void
buf_LRU_print(void)
/*===============*/
{
buf_block_t* block;
buf_frame_t* frame;
ulint len;
ut_ad(buf_pool);
mutex_enter(&(buf_pool->mutex));
fprintf(stderr, "Pool ulint clock %lu\n",
(ulong) buf_pool->ulint_clock);
block = UT_LIST_GET_FIRST(buf_pool->LRU);
len = 0;
while (block != NULL) {
fprintf(stderr, "BLOCK %lu ", (ulong) block->offset);
if (block->old) {
fputs("old ", stderr);
}
if (block->buf_fix_count) {
fprintf(stderr, "buffix count %lu ",
(ulong) block->buf_fix_count);
}
if (block->io_fix) {
fprintf(stderr, "io_fix %lu ", (ulong) block->io_fix);
}
if (ut_dulint_cmp(block->oldest_modification,
ut_dulint_zero) > 0) {
fputs("modif. ", stderr);
}
frame = buf_block_get_frame(block);
fprintf(stderr, "LRU pos %lu type %lu index id %lu ",
(ulong) block->LRU_position,
(ulong) fil_page_get_type(frame),
(ulong) ut_dulint_get_low
(btr_page_get_index_id(frame)));
block = UT_LIST_GET_NEXT(LRU, block);
if (++len == 10) {
len = 0;
putc('\n', stderr);
}
}
mutex_exit(&(buf_pool->mutex));
}
void
read_view_print(
/*============*/
read_view_t* view) /* in: read view */
{
ulint n_ids;
ulint i;
if (view->type == VIEW_HIGH_GRANULARITY) {
fprintf(stderr,
"High-granularity read view undo_n:o %lu %lu\n",
(ulong) ut_dulint_get_high(view->undo_no),
(ulong) ut_dulint_get_low(view->undo_no));
} else {
fprintf(stderr, "Normal read view\n");
}
fprintf(stderr, "Read view low limit trx n:o %lu %lu\n",
(ulong) ut_dulint_get_high(view->low_limit_no),
(ulong) ut_dulint_get_low(view->low_limit_no));
fprintf(stderr, "Read view up limit trx id %lu %lu\n",
(ulong) ut_dulint_get_high(view->up_limit_id),
(ulong) ut_dulint_get_low(view->up_limit_id));
fprintf(stderr, "Read view low limit trx id %lu %lu\n",
(ulong) ut_dulint_get_high(view->low_limit_id),
(ulong) ut_dulint_get_low(view->low_limit_id));
fprintf(stderr, "Read view individually stored trx ids:\n");
n_ids = view->n_trx_ids;
for (i = 0; i < n_ids; i++) {
fprintf(stderr, "Read view trx id %lu %lu\n",
(ulong) ut_dulint_get_high(
read_view_get_nth_trx_id(view, i)),
(ulong) ut_dulint_get_low(
read_view_get_nth_trx_id(view, i)));
}
}
void
trx_purge_sys_print(void)
/*=====================*/
{
fprintf(stderr, "InnoDB: Purge system view:\n");
read_view_print(purge_sys->view);
fprintf(stderr, "InnoDB: Purge trx n:o %lu %lu, undo n_o %lu %lu\n",
(ulong) ut_dulint_get_high(purge_sys->purge_trx_no),
(ulong) ut_dulint_get_low(purge_sys->purge_trx_no),
(ulong) ut_dulint_get_high(purge_sys->purge_undo_no),
(ulong) ut_dulint_get_low(purge_sys->purge_undo_no));
fprintf(stderr,
"InnoDB: Purge next stored %lu, page_no %lu, offset %lu,\n"
"InnoDB: Purge hdr_page_no %lu, hdr_offset %lu\n",
(ulong) purge_sys->next_stored,
(ulong) purge_sys->page_no,
(ulong) purge_sys->offset,
(ulong) purge_sys->hdr_page_no,
(ulong) purge_sys->hdr_offset);
}
/**********************************************************************//**
Prints info about a transaction to the given file. The caller must own the
kernel mutex. */
UNIV_INTERN
void
trx_print(
/*======*/
FILE* f, /*!< in: output stream */
trx_t* trx, /*!< in: transaction */
ulint max_query_len) /*!< in: max query length to print, or 0 to
use the default max length */
{
ibool newline;
fprintf(f, "TRANSACTION " TRX_ID_FMT, TRX_ID_PREP_PRINTF(trx->id));
