Re: Throwing away unnecessary spin-locks - Mailing list pgsql-hackers

Trying to find your snippet, is this the one?

  1262  ReserveXLogSwitch(XLogRecPtr *StartPos, XLogRecPtr *EndPos, XLogRecPtr *PrevPtr)
  1263  {
  1278          SpinLockAcquire(&Insert->insertpos_lck);
  1280          startbytepos = Insert->CurrBytePos;
  1282          ptr = XLogBytePosToEndRecPtr(startbytepos);
  1290          endbytepos = startbytepos + size;

  1303          Insert->CurrBytePos = endbytepos;
  1306          SpinLockRelease(&Insert->insertpos_lck);

This illustrates well the cases where things can't be written without a lock.

atomically. Because changes between line 1280 and line 1303 are rolled back.

However, this particular example doesn't stop us from reading because it will either

CurBytePos before line 1303, and that is the same as reading under a lock before that

block, or after the line 1303 and that is equivalent to reading under a lock after that block.