Age Owner Branch data TLA Line data Source code
1 : : /*-------------------------------------------------------------------------
2 : : *
3 : : * async.c
4 : : * Asynchronous notification: NOTIFY, LISTEN, UNLISTEN
5 : : *
6 : : * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group
7 : : * Portions Copyright (c) 1994, Regents of the University of California
8 : : *
9 : : * IDENTIFICATION
10 : : * src/backend/commands/async.c
11 : : *
12 : : *-------------------------------------------------------------------------
13 : : */
14 : :
15 : : /*-------------------------------------------------------------------------
16 : : * Async Notification Model as of v19:
17 : : *
18 : : * 1. Multiple backends on same machine. Multiple backends may be listening
19 : : * on each of several channels.
20 : : *
21 : : * 2. There is one central queue in disk-based storage (directory pg_notify/),
22 : : * with actively-used pages mapped into shared memory by the slru.c module.
23 : : * All notification messages are placed in the queue and later read out
24 : : * by listening backends. The single queue allows us to guarantee that
25 : : * notifications are received in commit order.
26 : : *
27 : : * Although there is only one queue, notifications are treated as being
28 : : * database-local; this is done by including the sender's database OID
29 : : * in each notification message. Listening backends ignore messages
30 : : * that don't match their database OID. This is important because it
31 : : * ensures senders and receivers have the same database encoding and won't
32 : : * misinterpret non-ASCII text in the channel name or payload string.
33 : : *
34 : : * Since notifications are not expected to survive database crashes,
35 : : * we can simply clean out the pg_notify data at any reboot, and there
36 : : * is no need for WAL support or fsync'ing.
37 : : *
38 : : * 3. Every backend that is listening on at least one channel registers by
39 : : * entering its PID into the array in AsyncQueueControl. It then scans all
40 : : * incoming notifications in the central queue and first compares the
41 : : * database OID of the notification with its own database OID and then
42 : : * compares the notified channel with the list of channels that it listens
43 : : * to. In case there is a match it delivers the notification event to its
44 : : * frontend. Non-matching events are simply skipped.
45 : : *
46 : : * 4. The NOTIFY statement (routine Async_Notify) stores the notification in
47 : : * a backend-local list which will not be processed until transaction end.
48 : : *
49 : : * Duplicate notifications from the same transaction are sent out as one
50 : : * notification only. This is done to save work when for example a trigger
51 : : * on a 2 million row table fires a notification for each row that has been
52 : : * changed. If the application needs to receive every single notification
53 : : * that has been sent, it can easily add some unique string into the extra
54 : : * payload parameter.
55 : : *
56 : : * When the transaction is ready to commit, PreCommit_Notify() adds the
57 : : * pending notifications to the head of the queue. The head pointer of the
58 : : * queue always points to the next free position and a position is just a
59 : : * page number and the offset in that page. This is done before marking the
60 : : * transaction as committed in clog. If we run into problems writing the
61 : : * notifications, we can still call elog(ERROR, ...) and the transaction
62 : : * will roll back safely.
63 : : *
64 : : * Once we have put all of the notifications into the queue, we return to
65 : : * CommitTransaction() which will then do the actual transaction commit.
66 : : *
67 : : * After commit we are called another time (AtCommit_Notify()). Here we
68 : : * make any required updates to the effective listen state (see below).
69 : : * Then we signal any backends that may be interested in our messages
70 : : * (including our own backend, if listening). This is done by
71 : : * SignalBackends(), which sends a PROCSIG_NOTIFY_INTERRUPT signal to
72 : : * each relevant backend, as described below.
73 : : *
74 : : * Finally, after we are out of the transaction altogether and about to go
75 : : * idle, we scan the queue for messages that need to be sent to our
76 : : * frontend (which might be notifies from other backends, or self-notifies
77 : : * from our own). This step is not part of the CommitTransaction sequence
78 : : * for two important reasons. First, we could get errors while sending
79 : : * data to our frontend, and it's really bad for errors to happen in
80 : : * post-commit cleanup. Second, in cases where a procedure issues commits
81 : : * within a single frontend command, we don't want to send notifies to our
82 : : * frontend until the command is done; but notifies to other backends
83 : : * should go out immediately after each commit.
84 : : *
85 : : * 5. Upon receipt of a PROCSIG_NOTIFY_INTERRUPT signal, the signal handler
86 : : * sets the process's latch, which triggers the event to be processed
87 : : * immediately if this backend is idle (i.e., it is waiting for a frontend
88 : : * command and is not within a transaction block. C.f.
89 : : * ProcessClientReadInterrupt()). Otherwise the handler may only set a
90 : : * flag, which will cause the processing to occur just before we next go
91 : : * idle.
92 : : *
93 : : * Inbound-notify processing consists of reading all of the notifications
94 : : * that have arrived since scanning last time. We read every notification
95 : : * until we reach either a notification from an uncommitted transaction or
96 : : * the head pointer's position.
97 : : *
98 : : * 6. To limit disk space consumption, the tail pointer needs to be advanced
99 : : * so that old pages can be truncated. This is relatively expensive
100 : : * (notably, it requires an exclusive lock), so we don't want to do it
101 : : * often. We make sending backends do this work if they advanced the queue
102 : : * head into a new page, but only once every QUEUE_CLEANUP_DELAY pages.
103 : : *
104 : : * 7. So far we have not discussed how backends change their listening state,
105 : : * nor how notification senders know which backends to awaken. To handle
106 : : * the latter, we maintain a global channel table (implemented as a dynamic
107 : : * shared hash table, or dshash) that maps channel names to the set of
108 : : * backends listening on each channel. This table is created lazily on the
109 : : * first LISTEN command and grows dynamically as needed. There is also a
110 : : * local channel table (a plain dynahash table) in each listening backend,
111 : : * tracking which channels that backend is listening to. The local table
112 : : * serves to reduce the number of accesses needed to the shared table.
113 : : *
114 : : * If the current transaction has executed any LISTEN/UNLISTEN actions,
115 : : * PreCommit_Notify() prepares to commit those. For LISTEN, it
116 : : * pre-allocates entries in both the per-backend localChannelTable and the
117 : : * shared globalChannelTable (with listening=false so that these entries
118 : : * are no-ops for the moment). It also records the final per-channel
119 : : * intent in pendingListenActions, so post-commit/abort processing can
120 : : * apply that in a single step. Since all these allocations happen before
121 : : * committing to clog, we can safely abort the transaction on failure.
122 : : *
123 : : * After commit, AtCommit_Notify() runs through pendingListenActions and
124 : : * updates the backend's per-channel listening flags to activate or
125 : : * deactivate listening. This happens before sending signals.
126 : : *
127 : : * SignalBackends() consults the shared global channel table to identify
128 : : * listeners for the channels that the current transaction sent
129 : : * notification(s) to. Each selected backend is marked as having a wakeup
130 : : * pending to avoid duplicate signals, and a PROCSIG_NOTIFY_INTERRUPT
131 : : * signal is sent to it.
132 : : *
133 : : * 8. While writing notifications, PreCommit_Notify() records the queue head
134 : : * position both before and after the write. Because all writers serialize
135 : : * on a cluster-wide heavyweight lock, no other backend can insert entries
136 : : * between these two points. SignalBackends() uses this fact to directly
137 : : * advance the queue pointer for any backend that is still positioned at
138 : : * the old head, or within the range written, but is not interested in any
139 : : * of our notifications. This avoids unnecessary wakeups for idle
140 : : * listeners that have nothing to read. Backends that are not interested
141 : : * in our notifications, but cannot be directly advanced, are signaled only
142 : : * if they are far behind the current queue head; that is to ensure that
143 : : * we can advance the queue tail without undue delay.
144 : : *
145 : : * An application that listens on the same channel it notifies will get
146 : : * NOTIFY messages for its own NOTIFYs. These can be ignored, if not useful,
147 : : * by comparing be_pid in the NOTIFY message to the application's own backend's
148 : : * PID. (As of FE/BE protocol 2.0, the backend's PID is provided to the
149 : : * frontend during startup.) The above design guarantees that notifies from
150 : : * other backends will never be missed by ignoring self-notifies.
151 : : *
152 : : * The amount of shared memory used for notify management (notify_buffers)
153 : : * can be varied without affecting anything but performance. The maximum
154 : : * amount of notification data that can be queued at one time is determined
155 : : * by the max_notify_queue_pages GUC.
156 : : *-------------------------------------------------------------------------
157 : : */
158 : :
159 : : #include "postgres.h"
160 : :
161 : : #include <limits.h>
162 : : #include <unistd.h>
163 : : #include <signal.h>
164 : :
165 : : #include "access/parallel.h"
166 : : #include "access/slru.h"
167 : : #include "access/transam.h"
168 : : #include "access/xact.h"
169 : : #include "catalog/pg_database.h"
170 : : #include "commands/async.h"
171 : : #include "common/hashfn.h"
172 : : #include "funcapi.h"
173 : : #include "lib/dshash.h"
174 : : #include "libpq/libpq.h"
175 : : #include "libpq/pqformat.h"
176 : : #include "miscadmin.h"
177 : : #include "storage/dsm_registry.h"
178 : : #include "storage/ipc.h"
179 : : #include "storage/latch.h"
180 : : #include "storage/lmgr.h"
181 : : #include "storage/procsignal.h"
182 : : #include "tcop/tcopprot.h"
183 : : #include "utils/builtins.h"
184 : : #include "utils/dsa.h"
185 : : #include "utils/guc_hooks.h"
186 : : #include "utils/memutils.h"
187 : : #include "utils/ps_status.h"
188 : : #include "utils/snapmgr.h"
189 : : #include "utils/timestamp.h"
190 : :
191 : :
192 : : /*
193 : : * Maximum size of a NOTIFY payload, including terminating NULL. This
194 : : * must be kept small enough so that a notification message fits on one
195 : : * SLRU page. The magic fudge factor here is noncritical as long as it's
196 : : * more than AsyncQueueEntryEmptySize --- we make it significantly bigger
197 : : * than that, so changes in that data structure won't affect user-visible
198 : : * restrictions.
199 : : */
200 : : #define NOTIFY_PAYLOAD_MAX_LENGTH (BLCKSZ - NAMEDATALEN - 128)
201 : :
202 : : /*
203 : : * Struct representing an entry in the global notify queue
204 : : *
205 : : * This struct declaration has the maximal length, but in a real queue entry
206 : : * the data area is only big enough for the actual channel and payload strings
207 : : * (each null-terminated). AsyncQueueEntryEmptySize is the minimum possible
208 : : * entry size, if both channel and payload strings are empty (but note it
209 : : * doesn't include alignment padding).
210 : : *
211 : : * The "length" field should always be rounded up to the next QUEUEALIGN
212 : : * multiple so that all fields are properly aligned.
213 : : */
214 : : typedef struct AsyncQueueEntry
215 : : {
216 : : int length; /* total allocated length of entry */
217 : : Oid dboid; /* sender's database OID */
218 : : TransactionId xid; /* sender's XID */
219 : : int32 srcPid; /* sender's PID */
220 : : char data[NAMEDATALEN + NOTIFY_PAYLOAD_MAX_LENGTH];
221 : : } AsyncQueueEntry;
222 : :
223 : : /* Currently, no field of AsyncQueueEntry requires more than int alignment */
224 : : #define QUEUEALIGN(len) INTALIGN(len)
225 : :
226 : : #define AsyncQueueEntryEmptySize (offsetof(AsyncQueueEntry, data) + 2)
227 : :
228 : : /*
229 : : * Struct describing a queue position, and assorted macros for working with it
230 : : */
231 : : typedef struct QueuePosition
232 : : {
233 : : int64 page; /* SLRU page number */
234 : : int offset; /* byte offset within page */
235 : : } QueuePosition;
236 : :
237 : : #define QUEUE_POS_PAGE(x) ((x).page)
238 : : #define QUEUE_POS_OFFSET(x) ((x).offset)
239 : :
240 : : #define SET_QUEUE_POS(x,y,z) \
241 : : do { \
242 : : (x).page = (y); \
243 : : (x).offset = (z); \
244 : : } while (0)
245 : :
246 : : #define QUEUE_POS_EQUAL(x,y) \
247 : : ((x).page == (y).page && (x).offset == (y).offset)
248 : :
249 : : #define QUEUE_POS_IS_ZERO(x) \
250 : : ((x).page == 0 && (x).offset == 0)
251 : :
252 : : /* choose logically smaller QueuePosition */
253 : : #define QUEUE_POS_MIN(x,y) \
254 : : (asyncQueuePagePrecedes((x).page, (y).page) ? (x) : \
255 : : (x).page != (y).page ? (y) : \
256 : : (x).offset < (y).offset ? (x) : (y))
257 : :
258 : : /* choose logically larger QueuePosition */
259 : : #define QUEUE_POS_MAX(x,y) \
260 : : (asyncQueuePagePrecedes((x).page, (y).page) ? (y) : \
261 : : (x).page != (y).page ? (x) : \
262 : : (x).offset > (y).offset ? (x) : (y))
263 : :
264 : : /* returns true if x comes before y in queue order */
265 : : #define QUEUE_POS_PRECEDES(x,y) \
266 : : (asyncQueuePagePrecedes((x).page, (y).page) || \
267 : : ((x).page == (y).page && (x).offset < (y).offset))
268 : :
269 : : /*
270 : : * Parameter determining how often we try to advance the tail pointer:
271 : : * we do that after every QUEUE_CLEANUP_DELAY pages of NOTIFY data. This is
272 : : * also the distance by which a backend that's not interested in our
273 : : * notifications needs to be behind before we'll decide we need to wake it
274 : : * up so it can advance its pointer.
275 : : *
276 : : * Resist the temptation to make this really large. While that would save
277 : : * work in some places, it would add cost in others. In particular, this
278 : : * should likely be less than notify_buffers, to ensure that backends
279 : : * catch up before the pages they'll need to read fall out of SLRU cache.
280 : : */
281 : : #define QUEUE_CLEANUP_DELAY 4
282 : :
283 : : /*
284 : : * Struct describing a listening backend's status
285 : : */
286 : : typedef struct QueueBackendStatus
287 : : {
288 : : int32 pid; /* either a PID or InvalidPid */
289 : : Oid dboid; /* backend's database OID, or InvalidOid */
290 : : ProcNumber nextListener; /* id of next listener, or INVALID_PROC_NUMBER */
291 : : QueuePosition pos; /* backend has read queue up to here */
292 : : bool wakeupPending; /* signal sent to backend, not yet processed */
293 : : bool isAdvancing; /* backend is advancing its position */
294 : : } QueueBackendStatus;
295 : :
296 : : /*
297 : : * Shared memory state for LISTEN/NOTIFY (excluding its SLRU stuff)
298 : : *
299 : : * The AsyncQueueControl structure is protected by the NotifyQueueLock and
300 : : * NotifyQueueTailLock.
301 : : *
302 : : * When holding NotifyQueueLock in SHARED mode, backends may only inspect
303 : : * their own entries as well as the head and tail pointers. Consequently we
304 : : * can allow a backend to update its own record while holding only SHARED lock
305 : : * (since no other backend will inspect it).
306 : : *
307 : : * When holding NotifyQueueLock in EXCLUSIVE mode, backends can inspect the
308 : : * entries of other backends and also change the head pointer. They can
309 : : * also advance other backends' queue positions, unless the other backend
310 : : * has isAdvancing set (i.e., is in process of doing that itself).
311 : : *
312 : : * When holding both NotifyQueueLock and NotifyQueueTailLock in EXCLUSIVE
313 : : * mode, backends can change the tail pointers.
314 : : *
315 : : * SLRU buffer pool is divided in banks and bank wise SLRU lock is used as
316 : : * the control lock for the pg_notify SLRU buffers.
317 : : * In order to avoid deadlocks, whenever we need multiple locks, we first get
318 : : * NotifyQueueTailLock, then NotifyQueueLock, then SLRU bank lock, and lastly
319 : : * globalChannelTable partition locks.
320 : : *
321 : : * Each backend uses the backend[] array entry with index equal to its
322 : : * ProcNumber. We rely on this to make SendProcSignal fast.
323 : : *
324 : : * The backend[] array entries for actively-listening backends are threaded
325 : : * together using firstListener and the nextListener links, so that we can
326 : : * scan them without having to iterate over inactive entries. We keep this
327 : : * list in order by ProcNumber so that the scan is cache-friendly when there
328 : : * are many active entries.
329 : : */
330 : : typedef struct AsyncQueueControl
331 : : {
332 : : QueuePosition head; /* head points to the next free location */
333 : : QueuePosition tail; /* tail must be <= the queue position of every
334 : : * listening backend */
335 : : int64 stopPage; /* oldest unrecycled page; must be <=
336 : : * tail.page */
337 : : ProcNumber firstListener; /* id of first listener, or
338 : : * INVALID_PROC_NUMBER */
339 : : TimestampTz lastQueueFillWarn; /* time of last queue-full msg */
340 : : dsa_handle globalChannelTableDSA; /* global channel table's DSA handle */
341 : : dshash_table_handle globalChannelTableDSH; /* and its dshash handle */
342 : : /* Array with room for MaxBackends entries: */
343 : : QueueBackendStatus backend[FLEXIBLE_ARRAY_MEMBER];
344 : : } AsyncQueueControl;
345 : :
346 : : static AsyncQueueControl *asyncQueueControl;
347 : :
348 : : #define QUEUE_HEAD (asyncQueueControl->head)
349 : : #define QUEUE_TAIL (asyncQueueControl->tail)
350 : : #define QUEUE_STOP_PAGE (asyncQueueControl->stopPage)
351 : : #define QUEUE_FIRST_LISTENER (asyncQueueControl->firstListener)
352 : : #define QUEUE_BACKEND_PID(i) (asyncQueueControl->backend[i].pid)
353 : : #define QUEUE_BACKEND_DBOID(i) (asyncQueueControl->backend[i].dboid)
354 : : #define QUEUE_NEXT_LISTENER(i) (asyncQueueControl->backend[i].nextListener)
355 : : #define QUEUE_BACKEND_POS(i) (asyncQueueControl->backend[i].pos)
356 : : #define QUEUE_BACKEND_WAKEUP_PENDING(i) (asyncQueueControl->backend[i].wakeupPending)
357 : : #define QUEUE_BACKEND_IS_ADVANCING(i) (asyncQueueControl->backend[i].isAdvancing)
358 : :
359 : : /*
360 : : * The SLRU buffer area through which we access the notification queue
361 : : */
362 : : static SlruCtlData NotifyCtlData;
363 : :
364 : : #define NotifyCtl (&NotifyCtlData)
365 : : #define QUEUE_PAGESIZE BLCKSZ
366 : :
367 : : #define QUEUE_FULL_WARN_INTERVAL 5000 /* warn at most once every 5s */
368 : :
369 : : /*
370 : : * Global channel table definitions
371 : : *
372 : : * This hash table maps (database OID, channel name) keys to arrays of
373 : : * ProcNumbers representing the backends listening or about to listen
374 : : * on each channel. The "listening" flags allow us to create hash table
375 : : * entries pre-commit and not have to assume that creating them post-commit
376 : : * will succeed.
377 : : */
378 : : #define INITIAL_LISTENERS_ARRAY_SIZE 4
379 : :
380 : : typedef struct GlobalChannelKey
381 : : {
382 : : Oid dboid;
383 : : char channel[NAMEDATALEN];
384 : : } GlobalChannelKey;
385 : :
386 : : typedef struct ListenerEntry
387 : : {
388 : : ProcNumber procNo; /* listener's ProcNumber */
389 : : bool listening; /* true if committed listener */
390 : : } ListenerEntry;
391 : :
392 : : typedef struct GlobalChannelEntry
393 : : {
394 : : GlobalChannelKey key; /* hash key */
395 : : dsa_pointer listenersArray; /* DSA pointer to ListenerEntry array */
396 : : int numListeners; /* Number of listeners currently stored */
397 : : int allocatedListeners; /* Allocated size of array */
398 : : } GlobalChannelEntry;
399 : :
400 : : static dshash_table *globalChannelTable = NULL;
401 : : static dsa_area *globalChannelDSA = NULL;
402 : :
403 : : /*
404 : : * localChannelTable caches the channel names this backend is listening on
405 : : * (including those we have staged to be listened on, but not yet committed).
406 : : * Used by IsListeningOn() for fast lookups when reading notifications.
407 : : */
408 : : static HTAB *localChannelTable = NULL;
409 : :
410 : : /* We test this condition to detect that we're not listening at all */
411 : : #define LocalChannelTableIsEmpty() \
412 : : (localChannelTable == NULL || hash_get_num_entries(localChannelTable) == 0)
413 : :
414 : : /*
415 : : * State for pending LISTEN/UNLISTEN actions consists of an ordered list of
416 : : * all actions requested in the current transaction. As explained above,
417 : : * we don't actually change listen state until we reach transaction commit.
418 : : *
419 : : * The list is kept in CurTransactionContext. In subtransactions, each
420 : : * subtransaction has its own list in its own CurTransactionContext, but
421 : : * successful subtransactions attach their lists to their parent's list.
