/********************************************************//**
Logs a write of a string to a file page buffered in the buffer pool.
Writes the corresponding log record to the mini-transaction log. */
UNIV_INTERN
void
mlog_log_string(
/*============*/
byte* ptr, /*!< in: pointer written to */
ulint len, /*!< in: string length */
mtr_t* mtr) /*!< in: mini-transaction handle */
{
byte* log_ptr;
ut_ad(ptr && mtr);
ut_ad(len <= UNIV_PAGE_SIZE);
log_ptr = mlog_open(mtr, 30);
/* If no logging is requested, we may return now */
if (log_ptr == NULL) {
return;
}
log_ptr = mlog_write_initial_log_record_fast(ptr, MLOG_WRITE_STRING,
log_ptr, mtr);
mach_write_to_2(log_ptr, page_offset(ptr));
log_ptr += 2;
mach_write_to_2(log_ptr, len);
log_ptr += 2;
mlog_close(mtr, log_ptr);
mlog_catenate_string(mtr, ptr, len);
}
/********************************************************//**
Writes 8 bytes to a file page buffered in the buffer pool.
Writes the corresponding log record to the mini-transaction log. */
UNIV_INTERN
void
mlog_write_dulint(
/*==============*/
byte* ptr, /*!< in: pointer where to write */
dulint val, /*!< in: value to write */
mtr_t* mtr) /*!< in: mini-transaction handle */
{
byte* log_ptr;
ut_ad(ptr && mtr);
mach_write_to_8(ptr, val);
log_ptr = mlog_open(mtr, 11 + 2 + 9);
/* If no logging is requested, we may return now */
if (log_ptr == NULL) {
return;
}
log_ptr = mlog_write_initial_log_record_fast(ptr, MLOG_8BYTES,
log_ptr, mtr);
mach_write_to_2(log_ptr, page_offset(ptr));
log_ptr += 2;
log_ptr += mach_dulint_write_compressed(log_ptr, val);
mlog_close(mtr, log_ptr);
}
/**********************************************************************//**
Set the next and previous pointers in the undo page for the undo record
that was written to ptr. Update the first free value by the number of bytes
written for this undo record.
@return offset of the inserted entry on the page if succeeded, 0 if fail */
static
ulint
trx_undo_page_set_next_prev_and_add(
/*================================*/
page_t* undo_page, /*!< in/out: undo log page */
byte* ptr, /*!< in: ptr up to where data has been
written on this undo page. */
mtr_t* mtr) /*!< in: mtr */
{
ulint first_free; /*!< offset within undo_page */
ulint end_of_rec; /*!< offset within undo_page */
byte* ptr_to_first_free;
/* pointer within undo_page
that points to the next free
offset value within undo_page.*/
ut_ad(ptr > undo_page);
ut_ad(ptr < undo_page + UNIV_PAGE_SIZE);
if (UNIV_UNLIKELY(trx_undo_left(undo_page, ptr) < 2)) {
return(0);
}
ptr_to_first_free = undo_page + TRX_UNDO_PAGE_HDR + TRX_UNDO_PAGE_FREE;
first_free = mach_read_from_2(ptr_to_first_free);
/* Write offset of the previous undo log record */
mach_write_to_2(ptr, first_free);
ptr += 2;
end_of_rec = ptr - undo_page;
/* Write offset of the next undo log record */
mach_write_to_2(undo_page + first_free, end_of_rec);
/* Update the offset to first free undo record */
mach_write_to_2(ptr_to_first_free, end_of_rec);
/* Write this log entry to the UNDO log */
trx_undof_page_add_undo_rec_log(undo_page, first_free,
end_of_rec, mtr);
return(first_free);
}
/***********************************************************//**
Parses a redo log record of adding an undo log record.
@return end of log record or NULL */
UNIV_INTERN
byte*
trx_undo_parse_add_undo_rec(
/*========================*/
byte* ptr, /*!< in: buffer */
byte* end_ptr,/*!< in: buffer end */
page_t* page) /*!< in: page or NULL */
{
ulint len;
byte* rec;
ulint first_free;
if (end_ptr < ptr + 2) {
return(NULL);
}
len = mach_read_from_2(ptr);
ptr += 2;
if (end_ptr < ptr + len) {
return(NULL);
}
if (page == NULL) {
return(ptr + len);
}
first_free = mach_read_from_2(page + TRX_UNDO_PAGE_HDR
+ TRX_UNDO_PAGE_FREE);
rec = page + first_free;
mach_write_to_2(rec, first_free + 4 + len);
mach_write_to_2(rec + 2 + len, first_free);
mach_write_to_2(page + TRX_UNDO_PAGE_HDR + TRX_UNDO_PAGE_FREE,
first_free + 4 + len);
ut_memcpy(rec + 2, ptr, len);
return(ptr + len);
}
void
mlog_write_ulint(
/*=============*/
byte* ptr, /* in: pointer where to write */
ulint val, /* in: value to write */
byte type, /* in: MLOG_1BYTE, MLOG_2BYTES, MLOG_4BYTES */
mtr_t* mtr) /* in: mini-transaction handle */
{
byte* log_ptr;
if (ptr < buf_pool->frame_zero || ptr >= buf_pool->high_end) {
fprintf(stderr,
"InnoDB: Error: trying to write to"
" a stray memory location %p\n", (void*) ptr);
ut_error;
}
if (type == MLOG_1BYTE) {
mach_write_to_1(ptr, val);
} else if (type == MLOG_2BYTES) {
mach_write_to_2(ptr, val);
} else {
ut_ad(type == MLOG_4BYTES);
mach_write_to_4(ptr, val);
}
log_ptr = mlog_open(mtr, 11 + 2 + 5);
/* If no logging is requested, we may return now */
if (log_ptr == NULL) {
return;
}
log_ptr = mlog_write_initial_log_record_fast(ptr, type, log_ptr, mtr);
mach_write_to_2(log_ptr, ptr - buf_frame_align(ptr));
log_ptr += 2;
log_ptr += mach_write_compressed(log_ptr, val);
mlog_close(mtr, log_ptr);
}
/********************************************************//**
Writes 1 - 4 bytes to a file page buffered in the buffer pool.
