Class B2IRowLocking3

java.lang.Object
org.apache.derby.impl.store.access.btree.index.B2IRowLocking3
All Implemented Interfaces:
BTreeLockingPolicy
Direct Known Subclasses:
B2IRowLockingRR

class B2IRowLocking3 extends Object implements BTreeLockingPolicy
Implements the jdbc serializable isolation level using row locks.

Holds read and write locks until end of transaction. Obtains previous key locks to protect from phantom reads.

  • Field Details

    • base_cc

      protected ConglomerateController base_cc
      The container id of the base container for this index. Used to build record handles to make lock calls on.
    • open_btree

      protected OpenBTree open_btree
      The OpenBtree to use if we have to lock anything in the btree vs. base row locking.
    • rawtran

      private Transaction rawtran
      The transaction to associate lock requests with.
  • Constructor Details

  • Method Details

    • lockPreviousToFirstKey

      private boolean lockPreviousToFirstKey(LeafControlRow current_leaf, LeafControlRow aux_leaf, int lock_operation, int lock_duration) throws StandardException
      Lock key previous to first key in btree.

      In the previous key locking protocol repeatable read and phantom protection is guaranteed by locking a range of keys in the btree. The range is defined by the key previous to the first key you look at and all subsequent keys you look at. The first key in the index is a special case, as there are no keys previous to it. In that case a special key is declared the "previous key" to the first key in the btree and is locked instead.

      In this implementation that first key is defined to be in the base container, page ContainerHandle.FIRST_PAGE_NUMBER, record id PREVIOUS_KEY_HANDLE.

      Note that the previous key is the same for all indexes on a given conglomerate. It seemed better for all locks on a base table to have the same containerid, rather than having some locks generated from a btree have a containerid from base table and some having a containerid from the btree. If this turns out to be a problem we could either have 2 different containerid's, be more creative with the record id, or even add more to the lock key.

      Parameters:
      aux_leaf - If non-null, this leaf is unlatched if the routine has to wait on the lock.
      lock_operation - Whether to lock exclusive or share.
      lock_duration - For what duration should the lock be held, if INSTANT_DURATION, then the routine will guarantee that lock was acquired while holding the latch, but then immediately release the lock. If COMMIT_DURATION or MANUAL_DURATION then the lock be held when routine returns successfully.
      Throws:
      StandardException - Standard exception policy.
    • lockRowOnPage

      private boolean lockRowOnPage(LeafControlRow current_leaf, LeafControlRow aux_leaf, int current_slot, BTreeRowPosition position, FetchDescriptor lock_fetch_desc, DataValueDescriptor[] lock_template, RowLocation lock_row_loc, int lock_operation, int lock_duration) throws StandardException
      Lock a btree row (row is at given slot in page).

      Lock the row at the given slot in the page. Meant to be used if caller only has the slot on the page to be locked, and has not read the row yet. This routine fetches the row location field from the page, and then locks that rowlocation in the base container.

      Lock a btree row, enforcing the standard lock/latch protocol. On return the row is locked. Return status indicates if the lock was waited for, which will mean a latch was dropped while waiting. In general a false status means that the caller will either have to research the tree unless some protocol has been implemented that insures that the row will not have moved while the latch was dropped.

      This routine request a row lock NOWAIT on the in-memory row "current_row.". If the lock is granted the routine will return true. If the lock cannot be granted NOWAIT, then the routine will release the latch on "current_leaf" and "aux_leaf" (if aux_leaf is non-null), and then it will request a WAIT lock on the row.

      Parameters:
      current_leaf - Latched current leaf where "current" key is.
      aux_leaf - If non-null, this leaf is unlatched if the routine has to wait on the lock.
      current_slot - Slot of row to lock.
      position - The position to lock if the lock is requested while performing a scan, null otherwise.
      lock_fetch_desc - Descriptor for fetching just the RowLocation, used for locking.
      lock_operation - Whether lock is for key prev to insert or not.
      lock_duration - For what duration should the lock be held, if INSTANT_DURATION, then the routine will guarantee that lock was acquired while holding the latch, but then immediately release the lock. If COMMIT_DURATION or MANUAL_DURATION then the lock be held when routine returns successfully.
      Throws:
      StandardException - Standard exception policy.
    • searchLeftAndLockPreviousKey

      private boolean searchLeftAndLockPreviousKey(LeafControlRow current_leaf, FetchDescriptor lock_fetch_desc, DataValueDescriptor[] lock_template, RowLocation lock_row_loc, OpenBTree open_btree, int lock_operation, int lock_duration) throws StandardException
      move left in btree and lock previous key.

