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 struct_parts = self.split_struct_proto(proto) 835 if not struct_parts: 836 self.emit_msg(ln, f"{proto} error: Cannot parse struct or union!") 837 return 838 decl_type, declaration_name, members = struct_parts 839 840 if self.entry.identifier != declaration_name: 841 self.emit_msg(ln, f"expecting prototype for {decl_type} {self.entry.identifier}. " 842 f"Prototype was for {decl_type} {declaration_name} instead") 843 return 844 # 845 # Go through the list of members applying all of our transformations. 846 # 847 members = self.xforms.apply("struct", members) 848 849 # 850 # Deal with embedded struct and union members, and drop enums entirely. 851 # 852 declaration = members 853 members = self.rewrite_struct_members(members) 854 members = re.sub(r'(\{[^\{\}]*\})', '', members) 855 # 856 # Output the result and we are done. 857 # 858 self.create_parameter_list(ln, decl_type, members, ';', 859 declaration_name) 860 self.check_sections(ln, declaration_name, decl_type) 861 self.output_declaration(decl_type, declaration_name, 862 source=source, 863 definition=self.format_struct_decl(declaration), 864 purpose=self.entry.declaration_purpose) 865 866 def dump_enum(self, ln, proto, source): 867 """ 868 Store an ``enum`` inside self.entries array. 869 """ 870 # 871 # Strip preprocessor directives. Note that this depends on the 872 # trailing semicolon we added in process_proto_type(). 873 # 874 source = source 875 proto = trim_private_members(proto) 876 proto = KernRe(r'#\s*((define|ifdef|if)\s+|endif)[^;]*;', flags=re.S).sub('', proto) 877 # 878 # Parse out the name and members of the enum. Typedef form first. 879 # 880 r = KernRe(r'typedef\s+enum\s*\{(.*)\}\s*(\w*)\s*;') 881 if r.search(proto): 882 declaration_name = r.group(2) 883 members = r.group(1) 884 # 885 # Failing that, look for a straight enum 886 # 887 else: 888 r = KernRe(r'enum\s+(\w*)\s*\{(.*)\}') 889 if r.match(proto): 890 declaration_name = r.group(1) 891 members = r.group(2) 892 # 893 # OK, this isn't going to work. 894 # 895 else: 896 self.emit_msg(ln, f"{proto}: error: Cannot parse enum!") 897 return 898 # 899 # Make sure we found what we were expecting. 900 # 901 if self.entry.identifier != declaration_name: 902 if self.entry.identifier == "": 903 self.emit_msg(ln, 904 f"{proto}: wrong kernel-doc identifier on prototype") 905 else: 906 self.emit_msg(ln, 907 f"expecting prototype for enum {self.entry.identifier}. " 908 f"Prototype was for enum {declaration_name} instead") 909 return 910 911 if not declaration_name: 912 declaration_name = "(anonymous)" 913 # 914 # Parse out the name of each enum member, and verify that we 915 # have a description for it. 916 # 917 member_set = set() 918 members = KernRe(r'\([^;)]*\)').sub('', members) 919 for arg in members.split(','): 920 arg = KernRe(r'^\s*(\w+).*').sub(r'\1', arg) 921 if not arg.strip(): 922 continue 923 924 self.entry.parameterlist.append(arg) 925 if arg not in self.entry.parameterdescs: 926 self.entry.parameterdescs[arg] = self.undescribed 927 self.emit_msg(ln, 928 f"Enum value '{arg}' not described in enum '{declaration_name}'") 929 member_set.add(arg) 930 # 931 # Ensure that every described member actually exists in the enum. 932 # 933 for k in self.entry.parameterdescs: 934 if k not in member_set: 935 self.emit_msg(ln, 936 f"Excess enum value '@{k}' description in '{declaration_name}'") 937 938 self.output_declaration('enum', declaration_name, 939 source=source, 940 purpose=self.entry.declaration_purpose) 941 942 def dump_var(self, ln, proto, source): 943 """ 944 Store variables that are part