Skip to content

M
MariaDB
Commit 3ce8a0fc authored by Marko Mäkelä's avatar Marko Mäkelä
Browse files
Options

MDEV-16136: Simplify trx_lock_t memory management 

Allocate trx->lock.rec_pool and trx->lock.table_pool directly from trx_t.
Remove unnecessary use of std::vector.

In order to do this, move some definitions from lock0priv.h to
lock0types.h, so that ib_lock_t will not be an opaque type.
parent b7951348
Hide whitespace changes
Inline Side-by-side
Showing with 216 additions and 275 deletions
storage/innobase/include/lock0lock.h
View file @ 3ce8a0fc
......@@ -834,69 +834,6 @@ lock_trx_has_rec_x_lock(
MY_ATTRIBUTE((warn_unused_result));
#endif /* UNIV_DEBUG */
/**
Allocate cached locks for the transaction.
@param trx allocate cached record locks for this transaction */
void
lock_trx_alloc_locks(trx_t* trx);
/** Lock modes and types */
/* @{ */
#define LOCK_MODE_MASK 0xFUL /*!< mask used to extract mode from the
type_mode field in a lock */
/** Lock types */
/* @{ */
#define LOCK_TABLE 16U /*!< table lock */
#define LOCK_REC 32U /*!< record lock */
#define LOCK_TYPE_MASK 0xF0UL /*!< mask used to extract lock type from the
type_mode field in a lock */
#if LOCK_MODE_MASK & LOCK_TYPE_MASK
# error "LOCK_MODE_MASK & LOCK_TYPE_MASK"
#endif
#define LOCK_WAIT 256U /*!< Waiting lock flag; when set, it
means that the lock has not yet been
granted, it is just waiting for its
turn in the wait queue */
/* Precise modes */
#define LOCK_ORDINARY 0 /*!< this flag denotes an ordinary
next-key lock in contrast to LOCK_GAP
or LOCK_REC_NOT_GAP */
#define LOCK_GAP 512U /*!< when this bit is set, it means that the
lock holds only on the gap before the record;
for instance, an x-lock on the gap does not
give permission to modify the record on which
the bit is set; locks of this type are created
when records are removed from the index chain
of records */
#define LOCK_REC_NOT_GAP 1024U /*!< this bit means that the lock is only on
the index record and does NOT block inserts
to the gap before the index record; this is
used in the case when we retrieve a record
with a unique key, and is also used in
locking plain SELECTs (not part of UPDATE
or DELETE) when the user has set the READ
COMMITTED isolation level */
#define LOCK_INSERT_INTENTION 2048U/*!< this bit is set when we place a waiting
gap type record lock request in order to let
an insert of an index record to wait until
there are no conflicting locks by other
transactions on the gap; note that this flag
remains set when the waiting lock is granted,
or if the lock is inherited to a neighboring
record */
#define LOCK_PREDICATE 8192U /*!< Predicate lock */
#define LOCK_PRDT_PAGE 16384U /*!< Page lock */
#if (LOCK_WAIT|LOCK_GAP|LOCK_REC_NOT_GAP|LOCK_INSERT_INTENTION|LOCK_PREDICATE|LOCK_PRDT_PAGE)&LOCK_MODE_MASK
# error
#endif
#if (LOCK_WAIT|LOCK_GAP|LOCK_REC_NOT_GAP|LOCK_INSERT_INTENTION|LOCK_PREDICATE|LOCK_PRDT_PAGE)&LOCK_TYPE_MASK
# error
#endif
/* @} */
/** Lock operation struct */
struct lock_op_t{
dict_table_t* table; /*!< table to be locked */
......
