/*--- This software is copyrighted by the Regents of the University of California, and other parties. The following terms apply to all files associated with the software unless explicitly disclaimed in individual files. The authors hereby grant permission to use, copy, modify, distribute, and license this software and its documentation for any purpose, provided that existing copyright notices are retained in all copies and that this notice is included verbatim in any distributions. No written agreement, license, or royalty fee is required for any of the authorized uses. Modifications to this software may be copyrighted by their authors and need not follow the licensing terms described here, provided that the new terms are clearly indicated on the first page of each file where they apply. IN NO EVENT SHALL THE AUTHORS OR DISTRIBUTORS BE LIABLE TO ANY PARTY FOR DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OF THIS SOFTWARE, ITS DOCUMENTATION, OR ANY DERIVATIVES THEREOF, EVEN IF THE AUTHORS HAVE BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. THE AUTHORS AND DISTRIBUTORS SPECIFICALLY DISCLAIM ANY WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, AND NON-INFRINGEMENT. THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, AND THE AUTHORS AND DISTRIBUTORS HAVE NO OBLIGATION TO PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS. GOVERNMENT USE: If you are acquiring this software on behalf of the U.S. government, the Government shall have only "Restricted Rights" in the software and related documentation as defined in the Federal Acquisition Regulations (FARs) in Clause 52.227.19 (c) (2). If you are acquiring the software on behalf of the Department of Defense, the software shall be classified as "Commercial Computer Software" and the Government shall have only "Restricted Rights" as defined in Clause 252.227-7013 (c) (1) of DFARs. Notwithstanding the foregoing, the authors grant the U.S. Government and others acting in its behalf permission to use and distribute the software in accordance with the terms specified in this license. ---*/ /** * @file * * interface for dealing with slotted pages * * This file provides a re-entrant interface for pages that contain * variable-size records. * * @ingroup LLADD_CORE * $Id$ * * @todo The slotted pages implementation, and the rest of the page * structure should be seperated, and each page should have a 'type' * slot so that we can implement multiple page types on top of LLADD. STRUCTURE OF A GENERIC PAGE
 +----------------------------------------------------------------------+
 |                                                                      |
 |  USABLE SPACE                                                        |
 |                                                                      |
 |                                                                      |
 |                                                                      |
 |                                                                      |
 |                                                                      |
 |                                                                      |
 |                                                                      |
 |                                                                      |
 |                                                                      |
 |                                                                      |
 |                                                                      |
 |                                                                      |
 |                                                    +-----------+-----+
 |                                                    | page type | LSN |
 +----------------------------------------------------+-----------+-----+
*/ #ifndef __PAGE_H__ #define __PAGE_H__ #include #include #include "latches.h" /** @todo page.h includes things that it shouldn't, and page.h should eventually be an installed header. */ #include #include BEGIN_C_DECLS #define UNINITIALIZED_PAGE 0 #define SLOTTED_PAGE 1 #define INDIRECT_PAGE 2 #define lsn_ptr(page) (((lsn_t *)(&((page)->memAddr[PAGE_SIZE])))-1) #define page_type_ptr(page) (((int*)lsn_ptr((page)))-1) #define end_of_usable_space_ptr(page) page_type_ptr((page)) #define shorts_from_end(page, count) (((short*)end_of_usable_space_ptr((page)))-(count)) #define bytes_from_start(page, count) (((byte*)((page)->memAddr))+(count)) #define ints_from_start(page, count) (((int*)((page)->memAddr))+(count)) #define USABLE_SIZE_OF_PAGE (PAGE_SIZE - sizeof(lsn_t) - sizeof(int)) /*#define invalidateSlot(page, n) (*slot_ptr((page), (n)) = INVALID_SLOT)*/ /** The page type contains in-memory information about pages. This information is used by LLADD to track the page while it is in memory, and is never written to disk. In particular, our current page replacement policy requires two doubly linked lists, @todo The Page struct should be tuned for better memory utilization. */ struct Page_s { /** @todo Shouldn't Page.id be a long? */ int id; /** @todo The Page.LSN field