of kAPI. 945 """ 946 VAR_ATTRIBS = [ 947 "extern", 948 "const", 949 ] 950 OPTIONAL_VAR_ATTR = r"^(?:\b(?:" +"|".join(VAR_ATTRIBS) +r")\b\s*)*" 951 952 # 953 # Store the full prototype before modifying it 954 # 955 source = source 956 full_proto = proto 957 declaration_name = None 958 959 # 960 # Handle macro definitions 961 # 962 macro_prefixes = [ 963 KernRe(r"DEFINE_[\w_]+\s*\(([\w_]+)\)"), 964 ] 965 966 for r in macro_prefixes: 967 match = r.search(proto) 968 if match: 969 declaration_name = match.group(1) 970 break 971 972 # 973 # Drop comments and macros to have a pure C prototype 974 # 975 if not declaration_name: 976 proto = self.xforms.apply("var", proto) 977 978 proto = proto.rstrip() 979 980 # 981 # Variable name is at the end of the declaration 982 # 983 984 default_val = None 985 986 r= KernRe(OPTIONAL_VAR_ATTR + r"\s*[\w_\s]*\s+(?:\*+)?([\w_]+)\s*[\d\]\[]*\s*(=.*)?") 987 if r.match(proto): 988 if not declaration_name: 989 declaration_name = r.group(1) 990 991 default_val = r.group(2) 992 else: 993 r= KernRe(OPTIONAL_VAR_ATTR + r"(?:[\w_\s]*)?\s+(?:\*+)?(?:[\w_]+)\s*[\d\]\[]*\s*(=.*)?") 994 995 if r.match(proto): 996 default_val = r.group(1) 997 if not declaration_name: 998 self.emit_msg(ln,f"{proto}: can't parse variable") 999 return 1000 1001 if default_val: 1002 default_val = default_val.lstrip("=").strip() 1003 1004 self.output_declaration("var", declaration_name, 1005 source=source, 1006 full_proto=full_proto, 1007 default_val=default_val, 1008 purpose=self.entry.declaration_purpose) 1009 1010 def dump_declaration(self, ln, prototype, source): 1011 """ 1012 Store a data declaration inside self.entries array. 1013 """ 1014 1015 if self.entry.decl_type == "enum": 1016 self.dump_enum(ln, prototype, source) 1017 elif self.entry.decl_type == "typedef": 1018 self.dump_typedef(ln, prototype, source) 1019 elif self.entry.decl_type in ["union", "struct"]: 1020 self.dump_struct(ln, prototype, source) 1021 elif self.entry.decl_type == "var": 1022 self.dump_var(ln, prototype, source) 1023 else: 1024 # This would be a bug 1025 self.emit_message(ln, f'Unknown declaration type: {self.entry.decl_type}') 1026 1027 def dump_function(self, ln, prototype, source): 1028 """ 1029 Store a function or function macro inside self.entries array. 1030 """ 1031 1032 source = source 1033 found = func_macro = False 1034 return_type = '' 1035 decl_type = 'function' 1036 1037 # 1038 # If we have a macro, remove the "#define" at the front. 1039 # 1040 new_proto = KernRe(r"^#\s*define\s+").sub("", prototype) 1041 if new_proto != prototype: 1042 prototype = new_proto 1043 # 1044 # Dispense with the simple "#define A B" case here; the key 1045 # is the space after the name of the symbol being defined. 1046 # NOTE that the seemingly misnamed "func_macro" indicates a 1047 # macro *without* arguments. 1048 # 1049 r = KernRe(r'^(\w+)\s+') 1050 if r.search(prototype): 1051 return_type = '' 1052 declaration_name = r.group(1) 1053 func_macro = True 1054 found = True 1055 else: 1056 # 1057 # Apply the initial transformations. 1058 # 1059 prototype = self.xforms.apply("func", prototype) 1060 1061 # Yes, this truly is vile. We are looking for: 1062 # 1. Return type (may be nothing if we're looking at a macro) 1063 # 2. Function name 1064 # 3. Function parameters. 1065 # 1066 # All the while we have to watch out for function pointer parameters 1067 # (which IIRC is what the two sections are for), C types (these 1068 # regexps don't even start to express all the possibilities), and 1069 # so on. 1070 # 1071 # If you mess with these regexps, it's a good idea to check that 1072 # the following functions' documentation still comes out right: 1073 # - parport_register_device (function pointer parameters) 1074 # - atomic_set (macro) 1075 # - pci_match_device, __copy_to_user (long return type) 1076 1077 name = r'\w+' 1078 type1 = r'(?:[\w\s]+)?' 