, end of page, or an
             * uncommitted message.
             *
             * Our stop position is what we found to be the head's position
             * when we entered this function. It might have changed already.
             * But if it has, we will receive (or have already received and
             * queued) another signal and come here again.
             *
             * We are not holding NotifyQueueLock here! The queue can only
             * extend beyond the head pointer (see above) and we leave our
             * backend's pointer where it is so nobody will truncate or
             * rewrite pages under us. Especially we don't want to hold a lock
             * while sending the notifications to the frontend.
             */
            reachedStop = asyncQueueProcessPageEntries(&pos, head, snapshot);
        } while (!reachedStop);
 
        /* Update shared state */
        LWLockAcquire(NotifyQueueLock, LW_SHARED);
        QUEUE_BACKEND_POS(MyProcNumber) = pos;
        QUEUE_BACKEND_IS_ADVANCING(MyProcNumber) = false;
        LWLockRelease(NotifyQueueLock);
 
        ExitOnAnyError = save_ExitOnAnyError;
    }
 
    /* Done with snapshot */
    UnregisterSnapshot(snapshot);
}

References Assert, asyncQueueProcessPageEntries(), ExitOnAnyError, fb(), GetLatestSnapshot(), LW_SHARED, LWLockAcquire(), LWLockRelease(), MyProcNumber, MyProcPid, QUEUE_BACKEND_IS_ADVANCING, QUEUE_BACKEND_PID, QUEUE_BACKEND_POS, QUEUE_BACKEND_WAKEUP_PENDING, QUEUE_HEAD, QUEUE_POS_EQUAL, RegisterSnapshot(), and UnregisterSnapshot().

Referenced by BecomeRegisteredListener(), and ProcessIncomingNotify().

asyncQueueUnregister()

static void asyncQueueUnregister ( void  )

static

Definition at line 1913 of file async.c.

{
    Assert(LocalChannelTableIsEmpty()); /* else caller error */
 
    if (!amRegisteredListener)  /* nothing to do */
        return;
 
    /*
     * Need exclusive lock here to manipulate list links.
     */
    LWLockAcquire(NotifyQueueLock, LW_EXCLUSIVE);
    /* Mark our entry as invalid */
    QUEUE_BACKEND_PID(MyProcNumber) = InvalidPid;
    QUEUE_BACKEND_DBOID(MyProcNumber) = InvalidOid;
    QUEUE_BACKEND_WAKEUP_PENDING(MyProcNumber) = false;
    QUEUE_BACKEND_IS_ADVANCING(MyProcNumber) = false;
    /* and remove it from the list */
    if (QUEUE_FIRST_LISTENER == MyProcNumber)
        QUEUE_FIRST_LISTENER = QUEUE_NEXT_LISTENER(MyProcNumber);
    else
    {
        for (ProcNumber i = QUEUE_FIRST_LISTENER; i != INVALID_PROC_NUMBER; i = QUEUE_NEXT_LISTENER(i))
        {
            if (QUEUE_NEXT_LISTENER(i) == MyProcNumber)
            {
                QUEUE_NEXT_LISTENER(i) = QUEUE_NEXT_LISTENER(MyProcNumber);
                break;
            }
        }
    }
    QUEUE_NEXT_LISTENER(MyProcNumber) = INVALID_PROC_NUMBER;
    LWLockRelease(NotifyQueueLock);
 
    /* mark ourselves as no longer listed in the global array */
    amRegisteredListener = false;
}

References amRegisteredListener, Assert, fb(), i, INVALID_PROC_NUMBER, InvalidOid, InvalidPid, LocalChannelTableIsEmpty, LW_EXCLUSIVE, LWLockAcquire(), LWLockRelease(), MyProcNumber, QUEUE_BACKEND_DBOID, QUEUE_BACKEND_IS_ADVANCING, QUEUE_BACKEND_PID, QUEUE_BACKEND_WAKEUP_PENDING, QUEUE_FIRST_LISTENER, and QUEUE_NEXT_LISTENER.

Referenced by Async_UnlistenOnExit(), AtAbort_Notify(), and AtCommit_Notify().

asyncQueueUsage()

static double asyncQueueUsage ( void  )

static

Definition at line 2190 of file async.c.

{
    int64       headPage = QUEUE_POS_PAGE(QUEUE_HEAD);
    int64       tailPage = QUEUE_POS_PAGE(QUEUE_TAIL);
    int64       occupied = headPage - tailPage;
 
    if (occupied == 0)
        return (double) 0;      /* fast exit for common case */
 
    return (double) occupied / (double) max_notify_queue_pages;
}

References fb(), max_notify_queue_pages, QUEUE_HEAD, QUEUE_POS_PAGE, and QUEUE_TAIL.

Referenced by asyncQueueFillWarning(), and pg_notification_queue_usage().

AsyncShmemInit()

static void AsyncShmemInit ( void *  arg )

static

Definition at line 829 of file async.c.

{
    SET_QUEUE_POS(QUEUE_HEAD, 0, 0);
    SET_QUEUE_POS(QUEUE_TAIL, 0, 0);
    QUEUE_STOP_PAGE = 0;
    QUEUE_FIRST_LISTENER = INVALID_PROC_NUMBER;
    asyncQueueControl->lastQueueFillWarn = 0;
    asyncQueueControl->globalChannelTableDSA = DSA_HANDLE_INVALID;
    asyncQueueControl->globalChannelTableDSH = DSHASH_HANDLE_INVALID;
    for (int i = 0; i < MaxBackends; i++)
    {
        QUEUE_BACKEND_PID(i) = InvalidPid;
        QUEUE_BACKEND_DBOID(i) = InvalidOid;
        QUEUE_NEXT_LISTENER(i) = INVALID_PROC_NUMBER;
        SET_QUEUE_POS(QUEUE_BACKEND_POS(i), 0, 0);
        QUEUE_BACKEND_WAKEUP_PENDING(i) = false;
        QUEUE_BACKEND_IS_ADVANCING(i) = false;
    }
 
    /*
     * During start or reboot, clean out the pg_notify directory.
     */
    (void) SlruScanDirectory(NotifyCtl, SlruScanDirCbDeleteAll, NULL);
}

References asyncQueueControl, DSA_HANDLE_INVALID, DSHASH_HANDLE_INVALID, fb(), AsyncQueueControl::globalChannelTableDSA, AsyncQueueControl::globalChannelTableDSH, i, INVALID_PROC_NUMBER, InvalidOid, InvalidPid, AsyncQueueControl::lastQueueFillWarn, MaxBackends, NotifyCtl, QUEUE_BACKEND_DBOID, QUEUE_BACKEND_IS_ADVANCING, QUEUE_BACKEND_PID, QUEUE_BACKEND_POS, QUEUE_BACKEND_WAKEUP_PENDING, QUEUE_FIRST_LISTENER, QUEUE_HEAD, QUEUE_NEXT_LISTENER, QUEUE_STOP_PAGE, QUEUE_TAIL, SET_QUEUE_POS, SlruScanDirCbDeleteAll(), and SlruScanDirectory().

AsyncShmemRequest()

static void AsyncShmemRequest ( void *  arg )

static

Definition at line 798 of file async.c.

{
    Size        size;
 
    size = mul_size(MaxBackends, sizeof(QueueBackendStatus));
    size = add_size(size, offsetof(AsyncQueueControl, backend));
 
    ShmemRequestStruct(.name = "Async Queue Control",
                       .size = size,
                       .ptr = (void **) &asyncQueueControl,
        );
 
    SimpleLruRequest(.desc = &NotifySlruDesc,
                     .name = "notify",
                     .Dir = "pg_notify",
 
    /* long segment names are used in order to avoid wraparound */
                     .long_segment_names = true,
 
                     .nslots = notify_buffers,
 
                     .sync_handler = SYNC_HANDLER_NONE,
                     .PagePrecedes = asyncQueuePagePrecedes,
                     .errdetail_for_io_error = asyncQueueErrdetailForIoError,
 
                     .buffer_tranche_id = LWTRANCHE_NOTIFY_BUFFER,
                     .bank_tranche_id = LWTRANCHE_NOTIFY_SLRU,
        );
}

References add_size(), asyncQueueControl, asyncQueueErrdetailForIoError(), asyncQueuePagePrecedes(), fb(), MaxBackends, mul_size(), name, notify_buffers, NotifySlruDesc, ShmemRequestStruct, SimpleLruRequest, and SYNC_HANDLER_NONE.

