Reference

MUTEX_PROFILING(9)

kernel mutex profiling support

NAME

MUTEX_PROFILING – kernel mutex profiling support

SYNOPSIS

options MUTEX_PROFILING

DESCRIPTION

The MUTEX_PROFILING kernel option adds support for measuring and reporting mutex use and contention statistics. These statistics are collated by “acquisition point”. Acquisition points are distinct places in the kernel source code (identified by source file name and line number) where a mutex is acquired.

For each acquisition point, the following statistics are accumulated:

  • The total number of non-recursive acquisitions.

  • The total time the mutex was held after being acquired at this point.

  • The longest time the mutex was ever continuously held after being acquired at this point.

  • The total number of times the mutex was already held by another thread when this point was reached, requiring a spin or a sleep.

  • The total number of time another thread tried to acquire the mutex while it was held after having been acquired at this point.

In addition, the average hold time is derived from the total hold time and the number of acquisitions.

The MUTEX_PROFILING kernel option also adds the following sysctl(8) variables to control and monitor the profiling code:

debug.mutex.prof.enable

Enable or disable the mutex profiling code. This defaults to 0 (off).

debug.mutex.prof.reset

Reset the current mutex profiling buffers.

debug.mutex.prof.acquisitions

The total number of mutex acquisitions recorded.

debug.mutex.prof.records

The total number of acquisition points recorded. Note that only active acquisition points (i.e., points that have been reached at least once) are counted.

debug.mutex.prof.maxrecords

The maximum number of acquisition points the profiling code is capable of monitoring. Since it would not be possible to call malloc(9) from within the mutex profiling code, this is a static limit. The number of records can be changed with the MPROF_BUFFERS kernel option.

debug.mutex.prof.rejected

The number of acquisition points that were ignored after the table filled up.

debug.mutex.prof.hashsize

The size of the hash table used to map acquisition points to statistics records. The hash size can be changed with the MPROF_HASH_SIZE kernel option.

debug.mutex.prof.collisions

The number of hash collisions in the acquisition point hash table.

debug.mutex.prof.stats

The actual profiling statistics in plain text. The columns are as follows, from left to right:

max

The longest continuous hold time in microseconds.

total

The total (accumulated) hold time in microseconds.

count

The total number of acquisitions.

avg

The average hold time in microseconds, derived from the total hold time and the number of acquisitions.

cnt_hold

The number of times the mutex was held and another thread attempted to lock the mutex.

cnt_lock

The number of times the mutex was already locked when this point was reached.

name

The name of the acquisition point, derived from the source file name and line number, followed by the name of the mutex in parentheses.

SEE ALSO

sysctl(8), mutex(9)

HISTORY

Mutex profiling support appeared in FreeBSD 5.0.

AUTHORS

The MUTEX_PROFILING code was written by Eivind Eklund ⟨[email protected]⟩, Dag-Erling Smørgrav ⟨[email protected]⟩ and Robert Watson ⟨[email protected]⟩. This manual page was written by Dag-Erling Smørgrav ⟨[email protected]⟩.

NOTES

The MUTEX_PROFILING option increases the size of struct mtx, so a kernel built with that option will not work with modules built without it.

The MUTEX_PROFILING option also prevents inlining of the mutex code, which results in a fairly severe performance penalty. It should therefore only be enabled on systems where mutex profiling is actually needed. MUTEX_PROFILING will introduce a substantial performance overhead that is easily monitorable using other profiling tools, so combining profiling tools with MUTEX_PROFILING is not recommended.

Measurements are made and stored in nanoseconds using nanotime(9), but are presented in microseconds. This should still be sufficient for the locks one would be most interested in profiling (those that are held long and/or acquired often).

MUTEX_PROFILING should generally not be used in combination with other debugging options, as the results may be strongly affected by interactions between the features. In particular, MUTEX_PROFILING will report higher than normal uma(9) lock contention when run with INVARIANTS due to extra locking that occurs when INVARIANTS is present; likewise, using it in combination with WITNESS will lead to much higher lock hold times and contention in profiling output.