xref: /linux/tools/lib/python/kdoc/kdoc_parser.py (revision 5181afcdf99527dd92a88f80fc4d0d8013e1b510) !
1#!/usr/bin/env python3
2# SPDX-License-Identifier: GPL-2.0
3# Copyright(c) 2025: Mauro Carvalho Chehab <mchehab@kernel.org>.
4#
5# pylint: disable=C0301,C0302,R0904,R0912,R0913,R0914,R0915,R0917,R1702
6
7"""
8Classes and functions related to reading a C language source or header FILE
9and extract embedded documentation comments from it.
10"""
11
12import sys
13import re
14from pprint import pformat
15
16from kdoc.c_lex import CTokenizer, tokenizer_set_log
17from kdoc.kdoc_re import KernRe
18from kdoc.kdoc_item import KdocItem
19
20#
21# Regular expressions used to parse kernel-doc markups at KernelDoc class.
22#
23# Let's declare them in lowercase outside any class to make it easier to
24# convert from the Perl script.
25#
26# As those are evaluated at the beginning, no need to cache them
27#
28
29# Allow whitespace at end of comment start.
30doc_start = KernRe(r'^/\*\*\s*$', cache=False)
31
32doc_end = KernRe(r'\*/', cache=False)
33doc_com = KernRe(r'\s*\*\s*', cache=False)
34doc_com_body = KernRe(r'\s*\* ?', cache=False)
35doc_decl = doc_com + KernRe(r'(\w+)', cache=False)
36
37# @params and a strictly limited set of supported section names
38# Specifically:
39#   Match @word:
40#         @...:
41#         @{section-name}:
42# while trying to not match literal block starts like "example::"
43#
44known_section_names = 'description|context|returns?|notes?|examples?'
45known_sections = KernRe(known_section_names, flags = re.I)
46doc_sect = doc_com + \
47    KernRe(r'\s*(@[.\w]+|@\.\.\.|' + known_section_names + r')\s*:([^:].*)?$',
48           flags=re.I, cache=False)
49
50doc_content = doc_com_body + KernRe(r'(.*)', cache=False)
51doc_inline_start = KernRe(r'^\s*/\*\*\s*$', cache=False)
52doc_inline_sect = KernRe(r'\s*\*\s*(@\s*[\w][\w\.]*\s*):(.*)', cache=False)
53doc_inline_end = KernRe(r'^\s*\*/\s*$', cache=False)
54doc_inline_oneline = KernRe(r'^\s*/\*\*\s*(@\s*[\w][\w\.]*\s*):\s*(.*)\s*\*/\s*$', cache=False)
55
56export_symbol = KernRe(r'^\s*EXPORT_SYMBOL(_GPL)?\s*\(\s*(\w+)\s*\)\s*', cache=False)
57export_symbol_ns = KernRe(r'^\s*EXPORT_SYMBOL_NS(_GPL)?\s*\(\s*(\w+)\s*,\s*"\S+"\)\s*', cache=False)
58
59type_param = KernRe(r"@(\w*((\.\w+)|(->\w+))*(\.\.\.)?)", cache=False)
60
61#
62# Tests for the beginning of a kerneldoc block in its various forms.
63#
64doc_block = doc_com + KernRe(r'DOC:\s*(.*)?', cache=False)
65doc_begin_data = KernRe(r"^\s*\*?\s*(struct|union|enum|typedef|var)\b\s*(\w*)", cache = False)
66doc_begin_func = KernRe(str(doc_com) +			# initial " * '
67                        r"(?:\w+\s*\*\s*)?" + 		# type (not captured)
68                        r'(?:define\s+)?' + 		# possible "define" (not captured)
69                        r'(\w+)\s*(?:\(\w*\))?\s*' +	# name and optional "(...)"
70                        r'(?:[-:].*)?$',		# description (not captured)
71                        cache = False)
72
73#
74# Ancillary functions
75#
76
77multi_space = KernRe(r'\s\s+')
78def trim_whitespace(s):
79    """
80    A little helper to get rid of excess white space.
81    """
82    return multi_space.sub(' ', s.strip())
83
84def trim_private_members(text):
85    """
86    Remove ``struct``/``enum`` members that have been marked "private".
87    """
88
89    tokens = CTokenizer(text)
90    return str(tokens)
91
92class state:
93    """
94    States used by the parser's state machine.
95    """
96
97    # Parser states
98    NORMAL        = 0        #: Normal code.
99    NAME          = 1        #: Looking for function name.
100    DECLARATION   = 2        #: We have seen a declaration which might not be done.
101    BODY          = 3        #: The body of the comment.
102    SPECIAL_SECTION = 4      #: Doc section ending with a blank line.
103    PROTO         = 5        #: Scanning prototype.
104    DOCBLOCK      = 6        #: Documentation block.
105    INLINE_NAME   = 7        #: Gathering doc outside main block.
106    INLINE_TEXT   = 8	     #: Reading the body of inline docs.
107
108    #: Names for each parser state.
109    name = [
110        "NORMAL",
111        "NAME",
112        "DECLARATION",
113        "BODY",
114        "SPECIAL_SECTION",
115        "PROTO",
116        "DOCBLOCK",
117        "INLINE_NAME",
118        "INLINE_TEXT",
119    ]
120
121
122SECTION_DEFAULT = "Description"  #: Default section.
123
124class KernelEntry:
125    """
126    Encapsulates a Kernel documentation entry.
127    """
128
129    def __init__(self, config, fname, ln):
130        self.config = config
131        self.fname = fname
132
133        self._contents = []
134        self.prototype = ""
135
136        self.warnings = []
137
138        self.parameterlist = []
139        self.parameterdescs = {}
140        self.parametertypes = {}
141        self.parameterdesc_start_lines = {}
142
143        self.sections_start_lines = {}
144        self.sections = {}
145
146        self.anon_struct_union = False
147
148        self.leading_space = None
149
150        self.fname = fname
151
152        # State flags
153        self.brcount = 0
154        self.declaration_start_line = ln + 1
155
156    #
157    # Management of section contents
158    #
159    def add_text(self, text):
160        """Add a new text to the entry contents list."""
161        self._contents.append(text)
162
163    def contents(self):
164        """Returns a string with all content texts that were added."""
165        return '\n'.join(self._contents) + '\n'
166
167    # TODO: rename to emit_message after removal of kernel-doc.pl
168    def emit_msg(self, ln, msg, *, warning=True):
169        """Emit a message."""
170
171        log_msg = f"{self.fname}:{ln} {msg}"
172
173        if not warning:
174            self.config.log.info(log_msg)
175            return
176
177        # Delegate warning output to output logic, as this way it
178        # will report warnings/info only for symbols that are output
179
180        self.warnings.append(log_msg)
181        return
182
183    def begin_section(self, line_no, title = SECTION_DEFAULT, dump = False):
184        """
185        Begin a new section.
186        """
187        if dump:
188            self.dump_section(start_new = True)
189        self.section = title
190        self.new_start_line = line_no
191
192    def dump_section(self, start_new=True):
193        """
194        Dumps section contents to arrays/hashes intended for that purpose.
195        """
196        #
197        # If we have accumulated no contents in the default ("description")
198        # section, don't bother.
