1berry/MariaDB-server
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

MDEV-34802 Recovery fails to note some log corruption

Browse Source 
recv_recovery_from_checkpoint_start(): Abort startup due to log
corruption if we were unable to parse the entire log between
the latest log checkpoint and the corresponding FILE_CHECKPOINT record.

Also, reduce some code bloat related to log output and log_sys.mutex.

Reviewed by: Debarun Banerjee
This commit is contained in:
Marko Mäkelä 2024-08-28 15:44:42 +03:00
parent e7bb9b7c55
commit 1ff6b6f0b4
1 changed files with 54 additions and 66 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

120
storage/innobase/log/log0recv.cc
View File

@ -3474,10 +3474,9 @@ recv_recovery_from_checkpoint_start(lsn_t flush_lsn)
ut_d(mysql_mutex_unlock(&buf_pool.flush_list_mutex));
if (srv_force_recovery >= SRV_FORCE_NO_LOG_REDO) {
ib::info() << "innodb_force_recovery=6 skips redo log apply";
return(DB_SUCCESS);
sql_print_information("InnoDB: innodb_force_recovery=6"
" skips redo log apply");
return err;
}
mysql_mutex_lock(&log_sys.mutex);
@ -3493,13 +3492,9 @@ recv_recovery_from_checkpoint_start(lsn_t flush_lsn)
ut_ad(RECV_SCAN_SIZE <= srv_log_buffer_size);
ut_ad(recv_sys.pages.empty());
contiguous_lsn = checkpoint_lsn;
switch (log_sys.log.format) {
case 0:
mysql_mutex_unlock(&log_sys.mutex);
return DB_SUCCESS;
default:
if (end_lsn == 0) {
break;
@ -3509,8 +3504,13 @@ recv_recovery_from_checkpoint_start(lsn_t flush_lsn)
break;
}
recv_sys.found_corrupt_log = true;
err_exit:
err = DB_ERROR;
/* fall through */
func_exit:
case 0:
mysql_mutex_unlock(&log_sys.mutex);
return(DB_ERROR);
return err;
}
size_t sizeof_checkpoint;
@ -3527,14 +3527,15 @@ recv_recovery_from_checkpoint_start(lsn_t flush_lsn)
ut_ad(!recv_sys.found_corrupt_fs || !srv_force_recovery);
if (srv_read_only_mode && recv_needed_recovery) {
mysql_mutex_unlock(&log_sys.mutex);
return(DB_READ_ONLY);
read_only:
err = DB_READ_ONLY;
goto func_exit;
}
if (recv_sys.found_corrupt_log && !srv_force_recovery) {
mysql_mutex_unlock(&log_sys.mutex);
ib::warn() << "Log scan aborted at LSN " << contiguous_lsn;
return(DB_ERROR);
sql_print_warning("InnoDB: Log scan aborted at LSN " LSN_PF,
contiguous_lsn);
goto err_exit;
}
/* If we fail to open a tablespace while looking for FILE_CHECKPOINT, we
@ -3542,14 +3543,12 @@ recv_recovery_from_checkpoint_start(lsn_t flush_lsn)
would not be able to open an encrypted tablespace and the flag could be
set. */
if (recv_sys.found_corrupt_fs) {
mysql_mutex_unlock(&log_sys.mutex);
return DB_ERROR;
goto err_exit;
}
if (recv_sys.mlog_checkpoint_lsn == 0) {
lsn_t scan_lsn = log_sys.log.scanned_lsn;
if (!srv_read_only_mode && scan_lsn != checkpoint_lsn) {
mysql_mutex_unlock(&log_sys.mutex);
ib::error err;
err << "Missing FILE_CHECKPOINT";
if (end_lsn) {
@ -3557,7 +3556,7 @@ recv_recovery_from_checkpoint_start(lsn_t flush_lsn)
}
err << " between the checkpoint " << checkpoint_lsn
<< " and the end " << scan_lsn << ".";
return(DB_ERROR);
goto err_exit;
}
log_sys.log.scanned_lsn = checkpoint_lsn;
@ -3568,8 +3567,7 @@ recv_recovery_from_checkpoint_start(lsn_t flush_lsn)
if ((recv_sys.found_corrupt_log && !srv_force_recovery)
|| recv_sys.found_corrupt_fs) {
mysql_mutex_unlock(&log_sys.mutex);
return(DB_ERROR);
goto err_exit;
}
}
@ -3599,19 +3597,17 @@ completed:
}
if (!recv_needed_recovery) {
ib::info()
<< "The log sequence number " << flush_lsn
<< " in the system tablespace does not match"
" the log sequence number "
