Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 1 | ================= |
Stefan Weil | 6576b74 | 2012-04-07 09:23:37 +0200 | [diff] [blame] | 2 | QEMU Coding Style |
aliguori | e68b98d | 2009-04-05 17:40:34 +0000 | [diff] [blame] | 3 | ================= |
| 4 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 5 | .. contents:: Table of Contents |
| 6 | |
Blue Swirl | b646968 | 2011-01-20 20:58:56 +0000 | [diff] [blame] | 7 | Please use the script checkpatch.pl in the scripts directory to check |
| 8 | patches before submitting. |
| 9 | |
Daniel P. Berrangé | 9f8efa7 | 2019-08-29 17:04:32 +0100 | [diff] [blame] | 10 | Formatting and style |
| 11 | ******************** |
| 12 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 13 | Whitespace |
| 14 | ========== |
aliguori | e68b98d | 2009-04-05 17:40:34 +0000 | [diff] [blame] | 15 | |
| 16 | Of course, the most important aspect in any coding style is whitespace. |
| 17 | Crusty old coders who have trouble spotting the glasses on their noses |
| 18 | can tell the difference between a tab and eight spaces from a distance |
Jonathan Neuschäfer | 56bef85 | 2016-09-30 02:04:28 +0200 | [diff] [blame] | 19 | of approximately fifteen parsecs. Many a flamewar has been fought and |
aliguori | e68b98d | 2009-04-05 17:40:34 +0000 | [diff] [blame] | 20 | lost on this issue. |
| 21 | |
| 22 | QEMU indents are four spaces. Tabs are never used, except in Makefiles |
edgar_igl | 1cb499f | 2009-04-07 02:10:16 +0000 | [diff] [blame] | 23 | where they have been irreversibly coded into the syntax. |
aliguori | e68b98d | 2009-04-05 17:40:34 +0000 | [diff] [blame] | 24 | Spaces of course are superior to tabs because: |
| 25 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 26 | * You have just one way to specify whitespace, not two. Ambiguity breeds |
| 27 | mistakes. |
| 28 | * The confusion surrounding 'use tabs to indent, spaces to justify' is gone. |
| 29 | * Tab indents push your code to the right, making your screen seriously |
| 30 | unbalanced. |
| 31 | * Tabs will be rendered incorrectly on editors who are misconfigured not |
| 32 | to use tab stops of eight positions. |
| 33 | * Tabs are rendered badly in patches, causing off-by-one errors in almost |
| 34 | every line. |
| 35 | * It is the QEMU coding style. |
aliguori | e68b98d | 2009-04-05 17:40:34 +0000 | [diff] [blame] | 36 | |
| 37 | Do not leave whitespace dangling off the ends of lines. |
| 38 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 39 | Multiline Indent |
| 40 | ---------------- |
Wei Yang | 6ac1fca | 2019-03-04 15:16:30 +0800 | [diff] [blame] | 41 | |
| 42 | There are several places where indent is necessary: |
| 43 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 44 | * if/else |
| 45 | * while/for |
| 46 | * function definition & call |
Wei Yang | 6ac1fca | 2019-03-04 15:16:30 +0800 | [diff] [blame] | 47 | |
| 48 | When breaking up a long line to fit within line width, we need a proper indent |
| 49 | for the following lines. |
| 50 | |
| 51 | In case of if/else, while/for, align the secondary lines just after the |
| 52 | opening parenthesis of the first. |
| 53 | |
| 54 | For example: |
| 55 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 56 | .. code-block:: c |
| 57 | |
Wei Yang | 6ac1fca | 2019-03-04 15:16:30 +0800 | [diff] [blame] | 58 | if (a == 1 && |
| 59 | b == 2) { |
| 60 | |
| 61 | while (a == 1 && |
| 62 | b == 2) { |
| 63 | |
| 64 | In case of function, there are several variants: |
| 65 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 66 | * 4 spaces indent from the beginning |
| 67 | * align the secondary lines just after the opening parenthesis of the first |
Wei Yang | 6ac1fca | 2019-03-04 15:16:30 +0800 | [diff] [blame] | 68 | |
| 69 | For example: |
| 70 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 71 | .. code-block:: c |
| 72 | |
Wei Yang | 6ac1fca | 2019-03-04 15:16:30 +0800 | [diff] [blame] | 73 | do_something(x, y, |
| 74 | z); |
| 75 | |
| 76 | do_something(x, y, |
| 77 | z); |
