man/man9/style.9

.\"-
.\" Copyright (c) 1995-2019 The FreeBSD Project
.\"
.\" 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 [your name] 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.
.\"
.\"	From: @(#)style	1.14 (Berkeley) 4/28/95
.\"
.Dd April 26, 2024
.Dt STYLE 9
.Os
.Sh NAME
.Nm style
.Nd "kernel source file style guide"
.Sh DESCRIPTION
This file specifies the preferred style for kernel source files in the
.Fx
source tree.
It is also a guide for the preferred userland code style.
Many of the style rules are implicit in the examples.
Be careful to check the examples before assuming that
.Nm
is silent on an issue.
.Bd -literal
/*
 * Style guide for FreeBSD.  Based on the CSRG's KNF (Kernel Normal Form).
 *
 *	@(#)style	1.14 (Berkeley) 4/28/95
 */

/*
 * VERY important single-line comments look like this.
 */

/* Most single-line comments look like this. */

/*
 * Multi-line comments look like this.  Make them real sentences.  Fill
 * them so they look like real paragraphs.
 */
.Ed
.Pp
The copyright header should be a multi-line comment, with the first
line of the comment having a dash after the star like so:
.Bd -literal
/*-
 * SPDX-License-Identifier: BSD-2-Clause
 *
 * Copyright (c) 1984-2025 John Q. Public
 *
 * Long, boring license goes here, but trimmed for brevity
 */
.Ed
.Pp
An automatic script collects license information from the tree for
all comments that start in the first column with
.Dq Li "/*-" .
If you desire to flag
.Xr indent 1
to not reformat a comment that starts in the first column which is not a
license or copyright notice, change the dash to a star for those
comments.
Comments starting in columns other than the first are never
considered license statements.
Use the appropriate SPDX-License-Identifier line before the copyright.
If the copyright assertion contains the phrase
.Dq Li "All Rights Reserved"
that should be on the same line as the word
.Dq Li "Copyright" .
You should not insert a new copyright line between an old
copyright line and this phrase.
Instead, you should insert a new copyright phrase after
a pre-existing
.Dq Li "All Rights Reserved"
line.
When making changes, it is acceptable to fold an
.Dq Li "All Rights Reserved"
line with each of the
.Dq Li "Copyright"
lines.
For files that have the
.Dq Li "All Rights Reserved"
line on the same line(s) as the word
.Dq Li "Copyright" ,
new copyright assertions should be added last.
New
.Dq Li "Copyright"
lines should only be added when making substantial changes to the file,
not for trivial changes.
.Pp
After any copyright and license comment, there is a blank line, and the
.Li $\&FreeBSD$
for non C/C++ language source files.
Version control system ID tags should only exist once in a file
(unlike in this one).
Non-C/C++ source files follow the example above, while C/C++ source files
follow the one below.
All VCS (version control system) revision identification in files obtained
from elsewhere should be maintained, including, where applicable, multiple IDs
showing a file's history.
In general, do not edit foreign IDs or their infrastructure.
Unless otherwise wrapped (such as
.Dq Li "#if defined(LIBC_SCCS)" ) ,
enclose both in
.Dq Li "#if 0 ... #endif"
to hide any uncompilable bits
and to keep the IDs out of object files.
Only add
.Dq Li "From: "
in front of foreign VCS IDs if the file is renamed.
.Bd -literal
/* From: @(#)style	1.14 (Berkeley) 4/28/95 */

