#include "postgres.h"
#include "access/xlog.h"
#include "access/xlog_internal.h"
#include "libpq/pqsignal.h"
#include "miscadmin.h"
#include "pgstat.h"
#include "postmaster/bgwriter.h"
#include "postmaster/interrupt.h"
#include "storage/buf_internals.h"
#include "storage/bufmgr.h"
#include "storage/condition_variable.h"
#include "storage/fd.h"
#include "storage/ipc.h"
#include "storage/lwlock.h"
#include "storage/proc.h"
#include "storage/procsignal.h"
#include "storage/shmem.h"
#include "storage/smgr.h"
#include "storage/spin.h"
#include "storage/standby.h"
#include "utils/guc.h"
#include "utils/memutils.h"
#include "utils/resowner.h"
#include "utils/timestamp.h"

Include dependency graph for bgwriter.c:

Macros
#define	HIBERNATE_FACTOR 50

#define	LOG_SNAPSHOT_INTERVAL_MS 15000

Functions
void	BackgroundWriterMain (void)

Variables
int	BgWriterDelay = 200

static TimestampTz	last_snapshot_ts

static XLogRecPtr	last_snapshot_lsn = InvalidXLogRecPtr

Macro Definition Documentation

◆ HIBERNATE_FACTOR

#define HIBERNATE_FACTOR 50

Definition at line 67 of file bgwriter.c.

◆ LOG_SNAPSHOT_INTERVAL_MS

#define LOG_SNAPSHOT_INTERVAL_MS 15000

Definition at line 73 of file bgwriter.c.

Function Documentation

◆ BackgroundWriterMain()

void BackgroundWriterMain ( void )

Definition at line 91 of file bgwriter.c.

 {
     sigjmp_buf  local_sigjmp_buf;
     MemoryContext bgwriter_context;
     bool        prev_hibernate;
     WritebackContext wb_context;
  
     /*
      * Properly accept or ignore signals that might be sent to us.
      */
     pqsignal(SIGHUP, SignalHandlerForConfigReload);
     pqsignal(SIGINT, SIG_IGN);
     pqsignal(SIGTERM, SignalHandlerForShutdownRequest);
     /* SIGQUIT handler was already set up by InitPostmasterChild */
     pqsignal(SIGALRM, SIG_IGN);
     pqsignal(SIGPIPE, SIG_IGN);
     pqsignal(SIGUSR1, procsignal_sigusr1_handler);
     pqsignal(SIGUSR2, SIG_IGN);
  
     /*
      * Reset some signals that are accepted by postmaster but not here
      */
     pqsignal(SIGCHLD, SIG_DFL);
  
     /*
      * We just started, assume there has been either a shutdown or
      * end-of-recovery snapshot.
      */
     last_snapshot_ts = GetCurrentTimestamp();
  
     /*
      * Create a memory context that we will do all our work in.  We do this so
      * that we can reset the context during error recovery and thereby avoid
      * possible memory leaks.  Formerly this code just ran in
      * TopMemoryContext, but resetting that would be a really bad idea.
      */
     bgwriter_context = AllocSetContextCreate(TopMemoryContext,
                                              "Background Writer",
                                              ALLOCSET_DEFAULT_SIZES);
     MemoryContextSwitchTo(bgwriter_context);
  
     WritebackContextInit(&wb_context, &bgwriter_flush_after);
  
     /*
      * If an exception is encountered, processing resumes here.
      *
      * You might wonder why this isn't coded as an infinite loop around a
      * PG_TRY construct.  The reason is that this is the bottom of the
      * exception stack, and so with PG_TRY there would be no exception handler
      * in force at all during the CATCH part.  By leaving the outermost setjmp
      * always active, we have at least some chance of recovering from an error
      * during error recovery.  (If we get into an infinite loop thereby, it
      * will soon be stopped by overflow of elog.c's internal state stack.)
      *
      * Note that we use sigsetjmp(..., 1), so that the prevailing signal mask
      * (to wit, BlockSig) will be restored when longjmp'ing to here.  Thus,
      * signals other than SIGQUIT will be blocked until we complete error
      * recovery.  It might seem that this policy makes the HOLD_INTERRUPTS()
      * call redundant, but it is not since InterruptPending might be set
      * already.
      */
     if (sigsetjmp(local_sigjmp_buf, 1) != 0)
     {
         /* Since not using PG_TRY, must reset error stack by hand */
         error_context_stack = NULL;
  
         /* Prevent interrupts while cleaning up */
         HOLD_INTERRUPTS();
  
         /* Report the error to the server log */
         EmitErrorReport();
  
         /*
          * These operations are really just a minimal subset of
          * AbortTransaction().  We don't have very many resources to worry
          * about in bgwriter, but we do have LWLocks, buffers, and temp files.
          */
         LWLockReleaseAll();
         ConditionVariableCancelSleep();
         UnlockBuffers();
         ReleaseAuxProcessResources(false);
         AtEOXact_Buffers(false);
         AtEOXact_SMgr();
         AtEOXact_Files(false);
         AtEOXact_HashTables(false);
  
         /*
          * Now return to normal top-level context and clear ErrorContext for
          * next time.
          */
         MemoryContextSwitchTo(bgwriter_context);
         FlushErrorState();
  
         /* Flush any leaked data in the top-level context */
         MemoryContextResetAndDeleteChildren(bgwriter_context);
  
         /* re-initialize to avoid repeated errors causing problems */
         WritebackContextInit(&wb_context, &bgwriter_flush_after);
  
         /* Now we can allow interrupts again */
         RESUME_INTERRUPTS();
  
