/*
* Wake up the oldest process sleeping on the CV, if there is any.
*
* Note: it's difficult to tell whether this has any real effect: we know
* whether we took an entry off the list, but the entry might only be a
* sentinel. Hence, think twice before proposing that this should return
* a flag telling whether it woke somebody.
*/
void
ConditionVariableSignal(ConditionVariable *cv)
{
PGPROC *proc = NULL;
/* Remove the first process from the wakeup queue (if any). */
SpinLockAcquire(&cv->mutex);
if (!proclist_is_empty(&cv->wakeup))
proc = proclist_pop_head_node(&cv->wakeup, cvWaitLink);
SpinLockRelease(&cv->mutex);
/* If we found someone sleeping, set their latch to wake them up. */
if (proc != NULL)
SetLatch(&proc->procLatch);
}
void
SubscribeTerminate()
{
LIST_ITEM *pTrapReferenceListItem;
_DBG_ENTER_LVL(_DBG_LVL_FUNC_TRACE, SubscribeTerminate);
SpinLockAcquire(&SubscribedTrapsListLock);
SubscribeInitialized = 0;
SpinLockRelease(&SubscribedTrapsListLock);
iba_sd_deregister(SdClientHandle);
/* no more SD callbacks will occur now, so OPQUEUED is not important */
SpinLockAcquire(&SubscribedTrapsListLock);
while ( NULL != (pTrapReferenceListItem = QListRemoveHead(&SubscribedTrapsList)))
{
TRAP_REFERENCE *pTrapReference = (TRAP_REFERENCE *)QListObj(pTrapReferenceListItem);
ASSERT(pTrapReference != NULL);
TimerStop(&pTrapReference->Timer);
// release lock so TimerDestroy can wait for callback
SpinLockRelease(&SubscribedTrapsListLock);
TimerDestroy(&pTrapReference->Timer);
MemoryDeallocate(pTrapReference);
SpinLockAcquire(&SubscribedTrapsListLock);
}
SpinLockRelease(&SubscribedTrapsListLock);
QListDestroy(&SubscribedTrapsList);
SpinLockDestroy(&SubscribedTrapsListLock);
_DBG_LEAVE_LVL( _DBG_LVL_FUNC_TRACE );
}
/*
* Set the identity of the process that will receive from a shared message
* queue.
*/
void
shm_mq_set_receiver(shm_mq *mq, PGPROC *proc)
{
volatile shm_mq *vmq = mq;
PGPROC *sender;
SpinLockAcquire(&mq->mq_mutex);
Assert(vmq->mq_receiver == NULL);
vmq->mq_receiver = proc;
sender = vmq->mq_sender;
SpinLockRelease(&mq->mq_mutex);
if (sender != NULL)
SetLatch(&sender->procLatch);
}
/* Set state for current walsender (only called in walsender) */
void
WalSndSetState(WalSndState state)
{
/* use volatile pointer to prevent code rearrangement */
volatile WalSnd *walsnd = MyWalSnd;
Assert(am_walsender);
if (walsnd->state == state)
return;
SpinLockAcquire(&walsnd->mutex);
walsnd->state = state;
SpinLockRelease(&walsnd->mutex);
}
/*
* Returns the last+1 byte position that walreceiver has written.
*
* Optionally, returns the previous chunk start, that is the first byte
* written in the most recent walreceiver flush cycle. Callers not
* interested in that value may pass NULL for latestChunkStart.
*/
XLogRecPtr
GetWalRcvWriteRecPtr(XLogRecPtr *latestChunkStart)
{
/* use volatile pointer to prevent code rearrangement */
volatile WalRcvData *walrcv = WalRcv;
XLogRecPtr recptr;
SpinLockAcquire(&walrcv->mutex);
recptr = walrcv->receivedUpto;
if (latestChunkStart)
*latestChunkStart = walrcv->latestChunkStart;
SpinLockRelease(&walrcv->mutex);
return recptr;
}
/*
* Convert a slot that's marked as RS_EPHEMERAL to a RS_PERSISTENT slot,
* guaranteeing it will be there after an eventual crash.
