9 - 8ch indent, no tabs, except for files in `man/` which are 2ch indent, and
10 still no tabs, and shell scripts, which are 4ch indent, and no tabs either.
12 - We prefer `/* comments */` over `// comments` in code you commit,
13 please. This way `// comments` are left for developers to use for local,
14 temporary commenting of code for debug purposes (i.e. uncommittable stuff),
15 making such comments easily discernible from explanatory, documenting code
16 comments (i.e. committable stuff).
18 - Don't break code lines too eagerly. We do **not** force line breaks at 80ch,
19 all of today's screens should be much larger than that. But then again, don't
20 overdo it, ~109ch should be enough really. The `.editorconfig`, `.vimrc` and
21 `.dir-locals.el` files contained in the repository will set this limit up for
22 you automatically, if you let them (as well as a few other things). Please
23 note that emacs loads `.dir-locals.el` automatically, but vim needs to be
24 configured to load `.vimrc`, see that file for instructions.
41 - Single-line `if` blocks should not be enclosed in `{}`. Write this:
56 - Do not write `foo ()`, write `foo()`.
58 - Preferably allocate local variables on the top of the block:
71 - structs in `PascalCase` (with exceptions, such as public API structs),
72 variables and functions in `snake_case`.
74 - To minimize strict aliasing violations, we prefer unions over casting.
76 - Do not issue NSS requests (that includes user name and host name
77 lookups) from PID 1 as this might trigger deadlocks when those
78 lookups involve synchronously talking to services that we would need
81 - Do not synchronously talk to any other service from PID 1, due to
84 - Be exceptionally careful when formatting and parsing floating point
85 numbers. Their syntax is locale dependent (i.e. `5.000` in en_US is
86 generally understood as 5, while in de_DE as 5000.).
88 - Do not mix function invocations with variable definitions in one
109 - Use `goto` for cleaning up, and only use it for that. i.e. you may
110 only jump to the end of a function, and little else. Never jump
113 - Avoid static variables, except for caches and very few other
114 cases. Think about thread-safety! While most of our code is never
115 used in threaded environments, at least the library code should make
116 sure it works correctly in them. Instead of doing a lot of locking
117 for that, we tend to prefer using TLS to do per-thread caching (which
118 only works for small, fixed-size cache objects), or we disable
119 caching for any thread that is not the main thread. Use
120 `is_main_thread()` to detect whether the calling thread is the main
123 - Do not write functions that clobber call-by-reference variables on
124 failure. Use temporary variables for these cases and change the
125 passed in variables only on success.
128 - Instead of using `memzero()`/`memset()` to initialize structs allocated
129 on the stack, please try to use c99 structure initializers. It's
130 short, prettier and actually even faster at execution. Hence:
148 - The order in which header files are included doesn't matter too
149 much. systemd-internal headers must not rely on an include order, so
150 it is safe to include them in any order possible.
151 However, to not clutter global includes, and to make sure internal
152 definitions will not affect global headers, please always include the
153 headers of external components first (these are all headers enclosed
154 in <>), followed by our own exported headers (usually everything
155 that's prefixed by `sd-`), and then followed by internal headers.
156 Furthermore, in all three groups, order all includes alphabetically
157 so duplicate includes can easily be detected.
159 - To implement an endless loop, use `for (;;)` rather than `while (1)`.
160 The latter is a bit ugly anyway, since you probably really
161 meant `while (true)`. To avoid the discussion what the right
162 always-true expression for an infinite while loop is, our
163 recommendation is to simply write it without any such expression by
166 - Avoid leaving long-running child processes around, i.e. `fork()`s that
167 are not followed quickly by an `execv()` in the child. Resource
168 management is unclear in this case, and memory CoW will result in
169 unexpected penalties in the parent much, much later on.
171 - Don't block execution for arbitrary amounts of time using `usleep()`
172 or a similar call, unless you really know what you do. Just "giving
173 something some time", or so is a lazy excuse. Always wait for the
174 proper event, instead of doing time-based poll loops.
