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 14from pprint import pformat 15 16from kdoc.c_lex import CTokenizer 17from kdoc.kdoc_re import NestedMatch, KernRe 18from kdoc.kdoc_item import KdocItem 19 20# 21# Regular expressions used to parse kernel-doc markups at KernelDoc class. 22# 23# Let's declare them in lowercase outside any class to make it easier to 24# convert from the Perl script. 25# 26# As those are evaluated at the beginning, no need to cache them 27# 28 29# Allow whitespace at end of comment start. 30doc_start = KernRe(r'^/\*\*\s*$', cache=False) 31 32doc_end = KernRe(r'\*/', cache=False) 33doc_com = KernRe(r'\s*\*\s*', cache=False) 34doc_com_body = KernRe(r'\s*\* ?', cache=False) 35doc_decl = doc_com + KernRe(r'(\w+)', cache=False) 36 37# @params and a strictly limited set of supported section names 38# Specifically: 39# Match @word: 40# @...: 41# @{section-name}: 42# while trying to not match literal block starts like "example::" 43# 44known_section_names = 'description|context|returns?|notes?|examples?' 45known_sections = KernRe(known_section_names, flags = re.I) 46doc_sect = doc_com + \ 47 KernRe(r'\s*(@[.\w]+|@\.\.\.|' + known_section_names + r')\s*:([^:].*)?$', 48 flags=re.I, cache=False) 49 50doc_content = doc_com_body + KernRe(r'(.*)', cache=False) 51doc_inline_start = KernRe(r'^\s*/\*\*\s*$', cache=False) 52doc_inline_sect = KernRe(r'\s*\*\s*(@\s*[\w][\w\.]*\s*):(.*)', cache=False) 53doc_inline_end = KernRe(r'^\s*\*/\s*$', cache=False) 54doc_inline_oneline = KernRe(r'^\s*/\*\*\s*(@\s*[\w][\w\.]*\s*):\s*(.*)\s*\*/\s*$', cache=False) 55 56export_symbol = KernRe(r'^\s*EXPORT_SYMBOL(_GPL)?\s*\(\s*(\w+)\s*\)\s*', cache=False) 57export_symbol_ns = KernRe(r'^\s*EXPORT_SYMBOL_NS(_GPL)?\s*\(\s*(\w+)\s*,\s*"\S+"\)\s*', cache=False) 58 59type_param = KernRe(r"@(\w*((\.\w+)|(->\w+))*(\.\.\.)?)", cache=False) 60 61# 62# Tests for the beginning of a kerneldoc block in its various forms. 63# 64doc_block = doc_com + KernRe(r'DOC:\s*(.*)?', cache=False) 65doc_begin_data = KernRe(r"^\s*\*?\s*(struct|union|enum|typedef|var)\b\s*(\w*)", cache = False) 66doc_begin_func = KernRe(str(doc_com) + # initial " * ' 67 r"(?:\w+\s*\*\s*)?" + # type (not captured) 68 r'(?:define\s+)?' + # possible "define" (not captured) 69 r'(\w+)\s*(?:\(\w*\))?\s*' + # name and optional "(...)" 70 r'(?:[-:].*)?$', # description (not captured) 71 cache = False) 72 73# 74# Ancillary functions 75# 76 77multi_space = KernRe(r'\s\s+') 78def trim_whitespace(s): 79 """ 80 A little helper to get rid of excess white space. 81 """ 82 return multi_space.sub(' ', s.strip()) 83 84def trim_private_members(text): 85 """ 86 Remove ``struct``/``enum`` members that have been marked "private". 87 """ 88 89 tokens = CTokenizer(text) 90 return str(tokens) 91 92class state: 93 """ 94 States used by the parser's state machine. 95 """ 96 97 # Parser states 98 NORMAL = 0 #: Normal code. 99 NAME = 1 #: Looking for function name. 100 DECLARATION = 2 #: We have seen a declaration which might not be done. 101 BODY = 3 #: The body of the comment. 102 SPECIAL_SECTION = 4 #: Doc section ending with a blank line. 103 PROTO = 5 #: Scanning prototype. 104 DOCBLOCK = 6 #: Documentation block. 105 INLINE_NAME = 7 #: Gathering doc outside main block. 106 INLINE_TEXT = 8 #: Reading the body of inline docs. 107 108 #: Names for each parser state. 109 name = [ 110 "NORMAL", 111 "NAME", 112 "DECLARATION", 113 "BODY", 114 "SPECIAL_SECTION", 115 "PROTO", 116 "DOCBLOCK", 117 "INLINE_NAME", 118 "INLINE_TEXT", 119 ] 120 121 122SECTION_DEFAULT = "Description" #: Default section. 123 124class KernelEntry: 125 """ 126 Encapsulates a Kernel documentation entry. 127 """ 128 129 def __init__(self, config, fname, ln): 130 self.config = config 131 self.fname = fname 132 133 self._contents = [] 134 self.prototype = "" 135 136 self.warnings = [] 137 138 self.parameterlist = [] 139 self.parameterdescs = {} 140 self.parametertypes = {} 141 self.parameterdesc_start_lines = {} 142 143 self.section_start_lines = {} 144 self.sections = {} 145 146 self.anon_struct_union = False 147 148 self.leading_space = None 149 150 self.fname = fname 151 152 # State flags 153 self.brcount = 0 154 self.declaration_start_line = ln + 1 155 156 # 157 # Management of section contents 158 # 159 def add_text(self, text): 160 """Add a new text to the entry contents list.""" 161 self._contents.append(text) 162 163 def contents(self): 164 """Returns a string with all content texts that were added.""" 165 return '\n'.join(self._contents) + '\n' 166 167 # TODO: rename to emit_message after removal of kernel-doc.pl 168 def emit_msg(self, ln, msg, *, warning=True): 169 """Emit a message.""" 170 171 log_msg = f"{self.fname}:{ln} {msg}" 172 173 if not warning: 174 self.config.log.info(log_msg) 175 return 176 177 # Delegate warning output to output logic, as this way it 178 # will report warnings/info only for symbols that are output 179 180 self.warnings.append(log_msg) 181 return 182 183 def begin_section(self, line_no, title = SECTION_DEFAULT, dump = False): 184 """ 185 Begin a new section. 186 """ 187 if dump: 188 self.dump_section(start_new = True) 189 self.section = title 190 self.new_start_line = line_no 191 192 def dump_section(self, start_new=True): 193 """ 194 Dumps section contents to arrays/hashes intended for that purpose. 195 """ 196 # 197 # If we have accumulated no contents in the default ("description") 198 # section, don't bother. 199 # 200 if self.section == SECTION_DEFAULT and not self._contents: 201 return 202 name = self.section 203 contents = self.contents() 204 205 if type_param.match(name): 206 name = type_param.group(1) 207 208 self.parameterdescs[name] = contents 209 self.parameterdesc_start_lines[name] = self.new_start_line 210 211 self.new_start_line = 0 212 213 else: 214 if name in self.sections and self.sections[name] != "": 215 # Only warn on user-specified duplicate section names 216 if name != SECTION_DEFAULT: 217 self.emit_msg(self.new_start_line, 218 f"duplicate section name '{name}'") 219 # Treat as a new paragraph - add a blank line 220 self.sections[name] += '\n' + contents 221 else: 222 self.sections[name] = contents 223 self.section_start_lines[name] = self.new_start_line 224 self.new_start_line = 0 225 226# self.config.log.debug("Section: %s : %s", name, pformat(vars(self))) 227 228 if start_new: 229 self.section = SECTION_DEFAULT 230 self._contents = [] 231 232python_warning = False 233 234class KernelDoc: 235 """ 236 Read a C language source or header FILE and extract embedded 237 documentation comments. 