*/
void
ReplicationSlotPersist(void)
{
ReplicationSlot *slot = MyReplicationSlot;
Assert(slot != NULL);
Assert(slot->data.persistency != RS_PERSISTENT);
SpinLockAcquire(&slot->mutex);
slot->data.persistency = RS_PERSISTENT;
SpinLockRelease(&slot->mutex);
ReplicationSlotMarkDirty();
ReplicationSlotSave();
}
/*
* Cancel any pending sleep operation.
*
* We just need to remove ourselves from the wait queue of any condition
* variable for which we have previously prepared a sleep.
*
* Do nothing if nothing is pending; this allows this function to be called
* during transaction abort to clean up any unfinished CV sleep.
*/
void
ConditionVariableCancelSleep(void)
{
ConditionVariable *cv = cv_sleep_target;
if (cv == NULL)
return;
SpinLockAcquire(&cv->mutex);
if (proclist_contains(&cv->wakeup, MyProc->pgprocno, cvWaitLink))
proclist_delete(&cv->wakeup, MyProc->pgprocno, cvWaitLink);
SpinLockRelease(&cv->mutex);
cv_sleep_target = NULL;
}
/*
* LWLockWaitCancel - cancel currently waiting on LW lock
*
* Used to clean up before immediate exit in certain very special situations
* like shutdown request to Filerep Resync Manger or Workers. Although this is
* not the best practice it is necessary to avoid any starvation situations
* during filerep transition situations (Resync Mode -> Changetracking mode)
*
* Note:- This function should not be used for normal situations. It is strictly
* written for very special situations. If you need to use this, you may want
* to re-think your design.
*/
void
LWLockWaitCancel(void)
{
volatile PGPROC *proc = MyProc;
volatile LWLock *lwWaitingLock = NULL;
/* We better have a PGPROC structure */
Assert(proc != NULL);
/* If we're not waiting on any LWLock then nothing doing here */
if (!proc->lwWaiting)
return;
lwWaitingLock = &(LWLockArray[lwWaitingLockId].lock);
/* Protect from other modifiers */
SpinLockAcquire(&lwWaitingLock->mutex);
PGPROC *currProc = lwWaitingLock->head;
/* Search our PROC in the waiters list and remove it */
if (proc == lwWaitingLock->head)
{
lwWaitingLock->head = currProc = proc->lwWaitLink;
proc->lwWaitLink = NULL;
}
else
{
while(currProc != NULL)
{
if (currProc->lwWaitLink == proc)
{
currProc->lwWaitLink = proc->lwWaitLink;
proc->lwWaitLink = NULL;
break;
}
currProc = currProc->lwWaitLink;
}
}
if (lwWaitingLock->tail == proc)
lwWaitingLock->tail = currProc;
/* Done with modification */
SpinLockRelease(&lwWaitingLock->mutex);
return;
}
/*
* ShmemAlloc -- allocate max-aligned chunk from shared memory
*
* Assumes ShmemLock and ShmemSegHdr are initialized.
*
* Returns: real pointer to memory or NULL if we are out
* of space. Has to return a real pointer in order
* to be compatible with malloc().
*/
void *
ShmemAlloc(Size size)
{
Size newStart;
Size newFree;
void *newSpace;
/*
* Ensure all space is adequately aligned. We used to only MAXALIGN this
* space but experience has proved that on modern systems that is not good
* enough. Many parts of the system are very sensitive to critical data
* structures getting split across cache line boundaries. To avoid that,
* attempt to align the beginning of the allocation to a cache line
* boundary. The calling code will still need to be careful about how it
* uses the allocated space - e.g. by padding each element in an array of
* structures out to a power-of-two size - but without this, even that
* won't be sufficient.