| 78 | |
| 79 | do_something(x, do_another(y, |
| 80 | z)); |
| 81 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 82 | Line width |
| 83 | ========== |
aliguori | e68b98d | 2009-04-05 17:40:34 +0000 | [diff] [blame] | 84 | |
Paolo Bonzini | 8fbe3d1 | 2015-09-07 11:53:02 +0200 | [diff] [blame] | 85 | Lines should be 80 characters; try not to make them longer. |
| 86 | |
| 87 | Sometimes it is hard to do, especially when dealing with QEMU subsystems |
| 88 | that use long function or symbol names. Even in that case, do not make |
| 89 | lines much longer than 80 characters. |
aliguori | e68b98d | 2009-04-05 17:40:34 +0000 | [diff] [blame] | 90 | |
| 91 | Rationale: |
aliguori | e68b98d | 2009-04-05 17:40:34 +0000 | [diff] [blame] | 92 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 93 | * Some people like to tile their 24" screens with a 6x4 matrix of 80x24 |
| 94 | xterms and use vi in all of them. The best way to punish them is to |
| 95 | let them keep doing it. |
| 96 | * Code and especially patches is much more readable if limited to a sane |
| 97 | line length. Eighty is traditional. |
| 98 | * The four-space indentation makes the most common excuse ("But look |
| 99 | at all that white space on the left!") moot. |
| 100 | * It is the QEMU coding style. |
| 101 | |
| 102 | Naming |
| 103 | ====== |
aliguori | e68b98d | 2009-04-05 17:40:34 +0000 | [diff] [blame] | 104 | |
Anthony Liguori | c227f09 | 2009-10-01 16:12:16 -0500 | [diff] [blame] | 105 | Variables are lower_case_with_underscores; easy to type and read. Structured |
Peter Maydell | e3c52bf | 2012-01-13 20:29:49 +0000 | [diff] [blame] | 106 | type names are in CamelCase; harder to type but standing out. Enum type |
| 107 | names and function type names should also be in CamelCase. Scalar type |
Anthony Liguori | c227f09 | 2009-10-01 16:12:16 -0500 | [diff] [blame] | 108 | names are lower_case_with_underscores_ending_with_a_t, like the POSIX |
| 109 | uint64_t and family. Note that this last convention contradicts POSIX |
| 110 | and is therefore likely to be changed. |
| 111 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 112 | When wrapping standard library functions, use the prefix ``qemu_`` to alert |
Avi Kivity | 77ac486 | 2010-03-11 16:48:43 +0200 | [diff] [blame] | 113 | readers that they are seeing a wrapped version; otherwise avoid this prefix. |
| 114 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 115 | Block structure |
| 116 | =============== |
aliguori | e68b98d | 2009-04-05 17:40:34 +0000 | [diff] [blame] | 117 | |
| 118 | Every indented statement is braced; even if the block contains just one |
| 119 | statement. The opening brace is on the line that contains the control |
| 120 | flow statement that introduces the new block; the closing brace is on the |
| 121 | same line as the else keyword, or on a line by itself if there is no else |
| 122 | keyword. Example: |
| 123 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 124 | .. code-block:: c |
| 125 | |
aliguori | e68b98d | 2009-04-05 17:40:34 +0000 | [diff] [blame] | 126 | if (a == 5) { |
| 127 | printf("a was 5.\n"); |
| 128 | } else if (a == 6) { |
| 129 | printf("a was 6.\n"); |
| 130 | } else { |
| 131 | printf("a was something else entirely.\n"); |
| 132 | } |
| 133 | |
Avi Kivity | 5f070c5 | 2011-07-25 18:55:53 +0300 | [diff] [blame] | 134 | Note that 'else if' is considered a single statement; otherwise a long if/ |
| 135 | else if/else if/.../else sequence would need an indent for every else |
| 136 | statement. |
| 137 | |
aliguori | e68b98d | 2009-04-05 17:40:34 +0000 | [diff] [blame] | 138 | An exception is the opening brace for a function; for reasons of tradition |
| 139 | and clarity it comes on a line by itself: |
| 140 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 141 | .. code-block:: c |
| 142 | |
aliguori | e68b98d | 2009-04-05 17:40:34 +0000 | [diff] [blame] | 143 | void a_function(void) |
| 144 | { |
| 145 | do_something(); |
| 146 | } |
| 147 | |
