xref: /linux/tools/lib/python/kdoc/kdoc_parser.py (revision d7758384ccb470aed7c7a86f0825b1b16bd7a288)
1#!/usr/bin/env python3
2# SPDX-License-Identifier: GPL-2.0
3# Copyright(c) 2025: Mauro Carvalho Chehab <mchehab@kernel.org>.
4#
5# pylint: disable=C0301,C0302,R0904,R0912,R0913,R0914,R0915,R0917,R1702
6
7"""
8Classes and functions related to reading a C language source or header FILE
9and extract embedded documentation comments from it.
10"""
11
12import sys
13import re
14import difflib
15from pprint import pformat
16
17from kdoc.c_lex import CTokenizer, tokenizer_set_log
18from kdoc.kdoc_re import KernRe
19from kdoc.kdoc_item import KdocItem
20
21#
22# Regular expressions used to parse kernel-doc markups at KernelDoc class.
23#
24# Let's declare them in lowercase outside any class to make it easier to
25# convert from the Perl script.
26#
27# As those are evaluated at the beginning, no need to cache them
28#
29
30# Allow whitespace at end of comment start.
31doc_start = KernRe(r'^/\*\*\s*$', cache=False)
32
33doc_end = KernRe(r'\*/', cache=False)
34doc_com = KernRe(r'\s*\*\s*', cache=False)
35doc_com_body = KernRe(r'\s*\* ?', cache=False)
36doc_decl = doc_com + KernRe(r'(\w+)', cache=False)
37
38# @params and a strictly limited set of supported section names
39# Specifically:
40#   Match @word:
41#         @...:
42#         @{section-name}:
43# while trying to not match literal block starts like "example::"
44#
45known_section_names = 'description|context|returns?|notes?|examples?'
46known_sections = KernRe(known_section_names, flags = re.I)
47doc_sect = doc_com + \
48    KernRe(r'\s*(@[.\w]+|@\.\.\.|' + known_section_names + r')\s*:([^:].*)?$',
49           flags=re.I, cache=False)
50
51doc_content = doc_com_body + KernRe(r'(.*)', cache=False)
52doc_inline_start = KernRe(r'^\s*/\*\*\s*$', cache=False)
53doc_inline_sect = KernRe(r'\s*\*\s*(@\s*[\w][\w\.]*\s*):(.*)', cache=False)
54doc_inline_end = KernRe(r'^\s*\*/\s*$', cache=False)
55doc_inline_oneline = KernRe(r'^\s*/\*\*\s*(@\s*[\w][\w\.]*\s*):\s*(.*)\s*\*/\s*$', cache=False)
56
57export_symbol = KernRe(r'^\s*EXPORT_SYMBOL(_GPL)?\s*\(\s*(\w+)\s*\)\s*', cache=False)
58export_symbol_ns = KernRe(r'^\s*EXPORT_SYMBOL_NS(_GPL)?\s*\(\s*(\w+)\s*,\s*"\S+"\)\s*', cache=False)
59
60type_param = KernRe(r"@(\w*((\.\w+)|(->\w+))*(\.\.\.)?)", cache=False)
61
62#
63# Tests for the beginning of a kerneldoc block in its various forms.
64#
65doc_block = doc_com + KernRe(r'DOC:\s*(.*)?', cache=False)
66doc_begin_data = KernRe(r"^\s*\*?\s*(struct|union|enum|typedef|var)\b\s*(\w*)", cache = False)
67doc_begin_func = KernRe(str(doc_com) +			# initial " * '
68                        r"(?:\w+\s*\*\s*)?" + 		# type (not captured)
69                        r'(?:define\s+)?' + 		# possible "define" (not captured)
70                        r'(\w+)\s*(?:\(\w*\))?\s*' +	# name and optional "(...)"
71                        r'(?:[-:].*)?$',		# description (not captured)
72                        cache = False)
73
74#
75# Ancillary functions
76#
77
78multi_space = KernRe(r'\s\s+')
79def trim_whitespace(s):
80    """
81    A little helper to get rid of excess white space.
82    """
83    return multi_space.sub(' ', s.strip())
84
85def trim_private_members(text):
86    """
87    Remove ``struct``/``enum`` members that have been marked "private".
88    """
89
90    tokens = CTokenizer(text)
91    return str(tokens)
92
93class state:
94    """
95    States used by the parser's state machine.
96    """
97
98    # Parser states
99    NORMAL        = 0        #: Normal code.
100    NAME          = 1        #: Looking for function name.
101    DECLARATION   = 2        #: We have seen a declaration which might not be done.
102    BODY          = 3        #: The body of the comment.
103    SPECIAL_SECTION = 4      #: Doc section ending with a blank line.
104    PROTO         = 5        #: Scanning prototype.
105    DOCBLOCK      = 6        #: Documentation block.
106    INLINE_NAME   = 7        #: Gathering doc outside main block.
107    INLINE_TEXT   = 8	     #: Reading the body of inline docs.
108
109    #: Names for each parser state.
110    name = [
111        "NORMAL",
112        "NAME",
113        "DECLARATION",
114        "BODY",
115        "SPECIAL_SECTION",
116        "PROTO",
117        "DOCBLOCK",
118        "INLINE_NAME",
119        "INLINE_TEXT",
120    ]
121
122
123SECTION_DEFAULT = "Description"  #: Default section.
124
125class KernelEntry:
126    """
127    Encapsulates a Kernel documentation entry.
128    """
129
130    def __init__(self, config, fname, ln):
131        self.config = config
132        self.fname = fname
133
134        self._contents = []
135        self.prototype = ""
136
137        self.warnings = []
138
139        self.parameterlist = []
140        self.parameterdescs = {}
141        self.parametertypes = {}
142        self.parameterdesc_start_lines = {}
143
144        self.sections_start_lines = {}
145        self.sections = {}
146
147        self.anon_struct_union = False
148
149        self.leading_space = None
150
151        self.fname = fname
152
153        # State flags
154        self.brcount = 0
155        self.declaration_start_line = ln + 1
156
157    #
158    # Management of section contents
159    #
160    def add_text(self, text):
161        """Add a new text to the entry contents list."""
162        self._contents.append(text)
163
164    def contents(self):
165        """Returns a string with all content texts that were added."""
166        return '\n'.join(self._contents) + '\n'
167
168    # TODO: rename to emit_message after removal of kernel-doc.pl
169    def emit_msg(self, ln, msg, *, warning=True):
170        """Emit a message."""
171
172        log_msg = f"{self.fname}:{ln} {msg}"
173
174        if not warning:
175            self.config.log.info(log_msg)
176            return
177
178        # Delegate warning output to output logic, as this way it
179        # will report warnings/info only for symbols that are output
180
181        self.warnings.append(log_msg)
182        return
183
184    def begin_section(self, line_no, title = SECTION_DEFAULT, dump = False):
185        """
186        Begin a new section.
187        """
188        if dump:
189            self.dump_section(start_new = True)
190        self.section = title
191        self.new_start_line = line_no
192
193    def dump_section(self, start_new=True):
194        """
195        Dumps section contents to arrays/hashes intended for that purpose.
196        """
197        #
198        # If we have accumulated no contents in the default ("description")
199        # section, don't bother.
200        #
201        if self.section == SECTION_DEFAULT and not self._contents:
202            return
203        name = self.section
204        contents = self.contents()
205
206        if type_param.match(name):
207            name = type_param.group(1)
208
209            self.parameterdescs[name] = contents
210            self.parameterdesc_start_lines[name] = self.new_start_line
211
212            self.new_start_line = 0
213
214        else:
215            if name in self.sections and self.sections[name] != "":
216                # Only warn on user-specified duplicate section names
217                if name != SECTION_DEFAULT:
218                    self.emit_msg(self.new_start_line,
219                                  f"duplicate section name '{name}'")
220                # Treat as a new paragraph - add a blank line
221                self.sections[name] += '\n' + contents
222            else:
223                self.sections[name] = contents
224                self.sections_start_lines[name] = self.new_start_line
225                self.new_start_line = 0
226
227#        self.config.log.debug("Section: %s : %s", name, pformat(vars(self)))
228
229        if start_new:
230            self.section = SECTION_DEFAULT
231            self._contents = []
232
233python_warning = False
234
235class KernelDoc:
236    """
237    Read a C language source or header FILE and extract embedded
238    documentation comments.