199        #
200        if self.section == SECTION_DEFAULT and not self._contents:
201            return
202        name = self.section
203        contents = self.contents()
204
205        if type_param.match(name):
206            name = type_param.group(1)
207
208            self.parameterdescs[name] = contents
209            self.parameterdesc_start_lines[name] = self.new_start_line
210
211            self.new_start_line = 0
212
213        else:
214            if name in self.sections and self.sections[name] != "":
215                # Only warn on user-specified duplicate section names
216                if name != SECTION_DEFAULT:
217                    self.emit_msg(self.new_start_line,
218                                  f"duplicate section name '{name}'")
219                # Treat as a new paragraph - add a blank line
220                self.sections[name] += '\n' + contents
221            else:
222                self.sections[name] = contents
223                self.sections_start_lines[name] = self.new_start_line
224                self.new_start_line = 0
225
226#        self.config.log.debug("Section: %s : %s", name, pformat(vars(self)))
227
228        if start_new:
229            self.section = SECTION_DEFAULT
230            self._contents = []
231
232python_warning = False
233
234class KernelDoc:
235    """
236    Read a C language source or header FILE and extract embedded
237    documentation comments.
238    """
239
240    #: Name of context section.
241    section_context = "Context"
242
243    #: Name of return section.
244    section_return = "Return"
245
246    #: String to write when a parameter is not described.
247    undescribed = "-- undescribed --"
248
249    def __init__(self, config, fname, xforms, store_src=False):
250        """Initialize internal variables"""
251
252        self.fname = fname
253        self.config = config
254        self.xforms = xforms
255        self.store_src = store_src
256
257        tokenizer_set_log(self.config.log, f"{self.fname}: CMatch: ")
258
259        # Initial state for the state machines
260        self.state = state.NORMAL
261
262        # Store entry currently being processed
263        self.entry = None
264
265        # Place all potential outputs into an array
266        self.entries = []
267
268        #
269        # We need Python 3.7 for its "dicts remember the insertion
270        # order" guarantee
271        #
272        global python_warning
273        if (not python_warning and
274            sys.version_info.major == 3 and sys.version_info.minor < 7):
275
276            self.emit_msg(0,
277                          'Python 3.7 or later is required for correct results')
278            python_warning = True
279
280    def emit_msg(self, ln, msg, *, warning=True):
281        """Emit a message"""
282
283        if self.entry:
284            self.entry.emit_msg(ln, msg, warning=warning)
285            return
286
287        log_msg = f"{self.fname}:{ln} {msg}"
288
289        if warning:
290            self.config.log.warning(log_msg)
291        else:
292            self.config.log.info(log_msg)
293
294    def dump_section(self, start_new=True):
295        """
296        Dump section contents to arrays/hashes intended for that purpose.
297        """
298
299        if self.entry:
300            self.entry.dump_section(start_new)
301
302    # TODO: rename it to store_declaration after removal of kernel-doc.pl
303    def output_declaration(self, dtype, name, **args):
304        """
305        Store the entry into an entry array.
306
307        The actual output and output filters will be handled elsewhere.
308        """
309
310        item = KdocItem(name, self.fname, dtype,
311                        self.entry.declaration_start_line, **args)
312        item.warnings = self.entry.warnings
313
314        # Drop empty sections
315        # TODO: improve empty sections logic to emit warnings
316        sections = self.entry.sections
317        for section in ["Description", "Return"]:
318            if section in sections and not sections[section].rstrip():
319                del sections[section]
320        item.set_sections(sections, self.entry.sections_start_lines)
321        item.set_params(self.entry.parameterlist, self.entry.parameterdescs,
322                        self.entry.parametertypes,
323                        self.entry.parameterdesc_start_lines)
324        self.entries.append(item)
325
326        self.config.log.debug("Output: %s:%s = %s", dtype, name, pformat(args))
327
328    def emit_unused_warnings(self):
329        """
330        When the parser fails to produce a valid entry, it places some
331        warnings under `entry.warnings` that will be discarded when resetting
332        the state.
333
334        Ensure that those warnings are not lost.
335
336        .. note::
337
338              Because we are calling `config.warning()` here, those
339              warnings are not filtered by the `-W` parameters: they will all
340              be produced even when `-Wreturn`, `-Wshort-desc`, and/or
341              `-Wcontents-before-sections` are used.
342
343              Allowing those warnings to be filtered is complex, because it
344              would require storing them in a buffer and then filtering them
345              during the output step of the code, depending on the
346              selected symbols.
347        """
348        if self.entry and self.entry not in self.entries:
349            for log_msg in self.entry.warnings:
350                self.config.warning(log_msg)
351
352    def reset_state(self, ln):
353        """
354        Ancillary routine to create a new entry. It initializes all
355        variables used by the state machine.
356        """
357
358        self.emit_unused_warnings()
359
360        self.entry = KernelEntry(self.config, self.fname, ln)
361
362        # State flags
363        self.state = state.NORMAL
364
365    def push_parameter(self, ln, decl_type, param, dtype,
366                       org_arg, declaration_name):
367        """
368        Store parameters and their descriptions at self.entry.
369        """
370
371        if self.entry.anon_struct_union and dtype == "" and param == "}":
372            return  # Ignore the ending }; from anonymous struct/union
373
374        self.entry.anon_struct_union = False
375
376        param = KernRe(r'[\[\)].*').sub('', param, count=1)
377
378        #
379        # Look at various "anonymous type" cases.
380        #
381        if dtype == '':
382            if param.endswith("..."):
383                if len(param) > 3: # there is a name provided, use that
384                    param = param[:-3]
385                if not self.entry.parameterdescs.get(param):
386                    self.entry.parameterdescs[param] = "variable arguments"
387
388            elif (not param) or param == "void":
389                param = "void"
390                self.entry.parameterdescs[param] = "no arguments"
391
392            elif param in ["struct", "union"]:
393                # Handle unnamed (anonymous) union or struct
394                dtype = param
395                param = "{unnamed_" + param + "}"
396                self.entry.parameterdescs[param] = "anonymous\n"
397                self.entry.anon_struct_union = True
398
399        # Warn if parameter has no description
400        # (but ignore ones starting with # as these are not parameters
401        # but inline preprocessor statements)
402        if param not in self.entry.parameterdescs and not param.startswith("#"):
403            self.entry.parameterdescs[param] = self.undescribed
404
405            if "." not in param:
406                if decl_type == 'function':
407                    dname = f"{decl_type} parameter"
408                else:
409                    dname = f"{decl_type} member"
410
411                self.emit_msg(ln,
412                              f"{dname} '{param}' not described in '{declaration_name}'")
413
414        # Strip spaces from param so that it is one continuous string on
415        # parameterlist. This fixes a problem where check_sections()
416        # cannot find a parameter like "addr[6 + 2]" because it actually
417        # appears as "addr[6", "+", "2]" on the parameter list.
418        # However, it's better to maintain the param string unchanged for
419        # output, so just weaken the string compare in check_sections()
420        # to ignore "[blah" in a parameter string.
