188 lines
6.3 KiB
Groff
188 lines
6.3 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 March 7, 2012
|
|
.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.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
|
|
.It Va debug.lock.prof.rejected
|
|
The number of acquisition points that were ignored after the table
|
|
filled up.
|
|
.It Va debug.lock.prof.skipspin
|
|
Disable or enable the lock profiling code for the spin locks.
|
|
This defaults to 0 (do profiling for the spin locks).
|
|
.It Va debug.lock.prof.skipcount
|
|
Do sampling approximately every N lock acquisitions.
|
|
.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.
|