*/
size = CACHELINEALIGN(size);
Assert(ShmemSegHdr != NULL);
SpinLockAcquire(ShmemLock);
newStart = ShmemSegHdr->freeoffset;
/* extra alignment for large requests, since they are probably buffers */
if (size >= BLCKSZ)
newStart = BUFFERALIGN(newStart);
newFree = newStart + size;
if (newFree <= ShmemSegHdr->totalsize)
{
newSpace = (void *) ((char *) ShmemBase + newStart);
ShmemSegHdr->freeoffset = newFree;
}
else
newSpace = NULL;
SpinLockRelease(ShmemLock);
if (!newSpace)
ereport(WARNING,
(errcode(ERRCODE_OUT_OF_MEMORY),
errmsg("out of shared memory")));
return newSpace;
}
/*
* Compute the oldest xmin across all slots and store it in the ProcArray.
*/
void
ReplicationSlotsComputeRequiredXmin(bool already_locked)
{
int i;
TransactionId agg_xmin = InvalidTransactionId;
TransactionId agg_catalog_xmin = InvalidTransactionId;
Assert(ReplicationSlotCtl != NULL);
if (!already_locked)
LWLockAcquire(ReplicationSlotControlLock, LW_SHARED);
for (i = 0; i < max_replication_slots; i++)
{
ReplicationSlot *s = &ReplicationSlotCtl->replication_slots[i];
TransactionId effective_xmin;
TransactionId effective_catalog_xmin;
if (!s->in_use)
continue;
{
volatile ReplicationSlot *vslot = s;
SpinLockAcquire(&s->mutex);
effective_xmin = vslot->effective_xmin;
effective_catalog_xmin = vslot->effective_catalog_xmin;
SpinLockRelease(&s->mutex);
}
/* check the data xmin */
if (TransactionIdIsValid(effective_xmin) &&
(!TransactionIdIsValid(agg_xmin) ||
TransactionIdPrecedes(effective_xmin, agg_xmin)))
agg_xmin = effective_xmin;
/* check the catalog xmin */
if (TransactionIdIsValid(effective_catalog_xmin) &&
(!TransactionIdIsValid(agg_catalog_xmin) ||
TransactionIdPrecedes(effective_catalog_xmin, agg_catalog_xmin)))
agg_catalog_xmin = effective_catalog_xmin;
}
if (!already_locked)
LWLockRelease(ReplicationSlotControlLock);
ProcArraySetReplicationSlotXmin(agg_xmin, agg_catalog_xmin, already_locked);
}
/*
* Release a replication slot, this or another backend can ReAcquire it
* later. Resources this slot requires will be preserved.
*/
void
ReplicationSlotRelease(void)
{
ReplicationSlot *slot = MyReplicationSlot;
Assert(slot != NULL && slot->active);
/* Mark slot inactive. We're not freeing it, just disconnecting. */
{
volatile ReplicationSlot *vslot = slot;
SpinLockAcquire(&slot->mutex);
vslot->active = false;
SpinLockRelease(&slot->mutex);
MyReplicationSlot = NULL;
}
}
/*
* Returns the last+1 byte position that walreceiver has written.
*
* Optionally, returns the previous chunk start, that is the first byte
* written in the most recent walreceiver flush cycle. Callers not
* interested in that value may pass NULL for latestChunkStart. Same for
* receiveTLI.