#include <sys/cdefs.h>
.Ed
.Pp
Leave one blank line before the header files.
.Pp
Kernel include files
.Pa ( sys/*.h )
come first.
If
.In sys/cdefs.h
is needed for
.Fn __FBSDID ,
include it first.
If either
.In sys/types.h
or
.In sys/param.h
is needed, include it before other include files.
.Po
.In sys/param.h
includes
.In sys/types.h ;
do not include both.
.Pc
Next, include
.In sys/systm.h ,
if needed.
The remaining kernel headers should be sorted alphabetically.
.Bd -literal
#include <sys/types.h>	/* Non-local includes in angle brackets. */
#include <sys/systm.h>
#include <sys/endian.h>
#include <sys/lock.h>
#include <sys/queue.h>
.Ed
.Pp
For a network program, put the network include files next.
.Bd -literal
#include <net/if.h>
#include <net/if_dl.h>
#include <net/route.h>
#include <netinet/in.h>
#include <protocols/rwhod.h>
.Ed
.Pp
Do not include files from
.Pa /usr/include
in the kernel.
.Pp
Leave a blank line before the next group, the
.Pa /usr/include
files,
which should be sorted alphabetically by name.
.Bd -literal
#include <stdio.h>
.Ed
.Pp
Global pathnames are defined in
.In paths.h .
Pathnames local
to the program go in
.Qq Pa pathnames.h
in the local directory.
.Bd -literal
#include <paths.h>
.Ed
.Pp
Leave another blank line before the local include files.
.Bd -literal
#include "pathnames.h"		/* Local includes in double quotes. */
.Ed
.Pp
Do not
.Ic #define
or declare names in the implementation namespace except
for implementing application interfaces.
.Pp
The names of
.Dq unsafe
macros (ones that have side effects), and the names of macros for
manifest constants, are all in uppercase.
The expansions of expression-like macros are either a single token
or have outer parentheses.
Put a single tab character between the
.Ic #define
and the macro name.
If a macro is an inline expansion of a function, the function name is
all in lowercase and the macro has the same name all in uppercase.
.\" XXX the above conflicts with ANSI style where the names are the
.\" same and you #undef the macro (if any) to get the function.
.\" It is not followed for MALLOC(), and not very common if inline
.\" functions are used.
Right-justify the
backslashes; it makes it easier to read.
If the macro encapsulates a compound statement, enclose it in a
.Ic do
loop,
so that it can safely be used in
.Ic if
statements.
Any final statement-terminating semicolon should be
supplied by the macro invocation rather than the macro, to make parsing easier
for pretty-printers and editors.
.Bd -literal
#define	MACRO(x, y) do {						\e
	variable = (x) + (y);						\e
	(y) += 2;							\e
} while (0)
.Ed
.Pp
When code is conditionally compiled using
.Ic #ifdef
or
.Ic #if ,
a comment may be added following the matching
.Ic #endif
or
.Ic #else
to permit the reader to easily discern where conditionally compiled code
regions end.
This comment should be used only for (subjectively) long regions, regions
greater than 20 lines, or where a series of nested
.Ic #ifdef 's
may be confusing to the reader.
The comment should be separated from the
.Ic #endif
or
.Ic #else
by a single space.
For short conditionally compiled regions, a closing comment should not be
used.
.Pp
The comment for
.Ic #endif
should match the expression used in the corresponding
.Ic #if
or
.Ic #ifdef .
The comment for
.Ic #else
and
.Ic #elif
should match the inverse of the expression(s) used in the preceding
.Ic #if
and/or
.Ic #elif
statements.
In the comments, the subexpression
.Dq Li defined(FOO)
is abbreviated as
.Dq Li FOO .
For the purposes of comments,
.Dq Ic #ifndef Li FOO
is treated as
.Dq Ic #if Li !defined(FOO) .
.Bd -literal
#ifdef KTRACE
#include <sys/ktrace.h>
#endif

#ifdef COMPAT_43
/* A large region here, or other conditional code. */
#else /* !COMPAT_43 */
/* Or here. */
#endif /* COMPAT_43 */