238 """ 239 240 #: Name of context section. 241 section_context = "Context" 242 243 #: Name of return section. 244 section_return = "Return" 245 246 #: String to write when a parameter is not described. 247 undescribed = "-- undescribed --" 248 249 def __init__(self, config, fname, xforms): 250 """Initialize internal variables""" 251 252 self.fname = fname 253 self.config = config 254 self.xforms = xforms 255 256 # Initial state for the state machines 257 self.state = state.NORMAL 258 259 # Store entry currently being processed 260 self.entry = None 261 262 # Place all potential outputs into an array 263 self.entries = [] 264 265 # 266 # We need Python 3.7 for its "dicts remember the insertion 267 # order" guarantee 268 # 269 global python_warning 270 if (not python_warning and 271 sys.version_info.major == 3 and sys.version_info.minor < 7): 272 273 self.emit_msg(0, 274 'Python 3.7 or later is required for correct results') 275 python_warning = True 276 277 def emit_msg(self, ln, msg, *, warning=True): 278 """Emit a message""" 279 280 if self.entry: 281 self.entry.emit_msg(ln, msg, warning=warning) 282 return 283 284 log_msg = f"{self.fname}:{ln} {msg}" 285 286 if warning: 287 self.config.log.warning(log_msg) 288 else: 289 self.config.log.info(log_msg) 290 291 def dump_section(self, start_new=True): 292 """ 293 Dump section contents to arrays/hashes intended for that purpose. 294 """ 295 296 if self.entry: 297 self.entry.dump_section(start_new) 298 299 # TODO: rename it to store_declaration after removal of kernel-doc.pl 300 def output_declaration(self, dtype, name, **args): 301 """ 302 Store the entry into an entry array. 303 304 The actual output and output filters will be handled elsewhere. 305 """ 306 307 item = KdocItem(name, self.fname, dtype, 308 self.entry.declaration_start_line, **args) 309 item.warnings = self.entry.warnings 310 311 # Drop empty sections 312 # TODO: improve empty sections logic to emit warnings 313 sections = self.entry.sections 314 for section in ["Description", "Return"]: 315 if section in sections and not sections[section].rstrip(): 316 del sections[section] 317 item.set_sections(sections, self.entry.section_start_lines) 318 item.set_params(self.entry.parameterlist, self.entry.parameterdescs, 319 self.entry.parametertypes, 320 self.entry.parameterdesc_start_lines) 321 self.entries.append(item) 322 323 self.config.log.debug("Output: %s:%s = %s", dtype, name, pformat(args)) 324 325 def emit_unused_warnings(self): 326 """ 327 When the parser fails to produce a valid entry, it places some 328 warnings under `entry.warnings` that will be discarded when resetting 329 the state. 330 331 Ensure that those warnings are not lost. 332 333 .. note:: 334 335 Because we are calling `config.warning()` here, those 336 warnings are not filtered by the `-W` parameters: they will all 337 be produced even when `-Wreturn`, `-Wshort-desc`, and/or 338 `-Wcontents-before-sections` are used. 339 340 Allowing those warnings to be filtered is complex, because it 341 would require storing them in a buffer and then filtering them 342 during the output step of the code, depending on the 343 selected symbols. 344 """ 345 if self.entry and self.entry not in self.entries: 346 for log_msg in self.entry.warnings: 347 self.config.warning(log_msg) 348 349 def reset_state(self, ln): 350 """ 351 Ancillary routine to create a new entry. It initializes all 352 variables used by the state machine. 353 """ 354 355 self.emit_unused_warnings() 356 357 self.entry = KernelEntry(self.config, self.fname, ln) 358 359 # State flags 360 self.state = state.NORMAL 361 362 def push_parameter(self, ln, decl_type, param, dtype, 363 org_arg, declaration_name): 364 """ 365 Store parameters and their descriptions at self.entry. 366 """ 367 368 if self.entry.anon_struct_union and dtype == "" and param == "}": 369 return # Ignore the ending }; from anonymous struct/union 370 371 self.entry.anon_struct_union = False 372 373 param = KernRe(r'[\[\)].*').sub('', param, count=1) 374 375 # 376 # Look at various "anonymous type" cases. 377 # 378 if dtype == '': 379 if param.endswith("..."): 380 if len(param) > 3: # there is a name provided, use that 381 param = param[:-3] 382 if not self.entry.parameterdescs.get(param): 383 self.entry.parameterdescs[param] = "variable arguments" 384 385 elif (not param) or param == "void": 386 param = "void" 387 self.entry.parameterdescs[param] = "no arguments" 388 389 elif param in ["struct", "union"]: 390 # Handle unnamed (anonymous) union or struct 391 dtype = param 392 param = "{unnamed_" + param + "}" 393 self.entry.parameterdescs[param] = "anonymous\n" 394 self.entry.anon_struct_union = True 395 396 # Warn if parameter has no description 397 # (but ignore ones starting with # as these are not parameters 398 # but inline preprocessor statements) 399 if param not in self.entry.parameterdescs and not param.startswith("#"): 400 self.entry.parameterdescs[param] = self.undescribed 401 402 if "." not in param: 403 if decl_type == 'function': 404 dname = f"{decl_type} parameter" 405 else: 406 dname = f"{decl_type} member" 407 408 self.emit_msg(ln, 409 f"{dname} '{param}' not described in '{declaration_name}'") 410 411 # Strip spaces from param so that it is one continuous string on 412 # parameterlist. This fixes a problem where check_sections() 413 # cannot find a parameter like "addr[6 + 2]" because it actually 414 # appears as "addr[6", "+", "2]" on the parameter list. 415 # However, it's better to maintain the param string unchanged for 416 # output, so just weaken the string compare in check_sections() 417 # to ignore "[blah" in a parameter string. 418 419 self.entry.parameterlist.append(param) 420 org_arg = KernRe(r'\s\s+').sub(' ', org_arg) 421 self.entry.parametertypes[param] = org_arg 422 423 424 def create_parameter_list(self, ln, decl_type, args, 425 splitter, declaration_name): 426 """ 427 Creates a list of parameters, storing them at self.entry. 