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""" 8Classes and functions related to reading a C language source or header FILE 9and extract embedded documentation comments from it. 10""" 11 12import sys 13import re 14import difflib 15from pprint import pformat 16 17from kdoc.c_lex import CTokenizer, tokenizer_set_log 18from kdoc.kdoc_re import KernRe 19from kdoc.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 it easier to 25# convert from the Perl 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*$', cache=False) 32 33doc_end = KernRe(r'\*/', cache=False) 34doc_com = KernRe(r'\s*\*\s*', cache=False) 35doc_com_body = KernRe(r'\s*\* ?', cache=False) 36doc_decl = doc_com + KernRe(r'(\w+)', cache=False) 37 38# @params and a strictly limited set of supported section names 39# Specifically: 40# Match @word: 41# @...: 42# @{section-name}: 43# while trying to not match literal block starts like "example::" 44# 45known_section_names = 'description|context|returns?|notes?|examples?' 46known_sections = KernRe(known_section_names, flags = re.I) 47doc_sect = doc_com + \ 48 KernRe(r'\s*(@[.\w]+|@\.\.\.|' + known_section_names + r')\s*:([^:].*)?$', 49 flags=re.I, cache=False) 50 51doc_content = doc_com_body + KernRe(r'(.*)', cache=False) 52doc_inline_start = KernRe(r'^\s*/\*\*\s*$', cache=False) 53doc_inline_sect = KernRe(r'\s*\*\s*(@\s*[\w][\w\.]*\s*):(.*)', cache=False) 54doc_inline_end = KernRe(r'^\s*\*/\s*$', cache=False) 55doc_inline_oneline = KernRe(r'^\s*/\*\*\s*(@\s*[\w][\w\.]*\s*):\s*(.*)\s*\*/\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|var)\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# Ancillary functions 76# 77 78multi_space = KernRe(r'\s\s+') 79def trim_whitespace(s): 80 """ 81 A little helper to get rid of excess white space. 82 """ 83 return multi_space.sub(' ', s.strip()) 84 85def trim_private_members(text): 86 """ 87 Remove ``struct``/``enum`` members that have been marked "private". 88 """ 89 90 tokens = CTokenizer(text) 91 return str(tokens) 92 93class state: 94 """ 95 States used by the parser's state machine. 96 """ 97 98 # Parser states 99 NORMAL = 0 #: Normal code. 100 NAME = 1 #: Looking for function name. 101 DECLARATION = 2 #: We have seen a declaration which might not be done. 102 BODY = 3 #: The body of the comment. 103 SPECIAL_SECTION = 4 #: Doc section ending with a blank line. 104 PROTO = 5 #: Scanning prototype. 105 DOCBLOCK = 6 #: Documentation block. 106 INLINE_NAME = 7 #: Gathering doc outside main block. 107 INLINE_TEXT = 8 #: Reading the body of inline docs. 108 109 #: Names for each parser state. 110 name = [ 111 "NORMAL", 112 "NAME", 113 "DECLARATION", 114 "BODY", 115 "SPECIAL_SECTION", 116 "PROTO", 117 "DOCBLOCK", 118 "INLINE_NAME", 119 "INLINE_TEXT", 120 ] 121 122 123SECTION_DEFAULT = "Description" #: Default section. 124 125class KernelEntry: 126 """ 127 Encapsulates a Kernel documentation entry. 128 """ 129 130 def __init__(self, config, fname, ln): 131 self.config = config 132 self.fname = fname 133 134 self._contents = [] 135 self.prototype = "" 136 137 self.warnings = [] 138 139 self.parameterlist = [] 140 self.parameterdescs = {} 141 self.parametertypes = {} 142 self.parameterdesc_start_lines = {} 143 144 self.sections_start_lines = {} 145 self.sections = {} 146 147 self.anon_struct_union = False 148 149 self.leading_space = None 150 151 self.fname = fname 152 153 # State flags 154 self.brcount = 0 155 self.declaration_start_line = ln + 1 156 157 # 158 # Management of section contents 159 # 160 def add_text(self, text): 161 """Add a new text to the entry contents list.""" 162 self._contents.append(text) 163 164 def contents(self): 165 """Returns a string with all content texts that were added.""" 166 return '\n'.join(self._contents) + '\n' 167 168 # TODO: rename to emit_message after removal of kernel-doc.pl 169 def emit_msg(self, ln, msg, *, warning=True): 170 """Emit a message.""" 171 172 log_msg = f"{self.fname}:{ln} {msg}" 173 174 if not warning: 175 self.config.log.info(log_msg) 176 return 177 178 # Delegate warning output to output logic, as this way it 179 # will report warnings/info only for symbols that are output 180 181 self.warnings.append(log_msg) 182 return 183 184 def begin_section(self, line_no, title = SECTION_DEFAULT, dump = False): 185 """ 186 Begin a new section. 187 """ 188 if dump: 189 self.dump_section(start_new = True) 190 self.section = title 191 self.new_start_line = line_no 192 193 def dump_section(self, start_new=True): 194 """ 195 Dumps section contents to arrays/hashes intended for that purpose. 196 """ 197 # 198 # If we have accumulated no contents in the default ("description") 199 # section, don't bother. 200 # 201 if self.section == SECTION_DEFAULT and not self._contents: 202 return 203 name = self.section 204 contents = self.contents() 205 206 if type_param.match(name): 207 name = type_param.group(1) 208 209 self.parameterdescs[name] = contents 210 self.parameterdesc_start_lines[name] = self.new_start_line 211 212 self.new_start_line = 0 213 214 else: 215 if name in self.sections and self.sections[name] != "": 216 # Only warn on user-specified duplicate section names 217 if name != SECTION_DEFAULT: 218 self.emit_msg(self.new_start_line, 219 f"duplicate section name '{name}'") 220 # Treat as a new paragraph - add a blank line 221 self.sections[name] += '\n' + contents 222 else: 223 self.sections[name] = contents 224 self.sections_start_lines[name] = self.new_start_line 225 self.new_start_line = 0 226 227# self.config.log.debug("Section: %s : %s", name, pformat(vars(self))) 228 229 if start_new: 230 self.section = SECTION_DEFAULT 231 self._contents = [] 232 233python_warning = False 234 235class KernelDoc: 236 """ 237 Read a C language source or header FILE and extract embedded 238 documentation comments. 239 """ 240 241 #: Name of context section. 242 section_context = "Context" 243 244 #: Name of return section. 245 section_return = "Return" 246 247 #: String to write when a parameter is not described. 248 undescribed = "-- undescribed --" 249 250 def __init__(self, config, fname, xforms, store_src=False): 251 """Initialize internal variables""" 252 253 self.fname = fname 254 self.config = config 255 self.xforms = xforms 256 self.store_src = store_src 257 258 tokenizer_set_log(self.config.log, f"{self.fname}: CMatch: ") 259 260 # Initial state for the state machines 261 self.state = state.NORMAL 262 263 # Store entry currently being processed 264 self.entry = None 265 266 # Place all potential outputs into an array 267 self.entries = [] 268 269 # 270 # We need Python 3.7 for its "dicts remember the insertion 271 # order" guarantee 272 # 273 global python_warning 274 if (not python_warning and 275 sys.version_info.major == 3 and sys.version_info.minor < 7): 276 277 self.emit_msg(0, 278 'Python 3.7 or later is required for correct results') 279 python_warning = True 280 281 def emit_msg(self, ln, msg, *, warning=True): 282 """Emit a message""" 283 284 if self.entry: 285 self.entry.emit_msg(ln, msg, warning=warning) 286 return 287 288 log_msg = f"{self.fname}:{ln} {msg}" 289 290 if warning: 291 self.config.log.warning(log_msg) 292 else: 293 self.config.log.info(log_msg) 294 295 def dump_section(self, start_new=True): 296 """ 297 Dump section contents to arrays/hashes intended for that purpose. 