      Enter routine with "current_leaf" latched. This routine implements the left travel ladder locking protocol to search the leaf pages from right to left for the previous key to 1st key on current_leaf. There are 2 cases: 1) the previous page has keys, in which case the last key on that page is locked, other wise search continues on the next page to the left. 2) there are no keys on the current page and there is no page to the left. In this case the special "leftmost key" lock is gotten by calling lockPreviousToFirstKey(). Left laddar locking is used if all latches can be obtained immediately with NOWAIT. This means that current latch is held while asking for left latch NOWAIT, and if left latch is granted then subsequently current latch can be released. If this protocol is followed and all latches are granted then caller is guaranteed that the correct previous key has been locked and current_page latch remains. The NOWAIT protocol is used to avoid latch/latch deadlocks. The overall protocol is that one never holds a latch while waiting on another unless the direction of travel is down and to the right.

      If along the search a latch has to be waited on then latches are released and a wait is performed, and "false" status is returned to caller. In this case the routine can no longer be sure of it's current position and may have to retry the whole operation.

      Returns:
      true if previous key found without ever waiting on a latch, false if latch released in order to wait for other latch.
      Throws:
      StandardException - Standard exception policy.
    • _lockScanRow

      protected boolean _lockScanRow(OpenBTree open_btree, BTreeRowPosition pos, boolean request_row_lock, FetchDescriptor lock_fetch_desc, DataValueDescriptor[] lock_template, RowLocation lock_row_loc, boolean previous_key_lock, boolean forUpdate, int lock_operation) throws StandardException
      Lock a row as part of doing the scan.

      Lock the row at the given slot (or the previous row if slot is 0).

      If this routine returns true all locks were acquired while maintaining the latch on leaf. If this routine returns false, locks may or may not have been acquired, and the routine should be called again after the client has researched the tree to reget the latch on the appropriate page.

      Parameters:
      open_btree - The open_btree to associate latches with - used if routine has to scan backward.
      pos - The position of the row to lock.
      request_row_lock - Whether to request the row lock, should only be requested once per page in the scan.
      lock_fetch_desc - The fetch descriptor to use to fetch the row location for the lock request.
      lock_template - A scratch area to use to read in rows.
      previous_key_lock - Is this a previous key lock call?
      forUpdate - Is the scan for update or for read only.
      Returns:
      Whether locks were acquired without releasing latch on leaf.
      Throws:
      StandardException - Standard exception policy.
    • lockScanCommittedDeletedRow

      public boolean lockScanCommittedDeletedRow(OpenBTree open_btree, LeafControlRow leaf, DataValueDescriptor[] template, FetchDescriptor lock_fetch_desc, int slot_no) throws StandardException
      Lock a btree row to determine if it is a committed deleted row.

      Specified by:
      lockScanCommittedDeletedRow in interface BTreeLockingPolicy
      Parameters:
      open_btree - The conglomerate we are locking.
      leaf - The leaf page with the row to lock.
      template - Empty full template row, to read row into.
      slot_no - The slot of row on "current_leaf"
      Throws:
      StandardException - Standard exception policy.
      See Also:
    • lockScanRow

      public boolean lockScanRow(OpenBTree open_btree, BTreeRowPosition pos, FetchDescriptor lock_fetch_desc, DataValueDescriptor[] lock_template, RowLocation lock_row_loc, boolean previous_key_lock, boolean forUpdate, int lock_operation) throws StandardException
      Lock a row as part of doing the scan.

      Lock the row at the given slot (or the previous row if slot is 0).

      If this routine returns true all locks were acquired while maintaining the latch on leaf. If this routine returns false, locks may or may not have been acquired, and the routine should be called again after the client has researched the tree to reget the latch on the appropriate page.