| 148 | Rationale: a consistent (except for functions...) bracing style reduces |
| 149 | ambiguity and avoids needless churn when lines are added or removed. |
| 150 | Furthermore, it is the QEMU coding style. |
Eduardo Habkost | e939c6e | 2014-03-17 15:26:31 -0300 | [diff] [blame] | 151 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 152 | Declarations |
| 153 | ============ |
Eduardo Habkost | e939c6e | 2014-03-17 15:26:31 -0300 | [diff] [blame] | 154 | |
Paolo Bonzini | 690a35e | 2015-06-19 09:28:13 +0200 | [diff] [blame] | 155 | Mixed declarations (interleaving statements and declarations within |
| 156 | blocks) are generally not allowed; declarations should be at the beginning |
| 157 | of blocks. |
| 158 | |
| 159 | Every now and then, an exception is made for declarations inside a |
| 160 | #ifdef or #ifndef block: if the code looks nicer, such declarations can |
| 161 | be placed at the top of the block even if there are statements above. |
| 162 | On the other hand, however, it's often best to move that #ifdef/#ifndef |
| 163 | block to a separate function altogether. |
Gonglei | 2bb0020 | 2014-08-11 21:00:51 +0800 | [diff] [blame] | 164 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 165 | Conditional statements |
| 166 | ====================== |
Gonglei | 2bb0020 | 2014-08-11 21:00:51 +0800 | [diff] [blame] | 167 | |
| 168 | When comparing a variable for (in)equality with a constant, list the |
| 169 | constant on the right, as in: |
| 170 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 171 | .. code-block:: c |
| 172 | |
Wei Yang | 25d68ff | 2019-03-04 15:16:31 +0800 | [diff] [blame] | 173 | if (a == 1) { |
| 174 | /* Reads like: "If a equals 1" */ |
| 175 | do_something(); |
| 176 | } |
Gonglei | 2bb0020 | 2014-08-11 21:00:51 +0800 | [diff] [blame] | 177 | |
| 178 | Rationale: Yoda conditions (as in 'if (1 == a)') are awkward to read. |
| 179 | Besides, good compilers already warn users when '==' is mis-typed as '=', |
| 180 | even when the constant is on the right. |
Peter Maydell | 25ac5bb | 2017-02-03 17:58:42 +0000 | [diff] [blame] | 181 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 182 | Comment style |
| 183 | ============= |
Peter Maydell | 25ac5bb | 2017-02-03 17:58:42 +0000 | [diff] [blame] | 184 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 185 | We use traditional C-style /``*`` ``*``/ comments and avoid // comments. |
Peter Maydell | 25ac5bb | 2017-02-03 17:58:42 +0000 | [diff] [blame] | 186 | |
| 187 | Rationale: The // form is valid in C99, so this is purely a matter of |
| 188 | consistency of style. The checkpatch script will warn you about this. |
Vladimir Sementsov-Ogievskiy | 44c6d63 | 2017-07-31 19:01:32 +0300 | [diff] [blame] | 189 | |
Peter Maydell | 2948f0c | 2018-06-15 14:57:14 +0100 | [diff] [blame] | 190 | Multiline comment blocks should have a row of stars on the left, |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 191 | and the initial /``*`` and terminating ``*``/ both on their own lines: |
| 192 | |
| 193 | .. code-block:: c |
| 194 | |
Peter Maydell | 2948f0c | 2018-06-15 14:57:14 +0100 | [diff] [blame] | 195 | /* |
| 196 | * like |
| 197 | * this |
| 198 | */ |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 199 | |
Peter Maydell | 2948f0c | 2018-06-15 14:57:14 +0100 | [diff] [blame] | 200 | This is the same format required by the Linux kernel coding style. |
| 201 | |
| 202 | (Some of the existing comments in the codebase use the GNU Coding |
| 203 | Standards form which does not have stars on the left, or other |
| 204 | variations; avoid these when writing new comments, but don't worry |
| 205 | about converting to the preferred form unless you're editing that |
| 206 | comment anyway.) |
| 207 | |
| 208 | Rationale: Consistency, and ease of visually picking out a multiline |
| 209 | comment from the surrounding code. |
| 210 | |
Daniel P. Berrangé | 9f8efa7 | 2019-08-29 17:04:32 +0100 | [diff] [blame] | 211 | Language usage |
| 212 | ************** |
| 213 | |
