xref: /linux/tools/lib/python/kdoc/kdoc_parser.py (revision 024e200e2a89d71dceff7d1bff8ae77b145726e0)
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.section_start_lines = {}
144        self.sections = {}
145
146        self.anon_struct_union = False
147
148        self.leading_space = None
149
150        self.fname = fname
151
152        # State flags
153        self.brcount = 0
154        self.declaration_start_line = ln + 1
155
156    #
157    # Management of section contents
158    #
159    def add_text(self, text):
160        """Add a new text to the entry contents list."""
161        self._contents.append(text)
162
163    def contents(self):
164        """Returns a string with all content texts that were added."""
165        return '\n'.join(self._contents) + '\n'
166
167    # TODO: rename to emit_message after removal of kernel-doc.pl
168    def emit_msg(self, ln, msg, *, warning=True):
169        """Emit a message."""
170
171        log_msg = f"{self.fname}:{ln} {msg}"
172
173        if not warning:
174            self.config.log.info(log_msg)
175            return
176
177        # Delegate warning output to output logic, as this way it
178        # will report warnings/info only for symbols that are output
179
180        self.warnings.append(log_msg)
181        return
182
183    def begin_section(self, line_no, title = SECTION_DEFAULT, dump = False):
184        """
185        Begin a new section.
186        """
187        if dump:
188            self.dump_section(start_new = True)
189        self.section = title
190        self.new_start_line = line_no
191
192    def dump_section(self, start_new=True):
193        """
194        Dumps section contents to arrays/hashes intended for that purpose.
195        """
196        #
197        # If we have accumulated no contents in the default ("description")
198        # section, don't bother.
199        #
200        if self.section == SECTION_DEFAULT and not self._contents:
201            return
202        name = self.section
203        contents = self.contents()
204
205        if type_param.match(name):
206            name = type_param.group(1)
207
208            self.parameterdescs[name] = contents
209            self.parameterdesc_start_lines[name] = self.new_start_line
210
211            self.new_start_line = 0
212
213        else:
214            if name in self.sections and self.sections[name] != "":
215                # Only warn on user-specified duplicate section names
216                if name != SECTION_DEFAULT:
217                    self.emit_msg(self.new_start_line,
218                                  f"duplicate section name '{name}'")
219                # Treat as a new paragraph - add a blank line
220                self.sections[name] += '\n' + contents
221            else:
222                self.sections[name] = contents
223                self.section_start_lines[name] = self.new_start_line
224                self.new_start_line = 0
225
226#        self.config.log.debug("Section: %s : %s", name, pformat(vars(self)))
227
228        if start_new:
229            self.section = SECTION_DEFAULT
230            self._contents = []
231
232python_warning = False
233
234class KernelDoc:
235    """
236    Read a C language source or header FILE and extract embedded
237    documentation comments.
238    """
239
240    #: Name of context section.
241    section_context = "Context"
242
243    #: Name of return section.
244    section_return = "Return"
245
246    #: String to write when a parameter is not described.
247    undescribed = "-- undescribed --"
248
249    def __init__(self, config, fname, xforms):
250        """Initialize internal variables"""
251
252        self.fname = fname
253        self.config = config
254        self.xforms = xforms
255
256        tokenizer_set_log(self.config.log, f"{self.fname}: CMatch: ")
257
258        # Initial state for the state machines
259        self.state = state.NORMAL
260
261        # Store entry currently being processed
262        self.entry = None
263
264        # Place all potential outputs into an array
265        self.entries = []
266
267        #
268        # We need Python 3.7 for its "dicts remember the insertion
269        # order" guarantee
270        #
271        global python_warning
272        if (not python_warning and
273            sys.version_info.major == 3 and sys.version_info.minor < 7):
274
275            self.emit_msg(0,
276                          'Python 3.7 or later is required for correct results')
277            python_warning = True
278
279    def emit_msg(self, ln, msg, *, warning=True):
280        """Emit a message"""
281
282        if self.entry:
283            self.entry.emit_msg(ln, msg, warning=warning)
284            return
285
286        log_msg = f"{self.fname}:{ln} {msg}"
287
288        if warning:
289            self.config.log.warning(log_msg)
290        else:
291            self.config.log.info(log_msg)
292
293    def dump_section(self, start_new=True):
294        """
295        Dump section contents to arrays/hashes intended for that purpose.
296        """
297
298        if self.entry:
299            self.entry.dump_section(start_new)
300
301    # TODO: rename it to store_declaration after removal of kernel-doc.pl
302    def output_declaration(self, dtype, name, **args):
303        """
304        Store the entry into an entry array.
305
306        The actual output and output filters will be handled elsewhere.
307        """
308
309        item = KdocItem(name, self.fname, dtype,
310                        self.entry.declaration_start_line, **args)
311        item.warnings = self.entry.warnings
312
313        # Drop empty sections
314        # TODO: improve empty sections logic to emit warnings
315        sections = self.entry.sections
316        for section in ["Description", "Return"]:
317            if section in sections and not sections[section].rstrip():
318                del sections[section]
319        item.set_sections(sections, self.entry.section_start_lines)
320        item.set_params(self.entry.parameterlist, self.entry.parameterdescs,
321                        self.entry.parametertypes,
322                        self.entry.parameterdesc_start_lines)
323        self.entries.append(item)
324
325        self.config.log.debug("Output: %s:%s = %s", dtype, name, pformat(args))
326
327    def emit_unused_warnings(self):
328        """
329        When the parser fails to produce a valid entry, it places some
330        warnings under `entry.warnings` that will be discarded when resetting
331        the state.
332
333        Ensure that those warnings are not lost.
334
335        .. note::
336
337              Because we are calling `config.warning()` here, those
338              warnings are not filtered by the `-W` parameters: they will all
339              be produced even when `-Wreturn`, `-Wshort-desc`, and/or
340              `-Wcontents-before-sections` are used.
341
342              Allowing those warnings to be filtered is complex, because it
343              would require storing them in a buffer and then filtering them
344              during the output step of the code, depending on the
345              selected symbols.
346        """
347        if self.entry and self.entry not in self.entries:
348            for log_msg in self.entry.warnings:
349                self.config.warning(log_msg)
350
351    def reset_state(self, ln):
352        """
353        Ancillary routine to create a new entry. It initializes all
354        variables used by the state machine.