421
422        self.entry.parameterlist.append(param)
423        org_arg = KernRe(r'\s\s+').sub(' ', org_arg)
424        self.entry.parametertypes[param] = org_arg
425
426
427    def create_parameter_list(self, ln, decl_type, args,
428                              splitter, declaration_name):
429        """
430        Creates a list of parameters, storing them at self.entry.
431        """
432
433        # temporarily replace all commas inside function pointer definition
434        arg_expr = KernRe(r'(\([^\),]+),')
435        while arg_expr.search(args):
436            args = arg_expr.sub(r"\1#", args)
437
438        for arg in args.split(splitter):
439            # Ignore argument attributes
440            arg = KernRe(r'\sPOS0?\s').sub(' ', arg)
441
442            # Strip leading/trailing spaces
443            arg = arg.strip()
444            arg = KernRe(r'\s+').sub(' ', arg, count=1)
445
446            if arg.startswith('#'):
447                # Treat preprocessor directive as a typeless variable just to fill
448                # corresponding data structures "correctly". Catch it later in
449                # output_* subs.
450
451                # Treat preprocessor directive as a typeless variable
452                self.push_parameter(ln, decl_type, arg, "",
453                                    "", declaration_name)
454            #
455            # The pointer-to-function case.
456            #
457            elif KernRe(r'\(.+\)\s*\(').search(arg):
458                arg = arg.replace('#', ',')
459                r = KernRe(r'[^\(]+\(\*?\s*'  # Everything up to "(*"
460                           r'([\w\[\].]*)'    # Capture the name and possible [array]
461                           r'\s*\)')	      # Make sure the trailing ")" is there
462                if r.match(arg):
463                    param = r.group(1)
464                else:
465                    self.emit_msg(ln, f"Invalid param: {arg}")
466                    param = arg
467                dtype = arg.replace(param, '')
468                self.push_parameter(ln, decl_type, param, dtype, arg, declaration_name)
469            #
470            # The array-of-pointers case.  Dig the parameter name out from the middle
471            # of the declaration.
472            #
473            elif KernRe(r'\(.+\)\s*\[').search(arg):
474                r = KernRe(r'[^\(]+\(\s*\*\s*'		# Up to "(" and maybe "*"
475                           r'([\w.]*?)'			# The actual pointer name
476                           r'\s*(\[\s*\w+\s*\]\s*)*\)') # The [array portion]
477                if r.match(arg):
478                    param = r.group(1)
479                else:
480                    self.emit_msg(ln, f"Invalid param: {arg}")
481                    param = arg
482                dtype = arg.replace(param, '')
483                self.push_parameter(ln, decl_type, param, dtype, arg, declaration_name)
484            elif arg:
485                #
486                # Clean up extraneous spaces and split the string at commas; the first
487                # element of the resulting list will also include the type information.
488                #
489                arg = KernRe(r'\s*:\s*').sub(":", arg)
490                arg = KernRe(r'\s*\[').sub('[', arg)
491                args = KernRe(r'\s*,\s*').split(arg)
492                args[0] = re.sub(r'(\*+)\s*', r' \1', args[0])
493                #
494                # args[0] has a string of "type a".  If "a" includes an [array]
495                # declaration, we want to not be fooled by any white space inside
496                # the brackets, so detect and handle that case specially.
497                #
498                r = KernRe(r'^([^[\]]*\s+)(.*)$')
499                if r.match(args[0]):
500                    args[0] = r.group(2)
501                    dtype = r.group(1)
502                else:
503                    # No space in args[0]; this seems wrong but preserves previous behavior
504                    dtype = ''
505
506                bitfield_re = KernRe(r'(.*?):(\w+)')
507                for param in args:
508                    #
509                    # For pointers, shift the star(s) from the variable name to the
510                    # type declaration.
511                    #
512                    r = KernRe(r'^(\*+)\s*(.*)')
513                    if r.match(param):
514                        self.push_parameter(ln, decl_type, r.group(2),
515                                            f"{dtype} {r.group(1)}",
516                                            arg, declaration_name)
517                    #
518                    # Perform a similar shift for bitfields.
519                    #
520                    elif bitfield_re.search(param):
521                        if dtype != "":  # Skip unnamed bit-fields
522                            self.push_parameter(ln, decl_type, bitfield_re.group(1),
523                                                f"{dtype}:{bitfield_re.group(2)}",
524                                                arg, declaration_name)
525                    else:
526                        self.push_parameter(ln, decl_type, param, dtype,
527                                            arg, declaration_name)
528
529    def check_sections(self, ln, decl_name, decl_type):
530        """
531        Check for errors inside sections, emitting warnings if not found
532        parameters are described.
533        """
534        for section in self.entry.sections:
535            if section not in self.entry.parameterlist and \
536               not known_sections.search(section):
537                if decl_type == 'function':
538                    dname = f"{decl_type} parameter"
539                else:
540                    dname = f"{decl_type} member"
541                self.emit_msg(ln,
542                              f"Excess {dname} '{section}' description in '{decl_name}'")
543
544    def check_return_section(self, ln, declaration_name, return_type):
545        """
546        If the function doesn't return void, warns about the lack of a
547        return description.
548        """
549
550        if not self.config.wreturn:
551            return
552
553        # Ignore an empty return type (It's a macro)
554        # Ignore functions with a "void" return type (but not "void *")
555        if not return_type or KernRe(r'void\s*\w*\s*$').search(return_type):
556            return
557
558        if not self.entry.sections.get("Return", None):
559            self.emit_msg(ln,
560                          f"No description found for return value of '{declaration_name}'")
561
562    def split_struct_proto(self, proto):
563        """
564        Split apart a structure prototype; returns (struct|union, name,
565        members) or ``None``.
566        """
567
568        type_pattern = r'(struct|union)'
569        qualifiers = [
570            "__attribute__",
571            "__packed",
572            "__aligned",
573            "____cacheline_aligned_in_smp",
574            "____cacheline_aligned",
575        ]
576        definition_body = r'\{(.*)\}\s*' + "(?:" + '|'.join(qualifiers) + ")?"
577
578        r = KernRe(type_pattern + r'\s+(\w+)\s*' + definition_body)
579        if r.search(proto):
580            return (r.group(1), r.group(2), r.group(3))
581        else:
582            r = KernRe(r'typedef\s+' + type_pattern + r'\s*' + definition_body + r'\s*(\w+)\s*;')
583            if r.search(proto):
584                return (r.group(1), r.group(3), r.group(2))
585        return None
586
587    def rewrite_struct_members(self, members):
588        """
589        Process ``struct``/``union`` members from the most deeply nested
590        outward.
591
592        Rewrite the members of a ``struct`` or ``union`` for easier formatting
593        later on. Among other things, this function will turn a member like::
594
595          struct { inner_members; } foo;
596
597        into::
598
599          struct foo; inner_members;
600        """
601
602        #
603        # The trick is in the ``^{`` below - it prevents a match of an outer
604        # ``struct``/``union`` until the inner one has been munged
605        # (removing the ``{`` in the process).
606        #
607        struct_members = KernRe(r'(struct|union)'   # 0: declaration type
608                                r'([^\{\};]+)' 	    # 1: possible name
609                                r'(\{)'
610                                r'([^\{\}]*)'       # 3: Contents of declaration
611                                r'(\})'
612                                r'([^\{\};]*)(;)')  # 5: Remaining stuff after declaration
613        tuples = struct_members.findall(members)
614        while tuples:
615            for t in tuples:
616                newmember = ""
617                oldmember = "".join(t) # Reconstruct the original formatting
618                dtype, name, lbr, content, rbr, rest, semi = t
619                #
620                # Pass through each field name, normalizing the form and formatting.