428 """ 429 430 # temporarily replace all commas inside function pointer definition 431 arg_expr = KernRe(r'(\([^\),]+),') 432 while arg_expr.search(args): 433 args = arg_expr.sub(r"\1#", args) 434 435 for arg in args.split(splitter): 436 # Ignore argument attributes 437 arg = KernRe(r'\sPOS0?\s').sub(' ', arg) 438 439 # Strip leading/trailing spaces 440 arg = arg.strip() 441 arg = KernRe(r'\s+').sub(' ', arg, count=1) 442 443 if arg.startswith('#'): 444 # Treat preprocessor directive as a typeless variable just to fill 445 # corresponding data structures "correctly". Catch it later in 446 # output_* subs. 447 448 # Treat preprocessor directive as a typeless variable 449 self.push_parameter(ln, decl_type, arg, "", 450 "", declaration_name) 451 # 452 # The pointer-to-function case. 453 # 454 elif KernRe(r'\(.+\)\s*\(').search(arg): 455 arg = arg.replace('#', ',') 456 r = KernRe(r'[^\(]+\(\*?\s*' # Everything up to "(*" 457 r'([\w\[\].]*)' # Capture the name and possible [array] 458 r'\s*\)') # Make sure the trailing ")" is there 459 if r.match(arg): 460 param = r.group(1) 461 else: 462 self.emit_msg(ln, f"Invalid param: {arg}") 463 param = arg 464 dtype = arg.replace(param, '') 465 self.push_parameter(ln, decl_type, param, dtype, arg, declaration_name) 466 # 467 # The array-of-pointers case. Dig the parameter name out from the middle 468 # of the declaration. 469 # 470 elif KernRe(r'\(.+\)\s*\[').search(arg): 471 r = KernRe(r'[^\(]+\(\s*\*\s*' # Up to "(" and maybe "*" 472 r'([\w.]*?)' # The actual pointer name 473 r'\s*(\[\s*\w+\s*\]\s*)*\)') # The [array portion] 474 if r.match(arg): 475 param = r.group(1) 476 else: 477 self.emit_msg(ln, f"Invalid param: {arg}") 478 param = arg 479 dtype = arg.replace(param, '') 480 self.push_parameter(ln, decl_type, param, dtype, arg, declaration_name) 481 elif arg: 482 # 483 # Clean up extraneous spaces and split the string at commas; the first 484 # element of the resulting list will also include the type information. 485 # 486 arg = KernRe(r'\s*:\s*').sub(":", arg) 487 arg = KernRe(r'\s*\[').sub('[', arg) 488 args = KernRe(r'\s*,\s*').split(arg) 489 args[0] = re.sub(r'(\*+)\s*', r' \1', args[0]) 490 # 491 # args[0] has a string of "type a". If "a" includes an [array] 492 # declaration, we want to not be fooled by any white space inside 493 # the brackets, so detect and handle that case specially. 494 # 495 r = KernRe(r'^([^[\]]*\s+)(.*)$') 496 if r.match(args[0]): 497 args[0] = r.group(2) 498 dtype = r.group(1) 499 else: 500 # No space in args[0]; this seems wrong but preserves previous behavior 501 dtype = '' 502 503 bitfield_re = KernRe(r'(.*?):(\w+)') 504 for param in args: 505 # 506 # For pointers, shift the star(s) from the variable name to the 507 # type declaration. 508 # 509 r = KernRe(r'^(\*+)\s*(.*)') 510 if r.match(param): 511 self.push_parameter(ln, decl_type, r.group(2), 512 f"{dtype} {r.group(1)}", 513 arg, declaration_name) 514 # 515 # Perform a similar shift for bitfields. 516 # 517 elif bitfield_re.search(param): 518 if dtype != "": # Skip unnamed bit-fields 519 self.push_parameter(ln, decl_type, bitfield_re.group(1), 520 f"{dtype}:{bitfield_re.group(2)}", 521 arg, declaration_name) 522 else: 523 self.push_parameter(ln, decl_type, param, dtype, 524 arg, declaration_name) 525 526 def check_sections(self, ln, decl_name, decl_type): 527 """ 528 Check for errors inside sections, emitting warnings if not found 529 parameters are described. 530 """ 531 for section in self.entry.sections: 532 if section not in self.entry.parameterlist and \ 533 not known_sections.search(section): 534 if decl_type == 'function': 535 dname = f"{decl_type} parameter" 536 else: 537 dname = f"{decl_type} member" 538 self.emit_msg(ln, 539 f"Excess {dname} '{section}' description in '{decl_name}'") 540 541 def check_return_section(self, ln, declaration_name, return_type): 542 """ 543 If the function doesn't return void, warns about the lack of a 544 return description. 545 """ 546 547 if not self.config.wreturn: 548 return 549 550 # Ignore an empty return type (It's a macro) 551 # Ignore functions with a "void" return type (but not "void *") 552 if not return_type or KernRe(r'void\s*\w*\s*$').search(return_type): 553 return 554 555 if not self.entry.sections.get("Return", None): 556 self.emit_msg(ln, 557 f"No description found for return value of '{declaration_name}'") 558 559 def split_struct_proto(self, proto): 560 """ 561 Split apart a structure prototype; returns (struct|union, name, 562 members) or ``None``. 563 """ 564 565 type_pattern = r'(struct|union)' 566 qualifiers = [ 567 "__attribute__", 568 "__packed", 569 "__aligned", 570 "____cacheline_aligned_in_smp", 571 "____cacheline_aligned", 572 ] 573 definition_body = r'\{(.*)\}\s*' + "(?:" + '|'.join(qualifiers) + ")?" 574 575 r = KernRe(type_pattern + r'\s+(\w+)\s*' + definition_body) 576 if r.search(proto): 577 return (r.group(1), r.group(2), r.group(3)) 578 else: 579 r = KernRe(r'typedef\s+' + type_pattern + r'\s*' + definition_body + r'\s*(\w+)\s*;') 580 if r.search(proto): 581 return (r.group(1), r.group(3), r.group(2)) 582 return None 583 584 def rewrite_struct_members(self, members): 585 """ 586 Process ``struct``/``union`` members from the most deeply nested 587 outward. 588 589 Rewrite the members of a ``struct`` or ``union`` for easier formatting 590 later on. Among other things, this function will turn a member like:: 591 592 struct { inner_members; } foo; 593 594 into:: 595 596 struct foo; inner_members; 597 """ 598 599 # 600 # The trick is in the ``^{`` below - it prevents a match of an outer 601 # ``struct``/``union`` until the inner one has been munged 602 # (removing the ``{`` in the process). 603 # 604 struct_members = KernRe(r'(struct|union)' # 0: declaration type 605 r'([^\{\};]+)' # 1: possible name 606 r'(\{)' 607 r'([^\{\}]*)' # 3: Contents of declaration 608 r'(\})' 609 r'([^\{\};]*)(;)') # 5: Remaining stuff after declaration 610 tuples = struct_members.findall(members) 611 while tuples: 612 for t in tuples: 613 newmember = "" 614 oldmember = "".join(t) # Reconstruct the original formatting 615 dtype, name, lbr, content, rbr, rest, semi = t 616 # 617 # Pass through each field name, normalizing the form and formatting. 