freebsd-skq/share/man/man9/LOCK_PROFILING.9

206 lines
6.9 KiB
Groff

.\"-
.\" Copyright (c) 2004 Dag-Erling Coïdan Smørgrav
.\" Copyright (c) 2005 Robert N. M. Watson
.\" Copyright (c) 2006 Kip Macy
.\" All rights reserved.
.\"
.\" Redistribution and use in source and binary forms, with or without
.\" modification, are permitted provided that the following conditions
.\" are met:
.\" 1. Redistributions of source code must retain the above copyright
.\" notice, this list of conditions and the following disclaimer.
.\" 2. Redistributions in binary form must reproduce the above copyright
.\" notice, this list of conditions and the following disclaimer in the
.\" documentation and/or other materials provided with the distribution.
.\" 3. The name of the author may not be used to endorse or promote products
.\" derived from this software without specific prior written permission.
.\"
.\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
.\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
.\" SUCH DAMAGE.
.\"
.\" $FreeBSD$
.\"
.Dd November 11, 2006
.Dt LOCK_PROFILING 9
.Os
.Sh NAME
.Nm LOCK_PROFILING
.Nd kernel lock profiling support
.Sh SYNOPSIS
.Cd "options LOCK_PROFILING"
.Sh DESCRIPTION
The
.Dv LOCK_PROFILING
kernel option adds support for measuring and reporting lock use and
contention statistics.
These statistics are collated by
.Dq acquisition point .
Acquisition points are
distinct places in the kernel source code (identified by source file
name and line number) where a lock is acquired.
.Pp
For each acquisition point, the following statistics are accumulated:
.Bl -bullet
.It
The longest time the lock was ever continuously held after being
acquired at this point.
.It
The total time the lock was held after being acquired at this point.
.It
The total time that threads have spent waiting to acquire the lock.
.It
The total number of non-recursive acquisitions.
.It
The total number of times the lock was already held by another thread
when this point was reached, requiring a spin or a sleep.
.It
The total number of times another thread tried to acquire the lock
while it was held after having been acquired at this point.
.El
.Pp
In addition, the average hold time and average wait time are derived
from the total hold time
and total wait time respectively and the number of acquisitions.
.Pp
The
.Dv LOCK_PROFILING
kernel option also adds the following
.Xr sysctl 8
variables to control and monitor the profiling code:
.Bl -tag -width indent
.It Va debug.lock.prof.enable
Enable or disable the lock profiling code.
This defaults to 0 (off).
.It Va debug.lock.prof.reset
Reset the current lock profiling buffers.
.It Va debug.lock.prof.acquisitions
The total number of lock acquisitions recorded.
.It Va debug.lock.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.
.It Va debug.lock.prof.maxrecords
The maximum number of acquisition points the profiling code is capable
of monitoring.
Since it would not be possible to call
.Xr malloc 9
from within the lock profiling code, this is a static limit.
The number of records can be changed with the
.Dv LPROF_BUFFERS
kernel option.
.It Va debug.lock.prof.rejected
The number of acquisition points that were ignored after the table
filled up.
.It Va debug.lock.prof.hashsize
The size of the hash table used to map acquisition points to
statistics records.
The hash size can be changed with the
.Dv LPROF_HASH_SIZE
kernel option.
.It Va debug.lock.prof.collisions
The number of hash collisions in the acquisition point hash table.
.It Va debug.lock.prof.stats
The actual profiling statistics in plain text.
The columns are as follows, from left to right:
.Bl -tag -width ".Va cnt_hold"
.It Va max
The longest continuous hold time in microseconds.
.It Va wait_max
The longest continuous wait time in microseconds.
.It Va total
The total (accumulated) hold time in microseconds.
.It Va wait_total
The total (accumulated) wait time in microseconds.
.It Va count
The total number of acquisitions.
.It Va avg
The average hold time in microseconds, derived from the total hold time
and the number of acquisitions.
.It Va wait_avg
The average wait time in microseconds, derived from the total wait time
and the number of acquisitions.
.It Va cnt_hold
The number of times the lock was held and another thread attempted to
acquire the lock.
.It Va cnt_lock
The number of times the lock was already held when this point was
reached.
.It Va name
The name of the acquisition point, derived from the source file name
and line number, followed by the name of the lock in parentheses.
.El
.El
.Sh SEE ALSO
.Xr sysctl 8 ,
.Xr mutex 9
.Sh HISTORY
Mutex profiling support appeared in
.Fx 5.0 .
Generalized lock profiling support appeared in
.Fx 7.0 .
.Sh AUTHORS
.An -nosplit
The
.Nm MUTEX_PROFILING
code was written by
.An Eivind Eklund Aq eivind@FreeBSD.org ,
.An Dag-Erling Sm\(/orgrav Aq des@FreeBSD.org
and
.An Robert Watson Aq rwatson@FreeBSD.org .
The
.Nm
code was written by
.An Kip Macy Aq kmacy@FreeBSD.org .
This manual page was written by
.An Dag-Erling Sm\(/orgrav Aq des@FreeBSD.org .
.Sh NOTES
The
.Dv LOCK_PROFILING
option increases the size of
.Vt "struct lock_object" ,
so a kernel built with that option will not work with modules built
without it.
.Pp
The
.Dv LOCK_PROFILING
option also prevents inlining of the mutex code, which can result in a
fairly severe performance penalty.
This is, however, not always the case.
.Dv LOCK_PROFILING
can introduce a substantial performance overhead that is easily
monitorable using other profiling tools, so combining profiling tools
with
.Dv LOCK_PROFILING
is not recommended.
.Pp
Measurements are made and stored in nanoseconds using
.Xr nanotime 9 ,
(on architectures without a synchronized TSC) 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).
.Pp
.Dv LOCK_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,
.Dv LOCK_PROFILING
will report higher than normal
.Xr uma 9
lock contention when run with
.Dv INVARIANTS
due to extra locking that occurs when
.Dv INVARIANTS
is present; likewise, using it in combination with
.Dv WITNESS
will lead to much higher lock hold times and contention in profiling output.