Daniel P. Berrangé | 637f395 | 2019-08-23 17:12:05 +0100 | [diff] [blame] | 214 | Preprocessor |
| 215 | ============ |
| 216 | |
| 217 | Variadic macros |
| 218 | --------------- |
| 219 | |
| 220 | For variadic macros, stick with this C99-like syntax: |
| 221 | |
| 222 | .. code-block:: c |
| 223 | |
| 224 | #define DPRINTF(fmt, ...) \ |
| 225 | do { printf("IRQ: " fmt, ## __VA_ARGS__); } while (0) |
| 226 | |
| 227 | Include directives |
| 228 | ------------------ |
| 229 | |
| 230 | Order include directives as follows: |
| 231 | |
| 232 | .. code-block:: c |
| 233 | |
| 234 | #include "qemu/osdep.h" /* Always first... */ |
| 235 | #include <...> /* then system headers... */ |
| 236 | #include "..." /* and finally QEMU headers. */ |
| 237 | |
| 238 | The "qemu/osdep.h" header contains preprocessor macros that affect the behavior |
| 239 | of core system headers like <stdint.h>. It must be the first include so that |
| 240 | core system headers included by external libraries get the preprocessor macros |
| 241 | that QEMU depends on. |
| 242 | |
| 243 | Do not include "qemu/osdep.h" from header files since the .c file will have |
| 244 | already included it. |
| 245 | |
| 246 | C types |
| 247 | ======= |
| 248 | |
| 249 | It should be common sense to use the right type, but we have collected |
| 250 | a few useful guidelines here. |
| 251 | |
| 252 | Scalars |
| 253 | ------- |
| 254 | |
| 255 | If you're using "int" or "long", odds are good that there's a better type. |
| 256 | If a variable is counting something, it should be declared with an |
| 257 | unsigned type. |
| 258 | |
| 259 | If it's host memory-size related, size_t should be a good choice (use |
| 260 | ssize_t only if required). Guest RAM memory offsets must use ram_addr_t, |
| 261 | but only for RAM, it may not cover whole guest address space. |
| 262 | |
| 263 | If it's file-size related, use off_t. |
| 264 | If it's file-offset related (i.e., signed), use off_t. |
| 265 | If it's just counting small numbers use "unsigned int"; |
| 266 | (on all but oddball embedded systems, you can assume that that |
| 267 | type is at least four bytes wide). |
| 268 | |
| 269 | In the event that you require a specific width, use a standard type |
| 270 | like int32_t, uint32_t, uint64_t, etc. The specific types are |
| 271 | mandatory for VMState fields. |
| 272 | |
| 273 | Don't use Linux kernel internal types like u32, __u32 or __le32. |
| 274 | |
| 275 | Use hwaddr for guest physical addresses except pcibus_t |
| 276 | for PCI addresses. In addition, ram_addr_t is a QEMU internal address |
| 277 | space that maps guest RAM physical addresses into an intermediate |
| 278 | address space that can map to host virtual address spaces. Generally |
| 279 | speaking, the size of guest memory can always fit into ram_addr_t but |
| 280 | it would not be correct to store an actual guest physical address in a |
| 281 | ram_addr_t. |
| 282 | |
| 283 | For CPU virtual addresses there are several possible types. |
| 284 | vaddr is the best type to use to hold a CPU virtual address in |
| 285 | target-independent code. It is guaranteed to be large enough to hold a |
| 286 | virtual address for any target, and it does not change size from target |
| 287 | to target. It is always unsigned. |
| 288 | target_ulong is a type the size of a virtual address on the CPU; this means |
| 289 | it may be 32 or 64 bits depending on which target is being built. It should |
| 290 | therefore be used only in target-specific code, and in some |
| 291 | performance-critical built-per-target core code such as the TLB code. |
| 292 | There is also a signed version, target_long. |
| 293 | abi_ulong is for the ``*``-user targets, and represents a type the size of |
| 294 | 'void ``*``' in that target's ABI. (This may not be the same as the size of a |
| 295 | full CPU virtual address in the case of target ABIs which use 32 bit pointers |
| 296 | on 64 bit CPUs, like sparc32plus.) Definitions of structures that must match |