618 # 619 for s_id in rest.split(','): 620 s_id = s_id.strip() 621 newmember += f"{dtype} {s_id}; " 622 # 623 # Remove bitfield/array/pointer info, getting the bare name. 624 # 625 s_id = KernRe(r'[:\[].*').sub('', s_id) 626 s_id = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', s_id) 627 # 628 # Pass through the members of this inner structure/union. 629 # 630 for arg in content.split(';'): 631 arg = arg.strip() 632 # 633 # Look for (type)(*name)(args) - pointer to function 634 # 635 r = KernRe(r'^([^\(]+\(\*?\s*)([\w.]*)(\s*\).*)') 636 if r.match(arg): 637 dtype, name, extra = r.group(1), r.group(2), r.group(3) 638 # Pointer-to-function 639 if not s_id: 640 # Anonymous struct/union 641 newmember += f"{dtype}{name}{extra}; " 642 else: 643 newmember += f"{dtype}{s_id}.{name}{extra}; " 644 # 645 # Otherwise a non-function member. 646 # 647 else: 648 # 649 # Remove bitmap and array portions and spaces around commas 650 # 651 arg = KernRe(r':\s*\d+\s*').sub('', arg) 652 arg = KernRe(r'\[.*\]').sub('', arg) 653 arg = KernRe(r'\s*,\s*').sub(',', arg) 654 # 655 # Look for a normal decl - "type name[,name...]" 656 # 657 r = KernRe(r'(.*)\s+([\S+,]+)') 658 if r.search(arg): 659 for name in r.group(2).split(','): 660 name = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', name) 661 if not s_id: 662 # Anonymous struct/union 663 newmember += f"{r.group(1)} {name}; " 664 else: 665 newmember += f"{r.group(1)} {s_id}.{name}; " 666 else: 667 newmember += f"{arg}; " 668 # 669 # At the end of the s_id loop, replace the original declaration with 670 # the munged version. 671 # 672 members = members.replace(oldmember, newmember) 673 # 674 # End of the tuple loop - search again and see if there are outer members 675 # that now turn up. 676 # 677 tuples = struct_members.findall(members) 678 return members 679 680 def format_struct_decl(self, declaration): 681 """ 682 Format the ``struct`` declaration into a standard form for inclusion 683 in the resulting docs. 684 """ 685 686 # 687 # Insert newlines, get rid of extra spaces. 688 # 689 declaration = KernRe(r'([\{;])').sub(r'\1\n', declaration) 690 declaration = KernRe(r'\}\s+;').sub('};', declaration) 691 # 692 # Format inline enums with each member on its own line. 693 # 694 r = KernRe(r'(enum\s+\{[^\}]+),([^\n])') 695 while r.search(declaration): 696 declaration = r.sub(r'\1,\n\2', declaration) 697 # 698 # Now go through and supply the right number of tabs 699 # for each line. 700 # 701 def_args = declaration.split('\n') 702 level = 1 703 declaration = "" 704 for clause in def_args: 705 clause = KernRe(r'\s+').sub(' ', clause.strip(), count=1) 706 if clause: 707 if '}' in clause and level > 1: 708 level -= 1 709 if not clause.startswith('#'): 710 declaration += "\t" * level 711 declaration += "\t" + clause + "\n" 712 if "{" in clause and "}" not in clause: 713 level += 1 714 return declaration 715 716 717 def dump_struct(self, ln, proto): 718 """ 719 Store an entry for a ``struct`` or ``union`` 720 """ 721 # 722 # Do the basic parse to get the pieces of the declaration. 723 # 724 struct_parts = self.split_struct_proto(proto) 725 if not struct_parts: 726 self.emit_msg(ln, f"{proto} error: Cannot parse struct or union!") 727 return 728 decl_type, declaration_name, members = struct_parts 729 730 if self.entry.identifier != declaration_name: 731 self.emit_msg(ln, f"expecting prototype for {decl_type} {self.entry.identifier}. " 732 f"Prototype was for {decl_type} {declaration_name} instead\n") 733 return 734 # 735 # Go through the list of members applying all of our transformations. 736 # 737 members = trim_private_members(members) 738 members = self.xforms.apply("struct", members) 739 740 # 741 # Deal with embedded struct and union members, and drop enums entirely. 742 # 743 declaration = members 744 members = self.rewrite_struct_members(members) 745 members = re.sub(r'(\{[^\{\}]*\})', '', members) 746 # 747 # Output the result and we are done. 748 # 749 self.create_parameter_list(ln, decl_type, members, ';', 750 declaration_name) 751 self.check_sections(ln, declaration_name, decl_type) 752 self.output_declaration(decl_type, declaration_name, 753 definition=self.format_struct_decl(declaration), 754 purpose=self.entry.declaration_purpose) 755 756 def dump_enum(self, ln, proto): 757 """ 758 Store an ``enum`` inside self.entries array. 759 """ 760 # 761 # Strip preprocessor directives. Note that this depends on the 762 # trailing semicolon we added in process_proto_type(). 763 # 764 proto = KernRe(r'#\s*((define|ifdef|if)\s+|endif)[^;]*;', flags=re.S).sub('', proto) 765 # 766 # Parse out the name and members of the enum. Typedef form first. 767 # 768 r = KernRe(r'typedef\s+enum\s*\{(.*)\}\s*(\w*)\s*;') 769 if r.search(proto): 770 declaration_name = r.group(2) 771 members = trim_private_members(r.group(1)) 772 # 773 # Failing that, look for a straight enum 774 # 775 else: 776 r = KernRe(r'enum\s+(\w*)\s*\{(.*)\}') 777 if r.match(proto): 778 declaration_name = r.group(1) 779 members = trim_private_members(r.group(2)) 780 # 781 # OK, this isn't going to work. 782 # 783 else: 784 self.emit_msg(ln, f"{proto}: error: Cannot parse enum!") 785 return 786 # 787 # Make sure we found what we were expecting. 788 # 789 if self.entry.identifier != declaration_name: 790 if self.entry.identifier == "": 791 self.emit_msg(ln, 792 f"{proto}: wrong kernel-doc identifier on prototype") 793 else: 794 self.emit_msg(ln, 795 f"expecting prototype for enum {self.entry.identifier}. " 796 f"Prototype was for enum {declaration_name} instead") 797 return 798 799 if not declaration_name: 800 declaration_name = "(anonymous)" 801 # 802 # Parse out the name of each enum member, and verify that we 803 # have a description for it. 804 # 805 member_set = set() 806 members = KernRe(r'\([^;)]*\)').sub('', members) 807 for arg in members.split(','): 808 arg = KernRe(r'^\s*(\w+).