1#!/usr/bin/env python3 2# SPDX-License-Identifier: GPL-2.0 3# Copyright(c) 2025: Mauro Carvalho Chehab <mchehab@kernel.org>. 4# 5# pylint: disable=C0301,C0302,R0904,R0912,R0913,R0914,R0915,R0917,R1702 6 7""" 8kdoc_parser 9=========== 10 11Read a C language source or header FILE and extract embedded 12documentation comments 13""" 14 15import re 16from pprint import pformat 17 18from kdoc_re import NestedMatch, KernRe 19from kdoc_item import KdocItem 20 21# 22# Regular expressions used to parse kernel-doc markups at KernelDoc class. 23# 24# Let's declare them in lowercase outside any class to make easier to 25# convert from the python script. 26# 27# As those are evaluated at the beginning, no need to cache them 28# 29 30# Allow whitespace at end of comment start. 31doc_start = KernRe(r'^/\*\*\s*$', cache=False) 32 33doc_end = KernRe(r'\*/', cache=False) 34doc_com = KernRe(r'\s*\*\s*', cache=False) 35doc_com_body = KernRe(r'\s*\* ?', cache=False) 36doc_decl = doc_com + KernRe(r'(\w+)', cache=False) 37 38# @params and a strictly limited set of supported section names 39# Specifically: 40# Match @word: 41# @...: 42# @{section-name}: 43# while trying to not match literal block starts like "example::" 44# 45doc_sect = doc_com + \ 46 KernRe(r'\s*(\@[.\w]+|\@\.\.\.|description|context|returns?|notes?|examples?)\s*:([^:].*)?$', 47 flags=re.I, cache=False) 48 49doc_content = doc_com_body + KernRe(r'(.*)', cache=False) 50doc_inline_start = KernRe(r'^\s*/\*\*\s*$', cache=False) 51doc_inline_sect = KernRe(r'\s*\*\s*(@\s*[\w][\w\.]*\s*):(.*)', cache=False) 52doc_inline_end = KernRe(r'^\s*\*/\s*$', cache=False) 53doc_inline_oneline = KernRe(r'^\s*/\*\*\s*(@[\w\s]+):\s*(.*)\s*\*/\s*$', cache=False) 54attribute = KernRe(r"__attribute__\s*\(\([a-z0-9,_\*\s\(\)]*\)\)", 55 flags=re.I | re.S, cache=False) 56 57export_symbol = KernRe(r'^\s*EXPORT_SYMBOL(_GPL)?\s*\(\s*(\w+)\s*\)\s*', cache=False) 58export_symbol_ns = KernRe(r'^\s*EXPORT_SYMBOL_NS(_GPL)?\s*\(\s*(\w+)\s*,\s*"\S+"\)\s*', cache=False) 59 60type_param = KernRe(r"\@(\w*((\.\w+)|(->\w+))*(\.\.\.)?)", cache=False) 61 62# 63# Tests for the beginning of a kerneldoc block in its various forms. 64# 65doc_block = doc_com + KernRe(r'DOC:\s*(.*)?', cache=False) 66doc_begin_data = KernRe(r"^\s*\*?\s*(struct|union|enum|typedef)\b\s*(\w*)", cache = False) 67doc_begin_func = KernRe(str(doc_com) + # initial " * ' 68 r"(?:\w+\s*\*\s*)?" + # type (not captured) 69 r'(?:define\s+)?' + # possible "define" (not captured) 70 r'(\w+)\s*(?:\(\w*\))?\s*' + # name and optional "(...)" 71 r'(?:[-:].*)?$', # description (not captured) 72 cache = False) 73 74# 75# A little helper to get rid of excess white space 76# 77multi_space = KernRe(r'\s\s+') 78def trim_whitespace(s): 79 return multi_space.sub(' ', s.strip()) 80 81class state: 82 """ 83 State machine enums 84 """ 85 86 # Parser states 87 NORMAL = 0 # normal code 88 NAME = 1 # looking for function name 89 DECLARATION = 2 # We have seen a declaration which might not be done 90 BODY = 3 # the body of the comment 91 SPECIAL_SECTION = 4 # doc section ending with a blank line 92 PROTO = 5 # scanning prototype 93 DOCBLOCK = 6 # documentation block 94 INLINE_NAME = 7 # gathering doc outside main block 95 INLINE_TEXT = 8 # reading the body of inline docs 96 97 name = [ 98 "NORMAL", 99 "NAME", 100 "DECLARATION", 101 "BODY", 102 "SPECIAL_SECTION", 103 "PROTO", 104 "DOCBLOCK", 105 "INLINE_NAME", 106 "INLINE_TEXT", 107 ] 108 109 110SECTION_DEFAULT = "Description" # default section 111 112class KernelEntry: 113 114 def __init__(self, config, ln): 115 self.config = config 116 117 self._contents = [] 118 self.sectcheck = "" 119 self.prototype = "" 120 121 self.warnings = [] 122 123 self.parameterlist = [] 124 self.parameterdescs = {} 125 self.parametertypes = {} 126 self.parameterdesc_start_lines = {} 127 128 self.section_start_lines = {} 129 self.sections = {} 130 131 self.anon_struct_union = False 132 133 self.leading_space = None 134 135 # State flags 136 self.brcount = 0 137 self.declaration_start_line = ln + 1 138 139 # 140 # Management of section contents 141 # 142 def add_text(self, text): 143 self._contents.append(text) 144 145 def contents(self): 146 return '\n'.join(self._contents) + '\n' 147 148 # TODO: rename to emit_message after removal of kernel-doc.pl 149 def emit_msg(self, log_msg, warning=True): 150 """Emit a message""" 151 152 if not warning: 153 self.config.log.info(log_msg) 154 return 155 156 # Delegate warning output to output logic, as this way it 157 # will report warnings/info only for symbols that are output 158 159 self.warnings.append(log_msg) 160 return 161 162 # 163 # Begin a new section. 164 # 165 def begin_section(self, line_no, title = SECTION_DEFAULT, dump = False): 166 if dump: 167 self.dump_section(start_new = True) 168 self.section = title 169 self.new_start_line = line_no 170 171 def dump_section(self, start_new=True): 172 """ 173 Dumps section contents to arrays/hashes intended for that purpose. 174 """ 175 # 176 # If we have accumulated no contents in the default ("description") 177 # section, don't bother. 178 # 179 if self.section == SECTION_DEFAULT and not self._contents: 180 return 181 name = self.section 182 contents = self.contents() 183 184 if type_param.match(name): 185 name = type_param.group(1) 186 187 self.parameterdescs[name] = contents 188 self.parameterdesc_start_lines[name] = self.new_start_line 189 190 self.sectcheck += name + " " 191 self.new_start_line = 0 192 193 else: 194 if name in self.sections and self.sections[name] != "": 195 # Only warn on user-specified duplicate section names 196 if name != SECTION_DEFAULT: 197 self.emit_msg(self.new_start_line, 198 f"duplicate section name '{name}'\n") 199 # Treat as a new paragraph - add a blank line 200 self.sections[name] += '\n' + contents 201 else: 202 self.sections[name] = contents 203 self.section_start_lines[name] = self.new_start_line 204 self.new_start_line = 0 205 206# self.config.log.debug("Section: %s : %s", name, pformat(vars(self))) 207 208 if start_new: 209 self.section = SECTION_DEFAULT 210 self._contents = [] 211 212 213class KernelDoc: 214 """ 215 Read a C language source or header FILE and extract embedded 216 documentation comments. 