1079 type2 = r'(?:[\w\s]+\*+)+' 1080 # 1081 # Attempt to match first on (args) with no internal parentheses; this 1082 # lets us easily filter out __acquires() and other post-args stuff. If 1083 # that fails, just grab the rest of the line to the last closing 1084 # parenthesis. 1085 # 1086 proto_args = r'\(([^\(]*|.*)\)' 1087 # 1088 # (Except for the simple macro case) attempt to split up the prototype 1089 # in the various ways we understand. 1090 # 1091 if not found: 1092 patterns = [ 1093 rf'^()({name})\s*{proto_args}', 1094 rf'^({type1})\s+({name})\s*{proto_args}', 1095 rf'^({type2})\s*({name})\s*{proto_args}', 1096 ] 1097 1098 for p in patterns: 1099 r = KernRe(p) 1100 if r.match(prototype): 1101 return_type = r.group(1) 1102 declaration_name = r.group(2) 1103 args = r.group(3) 1104 self.create_parameter_list(ln, decl_type, args, ',', 1105 declaration_name) 1106 found = True 1107 break 1108 # 1109 # Parsing done; make sure that things are as we expect. 1110 # 1111 if not found: 1112 self.emit_msg(ln, 1113 f"cannot understand function prototype: '{prototype}'") 1114 return 1115 if self.entry.identifier != declaration_name: 1116 self.emit_msg(ln, f"expecting prototype for {self.entry.identifier}(). " 1117 f"Prototype was for {declaration_name}() instead") 1118 return 1119 self.check_sections(ln, declaration_name, "function") 1120 self.check_return_section(ln, declaration_name, return_type) 1121 # 1122 # Store the result. 1123 # 1124 self.output_declaration(decl_type, declaration_name, 1125 source=source, 1126 typedef=('typedef' in return_type), 1127 functiontype=return_type, 1128 purpose=self.entry.declaration_purpose, 1129 func_macro=func_macro) 1130 1131 1132 def dump_typedef(self, ln, proto, source): 1133 """ 1134 Store a ``typedef`` inside self.entries array. 1135 """ 1136 # 1137 # We start by looking for function typedefs. 1138 # 1139 typedef_type = r'typedef((?:\s+[\w*]+\b){0,7}\s+(?:\w+\b|\*+))\s*' 1140 typedef_ident = r'\*?\s*(\w\S+)\s*' 1141 typedef_args = r'\s*\((.*)\);' 1142 1143 source = source 1144 1145 typedef1 = KernRe(typedef_type + r'\(' + typedef_ident + r'\)' + typedef_args) 1146 typedef2 = KernRe(typedef_type + typedef_ident + typedef_args) 1147 1148 # Parse function typedef prototypes 1149 for r in [typedef1, typedef2]: 1150 if not r.match(proto): 1151 continue 1152 1153 return_type = r.group(1).strip() 1154 declaration_name = r.group(2) 1155 args = r.group(3) 1156 1157 if self.entry.identifier != declaration_name: 1158 self.emit_msg(ln, 1159 f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead") 1160 return 1161 1162 self.create_parameter_list(ln, 'function', args, ',', declaration_name) 1163 1164 self.output_declaration('function', declaration_name, 1165 source=source, 1166 typedef=True, 1167 functiontype=return_type, 1168 purpose=self.entry.declaration_purpose) 1169 return 1170 # 1171 # Not a function, try to parse a simple typedef. 1172 # 1173 r = KernRe(r'typedef.*\s+(\w+)\s*;') 1174 if r.match(proto): 1175 declaration_name = r.group(1) 1176 1177 if self.entry.identifier != declaration_name: 1178 self.emit_msg(ln, 1179 f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead") 1180 return 1181 1182 self.output_declaration('typedef', declaration_name, 1183 source=source, 1184 purpose=self.entry.declaration_purpose) 1185 return 1186 1187 self.emit_msg(ln, "error: Cannot parse typedef!") 