176 - To determine the length of a constant string `"foo"`, don't bother with
177 `sizeof("foo")-1`, please use `strlen()` instead (both gcc and clang optimize
178 the call away for fixed strings). The only exception is when declaring an
179 array. In that case use STRLEN, which evaluates to a static constant and
180 doesn't force the compiler to create a VLA.
182 - Please avoid using global variables as much as you can. And if you
183 do use them make sure they are static at least, instead of
184 exported. Especially in library-like code it is important to avoid
185 global variables. Why are global variables bad? They usually hinder
186 generic reusability of code (since they break in threaded programs,
187 and usually would require locking there), and as the code using them
188 has side-effects make programs non-transparent. That said, there are
189 many cases where they explicitly make a lot of sense, and are OK to
190 use. For example, the log level and target in `log.c` is stored in a
191 global variable, and that's OK and probably expected by most. Also
192 in many cases we cache data in global variables. If you add more
193 caches like this, please be careful however, and think about
194 threading. Only use static variables if you are sure that
195 thread-safety doesn't matter in your case. Alternatively, consider
196 using TLS, which is pretty easy to use with gcc's `thread_local`
197 concept. It's also OK to store data that is inherently global in
198 global variables, for example data parsed from command lines, see
201 - Make sure to enforce limits on every user controllable resource. If the user
202 can allocate resources in your code, your code must enforce some form of
203 limits after which it will refuse operation. It's fine if it is hard-coded (at
204 least initially), but it needs to be there. This is particularly important
205 for objects that unprivileged users may allocate, but also matters for
206 everything else any user may allocated.
208 - You might wonder what kind of common code belongs in `src/shared/` and what
209 belongs in `src/basic/`. The split is like this: anything that is used to
210 implement the public shared object we provide (sd-bus, sd-login, sd-id128,
211 nss-systemd, nss-mymachines, nss-resolve, nss-myhostname, pam_systemd), must
212 be located in `src/basic` (those objects are not allowed to link to
213 libsystemd-shared.so). Conversely, anything which is shared between multiple
214 components and does not need to be in `src/basic/`, should be in
220 - may be used by all code in the tree
221 - may not use any code outside of `src/basic/`
224 - may be used by all code in the tree, except for code in `src/basic/`
225 - may not use any code outside of `src/basic/`, `src/libsystemd/`
228 - may be used by all code in the tree, except for code in `src/basic/`,
229 `src/libsystemd/`, `src/nss-*`, `src/login/pam_systemd.*`, and files under
230 `src/journal/` that end up in `libjournal-client.a` convenience library.
231 - may not use any code outside of `src/basic/`, `src/libsystemd/`, `src/shared/`
233 - Our focus is on the GNU libc (glibc), not any other libcs. If other libcs are
234 incompatible with glibc it's on them. However, if there are equivalent POSIX
235 and Linux/GNU-specific APIs, we generally prefer the POSIX APIs. If there
236 aren't, we are happy to use GNU or Linux APIs, and expect non-GNU
237 implementations of libc to catch up with glibc.
239 - Whenever installing a signal handler, make sure to set `SA_RESTART` for it, so
240 that interrupted system calls are automatically restarted, and we minimize
241 hassles with handling `EINTR` (in particular as `EINTR` handling is pretty broken
244 - When applying C-style unescaping as well as specifier expansion on the same
245 string, always apply the C-style unescaping fist, followed by the specifier
246 expansion. When doing the reverse, make sure to escape `%` in specifier-style
247 first (i.e. `%` → `%%`), and then do C-style escaping where necessary.
251 - The destructors always deregister the object from the next bigger object, not
252 the other way around.
254 - For robustness reasons, destructors should be able to destruct
255 half-initialized objects, too.
257 - When you define a destructor or `unref()` call for an object, please accept a
258 `NULL` object and simply treat this as NOP. This is similar to how libc
259 `free()` works, which accepts `NULL` pointers and becomes a NOP for them. By
260 following this scheme a lot of `if` checks can be removed before invoking
261 your destructor, which makes the code substantially more readable and robust.
