| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
directory
|
|
|
|
|
|
| |
The attached patch fixes the treatment of snprintf return.
Eric Leblond <eric@inl.fr>
|
|
|
|
|
|
|
| |
timeout. It is now not enabled by default. The following patch restores
the reconnection functionnality for the ulogd-mysql plugin.
Eric Leblond <eric@inl.fr>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
- included inttypes.h and changed to use PRId64/PRIu64 format definitions
to use correct number of l's on all platforms (WORDSIZE depending).
- added str flag to the struct _field - it is set in the pgsql_get_columns
function when the IP-Addr columns in the database are non-numeric/string
and used to insert proper converted value in pgsql_output.
See also the --with-pgsql-log-ip-as-string configure option and doc dir.
- added global stmt_siz variable for the allocated size of the stmt buffer;
obsoletes local size variable in pgsql_createstmt
- added guard at the beginning of pgsql_output for the case stmt_val == NULL
- changed to use snprintf instead of sprintf in pgsql_output to take care
about the allocated buffer length - using the STMT_ADD macro when the
string has to be appended.
- fixed to check in pgsql_output() if the buffer contains enough place for
PQescapeString calls -- see pgsql documentation.
- Improved PGSQL_GETCOLUMN_TEMPLATE and PGSQL_GETCOLUMN_TEMPLATE_SCHEMA sql
queries to return the typename of the column, so we can detect if the IP
address columns are strings or integers (str in struct _field).
See also doc/pgsql.table.
- Fixed to use strncat/strncpy in pgsql_open_db instead of strcpy/strcat
- Fixed to free connstr after PQconnectdb.
- Fixed to reset dbh handle after failure in pgsql_open_db,pgsql_init
- Fixed to check dbh handle before PQfinish call in pgsql_fini and to free
the stmt as well.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
- Cleaned up and unified some configure.in / Rules.make.in variables
(sometimes the variable contained -I or -L, sometimes not).
- Fixed to use AC_CHECK_FUNCS for mysql_real_escape_string detection
instead of "strings ..." call. In the past it never worked, because
the MYSQLLIBS variable contained "-L" and -l flags (see above) that
got passed to the strings call.
- Removed PGSQL_LIB="${DATABASE_LIB} -lpq " adding mysql libs to the
set of libs of sqlite3, ...
- Fixed sqlite3 installation check to work on both, 32 and 64 bit
platforms. The 64bit lib has to be checked first or the 32 bit lib
will be used while linking in 64 bit mode.
- Added $(LDFLAGS) to ulogd link rule in Makefile.in
- Somebody fixed the ugly "strings $MYSQLIBS" stuff already to AC_SEARCH_LIBS.
I changed it to my variant using "mysql_config --libs" to get the libs and
just check if the libs contain the mysql_real_escape_string function or not.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* sqlite3
- added global stmt_siz variable for the allocated size of the stmt buffer
(obsoletes local size variable in _sqlite3_createstmt).
- added guard at the beginning of _sqlite3_output so the output is disabled
if the plugin initialization (db open, ...) fails.
- changed many formats to use %zu for size_t variables, esp. in DEBUGP calls
- changed to use snprintf instead of sprintf in pgsql_output to take care
about the allocated buffer length - using the STMT_ADD macro when the
string has to be appended.
- fixed sqlite3_prepare() call error handling.
- fixed to use ULOGD_MAX_KEYLEN-1 in strncpy.
- added severals sqlite3_close(dbh); dbh = NULL; to reset it propelly.
* pcap
- fixed to use pcapsync_ce.u.value instead of pcapf_ce.u.value for flush
- changed to show pcap file name (so the user can see if the default or the
file specified in the config is used).
* mysql-table
- fixed id field definition in mysql.table.ipaddr-as-string; default value
can't be set when auto_increment is used.
* example configuration file:
- fixed to use an usefull default value for sqlite3 db file...
* user option
- added "user" config option, disabled by default (runs as root).
can be overwritten by the already existing "-u" cmd line option.
without this option, the customers have to edit init script or
source some sysconfig file to pass it via -u ... both not needed.
|
| |
|
| |
|
|
|
|
|
| |
- uncomment @DEFS@ to include compilation definitions
- check for autoconf 2.50 (required by AC_SYS_LARGE_FILE)
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
- included inttypes.h and changed to use PRId64/PRIu64 format definitions
to use correct number of l's on all platforms (WORDSIZE depending).
- added str flag to the struct _field - it is set in the mysql_get_columns
function when the IP-Addr columns in the database are non-numeric/string
and used to insert proper converted value.
See also the --with-mysql-log-ip-as-string configure option and doc dir.
- changed to use snprintf instead of sprintf to take care about the
allocated buffer length: STMT_ADD macro
- added global stmt_siz variable for the allocated size of the stmt buffer
(set in mysql_createstmt, used in mysql_output; obsoletes local size
variable in mysql_createstmt).
- Added guard at the beginning of mysql_output() triggering reconnect
if stmt_val == NULL.
- Fixed to check in mysql_output() if the buffer contains enough place for
mysql_*escape_string calls -- see mysql documentation.
- Fixed to close the old db handle before reconnect is initiated on failure
of mysql_real_query() in mysql_output.
- Increased size of allocated stmt buffer in mysql_createstmt() to +1.
- Fixed strncpy calls in mysql_createstmt and mysql_get_columns;
use ULOGD_MAX_KEYLEN-1, not ULOGD_MAX_KEYLEN and terminate the buffer
explicitelly after (in case the source string was longer / not term.).
- Fixed mysql_open_db() to close db handle after connect failure.
- Fixed _mysql_init_db() to close db handle after mysql_get_columns failure.
- Fixed _mysql_fini to call mysql_close(dbh) only if dbh not NULL and reset
it to NULL after.
|
| |
|
|
|
|
|
|
|
|
|
| |
- Fixed sigterm_handler to close logfile and then call output_plugin->fini. The output_plugin->fini function may still use the logger (e.g. sqlite3 does), what causes a SEGV (really happens in case of sqlite3 db init error - as far as I remember)
- Changed sighup_handler_print to not to call exit(2) on reopen failure, but try to reopen and fallback to continue using old descriptor on failure.
- Fixed mac output in ulogd_BASE.c
- Added explicit termination of the hostname buffer: "[...] if namelen is an insufficient length to hold the host name, then the returned name is truncated and it is unspecified whether the returned name is null-terminated. [...]" as SUSv2 specifies.
- Added to print "\n" after ULOGD_RET_NONE type as well after an unspecified type. See example above at extensions/ulogd_BASE.c.
- Fixed to set the '\0' at the end and not behind the buffer.
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
| |
(Vladimir Mosgalin) (Closes: #372)
|
|
|
|
| |
(Closes: #360)
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
cable (sigh) between the firewall and the logger machine, ulogd was unable
to connect to the mysql server. That resulted effectively a hang in the
boot process: being unable to connect mysql_real_connect did not return
(or the socket timeout was so long I was unable to wait out), ulogd could
not reach the point to fork and init could not start the daemons,
processes following ulogd.
The attached simple patch adds the connect_timeout parameter to the MYSQL
section and calls mysql_option when connect_timeout is set.
(Jozsef Kadlecsik)
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|