.\"- .\" Copyright (c) 1995-2022 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. .\" .Dd December 13, 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. The preferred line width is 80 characters, but some exceptions are made when a slightly longer line is clearer or easier to read. Anything that is frequently grepped for, such as diagnostic, error, or panic messages, should not be broken up over multiple lines despite this rule. 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). */ /* * VERY important single-line comments look like this. */ /* Most single-line comments look like this. */ // Although they may look like this. /* * Multi-line comments look like this. Make them real sentences. Fill * them so they look like real paragraphs. */ .Ed C++ comments may be used in C and C++ code. Single-line comments should be consistently either C or C++ within a file. Multi-line comments should also be consistently either C or C++, but may differ from single-line comments. .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. Non-C/C++ source files follow the example above, while C/C++ source files follow the one below. Version control system ID tags should only exist once in a file (unlike in this one). 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. Add .Dq Li "From: " and FreeBSD git hash with full path name if the file was derived from another FreeBSD file and include relevant copyright info from the original file. .Bd -literal .Ed .Pp Leave one blank line before the header files. .Pp Kernel include files .Pa ( sys/*.h ) come 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 /* Non-local includes in angle brackets. */ #include #include #include #include .Ed .Pp For a network program, put the network include files next. .Bd -literal #include #include #include #include #include .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 .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 .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 space or tab character between the .Ic #define and the macro name, but be consistent within a file. 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 #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 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 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 execute some code and then cascade to the next case 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 in any block, but must be placed before statements. 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 test without a comparison, or with unary .Ic \&! (except for booleans). For example, use: .Bd -literal if (*p == '\e0') .Ed .Pp not: .Bd -literal if (!*p) .Ed .Pp Prefer: .Bd -literal if (count != 0) .Ed .Pp over: .Bd -literal if (count) .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 Do not use K&R style declarations or definitions, they are obsolete and are forbidden in C23. Compilers warn of their use and some treat them as an error by default. When converting K&R style definitions 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 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) { .Ed .Pp Functions should have local variable declarations first, followed by one blank line, followed by the first statement. If no local variables are declared, the first line should be a statement. Older versions of this .Nm document required a blank line before code. Such lines should be removed when signficant changes are made to the code. .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 . .Pp .Fn __predict_true and .Fn __predict_false should only be used in frequently executed code when it makes the code measurably faster. It is wasteful to make predictions for infrequently run code, like subsystem initialization. When using branch prediction hints, atypical error conditions should use .Fn __predict_false (document the exceptions). Operations that almost always succeed use .Fn __predict_true . Only use the annotation for the entire if statement, rather than individual clauses. Do not add these annotations without empirical evidence of the likelihood of the branch. .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 was originally based on the .Pa src/admin/style/style file from the .Bx 4.4 Lite2 release, with frequent 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 .