| 297 | the target's ABI must use this type for anything that on the target is defined |
| 298 | to be an 'unsigned long' or a pointer type. |
| 299 | There is also a signed version, abi_long. |
| 300 | |
| 301 | Of course, take all of the above with a grain of salt. If you're about |
| 302 | to use some system interface that requires a type like size_t, pid_t or |
| 303 | off_t, use matching types for any corresponding variables. |
| 304 | |
| 305 | Also, if you try to use e.g., "unsigned int" as a type, and that |
| 306 | conflicts with the signedness of a related variable, sometimes |
| 307 | it's best just to use the *wrong* type, if "pulling the thread" |
| 308 | and fixing all related variables would be too invasive. |
| 309 | |
| 310 | Finally, while using descriptive types is important, be careful not to |
| 311 | go overboard. If whatever you're doing causes warnings, or requires |
| 312 | casts, then reconsider or ask for help. |
| 313 | |
| 314 | Pointers |
| 315 | -------- |
| 316 | |
| 317 | Ensure that all of your pointers are "const-correct". |
| 318 | Unless a pointer is used to modify the pointed-to storage, |
| 319 | give it the "const" attribute. That way, the reader knows |
| 320 | up-front that this is a read-only pointer. Perhaps more |
| 321 | importantly, if we're diligent about this, when you see a non-const |
| 322 | pointer, you're guaranteed that it is used to modify the storage |
| 323 | it points to, or it is aliased to another pointer that is. |
| 324 | |
| 325 | Typedefs |
| 326 | -------- |
| 327 | |
| 328 | Typedefs are used to eliminate the redundant 'struct' keyword, since type |
| 329 | names have a different style than other identifiers ("CamelCase" versus |
| 330 | "snake_case"). Each named struct type should have a CamelCase name and a |
| 331 | corresponding typedef. |
| 332 | |
| 333 | Since certain C compilers choke on duplicated typedefs, you should avoid |
| 334 | them and declare a typedef only in one header file. For common types, |
| 335 | you can use "include/qemu/typedefs.h" for example. However, as a matter |
| 336 | of convenience it is also perfectly fine to use forward struct |
| 337 | definitions instead of typedefs in headers and function prototypes; this |
| 338 | avoids problems with duplicated typedefs and reduces the need to include |
| 339 | headers from other headers. |
| 340 | |
| 341 | Reserved namespaces in C and POSIX |
| 342 | ---------------------------------- |
| 343 | |
| 344 | Underscore capital, double underscore, and underscore 't' suffixes should be |
| 345 | avoided. |
| 346 | |
| 347 | Low level memory management |
| 348 | =========================== |
| 349 | |
| 350 | Use of the malloc/free/realloc/calloc/valloc/memalign/posix_memalign |
| 351 | APIs is not allowed in the QEMU codebase. Instead of these routines, |
| 352 | use the GLib memory allocation routines g_malloc/g_malloc0/g_new/ |
| 353 | g_new0/g_realloc/g_free or QEMU's qemu_memalign/qemu_blockalign/qemu_vfree |
| 354 | APIs. |
| 355 | |
| 356 | Please note that g_malloc will exit on allocation failure, so there |
| 357 | is no need to test for failure (as you would have to with malloc). |
| 358 | Calling g_malloc with a zero size is valid and will return NULL. |
| 359 | |
| 360 | Prefer g_new(T, n) instead of g_malloc(sizeof(T) ``*`` n) for the following |
| 361 | reasons: |
| 362 | |
| 363 | * It catches multiplication overflowing size_t; |
| 364 | * It returns T ``*`` instead of void ``*``, letting compiler catch more type errors. |
| 365 | |
| 366 | Declarations like |
| 367 | |
| 368 | .. code-block:: c |
| 369 | |
| 370 | T *v = g_malloc(sizeof(*v)) |
| 371 | |
| 372 | are acceptable, though. |
| 373 | |
| 374 | Memory allocated by qemu_memalign or qemu_blockalign must be freed with |
| 375 | qemu_vfree, since breaking this will cause problems on Win32. |
| 376 | |
| 377 | String manipulation |
| 378 | =================== |
| 379 | |
| 380 | Do not use the strncpy function. As mentioned in the man page, it does *not* |