switch (trx->conc_state) {
case TRX_NOT_STARTED:
fputs(", not started", f);
break;
case TRX_ACTIVE:
fprintf(f, ", ACTIVE %lu sec",
(ulong)difftime(time(NULL), trx->start_time));
break;
case TRX_PREPARED:
fprintf(f, ", ACTIVE (PREPARED) %lu sec",
(ulong)difftime(time(NULL), trx->start_time));
break;
case TRX_COMMITTED_IN_MEMORY:
fputs(", COMMITTED IN MEMORY", f);
break;
default:
fprintf(f, " state %lu", (ulong) trx->conc_state);
}
#ifdef UNIV_LINUX
fprintf(f, ", process no %lu", trx->mysql_process_no);
#endif
fprintf(f, ", OS thread id %lu",
(ulong) os_thread_pf(trx->mysql_thread_id));
if (*trx->op_info) {
putc(' ', f);
fputs(trx->op_info, f);
}
if (trx->is_recovered) {
fputs(" recovered trx", f);
}
if (trx->is_purge) {
fputs(" purge trx", f);
}
if (trx->declared_to_be_inside_innodb) {
fprintf(f, ", thread declared inside InnoDB %lu",
(ulong) trx->n_tickets_to_enter_innodb);
}
putc('\n', f);
if (trx->n_mysql_tables_in_use > 0 || trx->mysql_n_tables_locked > 0) {
fprintf(f, "mysql tables in use %lu, locked %lu\n",
(ulong) trx->n_mysql_tables_in_use,
(ulong) trx->mysql_n_tables_locked);
}
newline = TRUE;
switch (trx->que_state) {
case TRX_QUE_RUNNING:
newline = FALSE; break;
case TRX_QUE_LOCK_WAIT:
fputs("LOCK WAIT ", f); break;
case TRX_QUE_ROLLING_BACK:
fputs("ROLLING BACK ", f); break;
case TRX_QUE_COMMITTING:
fputs("COMMITTING ", f); break;
default:
fprintf(f, "que state %lu ", (ulong) trx->que_state);
}
if (0 < UT_LIST_GET_LEN(trx->trx_locks)
|| mem_heap_get_size(trx->lock_heap) > 400) {
newline = TRUE;
fprintf(f, "%lu lock struct(s), heap size %lu,"
" %lu row lock(s)",
(ulong) UT_LIST_GET_LEN(trx->trx_locks),
(ulong) mem_heap_get_size(trx->lock_heap),
(ulong) lock_number_of_rows_locked(trx));
}
if (trx->has_search_latch) {
newline = TRUE;
fputs(", holds adaptive hash latch", f);
}
if (!ut_dulint_is_zero(trx->undo_no)) {
newline = TRUE;
fprintf(f, ", undo log entries %lu",
(ulong) ut_dulint_get_low(trx->undo_no));
}
if (newline) {
putc('\n', f);
}
if (trx->mysql_thd != NULL) {
innobase_mysql_print_thd(f, trx->mysql_thd, max_query_len);
}
}
ulint
dict_truncate_index_tree(
/*=====================*/
/* out: new root page number, or
FIL_NULL on failure */
dict_table_t* table, /* in: the table the index belongs to */
btr_pcur_t* pcur, /* in/out: persistent cursor pointing to
record in the clustered index of
SYS_INDEXES table. The cursor may be
repositioned in this call. */
mtr_t* mtr) /* in: mtr having the latch
on the record page. The mtr may be
committed and restarted in this call. */
{
ulint root_page_no;
ulint space;
ulint type;
dulint index_id;
rec_t* rec;
byte* ptr;
ulint len;
ulint comp;
dict_index_t* index;
ut_ad(mutex_own(&(dict_sys->mutex)));
ut_a(!dict_table_is_comp(dict_sys->sys_indexes));
rec = btr_pcur_get_rec(pcur);
ptr = rec_get_nth_field_old(rec, DICT_SYS_INDEXES_PAGE_NO_FIELD, &len);
ut_ad(len == 4);
root_page_no = mtr_read_ulint(ptr, MLOG_4BYTES, mtr);
if (root_page_no == FIL_NULL) {
/* The tree has been freed. */
ut_print_timestamp(stderr);
fprintf(stderr, " InnoDB: Trying to TRUNCATE"
" a missing index of table %s!\n", table->name);
return(FIL_NULL);
}
ptr = rec_get_nth_field_old(rec,
DICT_SYS_INDEXES_SPACE_NO_FIELD, &len);
ut_ad(len == 4);