355        """
356
357        self.emit_unused_warnings()
358
359        self.entry = KernelEntry(self.config, self.fname, ln)
360
361        # State flags
362        self.state = state.NORMAL
363
364    def push_parameter(self, ln, decl_type, param, dtype,
365                       org_arg, declaration_name):
366        """
367        Store parameters and their descriptions at self.entry.
368        """
369
370        if self.entry.anon_struct_union and dtype == "" and param == "}":
371            return  # Ignore the ending }; from anonymous struct/union
372
373        self.entry.anon_struct_union = False
374
375        param = KernRe(r'[\[\)].*').sub('', param, count=1)
376
377        #
378        # Look at various "anonymous type" cases.
379        #
380        if dtype == '':
381            if param.endswith("..."):
382                if len(param) > 3: # there is a name provided, use that
383                    param = param[:-3]
384                if not self.entry.parameterdescs.get(param):
385                    self.entry.parameterdescs[param] = "variable arguments"
386
387            elif (not param) or param == "void":
388                param = "void"
389                self.entry.parameterdescs[param] = "no arguments"
390
391            elif param in ["struct", "union"]:
392                # Handle unnamed (anonymous) union or struct
393                dtype = param
394                param = "{unnamed_" + param + "}"
395                self.entry.parameterdescs[param] = "anonymous\n"
396                self.entry.anon_struct_union = True
397
398        # Warn if parameter has no description
399        # (but ignore ones starting with # as these are not parameters
400        # but inline preprocessor statements)
401        if param not in self.entry.parameterdescs and not param.startswith("#"):
402            self.entry.parameterdescs[param] = self.undescribed
403
404            if "." not in param:
405                if decl_type == 'function':
406                    dname = f"{decl_type} parameter"
407                else:
408                    dname = f"{decl_type} member"
409
410                self.emit_msg(ln,
411                              f"{dname} '{param}' not described in '{declaration_name}'")
412
413        # Strip spaces from param so that it is one continuous string on
414        # parameterlist. This fixes a problem where check_sections()
415        # cannot find a parameter like "addr[6 + 2]" because it actually
416        # appears as "addr[6", "+", "2]" on the parameter list.
417        # However, it's better to maintain the param string unchanged for
418        # output, so just weaken the string compare in check_sections()
419        # to ignore "[blah" in a parameter string.
420
421        self.entry.parameterlist.append(param)
422        org_arg = KernRe(r'\s\s+').sub(' ', org_arg)
423        self.entry.parametertypes[param] = org_arg
424
425
426    def create_parameter_list(self, ln, decl_type, args,
427                              splitter, declaration_name):
428        """
429        Creates a list of parameters, storing them at self.entry.
430        """
431
432        # temporarily replace all commas inside function pointer definition
433        arg_expr = KernRe(r'(\([^\),]+),')
434        while arg_expr.search(args):
435            args = arg_expr.sub(r"\1#", args)
436
437        for arg in args.split(splitter):
438            # Ignore argument attributes
439            arg = KernRe(r'\sPOS0?\s').sub(' ', arg)
440
441            # Strip leading/trailing spaces
442            arg = arg.strip()
443            arg = KernRe(r'\s+').sub(' ', arg, count=1)
444
445            if arg.startswith('#'):
446                # Treat preprocessor directive as a typeless variable just to fill
447                # corresponding data structures "correctly". Catch it later in
448                # output_* subs.
449
450                # Treat preprocessor directive as a typeless variable
451                self.push_parameter(ln, decl_type, arg, "",
452                                    "", declaration_name)
453            #
454            # The pointer-to-function case.
455            #
456            elif KernRe(r'\(.+\)\s*\(').search(arg):
457                arg = arg.replace('#', ',')
458                r = KernRe(r'[^\(]+\(\*?\s*'  # Everything up to "(*"
459                           r'([\w\[\].]*)'    # Capture the name and possible [array]
460                           r'\s*\)')	      # Make sure the trailing ")" is there
461                if r.match(arg):
462                    param = r.group(1)
463                else:
464                    self.emit_msg(ln, f"Invalid param: {arg}")
465                    param = arg
466                dtype = arg.replace(param, '')
467                self.push_parameter(ln, decl_type, param, dtype, arg, declaration_name)
468            #
469            # The array-of-pointers case.  Dig the parameter name out from the middle
470            # of the declaration.
471            #
472            elif KernRe(r'\(.+\)\s*\[').search(arg):
473                r = KernRe(r'[^\(]+\(\s*\*\s*'		# Up to "(" and maybe "*"
474                           r'([\w.]*?)'			# The actual pointer name
475                           r'\s*(\[\s*\w+\s*\]\s*)*\)') # The [array portion]
476                if r.match(arg):
477                    param = r.group(1)
478                else:
479                    self.emit_msg(ln, f"Invalid param: {arg}")
480                    param = arg
481                dtype = arg.replace(param, '')
482                self.push_parameter(ln, decl_type, param, dtype, arg, declaration_name)
483            elif arg:
484                #
485                # Clean up extraneous spaces and split the string at commas; the first
486                # element of the resulting list will also include the type information.
487                #
488                arg = KernRe(r'\s*:\s*').sub(":", arg)
489                arg = KernRe(r'\s*\[').sub('[', arg)
490                args = KernRe(r'\s*,\s*').split(arg)
491                args[0] = re.sub(r'(\*+)\s*', r' \1', args[0])
492                #
493                # args[0] has a string of "type a".  If "a" includes an [array]
494                # declaration, we want to not be fooled by any white space inside
495                # the brackets, so detect and handle that case specially.
496                #
497                r = KernRe(r'^([^[\]]*\s+)(.*)$')
498                if r.match(args[0]):
499                    args[0] = r.group(2)
500                    dtype = r.group(1)
501                else:
502                    # No space in args[0]; this seems wrong but preserves previous behavior
503                    dtype = ''
504
505                bitfield_re = KernRe(r'(.*?):(\w+)')
506                for param in args:
507                    #
508                    # For pointers, shift the star(s) from the variable name to the
509                    # type declaration.
510                    #
511                    r = KernRe(r'^(\*+)\s*(.*)')
512                    if r.match(param):
513                        self.push_parameter(ln, decl_type, r.group(2),
514                                            f"{dtype} {r.group(1)}",
515                                            arg, declaration_name)
516                    #
517                    # Perform a similar shift for bitfields.
518                    #
519                    elif bitfield_re.search(param):
520                        if dtype != "":  # Skip unnamed bit-fields
521                            self.push_parameter(ln, decl_type, bitfield_re.group(1),
522                                                f"{dtype}:{bitfield_re.group(2)}",
523                                                arg, declaration_name)
524                    else:
525                        self.push_parameter(ln, decl_type, param, dtype,
526                                            arg, declaration_name)
527
528    def check_sections(self, ln, decl_name, decl_type):
529        """
530        Check for errors inside sections, emitting warnings if not found
531        parameters are described.