217 """ 218 219 # Section names 220 221 section_context = "Context" 222 section_return = "Return" 223 224 undescribed = "-- undescribed --" 225 226 def __init__(self, config, fname): 227 """Initialize internal variables""" 228 229 self.fname = fname 230 self.config = config 231 232 # Initial state for the state machines 233 self.state = state.NORMAL 234 235 # Store entry currently being processed 236 self.entry = None 237 238 # Place all potential outputs into an array 239 self.entries = [] 240 241 def emit_msg(self, ln, msg, warning=True): 242 """Emit a message""" 243 244 log_msg = f"{self.fname}:{ln} {msg}" 245 246 if self.entry: 247 self.entry.emit_msg(log_msg, warning) 248 return 249 250 if warning: 251 self.config.log.warning(log_msg) 252 else: 253 self.config.log.info(log_msg) 254 255 def dump_section(self, start_new=True): 256 """ 257 Dumps section contents to arrays/hashes intended for that purpose. 258 """ 259 260 if self.entry: 261 self.entry.dump_section(start_new) 262 263 # TODO: rename it to store_declaration after removal of kernel-doc.pl 264 def output_declaration(self, dtype, name, **args): 265 """ 266 Stores the entry into an entry array. 267 268 The actual output and output filters will be handled elsewhere 269 """ 270 271 item = KdocItem(name, dtype, self.entry.declaration_start_line, **args) 272 item.warnings = self.entry.warnings 273 274 # Drop empty sections 275 # TODO: improve empty sections logic to emit warnings 276 sections = self.entry.sections 277 for section in ["Description", "Return"]: 278 if section in sections and not sections[section].rstrip(): 279 del sections[section] 280 item.set_sections(sections, self.entry.section_start_lines) 281 item.set_params(self.entry.parameterlist, self.entry.parameterdescs, 282 self.entry.parametertypes, 283 self.entry.parameterdesc_start_lines) 284 self.entries.append(item) 285 286 self.config.log.debug("Output: %s:%s = %s", dtype, name, pformat(args)) 287 288 def reset_state(self, ln): 289 """ 290 Ancillary routine to create a new entry. It initializes all 291 variables used by the state machine. 292 """ 293 294 self.entry = KernelEntry(self.config, ln) 295 296 # State flags 297 self.state = state.NORMAL 298 299 def push_parameter(self, ln, decl_type, param, dtype, 300 org_arg, declaration_name): 301 """ 302 Store parameters and their descriptions at self.entry. 303 """ 304 305 if self.entry.anon_struct_union and dtype == "" and param == "}": 306 return # Ignore the ending }; from anonymous struct/union 307 308 self.entry.anon_struct_union = False 309 310 param = KernRe(r'[\[\)].*').sub('', param, count=1) 311 312 if dtype == "" and param.endswith("..."): 313 if KernRe(r'\w\.\.\.$').search(param): 314 # For named variable parameters of the form `x...`, 315 # remove the dots 316 param = param[:-3] 317 else: 318 # Handles unnamed variable parameters 319 param = "..." 320 321 if param not in self.entry.parameterdescs or \ 322 not self.entry.parameterdescs[param]: 323 324 self.entry.parameterdescs[param] = "variable arguments" 325 326 elif dtype == "" and (not param or param == "void"): 327 param = "void" 328 self.entry.parameterdescs[param] = "no arguments" 329 330 elif dtype == "" and param in ["struct", "union"]: 331 # Handle unnamed (anonymous) union or struct 332 dtype = param 333 param = "{unnamed_" + param + "}" 334 self.entry.parameterdescs[param] = "anonymous\n" 335 self.entry.anon_struct_union = True 336 337 # Handle cache group enforcing variables: they do not need 338 # to be described in header files 339 elif "__cacheline_group" in param: 340 # Ignore __cacheline_group_begin and __cacheline_group_end 341 return 342 343 # Warn if parameter has no description 344 # (but ignore ones starting with # as these are not parameters 345 # but inline preprocessor statements) 346 if param not in self.entry.parameterdescs and not param.startswith("#"): 347 self.entry.parameterdescs[param] = self.undescribed 348 349 if "." not in param: 350 if decl_type == 'function': 351 dname = f"{decl_type} parameter" 352 else: 353 dname = f"{decl_type} member" 354 355 self.emit_msg(ln, 356 f"{dname} '{param}' not described in '{declaration_name}'") 357 358 # Strip spaces from param so that it is one continuous string on 359 # parameterlist. This fixes a problem where check_sections() 360 # cannot find a parameter like "addr[6 + 2]" because it actually 361 # appears as "addr[6", "+", "2]" on the parameter list. 362 # However, it's better to maintain the param string unchanged for 363 # output, so just weaken the string compare in check_sections() 364 # to ignore "[blah" in a parameter string. 365 366 self.entry.parameterlist.append(param) 367 org_arg = KernRe(r'\s\s+').sub(' ', org_arg) 368 self.entry.parametertypes[param] = org_arg 369 370 371 def create_parameter_list(self, ln, decl_type, args, 372 splitter, declaration_name): 373 """ 374 Creates a list of parameters, storing them at self.entry. 375 """ 376 377 # temporarily replace all commas inside function pointer definition 378 arg_expr = KernRe(r'(\([^\),]+),') 379 while arg_expr.search(args): 380 args = arg_expr.sub(r"\1#", args) 381 382 for arg in args.split(splitter): 383 # Strip comments 384 arg = KernRe(r'\/\*.*\*\/').sub('', arg) 385 386 # Ignore argument attributes 387 arg = KernRe(r'\sPOS0?\s').sub(' ', arg) 388 389 # Strip leading/trailing spaces 390 arg = arg.strip() 391 arg = KernRe(r'\s+').sub(' ', arg, count=1) 392 393 if arg.startswith('#'): 394 # Treat preprocessor directive as a typeless variable just to fill 395 # corresponding data structures "correctly". Catch it later in 396 # output_* subs. 397 398 # Treat preprocessor directive as a typeless variable 399 self.push_parameter(ln, decl_type, arg, "", 400 "", declaration_name) 401 402 elif KernRe(r'\(.+\)\s*\(').search(arg): 403 # Pointer-to-function 404 405 arg = arg.replace('#', ',') 406 407 r = KernRe(r'[^\(]+\(\*?\s*([\w\[\]\.]*)\s*\)') 408 if r.match(arg): 409 param = r.group(1) 410 else: 411 self.emit_msg(ln, f"Invalid param: {arg}") 412 param = arg 413 414 dtype = KernRe(r'([^\(]+\(\*?)\s*' + re.escape(param)).sub(r'\1', arg) 415 self.push_parameter(ln, decl_type, param, dtype, 416 arg, declaration_name) 417 418 elif KernRe(r'\(.+\)\s*\[').search(arg): 419 # Array-of-pointers 420 421 arg = arg.replace('#', ',') 422 r = KernRe(r'[^\(]+\(\s*\*\s*([\w\[\]\.]*?)\s*(\s*\[\s*[\w]+\s*\]\s*)*\)') 423 if r.match(arg): 424 param = r.group(1) 425 else: 426 self.emit_msg(ln, f"Invalid param: {arg}") 427 param = arg 428 429 dtype = KernRe(r'([^\(]+\(\*?)