422 : : * Failed subtransactions simply discard their lists.
423 : : */
424 : : typedef enum
425 : : {
426 : : LISTEN_LISTEN,
427 : : LISTEN_UNLISTEN,
428 : : LISTEN_UNLISTEN_ALL,
429 : : } ListenActionKind;
430 : :
431 : : typedef struct
432 : : {
433 : : ListenActionKind action;
434 : : char channel[FLEXIBLE_ARRAY_MEMBER]; /* nul-terminated string */
435 : : } ListenAction;
436 : :
437 : : typedef struct ActionList
438 : : {
439 : : int nestingLevel; /* current transaction nesting depth */
440 : : List *actions; /* list of ListenAction structs */
441 : : struct ActionList *upper; /* details for upper transaction levels */
442 : : } ActionList;
443 : :
444 : : static ActionList *pendingActions = NULL;
445 : :
446 : : /*
447 : : * Hash table recording the final listen/unlisten intent per channel for
448 : : * the current transaction. Key is channel name, value is PENDING_LISTEN or
449 : : * PENDING_UNLISTEN. This keeps critical commit/abort processing to one step
450 : : * per channel instead of replaying every action. This is built from the
451 : : * pendingActions list by PreCommit_Notify, then used by AtCommit_Notify or
452 : : * AtAbort_Notify.
453 : : */
454 : : typedef enum
455 : : {
456 : : PENDING_LISTEN,
457 : : PENDING_UNLISTEN,
458 : : } PendingListenAction;
459 : :
460 : : typedef struct PendingListenEntry
461 : : {
462 : : char channel[NAMEDATALEN]; /* hash key */
463 : : PendingListenAction action; /* which action should we perform? */
464 : : } PendingListenEntry;
465 : :
466 : : static HTAB *pendingListenActions = NULL;
467 : :
468 : : /*
469 : : * State for outbound notifies consists of a list of all channels+payloads
470 : : * NOTIFYed in the current transaction. We do not actually perform a NOTIFY
471 : : * until and unless the transaction commits. pendingNotifies is NULL if no
472 : : * NOTIFYs have been done in the current (sub) transaction.
473 : : *
474 : : * We discard duplicate notify events issued in the same transaction.
475 : : * Hence, in addition to the list proper (which we need to track the order
476 : : * of the events, since we guarantee to deliver them in order), we build a
477 : : * hash table which we can probe to detect duplicates. Since building the
478 : : * hash table is somewhat expensive, we do so only once we have at least
479 : : * MIN_HASHABLE_NOTIFIES events queued in the current (sub) transaction;
480 : : * before that we just scan the events linearly.
481 : : *
482 : : * The list is kept in CurTransactionContext. In subtransactions, each
483 : : * subtransaction has its own list in its own CurTransactionContext, but
484 : : * successful subtransactions add their entries to their parent's list.
485 : : * Failed subtransactions simply discard their lists. Since these lists
486 : : * are independent, there may be notify events in a subtransaction's list
487 : : * that duplicate events in some ancestor (sub) transaction; we get rid of
488 : : * the dups when merging the subtransaction's list into its parent's.
489 : : *
490 : : * Note: the action and notify lists do not interact within a transaction.
491 : : * In particular, if a transaction does NOTIFY and then LISTEN on the same
492 : : * condition name, it will get a self-notify at commit. This is a bit odd
493 : : * but is consistent with our historical behavior.
494 : : */
495 : : typedef struct Notification
496 : : {
497 : : uint16 channel_len; /* length of channel-name string */
498 : : uint16 payload_len; /* length of payload string */
499 : : /* null-terminated channel name, then null-terminated payload follow */
500 : : char data[FLEXIBLE_ARRAY_MEMBER];
501 : : } Notification;
502 : :
503 : : typedef struct NotificationList
504 : : {
505 : : int nestingLevel; /* current transaction nesting depth */
506 : : List *events; /* list of Notification structs */
507 : : HTAB *hashtab; /* hash of NotificationHash structs, or NULL */
508 : : List *uniqueChannelNames; /* unique channel names being notified */
509 : : HTAB *uniqueChannelHash; /* hash of unique channel names, or NULL */
510 : : struct NotificationList *upper; /* details for upper transaction levels */
511 : : } NotificationList;
512 : :
513 : : #define MIN_HASHABLE_NOTIFIES 16 /* threshold to build hashtab */
514 : :
515 : : struct NotificationHash
516 : : {
517 : : Notification *event; /* => the actual Notification struct */
518 : : };
519 : :
520 : : static NotificationList *pendingNotifies = NULL;
521 : :
522 : : /*
523 : : * Hash entry in NotificationList.uniqueChannelHash or localChannelTable
524 : : * (both just carry the channel name, with no payload).
525 : : */
526 : : typedef struct ChannelName
527 : : {
528 : : char channel[NAMEDATALEN]; /* hash key */
529 : : } ChannelName;
530 : :
531 : : /*
532 : : * Inbound notifications are initially processed by HandleNotifyInterrupt(),
533 : : * called from inside a signal handler. That just sets the
534 : : * notifyInterruptPending flag and sets the process
535 : : * latch. ProcessNotifyInterrupt() will then be called whenever it's safe to
536 : : * actually deal with the interrupt.
537 : : */
538 : : volatile sig_atomic_t notifyInterruptPending = false;
539 : :
540 : : /* True if we've registered an on_shmem_exit cleanup */
541 : : static bool unlistenExitRegistered = false;
542 : :
543 : : /* True if we're currently registered as a listener in asyncQueueControl */
544 : : static bool amRegisteredListener = false;
545 : :
546 : : /*
547 : : * Queue head positions for direct advancement.
548 : : * These are captured during PreCommit_Notify while holding the heavyweight
549 : : * lock on database 0, ensuring no other backend can insert notifications
550 : : * between them. SignalBackends uses these to advance idle backends.
551 : : */
552 : : static QueuePosition queueHeadBeforeWrite;
553 : : static QueuePosition queueHeadAfterWrite;
554 : :
555 : : /*
556 : : * Workspace arrays for SignalBackends. These are preallocated in
557 : : * PreCommit_Notify to avoid needing memory allocation after committing to
558 : : * clog.
559 : : */
560 : : static int32 *signalPids = NULL;
561 : : static ProcNumber *signalProcnos = NULL;
562 : :
563 : : /* have we advanced to a page that's a multiple of QUEUE_CLEANUP_DELAY? */
564 : : static bool tryAdvanceTail = false;
565 : :
566 : : /* GUC parameters */
567 : : bool Trace_notify = false;
568 : :
569 : : /* For 8 KB pages this gives 8 GB of disk space */
570 : : int max_notify_queue_pages = 1048576;
571 : :
572 : : /* local function prototypes */
573 : : static int asyncQueueErrdetailForIoError(const void *opaque_data);
574 : : static inline int64 asyncQueuePageDiff(int64 p, int64 q);
575 : : static inline bool asyncQueuePagePrecedes(int64 p, int64 q);
576 : : static inline void GlobalChannelKeyInit(GlobalChannelKey *key, Oid dboid,
577 : : const char *channel);
578 : : static dshash_hash globalChannelTableHash(const void *key, size_t size,
579 : : void *arg);
580 : : static void initGlobalChannelTable(void);
581 : : static void initLocalChannelTable(void);
582 : : static void queue_listen(ListenActionKind action, const char *channel);
583 : : static void Async_UnlistenOnExit(int code, Datum arg);
584 : : static void BecomeRegisteredListener(void);
585 : : static void PrepareTableEntriesForListen(const char *channel);
586 : : static void PrepareTableEntriesForUnlisten(const char *channel);
587 : : static void PrepareTableEntriesForUnlistenAll(void);
588 : : static void RemoveListenerFromChannel(GlobalChannelEntry **entry_ptr,
589 : : ListenerEntry *listeners,
590 : : int idx);
591 : : static void ApplyPendingListenActions(bool isCommit);
592 : : static void CleanupListenersOnExit(void);
593 : : static bool IsListeningOn(const char *channel);
594 : : static void asyncQueueUnregister(void);
595 : : static bool asyncQueueIsFull(void);
596 : : static bool asyncQueueAdvance(volatile QueuePosition *position, int entryLength);
597 : : static void asyncQueueNotificationToEntry(Notification *n, AsyncQueueEntry *qe);
598 : : static ListCell *asyncQueueAddEntries(ListCell *nextNotify);
599 : : static double asyncQueueUsage(void);
600 : : static void asyncQueueFillWarning(void);
601 : : static void SignalBackends(void);
602 : : static void asyncQueueReadAllNotifications(void);
603 : : static bool asyncQueueProcessPageEntries(QueuePosition *current,
604 : : QueuePosition stop,
605 : : Snapshot snapshot);
606 : : static void asyncQueueAdvanceTail(void);
607 : : static void ProcessIncomingNotify(bool flush);
608 : : static bool AsyncExistsPendingNotify(Notification *n);
609 : : static void AddEventToPendingNotifies(Notification *n);
610 : : static uint32 notification_hash(const void *key, Size keysize);
611 : : static int notification_match(const void *key1, const void *key2, Size keysize);
612 : : static void ClearPendingActionsAndNotifies(void);
613 : :
614 : : static int
2 heikki.linnakangas@i 615 :UNC 0 : asyncQueueErrdetailForIoError(const void *opaque_data)
616 : : {
617 : 0 : const QueuePosition *position = opaque_data;
618 : :
619 : 0 : return errdetail("Could not access async queue at page %" PRId64 ", offset %d.",
620 : 0 : position->page, position->offset);
621 : : }
622 : :
623 : : /*
624 : : * Compute the difference between two queue page numbers.
625 : : * Previously this function accounted for a wraparound.
626 : : */
627 : : static inline int64
837 akorotkov@postgresql 628 :UBC 0 : asyncQueuePageDiff(int64 p, int64 q)
629 : : {
630 : 0 : return p - q;
631 : : }
632 : :
633 : : /*
634 : : * Determines whether p precedes q.
635 : : * Previously this function accounted for a wraparound.
636 : : */
637 : : static inline bool
837 akorotkov@postgresql 638 :CBC 144 : asyncQueuePagePrecedes(int64 p, int64 q)
639 : : {
640 : 144 : return p < q;
641 : : }
642 : :
643 : : /*
644 : : * GlobalChannelKeyInit
645 : : * Prepare a global channel table key for hashing.
646 : : */
647 : : static inline void
59 tgl@sss.pgh.pa.us 648 :GNC 206 : GlobalChannelKeyInit(GlobalChannelKey *key, Oid dboid, const char *channel)
649 : : {
650 : 206 : memset(key, 0, sizeof(GlobalChannelKey));
651 : 206 : key->dboid = dboid;
652 : 206 : strlcpy(key->channel, channel, NAMEDATALEN);
653 : 206 : }
654 : :
655 : : /*
656 : : * globalChannelTableHash
657 : : * Hash function for global channel table keys.
658 : : */
659 : : static dshash_hash
660 : 206 : globalChannelTableHash(const void *key, size_t size, void *arg)
661 : : {
662 : 206 : const GlobalChannelKey *k = (const GlobalChannelKey *) key;
663 : : dshash_hash h;
664 : :
665 : 206 : h = DatumGetUInt32(hash_uint32(k->dboid));
666 : 206 : h ^= DatumGetUInt32(hash_any((const unsigned char *) k->channel,
667 : 206 : strnlen(k->channel, NAMEDATALEN)));
668 : :
669 : 206 : return h;
670 : : }
671 : :
672 : : /* parameters for the global channel table */
673 : : static const dshash_parameters globalChannelTableDSHParams = {
674 : : sizeof(GlobalChannelKey),
675 : : sizeof(GlobalChannelEntry),
676 : : dshash_memcmp,
677 : : globalChannelTableHash,
678 : : dshash_memcpy,
679 : : LWTRANCHE_NOTIFY_CHANNEL_HASH
680 : : };
681 : :
682 : : /*
683 : : * initGlobalChannelTable
684 : : * Lazy initialization of the global channel table.
685 : : */
686 : : static void
687 : 146 : initGlobalChannelTable(void)
688 : : {
689 : : MemoryContext oldcontext;
690 : :
691 : : /* Quick exit if we already did this */
692 [ + + ]: 146 : if (asyncQueueControl->globalChannelTableDSH != DSHASH_HANDLE_INVALID &&
693 [ + + ]: 139 : globalChannelTable != NULL)
694 : 127 : return;
695 : :
696 : : /* Otherwise, use a lock to ensure only one process creates the table */
697 : 19 : LWLockAcquire(NotifyQueueLock, LW_EXCLUSIVE);
698 : :
699 : : /* Be sure any local memory allocated by DSA routines is persistent */
700 : 19 : oldcontext = MemoryContextSwitchTo(TopMemoryContext);
701 : :
702 [ + + ]: 19 : if (asyncQueueControl->globalChannelTableDSH == DSHASH_HANDLE_INVALID)
703 : : {
704 : : /* Initialize dynamic shared hash table for global channels */
705 : 7 : globalChannelDSA = dsa_create(LWTRANCHE_NOTIFY_CHANNEL_HASH);
706 : 7 : dsa_pin(globalChannelDSA);
707 : 7 : dsa_pin_mapping(globalChannelDSA);
708 : 7 : globalChannelTable = dshash_create(globalChannelDSA,
709 : : &globalChannelTableDSHParams,
710 : : NULL);
711 : :
712 : : /* Store handles in shared memory for other backends to use */
713 : 7 : asyncQueueControl->globalChannelTableDSA = dsa_get_handle(globalChannelDSA);
714 : 7 : asyncQueueControl->globalChannelTableDSH =
715 : 7 : dshash_get_hash_table_handle(globalChannelTable);
716 : : }
717 [ + - ]: 12 : else if (!globalChannelTable)
718 : : {
719 : : /* Attach to existing dynamic shared hash table */
720 : 12 : globalChannelDSA = dsa_attach(asyncQueueControl->globalChannelTableDSA);
721 : 12 : dsa_pin_mapping(globalChannelDSA);
722 : 12 : globalChannelTable = dshash_attach(globalChannelDSA,
723 : : &globalChannelTableDSHParams,
724 : 12 : asyncQueueControl->globalChannelTableDSH,
725 : : NULL);
726 : : }
727 : :
728 : 19 : MemoryContextSwitchTo(oldcontext);
729 : 19 : LWLockRelease(NotifyQueueLock);
730 : : }
731 : :
732 : : /*
733 : : * initLocalChannelTable
734 : : * Lazy initialization of the local channel table.
735 : : * Once created, this table lasts for the life of the session.
736 : : */
737 : : static void
738 : 91 : initLocalChannelTable(void)
739 : : {
740 : : HASHCTL hash_ctl;
741 : :
742 : : /* Quick exit if we already did this */
743 [ + + ]: 91 : if (localChannelTable != NULL)
744 : 74 : return;
745 : :
746 : : /* Initialize local hash table for this backend's listened channels */
747 : 17 : hash_ctl.keysize = NAMEDATALEN;
748 : 17 : hash_ctl.entrysize = sizeof(ChannelName);
749 : :
750 : 17 : localChannelTable =
751 : 17 : hash_create("Local Listen Channels",
752 : : 64,
753 : : &hash_ctl,
754 : : HASH_ELEM | HASH_STRINGS);
755 : : }
756 : :
757 : : /*
758 : : * initPendingListenActions
759 : : * Lazy initialization of the pending listen actions hash table.
760 : : * This is allocated in CurTransactionContext during PreCommit_Notify,
761 : : * and destroyed at transaction end.
762 : : */
763 : : static void
764 : 91 : initPendingListenActions(void)
765 : : {
766 : : HASHCTL hash_ctl;
767 : :
768 [ - + ]: 91 : if (pendingListenActions != NULL)
59 tgl@sss.pgh.pa.us 769 :UNC 0 : return;
770 : :
59 tgl@sss.pgh.pa.us 771 :GNC 91 : hash_ctl.keysize = NAMEDATALEN;
772 : 91 : hash_ctl.entrysize = sizeof(PendingListenEntry);
773 : 91 : hash_ctl.hcxt = CurTransactionContext;
774 : :
775 : 91 : pendingListenActions =
776 : 91 : hash_create("Pending Listen Actions",
777 : 91 : list_length(pendingActions->actions),
778 : : &hash_ctl,
779 : : HASH_ELEM | HASH_STRINGS | HASH_CONTEXT);
780 : : }
781 : :
782 : : /*
783 : : * Report space needed for our shared memory area
784 : : */
785 : : Size
5871 tgl@sss.pgh.pa.us 786 :CBC 2147 : AsyncShmemSize(void)
787 : : {
788 : : Size size;
789 : :
790 : : /* This had better match AsyncShmemInit */
742 heikki.linnakangas@i 791 : 2147 : size = mul_size(MaxBackends, sizeof(QueueBackendStatus));
4040 tgl@sss.pgh.pa.us 792 : 2147 : size = add_size(size, offsetof(AsyncQueueControl, backend));
793 : :
746 alvherre@alvh.no-ip. 794 : 2147 : size = add_size(size, SimpleLruShmemSize(notify_buffers, 0));
795 : :
5871 tgl@sss.pgh.pa.us 796 : 2147 : return size;
797 : : }
798 : :
799 : : /*
800 : : * Initialize our shared memory area
801 : : */
802 : : void
803 : 1150 : AsyncShmemInit(void)
804 : : {
805 : : bool found;
806 : : Size size;
807 : :
808 : : /*
809 : : * Create or attach to the AsyncQueueControl structure.
810 : : */
742 heikki.linnakangas@i 811 : 1150 : size = mul_size(MaxBackends, sizeof(QueueBackendStatus));
4040 tgl@sss.pgh.pa.us 812 : 1150 : size = add_size(size, offsetof(AsyncQueueControl, backend));
813 : :
5871 814 : 1150 : asyncQueueControl = (AsyncQueueControl *)
815 : 1150 : ShmemInitStruct("Async Queue Control", size, &found);
816 : :
817 [ + - ]: 1150 : if (!found)
818 : : {
819 : : /* First time through, so initialize it */
820 : 1150 : SET_QUEUE_POS(QUEUE_HEAD, 0, 0);
821 : 1150 : SET_QUEUE_POS(QUEUE_TAIL, 0, 0);
1933 822 : 1150 : QUEUE_STOP_PAGE = 0;
742 heikki.linnakangas@i 823 : 1150 : QUEUE_FIRST_LISTENER = INVALID_PROC_NUMBER;
5871 tgl@sss.pgh.pa.us 824 : 1150 : asyncQueueControl->lastQueueFillWarn = 0;
59 tgl@sss.pgh.pa.us 825 :GNC 1150 : asyncQueueControl->globalChannelTableDSA = DSA_HANDLE_INVALID;
826 : 1150 : asyncQueueControl->globalChannelTableDSH = DSHASH_HANDLE_INVALID;
742 heikki.linnakangas@i 827 [ + + ]:CBC 106834 : for (int i = 0; i < MaxBackends; i++)
828 : : {
5871 tgl@sss.pgh.pa.us 829 : 105684 : QUEUE_BACKEND_PID(i) = InvalidPid;
3819 830 : 105684 : QUEUE_BACKEND_DBOID(i) = InvalidOid;
742 heikki.linnakangas@i 831 : 105684 : QUEUE_NEXT_LISTENER(i) = INVALID_PROC_NUMBER;
5871 tgl@sss.pgh.pa.us 832 : 105684 : SET_QUEUE_POS(QUEUE_BACKEND_POS(i), 0, 0);
59 tgl@sss.pgh.pa.us 833 :GNC 105684 : QUEUE_BACKEND_WAKEUP_PENDING(i) = false;
834 : 105684 : QUEUE_BACKEND_IS_ADVANCING(i) = false;
835 : : }
836 : : }
837 : :
838 : : /*
839 : : * Set up SLRU management of the pg_notify data. Note that long segment
840 : : * names are used in order to avoid wraparound.
841 : : */
2130 tgl@sss.pgh.pa.us 842 :CBC 1150 : NotifyCtl->PagePrecedes = asyncQueuePagePrecedes;
2 heikki.linnakangas@i 843 :GNC 1150 : NotifyCtl->errdetail_for_io_error = asyncQueueErrdetailForIoError;
746 alvherre@alvh.no-ip. 844 :CBC 1150 : SimpleLruInit(NotifyCtl, "notify", notify_buffers, 0,
845 : : "pg_notify", LWTRANCHE_NOTIFY_BUFFER, LWTRANCHE_NOTIFY_SLRU,
846 : : SYNC_HANDLER_NONE, true);
847 : :
5871 tgl@sss.pgh.pa.us 848 [ + - ]: 1150 : if (!found)
849 : : {
850 : : /*
851 : : * During start or reboot, clean out the pg_notify directory.