space = mtr_read_ulint(ptr, MLOG_4BYTES, mtr);
if (!fil_tablespace_exists_in_mem(space)) {
/* It is a single table tablespace and the .ibd file is
missing: do nothing */
ut_print_timestamp(stderr);
fprintf(stderr, " InnoDB: Trying to TRUNCATE"
" a missing .ibd file of table %s!\n", table->name);
return(FIL_NULL);
}
ptr = rec_get_nth_field_old(rec,
DICT_SYS_INDEXES_TYPE_FIELD, &len);
ut_ad(len == 4);
type = mach_read_from_4(ptr);
ptr = rec_get_nth_field_old(rec, 1, &len);
ut_ad(len == 8);
index_id = mach_read_from_8(ptr);
/* We free all the pages but the root page first; this operation
may span several mini-transactions */
btr_free_but_not_root(space, root_page_no);
/* Then we free the root page in the same mini-transaction where
we create the b-tree and write its new root page number to the
appropriate field in the SYS_INDEXES record: this mini-transaction
marks the B-tree totally truncated */
comp = page_is_comp(btr_page_get(space, root_page_no, RW_X_LATCH,
mtr));
btr_free_root(space, root_page_no, mtr);
/* We will temporarily write FIL_NULL to the PAGE_NO field
in SYS_INDEXES, so that the database will not get into an
inconsistent state in case it crashes between the mtr_commit()
below and the following mtr_commit() call. */
page_rec_write_index_page_no(rec, DICT_SYS_INDEXES_PAGE_NO_FIELD,
FIL_NULL, mtr);
/* We will need to commit the mini-transaction in order to avoid
deadlocks in the btr_create() call, because otherwise we would
be freeing and allocating pages in the same mini-transaction. */
btr_pcur_store_position(pcur, mtr);
mtr_commit(mtr);
mtr_start(mtr);
btr_pcur_restore_position(BTR_MODIFY_LEAF, pcur, mtr);
/* Find the index corresponding to this SYS_INDEXES record. */
for (index = UT_LIST_GET_FIRST(table->indexes);
index;
index = UT_LIST_GET_NEXT(indexes, index)) {
if (!ut_dulint_cmp(index->id, index_id)) {
break;
}
}
root_page_no = btr_create(type, space, index_id, comp, mtr);
if (index) {
index->page = (unsigned int) root_page_no;
} else {
ut_print_timestamp(stderr);
fprintf(stderr,
" InnoDB: Index %lu %lu of table %s is missing\n"
"InnoDB: from the data dictionary during TRUNCATE!\n",
ut_dulint_get_high(index_id),
ut_dulint_get_low(index_id),
table->name);
}
return(root_page_no);
}
void
trx_lists_init_at_db_start(void)
/*============================*/
{
trx_rseg_t* rseg;
trx_undo_t* undo;
trx_t* trx;
UT_LIST_INIT(trx_sys->trx_list);
/* Look from the rollback segments if there exist undo logs for
transactions */
rseg = UT_LIST_GET_FIRST(trx_sys->rseg_list);
while (rseg != NULL) {
undo = UT_LIST_GET_FIRST(rseg->insert_undo_list);
while (undo != NULL) {
trx = trx_create(NULL);
trx->id = undo->trx_id;
trx->xid = undo->xid;
trx->insert_undo = undo;
trx->rseg = rseg;
if (undo->state != TRX_UNDO_ACTIVE) {
/* Prepared transactions are left in
the prepared state waiting for a
commit or abort decision from MySQL */
if (undo->state == TRX_UNDO_PREPARED) {
fprintf(stderr,
"InnoDB: Transaction %lu %lu"
" was in the"
" XA prepared state.\n",
ut_dulint_get_high(trx->id),
ut_dulint_get_low(trx->id));
if (srv_force_recovery == 0) {
trx->conc_state = TRX_PREPARED;
} else {
fprintf(stderr,
"InnoDB: Since"
" innodb_force_recovery"
" > 0, we will"
" rollback it"
" anyway.\n");
trx->conc_state = TRX_ACTIVE;