298 """ 299 300 if self.entry: 301 self.entry.dump_section(start_new) 302 303 # TODO: rename it to store_declaration after removal of kernel-doc.pl 304 def output_declaration(self, dtype, name, **args): 305 """ 306 Store the entry into an entry array. 307 308 The actual output and output filters will be handled elsewhere. 309 """ 310 311 item = KdocItem(name, self.fname, dtype, 312 self.entry.declaration_start_line, **args) 313 item.warnings = self.entry.warnings 314 315 # Drop empty sections 316 # TODO: improve empty sections logic to emit warnings 317 sections = self.entry.sections 318 for section in ["Description", "Return"]: 319 if section in sections and not sections[section].rstrip(): 320 del sections[section] 321 item.set_sections(sections, self.entry.sections_start_lines) 322 item.set_params(self.entry.parameterlist, self.entry.parameterdescs, 323 self.entry.parametertypes, 324 self.entry.parameterdesc_start_lines) 325 self.entries.append(item) 326 327 self.config.log.debug("Output: %s:%s = %s", dtype, name, pformat(args)) 328 329 def emit_unused_warnings(self): 330 """ 331 When the parser fails to produce a valid entry, it places some 332 warnings under `entry.warnings` that will be discarded when resetting 333 the state. 334 335 Ensure that those warnings are not lost. 336 337 .. note:: 338 339 Because we are calling `config.warning()` here, those 340 warnings are not filtered by the `-W` parameters: they will all 341 be produced even when `-Wreturn`, `-Wshort-desc`, and/or 342 `-Wcontents-before-sections` are used. 343 344 Allowing those warnings to be filtered is complex, because it 345 would require storing them in a buffer and then filtering them 346 during the output step of the code, depending on the 347 selected symbols. 348 """ 349 if self.entry and self.entry not in self.entries: 350 for log_msg in self.entry.warnings: 351 self.config.warning(log_msg) 352 353 def reset_state(self, ln): 354 """ 355 Ancillary routine to create a new entry. It initializes all 356 variables used by the state machine. 357 """ 358 359 self.emit_unused_warnings() 360 361 self.entry = KernelEntry(self.config, self.fname, ln) 362 363 # State flags 364 self.state = state.NORMAL 365 366 def push_parameter(self, ln, decl_type, param, dtype, 367 org_arg, declaration_name): 368 """ 369 Store parameters and their descriptions at self.entry. 370 """ 371 372 if self.entry.anon_struct_union and dtype == "" and param == "}": 373 return # Ignore the ending }; from anonymous struct/union 374 375 self.entry.anon_struct_union = False 376 377 param = KernRe(r'[\[\)].*').sub('', param, count=1) 378 379 # 380 # Look at various "anonymous type" cases. 381 # 382 if dtype == '': 383 if param.endswith("..."): 384 named_variadic = len(param) > 3 385 if named_variadic: # there is a name provided, use that 386 # 387 # If the user documented the parameter using the 388 # ``@name...:`` form, the description is stored in 389 # parameterdescs under the unstripped key. Migrate 390 # it to the stripped key so the user's text is not 391 # silently dropped during output, and so the new 392 # excess-parameter check in check_sections() does 393 # not flag the unstripped key as orphaned. 394 # 395 orig = self.entry.parameterdescs.pop(param, None) 396 param = param[:-3] 397 if orig is not None and \ 398 not self.entry.parameterdescs.get(param): 399 self.entry.parameterdescs[param] = orig 400 if not self.entry.parameterdescs.get(param): 401 # 402 # For a named variadic (e.g. ``args...``), emit the 403 # standard "not described" warning before auto-filling 404 # so a missing or mistyped ``@<name>:`` doc tag does 405 # not go undetected. The bare ``...`` form has no 406 # natural name for the user to document and so always 407 # gets the auto-generated text. 408 # 409 if named_variadic and decl_type == 'function': 410 self.emit_msg(ln, 411 f"function parameter '{param}' " 412 f"not described in " 413 f"'{declaration_name}'") 414 self.entry.parameterdescs[param] = "variable arguments" 415 416 elif (not param) or param == "void": 417 param = "void" 418 self.entry.parameterdescs[param] = "no arguments" 419 420 elif param in ["struct", "union"]: 421 # Handle unnamed (anonymous) union or struct 422 dtype = param 423 param = "{unnamed_" + param + "}" 424 self.entry.parameterdescs[param] = "anonymous\n" 425 self.entry.anon_struct_union = True 426 427 # Warn if parameter has no description 428 # (but ignore ones starting with # as these are not parameters 429 # but inline preprocessor statements) 430 if param not in self.entry.parameterdescs and not param.startswith("#"): 431 self.entry.parameterdescs[param] = self.undescribed 432 433 if "." not in param: 434 if decl_type == 'function': 435 dname = f"{decl_type} parameter" 436 else: 437 dname = f"{decl_type} member" 438 439 self.emit_msg(ln, 440 f"{dname} '{param}' not described in '{declaration_name}'") 441 442 # Strip spaces from param so that it is one continuous string on 443 # parameterlist. This fixes a problem where check_sections() 444 # cannot find a parameter like "addr[6 + 2]" because it actually 445 # appears as "addr[6", "+", "2]" on the parameter list. 446 # However, it's better to maintain the param string unchanged for 447 # output, so just weaken the string compare in check_sections() 448 # to ignore "[blah" in a parameter string. 449 450 self.entry.parameterlist.append(param) 451 org_arg = KernRe(r'\s\s+').sub(' ', org_arg) 452 self.entry.parametertypes[param] = org_arg 453 454 455 def create_parameter_list(self, ln, decl_type, args, 456 splitter, declaration_name): 457 """ 458 Creates a list of parameters, storing them at self.entry. 459 """ 460 461 # temporarily replace all commas inside function pointer definition 462 arg_expr = KernRe(r'(\([^\),]+),') 463 while arg_expr.search(args): 464 args = arg_expr.sub(r"\1#", args) 465 466 for arg in args.split(splitter): 467 # Ignore argument attributes 468 arg = KernRe(r'\sPOS0?\s').sub(' ', arg) 469 470 # Replace '[at_least ' with '[static '. This allows sphinx to parse 471 # array parameter declarations like 'char A[at_least 4]', where 472 # 'at_least' is #defined to 'static' by the kernel headers. 