/*
* Write a WAL record containing a full image of a page.
*
* Caller should initialize the buffer and mark it dirty before calling this
* function. This function will set the page LSN.
*
* If the page follows the standard page layout, with a PageHeader and unused
* space between pd_lower and pd_upper, set 'page_std' to TRUE. That allows
* the unused space to be left out from the WAL record, making it smaller.
*/
XLogRecPtr
log_newpage_buffer(Buffer buffer, bool page_std)
{
Page page = BufferGetPage(buffer);
RelFileNode rnode;
ForkNumber forkNum;
BlockNumber blkno;
/* Shared buffers should be modified in a critical section. */
Assert(CritSectionCount > 0);
BufferGetTag(buffer, &rnode, &forkNum, &blkno);
return log_newpage(&rnode, forkNum, blkno, page, page_std);
}
/*
* Register a reference to a buffer with the WAL record being constructed.
* This must be called for every page that the WAL-logged operation modifies.
*/
void
XLogRegisterBuffer(uint8 block_id, Buffer buffer, uint8 flags)
{
registered_buffer *regbuf;
/* NO_IMAGE doesn't make sense with FORCE_IMAGE */
Assert(!((flags & REGBUF_FORCE_IMAGE) && (flags & (REGBUF_NO_IMAGE))));
Assert(begininsert_called);
if (block_id >= max_registered_block_id)
{
if (block_id >= max_registered_buffers)
elog(ERROR, "too many registered buffers");
max_registered_block_id = block_id + 1;
}
regbuf = ®istered_buffers[block_id];
BufferGetTag(buffer, ®buf->rnode, ®buf->forkno, ®buf->block);
regbuf->page = BufferGetPage(buffer);
regbuf->flags = flags;
regbuf->rdata_tail = (XLogRecData *) ®buf->rdata_head;
regbuf->rdata_len = 0;
/*
* Check that this page hasn't already been registered with some other
* block_id.
*/
#ifdef USE_ASSERT_CHECKING
{
int i;
for (i = 0; i < max_registered_block_id; i++)
{
registered_buffer *regbuf_old = ®istered_buffers[i];
if (i == block_id || !regbuf_old->in_use)
continue;
Assert(!RelFileNodeEquals(regbuf_old->rnode, regbuf->rnode) ||
regbuf_old->forkno != regbuf->forkno ||
regbuf_old->block != regbuf->block);
}
}
#endif
regbuf->in_use = true;
}
static void
ginRedoInsertEntry(Buffer buffer, bool isLeaf, BlockNumber rightblkno, void *rdata)
{
Page page = BufferGetPage(buffer);
ginxlogInsertEntry *data = (ginxlogInsertEntry *) rdata;
OffsetNumber offset = data->offset;
IndexTuple itup;
if (rightblkno != InvalidBlockNumber)
{
/* update link to right page after split */
Assert(!GinPageIsLeaf(page));
Assert(offset >= FirstOffsetNumber && offset <= PageGetMaxOffsetNumber(page));
itup = (IndexTuple) PageGetItem(page, PageGetItemId(page, offset));
GinSetDownlink(itup, rightblkno);
}
if (data->isDelete)
{
Assert(GinPageIsLeaf(page));
Assert(offset >= FirstOffsetNumber && offset <= PageGetMaxOffsetNumber(page));
PageIndexTupleDelete(page, offset);
}
itup = &data->tuple;
if (PageAddItem(page, (Item) itup, IndexTupleSize(itup), offset, false, false) == InvalidOffsetNumber)
{
RelFileNode node;
ForkNumber forknum;
BlockNumber blknum;
BufferGetTag(buffer, &node, &forknum, &blknum);
elog(ERROR, "failed to add item to index page in %u/%u/%u",
node.spcNode, node.dbNode, node.relNode);
}
}
/*
* Write a backup block if needed when we are setting a hint. Note that
* this may be called for a variety of page types, not just heaps.
*
* Callable while holding just share lock on the buffer content.