}
} else {
trx->conc_state
= TRX_COMMITTED_IN_MEMORY;
}
/* We give a dummy value for the trx no;
this should have no relevance since purge
is not interested in committed transaction
numbers, unless they are in the history
list, in which case it looks the number
from the disk based undo log structure */
trx->no = trx->id;
} else {
trx->conc_state = TRX_ACTIVE;
/* A running transaction always has the number
field inited to ut_dulint_max */
trx->no = ut_dulint_max;
}
if (undo->dict_operation) {
trx->dict_operation = undo->dict_operation;
trx->table_id = undo->table_id;
}
if (!undo->empty) {
trx->undo_no = ut_dulint_add(undo->top_undo_no,
1);
}
trx_list_insert_ordered(trx);
undo = UT_LIST_GET_NEXT(undo_list, undo);
}
undo = UT_LIST_GET_FIRST(rseg->update_undo_list);
while (undo != NULL) {
trx = trx_get_on_id(undo->trx_id);
if (NULL == trx) {
trx = trx_create(NULL);
trx->id = undo->trx_id;
trx->xid = undo->xid;
if (undo->state != TRX_UNDO_ACTIVE) {
/* Prepared transactions are left in
the prepared state waiting for a
commit or abort decision from MySQL */
if (undo->state == TRX_UNDO_PREPARED) {
fprintf(stderr,
"InnoDB: Transaction"
" %lu %lu was in the"
" XA prepared state.\n",
ut_dulint_get_high(
trx->id),
ut_dulint_get_low(
trx->id));
if (srv_force_recovery == 0) {
trx->conc_state
= TRX_PREPARED;
} else {
fprintf(stderr,
"InnoDB: Since"
" innodb_force_recovery"
" > 0, we will"
" rollback it"
" anyway.\n");
trx->conc_state
= TRX_ACTIVE;
}
} else {
trx->conc_state
= TRX_COMMITTED_IN_MEMORY;
}
/* We give a dummy value for the trx
number */
trx->no = trx->id;
} else {
trx->conc_state = TRX_ACTIVE;
/* A running transaction always has
the number field inited to
ut_dulint_max */
trx->no = ut_dulint_max;
}
trx->rseg = rseg;
trx_list_insert_ordered(trx);
if (undo->dict_operation) {
trx->dict_operation
= undo->dict_operation;
trx->table_id = undo->table_id;
}
}
trx->update_undo = undo;
if ((!undo->empty)
&& (ut_dulint_cmp(undo->top_undo_no,
trx->undo_no) >= 0)) {
trx->undo_no = ut_dulint_add(undo->top_undo_no,
1);
}
undo = UT_LIST_GET_NEXT(undo_list, undo);
}
rseg = UT_LIST_GET_NEXT(rseg_list, rseg);
}
}
int
trx_recover_for_mysql(
/*==================*/
/* out: number of prepared transactions
stored in xid_list */
XID* xid_list, /* in/out: prepared transactions */
ulint len) /* in: number of slots in xid_list */
{
trx_t* trx;
ulint count = 0;
ut_ad(xid_list);
ut_ad(len);
/* We should set those transactions which are in the prepared state
to the xid_list */
mutex_enter(&kernel_mutex);
trx = UT_LIST_GET_FIRST(trx_sys->trx_list);
while (trx) {
if (trx->conc_state == TRX_PREPARED) {
xid_list[count] = trx->xid;
if (count == 0) {
ut_print_timestamp(stderr);
fprintf(stderr,
" InnoDB: Starting recovery for"
" XA transactions...\n");
}
ut_print_timestamp(stderr);
fprintf(stderr,
" InnoDB: Transaction %lu %lu in"
" prepared state after recovery\n",
(ulong) ut_dulint_get_high(trx->id),
(ulong) ut_dulint_get_low(trx->id));
ut_print_timestamp(stderr);
fprintf(stderr,
" InnoDB: Transaction contains changes"
" to %lu rows\n",
(ulong) ut_conv_dulint_to_longlong(
trx->undo_no));
count++;
if (count == len) {
break;
}
}
trx = UT_LIST_GET_NEXT(trx_list, trx);
}
mutex_exit(&kernel_mutex);
if (count > 0){