852 : : */
2130 853 : 1150 : (void) SlruScanDirectory(NotifyCtl, SlruScanDirCbDeleteAll, NULL);
854 : : }
5871 855 : 1150 : }
856 : :
857 : :
858 : : /*
859 : : * pg_notify -
860 : : * SQL function to send a notification event
861 : : */
862 : : Datum
863 : 1072 : pg_notify(PG_FUNCTION_ARGS)
864 : : {
865 : : const char *channel;
866 : : const char *payload;
867 : :
868 [ + + ]: 1072 : if (PG_ARGISNULL(0))
869 : 3 : channel = "";
870 : : else
871 : 1069 : channel = text_to_cstring(PG_GETARG_TEXT_PP(0));
872 : :
873 [ + + ]: 1072 : if (PG_ARGISNULL(1))
874 : 6 : payload = "";
875 : : else
876 : 1066 : payload = text_to_cstring(PG_GETARG_TEXT_PP(1));
877 : :
878 : : /* For NOTIFY as a statement, this is checked in ProcessUtility */
5867 879 : 1072 : PreventCommandDuringRecovery("NOTIFY");
880 : :
5871 881 : 1072 : Async_Notify(channel, payload);
882 : :
883 : 1063 : PG_RETURN_VOID();
884 : : }
885 : :
886 : :
887 : : /*
888 : : * Async_Notify
889 : : *
890 : : * This is executed by the SQL notify command.
891 : : *
892 : : * Adds the message to the list of pending notifies.
893 : : * Actual notification happens during transaction commit.
894 : : * ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
895 : : */
896 : : void
897 : 1128 : Async_Notify(const char *channel, const char *payload)
898 : : {
2354 rhaas@postgresql.org 899 : 1128 : int my_level = GetCurrentTransactionNestLevel();
900 : : size_t channel_len;
901 : : size_t payload_len;
902 : : Notification *n;
903 : : MemoryContext oldcontext;
904 : :
3797 905 [ - + ]: 1128 : if (IsParallelWorker())
3797 rhaas@postgresql.org 906 [ # # ]:UBC 0 : elog(ERROR, "cannot send notifications from a parallel worker");
907 : :
9419 peter_e@gmx.net 908 [ - + ]:CBC 1128 : if (Trace_notify)
5871 tgl@sss.pgh.pa.us 909 [ # # ]:UBC 0 : elog(DEBUG1, "Async_Notify(%s)", channel);
910 : :
2404 tgl@sss.pgh.pa.us 911 [ + - ]:CBC 1128 : channel_len = channel ? strlen(channel) : 0;
912 [ + + ]: 1128 : payload_len = payload ? strlen(payload) : 0;
913 : :
914 : : /* a channel name must be specified */
915 [ + + ]: 1128 : if (channel_len == 0)
5871 916 [ + - ]: 6 : ereport(ERROR,
917 : : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
918 : : errmsg("channel name cannot be empty")));
919 : :
920 : : /* enforce length limits */
2404 921 [ + + ]: 1122 : if (channel_len >= NAMEDATALEN)
5871 922 [ + - ]: 3 : ereport(ERROR,
923 : : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
924 : : errmsg("channel name too long")));
925 : :
2404 926 [ - + ]: 1119 : if (payload_len >= NOTIFY_PAYLOAD_MAX_LENGTH)
2404 tgl@sss.pgh.pa.us 927 [ # # ]:UBC 0 : ereport(ERROR,
928 : : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
929 : : errmsg("payload string too long")));
930 : :
931 : : /*
932 : : * We must construct the Notification entry, even if we end up not using
933 : : * it, in order to compare it cheaply to existing list entries.
934 : : *
935 : : * The notification list needs to live until end of transaction, so store
936 : : * it in the transaction context.
937 : : */
5871 tgl@sss.pgh.pa.us 938 :CBC 1119 : oldcontext = MemoryContextSwitchTo(CurTransactionContext);
939 : :
2404 940 : 1119 : n = (Notification *) palloc(offsetof(Notification, data) +
941 : 1119 : channel_len + payload_len + 2);
942 : 1119 : n->channel_len = channel_len;
943 : 1119 : n->payload_len = payload_len;
944 : 1119 : strcpy(n->data, channel);
5871 945 [ + + ]: 1119 : if (payload)
2404 946 : 1105 : strcpy(n->data + channel_len + 1, payload);
947 : : else
948 : 14 : n->data[channel_len + 1] = '\0';
949 : :
2354 rhaas@postgresql.org 950 [ + + + + ]: 1119 : if (pendingNotifies == NULL || my_level > pendingNotifies->nestingLevel)
2404 tgl@sss.pgh.pa.us 951 : 57 : {
952 : : NotificationList *notifies;
953 : :
954 : : /*
955 : : * First notify event in current (sub)xact. Note that we allocate the
956 : : * NotificationList in TopTransactionContext; the nestingLevel might
957 : : * get changed later by AtSubCommit_Notify.
958 : : */
959 : : notifies = (NotificationList *)
2354 rhaas@postgresql.org 960 : 57 : MemoryContextAlloc(TopTransactionContext,
961 : : sizeof(NotificationList));
962 : 57 : notifies->nestingLevel = my_level;
963 : 57 : notifies->events = list_make1(n);
964 : : /* We certainly don't need a hashtable yet */
965 : 57 : notifies->hashtab = NULL;
966 : : /* We won't build uniqueChannelNames/Hash till later, either */
59 tgl@sss.pgh.pa.us 967 :GNC 57 : notifies->uniqueChannelNames = NIL;
968 : 57 : notifies->uniqueChannelHash = NULL;
2354 rhaas@postgresql.org 969 :CBC 57 : notifies->upper = pendingNotifies;
970 : 57 : pendingNotifies = notifies;
971 : : }
972 : : else
973 : : {
974 : : /* Now check for duplicates */
975 [ + + ]: 1062 : if (AsyncExistsPendingNotify(n))
976 : : {
977 : : /* It's a dup, so forget it */
978 : 13 : pfree(n);
979 : 13 : MemoryContextSwitchTo(oldcontext);
980 : 13 : return;
981 : : }
982 : :
983 : : /* Append more events to existing list */
2404 tgl@sss.pgh.pa.us 984 : 1049 : AddEventToPendingNotifies(n);
985 : : }
986 : :
5871 987 : 1106 : MemoryContextSwitchTo(oldcontext);
988 : : }
989 : :
990 : : /*
991 : : * queue_listen
992 : : * Common code for listen, unlisten, unlisten all commands.
993 : : *
994 : : * Adds the request to the list of pending actions.
995 : : * Actual update of localChannelTable and globalChannelTable happens during
996 : : * PreCommit_Notify, with staged changes committed in AtCommit_Notify.
997 : : */
998 : : static void
999 : 110 : queue_listen(ListenActionKind action, const char *channel)
1000 : : {
1001 : : MemoryContext oldcontext;
1002 : : ListenAction *actrec;
2354 rhaas@postgresql.org 1003 : 110 : int my_level = GetCurrentTransactionNestLevel();
1004 : :
1005 : : /*
1006 : : * Unlike Async_Notify, we don't try to collapse out duplicates here. We
1007 : : * keep the ordered list to preserve interactions like UNLISTEN ALL; the
1008 : : * final per-channel intent is computed during PreCommit_Notify.
1009 : : */
6577 tgl@sss.pgh.pa.us 1010 : 110 : oldcontext = MemoryContextSwitchTo(CurTransactionContext);
1011 : :
1012 : : /* space for terminating null is included in sizeof(ListenAction) */
4040 1013 : 110 : actrec = (ListenAction *) palloc(offsetof(ListenAction, channel) +
1014 : 110 : strlen(channel) + 1);
6577 1015 : 110 : actrec->action = action;
5871 1016 : 110 : strcpy(actrec->channel, channel);
1017 : :
2354 rhaas@postgresql.org 1018 [ + + + + ]: 110 : if (pendingActions == NULL || my_level > pendingActions->nestingLevel)
1019 : 94 : {
1020 : : ActionList *actions;
1021 : :
1022 : : /*
1023 : : * First action in current sub(xact). Note that we allocate the
1024 : : * ActionList in TopTransactionContext; the nestingLevel might get
1025 : : * changed later by AtSubCommit_Notify.
1026 : : */
1027 : : actions = (ActionList *)
1028 : 94 : MemoryContextAlloc(TopTransactionContext, sizeof(ActionList));
1029 : 94 : actions->nestingLevel = my_level;
1030 : 94 : actions->actions = list_make1(actrec);
1031 : 94 : actions->upper = pendingActions;
1032 : 94 : pendingActions = actions;
1033 : : }
1034 : : else
1035 : 16 : pendingActions->actions = lappend(pendingActions->actions, actrec);
1036 : :
6577 tgl@sss.pgh.pa.us 1037 : 110 : MemoryContextSwitchTo(oldcontext);
1038 : 110 : }
1039 : :
1040 : : /*
1041 : : * Async_Listen
1042 : : *
1043 : : * This is executed by the SQL listen command.
1044 : : */
1045 : : void
5871 1046 : 58 : Async_Listen(const char *channel)
1047 : : {
6577 1048 [ - + ]: 58 : if (Trace_notify)
5871 tgl@sss.pgh.pa.us 1049 [ # # ]:UBC 0 : elog(DEBUG1, "Async_Listen(%s,%d)", channel, MyProcPid);
1050 : :
5871 tgl@sss.pgh.pa.us 1051 :CBC 58 : queue_listen(LISTEN_LISTEN, channel);
6577 1052 : 58 : }
1053 : :
1054 : : /*
1055 : : * Async_Unlisten
1056 : : *
1057 : : * This is executed by the SQL unlisten command.
1058 : : */
1059 : : void
5871 1060 : 3 : Async_Unlisten(const char *channel)
1061 : : {
6406 1062 [ - + ]: 3 : if (Trace_notify)
5871 tgl@sss.pgh.pa.us 1063 [ # # ]:UBC 0 : elog(DEBUG1, "Async_Unlisten(%s,%d)", channel, MyProcPid);
1064 : :
1065 : : /* If we couldn't possibly be listening, no need to queue anything */
2354 rhaas@postgresql.org 1066 [ + - - + ]:CBC 3 : if (pendingActions == NULL && !unlistenExitRegistered)
6239 tgl@sss.pgh.pa.us 1067 :UBC 0 : return;
1068 : :
5871 tgl@sss.pgh.pa.us 1069 :CBC 3 : queue_listen(LISTEN_UNLISTEN, channel);
1070 : : }
1071 : :
1072 : : /*
1073 : : * Async_UnlistenAll
1074 : : *
1075 : : * This is invoked by UNLISTEN * command, and also at backend exit.
1076 : : */
1077 : : void
6577 1078 : 78 : Async_UnlistenAll(void)
1079 : : {
1080 [ - + ]: 78 : if (Trace_notify)
6577 tgl@sss.pgh.pa.us 1081 [ # # ]:UBC 0 : elog(DEBUG1, "Async_UnlistenAll(%d)", MyProcPid);
1082 : :
1083 : : /* If we couldn't possibly be listening, no need to queue anything */
2354 rhaas@postgresql.org 1084 [ + + + + ]:CBC 78 : if (pendingActions == NULL && !unlistenExitRegistered)
6239 tgl@sss.pgh.pa.us 1085 : 29 : return;
1086 : :
6577 1087 : 49 : queue_listen(LISTEN_UNLISTEN_ALL, "");
1088 : : }
1089 : :
1090 : : /*
1091 : : * SQL function: return a set of the channel names this backend is actively
1092 : : * listening to.
1093 : : *
1094 : : * Note: this coding relies on the fact that the localChannelTable cannot
1095 : : * change within a transaction.
1096 : : */
1097 : : Datum
5871 1098 : 9 : pg_listening_channels(PG_FUNCTION_ARGS)
1099 : : {
1100 : : FuncCallContext *funcctx;
1101 : : HASH_SEQ_STATUS *status;
1102 : :
1103 : : /* stuff done only on the first call of the function */
1104 [ + + ]: 9 : if (SRF_IS_FIRSTCALL())
1105 : : {
1106 : : /* create a function context for cross-call persistence */
1107 : 6 : funcctx = SRF_FIRSTCALL_INIT();
1108 : :
1109 : : /* Initialize hash table iteration if we have any channels */
59 tgl@sss.pgh.pa.us 1110 [ + - ]:GNC 6 : if (localChannelTable != NULL)
1111 : : {
1112 : : MemoryContext oldcontext;
1113 : :
1114 : 6 : oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx);
1115 : 6 : status = (HASH_SEQ_STATUS *) palloc(sizeof(HASH_SEQ_STATUS));
1116 : 6 : hash_seq_init(status, localChannelTable);
1117 : 6 : funcctx->user_fctx = status;
1118 : 6 : MemoryContextSwitchTo(oldcontext);
1119 : : }
1120 : : else
1121 : : {
59 tgl@sss.pgh.pa.us 1122 :UNC 0 : funcctx->user_fctx = NULL;
1123 : : }
1124 : : }
1125 : :
1126 : : /* stuff done on every call of the function */
5871 tgl@sss.pgh.pa.us 1127 :CBC 9 : funcctx = SRF_PERCALL_SETUP();
59 tgl@sss.pgh.pa.us 1128 :GNC 9 : status = (HASH_SEQ_STATUS *) funcctx->user_fctx;
1129 : :
1130 [ + - ]: 9 : if (status != NULL)
1131 : : {
1132 : : ChannelName *entry;
1133 : :
1134 : 9 : entry = (ChannelName *) hash_seq_search(status);
1135 [ + + ]: 9 : if (entry != NULL)
1136 : 3 : SRF_RETURN_NEXT(funcctx, CStringGetTextDatum(entry->channel));
1137 : : }
1138 : :
5871 tgl@sss.pgh.pa.us 1139 :CBC 6 : SRF_RETURN_DONE(funcctx);
1140 : : }
1141 : :
1142 : : /*
1143 : : * Async_UnlistenOnExit
1144 : : *
1145 : : * This is executed at backend exit if we have done any LISTENs in this
1146 : : * backend. It might not be necessary anymore, if the user UNLISTENed
1147 : : * everything, but we don't try to detect that case.
1148 : : */
1149 : : static void
6577 1150 : 17 : Async_UnlistenOnExit(int code, Datum arg)
1151 : : {
59 tgl@sss.pgh.pa.us 1152 :GNC 17 : CleanupListenersOnExit();
4778 tgl@sss.pgh.pa.us 1153 :CBC 17 : asyncQueueUnregister();
6577 1154 : 17 : }
1155 : :
1156 : : /*
1157 : : * AtPrepare_Notify
1158 : : *
1159 : : * This is called at the prepare phase of a two-phase
1160 : : * transaction. Save the state for possible commit later.
1161 : : */
1162 : : void
1163 : 315 : AtPrepare_Notify(void)
1164 : : {
1165 : : /* It's not allowed to have any pending LISTEN/UNLISTEN/NOTIFY actions */
5871 1166 [ + - - + ]: 315 : if (pendingActions || pendingNotifies)
6577 tgl@sss.pgh.pa.us 1167 [ # # ]:UBC 0 : ereport(ERROR,
1168 : : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
1169 : : errmsg("cannot PREPARE a transaction that has executed LISTEN, UNLISTEN, or NOTIFY")));
5871 tgl@sss.pgh.pa.us 1170 :CBC 315 : }
1171 : :
1172 : : /*
1173 : : * PreCommit_Notify
1174 : : *
1175 : : * This is called at transaction commit, before actually committing to
1176 : : * clog.
1177 : : *
1178 : : * If there are pending LISTEN actions, make sure we are listed in the
1179 : : * shared-memory listener array. This must happen before commit to
1180 : : * ensure we don't miss any notifies from transactions that commit
1181 : : * just after ours.
1182 : : *
1183 : : * If there are outbound notify requests in the pendingNotifies list,
1184 : : * add them to the global queue. We do that before commit so that
1185 : : * we can still throw error if we run out of queue space.
1186 : : */
1187 : : void
1188 : 310613 : PreCommit_Notify(void)
1189 : : {
1190 : : ListCell *p;
1191 : :
2404 1192 [ + + + + ]: 310613 : if (!pendingActions && !pendingNotifies)
5871 1193 : 310467 : return; /* no relevant statements in this xact */
1194 : :
1195 [ - + ]: 146 : if (Trace_notify)
5871 tgl@sss.pgh.pa.us 1196 [ # # ]:UBC 0 : elog(DEBUG1, "PreCommit_Notify");
1197 : :
1198 : : /* Preflight for any pending listen/unlisten actions */
59 tgl@sss.pgh.pa.us 1199 :GNC 146 : initGlobalChannelTable();
1200 : :
2354 rhaas@postgresql.org 1201 [ + + ]:CBC 146 : if (pendingActions != NULL)
1202 : : {
1203 : : /* Ensure we have a local channel table */
59 tgl@sss.pgh.pa.us 1204 :GNC 91 : initLocalChannelTable();
1205 : : /* Create pendingListenActions hash table for this transaction */
1206 : 91 : initPendingListenActions();
1207 : :
1208 : : /* Stage all the actions this transaction wants to perform */
2354 rhaas@postgresql.org 1209 [ + - + + :CBC 198 : foreach(p, pendingActions->actions)
+ + ]
1210 : : {
1211 : 107 : ListenAction *actrec = (ListenAction *) lfirst(p);
1212 : :
1213 [ + + + - ]: 107 : switch (actrec->action)
1214 : : {
1215 : 56 : case LISTEN_LISTEN:
59 tgl@sss.pgh.pa.us 1216 :GNC 56 : BecomeRegisteredListener();
1217 : 56 : PrepareTableEntriesForListen(actrec->channel);
2354 rhaas@postgresql.org 1218 :CBC 56 : break;
1219 : 3 : case LISTEN_UNLISTEN:
59 tgl@sss.pgh.pa.us 1220 :GNC 3 : PrepareTableEntriesForUnlisten(actrec->channel);
2354 rhaas@postgresql.org 1221 :CBC 3 : break;
1222 : 48 : case LISTEN_UNLISTEN_ALL:
59 tgl@sss.pgh.pa.us 1223 :GNC 48 : PrepareTableEntriesForUnlistenAll();
2354 rhaas@postgresql.org 1224 :CBC 48 : break;
1225 : : }
1226 : : }
1227 : : }
1228 : :
1229 : : /* Queue any pending notifies (must happen after the above) */
5871 tgl@sss.pgh.pa.us 1230 [ + + ]: 146 : if (pendingNotifies)
1231 : : {
1232 : : ListCell *nextNotify;
59 tgl@sss.pgh.pa.us 1233 :GNC 55 : bool firstIteration = true;
1234 : :
1235 : : /*
1236 : : * Build list of unique channel names being notified for use by
1237 : : * SignalBackends().
1238 : : *
1239 : : * If uniqueChannelHash is available, use it to efficiently get the
1240 : : * unique channels. Otherwise, fall back to the O(N^2) approach.
1241 : : */
1242 : 55 : pendingNotifies->uniqueChannelNames = NIL;
1243 [ + + ]: 55 : if (pendingNotifies->uniqueChannelHash != NULL)
1244 : : {
1245 : : HASH_SEQ_STATUS status;
1246 : : ChannelName *channelEntry;
1247 : :
1248 : 2 : hash_seq_init(&status, pendingNotifies->uniqueChannelHash);
1249 [ + + ]: 4 : while ((channelEntry = (ChannelName *) hash_seq_search(&status)) != NULL)
1250 : 2 : pendingNotifies->uniqueChannelNames =
1251 : 2 : lappend(pendingNotifies->uniqueChannelNames,
1252 : 2 : channelEntry->channel);
1253 : : }
1254 : : else
1255 : : {
1256 : : /* O(N^2) approach is better for small number of notifications */
1257 [ + - + + : 189 : foreach_ptr(Notification, n, pendingNotifies->events)
+ + ]
1258 : : {
1259 : 83 : char *channel = n->data;
1260 : 83 : bool found = false;
1261 : :
1262 : : /* Name present in list? */
1263 [ + + + + : 170 : foreach_ptr(char, oldchan, pendingNotifies->uniqueChannelNames)
+ + ]
1264 : : {
1265 [ + + ]: 31 : if (strcmp(oldchan, channel) == 0)
1266 : : {
1267 : 27 : found = true;
1268 : 27 : break;
1269 : : }
1270 : : }
1271 : : /* Add if not already in list */
1272 [ + + ]: 83 : if (!found)
1273 : 56 : pendingNotifies->uniqueChannelNames =
1274 : 56 : lappend(pendingNotifies->uniqueChannelNames,
1275 : : channel);
1276 : : }
1277 : : }
1278 : :
1279 : : /* Preallocate workspace that will be needed by SignalBackends() */
1280 [ + + ]: 55 : if (signalPids == NULL)
1281 : 12 : signalPids = MemoryContextAlloc(TopMemoryContext,
1282 : : MaxBackends * sizeof(int32));
1283 : :
1284 [ + + ]: 55 : if (signalProcnos == NULL)
1285 : 12 : signalProcnos = MemoryContextAlloc(TopMemoryContext,
1286 : : MaxBackends * sizeof(ProcNumber));
1287 : :
1288 : : /*
1289 : : * Make sure that we have an XID assigned to the current transaction.
