diff --git a/share/man/man9/LOCK_PROFILING.9 b/share/man/man9/LOCK_PROFILING.9 new file mode 100644 index 000000000000..c8a3c419652a --- /dev/null +++ b/share/man/man9/LOCK_PROFILING.9 @@ -0,0 +1,201 @@ +.\"- +.\" 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 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 5.0 . +.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 LOCK_PROFILING +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.