\s*' + re.escape(param)).sub(r'\1', arg) 430 431 self.push_parameter(ln, decl_type, param, dtype, 432 arg, declaration_name) 433 434 elif arg: 435 arg = KernRe(r'\s*:\s*').sub(":", arg) 436 arg = KernRe(r'\s*\[').sub('[', arg) 437 438 args = KernRe(r'\s*,\s*').split(arg) 439 if args[0] and '*' in args[0]: 440 args[0] = re.sub(r'(\*+)\s*', r' \1', args[0]) 441 442 first_arg = [] 443 r = KernRe(r'^(.*\s+)(.*?\[.*\].*)$') 444 if args[0] and r.match(args[0]): 445 args.pop(0) 446 first_arg.extend(r.group(1)) 447 first_arg.append(r.group(2)) 448 else: 449 first_arg = KernRe(r'\s+').split(args.pop(0)) 450 451 args.insert(0, first_arg.pop()) 452 dtype = ' '.join(first_arg) 453 454 for param in args: 455 if KernRe(r'^(\*+)\s*(.*)').match(param): 456 r = KernRe(r'^(\*+)\s*(.*)') 457 if not r.match(param): 458 self.emit_msg(ln, f"Invalid param: {param}") 459 continue 460 461 param = r.group(1) 462 463 self.push_parameter(ln, decl_type, r.group(2), 464 f"{dtype} {r.group(1)}", 465 arg, declaration_name) 466 467 elif KernRe(r'(.*?):(\w+)').search(param): 468 r = KernRe(r'(.*?):(\w+)') 469 if not r.match(param): 470 self.emit_msg(ln, f"Invalid param: {param}") 471 continue 472 473 if dtype != "": # Skip unnamed bit-fields 474 self.push_parameter(ln, decl_type, r.group(1), 475 f"{dtype}:{r.group(2)}", 476 arg, declaration_name) 477 else: 478 self.push_parameter(ln, decl_type, param, dtype, 479 arg, declaration_name) 480 481 def check_sections(self, ln, decl_name, decl_type, sectcheck): 482 """ 483 Check for errors inside sections, emitting warnings if not found 484 parameters are described. 485 """ 486 487 sects = sectcheck.split() 488 489 for sx in range(len(sects)): # pylint: disable=C0200 490 err = True 491 for param in self.entry.parameterlist: 492 if param == sects[sx]: 493 err = False 494 break 495 496 if err: 497 if decl_type == 'function': 498 dname = f"{decl_type} parameter" 499 else: 500 dname = f"{decl_type} member" 501 502 self.emit_msg(ln, 503 f"Excess {dname} '{sects[sx]}' description in '{decl_name}'") 504 505 def check_return_section(self, ln, declaration_name, return_type): 506 """ 507 If the function doesn't return void, warns about the lack of a 508 return description. 509 """ 510 511 if not self.config.wreturn: 512 return 513 514 # Ignore an empty return type (It's a macro) 515 # Ignore functions with a "void" return type (but not "void *") 516 if not return_type or KernRe(r'void\s*\w*\s*$').search(return_type): 517 return 518 519 if not self.entry.sections.get("Return", None): 520 self.emit_msg(ln, 521 f"No description found for return value of '{declaration_name}'") 522 523 def dump_struct(self, ln, proto): 524 """ 525 Store an entry for an struct or union 526 """ 527 528 type_pattern = r'(struct|union)' 529 530 qualifiers = [ 531 "__attribute__", 532 "__packed", 533 "__aligned", 534 "____cacheline_aligned_in_smp", 535 "____cacheline_aligned", 536 ] 537 538 definition_body = r'\{(.*)\}\s*' + "(?:" + '|'.join(qualifiers) + ")?" 539 struct_members = KernRe(type_pattern + r'([^\{\};]+)(\{)([^\{\}]*)(\})([^\{\}\;]*)(\;)') 540 541 # Extract struct/union definition 542 members = None 543 declaration_name = None 544 decl_type = None 545 546 r = KernRe(type_pattern + r'\s+(\w+)\s*' + definition_body) 547 if r.search(proto): 548 decl_type = r.group(1) 549 declaration_name = r.group(2) 550 members = r.group(3) 551 else: 552 r = KernRe(r'typedef\s+' + type_pattern + r'\s*' + definition_body + r'\s*(\w+)\s*;') 553 554 if r.search(proto): 555 decl_type = r.group(1) 556 declaration_name = r.group(3) 557 members = r.group(2) 558 559 if not members: 560 self.emit_msg(ln, f"{proto} error: Cannot parse struct or union!") 561 return 562 563 if self.entry.identifier != declaration_name: 564 self.emit_msg(ln, 565 f"expecting prototype for {decl_type} {self.entry.identifier}. Prototype was for {decl_type} {declaration_name} instead\n") 566 return 567 568 args_pattern = r'([^,)]+)' 569 570 sub_prefixes = [ 571 (KernRe(r'\/\*\s*private:.*?\/\*\s*public:.*?\*\/', re.S | re.I), ''), 572 (KernRe(r'\/\*\s*private:.*', re.S | re.I), ''), 573 574 # Strip comments 575 (KernRe(r'\/\*.*?\*\/', re.S), ''), 576 577 # Strip attributes 578 (attribute, ' '), 579 (KernRe(r'\s*__aligned\s*\([^;]*\)', re.S), ' '), 580 (KernRe(r'\s*__counted_by\s*\([^;]*\)', re.S), ' '), 581 (KernRe(r'\s*__counted_by_(le|be)\s*\([^;]*\)', re.S), ' '), 582 (KernRe(r'\s*__packed\s*', re.S), ' '), 583 (KernRe(r'\s*CRYPTO_MINALIGN_ATTR', re.S), ' '), 584 (KernRe(r'\s*____cacheline_aligned_in_smp', re.S), ' '), 585 (KernRe(r'\s*____cacheline_aligned', re.S), ' '), 586 587 # Unwrap struct_group macros based on this definition: 588 # __struct_group(TAG, NAME, ATTRS, MEMBERS...) 589 # which has variants like: struct_group(NAME, MEMBERS...) 590 # Only MEMBERS arguments require documentation. 591 # 592 # Parsing them happens on two steps: 593 # 594 # 1. drop struct group arguments that aren't at MEMBERS, 595 # storing them as STRUCT_GROUP(MEMBERS) 596 # 597 # 2. remove STRUCT_GROUP() ancillary macro. 598 # 599 # The original logic used to remove STRUCT_GROUP() using an 600 # advanced regex: 601 # 602 # \bSTRUCT_GROUP(\(((?:(?>[^)(]+)|(?1))*)\))[^;]*; 603 # 604 # with two patterns that are incompatible with 605 # Python re module, as it has: 606 # 607 # - a recursive pattern: (?1) 608 # - an atomic grouping: (?>...) 609 # 610 # I tried a simpler version: but it didn't work either: 611 # \bSTRUCT_GROUP\(([^\)]+)\)[^;]*; 612 # 613 # As it doesn't properly match the end parenthesis on some cases. 614 # 615 # So, a better solution was crafted: there's now a NestedMatch 616 # class that ensures that delimiters after a search are properly 617 # matched. So, the implementation to drop STRUCT_GROUP() will be 618 # handled in separate. 619 620 (KernRe(r'\bstruct_group\s*\(([^,]*,)', re.S), r'STRUCT_GROUP('), 621 (KernRe(r'\bstruct_group_attr\s*\(([^,]*,){2}', re.S), r'STRUCT_GROUP('), 622 (KernRe(r'\bstruct_group_tagged\s*\(([^,]*),([^,]*),', re.S), r'struct \1 \2; STRUCT_GROUP('), 623 (KernRe(r'\b__struct_group\s*\(([^,]*,){3}', re.S), r'STRUCT_GROUP('), 624 625 # Replace macros 626 # 627 # TODO: use NestedMatch for FOO($1, $2, ...) matches 628 # 629 # it is better to also move those to the NestedMatch logic, 630 # to ensure that parenthesis will be properly matched. 