1290 : : * GetCurrentTransactionId is cheap if we already have an XID, but not
1291 : : * so cheap if we don't, and we'd prefer not to do that work while
1292 : : * holding NotifyQueueLock.
1293 : : */
5871 tgl@sss.pgh.pa.us 1294 :CBC 55 : (void) GetCurrentTransactionId();
1295 : :
1296 : : /*
1297 : : * Serialize writers by acquiring a special lock that we hold till
1298 : : * after commit. This ensures that queue entries appear in commit
1299 : : * order, and in particular that there are never uncommitted queue
1300 : : * entries ahead of committed ones, so an uncommitted transaction
1301 : : * can't block delivery of deliverable notifications.
1302 : : *
1303 : : * We use a heavyweight lock so that it'll automatically be released
1304 : : * after either commit or abort. This also allows deadlocks to be
1305 : : * detected, though really a deadlock shouldn't be possible here.
1306 : : *
1307 : : * The lock is on "database 0", which is pretty ugly but it doesn't
1308 : : * seem worth inventing a special locktag category just for this.
1309 : : * (Historical note: before PG 9.0, a similar lock on "database 0" was
1310 : : * used by the flatfiles mechanism.)
1311 : : */
1312 : 55 : LockSharedObject(DatabaseRelationId, InvalidOid, 0,
1313 : : AccessExclusiveLock);
1314 : :
1315 : : /*
1316 : : * For the direct advancement optimization in SignalBackends(), we
1317 : : * need to ensure that no other backend can insert queue entries
1318 : : * between queueHeadBeforeWrite and queueHeadAfterWrite. The
1319 : : * heavyweight lock above provides this guarantee, since it serializes
1320 : : * all writers.
1321 : : *
1322 : : * Note: if the heavyweight lock were ever removed for scalability
1323 : : * reasons, we could achieve the same guarantee by holding
1324 : : * NotifyQueueLock in EXCLUSIVE mode across all our insertions, rather
1325 : : * than releasing and reacquiring it for each page as we do below.
1326 : : */
1327 : :
1328 : : /* Initialize values to a safe default in case list is empty */
59 tgl@sss.pgh.pa.us 1329 :GNC 55 : SET_QUEUE_POS(queueHeadBeforeWrite, 0, 0);
1330 : 55 : SET_QUEUE_POS(queueHeadAfterWrite, 0, 0);
1331 : :
1332 : : /* Now push the notifications into the queue */
2404 tgl@sss.pgh.pa.us 1333 :CBC 55 : nextNotify = list_head(pendingNotifies->events);
5871 1334 [ + + ]: 145 : while (nextNotify != NULL)
1335 : : {
1336 : : /*
1337 : : * Add the pending notifications to the queue. We acquire and
1338 : : * release NotifyQueueLock once per page, which might be overkill
1339 : : * but it does allow readers to get in while we're doing this.
1340 : : *
1341 : : * A full queue is very uncommon and should really not happen,
1342 : : * given that we have so much space available in the SLRU pages.
1343 : : * Nevertheless we need to deal with this possibility. Note that
1344 : : * when we get here we are in the process of committing our
1345 : : * transaction, but we have not yet committed to clog, so at this
1346 : : * point in time we can still roll the transaction back.
1347 : : */
2130 1348 : 90 : LWLockAcquire(NotifyQueueLock, LW_EXCLUSIVE);
59 tgl@sss.pgh.pa.us 1349 [ + + ]:GNC 90 : if (firstIteration)
1350 : : {
1351 : 55 : queueHeadBeforeWrite = QUEUE_HEAD;
1352 : 55 : firstIteration = false;
1353 : : }
5871 tgl@sss.pgh.pa.us 1354 :CBC 90 : asyncQueueFillWarning();
1355 [ - + ]: 90 : if (asyncQueueIsFull())
5871 tgl@sss.pgh.pa.us 1356 [ # # ]:UBC 0 : ereport(ERROR,
1357 : : (errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
1358 : : errmsg("too many notifications in the NOTIFY queue")));
5871 tgl@sss.pgh.pa.us 1359 :CBC 90 : nextNotify = asyncQueueAddEntries(nextNotify);
59 tgl@sss.pgh.pa.us 1360 :GNC 90 : queueHeadAfterWrite = QUEUE_HEAD;
2130 tgl@sss.pgh.pa.us 1361 :CBC 90 : LWLockRelease(NotifyQueueLock);
1362 : : }
1363 : :
1364 : : /* Note that we don't clear pendingNotifies; AtCommit_Notify will. */
1365 : : }
1366 : : }
1367 : :
1368 : : /*
1369 : : * AtCommit_Notify
1370 : : *
1371 : : * This is called at transaction commit, after committing to clog.
1372 : : *
1373 : : * Apply pending listen/unlisten changes and clear transaction-local state.
1374 : : *
1375 : : * If we issued any notifications in the transaction, send signals to
1376 : : * listening backends (possibly including ourselves) to process them.
1377 : : * Also, if we filled enough queue pages with new notifies, try to
1378 : : * advance the queue tail pointer.
1379 : : */
1380 : : void
6577 1381 : 310458 : AtCommit_Notify(void)
1382 : : {
1383 : : /*
1384 : : * Allow transactions that have not executed LISTEN/UNLISTEN/NOTIFY to
1385 : : * return as soon as possible
1386 : : */
5871 1387 [ + + + + ]: 310458 : if (!pendingActions && !pendingNotifies)
6577 1388 : 310312 : return;
1389 : :
1390 [ - + ]: 146 : if (Trace_notify)
6577 tgl@sss.pgh.pa.us 1391 [ # # ]:UBC 0 : elog(DEBUG1, "AtCommit_Notify");
1392 : :
1393 : : /* Apply staged listen/unlisten changes */
59 tgl@sss.pgh.pa.us 1394 :GNC 146 : ApplyPendingListenActions(true);
1395 : :
1396 : : /* If no longer listening to anything, get out of listener array */
1397 [ + + + - : 146 : if (amRegisteredListener && LocalChannelTableIsEmpty())
+ + ]
4778 tgl@sss.pgh.pa.us 1398 :CBC 23 : asyncQueueUnregister();
1399 : :
1400 : : /*
1401 : : * Send signals to listening backends. We need do this only if there are
1402 : : * pending notifies, which were previously added to the shared queue by
1403 : : * PreCommit_Notify().
1404 : : */
1643 1405 [ + + ]: 146 : if (pendingNotifies != NULL)
1406 : 55 : SignalBackends();
1407 : :
1408 : : /*
1409 : : * If it's time to try to advance the global tail pointer, do that.
1410 : : *
1411 : : * (It might seem odd to do this in the sender, when more than likely the
1412 : : * listeners won't yet have read the messages we just sent. However,
1413 : : * there's less contention if only the sender does it, and there is little
1414 : : * need for urgency in advancing the global tail. So this typically will
1415 : : * be clearing out messages that were sent some time ago.)
1416 : : */
1417 [ + + ]: 146 : if (tryAdvanceTail)
1418 : : {
1419 : 8 : tryAdvanceTail = false;
1420 : 8 : asyncQueueAdvanceTail();
1421 : : }
1422 : :
1423 : : /* And clean up */
5871 1424 : 146 : ClearPendingActionsAndNotifies();
1425 : : }
1426 : :
1427 : : /*
1428 : : * BecomeRegisteredListener --- subroutine for PreCommit_Notify
1429 : : *
1430 : : * This function must make sure we are ready to catch any incoming messages.
1431 : : */
1432 : : static void
59 tgl@sss.pgh.pa.us 1433 :GNC 56 : BecomeRegisteredListener(void)
1434 : : {
1435 : : QueuePosition head;
1436 : : QueuePosition max;
1437 : : ProcNumber prevListener;
1438 : :
1439 : : /*
1440 : : * Nothing to do if we are already listening to something, nor if we
1441 : : * already ran this routine in this transaction.
1442 : : */
4778 tgl@sss.pgh.pa.us 1443 [ + + ]:CBC 56 : if (amRegisteredListener)
5871 1444 : 27 : return;
1445 : :
1446 [ - + ]: 29 : if (Trace_notify)
59 tgl@sss.pgh.pa.us 1447 [ # # ]:UNC 0 : elog(DEBUG1, "BecomeRegisteredListener(%d)", MyProcPid);
1448 : :
1449 : : /*
1450 : : * Before registering, make sure we will unlisten before dying. (Note:
1451 : : * this action does not get undone if we abort later.)
1452 : : */
5871 tgl@sss.pgh.pa.us 1453 [ + + ]:CBC 29 : if (!unlistenExitRegistered)
1454 : : {
4470 rhaas@postgresql.org 1455 : 17 : before_shmem_exit(Async_UnlistenOnExit, 0);
5871 tgl@sss.pgh.pa.us 1456 : 17 : unlistenExitRegistered = true;
1457 : : }
1458 : :
1459 : : /*
1460 : : * This is our first LISTEN, so establish our pointer.
1461 : : *
1462 : : * We set our pointer to the global tail pointer and then move it forward
1463 : : * over already-committed notifications. This ensures we cannot miss any
1464 : : * not-yet-committed notifications. We might get a few more but that
1465 : : * doesn't hurt.
1466 : : *
1467 : : * In some scenarios there might be a lot of committed notifications that
1468 : : * have not yet been pruned away (because some backend is being lazy about
1469 : : * reading them). To reduce our startup time, we can look at other
1470 : : * backends and adopt the maximum "pos" pointer of any backend that's in
1471 : : * our database; any notifications it's already advanced over are surely
1472 : : * committed and need not be re-examined by us. (We must consider only
1473 : : * backends connected to our DB, because others will not have bothered to
1474 : : * check committed-ness of notifications in our DB.)
1475 : : *
1476 : : * We need exclusive lock here so we can look at other backends' entries
1477 : : * and manipulate the list links.
1478 : : */
2130 1479 : 29 : LWLockAcquire(NotifyQueueLock, LW_EXCLUSIVE);
3819 1480 : 29 : head = QUEUE_HEAD;
1481 : 29 : max = QUEUE_TAIL;
742 heikki.linnakangas@i 1482 : 29 : prevListener = INVALID_PROC_NUMBER;
1483 [ + + ]: 44 : for (ProcNumber i = QUEUE_FIRST_LISTENER; i != INVALID_PROC_NUMBER; i = QUEUE_NEXT_LISTENER(i))
1484 : : {
2378 tgl@sss.pgh.pa.us 1485 [ + - ]: 15 : if (QUEUE_BACKEND_DBOID(i) == MyDatabaseId)
1486 [ - + + - : 15 : max = QUEUE_POS_MAX(max, QUEUE_BACKEND_POS(i));
+ - ]
1487 : : /* Also find last listening backend before this one */
742 heikki.linnakangas@i 1488 [ + + ]: 15 : if (i < MyProcNumber)
2378 tgl@sss.pgh.pa.us 1489 : 8 : prevListener = i;
1490 : : }
742 heikki.linnakangas@i 1491 : 29 : QUEUE_BACKEND_POS(MyProcNumber) = max;
1492 : 29 : QUEUE_BACKEND_PID(MyProcNumber) = MyProcPid;
1493 : 29 : QUEUE_BACKEND_DBOID(MyProcNumber) = MyDatabaseId;
59 tgl@sss.pgh.pa.us 1494 :GNC 29 : QUEUE_BACKEND_WAKEUP_PENDING(MyProcNumber) = false;
1495 : 29 : QUEUE_BACKEND_IS_ADVANCING(MyProcNumber) = false;
1496 : : /* Insert backend into list of listeners at correct position */
742 heikki.linnakangas@i 1497 [ + + ]:CBC 29 : if (prevListener != INVALID_PROC_NUMBER)
1498 : : {
1499 : 4 : QUEUE_NEXT_LISTENER(MyProcNumber) = QUEUE_NEXT_LISTENER(prevListener);
1500 : 4 : QUEUE_NEXT_LISTENER(prevListener) = MyProcNumber;
1501 : : }
1502 : : else
1503 : : {
1504 : 25 : QUEUE_NEXT_LISTENER(MyProcNumber) = QUEUE_FIRST_LISTENER;
1505 : 25 : QUEUE_FIRST_LISTENER = MyProcNumber;
1506 : : }
2130 tgl@sss.pgh.pa.us 1507 : 29 : LWLockRelease(NotifyQueueLock);
1508 : :
1509 : : /* Now we are listed in the global array, so remember we're listening */
4778 1510 : 29 : amRegisteredListener = true;
1511 : :
1512 : : /*
1513 : : * Try to move our pointer forward as far as possible. This will skip
1514 : : * over already-committed notifications, which we want to do because they
1515 : : * might be quite stale. Note that we are not yet listening on anything,
1516 : : * so we won't deliver such notifications to our frontend. Also, although
1517 : : * our transaction might have executed NOTIFY, those message(s) aren't
1518 : : * queued yet so we won't skip them here.
1519 : : */
3819 1520 [ + + + + ]: 29 : if (!QUEUE_POS_EQUAL(max, head))
1521 : 15 : asyncQueueReadAllNotifications();
1522 : : }
1523 : :
1524 : : /*
1525 : : * PrepareTableEntriesForListen --- subroutine for PreCommit_Notify
1526 : : *
1527 : : * Prepare a LISTEN by recording it in pendingListenActions, pre-allocating
1528 : : * an entry in localChannelTable, and pre-allocating an entry in the shared
1529 : : * globalChannelTable with listening=false. The listening flag will be set
1530 : : * to true in AtCommit_Notify. If we abort later, unwanted table entries
1531 : : * will be removed.
1532 : : */
1533 : : static void
59 tgl@sss.pgh.pa.us 1534 :GNC 56 : PrepareTableEntriesForListen(const char *channel)
1535 : : {
1536 : : GlobalChannelKey key;
1537 : : GlobalChannelEntry *entry;
1538 : : bool found;
1539 : : ListenerEntry *listeners;
1540 : : PendingListenEntry *pending;
1541 : :
1542 : : /*
1543 : : * Record in local pending hash that we want to LISTEN, overwriting any
1544 : : * earlier attempt to UNLISTEN.
1545 : : */
1546 : : pending = (PendingListenEntry *)
1547 : 56 : hash_search(pendingListenActions, channel, HASH_ENTER, NULL);
1548 : 56 : pending->action = PENDING_LISTEN;
1549 : :
1550 : : /*
1551 : : * Ensure that there is an entry for the channel in localChannelTable.
1552 : : * (Should this fail, we can just roll back.) If the transaction fails
1553 : : * after this point, we will remove the entry if appropriate during
1554 : : * ApplyPendingListenActions. Note that this entry allows IsListeningOn()
1555 : : * to return TRUE; we assume nothing is going to consult that before
1556 : : * AtCommit_Notify/AtAbort_Notify. However, if later actions attempt to
1557 : : * UNLISTEN this channel or UNLISTEN *, we need to have the local entry
1558 : : * present to ensure they do the right things; see
1559 : : * PrepareTableEntriesForUnlisten and PrepareTableEntriesForUnlistenAll.
1560 : : */
1561 : 56 : (void) hash_search(localChannelTable, channel, HASH_ENTER, NULL);
1562 : :
1563 : : /* Pre-allocate entry in shared globalChannelTable with listening=false */
1564 : 56 : GlobalChannelKeyInit(&key, MyDatabaseId, channel);
1565 : 56 : entry = dshash_find_or_insert(globalChannelTable, &key, &found);
1566 : :
1567 [ + + ]: 56 : if (!found)
1568 : : {
1569 : : /* New channel entry, so initialize it to a safe state */
1570 : 34 : entry->listenersArray = InvalidDsaPointer;
1571 : 34 : entry->numListeners = 0;
1572 : 34 : entry->allocatedListeners = 0;
1573 : : }
1574 : :
1575 : : /*
1576 : : * Create listenersArray if entry doesn't have one. It's tempting to fold
1577 : : * this into the !found case, but this coding allows us to cope in case
1578 : : * dsa_allocate() failed in an earlier attempt.
1579 : : */
1580 [ + + ]: 56 : if (!DsaPointerIsValid(entry->listenersArray))
1581 : : {
1582 : 34 : entry->listenersArray = dsa_allocate(globalChannelDSA,
1583 : : sizeof(ListenerEntry) * INITIAL_LISTENERS_ARRAY_SIZE);
1584 : 34 : entry->allocatedListeners = INITIAL_LISTENERS_ARRAY_SIZE;
1585 : : }
1586 : :
1587 : : listeners = (ListenerEntry *)
1588 : 56 : dsa_get_address(globalChannelDSA, entry->listenersArray);
1589 : :
1590 : : /*
1591 : : * Check if we already have a ListenerEntry (possibly from earlier in this
1592 : : * transaction)
1593 : : */
1594 [ + + ]: 75 : for (int i = 0; i < entry->numListeners; i++)
1595 : : {
1596 [ + + ]: 30 : if (listeners[i].procNo == MyProcNumber)
1597 : : {
1598 : : /* Already have an entry; listening flag stays as-is until commit */
1599 : 11 : dshash_release_lock(globalChannelTable, entry);
1600 : 11 : return;
1601 : : }
1602 : : }
1603 : :
1604 : : /* Need to add a new entry; grow array if necessary */
1605 [ + + ]: 45 : if (entry->numListeners >= entry->allocatedListeners)
1606 : : {
1607 : 1 : int new_size = entry->allocatedListeners * 2;
1608 : 1 : dsa_pointer old_array = entry->listenersArray;
1609 : 1 : dsa_pointer new_array = dsa_allocate(globalChannelDSA,
1610 : : sizeof(ListenerEntry) * new_size);
1611 : 1 : ListenerEntry *new_listeners = (ListenerEntry *) dsa_get_address(globalChannelDSA, new_array);
1612 : :
1613 : 1 : memcpy(new_listeners, listeners, sizeof(ListenerEntry) * entry->numListeners);
1614 : 1 : entry->listenersArray = new_array;
1615 : 1 : entry->allocatedListeners = new_size;
1616 : 1 : dsa_free(globalChannelDSA, old_array);
1617 : 1 : listeners = new_listeners;
1618 : : }
1619 : :
1620 : 45 : listeners[entry->numListeners].procNo = MyProcNumber;
1621 : 45 : listeners[entry->numListeners].listening = false; /* staged, not yet
1622 : : * committed */
1623 : 45 : entry->numListeners++;
1624 : :
1625 : 45 : dshash_release_lock(globalChannelTable, entry);
1626 : : }
1627 : :
1628 : : /*
1629 : : * PrepareTableEntriesForUnlisten --- subroutine for PreCommit_Notify
1630 : : *
1631 : : * Prepare an UNLISTEN by recording it in pendingListenActions, but only if
1632 : : * we're currently listening (committed or staged). We don't touch
1633 : : * globalChannelTable yet - the listener keeps receiving signals until
1634 : : * commit, when the entry is removed.
1635 : : */
1636 : : static void
1637 : 3 : PrepareTableEntriesForUnlisten(const char *channel)
1638 : : {
1639 : : PendingListenEntry *pending;
1640 : :
1641 : : /*
1642 : : * If the channel name is not in localChannelTable, then we are neither
1643 : : * listening on it nor preparing to listen on it, so we don't need to
1644 : : * record an UNLISTEN action.
1645 : : */
1646 [ - + ]: 3 : Assert(localChannelTable != NULL);
1647 [ - + ]: 3 : if (hash_search(localChannelTable, channel, HASH_FIND, NULL) == NULL)
5871 tgl@sss.pgh.pa.us 1648 :LBC (10) : return;
1649 : :
1650 : : /*
1651 : : * Record in local pending hash that we want to UNLISTEN, overwriting any
1652 : : * earlier attempt to LISTEN. Don't touch localChannelTable or
1653 : : * globalChannelTable yet - we keep receiving signals until commit.
1654 : : */
1655 : : pending = (PendingListenEntry *)
59 tgl@sss.pgh.pa.us 1656 :GNC 3 : hash_search(pendingListenActions, channel, HASH_ENTER, NULL);
1657 : 3 : pending->action = PENDING_UNLISTEN;
1658 : : }
1659 : :
1660 : : /*
1661 : : * PrepareTableEntriesForUnlistenAll --- subroutine for PreCommit_Notify
1662 : : *
1663 : : * Prepare UNLISTEN * by recording an UNLISTEN for all listened or
1664 : : * about-to-be-listened channels in pendingListenActions.
1665 : : */
1666 : : static void
1667 : 48 : PrepareTableEntriesForUnlistenAll(void)
1668 : : {
1669 : : HASH_SEQ_STATUS seq;
1670 : : ChannelName *channelEntry;
1671 : : PendingListenEntry *pending;
1672 : :
1673 : : /*
1674 : : * Scan localChannelTable, which will have the names of all channels that
1675 : : * we are listening on or have prepared to listen on. Record an UNLISTEN
1676 : : * action for each one, overwriting any earlier attempt to LISTEN.