AtAbort_Notify()

void AtAbort_Notify

(

void

)

Definition at line 2420 of file async.c.

{
    /* Revert staged listen/unlisten changes */
    ApplyPendingListenActions(false);
 
    /* If we're no longer listening on anything, unregister */
    if (amRegisteredListener && LocalChannelTableIsEmpty())
        asyncQueueUnregister();
 
    /* And clean up */
    ClearPendingActionsAndNotifies();
}

References amRegisteredListener, ApplyPendingListenActions(), asyncQueueUnregister(), ClearPendingActionsAndNotifies(), and LocalChannelTableIsEmpty.

Referenced by AbortTransaction().

AtCommit_Notify()

void AtCommit_Notify

(

void

)

Definition at line 1378 of file async.c.

{
    /*
     * Allow transactions that have not executed LISTEN/UNLISTEN/NOTIFY to
     * return as soon as possible
     */
    if (!pendingActions && !pendingNotifies)
        return;
 
    if (Trace_notify)
        elog(DEBUG1, "AtCommit_Notify");
 
    /* Apply staged listen/unlisten changes */
    ApplyPendingListenActions(true);
 
    /* If no longer listening to anything, get out of listener array */
    if (amRegisteredListener && LocalChannelTableIsEmpty())
        asyncQueueUnregister();
 
    /*
     * Send signals to listening backends.  We need do this only if there are
     * pending notifies, which were previously added to the shared queue by
     * PreCommit_Notify().
     */
    if (pendingNotifies != NULL)
        SignalBackends();
 
    /*
     * If it's time to try to advance the global tail pointer, do that.
     *
     * (It might seem odd to do this in the sender, when more than likely the
     * listeners won't yet have read the messages we just sent.  However,
     * there's less contention if only the sender does it, and there is little
     * need for urgency in advancing the global tail.  So this typically will
     * be clearing out messages that were sent some time ago.)
     */
    if (tryAdvanceTail)
    {
        tryAdvanceTail = false;
        asyncQueueAdvanceTail();
    }
 
    /* And clean up */
    ClearPendingActionsAndNotifies();
}

References amRegisteredListener, ApplyPendingListenActions(), asyncQueueAdvanceTail(), asyncQueueUnregister(), ClearPendingActionsAndNotifies(), DEBUG1, elog, fb(), LocalChannelTableIsEmpty, pendingActions, pendingNotifies, SignalBackends(), Trace_notify, and tryAdvanceTail.

Referenced by CommitTransaction().

AtPrepare_Notify()

void AtPrepare_Notify

(

void

)

Definition at line 1160 of file async.c.

{
    /* It's not allowed to have any pending LISTEN/UNLISTEN/NOTIFY actions */
    if (pendingActions || pendingNotifies)
        ereport(ERROR,
                (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
                 errmsg("cannot PREPARE a transaction that has executed LISTEN, UNLISTEN, or NOTIFY")));
}

References ereport, errcode(), errmsg, ERROR, fb(), pendingActions, and pendingNotifies.

Referenced by PrepareTransaction().

AtSubAbort_Notify()

void AtSubAbort_Notify

(

void

)

Definition at line 2509 of file async.c.

{
    int         my_level = GetCurrentTransactionNestLevel();
 
    /*
     * All we have to do is pop the stack --- the actions/notifies made in
     * this subxact are no longer interesting, and the space will be freed
     * when CurTransactionContext is recycled. We still have to free the
     * ActionList and NotificationList objects themselves, though, because
     * those are allocated in TopTransactionContext.
     *
     * Note that there might be no entries at all, or no entries for the
     * current subtransaction level, either because none were ever created, or
     * because we reentered this routine due to trouble during subxact abort.
     */
    while (pendingActions != NULL &&
           pendingActions->nestingLevel >= my_level)
    {
        ActionList *childPendingActions = pendingActions;
 
        pendingActions = pendingActions->upper;
        pfree(childPendingActions);
    }
 
    while (pendingNotifies != NULL &&
           pendingNotifies->nestingLevel >= my_level)
    {
        NotificationList *childPendingNotifies = pendingNotifies;
 
        pendingNotifies = pendingNotifies->upper;
        pfree(childPendingNotifies);
    }
}

References fb(), GetCurrentTransactionNestLevel(), ActionList::nestingLevel, NotificationList::nestingLevel, pendingActions, pendingNotifies, pfree(), ActionList::upper, and NotificationList::upper.

Referenced by AbortSubTransaction().

AtSubCommit_Notify()

void AtSubCommit_Notify

(

void

)

Definition at line 2439 of file async.c.

{
    int         my_level = GetCurrentTransactionNestLevel();
 
    /* If there are actions at our nesting level, we must reparent them. */
    if (pendingActions != NULL &&
        pendingActions->nestingLevel >= my_level)
    {
        if (pendingActions->upper == NULL ||
            pendingActions->upper->nestingLevel < my_level - 1)
        {
            /* nothing to merge; give the whole thing to the parent */
            --pendingActions->nestingLevel;
        }
        else
        {
            ActionList *childPendingActions = pendingActions;
 
            pendingActions = pendingActions->upper;
 
            /*
             * Mustn't try to eliminate duplicates here --- see queue_listen()
             */
            pendingActions->actions =
                list_concat(pendingActions->actions,
                            childPendingActions->actions);
            pfree(childPendingActions);
        }
    }
 
    /* If there are notifies at our nesting level, we must reparent them. */
    if (pendingNotifies != NULL &&
        pendingNotifies->nestingLevel >= my_level)
    {
        Assert(pendingNotifies->nestingLevel == my_level);
 
        if (pendingNotifies->upper == NULL ||
            pendingNotifies->upper->nestingLevel < my_level - 1)
        {
            /* nothing to merge; give the whole thing to the parent */
            --pendingNotifies->nestingLevel;
        }
        else
        {
            /*
             * Formerly, we didn't bother to eliminate duplicates here, but
             * now we must, else we fall foul of "Assert(!found)", either here
             * or during a later attempt to build the parent-level hashtable.
             */
            NotificationList *childPendingNotifies = pendingNotifies;
            ListCell   *l;
 
            pendingNotifies = pendingNotifies->upper;
            /* Insert all the subxact's events into parent, except for dups */
            foreach(l, childPendingNotifies->events)
            {
                Notification *childn = (Notification *) lfirst(l);
 
                if (!AsyncExistsPendingNotify(childn))
                    AddEventToPendingNotifies(childn);
            }
            pfree(childPendingNotifies);
        }
    }
}

References ActionList::actions, AddEventToPendingNotifies(), Assert, AsyncExistsPendingNotify(), fb(), GetCurrentTransactionNestLevel(), lfirst, list_concat(), ActionList::nestingLevel, NotificationList::nestingLevel, pendingActions, pendingNotifies, pfree(), ActionList::upper, and NotificationList::upper.

Referenced by CommitSubTransaction().

BecomeRegisteredListener()

static void BecomeRegisteredListener ( void  )

static

Definition at line 1430 of file async.c.

