wiki:Logging

Version 4 (modified by Vojtech Horky, 12 years ago) ( diff )

Start describing the rewritten logging

Logging functions in HelenOS

If you want to use some kind of logging - either for debugging purposes or to report errors - consider using the one provided in libc. The current implementation does not offer any spectacular features but shall be sufficient for most cases.

The logging messages are printed to screen (depending on kernel configuration) and are also written to files. To prevent flooding of the system with logging messages, the messages are assigned severity levels (see table below) and the user specifies how verbose the system shall be - i.e. to which detail are the messages actually printed. To allow more fine-grained tuning of what-to-print, the messages are directed to different logs that form a tree-like structure. It is possible to control reporting level (verbosity) of each log separately, giving user the ability to exactly specify which part of the system shall be monitored in the greatest detail.

For example, a (hypothetical) USB keyboard driver can offer a separate log for each keyboard that is plugged in. Thus, user trying to find out why his special button does not work can specify that all messages for the special keyboard shall be printed but also suppress any other messages - namely from the other keyboard that works and which is used to control the system.

Currently, following levels are recognised.

Name Typical usage
LVL_FATAL Fatal error, program is not able to recover at all.
LVL_ERROR Serious error but the program can recover from it.
E.g. driver cannot control one device but otherwise is healthy.
LVL_WARN Easily recoverable problem, such as one corrupted packet that can be skipped.
LVL_NOTE Message that does not indicate a problem, but should be printed at the default logging level.
LVL_DEBUG Debugging-purpose message. Not printed at the default logging level. Increasing logging level to LVL_DEBUG should not swamp the log.
LVL_DEBUG2 More detailed debugging message.

Going through existing logs

All logs are stored under the /log/ directory and file name corresponds to the server/application where the messages originated. To examine these files, one can use edit or cat.

The file contains the whole tree of the logs - each line is prefixed with severity level of the message and name of the log.

Example: running tester application with logger1 test shall create a /log/tester file with dummy messages. The test tries to print message at every severity level, what is actually stored into the file depends on current settings (see below for more examples).

Changing reported level at boot time

The following currently works only for GRUB-based platforms.

TODO - logger arguments

Changing reported level at run-time

TODO - logset

Using the C API

TODO log_msg, log_init

Behind the scene

TODO - how logger, logset and log_*() works

Obsoleted

Information below is obsoleted and will be rewritten soon.

To use logging functions, include io/log.h and initialize the logging

subsystem. After that, you can use the log_msg to do the logging.

...
#include <io/log.h>

#define NAME "myapp"
...

int main(int argc, char *argv[])
{
    log_init(NAME, LVL_NOTE);
    ...
    if (rc != EOK) {
        log_msg(LVL_ERROR, "Something failed (rc=%d), trying fallback...", rc);
        ...
    }
}

The first argument to log_init is application name, the second is the default logging level. Messages with higher level won't be printed at all.

Normally the (when an application is not being debugged) the logging level should be set to LVL_NOTE. That is, messages with level LVL_NOTE or higher are printed.

log_msg is a printf-like function where first argument is level of the message (its seriousness).

Note: See TracWiki for help on using the wiki.