man9: Add man page for kern_yield() and friends

The page is fairly simple and will be referenced by mi_switch(9).
Provide some usage notes so that the broader implications of how/when to
use these functions are understood.

Reviewed by:	kib, markj
MFC after:	1 week
Sponsored by:	The FreeBSD Foundation
Differential Revision:	https://reviews.freebsd.org/D38182
This commit is contained in:
Mitchell Horne 2023-02-09 11:35:39 -04:00
parent dc9b13736f
commit 30cd6fd75d
2 changed files with 137 additions and 0 deletions

View File

@ -196,6 +196,7 @@ MAN= accept_filter.9 \
KASSERT.9 \
kern_reboot.9 \
kern_testfrwk.9 \
kern_yield.9 \
kernacc.9 \
kernel_mount.9 \
khelp.9 \
@ -1332,6 +1333,8 @@ MLINKS+=intr_event.9 intr_event_add_handler.9 \
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 \

134
share/man/man9/kern_yield.9 Normal file
View File

@ -0,0 +1,134 @@
.\"
.\" Copyright (c) 2023 The FreeBSD Foundation
.\"
.\" This documentation was written by Mitchell Horne <mhorne@FreeBSD.org> 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.
Note that the requested priority will take effect until the thread returns to
usermode, after which its base user priority will be restored.
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
.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 are not guaranteed to preempt
each another.
Thus, threads executing in the kernel are expected to behave cooperatively
with respect to other threads in the system.
The yield 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 .