473 arg = arg.replace('[at_least ', '[static ') 474 475 # Strip leading/trailing spaces 476 arg = arg.strip() 477 arg = KernRe(r'\s+').sub(' ', arg, count=1) 478 479 if arg.startswith('#'): 480 # Treat preprocessor directive as a typeless variable just to fill 481 # corresponding data structures "correctly". Catch it later in 482 # output_* subs. 483 484 # Treat preprocessor directive as a typeless variable 485 self.push_parameter(ln, decl_type, arg, "", 486 "", declaration_name) 487 # 488 # The pointer-to-function case. 489 # 490 elif KernRe(r'\(.+\)\s*\(').search(arg): 491 arg = arg.replace('#', ',') 492 r = KernRe(r'[^\(]+\(\*?\s*' # Everything up to "(*" 493 r'([\w\[\].]*)' # Capture the name and possible [array] 494 r'\s*\)') # Make sure the trailing ")" is there 495 if r.match(arg): 496 param = r.group(1) 497 else: 498 self.emit_msg(ln, f"Invalid param: {arg}") 499 param = arg 500 dtype = arg.replace(param, '') 501 self.push_parameter(ln, decl_type, param, dtype, arg, declaration_name) 502 # 503 # The array-of-pointers case. Dig the parameter name out from the middle 504 # of the declaration. 505 # 506 elif KernRe(r'\(.+\)\s*\[').search(arg): 507 r = KernRe(r'[^\(]+\(\s*\*\s*' # Up to "(" and maybe "*" 508 r'([\w.]*?)' # The actual pointer name 509 r'\s*(\[\s*\w+\s*\]\s*)*\)') # The [array portion] 510 if r.match(arg): 511 param = r.group(1) 512 else: 513 self.emit_msg(ln, f"Invalid param: {arg}") 514 param = arg 515 dtype = arg.replace(param, '') 516 self.push_parameter(ln, decl_type, param, dtype, arg, declaration_name) 517 elif arg: 518 # 519 # Clean up extraneous spaces and split the string at commas; the first 520 # element of the resulting list will also include the type information. 521 # 522 arg = KernRe(r'\s*:\s*').sub(":", arg) 523 arg = KernRe(r'\s*\[').sub('[', arg) 524 args = KernRe(r'\s*,\s*').split(arg) 525 args[0] = re.sub(r'(\*+)\s*', r' \1', args[0]) 526 # 527 # args[0] has a string of "type a". If "a" includes an [array] 528 # declaration, we want to not be fooled by any white space inside 529 # the brackets, so detect and handle that case specially. 530 # 531 r = KernRe(r'^([^[\]]*\s+)(.*)$') 532 if r.match(args[0]): 533 args[0] = r.group(2) 534 dtype = r.group(1) 535 else: 536 # No space in args[0]; this seems wrong but preserves previous behavior 537 dtype = '' 538 539 bitfield_re = KernRe(r'(.*?):(\w+)') 540 for param in args: 541 # 542 # For pointers, shift the star(s) from the variable name to the 543 # type declaration. 544 # 545 r = KernRe(r'^(\*+)\s*(.*)') 546 if r.match(param): 547 self.push_parameter(ln, decl_type, r.group(2), 548 f"{dtype} {r.group(1)}", 549 arg, declaration_name) 550 # 551 # Perform a similar shift for bitfields. 552 # 553 elif bitfield_re.search(param): 554 if dtype != "": # Skip unnamed bit-fields 555 self.push_parameter(ln, decl_type, bitfield_re.group(1), 556 f"{dtype}:{bitfield_re.group(2)}", 557 arg, declaration_name) 558 else: 559 self.push_parameter(ln, decl_type, param, dtype, 560 arg, declaration_name) 561 562 def get_suggestions_hint(self, decl_name, possible_names): 563 # For decl name 'flags' or 'flgas', suggests 'substruct.flags' 564 submember_exact = [] 565 submember_substrings = [] 566 submember_suggestions = [] 567 for possible_name in possible_names: 568 parts = possible_name.strip().split('.') 569 if len(parts) < 2: 570 continue 571 572 final_part = parts[-1] 573 if decl_name == final_part: 574 submember_exact.append(possible_name) 575 elif decl_name in final_part: 576 submember_substrings.append(possible_name) 577 elif difflib.get_close_matches(decl_name, [final_part]): 578 submember_suggestions.append(possible_name) 579 580 # For decl name 'flgas', suggests 'flags' 581 full_suggestions = difflib.get_close_matches(decl_name, possible_names) 582 583 # For decl name 'member', suggests 'longer_member' 584 full_substrings = [name for name in possible_names if decl_name in name] 585 586 ordered_lists = [ 587 submember_exact, 588 submember_substrings, 589 submember_suggestions, 590 full_suggestions, 591 full_substrings, 592 ] 593 594 # Deduplicate but maintain order from most to least likely: 595 unique_suggestions = {} 596 for suggestion_list in ordered_lists: 597 for suggestion in suggestion_list: 598 unique_suggestions[suggestion] = None 599 600 suggestions = list(unique_suggestions.keys()) 601 if not suggestions: 602 return "" 603 604 joined_suggestions = "', '".join(suggestions) 605 return f"(did you mean one of: '{joined_suggestions}')" 606 607 def check_sections(self, ln, decl_name, decl_type): 608 """ 609 Check for errors inside sections, emitting warnings if not found 610 parameters are described. 611 """ 612 for section in self.entry.sections: 613 if section not in self.entry.parameterlist and \ 614 not known_sections.search(section): 615 hint = self.get_suggestions_hint(section, self.entry.parameterlist) 616 if decl_type == 'function': 617 dname = f"{decl_type} parameter" 618 else: 619 dname = f"{decl_type} member" 620 self.emit_msg(ln, 621 f"Excess {dname} '{section}' description in '{decl_name}' {hint}".strip()) 622 623 # 624 # Check that documented parameter names (from doc comments, including 625 # inline ``/** @member: */`` tags) actually match real members in 626 # the declaration. This catches mismatched or stale kernel-doc 627 # member tags that don't correspond to any actual struct/union 628 # member or function parameter. 629 # 630 for param_name, desc in self.entry.parameterdescs.items(): 631 # Skip auto-generated entries from push_parameter() 632 if desc == self.undescribed: 633 continue 634 if desc in ("no arguments", "anonymous\n", "variable arguments"): 635 continue 636 if param_name.startswith("{unnamed_"): 637 continue 638 if param_name in self.entry.parameterlist: 639 continue 640 641 hint = self.get_suggestions_hint(param_name, self.entry.parameterlist) 642 if decl_type == 'function': 643 dname = f"{decl_type} parameter" 644 else: 645 dname = f"{decl_type} member" 646 self.emit_msg(ln, 647 f"Excess {dname} '{param_name}' description in '{decl_name}' {hint}".strip()) 648 649 def check_return_section(self, ln, declaration_name, return_type): 650 """ 651 If the function doesn't return void, warns about the lack of a 652 return description. 653 """ 654 655 if not self.config.wreturn: 656 return 657 658 # Ignore an empty return type (It's a macro) 659 # Ignore functions with a "void" return type (but not "void *") 660 if not return_type or KernRe(r'void\s*\w*\s*$').search(return_type): 661 return 662 663 if not self.entry.sections.get("Return", None): 664 self.emit_msg(ln, 665 f"No description found for return value of '{declaration_name}'") 666 667 def split_struct_proto(self, proto): 668 """ 669 Split apart a structure prototype; returns (struct|union, name, 670 members) or ``None``. 