*').sub(r'\1', arg) 809 if not arg.strip(): 810 continue 811 812 self.entry.parameterlist.append(arg) 813 if arg not in self.entry.parameterdescs: 814 self.entry.parameterdescs[arg] = self.undescribed 815 self.emit_msg(ln, 816 f"Enum value '{arg}' not described in enum '{declaration_name}'") 817 member_set.add(arg) 818 # 819 # Ensure that every described member actually exists in the enum. 820 # 821 for k in self.entry.parameterdescs: 822 if k not in member_set: 823 self.emit_msg(ln, 824 f"Excess enum value '@{k}' description in '{declaration_name}'") 825 826 self.output_declaration('enum', declaration_name, 827 purpose=self.entry.declaration_purpose) 828 829 def dump_var(self, ln, proto): 830 """ 831 Store variables that are part of kAPI. 832 """ 833 VAR_ATTRIBS = [ 834 "extern", 835 "const", 836 ] 837 OPTIONAL_VAR_ATTR = r"^(?:\b(?:" +"|".join(VAR_ATTRIBS) +r")\b\s*)*" 838 839 # 840 # Store the full prototype before modifying it 841 # 842 full_proto = proto 843 declaration_name = None 844 845 # 846 # Handle macro definitions 847 # 848 macro_prefixes = [ 849 KernRe(r"DEFINE_[\w_]+\s*\(([\w_]+)\)"), 850 ] 851 852 for r in macro_prefixes: 853 match = r.search(proto) 854 if match: 855 declaration_name = match.group(1) 856 break 857 858 # 859 # Drop comments and macros to have a pure C prototype 860 # 861 if not declaration_name: 862 proto = self.xforms.apply("var", proto) 863 864 proto = proto.rstrip() 865 866 # 867 # Variable name is at the end of the declaration 868 # 869 870 default_val = None 871 872 r= KernRe(OPTIONAL_VAR_ATTR + r"\s*[\w_\s]*\s+(?:\*+)?([\w_]+)\s*[\d\]\[]*\s*(=.*)?") 873 if r.match(proto): 874 if not declaration_name: 875 declaration_name = r.group(1) 876 877 default_val = r.group(2) 878 else: 879 r= KernRe(OPTIONAL_VAR_ATTR + r"(?:[\w_\s]*)?\s+(?:\*+)?(?:[\w_]+)\s*[\d\]\[]*\s*(=.*)?") 880 881 if r.match(proto): 882 default_val = r.group(1) 883 if not declaration_name: 884 self.emit_msg(ln,f"{proto}: can't parse variable") 885 return 886 887 if default_val: 888 default_val = default_val.lstrip("=").strip() 889 890 self.output_declaration("var", declaration_name, 891 full_proto=full_proto, 892 default_val=default_val, 893 purpose=self.entry.declaration_purpose) 894 895 def dump_declaration(self, ln, prototype): 896 """ 897 Store a data declaration inside self.entries array. 898 """ 899 900 if self.entry.decl_type == "enum": 901 self.dump_enum(ln, prototype) 902 elif self.entry.decl_type == "typedef": 903 self.dump_typedef(ln, prototype) 904 elif self.entry.decl_type in ["union", "struct"]: 905 self.dump_struct(ln, prototype) 906 elif self.entry.decl_type == "var": 907 self.dump_var(ln, prototype) 908 else: 909 # This would be a bug 910 self.emit_message(ln, f'Unknown declaration type: {self.entry.decl_type}') 911 912 def dump_function(self, ln, prototype): 913 """ 914 Store a function or function macro inside self.entries array. 915 """ 916 917 found = func_macro = False 918 return_type = '' 919 decl_type = 'function' 920 921 # 922 # If we have a macro, remove the "#define" at the front. 923 # 924 new_proto = KernRe(r"^#\s*define\s+").sub("", prototype) 925 if new_proto != prototype: 926 prototype = new_proto 927 # 928 # Dispense with the simple "#define A B" case here; the key 929 # is the space after the name of the symbol being defined. 930 # NOTE that the seemingly misnamed "func_macro" indicates a 931 # macro *without* arguments. 932 # 933 r = KernRe(r'^(\w+)\s+') 934 if r.search(prototype): 935 return_type = '' 936 declaration_name = r.group(1) 937 func_macro = True 938 found = True 939 else: 940 # 941 # Apply the initial transformations. 942 # 943 prototype = self.xforms.apply("func", prototype) 944 945 # Yes, this truly is vile. We are looking for: 946 # 1. Return type (may be nothing if we're looking at a macro) 947 # 2. Function name 948 # 3. Function parameters. 949 # 950 # All the while we have to watch out for function pointer parameters 951 # (which IIRC is what the two sections are for), C types (these 952 # regexps don't even start to express all the possibilities), and 953 # so on. 954 # 955 # If you mess with these regexps, it's a good idea to check that 956 # the following functions' documentation still comes out right: 957 # - parport_register_device (function pointer parameters) 958 # - atomic_set (macro) 959 # - pci_match_device, __copy_to_user (long return type) 960 961 name = r'\w+' 962 type1 = r'(?:[\w\s]+)?' 963 type2 = r'(?:[\w\s]+\*+)+' 964 # 965 # Attempt to match first on (args) with no internal parentheses; this 966 # lets us easily filter out __acquires() and other post-args stuff. If 967 # that fails, just grab the rest of the line to the last closing 968 # parenthesis. 969 # 970 proto_args = r'\(([^\(]*|.*)\)' 971 # 972 # (Except for the simple macro case) attempt to split up the prototype 973 # in the various ways we understand. 974 # 975 if not found: 976 patterns = [ 977 rf'^()({name})\s*{proto_args}', 978 rf'^({type1})\s+({name})\s*{proto_args}', 979 rf'^({type2})\s*({name})\s*{proto_args}', 980 ] 981 982 for p in patterns: 983 r = KernRe(p) 984 if r.match(prototype): 985 return_type = r.group(1) 986 declaration_name = r.group(2) 987 args = r.group(3) 988 self.create_parameter_list(ln, decl_type, args, ',', 989 declaration_name) 990 found = True 991 break 992 # 993 # Parsing done; make sure that things are as we expect. 994 # 995 if not found: 996 self.emit_msg(ln, 997 f"cannot understand function prototype: '{prototype}'") 998 return 999 if self.entry.identifier != declaration_name: 1000 self.emit_msg(ln, f"expecting prototype for {self.entry.identifier}(). " 1001 f"Prototype was for {declaration_name}() instead") 1002 return 1003 self.check_sections(ln, declaration_name, "function") 1004 self.check_return_section(ln, declaration_name, return_type) 1005 # 1006 # Store the result. 1007 # 1008 self.output_declaration(decl_type, declaration_name, 1009 typedef=('typedef' in return_type), 1010 functiontype=return_type, 1011 purpose=self.entry.declaration_purpose, 1012 func_macro=func_macro) 1013 1014 1015 def dump_typedef(self, ln, proto): 1016 """ 1017 Store a ``typedef`` inside self.entries array. 1018 """ 1019 # 1020 # We start by looking for function typedefs. 1021 # 1022 typedef_type = r'typedef((?:\s+[\w*]+\b){0,7}\s+(?:\w+\b|\*+))\s*' 1023 typedef_ident = r'\*?\s*(\w\S+)\s*' 1024 typedef_args = r'\s*\((.