ut_print_timestamp(stderr);
fprintf(stderr,
" InnoDB: %lu transactions in prepared state"
" after recovery\n",
(ulong) count);
}
return ((int) count);
}
/*************************************************************//**
Pretty prints a dfield value according to its data type. Also the hex string
is printed if a string contains non-printable characters. */
UNIV_INTERN
void
dfield_print_also_hex(
/*==================*/
const dfield_t* dfield) /*!< in: dfield */
{
const byte* data;
ulint len;
ulint prtype;
ulint i;
ibool print_also_hex;
len = dfield_get_len(dfield);
data = dfield_get_data(dfield);
if (dfield_is_null(dfield)) {
fputs("NULL", stderr);
return;
}
prtype = dtype_get_prtype(dfield_get_type(dfield));
switch (dtype_get_mtype(dfield_get_type(dfield))) {
dulint id;
case DATA_INT:
switch (len) {
ulint val;
case 1:
val = mach_read_from_1(data);
if (!(prtype & DATA_UNSIGNED)) {
val &= ~0x80;
fprintf(stderr, "%ld", (long) val);
} else {
fprintf(stderr, "%lu", (ulong) val);
}
break;
case 2:
val = mach_read_from_2(data);
if (!(prtype & DATA_UNSIGNED)) {
val &= ~0x8000;
fprintf(stderr, "%ld", (long) val);
} else {
fprintf(stderr, "%lu", (ulong) val);
}
break;
case 3:
val = mach_read_from_3(data);
if (!(prtype & DATA_UNSIGNED)) {
val &= ~0x800000;
fprintf(stderr, "%ld", (long) val);
} else {
fprintf(stderr, "%lu", (ulong) val);
}
break;
case 4:
val = mach_read_from_4(data);
if (!(prtype & DATA_UNSIGNED)) {
val &= ~0x80000000;
fprintf(stderr, "%ld", (long) val);
} else {
fprintf(stderr, "%lu", (ulong) val);
}
break;
case 6:
id = mach_read_from_6(data);
fprintf(stderr, "{%lu %lu}",
ut_dulint_get_high(id),
ut_dulint_get_low(id));
break;
case 7:
id = mach_read_from_7(data);
fprintf(stderr, "{%lu %lu}",
ut_dulint_get_high(id),
ut_dulint_get_low(id));
break;
case 8:
id = mach_read_from_8(data);
fprintf(stderr, "{%lu %lu}",
ut_dulint_get_high(id),
ut_dulint_get_low(id));
break;
default:
goto print_hex;
}
break;
case DATA_SYS:
switch (prtype & DATA_SYS_PRTYPE_MASK) {
case DATA_TRX_ID:
id = mach_read_from_6(data);
fprintf(stderr, "trx_id " TRX_ID_FMT,
TRX_ID_PREP_PRINTF(id));
break;
case DATA_ROLL_PTR:
id = mach_read_from_7(data);
fprintf(stderr, "roll_ptr {%lu %lu}",
ut_dulint_get_high(id), ut_dulint_get_low(id));
break;
case DATA_ROW_ID:
id = mach_read_from_6(data);
fprintf(stderr, "row_id {%lu %lu}",
ut_dulint_get_high(id), ut_dulint_get_low(id));
break;
default:
id = mach_dulint_read_compressed(data);
fprintf(stderr, "mix_id {%lu %lu}",
ut_dulint_get_high(id), ut_dulint_get_low(id));
}
break;
case DATA_CHAR:
case DATA_VARCHAR:
print_also_hex = FALSE;
for (i = 0; i < len; i++) {
int c = *data++;
if (!isprint(c)) {
print_also_hex = TRUE;
fprintf(stderr, "\\x%02x", (unsigned char) c);
} else {
putc(c, stderr);
}
}
if (dfield_is_ext(dfield)) {
fputs("(external)", stderr);
}
if (!print_also_hex) {
break;
}
data = dfield_get_data(dfield);
/* fall through */
case DATA_BINARY:
default:
print_hex:
fputs(" Hex: ",stderr);
for (i = 0; i < len; i++) {
fprintf(stderr, "%02lx", (ulint) *data++);
}
if (dfield_is_ext(dfield)) {
fputs("(external)", stderr);
}
}
}
/************************************************************************
Loads definitions for table indexes. Adds them to the data dictionary cache.
