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