INNODB_COMPRESSION_BUDDY and INNODB_COMPRESSION_BUDDY_RESET.
buf_buddy_stat_struct, buf_buddy_stat_t, buf_buddy_stat[]:
Statistics of the buddy system grouped by block size.
i_s_innodb_compression_buddy, i_s_innodb_compression_buddy_reset:
New INFORMATION_SCHEMA plugins.
i_s_compression_buddy_fields_info[]: Define the fields:
size, used, free, relocated, relocated_sec.
i_s_compression_buddy_fill_low(), i_s_compression_buddy_fill(),
i_s_compression_buddy_reset_fill(): Fill the fields.
i_s_compression_buddy_init(), i_s_compression_buddy_reset_init():
Initialize the tables.
blocks that contains uncompressed and compressed frames. This patch was
designed by Heikki and Inaam, implemented by Inaam, and refined and reviewed
by Marko and Sunny.
buf_buddy_n_frames, buf_buddy_min_n_frames, buf_buddy_max_n_frames: Remove.
buf_page_belongs_to_unzip_LRU(): New predicate:
bpage->zip.data && buf_page_get_state(bpage) == BUF_BLOCK_FILE_PAGE.
buf_pool_t, buf_block_t: Add the linked list unzip_LRU. A block in the
regular LRU list is in unzip_LRU iff buf_page_belongs_to_unzip_LRU() holds.
buf_LRU_free_block(): Add a third return value to refine the case
"cannot free the block".
buf_LRU_search_and_free_block(): Update the documentation to reflect the
implementation.
buf_LRU_stat_t, buf_LRU_stat_cur, buf_LRU_stat_sum, buf_LRU_stat_arr[]:
Statistics for the unzip_LRU algorithm.
buf_LRU_stat_update(): New function: Update the statistics. Called once
per second by srv_error_monitor_thread().
buf_LRU_validate(): Validate the unzip_LRU list as well.
buf_LRU_evict_from_unzip_LRU(): New predicate: Use the unzip_LRU before
falling back to the regular LRU?
buf_LRU_free_from_unzip_LRU_list(), buf_LRU_free_from_common_LRU_list():
Subfunctions of buf_LRU_search_and_free_block().
buf_LRU_search_and_free_block(): Reimplement. Try to evict an uncompressed
page from the unzip_LRU list before falling back to evicting an entire block
from the common LRU list.
buf_unzip_LRU_remove_block_if_needed(): New function.
buf_unzip_LRU_add_block(): New function: Add a block to the unzip_LRU list.
buf_buddy_relocated_duration[],
page_zip_compress_duration[]
page_zip_decompress_duration[]: Record the total duration of the operations.
buf_buddy_relocate(), page_zip_compress(), page_zip_decompress():
Add ut_time_us() instrumentation.
i_s_zip_fields_info[], i_s_zip_fill_low(): Move the columns containing
cumulated statistics last. Add relocated_usec, compressed_usec, and
decompressed_usec.
symbols. Use it for all definitions of non-static variables and functions.
lexyy.c, make_flex.sh: Declare yylex as UNIV_INTERN, not static. It is
referenced from pars0grm.c.
Actually, according to
nm .libs/ha_innodb.so|grep -w '[ABCE-TVXYZ]'
the following symbols are still global:
* The vtable for class ha_innodb
* pars0grm.c: The function yyparse() and the variables yychar, yylval, yynerrs
The required changes to the Bison-generated file pars0grm.c will be addressed
in a separate commit, which will add a script similar to make_flex.sh.
The class ha_innodb is renamed from class ha_innobase by a #define. Thus,
there will be no clash with the builtin InnoDB. However, there will be some
overhead for invoking virtual methods of class ha_innodb. Ideas for making
the vtable hidden are welcome. -fvisibility=hidden is not available in GCC 3.
buf_pool->mutex: Rename to buf_pool_mutex, so that the wrappers will have
to be used when changes are merged from other source trees.
buf_pool->zip_mutex: Rename to buf_pool_zip_mutex.
buf_pool_mutex_own(), buf_pool_mutex_enter(), buf_pool_mutex_exit():
Wrappers for buf_pool_mutex.
buf_buddy_min_n_frames into two, also buf_buddy_max_n_frames.