621                #
622                for s_id in rest.split(','):
623                    s_id = s_id.strip()
624                    newmember += f"{dtype} {s_id}; "
625                    #
626                    # Remove bitfield/array/pointer info, getting the bare name.
627                    #
628                    s_id = KernRe(r'[:\[].*').sub('', s_id)
629                    s_id = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', s_id)
630                    #
631                    # Pass through the members of this inner structure/union.
632                    #
633                    for arg in content.split(';'):
634                        arg = arg.strip()
635                        #
636                        # Look for (type)(*name)(args) - pointer to function
637                        #
638                        r = KernRe(r'^([^\(]+\(\*?\s*)([\w.]*)(\s*\).*)')
639                        if r.match(arg):
640                            dtype, name, extra = r.group(1), r.group(2), r.group(3)
641                            # Pointer-to-function
642                            if not s_id:
643                                # Anonymous struct/union
644                                newmember += f"{dtype}{name}{extra}; "
645                            else:
646                                newmember += f"{dtype}{s_id}.{name}{extra}; "
647                        #
648                        # Otherwise a non-function member.
649                        #
650                        else:
651                            #
652                            # Remove bitmap and array portions and spaces around commas
653                            #
654                            arg = KernRe(r':\s*\d+\s*').sub('', arg)
655                            arg = KernRe(r'\[.*\]').sub('', arg)
656                            arg = KernRe(r'\s*,\s*').sub(',', arg)
657                            #
658                            # Look for a normal decl - "type name[,name...]"
659                            #
660                            r = KernRe(r'(.*)\s+([\S+,]+)')
661                            if r.search(arg):
662                                for name in r.group(2).split(','):
663                                    name = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', name)
664                                    if not s_id:
665                                        # Anonymous struct/union
666                                        newmember += f"{r.group(1)} {name}; "
667                                    else:
668                                        newmember += f"{r.group(1)} {s_id}.{name}; "
669                            else:
670                                newmember += f"{arg}; "
671                #
672                # At the end of the s_id loop, replace the original declaration with
673                # the munged version.
674                #
675                members = members.replace(oldmember, newmember)
676            #
677            # End of the tuple loop - search again and see if there are outer members
678            # that now turn up.
679            #
680            tuples = struct_members.findall(members)
681        return members
682
683    def format_struct_decl(self, declaration):
684        """
685        Format the ``struct`` declaration into a standard form for inclusion
686        in the resulting docs.
687        """
688
689        #
690        # Insert newlines, get rid of extra spaces.
691        #
692        declaration = KernRe(r'([\{;])').sub(r'\1\n', declaration)
693        declaration = KernRe(r'\}\s+;').sub('};', declaration)
694        #
695        # Format inline enums with each member on its own line.
696        #
697        r = KernRe(r'(enum\s+\{[^\}]+),([^\n])')
698        while r.search(declaration):
699            declaration = r.sub(r'\1,\n\2', declaration)
700        #
701        # Now go through and supply the right number of tabs
702        # for each line.
703        #
704        def_args = declaration.split('\n')
705        level = 1
706        declaration = ""
707        for clause in def_args:
708            clause = KernRe(r'\s+').sub(' ', clause.strip(), count=1)
709            if clause:
710                if '}' in clause and level > 1:
711                    level -= 1
712                if not clause.startswith('#'):
713                    declaration += "\t" * level
714                declaration += "\t" + clause + "\n"
715                if "{" in clause and "}" not in clause:
716                    level += 1
717        return declaration
718
719
720    def dump_struct(self, ln, proto, source):
721        """
722        Store an entry for a ``struct`` or ``union``
723        """
724        #
725        # Do the basic parse to get the pieces of the declaration.
726        #
727        source = source
728        proto = trim_private_members(proto)
729        struct_parts = self.split_struct_proto(proto)
730        if not struct_parts:
731            self.emit_msg(ln, f"{proto} error: Cannot parse struct or union!")
732            return
733        decl_type, declaration_name, members = struct_parts
734
735        if self.entry.identifier != declaration_name:
736            self.emit_msg(ln, f"expecting prototype for {decl_type} {self.entry.identifier}. "
737                          f"Prototype was for {decl_type} {declaration_name} instead\n")
738            return
739        #
740        # Go through the list of members applying all of our transformations.
741        #
742        members = self.xforms.apply("struct", members)
743
744        #
745        # Deal with embedded struct and union members, and drop enums entirely.
746        #
747        declaration = members
748        members = self.rewrite_struct_members(members)
749        members = re.sub(r'(\{[^\{\}]*\})', '', members)
750        #
751        # Output the result and we are done.
752        #
753        self.create_parameter_list(ln, decl_type, members, ';',
754                                   declaration_name)
755        self.check_sections(ln, declaration_name, decl_type)
756        self.output_declaration(decl_type, declaration_name,
757                                source=source,
758                                definition=self.format_struct_decl(declaration),
759                                purpose=self.entry.declaration_purpose)
760
761    def dump_enum(self, ln, proto, source):
762        """
763        Store an ``enum`` inside self.entries array.
764        """
765        #
766        # Strip preprocessor directives.  Note that this depends on the
767        # trailing semicolon we added in process_proto_type().
768        #
769        source = source
770        proto = trim_private_members(proto)
771        proto = KernRe(r'#\s*((define|ifdef|if)\s+|endif)[^;]*;', flags=re.S).sub('', proto)
772        #
773        # Parse out the name and members of the enum.  Typedef form first.
774        #
775        r = KernRe(r'typedef\s+enum\s*\{(.*)\}\s*(\w*)\s*;')
776        if r.search(proto):
777            declaration_name = r.group(2)
778            members = r.group(1)
779        #
780        # Failing that, look for a straight enum
781        #
782        else:
783            r = KernRe(r'enum\s+(\w*)\s*\{(.*)\}')
784            if r.match(proto):
785                declaration_name = r.group(1)
786                members = r.group(2)
787        #
788        # OK, this isn't going to work.
789        #
790            else:
791                self.emit_msg(ln, f"{proto}: error: Cannot parse enum!")
792                return
793        #
794        # Make sure we found what we were expecting.
795        #
796        if self.entry.identifier != declaration_name:
797            if self.entry.identifier == "":
798                self.emit_msg(ln,
799                              f"{proto}: wrong kernel-doc identifier on prototype")
800            else:
801                self.emit_msg(ln,
802                              f"expecting prototype for enum {self.entry.identifier}. "
803                              f"Prototype was for enum {declaration_name} instead")
804            return
805
806        if not declaration_name:
807            declaration_name = "(anonymous)"
808        #
809        # Parse out the name of each enum member, and verify that we
810        # have a description for it.
811        #
812        member_set = set()
813        members = KernRe(r'\([^;)]*\)').sub('', members)
814        for arg in members.split(','):
815            arg = KernRe(r'^\s*(\w+).*').sub(r'\1', arg)
816            if not arg.strip():
817                continue
818
819            self.entry.parameterlist.append(arg)
820            if arg not in self.entry.parameterdescs:
821                self.entry.parameterdescs[arg] = self.undescribed
822                self.emit_msg(ln,
823                              f"Enum value '{arg}' not described in enum '{declaration_name}'")
824            member_set.add(arg)
825        #
826        # Ensure that every described member actually exists in the enum.