Writes the corresponding log record to the mini-transaction log. */
UNIV_INTERN
void
mlog_write_ulint(
/*=============*/
byte* ptr, /*!< in: pointer where to write */
ulint val, /*!< in: value to write */
byte type, /*!< in: MLOG_1BYTE, MLOG_2BYTES, MLOG_4BYTES */
mtr_t* mtr) /*!< in: mini-transaction handle */
{
byte* log_ptr;
switch (type) {
case MLOG_1BYTE:
mach_write_to_1(ptr, val);
break;
case MLOG_2BYTES:
mach_write_to_2(ptr, val);
break;
case MLOG_4BYTES:
mach_write_to_4(ptr, val);
break;
default:
ut_error;
}
log_ptr = mlog_open(mtr, 11 + 2 + 5);
/* If no logging is requested, we may return now */
if (log_ptr == NULL) {
return;
}
log_ptr = mlog_write_initial_log_record_fast(ptr, type, log_ptr, mtr);
mach_write_to_2(log_ptr, page_offset(ptr));
log_ptr += 2;
log_ptr += mach_write_compressed(log_ptr, val);
mlog_close(mtr, log_ptr);
}
void
mlog_write_string(
/*==============*/
byte* ptr, /* in: pointer where to write */
const byte* str, /* in: string to write */
ulint len, /* in: string length */
mtr_t* mtr) /* in: mini-transaction handle */
{
byte* log_ptr;
if (UNIV_UNLIKELY(ptr < buf_pool->frame_zero)
|| UNIV_UNLIKELY(ptr >= buf_pool->high_end)) {
fprintf(stderr,
"InnoDB: Error: trying to write to"
" a stray memory location %p\n", (void*) ptr);
ut_error;
}
ut_ad(ptr && mtr);
ut_a(len < UNIV_PAGE_SIZE);
ut_memcpy(ptr, str, len);
log_ptr = mlog_open(mtr, 30);
/* If no logging is requested, we may return now */
if (log_ptr == NULL) {
return;
}
log_ptr = mlog_write_initial_log_record_fast(ptr, MLOG_WRITE_STRING,
log_ptr, mtr);
mach_write_to_2(log_ptr, ptr - buf_frame_align(ptr));
log_ptr += 2;
mach_write_to_2(log_ptr, len);
log_ptr += 2;
mlog_close(mtr, log_ptr);
mlog_catenate_string(mtr, str, len);
}
/************************************************************************
Initializes the fields in an undo log segment page. */
static
void
trx_undo_page_init(
/*================*/
page_t* undo_page, /* in: undo log segment page */
ulint type, /* in: undo log segment type */
mtr_t* mtr) /* in: mtr */
{
trx_upagef_t* page_hdr;
page_hdr = undo_page + TRX_UNDO_PAGE_HDR;
mach_write_to_2(page_hdr + TRX_UNDO_PAGE_TYPE, type);
mach_write_to_2(page_hdr + TRX_UNDO_PAGE_START,
TRX_UNDO_PAGE_HDR + TRX_UNDO_PAGE_HDR_SIZE);
mach_write_to_2(page_hdr + TRX_UNDO_PAGE_FREE,
TRX_UNDO_PAGE_HDR + TRX_UNDO_PAGE_HDR_SIZE);
fil_page_set_type(undo_page, FIL_PAGE_UNDO_LOG);
trx_undo_page_init_log(undo_page, type, mtr);
}
void
mlog_write_dulint(
/*==============*/
byte* ptr, /* in: pointer where to write */
dulint val, /* in: value to write */
mtr_t* mtr) /* in: mini-transaction handle */
{
byte* log_ptr;
if (UNIV_UNLIKELY(ptr < buf_pool->frame_zero)
|| UNIV_UNLIKELY(ptr >= buf_pool->high_end)) {
fprintf(stderr,
"InnoDB: Error: trying to write to"
" a stray memory location %p\n", (void*) ptr);
ut_error;
}
ut_ad(ptr && mtr);
mach_write_to_8(ptr, val);
log_ptr = mlog_open(mtr, 11 + 2 + 9);
/* If no logging is requested, we may return now */
if (log_ptr == NULL) {
return;
}
log_ptr = mlog_write_initial_log_record_fast(ptr, MLOG_8BYTES,
log_ptr, mtr);
mach_write_to_2(log_ptr, ptr - buf_frame_align(ptr));
log_ptr += 2;
log_ptr += mach_dulint_write_compressed(log_ptr, val);
mlog_close(mtr, log_ptr);
}
/**********************************************************************//**
Writes the mtr log entry of the inserted undo log record on the undo log
page. */
UNIV_INLINE
void
trx_undof_page_add_undo_rec_log(
/*============================*/
page_t* undo_page, /*!< in: undo log page */
ulint old_free, /*!< in: start offset of the inserted entry */
ulint new_free, /*!< in: end offset of the entry */
mtr_t* mtr) /*!< in: mtr */
{
byte* log_ptr;
const byte* log_end;
ulint len;
log_ptr = mlog_open(mtr, 11 + 13 + MLOG_BUF_MARGIN);
if (log_ptr == NULL) {
return;
}
log_end = &log_ptr[11 + 13 + MLOG_BUF_MARGIN];
log_ptr = mlog_write_initial_log_record_fast(
undo_page, MLOG_UNDO_INSERT, log_ptr, mtr);
len = new_free - old_free - 4;
mach_write_to_2(log_ptr, len);
log_ptr += 2;
if (log_ptr + len <= log_end) {
memcpy(log_ptr, undo_page + old_free + 2, len);
mlog_close(mtr, log_ptr + len);
} else {
mlog_close(mtr, log_ptr);
mlog_catenate_string(mtr, undo_page + old_free + 2, len);
}
}
/********************************************************//**
Parses a log record written by mlog_write_ulint or mlog_write_dulint.