1188 1189 @staticmethod 1190 def process_export(function_set, line): 1191 """ 1192 process ``EXPORT_SYMBOL*`` tags 1193 1194 This method doesn't use any variable from the class, so declare it 1195 with a staticmethod decorator. 1196 """ 1197 1198 # We support documenting some exported symbols with different 1199 # names. A horrible hack. 1200 suffixes = [ '_noprof' ] 1201 1202 # Note: it accepts only one EXPORT_SYMBOL* per line, as having 1203 # multiple export lines would violate Kernel coding style. 1204 1205 if export_symbol.search(line): 1206 symbol = export_symbol.group(2) 1207 elif export_symbol_ns.search(line): 1208 symbol = export_symbol_ns.group(2) 1209 else: 1210 return False 1211 # 1212 # Found an export, trim out any special suffixes 1213 # 1214 for suffix in suffixes: 1215 # Be backward compatible with Python < 3.9 1216 if symbol.endswith(suffix): 1217 symbol = symbol[:-len(suffix)] 1218 function_set.add(symbol) 1219 return True 1220 1221 def process_normal(self, ln, line, source): 1222 """ 1223 STATE_NORMAL: looking for the ``/**`` to begin everything. 1224 """ 1225 1226 if not doc_start.match(line): 1227 return 1228 1229 # start a new entry 1230 self.reset_state(ln) 1231 1232 # next line is always the function name 1233 self.state = state.NAME 1234 1235 def process_name(self, ln, line, source): 1236 """ 1237 STATE_NAME: Looking for the "name - description" line 1238 """ 1239 # 1240 # Check for a DOC: block and handle them specially. 1241 # 1242 if doc_block.search(line): 1243 1244 if not doc_block.group(1): 1245 self.entry.begin_section(ln, "Introduction") 1246 else: 1247 self.entry.begin_section(ln, doc_block.group(1)) 1248 1249 self.entry.identifier = self.entry.section 1250 self.state = state.DOCBLOCK 1251 # 1252 # Otherwise we're looking for a normal kerneldoc declaration line. 1253 # 1254 elif doc_decl.search(line): 1255 self.entry.identifier = doc_decl.group(1) 1256 1257 # Test for data declaration 1258 if doc_begin_data.search(line): 1259 self.entry.decl_type = doc_begin_data.group(1) 1260 self.entry.identifier = doc_begin_data.group(2) 1261 # 1262 # Look for a function description 1263 # 1264 elif doc_begin_func.search(line): 1265 self.entry.identifier = doc_begin_func.group(1) 1266 self.entry.decl_type = "function" 1267 # 1268 # We struck out. 1269 # 1270 else: 1271 self.emit_msg(ln, 1272 f"This comment starts with '/**', but isn't a kernel-doc comment. Refer to Documentation/doc-guide/kernel-doc.rst\n{line}") 1273 self.state = state.NORMAL 1274 return 1275 # 1276 # OK, set up for a new kerneldoc entry. 1277 # 1278 self.state = state.BODY 1279 self.entry.identifier = self.entry.identifier.strip(" ") 1280 # if there's no @param blocks need to set up default section here 1281 self.entry.begin_section(ln + 1) 1282 # 1283 # Find the description portion, which *should* be there but 1284 # isn't always. 1285 # (We should be able to capture this from the previous parsing - someday) 1286 # 1287 r = KernRe("[-:](.*)") 1288 if r.search(line): 1289 self.entry.declaration_purpose = trim_whitespace(r.group(1)) 1290 self.state = state.DECLARATION 1291 else: 1292 self.entry.declaration_purpose = "" 1293 1294 if not self.entry.declaration_purpose and self.config.wshort_desc: 1295 self.emit_msg(ln, 1296 f"missing initial short description on line:\n{line}") 1297 1298 if not self.entry.identifier and self.entry.decl_type != "enum": 1299 self.emit_msg(ln, 1300 f"wrong kernel-doc identifier on line:\n{line}") 1301 self.state = state.NORMAL 1302 1303 if self.config.verbose: 1304 self.emit_msg(ln, 1305 f"Scanning doc for {self.entry.decl_type} {self.entry.identifier}", 1306 warning=False) 1307 # 1308 # Failed to find an identifier. Emit a warning 1309 # 1310 else: 1311 self.emit_msg(ln, f"Cannot find identifier on line:\n{line}") 1312 1313 def is_new_section(self, ln, line): 1314 """ 1315 Helper function to determine if a new section is being started. 