827        #
828        for k in self.entry.parameterdescs:
829            if k not in member_set:
830                self.emit_msg(ln,
831                              f"Excess enum value '@{k}' description in '{declaration_name}'")
832
833        self.output_declaration('enum', declaration_name,
834                                source=source,
835                                purpose=self.entry.declaration_purpose)
836
837    def dump_var(self, ln, proto, source):
838        """
839        Store variables that are part of kAPI.
840        """
841        VAR_ATTRIBS = [
842            "extern",
843            "const",
844        ]
845        OPTIONAL_VAR_ATTR = r"^(?:\b(?:" +"|".join(VAR_ATTRIBS) +r")\b\s*)*"
846
847        #
848        # Store the full prototype before modifying it
849        #
850        source = source
851        full_proto = proto
852        declaration_name = None
853
854        #
855        # Handle macro definitions
856        #
857        macro_prefixes = [
858            KernRe(r"DEFINE_[\w_]+\s*\(([\w_]+)\)"),
859        ]
860
861        for r in macro_prefixes:
862            match = r.search(proto)
863            if match:
864                declaration_name = match.group(1)
865                break
866
867        #
868        # Drop comments and macros to have a pure C prototype
869        #
870        if not declaration_name:
871            proto = self.xforms.apply("var", proto)
872
873        proto = proto.rstrip()
874
875        #
876        # Variable name is at the end of the declaration
877        #
878
879        default_val = None
880
881        r= KernRe(OPTIONAL_VAR_ATTR + r"\s*[\w_\s]*\s+(?:\*+)?([\w_]+)\s*[\d\]\[]*\s*(=.*)?")
882        if r.match(proto):
883            if not declaration_name:
884                declaration_name = r.group(1)
885
886            default_val = r.group(2)
887        else:
888            r= KernRe(OPTIONAL_VAR_ATTR + r"(?:[\w_\s]*)?\s+(?:\*+)?(?:[\w_]+)\s*[\d\]\[]*\s*(=.*)?")
889
890            if r.match(proto):
891                default_val = r.group(1)
892        if not declaration_name:
893           self.emit_msg(ln,f"{proto}: can't parse variable")
894           return
895
896        if default_val:
897            default_val = default_val.lstrip("=").strip()
898
899        self.output_declaration("var", declaration_name,
900                                source=source,
901                                full_proto=full_proto,
902                                default_val=default_val,
903                                purpose=self.entry.declaration_purpose)
904
905    def dump_declaration(self, ln, prototype, source):
906        """
907        Store a data declaration inside self.entries array.
908        """
909
910        if self.entry.decl_type == "enum":
911            self.dump_enum(ln, prototype, source)
912        elif self.entry.decl_type == "typedef":
913            self.dump_typedef(ln, prototype, source)
914        elif self.entry.decl_type in ["union", "struct"]:
915            self.dump_struct(ln, prototype, source)
916        elif self.entry.decl_type == "var":
917            self.dump_var(ln, prototype, source)
918        else:
919            # This would be a bug
920            self.emit_message(ln, f'Unknown declaration type: {self.entry.decl_type}')
921
922    def dump_function(self, ln, prototype, source):
923        """
924        Store a function or function macro inside self.entries array.
925        """
926
927        source = source
928        found = func_macro = False
929        return_type = ''
930        decl_type = 'function'
931
932        #
933        # If we have a macro, remove the "#define" at the front.
934        #
935        new_proto = KernRe(r"^#\s*define\s+").sub("", prototype)
936        if new_proto != prototype:
937            prototype = new_proto
938            #
939            # Dispense with the simple "#define A B" case here; the key
940            # is the space after the name of the symbol being defined.
941            # NOTE that the seemingly misnamed "func_macro" indicates a
942            # macro *without* arguments.
943            #
944            r = KernRe(r'^(\w+)\s+')
945            if r.search(prototype):
946                return_type = ''
947                declaration_name = r.group(1)
948                func_macro = True
949                found = True
950        else:
951            #
952            # Apply the initial transformations.
953            #
954            prototype = self.xforms.apply("func", prototype)
955
956        # Yes, this truly is vile.  We are looking for:
957        # 1. Return type (may be nothing if we're looking at a macro)
958        # 2. Function name
959        # 3. Function parameters.
960        #
961        # All the while we have to watch out for function pointer parameters
962        # (which IIRC is what the two sections are for), C types (these
963        # regexps don't even start to express all the possibilities), and
964        # so on.
965        #
966        # If you mess with these regexps, it's a good idea to check that
967        # the following functions' documentation still comes out right:
968        # - parport_register_device (function pointer parameters)
969        # - atomic_set (macro)
970        # - pci_match_device, __copy_to_user (long return type)
971
972        name = r'\w+'
973        type1 = r'(?:[\w\s]+)?'
974        type2 = r'(?:[\w\s]+\*+)+'
975        #
976        # Attempt to match first on (args) with no internal parentheses; this
977        # lets us easily filter out __acquires() and other post-args stuff.  If
978        # that fails, just grab the rest of the line to the last closing
979        # parenthesis.
980        #
981        proto_args = r'\(([^\(]*|.*)\)'
982        #
983        # (Except for the simple macro case) attempt to split up the prototype
984        # in the various ways we understand.
985        #
986        if not found:
987            patterns = [
988                rf'^()({name})\s*{proto_args}',
989                rf'^({type1})\s+({name})\s*{proto_args}',
990                rf'^({type2})\s*({name})\s*{proto_args}',
991            ]
992
993            for p in patterns:
994                r = KernRe(p)
995                if r.match(prototype):
996                    return_type = r.group(1)
997                    declaration_name = r.group(2)
998                    args = r.group(3)
999                    self.create_parameter_list(ln, decl_type, args, ',',
1000                                               declaration_name)
1001                    found = True
1002                    break
1003        #
1004        # Parsing done; make sure that things are as we expect.
1005        #
1006        if not found:
1007            self.emit_msg(ln,
1008                          f"cannot understand function prototype: '{prototype}'")
1009            return
1010        if self.entry.identifier != declaration_name:
1011            self.emit_msg(ln, f"expecting prototype for {self.entry.identifier}(). "
1012                          f"Prototype was for {declaration_name}() instead")
1013            return
1014        self.check_sections(ln, declaration_name, "function")
1015        self.check_return_section(ln, declaration_name, return_type)
1016        #
1017        # Store the result.
1018        #
1019        self.output_declaration(decl_type, declaration_name,
1020                                source=source,
1021                                typedef=('typedef' in return_type),
1022                                functiontype=return_type,
1023                                purpose=self.entry.declaration_purpose,
1024                                func_macro=func_macro)
1025
1026
1027    def dump_typedef(self, ln, proto, source):
1028        """
1029        Store a ``typedef`` inside self.entries array.
1030        """
1031        #
1032        # We start by looking for function typedefs.
