diff --git a/share/man/man9/Makefile b/share/man/man9/Makefile --- a/share/man/man9/Makefile +++ b/share/man/man9/Makefile @@ -196,6 +196,7 @@ KASSERT.9 \ kern_reboot.9 \ kern_testfrwk.9 \ + kern_yield.9 \ kernacc.9 \ kernel_mount.9 \ khelp.9 \ @@ -1332,6 +1333,8 @@ intr_event.9 intr_event_handle.9 \ intr_event.9 intr_event_remove_handler.9 \ intr_event.9 intr_priority.9 +MLINKS+=kern_yield.9 maybe_yield.9 \ + kern_yield.9 should_yield.9 MLINKS+=kernacc.9 useracc.9 MLINKS+=kernel_mount.9 free_mntarg.9 \ kernel_mount.9 mount_arg.9 \ diff --git a/share/man/man9/kern_yield.9 b/share/man/man9/kern_yield.9 new file mode 100644 --- /dev/null +++ b/share/man/man9/kern_yield.9 @@ -0,0 +1,132 @@ +.\" +.\" Copyright (c) 2023 The FreeBSD Foundation +.\" +.\" This documentation was written by Mitchell Horne under +.\" sponsorship from the FreeBSD Foundation. +.\" +.\" 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. +.\" +.\" 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. +.\" +.Dd January 30, 2023 +.Dt KERN_YIELD 9 +.Os +.Sh NAME +.Nm kern_yield , +.Nm maybe_yield , +.Nm should_yield +.Nd "functions for yielding execution of the current thread" +.Sh SYNOPSIS +.Ft void +.Fn kern_yield "int prio" +.Ft void +.Fn maybe_yield +.Ft bool +.Fn should_yield +.Sh DESCRIPTION +The +.Fn kern_yield +function causes the currently running thread to voluntarily, but +unconditionally, surrender its execution to the scheduler. +The +.Va prio +argument specifies the scheduling priority to be assigned before the context +switch, which has an influence on when execution will resume. +Valid values for +.Va prio +are any of the +.Dv PRI_* +values defined in +.In sys/priority.h , +as well as the following special values: +.Bl -tag -width "PRI_UNCHANGED" +.It Dv PRI_USER +Schedule the thread with its base user priority; the value corresponding to +.Xr setpriority 2 / +.Xr nice 3 . +.It Dv PRI_UNCHANGED +Yield the thread without changing its priority. +.El +Note that the requested priority will take effect until the thread returns to +usermode, after which its base user priority will be restored. +.Pp +The +.Fn should_yield +function checks if sufficient time has passed since the thread's last voluntary +context switch that yielding would be a useful service to other threads. +It returns +.Va true +when this is the case. +See +.Sx USAGE NOTES +for an elaboration of what this means. +.Pp +The +.Fn maybe_yield +function is a helper function for the common task of optionally yielding the +processor. +Internally, +.Fn kern_yield "PRI_USER" +will be called if +.Fn should_yield +returns +.Va true . +.Sh USAGE NOTES +Although the kernel supports preemption, this is usually reserved for +high-priority realtime or interrupt threads. +Kernel worker threads and timesharing threads don't readily preempt each other. +Thus, some cooperation is required to help prevent starvation. +These functions are mostly intended to be used by threads which perform a lot +of work inside the kernel. +For example: +.Fn maybe_yield +is called by the +.Dv vlnru +process each time it reclaims a vnode. +.Pp +The scheduler aims to identify threads which monopolize the CPU, and will +schedule them with decreased priority. +Threads which regularly yield the processor will be given the chance to run +more often. +The possibly surprising effect of this is that, depending on the disposition of +other threads on the CPU's runqueue, a call to +.Fn kern_yield +does not guarantee that the yielding thread will be taken off the CPU. +.Pp +With the above considerations in mind, it is advised that code written using +.Fn kern_yield +be measured to confirm that its use has a positive effect on relevant +performance or responsiveness metrics. +Switching thread contexts has a non-zero cost, and thus yielding the processor +too eagerly could have a negative impact on performance. +.Pp +Additionally, since yielding is a cooperative action, it is advised that the +yielding thread release any locks that it may be holding, when possible. +Otherwise, threads which have been given the chance to run could end up waiting +on the yielding thread to release the lock, largely defeating the purpose of +the yield. +.Sh SEE ALSO +.Xr setpriority 2 , +.Xr nice 3 , +.Xr mi_switch 9 +.Sh AUTHORS +.An -nosplit +This manual page was written by +.An Mitchell Horne Aq Mt mhorne@FreeBSD.org .