{
    QueuePosition head;
    QueuePosition max;
    ProcNumber  prevListener;
 
    /*
     * Nothing to do if we are already listening to something, nor if we
     * already ran this routine in this transaction.
     */
    if (amRegisteredListener)
        return;
 
    if (Trace_notify)
        elog(DEBUG1, "BecomeRegisteredListener(%d)", MyProcPid);
 
    /*
     * Before registering, make sure we will unlisten before dying. (Note:
     * this action does not get undone if we abort later.)
     */
    if (!unlistenExitRegistered)
    {
        before_shmem_exit(Async_UnlistenOnExit, 0);
        unlistenExitRegistered = true;
    }
 
    /*
     * This is our first LISTEN, so establish our pointer.
     *
     * We set our pointer to the global tail pointer and then move it forward
     * over already-committed notifications.  This ensures we cannot miss any
     * not-yet-committed notifications.  We might get a few more but that
     * doesn't hurt.
     *
     * In some scenarios there might be a lot of committed notifications that
     * have not yet been pruned away (because some backend is being lazy about
     * reading them).  To reduce our startup time, we can look at other
     * backends and adopt the maximum "pos" pointer of any backend that's in
     * our database; any notifications it's already advanced over are surely
     * committed and need not be re-examined by us.  (We must consider only
     * backends connected to our DB, because others will not have bothered to
     * check committed-ness of notifications in our DB.)
     *
     * We need exclusive lock here so we can look at other backends' entries
     * and manipulate the list links.
     */
    LWLockAcquire(NotifyQueueLock, LW_EXCLUSIVE);
    head = QUEUE_HEAD;
    max = QUEUE_TAIL;
    prevListener = INVALID_PROC_NUMBER;
    for (ProcNumber i = QUEUE_FIRST_LISTENER; i != INVALID_PROC_NUMBER; i = QUEUE_NEXT_LISTENER(i))
    {
        if (QUEUE_BACKEND_DBOID(i) == MyDatabaseId)
            max = QUEUE_POS_MAX(max, QUEUE_BACKEND_POS(i));
        /* Also find last listening backend before this one */
        if (i < MyProcNumber)
            prevListener = i;
    }
    QUEUE_BACKEND_POS(MyProcNumber) = max;
    QUEUE_BACKEND_PID(MyProcNumber) = MyProcPid;
    QUEUE_BACKEND_DBOID(MyProcNumber) = MyDatabaseId;
    QUEUE_BACKEND_WAKEUP_PENDING(MyProcNumber) = false;
    QUEUE_BACKEND_IS_ADVANCING(MyProcNumber) = false;
    /* Insert backend into list of listeners at correct position */
    if (prevListener != INVALID_PROC_NUMBER)
    {
        QUEUE_NEXT_LISTENER(MyProcNumber) = QUEUE_NEXT_LISTENER(prevListener);
        QUEUE_NEXT_LISTENER(prevListener) = MyProcNumber;
    }
    else
    {
        QUEUE_NEXT_LISTENER(MyProcNumber) = QUEUE_FIRST_LISTENER;
        QUEUE_FIRST_LISTENER = MyProcNumber;
    }
    LWLockRelease(NotifyQueueLock);
 
    /* Now we are listed in the global array, so remember we're listening */
    amRegisteredListener = true;
 
    /*
     * Try to move our pointer forward as far as possible.  This will skip
     * over already-committed notifications, which we want to do because they
     * might be quite stale.  Note that we are not yet listening on anything,
     * so we won't deliver such notifications to our frontend.  Also, although
     * our transaction might have executed NOTIFY, those message(s) aren't
     * queued yet so we won't skip them here.
     */
    if (!QUEUE_POS_EQUAL(max, head))
        asyncQueueReadAllNotifications();
}

References amRegisteredListener, Async_UnlistenOnExit(), asyncQueueReadAllNotifications(), before_shmem_exit(), DEBUG1, elog, fb(), i, INVALID_PROC_NUMBER, LW_EXCLUSIVE, LWLockAcquire(), LWLockRelease(), MyDatabaseId, MyProcNumber, MyProcPid, QUEUE_BACKEND_DBOID, QUEUE_BACKEND_IS_ADVANCING, QUEUE_BACKEND_PID, QUEUE_BACKEND_POS, QUEUE_BACKEND_WAKEUP_PENDING, QUEUE_FIRST_LISTENER, QUEUE_HEAD, QUEUE_NEXT_LISTENER, QUEUE_POS_EQUAL, QUEUE_POS_MAX, QUEUE_TAIL, Trace_notify, and unlistenExitRegistered.

Referenced by PreCommit_Notify().

check_notify_buffers()

bool check_notify_buffers	(	int *	newval,
		void **	extra,
		GucSource	source
	)

Definition at line 3296 of file async.c.

{
    return check_slru_buffers("notify_buffers", newval);
}

References check_slru_buffers(), and newval.

CleanupListenersOnExit()

static void CleanupListenersOnExit ( void  )

static

Definition at line 1843 of file async.c.

{
    dshash_seq_status status;
    GlobalChannelEntry *entry;
 
    if (Trace_notify)
        elog(DEBUG1, "CleanupListenersOnExit(%d)", MyProcPid);
 
    /* Clear our local cache (not really necessary, but be consistent) */
    if (localChannelTable != NULL)
    {
        hash_destroy(localChannelTable);
        localChannelTable = NULL;
    }
 
    /* Now remove our entries from the shared globalChannelTable */
    if (globalChannelTable == NULL)
        return;
 
    dshash_seq_init(&status, globalChannelTable, true);
    while ((entry = dshash_seq_next(&status)) != NULL)
    {
        ListenerEntry *listeners;
 
        if (entry->key.dboid != MyDatabaseId)
            continue;           /* not relevant */
 
        listeners = (ListenerEntry *)
            dsa_get_address(globalChannelDSA, entry->listenersArray);
 
        for (int i = 0; i < entry->numListeners; i++)
        {
            if (listeners[i].procNo == MyProcNumber)
            {
                entry->numListeners--;
                if (i < entry->numListeners)
                    memmove(&listeners[i], &listeners[i + 1],
                            sizeof(ListenerEntry) * (entry->numListeners - i));
 
                if (entry->numListeners == 0)
                {
                    dsa_free(globalChannelDSA, entry->listenersArray);
                    dshash_delete_current(&status);
                }
                break;
            }
        }
    }
    dshash_seq_term(&status);
}

References GlobalChannelKey::dboid, DEBUG1, dsa_free(), dsa_get_address(), dshash_delete_current(), dshash_seq_init(), dshash_seq_next(), dshash_seq_term(), elog, fb(), globalChannelDSA, globalChannelTable, hash_destroy(), i, GlobalChannelEntry::key, GlobalChannelEntry::listenersArray, localChannelTable, MyDatabaseId, MyProcNumber, MyProcPid, GlobalChannelEntry::numListeners, and Trace_notify.

Referenced by Async_UnlistenOnExit().

ClearPendingActionsAndNotifies()

static void ClearPendingActionsAndNotifies ( void  )

static

Definition at line 3278 of file async.c.

{
    /*
     * Everything's allocated in either TopTransactionContext or the context
     * for the subtransaction to which it corresponds.  So, there's nothing to
     * do here except reset the pointers; the space will be reclaimed when the
     * contexts are deleted.
     */
    pendingActions = NULL;
    pendingNotifies = NULL;
    /* Also clear pendingListenActions, which is derived from pendingActions */
    pendingListenActions = NULL;
}

References fb(), pendingActions, pendingListenActions, and pendingNotifies.

Referenced by AtAbort_Notify(), and AtCommit_Notify().

GlobalChannelKeyInit()

static void GlobalChannelKeyInit ( GlobalChannelKey *  key, Oid  dboid, const char *  channel  )

inlinestatic

Definition at line 660 of file async.c.

{
    memset(key, 0, sizeof(GlobalChannelKey));
    key->dboid = dboid;
    strlcpy(key->channel, channel, NAMEDATALEN);
}

References fb(), NAMEDATALEN, and strlcpy().

Referenced by ApplyPendingListenActions(), PrepareTableEntriesForListen(), and SignalBackends().

globalChannelTableHash()

static dshash_hash globalChannelTableHash ( const void *  key, size_t  size, void *  arg  )

static

Definition at line 672 of file async.c.

{
    const GlobalChannelKey *k = (const GlobalChannelKey *) key;
    dshash_hash h;
 
    h = DatumGetUInt32(hash_uint32(k->dboid));
    h ^= DatumGetUInt32(hash_any((const unsigned char *) k->channel,
                                 strnlen(k->channel, NAMEDATALEN)));
 
    return h;
}

References GlobalChannelKey::channel, DatumGetUInt32(), GlobalChannelKey::dboid, fb(), hash_any(), hash_uint32(), and NAMEDATALEN.

HandleNotifyInterrupt()

void HandleNotifyInterrupt

(

void

)

Definition at line 2552 of file async.c.

{
    /*
     * Note: this is called by a SIGNAL HANDLER. You must be very wary what
     * you do here.
     */
 
    /* signal that work needs to be done */
    notifyInterruptPending = true;
 
    /* latch will be set by procsignal_sigusr1_handler */
}

References notifyInterruptPending.

Referenced by procsignal_sigusr1_handler().

initGlobalChannelTable()

static void initGlobalChannelTable ( void  )

static

Definition at line 699 of file async.c.