storage/innobase/include/lock0priv.h
View file @ 3ce8a0fc
......@@ -42,19 +42,6 @@ those functions in lock/ */
#define UINT32_MAX (4294967295U)
#endif
/** A table lock */
struct lock_table_t {
dict_table_t* table; /*!< database table in dictionary
cache */
UT_LIST_NODE_T(lock_t)
locks; /*!< list of locks on the same
table */
/** Print the table lock into the given output stream
@param[in,out] out the output stream
@return the given output stream. */
std::ostream& print(std::ostream& out) const;
};
/** Print the table lock into the given output stream
@param[in,out] out the output stream
@return the given output stream. */
......@@ -77,131 +64,11 @@ operator<<(std::ostream& out, const lock_table_t& lock)
return(lock.print(out));
}
/** Record lock for a page */
struct lock_rec_t {
ib_uint32_t space; /*!< space id */
ib_uint32_t page_no; /*!< page number */
ib_uint32_t n_bits; /*!< number of bits in the lock
bitmap; NOTE: the lock bitmap is
placed immediately after the
lock struct */
/** Print the record lock into the given output stream
@param[in,out] out the output stream
@return the given output stream. */
std::ostream& print(std::ostream& out) const;
};
/** Print the record lock into the given output stream
@param[in,out] out the output stream
@return the given output stream. */
inline
std::ostream& lock_rec_t::print(std::ostream& out) const
{
out << "[lock_rec_t: space=" << space << ", page_no=" << page_no
<< ", n_bits=" << n_bits << "]";
return(out);
}
inline
std::ostream&
operator<<(std::ostream& out, const lock_rec_t& lock)
{
return(lock.print(out));
}
/** Lock struct; protected by lock_sys->mutex */
struct lock_t {
trx_t* trx; /*!< transaction owning the
lock */
UT_LIST_NODE_T(lock_t)
trx_locks; /*!< list of the locks of the
transaction */
dict_index_t* index; /*!< index for a record lock */
lock_t* hash; /*!< hash chain node for a record
lock. The link node in a singly linked
list, used during hashing. */
/* Statistics for how long lock has been held and time
how long this lock had to be waited before it was granted */
time_t requested_time; /*!< Lock request time */
ulint wait_time; /*!< Time waited this lock or 0 */
union {
lock_table_t tab_lock;/*!< table lock */
lock_rec_t rec_lock;/*!< record lock */
} un_member; /*!< lock details */
ib_uint32_t type_mode; /*!< lock type, mode, LOCK_GAP or
LOCK_REC_NOT_GAP,
LOCK_INSERT_INTENTION,
wait flag, ORed */
/** Determine if the lock object is a record lock.
@return true if record lock, false otherwise. */
bool is_record_lock() const
{
return(type() == LOCK_REC);
}
bool is_waiting() const
{
return(type_mode & LOCK_WAIT);
}
bool is_gap() const
{
return(type_mode & LOCK_GAP);
}
bool is_record_not_gap() const
{
return(type_mode & LOCK_REC_NOT_GAP);
}
bool is_insert_intention() const
{
return(type_mode & LOCK_INSERT_INTENTION);
}
ulint type() const {
return(type_mode & LOCK_TYPE_MASK);
}
enum lock_mode mode() const
{
return(static_cast<enum lock_mode>(type_mode & LOCK_MODE_MASK));
}
/** Print the lock object into the given output stream.
@param[in,out] out the output stream
@return the given output stream. */
std::ostream& print(std::ostream& out) const;
/** Convert the member 'type_mode' into a human readable string.
@return human readable string */
std::string type_mode_string() const;
const char* type_string() const
{
switch (type_mode & LOCK_TYPE_MASK) {
case LOCK_REC:
return("LOCK_REC");
case LOCK_TABLE:
return("LOCK_TABLE");
default:
ut_error;
}
}
};
/** Convert the member 'type_mode' into a human readable string.
@return human readable string */
inline
std::string
lock_t::type_mode_string() const
ib_lock_t::type_mode_string() const
{
std::ostringstream sout;
sout << type_string();
......@@ -227,7 +94,7 @@ lock_t::type_mode_string() const
inline
std::ostream&
lock_t::print(std::ostream& out) const
ib_lock_t::print(std::ostream& out) const
{
out << "[lock_t: type_mode=" << type_mode << "("
<< type_mode_string() << ")";
......@@ -244,7 +111,7 @@ lock_t::print(std::ostream& out) const
inline
std::ostream&
operator<<(std::ostream& out, const lock_t& lock)
operator<<(std::ostream& out, const ib_lock_t& lock)
{
return(lock.print(out));
}
......
storage/innobase/include/lock0types.h
View file @ 3ce8a0fc
/*****************************************************************************
Copyright (c) 1996, 2015, Oracle and/or its affiliates. All Rights Reserved.