1677 : : */
1678 : 48 : hash_seq_init(&seq, localChannelTable);
1679 [ + + ]: 82 : while ((channelEntry = (ChannelName *) hash_seq_search(&seq)) != NULL)
1680 : : {
1681 : : pending = (PendingListenEntry *)
1682 : 34 : hash_search(pendingListenActions, channelEntry->channel, HASH_ENTER, NULL);
1683 : 34 : pending->action = PENDING_UNLISTEN;
1684 : : }
1685 : 48 : }
1686 : :
1687 : : /*
1688 : : * RemoveListenerFromChannel --- remove idx'th listener from global channel entry
1689 : : *
1690 : : * Decrements numListeners, compacts the array, and frees the entry if empty.
1691 : : * Sets *entry_ptr to NULL if the entry was deleted.
1692 : : *
1693 : : * We could get the listeners pointer from the entry, but all callers
1694 : : * already have it at hand.
1695 : : */
1696 : : static void
1697 : 37 : RemoveListenerFromChannel(GlobalChannelEntry **entry_ptr,
1698 : : ListenerEntry *listeners,
1699 : : int idx)
1700 : : {
1701 : 37 : GlobalChannelEntry *entry = *entry_ptr;
1702 : :
1703 : 37 : entry->numListeners--;
1704 [ + + ]: 37 : if (idx < entry->numListeners)
1705 : 8 : memmove(&listeners[idx], &listeners[idx + 1],
1706 : 8 : sizeof(ListenerEntry) * (entry->numListeners - idx));
1707 : :
1708 [ + + ]: 37 : if (entry->numListeners == 0)
1709 : : {
1710 : 26 : dsa_free(globalChannelDSA, entry->listenersArray);
1711 : 26 : dshash_delete_entry(globalChannelTable, entry);
1712 : : /* tells caller not to release the entry's lock: */
1713 : 26 : *entry_ptr = NULL;
1714 : : }
1715 : 37 : }
1716 : :
1717 : : /*
1718 : : * ApplyPendingListenActions
1719 : : *
1720 : : * Apply, or revert, staged listen/unlisten changes to the local and global
1721 : : * hash tables.
1722 : : */
1723 : : static void
1724 : 26888 : ApplyPendingListenActions(bool isCommit)
1725 : : {
1726 : : HASH_SEQ_STATUS seq;
1727 : : PendingListenEntry *pending;
1728 : :
1729 : : /* Quick exit if nothing to do */
1730 [ + + ]: 26888 : if (pendingListenActions == NULL)
1731 : 26797 : return;
1732 : :
1733 : : /* We made a globalChannelTable before building pendingListenActions */
1734 [ - + ]: 91 : if (globalChannelTable == NULL)
59 tgl@sss.pgh.pa.us 1735 [ # # ]:UNC 0 : elog(PANIC, "global channel table missing post-commit/abort");
1736 : :
1737 : : /* For each staged action ... */
59 tgl@sss.pgh.pa.us 1738 :GNC 91 : hash_seq_init(&seq, pendingListenActions);
1739 [ + + ]: 183 : while ((pending = (PendingListenEntry *) hash_seq_search(&seq)) != NULL)
1740 : : {
1741 : : GlobalChannelKey key;
1742 : : GlobalChannelEntry *entry;
1743 : 92 : bool removeLocal = true;
1744 : 92 : bool foundListener = false;
1745 : :
1746 : : /*
1747 : : * Find the global entry for this channel. If isCommit, it had better
1748 : : * exist (it was created in PreCommit). In an abort, it might not
1749 : : * exist, in which case we are not listening and should discard any
1750 : : * local entry that PreCommit may have managed to create.
1751 : : */
1752 : 92 : GlobalChannelKeyInit(&key, MyDatabaseId, pending->channel);
1753 : 92 : entry = dshash_find(globalChannelTable, &key, true);
1754 [ + - ]: 92 : if (entry != NULL)
1755 : : {
1756 : : /* Scan entry to find the ListenerEntry for this backend */
1757 : : ListenerEntry *listeners;
1758 : :
1759 : : listeners = (ListenerEntry *)
1760 : 92 : dsa_get_address(globalChannelDSA, entry->listenersArray);
1761 : :
1762 [ + - ]: 116 : for (int i = 0; i < entry->numListeners; i++)
1763 : : {
1764 [ + + ]: 116 : if (listeners[i].procNo != MyProcNumber)
1765 : 24 : continue;
1766 : 92 : foundListener = true;
1767 [ + - ]: 92 : if (isCommit)
1768 : : {
1769 [ + + ]: 92 : if (pending->action == PENDING_LISTEN)
1770 : : {
1771 : : /*
1772 : : * LISTEN being committed: set listening=true.
1773 : : * localChannelTable entry was created during
1774 : : * PreCommit and should be kept.
1775 : : */
1776 : 55 : listeners[i].listening = true;
1777 : 55 : removeLocal = false;
1778 : : }
1779 : : else
1780 : : {
1781 : : /*
1782 : : * UNLISTEN being committed: remove pre-allocated
1783 : : * entries from both tables.
1784 : : */
1785 : 37 : RemoveListenerFromChannel(&entry, listeners, i);
1786 : : }
1787 : : }
1788 : : else
1789 : : {
1790 : : /*
1791 : : * Note: this part is reachable only if the transaction
1792 : : * aborts after PreCommit_Notify() has made some
1793 : : * pendingListenActions entries, so it's pretty hard to
1794 : : * test.
1795 : : */
59 tgl@sss.pgh.pa.us 1796 [ # # ]:UNC 0 : if (!listeners[i].listening)
1797 : : {
1798 : : /*
1799 : : * Staged LISTEN (or LISTEN+UNLISTEN) being aborted,
1800 : : * and we weren't listening before, so remove
1801 : : * pre-allocated entries from both tables.
1802 : : */
1803 : 0 : RemoveListenerFromChannel(&entry, listeners, i);
1804 : : }
1805 : : else
1806 : : {
1807 : : /*
1808 : : * We're aborting, but the previous state was that
1809 : : * we're listening, so keep localChannelTable entry.
1810 : : */
1811 : 0 : removeLocal = false;
1812 : : }
1813 : : }
59 tgl@sss.pgh.pa.us 1814 :GNC 92 : break; /* there shouldn't be another match */
1815 : : }
1816 : :
1817 : : /* We might have already released the entry by removing it */
1818 [ + + ]: 92 : if (entry != NULL)
1819 : 66 : dshash_release_lock(globalChannelTable, entry);
1820 : : }
1821 : :
1822 : : /*
1823 : : * If we're committing a LISTEN action, we should have found a
1824 : : * matching ListenerEntry, but otherwise it's okay if we didn't.
1825 : : */
1826 [ + - + + : 92 : if (isCommit && pending->action == PENDING_LISTEN && !foundListener)
- + ]
59 tgl@sss.pgh.pa.us 1827 [ # # ]:UNC 0 : elog(PANIC, "could not find listener entry for channel \"%s\" backend %d",
1828 : : pending->channel, MyProcNumber);
1829 : :
1830 : : /*
1831 : : * If we did not find a globalChannelTable entry for our backend, or
1832 : : * if we are unlistening, remove any localChannelTable entry that may
1833 : : * exist. (Note in particular that this cleans up if we created a
1834 : : * localChannelTable entry and then failed while trying to create a
1835 : : * globalChannelTable entry.)
1836 : : */
59 tgl@sss.pgh.pa.us 1837 [ + + + - ]:GNC 92 : if (removeLocal && localChannelTable != NULL)
1838 : 37 : (void) hash_search(localChannelTable, pending->channel,
1839 : : HASH_REMOVE, NULL);
1840 : : }
5871 tgl@sss.pgh.pa.us 1841 :ECB (3) : }
1842 : :
1843 : : /*
1844 : : * CleanupListenersOnExit --- called from Async_UnlistenOnExit
1845 : : *
1846 : : * Remove this backend from all channels in the shared global table.
1847 : : */
1848 : : static void
59 tgl@sss.pgh.pa.us 1849 :GNC 17 : CleanupListenersOnExit(void)
1850 : : {
1851 : : dshash_seq_status status;
1852 : : GlobalChannelEntry *entry;
1853 : :
5871 tgl@sss.pgh.pa.us 1854 [ - + ]:CBC 17 : if (Trace_notify)
59 tgl@sss.pgh.pa.us 1855 [ # # ]:UNC 0 : elog(DEBUG1, "CleanupListenersOnExit(%d)", MyProcPid);
1856 : :
1857 : : /* Clear our local cache (not really necessary, but be consistent) */
59 tgl@sss.pgh.pa.us 1858 [ + - ]:GNC 17 : if (localChannelTable != NULL)
1859 : : {
1860 : 17 : hash_destroy(localChannelTable);
1861 : 17 : localChannelTable = NULL;
1862 : : }
1863 : :
1864 : : /* Now remove our entries from the shared globalChannelTable */
1865 [ - + ]: 17 : if (globalChannelTable == NULL)
59 tgl@sss.pgh.pa.us 1866 :UNC 0 : return;
1867 : :
59 tgl@sss.pgh.pa.us 1868 :GNC 17 : dshash_seq_init(&status, globalChannelTable, true);
1869 [ + + ]: 25 : while ((entry = dshash_seq_next(&status)) != NULL)
1870 : : {
1871 : : ListenerEntry *listeners;
1872 : :
1873 [ - + ]: 8 : if (entry->key.dboid != MyDatabaseId)
59 tgl@sss.pgh.pa.us 1874 :UNC 0 : continue; /* not relevant */
1875 : :
1876 : : listeners = (ListenerEntry *)
59 tgl@sss.pgh.pa.us 1877 :GNC 8 : dsa_get_address(globalChannelDSA, entry->listenersArray);
1878 : :
1879 [ + - ]: 8 : for (int i = 0; i < entry->numListeners; i++)
1880 : : {
1881 [ + - ]: 8 : if (listeners[i].procNo == MyProcNumber)
1882 : : {
1883 : 8 : entry->numListeners--;
1884 [ - + ]: 8 : if (i < entry->numListeners)
59 tgl@sss.pgh.pa.us 1885 :UNC 0 : memmove(&listeners[i], &listeners[i + 1],
1886 : 0 : sizeof(ListenerEntry) * (entry->numListeners - i));
1887 : :
59 tgl@sss.pgh.pa.us 1888 [ + - ]:GNC 8 : if (entry->numListeners == 0)
1889 : : {
1890 : 8 : dsa_free(globalChannelDSA, entry->listenersArray);
1891 : 8 : dshash_delete_current(&status);
1892 : : }
1893 : 8 : break;
1894 : : }
1895 : : }
1896 : : }
1897 : 17 : dshash_seq_term(&status);
5871 tgl@sss.pgh.pa.us 1898 :ECB (30) : }
1899 : :
1900 : : /*
1901 : : * Test whether we are actively listening on the given channel name.
1902 : : *
1903 : : * Note: this function is executed for every notification found in the queue.
1904 : : */
1905 : : static bool
5871 tgl@sss.pgh.pa.us 1906 :CBC 55 : IsListeningOn(const char *channel)
1907 : : {
59 tgl@sss.pgh.pa.us 1908 [ - + ]:GNC 55 : if (localChannelTable == NULL)
59 tgl@sss.pgh.pa.us 1909 :UNC 0 : return false;
1910 : :
59 tgl@sss.pgh.pa.us 1911 :GNC 55 : return (hash_search(localChannelTable, channel, HASH_FIND, NULL) != NULL);
1912 : : }
1913 : :
1914 : : /*
1915 : : * Remove our entry from the listeners array when we are no longer listening
1916 : : * on any channel. NB: must not fail if we're already not listening.
1917 : : */
1918 : : static void
5871 tgl@sss.pgh.pa.us 1919 :CBC 40 : asyncQueueUnregister(void)
1920 : : {
59 tgl@sss.pgh.pa.us 1921 [ + + - + ]:GNC 40 : Assert(LocalChannelTableIsEmpty()); /* else caller error */
1922 : :
4673 bruce@momjian.us 1923 [ + + ]:CBC 40 : if (!amRegisteredListener) /* nothing to do */
4778 tgl@sss.pgh.pa.us 1924 : 11 : return;
1925 : :
1926 : : /*
1927 : : * Need exclusive lock here to manipulate list links.
1928 : : */
2130 1929 : 29 : LWLockAcquire(NotifyQueueLock, LW_EXCLUSIVE);
1930 : : /* Mark our entry as invalid */
742 heikki.linnakangas@i 1931 : 29 : QUEUE_BACKEND_PID(MyProcNumber) = InvalidPid;
1932 : 29 : QUEUE_BACKEND_DBOID(MyProcNumber) = InvalidOid;
59 tgl@sss.pgh.pa.us 1933 :GNC 29 : QUEUE_BACKEND_WAKEUP_PENDING(MyProcNumber) = false;
1934 : 29 : QUEUE_BACKEND_IS_ADVANCING(MyProcNumber) = false;
1935 : : /* and remove it from the list */
742 heikki.linnakangas@i 1936 [ + + ]:CBC 29 : if (QUEUE_FIRST_LISTENER == MyProcNumber)
1937 : 24 : QUEUE_FIRST_LISTENER = QUEUE_NEXT_LISTENER(MyProcNumber);
1938 : : else
1939 : : {
1940 [ + - ]: 8 : for (ProcNumber i = QUEUE_FIRST_LISTENER; i != INVALID_PROC_NUMBER; i = QUEUE_NEXT_LISTENER(i))
1941 : : {
1942 [ + + ]: 8 : if (QUEUE_NEXT_LISTENER(i) == MyProcNumber)
1943 : : {
1944 : 5 : QUEUE_NEXT_LISTENER(i) = QUEUE_NEXT_LISTENER(MyProcNumber);
2378 tgl@sss.pgh.pa.us 1945 : 5 : break;
1946 : : }
1947 : : }
1948 : : }
742 heikki.linnakangas@i 1949 : 29 : QUEUE_NEXT_LISTENER(MyProcNumber) = INVALID_PROC_NUMBER;
2130 tgl@sss.pgh.pa.us 1950 : 29 : LWLockRelease(NotifyQueueLock);
1951 : :
1952 : : /* mark ourselves as no longer listed in the global array */
4778 1953 : 29 : amRegisteredListener = false;
1954 : : }
1955 : :
1956 : : /*
1957 : : * Test whether there is room to insert more notification messages.
1958 : : *
1959 : : * Caller must hold at least shared NotifyQueueLock.
1960 : : */
1961 : : static bool
5871 1962 : 90 : asyncQueueIsFull(void)
1963 : : {
600 michael@paquier.xyz 1964 : 90 : int64 headPage = QUEUE_POS_PAGE(QUEUE_HEAD);
1965 : 90 : int64 tailPage = QUEUE_POS_PAGE(QUEUE_TAIL);
1966 : 90 : int64 occupied = headPage - tailPage;
1967 : :
837 akorotkov@postgresql 1968 : 90 : return occupied >= max_notify_queue_pages;
1969 : : }
1970 : :
1971 : : /*
1972 : : * Advance the QueuePosition to the next entry, assuming that the current
1973 : : * entry is of length entryLength. If we jump to a new page the function
1974 : : * returns true, else false.
1975 : : */
1976 : : static bool
4066 tgl@sss.pgh.pa.us 1977 : 2413 : asyncQueueAdvance(volatile QueuePosition *position, int entryLength)
1978 : : {
837 akorotkov@postgresql 1979 : 2413 : int64 pageno = QUEUE_POS_PAGE(*position);
5861 bruce@momjian.us 1980 : 2413 : int offset = QUEUE_POS_OFFSET(*position);
1981 : 2413 : bool pageJump = false;
1982 : :
1983 : : /*
1984 : : * Move to the next writing position: First jump over what we have just
1985 : : * written or read.
1986 : : */
5871 tgl@sss.pgh.pa.us 1987 : 2413 : offset += entryLength;
1988 [ - + ]: 2413 : Assert(offset <= QUEUE_PAGESIZE);
1989 : :
1990 : : /*
1991 : : * In a second step check if another entry can possibly be written to the
1992 : : * page. If so, stay here, we have reached the next position. If not, then
1993 : : * we need to move on to the next page.
1994 : : */
1995 [ + + ]: 2413 : if (offset + QUEUEALIGN(AsyncQueueEntryEmptySize) > QUEUE_PAGESIZE)
1996 : : {
1997 : 38 : pageno++;
1998 : 38 : offset = 0;
1999 : 38 : pageJump = true;
2000 : : }
2001 : :
2002 : 2413 : SET_QUEUE_POS(*position, pageno, offset);
2003 : 2413 : return pageJump;
2004 : : }
2005 : :
2006 : : /*
2007 : : * Fill the AsyncQueueEntry at *qe with an outbound notification message.
2008 : : */
2009 : : static void
2010 : 1132 : asyncQueueNotificationToEntry(Notification *n, AsyncQueueEntry *qe)
2011 : : {
2404 2012 : 1132 : size_t channellen = n->channel_len;
2013 : 1132 : size_t payloadlen = n->payload_len;
2014 : : int entryLength;
2015 : :
5871 2016 [ - + ]: 1132 : Assert(channellen < NAMEDATALEN);
2017 [ - + ]: 1132 : Assert(payloadlen < NOTIFY_PAYLOAD_MAX_LENGTH);
2018 : :
2019 : : /* The terminators are already included in AsyncQueueEntryEmptySize */
2020 : 1132 : entryLength = AsyncQueueEntryEmptySize + payloadlen + channellen;
2021 : 1132 : entryLength = QUEUEALIGN(entryLength);
2022 : 1132 : qe->length = entryLength;
2023 : 1132 : qe->dboid = MyDatabaseId;
2024 : 1132 : qe->xid = GetCurrentTransactionId();
2025 : 1132 : qe->srcPid = MyProcPid;
2404 2026 : 1132 : memcpy(qe->data, n->data, channellen + payloadlen + 2);
5871 2027 : 1132 : }
2028 : :
2029 : : /*
2030 : : * Add pending notifications to the queue.
2031 : : *
2032 : : * We go page by page here, i.e. we stop once we have to go to a new page but
2033 : : * we will be called again and then fill that next page. If an entry does not
2034 : : * fit into the current page, we write a dummy entry with an InvalidOid as the
2035 : : * database OID in order to fill the page. So every page is always used up to
2036 : : * the last byte which simplifies reading the page later.
2037 : : *
2038 : : * We are passed the list cell (in pendingNotifies->events) containing the next
2039 : : * notification to write and return the first still-unwritten cell back.
2040 : : * Eventually we will return NULL indicating all is done.
2041 : : *
2042 : : * We are holding NotifyQueueLock already from the caller and grab
2043 : : * page specific SLRU bank lock locally in this function.
2044 : : */
2045 : : static ListCell *
2046 : 90 : asyncQueueAddEntries(ListCell *nextNotify)
2047 : : {
2048 : : AsyncQueueEntry qe;
2049 : : QueuePosition queue_head;
2050 : : int64 pageno;
2051 : : int offset;
2052 : : int slotno;
2053 : : LWLock *prevlock;
2054 : :
2055 : : /*
2056 : : * We work with a local copy of QUEUE_HEAD, which we write back to shared
2057 : : * memory upon exiting. The reason for this is that if we have to advance
2058 : : * to a new page, SimpleLruZeroPage might fail (out of disk space, for
2059 : : * instance), and we must not advance QUEUE_HEAD if it does. (Otherwise,
2060 : : * subsequent insertions would try to put entries into a page that slru.c
2061 : : * thinks doesn't exist yet.) So, use a local position variable. Note
2062 : : * that if we do fail, any already-inserted queue entries are forgotten;
2063 : : * this is okay, since they'd be useless anyway after our transaction
2064 : : * rolls back.
2065 : : */
5007 2066 : 90 : queue_head = QUEUE_HEAD;
2067 : :
2068 : : /*
2069 : : * If this is the first write since the postmaster started, we need to
2070 : : * initialize the first page of the async SLRU. Otherwise, the current
2071 : : * page should be initialized already, so just fetch it.
2072 : : */
2073 : 90 : pageno = QUEUE_POS_PAGE(queue_head);
746 alvherre@alvh.no-ip. 2074 : 90 : prevlock = SimpleLruGetBankLock(NotifyCtl, pageno);
2075 : :
2076 : : /* We hold both NotifyQueueLock and SLRU bank lock during this operation */
2077 : 90 : LWLockAcquire(prevlock, LW_EXCLUSIVE);
2078 : :
2132 tgl@sss.pgh.pa.us 2079 [ + + + + ]: 90 : if (QUEUE_POS_IS_ZERO(queue_head))
2130 2080 : 7 : slotno = SimpleLruZeroPage(NotifyCtl, pageno);
2081 : : else
2 heikki.linnakangas@i 2082 :GNC 83 : slotno = SimpleLruReadPage(NotifyCtl, pageno, true, &queue_head);
2083 : :
2084 : : /* Note we mark the page dirty before writing in it */
2130 tgl@sss.pgh.pa.us 2085 :CBC 90 : NotifyCtl->shared->page_dirty[slotno] = true;
2086 : :
5871 2087 [ + + ]: 1187 : while (nextNotify != NULL)
2088 : : {
5861 bruce@momjian.us 2089 : 1132 : Notification *n = (Notification *) lfirst(nextNotify);
2090 : :
2091 : : /* Construct a valid queue entry in local variable qe */
5871 tgl@sss.pgh.pa.us 2092 : 1132 : asyncQueueNotificationToEntry(n, &qe);
2093 : :
5007 2094 : 1132 : offset = QUEUE_POS_OFFSET(queue_head);
2095 : :
2096 : : /* Check whether the entry really fits on the current page */
5871 2097 [ + + ]: 1132 : if (offset + qe.length <= QUEUE_PAGESIZE)
2098 : : {
2099 : : /* OK, so advance nextNotify past this item */
2404 2100 : 1100 : nextNotify = lnext(pendingNotifies->events, nextNotify);
2101 : : }
2102 : : else
2103 : : {
2104 : : /*
2105 : : * Write a dummy entry to fill up the page. Actually readers will
2106 : : * only check dboid and since it won't match any reader's database
2107 : : * OID, they will ignore this entry and move on.