{
    MemoryContext oldcontext;
 
    /* Quick exit if we already did this */
    if (asyncQueueControl->globalChannelTableDSH != DSHASH_HANDLE_INVALID &&
        globalChannelTable != NULL)
        return;
 
    /* Otherwise, use a lock to ensure only one process creates the table */
    LWLockAcquire(NotifyQueueLock, LW_EXCLUSIVE);
 
    /* Be sure any local memory allocated by DSA routines is persistent */
    oldcontext = MemoryContextSwitchTo(TopMemoryContext);
 
    if (asyncQueueControl->globalChannelTableDSH == DSHASH_HANDLE_INVALID)
    {
        /* Initialize dynamic shared hash table for global channels */
        globalChannelDSA = dsa_create(LWTRANCHE_NOTIFY_CHANNEL_HASH);
        dsa_pin(globalChannelDSA);
        dsa_pin_mapping(globalChannelDSA);
        globalChannelTable = dshash_create(globalChannelDSA,
                                           &globalChannelTableDSHParams,
                                           NULL);
 
        /* Store handles in shared memory for other backends to use */
        asyncQueueControl->globalChannelTableDSA = dsa_get_handle(globalChannelDSA);
        asyncQueueControl->globalChannelTableDSH =
            dshash_get_hash_table_handle(globalChannelTable);
    }
    else if (!globalChannelTable)
    {
        /* Attach to existing dynamic shared hash table */
        globalChannelDSA = dsa_attach(asyncQueueControl->globalChannelTableDSA);
        dsa_pin_mapping(globalChannelDSA);
        globalChannelTable = dshash_attach(globalChannelDSA,
                                           &globalChannelTableDSHParams,
                                           asyncQueueControl->globalChannelTableDSH,
                                           NULL);
    }
 
    MemoryContextSwitchTo(oldcontext);
    LWLockRelease(NotifyQueueLock);
}

References asyncQueueControl, dsa_attach(), dsa_create, dsa_get_handle(), dsa_pin(), dsa_pin_mapping(), dshash_attach(), dshash_create(), dshash_get_hash_table_handle(), DSHASH_HANDLE_INVALID, fb(), globalChannelDSA, globalChannelTable, AsyncQueueControl::globalChannelTableDSA, AsyncQueueControl::globalChannelTableDSH, globalChannelTableDSHParams, LW_EXCLUSIVE, LWLockAcquire(), LWLockRelease(), MemoryContextSwitchTo(), and TopMemoryContext.

Referenced by PreCommit_Notify().

initLocalChannelTable()

static void initLocalChannelTable ( void  )

static

Definition at line 750 of file async.c.

{
    HASHCTL     hash_ctl;
 
    /* Quick exit if we already did this */
    if (localChannelTable != NULL)
        return;
 
    /* Initialize local hash table for this backend's listened channels */
    hash_ctl.keysize = NAMEDATALEN;
    hash_ctl.entrysize = sizeof(ChannelName);
 
    localChannelTable =
        hash_create("Local Listen Channels",
                    64,
                    &hash_ctl,
                    HASH_ELEM | HASH_STRINGS);
}

References fb(), hash_create(), HASH_ELEM, HASH_STRINGS, HASHCTL::keysize, localChannelTable, and NAMEDATALEN.

Referenced by PreCommit_Notify().

initPendingListenActions()

static void initPendingListenActions ( void  )

static

Definition at line 776 of file async.c.

{
    HASHCTL     hash_ctl;
 
    if (pendingListenActions != NULL)
        return;
 
    hash_ctl.keysize = NAMEDATALEN;
    hash_ctl.entrysize = sizeof(PendingListenEntry);
    hash_ctl.hcxt = CurTransactionContext;
 
    pendingListenActions =
        hash_create("Pending Listen Actions",
                    list_length(pendingActions->actions),
                    &hash_ctl,
                    HASH_ELEM | HASH_STRINGS | HASH_CONTEXT);
}

References ActionList::actions, CurTransactionContext, fb(), HASH_CONTEXT, hash_create(), HASH_ELEM, HASH_STRINGS, HASHCTL::keysize, list_length(), NAMEDATALEN, pendingActions, and pendingListenActions.

Referenced by PreCommit_Notify().

IsListeningOn()

static bool IsListeningOn ( const char *  channel )

static

Definition at line 1900 of file async.c.

{
    if (localChannelTable == NULL)
        return false;
 
    return (hash_search(localChannelTable, channel, HASH_FIND, NULL) != NULL);
}

References fb(), HASH_FIND, hash_search(), and localChannelTable.

Referenced by asyncQueueProcessPageEntries().

notification_hash()

static uint32 notification_hash ( const void *  key, Size  keysize  )

static

Definition at line 3248 of file async.c.

{
    const Notification *k = *(const Notification *const *) key;
 
    Assert(keysize == sizeof(Notification *));
    /* We don't bother to include the payload's trailing null in the hash */
    return DatumGetUInt32(hash_any((const unsigned char *) k->data,
                                   k->channel_len + k->payload_len + 1));
}

References Assert, Notification::channel_len, Notification::data, DatumGetUInt32(), hash_any(), and Notification::payload_len.

Referenced by AddEventToPendingNotifies().

notification_match()

static int notification_match ( const void *  key1, const void *  key2, Size  keysize  )

static

Definition at line 3262 of file async.c.

{
    const Notification *k1 = *(const Notification *const *) key1;
    const Notification *k2 = *(const Notification *const *) key2;
 
    Assert(keysize == sizeof(Notification *));
    if (k1->channel_len == k2->channel_len &&
        k1->payload_len == k2->payload_len &&
        memcmp(k1->data, k2->data,
               k1->channel_len + k1->payload_len + 2) == 0)
        return 0;               /* equal */
    return 1;                   /* not equal */
}

References Assert, and fb().

Referenced by AddEventToPendingNotifies().

NotifyMyFrontEnd()

void NotifyMyFrontEnd	(	const char *	channel,
		const char *	payload,
		int32	srcPid
	)

Definition at line 3099 of file async.c.

{
    if (whereToSendOutput == DestRemote)
    {
        StringInfoData buf;
 
        pq_beginmessage(&buf, PqMsg_NotificationResponse);
        pq_sendint32(&buf, srcPid);
        pq_sendstring(&buf, channel);
        pq_sendstring(&buf, payload);
        pq_endmessage(&buf);
 
        /*
         * NOTE: we do not do pq_flush() here.  Some level of caller will
         * handle it later, allowing this message to be combined into a packet
         * with other ones.
         */
    }
    else
        elog(INFO, "NOTIFY for \"%s\" payload \"%s\"", channel, payload);
}

References buf, DestRemote, elog, INFO, pq_beginmessage(), pq_endmessage(), pq_sendint32(), pq_sendstring(), PqMsg_NotificationResponse, and whereToSendOutput.

Referenced by asyncQueueProcessPageEntries(), and ProcessParallelMessage().

pg_listening_channels()

Datum pg_listening_channels

(

PG_FUNCTION_ARGS

)

Definition at line 1095 of file async.c.

{
    FuncCallContext *funcctx;
    HASH_SEQ_STATUS *status;
 
    /* stuff done only on the first call of the function */
    if (SRF_IS_FIRSTCALL())
    {
        /* create a function context for cross-call persistence */
        funcctx = SRF_FIRSTCALL_INIT();
 
        /* Initialize hash table iteration if we have any channels */
        if (localChannelTable != NULL)
        {
            MemoryContext oldcontext;
 
            oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx);
            status = (HASH_SEQ_STATUS *) palloc(sizeof(HASH_SEQ_STATUS));
            hash_seq_init(status, localChannelTable);
            funcctx->user_fctx = status;
            MemoryContextSwitchTo(oldcontext);
        }
        else
        {
            funcctx->user_fctx = NULL;
        }
    }
 
    /* stuff done on every call of the function */
    funcctx = SRF_PERCALL_SETUP();
    status = (HASH_SEQ_STATUS *) funcctx->user_fctx;
 
    if (status != NULL)
    {
        ChannelName *entry;
 
        entry = (ChannelName *) hash_seq_search(status);
        if (entry != NULL)
            SRF_RETURN_NEXT(funcctx, CStringGetTextDatum(entry->channel));
    }
 
    SRF_RETURN_DONE(funcctx);
}

References ChannelName::channel, CStringGetTextDatum, fb(), hash_seq_init(), hash_seq_search(), localChannelTable, MemoryContextSwitchTo(), palloc(), SRF_FIRSTCALL_INIT, SRF_IS_FIRSTCALL, SRF_PERCALL_SETUP, SRF_RETURN_DONE, and SRF_RETURN_NEXT.

pg_notification_queue_usage()

Datum pg_notification_queue_usage

(

PG_FUNCTION_ARGS

)

Definition at line 2165 of file async.c.