Copyright (c) 2018, MariaDB Corporation.
This program is free software; you can redistribute it and/or modify it under
the terms of the GNU General Public License as published by the Free Software
......@@ -73,6 +74,195 @@ const char* lock_mode_string(enum lock_mode mode)
}
}
typedef UT_LIST_BASE_NODE_T(lock_t) trx_lock_list_t;
/** A table lock */
struct lock_table_t {
dict_table_t* table; /*!< database table in dictionary
cache */
UT_LIST_NODE_T(ib_lock_t)
locks; /*!< list of locks on the same
table */
/** Print the table lock into the given output stream
@param[in,out] out the output stream
@return the given output stream. */
std::ostream& print(std::ostream& out) const;
};
/** Record lock for a page */
struct lock_rec_t {
ib_uint32_t space; /*!< space id */
ib_uint32_t page_no; /*!< page number */
ib_uint32_t n_bits; /*!< number of bits in the lock
bitmap; NOTE: the lock bitmap is
placed immediately after the
lock struct */
/** Print the record lock into the given output stream
@param[in,out] out the output stream
@return the given output stream. */
std::ostream& print(std::ostream& out) const;
};
/** Print the record lock into the given output stream
@param[in,out] out the output stream
@return the given output stream. */
inline
std::ostream& lock_rec_t::print(std::ostream& out) const
{
out << "[lock_rec_t: space=" << space << ", page_no=" << page_no
<< ", n_bits=" << n_bits << "]";
return(out);
}
inline
std::ostream&
operator<<(std::ostream& out, const lock_rec_t& lock)
{
return(lock.print(out));
}
#define LOCK_MODE_MASK 0xFUL /*!< mask used to extract mode from the
type_mode field in a lock */
/** Lock types */
/* @{ */
#define LOCK_TABLE 16U /*!< table lock */
#define LOCK_REC 32U /*!< record lock */
#define LOCK_TYPE_MASK 0xF0UL /*!< mask used to extract lock type from the
type_mode field in a lock */
#if LOCK_MODE_MASK & LOCK_TYPE_MASK
# error "LOCK_MODE_MASK & LOCK_TYPE_MASK"
#endif
#define LOCK_WAIT 256U /*!< Waiting lock flag; when set, it
means that the lock has not yet been
granted, it is just waiting for its
turn in the wait queue */
/* Precise modes */
#define LOCK_ORDINARY 0 /*!< this flag denotes an ordinary
next-key lock in contrast to LOCK_GAP
or LOCK_REC_NOT_GAP */
#define LOCK_GAP 512U /*!< when this bit is set, it means that the
lock holds only on the gap before the record;
for instance, an x-lock on the gap does not
give permission to modify the record on which
the bit is set; locks of this type are created
when records are removed from the index chain
of records */
#define LOCK_REC_NOT_GAP 1024U /*!< this bit means that the lock is only on
the index record and does NOT block inserts
to the gap before the index record; this is
used in the case when we retrieve a record
with a unique key, and is also used in
locking plain SELECTs (not part of UPDATE
or DELETE) when the user has set the READ
COMMITTED isolation level */
#define LOCK_INSERT_INTENTION 2048U/*!< this bit is set when we place a waiting
gap type record lock request in order to let
an insert of an index record to wait until
there are no conflicting locks by other
transactions on the gap; note that this flag
remains set when the waiting lock is granted,
or if the lock is inherited to a neighboring
record */
#define LOCK_PREDICATE 8192U /*!< Predicate lock */
#define LOCK_PRDT_PAGE 16384U /*!< Page lock */
#if (LOCK_WAIT|LOCK_GAP|LOCK_REC_NOT_GAP|LOCK_INSERT_INTENTION|LOCK_PREDICATE|LOCK_PRDT_PAGE)&LOCK_MODE_MASK
# error
#endif
#if (LOCK_WAIT|LOCK_GAP|LOCK_REC_NOT_GAP|LOCK_INSERT_INTENTION|LOCK_PREDICATE|LOCK_PRDT_PAGE)&LOCK_TYPE_MASK
# error
#endif
/* @} */
/** Lock struct; protected by lock_sys->mutex */
struct ib_lock_t
{
trx_t* trx; /*!< transaction owning the
lock */
UT_LIST_NODE_T(ib_lock_t)
trx_locks; /*!< list of the locks of the
transaction */
dict_index_t* index; /*!< index for a record lock */
ib_lock_t* hash; /*!< hash chain node for a record
lock. The link node in a singly linked
list, used during hashing. */
/* Statistics for how long lock has been held and time
how long this lock had to be waited before it was granted */
time_t requested_time; /*!< Lock request time */
ulint wait_time; /*!< Time waited this lock or 0 */
union {
lock_table_t tab_lock;/*!< table lock */
lock_rec_t rec_lock;/*!< record lock */
} un_member; /*!< lock details */
ib_uint32_t type_mode; /*!< lock type, mode, LOCK_GAP or
LOCK_REC_NOT_GAP,
LOCK_INSERT_INTENTION,
wait flag, ORed */
/** Determine if the lock object is a record lock.
