Date: Mon, 29 Jul 2019 21:53:02 +0000 (UTC) From: Mariusz Zaborski <oshogbo@FreeBSD.org> To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: svn commit: r350430 - in head: share/man/man9 sys/sys Message-ID: <201907292153.x6TLr2s8044106@repo.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: oshogbo Date: Mon Jul 29 21:53:02 2019 New Revision: 350430 URL: https://svnweb.freebsd.org/changeset/base/350430 Log: seqc: add man page Reviewed by: markj Earlier version reviewed by: emaste, mjg, bcr, 0mp Differential Revision: https://reviews.freebsd.org/D16744 Added: head/share/man/man9/seqc.9 (contents, props changed) Modified: head/share/man/man9/Makefile head/sys/sys/seqc.h Modified: head/share/man/man9/Makefile ============================================================================== --- head/share/man/man9/Makefile Mon Jul 29 21:42:57 2019 (r350429) +++ head/share/man/man9/Makefile Mon Jul 29 21:53:02 2019 (r350430) @@ -289,6 +289,7 @@ MAN= accept_filter.9 \ securelevel_gt.9 \ selrecord.9 \ sema.9 \ + seqc.9 \ sf_buf.9 \ sglist.9 \ shm_map.9 \ @@ -1825,6 +1826,10 @@ MLINKS+=sema.9 sema_destroy.9 \ sema.9 sema_trywait.9 \ sema.9 sema_value.9 \ sema.9 sema_wait.9 +MLINKS+=seqc.9 seqc_consistent.9 \ + seqc.9 seqc_read.9 \ + seqc.9 seqc_write_begin.9 \ + seqc.9 seqc_write_end.9 MLINKS+=sf_buf.9 sf_buf_alloc.9 \ sf_buf.9 sf_buf_free.9 \ sf_buf.9 sf_buf_kva.9 \ Added: head/share/man/man9/seqc.9 ============================================================================== --- /dev/null 00:00:00 1970 (empty, because file is newly added) +++ head/share/man/man9/seqc.9 Mon Jul 29 21:53:02 2019 (r350430) @@ -0,0 +1,138 @@ +.\" +.\" Copyright (C) 2019 Mariusz Zaborski <oshogbo@FreeBSD.org> +.\" +.\" 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(s), this list of conditions and the following disclaimer as +.\" the first lines of this file unmodified other than the possible +.\" addition of one or more copyright notices. +.\" 2. Redistributions in binary form must reproduce the above copyright +.\" notice(s), 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 COPYRIGHT HOLDER(S) ``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 COPYRIGHT HOLDER(S) 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 July 29, 2019 +.Dt SEQC 9 +.Os +.Sh NAME +.Nm seqc_consistent , +.Nm seqc_read , +.Nm seqc_write_begin , +.Nm seqc_write_end +.Nd "lockless read algorithm" +.Sh SYNOPSIS +.In sys/seqc.h +.Ft void +.Fn seqc_write_begin "seqc_t *seqcp" +.Ft void +.Fn seqc_write_end "seqc_t *seqcp" +.Ft seqc_t +.Fn seqc_read "seqc_t *seqcp" +.Ft seqc_t +.Fn seqc_consistent "const seqc_t *seqcp" "seqc_t oldseqc" +.Sh DESCRIPTION +The +.Nm seqc +allows zero or more readers and zero or one writer to concurrently access +an object, providing a consistent snapshot of the object for readers. +No mutual exclusion between readers and writers is required, +but readers may be starved indefinitely by writers. +.Pp +The functions +.Fn seqc_write_begin +and +.Fn seqc_write_end +are used to create a transaction for writer, and notify the readers that the +object will be modified. +.Pp +The +.Fn seqc_read +function returns the current sequence number. +If a writer has started a transaction, this function will spin until the +transaction has ended. +.Pp +The +.Fn seqc_consistent +function compares the sequence number with a previously fetched value. +The +.Fa oldseqc +variable should contain a sequence number from the beginning of read +transaction. +.Pp +The reader at the end of a transaction checks if the sequence number has +changed. +If the sequence number didn't change the object wasn't modified, and fetched +variables are valid. +If the sequence number changed the object was modified and the fetch should be +repeated. +In case when