1033        #
1034        typedef_type = r'typedef((?:\s+[\w*]+\b){0,7}\s+(?:\w+\b|\*+))\s*'
1035        typedef_ident = r'\*?\s*(\w\S+)\s*'
1036        typedef_args = r'\s*\((.*)\);'
1037
1038        source = source
1039
1040        typedef1 = KernRe(typedef_type + r'\(' + typedef_ident + r'\)' + typedef_args)
1041        typedef2 = KernRe(typedef_type + typedef_ident + typedef_args)
1042
1043        # Parse function typedef prototypes
1044        for r in [typedef1, typedef2]:
1045            if not r.match(proto):
1046                continue
1047
1048            return_type = r.group(1).strip()
1049            declaration_name = r.group(2)
1050            args = r.group(3)
1051
1052            if self.entry.identifier != declaration_name:
1053                self.emit_msg(ln,
1054                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1055                return
1056
1057            self.create_parameter_list(ln, 'function', args, ',', declaration_name)
1058
1059            self.output_declaration('function', declaration_name,
1060                                    source=source,
1061                                    typedef=True,
1062                                    functiontype=return_type,
1063                                    purpose=self.entry.declaration_purpose)
1064            return
1065        #
1066        # Not a function, try to parse a simple typedef.
1067        #
1068        r = KernRe(r'typedef.*\s+(\w+)\s*;')
1069        if r.match(proto):
1070            declaration_name = r.group(1)
1071
1072            if self.entry.identifier != declaration_name:
1073                self.emit_msg(ln,
1074                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1075                return
1076
1077            self.output_declaration('typedef', declaration_name,
1078                                    source=source,
1079                                    purpose=self.entry.declaration_purpose)
1080            return
1081
1082        self.emit_msg(ln, "error: Cannot parse typedef!")
1083
1084    @staticmethod
1085    def process_export(function_set, line):
1086        """
1087        process ``EXPORT_SYMBOL*`` tags
1088
1089        This method doesn't use any variable from the class, so declare it
1090        with a staticmethod decorator.
1091        """
1092
1093        # We support documenting some exported symbols with different
1094        # names.  A horrible hack.
1095        suffixes = [ '_noprof' ]
1096
1097        # Note: it accepts only one EXPORT_SYMBOL* per line, as having
1098        # multiple export lines would violate Kernel coding style.
1099
1100        if export_symbol.search(line):
1101            symbol = export_symbol.group(2)
1102        elif export_symbol_ns.search(line):
1103            symbol = export_symbol_ns.group(2)
1104        else:
1105            return False
1106        #
1107        # Found an export, trim out any special suffixes
1108        #
1109        for suffix in suffixes:
1110            # Be backward compatible with Python < 3.9
1111            if symbol.endswith(suffix):
1112                symbol = symbol[:-len(suffix)]
1113        function_set.add(symbol)
1114        return True
1115
1116    def process_normal(self, ln, line, source):
1117        """
1118        STATE_NORMAL: looking for the ``/**`` to begin everything.
1119        """
1120
1121        if not doc_start.match(line):
1122            return
1123
1124        # start a new entry
1125        self.reset_state(ln)
1126
1127        # next line is always the function name
1128        self.state = state.NAME
1129
1130    def process_name(self, ln, line, source):
1131        """
1132        STATE_NAME: Looking for the "name - description" line
1133        """
1134        #
1135        # Check for a DOC: block and handle them specially.
1136        #
1137        if doc_block.search(line):
1138
1139            if not doc_block.group(1):
1140                self.entry.begin_section(ln, "Introduction")
1141            else:
1142                self.entry.begin_section(ln, doc_block.group(1))
1143
1144            self.entry.identifier = self.entry.section
1145            self.state = state.DOCBLOCK
1146        #
1147        # Otherwise we're looking for a normal kerneldoc declaration line.
1148        #
1149        elif doc_decl.search(line):
1150            self.entry.identifier = doc_decl.group(1)
1151
1152            # Test for data declaration
1153            if doc_begin_data.search(line):
1154                self.entry.decl_type = doc_begin_data.group(1)
1155                self.entry.identifier = doc_begin_data.group(2)
1156            #
1157            # Look for a function description
1158            #
1159            elif doc_begin_func.search(line):
1160                self.entry.identifier = doc_begin_func.group(1)
1161                self.entry.decl_type = "function"
1162            #
1163            # We struck out.
1164            #
1165            else:
1166                self.emit_msg(ln,
1167                              f"This comment starts with '/**', but isn't a kernel-doc comment. Refer to Documentation/doc-guide/kernel-doc.rst\n{line}")
1168                self.state = state.NORMAL
1169                return
1170            #
1171            # OK, set up for a new kerneldoc entry.
1172            #
1173            self.state = state.BODY
1174            self.entry.identifier = self.entry.identifier.strip(" ")
1175            # if there's no @param blocks need to set up default section here
1176            self.entry.begin_section(ln + 1)
1177            #
1178            # Find the description portion, which *should* be there but
1179            # isn't always.
1180            # (We should be able to capture this from the previous parsing - someday)
1181            #
1182            r = KernRe("[-:](.*)")
1183            if r.search(line):
1184                self.entry.declaration_purpose = trim_whitespace(r.group(1))
1185                self.state = state.DECLARATION
1186            else:
1187                self.entry.declaration_purpose = ""
1188
1189            if not self.entry.declaration_purpose and self.config.wshort_desc:
1190                self.emit_msg(ln,
1191                              f"missing initial short description on line:\n{line}")
1192
1193            if not self.entry.identifier and self.entry.decl_type != "enum":
1194                self.emit_msg(ln,
1195                              f"wrong kernel-doc identifier on line:\n{line}")
1196                self.state = state.NORMAL
1197
1198            if self.config.verbose:
1199                self.emit_msg(ln,
1200                              f"Scanning doc for {self.entry.decl_type} {self.entry.identifier}",
1201                                  warning=False)
1202        #
1203        # Failed to find an identifier. Emit a warning
1204        #
1205        else:
1206            self.emit_msg(ln, f"Cannot find identifier on line:\n{line}")
1207
1208    def is_new_section(self, ln, line):
1209        """
1210        Helper function to determine if a new section is being started.
1211        """
1212        if doc_sect.search(line):
1213            self.state = state.BODY
1214            #
1215            # Pick out the name of our new section, tweaking it if need be.
1216            #
1217            newsection = doc_sect.group(1)
1218            if newsection.lower() == 'description':
1219                newsection = 'Description'
1220            elif newsection.lower() == 'context':
1221                newsection = 'Context'
1222                self.state = state.SPECIAL_SECTION
1223            elif newsection.lower() in ["@return", "@returns",
1224                                        "return", "returns"]:
1225                newsection = "Return"
1226                self.state = state.SPECIAL_SECTION
1227            elif newsection[0] == '@':
1228                self.state = state.SPECIAL_SECTION
1229            #
1230            # Initialize the contents, and get the new section going.
1231            #
1232            newcontents = doc_sect.group(2)
1233            if not newcontents:
1234                newcontents = ""
1235            self.dump_section()
1236            self.entry.begin_section(ln, newsection)
1237            self.entry.leading_space = None
1238
1239            self.entry.add_text(newcontents.lstrip())
1240            return True
1241        return False
1242
1243    def is_comment_end(self, ln, line):
1244        """
1245        Helper function to detect (and effect) the end of a kerneldoc comment.
1246        """
1247        if doc_end.search(line):
1248            self.dump_section()
1249
1250            # Look for doc_com + <text> + doc_end:
1251            r = KernRe(r'\s*\*\s*[a-zA-Z_0-9:.]+\*/')
1252            if r.match(line):
1253                self.emit_msg(ln, f"suspicious ending line: {line}")
1254
1255            self.entry.prototype = ""
1256            self.entry.new_start_line = ln + 1
1257
1258            self.state = state.PROTO
1259            return True
1260        return False
1261
1262
1263    def process_decl(self, ln, line, source):
1264        """
1265        STATE_DECLARATION: We've seen the beginning of a declaration.
1266        """
1267        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1268            return
1269        #
1270        # Look for anything with the " * " line beginning.