239    """
240
241    #: Name of context section.
242    section_context = "Context"
243
244    #: Name of return section.
245    section_return = "Return"
246
247    #: String to write when a parameter is not described.
248    undescribed = "-- undescribed --"
249
250    def __init__(self, config, fname, xforms, store_src=False):
251        """Initialize internal variables"""
252
253        self.fname = fname
254        self.config = config
255        self.xforms = xforms
256        self.store_src = store_src
257
258        tokenizer_set_log(self.config.log, f"{self.fname}: CMatch: ")
259
260        # Initial state for the state machines
261        self.state = state.NORMAL
262
263        # Store entry currently being processed
264        self.entry = None
265
266        # Place all potential outputs into an array
267        self.entries = []
268
269        #
270        # We need Python 3.7 for its "dicts remember the insertion
271        # order" guarantee
272        #
273        global python_warning
274        if (not python_warning and
275            sys.version_info.major == 3 and sys.version_info.minor < 7):
276
277            self.emit_msg(0,
278                          'Python 3.7 or later is required for correct results')
279            python_warning = True
280
281    def emit_msg(self, ln, msg, *, warning=True):
282        """Emit a message"""
283
284        if self.entry:
285            self.entry.emit_msg(ln, msg, warning=warning)
286            return
287
288        log_msg = f"{self.fname}:{ln} {msg}"
289
290        if warning:
291            self.config.log.warning(log_msg)
292        else:
293            self.config.log.info(log_msg)
294
295    def dump_section(self, start_new=True):
296        """
297        Dump section contents to arrays/hashes intended for that purpose.
298        """
299
300        if self.entry:
301            self.entry.dump_section(start_new)
302
303    # TODO: rename it to store_declaration after removal of kernel-doc.pl
304    def output_declaration(self, dtype, name, **args):
305        """
306        Store the entry into an entry array.
307
308        The actual output and output filters will be handled elsewhere.
309        """
310
311        item = KdocItem(name, self.fname, dtype,
312                        self.entry.declaration_start_line, **args)
313        item.warnings = self.entry.warnings
314
315        # Drop empty sections
316        # TODO: improve empty sections logic to emit warnings
317        sections = self.entry.sections
318        for section in ["Description", "Return"]:
319            if section in sections and not sections[section].rstrip():
320                del sections[section]
321        item.set_sections(sections, self.entry.sections_start_lines)
322        item.set_params(self.entry.parameterlist, self.entry.parameterdescs,
323                        self.entry.parametertypes,
324                        self.entry.parameterdesc_start_lines)
325        self.entries.append(item)
326
327        self.config.log.debug("Output: %s:%s = %s", dtype, name, pformat(args))
328
329    def emit_unused_warnings(self):
330        """
331        When the parser fails to produce a valid entry, it places some
332        warnings under `entry.warnings` that will be discarded when resetting
333        the state.
334
335        Ensure that those warnings are not lost.
336
337        .. note::
338
339              Because we are calling `config.warning()` here, those
340              warnings are not filtered by the `-W` parameters: they will all
341              be produced even when `-Wreturn`, `-Wshort-desc`, and/or
342              `-Wcontents-before-sections` are used.
343
344              Allowing those warnings to be filtered is complex, because it
345              would require storing them in a buffer and then filtering them
346              during the output step of the code, depending on the
347              selected symbols.
348        """
349        if self.entry and self.entry not in self.entries:
350            for log_msg in self.entry.warnings:
351                self.config.warning(log_msg)
352
353    def reset_state(self, ln):
354        """
355        Ancillary routine to create a new entry. It initializes all
356        variables used by the state machine.
357        """
358
359        self.emit_unused_warnings()
360
361        self.entry = KernelEntry(self.config, self.fname, ln)
362
363        # State flags
364        self.state = state.NORMAL
365
366    def push_parameter(self, ln, decl_type, param, dtype,
367                       org_arg, declaration_name):
368        """
369        Store parameters and their descriptions at self.entry.
370        """
371
372        if self.entry.anon_struct_union and dtype == "" and param == "}":
373            return  # Ignore the ending }; from anonymous struct/union
374
375        self.entry.anon_struct_union = False
376
377        param = KernRe(r'[\[\)].*').sub('', param, count=1)
378
379        #
380        # Look at various "anonymous type" cases.
381        #
382        if dtype == '':
383            if param.endswith("..."):
384                named_variadic = len(param) > 3
385                if named_variadic: # there is a name provided, use that
386                    #
387                    # If the user documented the parameter using the
388                    # ``@name...:`` form, the description is stored in
389                    # parameterdescs under the unstripped key.  Migrate
390                    # it to the stripped key so the user's text is not
391                    # silently dropped during output, and so the new
392                    # excess-parameter check in check_sections() does
393                    # not flag the unstripped key as orphaned.
394                    #
395                    orig = self.entry.parameterdescs.pop(param, None)
396                    param = param[:-3]
397                    if orig is not None and \
398                       not self.entry.parameterdescs.get(param):
399                        self.entry.parameterdescs[param] = orig
400                if not self.entry.parameterdescs.get(param):
401                    #
402                    # For a named variadic (e.g. ``args...``), emit the
403                    # standard "not described" warning before auto-filling
404                    # so a missing or mistyped ``@<name>:`` doc tag does
405                    # not go undetected.  The bare ``...`` form has no
406                    # natural name for the user to document and so always
407                    # gets the auto-generated text.
408                    #
409                    if named_variadic and decl_type == 'function':
410                        self.emit_msg(ln,
411                                      f"function parameter '{param}' "
412                                      f"not described in "
413                                      f"'{declaration_name}'")
414                    self.entry.parameterdescs[param] = "variable arguments"
415
416            elif (not param) or param == "void":
417                param = "void"
418                self.entry.parameterdescs[param] = "no arguments"
419
420            elif param in ["struct", "union"]:
421                # Handle unnamed (anonymous) union or struct
422                dtype = param
423                param = "{unnamed_" + param + "}"
424                self.entry.parameterdescs[param] = "anonymous\n"
425                self.entry.anon_struct_union = True
426
427        # Warn if parameter has no description
428        # (but ignore ones starting with # as these are not parameters
429        # but inline preprocessor statements)
430        if param not in self.entry.parameterdescs and not param.startswith("#"):
431            self.entry.parameterdescs[param] = self.undescribed
432
433            if "." not in param:
434                if decl_type == 'function':
435                    dname = f"{decl_type} parameter"
436                else:
437                    dname = f"{decl_type} member"
438
439                self.emit_msg(ln,
440                              f"{dname} '{param}' not described in '{declaration_name}'")
441
442        # Strip spaces from param so that it is one continuous string on
443        # parameterlist. This fixes a problem where check_sections()
444        # cannot find a parameter like "addr[6 + 2]" because it actually
445        # appears as "addr[6", "+", "2]" on the parameter list.
446        # However, it's better to maintain the param string unchanged for
447        # output, so just weaken the string compare in check_sections()
448        # to ignore "[blah" in a parameter string.