#ifndef COMPAT_43
/* Yet another large region here, or other conditional code. */
#else /* COMPAT_43 */
/* Or here. */
#endif /* !COMPAT_43 */
.Ed
.Pp
The project prefers the use of
.St -isoC-99
unsigned integer identifiers of the form
.Vt uintXX_t
rather than the older
.Bx Ns -style
integer identifiers of the form
.Vt u_intXX_t .
New code should use the former, and old code should be converted to
the new form if other major work is being done in that area and
there is no overriding reason to prefer the older
.Bx Ns -style .
Like white-space commits, care should be taken in making
.Vt uintXX_t
only commits.
.Pp
Similarly, the project prefers the use of
ISO C99
.Vt bool
rather than the older
.Vt int
or
.Vt boolean_t .
New code should use
.Vt bool ,
and old code may be converted if it is
reasonable to do so.
Literal values are named
.Dv true
and
.Dv false .
These are preferred to the old spellings
.Dv TRUE
and
.Dv FALSE .
Userspace code should include
.In stdbool.h ,
while kernel code should include
.In sys/types.h .
.Pp
Likewise, the project prefers
ISO C99
designated initializers when it makes sense to do so.
.Pp
Enumeration values are all uppercase.
.Bd -literal
enum enumtype { ONE, TWO } et;
.Ed
.Pp
The use of internal_underscores in identifiers is preferred over
camelCase or TitleCase.
.Pp
In declarations, do not put any whitespace between asterisks and
adjacent tokens, except for tokens that are identifiers related to
types.
(These identifiers are the names of basic types, type
qualifiers, and
.Ic typedef Ns -names
other than the one being declared.)
Separate these identifiers from asterisks using a single space.
.Pp
When declaring variables in structures, declare them sorted by use, then
by size (largest to smallest), and then in alphabetical order.
The first category normally does not apply, but there are exceptions.
Each one gets its own line.
Try to make the structure
readable by aligning the member names using either one or two tabs
depending upon your judgment.
You should use one tab only if it suffices to align at least 90% of
the member names.
Names following extremely long types
should be separated by a single space.
.Pp
Major structures should be declared at the top of the file in which they
are used, or in separate header files if they are used in multiple
source files.
Use of the structures should be by separate declarations
and should be
.Ic extern
if they are declared in a header file.
.Bd -literal
struct foo {
	struct foo	*next;		/* List of active foo. */
	struct mumble	amumble;	/* Comment for mumble. */
	int		bar;		/* Try to align the comments. */
	struct verylongtypename *baz;	/* Does not fit in 2 tabs. */
};
struct foo *foohead;			/* Head of global foo list. */
.Ed
.Pp
Use
.Xr queue 3
macros rather than rolling your own lists, whenever possible.
Thus,
the previous example would be better written:
.Bd -literal
#include <sys/queue.h>

struct foo {
	LIST_ENTRY(foo)	link;		/* Use queue macros for foo lists. */
	struct mumble	amumble;	/* Comment for mumble. */
	int		bar;		/* Try to align the comments. */
	struct verylongtypename *baz;	/* Does not fit in 2 tabs. */
};
LIST_HEAD(, foo) foohead;		/* Head of global foo list. */
.Ed
.Pp
Avoid using typedefs for structure types.
Typedefs are problematic because they do not properly hide their
underlying type; for example you need to know if the typedef is
the structure itself or a pointer to the structure.
In addition they must be declared exactly once, whereas an
incomplete structure type can be mentioned as many times as
necessary.
Typedefs are difficult to use in stand-alone header files:
the header that defines the typedef must be included
before the header that uses it, or by the header that uses
it (which causes namespace pollution), or there must be a
back-door mechanism for obtaining the typedef.
.Pp
When convention requires a
.Ic typedef ,
make its name match the struct tag.
Avoid typedefs ending in
.Dq Li _t ,
except as specified in Standard C or by POSIX.
.Bd -literal
/* Make the structure name match the typedef. */
typedef	struct bar {
	int	level;
} BAR;
typedef	int		foo;		/* This is foo. */
typedef	const long	baz;		/* This is baz. */
.Ed
.Pp
All functions are prototyped somewhere.
.Pp
Function prototypes for private functions (i.e., functions not used
elsewhere) go at the top of the first source module.
Functions
local to one source module should be declared
.Ic static .
.Pp
Functions used from other parts of the kernel are prototyped in the
relevant include file.
Function prototypes should be listed in a logical order, preferably
alphabetical unless there is a compelling reason to use a different
ordering.
.Pp
Functions that are used locally in more than one module go into a
separate header file, e.g.,
.Qq Pa extern.h .
.Pp
Do not use the
.Dv __P
macro.
.Pp
In general code can be considered
.Dq "new code"
when it makes up about 50% or more of the file(s) involved.
This is enough
to break precedents in the existing code and use the current
.Nm
guidelines.
.Pp
The kernel has a name associated with parameter types, e.g., in the kernel
use:
.Bd -literal
void	function(int fd);
.Ed
.Pp
In header files visible to userland applications, prototypes that are
visible must use either
.Dq protected
names (ones beginning with an underscore)
or no names with the types.
It is preferable to use protected names.
E.g., use:
.Bd -literal
void	function(int);
.Ed
.Pp
or:
.Bd -literal
void	function(int _fd);
.Ed
.Pp
Prototypes may have an extra space after a tab to enable function names
to line up:
.Bd -literal
static char	*function(int _arg, const char *_arg2, struct foo *_arg3,
		    struct bar *_arg4);
static void	 usage(void);