@return parsed record end, NULL if not a complete record or a corrupt record */
UNIV_INTERN
byte*
mlog_parse_nbytes(
/*==============*/
ulint type, /*!< in: log record type: MLOG_1BYTE, ... */
byte* ptr, /*!< in: buffer */
byte* end_ptr,/*!< in: buffer end */
byte* page, /*!< in: page where to apply the log record, or NULL */
void* page_zip)/*!< in/out: compressed page, or NULL */
{
ulint offset;
ulint val;
dulint dval;
ut_a(type <= MLOG_8BYTES);
ut_a(!page || !page_zip || fil_page_get_type(page) != FIL_PAGE_INDEX);
if (end_ptr < ptr + 2) {
return(NULL);
}
offset = mach_read_from_2(ptr);
ptr += 2;
if (offset >= UNIV_PAGE_SIZE) {
recv_sys->found_corrupt_log = TRUE;
return(NULL);
}
if (type == MLOG_8BYTES) {
ptr = mach_dulint_parse_compressed(ptr, end_ptr, &dval);
if (ptr == NULL) {
return(NULL);
}
if (page) {
if (UNIV_LIKELY_NULL(page_zip)) {
mach_write_to_8
(((page_zip_des_t*) page_zip)->data
+ offset, dval);
}
mach_write_to_8(page + offset, dval);
}
return(ptr);
}
ptr = mach_parse_compressed(ptr, end_ptr, &val);
if (ptr == NULL) {
return(NULL);
}
switch (type) {
case MLOG_1BYTE:
if (UNIV_UNLIKELY(val > 0xFFUL)) {
goto corrupt;
}
if (page) {
if (UNIV_LIKELY_NULL(page_zip)) {
mach_write_to_1
(((page_zip_des_t*) page_zip)->data
+ offset, val);
}
mach_write_to_1(page + offset, val);
}
break;
case MLOG_2BYTES:
if (UNIV_UNLIKELY(val > 0xFFFFUL)) {
goto corrupt;
}
if (page) {
if (UNIV_LIKELY_NULL(page_zip)) {
mach_write_to_2
(((page_zip_des_t*) page_zip)->data
+ offset, val);
}
mach_write_to_2(page + offset, val);
}
break;
case MLOG_4BYTES:
if (page) {
if (UNIV_LIKELY_NULL(page_zip)) {
mach_write_to_4
(((page_zip_des_t*) page_zip)->data
+ offset, val);
}
mach_write_to_4(page + offset, val);
}
break;
default:
corrupt:
recv_sys->found_corrupt_log = TRUE;
ptr = NULL;
}
return(ptr);
}
/**********************************************************************//**
Reports in the undo log of an update or delete marking of a clustered index
record.
@return byte offset of the inserted undo log entry on the page if
succeed, 0 if fail */
static
ulint
trx_undo_page_report_modify(
/*========================*/
page_t* undo_page, /*!< in: undo log page */
trx_t* trx, /*!< in: transaction */
dict_index_t* index, /*!< in: clustered index where update or
delete marking is done */
const rec_t* rec, /*!< in: clustered index record which
has NOT yet been modified */
const ulint* offsets, /*!< in: rec_get_offsets(rec, index) */
const upd_t* update, /*!< in: update vector which tells the
columns to be updated; in the case of
a delete, this should be set to NULL */
ulint cmpl_info, /*!< in: compiler info on secondary
index updates */
mtr_t* mtr) /*!< in: mtr */
{
dict_table_t* table;
ulint first_free;
byte* ptr;
const byte* field;
ulint flen;
ulint col_no;
ulint type_cmpl;
byte* type_cmpl_ptr;
ulint i;
trx_id_t trx_id;
ibool ignore_prefix = FALSE;
byte ext_buf[REC_MAX_INDEX_COL_LEN
+ BTR_EXTERN_FIELD_REF_SIZE];
ut_a(dict_index_is_clust(index));
ut_ad(rec_offs_validate(rec, index, offsets));
ut_ad(mach_read_from_2(undo_page + TRX_UNDO_PAGE_HDR
+ TRX_UNDO_PAGE_TYPE) == TRX_UNDO_UPDATE);
table = index->table;
first_free = mach_read_from_2(undo_page + TRX_UNDO_PAGE_HDR
+ TRX_UNDO_PAGE_FREE);
ptr = undo_page + first_free;
ut_ad(first_free <= UNIV_PAGE_SIZE);
if (trx_undo_left(undo_page, ptr) < 50) {
/* NOTE: the value 50 must be big enough so that the general
fields written below fit on the undo log page */
return(0);
}
/* Reserve 2 bytes for the pointer to the next undo log record */
ptr += 2;
/* Store first some general parameters to the undo log */
if (!update) {
type_cmpl = TRX_UNDO_DEL_MARK_REC;
} else if (rec_get_deleted_flag(rec, dict_table_is_comp(table))) {
type_cmpl = TRX_UNDO_UPD_DEL_REC;
/* We are about to update a delete marked record.