1271        #
1272        if doc_content.search(line):
1273            cont = doc_content.group(1)
1274            #
1275            # A blank line means that we have moved out of the declaration
1276            # part of the comment (without any "special section" parameter
1277            # descriptions).
1278            #
1279            if cont == "":
1280                self.state = state.BODY
1281            #
1282            # Otherwise we have more of the declaration section to soak up.
1283            #
1284            else:
1285                self.entry.declaration_purpose = \
1286                    trim_whitespace(self.entry.declaration_purpose + ' ' + cont)
1287        else:
1288            # Unknown line, ignore
1289            self.emit_msg(ln, f"bad line: {line}")
1290
1291
1292    def process_special(self, ln, line, source):
1293        """
1294        STATE_SPECIAL_SECTION: a section ending with a blank line.
1295        """
1296        #
1297        # If we have hit a blank line (only the " * " marker), then this
1298        # section is done.
1299        #
1300        if KernRe(r"\s*\*\s*$").match(line):
1301            self.entry.begin_section(ln, dump = True)
1302            self.state = state.BODY
1303            return
1304        #
1305        # Not a blank line, look for the other ways to end the section.
1306        #
1307        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1308            return
1309        #
1310        # OK, we should have a continuation of the text for this section.
1311        #
1312        if doc_content.search(line):
1313            cont = doc_content.group(1)
1314            #
1315            # If the lines of text after the first in a special section have
1316            # leading white space, we need to trim it out or Sphinx will get
1317            # confused.  For the second line (the None case), see what we
1318            # find there and remember it.
1319            #
1320            if self.entry.leading_space is None:
1321                r = KernRe(r'^(\s+)')
1322                if r.match(cont):
1323                    self.entry.leading_space = len(r.group(1))
1324                else:
1325                    self.entry.leading_space = 0
1326            #
1327            # Otherwise, before trimming any leading chars, be *sure*
1328            # that they are white space.  We should maybe warn if this
1329            # isn't the case.
1330            #
1331            for i in range(0, self.entry.leading_space):
1332                if cont[i] != " ":
1333                    self.entry.leading_space = i
1334                    break
1335            #
1336            # Add the trimmed result to the section and we're done.
1337            #
1338            self.entry.add_text(cont[self.entry.leading_space:])
1339        else:
1340            # Unknown line, ignore
1341            self.emit_msg(ln, f"bad line: {line}")
1342
1343    def process_body(self, ln, line, source):
1344        """
1345        STATE_BODY: the bulk of a kerneldoc comment.
1346        """
1347        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1348            return
1349
1350        if doc_content.search(line):
1351            cont = doc_content.group(1)
1352            self.entry.add_text(cont)
1353        else:
1354            # Unknown line, ignore
1355            self.emit_msg(ln, f"bad line: {line}")
1356
1357    def process_inline_name(self, ln, line, source):
1358        """STATE_INLINE_NAME: beginning of docbook comments within a prototype."""
1359
1360        if doc_inline_sect.search(line):
1361            self.entry.begin_section(ln, doc_inline_sect.group(1))
1362            self.entry.add_text(doc_inline_sect.group(2).lstrip())
1363            self.state = state.INLINE_TEXT
1364        elif doc_inline_end.search(line):
1365            self.dump_section()
1366            self.state = state.PROTO
1367        elif doc_content.search(line):
1368            self.emit_msg(ln, f"Incorrect use of kernel-doc format: {line}")
1369            self.state = state.PROTO
1370
1371            #
1372            # Don't let it add partial comments at the code, as breaks the
1373            # logic meant to remove comments from prototypes.
1374            #
1375            self.process_proto_type(ln, "/**\n" + line, source)
1376        # else ... ??
1377
1378    def process_inline_text(self, ln, line, source):
1379        """STATE_INLINE_TEXT: docbook comments within a prototype."""
1380
1381        if doc_inline_end.search(line):
1382            self.dump_section()
1383            self.state = state.PROTO
1384        elif doc_content.search(line):
1385            self.entry.add_text(doc_content.group(1))
1386        # else ... ??
1387
1388    def syscall_munge(self, ln, proto):         # pylint: disable=W0613
1389        """
1390        Handle syscall definitions.
1391        """
1392
1393        is_void = False
1394
1395        # Strip newlines/CR's
1396        proto = re.sub(r'[\r\n]+', ' ', proto)
1397
1398        # Check if it's a SYSCALL_DEFINE0
1399        if 'SYSCALL_DEFINE0' in proto:
1400            is_void = True
1401
1402        # Replace SYSCALL_DEFINE with correct return type & function name
1403        proto = KernRe(r'SYSCALL_DEFINE.*\(').sub('long sys_', proto)
1404
1405        r = KernRe(r'long\s+(sys_.*?),')
1406        if r.search(proto):
1407            proto = KernRe(',').sub('(', proto, count=1)
1408        elif is_void:
1409            proto = KernRe(r'\)').sub('(void)', proto, count=1)
1410
1411        # Now delete all of the odd-numbered commas in the proto
1412        # so that argument types & names don't have a comma between them
1413        count = 0
1414        length = len(proto)
1415
1416        if is_void:
1417            length = 0  # skip the loop if is_void
1418
1419        for ix in range(length):
1420            if proto[ix] == ',':
1421                count += 1
1422                if count % 2 == 1:
1423                    proto = proto[:ix] + ' ' + proto[ix + 1:]
1424
1425        return proto
1426
1427    def tracepoint_munge(self, ln, proto):
1428        """
1429        Handle tracepoint definitions.
1430        """
1431
1432        tracepointname = None
1433        tracepointargs = None
1434
1435        # Match tracepoint name based on different patterns
1436        r = KernRe(r'TRACE_EVENT\((.*?),')
1437        if r.search(proto):
1438            tracepointname = r.group(1)
1439
1440        r = KernRe(r'DEFINE_SINGLE_EVENT\((.*?),')
1441        if r.search(proto):
1442            tracepointname = r.group(1)
1443
1444        r = KernRe(r'DEFINE_EVENT\((.*?),(.*?),')
1445        if r.search(proto):
1446            tracepointname = r.group(2)
1447
1448        if tracepointname:
1449            tracepointname = tracepointname.lstrip()
1450
1451        r = KernRe(r'TP_PROTO\((.*?)\)')
1452        if r.search(proto):
1453            tracepointargs = r.group(1)
1454
1455        if not tracepointname or not tracepointargs:
1456            self.emit_msg(ln,
1457                          f"Unrecognized tracepoint format:\n{proto}\n")
1458        else:
1459            proto = f"static inline void trace_{tracepointname}({tracepointargs})"
1460            self.entry.identifier = f"trace_{self.entry.identifier}"
1461
1462        return proto
1463
1464    def process_proto_function(self, ln, line, source):
1465        """Ancillary routine to process a function prototype."""
1466
1467        # strip C99-style comments to end of line
1468        line = KernRe(r"//.*$", re.S).sub('', line)
1469        #
1470        # Soak up the line's worth of prototype text, stopping at { or ; if present.