449
450        self.entry.parameterlist.append(param)
451        org_arg = KernRe(r'\s\s+').sub(' ', org_arg)
452        self.entry.parametertypes[param] = org_arg
453
454
455    def create_parameter_list(self, ln, decl_type, args,
456                              splitter, declaration_name):
457        """
458        Creates a list of parameters, storing them at self.entry.
459        """
460
461        # temporarily replace all commas inside function pointer definition
462        arg_expr = KernRe(r'(\([^\),]+),')
463        while arg_expr.search(args):
464            args = arg_expr.sub(r"\1#", args)
465
466        for arg in args.split(splitter):
467            # Ignore argument attributes
468            arg = KernRe(r'\sPOS0?\s').sub(' ', arg)
469
470            # Replace '[at_least ' with '[static '.  This allows sphinx to parse
471            # array parameter declarations like 'char A[at_least 4]', where
472            # 'at_least' is #defined to 'static' by the kernel headers.
473            arg = arg.replace('[at_least ', '[static ')
474
475            # Strip leading/trailing spaces
476            arg = arg.strip()
477            arg = KernRe(r'\s+').sub(' ', arg, count=1)
478
479            if arg.startswith('#'):
480                # Treat preprocessor directive as a typeless variable just to fill
481                # corresponding data structures "correctly". Catch it later in
482                # output_* subs.
483
484                # Treat preprocessor directive as a typeless variable
485                self.push_parameter(ln, decl_type, arg, "",
486                                    "", declaration_name)
487            #
488            # The pointer-to-function case.
489            #
490            elif KernRe(r'\(.+\)\s*\(').search(arg):
491                arg = arg.replace('#', ',')
492                r = KernRe(r'[^\(]+\(\*?\s*'  # Everything up to "(*"
493                           r'([\w\[\].]*)'    # Capture the name and possible [array]
494                           r'\s*\)')	      # Make sure the trailing ")" is there
495                if r.match(arg):
496                    param = r.group(1)
497                else:
498                    self.emit_msg(ln, f"Invalid param: {arg}")
499                    param = arg
500                dtype = arg.replace(param, '')
501                self.push_parameter(ln, decl_type, param, dtype, arg, declaration_name)
502            #
503            # The array-of-pointers case.  Dig the parameter name out from the middle
504            # of the declaration.
505            #
506            elif KernRe(r'\(.+\)\s*\[').search(arg):
507                r = KernRe(r'[^\(]+\(\s*\*\s*'		# Up to "(" and maybe "*"
508                           r'([\w.]*?)'			# The actual pointer name
509                           r'\s*(\[\s*\w+\s*\]\s*)*\)') # The [array portion]
510                if r.match(arg):
511                    param = r.group(1)
512                else:
513                    self.emit_msg(ln, f"Invalid param: {arg}")
514                    param = arg
515                dtype = arg.replace(param, '')
516                self.push_parameter(ln, decl_type, param, dtype, arg, declaration_name)
517            elif arg:
518                #
519                # Clean up extraneous spaces and split the string at commas; the first
520                # element of the resulting list will also include the type information.
521                #
522                arg = KernRe(r'\s*:\s*').sub(":", arg)
523                arg = KernRe(r'\s*\[').sub('[', arg)
524                args = KernRe(r'\s*,\s*').split(arg)
525                args[0] = re.sub(r'(\*+)\s*', r' \1', args[0])
526                #
527                # args[0] has a string of "type a".  If "a" includes an [array]
528                # declaration, we want to not be fooled by any white space inside
529                # the brackets, so detect and handle that case specially.
530                #
531                r = KernRe(r'^([^[\]]*\s+)(.*)$')
532                if r.match(args[0]):
533                    args[0] = r.group(2)
534                    dtype = r.group(1)
535                else:
536                    # No space in args[0]; this seems wrong but preserves previous behavior
537                    dtype = ''
538
539                bitfield_re = KernRe(r'(.*?):(\w+)')
540                for param in args:
541                    #
542                    # For pointers, shift the star(s) from the variable name to the
543                    # type declaration.
544                    #
545                    r = KernRe(r'^(\*+)\s*(.*)')
546                    if r.match(param):
547                        self.push_parameter(ln, decl_type, r.group(2),
548                                            f"{dtype} {r.group(1)}",
549                                            arg, declaration_name)
550                    #
551                    # Perform a similar shift for bitfields.
552                    #
553                    elif bitfield_re.search(param):
554                        if dtype != "":  # Skip unnamed bit-fields
555                            self.push_parameter(ln, decl_type, bitfield_re.group(1),
556                                                f"{dtype}:{bitfield_re.group(2)}",
557                                                arg, declaration_name)
558                    else:
559                        self.push_parameter(ln, decl_type, param, dtype,
560                                            arg, declaration_name)
561
562    def get_suggestions_hint(self, decl_name, possible_names):
563        # For decl name 'flags' or 'flgas', suggests 'substruct.flags'
564        submember_exact = []
565        submember_substrings = []
566        submember_suggestions = []
567        for possible_name in possible_names:
568            parts = possible_name.strip().split('.')
569            if len(parts) < 2:
570                continue
571
572            final_part = parts[-1]
573            if decl_name == final_part:
574                submember_exact.append(possible_name)
575            elif decl_name in final_part:
576                submember_substrings.append(possible_name)
577            elif difflib.get_close_matches(decl_name, [final_part]):
578                submember_suggestions.append(possible_name)
579
580        # For decl name 'flgas', suggests 'flags'
581        full_suggestions = difflib.get_close_matches(decl_name, possible_names)
582
583        # For decl name 'member', suggests 'longer_member'
584        full_substrings = [name for name in possible_names if decl_name in name]
585
586        ordered_lists = [
587            submember_exact,
588            submember_substrings,
589            submember_suggestions,
590            full_suggestions,
591            full_substrings,
592        ]
593
594        # Deduplicate but maintain order from most to least likely:
595        unique_suggestions = {}
596        for suggestion_list in ordered_lists:
597            for suggestion in suggestion_list:
598                unique_suggestions[suggestion] = None
599
600        suggestions = list(unique_suggestions.keys())
601        if not suggestions:
602            return ""
603
604        return f"(did you mean one of: '{"', '".join(suggestions)}')"
605
606    def check_sections(self, ln, decl_name, decl_type):
607        """
608        Check for errors inside sections, emitting warnings if not found
609        parameters are described.
610        """
611        for section in self.entry.sections:
612            if section not in self.entry.parameterlist and \
613               not known_sections.search(section):
614                hint = self.get_suggestions_hint(section, self.entry.parameterlist)
615                if decl_type == 'function':
616                    dname = f"{decl_type} parameter"
617                else:
618                    dname = f"{decl_type} member"
619                self.emit_msg(ln,
620                              f"Excess {dname} '{section}' description in '{decl_name}' {hint}".strip())
621
622        #
623        # Check that documented parameter names (from doc comments, including
624        # inline ``/** @member: */`` tags) actually match real members in
625        # the declaration.  This catches mismatched or stale kernel-doc
626        # member tags that don't correspond to any actual struct/union
627        # member or function parameter.
628        #
629        for param_name, desc in self.entry.parameterdescs.items():
630            # Skip auto-generated entries from push_parameter()
631            if desc == self.undescribed:
632                continue
633            if desc in ("no arguments", "anonymous\n", "variable arguments"):
634                continue
635            if param_name.startswith("{unnamed_"):
636                continue
637            if param_name in self.entry.parameterlist:
638                continue
639
640            hint = self.get_suggestions_hint(param_name, self.entry.parameterlist)
641            if decl_type == 'function':
642                dname = f"{decl_type} parameter"
643            else:
644                dname = f"{decl_type} member"
645            self.emit_msg(ln,
646                          f"Excess {dname} '{param_name}' description in '{decl_name}' {hint}".strip())
647
648    def check_return_section(self, ln, declaration_name, return_type):
649        """
650        If the function doesn't return void, warns about the lack of a
651        return description.