{
    double      usage;
 
    /* Advance the queue tail so we don't report a too-large result */
    asyncQueueAdvanceTail();
 
    LWLockAcquire(NotifyQueueLock, LW_SHARED);
    usage = asyncQueueUsage();
    LWLockRelease(NotifyQueueLock);
 
    PG_RETURN_FLOAT8(usage);
}

References asyncQueueAdvanceTail(), asyncQueueUsage(), fb(), LW_SHARED, LWLockAcquire(), LWLockRelease(), PG_RETURN_FLOAT8, and usage().

pg_notify()

Datum pg_notify

(

PG_FUNCTION_ARGS

)

Definition at line 860 of file async.c.

{
    const char *channel;
    const char *payload;
 
    if (PG_ARGISNULL(0))
        channel = "";
    else
        channel = text_to_cstring(PG_GETARG_TEXT_PP(0));
 
    if (PG_ARGISNULL(1))
        payload = "";
    else
        payload = text_to_cstring(PG_GETARG_TEXT_PP(1));
 
    /* For NOTIFY as a statement, this is checked in ProcessUtility */
    PreventCommandDuringRecovery("NOTIFY");
 
    Async_Notify(channel, payload);
 
    PG_RETURN_VOID();
}

References Async_Notify(), PG_ARGISNULL, PG_GETARG_TEXT_PP, PG_RETURN_VOID, PreventCommandDuringRecovery(), and text_to_cstring().

PreCommit_Notify()

void PreCommit_Notify

(

void

)

Definition at line 1185 of file async.c.

{
    ListCell   *p;
 
    if (!pendingActions && !pendingNotifies)
        return;                 /* no relevant statements in this xact */
 
    if (Trace_notify)
        elog(DEBUG1, "PreCommit_Notify");
 
    /* Preflight for any pending listen/unlisten actions */
    initGlobalChannelTable();
 
    if (pendingActions != NULL)
    {
        /* Ensure we have a local channel table */
        initLocalChannelTable();
        /* Create pendingListenActions hash table for this transaction */
        initPendingListenActions();
 
        /* Stage all the actions this transaction wants to perform */
        foreach(p, pendingActions->actions)
        {
            ListenAction *actrec = (ListenAction *) lfirst(p);
 
            switch (actrec->action)
            {
                case LISTEN_LISTEN:
                    BecomeRegisteredListener();
                    PrepareTableEntriesForListen(actrec->channel);
                    break;
                case LISTEN_UNLISTEN:
                    PrepareTableEntriesForUnlisten(actrec->channel);
                    break;
                case LISTEN_UNLISTEN_ALL:
                    PrepareTableEntriesForUnlistenAll();
                    break;
            }
        }
    }
 
    /* Queue any pending notifies (must happen after the above) */
    if (pendingNotifies)
    {
        ListCell   *nextNotify;
        bool        firstIteration = true;
 
        /*
         * Build list of unique channel names being notified for use by
         * SignalBackends().
         *
         * If uniqueChannelHash is available, use it to efficiently get the
         * unique channels.  Otherwise, fall back to the O(N^2) approach.
         */
        pendingNotifies->uniqueChannelNames = NIL;
        if (pendingNotifies->uniqueChannelHash != NULL)
        {
            HASH_SEQ_STATUS status;
            ChannelName *channelEntry;
 
            hash_seq_init(&status, pendingNotifies->uniqueChannelHash);
            while ((channelEntry = (ChannelName *) hash_seq_search(&status)) != NULL)
                pendingNotifies->uniqueChannelNames =
                    lappend(pendingNotifies->uniqueChannelNames,
                            channelEntry->channel);
        }
        else
        {
            /* O(N^2) approach is better for small number of notifications */
            foreach_ptr(Notification, n, pendingNotifies->events)
            {
                char       *channel = n->data;
                bool        found = false;
 
                /* Name present in list? */
                foreach_ptr(char, oldchan, pendingNotifies->uniqueChannelNames)
                {
                    if (strcmp(oldchan, channel) == 0)
                    {
                        found = true;
                        break;
                    }
                }
                /* Add if not already in list */
                if (!found)
                    pendingNotifies->uniqueChannelNames =
                        lappend(pendingNotifies->uniqueChannelNames,
                                channel);
            }
        }
 
        /* Preallocate workspace that will be needed by SignalBackends() */
        if (signalPids == NULL)
            signalPids = MemoryContextAlloc(TopMemoryContext,
                                            MaxBackends * sizeof(int32));
 
        if (signalProcnos == NULL)
            signalProcnos = MemoryContextAlloc(TopMemoryContext,
                                               MaxBackends * sizeof(ProcNumber));
 
        /*
         * Make sure that we have an XID assigned to the current transaction.
         * GetCurrentTransactionId is cheap if we already have an XID, but not
         * so cheap if we don't, and we'd prefer not to do that work while
         * holding NotifyQueueLock.
         */
        (void) GetCurrentTransactionId();
 
        /*
         * Serialize writers by acquiring a special lock that we hold till
         * after commit.  This ensures that queue entries appear in commit
         * order, and in particular that there are never uncommitted queue
         * entries ahead of committed ones, so an uncommitted transaction
         * can't block delivery of deliverable notifications.
         *
         * We use a heavyweight lock so that it'll automatically be released
         * after either commit or abort.  This also allows deadlocks to be
         * detected, though really a deadlock shouldn't be possible here.
         *
         * The lock is on "database 0", which is pretty ugly but it doesn't
         * seem worth inventing a special locktag category just for this.
         * (Historical note: before PG 9.0, a similar lock on "database 0" was
         * used by the flatfiles mechanism.)
         */
        LockSharedObject(DatabaseRelationId, InvalidOid, 0,
                         AccessExclusiveLock);
 
        /*
         * For the direct advancement optimization in SignalBackends(), we
         * need to ensure that no other backend can insert queue entries
         * between queueHeadBeforeWrite and queueHeadAfterWrite.  The
         * heavyweight lock above provides this guarantee, since it serializes
         * all writers.
         *
         * Note: if the heavyweight lock were ever removed for scalability
         * reasons, we could achieve the same guarantee by holding
         * NotifyQueueLock in EXCLUSIVE mode across all our insertions, rather
         * than releasing and reacquiring it for each page as we do below.
         */
 
        /* Initialize values to a safe default in case list is empty */
        SET_QUEUE_POS(queueHeadBeforeWrite, 0, 0);
        SET_QUEUE_POS(queueHeadAfterWrite, 0, 0);
 
        /* Now push the notifications into the queue */
        nextNotify = list_head(pendingNotifies->events);
        while (nextNotify != NULL)
        {
            /*
             * Add the pending notifications to the queue.  We acquire and
             * release NotifyQueueLock once per page, which might be overkill
             * but it does allow readers to get in while we're doing this.
             *
             * A full queue is very uncommon and should really not happen,
             * given that we have so much space available in the SLRU pages.
             * Nevertheless we need to deal with this possibility. Note that
             * when we get here we are in the process of committing our
             * transaction, but we have not yet committed to clog, so at this
             * point in time we can still roll the transaction back.
             */
            LWLockAcquire(NotifyQueueLock, LW_EXCLUSIVE);
            if (firstIteration)
            {
                queueHeadBeforeWrite = QUEUE_HEAD;
                firstIteration = false;
            }
            asyncQueueFillWarning();
            if (asyncQueueIsFull())
                ereport(ERROR,
                        (errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
                         errmsg("too many notifications in the NOTIFY queue")));
            nextNotify = asyncQueueAddEntries(nextNotify);
            queueHeadAfterWrite = QUEUE_HEAD;
            LWLockRelease(NotifyQueueLock);
        }
 
        /* Note that we don't clear pendingNotifies; AtCommit_Notify will. */
    }
}

References AccessExclusiveLock, ActionList::actions, asyncQueueAddEntries(), asyncQueueFillWarning(), asyncQueueIsFull(), BecomeRegisteredListener(), DEBUG1, elog, ereport, errcode(), errmsg, ERROR, NotificationList::events, fb(), foreach_ptr, GetCurrentTransactionId(), hash_seq_init(), hash_seq_search(), initGlobalChannelTable(), initLocalChannelTable(), initPendingListenActions(), InvalidOid, lappend(), lfirst, list_head(), LISTEN_LISTEN, LISTEN_UNLISTEN, LISTEN_UNLISTEN_ALL, LockSharedObject(), LW_EXCLUSIVE, LWLockAcquire(), LWLockRelease(), MaxBackends, MemoryContextAlloc(), NIL, pendingActions, pendingNotifies, PrepareTableEntriesForListen(), PrepareTableEntriesForUnlisten(), PrepareTableEntriesForUnlistenAll(), QUEUE_HEAD, queueHeadAfterWrite, queueHeadBeforeWrite, SET_QUEUE_POS, signalPids, signalProcnos, TopMemoryContext, Trace_notify, NotificationList::uniqueChannelHash, and NotificationList::uniqueChannelNames.