*/
XLogRecPtr
GetWalRcvWriteRecPtr(XLogRecPtr *latestChunkStart, TimeLineID *receiveTLI)
{
WalRcvData *walrcv = WalRcv;
XLogRecPtr recptr;
SpinLockAcquire(&walrcv->mutex);
recptr = walrcv->receivedUpto;
if (latestChunkStart)
*latestChunkStart = walrcv->latestChunkStart;
if (receiveTLI)
*receiveTLI = walrcv->receivedTLI;
SpinLockRelease(&walrcv->mutex);
return recptr;
}
// callback when SA query completes or fails
static void
ProcessTrapSubscribeResponse(void* Context, FABRIC_OPERATION_DATA* pFabOp, FSTATUS Status, uint32 MadStatus)
{
TRAP_REFERENCE *pTrapReference = (TRAP_REFERENCE *)Context;
_DBG_ENTER_LVL(_DBG_LVL_FUNC_TRACE, ProcessTrapSubscribeResponse);
SpinLockAcquire(&SubscribedTrapsListLock);
if (SubscribeInitialized == 0)
goto unlock; // we are shutting down, don't start anything new
pTrapReference->Flags = (TRAP_REFERENCE_FLAGS)(pTrapReference->Flags & ~TRAP_REF_FLAGS_OPQUEUED);
if (Status != FSUCCESS)
{
// our attempt failed
if (pFabOp->Value.InformInfo.Subscribe == 0)
{
_DBG_WARN( ("Trap Unsubscribe Failure. Status = %s, MadStatus = 0x%x: %s.\n",
_DBG_PTR(iba_fstatus_msg(Status)), MadStatus, _DBG_PTR(iba_sd_mad_status_msg(MadStatus))) );
} else {
_DBG_WARN( ("Trap Subscribe Failure. Status = %s, MadStatus = 0x%x: %s.\n",
_DBG_PTR(iba_fstatus_msg(Status)), MadStatus, _DBG_PTR(iba_sd_mad_status_msg(MadStatus))) );
}
// delay any retry or adjustment so we don't beat on SA too hard
if (MadStatus == MAD_STATUS_BUSY)
{
ProcessTrapStateChange(pTrapReference, (SystemGetRandom() % PORT_DOWN_DELAY));
_DBG_PRINT(_DBG_LVL_INFO, ("SM Returned Busy. Delaying Subscription By %ld MS.\n", (SystemGetRandom() % PORT_DOWN_DELAY)));
} else {
ProcessTrapStateChange(pTrapReference, RETRY_DELAY);
}
} else {
if (pFabOp->Value.InformInfo.Subscribe == 1)
{
/* We are now subscribed with the SA */
pTrapReference->Flags = (TRAP_REFERENCE_FLAGS)(pTrapReference->Flags | TRAP_REF_FLAGS_SUBSCRIBED);
} else {
/* We are now completely unsubscribed from the SA */
// flag was cleared when we issued unsubscribe request
}
ProcessTrapStateChange(pTrapReference, 0);
}
unlock:
SpinLockRelease(&SubscribedTrapsListLock);
_DBG_LEAVE_LVL( _DBG_LVL_FUNC_TRACE );
}
/*
* Prepare to wait on a given condition variable.
*
* This can optionally be called before entering a test/sleep loop.
* Doing so is more efficient if we'll need to sleep at least once.
* However, if the first test of the exit condition is likely to succeed,
* it's more efficient to omit the ConditionVariablePrepareToSleep call.
* See comments in ConditionVariableSleep for more detail.
*
* Caution: "before entering the loop" means you *must* test the exit
* condition between calling ConditionVariablePrepareToSleep and calling
* ConditionVariableSleep. If that is inconvenient, omit calling
* ConditionVariablePrepareToSleep.
*/
void
ConditionVariablePrepareToSleep(ConditionVariable *cv)
{
int pgprocno = MyProc->pgprocno;
/*
* If first time through in this process, create a WaitEventSet, which
* we'll reuse for all condition variable sleeps.
*/
if (cv_wait_event_set == NULL)
{
WaitEventSet *new_event_set;
new_event_set = CreateWaitEventSet(TopMemoryContext, 2);
AddWaitEventToSet(new_event_set, WL_LATCH_SET, PGINVALID_SOCKET,
MyLatch, NULL);
AddWaitEventToSet(new_event_set, WL_EXIT_ON_PM_DEATH, PGINVALID_SOCKET,
NULL, NULL);
/* Don't set cv_wait_event_set until we have a correct WES. */
cv_wait_event_set = new_event_set;
}
/*
* If some other sleep is already prepared, cancel it; this is necessary
* because we have just one static variable tracking the prepared sleep,
* and also only one cvWaitLink in our PGPROC. It's okay to do this
* because whenever control does return to the other test-and-sleep loop,
* its ConditionVariableSleep call will just re-establish that sleep as
* the prepared one.
*/
if (cv_sleep_target != NULL)
ConditionVariableCancelSleep();
/* Record the condition variable on which we will sleep. */
cv_sleep_target = cv;
/*
* Reset my latch before adding myself to the queue, to ensure that we
* don't miss a wakeup that occurs immediately.