631 632 (KernRe(r'__ETHTOOL_DECLARE_LINK_MODE_MASK\s*\(([^\)]+)\)', re.S), r'DECLARE_BITMAP(\1, __ETHTOOL_LINK_MODE_MASK_NBITS)'), 633 (KernRe(r'DECLARE_PHY_INTERFACE_MASK\s*\(([^\)]+)\)', re.S), r'DECLARE_BITMAP(\1, PHY_INTERFACE_MODE_MAX)'), 634 (KernRe(r'DECLARE_BITMAP\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'unsigned long \1[BITS_TO_LONGS(\2)]'), 635 (KernRe(r'DECLARE_HASHTABLE\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'unsigned long \1[1 << ((\2) - 1)]'), 636 (KernRe(r'DECLARE_KFIFO\s*\(' + args_pattern + r',\s*' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\2 *\1'), 637 (KernRe(r'DECLARE_KFIFO_PTR\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\2 *\1'), 638 (KernRe(r'(?:__)?DECLARE_FLEX_ARRAY\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\1 \2[]'), 639 (KernRe(r'DEFINE_DMA_UNMAP_ADDR\s*\(' + args_pattern + r'\)', re.S), r'dma_addr_t \1'), 640 (KernRe(r'DEFINE_DMA_UNMAP_LEN\s*\(' + args_pattern + r'\)', re.S), r'__u32 \1'), 641 ] 642 643 # Regexes here are guaranteed to have the end limiter matching 644 # the start delimiter. Yet, right now, only one replace group 645 # is allowed. 646 647 sub_nested_prefixes = [ 648 (re.compile(r'\bSTRUCT_GROUP\('), r'\1'), 649 ] 650 651 for search, sub in sub_prefixes: 652 members = search.sub(sub, members) 653 654 nested = NestedMatch() 655 656 for search, sub in sub_nested_prefixes: 657 members = nested.sub(search, sub, members) 658 659 # Keeps the original declaration as-is 660 declaration = members 661 662 # Split nested struct/union elements 663 # 664 # This loop was simpler at the original kernel-doc perl version, as 665 # while ($members =~ m/$struct_members/) { ... } 666 # reads 'members' string on each interaction. 667 # 668 # Python behavior is different: it parses 'members' only once, 669 # creating a list of tuples from the first interaction. 670 # 671 # On other words, this won't get nested structs. 672 # 673 # So, we need to have an extra loop on Python to override such 674 # re limitation. 675 676 while True: 677 tuples = struct_members.findall(members) 678 if not tuples: 679 break 680 681 for t in tuples: 682 newmember = "" 683 maintype = t[0] 684 s_ids = t[5] 685 content = t[3] 686 687 oldmember = "".join(t) 688 689 for s_id in s_ids.split(','): 690 s_id = s_id.strip() 691 692 newmember += f"{maintype} {s_id}; " 693 s_id = KernRe(r'[:\[].*').sub('', s_id) 694 s_id = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', s_id) 695 696 for arg in content.split(';'): 697 arg = arg.strip() 698 699 if not arg: 700 continue 701 702 r = KernRe(r'^([^\(]+\(\*?\s*)([\w\.]*)(\s*\).*)') 703 if r.match(arg): 704 # Pointer-to-function 705 dtype = r.group(1) 706 name = r.group(2) 707 extra = r.group(3) 708 709 if not name: 710 continue 711 712 if not s_id: 713 # Anonymous struct/union 714 newmember += f"{dtype}{name}{extra}; " 715 else: 716 newmember += f"{dtype}{s_id}.{name}{extra}; " 717 718 else: 719 arg = arg.strip() 720 # Handle bitmaps 721 arg = KernRe(r':\s*\d+\s*').sub('', arg) 722 723 # Handle arrays 724 arg = KernRe(r'\[.*\]').sub('', arg) 725 726 # Handle multiple IDs 727 arg = KernRe(r'\s*,\s*').sub(',', arg) 728 729 r = KernRe(r'(.*)\s+([\S+,]+)') 730 731 if r.search(arg): 732 dtype = r.group(1) 733 names = r.group(2) 734 else: 735 newmember += f"{arg}; " 736 continue 737 738 for name in names.split(','): 739 name = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', name).strip() 740 741 if not name: 742 continue 743 744 if not s_id: 745 # Anonymous struct/union 746 newmember += f"{dtype} {name}; " 747 else: 748 newmember += f"{dtype} {s_id}.{name}; " 749 750 members = members.replace(oldmember, newmember) 751 752 # Ignore other nested elements, like enums 753 members = re.sub(r'(\{[^\{\}]*\})', '', members) 754 755 self.create_parameter_list(ln, decl_type, members, ';', 756 declaration_name) 757 self.check_sections(ln, declaration_name, decl_type, self.entry.sectcheck) 758 759 # Adjust declaration for better display 760 declaration = KernRe(r'([\{;])').sub(r'\1\n', declaration) 761 declaration = KernRe(r'\}\s+;').sub('};', declaration) 762 763 # Better handle inlined enums 764 while True: 765 r = KernRe(r'(enum\s+\{[^\}]+),([^\n])') 766 if not r.search(declaration): 767 break 768 769 declaration = r.sub(r'\1,\n\2', declaration) 770 771 def_args = declaration.split('\n') 772 level = 1 773 declaration = "" 774 for clause in def_args: 775 776 clause = clause.strip() 777 clause = KernRe(r'\s+').sub(' ', clause, count=1) 778 779 if not clause: 780 continue 781 782 if '}' in clause and level > 1: 783 level -= 1 784 785 if not KernRe(r'^\s*#').match(clause): 786 declaration += "\t" * level 787 788 declaration += "\t" + clause + "\n" 789 if "{" in clause and "}" not in clause: 790 level += 1 791 792 self.output_declaration(decl_type, declaration_name, 793 struct=declaration_name, 794 definition=declaration, 795 purpose=self.entry.declaration_purpose) 796 797 def dump_enum(self, ln, proto): 798 """ 799 Stores an enum inside self.entries array. 800 """ 801 802 # Ignore members marked private 803 proto = KernRe(r'\/\*\s*private:.*?\/\*\s*public:.*?\*\/', flags=re.S).sub('', proto) 804 proto = KernRe(r'\/\*\s*private:.*}', flags=re.S).sub('}', proto) 805 806 # Strip comments 807 proto = KernRe(r'\/\*.*?\*\/', flags=re.S).sub('', proto) 808 809 # Strip #define macros inside enums 810 proto = KernRe(r'#\s*((define|ifdef|if)\s+|endif)[^;]*;', flags=re.S).sub('', proto) 811 812 # 813 # Parse out the name and members of the enum. Typedef form first. 814 # 815 r = KernRe(r'typedef\s+enum\s*\{(.*)\}\s*(\w*)\s*;') 816 if r.search(proto): 817 declaration_name = r.group(2) 818 members = r.group(1).rstrip() 819 # 820 # Failing that, look for a straight enum 821 # 822 else: 823 r = KernRe(r'enum\s+(\w*)\s*\{(.*)\}') 824 if r.match(proto): 825 declaration_name = r.group(1) 826 members = r.group(2).rstrip() 827 # 828 # OK, this isn't going to work. 829 # 830 else: 831 self.emit_msg(ln, f"{proto}: error: Cannot parse enum!") 832 return 833 # 834 # Make sure we found what we were expecting. 