2108 : : */
5871 2109 : 32 : qe.length = QUEUE_PAGESIZE - offset;
2110 : 32 : qe.dboid = InvalidOid;
123 heikki.linnakangas@i 2111 : 32 : qe.xid = InvalidTransactionId;
5861 bruce@momjian.us 2112 : 32 : qe.data[0] = '\0'; /* empty channel */
2113 : 32 : qe.data[1] = '\0'; /* empty payload */
2114 : : }
2115 : :
2116 : : /* Now copy qe into the shared buffer page */
2130 tgl@sss.pgh.pa.us 2117 : 1132 : memcpy(NotifyCtl->shared->page_buffer[slotno] + offset,
2118 : : &qe,
5871 2119 : 1132 : qe.length);
2120 : :
2121 : : /* Advance queue_head appropriately, and detect if page is full */
5007 2122 [ + + ]: 1132 : if (asyncQueueAdvance(&(queue_head), qe.length))
2123 : : {
2124 : : LWLock *lock;
2125 : :
746 alvherre@alvh.no-ip. 2126 : 35 : pageno = QUEUE_POS_PAGE(queue_head);
2127 : 35 : lock = SimpleLruGetBankLock(NotifyCtl, pageno);
2128 [ - + ]: 35 : if (lock != prevlock)
2129 : : {
746 alvherre@alvh.no-ip. 2130 :UBC 0 : LWLockRelease(prevlock);
2131 : 0 : LWLockAcquire(lock, LW_EXCLUSIVE);
2132 : 0 : prevlock = lock;
2133 : : }
2134 : :
2135 : : /*
2136 : : * Page is full, so we're done here, but first fill the next page
2137 : : * with zeroes. The reason to do this is to ensure that slru.c's
2138 : : * idea of the head page is always the same as ours, which avoids
2139 : : * boundary problems in SimpleLruTruncate. The test in
2140 : : * asyncQueueIsFull() ensured that there is room to create this
2141 : : * page without overrunning the queue.
2142 : : */
2130 tgl@sss.pgh.pa.us 2143 :CBC 35 : slotno = SimpleLruZeroPage(NotifyCtl, QUEUE_POS_PAGE(queue_head));
2144 : :
2145 : : /*
2146 : : * If the new page address is a multiple of QUEUE_CLEANUP_DELAY,
2147 : : * set flag to remember that we should try to advance the tail
2148 : : * pointer (we don't want to actually do that right here).
2149 : : */
2366 2150 [ + + ]: 35 : if (QUEUE_POS_PAGE(queue_head) % QUEUE_CLEANUP_DELAY == 0)
1643 2151 : 8 : tryAdvanceTail = true;
2152 : :
2153 : : /* And exit the loop */
9042 2154 : 35 : break;
2155 : : }
2156 : : }
2157 : :
2158 : : /* Success, so update the global QUEUE_HEAD */
5007 2159 : 90 : QUEUE_HEAD = queue_head;
2160 : :
746 alvherre@alvh.no-ip. 2161 : 90 : LWLockRelease(prevlock);
2162 : :
5871 tgl@sss.pgh.pa.us 2163 : 90 : return nextNotify;
2164 : : }
2165 : :
2166 : : /*
2167 : : * SQL function to return the fraction of the notification queue currently
2168 : : * occupied.
2169 : : */
2170 : : Datum
3894 rhaas@postgresql.org 2171 : 5 : pg_notification_queue_usage(PG_FUNCTION_ARGS)
2172 : : {
2173 : : double usage;
2174 : :
2175 : : /* Advance the queue tail so we don't report a too-large result */
2303 tgl@sss.pgh.pa.us 2176 : 5 : asyncQueueAdvanceTail();
2177 : :
2130 2178 : 5 : LWLockAcquire(NotifyQueueLock, LW_SHARED);
3894 rhaas@postgresql.org 2179 : 5 : usage = asyncQueueUsage();
2130 tgl@sss.pgh.pa.us 2180 : 5 : LWLockRelease(NotifyQueueLock);
2181 : :
3894 rhaas@postgresql.org 2182 : 5 : PG_RETURN_FLOAT8(usage);
2183 : : }
2184 : :
2185 : : /*
2186 : : * Return the fraction of the queue that is currently occupied.
2187 : : *
2188 : : * The caller must hold NotifyQueueLock in (at least) shared mode.
2189 : : *
2190 : : * Note: we measure the distance to the logical tail page, not the physical
2191 : : * tail page. In some sense that's wrong, but the relative position of the
2192 : : * physical tail is affected by details such as SLRU segment boundaries,
2193 : : * so that a result based on that is unpleasantly unstable.
2194 : : */
2195 : : static double
2196 : 95 : asyncQueueUsage(void)
2197 : : {
600 michael@paquier.xyz 2198 : 95 : int64 headPage = QUEUE_POS_PAGE(QUEUE_HEAD);
2199 : 95 : int64 tailPage = QUEUE_POS_PAGE(QUEUE_TAIL);
2200 : 95 : int64 occupied = headPage - tailPage;
2201 : :
5871 tgl@sss.pgh.pa.us 2202 [ + + ]: 95 : if (occupied == 0)
3894 rhaas@postgresql.org 2203 : 56 : return (double) 0; /* fast exit for common case */
2204 : :
837 akorotkov@postgresql 2205 : 39 : return (double) occupied / (double) max_notify_queue_pages;
2206 : : }
2207 : :
2208 : : /*
2209 : : * Check whether the queue is at least half full, and emit a warning if so.
2210 : : *
2211 : : * This is unlikely given the size of the queue, but possible.
2212 : : * The warnings show up at most once every QUEUE_FULL_WARN_INTERVAL.
2213 : : *
2214 : : * Caller must hold exclusive NotifyQueueLock.
2215 : : */
2216 : : static void
3894 rhaas@postgresql.org 2217 : 90 : asyncQueueFillWarning(void)
2218 : : {
2219 : : double fillDegree;
2220 : : TimestampTz t;
2221 : :
2222 : 90 : fillDegree = asyncQueueUsage();
5871 tgl@sss.pgh.pa.us 2223 [ + - ]: 90 : if (fillDegree < 0.5)
2224 : 90 : return;
2225 : :
5871 tgl@sss.pgh.pa.us 2226 :UBC 0 : t = GetCurrentTimestamp();
2227 : :
2228 [ # # ]: 0 : if (TimestampDifferenceExceeds(asyncQueueControl->lastQueueFillWarn,
2229 : : t, QUEUE_FULL_WARN_INTERVAL))
2230 : : {
5861 bruce@momjian.us 2231 : 0 : QueuePosition min = QUEUE_HEAD;
2232 : 0 : int32 minPid = InvalidPid;
2233 : :
742 heikki.linnakangas@i 2234 [ # # ]: 0 : for (ProcNumber i = QUEUE_FIRST_LISTENER; i != INVALID_PROC_NUMBER; i = QUEUE_NEXT_LISTENER(i))
2235 : : {
2378 tgl@sss.pgh.pa.us 2236 [ # # ]: 0 : Assert(QUEUE_BACKEND_PID(i) != InvalidPid);
2237 [ # # # # : 0 : min = QUEUE_POS_MIN(min, QUEUE_BACKEND_POS(i));
# # ]
2238 [ # # # # ]: 0 : if (QUEUE_POS_EQUAL(min, QUEUE_BACKEND_POS(i)))
2239 : 0 : minPid = QUEUE_BACKEND_PID(i);
2240 : : }
2241 : :
5871 2242 [ # # # # : 0 : ereport(WARNING,
# # ]
2243 : : (errmsg("NOTIFY queue is %.0f%% full", fillDegree * 100),
2244 : : (minPid != InvalidPid ?
2245 : : errdetail("The server process with PID %d is among those with the oldest transactions.", minPid)
2246 : : : 0),
2247 : : (minPid != InvalidPid ?
2248 : : errhint("The NOTIFY queue cannot be emptied until that process ends its current transaction.")
2249 : : : 0)));
2250 : :
2251 : 0 : asyncQueueControl->lastQueueFillWarn = t;
2252 : : }
2253 : : }
2254 : :
2255 : : /*
2256 : : * Send signals to listening backends.
2257 : : *
2258 : : * Normally we signal only backends that are interested in the notifies that
2259 : : * we just sent. However, that will leave idle listeners falling further and
2260 : : * further behind. Waken them anyway if they're far enough behind, so they'll
2261 : : * advance their queue position pointers, allowing the global tail to advance.
2262 : : *
2263 : : * Since we know the ProcNumber and the Pid the signaling is quite cheap.
2264 : : *
2265 : : * This is called during CommitTransaction(), so it's important for it
2266 : : * to have very low probability of failure.
2267 : : */
2268 : : static void
5871 tgl@sss.pgh.pa.us 2269 :CBC 55 : SignalBackends(void)
2270 : : {
2271 : : int count;
2272 : :
2273 : : /* Can't get here without PreCommit_Notify having made the global table */
59 tgl@sss.pgh.pa.us 2274 [ - + ]:GNC 55 : Assert(globalChannelTable != NULL);
2275 : :
2276 : : /* It should have set up these arrays, too */
2277 [ + - - + ]: 55 : Assert(signalPids != NULL && signalProcnos != NULL);
2278 : :
2279 : : /*
2280 : : * Identify backends that we need to signal. We don't want to send
2281 : : * signals while holding the NotifyQueueLock, so this part just builds a
2282 : : * list of target PIDs in signalPids[] and signalProcnos[].
2283 : : */
5871 tgl@sss.pgh.pa.us 2284 :CBC 55 : count = 0;
2285 : :
2130 2286 : 55 : LWLockAcquire(NotifyQueueLock, LW_EXCLUSIVE);
2287 : :
2288 : : /* Scan each channel name that we notified in this transaction */
59 tgl@sss.pgh.pa.us 2289 [ + - + + :GNC 168 : foreach_ptr(char, channel, pendingNotifies->uniqueChannelNames)
+ + ]
2290 : : {
2291 : : GlobalChannelKey key;
2292 : : GlobalChannelEntry *entry;
2293 : : ListenerEntry *listeners;
2294 : :
2295 : 58 : GlobalChannelKeyInit(&key, MyDatabaseId, channel);
2296 : 58 : entry = dshash_find(globalChannelTable, &key, false);
2297 [ + + ]: 58 : if (entry == NULL)
2298 : 28 : continue; /* nobody is listening */
2299 : :
2300 : 30 : listeners = (ListenerEntry *) dsa_get_address(globalChannelDSA,
2301 : : entry->listenersArray);
2302 : :
2303 : : /* Identify listeners that now need waking, add them to arrays */
2304 [ + + ]: 65 : for (int j = 0; j < entry->numListeners; j++)
2305 : : {
2306 : : ProcNumber i;
2307 : : int32 pid;
2308 : : QueuePosition pos;
2309 : :
2310 [ - + ]: 35 : if (!listeners[j].listening)
2311 : 3 : continue; /* ignore not-yet-committed listeners */
2312 : :
2313 : 35 : i = listeners[j].procNo;
2314 : :
2315 [ + + ]: 35 : if (QUEUE_BACKEND_WAKEUP_PENDING(i))
2316 : 3 : continue; /* already signaled, no need to repeat */
2317 : :
2318 : 32 : pid = QUEUE_BACKEND_PID(i);
2319 : 32 : pos = QUEUE_BACKEND_POS(i);
2320 : :
2321 [ + + - + ]: 32 : if (QUEUE_POS_EQUAL(pos, QUEUE_HEAD))
59 tgl@sss.pgh.pa.us 2322 :UNC 0 : continue; /* it's fully caught up already */
2323 : :
59 tgl@sss.pgh.pa.us 2324 [ - + ]:GNC 32 : Assert(pid != InvalidPid);
2325 : :
2326 : 32 : QUEUE_BACKEND_WAKEUP_PENDING(i) = true;
2327 : 32 : signalPids[count] = pid;
2328 : 32 : signalProcnos[count] = i;
2329 : 32 : count++;
2330 : : }
2331 : :
2332 : 30 : dshash_release_lock(globalChannelTable, entry);
2333 : : }
2334 : :
2335 : : /*
2336 : : * Scan all listeners. Any that are not already pending wakeup must not
2337 : : * be interested in our notifications (else we'd have set their wakeup
2338 : : * flags above). Check to see if we can directly advance their queue
2339 : : * pointers to save a wakeup. Otherwise, if they are far behind, wake
2340 : : * them anyway so they will catch up.
2341 : : */
742 heikki.linnakangas@i 2342 [ + + ]:CBC 108 : for (ProcNumber i = QUEUE_FIRST_LISTENER; i != INVALID_PROC_NUMBER; i = QUEUE_NEXT_LISTENER(i))
2343 : : {
2344 : : int32 pid;
2345 : : QueuePosition pos;
2346 : :
59 tgl@sss.pgh.pa.us 2347 [ + + ]:GNC 53 : if (QUEUE_BACKEND_WAKEUP_PENDING(i))
2348 : 32 : continue;
2349 : :
2350 : : /* If it's currently advancing, we should not touch it */
2351 [ - + ]: 21 : if (QUEUE_BACKEND_IS_ADVANCING(i))
59 tgl@sss.pgh.pa.us 2352 :UNC 0 : continue;
2353 : :
59 tgl@sss.pgh.pa.us 2354 :GNC 21 : pid = QUEUE_BACKEND_PID(i);
2366 tgl@sss.pgh.pa.us 2355 :CBC 21 : pos = QUEUE_BACKEND_POS(i);
2356 : :
2357 : : /*
2358 : : * We can directly advance the other backend's queue pointer if it's
2359 : : * not currently advancing (else there are race conditions), and its
2360 : : * current pointer is not behind queueHeadBeforeWrite (else we'd make
2361 : : * it miss some older messages), and we'd not be moving the pointer
2362 : : * backward.
2363 : : */
59 tgl@sss.pgh.pa.us 2364 [ + - + - :GNC 42 : if (!QUEUE_POS_PRECEDES(pos, queueHeadBeforeWrite) &&
+ - + + ]
2365 [ + - + - ]: 31 : QUEUE_POS_PRECEDES(pos, queueHeadAfterWrite))
2366 : : {
2367 : : /* We can directly advance its pointer past what we wrote */
2368 : 21 : QUEUE_BACKEND_POS(i) = queueHeadAfterWrite;
2369 : : }
59 tgl@sss.pgh.pa.us 2370 [ # # ]:UNC 0 : else if (asyncQueuePageDiff(QUEUE_POS_PAGE(QUEUE_HEAD),
2371 : : QUEUE_POS_PAGE(pos)) >= QUEUE_CLEANUP_DELAY)
2372 : : {
2373 : : /* It's idle and far behind, so wake it up */
2374 [ # # ]: 0 : Assert(pid != InvalidPid);
2375 : :
2376 : 0 : QUEUE_BACKEND_WAKEUP_PENDING(i) = true;
2377 : 0 : signalPids[count] = pid;
2378 : 0 : signalProcnos[count] = i;
2379 : 0 : count++;
2380 : : }
2381 : : }
2382 : :
2130 tgl@sss.pgh.pa.us 2383 :CBC 55 : LWLockRelease(NotifyQueueLock);
2384 : :
2385 : : /* Now send signals */
2378 2386 [ + + ]: 87 : for (int i = 0; i < count; i++)
2387 : : {
59 tgl@sss.pgh.pa.us 2388 :GNC 32 : int32 pid = signalPids[i];
2389 : :
2390 : : /*
2391 : : * If we are signaling our own process, no need to involve the kernel;
2392 : : * just set the flag directly.
2393 : : */
1643 tgl@sss.pgh.pa.us 2394 [ + + ]:CBC 32 : if (pid == MyProcPid)
2395 : : {
2396 : 21 : notifyInterruptPending = true;
2397 : 21 : continue;
2398 : : }
2399 : :
2400 : : /*
2401 : : * Note: assuming things aren't broken, a signal failure here could
2402 : : * only occur if the target backend exited since we released
2403 : : * NotifyQueueLock; which is unlikely but certainly possible. So we
2404 : : * just log a low-level debug message if it happens.
2405 : : */
59 tgl@sss.pgh.pa.us 2406 [ - + ]:GNC 11 : if (SendProcSignal(pid, PROCSIG_NOTIFY_INTERRUPT, signalProcnos[i]) < 0)
5871 tgl@sss.pgh.pa.us 2407 [ # # ]:UBC 0 : elog(DEBUG3, "could not signal backend with PID %d: %m", pid);
2408 : : }
10022 tgl@sss.pgh.pa.us 2409 :CBC 55 : }
2410 : :
2411 : : /*
2412 : : * AtAbort_Notify
2413 : : *
2414 : : * This is called at transaction abort.
2415 : : *
2416 : : * Revert any staged listen/unlisten changes and clean up transaction state.
2417 : : * This only does anything if we abort after PreCommit_Notify has staged
2418 : : * some entries.
2419 : : */
2420 : : void
9037 2421 : 26742 : AtAbort_Notify(void)
2422 : : {
2423 : : /* Revert staged listen/unlisten changes */
59 tgl@sss.pgh.pa.us 2424 :GNC 26742 : ApplyPendingListenActions(false);
2425 : :
2426 : : /* If we're no longer listening on anything, unregister */
2427 [ - + - - : 26742 : if (amRegisteredListener && LocalChannelTableIsEmpty())
- - ]
4778 tgl@sss.pgh.pa.us 2428 :UBC 0 : asyncQueueUnregister();
2429 : :
2430 : : /* And clean up */
6577 tgl@sss.pgh.pa.us 2431 :CBC 26742 : ClearPendingActionsAndNotifies();
10022 2432 : 26742 : }
2433 : :
2434 : : /*
2435 : : * AtSubCommit_Notify() --- Take care of subtransaction commit.
2436 : : *
2437 : : * Reassign all items in the pending lists to the parent transaction.
2438 : : */
2439 : : void
7927 2440 : 6982 : AtSubCommit_Notify(void)
2441 : : {
2354 rhaas@postgresql.org 2442 : 6982 : int my_level = GetCurrentTransactionNestLevel();
2443 : :
2444 : : /* If there are actions at our nesting level, we must reparent them. */
2445 [ + + ]: 6982 : if (pendingActions != NULL &&
2354 rhaas@postgresql.org 2446 [ + - ]:GBC 2 : pendingActions->nestingLevel >= my_level)
2447 : : {
2448 [ + + ]: 2 : if (pendingActions->upper == NULL ||
2449 [ - + ]: 1 : pendingActions->upper->nestingLevel < my_level - 1)
2450 : : {
2451 : : /* nothing to merge; give the whole thing to the parent */
2452 : 1 : --pendingActions->nestingLevel;
2453 : : }
2454 : : else
2455 : : {
2456 : 1 : ActionList *childPendingActions = pendingActions;
2457 : :
2458 : 1 : pendingActions = pendingActions->upper;
2459 : :
2460 : : /*
2461 : : * Mustn't try to eliminate duplicates here --- see queue_listen()
2462 : : */
2463 : 2 : pendingActions->actions =
2464 : 1 : list_concat(pendingActions->actions,
2465 : 1 : childPendingActions->actions);
2466 : 1 : pfree(childPendingActions);
2467 : : }
2468 : : }
2469 : :
2470 : : /* If there are notifies at our nesting level, we must reparent them. */
2354 rhaas@postgresql.org 2471 [ + + ]:CBC 6982 : if (pendingNotifies != NULL &&
2472 [ + + ]: 3 : pendingNotifies->nestingLevel >= my_level)
2473 : : {
2474 [ - + ]: 2 : Assert(pendingNotifies->nestingLevel == my_level);
2475 : :
2476 [ + + ]: 2 : if (pendingNotifies->upper == NULL ||
2477 [ - + ]: 1 : pendingNotifies->upper->nestingLevel < my_level - 1)
2478 : : {
2479 : : /* nothing to merge; give the whole thing to the parent */
2354 rhaas@postgresql.org 2480 :GBC 1 : --pendingNotifies->nestingLevel;
2481 : : }
2482 : : else
2483 : : {
2484 : : /*
2485 : : * Formerly, we didn't bother to eliminate duplicates here, but
2486 : : * now we must, else we fall foul of "Assert(!found)", either here
2487 : : * or during a later attempt to build the parent-level hashtable.
