Move part of the code from lock_rec_print() in a separate function
buf_page_try_get() because the same functionality is needed in
INFORMATION SCHEMA code.
Approved by: Heikki
only after the buffer_cache has warmed up. During the initial warmup
phase these values are set to zero. We should recommend to move the
block to the start of the LRU in this case.
Also added some comments to buf0buf.c about the LRU logic.
reviewed by: Sunny
mtr_commit(): Do not acquire the log mutex if no log records are to be written.
ibuf_set_free_bits_func(): Do not clear mtr.modifications, because that would
prevent the dirty insert buffer bitmap block from being moved to the flush
list. Instead, assume that mtr_commit() will not acquire log_sys->mutex
because redo logging is disabled.
buf_page_init_for_read(): I/O-fix the block before buf_buddy_alloc(), because
the newly allocated block could otherwise be reused by the compressed page.
page_zip_max_ins_size(): New function.
btr_cur_optimistic_insert(), btr_cur_optimistic_delete(),
btr_page_split_and_insert(), btr_compress(): Do not update the
ibuf free bits for non-leaf pages or pages belonging to a clustered index.
The insert buffer only covers operations on leaf pages of secondary indexes.
For pages covered by the insert buffer, limit the max_ins_size to
page_zip_max_ins_size().
buf_page_get_gen(): Merge the insert buffer after decompressing the page.
buf_page_io_complete(): Relax the assertion about ibuf_count. For
compressed-only pages, the insert buffer merge takes place
in buf_page_get_gen().
ibuf_index_page_calc_free_bits(), ibuf_index_page_calc_free_from_bits(),
ibuf_index_page_calc_free(), ibuf_update_free_bits_if_full(),
ibuf_update_free_bits_low(), ibuf_update_free_bits_for_two_pages_low(),
ibuf_set_free_bits_low(): Add the parameter zip_size. Limit the maximum
insert size to page_zip_max_ins_size().
buf_buddy_alloc() until after the block has been found not to exist
in the buffer pool.
buf_buddy_alloc(), buf_buddy_alloc_low(): Make lru an output parameter.
Improve documentation.
buf_page_init_for_read(): Simplify the code. Check buf_page_hash_get()
again if buf_buddy_alloc() released buf_pool->mutex.
frame when reading compressed pages, unless crash recovery is in progress.
buf_page_read_low(): Adapt for buf_page_init_for_read() returning
buf_page_t* instead of buf_block_t*.
but in buf_page_get_gen(). Adjust comments. This prevents
unnecessary decompression in read-ahead when the compressed block
exists in the buffer pool.
If the block does not exist in the buffer pool, read-ahead will still
allocate an uncompressed page and decompress the block.
Move buf_block_init_low() and buf_zip_decompress() earlier in the file,
because some compilers are unable to inline functions that are defined
after the invocation.
buf_page_get_gen(): Decompress the block if needed.
buf_page_init_for_read(): Do not decompress.
of HASH_DELETE().
HASH_ASSERT_VALID(), HASH_INVALIDATE(): New macros, used in HASH_DELETE()
and HASH_SEARCH().
buf_page_init_low(): Add HASH_INVALIDATE(bpage, hash);
buf_page_init(): Call buf_page_init_low() before HASH_INSERT(), so that
bpage->hash will not be invalidated.
block->page.buf_fix_count. This mistake was made in r1255.
When access to the block is requested by some other thread,
buf_fix_count can be incremented during the execution of
buf_zip_decompress().
initialize all the fields of the control block.
buf_block_init_low(), buf_page_init_low(): New functions, used in
buf_page_init_for_backup_restore(), buf_page_init(), and
buf_page_init_for_read().
buf_buddy_relocate_block(): New function, split from buf_buddy_relocate().
Return FALSE when buf_page_can_relocate() returns FALSE.
