Autofs Stuff

Filing bug reports

• autofs rpm version, obtained via 'rpm -q autofs'
• kernel version, obtained via 'uname -r'
• contents of your autofs maps. This includes auto.master and at least the map which is problematic for you.
• contents of /etc/nsswitch.conf
• contents of /etc/sysconfig/autofs
• Steps to reproduce the problem, or circumstances surrounding the problem.
• debug output.
  • Add a line like the following to your /etc/syslog.conf:
    

    daemon.*			/var/log/debug

    In newer distributions (like Fedora 9), the file to edit is /etc/rsyslog.conf, but the format remains the same.
  • Tell autofs to log debugging information for the problematic map. This can be done by adding a --debug to the entry for this map in /etc/auto.master. For example:

    /misc		/etc/auto.misc --timeout=60 --debug

    If, instead, you need to log debug output for all maps, you can do the following:
    For v4, add the --debug option to the DAEMONOPTIONS variable specified in /etc/sysconfig/autofs.
    For v5, the /etc/sysconfig/autofs file format may use either LOGGING or DEFAULT_LOGGING. Older versions require DEFAULT_LOGGING, while newer versions will accept either option. You will want to set whichever one you have to debug, like so:

    DEFAULT_LOGGING="debug"

    For autofs version 4.1.3-231.jmoyer.1 and later, you can toggle debug logging while the daemon is running using the following command line:

    /usr/sbin/automount --set-log-priority N <path>

    where 'N' can either be a number from 0-7: 0 logging the least information and 7 logging the most, or the textual debug level, like "debug". See the man page for more options, but really you should just set the log level to 7 to gather debug information. The path specified determines for which automount-managed mount point logging will be changed. In this respect, it functions very similarly to adding the --debug option to an entry in the master map. Note that this change is temporary. The next time the daemon is restarted it will switch back to log level as defined in the static configuration.
  • Restart syslogd or rsyslogd, depending on which is shipped with your operating system (or send it a HUP signal).
  • If you modified the automount configuration to enable debugging (instead of using the run-time switch), then restart autofs. Make sure it restarted! Check the logs, and ensure that the problematic mount point was shut down properly and subsequently restarted. Otherwise, you will not get any debug output in the logs.
  • Now you're set to get debug output. Reproduce the problem, and attach the logs with your bug report.
• If it's not a readily reproducible problem, then please gather some system state when it does occur. The following are useful:
  • Current automount status:
    

    # /etc/init.d/autofs status
    # ps auxwww | grep automount

  • Output from sysrq-t:
    

    # echo t > /proc/sysrq-trigger

    You will find the output in /var/log/messages.

AIO/DIO Stuff

• git access to the libaio sources

$ git clone git://git.kernel.org/pub/scm/libs/libaio/libaio.git

• Test "Harness"
Zach has a git repository setup for AIO/DIO regression tests here: http://git.kernel.org/?p=linux/kernel/git/zab/aio-dio-regress.git;a=summary
• My aio/dio test kernels, built for the various versions of RHEL
• A version of pipetest that exercises the IOCB_CMD_EPOLL_WAIT command can be found here.

• Deploying Autofs

• Hacking the Linux Automounter: Current Limitations and Future Directions
• Programming with the Netpoll API

• Capturing System Crash Information for Postmortem Analysis

• Kernel Debugging with Netdump and Crash