*/
ResetLatch(MyLatch);
/* Add myself to the wait queue. */
SpinLockAcquire(&cv->mutex);
proclist_push_tail(&cv->wakeup, pgprocno, cvWaitLink);
SpinLockRelease(&cv->mutex);
}
/*
* StrategySyncStart -- tell BufferSync where to start syncing
*
* The result is the buffer index of the best buffer to sync first.
* BufferSync() will proceed circularly around the buffer array from there.
*
* In addition, we return the completed-pass count (which is effectively
* the higher-order bits of nextVictimBuffer) and the count of recent buffer
* allocs if non-NULL pointers are passed. The alloc count is reset after
* being read.
*/
int
StrategySyncStart(uint32 *complete_passes, uint32 *num_buf_alloc)
{
int result;
SpinLockAcquire(&StrategyControl->buffer_strategy_lock);
result = StrategyControl->nextVictimBuffer;
if (complete_passes)
*complete_passes = StrategyControl->completePasses;
if (num_buf_alloc)
{
*num_buf_alloc = StrategyControl->numBufferAllocs;
StrategyControl->numBufferAllocs = 0;
}
SpinLockRelease(&StrategyControl->buffer_strategy_lock);
return result;
}
/*
* UnSetDistributedTransactionId simply acquires the mutex and resets the backend's
* distributed transaction data in shared memory to the initial values.
*/
void
UnSetDistributedTransactionId(void)
{
/* backend does not exist if the extension is not created */
if (MyBackendData)
{
SpinLockAcquire(&MyBackendData->mutex);
MyBackendData->databaseId = 0;
MyBackendData->transactionId.initiatorNodeIdentifier = 0;
MyBackendData->transactionId.transactionOriginator = false;
MyBackendData->transactionId.transactionNumber = 0;
MyBackendData->transactionId.timestamp = 0;
SpinLockRelease(&MyBackendData->mutex);
}
}
/*
* Mark us as STOPPED in shared memory at exit.
*/
static void
WalRcvDie(int code, Datum arg)
{
/* use volatile pointer to prevent code rearrangement */
volatile WalRcvData *walrcv = WalRcv;
SpinLockAcquire(&walrcv->mutex);
Assert(walrcv->walRcvState == WALRCV_RUNNING ||
walrcv->walRcvState == WALRCV_STOPPING);
walrcv->walRcvState = WALRCV_STOPPED;
walrcv->pid = 0;
SpinLockRelease(&walrcv->mutex);
/* Terminate the connection gracefully. */
if (walrcv_disconnect != NULL)
walrcv_disconnect();
}
/*
* allocate some new elements and link them into the free list
*/
static bool
element_alloc(HTAB *hashp, int nelem)
{
/* use volatile pointer to prevent code rearrangement */
volatile HASHHDR *hctlv = hashp->hctl;
Size elementSize;
HASHELEMENT *firstElement;
HASHELEMENT *tmpElement;
HASHELEMENT *prevElement;
int i;
if (hashp->isfixed)
return false;
/* Each element has a HASHELEMENT header plus user data. */
elementSize = MAXALIGN(sizeof(HASHELEMENT)) + MAXALIGN(hctlv->entrysize);
CurrentDynaHashCxt = hashp->hcxt;
firstElement = (HASHELEMENT *) hashp->alloc(nelem * elementSize);
if (!firstElement)
return false;
/* prepare to link all the new entries into the freelist */
prevElement = NULL;
tmpElement = firstElement;
for (i = 0; i < nelem; i++)
{
tmpElement->link = prevElement;
prevElement = tmpElement;
tmpElement = (HASHELEMENT *) (((char *) tmpElement) + elementSize);
}
/* if partitioned, must lock to touch freeList */
if (IS_PARTITIONED(hctlv))
SpinLockAcquire(&hctlv->mutex);
/* freelist could be nonempty if two backends did this concurrently */
firstElement->link = hctlv->freeList;
hctlv->freeList = prevElement;
if (IS_PARTITIONED(hctlv))
SpinLockRelease(&hctlv->mutex);
return true;
}
/*
* Find a previously created slot and mark it as used by this backend.