We don't typically need the prefix in this case unless
the delete marking is done by the same transaction
(which we check below). */
ignore_prefix = TRUE;
} else {
type_cmpl = TRX_UNDO_UPD_EXIST_REC;
}
type_cmpl |= cmpl_info * TRX_UNDO_CMPL_INFO_MULT;
type_cmpl_ptr = ptr;
*ptr++ = (byte) type_cmpl;
ptr += mach_dulint_write_much_compressed(ptr, trx->undo_no);
ptr += mach_dulint_write_much_compressed(ptr, table->id);
/*----------------------------------------*/
/* Store the state of the info bits */
*ptr++ = (byte) rec_get_info_bits(rec, dict_table_is_comp(table));
/* Store the values of the system columns */
field = rec_get_nth_field(rec, offsets,
dict_index_get_sys_col_pos(
index, DATA_TRX_ID), &flen);
ut_ad(flen == DATA_TRX_ID_LEN);
trx_id = trx_read_trx_id(field);
/* If it is an update of a delete marked record, then we are
allowed to ignore blob prefixes if the delete marking was done
by some other trx as it must have committed by now for us to
allow an over-write. */
if (ignore_prefix) {
ignore_prefix = ut_dulint_cmp(trx_id, trx->id) != 0;
}
ptr += mach_dulint_write_compressed(ptr, trx_id);
field = rec_get_nth_field(rec, offsets,
dict_index_get_sys_col_pos(
index, DATA_ROLL_PTR), &flen);
ut_ad(flen == DATA_ROLL_PTR_LEN);
ptr += mach_dulint_write_compressed(ptr, trx_read_roll_ptr(field));
/*----------------------------------------*/
/* Store then the fields required to uniquely determine the
record which will be modified in the clustered index */
for (i = 0; i < dict_index_get_n_unique(index); i++) {
field = rec_get_nth_field(rec, offsets, i, &flen);
/* The ordering columns must not be stored externally. */
ut_ad(!rec_offs_nth_extern(offsets, i));
ut_ad(dict_index_get_nth_col(index, i)->ord_part);
if (trx_undo_left(undo_page, ptr) < 5) {
return(0);
}
ptr += mach_write_compressed(ptr, flen);
if (flen != UNIV_SQL_NULL) {
if (trx_undo_left(undo_page, ptr) < flen) {
return(0);
}
ut_memcpy(ptr, field, flen);
ptr += flen;
}
}
/*----------------------------------------*/
/* Save to the undo log the old values of the columns to be updated. */
if (update) {
if (trx_undo_left(undo_page, ptr) < 5) {
return(0);
}
ptr += mach_write_compressed(ptr, upd_get_n_fields(update));
for (i = 0; i < upd_get_n_fields(update); i++) {
ulint pos = upd_get_nth_field(update, i)->field_no;
/* Write field number to undo log */
if (trx_undo_left(undo_page, ptr) < 5) {
return(0);
}
ptr += mach_write_compressed(ptr, pos);
/* Save the old value of field */
field = rec_get_nth_field(rec, offsets, pos, &flen);
if (trx_undo_left(undo_page, ptr) < 15) {
return(0);
}
if (rec_offs_nth_extern(offsets, pos)) {
ptr = trx_undo_page_report_modify_ext(
ptr,
dict_index_get_nth_col(index, pos)
->ord_part
&& !ignore_prefix
&& flen < REC_MAX_INDEX_COL_LEN
? ext_buf : NULL,
dict_table_zip_size(table),
&field, &flen);
/* Notify purge that it eventually has to
free the old externally stored field */
trx->update_undo->del_marks = TRUE;
*type_cmpl_ptr |= TRX_UNDO_UPD_EXTERN;
} else {
ptr += mach_write_compressed(ptr, flen);
}
if (flen != UNIV_SQL_NULL) {
if (trx_undo_left(undo_page, ptr) < flen) {
return(0);
}
ut_memcpy(ptr, field, flen);
ptr += flen;
}
}
}
/*----------------------------------------*/
/* In the case of a delete marking, and also in the case of an update
where any ordering field of any index changes, store the values of all
columns which occur as ordering fields in any index. This info is used
in the purge of old versions where we use it to build and search the
delete marked index records, to look if we can remove them from the
index tree. Note that starting from 4.0.14 also externally stored
fields can be ordering in some index. Starting from 5.2, we no longer
store REC_MAX_INDEX_COL_LEN first bytes to the undo log record,
but we can construct the column prefix fields in the index by
fetching the first page of the BLOB that is pointed to by the
clustered index. This works also in crash recovery, because all pages
(including BLOBs) are recovered before anything is rolled back. */
if (!update || !(cmpl_info & UPD_NODE_NO_ORD_CHANGE)) {
byte* old_ptr = ptr;
trx->update_undo->del_marks = TRUE;
if (trx_undo_left(undo_page, ptr) < 5) {
return(0);
}
/* Reserve 2 bytes to write the number of bytes the stored
fields take in this undo record */
ptr += 2;
for (col_no = 0; col_no < dict_table_get_n_cols(table);
col_no++) {
const dict_col_t* col
= dict_table_get_nth_col(table, col_no);
if (col->ord_part) {
ulint pos;
/* Write field number to undo log */
if (trx_undo_left(undo_page, ptr) < 5 + 15) {
return(0);
}
pos = dict_index_get_nth_col_pos(index,
col_no);
ptr += mach_write_compressed(ptr, pos);
/* Save the old value of field */
field = rec_get_nth_field(rec, offsets, pos,