*
* We can't use the plain backup block mechanism since that relies on the
* Buffer being exclusively locked. Since some modifications (setting LSN, hint
* bits) are allowed in a sharelocked buffer that can lead to wal checksum
* failures. So instead we copy the page and insert the copied data as normal
* record data.
*
* We only need to do something if page has not yet been full page written in
* this checkpoint round. The LSN of the inserted wal record is returned if we
* had to write, InvalidXLogRecPtr otherwise.
*
* It is possible that multiple concurrent backends could attempt to write WAL
* records. In that case, multiple copies of the same block would be recorded
* in separate WAL records by different backends, though that is still OK from
* a correctness perspective.
*/
XLogRecPtr
XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
{
XLogRecPtr recptr = InvalidXLogRecPtr;
XLogRecPtr lsn;
XLogRecPtr RedoRecPtr;
/*
* Ensure no checkpoint can change our view of RedoRecPtr.
*/
Assert(MyPgXact->delayChkpt);
/*
* Update RedoRecPtr so that we can make the right decision
*/
RedoRecPtr = GetRedoRecPtr();
/*
* We assume page LSN is first data on *every* page that can be passed to
* XLogInsert, whether it has the standard page layout or not. Since we're
* only holding a share-lock on the page, we must take the buffer header
* lock when we look at the LSN.
*/
lsn = BufferGetLSNAtomic(buffer);
if (lsn <= RedoRecPtr)
{
int flags;
char copied_buffer[BLCKSZ];
char *origdata = (char *) BufferGetBlock(buffer);
RelFileNode rnode;
ForkNumber forkno;
BlockNumber blkno;
/*
* Copy buffer so we don't have to worry about concurrent hint bit or
* lsn updates. We assume pd_lower/upper cannot be changed without an
* exclusive lock, so the contents bkp are not racy.
*/
if (buffer_std)
{
/* Assume we can omit data between pd_lower and pd_upper */
Page page = BufferGetPage(buffer);
uint16 lower = ((PageHeader) page)->pd_lower;
uint16 upper = ((PageHeader) page)->pd_upper;
memcpy(copied_buffer, origdata, lower);
memcpy(copied_buffer + upper, origdata + upper, BLCKSZ - upper);
}
else
memcpy(copied_buffer, origdata, BLCKSZ);
XLogBeginInsert();
flags = REGBUF_FORCE_IMAGE;
if (buffer_std)
flags |= REGBUF_STANDARD;
BufferGetTag(buffer, &rnode, &forkno, &blkno);
XLogRegisterBlock(0, &rnode, forkno, blkno, copied_buffer, flags);
recptr = XLogInsert(RM_XLOG_ID, XLOG_FPI_FOR_HINT);
}
return recptr;
}
/*
* Searches for a slot with category at least minvalue.
* Returns slot number, or -1 if none found.
*
* The caller must hold at least a shared lock on the page, and this
* function can unlock and lock the page again in exclusive mode if it
* needs to be updated. exclusive_lock_held should be set to true if the
* caller is already holding an exclusive lock, to avoid extra work.
*
* If advancenext is false, fp_next_slot is set to point to the returned
* slot, and if it's true, to the slot after the returned slot.
*/
int
fsm_search_avail(Buffer buf, uint8 minvalue, bool advancenext,
bool exclusive_lock_held)
{
Page page = BufferGetPage(buf);
FSMPage fsmpage = (FSMPage) PageGetContents(page);
int nodeno;
int target;
uint16 slot;
restart:
/*
* Check the root first, and exit quickly if there's no leaf with enough
* free space
*/
if (fsmpage->fp_nodes[0] < minvalue)
return -1;
/*
* Start search using fp_next_slot. It's just a hint, so check that it's
* sane. (This also handles wrapping around when the prior call returned
* the last slot on the page.)