*/
void
ReplicationSlotAcquire(const char *name)
{
ReplicationSlot *slot = NULL;
int i;
int active_pid = 0;
Assert(MyReplicationSlot == NULL);
ReplicationSlotValidateName(name, ERROR);
/* Search for the named slot and mark it active if we find it. */
LWLockAcquire(ReplicationSlotControlLock, LW_SHARED);
for (i = 0; i < max_replication_slots; i++)
{
ReplicationSlot *s = &ReplicationSlotCtl->replication_slots[i];
if (s->in_use && strcmp(name, NameStr(s->data.name)) == 0)
{
volatile ReplicationSlot *vslot = s;
SpinLockAcquire(&s->mutex);
active_pid = vslot->active_pid;
if (active_pid == 0)
vslot->active_pid = MyProcPid;
SpinLockRelease(&s->mutex);
slot = s;
break;
}
}
LWLockRelease(ReplicationSlotControlLock);
/* If we did not find the slot or it was already active, error out. */
if (slot == NULL)
ereport(ERROR,
(errcode(ERRCODE_UNDEFINED_OBJECT),
errmsg("replication slot \"%s\" does not exist", name)));
if (active_pid != 0)
ereport(ERROR,
(errcode(ERRCODE_OBJECT_IN_USE),
errmsg("replication slot \"%s\" is already active for pid %d",
name, active_pid)));
/* We made this slot active, so it's ours now. */
MyReplicationSlot = slot;
}
/*
* Set sender's latch, unless queue is detached.
*/
static shm_mq_result
shm_mq_notify_receiver(volatile shm_mq *mq)
{
PGPROC *receiver;
bool detached;
SpinLockAcquire(&mq->mq_mutex);
detached = mq->mq_detached;
receiver = mq->mq_receiver;
SpinLockRelease(&mq->mq_mutex);
if (detached)
return SHM_MQ_DETACHED;
if (receiver)
SetLatch(&receiver->procLatch);
return SHM_MQ_SUCCESS;
}
/*
* Flush the log to disk.
*
* If we're in the midst of dying, it's unwise to do anything that might throw
* an error, so we skip sending a reply in that case.
*/
static void
XLogWalRcvFlush(bool dying)
{
if (XLByteLT(LogstreamResult.Flush, LogstreamResult.Write))
{
/* use volatile pointer to prevent code rearrangement */
volatile WalRcvData *walrcv = WalRcv;
issue_xlog_fsync(recvFile, recvId, recvSeg);
LogstreamResult.Flush = LogstreamResult.Write;
/* Update shared-memory status */
SpinLockAcquire(&walrcv->mutex);
if (XLByteLT(walrcv->receivedUpto, LogstreamResult.Flush))
{
walrcv->latestChunkStart = walrcv->receivedUpto;
walrcv->receivedUpto = LogstreamResult.Flush;
}
SpinLockRelease(&walrcv->mutex);
/* Signal the startup process and walsender that new WAL has arrived */
WakeupRecovery();
if (AllowCascadeReplication())
WalSndWakeup();
/* Report XLOG streaming progress in PS display */
if (update_process_title)
{
char activitymsg[50];
snprintf(activitymsg, sizeof(activitymsg), "streaming %X/%X",
LogstreamResult.Write.xlogid,
LogstreamResult.Write.xrecoff);
set_ps_display(activitymsg, false);
}
/* Also let the master know that we made some progress */
if (!dying)
{
XLogWalRcvSendReply();
XLogWalRcvSendHSFeedback();
}
}
}
/*
* Request walsenders to reload the currently-open WAL file
*/
void
WalSndRqstFileReload(void)
{
int i;
for (i = 0; i < max_wal_senders; i++)
{
/* use volatile pointer to prevent code rearrangement */
volatile WalSnd *walsnd = &WalSndCtl->walsnds[i];
if (walsnd->pid == 0)
continue;
SpinLockAcquire(&walsnd->mutex);
walsnd->needreload = true;
SpinLockRelease(&walsnd->mutex);
}
}
/*
* FirstCallSinceLastCheckpoint allows a process to take an action once
* per checkpoint cycle by asynchronously checking for checkpoint completion.