532        """
533        for section in self.entry.sections:
534            if section not in self.entry.parameterlist and \
535               not known_sections.search(section):
536                if decl_type == 'function':
537                    dname = f"{decl_type} parameter"
538                else:
539                    dname = f"{decl_type} member"
540                self.emit_msg(ln,
541                              f"Excess {dname} '{section}' description in '{decl_name}'")
542
543    def check_return_section(self, ln, declaration_name, return_type):
544        """
545        If the function doesn't return void, warns about the lack of a
546        return description.
547        """
548
549        if not self.config.wreturn:
550            return
551
552        # Ignore an empty return type (It's a macro)
553        # Ignore functions with a "void" return type (but not "void *")
554        if not return_type or KernRe(r'void\s*\w*\s*$').search(return_type):
555            return
556
557        if not self.entry.sections.get("Return", None):
558            self.emit_msg(ln,
559                          f"No description found for return value of '{declaration_name}'")
560
561    def split_struct_proto(self, proto):
562        """
563        Split apart a structure prototype; returns (struct|union, name,
564        members) or ``None``.
565        """
566
567        type_pattern = r'(struct|union)'
568        qualifiers = [
569            "__attribute__",
570            "__packed",
571            "__aligned",
572            "____cacheline_aligned_in_smp",
573            "____cacheline_aligned",
574        ]
575        definition_body = r'\{(.*)\}\s*' + "(?:" + '|'.join(qualifiers) + ")?"
576
577        r = KernRe(type_pattern + r'\s+(\w+)\s*' + definition_body)
578        if r.search(proto):
579            return (r.group(1), r.group(2), r.group(3))
580        else:
581            r = KernRe(r'typedef\s+' + type_pattern + r'\s*' + definition_body + r'\s*(\w+)\s*;')
582            if r.search(proto):
583                return (r.group(1), r.group(3), r.group(2))
584        return None
585
586    def rewrite_struct_members(self, members):
587        """
588        Process ``struct``/``union`` members from the most deeply nested
589        outward.
590
591        Rewrite the members of a ``struct`` or ``union`` for easier formatting
592        later on. Among other things, this function will turn a member like::
593
594          struct { inner_members; } foo;
595
596        into::
597
598          struct foo; inner_members;
599        """
600
601        #
602        # The trick is in the ``^{`` below - it prevents a match of an outer
603        # ``struct``/``union`` until the inner one has been munged
604        # (removing the ``{`` in the process).
605        #
606        struct_members = KernRe(r'(struct|union)'   # 0: declaration type
607                                r'([^\{\};]+)' 	    # 1: possible name
608                                r'(\{)'
609                                r'([^\{\}]*)'       # 3: Contents of declaration
610                                r'(\})'
611                                r'([^\{\};]*)(;)')  # 5: Remaining stuff after declaration
612        tuples = struct_members.findall(members)
613        while tuples:
614            for t in tuples:
615                newmember = ""
616                oldmember = "".join(t) # Reconstruct the original formatting
617                dtype, name, lbr, content, rbr, rest, semi = t
618                #
619                # Pass through each field name, normalizing the form and formatting.
620                #
621                for s_id in rest.split(','):
622                    s_id = s_id.strip()
623                    newmember += f"{dtype} {s_id}; "
624                    #
625                    # Remove bitfield/array/pointer info, getting the bare name.
626                    #
627                    s_id = KernRe(r'[:\[].*').sub('', s_id)
628                    s_id = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', s_id)
629                    #
630                    # Pass through the members of this inner structure/union.
631                    #
632                    for arg in content.split(';'):
633                        arg = arg.strip()
634                        #
635                        # Look for (type)(*name)(args) - pointer to function
636                        #
637                        r = KernRe(r'^([^\(]+\(\*?\s*)([\w.]*)(\s*\).*)')
638                        if r.match(arg):
639                            dtype, name, extra = r.group(1), r.group(2), r.group(3)
640                            # Pointer-to-function
641                            if not s_id:
642                                # Anonymous struct/union
643                                newmember += f"{dtype}{name}{extra}; "
644                            else:
645                                newmember += f"{dtype}{s_id}.{name}{extra}; "
646                        #
647                        # Otherwise a non-function member.
648                        #
649                        else:
650                            #
651                            # Remove bitmap and array portions and spaces around commas
652                            #
653                            arg = KernRe(r':\s*\d+\s*').sub('', arg)
654                            arg = KernRe(r'\[.*\]').sub('', arg)
655                            arg = KernRe(r'\s*,\s*').sub(',', arg)
656                            #
657                            # Look for a normal decl - "type name[,name...]"
658                            #
659                            r = KernRe(r'(.*)\s+([\S+,]+)')
660                            if r.search(arg):
661                                for name in r.group(2).split(','):
662                                    name = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', name)
663                                    if not s_id:
664                                        # Anonymous struct/union
665                                        newmember += f"{r.group(1)} {name}; "
666                                    else:
667                                        newmember += f"{r.group(1)} {s_id}.{name}; "
668                            else:
669                                newmember += f"{arg}; "
670                #
671                # At the end of the s_id loop, replace the original declaration with
672                # the munged version.
673                #
674                members = members.replace(oldmember, newmember)
675            #
676            # End of the tuple loop - search again and see if there are outer members
677            # that now turn up.
678            #
679            tuples = struct_members.findall(members)
680        return members
681
682    def format_struct_decl(self, declaration):
683        """
684        Format the ``struct`` declaration into a standard form for inclusion
685        in the resulting docs.
686        """
687
688        #
689        # Insert newlines, get rid of extra spaces.
690        #
691        declaration = KernRe(r'([\{;])').sub(r'\1\n', declaration)
692        declaration = KernRe(r'\}\s+;').sub('};', declaration)
693        #
694        # Format inline enums with each member on its own line.
695        #
696        r = KernRe(r'(enum\s+\{[^\}]+),([^\n])')
697        while r.search(declaration):
698            declaration = r.sub(r'\1,\n\2', declaration)
699        #
700        # Now go through and supply the right number of tabs
701        # for each line.