*/
target = fsmpage->fp_next_slot;
if (target < 0 || target >= LeafNodesPerPage)
target = 0;
target += NonLeafNodesPerPage;
/*----------
* Start the search from the target slot. At every step, move one
* node to the right, then climb up to the parent. Stop when we reach
* a node with enough free space (as we must, since the root has enough
* space).
*
* The idea is to gradually expand our "search triangle", that is, all
* nodes covered by the current node, and to be sure we search to the
* right from the start point. At the first step, only the target slot
* is examined. When we move up from a left child to its parent, we are
* adding the right-hand subtree of that parent to the search triangle.
* When we move right then up from a right child, we are dropping the
* current search triangle (which we know doesn't contain any suitable
* page) and instead looking at the next-larger-size triangle to its
* right. So we never look left from our original start point, and at
* each step the size of the search triangle doubles, ensuring it takes
* only log2(N) work to search N pages.
*
* The "move right" operation will wrap around if it hits the right edge
* of the tree, so the behavior is still good if we start near the right.
* Note also that the move-and-climb behavior ensures that we can't end
* up on one of the missing nodes at the right of the leaf level.
*
* For example, consider this tree:
*
* 7
* 7 6
* 5 7 6 5
* 4 5 5 7 2 6 5 2
* T
*
* Assume that the target node is the node indicated by the letter T,
* and we're searching for a node with value of 6 or higher. The search
* begins at T. At the first iteration, we move to the right, then to the
* parent, arriving at the rightmost 5. At the second iteration, we move
* to the right, wrapping around, then climb up, arriving at the 7 on the
* third level. 7 satisfies our search, so we descend down to the bottom,
* following the path of sevens. This is in fact the first suitable page
* to the right of (allowing for wraparound) our start point.
*----------
*/
nodeno = target;
while (nodeno > 0)
{
if (fsmpage->fp_nodes[nodeno] >= minvalue)
break;
/*
* Move to the right, wrapping around on same level if necessary, then
* climb up.
*/
nodeno = parentof(rightneighbor(nodeno));
}
/*
* We're now at a node with enough free space, somewhere in the middle of
* the tree. Descend to the bottom, following a path with enough free
* space, preferring to move left if there's a choice.
*/
while (nodeno < NonLeafNodesPerPage)
{
int childnodeno = leftchild(nodeno);
if (childnodeno < NodesPerPage &&
fsmpage->fp_nodes[childnodeno] >= minvalue)
{
nodeno = childnodeno;
continue;
}
childnodeno++; /* point to right child */
if (childnodeno < NodesPerPage &&
fsmpage->fp_nodes[childnodeno] >= minvalue)
{
nodeno = childnodeno;
}
else
{
/*
* Oops. The parent node promised that either left or right child
* has enough space, but neither actually did. This can happen in
* case of a "torn page", IOW if we crashed earlier while writing
* the page to disk, and only part of the page made it to disk.
*
* Fix the corruption and restart.
*/
RelFileNode rnode;
ForkNumber forknum;
BlockNumber blknum;
BufferGetTag(buf, &rnode, &forknum, &blknum);
elog(DEBUG1, "fixing corrupt FSM block %u, relation %u/%u/%u",
blknum, rnode.spcNode, rnode.dbNode, rnode.relNode);
/* make sure we hold an exclusive lock */
if (!exclusive_lock_held)
{
LockBuffer(buf, BUFFER_LOCK_UNLOCK);
LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
exclusive_lock_held = true;
}
fsm_rebuild_page(page);
MarkBufferDirty(buf);
goto restart;
}
}
/* We're now at the bottom level, at a node with enough space. */
slot = nodeno - NonLeafNodesPerPage;
/*
* Update the next-target pointer. Note that we do this even if we're only
* holding a shared lock, on the grounds that it's better to use a shared
* lock and get a garbled next pointer every now and then, than take the
* concurrency hit of an exclusive lock.
*
* Wrap-around is handled at the beginning of this function.
*/
fsmpage->fp_next_slot = slot + (advancenext ? 1 : 0);
return slot;
}