835 # 836 if self.entry.identifier != declaration_name: 837 if self.entry.identifier == "": 838 self.emit_msg(ln, 839 f"{proto}: wrong kernel-doc identifier on prototype") 840 else: 841 self.emit_msg(ln, 842 f"expecting prototype for enum {self.entry.identifier}. " 843 f"Prototype was for enum {declaration_name} instead") 844 return 845 846 if not declaration_name: 847 declaration_name = "(anonymous)" 848 # 849 # Parse out the name of each enum member, and verify that we 850 # have a description for it. 851 # 852 member_set = set() 853 members = KernRe(r'\([^;)]*\)').sub('', members) 854 for arg in members.split(','): 855 if not arg: 856 continue 857 arg = KernRe(r'^\s*(\w+).*').sub(r'\1', arg) 858 self.entry.parameterlist.append(arg) 859 if arg not in self.entry.parameterdescs: 860 self.entry.parameterdescs[arg] = self.undescribed 861 self.emit_msg(ln, 862 f"Enum value '{arg}' not described in enum '{declaration_name}'") 863 member_set.add(arg) 864 # 865 # Ensure that every described member actually exists in the enum. 866 # 867 for k in self.entry.parameterdescs: 868 if k not in member_set: 869 self.emit_msg(ln, 870 f"Excess enum value '%{k}' description in '{declaration_name}'") 871 872 self.output_declaration('enum', declaration_name, 873 enum=declaration_name, 874 purpose=self.entry.declaration_purpose) 875 876 def dump_declaration(self, ln, prototype): 877 """ 878 Stores a data declaration inside self.entries array. 879 """ 880 881 if self.entry.decl_type == "enum": 882 self.dump_enum(ln, prototype) 883 return 884 885 if self.entry.decl_type == "typedef": 886 self.dump_typedef(ln, prototype) 887 return 888 889 if self.entry.decl_type in ["union", "struct"]: 890 self.dump_struct(ln, prototype) 891 return 892 893 self.output_declaration(self.entry.decl_type, prototype, 894 entry=self.entry) 895 896 def dump_function(self, ln, prototype): 897 """ 898 Stores a function of function macro inside self.entries array. 899 """ 900 901 func_macro = False 902 return_type = '' 903 decl_type = 'function' 904 905 # Prefixes that would be removed 906 sub_prefixes = [ 907 (r"^static +", "", 0), 908 (r"^extern +", "", 0), 909 (r"^asmlinkage +", "", 0), 910 (r"^inline +", "", 0), 911 (r"^__inline__ +", "", 0), 912 (r"^__inline +", "", 0), 913 (r"^__always_inline +", "", 0), 914 (r"^noinline +", "", 0), 915 (r"^__FORTIFY_INLINE +", "", 0), 916 (r"__init +", "", 0), 917 (r"__init_or_module +", "", 0), 918 (r"__deprecated +", "", 0), 919 (r"__flatten +", "", 0), 920 (r"__meminit +", "", 0), 921 (r"__must_check +", "", 0), 922 (r"__weak +", "", 0), 923 (r"__sched +", "", 0), 924 (r"_noprof", "", 0), 925 (r"__printf\s*\(\s*\d*\s*,\s*\d*\s*\) +", "", 0), 926 (r"__(?:re)?alloc_size\s*\(\s*\d+\s*(?:,\s*\d+\s*)?\) +", "", 0), 927 (r"__diagnose_as\s*\(\s*\S+\s*(?:,\s*\d+\s*)*\) +", "", 0), 928 (r"DECL_BUCKET_PARAMS\s*\(\s*(\S+)\s*,\s*(\S+)\s*\)", r"\1, \2", 0), 929 (r"__attribute_const__ +", "", 0), 930 931 # It seems that Python support for re.X is broken: 932 # At least for me (Python 3.13), this didn't work 933# (r""" 934# __attribute__\s*\(\( 935# (?: 936# [\w\s]+ # attribute name 937# (?:\([^)]*\))? # attribute arguments 938# \s*,? # optional comma at the end 939# )+ 940# \)\)\s+ 941# """, "", re.X), 942 943 # So, remove whitespaces and comments from it 944 (r"__attribute__\s*\(\((?:[\w\s]+(?:\([^)]*\))?\s*,?)+\)\)\s+", "", 0), 945 ] 946 947 for search, sub, flags in sub_prefixes: 948 prototype = KernRe(search, flags).sub(sub, prototype) 949 950 # Macros are a special case, as they change the prototype format 951 new_proto = KernRe(r"^#\s*define\s+").sub("", prototype) 952 if new_proto != prototype: 953 is_define_proto = True 954 prototype = new_proto 955 else: 956 is_define_proto = False 957 958 # Yes, this truly is vile. We are looking for: 959 # 1. Return type (may be nothing if we're looking at a macro) 960 # 2. Function name 961 # 3. Function parameters. 962 # 963 # All the while we have to watch out for function pointer parameters 964 # (which IIRC is what the two sections are for), C types (these 965 # regexps don't even start to express all the possibilities), and 966 # so on. 967 # 968 # If you mess with these regexps, it's a good idea to check that 969 # the following functions' documentation still comes out right: 970 # - parport_register_device (function pointer parameters) 971 # - atomic_set (macro) 972 # - pci_match_device, __copy_to_user (long return type) 973 974 name = r'[a-zA-Z0-9_~:]+' 975 prototype_end1 = r'[^\(]*' 976 prototype_end2 = r'[^\{]*' 977 prototype_end = fr'\(({prototype_end1}|{prototype_end2})\)' 978 979 # Besides compiling, Perl qr{[\w\s]+} works as a non-capturing group. 980 # So, this needs to be mapped in Python with (?:...)? or (?:...)+ 981 982 type1 = r'(?:[\w\s]+)?' 983 type2 = r'(?:[\w\s]+\*+)+' 984 985 found = False 986 987 if is_define_proto: 988 r = KernRe(r'^()(' + name + r')\s+') 989 990 if r.search(prototype): 991 return_type = '' 992 declaration_name = r.group(2) 993 func_macro = True 994 995 found = True 996 997 if not found: 998 patterns = [ 999 rf'^()({name})\s*{prototype_end}', 1000 rf'^({type1})\s+({name})\s*{prototype_end}', 1001 rf'^({type2})\s*({name})\s*{prototype_end}', 1002 ] 1003 1004 for p in patterns: 1005 r = KernRe(p) 1006 1007 if r.match(prototype): 1008 1009 return_type = r.group(1) 1010 declaration_name = r.group(2) 1011 args = r.group(3) 1012 1013 self.create_parameter_list(ln, decl_type, args, ',', 1014 declaration_name) 1015 1016 found = True 1017 break 1018 if not found: 1019 self.emit_msg(ln, 1020 f"cannot understand function prototype: '{prototype}'") 1021 return 1022 1023 if self.entry.identifier != declaration_name: 1024 self.emit_msg(ln, 1025 f"expecting prototype for {self.entry.identifier}(). Prototype was for {declaration_name}() instead") 1026 return 1027 1028 self.check_sections(ln, declaration_name, "function", self.entry.sectcheck) 1029 1030 self.check_return_section(ln, declaration_name, return_type) 1031 1032 if 'typedef' in return_type: 1033 self.output_declaration(decl_type, declaration_name, 1034 function=declaration_name, 1035 typedef=True, 1036 functiontype=return_type, 1037 purpose=self.entry.declaration_purpose, 1038 func_macro=func_macro) 1039 else: 1040 self.output_declaration(decl_type, declaration_name, 1041 function=declaration_name, 1042 typedef=False, 1043 functiontype=return_type, 1044 purpose=self.entry.declaration_purpose, 1045 func_macro=func_macro) 1046 1047 def dump_typedef(self, ln, proto): 1048 """ 1049 Stores a typedef inside self.entries array. 