671 """ 672 673 type_pattern = r'(struct|union)' 674 qualifiers = [ 675 "__attribute__", 676 "__packed", 677 "__aligned", 678 "____cacheline_aligned_in_smp", 679 "____cacheline_aligned", 680 ] 681 definition_body = r'\{(.*)\}\s*' + "(?:" + '|'.join(qualifiers) + ")?" 682 683 r = KernRe(type_pattern + r'\s+(\w+)\s*' + definition_body) 684 if r.search(proto): 685 return (r.group(1), r.group(2), r.group(3)) 686 else: 687 r = KernRe(r'typedef\s+' + type_pattern + r'\s*' + definition_body + r'\s*(\w+)\s*;') 688 if r.search(proto): 689 return (r.group(1), r.group(3), r.group(2)) 690 return None 691 692 def rewrite_struct_members(self, members): 693 """ 694 Process ``struct``/``union`` members from the most deeply nested 695 outward. 696 697 Rewrite the members of a ``struct`` or ``union`` for easier formatting 698 later on. Among other things, this function will turn a member like:: 699 700 struct { inner_members; } foo; 701 702 into:: 703 704 struct foo; inner_members; 705 """ 706 707 # 708 # The trick is in the ``^{`` below - it prevents a match of an outer 709 # ``struct``/``union`` until the inner one has been munged 710 # (removing the ``{`` in the process). 711 # 712 struct_members = KernRe(r'(struct|union)' # 0: declaration type 713 r'([^\{\};]+)' # 1: possible name 714 r'(\{)' 715 r'([^\{\}]*)' # 3: Contents of declaration 716 r'(\})' 717 r'([^\{\};]*)(;)') # 5: Remaining stuff after declaration 718 tuples = struct_members.findall(members) 719 while tuples: 720 for t in tuples: 721 newmember = "" 722 oldmember = "".join(t) # Reconstruct the original formatting 723 dtype, name, lbr, content, rbr, rest, semi = t 724 # 725 # Pass through each field name, normalizing the form and formatting. 726 # 727 for s_id in rest.split(','): 728 s_id = s_id.strip() 729 newmember += f"{dtype} {s_id}; " 730 # 731 # Remove bitfield/array/pointer info, getting the bare name. 732 # 733 s_id = KernRe(r'[:\[].*').sub('', s_id) 734 s_id = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', s_id) 735 # 736 # Pass through the members of this inner structure/union. 737 # 738 for arg in content.split(';'): 739 arg = arg.strip() 740 # 741 # Look for (type)(*name)(args) - pointer to function 742 # 743 r = KernRe(r'^([^\(]+\(\*?\s*)([\w.]*)(\s*\).*)') 744 if r.match(arg): 745 dtype, name, extra = r.group(1), r.group(2), r.group(3) 746 # Pointer-to-function 747 if not s_id: 748 # Anonymous struct/union 749 newmember += f"{dtype}{name}{extra}; " 750 else: 751 newmember += f"{dtype}{s_id}.{name}{extra}; " 752 # 753 # Otherwise a non-function member. 754 # 755 else: 756 # 757 # Remove bitmap and array portions and spaces around commas 758 # 759 arg = KernRe(r':\s*\d+\s*').sub('', arg) 760 arg = KernRe(r'\[.*\]').sub('', arg) 761 arg = KernRe(r'\s*,\s*').sub(',', arg) 762 # 763 # Look for a normal decl - "type name[,name...]" 764 # 765 r = KernRe(r'(.*)\s+([\S+,]+)') 766 if r.search(arg): 767 for name in r.group(2).split(','): 768 name = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', name) 769 if not s_id: 770 # Anonymous struct/union 771 newmember += f"{r.group(1)} {name}; " 772 else: 773 newmember += f"{r.group(1)} {s_id}.{name}; " 774 else: 775 newmember += f"{arg}; " 776 # 777 # At the end of the s_id loop, replace the original declaration with 778 # the munged version. 779 # 780 members = members.replace(oldmember, newmember) 781 # 782 # End of the tuple loop - search again and see if there are outer members 783 # that now turn up. 784 # 785 tuples = struct_members.findall(members) 786 return members 787 788 def format_struct_decl(self, declaration): 789 """ 790 Format the ``struct`` declaration into a standard form for inclusion 791 in the resulting docs. 792 """ 793 794 # 795 # Insert newlines, get rid of extra spaces. 796 # 797 declaration = KernRe(r'([\{;])').sub(r'\1\n', declaration) 798 declaration = KernRe(r'\}\s+;').sub('};', declaration) 799 # 800 # Format inline enums with each member on its own line. 801 # 802 r = KernRe(r'(enum\s+\{[^\}]+),([^\n])') 803 while r.search(declaration): 804 declaration = r.sub(r'\1,\n\2', declaration) 805 # 806 # Now go through and supply the right number of tabs 807 # for each line. 808 # 809 def_args = declaration.split('\n') 810 level = 1 811 declaration = "" 812 for clause in def_args: 813 clause = KernRe(r'\s+').sub(' ', clause.strip(), count=1) 814 if clause: 815 if '}' in clause and level > 1: 816 level -= 1 817 if not clause.startswith('#'): 818 declaration += "\t" * level 819 declaration += "\t" + clause + "\n" 820 if "{" in clause and "}" not in clause: 821 level += 1 822 return declaration 823 824 825 def dump_struct(self, ln, proto, source): 826 """ 827 Store an entry for a ``struct`` or ``union`` 828 """ 829 # 830 # Do the basic parse to get the pieces of the declaration. 831 # 832 source = source 833 proto = trim_private_members(proto) 834 835 # 836 # Expand macro-based declarations, like context_lock_struct(), 837 # into a plain struct/union declaration before splitting it, 838 # as dump_var() and dump_function() already do. 839 # 840 proto = self.xforms.apply("struct", proto) 841 struct_parts = self.split_struct_proto(proto) 842 if not struct_parts: 843 self.emit_msg(ln, f"{proto} error: Cannot parse struct or union!") 844 return 845 decl_type, declaration_name, members = struct_parts 846 847 if self.entry.identifier != declaration_name: 848 self.emit_msg(ln, f"expecting prototype for {decl_type} {self.entry.identifier}. " 849 f"Prototype was for {decl_type} {declaration_name} instead") 850 return 851 852 # 853 # Deal with embedded struct and union members, and drop enums entirely. 854 # 855 declaration = members 856 members = self.rewrite_struct_members(members) 857 members = re.sub(r'(\{[^\{\}]*\})', '', members) 858 # 859 # Output the result and we are done. 860 # 861 self.create_parameter_list(ln, decl_type, members, ';', 862 declaration_name) 863 self.check_sections(ln, declaration_name, decl_type) 864 self.output_declaration(decl_type, declaration_name, 865 source=source, 866 definition=self.format_struct_decl(declaration), 867 purpose=self.entry.declaration_purpose) 868 869 def dump_enum(self, ln, proto, source): 870 """ 871 Store an ``enum`` inside self.entries array. 872 """ 873 # 874 # Strip preprocessor directives. Note that this depends on the 875 # trailing semicolon we added in process_proto_type(). 876 # 877 source = source 878 proto = trim_private_members(proto) 879 proto = KernRe(r'#\s*((define|ifdef|if)\s+|endif)[^;]*;', flags=re.S).sub('', proto) 880 # 881 # Parse out the name and members of the enum. Typedef form first. 882 # 883 r = KernRe(r'typedef\s+enum\s*\{(.*)\}\s*(\w*)\s*;') 884 if r.search(proto): 885 declaration_name = r.group(2) 886 members = r.group(1) 887 # 888 # Failing that, look for a straight enum 889 # 890 else: 891 r = KernRe(r'enum\s+(\w*)\s*\{(.