clockintr_bind.9: document clockintr, clockrequest APIs

This manpage only covers the frontend APIs.  The backend API is still
moving around and will be documented separately.

Tons of input, editing, and suggestions from jmc@, schwarze@, mpi@,
claudio@, mlarkin@, and probably several others I'm forgetting.

ok jmc@ schwarze@ claudio@ mpi@

diff --git a/share/man/man9/Makefile b/share/man/man9/Makefile
index c990a8a..4bf92b2 100644 (file)
--- a/share/man/man9/Makefile
+++ b/share/man/man9/Makefile
@@ -1,4 +1,4 @@
-#      $OpenBSD: Makefile,v 1.309 2023/12/21 02:57:14 jsg Exp $
+#      $OpenBSD: Makefile,v 1.310 2024/02/24 16:21:32 cheloha Exp $
 #      $NetBSD: Makefile,v 1.4 1996/01/09 03:23:01 thorpej Exp $
 
 #      Makefile for section 9 (kernel function and variable) manual pages.
@@ -9,7 +9,7 @@ MAN=    aml_evalnode.9 atomic_add_int.9 atomic_cas_uint.9 \
        audio.9 autoconf.9 \
        bemtoh32.9 bio_register.9 bintimeadd.9 boot.9 bpf_mtap.9 buffercache.9 \
        bufq_init.9 bus_dma.9 bus_space.9 \
-       clockintr.9 \
+       clockintr.9 clockintr_bind.9 \
        copy.9 cond_init.9 config_attach.9 config_defer.9 counters_alloc.9 \
        cpumem_get.9 crypto.9 \
        delay.9 disk.9 disklabel.9 dma_alloc.9 dohooks.9 \