*/
bool
FirstCallSinceLastCheckpoint(void)
{
static int ckpt_done = 0;
int new_done;
bool FirstCall = false;
SpinLockAcquire(&CheckpointerShmem->ckpt_lck);
new_done = CheckpointerShmem->ckpt_done;
SpinLockRelease(&CheckpointerShmem->ckpt_lck);
if (new_done != ckpt_done)
FirstCall = true;
ckpt_done = new_done;
return FirstCall;
}
/*
* Request postmaster to start walreceiver.
*
* recptr indicates the position where streaming should begin, and conninfo
* is a libpq connection string to use.
*/
void
RequestXLogStreaming(XLogRecPtr recptr, const char *conninfo)
{
/* use volatile pointer to prevent code rearrangement */
volatile WalRcvData *walrcv = WalRcv;
pg_time_t now = (pg_time_t) time(NULL);
/*
* We always start at the beginning of the segment. That prevents a broken
* segment (i.e., with no records in the first half of a segment) from
* being created by XLOG streaming, which might cause trouble later on if
* the segment is e.g archived.
*/
if (recptr.xrecoff % XLogSegSize != 0)
recptr.xrecoff -= recptr.xrecoff % XLogSegSize;
SpinLockAcquire(&walrcv->mutex);
/* It better be stopped before we try to restart it */
Assert(walrcv->walRcvState == WALRCV_STOPPED);
if (conninfo != NULL)
strlcpy((char *) walrcv->conninfo, conninfo, MAXCONNINFO);
else
walrcv->conninfo[0] = '\0';
walrcv->walRcvState = WALRCV_STARTING;
walrcv->startTime = now;
/*
* If this is the first startup of walreceiver, we initialize receivedUpto
* and latestChunkStart to receiveStart.
*/
if (walrcv->receiveStart.xlogid == 0 &&
walrcv->receiveStart.xrecoff == 0)
{
walrcv->receivedUpto = recptr;
walrcv->latestChunkStart = recptr;
}
walrcv->receiveStart = recptr;
SpinLockRelease(&walrcv->mutex);
SendPostmasterSignal(PMSIGNAL_START_WALRECEIVER);
}
/*
* Return the number of bytes that can still be allocated.
*/
extern Size
shm_toc_freespace(shm_toc *toc)
{
volatile shm_toc *vtoc = toc;
Size total_bytes;
Size allocated_bytes;
Size nentry;
Size toc_bytes;
SpinLockAcquire(&toc->toc_mutex);
total_bytes = vtoc->toc_total_bytes;
allocated_bytes = vtoc->toc_allocated_bytes;
nentry = vtoc->toc_nentry;
SpinLockRelease(&toc->toc_mutex);
toc_bytes = offsetof(shm_toc, toc_entry) +nentry * sizeof(shm_toc_entry);
Assert(allocated_bytes + BUFFERALIGN(toc_bytes) <= total_bytes);
return total_bytes - (allocated_bytes + BUFFERALIGN(toc_bytes));
}
/*
* ShmemAlloc -- allocate max-aligned chunk from shared memory
*
* Assumes ShmemLock and ShmemSegHdr are initialized.
*
* Returns: real pointer to memory or NULL if we are out
* of space. Has to return a real pointer in order
* to be compatible with malloc().
*/
void *
ShmemAlloc(Size size)
{
Size newStart;
Size newFree;
void *newSpace;
/* use volatile pointer to prevent code rearrangement */
volatile PGShmemHeader *shmemseghdr = ShmemSegHdr;
/*
* ensure all space is adequately aligned.