*)\}') 892 if r.match(proto): 893 declaration_name = r.group(1) 894 members = r.group(2) 895 # 896 # OK, this isn't going to work. 897 # 898 else: 899 self.emit_msg(ln, f"{proto}: error: Cannot parse enum!") 900 return 901 # 902 # Make sure we found what we were expecting. 903 # 904 if self.entry.identifier != declaration_name: 905 if self.entry.identifier == "": 906 self.emit_msg(ln, 907 f"{proto}: wrong kernel-doc identifier on prototype") 908 else: 909 self.emit_msg(ln, 910 f"expecting prototype for enum {self.entry.identifier}. " 911 f"Prototype was for enum {declaration_name} instead") 912 return 913 914 if not declaration_name: 915 declaration_name = "(anonymous)" 916 # 917 # Parse out the name of each enum member, and verify that we 918 # have a description for it. 919 # 920 member_set = set() 921 members = KernRe(r'\([^;)]*\)').sub('', members) 922 for arg in members.split(','): 923 arg = KernRe(r'^\s*(\w+).*').sub(r'\1', arg) 924 if not arg.strip(): 925 continue 926 927 self.entry.parameterlist.append(arg) 928 if arg not in self.entry.parameterdescs: 929 self.entry.parameterdescs[arg] = self.undescribed 930 self.emit_msg(ln, 931 f"Enum value '{arg}' not described in enum '{declaration_name}'") 932 member_set.add(arg) 933 # 934 # Ensure that every described member actually exists in the enum. 935 # 936 for k in self.entry.parameterdescs: 937 if k not in member_set: 938 self.emit_msg(ln, 939 f"Excess enum value '@{k}' description in '{declaration_name}'") 940 941 self.output_declaration('enum', declaration_name, 942 source=source, 943 purpose=self.entry.declaration_purpose) 944 945 def dump_var(self, ln, proto, source): 946 """ 947 Store variables that are part of kAPI. 948 """ 949 VAR_ATTRIBS = [ 950 "extern", 951 "const", 952 ] 953 OPTIONAL_VAR_ATTR = r"^(?:\b(?:" +"|".join(VAR_ATTRIBS) +r")\b\s*)*" 954 955 # 956 # Store the full prototype before modifying it 957 # 958 source = source 959 full_proto = proto 960 declaration_name = None 961 962 # 963 # Handle macro definitions 964 # 965 macro_prefixes = [ 966 KernRe(r"DEFINE_[\w_]+\s*\(([\w_]+)\)"), 967 ] 968 969 for r in macro_prefixes: 970 match = r.search(proto) 971 if match: 972 declaration_name = match.group(1) 973 break 974 975 # 976 # Drop comments and macros to have a pure C prototype 977 # 978 if not declaration_name: 979 proto = self.xforms.apply("var", proto) 980 981 proto = proto.rstrip() 982 983 # 984 # Variable name is at the end of the declaration 985 # 986 987 default_val = None 988 989 r= KernRe(OPTIONAL_VAR_ATTR + r"\s*[\w_\s]*\s+(?:\*+)?([\w_]+)\s*[\d\]\[]*\s*(=.*)?") 990 if r.match(proto): 991 if not declaration_name: 992 declaration_name = r.group(1) 993 994 default_val = r.group(2) 995 else: 996 r= KernRe(OPTIONAL_VAR_ATTR + r"(?:[\w_\s]*)?\s+(?:\*+)?(?:[\w_]+)\s*[\d\]\[]*\s*(=.*)?") 997 998 if r.match(proto): 999 default_val = r.group(1) 1000 if not declaration_name: 1001 self.emit_msg(ln,f"{proto}: can't parse variable") 1002 return 1003 1004 if default_val: 1005 default_val = default_val.lstrip("=").strip() 1006 1007 self.output_declaration("var", declaration_name, 1008 source=source, 1009 full_proto=full_proto, 1010 default_val=default_val, 1011 purpose=self.entry.declaration_purpose) 1012 1013 def dump_declaration(self, ln, prototype, source): 1014 """ 1015 Store a data declaration inside self.entries array. 1016 """ 1017 1018 if self.entry.decl_type == "enum": 1019 self.dump_enum(ln, prototype, source) 1020 elif self.entry.decl_type == "typedef": 1021 self.dump_typedef(ln, prototype, source) 1022 elif self.entry.decl_type in ["union", "struct"]: 1023 self.dump_struct(ln, prototype, source) 1024 elif self.entry.decl_type == "var": 1025 self.dump_var(ln, prototype, source) 1026 else: 1027 # This would be a bug 1028 self.emit_message(ln, f'Unknown declaration type: {self.entry.decl_type}') 1029 1030 def dump_function(self, ln, prototype, source): 1031 """ 1032 Store a function or function macro inside self.entries array. 1033 """ 1034 1035 source = source 1036 found = func_macro = False 1037 return_type = '' 1038 decl_type = 'function' 1039 1040 # 1041 # If we have a macro, remove the "#define" at the front. 1042 # 1043 new_proto = KernRe(r"^#\s*define\s+").sub("", prototype) 1044 if new_proto != prototype: 1045 prototype = new_proto 1046 # 1047 # Dispense with the simple "#define A B" case here; the key 1048 # is the space after the name of the symbol being defined. 1049 # NOTE that the seemingly misnamed "func_macro" indicates a 1050 # macro *without* arguments. 1051 # 1052 r = KernRe(r'^(\w+)\s+') 1053 if r.search(prototype): 1054 return_type = '' 1055 declaration_name = r.group(1) 1056 func_macro = True 1057 found = True 1058 else: 1059 # 1060 # Apply the initial transformations. 1061 # 1062 prototype = self.xforms.apply("func", prototype) 1063 1064 # Yes, this truly is vile. We are looking for: 1065 # 1. Return type (may be nothing if we're looking at a macro) 1066 # 2. Function name 1067 # 3. Function parameters. 1068 # 1069 # All the while we have to watch out for function pointer parameters 1070 # (which IIRC is what the two sections are for), C types (these 1071 # regexps don't even start to express all the possibilities), and 1072 # so on. 1073 # 1074 # If you mess with these regexps, it's a good idea to check that 1075 # the following functions' documentation still comes out right: 1076 # - parport_register_device (function pointer parameters) 1077 # - atomic_set (macro) 1078 # - pci_match_device, __copy_to_user (long return type) 1079 1080 name = r'\w+' 1081 type1 = r'(?:[\w\s]+)?' 1082 type2 = r'(?:[\w\s]+\*+)+' 1083 # 1084 # Attempt to match first on (args) with no internal parentheses; this 1085 # lets us easily filter out __acquires() and other post-args stuff. If 1086 # that fails, just grab the rest of the line to the last closing 1087 # parenthesis. 1088 # 1089 proto_args = r'\(([^\(]*|.*)\)' 1090 # 1091 # (Except for the simple macro case) attempt to split up the prototype 1092 # in the various ways we understand. 1093 # 1094 if not found: 1095 patterns = [ 1096 rf'^()({name})\s*{proto_args}', 1097 rf'^({type1})\s+({name})\s*{proto_args}', 1098 rf'^({type2})\s*({name})\s*{proto_args}', 1099 ] 1100 1101 for p in patterns: 1102 r = KernRe(p) 1103 if r.match(prototype): 1104 return_type = r.group(1) 1105 declaration_name = r.group(2) 1106 args = r.group(3) 1107 self.create_parameter_list(ln, decl_type, args, ',', 1108 declaration_name) 1109 found = True 1110 break 1111 # 1112 # Parsing done; make sure that things are as we expect. 