1316 """ 1317 if doc_sect.search(line): 1318 self.state = state.BODY 1319 # 1320 # Pick out the name of our new section, tweaking it if need be. 1321 # 1322 newsection = doc_sect.group(1) 1323 if newsection.lower() == 'description': 1324 newsection = 'Description' 1325 elif newsection.lower() == 'context': 1326 newsection = 'Context' 1327 self.state = state.SPECIAL_SECTION 1328 elif newsection.lower() in ["@return", "@returns", 1329 "return", "returns"]: 1330 newsection = "Return" 1331 self.state = state.SPECIAL_SECTION 1332 elif newsection[0] == '@': 1333 self.state = state.SPECIAL_SECTION 1334 # 1335 # Initialize the contents, and get the new section going. 1336 # 1337 newcontents = doc_sect.group(2) 1338 if not newcontents: 1339 newcontents = "" 1340 self.dump_section() 1341 self.entry.begin_section(ln, newsection) 1342 self.entry.leading_space = None 1343 1344 self.entry.add_text(newcontents.lstrip()) 1345 return True 1346 return False 1347 1348 def is_comment_end(self, ln, line): 1349 """ 1350 Helper function to detect (and effect) the end of a kerneldoc comment. 1351 """ 1352 if doc_end.search(line): 1353 self.dump_section() 1354 1355 # Look for doc_com + <text> + doc_end: 1356 r = KernRe(r'\s*\*\s*[a-zA-Z_0-9:.]+\*/') 1357 if r.match(line): 1358 self.emit_msg(ln, f"suspicious ending line: {line}") 1359 1360 self.entry.prototype = "" 1361 self.entry.new_start_line = ln + 1 1362 1363 self.state = state.PROTO 1364 return True 1365 return False 1366 1367 1368 def process_decl(self, ln, line, source): 1369 """ 1370 STATE_DECLARATION: We've seen the beginning of a declaration. 1371 """ 1372 if self.is_new_section(ln, line) or self.is_comment_end(ln, line): 1373 return 1374 # 1375 # Look for anything with the " * " line beginning. 1376 # 1377 if doc_content.search(line): 1378 cont = doc_content.group(1) 1379 # 1380 # A blank line means that we have moved out of the declaration 1381 # part of the comment (without any "special section" parameter 1382 # descriptions). 1383 # 1384 if cont == "": 1385 self.state = state.BODY 1386 # 1387 # Otherwise we have more of the declaration section to soak up. 1388 # 1389 else: 1390 self.entry.declaration_purpose = \ 1391 trim_whitespace(self.entry.declaration_purpose + ' ' + cont) 1392 else: 1393 # Unknown line, ignore 1394 self.emit_msg(ln, f"bad line: {line}") 1395 1396 1397 def process_special(self, ln, line, source): 1398 """ 1399 STATE_SPECIAL_SECTION: a section ending with a blank line. 1400 """ 1401 # 1402 # If we have hit a blank line (only the " * " marker), then this 1403 # section is done. 1404 # 1405 if KernRe(r"\s*\*\s*$").match(line): 1406 self.entry.begin_section(ln, dump = True) 1407 self.state = state.BODY 1408 return 1409 # 1410 # Not a blank line, look for the other ways to end the section. 1411 # 1412 if self.is_new_section(ln, line) or self.is_comment_end(ln, line): 1413 return 1414 # 1415 # OK, we should have a continuation of the text for this section. 