263 - Related to this: when you define a destructor or `unref()` call for an
264 object, please make it return the same type it takes and always return `NULL`
265 from it. This allows writing code like this:
271 which will always work regardless if `p` is initialized or not,x and
272 guarantees that `p` is `NULL` afterwards, all in just one line.
276 - Error codes are returned as negative `Exxx`. e.g. `return -EINVAL`. There are
277 some exceptions: for constructors, it is OK to return `NULL` on OOM. For
278 lookup functions, `NULL` is fine too for "not found".
280 Be strict with this. When you write a function that can fail due to more than
281 one cause, it *really* should have an `int` as the return value for the error
284 - Do not bother with error checking whether writing to stdout/stderr worked.
286 - Do not log errors from "library" code, only do so from "main program"
287 code. (With one exception: it is OK to log with DEBUG level from any code,
288 with the exception of maybe inner loops).
290 - In public API calls, you **must** validate all your input arguments for
291 programming error with `assert_return()` and return a sensible return
292 code. In all other calls, it is recommended to check for programming errors
293 with a more brutal `assert()`. We are more forgiving to public users than for
294 ourselves! Note that `assert()` and `assert_return()` really only should be
295 used for detecting programming errors, not for runtime errors. `assert()` and
296 `assert_return()` by usage of `_likely_()` inform the compiler that he should
297 not expect these checks to fail, and they inform fellow programmers about the
298 expected validity and range of parameters.
300 - When you invoke certain calls like `unlink()`, or `mkdir_p()` and you know it
301 is safe to ignore the error it might return (because a later call would
302 detect the failure anyway, or because the error is in an error path and you
303 thus couldn't do anything about it anyway), then make this clear by casting
304 the invocation explicitly to `(void)`. Code checks like Coverity understand
305 that, and will not complain about ignored error codes. Hence, please use
309 (void) unlink("/foo/bar/baz");
312 instead of just this:
315 unlink("/foo/bar/baz");
318 Don't cast function calls to `(void)` that return no error
319 conditions. Specifically, the various `xyz_unref()` calls that return a
320 `NULL` object shouldn't be cast to `(void)`, since not using the return value
321 does not hide any errors.
323 - When returning a return code from `main()`, please preferably use
324 `EXIT_FAILURE` and `EXIT_SUCCESS` as defined by libc.
328 - For every function you add, think about whether it is a "logging" function or
329 a "non-logging" function. "Logging" functions do logging on their own,
330 "non-logging" function never log on their own and expect their callers to
331 log. All functions in "library" code, i.e. in `src/shared/` and suchlike must
332 be "non-logging". Every time a "logging" function calls a "non-logging"
333 function, it should log about the resulting errors. If a "logging" function
334 calls another "logging" function, then it should not generate log messages,
335 so that log messages are not generated twice for the same errors.
337 - If possible, do a combined log & return operation:
342 return log_(error|warning|notice|...)_errno(r, "Failed to ...: %m");
345 If the error value is "synthetic", i.e. it was not received from
346 the called function, use `SYNTHETIC_ERRNO` wrapper to tell the logging
347 system to not log the errno value, but still return it:
350 n = read(..., s, sizeof s);
352 return log_error_errno(SYNTHETIC_ERRNO(EIO), "Failed to read ...");
357 - Always check OOM. There is no excuse. In program code, you can use
358 `log_oom()` for then printing a short message, but not in "library" code.
360 - Avoid fixed-size string buffers, unless you really know the maximum size and
361 that maximum size is small. They are a source of errors, since they possibly
362 result in truncated strings. It is often nicer to use dynamic memory,
363 `alloca()` or VLAs. If you do allocate fixed-size strings on the stack, then
364 it is probably only OK if you either use a maximum size such as `LINE_MAX`,
365 or count in detail the maximum size a string can have. (`DECIMAL_STR_MAX` and
366 `DECIMAL_STR_WIDTH` macros are your friends for this!)