Fix some formatting in buf_buddy_free_low() and buf_page_init_for_read().
uncompressed page frames.
buf_page_get_zip(), buf_page_release_zip(): New functions, used
by btr_copy_zblob_prefix().
btr_copy_zblob_prefix(): Do not call inflateEnd().
buf_LRU_free_block(): Do not allocate the control block when
a compressed-only page is being freed.
buf_validate(): Note that buf_pool->zip_clean may contain blocks
with io_fix == BUF_IO_READ.
buf_page_get_gen(). This saves one mutex operation per block request.
buf_page_get_gen(), various macros and functions: Add parameter zip_size.
btr_node_ptr_get_child(): Add parameter index.
fil_space_get_latch(): Add optional output parameter zip_size.
fil_space_get_zip_size(): Return 0 for space id==0, because the
system tablespace is never compressed.
fsp_header_init(): Remove the parameter zip_size.
ibuf_free_excess_pages(): Remove the parameter zip_size.
trx_rseg_t, trx_undo_t: Add field zip_size.
xdes_lst_get_next(): Remove, unused.
invoking buf_buddy_alloc(). An IO-fixed block must also be x-latched
when block->mutex and buf_pool->mutex are released.
buf_page_create(): Protect the block during buf_buddy_alloc() by
IO-fixing and X-latching it.
buf_zip_decompress(). This will prevent btr_cur_search_to_nth_level()
from accessing a half-decompressed leaf page. Internal B-tree nodes
(non-leaf pages) in B-tree searches are only protected by the tree
latch (index->lock) and by buffer-fixing.
buf_pool->page_hash and buf_pool->LRU before releasing buf_pool->mutex.
buf_page_init_for_read(), buf_page_create(): Allocate the compressed
page after the block has been added to the buffer pool. Document the
reason for this.
buf_page_init(): Declare block->frame valid only in the system tablespace.
trx_sysf_create(): Declare the rest of the page valid, although it is
uninitialized.
UNIV_MEM_FREE(): Declare a memory area free.
UNIV_MEM_ALLOC(): Declare a memory area allocated (but uninitialized).
UNIV_MEM_DESC(): Associate a memory area with a control block.
UNIV_MEM_UNDESC(): Unassociate a control block.
trx_sys_create_doublewrite_buf(): Clear the buffer with memset().
buf_page_init(): Add a bogus UNIV_MEM_VALID(block->frame) to silence
valid warnings about InnoDB data pages containing uninitialized data.
buf_LRU_get_free_only(): Add UNIV_MEM_ALLOC(block->frame).
buf_LRU_get_free_block(): Add UNIV_MEM_DESC(block->page.zip.data, block).
buf_LRU_free_block(): Add UNIV_MEM_DESC(b->zip.data, b) when allocating
a compressed-only control block for a compressed page.
buf_LRU_block_free_non_file_page(): Replace UNIV_MEM_INVALID() with
UNIV_MEM_FREE().
buf_LRU_block_remove_hashed_page(): Add UNIV_MEM_UNDESC(bpage) when
deallocating a compressed-only control block. Add
UNIV_MEM_INVALID(block->frame). (The frame should be flagged free
by buf_LRU_block_free_non_file_page() moments later.)
buf0buddy.c: Disable some extra checks in Valgrind-instrumented builds.
Add UNIV_MEM_VALID(), UNIV_MEM_INVALID(), UNIV_MEM_FREE(), UNIV_MEM_ALLOC()
as necessary.
buf_buddy_block_free(), buf_buddy_free_low(): Flag the freed bytes
uninitialized.
buf_page_init_for_read(), buf_buddy_relocate(): Do not refer to bpage
after invoking buf_relocate().
buf_buddy_relocate(): Flag src uninitialized.
during the buf_zip_decompress() call. Otherwise, the block may end
up being freed by another thread. This bug was caught by the
page_zip_validate() check in buf_LRU_block_remove_hashed_page().
buf_page_init_for_read(): Protect the buf_zip_decompress() call with
an X-lock on the block.
page_zip_validate(): Move the page_is_comp() assertion after the
page header comparison.
the buffer pool.
buf_LRU_free_block(): When preserving the compressed page, compute the
checksum. Also assert page_zip_validate().
buf_zip_decompress(): Add parameter "check" for enabling checksum tests.
Before decompressing a compressed-only page from the buffer pool, verify
the compressed page checksum.
buf_page_get_gen(): Note that buf_read_page() may uncompress
a compressed-only page.
buf_page_create(): Test buf_page_in_file() instead of BUF_BLOCK_FILE_PAGE.
and add compressed-only pages to buf_pool->page_hash.