1416 # 1417 if doc_content.search(line): 1418 cont = doc_content.group(1) 1419 # 1420 # If the lines of text after the first in a special section have 1421 # leading white space, we need to trim it out or Sphinx will get 1422 # confused. For the second line (the None case), see what we 1423 # find there and remember it. 1424 # 1425 if self.entry.leading_space is None: 1426 r = KernRe(r'^(\s+)') 1427 if r.match(cont): 1428 self.entry.leading_space = len(r.group(1)) 1429 else: 1430 self.entry.leading_space = 0 1431 # 1432 # Otherwise, before trimming any leading chars, be *sure* 1433 # that they are white space. We should maybe warn if this 1434 # isn't the case. 1435 # 1436 for i in range(0, self.entry.leading_space): 1437 if cont[i] != " ": 1438 self.entry.leading_space = i 1439 break 1440 # 1441 # Add the trimmed result to the section and we're done. 1442 # 1443 self.entry.add_text(cont[self.entry.leading_space:]) 1444 else: 1445 # Unknown line, ignore 1446 self.emit_msg(ln, f"bad line: {line}") 1447 1448 def process_body(self, ln, line, source): 1449 """ 1450 STATE_BODY: the bulk of a kerneldoc comment. 1451 """ 1452 if self.is_new_section(ln, line) or self.is_comment_end(ln, line): 1453 return 1454 1455 if doc_content.search(line): 1456 cont = doc_content.group(1) 1457 self.entry.add_text(cont) 1458 else: 1459 # Unknown line, ignore 1460 self.emit_msg(ln, f"bad line: {line}") 1461 1462 def process_inline_name(self, ln, line, source): 1463 """STATE_INLINE_NAME: beginning of docbook comments within a prototype.""" 1464 1465 if doc_inline_sect.search(line): 1466 self.entry.begin_section(ln, doc_inline_sect.group(1)) 1467 self.entry.add_text(doc_inline_sect.group(2).lstrip()) 1468 self.state = state.INLINE_TEXT 1469 elif doc_inline_end.search(line): 1470 self.dump_section() 1471 self.state = state.PROTO 1472 elif doc_content.search(line): 1473 self.emit_msg(ln, f"Incorrect use of kernel-doc format: {line}") 1474 self.state = state.PROTO 1475 1476 # 1477 # Don't let it add partial comments at the code, as breaks the 1478 # logic meant to remove comments from prototypes. 1479 # 1480 self.process_proto_type(ln, "/**\n" + line, source) 1481 # else ... ?? 1482 1483 def process_inline_text(self, ln, line, source): 1484 """STATE_INLINE_TEXT: docbook comments within a prototype.""" 1485 1486 if doc_inline_end.search(line): 1487 self.dump_section() 1488 self.state = state.PROTO 1489 elif doc_content.search(line): 1490 self.entry.add_text(doc_content.group(1)) 1491 # else ... ?? 1492 1493 def syscall_munge(self, ln, proto): # pylint: disable=W0613 1494 """ 1495 Handle syscall definitions. 1496 """ 1497 1498 is_void = False 1499 1500 # Strip newlines/CR's 1501 proto = re.sub(r'[\r\n]+', ' ', proto) 1502 1503 # Check if it's a SYSCALL_DEFINE0 1504 if 'SYSCALL_DEFINE0' in proto: 1505 is_void = True 1506 1507 # Replace SYSCALL_DEFINE with correct return type & function name 1508 proto = KernRe(r'SYSCALL_DEFINE.*\(').sub('long sys_', proto) 1509 1510 r = KernRe(r'long\s+(sys_.*?),') 1511 if r.search(proto): 1512 proto = KernRe(',').sub('(', proto, count=1) 1513 elif is_void: 1514 proto = KernRe(r'\)').sub('(void)', proto, count=1) 1515 1516 # Now delete all of the odd-numbered commas in the proto 1517 # so that argument types & names don't have a comma between them 1518 count = 0 1519 length = len(proto) 1520 1521 if is_void: 1522 length = 0 # skip the loop if is_void 1523 1524 for ix in range(length): 1525 if proto[ix] == ',': 1526 count += 1 1527 if count % 2 == 1: 1528 proto = proto[:ix] + ' ' + proto[ix + 1:] 1529 1530 return proto 1531 1532 def tracepoint_munge(self, ln, proto): 1533 """ 1534 Handle tracepoint definitions. 1535 """ 1536 1537 tracepointname = None 1538 tracepointargs = None 1539 1540 # Match tracepoint name based on different patterns 1541 r = KernRe(r'TRACE_EVENT\((.*?),') 1542 if r.search(proto): 1543 tracepointname = r.group(1) 1544 1545 r = KernRe(r'DEFINE_SINGLE_EVENT\((.