1113 # 1114 if not found: 1115 self.emit_msg(ln, 1116 f"cannot understand function prototype: '{prototype}'") 1117 return 1118 if self.entry.identifier != declaration_name: 1119 self.emit_msg(ln, f"expecting prototype for {self.entry.identifier}(). " 1120 f"Prototype was for {declaration_name}() instead") 1121 return 1122 self.check_sections(ln, declaration_name, "function") 1123 self.check_return_section(ln, declaration_name, return_type) 1124 # 1125 # Store the result. 1126 # 1127 self.output_declaration(decl_type, declaration_name, 1128 source=source, 1129 typedef=('typedef' in return_type), 1130 functiontype=return_type, 1131 purpose=self.entry.declaration_purpose, 1132 func_macro=func_macro) 1133 1134 1135 def dump_typedef(self, ln, proto, source): 1136 """ 1137 Store a ``typedef`` inside self.entries array. 1138 """ 1139 # 1140 # We start by looking for function typedefs. 1141 # 1142 typedef_type = r'typedef((?:\s+[\w*]+\b){0,7}\s+(?:\w+\b|\*+))\s*' 1143 typedef_ident = r'\*?\s*(\w\S+)\s*' 1144 typedef_args = r'\s*\((.*)\);' 1145 1146 source = source 1147 1148 typedef1 = KernRe(typedef_type + r'\(' + typedef_ident + r'\)' + typedef_args) 1149 typedef2 = KernRe(typedef_type + typedef_ident + typedef_args) 1150 1151 # Parse function typedef prototypes 1152 for r in [typedef1, typedef2]: 1153 if not r.match(proto): 1154 continue 1155 1156 return_type = r.group(1).strip() 1157 declaration_name = r.group(2) 1158 args = r.group(3) 1159 1160 if self.entry.identifier != declaration_name: 1161 self.emit_msg(ln, 1162 f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead") 1163 return 1164 1165 self.create_parameter_list(ln, 'function', args, ',', declaration_name) 1166 1167 self.output_declaration('function', declaration_name, 1168 source=source, 1169 typedef=True, 1170 functiontype=return_type, 1171 purpose=self.entry.declaration_purpose) 1172 return 1173 # 1174 # Not a function, try to parse a simple typedef. 1175 # 1176 r = KernRe(r'typedef.*\s+(\w+)\s*;') 1177 if r.match(proto): 1178 declaration_name = r.group(1) 1179 1180 if self.entry.identifier != declaration_name: 1181 self.emit_msg(ln, 1182 f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead") 1183 return 1184 1185 self.output_declaration('typedef', declaration_name, 1186 source=source, 1187 purpose=self.entry.declaration_purpose) 1188 return 1189 1190 self.emit_msg(ln, "error: Cannot parse typedef!") 1191 1192 @staticmethod 1193 def process_export(function_set, line): 1194 """ 1195 process ``EXPORT_SYMBOL*`` tags 1196 1197 This method doesn't use any variable from the class, so declare it 1198 with a staticmethod decorator. 1199 """ 1200 1201 # We support documenting some exported symbols with different 1202 # names. A horrible hack. 1203 suffixes = [ '_noprof' ] 1204 1205 # Note: it accepts only one EXPORT_SYMBOL* per line, as having 1206 # multiple export lines would violate Kernel coding style. 1207 1208 if export_symbol.search(line): 1209 symbol = export_symbol.group(2) 1210 elif export_symbol_ns.search(line): 1211 symbol = export_symbol_ns.group(2) 1212 else: 1213 return False 1214 # 1215 # Found an export, trim out any special suffixes 1216 # 1217 for suffix in suffixes: 1218 # Be backward compatible with Python < 3.9 1219 if symbol.endswith(suffix): 1220 symbol = symbol[:-len(suffix)] 1221 function_set.add(symbol) 1222 return True 1223 1224 def process_normal(self, ln, line, source): 1225 """ 1226 STATE_NORMAL: looking for the ``/**`` to begin everything. 1227 """ 1228 1229 if not doc_start.match(line): 1230 return 1231 1232 # start a new entry 1233 self.reset_state(ln) 1234 1235 # next line is always the function name 1236 self.state = state.NAME 1237 1238 def process_name(self, ln, line, source): 1239 """ 1240 STATE_NAME: Looking for the "name - description" line 1241 """ 1242 # 1243 # Check for a DOC: block and handle them specially. 1244 # 1245 if doc_block.search(line): 1246 1247 if not doc_block.group(1): 1248 self.entry.begin_section(ln, "Introduction") 1249 else: 1250 self.entry.begin_section(ln, doc_block.group(1)) 1251 1252 self.entry.identifier = self.entry.section 1253 self.state = state.DOCBLOCK 1254 # 1255 # Otherwise we're looking for a normal kerneldoc declaration line. 1256 # 1257 elif doc_decl.search(line): 1258 self.entry.identifier = doc_decl.group(1) 1259 1260 # Test for data declaration 1261 if doc_begin_data.search(line): 1262 self.entry.decl_type = doc_begin_data.group(1) 1263 self.entry.identifier = doc_begin_data.group(2) 1264 # 1265 # Look for a function description 1266 # 1267 elif doc_begin_func.search(line): 1268 self.entry.identifier = doc_begin_func.group(1) 1269 self.entry.decl_type = "function" 1270 # 1271 # We struck out. 1272 # 1273 else: 1274 self.emit_msg(ln, 1275 f"This comment starts with '/**', but isn't a kernel-doc comment. Refer to Documentation/doc-guide/kernel-doc.rst\n{line}") 1276 self.state = state.NORMAL 1277 return 1278 # 1279 # OK, set up for a new kerneldoc entry. 1280 # 1281 self.state = state.BODY 1282 self.entry.identifier = self.entry.identifier.strip(" ") 1283 # if there's no @param blocks need to set up default section here 1284 self.entry.begin_section(ln + 1) 1285 # 1286 # Find the description portion, which *should* be there but 1287 # isn't always. 1288 # (We should be able to capture this from the previous parsing - someday) 1289 # 1290 r = KernRe("[-:](.*)") 1291 if r.search(line): 1292 self.entry.declaration_purpose = trim_whitespace(r.group(1)) 1293 self.state = state.DECLARATION 1294 else: 1295 self.entry.declaration_purpose = "" 1296 1297 if not self.entry.declaration_purpose and self.config.wshort_desc: 1298 self.emit_msg(ln, 1299 f"missing initial short description on line:\n{line}") 1300 1301 if not self.entry.identifier and self.entry.decl_type != "enum": 1302 self.emit_msg(ln, 1303 f"wrong kernel-doc identifier on line:\n{line}") 1304 self.state = state.NORMAL 1305 1306 if self.config.verbose: 1307 self.emit_msg(ln, 1308 f"Scanning doc for {self.entry.decl_type} {self.entry.identifier}", 1309 warning=False) 1310 # 1311 # Failed to find an identifier. Emit a warning 1312 # 1313 else: 1314 self.emit_msg(ln, f"Cannot find identifier on line:\n{line}") 1315 1316 def is_new_section(self, ln, line): 1317 """ 1318 Helper function to determine if a new section is being started. 1319 """ 1320 if doc_sect.search(line): 1321 self.state = state.BODY 1322 # 1323 # Pick out the name of our new section, tweaking it if need be. 1324 # 1325 newsection = doc_sect.group(1) 1326 if newsection.lower() == 'description': 1327 newsection = 'Description' 1328 elif newsection.lower() == 'context': 1329 newsection = 'Context' 1330 self.state = state.SPECIAL_SECTION 1331 elif newsection.lower() in ["@return", "@returns", 1332 "return", "returns"]: 1333 newsection = "Return" 1334 self.state = state.SPECIAL_SECTION 1335 elif newsection[0] == '@': 1336 self.state = state.SPECIAL_SECTION 1337 # 1338 # Initialize the contents, and get the new section going. 