1050 """ 1051 1052 typedef_type = r'((?:\s+[\w\*]+\b){0,7}\s+(?:\w+\b|\*+))\s*' 1053 typedef_ident = r'\*?\s*(\w\S+)\s*' 1054 typedef_args = r'\s*\((.*)\);' 1055 1056 typedef1 = KernRe(r'typedef' + typedef_type + r'\(' + typedef_ident + r'\)' + typedef_args) 1057 typedef2 = KernRe(r'typedef' + typedef_type + typedef_ident + typedef_args) 1058 1059 # Strip comments 1060 proto = KernRe(r'/\*.*?\*/', flags=re.S).sub('', proto) 1061 1062 # Parse function typedef prototypes 1063 for r in [typedef1, typedef2]: 1064 if not r.match(proto): 1065 continue 1066 1067 return_type = r.group(1).strip() 1068 declaration_name = r.group(2) 1069 args = r.group(3) 1070 1071 if self.entry.identifier != declaration_name: 1072 self.emit_msg(ln, 1073 f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n") 1074 return 1075 1076 decl_type = 'function' 1077 self.create_parameter_list(ln, decl_type, args, ',', declaration_name) 1078 1079 self.output_declaration(decl_type, declaration_name, 1080 function=declaration_name, 1081 typedef=True, 1082 functiontype=return_type, 1083 purpose=self.entry.declaration_purpose) 1084 return 1085 1086 # Handle nested parentheses or brackets 1087 r = KernRe(r'(\(*.\)\s*|\[*.\]\s*);$') 1088 while r.search(proto): 1089 proto = r.sub('', proto) 1090 1091 # Parse simple typedefs 1092 r = KernRe(r'typedef.*\s+(\w+)\s*;') 1093 if r.match(proto): 1094 declaration_name = r.group(1) 1095 1096 if self.entry.identifier != declaration_name: 1097 self.emit_msg(ln, 1098 f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n") 1099 return 1100 1101 self.output_declaration('typedef', declaration_name, 1102 typedef=declaration_name, 1103 purpose=self.entry.declaration_purpose) 1104 return 1105 1106 self.emit_msg(ln, "error: Cannot parse typedef!") 1107 1108 @staticmethod 1109 def process_export(function_set, line): 1110 """ 1111 process EXPORT_SYMBOL* tags 1112 1113 This method doesn't use any variable from the class, so declare it 1114 with a staticmethod decorator. 1115 """ 1116 1117 # We support documenting some exported symbols with different 1118 # names. A horrible hack. 1119 suffixes = [ '_noprof' ] 1120 1121 # Note: it accepts only one EXPORT_SYMBOL* per line, as having 1122 # multiple export lines would violate Kernel coding style. 1123 1124 if export_symbol.search(line): 1125 symbol = export_symbol.group(2) 1126 elif export_symbol_ns.search(line): 1127 symbol = export_symbol_ns.group(2) 1128 else: 1129 return False 1130 # 1131 # Found an export, trim out any special suffixes 1132 # 1133 for suffix in suffixes: 1134 symbol = symbol.removesuffix(suffix) 1135 function_set.add(symbol) 1136 return True 1137 1138 def process_normal(self, ln, line): 1139 """ 1140 STATE_NORMAL: looking for the /** to begin everything. 1141 """ 1142 1143 if not doc_start.match(line): 1144 return 1145 1146 # start a new entry 1147 self.reset_state(ln) 1148 1149 # next line is always the function name 1150 self.state = state.NAME 1151 1152 def process_name(self, ln, line): 1153 """ 1154 STATE_NAME: Looking for the "name - description" line 1155 """ 1156 # 1157 # Check for a DOC: block and handle them specially. 1158 # 1159 if doc_block.search(line): 1160 1161 if not doc_block.group(1): 1162 self.entry.begin_section(ln, "Introduction") 1163 else: 1164 self.entry.begin_section(ln, doc_block.group(1)) 1165 1166 self.entry.identifier = self.entry.section 1167 self.state = state.DOCBLOCK 1168 # 1169 # Otherwise we're looking for a normal kerneldoc declaration line. 1170 # 1171 elif doc_decl.search(line): 1172 self.entry.identifier = doc_decl.group(1) 1173 1174 # Test for data declaration 1175 if doc_begin_data.search(line): 1176 self.entry.decl_type = doc_begin_data.group(1) 1177 self.entry.identifier = doc_begin_data.group(2) 1178 # 1179 # Look for a function description 1180 # 1181 elif doc_begin_func.search(line): 1182 self.entry.identifier = doc_begin_func.group(1) 1183 self.entry.decl_type = "function" 1184 # 1185 # We struck out. 1186 # 1187 else: 1188 self.emit_msg(ln, 1189 f"This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst\n{line}") 1190 self.state = state.NORMAL 1191 return 1192 # 1193 # OK, set up for a new kerneldoc entry. 1194 # 1195 self.state = state.BODY 1196 self.entry.identifier = self.entry.identifier.strip(" ") 1197 # if there's no @param blocks need to set up default section here 1198 self.entry.begin_section(ln + 1) 1199 # 1200 # Find the description portion, which *should* be there but 1201 # isn't always. 1202 # (We should be able to capture this from the previous parsing - someday) 1203 # 1204 r = KernRe("[-:](.*)") 1205 if r.search(line): 1206 self.entry.declaration_purpose = trim_whitespace(r.group(1)) 1207 self.state = state.DECLARATION 1208 else: 1209 self.entry.declaration_purpose = "" 1210 1211 if not self.entry.declaration_purpose and self.config.wshort_desc: 1212 self.emit_msg(ln, 1213 f"missing initial short description on line:\n{line}") 1214 1215 if not self.entry.identifier and self.entry.decl_type != "enum": 1216 self.emit_msg(ln, 1217 f"wrong kernel-doc identifier on line:\n{line}") 1218 self.state = state.NORMAL 1219 1220 if self.config.verbose: 1221 self.emit_msg(ln, 1222 f"Scanning doc for {self.entry.decl_type} {self.entry.identifier}", 1223 warning=False) 1224 # 1225 # Failed to find an identifier. Emit a warning 1226 # 1227 else: 1228 self.emit_msg(ln, f"Cannot find identifier on line:\n{line}") 1229 1230 # 1231 # Helper function to determine if a new section is being started. 1232 # 1233 def is_new_section(self, ln, line): 1234 if doc_sect.search(line): 1235 self.state = state.BODY 1236 # 1237 # Pick out the name of our new section, tweaking it if need be. 1238 # 1239 newsection = doc_sect.group(1) 1240 if newsection.lower() == 'description': 1241 newsection = 'Description' 1242 elif newsection.lower() == 'context': 1243 newsection = 'Context' 1244 self.state = state.SPECIAL_SECTION 1245 elif newsection.lower() in ["@return", "@returns", 1246 "return", "returns"]: 1247 newsection = "Return" 1248 self.state = state.SPECIAL_SECTION 1249 elif newsection[0] == '@': 1250 self.state = state.SPECIAL_SECTION 1251 # 1252 # Initialize the contents, and get the new section going. 