*?),') 1546 if r.search(proto): 1547 tracepointname = r.group(1) 1548 1549 r = KernRe(r'DEFINE_EVENT\((.*?),(.*?),') 1550 if r.search(proto): 1551 tracepointname = r.group(2) 1552 1553 if tracepointname: 1554 tracepointname = tracepointname.lstrip() 1555 1556 r = KernRe(r'TP_PROTO\((.*?)\)') 1557 if r.search(proto): 1558 tracepointargs = r.group(1) 1559 1560 if not tracepointname or not tracepointargs: 1561 self.emit_msg(ln, 1562 f"Unrecognized tracepoint format:\n{proto}\n") 1563 else: 1564 proto = f"static inline void trace_{tracepointname}({tracepointargs})" 1565 self.entry.identifier = f"trace_{self.entry.identifier}" 1566 1567 return proto 1568 1569 def process_proto_function(self, ln, line, source): 1570 """Ancillary routine to process a function prototype.""" 1571 1572 # strip C99-style comments to end of line 1573 line = KernRe(r"//.*$", re.S).sub('', line) 1574 # 1575 # Soak up the line's worth of prototype text, stopping at { or ; if present. 1576 # 1577 if KernRe(r'\s*#\s*define').match(line): 1578 self.entry.prototype = line 1579 elif not line.startswith('#'): # skip other preprocessor stuff 1580 r = KernRe(r'([^\{]*)') 1581 if r.match(line): 1582 self.entry.prototype += r.group(1) + " " 1583 # 1584 # If we now have the whole prototype, clean it up and declare victory. 1585 # 1586 if '{' in line or ';' in line or KernRe(r'\s*#\s*define').match(line): 1587 # strip comments and surrounding spaces 1588 self.entry.prototype = KernRe(r'/\*.*\*/').sub('', self.entry.prototype).strip() 1589 # 1590 # Handle self.entry.prototypes for function pointers like: 1591 # int (*pcs_config)(struct foo) 1592 # by turning it into 1593 # int pcs_config(struct foo) 1594 # 1595 r = KernRe(r'^(\S+\s+)\(\s*\*(\S+)\)') 1596 self.entry.prototype = r.sub(r'\1\2', self.entry.prototype) 1597 # 1598 # Handle special declaration syntaxes 1599 # 1600 if 'SYSCALL_DEFINE' in self.entry.prototype: 1601 self.entry.prototype = self.syscall_munge(ln, 1602 self.entry.prototype) 1603 else: 1604 r = KernRe(r'TRACE_EVENT|DEFINE_EVENT|DEFINE_SINGLE_EVENT') 1605 if r.search(self.entry.prototype): 1606 self.entry.prototype = self.tracepoint_munge(ln, 1607 self.entry.prototype) 1608 # 1609 # ... and we're done 1610 # 1611 self.dump_function(ln, self.entry.prototype, source) 1612 self.reset_state(ln) 1613 1614 def process_proto_type(self, ln, line, source): 1615 """ 1616 Ancillary routine to process a type. 1617 """ 1618 1619 # Strip C99-style comments and surrounding whitespace 1620 line = KernRe(r"//.*$", re.S).sub('', line).strip() 1621 if not line: 1622 return # nothing to see here 1623 1624 # To distinguish preprocessor directive from regular declaration later. 1625 if line.startswith('#'): 1626 line += ";" 1627 # 1628 # Split the declaration on any of { } or ;, and accumulate pieces 1629 # until we hit a semicolon while not inside {brackets} 1630 # 1631 r = KernRe(r'(.*?)([{};])') 1632 for chunk in r.split(line): 1633 if chunk: # Ignore empty matches 1634 self.entry.prototype += chunk 1635 # 1636 # This cries out for a match statement ... someday after we can 1637 # drop Python 3.9 ... 1638 # 1639 if chunk == '{': 1640 self.entry.brcount += 1 1641 elif chunk == '}': 1642 self.entry.brcount -= 1 1643 elif chunk == ';' and self.entry.brcount <= 0: 1644 self.dump_declaration(ln, self.entry.prototype, source) 1645 self.reset_state(ln) 1646 return 1647 # 1648 # We hit the end of the line while still in the declaration; put 1649 # in a space to represent the newline. 1650 # 1651 self.entry.prototype += ' ' 1652 1653 def process_proto(self, ln, line, source): 1654 """STATE_PROTO: reading a function/whatever prototype.""" 