Set the default values in such a way that a strict LRU policy will apply
for replacing compressed or uncompressed page frames in the buffer pool.
These parameters have not yet been exposed to the MySQL layer.
allocating compressed page frames or their control blocks. Also note
that if buf_buddy_alloc() is used for allocating a control block,
it must be initialized before releasing buf_pool->mutex.
buf_page_init_for_read(): When the page hash check fails after
buf_buddy_alloc(), free the uninitialized control block before freeing
the compressed page frame. This fixes a potential error in
buf_buddy_relocate_block().
mutex is temporarily released.
buf_LRU_free_block(), buf_buddy_alloc_clean(): Add an output parameter that
will be assigned TRUE when the buffer pool mutex is released.
This bug was spotted by and fix provided by Sunny.
buf_buddy_relocate(): Allow the source block to contain uninitialized data.
buf0buddy.c: Replace the remaining VALGRIND_CHECK_ macros with the wrappers
defined in univ.i.
buf_LRU_free_block(): Check that the block descriptor contains valid data.
buf_buddy_relocate(): Check that the source block contains valid data.
buf_page_get_gen(): Do not dereference bpage after calling buf_relocate().
This avoids a bogus Valgrind warning; the memory itself was valid.
buf_page_hash_get(): Check that the returned block descriptor contains
valid data.
UNIV_MEM_ASSERT_RW(): New macro, to check that the contents of a memory
area is defined.
UNIV_MEM_ASSERT_W(): New macro, to check that a memory area is writable.
UNIV_MEM_ASSERT_AND_FREE(): New macro, to check that the memory is
writable before declaring it free (unwritable). This replaces UNIV_MEM_FREE()
in many places.
mem_init_buf(): Declare the memory undefined.
mem_erase_buf(): Declare the memory freed.
C++ requires explicit type casts when converting from void*, and *.ic files
are now inlined in ha_innodb.cc.
fil_space_get_by_id(), fil_space_get_by_name(): New inline functions, to
replace occurrences of the HASH_SEARCH() macro in fil0fil.c. This should
improve readability.
for displaying buddy allocator statistics.
buf_buddy_used[], buf_buddy_relocated[]: Declare as global symbols.
mysql_declare_plugin(innobase): Add MYSQL_INFORMATION_SCHEMA_PLUGIN
for innobase_is_buddy ("INNODB_BUDDY").
buf_buddy_n_frames: Number of buffer frames allocated for the buddy system
buf_buddy_used[]: Counts of allocated blocks, by buddy size.
buf_buddy_relocated[]: Counts of blocks relocated, bu buddy size.
buf_buddy_alloc_clean(): Note the added configuration parameter
buf_buddy_min_n_frames: Preferred minimum buf_buddy_n_frames
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.
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().
buf_buddy_relocate(): Document that src may have been split and some of
it may be in buf_pool->zip_free[j] for some j<i, thus mach_read_from_4()
may be reading from free memory. Check buffer pages only for addressability,
not for being completely initialized. Ensure that block descriptors are
completely initialized.
buf_buddy_remove_from_free(): Check prev->state and next->state in
UNIV_DEBUG && UNIV_DEBUG_VALGRIND builds.
buf_buddy_free_low(): Flag the entire block valid prior to invoking
buf_buddy_remove_from_free() in order to avoid a Valgrind warning about
the debug assertion on bpage->state. Remove the Valgrind check for
buddy being completely allocated. If it has been split to smaller blocks,
some of those blocks may be free.
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.