652        """
653
654        if not self.config.wreturn:
655            return
656
657        # Ignore an empty return type (It's a macro)
658        # Ignore functions with a "void" return type (but not "void *")
659        if not return_type or KernRe(r'void\s*\w*\s*$').search(return_type):
660            return
661
662        if not self.entry.sections.get("Return", None):
663            self.emit_msg(ln,
664                          f"No description found for return value of '{declaration_name}'")
665
666    def split_struct_proto(self, proto):
667        """
668        Split apart a structure prototype; returns (struct|union, name,
669        members) or ``None``.
670        """
671
672        type_pattern = r'(struct|union)'
673        qualifiers = [
674            "__attribute__",
675            "__packed",
676            "__aligned",
677            "____cacheline_aligned_in_smp",
678            "____cacheline_aligned",
679        ]
680        definition_body = r'\{(.*)\}\s*' + "(?:" + '|'.join(qualifiers) + ")?"
681
682        r = KernRe(type_pattern + r'\s+(\w+)\s*' + definition_body)
683        if r.search(proto):
684            return (r.group(1), r.group(2), r.group(3))
685        else:
686            r = KernRe(r'typedef\s+' + type_pattern + r'\s*' + definition_body + r'\s*(\w+)\s*;')
687            if r.search(proto):
688                return (r.group(1), r.group(3), r.group(2))
689        return None
690
691    def rewrite_struct_members(self, members):
692        """
693        Process ``struct``/``union`` members from the most deeply nested
694        outward.
695
696        Rewrite the members of a ``struct`` or ``union`` for easier formatting
697        later on. Among other things, this function will turn a member like::
698
699          struct { inner_members; } foo;
700
701        into::
702
703          struct foo; inner_members;
704        """
705
706        #
707        # The trick is in the ``^{`` below - it prevents a match of an outer
708        # ``struct``/``union`` until the inner one has been munged
709        # (removing the ``{`` in the process).
710        #
711        struct_members = KernRe(r'(struct|union)'   # 0: declaration type
712                                r'([^\{\};]+)' 	    # 1: possible name
713                                r'(\{)'
714                                r'([^\{\}]*)'       # 3: Contents of declaration
715                                r'(\})'
716                                r'([^\{\};]*)(;)')  # 5: Remaining stuff after declaration
717        tuples = struct_members.findall(members)
718        while tuples:
719            for t in tuples:
720                newmember = ""
721                oldmember = "".join(t) # Reconstruct the original formatting
722                dtype, name, lbr, content, rbr, rest, semi = t
723                #
724                # Pass through each field name, normalizing the form and formatting.
725                #
726                for s_id in rest.split(','):
727                    s_id = s_id.strip()
728                    newmember += f"{dtype} {s_id}; "
729                    #
730                    # Remove bitfield/array/pointer info, getting the bare name.
731                    #
732                    s_id = KernRe(r'[:\[].*').sub('', s_id)
733                    s_id = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', s_id)
734                    #
735                    # Pass through the members of this inner structure/union.
736                    #
737                    for arg in content.split(';'):
738                        arg = arg.strip()
739                        #
740                        # Look for (type)(*name)(args) - pointer to function
741                        #
742                        r = KernRe(r'^([^\(]+\(\*?\s*)([\w.]*)(\s*\).*)')
743                        if r.match(arg):
744                            dtype, name, extra = r.group(1), r.group(2), r.group(3)
745                            # Pointer-to-function
746                            if not s_id:
747                                # Anonymous struct/union
748                                newmember += f"{dtype}{name}{extra}; "
749                            else:
750                                newmember += f"{dtype}{s_id}.{name}{extra}; "
751                        #
752                        # Otherwise a non-function member.
753                        #
754                        else:
755                            #
756                            # Remove bitmap and array portions and spaces around commas
757                            #
758                            arg = KernRe(r':\s*\d+\s*').sub('', arg)
759                            arg = KernRe(r'\[.*\]').sub('', arg)
760                            arg = KernRe(r'\s*,\s*').sub(',', arg)
761                            #
762                            # Look for a normal decl - "type name[,name...]"
763                            #
764                            r = KernRe(r'(.*)\s+([\S+,]+)')
765                            if r.search(arg):
766                                for name in r.group(2).split(','):
767                                    name = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', name)
768                                    if not s_id:
769                                        # Anonymous struct/union
770                                        newmember += f"{r.group(1)} {name}; "
771                                    else:
772                                        newmember += f"{r.group(1)} {s_id}.{name}; "
773                            else:
774                                newmember += f"{arg}; "
775                #
776                # At the end of the s_id loop, replace the original declaration with
777                # the munged version.
778                #
779                members = members.replace(oldmember, newmember)
780            #
781            # End of the tuple loop - search again and see if there are outer members
782            # that now turn up.
783            #
784            tuples = struct_members.findall(members)
785        return members
786
787    def format_struct_decl(self, declaration):
788        """
789        Format the ``struct`` declaration into a standard form for inclusion
790        in the resulting docs.
791        """
792
793        #
794        # Insert newlines, get rid of extra spaces.
795        #
796        declaration = KernRe(r'([\{;])').sub(r'\1\n', declaration)
797        declaration = KernRe(r'\}\s+;').sub('};', declaration)
798        #
799        # Format inline enums with each member on its own line.
800        #
801        r = KernRe(r'(enum\s+\{[^\}]+),([^\n])')
802        while r.search(declaration):
803            declaration = r.sub(r'\1,\n\2', declaration)
804        #
805        # Now go through and supply the right number of tabs
806        # for each line.
807        #
808        def_args = declaration.split('\n')
809        level = 1
810        declaration = ""
811        for clause in def_args:
812            clause = KernRe(r'\s+').sub(' ', clause.strip(), count=1)
813            if clause:
814                if '}' in clause and level > 1:
815                    level -= 1
816                if not clause.startswith('#'):
817                    declaration += "\t" * level
818                declaration += "\t" + clause + "\n"
819                if "{" in clause and "}" not in clause:
820                    level += 1
821        return declaration
822
823
824    def dump_struct(self, ln, proto, source):
825        """
826        Store an entry for a ``struct`` or ``union``
827        """
828        #
829        # Do the basic parse to get the pieces of the declaration.
830        #
831        source = source
832        proto = trim_private_members(proto)
833        struct_parts = self.split_struct_proto(proto)
834        if not struct_parts:
835            self.emit_msg(ln, f"{proto} error: Cannot parse struct or union!")
836            return
837        decl_type, declaration_name, members = struct_parts
838
839        if self.entry.identifier != declaration_name:
840            self.emit_msg(ln, f"expecting prototype for {decl_type} {self.entry.identifier}. "
841                          f"Prototype was for {decl_type} {declaration_name} instead\n")
842            return
843        #
844        # Go through the list of members applying all of our transformations.
845        #
846        members = self.xforms.apply("struct", members)
847
848        #
849        # Deal with embedded struct and union members, and drop enums entirely.
850        #
851        declaration = members
852        members = self.rewrite_struct_members(members)
853        members = re.sub(r'(\{[^\{\}]*\})', '', members)
854        #
855        # Output the result and we are done.
856        #
857        self.create_parameter_list(ln, decl_type, members, ';',
858                                   declaration_name)
859        self.check_sections(ln, declaration_name, decl_type)
860        self.output_declaration(decl_type, declaration_name,
861                                source=source,
862                                definition=self.format_struct_decl(declaration),
863                                purpose=self.entry.declaration_purpose)
864
865    def dump_enum(self, ln, proto, source):
866        """
867        Store an ``enum`` inside self.entries array.
868        """
869        #
870        # Strip preprocessor directives.  Note that this depends on the
871        # trailing semicolon we added in process_proto_type().
872        #
873        source = source
874        proto = trim_private_members(proto)
875        proto = KernRe(r'#\s*((define|ifdef|if)\s+|endif)[^;]*;', flags=re.S).sub('', proto)
876        #
877        # Parse out the name and members of the enum.  Typedef form first.