702        #
703        def_args = declaration.split('\n')
704        level = 1
705        declaration = ""
706        for clause in def_args:
707            clause = KernRe(r'\s+').sub(' ', clause.strip(), count=1)
708            if clause:
709                if '}' in clause and level > 1:
710                    level -= 1
711                if not clause.startswith('#'):
712                    declaration += "\t" * level
713                declaration += "\t" + clause + "\n"
714                if "{" in clause and "}" not in clause:
715                    level += 1
716        return declaration
717
718
719    def dump_struct(self, ln, proto):
720        """
721        Store an entry for a ``struct`` or ``union``
722        """
723        #
724        # Do the basic parse to get the pieces of the declaration.
725        #
726        struct_parts = self.split_struct_proto(proto)
727        if not struct_parts:
728            self.emit_msg(ln, f"{proto} error: Cannot parse struct or union!")
729            return
730        decl_type, declaration_name, members = struct_parts
731
732        if self.entry.identifier != declaration_name:
733            self.emit_msg(ln, f"expecting prototype for {decl_type} {self.entry.identifier}. "
734                          f"Prototype was for {decl_type} {declaration_name} instead\n")
735            return
736        #
737        # Go through the list of members applying all of our transformations.
738        #
739        members = trim_private_members(members)
740        members = self.xforms.apply("struct", members)
741
742        #
743        # Deal with embedded struct and union members, and drop enums entirely.
744        #
745        declaration = members
746        members = self.rewrite_struct_members(members)
747        members = re.sub(r'(\{[^\{\}]*\})', '', members)
748        #
749        # Output the result and we are done.
750        #
751        self.create_parameter_list(ln, decl_type, members, ';',
752                                   declaration_name)
753        self.check_sections(ln, declaration_name, decl_type)
754        self.output_declaration(decl_type, declaration_name,
755                                definition=self.format_struct_decl(declaration),
756                                purpose=self.entry.declaration_purpose)
757
758    def dump_enum(self, ln, proto):
759        """
760        Store an ``enum`` inside self.entries array.
761        """
762        #
763        # Strip preprocessor directives.  Note that this depends on the
764        # trailing semicolon we added in process_proto_type().
765        #
766        proto = KernRe(r'#\s*((define|ifdef|if)\s+|endif)[^;]*;', flags=re.S).sub('', proto)
767        #
768        # Parse out the name and members of the enum.  Typedef form first.
769        #
770        r = KernRe(r'typedef\s+enum\s*\{(.*)\}\s*(\w*)\s*;')
771        if r.search(proto):
772            declaration_name = r.group(2)
773            members = trim_private_members(r.group(1))
774        #
775        # Failing that, look for a straight enum
776        #
777        else:
778            r = KernRe(r'enum\s+(\w*)\s*\{(.*)\}')
779            if r.match(proto):
780                declaration_name = r.group(1)
781                members = trim_private_members(r.group(2))
782        #
783        # OK, this isn't going to work.
784        #
785            else:
786                self.emit_msg(ln, f"{proto}: error: Cannot parse enum!")
787                return
788        #
789        # Make sure we found what we were expecting.
790        #
791        if self.entry.identifier != declaration_name:
792            if self.entry.identifier == "":
793                self.emit_msg(ln,
794                              f"{proto}: wrong kernel-doc identifier on prototype")
795            else:
796                self.emit_msg(ln,
797                              f"expecting prototype for enum {self.entry.identifier}. "
798                              f"Prototype was for enum {declaration_name} instead")
799            return
800
801        if not declaration_name:
802            declaration_name = "(anonymous)"
803        #
804        # Parse out the name of each enum member, and verify that we
805        # have a description for it.
806        #
807        member_set = set()
808        members = KernRe(r'\([^;)]*\)').sub('', members)
809        for arg in members.split(','):
810            arg = KernRe(r'^\s*(\w+).*').sub(r'\1', arg)
811            if not arg.strip():
812                continue
813
814            self.entry.parameterlist.append(arg)
815            if arg not in self.entry.parameterdescs:
816                self.entry.parameterdescs[arg] = self.undescribed
817                self.emit_msg(ln,
818                              f"Enum value '{arg}' not described in enum '{declaration_name}'")
819            member_set.add(arg)
820        #
821        # Ensure that every described member actually exists in the enum.
822        #
823        for k in self.entry.parameterdescs:
824            if k not in member_set:
825                self.emit_msg(ln,
826                              f"Excess enum value '@{k}' description in '{declaration_name}'")
827
828        self.output_declaration('enum', declaration_name,
829                                purpose=self.entry.declaration_purpose)
830
831    def dump_var(self, ln, proto):
832        """
833        Store variables that are part of kAPI.
834        """
835        VAR_ATTRIBS = [
836            "extern",
837            "const",
838        ]
839        OPTIONAL_VAR_ATTR = r"^(?:\b(?:" +"|".join(VAR_ATTRIBS) +r")\b\s*)*"
840
841        #
842        # Store the full prototype before modifying it
843        #
844        full_proto = proto
845        declaration_name = None
846
847        #
848        # Handle macro definitions
849        #
850        macro_prefixes = [
851            KernRe(r"DEFINE_[\w_]+\s*\(([\w_]+)\)"),
852        ]
853
854        for r in macro_prefixes:
855            match = r.search(proto)
856            if match:
857                declaration_name = match.group(1)
858                break
859
860        #
861        # Drop comments and macros to have a pure C prototype
862        #
863        if not declaration_name:
864            proto = self.xforms.apply("var", proto)
865
866        proto = proto.rstrip()
867
868        #
869        # Variable name is at the end of the declaration
870        #
871
872        default_val = None
873
874        r= KernRe(OPTIONAL_VAR_ATTR + r"\s*[\w_\s]*\s+(?:\*+)?([\w_]+)\s*[\d\]\[]*\s*(=.*)?")
875        if r.match(proto):
876            if not declaration_name:
877                declaration_name = r.group(1)
878
879            default_val = r.group(2)
880        else:
881            r= KernRe(OPTIONAL_VAR_ATTR + r"(?:[\w_\s]*)?\s+(?:\*+)?(?:[\w_]+)\s*[\d\]\[]*\s*(=.*)?")