2488 : : */
2354 rhaas@postgresql.org 2489 :CBC 1 : NotificationList *childPendingNotifies = pendingNotifies;
2490 : : ListCell *l;
2491 : :
2492 : 1 : pendingNotifies = pendingNotifies->upper;
2493 : : /* Insert all the subxact's events into parent, except for dups */
2494 [ + - + + : 5 : foreach(l, childPendingNotifies->events)
+ + ]
2495 : : {
2496 : 4 : Notification *childn = (Notification *) lfirst(l);
2497 : :
2498 [ + + ]: 4 : if (!AsyncExistsPendingNotify(childn))
2499 : 2 : AddEventToPendingNotifies(childn);
2500 : : }
2501 : 1 : pfree(childPendingNotifies);
2502 : : }
2503 : : }
7927 tgl@sss.pgh.pa.us 2504 : 6982 : }
2505 : :
2506 : : /*
2507 : : * AtSubAbort_Notify() --- Take care of subtransaction abort.
2508 : : */
2509 : : void
2510 : 4703 : AtSubAbort_Notify(void)
2511 : : {
7860 2512 : 4703 : int my_level = GetCurrentTransactionNestLevel();
2513 : :
2514 : : /*
2515 : : * All we have to do is pop the stack --- the actions/notifies made in
2516 : : * this subxact are no longer interesting, and the space will be freed
2517 : : * when CurTransactionContext is recycled. We still have to free the
2518 : : * ActionList and NotificationList objects themselves, though, because
2519 : : * those are allocated in TopTransactionContext.
2520 : : *
2521 : : * Note that there might be no entries at all, or no entries for the
2522 : : * current subtransaction level, either because none were ever created, or
2523 : : * because we reentered this routine due to trouble during subxact abort.
2524 : : */
2354 rhaas@postgresql.org 2525 [ + + ]: 4704 : while (pendingActions != NULL &&
2354 rhaas@postgresql.org 2526 [ + - ]:GBC 1 : pendingActions->nestingLevel >= my_level)
2527 : : {
2528 : 1 : ActionList *childPendingActions = pendingActions;
2529 : :
2530 : 1 : pendingActions = pendingActions->upper;
2531 : 1 : pfree(childPendingActions);
2532 : : }
2533 : :
2354 rhaas@postgresql.org 2534 [ + + ]:CBC 4704 : while (pendingNotifies != NULL &&
2535 [ + + ]: 2 : pendingNotifies->nestingLevel >= my_level)
2536 : : {
2537 : 1 : NotificationList *childPendingNotifies = pendingNotifies;
2538 : :
2539 : 1 : pendingNotifies = pendingNotifies->upper;
2540 : 1 : pfree(childPendingNotifies);
2541 : : }
7927 tgl@sss.pgh.pa.us 2542 : 4703 : }
2543 : :
2544 : : /*
2545 : : * HandleNotifyInterrupt
2546 : : *
2547 : : * Signal handler portion of interrupt handling. Let the backend know
2548 : : * that there's a pending notify interrupt. If we're currently reading
2549 : : * from the client, this will interrupt the read and
2550 : : * ProcessClientReadInterrupt() will call ProcessNotifyInterrupt().
2551 : : */
2552 : : void
6071 2553 : 11 : HandleNotifyInterrupt(void)
2554 : : {
2555 : : /*
2556 : : * Note: this is called by a SIGNAL HANDLER. You must be very wary what
2557 : : * you do here.
2558 : : */
2559 : :
2560 : : /* signal that work needs to be done */
4058 andres@anarazel.de 2561 : 11 : notifyInterruptPending = true;
2562 : :
2563 : : /* make sure the event is processed in due course */
2564 : 11 : SetLatch(MyLatch);
10022 tgl@sss.pgh.pa.us 2565 : 11 : }
2566 : :
2567 : : /*
2568 : : * ProcessNotifyInterrupt
2569 : : *
2570 : : * This is called if we see notifyInterruptPending set, just before
2571 : : * transmitting ReadyForQuery at the end of a frontend command, and
2572 : : * also if a notify signal occurs while reading from the frontend.
2573 : : * HandleNotifyInterrupt() will cause the read to be interrupted
2574 : : * via the process's latch, and this routine will get called.
2575 : : * If we are truly idle (ie, *not* inside a transaction block),
2576 : : * process the incoming notifies.
2577 : : *
2578 : : * If "flush" is true, force any frontend messages out immediately.
2579 : : * This can be false when being called at the end of a frontend command,
2580 : : * since we'll flush after sending ReadyForQuery.
2581 : : */
2582 : : void
1643 2583 : 38 : ProcessNotifyInterrupt(bool flush)
2584 : : {
8186 2585 [ + + ]: 38 : if (IsTransactionOrTransactionBlock())
10022 2586 : 7 : return; /* not really idle */
2587 : :
2588 : : /* Loop in case another signal arrives while sending messages */
4058 andres@anarazel.de 2589 [ + + ]: 62 : while (notifyInterruptPending)
1643 tgl@sss.pgh.pa.us 2590 : 31 : ProcessIncomingNotify(flush);
2591 : : }
2592 : :
2593 : :
2594 : : /*
2595 : : * Read all pending notifications from the queue, and deliver appropriate
2596 : : * ones to my frontend. Stop when we reach queue head or an uncommitted
2597 : : * notification.
2598 : : */
2599 : : static void
5871 2600 : 46 : asyncQueueReadAllNotifications(void)
2601 : : {
2602 : : QueuePosition pos;
2603 : : QueuePosition head;
2604 : : Snapshot snapshot;
2605 : :
2606 : : /*
2607 : : * Fetch current state, indicate to others that we have woken up, and that
2608 : : * we are in process of advancing our position.
2609 : : */
2130 2610 : 46 : LWLockAcquire(NotifyQueueLock, LW_SHARED);
2611 : : /* Assert checks that we have a valid state entry */
742 heikki.linnakangas@i 2612 [ - + ]: 46 : Assert(MyProcPid == QUEUE_BACKEND_PID(MyProcNumber));
59 tgl@sss.pgh.pa.us 2613 :GNC 46 : QUEUE_BACKEND_WAKEUP_PENDING(MyProcNumber) = false;
742 heikki.linnakangas@i 2614 :CBC 46 : pos = QUEUE_BACKEND_POS(MyProcNumber);
5871 tgl@sss.pgh.pa.us 2615 : 46 : head = QUEUE_HEAD;
2616 : :
2617 [ + + - + ]: 46 : if (QUEUE_POS_EQUAL(pos, head))
2618 : : {
2619 : : /* Nothing to do, we have read all notifications already. */
59 tgl@sss.pgh.pa.us 2620 :UNC 0 : LWLockRelease(NotifyQueueLock);
5871 tgl@sss.pgh.pa.us 2621 :UBC 0 : return;
2622 : : }
2623 : :
59 tgl@sss.pgh.pa.us 2624 :GNC 46 : QUEUE_BACKEND_IS_ADVANCING(MyProcNumber) = true;
2625 : 46 : LWLockRelease(NotifyQueueLock);
2626 : :
2627 : : /*----------
2628 : : * Get snapshot we'll use to decide which xacts are still in progress.
2629 : : * This is trickier than it might seem, because of race conditions.
2630 : : * Consider the following example:
2631 : : *
2632 : : * Backend 1: Backend 2:
2633 : : *
2634 : : * transaction starts
2635 : : * UPDATE foo SET ...;
2636 : : * NOTIFY foo;
2637 : : * commit starts
2638 : : * queue the notify message
2639 : : * transaction starts
2640 : : * LISTEN foo; -- first LISTEN in session
2641 : : * SELECT * FROM foo WHERE ...;
2642 : : * commit to clog
2643 : : * commit starts
2644 : : * add backend 2 to array of listeners
2645 : : * advance to queue head (this code)
2646 : : * commit to clog
2647 : : *
2648 : : * Transaction 2's SELECT has not seen the UPDATE's effects, since that
2649 : : * wasn't committed yet. Ideally we'd ensure that client 2 would
2650 : : * eventually get transaction 1's notify message, but there's no way
2651 : : * to do that; until we're in the listener array, there's no guarantee
2652 : : * that the notify message doesn't get removed from the queue.
2653 : : *
2654 : : * Therefore the coding technique transaction 2 is using is unsafe:
2655 : : * applications must commit a LISTEN before inspecting database state,
2656 : : * if they want to ensure they will see notifications about subsequent
2657 : : * changes to that state.
2658 : : *
2659 : : * What we do guarantee is that we'll see all notifications from
2660 : : * transactions committing after the snapshot we take here.
2661 : : * BecomeRegisteredListener has already added us to the listener array,
2662 : : * so no not-yet-committed messages can be removed from the queue
2663 : : * before we see them.
2664 : : *----------
2665 : : */
2303 tgl@sss.pgh.pa.us 2666 :CBC 46 : snapshot = RegisterSnapshot(GetLatestSnapshot());
2667 : :
2668 : : /*
2669 : : * It is possible that we fail while trying to send a message to our
2670 : : * frontend (for example, because of encoding conversion failure). If
2671 : : * that happens it is critical that we not try to send the same message
2672 : : * over and over again. Therefore, we set ExitOnAnyError to upgrade any
2673 : : * ERRORs to FATAL, causing the client connection to be closed on error.
2674 : : *
2675 : : * We used to only skip over the offending message and try to soldier on,
2676 : : * but it was somewhat questionable to lose a notification and give the
2677 : : * client an ERROR instead. A client application is not be prepared for
2678 : : * that and can't tell that a notification was missed. It was also not
2679 : : * very useful in practice because notifications are often processed while
2680 : : * a connection is idle and reading a message from the client, and in that
2681 : : * state, any error is upgraded to FATAL anyway. Closing the connection
2682 : : * is a clear signal to the application that it might have missed
2683 : : * notifications.
2684 : : */
2685 : : {
123 heikki.linnakangas@i 2686 : 46 : bool save_ExitOnAnyError = ExitOnAnyError;
2687 : : bool reachedStop;
2688 : :
2689 : 46 : ExitOnAnyError = true;
2690 : :
2691 : : do
2692 : : {
2693 : : /*
2694 : : * Process messages up to the stop position, end of page, or an
2695 : : * uncommitted message.
2696 : : *
2697 : : * Our stop position is what we found to be the head's position
2698 : : * when we entered this function. It might have changed already.
2699 : : * But if it has, we will receive (or have already received and
2700 : : * queued) another signal and come here again.
2701 : : *
2702 : : * We are not holding NotifyQueueLock here! The queue can only
2703 : : * extend beyond the head pointer (see above) and we leave our
2704 : : * backend's pointer where it is so nobody will truncate or
2705 : : * rewrite pages under us. Especially we don't want to hold a lock
2706 : : * while sending the notifications to the frontend.
2707 : : */
2708 : 49 : reachedStop = asyncQueueProcessPageEntries(&pos, head, snapshot);
5871 tgl@sss.pgh.pa.us 2709 [ + + ]: 49 : } while (!reachedStop);
2710 : :
2711 : : /* Update shared state */
2130 2712 : 46 : LWLockAcquire(NotifyQueueLock, LW_SHARED);
742 heikki.linnakangas@i 2713 : 46 : QUEUE_BACKEND_POS(MyProcNumber) = pos;
59 tgl@sss.pgh.pa.us 2714 :GNC 46 : QUEUE_BACKEND_IS_ADVANCING(MyProcNumber) = false;
2130 tgl@sss.pgh.pa.us 2715 :CBC 46 : LWLockRelease(NotifyQueueLock);
2716 : :
123 heikki.linnakangas@i 2717 : 46 : ExitOnAnyError = save_ExitOnAnyError;
2718 : : }
2719 : :
2720 : : /* Done with snapshot */
3077 tgl@sss.pgh.pa.us 2721 : 46 : UnregisterSnapshot(snapshot);
2722 : : }
2723 : :
2724 : : /*
2725 : : * Fetch notifications from the shared queue, beginning at position current,
2726 : : * and deliver relevant ones to my frontend.
2727 : : *
2728 : : * The function returns true once we have reached the stop position or an
2729 : : * uncommitted notification, and false if we have finished with the page.
2730 : : * In other words: once it returns true there is no need to look further.
2731 : : * The QueuePosition *current is advanced past all processed messages.
2732 : : */
2733 : : static bool
123 heikki.linnakangas@i 2734 : 49 : asyncQueueProcessPageEntries(QueuePosition *current,
2735 : : QueuePosition stop,
2736 : : Snapshot snapshot)
2737 : : {
2738 : 49 : int64 curpage = QUEUE_POS_PAGE(*current);
2739 : : int slotno;
2740 : : char *page_buffer;
5871 tgl@sss.pgh.pa.us 2741 : 49 : bool reachedStop = false;
2742 : : bool reachedEndOfPage;
2743 : :
2744 : : /*
2745 : : * We copy the entries into a local buffer to avoid holding the SLRU lock
2746 : : * while we transmit them to our frontend. The local buffer must be
2747 : : * adequately aligned.
2748 : : */
2749 : : alignas(AsyncQueueEntry) char local_buf[QUEUE_PAGESIZE];
114 peter@eisentraut.org 2750 :GNC 49 : char *local_buf_end = local_buf;
2751 : :
2 heikki.linnakangas@i 2752 : 49 : slotno = SimpleLruReadPage_ReadOnly(NotifyCtl, curpage, current);
123 heikki.linnakangas@i 2753 :CBC 49 : page_buffer = NotifyCtl->shared->page_buffer[slotno];
2754 : :
2755 : : do
2756 : : {
5861 bruce@momjian.us 2757 : 1327 : QueuePosition thisentry = *current;
2758 : : AsyncQueueEntry *qe;
2759 : :
5870 tgl@sss.pgh.pa.us 2760 [ + + + + ]: 1327 : if (QUEUE_POS_EQUAL(thisentry, stop))
5871 2761 : 46 : break;
2762 : :
5870 2763 : 1281 : qe = (AsyncQueueEntry *) (page_buffer + QUEUE_POS_OFFSET(thisentry));
2764 : :
2765 : : /*
2766 : : * Advance *current over this message, possibly to the next page. As
2767 : : * noted in the comments for asyncQueueReadAllNotifications, we must
2768 : : * do this before possibly failing while processing the message.
2769 : : */
5871 2770 : 1281 : reachedEndOfPage = asyncQueueAdvance(current, qe->length);
2771 : :
2772 : : /* Ignore messages destined for other databases */
2773 [ + - ]: 1281 : if (qe->dboid == MyDatabaseId)
2774 : : {
3077 2775 [ - + ]: 1281 : if (XidInMVCCSnapshot(qe->xid, snapshot))
2776 : : {
2777 : : /*
2778 : : * The source transaction is still in progress, so we can't
2779 : : * process this message yet. Break out of the loop, but first
2780 : : * back up *current so we will reprocess the message next
2781 : : * time. (Note: it is unlikely but not impossible for
2782 : : * TransactionIdDidCommit to fail, so we can't really avoid
2783 : : * this advance-then-back-up behavior when dealing with an
2784 : : * uncommitted message.)
2785 : : *
2786 : : * Note that we must test XidInMVCCSnapshot before we test
2787 : : * TransactionIdDidCommit, else we might return a message from
2788 : : * a transaction that is not yet visible to snapshots; compare
2789 : : * the comments at the head of heapam_visibility.c.
2790 : : *
2791 : : * Also, while our own xact won't be listed in the snapshot,
2792 : : * we need not check for TransactionIdIsCurrentTransactionId
2793 : : * because our transaction cannot (yet) have queued any
2794 : : * messages.
2795 : : */
4385 tgl@sss.pgh.pa.us 2796 :UBC 0 : *current = thisentry;
2797 : 0 : reachedStop = true;
2798 : 0 : break;
2799 : : }
2800 : :
2801 : : /*
2802 : : * Quick check for the case that we're not listening on any
2803 : : * channels, before calling TransactionIdDidCommit(). This makes
2804 : : * that case a little faster, but more importantly, it ensures
2805 : : * that if there's a bad entry in the queue for which
2806 : : * TransactionIdDidCommit() fails for some reason, we can skip
2807 : : * over it on the first LISTEN in a session, and not get stuck on
2808 : : * it indefinitely. (This is a little trickier than it looks: it
2809 : : * works because BecomeRegisteredListener runs this code before we
2810 : : * have made the first entry in localChannelTable.)
2811 : : */
59 tgl@sss.pgh.pa.us 2812 [ + - + + ]:GNC 1281 : if (LocalChannelTableIsEmpty())
123 heikki.linnakangas@i 2813 :CBC 1226 : continue;
2814 : :
2815 [ + - ]: 55 : if (TransactionIdDidCommit(qe->xid))
2816 : : {
2817 : 55 : memcpy(local_buf_end, qe, qe->length);
2818 : 55 : local_buf_end += qe->length;
2819 : : }
2820 : : else
2821 : : {
2822 : : /*
2823 : : * The source transaction aborted or crashed, so we just
2824 : : * ignore its notifications.
2825 : : */
2826 : : }
2827 : : }
2828 : :
2829 : : /* Loop back if we're not at end of page */
5871 tgl@sss.pgh.pa.us 2830 [ + + ]: 1281 : } while (!reachedEndOfPage);
2831 : :
2832 : : /* Release lock that we got from SimpleLruReadPage_ReadOnly() */
123 heikki.linnakangas@i 2833 : 49 : LWLockRelease(SimpleLruGetBankLock(NotifyCtl, curpage));
2834 : :
2835 : : /*
2836 : : * Now that we have let go of the SLRU bank lock, send the notifications
2837 : : * to our backend
2838 : : */
114 peter@eisentraut.org 2839 [ - + ]:GNC 49 : Assert(local_buf_end - local_buf <= BLCKSZ);
2840 [ + + ]: 104 : for (char *p = local_buf; p < local_buf_end;)
2841 : : {
123 heikki.linnakangas@i 2842 :CBC 55 : AsyncQueueEntry *qe = (AsyncQueueEntry *) p;
2843 : :
2844 : : /* qe->data is the null-terminated channel name */
2845 : 55 : char *channel = qe->data;
2846 : :
2847 [ + - ]: 55 : if (IsListeningOn(channel))
2848 : : {
2849 : : /* payload follows channel name */
2850 : 55 : char *payload = qe->data + strlen(channel) + 1;
2851 : :
2852 : 55 : NotifyMyFrontEnd(channel, payload, qe->srcPid);
2853 : : }
2854 : :
2855 : 55 : p += qe->length;
2856 : : }
2857 : :
5871 tgl@sss.pgh.pa.us 2858 [ + + + + ]: 49 : if (QUEUE_POS_EQUAL(*current, stop))
2859 : 46 : reachedStop = true;
2860 : :
2861 : 49 : return reachedStop;
2862 : : }
2863 : :
2864 : : /*
2865 : : * Advance the shared queue tail variable to the minimum of all the
2866 : : * per-backend tail pointers. Truncate pg_notify space if possible.
2867 : : *
2868 : : * This is (usually) called during CommitTransaction(), so it's important for
2869 : : * it to have very low probability of failure.
2870 : : */
2871 : : static void
2872 : 13 : asyncQueueAdvanceTail(void)
2873 : : {
2874 : : QueuePosition min;
2875 : : int64 oldtailpage;
2876 : : int64 newtailpage;
2877 : : int64 boundary;
2878 : :
2879 : : /* Restrict task to one backend per cluster; see SimpleLruTruncate(). */
2038 noah@leadboat.com 2880 : 13 : LWLockAcquire(NotifyQueueTailLock, LW_EXCLUSIVE);
2881 : :
2882 : : /*
2883 : : * Compute the new tail. Pre-v13, it's essential that QUEUE_TAIL be exact
2884 : : * (ie, exactly match at least one backend's queue position), so it must
2885 : : * be updated atomically with the actual computation. Since v13, we could
2886 : : * get away with not doing it like that, but it seems prudent to keep it
2887 : : * so.
2888 : : *
2889 : : * Also, because incoming backends will scan forward from QUEUE_TAIL, that
2890 : : * must be advanced before we can truncate any data. Thus, QUEUE_TAIL is
2891 : : * the logical tail, while QUEUE_STOP_PAGE is the physical tail, or oldest
2892 : : * un-truncated page. When QUEUE_STOP_PAGE != QUEUE_POS_PAGE(QUEUE_TAIL),
2893 : : * there are pages we can truncate but haven't yet finished doing so.
2894 : : *
2895 : : * For concurrency's sake, we don't want to hold NotifyQueueLock while
2896 : : * performing SimpleLruTruncate. This is OK because no backend will try
2897 : : * to access the pages we are in the midst of truncating.
