1.\"- 2.\" Copyright (c) 2013 Gleb Smirnoff <glebius@FreeBSD.org> 3.\" All rights reserved. 4.\" 5.\" Redistribution and use in source and binary forms, with or without 6.\" modification, are permitted provided that the following conditions 7.\" are met: 8.\" 1. Redistributions of source code must retain the above copyright 9.\" notice, this list of conditions and the following disclaimer. 10.\" 2. Redistributions in binary form must reproduce the above copyright 11.\" notice, this list of conditions and the following disclaimer in the 12.\" documentation and/or other materials provided with the distribution. 13.\" 14.\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 15.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 16.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 17.\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 18.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 19.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 20.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 21.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 22.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 23.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 24.\" SUCH DAMAGE. 25.\" 26.\" $FreeBSD$ 27.\" 28.Dd February 7, 2014 29.Dt COUNTER 9 30.Os 31.Sh NAME 32.Nm counter 33.Nd "SMP-friendly kernel counter implementation" 34.Sh SYNOPSIS 35.In sys/types.h 36.In sys/systm.h 37.In sys/counter.h 38.Ft counter_u64_t 39.Fn counter_u64_alloc "int wait" 40.Ft void 41.Fn counter_u64_free "counter_u64_t c" 42.Ft void 43.Fn counter_u64_add "counter_u64_t c" "int64_t v" 44.Ft void 45.Fn counter_enter 46.Ft void 47.Fn counter_exit 48.Ft void 49.Fn counter_u64_add_protected "counter_u64_t c" "int64_t v" 50.Ft uint64_t 51.Fn counter_u64_fetch "counter_u64_t c" 52.Ft void 53.Fn counter_u64_zero "counter_u64_t c" 54.In sys/sysctl.h 55.Fn SYSCTL_COUNTER_U64 parent nbr name access ptr descr 56.Fn SYSCTL_ADD_COUNTER_U64 ctx parent nbr name access ptr descr 57.Sh DESCRIPTION 58.Nm 59is a generic facility to create counters 60that can be utilized for any purpose (such as collecting statistical 61data). 62A 63.Nm 64is guaranteed to be lossless when several kernel threads do simultaneous 65updates. 66However, 67.Nm 68does not block the calling thread, 69also no 70.Xr atomic 9 71operations are used for the update, therefore the counters 72can be used in any non-interrupt context. 73Moreover, 74.Nm 75has special optimisations for SMP environments, making 76.Nm 77update faster than simple arithmetic on the global variable. 78Thus 79.Nm 80is considered suitable for accounting in the performance-critical 81code pathes. 82.Bl -tag -width indent 83.It Fn counter_u64_alloc how 84Allocate a new 64-bit unsigned counter. 85The 86.Fa wait 87argument is the 88.Xr malloc 9 89wait flag, should be either 90.Va M_NOWAIT 91or 92.Va M_WAITOK . 93If 94.Va M_NOWAIT 95is specified the operation may fail. 96.It Fn counter_u64_free c 97Free the previously allocated counter 98.Fa c . 99.It Fn counter_u64_add c v 100Add 101.Fa v 102to 103.Fa c . 104The KPI does not guarantee any protection from wraparound. 105.It Fn counter_enter 106Enter mode that would allow to safely update several counters via 107.Fn counter_u64_add_protected . 108On some machines this expands to 109.Xr critical 9 110section, while on other is a nop. 111See 112.Sx IMPLEMENTATION DETAILS . 113.It Fn counter_exit 114Exit mode for updating several counters. 115.It Fn counter_u64_add_protected c v 116Same as 117.Fn counter_u64_add , 118but should be preceded by 119.Fn counter_enter . 120.It Fn counter_u64_fetch c 121Take a snapshot of counter 122.Fa c . 123The data obtained is not guaranteed to reflect the real cumulative 124value for any moment. 125.It Fn counter_u64_zero c 126Clear the counter 127.Fa c 128and set it to zero. 129.It Fn SYSCTL_COUNTER_U64 parent nbr name access ptr descr 130Declare a static 131.Xr sysctl 132oid that would represent a 133.Nm . 134The 135.Fa ptr 136argument should be a pointer to allocated 137.Vt counter_u64_t . 138A read of the oid returns value obtained through 139.Fn counter_u64_fetch . 140Any write to the oid zeroes it. 141.It Fn SYSCTL_ADD_COUNTER_U64 ctx parent nbr name access ptr descr 142Create a 143.Xr sysctl 144oid that would represent a 145.Nm . 146The 147.Fa ptr 148argument should be a pointer to allocated 149.Vt counter_u64_t . 150A read of the oid returns value obtained through 151.Fn counter_u64_fetch . 152Any write to the oid zeroes it. 153.El 154.Sh IMPLEMENTATION DETAILS 155On all architectures 156.Nm 157is implemented using per-CPU data fields that are specially aligned 158in memory, to avoid inter-CPU bus traffic due to shared use 159of the variables between CPUs. 160These are allocated using 161.Va UMA_ZONE_PCPU 162.Xr uma 9 163zone. 164The update operation only touches the field that is private to current CPU. 165Fetch operation loops through all per-CPU fields and obtains a snapshot 166sum of all fields. 167.Pp 168On amd64 a 169.Nm counter 170update is implemented as a single instruction without lock semantics, 171operating on the private data for the current CPU, 172which is safe against preemption and interrupts. 173.Pp 174On i386 architecture, when machine supports the cmpxchg8 instruction, 175this instruction is used. 176The multi-instruction sequence provides the same guarantees as the 177amd64 single-instruction implementation. 178.Pp 179On some architectures updating a counter require a 180.Xr critical 9 181section. 182.Sh SEE ALSO 183.Xr atomic 9 , 184.Xr critical 9 , 185.Xr locking 9 , 186.Xr malloc 9 , 187.Xr sysctl 9 , 188.Xr uma 9 189.Sh HISTORY 190The 191.Nm 192facility first appeared in 193.Fx 10.0 . 194.Sh AUTHORS 195.An -nosplit 196The 197.Nm 198facility was written by 199.An Gleb Smirnoff 200and 201.An Konstantin Belousov . 202