Referenced by CommitTransaction().

PrepareTableEntriesForListen()

static void PrepareTableEntriesForListen ( const char *  channel )

static

Definition at line 1530 of file async.c.

{
    GlobalChannelKey key;
    GlobalChannelEntry *entry;
    bool        found;
    ListenerEntry *listeners;
    PendingListenEntry *pending;
 
    /*
     * Record in local pending hash that we want to LISTEN, overwriting any
     * earlier attempt to UNLISTEN.
     */
    pending = (PendingListenEntry *)
        hash_search(pendingListenActions, channel, HASH_ENTER, NULL);
    pending->action = PENDING_LISTEN;
 
    /*
     * Ensure that there is an entry for the channel in localChannelTable.
     * (Should this fail, we can just roll back.)  If the transaction fails
     * after this point, we will remove the entry if appropriate during
     * ApplyPendingListenActions.  Note that this entry allows IsListeningOn()
     * to return TRUE; we assume nothing is going to consult that before
     * AtCommit_Notify/AtAbort_Notify.  However, if later actions attempt to
     * UNLISTEN this channel or UNLISTEN *, we need to have the local entry
     * present to ensure they do the right things; see
     * PrepareTableEntriesForUnlisten and PrepareTableEntriesForUnlistenAll.
     */
    (void) hash_search(localChannelTable, channel, HASH_ENTER, NULL);
 
    /* Pre-allocate entry in shared globalChannelTable */
    GlobalChannelKeyInit(&key, MyDatabaseId, channel);
    entry = dshash_find_or_insert(globalChannelTable, &key, &found);
 
    if (!found)
    {
        /* New channel entry, so initialize it to a safe state */
        entry->listenersArray = InvalidDsaPointer;
        entry->numListeners = 0;
        entry->allocatedListeners = 0;
    }
 
    /*
     * Create listenersArray if entry doesn't have one.  It's tempting to fold
     * this into the !found case, but this coding allows us to cope in case
     * dsa_allocate() failed in an earlier attempt.
     */
    if (!DsaPointerIsValid(entry->listenersArray))
    {
        entry->listenersArray = dsa_allocate(globalChannelDSA,
                                             sizeof(ListenerEntry) * INITIAL_LISTENERS_ARRAY_SIZE);
        entry->allocatedListeners = INITIAL_LISTENERS_ARRAY_SIZE;
    }
 
    listeners = (ListenerEntry *)
        dsa_get_address(globalChannelDSA, entry->listenersArray);
 
    /*
     * Check if we already have a ListenerEntry (possibly from earlier in this
     * transaction)
     */
    for (int i = 0; i < entry->numListeners; i++)
    {
        if (listeners[i].procNo == MyProcNumber)
        {
            /* Already have an entry; leave removeOnAbort as-is */
            dshash_release_lock(globalChannelTable, entry);
            return;
        }
    }
 
    /* Need to add a new entry; grow array if necessary */
    if (entry->numListeners >= entry->allocatedListeners)
    {
        int         new_size = entry->allocatedListeners * 2;
        dsa_pointer old_array = entry->listenersArray;
        dsa_pointer new_array = dsa_allocate(globalChannelDSA,
                                             sizeof(ListenerEntry) * new_size);
        ListenerEntry *new_listeners = (ListenerEntry *) dsa_get_address(globalChannelDSA, new_array);
 
        memcpy(new_listeners, listeners, sizeof(ListenerEntry) * entry->numListeners);
        entry->listenersArray = new_array;
        entry->allocatedListeners = new_size;
        dsa_free(globalChannelDSA, old_array);
        listeners = new_listeners;
    }
 
    listeners[entry->numListeners].procNo = MyProcNumber;
    listeners[entry->numListeners].removeOnAbort = true;
    entry->numListeners++;
 
    dshash_release_lock(globalChannelTable, entry);
}

References PendingListenEntry::action, GlobalChannelEntry::allocatedListeners, dsa_allocate, dsa_free(), dsa_get_address(), DsaPointerIsValid, dshash_find_or_insert, dshash_release_lock(), fb(), globalChannelDSA, GlobalChannelKeyInit(), globalChannelTable, HASH_ENTER, hash_search(), i, INITIAL_LISTENERS_ARRAY_SIZE, InvalidDsaPointer, GlobalChannelEntry::listenersArray, localChannelTable, memcpy(), MyDatabaseId, MyProcNumber, GlobalChannelEntry::numListeners, PENDING_LISTEN, and pendingListenActions.

Referenced by PreCommit_Notify().

PrepareTableEntriesForUnlisten()

static void PrepareTableEntriesForUnlisten ( const char *  channel )

static

Definition at line 1632 of file async.c.

{
    PendingListenEntry *pending;
 
    /*
     * If the channel name is not in localChannelTable, then we are neither
     * listening on it nor preparing to listen on it, so we don't need to
     * record an UNLISTEN action.
     */
    Assert(localChannelTable != NULL);
    if (hash_search(localChannelTable, channel, HASH_FIND, NULL) == NULL)
        return;
 
    /*
     * Record in local pending hash that we want to UNLISTEN, overwriting any
     * earlier attempt to LISTEN.  Don't touch localChannelTable or
     * globalChannelTable yet - we keep receiving signals until commit.
     */
    pending = (PendingListenEntry *)
        hash_search(pendingListenActions, channel, HASH_ENTER, NULL);
    pending->action = PENDING_UNLISTEN;
}

References PendingListenEntry::action, Assert, fb(), HASH_ENTER, HASH_FIND, hash_search(), localChannelTable, PENDING_UNLISTEN, and pendingListenActions.

Referenced by PreCommit_Notify().

PrepareTableEntriesForUnlistenAll()

static void PrepareTableEntriesForUnlistenAll ( void  )

static

Definition at line 1662 of file async.c.

{
    HASH_SEQ_STATUS seq;
    ChannelName *channelEntry;
    PendingListenEntry *pending;
 
    /*
     * Scan localChannelTable, which will have the names of all channels that
     * we are listening on or have prepared to listen on.  Record an UNLISTEN
     * action for each one, overwriting any earlier attempt to LISTEN.
     */
    hash_seq_init(&seq, localChannelTable);
    while ((channelEntry = (ChannelName *) hash_seq_search(&seq)) != NULL)
    {
        pending = (PendingListenEntry *)
            hash_search(pendingListenActions, channelEntry->channel, HASH_ENTER, NULL);
        pending->action = PENDING_UNLISTEN;
    }
}

References PendingListenEntry::action, fb(), HASH_ENTER, hash_search(), hash_seq_init(), hash_seq_search(), localChannelTable, PENDING_UNLISTEN, and pendingListenActions.

Referenced by PreCommit_Notify().

ProcessIncomingNotify()

static void ProcessIncomingNotify ( bool  flush )

static

Definition at line 3058 of file async.c.