1253 # 1254 newcontents = doc_sect.group(2) 1255 if not newcontents: 1256 newcontents = "" 1257 self.dump_section() 1258 self.entry.begin_section(ln, newsection) 1259 self.entry.leading_space = None 1260 1261 self.entry.add_text(newcontents.lstrip()) 1262 return True 1263 return False 1264 1265 # 1266 # Helper function to detect (and effect) the end of a kerneldoc comment. 1267 # 1268 def is_comment_end(self, ln, line): 1269 if doc_end.search(line): 1270 self.dump_section() 1271 1272 # Look for doc_com + <text> + doc_end: 1273 r = KernRe(r'\s*\*\s*[a-zA-Z_0-9:\.]+\*/') 1274 if r.match(line): 1275 self.emit_msg(ln, f"suspicious ending line: {line}") 1276 1277 self.entry.prototype = "" 1278 self.entry.new_start_line = ln + 1 1279 1280 self.state = state.PROTO 1281 return True 1282 return False 1283 1284 1285 def process_decl(self, ln, line): 1286 """ 1287 STATE_DECLARATION: We've seen the beginning of a declaration 1288 """ 1289 if self.is_new_section(ln, line) or self.is_comment_end(ln, line): 1290 return 1291 # 1292 # Look for anything with the " * " line beginning. 1293 # 1294 if doc_content.search(line): 1295 cont = doc_content.group(1) 1296 # 1297 # A blank line means that we have moved out of the declaration 1298 # part of the comment (without any "special section" parameter 1299 # descriptions). 1300 # 1301 if cont == "": 1302 self.state = state.BODY 1303 # 1304 # Otherwise we have more of the declaration section to soak up. 1305 # 1306 else: 1307 self.entry.declaration_purpose = \ 1308 trim_whitespace(self.entry.declaration_purpose + ' ' + cont) 1309 else: 1310 # Unknown line, ignore 1311 self.emit_msg(ln, f"bad line: {line}") 1312 1313 1314 def process_special(self, ln, line): 1315 """ 1316 STATE_SPECIAL_SECTION: a section ending with a blank line 1317 """ 1318 # 1319 # If we have hit a blank line (only the " * " marker), then this 1320 # section is done. 1321 # 1322 if KernRe(r"\s*\*\s*$").match(line): 1323 self.entry.begin_section(ln, dump = True) 1324 self.state = state.BODY 1325 return 1326 # 1327 # Not a blank line, look for the other ways to end the section. 1328 # 1329 if self.is_new_section(ln, line) or self.is_comment_end(ln, line): 1330 return 1331 # 1332 # OK, we should have a continuation of the text for this section. 1333 # 1334 if doc_content.search(line): 1335 cont = doc_content.group(1) 1336 # 1337 # If the lines of text after the first in a special section have 1338 # leading white space, we need to trim it out or Sphinx will get 1339 # confused. For the second line (the None case), see what we 1340 # find there and remember it. 1341 # 1342 if self.entry.leading_space is None: 1343 r = KernRe(r'^(\s+)') 1344 if r.match(cont): 1345 self.entry.leading_space = len(r.group(1)) 1346 else: 1347 self.entry.leading_space = 0 1348 # 1349 # Otherwise, before trimming any leading chars, be *sure* 1350 # that they are white space. We should maybe warn if this 1351 # isn't the case. 1352 # 1353 for i in range(0, self.entry.leading_space): 1354 if cont[i] != " ": 1355 self.entry.leading_space = i 1356 break 1357 # 1358 # Add the trimmed result to the section and we're done. 1359 # 1360 self.entry.add_text(cont[self.entry.leading_space:]) 1361 else: 1362 # Unknown line, ignore 1363 self.emit_msg(ln, f"bad line: {line}") 1364 1365 def process_body(self, ln, line): 1366 """ 1367 STATE_BODY: the bulk of a kerneldoc comment. 1368 """ 1369 if self.is_new_section(ln, line) or self.is_comment_end(ln, line): 1370 return 1371 1372 if doc_content.search(line): 1373 cont = doc_content.group(1) 1374 self.entry.add_text(cont) 1375 else: 1376 # Unknown line, ignore 1377 self.emit_msg(ln, f"bad line: {line}") 1378 1379 def process_inline_name(self, ln, line): 1380 """STATE_INLINE_NAME: beginning of docbook comments within a prototype.""" 1381 1382 if doc_inline_sect.search(line): 1383 self.entry.begin_section(ln, doc_inline_sect.group(1)) 1384 self.entry.add_text(doc_inline_sect.group(2).lstrip()) 1385 self.state = state.INLINE_TEXT 1386 elif doc_inline_end.search(line): 1387 self.dump_section() 1388 self.state = state.PROTO 1389 elif doc_content.search(line): 1390 self.emit_msg(ln, f"Incorrect use of kernel-doc format: {line}") 1391 self.state = state.PROTO 1392 # else ... ?? 1393 1394 def process_inline_text(self, ln, line): 1395 """STATE_INLINE_TEXT: docbook comments within a prototype.""" 1396 1397 if doc_inline_end.search(line): 1398 self.dump_section() 1399 self.state = state.PROTO 1400 elif doc_content.search(line): 1401 self.entry.add_text(doc_content.group(1)) 1402 # else ... ?? 1403 1404 def syscall_munge(self, ln, proto): # pylint: disable=W0613 1405 """ 1406 Handle syscall definitions 1407 """ 1408 1409 is_void = False 1410 1411 # Strip newlines/CR's 1412 proto = re.sub(r'[\r\n]+', ' ', proto) 1413 1414 # Check if it's a SYSCALL_DEFINE0 1415 if 'SYSCALL_DEFINE0' in proto: 1416 is_void = True 1417 1418 # Replace SYSCALL_DEFINE with correct return type & function name 1419 proto = KernRe(r'SYSCALL_DEFINE.*\(').sub('long sys_', proto) 1420 1421 r = KernRe(r'long\s+(sys_.*?),') 1422 if r.search(proto): 1423 proto = KernRe(',').sub('(', proto, count=1) 1424 elif is_void: 1425 proto = KernRe(r'\)').sub('(void)', proto, count=1) 1426 1427 # Now delete all of the odd-numbered commas in the proto 1428 # so that argument types & names don't have a comma between them 1429 count = 0 1430 length = len(proto) 1431 1432 if is_void: 1433 length = 0 # skip the loop if is_void 1434 1435 for ix in range(length): 1436 if proto[ix] == ',': 1437 count += 1 1438 if count % 2 == 1: 1439 proto = proto[:ix] + ' ' + proto[ix + 1:] 1440 1441 return proto 1442 1443 def tracepoint_munge(self, ln, proto): 1444 """ 1445 Handle tracepoint definitions 1446 """ 1447 1448 tracepointname = None 1449 tracepointargs = None 1450 1451 # Match tracepoint name based on different patterns 1452 r = KernRe(r'TRACE_EVENT\((.*?),') 1453 if r.search(proto): 1454 tracepointname = r.group(1) 1455 1456 r = KernRe(r'DEFINE_SINGLE_EVENT\((.*?),') 1457 if r.search(proto): 1458 tracepointname = r.group(1) 1459 1460 r = KernRe(r'DEFINE_EVENT\((.*?),(.*?),') 1461 if r.search(proto): 1462 tracepointname = r.group(2) 1463 1464 if tracepointname: 1465 tracepointname = tracepointname.lstrip() 1466 1467 r = KernRe(r'TP_PROTO\((.*?)