@return true if record lock, false otherwise. */
bool is_record_lock() const
{
return(type() == LOCK_REC);
}
bool is_waiting() const
{
return(type_mode & LOCK_WAIT);
}
bool is_gap() const
{
return(type_mode & LOCK_GAP);
}
bool is_record_not_gap() const
{
return(type_mode & LOCK_REC_NOT_GAP);
}
bool is_insert_intention() const
{
return(type_mode & LOCK_INSERT_INTENTION);
}
ulint type() const {
return(type_mode & LOCK_TYPE_MASK);
}
enum lock_mode mode() const
{
return(static_cast<enum lock_mode>(type_mode & LOCK_MODE_MASK));
}
/** Print the lock object into the given output stream.
@param[in,out] out the output stream
@return the given output stream. */
std::ostream& print(std::ostream& out) const;
/** Convert the member 'type_mode' into a human readable string.
@return human readable string */
std::string type_mode_string() const;
const char* type_string() const
{
switch (type_mode & LOCK_TYPE_MASK) {
case LOCK_REC:
return("LOCK_REC");
case LOCK_TABLE:
return("LOCK_TABLE");
default:
ut_error;
}
}
};
typedef UT_LIST_BASE_NODE_T(ib_lock_t) trx_lock_list_t;
#endif /* lock0types_h */
storage/innobase/include/trx0trx.h
View file @ 3ce8a0fc
......@@ -704,13 +704,19 @@ struct trx_lock_t {
only be modified by the thread that is
serving the running transaction. */
lock_pool_t rec_pool; /*!< Pre-allocated record locks */
/** Pre-allocated record locks */
struct {
ib_lock_t lock; byte pad[256];
} rec_pool[8];
lock_pool_t table_pool; /*!< Pre-allocated table locks */
/** Pre-allocated table locks */
ib_lock_t table_pool[8];
ulint rec_cached; /*!< Next free rec lock in pool */
/** Next available rec_pool[] entry */
unsigned rec_cached;
ulint table_cached; /*!< Next free table lock in pool */
/** Next available table_pool[] entry */
unsigned table_cached;
mem_heap_t* lock_heap; /*!< memory heap for trx_locks;
protected by lock_sys->mutex */
......
storage/innobase/lock/lock0lock.cc
View file @ 3ce8a0fc
......@@ -59,18 +59,6 @@ ulong innodb_lock_schedule_algorithm;
/** The value of innodb_deadlock_detect */
my_bool innobase_deadlock_detect;
/** Total number of cached record locks */
static const ulint REC_LOCK_CACHE = 8;
/** Maximum record lock size in bytes */
static const ulint REC_LOCK_SIZE = sizeof(ib_lock_t) + 256;
/** Total number of cached table locks */
static const ulint TABLE_LOCK_CACHE = 8;
/** Size in bytes, of the table lock instance */
static const ulint TABLE_LOCK_SIZE = sizeof(ib_lock_t);
/*********************************************************************//**
Checks if a waiting record lock request still has to wait in a queue.