{
    /* We *must* reset the flag */
    notifyInterruptPending = false;
 
    /* Do nothing else if we aren't actively listening */
    if (LocalChannelTableIsEmpty())
        return;
 
    if (Trace_notify)
        elog(DEBUG1, "ProcessIncomingNotify");
 
    set_ps_display("notify interrupt");
 
    /*
     * We must run asyncQueueReadAllNotifications inside a transaction, else
     * bad things happen if it gets an error.
     */
    StartTransactionCommand();
 
    asyncQueueReadAllNotifications();
 
    CommitTransactionCommand();
 
    /*
     * If this isn't an end-of-command case, we must flush the notify messages
     * to ensure frontend gets them promptly.
     */
    if (flush)
        pq_flush();
 
    set_ps_display("idle");
 
    if (Trace_notify)
        elog(DEBUG1, "ProcessIncomingNotify: done");
}

References asyncQueueReadAllNotifications(), CommitTransactionCommand(), DEBUG1, elog, LocalChannelTableIsEmpty, notifyInterruptPending, pq_flush, set_ps_display(), StartTransactionCommand(), and Trace_notify.

Referenced by ProcessNotifyInterrupt().

ProcessNotifyInterrupt()

void ProcessNotifyInterrupt

(

bool

flush

)

Definition at line 2581 of file async.c.

{
    if (IsTransactionOrTransactionBlock())
        return;                 /* not really idle */
 
    /* Loop in case another signal arrives while sending messages */
    while (notifyInterruptPending)
        ProcessIncomingNotify(flush);
}

References IsTransactionOrTransactionBlock(), notifyInterruptPending, and ProcessIncomingNotify().

Referenced by PostgresMain(), and ProcessClientReadInterrupt().

queue_listen()

static void queue_listen ( ListenActionKind  action, const char *  channel  )

static

Definition at line 996 of file async.c.

{
    MemoryContext oldcontext;
    ListenAction *actrec;
    int         my_level = GetCurrentTransactionNestLevel();
 
    /*
     * Unlike Async_Notify, we don't try to collapse out duplicates here. We
     * keep the ordered list to preserve interactions like UNLISTEN ALL; the
     * final per-channel intent is computed during PreCommit_Notify.
     */
    oldcontext = MemoryContextSwitchTo(CurTransactionContext);
 
    /* space for terminating null is included in sizeof(ListenAction) */
    actrec = (ListenAction *) palloc(offsetof(ListenAction, channel) +
                                     strlen(channel) + 1);
    actrec->action = action;
    strcpy(actrec->channel, channel);
 
    if (pendingActions == NULL || my_level > pendingActions->nestingLevel)
    {
        ActionList *actions;
 
        /*
         * First action in current sub(xact). Note that we allocate the
         * ActionList in TopTransactionContext; the nestingLevel might get
         * changed later by AtSubCommit_Notify.
         */
        actions = (ActionList *)
            MemoryContextAlloc(TopTransactionContext, sizeof(ActionList));
        actions->nestingLevel = my_level;
        actions->actions = list_make1(actrec);
        actions->upper = pendingActions;
        pendingActions = actions;
    }
    else
        pendingActions->actions = lappend(pendingActions->actions, actrec);
 
    MemoryContextSwitchTo(oldcontext);
}

References ActionList::actions, CurTransactionContext, fb(), GetCurrentTransactionNestLevel(), lappend(), list_make1, MemoryContextAlloc(), MemoryContextSwitchTo(), ActionList::nestingLevel, palloc(), pendingActions, TopTransactionContext, and ActionList::upper.

Referenced by Async_Listen(), Async_Unlisten(), and Async_UnlistenAll().

RemoveListenerFromChannel()

static void RemoveListenerFromChannel ( GlobalChannelEntry **  entry_ptr, ListenerEntry *  listeners, int  idx  )

static

Definition at line 1692 of file async.c.

{
    GlobalChannelEntry *entry = *entry_ptr;
 
    entry->numListeners--;
    if (idx < entry->numListeners)
        memmove(&listeners[idx], &listeners[idx + 1],
                sizeof(ListenerEntry) * (entry->numListeners - idx));
 
    if (entry->numListeners == 0)
    {
        dsa_free(globalChannelDSA, entry->listenersArray);
        dshash_delete_entry(globalChannelTable, entry);
        /* tells caller not to release the entry's lock: */
        *entry_ptr = NULL;
    }
}

References dsa_free(), dshash_delete_entry(), fb(), globalChannelDSA, globalChannelTable, idx(), GlobalChannelEntry::listenersArray, and GlobalChannelEntry::numListeners.

Referenced by ApplyPendingListenActions().

SignalBackends()

static void SignalBackends ( void  )

static

Definition at line 2263 of file async.c.

{
    int         count;
 
    /* Can't get here without PreCommit_Notify having made the global table */
    Assert(globalChannelTable != NULL);
 
    /* It should have set up these arrays, too */
    Assert(signalPids != NULL && signalProcnos != NULL);
 
    /*
     * Identify backends that we need to signal.  We don't want to send
     * signals while holding the NotifyQueueLock, so this part just builds a
     * list of target PIDs in signalPids[] and signalProcnos[].
     */
    count = 0;
 
    LWLockAcquire(NotifyQueueLock, LW_EXCLUSIVE);
 
    /* Scan each channel name that we notified in this transaction */
    foreach_ptr(char, channel, pendingNotifies->uniqueChannelNames)
    {
        GlobalChannelKey key;
        GlobalChannelEntry *entry;
        ListenerEntry *listeners;
 
        GlobalChannelKeyInit(&key, MyDatabaseId, channel);
        entry = dshash_find(globalChannelTable, &key, false);
        if (entry == NULL)
            continue;           /* nobody is listening */
 
        listeners = (ListenerEntry *) dsa_get_address(globalChannelDSA,
                                                      entry->listenersArray);
 
        /*
         * Identify listeners that now need waking, add them to arrays.
         *
         * Note that we signal listeners regardless of the state of their
         * removeOnAbort flags.  Hence a new listener that reached PreCommit,
         * but then failed before AtCommit_Notify, can receive a signal even
         * though it was never really listening.  This is okay because it will
         * not do anything in response to that signal.  If we did not do it
         * like this then a new listener might miss some messages due to the
         * direct-advance logic below.
         */
        for (int j = 0; j < entry->numListeners; j++)
        {
            ProcNumber  i = listeners[j].procNo;
            int32       pid;
            QueuePosition pos;
 
            if (QUEUE_BACKEND_WAKEUP_PENDING(i))
                continue;       /* already signaled, no need to repeat */
 
            pid = QUEUE_BACKEND_PID(i);
            pos = QUEUE_BACKEND_POS(i);
 
            if (QUEUE_POS_EQUAL(pos, QUEUE_HEAD))
                continue;       /* it's fully caught up already */
 
            Assert(pid != InvalidPid);
 
            QUEUE_BACKEND_WAKEUP_PENDING(i) = true;
            signalPids[count] = pid;
            signalProcnos[count] = i;
            count++;
        }
 
        dshash_release_lock(globalChannelTable, entry);
    }
 
    /*
     * Scan all listeners.  Any that are not already pending wakeup must not
     * be interested in our notifications (else we'd have set their wakeup
     * flags above).  Check to see if we can directly advance their queue
     * pointers to save a wakeup.  Otherwise, if they are far behind, wake
     * them anyway so they will catch up.
     */
    for (ProcNumber i = QUEUE_FIRST_LISTENER; i != INVALID_PROC_NUMBER; i = QUEUE_NEXT_LISTENER(i))
    {
        int32       pid;
        QueuePosition pos;
 
        if (QUEUE_BACKEND_WAKEUP_PENDING(i))
            continue;
 
        /* If it's currently advancing, we should not touch it */
        if (QUEUE_BACKEND_IS_ADVANCING(i))
            continue;
 
        pid = QUEUE_BACKEND_PID(i);
        pos = QUEUE_BACKEND_POS(i);
 
        /*
         * We can directly advance the other backend's queue pointer if it's
         * not currently advancing (else there are race conditions), and its
         * current pointer is not behind queueHeadBeforeWrite (else we'd make
         * it miss some older messages), and we'd not be moving the pointer
         * backward.
         */
        if (!QUEUE_POS_PRECEDES(pos, queueHeadBeforeWrite) &&
            QUEUE_POS_PRECEDES(pos, queueHeadAfterWrite))
        {
            /* We can directly advance its pointer past what we wrote */
            QUEUE_BACKEND_POS(i) = queueHeadAfterWrite;
        }
        else if (asyncQueuePageDiff(QUEUE_POS_PAGE(QUEUE_HEAD),
                                    QUEUE_POS_PAGE(pos)) >= QUEUE_CLEANUP_DELAY)
        {
            /* It's idle and far behind, so wake it up */
            Assert(pid != InvalidPid);
 