368 Or in other words, if you use `char buf[256]` then you are likely doing
371 - Make use of `_cleanup_free_` and friends. It makes your code much nicer to
374 - Use `alloca()`, but never forget that it is not OK to invoke `alloca()`
375 within a loop or within function call parameters. `alloca()` memory is
376 released at the end of a function, and not at the end of a `{}` block. Thus,
377 if you invoke it in a loop, you keep increasing the stack pointer without
378 ever releasing memory again. (VLAs have better behavior in this case, so
379 consider using them as an alternative.) Regarding not using `alloca()`
380 within function parameters, see the BUGS section of the `alloca(3)` man page.
382 - If you want to concatenate two or more strings, consider using `strjoina()`
383 or `strjoin()` rather than `asprintf()`, as the latter is a lot slower. This
384 matters particularly in inner loops (but note that `strjoina()` cannot be
389 - Think about the types you use. If a value cannot sensibly be negative, do not
390 use `int`, but use `unsigned`.
392 - Use `char` only for actual characters. Use `uint8_t` or `int8_t` when you
393 actually mean a byte-sized signed or unsigned integers. When referring to a
394 generic byte, we generally prefer the unsigned variant `uint8_t`. Do not use
395 types based on `short`. They *never* make sense. Use `int`, `long`, `long
396 long`, all in unsigned and signed fashion, and the fixed-size types
397 `uint8_t`, `uint16_t`, `uint32_t`, `uint64_t`, `int8_t`, `int16_t`, `int32_t`
398 and so on, as well as `size_t`, but nothing else. Do not use kernel types
399 like `u32` and so on, leave that to the kernel.
401 - Stay uniform. For example, always use `usec_t` for time values. Do not mix
402 `usec` and `msec`, and `usec` and whatnot.
404 - Never use the `off_t` type, and particularly avoid it in public APIs. It's
405 really weirdly defined, as it usually is 64-bit and we don't support it any
406 other way, but it could in theory also be 32-bit. Which one it is depends on
407 a compiler switch chosen by the compiled program, which hence corrupts APIs
408 using it unless they can also follow the program's choice. Moreover, in
409 systemd we should parse values the same way on all architectures and cannot
410 expose `off_t` values over D-Bus. To avoid any confusion regarding conversion
411 and ABIs, always use simply `uint64_t` directly.
413 - Unless you allocate an array, `double` is always a better choice than
414 `float`. Processors speak `double` natively anyway, so there is no speed
415 benefit, and on calls like `printf()` `float`s get promoted to `double`s
416 anyway, so there is no point.
418 - Use the bool type for booleans, not integers. One exception: in public
419 headers (i.e those in `src/systemd/sd-*.h`) use integers after all, as `bool`
420 is C99 and in our public APIs we try to stick to C89 (with a few extension).
424 - When you allocate a file descriptor, it should be made `O_CLOEXEC` right from
425 the beginning, as none of our files should leak to forked binaries by
426 default. Hence, whenever you open a file, `O_CLOEXEC` must be specified,
427 right from the beginning. This also applies to sockets. Effectively, this
428 means that all invocations to:
430 - `open()` must get `O_CLOEXEC` passed,
431 - `socket()` and `socketpair()` must get `SOCK_CLOEXEC` passed,
432 - `recvmsg()` must get `MSG_CMSG_CLOEXEC` set,
433 - `F_DUPFD_CLOEXEC` should be used instead of `F_DUPFD`, and so on,
434 - invocations of `fopen()` should take `e`.
436 - It's a good idea to use `O_NONBLOCK` when opening 'foreign' regular files,
437 i.e. file system objects that are supposed to be regular files whose paths
438 where specified by the user and hence might actually refer to other types of
439 file system objects. This is a good idea so that we don't end up blocking on
440 'strange' file nodes, for example if the user pointed us to a FIFO or device
441 node which may block when opening. Moreover even for actual regular files
442 `O_NONBLOCK` has a benefit: it bypasses any mandatory lock that might be in
443 effect on the regular file. If in doubt consider turning off `O_NONBLOCK`
448 - If you parse a command line, and want to store the parsed parameters in
449 global variables, please consider prefixing their names with `arg_`. We have
450 been following this naming rule in most of our tools, and we should continue
451 to do so, as it makes it easy to identify command line parameter variables,
452 and makes it clear why it is OK that they are global variables.