diff --git a/share/man/man9/clockintr_bind.9 b/share/man/man9/clockintr_bind.9
new file mode 100644 (file)
index 0000000..38b89fb
--- /dev/null
+++ b/share/man/man9/clockintr_bind.9
@@ -0,0 +1,296 @@
+.\"    $OpenBSD: clockintr_bind.9,v 1.1 2024/02/24 16:21:32 cheloha Exp $
+.\"
+.\" Copyright (c) 2023-2024 Scott Cheloha <cheloha@openbsd.org>
+.\"
+.\" Permission to use, copy, modify, and distribute this software for any
+.\" purpose with or without fee is hereby granted, provided that the above
+.\" copyright notice and this permission notice appear in all copies.
+.\"
+.\" THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+.\" WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+.\" MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+.\" ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+.\" WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+.\" ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+.\" OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+.\"
+.Dd $Mdocdate: February 24 2024 $
+.Dt CLOCKINTR_BIND 9
+.Os
+.Sh NAME
+.Nm clockintr_bind ,
+.Nm clockintr_schedule ,
+.Nm clockintr_advance ,
+.Nm clockintr_cancel ,
+.Nm clockintr_unbind ,
+.Nm clockintr_stagger ,
+.Nm clockrequest_advance
+.Nd execute a function in a clock interrupt context
+.Sh SYNOPSIS
+.In sys/clockintr.h
+.Ft void
+.Fo clockintr_bind
+.Fa "struct clockintr *cl"
+.Fa "struct cpu_info *cpu"
+.Fa "void (*callback)(struct clockrequest *cr, void *cf, void *arg)"
+.Fa "void *arg"
+.Fc
+.Ft void
+.Fo clockintr_schedule
+.Fa "struct clockintr *cl"
+.Fa "uint64_t abs"
+.Fc
+.Ft uint64_t
+.Fo clockintr_advance
+.Fa "struct clockintr *cl"
+.Fa "uint64_t interval"
+.Fc
+.Ft void
+.Fo clockintr_cancel
+.Fa "struct clockintr *cl"
+.Fc
+.Ft void
+.Fo clockintr_unbind
+.Fa "struct clockintr *cl"
+.Fa "uint32_t flags"
+.Fc
+.Ft void
+.Fo clockintr_stagger
+.Fa "struct clockintr *cl"
+.Fa "uint64_t interval"
+.Fa "uint32_t numer"
+.Fa "uint32_t denom"
+.Fc
+.Ft uint64_t
+.Fo clockrequest_advance
+.Fa "struct clockrequest *cr"
+.Fa "uint64_t interval"
+.Fc
+.\" .Fn clockrequest_advance_random is intentionally undocumented.
+.\"  It may be removed in the future.  New code should not use it.
+.Sh DESCRIPTION
+The clock interrupt subsystem schedules functions for asynchronous execution
+from the clock interrupt context on a particular CPU.
+.Pp
+Clock interrupts are well-suited for timekeeping,
+scheduling,
+and statistical profiling.
+Applications with more relaxed latency requirements should use timeouts
+to schedule asynchronous execution;
+see
+.Xr timeout_add 9
+for details.
+.Pp
+The
+.Fn clockintr_bind
+function initializes the clock interrupt object
+.Fa cl .
+When
+.Fa cl
+expires,
+its
+.Fa callback
+function is executed from the
+.Dv IPL_CLOCK
+context on its host
+.Fa cpu
+without any locks or mutexes.
+The callback function must not block.
+Its parameters are as follows:
+.Bl -tag -width indent
+.It Fa cr
+A private
+.Vt clockrequest
+object.
+May be used to request rescheduling;
+see
+.Fn clockrequest_advance
+below.
+.It Fa cf
+The
+.Fa cpu Ns 's
+current machine-dependent clockframe.
+.It Fa arg
+The
+.Fa arg
+given to
+.Fn clockintr_bind .
+.El
+.Pp
+The memory pointed to by
+.Fa cl
+must be zeroed before it is first bound.
+It is an error to use
+.Fa cl
+as argument to any other function in the
+.Vt clockintr
+API before it is bound.
+It is an error to rebind
+.Fa cl
+without first unbinding it;
+see
+.Fn clockintr_unbind
+below.
+.Pp
+The
+.Fn clockintr_schedule
+function schedules
+.Fa cl
+to expire at the absolute time
+.Fa abs
+on the system uptime clock.
+The subsystem will never execute
+.Fa cl Ns 's
+callback function before this expiration time,
+though its execution may be delayed by other activity on the system.
+.Pp
+The
+.Fn clockintr_advance
+function schedules
+.Fa cl
+to expire at the next terminus of the given
+.Fa interval ,
+a non-zero count of nanoseconds,
+relative to
+.Fa cl Ns 's
+current expiration time.
+Periodic clock interrupts should be scheduled with
+.Fn clockintr_advance
+to keep the execution period from drifting.
+.Pp
+The
+.Fn clockintr_cancel
+function cancels any pending expiration of
+.Fa cl .
+.Pp
+The
+.Fn clockintr_unbind
+function cancels any pending expiration of
+.Fa cl
+and severs the binding between
+.Fa cl
+and its host
+.Fa cpu .
+Upon return,
+.Fa cl
+may be rebound with
+.Fn clockintr_bind .
+The call may be configured with zero or more of the following
+.Fa flags :
+.Bl -tag -width CL_BARRIER
+.It Dv CL_BARRIER
+If
+.Fa cl Ns 's
+callback function is executing,
+block until it returns.
+By default,
+the caller does not block.
+Useful when
+.Fa arg
+is a shared reference.
+.El
+.Pp
+The
+.Fn clockintr_stagger
+function resets
+.Fa cl Ns 's
+expiration time to a fraction of the given
+.Fa interval ,
+a count of nanoseconds.
+Specifically,
+.Fa cl Ns 's
+expiration time is reset to
+.Pq Fa interval Ms / Fa denom Ms * Fa numer .
+Periodic clock interrupts bound to multiple CPUs may be staggered
+to reduce the likelihood that their callback functions will execute
+simultaneously and compete for a shared resource.
+It is an error if
+.Fa numer
+is greater than or equal to
+.Fa denom .
+It is an error if
+.Fa cl
+is already scheduled to expire.
+.Pp
+The
+.Fn clockrequest_advance
+function is similar to
+.Fn clockintr_advance ,
+except that
+(a) it may only be called during the execution of a
+.Fa callback
+function,
+(b) it accepts a
+.Vt clockrequest
+pointer as argument,
+and (c) scheduling requests submitted with the interface are not fulfilled
+until the callback function returns.
+When the callback function returns,
+scheduling requests are only committed to the underlying clock interrupt
+object if that object was not manipulated during the execution of the
+callback function.
+Otherwise,
+outstanding requests are discarded.
+.Sh CONTEXT
+The
+.Fn clockintr_bind
+function may only be called from process context.
+.Pp
+The
+.Fn clockintr_advance ,
+.Fn clockintr_cancel ,
+.Fn clockintr_schedule ,
+and
+.Fn clockintr_stagger
+functions may be called from process context or from interrupt context.
+.Pp
+The
+.Fn clockintr_unbind
+function may normally be called from process context or from interrupt context.
+However,
+if the
+.Dv CL_BARRIER
+flag is set in
+.Fa flags ,
+the function may only be called from process context.
+.Pp
+The
+.Fn clockrequest_advance
+function may only be called during execution of a
+.Fa callback
+function.
+.Sh RETURN VALUES
+The
+.Fn clockintr_advance
+and
+.Fn clockrequest_advance
+functions return the number of
+.Fa interval Ns s
+that have elapsed since
+.Fa cl
+was scheduled to expire,
+or zero if
+.Fa cl
+has not yet expired.
+.Sh CODE REFERENCES
+.Pa sys/kern/kern_clockintr.c
+.Sh SEE ALSO
+.Xr microtime 9 ,
+.Xr spl 9 ,
+.Xr timeout 9
+.Rs
+.%A Richard McDougall
+.%A Jim Mauro
+.%B Solaris Internals: Solaris 10 and OpenSolaris Kernel Architecture
+.%I Prentice Hall
+.%I Sun Microsystems Press
+.%D 2nd Edition, 2007
+.%P pp. 912\(en925
+.Re
+.Sh HISTORY
+The
+.Vt clockintr
+and
+.Vt clockrequest
+APIs first appeared in
+.Ox 7.5 .