&flen);
if (rec_offs_nth_extern(offsets, pos)) {
ptr = trx_undo_page_report_modify_ext(
ptr,
flen < REC_MAX_INDEX_COL_LEN
&& !ignore_prefix
? ext_buf : NULL,
dict_table_zip_size(table),
&field, &flen);
} else {
ptr += mach_write_compressed(
ptr, flen);
}
if (flen != UNIV_SQL_NULL) {
if (trx_undo_left(undo_page, ptr)
< flen) {
return(0);
}
ut_memcpy(ptr, field, flen);
ptr += flen;
}
}
}
mach_write_to_2(old_ptr, ptr - old_ptr);
}
/*----------------------------------------*/
/* Write pointers to the previous and the next undo log records */
if (trx_undo_left(undo_page, ptr) < 2) {
return(0);
}
mach_write_to_2(ptr, first_free);
ptr += 2;
mach_write_to_2(undo_page + first_free, ptr - undo_page);
mach_write_to_2(undo_page + TRX_UNDO_PAGE_HDR + TRX_UNDO_PAGE_FREE,
ptr - undo_page);
/* Write to the REDO log about this change in the UNDO log */
trx_undof_page_add_undo_rec_log(undo_page, first_free,
ptr - undo_page, mtr);
return(first_free);
}
byte*
mlog_parse_nbytes(
/*==============*/
/* out: parsed record end, NULL if not a complete
record or a corrupt record */
ulint type, /* in: log record type: MLOG_1BYTE, ... */
byte* ptr, /* in: buffer */
byte* end_ptr,/* in: buffer end */
byte* page) /* in: page where to apply the log record, or NULL */
{
ulint offset;
ulint val;
dulint dval;
ut_a(type <= MLOG_8BYTES);
if (end_ptr < ptr + 2) {
return(NULL);
}
offset = mach_read_from_2(ptr);
ptr += 2;
if (offset >= UNIV_PAGE_SIZE) {
recv_sys->found_corrupt_log = TRUE;
return(NULL);
}
if (type == MLOG_8BYTES) {
ptr = mach_dulint_parse_compressed(ptr, end_ptr, &dval);
if (ptr == NULL) {
return(NULL);
}
if (page) {
mach_write_to_8(page + offset, dval);
}
return(ptr);
}
ptr = mach_parse_compressed(ptr, end_ptr, &val);
if (ptr == NULL) {
return(NULL);
}
if (type == MLOG_1BYTE) {
if (val > 0xFFUL) {
recv_sys->found_corrupt_log = TRUE;
return(NULL);
}
} else if (type == MLOG_2BYTES) {
if (val > 0xFFFFUL) {
recv_sys->found_corrupt_log = TRUE;
return(NULL);
}
} else {
if (type != MLOG_4BYTES) {
recv_sys->found_corrupt_log = TRUE;
return(NULL);
}
}
if (page) {
if (type == MLOG_1BYTE) {
mach_write_to_1(page + offset, val);
} else if (type == MLOG_2BYTES) {
mach_write_to_2(page + offset, val);
} else {
ut_a(type == MLOG_4BYTES);
mach_write_to_4(page + offset, val);
}
}
return(ptr);
}
/**************************************************************************
Reports in the undo log of an update or delete marking of a clustered index
record. */
static
ulint
trx_undo_page_report_modify(
/*========================*/
/* out: byte offset of the inserted
undo log entry on the page if succeed,
0 if fail */
page_t* undo_page, /* in: undo log page */
trx_t* trx, /* in: transaction */
dict_index_t* index, /* in: clustered index where update or
delete marking is done */
rec_t* rec, /* in: clustered index record which
has NOT yet been modified */
const ulint* offsets, /* in: rec_get_offsets(rec, index) */
upd_t* update, /* in: update vector which tells the
columns to be updated; in the case of
a delete, this should be set to NULL */
ulint cmpl_info, /* in: compiler info on secondary
index updates */
mtr_t* mtr) /* in: mtr */
{
dict_table_t* table;
upd_field_t* upd_field;
ulint first_free;
byte* ptr;
ulint len;
byte* field;
ulint flen;
ulint pos;
dulint roll_ptr;
dulint trx_id;
ulint bits;
ulint col_no;
byte* old_ptr;
ulint type_cmpl;
byte* type_cmpl_ptr;
ulint i;
ut_a(index->type & DICT_CLUSTERED);
ut_ad(rec_offs_validate(rec, index, offsets));
ut_ad(mach_read_from_2(undo_page + TRX_UNDO_PAGE_HDR
+ TRX_UNDO_PAGE_TYPE) == TRX_UNDO_UPDATE);
table = index->table;
first_free = mach_read_from_2(undo_page + TRX_UNDO_PAGE_HDR
+ TRX_UNDO_PAGE_FREE);
ptr = undo_page + first_free;
ut_ad(first_free <= UNIV_PAGE_SIZE);
if (trx_undo_left(undo_page, ptr) < 50) {
/* NOTE: the value 50 must be big enough so that the general
fields written below fit on the undo log page */
return(0);
}
/* Reserve 2 bytes for the pointer to the next undo log record */
ptr += 2;
/* Store first some general parameters to the undo log */
if (update) {
if (rec_get_deleted_flag(rec, dict_table_is_comp(table))) {
type_cmpl = TRX_UNDO_UPD_DEL_REC;
} else {
type_cmpl = TRX_UNDO_UPD_EXIST_REC;
}
} else {
type_cmpl = TRX_UNDO_DEL_MARK_REC;
}
type_cmpl = type_cmpl | (cmpl_info * TRX_UNDO_CMPL_INFO_MULT);
mach_write_to_1(ptr, type_cmpl);
type_cmpl_ptr = ptr;
ptr++;
len = mach_dulint_write_much_compressed(ptr, trx->undo_no);
ptr += len;
len = mach_dulint_write_much_compressed(ptr, table->id);
ptr += len;
/*----------------------------------------*/
/* Store the state of the info bits */
bits = rec_get_info_bits(rec, dict_table_is_comp(table));
mach_write_to_1(ptr, bits);
ptr += 1;
/* Store the values of the system columns */