*)\);' 1025 1026 typedef1 = KernRe(typedef_type + r'\(' + typedef_ident + r'\)' + typedef_args) 1027 typedef2 = KernRe(typedef_type + typedef_ident + typedef_args) 1028 1029 # Parse function typedef prototypes 1030 for r in [typedef1, typedef2]: 1031 if not r.match(proto): 1032 continue 1033 1034 return_type = r.group(1).strip() 1035 declaration_name = r.group(2) 1036 args = r.group(3) 1037 1038 if self.entry.identifier != declaration_name: 1039 self.emit_msg(ln, 1040 f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n") 1041 return 1042 1043 self.create_parameter_list(ln, 'function', args, ',', declaration_name) 1044 1045 self.output_declaration('function', declaration_name, 1046 typedef=True, 1047 functiontype=return_type, 1048 purpose=self.entry.declaration_purpose) 1049 return 1050 # 1051 # Not a function, try to parse a simple typedef. 1052 # 1053 r = KernRe(r'typedef.*\s+(\w+)\s*;') 1054 if r.match(proto): 1055 declaration_name = r.group(1) 1056 1057 if self.entry.identifier != declaration_name: 1058 self.emit_msg(ln, 1059 f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n") 1060 return 1061 1062 self.output_declaration('typedef', declaration_name, 1063 purpose=self.entry.declaration_purpose) 1064 return 1065 1066 self.emit_msg(ln, "error: Cannot parse typedef!") 1067 1068 @staticmethod 1069 def process_export(function_set, line): 1070 """ 1071 process ``EXPORT_SYMBOL*`` tags 1072 1073 This method doesn't use any variable from the class, so declare it 1074 with a staticmethod decorator. 1075 """ 1076 1077 # We support documenting some exported symbols with different 1078 # names. A horrible hack. 1079 suffixes = [ '_noprof' ] 1080 1081 # Note: it accepts only one EXPORT_SYMBOL* per line, as having 1082 # multiple export lines would violate Kernel coding style. 1083 1084 if export_symbol.search(line): 1085 symbol = export_symbol.group(2) 1086 elif export_symbol_ns.search(line): 1087 symbol = export_symbol_ns.group(2) 1088 else: 1089 return False 1090 # 1091 # Found an export, trim out any special suffixes 1092 # 1093 for suffix in suffixes: 1094 # Be backward compatible with Python < 3.9 1095 if symbol.endswith(suffix): 1096 symbol = symbol[:-len(suffix)] 1097 function_set.add(symbol) 1098 return True 1099 1100 def process_normal(self, ln, line): 1101 """ 1102 STATE_NORMAL: looking for the ``/**`` to begin everything. 1103 """ 1104 1105 if not doc_start.match(line): 1106 return 1107 1108 # start a new entry 1109 self.reset_state(ln) 1110 1111 # next line is always the function name 1112 self.state = state.NAME 1113 1114 def process_name(self, ln, line): 1115 """ 1116 STATE_NAME: Looking for the "name - description" line 1117 """ 1118 # 1119 # Check for a DOC: block and handle them specially. 1120 # 1121 if doc_block.search(line): 1122 1123 if not doc_block.group(1): 1124 self.entry.begin_section(ln, "Introduction") 1125 else: 1126 self.entry.begin_section(ln, doc_block.group(1)) 1127 1128 self.entry.identifier = self.entry.section 1129 self.state = state.DOCBLOCK 1130 # 1131 # Otherwise we're looking for a normal kerneldoc declaration line. 1132 # 1133 elif doc_decl.search(line): 1134 self.entry.identifier = doc_decl.group(1) 1135 1136 # Test for data declaration 1137 if doc_begin_data.search(line): 1138 self.entry.decl_type = doc_begin_data.group(1) 1139 self.entry.identifier = doc_begin_data.group(2) 1140 # 1141 # Look for a function description 1142 # 1143 elif doc_begin_func.search(line): 1144 self.entry.identifier = doc_begin_func.group(1) 1145 self.entry.decl_type = "function" 1146 # 1147 # We struck out. 1148 # 1149 else: 1150 self.emit_msg(ln, 1151 f"This comment starts with '/**', but isn't a kernel-doc comment. Refer to Documentation/doc-guide/kernel-doc.rst\n{line}") 1152 self.state = state.NORMAL 1153 return 1154 # 1155 # OK, set up for a new kerneldoc entry. 1156 # 1157 self.state = state.BODY 1158 self.entry.identifier = self.entry.identifier.strip(" ") 1159 # if there's no @param blocks need to set up default section here 1160 self.entry.begin_section(ln + 1) 1161 # 1162 # Find the description portion, which *should* be there but 1163 # isn't always. 1164 # (We should be able to capture this from the previous parsing - someday) 1165 # 1166 r = KernRe("[-:](.*)") 1167 if r.search(line): 1168 self.entry.declaration_purpose = trim_whitespace(r.group(1)) 1169 self.state = state.DECLARATION 1170 else: 1171 self.entry.declaration_purpose = "" 1172 1173 if not self.entry.declaration_purpose and self.config.wshort_desc: 1174 self.emit_msg(ln, 1175 f"missing initial short description on line:\n{line}") 1176 1177 if not self.entry.identifier and self.entry.decl_type != "enum": 1178 self.emit_msg(ln, 1179 f"wrong kernel-doc identifier on line:\n{line}") 1180 self.state = state.NORMAL 1181 1182 if self.config.verbose: 1183 self.emit_msg(ln, 1184 f"Scanning doc for {self.entry.decl_type} {self.entry.identifier}", 1185 warning=False) 1186 # 1187 # Failed to find an identifier. Emit a warning 1188 # 1189 else: 1190 self.emit_msg(ln, f"Cannot find identifier on line:\n{line}") 1191 1192 def is_new_section(self, ln, line): 1193 """ 1194 Helper function to determine if a new section is being started. 1195 """ 1196 if doc_sect.search(line): 1197 self.state = state.BODY 1198 # 1199 # Pick out the name of our new section, tweaking it if need be. 1200 # 1201 newsection = doc_sect.group(1) 1202 if newsection.lower() == 'description': 1203 newsection = 'Description' 1204 elif newsection.lower() == 'context': 1205 newsection = 'Context' 1206 self.state = state.SPECIAL_SECTION 1207 elif newsection.lower() in ["@return", "@returns", 1208 "return", "returns"]: 1209 newsection = "Return" 1210 self.state = state.SPECIAL_SECTION 1211 elif newsection[0] == '@': 1212 self.state = state.SPECIAL_SECTION 1213 # 1214 # Initialize the contents, and get the new section going. 1215 # 1216 newcontents = doc_sect.group(2) 1217 if not newcontents: 1218 newcontents = "" 1219 self.dump_section() 1220 self.entry.begin_section(ln, newsection) 1221 self.entry.leading_space = None 1222 1223 self.entry.add_text(newcontents.lstrip()) 1224 return True 1225 return False 1226 1227 def is_comment_end(self, ln, line): 1228 """ 1229 Helper function to detect (and effect) the end of a kerneldoc comment. 1230 """ 1231 if doc_end.search(line): 1232 self.dump_section() 1233 1234 # Look for doc_com + <text> + doc_end: 1235 r = KernRe(r'\s*\*\s*[a-zA-Z_0-9:.]