878        #
879        r = KernRe(r'typedef\s+enum\s*\{(.*)\}\s*(\w*)\s*;')
880        if r.search(proto):
881            declaration_name = r.group(2)
882            members = r.group(1)
883        #
884        # Failing that, look for a straight enum
885        #
886        else:
887            r = KernRe(r'enum\s+(\w*)\s*\{(.*)\}')
888            if r.match(proto):
889                declaration_name = r.group(1)
890                members = r.group(2)
891        #
892        # OK, this isn't going to work.
893        #
894            else:
895                self.emit_msg(ln, f"{proto}: error: Cannot parse enum!")
896                return
897        #
898        # Make sure we found what we were expecting.
899        #
900        if self.entry.identifier != declaration_name:
901            if self.entry.identifier == "":
902                self.emit_msg(ln,
903                              f"{proto}: wrong kernel-doc identifier on prototype")
904            else:
905                self.emit_msg(ln,
906                              f"expecting prototype for enum {self.entry.identifier}. "
907                              f"Prototype was for enum {declaration_name} instead")
908            return
909
910        if not declaration_name:
911            declaration_name = "(anonymous)"
912        #
913        # Parse out the name of each enum member, and verify that we
914        # have a description for it.
915        #
916        member_set = set()
917        members = KernRe(r'\([^;)]*\)').sub('', members)
918        for arg in members.split(','):
919            arg = KernRe(r'^\s*(\w+).*').sub(r'\1', arg)
920            if not arg.strip():
921                continue
922
923            self.entry.parameterlist.append(arg)
924            if arg not in self.entry.parameterdescs:
925                self.entry.parameterdescs[arg] = self.undescribed
926                self.emit_msg(ln,
927                              f"Enum value '{arg}' not described in enum '{declaration_name}'")
928            member_set.add(arg)
929        #
930        # Ensure that every described member actually exists in the enum.
931        #
932        for k in self.entry.parameterdescs:
933            if k not in member_set:
934                self.emit_msg(ln,
935                              f"Excess enum value '@{k}' description in '{declaration_name}'")
936
937        self.output_declaration('enum', declaration_name,
938                                source=source,
939                                purpose=self.entry.declaration_purpose)
940
941    def dump_var(self, ln, proto, source):
942        """
943        Store variables that are part of kAPI.
944        """
945        VAR_ATTRIBS = [
946            "extern",
947            "const",
948        ]
949        OPTIONAL_VAR_ATTR = r"^(?:\b(?:" +"|".join(VAR_ATTRIBS) +r")\b\s*)*"
950
951        #
952        # Store the full prototype before modifying it
953        #
954        source = source
955        full_proto = proto
956        declaration_name = None
957
958        #
959        # Handle macro definitions
960        #
961        macro_prefixes = [
962            KernRe(r"DEFINE_[\w_]+\s*\(([\w_]+)\)"),
963        ]
964
965        for r in macro_prefixes:
966            match = r.search(proto)
967            if match:
968                declaration_name = match.group(1)
969                break
970
971        #
972        # Drop comments and macros to have a pure C prototype
973        #
974        if not declaration_name:
975            proto = self.xforms.apply("var", proto)
976
977        proto = proto.rstrip()
978
979        #
980        # Variable name is at the end of the declaration
981        #
982
983        default_val = None
984
985        r= KernRe(OPTIONAL_VAR_ATTR + r"\s*[\w_\s]*\s+(?:\*+)?([\w_]+)\s*[\d\]\[]*\s*(=.*)?")
986        if r.match(proto):
987            if not declaration_name:
988                declaration_name = r.group(1)
989
990            default_val = r.group(2)
991        else:
992            r= KernRe(OPTIONAL_VAR_ATTR + r"(?:[\w_\s]*)?\s+(?:\*+)?(?:[\w_]+)\s*[\d\]\[]*\s*(=.*)?")
993
994            if r.match(proto):
995                default_val = r.group(1)
996        if not declaration_name:
997           self.emit_msg(ln,f"{proto}: can't parse variable")
998           return
999
1000        if default_val:
1001            default_val = default_val.lstrip("=").strip()
1002
1003        self.output_declaration("var", declaration_name,
1004                                source=source,
1005                                full_proto=full_proto,
1006                                default_val=default_val,
1007                                purpose=self.entry.declaration_purpose)
1008
1009    def dump_declaration(self, ln, prototype, source):
1010        """
1011        Store a data declaration inside self.entries array.
1012        """
1013
1014        if self.entry.decl_type == "enum":
1015            self.dump_enum(ln, prototype, source)
1016        elif self.entry.decl_type == "typedef":
1017            self.dump_typedef(ln, prototype, source)
1018        elif self.entry.decl_type in ["union", "struct"]:
1019            self.dump_struct(ln, prototype, source)
1020        elif self.entry.decl_type == "var":
1021            self.dump_var(ln, prototype, source)
1022        else:
1023            # This would be a bug
1024            self.emit_message(ln, f'Unknown declaration type: {self.entry.decl_type}')
1025
1026    def dump_function(self, ln, prototype, source):
1027        """
1028        Store a function or function macro inside self.entries array.
1029        """
1030
1031        source = source
1032        found = func_macro = False
1033        return_type = ''
1034        decl_type = 'function'
1035
1036        #
1037        # If we have a macro, remove the "#define" at the front.
1038        #
1039        new_proto = KernRe(r"^#\s*define\s+").sub("", prototype)
1040        if new_proto != prototype:
1041            prototype = new_proto
1042            #
1043            # Dispense with the simple "#define A B" case here; the key
1044            # is the space after the name of the symbol being defined.
1045            # NOTE that the seemingly misnamed "func_macro" indicates a
1046            # macro *without* arguments.
1047            #
1048            r = KernRe(r'^(\w+)\s+')
1049            if r.search(prototype):
1050                return_type = ''
1051                declaration_name = r.group(1)
1052                func_macro = True
1053                found = True
1054        else:
1055            #
1056            # Apply the initial transformations.
1057            #
1058            prototype = self.xforms.apply("func", prototype)
1059
1060        # Yes, this truly is vile.  We are looking for:
1061        # 1. Return type (may be nothing if we're looking at a macro)
1062        # 2. Function name
1063        # 3. Function parameters.
1064        #
1065        # All the while we have to watch out for function pointer parameters
1066        # (which IIRC is what the two sections are for), C types (these
1067        # regexps don't even start to express all the possibilities), and
1068        # so on.
1069        #
1070        # If you mess with these regexps, it's a good idea to check that
1071        # the following functions' documentation still comes out right:
1072        # - parport_register_device (function pointer parameters)
1073        # - atomic_set (macro)
1074        # - pci_match_device, __copy_to_user (long return type)
1075
1076        name = r'\w+'
1077        type1 = r'(?:[\w\s]+)?'
1078        type2 = r'(?:[\w\s]+\*+)+'
1079        #
1080        # Attempt to match first on (args) with no internal parentheses; this
1081        # lets us easily filter out __acquires() and other post-args stuff.  If
1082        # that fails, just grab the rest of the line to the last closing
1083        # parenthesis.
1084        #
1085        proto_args = r'\(([^\(]*|.*)\)'
1086        #
1087        # (Except for the simple macro case) attempt to split up the prototype
1088        # in the various ways we understand.
1089        #
1090        if not found:
1091            patterns = [
1092                rf'^()({name})\s*{proto_args}',
1093                rf'^({type1})\s+({name})\s*{proto_args}',
1094                rf'^({type2})\s*({name})\s*{proto_args}',
1095            ]
1096
1097            for p in patterns:
1098                r = KernRe(p)
1099                if r.match(prototype):
1100                    return_type = r.group(1)
1101                    declaration_name = r.group(2)
1102                    args = r.group(3)
1103                    self.create_parameter_list(ln, decl_type, args, ',',
1104                                               declaration_name)
1105                    found = True
1106                    break
1107        #
1108        # Parsing done; make sure that things are as we expect.