<< checkpoint_lsn << " in the "
<< LOG_FILE_NAME << "!";
sql_print_information(
"InnoDB: The log sequence number " LSN_PF
" in the system tablespace does not match"
" the log sequence number " LSN_PF
" in the ib_logfile0!",
flush_lsn, checkpoint_lsn);
if (srv_read_only_mode) {
ib::error() << "innodb_read_only"
" prevents crash recovery";
mysql_mutex_unlock(&log_sys.mutex);
return(DB_READ_ONLY);
sql_print_error("InnoDB: innodb_read_only"
" prevents crash recovery");
goto read_only;
}
recv_needed_recovery = true;
@ -3632,8 +3628,7 @@ completed:
rescan, missing_tablespace);
if (err != DB_SUCCESS) {
mysql_mutex_unlock(&log_sys.mutex);
return(err);
goto func_exit;
}
/* If there is any missing tablespace and rescan is needed
@ -3662,8 +3657,7 @@ completed:
rescan, missing_tablespace);
if (err != DB_SUCCESS) {
mysql_mutex_unlock(&log_sys.mutex);
return err;
goto func_exit;
}
rescan = true;
@ -3686,8 +3680,7 @@ completed:
if ((recv_sys.found_corrupt_log
&& !srv_force_recovery)
|| recv_sys.found_corrupt_fs) {
mysql_mutex_unlock(&log_sys.mutex);
return(DB_ERROR);
goto err_exit;
}
/* In case of multi-batch recovery,
@ -3699,26 +3692,26 @@ completed:
ut_ad(!rescan || recv_sys.pages.empty());
}
if (log_sys.is_physical()
&& (log_sys.log.scanned_lsn < checkpoint_lsn
|| log_sys.log.scanned_lsn < recv_max_page_lsn)) {
ib::error() << "We scanned the log up to "
<< log_sys.log.scanned_lsn
<< ". A checkpoint was at " << checkpoint_lsn << " and"
" the maximum LSN on a database page was "
<< recv_max_page_lsn << ". It is possible that the"
" database is now corrupt!";
}
if (recv_sys.recovered_lsn < checkpoint_lsn) {
mysql_mutex_unlock(&log_sys.mutex);
ib::error() << "Recovered only to lsn:"
<< recv_sys.recovered_lsn
<< " checkpoint_lsn: " << checkpoint_lsn;
return(DB_ERROR);
if (!log_sys.is_physical()) {
} else if (recv_sys.recovered_lsn < checkpoint_lsn
|| recv_sys.recovered_lsn < end_lsn) {
sql_print_error("InnoDB: The log was only scanned up to "
LSN_PF ", while the current LSN at the "
"time of the latest checkpoint " LSN_PF
" was " LSN_PF "!",
recv_sys.recovered_lsn,
checkpoint_lsn, end_lsn);
goto err_exit;
} else if (log_sys.log.scanned_lsn < checkpoint_lsn
|| log_sys.log.scanned_lsn < end_lsn
|| log_sys.log.scanned_lsn < recv_max_page_lsn) {
sql_print_error("InnoDB: We scanned the log up to " LSN_PF
". A checkpoint was at " LSN_PF
" and the maximum LSN on a database page was "
LSN_PF ". It is possible that the"
" database is now corrupt!",
log_sys.log.scanned_lsn, checkpoint_lsn,
recv_max_page_lsn);
}
log_sys.next_checkpoint_lsn = checkpoint_lsn;
@ -3750,9 +3743,7 @@ completed:
log_sys.next_checkpoint_no = ++checkpoint_no;
DBUG_EXECUTE_IF("before_final_redo_apply",
mysql_mutex_unlock(&log_sys.mutex);
return DB_ERROR;);
DBUG_EXECUTE_IF("before_final_redo_apply", goto err_exit;);
mutex_enter(&recv_sys.mutex);
recv_sys.apply_log_recs = true;
@ -3760,17 +3751,14 @@ completed:
ut_d(recv_no_log_write = srv_operation == SRV_OPERATION_RESTORE
|| srv_operation == SRV_OPERATION_RESTORE_EXPORT);
mutex_exit(&recv_sys.mutex);
mysql_mutex_unlock(&log_sys.mutex);
recv_lsn_checks_on = true;
mutex_exit(&recv_sys.mutex);
/* The database is now ready to start almost normal processing of user
transactions: transaction rollbacks and the application of the log
records in the hash table can be run in background. */
return(DB_SUCCESS);
ut_ad(err == DB_SUCCESS);
goto func_exit;
}
bool recv_dblwr_t::validate_page(const page_id_t page_id,