+\*/') 1236 if r.match(line): 1237 self.emit_msg(ln, f"suspicious ending line: {line}") 1238 1239 self.entry.prototype = "" 1240 self.entry.new_start_line = ln + 1 1241 1242 self.state = state.PROTO 1243 return True 1244 return False 1245 1246 1247 def process_decl(self, ln, line): 1248 """ 1249 STATE_DECLARATION: We've seen the beginning of a declaration. 1250 """ 1251 if self.is_new_section(ln, line) or self.is_comment_end(ln, line): 1252 return 1253 # 1254 # Look for anything with the " * " line beginning. 1255 # 1256 if doc_content.search(line): 1257 cont = doc_content.group(1) 1258 # 1259 # A blank line means that we have moved out of the declaration 1260 # part of the comment (without any "special section" parameter 1261 # descriptions). 1262 # 1263 if cont == "": 1264 self.state = state.BODY 1265 # 1266 # Otherwise we have more of the declaration section to soak up. 1267 # 1268 else: 1269 self.entry.declaration_purpose = \ 1270 trim_whitespace(self.entry.declaration_purpose + ' ' + cont) 1271 else: 1272 # Unknown line, ignore 1273 self.emit_msg(ln, f"bad line: {line}") 1274 1275 1276 def process_special(self, ln, line): 1277 """ 1278 STATE_SPECIAL_SECTION: a section ending with a blank line. 1279 """ 1280 # 1281 # If we have hit a blank line (only the " * " marker), then this 1282 # section is done. 1283 # 1284 if KernRe(r"\s*\*\s*$").match(line): 1285 self.entry.begin_section(ln, dump = True) 1286 self.state = state.BODY 1287 return 1288 # 1289 # Not a blank line, look for the other ways to end the section. 1290 # 1291 if self.is_new_section(ln, line) or self.is_comment_end(ln, line): 1292 return 1293 # 1294 # OK, we should have a continuation of the text for this section. 1295 # 1296 if doc_content.search(line): 1297 cont = doc_content.group(1) 1298 # 1299 # If the lines of text after the first in a special section have 1300 # leading white space, we need to trim it out or Sphinx will get 1301 # confused. For the second line (the None case), see what we 1302 # find there and remember it. 1303 # 1304 if self.entry.leading_space is None: 1305 r = KernRe(r'^(\s+)') 1306 if r.match(cont): 1307 self.entry.leading_space = len(r.group(1)) 1308 else: 1309 self.entry.leading_space = 0 1310 # 1311 # Otherwise, before trimming any leading chars, be *sure* 1312 # that they are white space. We should maybe warn if this 1313 # isn't the case. 1314 # 1315 for i in range(0, self.entry.leading_space): 1316 if cont[i] != " ": 1317 self.entry.leading_space = i 1318 break 1319 # 1320 # Add the trimmed result to the section and we're done. 1321 # 1322 self.entry.add_text(cont[self.entry.leading_space:]) 1323 else: 1324 # Unknown line, ignore 1325 self.emit_msg(ln, f"bad line: {line}") 1326 1327 def process_body(self, ln, line): 1328 """ 1329 STATE_BODY: the bulk of a kerneldoc comment. 1330 """ 1331 if self.is_new_section(ln, line) or self.is_comment_end(ln, line): 1332 return 1333 1334 if doc_content.search(line): 1335 cont = doc_content.group(1) 1336 self.entry.add_text(cont) 1337 else: 1338 # Unknown line, ignore 1339 self.emit_msg(ln, f"bad line: {line}") 1340 1341 def process_inline_name(self, ln, line): 1342 """STATE_INLINE_NAME: beginning of docbook comments within a prototype.""" 1343 1344 if doc_inline_sect.search(line): 1345 self.entry.begin_section(ln, doc_inline_sect.group(1)) 1346 self.entry.add_text(doc_inline_sect.group(2).lstrip()) 1347 self.state = state.INLINE_TEXT 1348 elif doc_inline_end.search(line): 1349 self.dump_section() 1350 self.state = state.PROTO 1351 elif doc_content.search(line): 1352 self.emit_msg(ln, f"Incorrect use of kernel-doc format: {line}") 1353 self.state = state.PROTO 1354 1355 # 1356 # Don't let it add partial comments at the code, as breaks the 1357 # logic meant to remove comments from prototypes. 1358 # 1359 self.process_proto_type(ln, "/**\n" + line) 1360 # else ... ?? 1361 1362 def process_inline_text(self, ln, line): 1363 """STATE_INLINE_TEXT: docbook comments within a prototype.""" 1364 1365 if doc_inline_end.search(line): 1366 self.dump_section() 1367 self.state = state.PROTO 1368 elif doc_content.search(line): 1369 self.entry.add_text(doc_content.group(1)) 1370 # else ... ?? 1371 1372 def syscall_munge(self, ln, proto): # pylint: disable=W0613 1373 """ 1374 Handle syscall definitions. 1375 """ 1376 1377 is_void = False 1378 1379 # Strip newlines/CR's 1380 proto = re.sub(r'[\r\n]+', ' ', proto) 1381 1382 # Check if it's a SYSCALL_DEFINE0 1383 if 'SYSCALL_DEFINE0' in proto: 1384 is_void = True 1385 1386 # Replace SYSCALL_DEFINE with correct return type & function name 1387 proto = KernRe(r'SYSCALL_DEFINE.*\(').sub('long sys_', proto) 1388 1389 r = KernRe(r'long\s+(sys_.*?),') 1390 if r.search(proto): 1391 proto = KernRe(',').sub('(', proto, count=1) 1392 elif is_void: 1393 proto = KernRe(r'\)').sub('(void)', proto, count=1) 1394 1395 # Now delete all of the odd-numbered commas in the proto 1396 # so that argument types & names don't have a comma between them 1397 count = 0 1398 length = len(proto) 1399 1400 if is_void: 1401 length = 0 # skip the loop if is_void 1402 1403 for ix in range(length): 1404 if proto[ix] == ',': 1405 count += 1 1406 if count % 2 == 1: 1407 proto = proto[:ix] + ' ' + proto[ix + 1:] 1408 1409 return proto 1410 1411 def tracepoint_munge(self, ln, proto): 1412 """ 1413 Handle tracepoint definitions. 1414 """ 1415 1416 tracepointname = None 1417 tracepointargs = None 1418 1419 # Match tracepoint name based on different patterns 1420 r = KernRe(r'TRACE_EVENT\((.*?),') 1421 if r.search(proto): 1422 tracepointname = r.group(1) 1423 1424 r = KernRe(r'DEFINE_SINGLE_EVENT\((.*?),') 1425 if r.search(proto): 1426 tracepointname = r.group(1) 1427 1428 r = KernRe(r'DEFINE_EVENT\((.*?),(.*?),') 1429 if r.search(proto): 1430 tracepointname = r.group(2) 1431 1432 if tracepointname: 1433 tracepointname = tracepointname.lstrip() 1434 1435 r = KernRe(r'TP_PROTO\((.*?)\)') 1436 if r.search(proto): 1437 tracepointargs = r.group(1) 1438 1439 if not tracepointname or not tracepointargs: 1440 self.emit_msg(ln, 1441 f"Unrecognized tracepoint format:\n{proto}\n") 1442 else: 1443 proto = f"static inline void trace_{tracepointname}({tracepointargs})" 1444 self.entry.identifier = f"trace_{self.entry.identifier}" 1445 1446 return proto 1447 1448 def process_proto_function(self, ln, line): 1449 """Ancillary routine to process a function prototype.""" 