field = rec_get_nth_field(rec, offsets,
dict_index_get_sys_col_pos(
index, DATA_TRX_ID), &len);
ut_ad(len == DATA_TRX_ID_LEN);
trx_id = trx_read_trx_id(field);
field = rec_get_nth_field(rec, offsets,
dict_index_get_sys_col_pos(
index, DATA_ROLL_PTR), &len);
ut_ad(len == DATA_ROLL_PTR_LEN);
roll_ptr = trx_read_roll_ptr(field);
len = mach_dulint_write_compressed(ptr, trx_id);
ptr += len;
len = mach_dulint_write_compressed(ptr, roll_ptr);
ptr += len;
/*----------------------------------------*/
/* Store then the fields required to uniquely determine the
record which will be modified in the clustered index */
for (i = 0; i < dict_index_get_n_unique(index); i++) {
field = rec_get_nth_field(rec, offsets, i, &flen);
if (trx_undo_left(undo_page, ptr) < 4) {
return(0);
}
len = mach_write_compressed(ptr, flen);
ptr += len;
if (flen != UNIV_SQL_NULL) {
if (trx_undo_left(undo_page, ptr) < flen) {
return(0);
}
ut_memcpy(ptr, field, flen);
ptr += flen;
}
}
/*----------------------------------------*/
/* Save to the undo log the old values of the columns to be updated. */
if (update) {
if (trx_undo_left(undo_page, ptr) < 5) {
return(0);
}
len = mach_write_compressed(ptr, upd_get_n_fields(update));
ptr += len;
for (i = 0; i < upd_get_n_fields(update); i++) {
upd_field = upd_get_nth_field(update, i);
pos = upd_field->field_no;
/* Write field number to undo log */
if (trx_undo_left(undo_page, ptr) < 5) {
return(0);
}
len = mach_write_compressed(ptr, pos);
ptr += len;
/* Save the old value of field */
field = rec_get_nth_field(rec, offsets, pos, &flen);
if (trx_undo_left(undo_page, ptr) < 5) {
return(0);
}
if (rec_offs_nth_extern(offsets, pos)) {
/* If a field has external storage, we add
to flen the flag */
len = mach_write_compressed(
ptr,
UNIV_EXTERN_STORAGE_FIELD + flen);
/* Notify purge that it eventually has to
free the old externally stored field */
trx->update_undo->del_marks = TRUE;
*type_cmpl_ptr = *type_cmpl_ptr
| TRX_UNDO_UPD_EXTERN;
} else {
len = mach_write_compressed(ptr, flen);
}
ptr += len;
if (flen != UNIV_SQL_NULL) {
if (trx_undo_left(undo_page, ptr) < flen) {
return(0);
}
ut_memcpy(ptr, field, flen);
ptr += flen;
}
}
}
/*----------------------------------------*/
/* In the case of a delete marking, and also in the case of an update
where any ordering field of any index changes, store the values of all
columns which occur as ordering fields in any index. This info is used
in the purge of old versions where we use it to build and search the
delete marked index records, to look if we can remove them from the
index tree. Note that starting from 4.0.14 also externally stored
fields can be ordering in some index. But we always store at least
384 first bytes locally to the clustered index record, which means
we can construct the column prefix fields in the index from the
stored data. */
if (!update || !(cmpl_info & UPD_NODE_NO_ORD_CHANGE)) {
trx->update_undo->del_marks = TRUE;
if (trx_undo_left(undo_page, ptr) < 5) {
return(0);
}
old_ptr = ptr;
/* Reserve 2 bytes to write the number of bytes the stored
fields take in this undo record */
ptr += 2;
for (col_no = 0; col_no < dict_table_get_n_cols(table);
col_no++) {
const dict_col_t* col
= dict_table_get_nth_col(table, col_no);
if (col->ord_part > 0) {
pos = dict_index_get_nth_col_pos(index,
col_no);
/* Write field number to undo log */
if (trx_undo_left(undo_page, ptr) < 5) {
return(0);
}
len = mach_write_compressed(ptr, pos);
ptr += len;
/* Save the old value of field */
field = rec_get_nth_field(rec, offsets, pos,
&flen);
if (trx_undo_left(undo_page, ptr) < 5) {
return(0);
}
len = mach_write_compressed(ptr, flen);
ptr += len;
if (flen != UNIV_SQL_NULL) {
if (trx_undo_left(undo_page, ptr)
< flen) {
return(0);
}
ut_memcpy(ptr, field, flen);
ptr += flen;
}
}
}
mach_write_to_2(old_ptr, ptr - old_ptr);
}
/*----------------------------------------*/
/* Write pointers to the previous and the next undo log records */
if (trx_undo_left(undo_page, ptr) < 2) {
return(0);
}
mach_write_to_2(ptr, first_free);
ptr += 2;
mach_write_to_2(undo_page + first_free, ptr - undo_page);
mach_write_to_2(undo_page + TRX_UNDO_PAGE_HDR + TRX_UNDO_PAGE_FREE,
ptr - undo_page);
/* Write to the REDO log about this change in the UNDO log */
trx_undof_page_add_undo_rec_log(undo_page, first_free,
ptr - undo_page, mtr);
return(first_free);
}
/**************************************************************************
Reports in the undo log of an insert of a clustered index record. */
static
ulint
trx_undo_page_report_insert(
/*========================*/
/* out: offset of the inserted entry
on the page if succeed, 0 if fail */
page_t* undo_page, /* in: undo log page */
trx_t* trx, /* in: transaction */