*/
size = MAXALIGN(size);
Assert(shmemseghdr != NULL);
SpinLockAcquire(ShmemLock);
newStart = shmemseghdr->freeoffset;
/* extra alignment for large requests, since they are probably buffers */
if (size >= BLCKSZ)
newStart = BUFFERALIGN(newStart);
newFree = newStart + size;
if (newFree <= shmemseghdr->totalsize)
{
newSpace = (void *) ((char *) ShmemBase + newStart);
shmemseghdr->freeoffset = newFree;
}
else
newSpace = NULL;
SpinLockRelease(ShmemLock);
if (!newSpace)
ereport(WARNING,
(errcode(ERRCODE_OUT_OF_MEMORY),
errmsg("out of shared memory")));
return newSpace;
}
/*
* StrategyFreeBuffer: put a buffer on the freelist
*/
void
StrategyFreeBuffer(volatile BufferDesc *buf)
{
SpinLockAcquire(&StrategyControl->buffer_strategy_lock);
/*
* It is possible that we are told to put something in the freelist that
* is already in it; don't screw up the list if so.
*/
if (buf->freeNext == FREENEXT_NOT_IN_LIST)
{
buf->freeNext = StrategyControl->firstFreeBuffer;
if (buf->freeNext < 0)
StrategyControl->lastFreeBuffer = buf->buf_id;
StrategyControl->firstFreeBuffer = buf->buf_id;
}
SpinLockRelease(&StrategyControl->buffer_strategy_lock);
}
/*
* GetBackendDataForProc writes the backend data for the given process to
* result. If the process is part of a lock group (parallel query) it
* returns the leader data instead.
*/
void
GetBackendDataForProc(PGPROC *proc, BackendData *result)
{
BackendData *backendData = NULL;
int pgprocno = proc->pgprocno;
if (proc->lockGroupLeader != NULL)
{
pgprocno = proc->lockGroupLeader->pgprocno;
}
backendData = &backendManagementShmemData->backends[pgprocno];
SpinLockAcquire(&backendData->mutex);
memcpy(result, backendData, sizeof(BackendData));
SpinLockRelease(&backendData->mutex);
}
/*
* Link an entry back in the cache freelist
*
* The entry must be already marked as free by the caller.
*/
void
Cache_AddToFreelist(Cache *cache, CacheEntry *entry)
{
Assert(NULL != cache);
Assert(NULL != entry);
CACHE_ASSERT_WIPED(entry);
Assert(entry->state == CACHE_ENTRY_FREE);
CacheHdr *cacheHdr = cache->cacheHdr;
/* Must lock to touch freeList */
SpinLockAcquire(&cacheHdr->spinlock);
entry->nextEntry = cacheHdr->freeList;
cacheHdr->freeList = entry;
Cache_UpdatePerfCounter(&cacheHdr->cacheStats.noFreeEntries, 1 /* delta */);
SpinLockRelease(&cacheHdr->spinlock);
}
/*
* ShmemAllocNoError -- allocate max-aligned chunk from shared memory
*
* As ShmemAlloc, but returns NULL if out of space, rather than erroring.
*/
void *
ShmemAllocNoError(Size size)
{
Size newStart;
Size newFree;
void *newSpace;
/*
* Ensure all space is adequately aligned. We used to only MAXALIGN this
* space but experience has proved that on modern systems that is not good
* enough. Many parts of the system are very sensitive to critical data
* structures getting split across cache line boundaries. To avoid that,
* attempt to align the beginning of the allocation to a cache line
* boundary. The calling code will still need to be careful about how it
* uses the allocated space - e.g. by padding each element in an array of
* structures out to a power-of-two size - but without this, even that
* won't be sufficient.
*/
size = CACHELINEALIGN(size);
Assert(ShmemSegHdr != NULL);
SpinLockAcquire(ShmemLock);
newStart = ShmemSegHdr->freeoffset;
newFree = newStart + size;
if (newFree <= ShmemSegHdr->totalsize)
{
newSpace = (void *) ((char *) ShmemBase + newStart);
ShmemSegHdr->freeoffset = newFree;
}
else
newSpace = NULL;
SpinLockRelease(ShmemLock);
/* note this assert is okay with newSpace == NULL */
Assert(newSpace == (void *) CACHELINEALIGN(newSpace));
return newSpace;
}