1450 1451 # strip C99-style comments to end of line 1452 line = KernRe(r"//.*$", re.S).sub('', line) 1453 # 1454 # Soak up the line's worth of prototype text, stopping at { or ; if present. 1455 # 1456 if KernRe(r'\s*#\s*define').match(line): 1457 self.entry.prototype = line 1458 elif not line.startswith('#'): # skip other preprocessor stuff 1459 r = KernRe(r'([^\{]*)') 1460 if r.match(line): 1461 self.entry.prototype += r.group(1) + " " 1462 # 1463 # If we now have the whole prototype, clean it up and declare victory. 1464 # 1465 if '{' in line or ';' in line or KernRe(r'\s*#\s*define').match(line): 1466 # strip comments and surrounding spaces 1467 self.entry.prototype = KernRe(r'/\*.*\*/').sub('', self.entry.prototype).strip() 1468 # 1469 # Handle self.entry.prototypes for function pointers like: 1470 # int (*pcs_config)(struct foo) 1471 # by turning it into 1472 # int pcs_config(struct foo) 1473 # 1474 r = KernRe(r'^(\S+\s+)\(\s*\*(\S+)\)') 1475 self.entry.prototype = r.sub(r'\1\2', self.entry.prototype) 1476 # 1477 # Handle special declaration syntaxes 1478 # 1479 if 'SYSCALL_DEFINE' in self.entry.prototype: 1480 self.entry.prototype = self.syscall_munge(ln, 1481 self.entry.prototype) 1482 else: 1483 r = KernRe(r'TRACE_EVENT|DEFINE_EVENT|DEFINE_SINGLE_EVENT') 1484 if r.search(self.entry.prototype): 1485 self.entry.prototype = self.tracepoint_munge(ln, 1486 self.entry.prototype) 1487 # 1488 # ... and we're done 1489 # 1490 self.dump_function(ln, self.entry.prototype) 1491 self.reset_state(ln) 1492 1493 def process_proto_type(self, ln, line): 1494 """ 1495 Ancillary routine to process a type. 1496 """ 1497 1498 # Strip C99-style comments and surrounding whitespace 1499 line = KernRe(r"//.*$", re.S).sub('', line).strip() 1500 if not line: 1501 return # nothing to see here 1502 1503 # To distinguish preprocessor directive from regular declaration later. 1504 if line.startswith('#'): 1505 line += ";" 1506 # 1507 # Split the declaration on any of { } or ;, and accumulate pieces 1508 # until we hit a semicolon while not inside {brackets} 1509 # 1510 r = KernRe(r'(.*?)([{};])') 1511 for chunk in r.split(line): 1512 if chunk: # Ignore empty matches 1513 self.entry.prototype += chunk 1514 # 1515 # This cries out for a match statement ... someday after we can 1516 # drop Python 3.9 ... 1517 # 1518 if chunk == '{': 1519 self.entry.brcount += 1 1520 elif chunk == '}': 1521 self.entry.brcount -= 1 1522 elif chunk == ';' and self.entry.brcount <= 0: 1523 self.dump_declaration(ln, self.entry.prototype) 1524 self.reset_state(ln) 1525 return 1526 # 1527 # We hit the end of the line while still in the declaration; put 1528 # in a space to represent the newline. 1529 # 1530 self.entry.prototype += ' ' 1531 1532 def process_proto(self, ln, line): 1533 """STATE_PROTO: reading a function/whatever prototype.""" 1534 1535 if doc_inline_oneline.search(line): 1536 self.entry.begin_section(ln, doc_inline_oneline.group(1)) 1537 self.entry.add_text(doc_inline_oneline.group(2)) 1538 self.dump_section() 1539 1540 elif doc_inline_start.search(line): 1541 self.state = state.INLINE_NAME 1542 1543 elif self.entry.decl_type == 'function': 1544 self.process_proto_function(ln, line) 1545 1546 else: 1547 self.process_proto_type(ln, line) 1548 1549 def process_docblock(self, ln, line): 1550 """STATE_DOCBLOCK: within a ``DOC:`` block.""" 1551 1552 if doc_end.search(line): 1553 self.dump_section() 1554 self.output_declaration("doc", self.entry.identifier) 1555 self.reset_state(ln) 1556 1557 elif doc_content.search(line): 1558 self.entry.add_text(doc_content.group(1)) 1559 1560 def parse_export(self): 1561 """ 1562 Parses ``EXPORT_SYMBOL*`` macros from a single Kernel source file. 1563 """ 1564 1565 export_table = set() 1566 1567 try: 1568 with open(self.fname, "r", encoding="utf8", 1569 errors="backslashreplace") as fp: 1570 1571 for line in fp: 1572 self.process_export(export_table, line) 1573 1574 except IOError: 1575 return None 1576 1577 return export_table 1578 1579 #: The state/action table telling us which function to invoke in each state. 1580 state_actions = { 1581 state.NORMAL: process_normal, 1582 state.NAME: process_name, 1583 state.BODY: process_body, 1584 state.DECLARATION: process_decl, 1585 state.SPECIAL_SECTION: process_special, 1586 state.INLINE_NAME: process_inline_name, 1587 state.INLINE_TEXT: process_inline_text, 1588 state.PROTO: process_proto, 1589 state.DOCBLOCK: process_docblock, 1590 } 1591 1592 def parse_kdoc(self): 1593 """ 1594 Open and process each line of a C source file. 1595 The parsing is controlled via a state machine, and the line is passed 1596 to a different process function depending on the state. The process 1597 function may update the state as needed. 1598 1599 Besides parsing kernel-doc tags, it also parses export symbols. 1600 """ 1601 1602 prev = "" 1603 prev_ln = None 1604 export_table = set() 1605 1606 try: 1607 with open(self.fname, "r", encoding="utf8", 1608 errors="backslashreplace") as fp: 1609 for ln, line in enumerate(fp): 1610 1611 line = line.expandtabs().strip("\n") 1612 1613 # Group continuation lines on prototypes 1614 if self.state == state.PROTO: 1615 if line.endswith("\\"): 1616 prev += line.rstrip("\\") 1617 if not prev_ln: 1618 prev_ln = ln 1619 continue 1620 1621 if prev: 1622 ln = prev_ln 1623 line = prev + line 1624 prev = "" 1625 prev_ln = None 1626 1627 self.config.log.debug("%d %s: %s", 1628 ln, state.name[self.state], 1629 line) 1630 1631 # This is an optimization over the original script. 1632 # There, when export_file was used for the same file, 1633 # it was read twice. Here, we use the already-existing 1634 # loop to parse exported symbols as well. 1635 # 1636 if (self.state != state.NORMAL) or \ 1637 not self.process_export(export_table, line): 1638 # Hand this line to the appropriate state handler 1639 self.state_actions[self.state](self, ln, line) 1640 1641 self.emit_unused_warnings() 1642 1643 except OSError: 1644 self.config.log.error(f"Error: Cannot open file {self.fname}") 1645 1646 return export_table, self.entries 1647