2898 : : */
2130 tgl@sss.pgh.pa.us 2899 : 13 : LWLockAcquire(NotifyQueueLock, LW_EXCLUSIVE);
5871 2900 : 13 : min = QUEUE_HEAD;
742 heikki.linnakangas@i 2901 [ + + ]: 23 : for (ProcNumber i = QUEUE_FIRST_LISTENER; i != INVALID_PROC_NUMBER; i = QUEUE_NEXT_LISTENER(i))
2902 : : {
2378 tgl@sss.pgh.pa.us 2903 [ - + ]: 10 : Assert(QUEUE_BACKEND_PID(i) != InvalidPid);
2904 [ + - + + : 10 : min = QUEUE_POS_MIN(min, QUEUE_BACKEND_POS(i));
+ - ]
2905 : : }
1933 2906 : 13 : QUEUE_TAIL = min;
2907 : 13 : oldtailpage = QUEUE_STOP_PAGE;
2130 2908 : 13 : LWLockRelease(NotifyQueueLock);
2909 : :
2910 : : /*
2911 : : * We can truncate something if the global tail advanced across an SLRU
2912 : : * segment boundary.
2913 : : *
2914 : : * XXX it might be better to truncate only once every several segments, to
2915 : : * reduce the number of directory scans.
2916 : : */
5871 2917 : 13 : newtailpage = QUEUE_POS_PAGE(min);
2918 : 13 : boundary = newtailpage - (newtailpage % SLRU_PAGES_PER_SEGMENT);
5282 alvherre@alvh.no-ip. 2919 [ + + ]: 13 : if (asyncQueuePagePrecedes(oldtailpage, boundary))
2920 : : {
2921 : : /*
2922 : : * SimpleLruTruncate() will ask for SLRU bank locks but will also
2923 : : * release the lock again.
2924 : : */
2130 tgl@sss.pgh.pa.us 2925 :GBC 1 : SimpleLruTruncate(NotifyCtl, newtailpage);
2926 : :
1933 2927 : 1 : LWLockAcquire(NotifyQueueLock, LW_EXCLUSIVE);
2928 : 1 : QUEUE_STOP_PAGE = newtailpage;
2929 : 1 : LWLockRelease(NotifyQueueLock);
2930 : : }
2931 : :
2038 noah@leadboat.com 2932 :CBC 13 : LWLockRelease(NotifyQueueTailLock);
5871 tgl@sss.pgh.pa.us 2933 : 13 : }
2934 : :
2935 : : /*
2936 : : * AsyncNotifyFreezeXids
2937 : : *
2938 : : * Prepare the async notification queue for CLOG truncation by freezing
2939 : : * transaction IDs that are about to become inaccessible.
2940 : : *
2941 : : * This function is called by VACUUM before advancing datfrozenxid. It scans
2942 : : * the notification queue and replaces XIDs that would become inaccessible
2943 : : * after CLOG truncation with special markers:
2944 : : * - Committed transactions are set to FrozenTransactionId
2945 : : * - Aborted/crashed transactions are set to InvalidTransactionId
2946 : : *
2947 : : * Only XIDs < newFrozenXid are processed, as those are the ones whose CLOG
2948 : : * pages will be truncated. If XID < newFrozenXid, it cannot still be running
2949 : : * (or it would have held back newFrozenXid through ProcArray).
2950 : : * Therefore, if TransactionIdDidCommit returns false, we know the transaction
2951 : : * either aborted explicitly or crashed, and we can safely mark it invalid.
2952 : : */
2953 : : void
123 heikki.linnakangas@i 2954 : 103 : AsyncNotifyFreezeXids(TransactionId newFrozenXid)
2955 : : {
2956 : : QueuePosition pos;
2957 : : QueuePosition head;
2958 : 103 : int64 curpage = -1;
2959 : 103 : int slotno = -1;
2960 : 103 : char *page_buffer = NULL;
2961 : 103 : bool page_dirty = false;
2962 : :
2963 : : /*
2964 : : * Acquire locks in the correct order to avoid deadlocks. As per the
2965 : : * locking protocol: NotifyQueueTailLock, then NotifyQueueLock, then SLRU
2966 : : * bank locks.
2967 : : *
2968 : : * We only need SHARED mode since we're just reading the head/tail
2969 : : * positions, not modifying them.
2970 : : */
2971 : 103 : LWLockAcquire(NotifyQueueTailLock, LW_SHARED);
2972 : 103 : LWLockAcquire(NotifyQueueLock, LW_SHARED);
2973 : :
2974 : 103 : pos = QUEUE_TAIL;
2975 : 103 : head = QUEUE_HEAD;
2976 : :
2977 : : /* Release NotifyQueueLock early, we only needed to read the positions */
2978 : 103 : LWLockRelease(NotifyQueueLock);
2979 : :
2980 : : /*
2981 : : * Scan the queue from tail to head, freezing XIDs as needed. We hold
2982 : : * NotifyQueueTailLock throughout to ensure the tail doesn't move while
2983 : : * we're working.
2984 : : */
2985 [ - + - + ]: 103 : while (!QUEUE_POS_EQUAL(pos, head))
2986 : : {
2987 : : AsyncQueueEntry *qe;
2988 : : TransactionId xid;
123 heikki.linnakangas@i 2989 :UBC 0 : int64 pageno = QUEUE_POS_PAGE(pos);
2990 : 0 : int offset = QUEUE_POS_OFFSET(pos);
2991 : :
2992 : : /* If we need a different page, release old lock and get new one */
2993 [ # # ]: 0 : if (pageno != curpage)
2994 : : {
2995 : : LWLock *lock;
2996 : :
2997 : : /* Release previous page if any */
2998 [ # # ]: 0 : if (slotno >= 0)
2999 : : {
3000 [ # # ]: 0 : if (page_dirty)
3001 : : {
3002 : 0 : NotifyCtl->shared->page_dirty[slotno] = true;
3003 : 0 : page_dirty = false;
3004 : : }
3005 : 0 : LWLockRelease(SimpleLruGetBankLock(NotifyCtl, curpage));
3006 : : }
3007 : :
3008 : 0 : lock = SimpleLruGetBankLock(NotifyCtl, pageno);
3009 : 0 : LWLockAcquire(lock, LW_EXCLUSIVE);
2 heikki.linnakangas@i 3010 :UNC 0 : slotno = SimpleLruReadPage(NotifyCtl, pageno, true, &pos);
123 heikki.linnakangas@i 3011 :UBC 0 : page_buffer = NotifyCtl->shared->page_buffer[slotno];
3012 : 0 : curpage = pageno;
3013 : : }
3014 : :
3015 : 0 : qe = (AsyncQueueEntry *) (page_buffer + offset);
3016 : 0 : xid = qe->xid;
3017 : :
3018 [ # # # # ]: 0 : if (TransactionIdIsNormal(xid) &&
3019 : 0 : TransactionIdPrecedes(xid, newFrozenXid))
3020 : : {
3021 [ # # ]: 0 : if (TransactionIdDidCommit(xid))
3022 : : {
3023 : 0 : qe->xid = FrozenTransactionId;
3024 : 0 : page_dirty = true;
3025 : : }
3026 : : else
3027 : : {
3028 : 0 : qe->xid = InvalidTransactionId;
3029 : 0 : page_dirty = true;
3030 : : }
3031 : : }
3032 : :
3033 : : /* Advance to next entry */
3034 : 0 : asyncQueueAdvance(&pos, qe->length);
3035 : : }
3036 : :
3037 : : /* Release final page lock if we acquired one */
123 heikki.linnakangas@i 3038 [ - + ]:CBC 103 : if (slotno >= 0)
3039 : : {
123 heikki.linnakangas@i 3040 [ # # ]:UBC 0 : if (page_dirty)
3041 : 0 : NotifyCtl->shared->page_dirty[slotno] = true;
3042 : 0 : LWLockRelease(SimpleLruGetBankLock(NotifyCtl, curpage));
3043 : : }
3044 : :
123 heikki.linnakangas@i 3045 :CBC 103 : LWLockRelease(NotifyQueueTailLock);
3046 : 103 : }
3047 : :
3048 : : /*
3049 : : * ProcessIncomingNotify
3050 : : *
3051 : : * Scan the queue for arriving notifications and report them to the front
3052 : : * end. The notifications might be from other sessions, or our own;
3053 : : * there's no need to distinguish here.
3054 : : *
3055 : : * If "flush" is true, force any frontend messages out immediately.
3056 : : *
3057 : : * NOTE: since we are outside any transaction, we must create our own.
3058 : : */
3059 : : static void
1643 tgl@sss.pgh.pa.us 3060 : 31 : ProcessIncomingNotify(bool flush)
3061 : : {
3062 : : /* We *must* reset the flag */
4058 andres@anarazel.de 3063 : 31 : notifyInterruptPending = false;
3064 : :
3065 : : /* Do nothing else if we aren't actively listening */
59 tgl@sss.pgh.pa.us 3066 [ + - - + ]:GNC 31 : if (LocalChannelTableIsEmpty())
5871 tgl@sss.pgh.pa.us 3067 :UBC 0 : return;
3068 : :
9419 peter_e@gmx.net 3069 [ - + ]:CBC 31 : if (Trace_notify)
8328 bruce@momjian.us 3070 [ # # ]:UBC 0 : elog(DEBUG1, "ProcessIncomingNotify");
3071 : :
2195 peter@eisentraut.org 3072 :CBC 31 : set_ps_display("notify interrupt");
3073 : :
3074 : : /*
3075 : : * We must run asyncQueueReadAllNotifications inside a transaction, else
3076 : : * bad things happen if it gets an error.
3077 : : */
5871 tgl@sss.pgh.pa.us 3078 : 31 : StartTransactionCommand();
3079 : :
3080 : 31 : asyncQueueReadAllNotifications();
3081 : :
8341 3082 : 31 : CommitTransactionCommand();
3083 : :
3084 : : /*
3085 : : * If this isn't an end-of-command case, we must flush the notify messages
3086 : : * to ensure frontend gets them promptly.
3087 : : */
1643 3088 [ + + ]: 31 : if (flush)
3089 : 9 : pq_flush();
3090 : :
2195 peter@eisentraut.org 3091 : 31 : set_ps_display("idle");
3092 : :
9419 peter_e@gmx.net 3093 [ - + ]: 31 : if (Trace_notify)
8328 bruce@momjian.us 3094 [ # # ]:UBC 0 : elog(DEBUG1, "ProcessIncomingNotify: done");
3095 : : }
3096 : :
3097 : : /*
3098 : : * Send NOTIFY message to my front end.
3099 : : */
3100 : : void
5871 tgl@sss.pgh.pa.us 3101 :CBC 55 : NotifyMyFrontEnd(const char *channel, const char *payload, int32 srcPid)
3102 : : {
7437 alvherre@alvh.no-ip. 3103 [ + - ]: 55 : if (whereToSendOutput == DestRemote)
3104 : : {
3105 : : StringInfoData buf;
3106 : :
936 nathan@postgresql.or 3107 : 55 : pq_beginmessage(&buf, PqMsg_NotificationResponse);
3077 andres@anarazel.de 3108 : 55 : pq_sendint32(&buf, srcPid);
5871 tgl@sss.pgh.pa.us 3109 : 55 : pq_sendstring(&buf, channel);
1837 heikki.linnakangas@i 3110 : 55 : pq_sendstring(&buf, payload);
9821 tgl@sss.pgh.pa.us 3111 : 55 : pq_endmessage(&buf);
3112 : :
3113 : : /*
3114 : : * NOTE: we do not do pq_flush() here. Some level of caller will
3115 : : * handle it later, allowing this message to be combined into a packet
3116 : : * with other ones.
3117 : : */
3118 : : }
3119 : : else
5871 tgl@sss.pgh.pa.us 3120 [ # # ]:UBC 0 : elog(INFO, "NOTIFY for \"%s\" payload \"%s\"", channel, payload);
10022 tgl@sss.pgh.pa.us 3121 :CBC 55 : }
3122 : :
3123 : : /* Does pendingNotifies include a match for the given event? */
3124 : : static bool
2404 3125 : 1066 : AsyncExistsPendingNotify(Notification *n)
3126 : : {
3127 [ - + ]: 1066 : if (pendingNotifies == NULL)
5871 tgl@sss.pgh.pa.us 3128 :UBC 0 : return false;
3129 : :
2404 tgl@sss.pgh.pa.us 3130 [ + + ]:CBC 1066 : if (pendingNotifies->hashtab != NULL)
3131 : : {
3132 : : /* Use the hash table to probe for a match */
3133 [ + + ]: 984 : if (hash_search(pendingNotifies->hashtab,
3134 : : &n,
3135 : : HASH_FIND,
3136 : : NULL))
2404 tgl@sss.pgh.pa.us 3137 :GBC 1 : return true;
3138 : : }
3139 : : else
3140 : : {
3141 : : /* Must scan the event list */
3142 : : ListCell *l;
3143 : :
2404 tgl@sss.pgh.pa.us 3144 [ + - + + :CBC 425 : foreach(l, pendingNotifies->events)
+ + ]
3145 : : {
3146 : 357 : Notification *oldn = (Notification *) lfirst(l);
3147 : :
3148 [ + - ]: 357 : if (n->channel_len == oldn->channel_len &&
3149 [ + + ]: 357 : n->payload_len == oldn->payload_len &&
3150 : 190 : memcmp(n->data, oldn->data,
3151 [ + + ]: 190 : n->channel_len + n->payload_len + 2) == 0)
3152 : 14 : return true;
3153 : : }
3154 : : }
3155 : :
3156 : 1051 : return false;
3157 : : }
3158 : :
3159 : : /*
3160 : : * Add a notification event to a pre-existing pendingNotifies list.
3161 : : *
3162 : : * Because pendingNotifies->events is already nonempty, this works
3163 : : * correctly no matter what CurrentMemoryContext is.
3164 : : */
3165 : : static void
3166 : 1051 : AddEventToPendingNotifies(Notification *n)
3167 : : {
3168 [ - + ]: 1051 : Assert(pendingNotifies->events != NIL);
3169 : :
3170 : : /* Create the hash tables if it's time to */
3171 [ + + ]: 1051 : if (list_length(pendingNotifies->events) >= MIN_HASHABLE_NOTIFIES &&
3172 [ + + ]: 985 : pendingNotifies->hashtab == NULL)
3173 : : {
3174 : : HASHCTL hash_ctl;
3175 : : ListCell *l;
3176 : :
3177 : : /* Create the hash table */
3178 : 2 : hash_ctl.keysize = sizeof(Notification *);
667 peter@eisentraut.org 3179 : 2 : hash_ctl.entrysize = sizeof(struct NotificationHash);
2404 tgl@sss.pgh.pa.us 3180 : 2 : hash_ctl.hash = notification_hash;
3181 : 2 : hash_ctl.match = notification_match;
3182 : 2 : hash_ctl.hcxt = CurTransactionContext;
3183 : 4 : pendingNotifies->hashtab =
3184 : 2 : hash_create("Pending Notifies",
3185 : : 256L,
3186 : : &hash_ctl,
3187 : : HASH_ELEM | HASH_FUNCTION | HASH_COMPARE | HASH_CONTEXT);
3188 : :
3189 : : /* Create the unique channel name table */
59 tgl@sss.pgh.pa.us 3190 [ - + ]:GNC 2 : Assert(pendingNotifies->uniqueChannelHash == NULL);
3191 : 2 : hash_ctl.keysize = NAMEDATALEN;
3192 : 2 : hash_ctl.entrysize = sizeof(ChannelName);
3193 : 2 : hash_ctl.hcxt = CurTransactionContext;
3194 : 4 : pendingNotifies->uniqueChannelHash =
3195 : 2 : hash_create("Pending Notify Channel Names",
3196 : : 64L,
3197 : : &hash_ctl,
3198 : : HASH_ELEM | HASH_STRINGS | HASH_CONTEXT);
3199 : :
3200 : : /* Insert all the already-existing events */
2404 tgl@sss.pgh.pa.us 3201 [ + - + + :CBC 34 : foreach(l, pendingNotifies->events)
+ + ]
3202 : : {
3203 : 32 : Notification *oldn = (Notification *) lfirst(l);
59 tgl@sss.pgh.pa.us 3204 :GNC 32 : char *channel = oldn->data;
3205 : : bool found;
3206 : :
836 john.naylor@postgres 3207 :CBC 32 : (void) hash_search(pendingNotifies->hashtab,
3208 : : &oldn,
3209 : : HASH_ENTER,
3210 : : &found);
2404 tgl@sss.pgh.pa.us 3211 [ - + ]: 32 : Assert(!found);
3212 : :
3213 : : /* Add channel name to uniqueChannelHash; might be there already */
59 tgl@sss.pgh.pa.us 3214 :GNC 32 : (void) hash_search(pendingNotifies->uniqueChannelHash,
3215 : : channel,
3216 : : HASH_ENTER,
3217 : : NULL);
3218 : : }
3219 : : }
3220 : :
3221 : : /* Add new event to the list, in order */
2404 tgl@sss.pgh.pa.us 3222 :CBC 1051 : pendingNotifies->events = lappend(pendingNotifies->events, n);
3223 : :
3224 : : /* Add event to the hash tables if needed */
3225 [ + + ]: 1051 : if (pendingNotifies->hashtab != NULL)
3226 : : {
59 tgl@sss.pgh.pa.us 3227 :GNC 985 : char *channel = n->data;
3228 : : bool found;
3229 : :
836 john.naylor@postgres 3230 :CBC 985 : (void) hash_search(pendingNotifies->hashtab,
3231 : : &n,
3232 : : HASH_ENTER,
3233 : : &found);
2404 tgl@sss.pgh.pa.us 3234 [ - + ]: 985 : Assert(!found);
3235 : :
3236 : : /* Add channel name to uniqueChannelHash; might be there already */
59 tgl@sss.pgh.pa.us 3237 :GNC 985 : (void) hash_search(pendingNotifies->uniqueChannelHash,
3238 : : channel,
3239 : : HASH_ENTER,
3240 : : NULL);
3241 : : }
2404 tgl@sss.pgh.pa.us 3242 :CBC 1051 : }
3243 : :
3244 : : /*
3245 : : * notification_hash: hash function for notification hash table
3246 : : *
3247 : : * The hash "keys" are pointers to Notification structs.
3248 : : */
3249 : : static uint32
3250 : 2001 : notification_hash(const void *key, Size keysize)
3251 : : {
3252 : 2001 : const Notification *k = *(const Notification *const *) key;
3253 : :
3254 [ - + ]: 2001 : Assert(keysize == sizeof(Notification *));
3255 : : /* We don't bother to include the payload's trailing null in the hash */
3256 : 2001 : return DatumGetUInt32(hash_any((const unsigned char *) k->data,
3257 : 2001 : k->channel_len + k->payload_len + 1));
3258 : : }
3259 : :
3260 : : /*
3261 : : * notification_match: match function to use with notification_hash
3262 : : */
3263 : : static int
2404 tgl@sss.pgh.pa.us 3264 :GBC 1 : notification_match(const void *key1, const void *key2, Size keysize)
3265 : : {
3266 : 1 : const Notification *k1 = *(const Notification *const *) key1;
3267 : 1 : const Notification *k2 = *(const Notification *const *) key2;
3268 : :
3269 [ - + ]: 1 : Assert(keysize == sizeof(Notification *));
3270 [ + - ]: 1 : if (k1->channel_len == k2->channel_len &&
3271 [ + - ]: 1 : k1->payload_len == k2->payload_len &&
3272 : 1 : memcmp(k1->data, k2->data,
3273 [ + - ]: 1 : k1->channel_len + k1->payload_len + 2) == 0)
3274 : 1 : return 0; /* equal */
2404 tgl@sss.pgh.pa.us 3275 :UBC 0 : return 1; /* not equal */
3276 : : }
3277 : :
3278 : : /* Clear the pendingActions and pendingNotifies lists. */
3279 : : static void
6577 tgl@sss.pgh.pa.us 3280 :CBC 26888 : ClearPendingActionsAndNotifies(void)
3281 : : {
3282 : : /*
3283 : : * Everything's allocated in either TopTransactionContext or the context
3284 : : * for the subtransaction to which it corresponds. So, there's nothing to
3285 : : * do here except reset the pointers; the space will be reclaimed when the
3286 : : * contexts are deleted.
3287 : : */
2354 rhaas@postgresql.org 3288 : 26888 : pendingActions = NULL;
2404 tgl@sss.pgh.pa.us 3289 : 26888 : pendingNotifies = NULL;
3290 : : /* Also clear pendingListenActions, which is derived from pendingActions */
59 tgl@sss.pgh.pa.us 3291 :GNC 26888 : pendingListenActions = NULL;
10416 bruce@momjian.us 3292 :CBC 26888 : }
3293 : :
3294 : : /*
3295 : : * GUC check_hook for notify_buffers
3296 : : */
3297 : : bool
746 alvherre@alvh.no-ip. 3298 : 1184 : check_notify_buffers(int *newval, void **extra, GucSource source)
3299 : : {
3300 : 1184 : return check_slru_buffers("notify_buffers", newval);
3301 : : }
|