1109        #
1110        if not found:
1111            self.emit_msg(ln,
1112                          f"cannot understand function prototype: '{prototype}'")
1113            return
1114        if self.entry.identifier != declaration_name:
1115            self.emit_msg(ln, f"expecting prototype for {self.entry.identifier}(). "
1116                          f"Prototype was for {declaration_name}() instead")
1117            return
1118        self.check_sections(ln, declaration_name, "function")
1119        self.check_return_section(ln, declaration_name, return_type)
1120        #
1121        # Store the result.
1122        #
1123        self.output_declaration(decl_type, declaration_name,
1124                                source=source,
1125                                typedef=('typedef' in return_type),
1126                                functiontype=return_type,
1127                                purpose=self.entry.declaration_purpose,
1128                                func_macro=func_macro)
1129
1130
1131    def dump_typedef(self, ln, proto, source):
1132        """
1133        Store a ``typedef`` inside self.entries array.
1134        """
1135        #
1136        # We start by looking for function typedefs.
1137        #
1138        typedef_type = r'typedef((?:\s+[\w*]+\b){0,7}\s+(?:\w+\b|\*+))\s*'
1139        typedef_ident = r'\*?\s*(\w\S+)\s*'
1140        typedef_args = r'\s*\((.*)\);'
1141
1142        source = source
1143
1144        typedef1 = KernRe(typedef_type + r'\(' + typedef_ident + r'\)' + typedef_args)
1145        typedef2 = KernRe(typedef_type + typedef_ident + typedef_args)
1146
1147        # Parse function typedef prototypes
1148        for r in [typedef1, typedef2]:
1149            if not r.match(proto):
1150                continue
1151
1152            return_type = r.group(1).strip()
1153            declaration_name = r.group(2)
1154            args = r.group(3)
1155
1156            if self.entry.identifier != declaration_name:
1157                self.emit_msg(ln,
1158                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1159                return
1160
1161            self.create_parameter_list(ln, 'function', args, ',', declaration_name)
1162
1163            self.output_declaration('function', declaration_name,
1164                                    source=source,
1165                                    typedef=True,
1166                                    functiontype=return_type,
1167                                    purpose=self.entry.declaration_purpose)
1168            return
1169        #
1170        # Not a function, try to parse a simple typedef.
1171        #
1172        r = KernRe(r'typedef.*\s+(\w+)\s*;')
1173        if r.match(proto):
1174            declaration_name = r.group(1)
1175
1176            if self.entry.identifier != declaration_name:
1177                self.emit_msg(ln,
1178                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1179                return
1180
1181            self.output_declaration('typedef', declaration_name,
1182                                    source=source,
1183                                    purpose=self.entry.declaration_purpose)
1184            return
1185
1186        self.emit_msg(ln, "error: Cannot parse typedef!")
1187
1188    @staticmethod
1189    def process_export(function_set, line):
1190        """
1191        process ``EXPORT_SYMBOL*`` tags
1192
1193        This method doesn't use any variable from the class, so declare it
1194        with a staticmethod decorator.
1195        """
1196
1197        # We support documenting some exported symbols with different
1198        # names.  A horrible hack.
1199        suffixes = [ '_noprof' ]
1200
1201        # Note: it accepts only one EXPORT_SYMBOL* per line, as having
1202        # multiple export lines would violate Kernel coding style.
1203
1204        if export_symbol.search(line):
1205            symbol = export_symbol.group(2)
1206        elif export_symbol_ns.search(line):
1207            symbol = export_symbol_ns.group(2)
1208        else:
1209            return False
1210        #
1211        # Found an export, trim out any special suffixes
1212        #
1213        for suffix in suffixes:
1214            # Be backward compatible with Python < 3.9
1215            if symbol.endswith(suffix):
1216                symbol = symbol[:-len(suffix)]
1217        function_set.add(symbol)
1218        return True
1219
1220    def process_normal(self, ln, line, source):
1221        """
1222        STATE_NORMAL: looking for the ``/**`` to begin everything.
1223        """
1224
1225        if not doc_start.match(line):
1226            return
1227
1228        # start a new entry
1229        self.reset_state(ln)
1230
1231        # next line is always the function name
1232        self.state = state.NAME
1233
1234    def process_name(self, ln, line, source):
1235        """
1236        STATE_NAME: Looking for the "name - description" line
1237        """
1238        #
1239        # Check for a DOC: block and handle them specially.
1240        #
1241        if doc_block.search(line):
1242
1243            if not doc_block.group(1):
1244                self.entry.begin_section(ln, "Introduction")
1245            else:
1246                self.entry.begin_section(ln, doc_block.group(1))
1247
1248            self.entry.identifier = self.entry.section
1249            self.state = state.DOCBLOCK
1250        #
1251        # Otherwise we're looking for a normal kerneldoc declaration line.
1252        #
1253        elif doc_decl.search(line):
1254            self.entry.identifier = doc_decl.group(1)
1255
1256            # Test for data declaration
1257            if doc_begin_data.search(line):
1258                self.entry.decl_type = doc_begin_data.group(1)
1259                self.entry.identifier = doc_begin_data.group(2)
1260            #
1261            # Look for a function description
1262            #
1263            elif doc_begin_func.search(line):
1264                self.entry.identifier = doc_begin_func.group(1)
1265                self.entry.decl_type = "function"
1266            #
1267            # We struck out.
1268            #
1269            else:
1270                self.emit_msg(ln,
1271                              f"This comment starts with '/**', but isn't a kernel-doc comment. Refer to Documentation/doc-guide/kernel-doc.rst\n{line}")
1272                self.state = state.NORMAL
1273                return
1274            #
1275            # OK, set up for a new kerneldoc entry.
1276            #
1277            self.state = state.BODY
1278            self.entry.identifier = self.entry.identifier.strip(" ")
1279            # if there's no @param blocks need to set up default section here
1280            self.entry.begin_section(ln + 1)
1281            #
1282            # Find the description portion, which *should* be there but
1283            # isn't always.
1284            # (We should be able to capture this from the previous parsing - someday)
1285            #
1286            r = KernRe("[-:](.*)")
1287            if r.search(line):
1288                self.entry.declaration_purpose = trim_whitespace(r.group(1))
1289                self.state = state.DECLARATION
1290            else:
1291                self.entry.declaration_purpose = ""
1292
1293            if not self.entry.declaration_purpose and self.config.wshort_desc:
1294                self.emit_msg(ln,
1295                              f"missing initial short description on line:\n{line}")
1296
1297            if not self.entry.identifier and self.entry.decl_type != "enum":
1298                self.emit_msg(ln,
1299                              f"wrong kernel-doc identifier on line:\n{line}")
1300                self.state = state.NORMAL
1301
1302            if self.config.verbose:
1303                self.emit_msg(ln,
1304                              f"Scanning doc for {self.entry.decl_type} {self.entry.identifier}",
1305                                  warning=False)
1306        #
1307        # Failed to find an identifier. Emit a warning
1308        #
1309        else:
1310            self.emit_msg(ln, f"Cannot find identifier on line:\n{line}")
1311
1312    def is_new_section(self, ln, line):
1313        """
1314        Helper function to determine if a new section is being started.
1315        """
1316        if doc_sect.search(line):
1317            self.state = state.BODY
1318            #
1319            # Pick out the name of our new section, tweaking it if need be.
1320            #
1321            newsection = doc_sect.group(1)
1322            if newsection.lower() == 'description':
1323                newsection = 'Description'
1324            elif newsection.lower() == 'context':
1325                newsection = 'Context'
1326                self.state = state.SPECIAL_SECTION
1327            elif newsection.lower() in ["@return", "@returns",
1328                                        "return", "returns"]:
1329                newsection = "Return"
1330                self.state = state.SPECIAL_SECTION
1331            elif newsection[0] == '@':
1332                self.state = state.SPECIAL_SECTION
1333            #
1334            # Initialize the contents, and get the new section going.