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.
buf_buddy_get(): New function in buf0buddy.c, to replace the macro
in buf0buddy.h.
buf_buddy_get_offset(): Remove.
buf_buddy_get_slot(): Remove the prototype from buf0buddy.h. This function
is for internal use in buf0buddy.ic.
buf_buddy_block_free(): Remove the block from buf_pool->zip_hash.
buf_buddy_alloc_clean(): Test for buf_page_in_file() before
invoking buf_LRU_free_block(). The LRU list also contains "half-freed"
blocks (state == BUF_BLOCK_REMOVE_HASH).
twice the maximum block size of the buddy system.
buf_page_t: Note that state may change from BUF_BLOCK_READY_FOR_USE
to BUF_BLOCK_MEMORY without buf_page_get_mutex() protection
[only buf_pool->mutex].
buf_buddy_get_slot(): Extend the output to BUF_BUDDY_SIZES.
buf_buddy_alloc(), buf_buddy_free(): Allow sizes up to UNIV_PAGE_SIZE.
buf_buddy_alloc_low(), buf_buddy_free_low(), buf_buddy_alloc_clean():
Allow i==BUF_BUDDY_SIZES.
buf_buddy_free(), buf_buddy_free_low(): Document that the data must
not be pointed to by the buffer pool. Add ut_ad(!buf_pool_contains_zip())
assertions to catch pointers to freed blocks.
Validate the zip_free[] lists #ifdef UNIV_DEBUG.
buf_buddy_relocate(): Ensure that the size of the compressed page
matches. The buddy block can be split, and the control block can be
pointing to a smaller compressed page than the one whose relocation
is being attempted.
buf_buddy_alloc_zip() [UNIV_DEBUG]: Byte-fill the allocated blocks with ~i.
buf_buddy_block_free() [UNIV_DEBUG]: Clear the frame before releasing it
to the buffer pool.
buf_buddy_alloc_from() [UNIV_DEBUG]: Byte-fill the free blocks with j.
buf_buddy_relocate(): Allow the buf_page_hash_get() lookup to fail,
and note the reason.
buf_buddy_free_low(): Correctly compute the address of the combined free block.
buf_page_struct: Clarify the mutex protection rules. Note that
the pointer zip.data is also protected by buf_pool->mutex.
buf_buddy_alloc_zip(): Calculate the address with char* aritmetics.
buf_buddy_alloc_clean(): Acquire a mutex on the block before
calling buf_LRU_free_block().
buf_buddy_relocate(): Remove the bogus assertion about src and dst
being buddies of each other. Use mach_read_from_4() instead of
page_get_space_id() and page_get_page_no() to avoid page alignment
assertions.
buf_buddy_free_low(): Replace bogus assertions.
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).
buf_LRU_get_free_block(): Document that the block is in the state
BUF_BLOCK_READY_FOR_USE.
buf_block_alloc(): Change the state of the block to BUF_BLOCK_MEMORY
and document it.
Replace those invocations of buf_block_alloc() with
buf_LRU_get_free_block() where the allocated block is used for buffer
pool input and output. However, temporary copies of B-tree pages
during reorganization are not used for file I/O, and such blocks are
still allocated with buf_block_alloc().
the LRU list regardless of their size.
buf_buddy_alloc_clean_zip(): Rename to buf_buddy_alloc_clean(),
now that also uncompressed blocks can be reclaimed.
Use buf_buddy_alloc_zip() instead of reusing the freed block
directly, since buf_buddy_free() may combine blocks.
buf_buddy_alloc(), buf_buddy_alloc_low(): Add parameter "lru" for
enabling allocation from the list of least-recently-used blocks.
buf_buddy_alloc_low(): Release buf_pool->mutex while calling
buf_LRU_get_free_block().
buf_LRU_free_block(): Free compressed-only pages and their
descriptors with buf_buddy_free().
buf_LRU_get_free_only(): New function for returning a block from
buf_pool->free if one is available. Split from buf_LRU_get_free_block().
buf_buddy_alloc_zip(): Rename from buf_buddy_alloc_low() and make static.
Remove parameter "split". Always try to split.
buf_buddy_free_block(): Rename to buf_buddy_block_free().
buf_buddy_block_register(): New function for registering buf_block_t
objects reserved by the allocator.
buf_buddy_alloc_from(): New function for allocating an object from a
bigger object, and putting the unused parts on the free list.
buf_buddy_alloc_clean_zip(): New function for allocating memory by
replacing an unmodified compressed page.
buf_buddy_alloc_low(): New function for allocating memory, either from
the free lists of compressed pages, from the global free list, or from
unmodified pages in the buffer pool.