      Specified by:
      lockScanRow in interface BTreeLockingPolicy
      Parameters:
      open_btree - The open_btree to associate latches with - used if routine has to scan backward.
      pos - The position of the row to lock.
      lock_template - A scratch area to use to read in rows.
      previous_key_lock - Is this a previous key lock call?
      forUpdate - Is the scan for update or for read only.
      lock_operation - For what operation are we requesting the lock, this should be one of the following 4 options: LOCK_READ [read lock], (LOCK_INS | LOCK_UPD) [ lock for insert], (LOCK_INSERT_PREVKEY | LOCK_UPD) [lock for previous key to insert], (LOCK_UPD) [lock for delete or replace]
      Returns:
      Whether locks were acquired without releasing latch on leaf.
      Throws:
      StandardException - Standard exception policy.
    • unlockScanRecordAfterRead

      public void unlockScanRecordAfterRead(BTreeRowPosition pos, boolean forUpdate) throws StandardException
      Release read lock on a row. For serializable, there is no work to do.
      Specified by:
      unlockScanRecordAfterRead in interface BTreeLockingPolicy
      Parameters:
      pos - Data structure that defines the current position in the scan to be unlocked.
      forUpdate - Is the scan for update or for read only.
      Throws:
      StandardException - Standard exception policy.
    • lockNonScanPreviousRow

      public boolean lockNonScanPreviousRow(LeafControlRow current_leaf, int current_slot, FetchDescriptor lock_fetch_desc, DataValueDescriptor[] lock_template, RowLocation lock_row_loc, OpenBTree open_btree, int lock_operation, int lock_duration) throws StandardException
      Lock the row previous to the input row.

      See BTreeLockingPolicy.lockNonScanPreviousRow

      Specified by:
      lockNonScanPreviousRow in interface BTreeLockingPolicy
      Parameters:
      current_leaf - Latched current leaf where "current" key is.
      current_slot - The slot of row on "current_leaf"
      lock_template - Empty full template row, to read row into.
      open_btree - The open_btree to associate latches with - used if routine has to scan backward.
      lock_operation - For what operation are we requesting the lock, this should be one of the following 4 options: LOCK_READ [read lock], (LOCK_INS | LOCK_UPD) [ lock for insert], (LOCK_INSERT_PREVKEY | LOCK_UPD) [lock for previous key to insert], (LOCK_UPD) [lock for delete or replace]
      lock_duration - For what duration should the lock be held, if INSTANT_DURATION, then the routine will guarantee that lock was acquired while holding the latch, but then immediately release the lock. If COMMIT_DURATION or MANUAL_DURATION then the lock be held when routine returns successfully.
      Throws:
      StandardException - Standard exception policy.
    • lockNonScanRow

      public boolean lockNonScanRow(BTree btree, LeafControlRow current_leaf, LeafControlRow aux_leaf, DataValueDescriptor[] current_row, int lock_operation) throws StandardException
      Lock the in memory row.

      See BTree.lockRow() for more info.

      Specified by:
      lockNonScanRow in interface BTreeLockingPolicy
      Parameters:
      btree - The conglomerate we are locking.
      current_leaf - If non-null, this leaf is unlatched if the routine has to wait on the lock.
      aux_leaf - If non-null, this leaf is unlatched if the routine has to wait on the lock.
      current_row - In memory, objectified "current" row.
      lock_operation - For what operation are we requesting the lock, this should be one of the following 4 options: LOCK_READ [read lock], (LOCK_INS | LOCK_UPD) [ lock for insert], (LOCK_INSERT_PREVKEY | LOCK_UPD) [lock for previous key to insert], (LOCK_UPD) [lock for delete or replace]
      Throws:
      StandardException - Standard exception policy.
    • lockNonScanRowOnPage

      public boolean lockNonScanRowOnPage(LeafControlRow current_leaf, int current_slot, FetchDescriptor lock_fetch_desc, DataValueDescriptor[] lock_template, RowLocation lock_row_loc, int lock_operation) throws StandardException
      Description copied from interface: BTreeLockingPolicy
      Lock the row at the given slot.

      If this routine returns true all locks were acquired while maintaining the latch on leaf. If this routine returns false, locks may or may not have been acquired, and the routine should be called again after the client has researched the tree to reget the latch on the appropriate page.

      Specified by:
      lockNonScanRowOnPage in interface BTreeLockingPolicy
      Parameters:
      current_leaf - The control row of the current leaf to lock.
      current_slot - The slot position of the row to lock.
      lock_template - A scratch area to use to read in rows.
      lock_operation - For what operation are we requesting the lock, this should be one of the following 4 options: LOCK_READ [read lock], (LOCK_INS | LOCK_UPD) [ lock for insert], (LOCK_INSERT_PREVKEY | LOCK_UPD) [lock for previous key to insert], (LOCK_UPD) [lock for delete or replace]
      Returns:
      Whether locks were acquired without releasing latch on leaf.
      Throws:
      StandardException - Standard exception policy.