/*
 * All major routines should have a comment briefly describing what
 * they do.  The comment before the "main" routine should describe
 * what the program does.
 */
int
main(int argc, char *argv[])
{
	char *ep;
	long num;
	int ch;
.Ed
.Pp
For consistency,
.Xr getopt 3
should be used to parse options.
Options
should be sorted in the
.Xr getopt 3
call and the
.Ic switch
statement, unless
parts of the
.Ic switch
cascade.
Elements in a
.Ic switch
statement that cascade should have a
.Li FALLTHROUGH
comment.
Numerical arguments should be checked for accuracy.
Code which is unreachable for non-obvious reasons may be marked /*
.Li NOTREACHED
*/.
.Bd -literal
	while ((ch = getopt(argc, argv, "abNn:")) != -1)
		switch (ch) {		/* Indent the switch. */
		case 'a':		/* Do not indent the case. */
			aflag = 1;	/* Indent case body one tab. */
			/* FALLTHROUGH */
		case 'b':
			bflag = 1;
			break;
		case 'N':
			Nflag = 1;
			break;
		case 'n':
			num = strtol(optarg, &ep, 10);
			if (num <= 0 || *ep != '\e0') {
				warnx("illegal number, -n argument -- %s",
				    optarg);
				usage();
			}
			break;
		case '?':
		default:
			usage();
		}
	argc -= optind;
	argv += optind;
.Ed
.Pp
Space after keywords
.Pq Ic if , while , for , return , switch .
Two styles of braces
.Ql ( \&{
and
.Ql \&} )
are allowed for single line statements.
Either they are used for all single statements, or
they are used only where needed for clarity.
Usage within a function should be consistent.
Forever loops are done with
.Ic for Ns 's ,
not
.Ic while Ns 's .
.Bd -literal
	for (p = buf; *p != '\e0'; ++p)
		;	/* nothing */
	for (;;)
		stmt;
	for (;;) {
		z = a + really + long + statement + that + needs +
		    two + lines + gets + indented + four + spaces +
		    on + the + second + and + subsequent + lines;
	}
	for (;;) {
		if (cond)
			stmt;
	}
	if (val != NULL)
		val = realloc(val, newsize);
.Ed
.Pp
Parts of a
.Ic for
loop may be left empty.
.Bd -literal
	for (; cnt < 15; cnt++) {
		stmt1;
		stmt2;
	}
.Ed
.Pp
A
.Ic for
loop may declare and initialize its counting variable.
.Bd -literal
	for (int i = 0; i < 15; i++) {
		stmt1;
	}
.Ed
.Pp
Indentation is an 8 character tab.
Second level indents are four spaces.
If you have to wrap a long statement, put the operator at the end of the
line.
.Bd -literal
	while (cnt < 20 && this_variable_name_is_too_long &&
	    ep != NULL)
		z = a + really + long + statement + that + needs +
		    two + lines + gets + indented + four + spaces +
		    on + the + second + and + subsequent + lines;
.Ed
.Pp
Do not add whitespace at the end of a line, and only use tabs
followed by spaces
to form the indentation.
Do not use more spaces than a tab will produce
and do not use spaces in front of tabs.
.Pp
Closing and opening braces go on the same line as the
.Ic else .
Braces that are not necessary may be left out.
.Bd -literal
	if (test)
		stmt;
	else if (bar) {
		stmt;
		stmt;
	} else
		stmt;
.Ed
.Pp
No spaces after function names.
Commas have a space after them.
No spaces
after
.Ql \&(
or
.Ql \&[
or preceding
.Ql \&]
or
.Ql \&)
characters.
.Bd -literal
	error = function(a1, a2);
	if (error != 0)
		exit(error);
.Ed
.Pp
Unary operators do not require spaces, binary operators do.
Do not use parentheses unless they are required for precedence or unless the
statement is confusing without them.
Remember that other people may
confuse easier than you.
Do YOU understand the following?
.Bd -literal
	a = b->c[0] + ~d == (e || f) || g && h ? i : j >> 1;
	k = !(l & FLAGS);
.Ed
.Pp
Exits should be 0 on success, or 1 on failure.
.Bd -literal
	exit(0);	/*
			 * Avoid obvious comments such as
			 * "Exit 0 on success."
			 */
}
.Ed
.Pp
The function type should be on a line by itself
preceding the function.
The opening brace of the function body should be
on a line by itself.
.Bd -literal
static char *
function(int a1, int a2, float fl, int a4, struct bar *bar)
{
.Ed
.Pp
When declaring variables in functions declare them sorted by size,
then in alphabetical order; multiple ones per line are okay.
If a line overflows reuse the type keyword.
Variables may be initialized where declared especially when they
are constant for the rest of the scope.
Declarations may be placed before executable lines at the start
of any block.
Calls to complicated functions should be avoided when initializing variables.
.Bd -literal
	struct foo one, *two;
	struct baz *three = bar_get_baz(bar);
	double four;
	int *five, six;
	char *seven, eight, nine, ten, eleven, twelve;