1339 # 1340 newcontents = doc_sect.group(2) 1341 if not newcontents: 1342 newcontents = "" 1343 self.dump_section() 1344 self.entry.begin_section(ln, newsection) 1345 self.entry.leading_space = None 1346 1347 self.entry.add_text(newcontents.lstrip()) 1348 return True 1349 return False 1350 1351 def is_comment_end(self, ln, line): 1352 """ 1353 Helper function to detect (and effect) the end of a kerneldoc comment. 1354 """ 1355 if doc_end.search(line): 1356 self.dump_section() 1357 1358 # Look for doc_com + <text> + doc_end: 1359 r = KernRe(r'\s*\*\s*[a-zA-Z_0-9:.]+\*/') 1360 if r.match(line): 1361 self.emit_msg(ln, f"suspicious ending line: {line}") 1362 1363 self.entry.prototype = "" 1364 self.entry.new_start_line = ln + 1 1365 1366 self.state = state.PROTO 1367 return True 1368 return False 1369 1370 1371 def process_decl(self, ln, line, source): 1372 """ 1373 STATE_DECLARATION: We've seen the beginning of a declaration. 1374 """ 1375 if self.is_new_section(ln, line) or self.is_comment_end(ln, line): 1376 return 1377 # 1378 # Look for anything with the " * " line beginning. 1379 # 1380 if doc_content.search(line): 1381 cont = doc_content.group(1) 1382 # 1383 # A blank line means that we have moved out of the declaration 1384 # part of the comment (without any "special section" parameter 1385 # descriptions). 1386 # 1387 if cont == "": 1388 self.state = state.BODY 1389 # 1390 # Otherwise we have more of the declaration section to soak up. 1391 # 1392 else: 1393 self.entry.declaration_purpose = \ 1394 trim_whitespace(self.entry.declaration_purpose + ' ' + cont) 1395 else: 1396 # Unknown line, ignore 1397 self.emit_msg(ln, f"bad line: {line}") 1398 1399 1400 def process_special(self, ln, line, source): 1401 """ 1402 STATE_SPECIAL_SECTION: a section ending with a blank line. 1403 """ 1404 # 1405 # If we have hit a blank line (only the " * " marker), then this 1406 # section is done. 1407 # 1408 if KernRe(r"\s*\*\s*$").match(line): 1409 self.entry.begin_section(ln, dump = True) 1410 self.state = state.BODY 1411 return 1412 # 1413 # Not a blank line, look for the other ways to end the section. 1414 # 1415 if self.is_new_section(ln, line) or self.is_comment_end(ln, line): 1416 return 1417 # 1418 # OK, we should have a continuation of the text for this section. 1419 # 1420 if doc_content.search(line): 1421 cont = doc_content.group(1) 1422 # 1423 # If the lines of text after the first in a special section have 1424 # leading white space, we need to trim it out or Sphinx will get 1425 # confused. For the second line (the None case), see what we 1426 # find there and remember it. 1427 # 1428 if self.entry.leading_space is None: 1429 r = KernRe(r'^(\s+)') 1430 if r.match(cont): 1431 self.entry.leading_space = len(r.group(1)) 1432 else: 1433 self.entry.leading_space = 0 1434 # 1435 # Otherwise, before trimming any leading chars, be *sure* 1436 # that they are white space. We should maybe warn if this 1437 # isn't the case. 1438 # 1439 for i in range(0, self.entry.leading_space): 1440 if cont[i] != " ": 1441 self.entry.leading_space = i 1442 break 1443 # 1444 # Add the trimmed result to the section and we're done. 1445 # 1446 self.entry.add_text(cont[self.entry.leading_space:]) 1447 else: 1448 # Unknown line, ignore 1449 self.emit_msg(ln, f"bad line: {line}") 1450 1451 def process_body(self, ln, line, source): 1452 """ 1453 STATE_BODY: the bulk of a kerneldoc comment. 1454 """ 1455 if self.is_new_section(ln, line) or self.is_comment_end(ln, line): 1456 return 1457 1458 if doc_content.search(line): 1459 cont = doc_content.group(1) 1460 self.entry.add_text(cont) 1461 else: 1462 # Unknown line, ignore 1463 self.emit_msg(ln, f"bad line: {line}") 1464 1465 def process_inline_name(self, ln, line, source): 1466 """STATE_INLINE_NAME: beginning of docbook comments within a prototype.""" 1467 1468 if doc_inline_sect.search(line): 1469 self.entry.begin_section(ln, doc_inline_sect.group(1)) 1470 self.entry.add_text(doc_inline_sect.group(2).lstrip()) 1471 self.state = state.INLINE_TEXT 1472 elif doc_inline_end.search(line): 1473 self.dump_section() 1474 self.state = state.PROTO 1475 elif doc_content.search(line): 1476 self.emit_msg(ln, f"Incorrect use of kernel-doc format: {line}") 1477 self.state = state.PROTO 1478 1479 # 1480 # Don't let it add partial comments at the code, as breaks the 1481 # logic meant to remove comments from prototypes. 1482 # 1483 self.process_proto_type(ln, "/**\n" + line, source) 1484 # else ... ?? 1485 1486 def process_inline_text(self, ln, line, source): 1487 """STATE_INLINE_TEXT: docbook comments within a prototype.""" 1488 1489 if doc_inline_end.search(line): 1490 self.dump_section() 1491 self.state = state.PROTO 1492 elif doc_content.search(line): 1493 self.entry.add_text(doc_content.group(1)) 1494 # else ... ?? 1495 1496 def syscall_munge(self, ln, proto): # pylint: disable=W0613 1497 """ 1498 Handle syscall definitions. 1499 """ 1500 1501 is_void = False 1502 1503 # Strip newlines/CR's 1504 proto = re.sub(r'[\r\n]+', ' ', proto) 1505 1506 # Check if it's a SYSCALL_DEFINE0 1507 if 'SYSCALL_DEFINE0' in proto: 1508 is_void = True 1509 1510 # Replace SYSCALL_DEFINE with correct return type & function name 1511 proto = KernRe(r'SYSCALL_DEFINE.*\(').sub('long sys_', proto) 1512 1513 r = KernRe(r'long\s+(sys_.*?),') 1514 if r.search(proto): 1515 proto = KernRe(',').sub('(', proto, count=1) 1516 elif is_void: 1517 proto = KernRe(r'\)').sub('(void)', proto, count=1) 1518 1519 # Now delete all of the odd-numbered commas in the proto 1520 # so that argument types & names don't have a comma between them 1521 count = 0 1522 length = len(proto) 1523 1524 if is_void: 1525 length = 0 # skip the loop if is_void 1526 1527 for ix in range(length): 1528 if proto[ix] == ',': 1529 count += 1 1530 if count % 2 == 1: 1531 proto = proto[:ix] + ' ' + proto[ix + 1:] 1532 1533 return proto 1534 1535 def tracepoint_munge(self, ln, proto): 1536 """ 1537 Handle tracepoint definitions. 1538 """ 1539 1540 tracepointname = None 1541 tracepointargs = None 1542 1543 # Match tracepoint name based on different patterns 1544 r = KernRe(r'TRACE_EVENT\((.*?),') 1545 if r.search(proto): 1546 tracepointname = r.group(1) 1547 1548 r = KernRe(r'DEFINE_SINGLE_EVENT\((.*?),') 1549 if r.search(proto): 1550 tracepointname = r.group(1) 1551 1552 r = KernRe(r'DEFINE_EVENT\((.*?),(.*?),') 1553 if r.search(proto): 1554 tracepointname = r.group(2) 1555 1556 if tracepointname: 1557 tracepointname = tracepointname.lstrip() 1558 1559 r = KernRe(r'TP_PROTO\((.*?)