| 381 | guarantee a NULL-terminated buffer, which makes it extremely dangerous to use. |
| 382 | It also zeros trailing destination bytes out to the specified length. Instead, |
| 383 | use this similar function when possible, but note its different signature: |
| 384 | |
| 385 | .. code-block:: c |
| 386 | |
| 387 | void pstrcpy(char *dest, int dest_buf_size, const char *src) |
| 388 | |
| 389 | Don't use strcat because it can't check for buffer overflows, but: |
| 390 | |
| 391 | .. code-block:: c |
| 392 | |
| 393 | char *pstrcat(char *buf, int buf_size, const char *s) |
| 394 | |
| 395 | The same limitation exists with sprintf and vsprintf, so use snprintf and |
| 396 | vsnprintf. |
| 397 | |
| 398 | QEMU provides other useful string functions: |
| 399 | |
| 400 | .. code-block:: c |
| 401 | |
| 402 | int strstart(const char *str, const char *val, const char **ptr) |
| 403 | int stristart(const char *str, const char *val, const char **ptr) |
| 404 | int qemu_strnlen(const char *s, int max_len) |
| 405 | |
| 406 | There are also replacement character processing macros for isxyz and toxyz, |
| 407 | so instead of e.g. isalnum you should use qemu_isalnum. |
| 408 | |
| 409 | Because of the memory management rules, you must use g_strdup/g_strndup |
| 410 | instead of plain strdup/strndup. |
| 411 | |
| 412 | Printf-style functions |
| 413 | ====================== |
| 414 | |
| 415 | Whenever you add a new printf-style function, i.e., one with a format |
| 416 | string argument and following "..." in its prototype, be sure to use |
| 417 | gcc's printf attribute directive in the prototype. |
| 418 | |
| 419 | This makes it so gcc's -Wformat and -Wformat-security options can do |
| 420 | their jobs and cross-check format strings with the number and types |
| 421 | of arguments. |
| 422 | |
| 423 | C standard, implementation defined and undefined behaviors |
| 424 | ========================================================== |
| 425 | |
| 426 | C code in QEMU should be written to the C99 language specification. A copy |
| 427 | of the final version of the C99 standard with corrigenda TC1, TC2, and TC3 |
| 428 | included, formatted as a draft, can be downloaded from: |
| 429 | |
| 430 | `<http://www.open-std.org/jtc1/sc22/WG14/www/docs/n1256.pdf>`_ |
| 431 | |
| 432 | The C language specification defines regions of undefined behavior and |
| 433 | implementation defined behavior (to give compiler authors enough leeway to |
| 434 | produce better code). In general, code in QEMU should follow the language |
| 435 | specification and avoid both undefined and implementation defined |
| 436 | constructs. ("It works fine on the gcc I tested it with" is not a valid |
| 437 | argument...) However there are a few areas where we allow ourselves to |
| 438 | assume certain behaviors because in practice all the platforms we care about |
| 439 | behave in the same way and writing strictly conformant code would be |
| 440 | painful. These are: |
| 441 | |
| 442 | * you may assume that integers are 2s complement representation |
| 443 | * you may assume that right shift of a signed integer duplicates |
| 444 | the sign bit (ie it is an arithmetic shift, not a logical shift) |
| 445 | |
| 446 | In addition, QEMU assumes that the compiler does not use the latitude |
| 447 | given in C99 and C11 to treat aspects of signed '<<' as undefined, as |
| 448 | documented in the GNU Compiler Collection manual starting at version 4.0. |
| 449 | |
Daniel P. Berrangé | 821f296 | 2019-08-23 17:31:35 +0100 | [diff] [blame] | 450 | Automatic memory deallocation |
| 451 | ============================= |
| 452 | |
| 453 | QEMU has a mandatory dependency either the GCC or CLang compiler. As |
| 454 | such it has the freedom to make use of a C language extension for |
| 455 | automatically running a cleanup function when a stack variable goes |
| 456 | out of scope. This can be used to simplify function cleanup paths, |
| 457 | often allowing many goto jumps to be eliminated, through automatic |