dict_index_t* index, /* in: clustered index */
dtuple_t* clust_entry, /* in: index entry which will be
inserted to the clustered index */
mtr_t* mtr) /* in: mtr */
{
ulint first_free;
byte* ptr;
ulint len;
dfield_t* field;
ulint flen;
ulint i;
ut_ad(mach_read_from_2(undo_page + TRX_UNDO_PAGE_HDR
+ TRX_UNDO_PAGE_TYPE) == TRX_UNDO_INSERT);
first_free = mach_read_from_2(undo_page + TRX_UNDO_PAGE_HDR
+ TRX_UNDO_PAGE_FREE);
ptr = undo_page + first_free;
ut_ad(first_free <= UNIV_PAGE_SIZE);
if (trx_undo_left(undo_page, ptr) < 30) {
/* NOTE: the value 30 must be big enough such that the general
fields written below fit on the undo log page */
return(0);
}
/* Reserve 2 bytes for the pointer to the next undo log record */
ptr += 2;
/* Store first some general parameters to the undo log */
mach_write_to_1(ptr, TRX_UNDO_INSERT_REC);
ptr++;
len = mach_dulint_write_much_compressed(ptr, trx->undo_no);
ptr += len;
len = mach_dulint_write_much_compressed(ptr, (index->table)->id);
ptr += len;
/*----------------------------------------*/
/* Store then the fields required to uniquely determine the record
to be inserted in the clustered index */
for (i = 0; i < dict_index_get_n_unique(index); i++) {
field = dtuple_get_nth_field(clust_entry, i);
flen = dfield_get_len(field);
if (trx_undo_left(undo_page, ptr) < 5) {
return(0);
}
len = mach_write_compressed(ptr, flen);
ptr += len;
if (flen != UNIV_SQL_NULL) {
if (trx_undo_left(undo_page, ptr) < flen) {
return(0);
}
ut_memcpy(ptr, dfield_get_data(field), flen);
ptr += flen;
}
}
if (trx_undo_left(undo_page, ptr) < 2) {
return(0);
}
/*----------------------------------------*/
/* Write pointers to the previous and the next undo log records */
if (trx_undo_left(undo_page, ptr) < 2) {
return(0);
}
mach_write_to_2(ptr, first_free);
ptr += 2;
mach_write_to_2(undo_page + first_free, ptr - undo_page);
mach_write_to_2(undo_page + TRX_UNDO_PAGE_HDR + TRX_UNDO_PAGE_FREE,
ptr - undo_page);
/* Write the log entry to the REDO log of this change in the UNDO
log */
trx_undof_page_add_undo_rec_log(undo_page, first_free,
ptr - undo_page, mtr);
return(first_free);
}
/********************************************************//**
Opens a buffer for mlog, writes the initial log record and,
if needed, the field lengths of an index.
@return buffer, NULL if log mode MTR_LOG_NONE */
UNIV_INTERN
byte*
mlog_open_and_write_index(
/*======================*/
mtr_t* mtr, /*!< in: mtr */
const byte* rec, /*!< in: index record or page */
dict_index_t* index, /*!< in: record descriptor */
byte type, /*!< in: log item type */
ulint size) /*!< in: requested buffer size in bytes
(if 0, calls mlog_close() and returns NULL) */
{
byte* log_ptr;
const byte* log_start;
const byte* log_end;
ut_ad(!!page_rec_is_comp(rec) == dict_table_is_comp(index->table));
if (!page_rec_is_comp(rec)) {
log_start = log_ptr = mlog_open(mtr, 11 + size);
if (!log_ptr) {
return(NULL); /* logging is disabled */
}
log_ptr = mlog_write_initial_log_record_fast(rec, type,
log_ptr, mtr);
log_end = log_ptr + 11 + size;
} else {
ulint i;
ulint n = dict_index_get_n_fields(index);
/* total size needed */
ulint total = 11 + size + (n + 2) * 2;
ulint alloc = total;
/* allocate at most DYN_ARRAY_DATA_SIZE at a time */
if (alloc > DYN_ARRAY_DATA_SIZE) {
alloc = DYN_ARRAY_DATA_SIZE;
}
log_start = log_ptr = mlog_open(mtr, alloc);
if (!log_ptr) {
return(NULL); /* logging is disabled */
}
log_end = log_ptr + alloc;
log_ptr = mlog_write_initial_log_record_fast(rec, type,
log_ptr, mtr);
mach_write_to_2(log_ptr, n);
log_ptr += 2;
mach_write_to_2(log_ptr,
dict_index_get_n_unique_in_tree(index));
log_ptr += 2;
for (i = 0; i < n; i++) {
dict_field_t* field;
const dict_col_t* col;
ulint len;
field = dict_index_get_nth_field(index, i);
col = dict_field_get_col(field);
len = field->fixed_len;
ut_ad(len < 0x7fff);
if (len == 0
&& (col->len > 255 || col->mtype == DATA_BLOB)) {
/* variable-length field
with maximum length > 255 */
len = 0x7fff;
}
if (col->prtype & DATA_NOT_NULL) {
len |= 0x8000;
}
if (log_ptr + 2 > log_end) {
mlog_close(mtr, log_ptr);
ut_a(total > (ulint) (log_ptr - log_start));
total -= log_ptr - log_start;
alloc = total;
if (alloc > DYN_ARRAY_DATA_SIZE) {
alloc = DYN_ARRAY_DATA_SIZE;
}
log_start = log_ptr = mlog_open(mtr, alloc);
if (!log_ptr) {
return(NULL); /* logging is disabled */
}
log_end = log_ptr + alloc;
}
mach_write_to_2(log_ptr, len);
log_ptr += 2;
}
}
if (size == 0) {
mlog_close(mtr, log_ptr);
log_ptr = NULL;
} else if (log_ptr + size > log_end) {
mlog_close(mtr, log_ptr);
log_ptr = mlog_open(mtr, size);
}
return(log_ptr);
}
/********************************************************//**
Opens a buffer for mlog, writes the initial log record and,
if needed, the field lengths of an index.