            QUEUE_BACKEND_WAKEUP_PENDING(i) = true;
            signalPids[count] = pid;
            signalProcnos[count] = i;
            count++;
        }
    }
 
    LWLockRelease(NotifyQueueLock);
 
    /* Now send signals */
    for (int i = 0; i < count; i++)
    {
        int32       pid = signalPids[i];
 
        /*
         * If we are signaling our own process, no need to involve the kernel;
         * just set the flag directly.
         */
        if (pid == MyProcPid)
        {
            notifyInterruptPending = true;
            continue;
        }
 
        /*
         * Note: assuming things aren't broken, a signal failure here could
         * only occur if the target backend exited since we released
         * NotifyQueueLock; which is unlikely but certainly possible. So we
         * just log a low-level debug message if it happens.
         */
        if (SendProcSignal(pid, PROCSIG_NOTIFY_INTERRUPT, signalProcnos[i]) < 0)
            elog(DEBUG3, "could not signal backend with PID %d: %m", pid);
    }
}

References Assert, asyncQueuePageDiff(), DEBUG3, dsa_get_address(), dshash_find(), dshash_release_lock(), elog, fb(), foreach_ptr, globalChannelDSA, GlobalChannelKeyInit(), globalChannelTable, i, INVALID_PROC_NUMBER, InvalidPid, j, GlobalChannelEntry::listenersArray, LW_EXCLUSIVE, LWLockAcquire(), LWLockRelease(), MyDatabaseId, MyProcPid, notifyInterruptPending, GlobalChannelEntry::numListeners, pendingNotifies, PROCSIG_NOTIFY_INTERRUPT, QUEUE_BACKEND_IS_ADVANCING, QUEUE_BACKEND_PID, QUEUE_BACKEND_POS, QUEUE_BACKEND_WAKEUP_PENDING, QUEUE_CLEANUP_DELAY, QUEUE_FIRST_LISTENER, QUEUE_HEAD, QUEUE_NEXT_LISTENER, QUEUE_POS_EQUAL, QUEUE_POS_PAGE, QUEUE_POS_PRECEDES, queueHeadAfterWrite, queueHeadBeforeWrite, SendProcSignal(), signalPids, signalProcnos, and NotificationList::uniqueChannelNames.

Referenced by AtCommit_Notify().

Variable Documentation

amRegisteredListener

bool amRegisteredListener = false

static

Definition at line 558 of file async.c.

Referenced by asyncQueueUnregister(), AtAbort_Notify(), AtCommit_Notify(), and BecomeRegisteredListener().

asyncQueueControl

AsyncQueueControl* asyncQueueControl

static

Definition at line 347 of file async.c.

Referenced by asyncQueueFillWarning(), AsyncShmemInit(), AsyncShmemRequest(), and initGlobalChannelTable().

AsyncShmemCallbacks

const ShmemCallbacks AsyncShmemCallbacks

Initial value:

= {
    .request_fn = AsyncShmemRequest,
    .init_fn = AsyncShmemInit,
}

Definition at line 352 of file async.c.

                                           {
    .request_fn = AsyncShmemRequest,
    .init_fn = AsyncShmemInit,
};

globalChannelDSA

dsa_area* globalChannelDSA = NULL

static

Definition at line 415 of file async.c.

Referenced by ApplyPendingListenActions(), CleanupListenersOnExit(), initGlobalChannelTable(), PrepareTableEntriesForListen(), RemoveListenerFromChannel(), and SignalBackends().

globalChannelTable

dshash_table* globalChannelTable = NULL

static

Definition at line 414 of file async.c.

Referenced by ApplyPendingListenActions(), CleanupListenersOnExit(), initGlobalChannelTable(), PrepareTableEntriesForListen(), RemoveListenerFromChannel(), and SignalBackends().

globalChannelTableDSHParams

const dshash_parameters globalChannelTableDSHParams

static

Initial value:

= {
    sizeof(GlobalChannelKey),
    sizeof(GlobalChannelEntry),
    dshash_memcmp,
    globalChannelTableHash,
    dshash_memcpy,
    LWTRANCHE_NOTIFY_CHANNEL_HASH
}

Definition at line 685 of file async.c.

                                                             {
    sizeof(GlobalChannelKey),
    sizeof(GlobalChannelEntry),
    dshash_memcmp,
    globalChannelTableHash,
    dshash_memcpy,
    LWTRANCHE_NOTIFY_CHANNEL_HASH
};

Referenced by initGlobalChannelTable().

localChannelTable

HTAB* localChannelTable = NULL

static

Definition at line 422 of file async.c.

Referenced by ApplyPendingListenActions(), CleanupListenersOnExit(), initLocalChannelTable(), IsListeningOn(), pg_listening_channels(), PrepareTableEntriesForListen(), PrepareTableEntriesForUnlisten(), and PrepareTableEntriesForUnlistenAll().

max_notify_queue_pages

int max_notify_queue_pages = 1048576

Definition at line 584 of file async.c.

Referenced by asyncQueueIsFull(), and asyncQueueUsage().

notifyInterruptPending

volatile sig_atomic_t notifyInterruptPending = false

Definition at line 552 of file async.c.

Referenced by HandleNotifyInterrupt(), PostgresMain(), ProcessClientReadInterrupt(), ProcessIncomingNotify(), ProcessNotifyInterrupt(), and SignalBackends().

NotifySlruDesc

SlruDesc NotifySlruDesc

static

Definition at line 375 of file async.c.

Referenced by AsyncShmemRequest().

pendingActions

ActionList* pendingActions = NULL

static

Definition at line 458 of file async.c.

Referenced by Async_Unlisten(), Async_UnlistenAll(), AtCommit_Notify(), AtPrepare_Notify(), AtSubAbort_Notify(), AtSubCommit_Notify(), ClearPendingActionsAndNotifies(), initPendingListenActions(), PreCommit_Notify(), and queue_listen().

pendingListenActions

HTAB* pendingListenActions = NULL

static

Definition at line 480 of file async.c.

Referenced by ApplyPendingListenActions(), ClearPendingActionsAndNotifies(), initPendingListenActions(), PrepareTableEntriesForListen(), PrepareTableEntriesForUnlisten(), and PrepareTableEntriesForUnlistenAll().

pendingNotifies

NotificationList* pendingNotifies = NULL

static

Definition at line 534 of file async.c.

Referenced by AddEventToPendingNotifies(), Async_Notify(), AsyncExistsPendingNotify(), asyncQueueAddEntries(), AtCommit_Notify(), AtPrepare_Notify(), AtSubAbort_Notify(), AtSubCommit_Notify(), ClearPendingActionsAndNotifies(), PreCommit_Notify(), and SignalBackends().

queueHeadAfterWrite

QueuePosition queueHeadAfterWrite

static

Definition at line 567 of file async.c.

Referenced by PreCommit_Notify(), and SignalBackends().

queueHeadBeforeWrite

QueuePosition queueHeadBeforeWrite

static

Definition at line 566 of file async.c.

Referenced by PreCommit_Notify(), and SignalBackends().

signalPids

int32* signalPids = NULL

static

Definition at line 574 of file async.c.

Referenced by PreCommit_Notify(), and SignalBackends().

signalProcnos

ProcNumber* signalProcnos = NULL

static

Definition at line 575 of file async.c.

Referenced by PreCommit_Notify(), and SignalBackends().

Trace_notify

bool Trace_notify = false

Definition at line 581 of file async.c.

Referenced by Async_Listen(), Async_Notify(), Async_Unlisten(), Async_UnlistenAll(), AtCommit_Notify(), BecomeRegisteredListener(), CleanupListenersOnExit(), PreCommit_Notify(), and ProcessIncomingNotify().

tryAdvanceTail

bool tryAdvanceTail = false

static

Definition at line 578 of file async.c.

Referenced by asyncQueueAddEntries(), and AtCommit_Notify().

unlistenExitRegistered

bool unlistenExitRegistered = false

static

Definition at line 555 of file async.c.

Referenced by Async_Unlisten(), Async_UnlistenAll(), and BecomeRegisteredListener().