1655 1656 if doc_inline_oneline.search(line): 1657 self.entry.begin_section(ln, doc_inline_oneline.group(1)) 1658 self.entry.add_text(doc_inline_oneline.group(2)) 1659 self.dump_section() 1660 1661 elif doc_inline_start.search(line): 1662 self.state = state.INLINE_NAME 1663 1664 elif self.entry.decl_type == 'function': 1665 self.process_proto_function(ln, line, source) 1666 1667 else: 1668 self.process_proto_type(ln, line, source) 1669 1670 def process_docblock(self, ln, line, source): 1671 """STATE_DOCBLOCK: within a ``DOC:`` block.""" 1672 1673 if doc_end.search(line): 1674 self.dump_section() 1675 self.output_declaration("doc", self.entry.identifier, 1676 source=source) 1677 self.reset_state(ln) 1678 1679 elif doc_content.search(line): 1680 self.entry.add_text(doc_content.group(1)) 1681 1682 def parse_export(self): 1683 """ 1684 Parses ``EXPORT_SYMBOL*`` macros from a single Kernel source file. 1685 """ 1686 1687 export_table = set() 1688 1689 try: 1690 with open(self.fname, "r", encoding="utf8", 1691 errors="backslashreplace") as fp: 1692 1693 for line in fp: 1694 self.process_export(export_table, line) 1695 1696 except IOError: 1697 return None 1698 1699 return export_table 1700 1701 #: The state/action table telling us which function to invoke in each state. 1702 state_actions = { 1703 state.NORMAL: process_normal, 1704 state.NAME: process_name, 1705 state.BODY: process_body, 1706 state.DECLARATION: process_decl, 1707 state.SPECIAL_SECTION: process_special, 1708 state.INLINE_NAME: process_inline_name, 1709 state.INLINE_TEXT: process_inline_text, 1710 state.PROTO: process_proto, 1711 state.DOCBLOCK: process_docblock, 1712 } 1713 1714 def parse_kdoc(self): 1715 """ 1716 Open and process each line of a C source file. 1717 The parsing is controlled via a state machine, and the line is passed 1718 to a different process function depending on the state. The process 1719 function may update the state as needed. 1720 1721 Besides parsing kernel-doc tags, it also parses export symbols. 1722 """ 1723 1724 prev = "" 1725 prev_ln = None 1726 export_table = set() 1727 self.state = state.NORMAL 1728 source = "" 1729 1730 try: 1731 with open(self.fname, "r", encoding="utf8", 1732 errors="backslashreplace") as fp: 1733 for ln, line in enumerate(fp): 1734 1735 line = line.expandtabs().strip("\n") 1736 1737 # Group continuation lines on prototypes 1738 if self.state == state.PROTO: 1739 if line.endswith("\\"): 1740 prev += line.rstrip("\\") 1741 if not prev_ln: 1742 prev_ln = ln 1743 continue 1744 1745 if prev: 1746 ln = prev_ln 1747 line = prev + line 1748 prev = "" 1749 prev_ln = None 1750 1751 self.config.log.debug("%d %s: %s", 1752 ln, state.name[self.state], 1753 line) 1754 1755 if self.store_src: 1756 if source and self.state == state.NORMAL: 1757 source = "" 1758 elif self.state != state.NORMAL: 1759 source += line + "\n" 1760 1761 # This is an optimization over the original script. 1762 # There, when export_file was used for the same file, 1763 # it was read twice. Here, we use the already-existing 1764 # loop to parse exported symbols as well. 1765 # 1766 if (self.state != state.NORMAL) or \ 1767 not self.process_export(export_table, line): 1768 prev_state = self.state 1769 # Hand this line to the appropriate state handler 1770 self.state_actions[self.state](self, ln, line, source) 1771 if prev_state == state.NORMAL and self.state != state.NORMAL: 1772 source += line + "\n" 1773 1774 self.emit_unused_warnings() 1775 1776 except OSError: 1777 self.config.log.error(f"Error: Cannot open file {self.fname}") 1778 1779 return export_table, self.entries 1780