@return lock that is causing the wait */
......@@ -1469,13 +1457,13 @@ lock_rec_create_low(
}
}
if (trx->lock.rec_cached >= trx->lock.rec_pool.size()
|| sizeof *lock + n_bytes > REC_LOCK_SIZE) {
if (trx->lock.rec_cached >= UT_ARR_SIZE(trx->lock.rec_pool)
|| sizeof *lock + n_bytes > sizeof *trx->lock.rec_pool) {
lock = static_cast<lock_t*>(
mem_heap_alloc(trx->lock.lock_heap,
sizeof *lock + n_bytes));
} else {
lock = trx->lock.rec_pool[trx->lock.rec_cached++];
lock = &trx->lock.rec_pool[trx->lock.rec_cached++].lock;
}
lock->trx = trx;
......@@ -3653,8 +3641,9 @@ lock_table_create(
ib_vector_push(trx->autoinc_locks, &lock);
} else if (trx->lock.table_cached < trx->lock.table_pool.size()) {
lock = trx->lock.table_pool[trx->lock.table_cached++];
} else if (trx->lock.table_cached
< UT_ARR_SIZE(trx->lock.table_pool)) {
lock = &trx->lock.table_pool[trx->lock.table_cached++];
} else {
lock = static_cast<lock_t*>(
......@@ -7678,33 +7667,6 @@ DeadlockChecker::check_and_resolve(const lock_t* lock, trx_t* trx)
return(victim_trx);
}
/**
Allocate cached locks for the transaction.
@param trx allocate cached record locks for this transaction */
void
lock_trx_alloc_locks(trx_t* trx)
{
ulint sz = REC_LOCK_SIZE * REC_LOCK_CACHE;
byte* ptr = reinterpret_cast<byte*>(ut_malloc_nokey(sz));
/* We allocate one big chunk and then distribute it among
the rest of the elements. The allocated chunk pointer is always
at index 0. */
for (ulint i = 0; i < REC_LOCK_CACHE; ++i, ptr += REC_LOCK_SIZE) {
trx->lock.rec_pool.push_back(
reinterpret_cast<ib_lock_t*>(ptr));
}
sz = TABLE_LOCK_SIZE * TABLE_LOCK_CACHE;
ptr = reinterpret_cast<byte*>(ut_malloc_nokey(sz));
for (ulint i = 0; i < TABLE_LOCK_CACHE; ++i, ptr += TABLE_LOCK_SIZE) {
trx->lock.table_pool.push_back(
reinterpret_cast<ib_lock_t*>(ptr));
}
}
/*************************************************************//**
Updates the lock table when a page is split and merged to
two pages. */
......
storage/innobase/trx/trx0trx.cc
View file @ 3ce8a0fc
......@@ -190,10 +190,6 @@ struct TrxFactory {
the constructors of the trx_t members. */
new(&trx->mod_tables) trx_mod_tables_t();
new(&trx->lock.rec_pool) lock_pool_t();
new(&trx->lock.table_pool) lock_pool_t();
new(&trx->lock.table_locks) lock_pool_t();
trx_init(trx);
......@@ -216,8 +212,6 @@ struct TrxFactory {
mutex_create(LATCH_ID_TRX, &trx->mutex);
mutex_create(LATCH_ID_TRX_UNDO, &trx->undo_mutex);
lock_trx_alloc_locks(trx);
}
/** Release resources held by the transaction object.
......@@ -249,26 +243,6 @@ struct TrxFactory {
ut_ad(trx->read_view == NULL);
if (!trx->lock.rec_pool.empty()) {
/* See lock_trx_alloc_locks() why we only free
the first element. */
ut_free(trx->lock.rec_pool[0]);
}
if (!trx->lock.table_pool.empty()) {
/* See lock_trx_alloc_locks() why we only free
the first element. */
ut_free(trx->lock.table_pool[0]);
}
trx->lock.rec_pool.~lock_pool_t();
trx->lock.table_pool.~lock_pool_t();
trx->lock.table_locks.~lock_pool_t();
}
......@@ -407,7 +381,12 @@ trx_create_low()
/* Should have been either just initialized or .clear()ed by
trx_free(). */
ut_a(trx->mod_tables.size() == 0);
ut_ad(trx->mod_tables.empty());
ut_ad(trx->lock.table_locks.empty());
ut_ad(UT_LIST_GET_LEN(trx->lock.trx_locks) == 0);
ut_ad(trx->lock.n_rec_locks == 0);
ut_ad(trx->lock.table_cached == 0);
ut_ad(trx->lock.rec_cached == 0);
#ifdef WITH_WSREP
trx->wsrep_event = NULL;
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7