BUF_POOL_ZIP_FOLD(), BUF_POOL_ZIP_FOLD_PTR(), BUF_POOL_ZIP_FOLD_BPAGE():
Macros for computing buf_pool->zip_hash fold values. Make use of these
in buf0buddy.c.
buf0buddy.c: Replace some ptr == ut_align_down(ptr, ...) in assertions with
!ut_align_offset(ptr, ...).
to BUF_BLOCK_ZIP_DIRTY (discarding the uncompressed page corresponding
to a modified compressed page that has not been flushed to disk).
buf_page_can_relocate(): New function, modelled after
buf_flush_ready_for_replace().
buf_LRU_free_block(): Allow the transition to BUF_BLOCK_ZIP_DIRTY.
buf_flush_insert_into_flush_list(): Make the prototype public.
buf_buddy_relocate(): Remove an unnecessary switch statement.
Use buf_page_can_relocate() instead of buf_flush_ready_for_replace().
buf_page_peek(): Made UNIV_INLINE.
Document UNIV_ZIP_DEBUG.
Improve diagnostics.
buf_buddy_relocate(): Do not assume all small blocks to contain only one
buf_page_t object. Require an exact size match, as in the case of
compressed pages.
buf_relocate(): Relocate buf_pool->LRU_old if needed.
Validate the LRU list in debug builds.
buf_LRU_block_remove_hashed_page(): Do not overwrite FIL_PAGE_OFFSET
or FIL_PAGE_ARCH_LOG_NO_OR_SPACE_ID. The memory will be invalidated
in debug builds of buf_buddy_free().
buf0buddy.c: Add assertions for not holding buf_pool->zip_mutex.
The calling thread should not be holding any block->mutex, but it might
be too time-consuming to assert against that, even #ifdef UNIV_SYNC_DEBUG.
There is something wrong with the management of compressed-only blocks
(BUF_BLOCK_ZIP_PAGE). To disable the creation of such blocks, set zip=TRUE
in buf_LRU_block_remove_hashed_page().
buf_LRU_block_remove_hashed_page(): Release buf_pool->zip_mutex when
freeing a compressed-only page and its control block, with buf_buddy_free().
Adapt callers.
buf_LRU_block_free_hashed_page(): Change the parameter type from buf_page_t*
to buf_block_t*.
buf_LRU_free_block(): Move below the definition of buf_LRU_add_block_low().
Allocate block descriptor for compressed-only blocks.
buf_LRU_block_free_non_file_page(): Replace ut_free() with buf_buddy_free().
buf_zip_decompress(): New function, split from buf_page_io_complete().
buf_page_init_for_read(): Do not allocate the compressed page until it is
really needed, to avoid the overhead of the buddy allocator. Decompress
compressed-only blocks that are needed again.
buf_page_create(): Do not allocate the compressed page until it is
really needed, to avoid the overhead of the buddy allocator.
buf_validate(): Replace some equality tests on the lengths of the LRU
list and the flush lists with greater-or-equal tests, since the counted
numbers do not include control blocks for compressed-only pages.
buf_page_get_gen(): Cache the result of fil_space_get_zip_size().
trx_commit_off_kernel(): Eliminate the flag must_flush_log. Initialize lsn = 0
to signify must_flush_log == FALSE.
log_flush_margin(): Eliminate the flag do_flush. Initialize lsn = 0
to signify do_flush == FALSE.
buf_LRU_block_remove_hashed_page(): Return the new state of the block.
Only call buf_LRU_block_free_hashed_page()
if buf_LRU_block_remove_hashed_page() did not return BUF_BLOCK_ZIP_FREE,
that is, the control block was not freed.
buf_LRU_insert_zip_clean(): New function, for inserting a compressed-only
block into buf_pool->zip_clean in the LRU order.
buf_LRU_block_remove_hashed_page(), buf_LRU_free_block():
Add the flag "zip" for controlling if the compressed page of an uncompressed
page should be removed. For now, assume zip==TRUE.
buf_LRU_get_free_block(): Replace the test for UT_LIST_GET_LEN(buf_pool->free)
with a test for the return value of buf_LRU_get_free_only(). Do not
free zip.data, as it must already have been freed.
buf_flush_insert_into_flush_list(), buf_flush_insert_sorted_into_flush_list():
Remove compressed-only blocks from the buf_pool->zip_clean list.
buf_flush_remove(): Restore compressed-only blocks to
the buf_pool->zip_clean list.
buf_page_init_for_read(): Uncompress compressed-only blocks when possible.
Currently, there cannot be any compressed-only blocks in the buffer pool;
they would be inserted by buf_LRU_free_block(bpage, zip=FALSE).