882
883            if r.match(proto):
884                default_val = r.group(1)
885        if not declaration_name:
886           self.emit_msg(ln,f"{proto}: can't parse variable")
887           return
888
889        if default_val:
890            default_val = default_val.lstrip("=").strip()
891
892        self.output_declaration("var", declaration_name,
893                                full_proto=full_proto,
894                                default_val=default_val,
895                                purpose=self.entry.declaration_purpose)
896
897    def dump_declaration(self, ln, prototype):
898        """
899        Store a data declaration inside self.entries array.
900        """
901
902        if self.entry.decl_type == "enum":
903            self.dump_enum(ln, prototype)
904        elif self.entry.decl_type == "typedef":
905            self.dump_typedef(ln, prototype)
906        elif self.entry.decl_type in ["union", "struct"]:
907            self.dump_struct(ln, prototype)
908        elif self.entry.decl_type == "var":
909            self.dump_var(ln, prototype)
910        else:
911            # This would be a bug
912            self.emit_message(ln, f'Unknown declaration type: {self.entry.decl_type}')
913
914    def dump_function(self, ln, prototype):
915        """
916        Store a function or function macro inside self.entries array.
917        """
918
919        found = func_macro = False
920        return_type = ''
921        decl_type = 'function'
922
923        #
924        # If we have a macro, remove the "#define" at the front.
925        #
926        new_proto = KernRe(r"^#\s*define\s+").sub("", prototype)
927        if new_proto != prototype:
928            prototype = new_proto
929            #
930            # Dispense with the simple "#define A B" case here; the key
931            # is the space after the name of the symbol being defined.
932            # NOTE that the seemingly misnamed "func_macro" indicates a
933            # macro *without* arguments.
934            #
935            r = KernRe(r'^(\w+)\s+')
936            if r.search(prototype):
937                return_type = ''
938                declaration_name = r.group(1)
939                func_macro = True
940                found = True
941        else:
942            #
943            # Apply the initial transformations.
944            #
945            prototype = self.xforms.apply("func", prototype)
946
947        # Yes, this truly is vile.  We are looking for:
948        # 1. Return type (may be nothing if we're looking at a macro)
949        # 2. Function name
950        # 3. Function parameters.
951        #
952        # All the while we have to watch out for function pointer parameters
953        # (which IIRC is what the two sections are for), C types (these
954        # regexps don't even start to express all the possibilities), and
955        # so on.
956        #
957        # If you mess with these regexps, it's a good idea to check that
958        # the following functions' documentation still comes out right:
959        # - parport_register_device (function pointer parameters)
960        # - atomic_set (macro)
961        # - pci_match_device, __copy_to_user (long return type)
962
963        name = r'\w+'
964        type1 = r'(?:[\w\s]+)?'
965        type2 = r'(?:[\w\s]+\*+)+'
966        #
967        # Attempt to match first on (args) with no internal parentheses; this
968        # lets us easily filter out __acquires() and other post-args stuff.  If
969        # that fails, just grab the rest of the line to the last closing
970        # parenthesis.
971        #
972        proto_args = r'\(([^\(]*|.*)\)'
973        #
974        # (Except for the simple macro case) attempt to split up the prototype
975        # in the various ways we understand.
976        #
977        if not found:
978            patterns = [
979                rf'^()({name})\s*{proto_args}',
980                rf'^({type1})\s+({name})\s*{proto_args}',
981                rf'^({type2})\s*({name})\s*{proto_args}',
982            ]
983
984            for p in patterns:
985                r = KernRe(p)
986                if r.match(prototype):
987                    return_type = r.group(1)
988                    declaration_name = r.group(2)
989                    args = r.group(3)
990                    self.create_parameter_list(ln, decl_type, args, ',',
991                                               declaration_name)
992                    found = True
993                    break
994        #
995        # Parsing done; make sure that things are as we expect.
996        #
997        if not found:
998            self.emit_msg(ln,
999                          f"cannot understand function prototype: '{prototype}'")
1000            return
1001        if self.entry.identifier != declaration_name:
1002            self.emit_msg(ln, f"expecting prototype for {self.entry.identifier}(). "
1003                          f"Prototype was for {declaration_name}() instead")
1004            return
1005        self.check_sections(ln, declaration_name, "function")
1006        self.check_return_section(ln, declaration_name, return_type)
1007        #
1008        # Store the result.
1009        #
1010        self.output_declaration(decl_type, declaration_name,
1011                                typedef=('typedef' in return_type),
1012                                functiontype=return_type,
1013                                purpose=self.entry.declaration_purpose,
1014                                func_macro=func_macro)
1015
1016
1017    def dump_typedef(self, ln, proto):
1018        """
1019        Store a ``typedef`` inside self.entries array.
1020        """
1021        #
1022        # We start by looking for function typedefs.
1023        #
1024        typedef_type = r'typedef((?:\s+[\w*]+\b){0,7}\s+(?:\w+\b|\*+))\s*'
1025        typedef_ident = r'\*?\s*(\w\S+)\s*'
1026        typedef_args = r'\s*\((.*)\);'
1027
1028        typedef1 = KernRe(typedef_type + r'\(' + typedef_ident + r'\)' + typedef_args)
1029        typedef2 = KernRe(typedef_type + typedef_ident + typedef_args)
1030
1031        # Parse function typedef prototypes
1032        for r in [typedef1, typedef2]:
1033            if not r.match(proto):
1034                continue
1035
1036            return_type = r.group(1).strip()
1037            declaration_name = r.group(2)
1038            args = r.group(3)
1039
1040            if self.entry.identifier != declaration_name:
1041                self.emit_msg(ln,
1042                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1043                return
1044
1045            self.create_parameter_list(ln, 'function', args, ',', declaration_name)
1046
1047            self.output_declaration('function', declaration_name,
1048                                    typedef=True,
1049                                    functiontype=return_type,
1050                                    purpose=self.entry.declaration_purpose)
1051            return
1052        #
1053        # Not a function, try to parse a simple typedef.
1054        #
1055        r = KernRe(r'typedef.*\s+(\w+)\s*;')
1056        if r.match(proto):
1057            declaration_name = r.group(1)
1058
1059            if self.entry.identifier != declaration_name:
1060                self.emit_msg(ln,
1061                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1062                return
1063
1064            self.output_declaration('typedef', declaration_name,
1065                                    purpose=self.entry.declaration_purpose)
1066            return
1067
1068        self.emit_msg(ln, "error: Cannot parse typedef!")
1069
1070    @staticmethod
1071    def process_export(function_set, line):
1072        """
1073        process ``EXPORT_SYMBOL*`` tags
1074
1075        This method doesn't use any variable from the class, so declare it
1076        with a staticmethod decorator.
1077        """
1078
1079        # We support documenting some exported symbols with different
1080        # names.  A horrible hack.
1081        suffixes = [ '_noprof' ]
1082
1083        # Note: it accepts only one EXPORT_SYMBOL* per line, as having
1084        # multiple export lines would violate Kernel coding style.