	four = my_complicated_function(a1, f1, a4);
.Ed
.Pp
Do not declare functions inside other functions; ANSI C says that
such declarations have file scope regardless of the nesting of the
declaration.
Hiding file declarations in what appears to be a local
scope is undesirable and will elicit complaints from a good compiler.
.Pp
Casts and
.Ic sizeof Ns 's
are not followed by a space.
.Ic sizeof Ns 's
are written with parenthesis always.
The redundant parenthesis rules do not apply to
.Fn sizeof var
instances.
.Pp
.Dv NULL
is the preferred null pointer constant.
Use
.Dv NULL
instead of
.Vt ( "type *" ) Ns 0
or
.Vt ( "type *" ) Ns Dv NULL
in contexts where the compiler knows the
type, e.g., in assignments.
Use
.Vt ( "type *" ) Ns Dv NULL
in other contexts,
in particular for all function args.
(Casting is essential for
variadic args and is necessary for other args if the function prototype
might not be in scope.)
Test pointers against
.Dv NULL ,
e.g., use:
.Bd -literal
(p = f()) == NULL
.Ed
.Pp
not:
.Bd -literal
!(p = f())
.Ed
.Pp
Do not use
.Ic \&!
for tests unless it is a boolean, e.g., use:
.Bd -literal
if (*p == '\e0')
.Ed
.Pp
not:
.Bd -literal
if (!*p)
.Ed
.Pp
Routines returning
.Vt "void *"
should not have their return values cast
to any pointer type.
.Pp
Values in
.Ic return
statements should be enclosed in parentheses.
.Pp
Use
.Xr err 3
or
.Xr warn 3 ,
do not roll your own.
.Bd -literal
	if ((four = malloc(sizeof(struct foo))) == NULL)
		err(1, (char *)NULL);
	if ((six = (int *)overflow()) == NULL)
		errx(1, "number overflowed");
	return (eight);
}
.Ed
.Pp
When converting K&R style declarations to ANSI style, preserve
any comments about parameters.
.Pp
Long parameter lists are wrapped with a normal four space indent.
.Pp
Variable numbers of arguments should look like this:
.Bd -literal
#include <stdarg.h>