1471        #
1472        if KernRe(r'\s*#\s*define').match(line):
1473            self.entry.prototype = line
1474        elif not line.startswith('#'):   # skip other preprocessor stuff
1475            r = KernRe(r'([^\{]*)')
1476            if r.match(line):
1477                self.entry.prototype += r.group(1) + " "
1478        #
1479        # If we now have the whole prototype, clean it up and declare victory.
1480        #
1481        if '{' in line or ';' in line or KernRe(r'\s*#\s*define').match(line):
1482            # strip comments and surrounding spaces
1483            self.entry.prototype = KernRe(r'/\*.*\*/').sub('', self.entry.prototype).strip()
1484            #
1485            # Handle self.entry.prototypes for function pointers like:
1486            #       int (*pcs_config)(struct foo)
1487            # by turning it into
1488            #	    int pcs_config(struct foo)
1489            #
1490            r = KernRe(r'^(\S+\s+)\(\s*\*(\S+)\)')
1491            self.entry.prototype = r.sub(r'\1\2', self.entry.prototype)
1492            #
1493            # Handle special declaration syntaxes
1494            #
1495            if 'SYSCALL_DEFINE' in self.entry.prototype:
1496                self.entry.prototype = self.syscall_munge(ln,
1497                                                          self.entry.prototype)
1498            else:
1499                r = KernRe(r'TRACE_EVENT|DEFINE_EVENT|DEFINE_SINGLE_EVENT')
1500                if r.search(self.entry.prototype):
1501                    self.entry.prototype = self.tracepoint_munge(ln,
1502                                                                 self.entry.prototype)
1503            #
1504            # ... and we're done
1505            #
1506            self.dump_function(ln, self.entry.prototype, source)
1507            self.reset_state(ln)
1508
1509    def process_proto_type(self, ln, line, source):
1510        """
1511        Ancillary routine to process a type.
1512        """
1513
1514        # Strip C99-style comments and surrounding whitespace
1515        line = KernRe(r"//.*$", re.S).sub('', line).strip()
1516        if not line:
1517            return # nothing to see here
1518
1519        # To distinguish preprocessor directive from regular declaration later.
1520        if line.startswith('#'):
1521            line += ";"
1522        #
1523        # Split the declaration on any of { } or ;, and accumulate pieces
1524        # until we hit a semicolon while not inside {brackets}
1525        #
1526        r = KernRe(r'(.*?)([{};])')
1527        for chunk in r.split(line):
1528            if chunk:  # Ignore empty matches
1529                self.entry.prototype += chunk
1530                #
1531                # This cries out for a match statement ... someday after we can
1532                # drop Python 3.9 ...
1533                #
1534                if chunk == '{':
1535                    self.entry.brcount += 1
1536                elif chunk == '}':
1537                    self.entry.brcount -= 1
1538                elif chunk == ';' and self.entry.brcount <= 0:
1539                    self.dump_declaration(ln, self.entry.prototype, source)
1540                    self.reset_state(ln)
1541                    return
1542        #
1543        # We hit the end of the line while still in the declaration; put
1544        # in a space to represent the newline.
1545        #
1546        self.entry.prototype += ' '
1547
1548    def process_proto(self, ln, line, source):
1549        """STATE_PROTO: reading a function/whatever prototype."""
1550
1551        if doc_inline_oneline.search(line):
1552            self.entry.begin_section(ln, doc_inline_oneline.group(1))
1553            self.entry.add_text(doc_inline_oneline.group(2))
1554            self.dump_section()
1555
1556        elif doc_inline_start.search(line):
1557            self.state = state.INLINE_NAME
1558
1559        elif self.entry.decl_type == 'function':
1560            self.process_proto_function(ln, line, source)
1561
1562        else:
1563            self.process_proto_type(ln, line, source)
1564
1565    def process_docblock(self, ln, line, source):
1566        """STATE_DOCBLOCK: within a ``DOC:`` block."""
1567
1568        if doc_end.search(line):
1569            self.dump_section()
1570            self.output_declaration("doc", self.entry.identifier,
1571                                    source=source)
1572            self.reset_state(ln)
1573
1574        elif doc_content.search(line):
1575            self.entry.add_text(doc_content.group(1))
1576
1577    def parse_export(self):
1578        """
1579        Parses ``EXPORT_SYMBOL*`` macros from a single Kernel source file.
1580        """
1581
1582        export_table = set()
1583
1584        try:
1585            with open(self.fname, "r", encoding="utf8",
1586                      errors="backslashreplace") as fp:
1587
1588                for line in fp:
1589                    self.process_export(export_table, line)
1590
1591        except IOError:
1592            return None
1593
1594        return export_table
1595
1596    #: The state/action table telling us which function to invoke in each state.
1597    state_actions = {
1598        state.NORMAL:			process_normal,
1599        state.NAME:			process_name,
1600        state.BODY:			process_body,
1601        state.DECLARATION:		process_decl,
1602        state.SPECIAL_SECTION:		process_special,
1603        state.INLINE_NAME:		process_inline_name,
1604        state.INLINE_TEXT:		process_inline_text,
1605        state.PROTO:			process_proto,
1606        state.DOCBLOCK:			process_docblock,
1607        }
1608
1609    def parse_kdoc(self):
1610        """
1611        Open and process each line of a C source file.
1612        The parsing is controlled via a state machine, and the line is passed
1613        to a different process function depending on the state. The process
1614        function may update the state as needed.
1615
1616        Besides parsing kernel-doc tags, it also parses export symbols.
1617        """
1618
1619        prev = ""
1620        prev_ln = None
1621        export_table = set()
1622        self.state = state.NORMAL
1623        source = ""
1624
1625        try:
1626            with open(self.fname, "r", encoding="utf8",
1627                      errors="backslashreplace") as fp:
1628                for ln, line in enumerate(fp):
1629
1630                    line = line.expandtabs().strip("\n")
1631
1632                    # Group continuation lines on prototypes
1633                    if self.state == state.PROTO:
1634                        if line.endswith("\\"):
1635                            prev += line.rstrip("\\")
1636                            if not prev_ln:
1637                                prev_ln = ln
1638                            continue
1639
1640                        if prev:
1641                            ln = prev_ln
1642                            line = prev + line
1643                            prev = ""
1644                            prev_ln = None
1645
1646                    self.config.log.debug("%d %s: %s",
1647                                          ln, state.name[self.state],
1648                                          line)
1649
1650                    if self.store_src:
1651                        if source and self.state == state.NORMAL:
1652                            source = ""
1653                        elif self.state != state.NORMAL:
1654                            source += line + "\n"
1655
1656                    # This is an optimization over the original script.
1657                    # There, when export_file was used for the same file,
1658                    # it was read twice. Here, we use the already-existing
1659                    # loop to parse exported symbols as well.
1660                    #
1661                    if (self.state != state.NORMAL) or \
1662                       not self.process_export(export_table, line):
1663                        prev_state = self.state
1664                        # Hand this line to the appropriate state handler
1665                        self.state_actions[self.state](self, ln, line, source)
1666                        if prev_state == state.NORMAL and self.state != state.NORMAL:
1667                            source += line + "\n"
1668
1669            self.emit_unused_warnings()
1670
1671        except OSError:
1672            self.config.log.error(f"Error: Cannot open file {self.fname}")
1673
1674        return export_table, self.entries
1675