1085
1086        if export_symbol.search(line):
1087            symbol = export_symbol.group(2)
1088        elif export_symbol_ns.search(line):
1089            symbol = export_symbol_ns.group(2)
1090        else:
1091            return False
1092        #
1093        # Found an export, trim out any special suffixes
1094        #
1095        for suffix in suffixes:
1096            # Be backward compatible with Python < 3.9
1097            if symbol.endswith(suffix):
1098                symbol = symbol[:-len(suffix)]
1099        function_set.add(symbol)
1100        return True
1101
1102    def process_normal(self, ln, line):
1103        """
1104        STATE_NORMAL: looking for the ``/**`` to begin everything.
1105        """
1106
1107        if not doc_start.match(line):
1108            return
1109
1110        # start a new entry
1111        self.reset_state(ln)
1112
1113        # next line is always the function name
1114        self.state = state.NAME
1115
1116    def process_name(self, ln, line):
1117        """
1118        STATE_NAME: Looking for the "name - description" line
1119        """
1120        #
1121        # Check for a DOC: block and handle them specially.
1122        #
1123        if doc_block.search(line):
1124
1125            if not doc_block.group(1):
1126                self.entry.begin_section(ln, "Introduction")
1127            else:
1128                self.entry.begin_section(ln, doc_block.group(1))
1129
1130            self.entry.identifier = self.entry.section
1131            self.state = state.DOCBLOCK
1132        #
1133        # Otherwise we're looking for a normal kerneldoc declaration line.
1134        #
1135        elif doc_decl.search(line):
1136            self.entry.identifier = doc_decl.group(1)
1137
1138            # Test for data declaration
1139            if doc_begin_data.search(line):
1140                self.entry.decl_type = doc_begin_data.group(1)
1141                self.entry.identifier = doc_begin_data.group(2)
1142            #
1143            # Look for a function description
1144            #
1145            elif doc_begin_func.search(line):
1146                self.entry.identifier = doc_begin_func.group(1)
1147                self.entry.decl_type = "function"
1148            #
1149            # We struck out.
1150            #
1151            else:
1152                self.emit_msg(ln,
1153                              f"This comment starts with '/**', but isn't a kernel-doc comment. Refer to Documentation/doc-guide/kernel-doc.rst\n{line}")
1154                self.state = state.NORMAL
1155                return
1156            #
1157            # OK, set up for a new kerneldoc entry.
1158            #
1159            self.state = state.BODY
1160            self.entry.identifier = self.entry.identifier.strip(" ")
1161            # if there's no @param blocks need to set up default section here
1162            self.entry.begin_section(ln + 1)
1163            #
1164            # Find the description portion, which *should* be there but
1165            # isn't always.
1166            # (We should be able to capture this from the previous parsing - someday)
1167            #
1168            r = KernRe("[-:](.*)")
1169            if r.search(line):
1170                self.entry.declaration_purpose = trim_whitespace(r.group(1))
1171                self.state = state.DECLARATION
1172            else:
1173                self.entry.declaration_purpose = ""
1174
1175            if not self.entry.declaration_purpose and self.config.wshort_desc:
1176                self.emit_msg(ln,
1177                              f"missing initial short description on line:\n{line}")
1178
1179            if not self.entry.identifier and self.entry.decl_type != "enum":
1180                self.emit_msg(ln,
1181                              f"wrong kernel-doc identifier on line:\n{line}")
1182                self.state = state.NORMAL
1183
1184            if self.config.verbose:
1185                self.emit_msg(ln,
1186                              f"Scanning doc for {self.entry.decl_type} {self.entry.identifier}",
1187                                  warning=False)
1188        #
1189        # Failed to find an identifier. Emit a warning
1190        #
1191        else:
1192            self.emit_msg(ln, f"Cannot find identifier on line:\n{line}")
1193
1194    def is_new_section(self, ln, line):
1195        """
1196        Helper function to determine if a new section is being started.
1197        """
1198        if doc_sect.search(line):
1199            self.state = state.BODY
1200            #
1201            # Pick out the name of our new section, tweaking it if need be.
1202            #
1203            newsection = doc_sect.group(1)
1204            if newsection.lower() == 'description':
1205                newsection = 'Description'
1206            elif newsection.lower() == 'context':
1207                newsection = 'Context'
1208                self.state = state.SPECIAL_SECTION
1209            elif newsection.lower() in ["@return", "@returns",
1210                                        "return", "returns"]:
1211                newsection = "Return"
1212                self.state = state.SPECIAL_SECTION
1213            elif newsection[0] == '@':
1214                self.state = state.SPECIAL_SECTION
1215            #
1216            # Initialize the contents, and get the new section going.
1217            #
1218            newcontents = doc_sect.group(2)
1219            if not newcontents:
1220                newcontents = ""
1221            self.dump_section()
1222            self.entry.begin_section(ln, newsection)
1223            self.entry.leading_space = None
1224
1225            self.entry.add_text(newcontents.lstrip())
1226            return True
1227        return False
1228
1229    def is_comment_end(self, ln, line):
1230        """
1231        Helper function to detect (and effect) the end of a kerneldoc comment.
1232        """
1233        if doc_end.search(line):
1234            self.dump_section()
1235
1236            # Look for doc_com + <text> + doc_end:
1237            r = KernRe(r'\s*\*\s*[a-zA-Z_0-9:.]+\*/')
1238            if r.match(line):
1239                self.emit_msg(ln, f"suspicious ending line: {line}")
1240
1241            self.entry.prototype = ""
1242            self.entry.new_start_line = ln + 1
1243
1244            self.state = state.PROTO
1245            return True
1246        return False
1247
1248
1249    def process_decl(self, ln, line):
1250        """
1251        STATE_DECLARATION: We've seen the beginning of a declaration.
1252        """
1253        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1254            return
1255        #
1256        # Look for anything with the " * " line beginning.
1257        #
1258        if doc_content.search(line):
1259            cont = doc_content.group(1)
1260            #
1261            # A blank line means that we have moved out of the declaration
1262            # part of the comment (without any "special section" parameter
1263            # descriptions).
1264            #
1265            if cont == "":
1266                self.state = state.BODY
1267            #
1268            # Otherwise we have more of the declaration section to soak up.
1269            #
1270            else:
1271                self.entry.declaration_purpose = \
1272                    trim_whitespace(self.entry.declaration_purpose + ' ' + cont)
1273        else:
1274            # Unknown line, ignore
1275            self.emit_msg(ln, f"bad line: {line}")
1276
1277
1278    def process_special(self, ln, line):
1279        """
1280        STATE_SPECIAL_SECTION: a section ending with a blank line.
1281        """
1282        #
1283        # If we have hit a blank line (only the " * " marker), then this
1284        # section is done.