1335            #
1336            newcontents = doc_sect.group(2)
1337            if not newcontents:
1338                newcontents = ""
1339            self.dump_section()
1340            self.entry.begin_section(ln, newsection)
1341            self.entry.leading_space = None
1342
1343            self.entry.add_text(newcontents.lstrip())
1344            return True
1345        return False
1346
1347    def is_comment_end(self, ln, line):
1348        """
1349        Helper function to detect (and effect) the end of a kerneldoc comment.
1350        """
1351        if doc_end.search(line):
1352            self.dump_section()
1353
1354            # Look for doc_com + <text> + doc_end:
1355            r = KernRe(r'\s*\*\s*[a-zA-Z_0-9:.]+\*/')
1356            if r.match(line):
1357                self.emit_msg(ln, f"suspicious ending line: {line}")
1358
1359            self.entry.prototype = ""
1360            self.entry.new_start_line = ln + 1
1361
1362            self.state = state.PROTO
1363            return True
1364        return False
1365
1366
1367    def process_decl(self, ln, line, source):
1368        """
1369        STATE_DECLARATION: We've seen the beginning of a declaration.
1370        """
1371        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1372            return
1373        #
1374        # Look for anything with the " * " line beginning.
1375        #
1376        if doc_content.search(line):
1377            cont = doc_content.group(1)
1378            #
1379            # A blank line means that we have moved out of the declaration
1380            # part of the comment (without any "special section" parameter
1381            # descriptions).
1382            #
1383            if cont == "":
1384                self.state = state.BODY
1385            #
1386            # Otherwise we have more of the declaration section to soak up.
1387            #
1388            else:
1389                self.entry.declaration_purpose = \
1390                    trim_whitespace(self.entry.declaration_purpose + ' ' + cont)
1391        else:
1392            # Unknown line, ignore
1393            self.emit_msg(ln, f"bad line: {line}")
1394
1395
1396    def process_special(self, ln, line, source):
1397        """
1398        STATE_SPECIAL_SECTION: a section ending with a blank line.
1399        """
1400        #
1401        # If we have hit a blank line (only the " * " marker), then this
1402        # section is done.
1403        #
1404        if KernRe(r"\s*\*\s*$").match(line):
1405            self.entry.begin_section(ln, dump = True)
1406            self.state = state.BODY
1407            return
1408        #
1409        # Not a blank line, look for the other ways to end the section.
1410        #
1411        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1412            return
1413        #
1414        # OK, we should have a continuation of the text for this section.
1415        #
1416        if doc_content.search(line):
1417            cont = doc_content.group(1)
1418            #
1419            # If the lines of text after the first in a special section have
1420            # leading white space, we need to trim it out or Sphinx will get
1421            # confused.  For the second line (the None case), see what we
1422            # find there and remember it.
1423            #
1424            if self.entry.leading_space is None:
1425                r = KernRe(r'^(\s+)')
1426                if r.match(cont):
1427                    self.entry.leading_space = len(r.group(1))
1428                else:
1429                    self.entry.leading_space = 0
1430            #
1431            # Otherwise, before trimming any leading chars, be *sure*
1432            # that they are white space.  We should maybe warn if this
1433            # isn't the case.
1434            #
1435            for i in range(0, self.entry.leading_space):
1436                if cont[i] != " ":
1437                    self.entry.leading_space = i
1438                    break
1439            #
1440            # Add the trimmed result to the section and we're done.
1441            #
1442            self.entry.add_text(cont[self.entry.leading_space:])
1443        else:
1444            # Unknown line, ignore
1445            self.emit_msg(ln, f"bad line: {line}")
1446
1447    def process_body(self, ln, line, source):
1448        """
1449        STATE_BODY: the bulk of a kerneldoc comment.
1450        """
1451        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1452            return
1453
1454        if doc_content.search(line):
1455            cont = doc_content.group(1)
1456            self.entry.add_text(cont)
1457        else:
1458            # Unknown line, ignore
1459            self.emit_msg(ln, f"bad line: {line}")
1460
1461    def process_inline_name(self, ln, line, source):
1462        """STATE_INLINE_NAME: beginning of docbook comments within a prototype."""
1463
1464        if doc_inline_sect.search(line):
1465            self.entry.begin_section(ln, doc_inline_sect.group(1))
1466            self.entry.add_text(doc_inline_sect.group(2).lstrip())
1467            self.state = state.INLINE_TEXT
1468        elif doc_inline_end.search(line):
1469            self.dump_section()
1470            self.state = state.PROTO
1471        elif doc_content.search(line):
1472            self.emit_msg(ln, f"Incorrect use of kernel-doc format: {line}")
1473            self.state = state.PROTO
1474
1475            #
1476            # Don't let it add partial comments at the code, as breaks the
1477            # logic meant to remove comments from prototypes.
1478            #
1479            self.process_proto_type(ln, "/**\n" + line, source)
1480        # else ... ??
1481
1482    def process_inline_text(self, ln, line, source):
1483        """STATE_INLINE_TEXT: docbook comments within a prototype."""
1484
1485        if doc_inline_end.search(line):
1486            self.dump_section()
1487            self.state = state.PROTO
1488        elif doc_content.search(line):
1489            self.entry.add_text(doc_content.group(1))
1490        # else ... ??
1491
1492    def syscall_munge(self, ln, proto):         # pylint: disable=W0613
1493        """
1494        Handle syscall definitions.
1495        """
1496
1497        is_void = False
1498
1499        # Strip newlines/CR's
1500        proto = re.sub(r'[\r\n]+', ' ', proto)
1501
1502        # Check if it's a SYSCALL_DEFINE0
1503        if 'SYSCALL_DEFINE0' in proto:
1504            is_void = True
1505
1506        # Replace SYSCALL_DEFINE with correct return type & function name
1507        proto = KernRe(r'SYSCALL_DEFINE.*\(').sub('long sys_', proto)
1508
1509        r = KernRe(r'long\s+(sys_.*?),')
1510        if r.search(proto):
1511            proto = KernRe(',').sub('(', proto, count=1)
1512        elif is_void:
1513            proto = KernRe(r'\)').sub('(void)', proto, count=1)
1514
1515        # Now delete all of the odd-numbered commas in the proto
1516        # so that argument types & names don't have a comma between them
1517        count = 0
1518        length = len(proto)
1519
1520        if is_void:
1521            length = 0  # skip the loop if is_void
1522
1523        for ix in range(length):
1524            if proto[ix] == ',':
1525                count += 1
1526                if count % 2 == 1:
1527                    proto = proto[:ix] + ' ' + proto[ix + 1:]
1528
1529        return proto
1530
1531    def tracepoint_munge(self, ln, proto):
1532        """
1533        Handle tracepoint definitions.
1534        """
1535
1536        tracepointname = None
1537        tracepointargs = None
1538
1539        # Match tracepoint name based on different patterns
1540        r = KernRe(r'TRACE_EVENT\((.*?),')
1541        if r.search(proto):
1542            tracepointname = r.group(1)
1543
1544        r = KernRe(r'DEFINE_SINGLE_EVENT\((.*?),')
1545        if r.search(proto):
1546            tracepointname = r.group(1)
1547
1548        r = KernRe(r'DEFINE_EVENT\((.*?),(.*?),')
1549        if r.search(proto):
1550            tracepointname = r.group(2)
1551
1552        if tracepointname:
1553            tracepointname = tracepointname.lstrip()
1554
1555        r = KernRe(r'TP_PROTO\((.*?)\)')
1556        if r.search(proto):
1557            tracepointargs = r.group(1)
1558
1559        if not tracepointname or not tracepointargs:
1560            self.emit_msg(ln,
1561                          f"Unrecognized tracepoint format:\n{proto}\n")
1562        else:
1563            proto = f"static inline void trace_{tracepointname}({tracepointargs})"
1564            self.entry.identifier = f"trace_{self.entry.identifier}"
1565
1566        return proto
1567
1568    def process_proto_function(self, ln, line, source):
1569        """Ancillary routine to process a function prototype."""