*/
static
void
dict_load_indexes(
/*==============*/
dict_table_t* table, /* in: table */
mem_heap_t* heap) /* in: memory heap for temporary storage */
{
dict_table_t* sys_indexes;
dict_index_t* sys_index;
dict_index_t* index;
btr_pcur_t pcur;
dtuple_t* tuple;
dfield_t* dfield;
rec_t* rec;
byte* field;
ulint len;
ulint name_len;
char* name_buf;
ulint type;
ulint space;
ulint page_no;
ulint n_fields;
byte* buf;
ibool is_sys_table;
dulint id;
mtr_t mtr;
ut_ad(mutex_own(&(dict_sys->mutex)));
if ((ut_dulint_get_high(table->id) == 0)
&& (ut_dulint_get_low(table->id) < DICT_HDR_FIRST_ID)) {
is_sys_table = TRUE;
} else {
is_sys_table = FALSE;
}
mtr_start(&mtr);
sys_indexes = dict_table_get_low("SYS_INDEXES");
sys_index = UT_LIST_GET_FIRST(sys_indexes->indexes);
tuple = dtuple_create(heap, 1);
dfield = dtuple_get_nth_field(tuple, 0);
buf = mem_heap_alloc(heap, 8);
mach_write_to_8(buf, table->id);
dfield_set_data(dfield, buf, 8);
dict_index_copy_types(tuple, sys_index, 1);
btr_pcur_open_on_user_rec(sys_index, tuple, PAGE_CUR_GE,
BTR_SEARCH_LEAF, &pcur, &mtr);
for (;;) {
if (!btr_pcur_is_on_user_rec(&pcur, &mtr)) {
break;
}
rec = btr_pcur_get_rec(&pcur);
field = rec_get_nth_field(rec, 0, &len);
ut_ad(len == 8);
if (ut_memcmp(buf, field, len) != 0) {
break;
}
ut_a(!rec_get_deleted_flag(rec));
field = rec_get_nth_field(rec, 1, &len);
ut_ad(len == 8);
id = mach_read_from_8(field);
ut_a(0 == ut_strcmp("NAME",
dict_field_get_col(
dict_index_get_nth_field(
dict_table_get_first_index(sys_indexes), 4))->name));
field = rec_get_nth_field(rec, 4, &name_len);
name_buf = mem_heap_alloc(heap, name_len + 1);
ut_memcpy(name_buf, field, name_len);
name_buf[name_len] = '\0';
field = rec_get_nth_field(rec, 5, &len);
n_fields = mach_read_from_4(field);
field = rec_get_nth_field(rec, 6, &len);
type = mach_read_from_4(field);
field = rec_get_nth_field(rec, 7, &len);
space = mach_read_from_4(field);
ut_a(0 == ut_strcmp("PAGE_NO",
dict_field_get_col(
dict_index_get_nth_field(
dict_table_get_first_index(sys_indexes), 8))->name));
field = rec_get_nth_field(rec, 8, &len);
page_no = mach_read_from_4(field);
if (is_sys_table
&& ((type & DICT_CLUSTERED)
|| ((table == dict_sys->sys_tables)
&& (name_len == ut_strlen("ID_IND"))
&& (0 == ut_memcmp(name_buf, "ID_IND",
name_len))))) {
/* The index was created in memory already in
booting */
} else {
index = dict_mem_index_create(table->name, name_buf,
space, type, n_fields);
index->page_no = page_no;
index->id = id;
dict_load_fields(table, index, heap);
dict_index_add_to_cache(table, index);
}
btr_pcur_move_to_next_user_rec(&pcur, &mtr);
}
btr_pcur_close(&pcur);
mtr_commit(&mtr);
}