1285        #
1286        if KernRe(r"\s*\*\s*$").match(line):
1287            self.entry.begin_section(ln, dump = True)
1288            self.state = state.BODY
1289            return
1290        #
1291        # Not a blank line, look for the other ways to end the section.
1292        #
1293        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1294            return
1295        #
1296        # OK, we should have a continuation of the text for this section.
1297        #
1298        if doc_content.search(line):
1299            cont = doc_content.group(1)
1300            #
1301            # If the lines of text after the first in a special section have
1302            # leading white space, we need to trim it out or Sphinx will get
1303            # confused.  For the second line (the None case), see what we
1304            # find there and remember it.
1305            #
1306            if self.entry.leading_space is None:
1307                r = KernRe(r'^(\s+)')
1308                if r.match(cont):
1309                    self.entry.leading_space = len(r.group(1))
1310                else:
1311                    self.entry.leading_space = 0
1312            #
1313            # Otherwise, before trimming any leading chars, be *sure*
1314            # that they are white space.  We should maybe warn if this
1315            # isn't the case.
1316            #
1317            for i in range(0, self.entry.leading_space):
1318                if cont[i] != " ":
1319                    self.entry.leading_space = i
1320                    break
1321            #
1322            # Add the trimmed result to the section and we're done.
1323            #
1324            self.entry.add_text(cont[self.entry.leading_space:])
1325        else:
1326            # Unknown line, ignore
1327            self.emit_msg(ln, f"bad line: {line}")
1328
1329    def process_body(self, ln, line):
1330        """
1331        STATE_BODY: the bulk of a kerneldoc comment.
1332        """
1333        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1334            return
1335
1336        if doc_content.search(line):
1337            cont = doc_content.group(1)
1338            self.entry.add_text(cont)
1339        else:
1340            # Unknown line, ignore
1341            self.emit_msg(ln, f"bad line: {line}")
1342
1343    def process_inline_name(self, ln, line):
1344        """STATE_INLINE_NAME: beginning of docbook comments within a prototype."""
1345
1346        if doc_inline_sect.search(line):
1347            self.entry.begin_section(ln, doc_inline_sect.group(1))
1348            self.entry.add_text(doc_inline_sect.group(2).lstrip())
1349            self.state = state.INLINE_TEXT
1350        elif doc_inline_end.search(line):
1351            self.dump_section()
1352            self.state = state.PROTO
1353        elif doc_content.search(line):
1354            self.emit_msg(ln, f"Incorrect use of kernel-doc format: {line}")
1355            self.state = state.PROTO
1356
1357            #
1358            # Don't let it add partial comments at the code, as breaks the
1359            # logic meant to remove comments from prototypes.
1360            #
1361            self.process_proto_type(ln, "/**\n" + line)
1362        # else ... ??
1363
1364    def process_inline_text(self, ln, line):
1365        """STATE_INLINE_TEXT: docbook comments within a prototype."""
1366
1367        if doc_inline_end.search(line):
1368            self.dump_section()
1369            self.state = state.PROTO
1370        elif doc_content.search(line):
1371            self.entry.add_text(doc_content.group(1))
1372        # else ... ??
1373
1374    def syscall_munge(self, ln, proto):         # pylint: disable=W0613
1375        """
1376        Handle syscall definitions.
1377        """
1378
1379        is_void = False
1380
1381        # Strip newlines/CR's
1382        proto = re.sub(r'[\r\n]+', ' ', proto)
1383
1384        # Check if it's a SYSCALL_DEFINE0
1385        if 'SYSCALL_DEFINE0' in proto:
1386            is_void = True
1387
1388        # Replace SYSCALL_DEFINE with correct return type & function name
1389        proto = KernRe(r'SYSCALL_DEFINE.*\(').sub('long sys_', proto)
1390
1391        r = KernRe(r'long\s+(sys_.*?),')
1392        if r.search(proto):
1393            proto = KernRe(',').sub('(', proto, count=1)
1394        elif is_void:
1395            proto = KernRe(r'\)').sub('(void)', proto, count=1)
1396
1397        # Now delete all of the odd-numbered commas in the proto
1398        # so that argument types & names don't have a comma between them
1399        count = 0
1400        length = len(proto)
1401
1402        if is_void:
1403            length = 0  # skip the loop if is_void
1404
1405        for ix in range(length):
1406            if proto[ix] == ',':
1407                count += 1
1408                if count % 2 == 1:
1409                    proto = proto[:ix] + ' ' + proto[ix + 1:]
1410
1411        return proto
1412
1413    def tracepoint_munge(self, ln, proto):
1414        """
1415        Handle tracepoint definitions.
1416        """
1417
1418        tracepointname = None
1419        tracepointargs = None
1420
1421        # Match tracepoint name based on different patterns
1422        r = KernRe(r'TRACE_EVENT\((.*?),')
1423        if r.search(proto):
1424            tracepointname = r.group(1)
1425
1426        r = KernRe(r'DEFINE_SINGLE_EVENT\((.*?),')
1427        if r.search(proto):
1428            tracepointname = r.group(1)
1429
1430        r = KernRe(r'DEFINE_EVENT\((.*?),(.*?),')
1431        if r.search(proto):
1432            tracepointname = r.group(2)
1433
1434        if tracepointname:
1435            tracepointname = tracepointname.lstrip()
1436
1437        r = KernRe(r'TP_PROTO\((.*?)\)')
1438        if r.search(proto):
1439            tracepointargs = r.group(1)
1440
1441        if not tracepointname or not tracepointargs:
1442            self.emit_msg(ln,
1443                          f"Unrecognized tracepoint format:\n{proto}\n")
1444        else:
1445            proto = f"static inline void trace_{tracepointname}({tracepointargs})"
1446            self.entry.identifier = f"trace_{self.entry.identifier}"
1447
1448        return proto
1449
1450    def process_proto_function(self, ln, line):
1451        """Ancillary routine to process a function prototype."""
1452
1453        # strip C99-style comments to end of line
1454        line = KernRe(r"//.*$", re.S).sub('', line)
1455        #
1456        # Soak up the line's worth of prototype text, stopping at { or ; if present.
1457        #
1458        if KernRe(r'\s*#\s*define').match(line):
1459            self.entry.prototype = line
1460        elif not line.startswith('#'):   # skip other preprocessor stuff
1461            r = KernRe(r'([^\{]*)')
1462            if r.match(line):
1463                self.entry.prototype += r.group(1) + " "
1464        #
1465        # If we now have the whole prototype, clean it up and declare victory.