1570
1571        # strip C99-style comments to end of line
1572        line = KernRe(r"//.*$", re.S).sub('', line)
1573        #
1574        # Soak up the line's worth of prototype text, stopping at { or ; if present.
1575        #
1576        if KernRe(r'\s*#\s*define').match(line):
1577            self.entry.prototype = line
1578        elif not line.startswith('#'):   # skip other preprocessor stuff
1579            r = KernRe(r'([^\{]*)')
1580            if r.match(line):
1581                self.entry.prototype += r.group(1) + " "
1582        #
1583        # If we now have the whole prototype, clean it up and declare victory.
1584        #
1585        if '{' in line or ';' in line or KernRe(r'\s*#\s*define').match(line):
1586            # strip comments and surrounding spaces
1587            self.entry.prototype = KernRe(r'/\*.*\*/').sub('', self.entry.prototype).strip()
1588            #
1589            # Handle self.entry.prototypes for function pointers like:
1590            #       int (*pcs_config)(struct foo)
1591            # by turning it into
1592            #	    int pcs_config(struct foo)
1593            #
1594            r = KernRe(r'^(\S+\s+)\(\s*\*(\S+)\)')
1595            self.entry.prototype = r.sub(r'\1\2', self.entry.prototype)
1596            #
1597            # Handle special declaration syntaxes
1598            #
1599            if 'SYSCALL_DEFINE' in self.entry.prototype:
1600                self.entry.prototype = self.syscall_munge(ln,
1601                                                          self.entry.prototype)
1602            else:
1603                r = KernRe(r'TRACE_EVENT|DEFINE_EVENT|DEFINE_SINGLE_EVENT')
1604                if r.search(self.entry.prototype):
1605                    self.entry.prototype = self.tracepoint_munge(ln,
1606                                                                 self.entry.prototype)
1607            #
1608            # ... and we're done
1609            #
1610            self.dump_function(ln, self.entry.prototype, source)
1611            self.reset_state(ln)
1612
1613    def process_proto_type(self, ln, line, source):
1614        """
1615        Ancillary routine to process a type.
1616        """
1617
1618        # Strip C99-style comments and surrounding whitespace
1619        line = KernRe(r"//.*$", re.S).sub('', line).strip()
1620        if not line:
1621            return # nothing to see here
1622
1623        # To distinguish preprocessor directive from regular declaration later.
1624        if line.startswith('#'):
1625            line += ";"
1626        #
1627        # Split the declaration on any of { } or ;, and accumulate pieces
1628        # until we hit a semicolon while not inside {brackets}
1629        #
1630        r = KernRe(r'(.*?)([{};])')
1631        for chunk in r.split(line):
1632            if chunk:  # Ignore empty matches
1633                self.entry.prototype += chunk
1634                #
1635                # This cries out for a match statement ... someday after we can
1636                # drop Python 3.9 ...
1637                #
1638                if chunk == '{':
1639                    self.entry.brcount += 1
1640                elif chunk == '}':
1641                    self.entry.brcount -= 1
1642                elif chunk == ';' and self.entry.brcount <= 0:
1643                    self.dump_declaration(ln, self.entry.prototype, source)
1644                    self.reset_state(ln)
1645                    return
1646        #
1647        # We hit the end of the line while still in the declaration; put
1648        # in a space to represent the newline.
1649        #
1650        self.entry.prototype += ' '
1651
1652    def process_proto(self, ln, line, source):
1653        """STATE_PROTO: reading a function/whatever prototype."""
1654
1655        if doc_inline_oneline.search(line):
1656            self.entry.begin_section(ln, doc_inline_oneline.group(1))
1657            self.entry.add_text(doc_inline_oneline.group(2))
1658            self.dump_section()
1659
1660        elif doc_inline_start.search(line):
1661            self.state = state.INLINE_NAME
1662
1663        elif self.entry.decl_type == 'function':
1664            self.process_proto_function(ln, line, source)
1665
1666        else:
1667            self.process_proto_type(ln, line, source)
1668
1669    def process_docblock(self, ln, line, source):
1670        """STATE_DOCBLOCK: within a ``DOC:`` block."""
1671
1672        if doc_end.search(line):
1673            self.dump_section()
1674            self.output_declaration("doc", self.entry.identifier,
1675                                    source=source)
1676            self.reset_state(ln)
1677
1678        elif doc_content.search(line):
1679            self.entry.add_text(doc_content.group(1))
1680
1681    def parse_export(self):
1682        """
1683        Parses ``EXPORT_SYMBOL*`` macros from a single Kernel source file.
1684        """
1685
1686        export_table = set()
1687
1688        try:
1689            with open(self.fname, "r", encoding="utf8",
1690                      errors="backslashreplace") as fp:
1691
1692                for line in fp:
1693                    self.process_export(export_table, line)
1694
1695        except IOError:
1696            return None
1697
1698        return export_table
1699
1700    #: The state/action table telling us which function to invoke in each state.
1701    state_actions = {
1702        state.NORMAL:			process_normal,
1703        state.NAME:			process_name,
1704        state.BODY:			process_body,
1705        state.DECLARATION:		process_decl,
1706        state.SPECIAL_SECTION:		process_special,
1707        state.INLINE_NAME:		process_inline_name,
1708        state.INLINE_TEXT:		process_inline_text,
1709        state.PROTO:			process_proto,
1710        state.DOCBLOCK:			process_docblock,
1711        }
1712
1713    def parse_kdoc(self):
1714        """
1715        Open and process each line of a C source file.
1716        The parsing is controlled via a state machine, and the line is passed
1717        to a different process function depending on the state. The process
1718        function may update the state as needed.
1719
1720        Besides parsing kernel-doc tags, it also parses export symbols.
1721        """
1722
1723        prev = ""
1724        prev_ln = None
1725        export_table = set()
1726        self.state = state.NORMAL
1727        source = ""
1728
1729        try:
1730            with open(self.fname, "r", encoding="utf8",
1731                      errors="backslashreplace") as fp:
1732                for ln, line in enumerate(fp):
1733
1734                    line = line.expandtabs().strip("\n")
1735
1736                    # Group continuation lines on prototypes
1737                    if self.state == state.PROTO:
1738                        if line.endswith("\\"):
1739                            prev += line.rstrip("\\")
1740                            if not prev_ln:
1741                                prev_ln = ln
1742                            continue
1743
1744                        if prev:
1745                            ln = prev_ln
1746                            line = prev + line
1747                            prev = ""
1748                            prev_ln = None
1749
1750                    self.config.log.debug("%d %s: %s",
1751                                          ln, state.name[self.state],
1752                                          line)
1753
1754                    if self.store_src:
1755                        if source and self.state == state.NORMAL:
1756                            source = ""
1757                        elif self.state != state.NORMAL:
1758                            source += line + "\n"
1759
1760                    # This is an optimization over the original script.
1761                    # There, when export_file was used for the same file,
1762                    # it was read twice. Here, we use the already-existing
1763                    # loop to parse exported symbols as well.
1764                    #
1765                    if (self.state != state.NORMAL) or \
1766                       not self.process_export(export_table, line):
1767                        prev_state = self.state
1768                        # Hand this line to the appropriate state handler
1769                        self.state_actions[self.state](self, ln, line, source)
1770                        if prev_state == state.NORMAL and self.state != state.NORMAL:
1771                            source += line + "\n"
1772
1773            self.emit_unused_warnings()
1774
1775        except OSError:
1776            self.config.log.error(f"Error: Cannot open file {self.fname}")
1777
1778        return export_table, self.entries
1779