         /*
          * Sleep at least 1 second after any error.  A write error is likely
          * to be repeated, and we don't want to be filling the error logs as
          * fast as we can.
          */
         pg_usleep(1000000L);
  
         /*
          * Close all open files after any error.  This is helpful on Windows,
          * where holding deleted files open causes various strange errors.
          * It's not clear we need it elsewhere, but shouldn't hurt.
          */
         smgrcloseall();
  
         /* Report wait end here, when there is no further possibility of wait */
         pgstat_report_wait_end();
     }
  
     /* We can now handle ereport(ERROR) */
     PG_exception_stack = &local_sigjmp_buf;
  
     /*
      * Unblock signals (they were blocked when the postmaster forked us)
      */
     sigprocmask(SIG_SETMASK, &UnBlockSig, NULL);
  
     /*
      * Reset hibernation state after any error.
      */
     prev_hibernate = false;
  
     /*
      * Loop forever
      */
     for (;;)
     {
         bool        can_hibernate;
         int         rc;
  
         /* Clear any already-pending wakeups */
         ResetLatch(MyLatch);
  
         HandleMainLoopInterrupts();
  
         /*
          * Do one cycle of dirty-buffer writing.
          */
         can_hibernate = BgBufferSync(&wb_context);
  
         /* Report pending statistics to the cumulative stats system */
         pgstat_report_bgwriter();
  
         if (FirstCallSinceLastCheckpoint())
         {
             /*
              * After any checkpoint, close all smgr files.  This is so we
              * won't hang onto smgr references to deleted files indefinitely.
              */
             smgrcloseall();
         }
  
         /*
          * Log a new xl_running_xacts every now and then so replication can
          * get into a consistent state faster (think of suboverflowed
          * snapshots) and clean up resources (locks, KnownXids*) more
          * frequently. The costs of this are relatively low, so doing it 4
          * times (LOG_SNAPSHOT_INTERVAL_MS) a minute seems fine.
          *
          * We assume the interval for writing xl_running_xacts is
          * significantly bigger than BgWriterDelay, so we don't complicate the
          * overall timeout handling but just assume we're going to get called
          * often enough even if hibernation mode is active. It's not that
          * important that LOG_SNAPSHOT_INTERVAL_MS is met strictly. To make
          * sure we're not waking the disk up unnecessarily on an idle system
          * we check whether there has been any WAL inserted since the last
          * time we've logged a running xacts.
          *
          * We do this logging in the bgwriter as it is the only process that
          * is run regularly and returns to its mainloop all the time. E.g.
          * Checkpointer, when active, is barely ever in its mainloop and thus
          * makes it hard to log regularly.
          */
         if (XLogStandbyInfoActive() && !RecoveryInProgress())
         {
             TimestampTz timeout = 0;
             TimestampTz now = GetCurrentTimestamp();
  
             timeout = TimestampTzPlusMilliseconds(last_snapshot_ts,
                                                   LOG_SNAPSHOT_INTERVAL_MS);
  
             /*
              * Only log if enough time has passed and interesting records have
              * been inserted since the last snapshot.  Have to compare with <=
              * instead of < because GetLastImportantRecPtr() points at the
              * start of a record, whereas last_snapshot_lsn points just past
              * the end of the record.
              */
             if (now >= timeout &&
                 last_snapshot_lsn <= GetLastImportantRecPtr())
             {
                 last_snapshot_lsn = LogStandbySnapshot();
                 last_snapshot_ts = now;
             }
         }
  
         /*
          * Sleep until we are signaled or BgWriterDelay has elapsed.
          *
          * Note: the feedback control loop in BgBufferSync() expects that we
          * will call it every BgWriterDelay msec.  While it's not critical for
          * correctness that that be exact, the feedback loop might misbehave
          * if we stray too far from that.  Hence, avoid loading this process
          * down with latch events that are likely to happen frequently during
          * normal operation.
          */
         rc = WaitLatch(MyLatch,
                        WL_LATCH_SET | WL_TIMEOUT | WL_EXIT_ON_PM_DEATH,
                        BgWriterDelay /* ms */ , WAIT_EVENT_BGWRITER_MAIN);
  
         /*
          * If no latch event and BgBufferSync says nothing's happening, extend
          * the sleep in "hibernation" mode, where we sleep for much longer
          * than bgwriter_delay says.  Fewer wakeups save electricity.  When a
          * backend starts using buffers again, it will wake us up by setting
          * our latch.  Because the extra sleep will persist only as long as no
          * buffer allocations happen, this should not distort the behavior of
          * BgBufferSync's control loop too badly; essentially, it will think
          * that the system-wide idle interval didn't exist.
          *
          * There is a race condition here, in that a backend might allocate a
          * buffer between the time BgBufferSync saw the alloc count as zero
          * and the time we call StrategyNotifyBgWriter.  While it's not
          * critical that we not hibernate anyway, we try to reduce the odds of
          * that by only hibernating when BgBufferSync says nothing's happening
          * for two consecutive cycles.  Also, we mitigate any possible
          * consequences of a missed wakeup by not hibernating forever.
          */
         if (rc == WL_TIMEOUT && can_hibernate && prev_hibernate)
         {
             /* Ask for notification at next buffer allocation */
             StrategyNotifyBgWriter(MyProc->pgprocno);
             /* Sleep ... */
             (void) WaitLatch(MyLatch,
                              WL_LATCH_SET | WL_TIMEOUT | WL_EXIT_ON_PM_DEATH,
                              BgWriterDelay * HIBERNATE_FACTOR,
                              WAIT_EVENT_BGWRITER_HIBERNATE);
             /* Reset the notification request in case we timed out */
             StrategyNotifyBgWriter(-1);
         }
  
         prev_hibernate = can_hibernate;
     }
 }