void
vaf(const char *fmt, ...)
{
	va_list ap;

	va_start(ap, fmt);
	STUFF;
	va_end(ap);
	/* No return needed for void functions. */
}

static void
usage(void)
{
	/* Optional blank line goes here. */
.Ed
.Pp
Optionally, insert a blank line at the beginning of functions with no local
variables.
Older versions of this
.Nm
document required the blank line convention, so it is widely used in existing
code.
.Pp
Do not insert a blank line at the beginning of functions with local variables.
Instead, these should have local variable declarations first, followed by one
blank line, followed by the first statement.
.Pp
Use
.Xr printf 3 ,
not
.Xr fputs 3 ,
.Xr puts 3 ,
.Xr putchar 3 ,
whatever; it is faster and usually cleaner, not
to mention avoiding stupid bugs.
.Pp
Usage statements should look like the manual pages
.Sx SYNOPSIS .
The usage statement should be structured in the following order:
.Bl -enum
.It
Options without operands come first,
in alphabetical order,
inside a single set of brackets
.Ql ( \&[
and
.Ql \&] ) .
.It
Options with operands come next,
also in alphabetical order,
with each option and its argument inside its own pair of brackets.
.It
Required arguments
(if any)
are next,
listed in the order they should be specified on the command line.
.It
Finally,
any optional arguments should be listed,
listed in the order they should be specified,
and all inside brackets.
.El
.Pp
A bar
.Pq Ql \&|
separates
.Dq either-or
options/arguments,
and multiple options/arguments which are specified together are
placed in a single set of brackets.
.Bd -literal -offset 4n
"usage: f [-aDde] [-b b_arg] [-m m_arg] req1 req2 [opt1 [opt2]]\en"
"usage: f [-a | -b] [-c [-dEe] [-n number]]\en"
.Ed
.Bd -literal
	(void)fprintf(stderr, "usage: f [-ab]\en");
	exit(1);
}
.Ed
.Pp
Note that the manual page options description should list the options in
pure alphabetical order.
That is, without regard to whether an option takes arguments or not.
The alphabetical ordering should take into account the case ordering
shown above.
.Pp
New core kernel code should be reasonably compliant with the
.Nm
guides.
The guidelines for third-party maintained modules and device drivers are more
relaxed but at a minimum should be internally consistent with their style.
.Pp
Stylistic changes (including whitespace changes) are hard on the source
repository and are to be avoided without good reason.
Code that is approximately
.Fx
KNF
.Nm
compliant in the repository must not diverge from compliance.
.Pp
Whenever possible, code should be run through a code checker
(e.g., various static analyzers or
.Nm cc Fl Wall )
and produce minimal warnings.
.Pp
New code should use
.Fn _Static_assert
instead of the older
.Fn CTASSERT .
.Sh FILES
.Bl -tag -width indent
.It Pa /usr/src/tools/build/checkstyle9.pl
A script to check for violations of
.Nm
in a source file.
.It Pa /usr/src/tools/tools/editing/freebsd.el
An Emacs plugin to follow the
.Fx
.Nm
indentation rules.
.It Pa /usr/src/tools/tools/editing/freebsd.vim
A Vim plugin to follow the
.Fx
.Nm
indentation rules.
.El
.Sh SEE ALSO
.Xr indent 1 ,
.Xr err 3 ,
.Xr warn 3 ,
.Xr style.Makefile 5 ,
.Xr style.mdoc 5 ,
.Xr style.lua 9
.Sh HISTORY
This manual page is largely based on the
.Pa src/admin/style/style
file from the
.Bx 4.4 Lite2
release, with occasional updates to reflect the current practice and
desire of the
.Fx
project.
.Pa src/admin/style/style
is a codification by the CSRG of the programming style of Ken Thompson and
Dennis Ritchie in
.At v6 .