seems extraneous. Why do we need it? */ long LSN; byte *memAddr; /** @todo dirty pages currently aren't marked dirty! */ int dirty; /** The next item in the replacement policy's queue */ struct Page_s *next; /** The previous item in the replacement policy's queue. */ struct Page_s *prev; /** Which queue is the page in? */ int queue; /** Is the page in the cache at all? */ int inCache; /** Used for page-level latching. Each page has an associated read/write lock. This lock only protects the internal layout of the page, and the members of the page struct. Here is how rwlatch is held in various circumstances: Record allocation: Write lock Record read: Read lock Read LSN Read lock Record write *READ LOCK* Write LSN Write lock Write page to disk No lock Read page from disk No lock Any circumstance where one these locks are held during an I/O operation is a bug. For the 'no lock' cases, see @loadlatch */ rwl * rwlatch; /** Since the bufferManager re-uses page structs, this lock is used to ensure that the page is in one of two consistent states, depending on whether a read lock or a write lock is being held. If a read lock is held, then the page is managed by the rwlatch also defined in this struct. Therefore, it cannot be read from or written to disk. Furthermore, since we do not impose an order on operations, the holder of a readlock may not use the lsn field to determine whether a particular operation has completed on the page. The write lock is used to block all writers (other than the one holding the page), and to ensure that all updates with lsn less than or equal to the page's lsn have been applied. Therefore, threads that write the page to disk must hold this lock. Since it precludes access by all other threads, a write lock also allows the holder to evict the current page, and replace it. Examples: Write page to disk Write lock Read page from disk Write lock Allocate a new record Read lock Write to a record Read lock Read from a record Read lock @see rwlatch, getPage(), pageRalloc(), pageRead() */ rwl * loadlatch; }; /** * initializes all the important variables needed in all the * functions dealing with pages. */ void pageInit(); void pageDeInit(); /** * assumes that the page is already loaded in memory. It takes * as a parameter a Page. The Page struct contains the new LSN and the page * number to which the new LSN must be written to. */ void pageWriteLSN(Page * page, lsn_t lsn); /** * assumes that the page is already loaded in memory. It takes * as a parameter a Page and returns the LSN that is currently written on that * page in memory. */ lsn_t pageReadLSN(const Page * page); /** * @param xid transaction id @param lsn the lsn that the updated * record will reflect. This is needed by recovery, and undo. (The * lsn of a page must always increase. Undos are handled by passing * in the LSN of the CLR that records the undo.) * * @param rid recordid where you want to write @param dat data you * wish to write */ void writeRecord(int xid, Page * page, lsn_t lsn, recordid rid, const void *dat); /** * @param xid transaction ID * @param rid * @param dat buffer for data */ void readRecord(int xid, Page * page, recordid rid, void *dat); /** * allocate a record. This must be done in two phases. The first * phase reserves a slot, and produces a log entry. The second phase * sets up the slot according to the contents of the log entry. * * Ralloc implements the first phase. * * @param xid The active transaction. * @param size The size of the new record * @return allocated record * * @see slotRalloc the implementation of the second phase. */ recordid ralloc(int xid, long size); /** * assumes that the page is already loaded in memory. It takes as * parameters a Page and the size in bytes of the new record. pageRalloc() * returns a recordid representing the newly allocated record. * * If you call this function, you probably need to be holding lastFreepage_mutex. * * @see lastFreepage_mutex * * NOTE: might want to pad records to be multiple of words in length, or, simply * make sure all records start word aligned, but not necessarily having * a length that is a multiple of words. (Since Tread(), Twrite() ultimately * call memcpy(), this shouldn't be an issue) * * NOTE: pageRalloc() assumes that the caller already made sure that sufficient * amount of freespace exists in this page. (@see freespace()) * * @todo Makes no attempt to reuse old recordid's. */ recordid pageRalloc(Page * page, int size); recordid pageSlotRalloc(Page * page, lsn_t lsn, recordid rid); void pageDeRalloc(Page * page, recordid rid); void pageCommit(int xid); void pageAbort(int xid); Page* pageAlloc(int id); void pageRealloc(Page * p, int id); /** Allocates a set of contiguous pages on disk. Has nothing to do with pageAlloc. @todo need a better naming convention for pageAlloc (alloc's memory) and pageAllocMultiple (alloc's disk) @todo is there any case where this function can safely be called with newPageCount > 1? */ int pageAllocMultiple(int newPageCount) ; int pageGetSlotType(Page * p, int slot, int type); void pageSetSlotType(Page * p, int slot, int type); END_C_DECLS #endif