sequence number is odd the object change is in progress and the +reader will wait until the write will the sequence number will become even. +.Sh EXAMPLES +The following example for a writer changees the +.Va var1 +and +.Va var2 +variables in the +.Va obj +structure: +.Bd -literal +lock_exclusive(&obj->lock); +seqc_write_begin(&obj->seqc); +obj->var1 = 1; +obj->var2 = 2; +seqc_write_end(&obj->seqc); +unlock_exclusive(&obj->lock); +.Ed +The following example for a reader reads the +.Va var1 +and +.Va var2 +variables from the +.Va obj +structure. +In the case where the sequence number was changed it restarts the whole process. +.Bd -literal +int var1, var2; +seqc_t seqc; + +for (;;) { + seqc = seqc_read(&obj->seqc); + var1 = obj->var1; + var2 = obj->var2; + if (seqc_consistent(&obj->seqc, seqc)) + break; +} +.Ed +.Sh AUTHORS +The +.Nm seqc +functions was implemented by +.An Mateusz Guzik Aq Mt mjg@FreeBSD.org . +This manual page was written by +.An Mariusz Zaborski Aq Mt oshogbo@FreeBSD.org . +.Sh CAVEATS +There is no guarantee of progress for readers. +In case when there are a lot of writers the reader can be starved. +This concern may be solved by returning error after a few attempts. +.Pp +Theoretically if reading takes a very long time, and when there are many writers +the counter may overflow and wrap around to the same value. +In that case the reader will not notice that the object was changed. +Given that this needs 4 billion transactional writes across a single contended +reader, it is unlikely to ever happen. +This could be avoided by extending the interface to allow 64-bit counters. Modified: head/sys/sys/seqc.h ============================================================================== --- head/sys/sys/seqc.h Mon Jul 29 21:42:57 2019 (r350429) +++ head/sys/sys/seqc.h Mon Jul 29 21:53:02 2019 (r350430) @@ -40,55 +40,6 @@ typedef uint32_t seqc_t; #ifdef _KERNEL -/* - * seqc allows readers and writers to work with a consistent snapshot. Modifying - * operations must be enclosed within a transaction delineated by - * seqc_write_beg/seqc_write_end. The trick works by having the writer increment - * the sequence number twice, at the beginning and end of the transaction. - * The reader detects that the sequence number has not changed between its start - * and end, and that the sequence number is even, to validate consistency. - * - * Some fencing (both hard fencing and compiler barriers) may be needed, - * depending on the cpu. Modern AMD cpus provide strong enough guarantees to not - * require any fencing by the reader or writer. - * - * Example usage: - * - * writers: - * lock_exclusive(&obj->lock); - * seqc_write_begin(&obj->seqc); - * obj->var1 = ...; - * obj->var2 = ...; - * seqc_write_end(&obj->seqc); - * unlock_exclusive(&obj->lock); - * - * readers: - * int var1, var2; - * seqc_t seqc; - * - * for (;;) { - * seqc = seqc_read(&obj->seqc); - * var1 = obj->var1; - * var2 = obj->var2; - * if (seqc_consistent(&obj->seqc, seqc)) - * break; - * } - * ..... - * - * Writers may not block or sleep in any way. - * - * There are 2 minor caveats in this implementation: - * - * 1. There is no guarantee of progress. That is, a large number of writers can - * interfere with the execution of the readers and cause the code to live-lock - * in a loop trying to acquire a consistent snapshot. - * - * 2. If the reader loops long enough, the counter may overflow and eventually - * wrap back to its initial value, fooling the reader into accepting the - * snapshot. Given that this needs 4 billion transactional writes across a - * single contended reader, it is unlikely to ever happen. - */ - /* A hack to get MPASS macro */ #include <sys/lock.h>
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201907292153.x6TLr2s8044106>