1466        #
1467        if '{' in line or ';' in line or KernRe(r'\s*#\s*define').match(line):
1468            # strip comments and surrounding spaces
1469            self.entry.prototype = KernRe(r'/\*.*\*/').sub('', self.entry.prototype).strip()
1470            #
1471            # Handle self.entry.prototypes for function pointers like:
1472            #       int (*pcs_config)(struct foo)
1473            # by turning it into
1474            #	    int pcs_config(struct foo)
1475            #
1476            r = KernRe(r'^(\S+\s+)\(\s*\*(\S+)\)')
1477            self.entry.prototype = r.sub(r'\1\2', self.entry.prototype)
1478            #
1479            # Handle special declaration syntaxes
1480            #
1481            if 'SYSCALL_DEFINE' in self.entry.prototype:
1482                self.entry.prototype = self.syscall_munge(ln,
1483                                                          self.entry.prototype)
1484            else:
1485                r = KernRe(r'TRACE_EVENT|DEFINE_EVENT|DEFINE_SINGLE_EVENT')
1486                if r.search(self.entry.prototype):
1487                    self.entry.prototype = self.tracepoint_munge(ln,
1488                                                                 self.entry.prototype)
1489            #
1490            # ... and we're done
1491            #
1492            self.dump_function(ln, self.entry.prototype)
1493            self.reset_state(ln)
1494
1495    def process_proto_type(self, ln, line):
1496        """
1497        Ancillary routine to process a type.
1498        """
1499
1500        # Strip C99-style comments and surrounding whitespace
1501        line = KernRe(r"//.*$", re.S).sub('', line).strip()
1502        if not line:
1503            return # nothing to see here
1504
1505        # To distinguish preprocessor directive from regular declaration later.
1506        if line.startswith('#'):
1507            line += ";"
1508        #
1509        # Split the declaration on any of { } or ;, and accumulate pieces
1510        # until we hit a semicolon while not inside {brackets}
1511        #
1512        r = KernRe(r'(.*?)([{};])')
1513        for chunk in r.split(line):
1514            if chunk:  # Ignore empty matches
1515                self.entry.prototype += chunk
1516                #
1517                # This cries out for a match statement ... someday after we can
1518                # drop Python 3.9 ...
1519                #
1520                if chunk == '{':
1521                    self.entry.brcount += 1
1522                elif chunk == '}':
1523                    self.entry.brcount -= 1
1524                elif chunk == ';' and self.entry.brcount <= 0:
1525                    self.dump_declaration(ln, self.entry.prototype)
1526                    self.reset_state(ln)
1527                    return
1528        #
1529        # We hit the end of the line while still in the declaration; put
1530        # in a space to represent the newline.
1531        #
1532        self.entry.prototype += ' '
1533
1534    def process_proto(self, ln, line):
1535        """STATE_PROTO: reading a function/whatever prototype."""
1536
1537        if doc_inline_oneline.search(line):
1538            self.entry.begin_section(ln, doc_inline_oneline.group(1))
1539            self.entry.add_text(doc_inline_oneline.group(2))
1540            self.dump_section()
1541
1542        elif doc_inline_start.search(line):
1543            self.state = state.INLINE_NAME
1544
1545        elif self.entry.decl_type == 'function':
1546            self.process_proto_function(ln, line)
1547
1548        else:
1549            self.process_proto_type(ln, line)
1550
1551    def process_docblock(self, ln, line):
1552        """STATE_DOCBLOCK: within a ``DOC:`` block."""
1553
1554        if doc_end.search(line):
1555            self.dump_section()
1556            self.output_declaration("doc", self.entry.identifier)
1557            self.reset_state(ln)
1558
1559        elif doc_content.search(line):
1560            self.entry.add_text(doc_content.group(1))
1561
1562    def parse_export(self):
1563        """
1564        Parses ``EXPORT_SYMBOL*`` macros from a single Kernel source file.
1565        """
1566
1567        export_table = set()
1568
1569        try:
1570            with open(self.fname, "r", encoding="utf8",
1571                      errors="backslashreplace") as fp:
1572
1573                for line in fp:
1574                    self.process_export(export_table, line)
1575
1576        except IOError:
1577            return None
1578
1579        return export_table
1580
1581    #: The state/action table telling us which function to invoke in each state.
1582    state_actions = {
1583        state.NORMAL:			process_normal,
1584        state.NAME:			process_name,
1585        state.BODY:			process_body,
1586        state.DECLARATION:		process_decl,
1587        state.SPECIAL_SECTION:		process_special,
1588        state.INLINE_NAME:		process_inline_name,
1589        state.INLINE_TEXT:		process_inline_text,
1590        state.PROTO:			process_proto,
1591        state.DOCBLOCK:			process_docblock,
1592        }
1593
1594    def parse_kdoc(self):
1595        """
1596        Open and process each line of a C source file.
1597        The parsing is controlled via a state machine, and the line is passed
1598        to a different process function depending on the state. The process
1599        function may update the state as needed.
1600
1601        Besides parsing kernel-doc tags, it also parses export symbols.
1602        """
1603
1604        prev = ""
1605        prev_ln = None
1606        export_table = set()
1607
1608        try:
1609            with open(self.fname, "r", encoding="utf8",
1610                      errors="backslashreplace") as fp:
1611                for ln, line in enumerate(fp):
1612
1613                    line = line.expandtabs().strip("\n")
1614
1615                    # Group continuation lines on prototypes
1616                    if self.state == state.PROTO:
1617                        if line.endswith("\\"):
1618                            prev += line.rstrip("\\")
1619                            if not prev_ln:
1620                                prev_ln = ln
1621                            continue
1622
1623                        if prev:
1624                            ln = prev_ln
1625                            line = prev + line
1626                            prev = ""
1627                            prev_ln = None
1628
1629                    self.config.log.debug("%d %s: %s",
1630                                          ln, state.name[self.state],
1631                                          line)
1632
1633                    # This is an optimization over the original script.
1634                    # There, when export_file was used for the same file,
1635                    # it was read twice. Here, we use the already-existing
1636                    # loop to parse exported symbols as well.
1637                    #
1638                    if (self.state != state.NORMAL) or \
1639                       not self.process_export(export_table, line):
1640                        # Hand this line to the appropriate state handler
1641                        self.state_actions[self.state](self, ln, line)
1642
1643            self.emit_unused_warnings()
1644
1645        except OSError:
1646            self.config.log.error(f"Error: Cannot open file {self.fname}")
1647
1648        return export_table, self.entries
1649