@return buffer, NULL if log mode MTR_LOG_NONE */
UNIV_INTERN
byte*
mlog_open_and_write_index(
/*======================*/
mtr_t* mtr, /*!< in: mtr */
const byte* rec, /*!< in: index record or page */
dict_index_t* index, /*!< in: record descriptor */
byte type, /*!< in: log item type */
ulint size) /*!< in: requested buffer size in bytes
(if 0, calls mlog_close() and returns NULL) */
{
byte* log_ptr;
const byte* log_start;
const byte* log_end;
ut_ad(!!page_rec_is_comp(rec) == dict_table_is_comp(index->table));
if (!page_rec_is_comp(rec)) {
log_start = log_ptr = mlog_open(mtr, 11 + size);
if (!log_ptr) {
return(NULL); /* logging is disabled */
}
log_ptr = mlog_write_initial_log_record_fast(rec, type,
log_ptr, mtr);
log_end = log_ptr + 11 + size;
} else {
ulint i;
ulint n = dict_index_get_n_fields(index);
ibool is_gcs_cluster = dict_index_is_gcs_clust_after_alter_table(index);
/* total size needed */
/* redo日志有可能需要多两字节，total需要根据实际情况分配空间！ */
ulint total = 11 + size + (n + 2 + (is_gcs_cluster ? 1 : 0)) * 2;
ulint alloc = total;
/* allocate at most DYN_ARRAY_DATA_SIZE at a time */
if (alloc > DYN_ARRAY_DATA_SIZE) {
alloc = DYN_ARRAY_DATA_SIZE;
}
log_start = log_ptr = mlog_open(mtr, alloc);
if (!log_ptr) {
return(NULL); /* logging is disabled */
}
log_end = log_ptr + alloc;
log_ptr = mlog_write_initial_log_record_fast(rec, type,
log_ptr, mtr);
/* 在第一次alter table前，所有gcs表可以当成compact表，即使是redo过程，这样redo log也兼容！ */
if (is_gcs_cluster)
{
//ut_ad(rec_is_gcs(rec) || rec == page_align(rec) || rec_get_status(rec) & REC_STATUS_NODE_PTR); /* rec有可能就是页头，如日志MLOG_COMP_LIST_END_COPY_CREATED */
mach_write_to_2(log_ptr, n | 0x8000); /* 标记是gcs表 */
}
else
{
ut_ad(rec == page_align(rec) || !rec_is_gcs(rec));
mach_write_to_2(log_ptr, n);
}
/* 对于gcs聚集索引，记录第一次alter table前聚集索引的字段数 */
if (is_gcs_cluster)
{
log_ptr += 2;
mach_write_to_2(log_ptr, (ulint)index->n_fields_before_alter);
ut_ad(!index->n_fields_before_alter == !dict_index_is_gcs_clust_after_alter_table(index));
}
log_ptr += 2;
mach_write_to_2(log_ptr,
dict_index_get_n_unique_in_tree(index));
log_ptr += 2;
for (i = 0; i < n; i++) {
dict_field_t* field;
const dict_col_t* col;
ulint len;
field = dict_index_get_nth_field(index, i);
col = dict_field_get_col(field);
len = field->fixed_len;
ut_ad(len < 0x7fff);
if (len == 0
&& (col->len > 255 || col->mtype == DATA_BLOB)) {
/* variable-length field
with maximum length > 255 */
len = 0x7fff;
}
if (col->prtype & DATA_NOT_NULL) {
len |= 0x8000;
}
if (log_ptr + 2 > log_end) {
mlog_close(mtr, log_ptr);
ut_a(total > (ulint) (log_ptr - log_start));
total -= log_ptr - log_start;
alloc = total;
if (alloc > DYN_ARRAY_DATA_SIZE) {
alloc = DYN_ARRAY_DATA_SIZE;
}
log_start = log_ptr = mlog_open(mtr, alloc);
if (!log_ptr) {
return(NULL); /* logging is disabled */
}
log_end = log_ptr + alloc;
}
mach_write_to_2(log_ptr, len);
log_ptr += 2;
}
}
if (size == 0) {
mlog_close(mtr, log_ptr);
log_ptr = NULL;
} else if (log_ptr + size > log_end) {
mlog_close(mtr, log_ptr);
log_ptr = mlog_open(mtr, size);
}
return(log_ptr);
}