| 458 | free'ing of memory. |
| 459 | |
| 460 | The GLib2 library provides a number of functions/macros for enabling |
| 461 | automatic cleanup: |
| 462 | |
| 463 | `<https://developer.gnome.org/glib/stable/glib-Miscellaneous-Macros.html>`_ |
| 464 | |
| 465 | Most notably: |
| 466 | |
| 467 | * g_autofree - will invoke g_free() on the variable going out of scope |
| 468 | |
| 469 | * g_autoptr - for structs / objects, will invoke the cleanup func created |
| 470 | by a previous use of G_DEFINE_AUTOPTR_CLEANUP_FUNC. This is |
| 471 | supported for most GLib data types and GObjects |
| 472 | |
| 473 | For example, instead of |
| 474 | |
| 475 | .. code-block:: c |
| 476 | |
| 477 | int somefunc(void) { |
| 478 | int ret = -1; |
| 479 | char *foo = g_strdup_printf("foo%", "wibble"); |
| 480 | GList *bar = ..... |
| 481 | |
| 482 | if (eek) { |
| 483 | goto cleanup; |
| 484 | } |
| 485 | |
| 486 | ret = 0; |
| 487 | |
| 488 | cleanup: |
| 489 | g_free(foo); |
| 490 | g_list_free(bar); |
| 491 | return ret; |
| 492 | } |
| 493 | |
| 494 | Using g_autofree/g_autoptr enables the code to be written as: |
| 495 | |
| 496 | .. code-block:: c |
| 497 | |
| 498 | int somefunc(void) { |
| 499 | g_autofree char *foo = g_strdup_printf("foo%", "wibble"); |
| 500 | g_autoptr (GList) bar = ..... |
| 501 | |
| 502 | if (eek) { |
| 503 | return -1; |
| 504 | } |
| 505 | |
| 506 | return 0; |
| 507 | } |
| 508 | |
| 509 | While this generally results in simpler, less leak-prone code, there |
| 510 | are still some caveats to beware of |
| 511 | |
| 512 | * Variables declared with g_auto* MUST always be initialized, |
| 513 | otherwise the cleanup function will use uninitialized stack memory |
| 514 | |
| 515 | * If a variable declared with g_auto* holds a value which must |
| 516 | live beyond the life of the function, that value must be saved |
| 517 | and the original variable NULL'd out. This can be simpler using |
| 518 | g_steal_pointer |
| 519 | |
| 520 | |
| 521 | .. code-block:: c |
| 522 | |
| 523 | char *somefunc(void) { |
| 524 | g_autofree char *foo = g_strdup_printf("foo%", "wibble"); |
| 525 | g_autoptr (GList) bar = ..... |
| 526 | |
| 527 | if (eek) { |
| 528 | return NULL; |
| 529 | } |
| 530 | |
| 531 | return g_steal_pointer(&foo); |
| 532 | } |
| 533 | |
| 534 | |
Daniel P. Berrangé | 9f8efa7 | 2019-08-29 17:04:32 +0100 | [diff] [blame] | 535 | QEMU Specific Idioms |
| 536 | ******************** |
| 537 | |
Daniel P. Berrangé | 637f395 | 2019-08-23 17:12:05 +0100 | [diff] [blame] | 538 | Error handling and reporting |
| 539 | ============================ |
| 540 | |
| 541 | Reporting errors to the human user |
| 542 | ---------------------------------- |
| 543 | |
| 544 | Do not use printf(), fprintf() or monitor_printf(). Instead, use |
| 545 | error_report() or error_vreport() from error-report.h. This ensures the |
| 546 | error is reported in the right place (current monitor or stderr), and in |
| 547 | a uniform format. |
| 548 | |
| 549 | Use error_printf() & friends to print additional information. |
| 550 | |
| 551 | error_report() prints the current location. In certain common cases |
| 552 | like command line parsing, the current location is tracked |
| 553 | automatically. To manipulate it manually, use the loc_``*``() from |
| 554 | error-report.h. |
| 555 | |
| 556 | Propagating errors |
| 557 | ------------------ |
| 558 | |
| 559 | An error can't always be reported to the user right where it's detected, |
| 560 | but often needs to be propagated up the call chain to a place that can |
| 561 | handle it. This can be done in various ways. |
| 562 | |
| 563 | The most flexible one is Error objects. See error.h for usage |
| 564 | information. |
| 565 | |
| 566 | Use the simplest suitable method to communicate success / failure to |
| 567 | callers. Stick to common methods: non-negative on success / -1 on |
| 568 | error, non-negative / -errno, non-null / null, or Error objects. |
| 569 | |
| 570 | Example: when a function returns a non-null pointer on success, and it |