\)') 1468 if r.search(proto): 1469 tracepointargs = r.group(1) 1470 1471 if not tracepointname or not tracepointargs: 1472 self.emit_msg(ln, 1473 f"Unrecognized tracepoint format:\n{proto}\n") 1474 else: 1475 proto = f"static inline void trace_{tracepointname}({tracepointargs})" 1476 self.entry.identifier = f"trace_{self.entry.identifier}" 1477 1478 return proto 1479 1480 def process_proto_function(self, ln, line): 1481 """Ancillary routine to process a function prototype""" 1482 1483 # strip C99-style comments to end of line 1484 line = KernRe(r"\/\/.*$", re.S).sub('', line) 1485 # 1486 # Soak up the line's worth of prototype text, stopping at { or ; if present. 1487 # 1488 if KernRe(r'\s*#\s*define').match(line): 1489 self.entry.prototype = line 1490 elif not line.startswith('#'): # skip other preprocessor stuff 1491 r = KernRe(r'([^\{]*)') 1492 if r.match(line): 1493 self.entry.prototype += r.group(1) + " " 1494 # 1495 # If we now have the whole prototype, clean it up and declare victory. 1496 # 1497 if '{' in line or ';' in line or KernRe(r'\s*#\s*define').match(line): 1498 # strip comments and surrounding spaces 1499 self.entry.prototype = KernRe(r'/\*.*\*/').sub('', self.entry.prototype).strip() 1500 # 1501 # Handle self.entry.prototypes for function pointers like: 1502 # int (*pcs_config)(struct foo) 1503 # by turning it into 1504 # int pcs_config(struct foo) 1505 # 1506 r = KernRe(r'^(\S+\s+)\(\s*\*(\S+)\)') 1507 self.entry.prototype = r.sub(r'\1\2', self.entry.prototype) 1508 # 1509 # Handle special declaration syntaxes 1510 # 1511 if 'SYSCALL_DEFINE' in self.entry.prototype: 1512 self.entry.prototype = self.syscall_munge(ln, 1513 self.entry.prototype) 1514 else: 1515 r = KernRe(r'TRACE_EVENT|DEFINE_EVENT|DEFINE_SINGLE_EVENT') 1516 if r.search(self.entry.prototype): 1517 self.entry.prototype = self.tracepoint_munge(ln, 1518 self.entry.prototype) 1519 # 1520 # ... and we're done 1521 # 1522 self.dump_function(ln, self.entry.prototype) 1523 self.reset_state(ln) 1524 1525 def process_proto_type(self, ln, line): 1526 """Ancillary routine to process a type""" 1527 1528 # Strip C99-style comments and surrounding whitespace 1529 line = KernRe(r"//.*$", re.S).sub('', line).strip() 1530 if not line: 1531 return # nothing to see here 1532 1533 # To distinguish preprocessor directive from regular declaration later. 1534 if line.startswith('#'): 1535 line += ";" 1536 # 1537 # Split the declaration on any of { } or ;, and accumulate pieces 1538 # until we hit a semicolon while not inside {brackets} 1539 # 1540 r = KernRe(r'(.*?)([{};])') 1541 for chunk in r.split(line): 1542 if chunk: # Ignore empty matches 1543 self.entry.prototype += chunk 1544 # 1545 # This cries out for a match statement ... someday after we can 1546 # drop Python 3.9 ... 1547 # 1548 if chunk == '{': 1549 self.entry.brcount += 1 1550 elif chunk == '}': 1551 self.entry.brcount -= 1 1552 elif chunk == ';' and self.entry.brcount <= 0: 1553 self.dump_declaration(ln, self.entry.prototype) 1554 self.reset_state(ln) 1555 return 1556 # 1557 # We hit the end of the line while still in the declaration; put 1558 # in a space to represent the newline. 1559 # 1560 self.entry.prototype += ' ' 1561 1562 def process_proto(self, ln, line): 1563 """STATE_PROTO: reading a function/whatever prototype.""" 1564 1565 if doc_inline_oneline.search(line): 1566 self.entry.begin_section(ln, doc_inline_oneline.group(1)) 1567 self.entry.add_text(doc_inline_oneline.group(2)) 1568 self.dump_section() 1569 1570 elif doc_inline_start.search(line): 1571 self.state = state.INLINE_NAME 1572 1573 elif self.entry.decl_type == 'function': 1574 self.process_proto_function(ln, line) 1575 1576 else: 1577 self.process_proto_type(ln, line) 1578 1579 def process_docblock(self, ln, line): 1580 """STATE_DOCBLOCK: within a DOC: block.""" 1581 1582 if doc_end.search(line): 1583 self.dump_section() 1584 self.output_declaration("doc", self.entry.identifier) 1585 self.reset_state(ln) 1586 1587 elif doc_content.search(line): 1588 self.entry.add_text(doc_content.group(1)) 1589 1590 def parse_export(self): 1591 """ 1592 Parses EXPORT_SYMBOL* macros from a single Kernel source file. 1593 """ 1594 1595 export_table = set() 1596 1597 try: 1598 with open(self.fname, "r", encoding="utf8", 1599 errors="backslashreplace") as fp: 1600 1601 for line in fp: 1602 self.process_export(export_table, line) 1603 1604 except IOError: 1605 return None 1606 1607 return export_table 1608 1609 # 1610 # The state/action table telling us which function to invoke in 1611 # each state. 1612 # 1613 state_actions = { 1614 state.NORMAL: process_normal, 1615 state.NAME: process_name, 1616 state.BODY: process_body, 1617 state.DECLARATION: process_decl, 1618 state.SPECIAL_SECTION: process_special, 1619 state.INLINE_NAME: process_inline_name, 1620 state.INLINE_TEXT: process_inline_text, 1621 state.PROTO: process_proto, 1622 state.DOCBLOCK: process_docblock, 1623 } 1624 1625 def parse_kdoc(self): 1626 """ 1627 Open and process each line of a C source file. 1628 The parsing is controlled via a state machine, and the line is passed 1629 to a different process function depending on the state. The process 1630 function may update the state as needed. 1631 1632 Besides parsing kernel-doc tags, it also parses export symbols. 1633 """ 1634 1635 prev = "" 1636 prev_ln = None 1637 export_table = set() 1638 1639 try: 1640 with open(self.fname, "r", encoding="utf8", 1641 errors="backslashreplace") as fp: 1642 for ln, line in enumerate(fp): 1643 1644 line = line.expandtabs().strip("\n") 1645 1646 # Group continuation lines on prototypes 1647 if self.state == state.PROTO: 1648 if line.endswith("\\"): 1649 prev += line.rstrip("\\") 1650 if not prev_ln: 1651 prev_ln = ln 1652 continue 1653 1654 if prev: 1655 ln = prev_ln 1656 line = prev + line 1657 prev = "" 1658 prev_ln = None 1659 1660 self.config.log.debug("%d %s: %s", 1661 ln, state.name[self.state], 1662 line) 1663 1664 # This is an optimization over the original script. 1665 # There, when export_file was used for the same file, 1666 # it was read twice. Here, we use the already-existing 1667 # loop to parse exported symbols as well. 1668 # 1669 if (self.state != state.NORMAL) or \ 1670 not self.process_export(export_table, line): 1671 # Hand this line to the appropriate state handler 1672 self.state_actions[self.state](self, ln, line) 1673 1674 except OSError: 1675 self.config.log.error(f"Error: Cannot open file {self.fname}") 1676 1677 return export_table, self.entries 1678