454 - Command line option parsing:
455 - Do not print full `help()` on error, be specific about the error.
456 - Do not print messages to stdout on error.
457 - Do not POSIX_ME_HARDER unless necessary, i.e. avoid `+` in option string.
461 - Variables and functions **must** be static, unless they have a prototype, and
462 are supposed to be exported.
464 - Public API calls (i.e. functions exported by our shared libraries)
465 must be marked `_public_` and need to be prefixed with `sd_`. No
466 other functions should be prefixed like that.
468 - When exposing public C APIs, be careful what function parameters you make
469 `const`. For example, a parameter taking a context object should probably not
470 be `const`, even if you are writing an otherwise read-only accessor function
471 for it. The reason is that making it `const` fixates the contract that your
472 call won't alter the object ever, as part of the API. However, that's often
473 quite a promise, given that this even prohibits object-internal caching or
474 lazy initialization of object variables. Moreover, it's usually not too
475 useful for client applications. Hence, please be careful and avoid `const` on
476 object parameters, unless you are very sure `const` is appropriate.
478 ## Referencing Concepts
480 - When referring to a configuration file option in the documentation and such,
481 please always suffix it with `=`, to indicate that it is a configuration file
484 - When referring to a command line option in the documentation and such, please
485 always prefix with `--` or `-` (as appropriate), to indicate that it is a
488 - When referring to a file system path that is a directory, please always
489 suffix it with `/`, to indicate that it is a directory, not a regular file
490 (or other file system object).
492 ## Functions to Avoid
494 - Use `memzero()` or even better `zero()` instead of `memset(..., 0, ...)`
496 - Please use `streq()` and `strneq()` instead of `strcmp()`, `strncmp()` where
497 applicable (i.e. wherever you just care about equality/inequality, not about
500 - Never use `strtol()`, `atoi()` and similar calls. Use `safe_atoli()`,
501 `safe_atou32()` and suchlike instead. They are much nicer to use in most
502 cases and correctly check for parsing errors.
504 - `htonl()`/`ntohl()` and `htons()`/`ntohs()` are weird. Please use `htobe32()`
505 and `htobe16()` instead, it's much more descriptive, and actually says what
506 really is happening, after all `htonl()` and `htons()` don't operate on
507 `long`s and `short`s as their name would suggest, but on `uint32_t` and
508 `uint16_t`. Also, "network byte order" is just a weird name for "big endian",
509 hence we might want to call it "big endian" right-away.
511 - Please never use `dup()`. Use `fcntl(fd, F_DUPFD_CLOEXEC, 3)` instead. For
512 two reason: first, you want `O_CLOEXEC` set on the new `fd` (see
513 above). Second, `dup()` will happily duplicate your `fd` as 0, 1, 2,
514 i.e. stdin, stdout, stderr, should those `fd`s be closed. Given the special
515 semantics of those `fd`s, it's probably a good idea to avoid
516 them. `F_DUPFD_CLOEXEC` with `3` as parameter avoids them.
518 - Don't use `fgets()`, it's too hard to properly handle errors such as overly
519 long lines. Use `read_line()` instead, which is our own function that handles
522 - Don't invoke `exit()`, ever. It is not replacement for proper error
523 handling. Please escalate errors up your call chain, and use normal `return`
524 to exit from the main function of a process. If you `fork()`ed off a child
525 process, please use `_exit()` instead of `exit()`, so that the exit handlers
528 - We never use the POSIX version of `basename()` (which glibc defines it in
529 `libgen.h`), only the GNU version (which glibc defines in `string.h`). The
530 only reason to include `libgen.h` is because `dirname()` is needed. Every
531 time you need that please immediately undefine `basename()`, and add a
532 comment about it, so that no code ever ends up using the POSIX version!
536 - Commit message subject lines should be prefixed with an appropriate component
537 name of some kind. For example "journal: ", "nspawn: " and so on.
539 - Do not use "Signed-Off-By:" in your commit messages. That's a kernel thing we
540 don't do in the systemd project.