\)') 1560 if r.search(proto): 1561 tracepointargs = r.group(1) 1562 1563 if not tracepointname or not tracepointargs: 1564 self.emit_msg(ln, 1565 f"Unrecognized tracepoint format:\n{proto}\n") 1566 else: 1567 proto = f"static inline void trace_{tracepointname}({tracepointargs})" 1568 self.entry.identifier = f"trace_{self.entry.identifier}" 1569 1570 return proto 1571 1572 def process_proto_function(self, ln, line, source): 1573 """Ancillary routine to process a function prototype.""" 1574 1575 # strip C99-style comments to end of line 1576 line = KernRe(r"//.*$", re.S).sub('', line) 1577 # 1578 # Soak up the line's worth of prototype text, stopping at { or ; if present. 1579 # 1580 if KernRe(r'\s*#\s*define').match(line): 1581 self.entry.prototype = line 1582 elif not line.startswith('#'): # skip other preprocessor stuff 1583 r = KernRe(r'([^\{]*)') 1584 if r.match(line): 1585 self.entry.prototype += r.group(1) + " " 1586 # 1587 # If we now have the whole prototype, clean it up and declare victory. 1588 # 1589 if '{' in line or ';' in line or KernRe(r'\s*#\s*define').match(line): 1590 # strip comments and surrounding spaces 1591 self.entry.prototype = KernRe(r'/\*.*\*/').sub('', self.entry.prototype).strip() 1592 # 1593 # Handle self.entry.prototypes for function pointers like: 1594 # int (*pcs_config)(struct foo) 1595 # by turning it into 1596 # int pcs_config(struct foo) 1597 # 1598 r = KernRe(r'^(\S+\s+)\(\s*\*(\S+)\)') 1599 self.entry.prototype = r.sub(r'\1\2', self.entry.prototype) 1600 # 1601 # Handle special declaration syntaxes 1602 # 1603 if 'SYSCALL_DEFINE' in self.entry.prototype: 1604 self.entry.prototype = self.syscall_munge(ln, 1605 self.entry.prototype) 1606 else: 1607 r = KernRe(r'TRACE_EVENT|DEFINE_EVENT|DEFINE_SINGLE_EVENT') 1608 if r.search(self.entry.prototype): 1609 self.entry.prototype = self.tracepoint_munge(ln, 1610 self.entry.prototype) 1611 # 1612 # ... and we're done 1613 # 1614 self.dump_function(ln, self.entry.prototype, source) 1615 self.reset_state(ln) 1616 1617 def process_proto_type(self, ln, line, source): 1618 """ 1619 Ancillary routine to process a type. 1620 """ 1621 1622 # Strip C99-style comments and surrounding whitespace 1623 line = KernRe(r"//.*$", re.S).sub('', line).strip() 1624 if not line: 1625 return # nothing to see here 1626 1627 # To distinguish preprocessor directive from regular declaration later. 1628 if line.startswith('#'): 1629 line += ";" 1630 # 1631 # Split the declaration on any of { } or ;, and accumulate pieces 1632 # until we hit a semicolon while not inside {brackets} 1633 # 1634 r = KernRe(r'(.*?)([{};])') 1635 for chunk in r.split(line): 1636 if chunk: # Ignore empty matches 1637 self.entry.prototype += chunk 1638 # 1639 # This cries out for a match statement ... someday after we can 1640 # drop Python 3.9 ... 1641 # 1642 if chunk == '{': 1643 self.entry.brcount += 1 1644 elif chunk == '}': 1645 self.entry.brcount -= 1 1646 elif chunk == ';' and self.entry.brcount <= 0: 1647 self.dump_declaration(ln, self.entry.prototype, source) 1648 self.reset_state(ln) 1649 return 1650 # 1651 # We hit the end of the line while still in the declaration; put 1652 # in a space to represent the newline. 1653 # 1654 self.entry.prototype += ' ' 1655 1656 def process_proto(self, ln, line, source): 1657 """STATE_PROTO: reading a function/whatever prototype.""" 1658 1659 if doc_inline_oneline.search(line): 1660 self.entry.begin_section(ln, doc_inline_oneline.group(1)) 1661 self.entry.add_text(doc_inline_oneline.group(2)) 1662 self.dump_section() 1663 1664 elif doc_inline_start.search(line): 1665 self.state = state.INLINE_NAME 1666 1667 elif self.entry.decl_type == 'function': 1668 self.process_proto_function(ln, line, source) 1669 1670 else: 1671 self.process_proto_type(ln, line, source) 1672 1673 def process_docblock(self, ln, line, source): 1674 """STATE_DOCBLOCK: within a ``DOC:`` block.""" 1675 1676 if doc_end.search(line): 1677 self.dump_section() 1678 self.output_declaration("doc", self.entry.identifier, 1679 source=source) 1680 self.reset_state(ln) 1681 1682 elif doc_content.search(line): 1683 self.entry.add_text(doc_content.group(1)) 1684 1685 def parse_export(self): 1686 """ 1687 Parses ``EXPORT_SYMBOL*`` macros from a single Kernel source file. 1688 """ 1689 1690 export_table = set() 1691 1692 try: 1693 with open(self.fname, "r", encoding="utf8", 1694 errors="backslashreplace") as fp: 1695 1696 for line in fp: 1697 self.process_export(export_table, line) 1698 1699 except IOError: 1700 return None 1701 1702 return export_table 1703 1704 #: The state/action table telling us which function to invoke in each state. 1705 state_actions = { 1706 state.NORMAL: process_normal, 1707 state.NAME: process_name, 1708 state.BODY: process_body, 1709 state.DECLARATION: process_decl, 1710 state.SPECIAL_SECTION: process_special, 1711 state.INLINE_NAME: process_inline_name, 1712 state.INLINE_TEXT: process_inline_text, 1713 state.PROTO: process_proto, 1714 state.DOCBLOCK: process_docblock, 1715 } 1716 1717 def parse_kdoc(self): 1718 """ 1719 Open and process each line of a C source file. 1720 The parsing is controlled via a state machine, and the line is passed 1721 to a different process function depending on the state. The process 1722 function may update the state as needed. 1723 1724 Besides parsing kernel-doc tags, it also parses export symbols. 1725 """ 1726 1727 prev = "" 1728 prev_ln = None 1729 export_table = set() 1730 self.state = state.NORMAL 1731 source = "" 1732 1733 try: 1734 with open(self.fname, "r", encoding="utf8", 1735 errors="backslashreplace") as fp: 1736 for ln, line in enumerate(fp): 1737 1738 line = line.expandtabs().strip("\n") 1739 1740 # Group continuation lines on prototypes 1741 if self.state == state.PROTO: 1742 if line.endswith("\\"): 1743 prev += line.rstrip("\\") 1744 if not prev_ln: 1745 prev_ln = ln 1746 continue 1747 1748 if prev: 1749 ln = prev_ln 1750 line = prev + line 1751 prev = "" 1752 prev_ln = None 1753 1754 self.config.log.debug("%d %s: %s", 1755 ln, state.name[self.state], 1756 line) 1757 1758 if self.store_src: 1759 if source and self.state == state.NORMAL: 1760 source = "" 1761 elif self.state != state.NORMAL: 1762 source += line + "\n" 1763 1764 # This is an optimization over the original script. 1765 # There, when export_file was used for the same file, 1766 # it was read twice. Here, we use the already-existing 1767 # loop to parse exported symbols as well. 1768 # 1769 if (self.state != state.NORMAL) or \ 1770 not self.process_export(export_table, line): 1771 prev_state = self.state 1772 # Hand this line to the appropriate state handler 1773 self.state_actions[self.state](self, ln, line, source) 1774 if prev_state == state.NORMAL and self.state != state.NORMAL: 1775 source += line + "\n" 1776 1777 self.emit_unused_warnings() 1778 1779 except OSError: 1780 self.config.log.error(f"Error: Cannot open file {self.fname}") 1781 1782 return export_table, self.entries 1783