| 571 | can fail only in one way (as far as the caller is concerned), returning |
| 572 | null on failure is just fine, and certainly simpler and a lot easier on |
| 573 | the eyes than propagating an Error object through an Error ``*````*`` parameter. |
| 574 | |
| 575 | Example: when a function's callers need to report details on failure |
| 576 | only the function really knows, use Error ``*````*``, and set suitable errors. |
| 577 | |
| 578 | Do not report an error to the user when you're also returning an error |
| 579 | for somebody else to handle. Leave the reporting to the place that |
| 580 | consumes the error returned. |
| 581 | |
| 582 | Handling errors |
| 583 | --------------- |
| 584 | |
| 585 | Calling exit() is fine when handling configuration errors during |
| 586 | startup. It's problematic during normal operation. In particular, |
| 587 | monitor commands should never exit(). |
| 588 | |
| 589 | Do not call exit() or abort() to handle an error that can be triggered |
| 590 | by the guest (e.g., some unimplemented corner case in guest code |
| 591 | translation or device emulation). Guests should not be able to |
| 592 | terminate QEMU. |
| 593 | |
| 594 | Note that &error_fatal is just another way to exit(1), and &error_abort |
| 595 | is just another way to abort(). |
| 596 | |
| 597 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 598 | trace-events style |
| 599 | ================== |
Vladimir Sementsov-Ogievskiy | 44c6d63 | 2017-07-31 19:01:32 +0300 | [diff] [blame] | 600 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 601 | 0x prefix |
| 602 | --------- |
Vladimir Sementsov-Ogievskiy | 44c6d63 | 2017-07-31 19:01:32 +0300 | [diff] [blame] | 603 | |
| 604 | In trace-events files, use a '0x' prefix to specify hex numbers, as in: |
| 605 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 606 | .. code-block:: |
| 607 | |
| 608 | some_trace(unsigned x, uint64_t y) "x 0x%x y 0x" PRIx64 |
Vladimir Sementsov-Ogievskiy | 44c6d63 | 2017-07-31 19:01:32 +0300 | [diff] [blame] | 609 | |
| 610 | An exception is made for groups of numbers that are hexadecimal by |
| 611 | convention and separated by the symbols '.', '/', ':', or ' ' (such as |
| 612 | PCI bus id): |
| 613 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 614 | .. code-block:: |
| 615 | |
| 616 | another_trace(int cssid, int ssid, int dev_num) "bus id: %x.%x.%04x" |
Vladimir Sementsov-Ogievskiy | 44c6d63 | 2017-07-31 19:01:32 +0300 | [diff] [blame] | 617 | |
| 618 | However, you can use '0x' for such groups if you want. Anyway, be sure that |
| 619 | it is obvious that numbers are in hex, ex.: |
| 620 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 621 | .. code-block:: |
| 622 | |
| 623 | data_dump(uint8_t c1, uint8_t c2, uint8_t c3) "bytes (in hex): %02x %02x %02x" |
Vladimir Sementsov-Ogievskiy | 44c6d63 | 2017-07-31 19:01:32 +0300 | [diff] [blame] | 624 | |
| 625 | Rationale: hex numbers are hard to read in logs when there is no 0x prefix, |
| 626 | especially when (occasionally) the representation doesn't contain any letters |
| 627 | and especially in one line with other decimal numbers. Number groups are allowed |
| 628 | to not use '0x' because for some things notations like %x.%x.%x are used not |
| 629 | only in Qemu. Also dumping raw data bytes with '0x' is less readable. |
| 630 | |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 631 | '#' printf flag |
| 632 | --------------- |
Vladimir Sementsov-Ogievskiy | 44c6d63 | 2017-07-31 19:01:32 +0300 | [diff] [blame] | 633 | |
| 634 | Do not use printf flag '#', like '%#x'. |
| 635 | |
| 636 | Rationale: there are two ways to add a '0x' prefix to printed number: '0x%...' |
| 637 | and '%#...'. For consistency the only one way should be used. Arguments for |
| 638 | '0x%' are: |
Daniel P. Berrangé | 336a745 | 2019-08-23 17:09:24 +0100 | [diff] [blame] | 639 | |
| 640 | * it is more popular |
| 641 | * '%#' omits the 0x for the value 0 which makes output inconsistent |