xref: /linux/tools/lib/python/kdoc/kdoc_parser.py (revision d842057c4a205084fb3036122c7426963f04e826)
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.kdoc_re import NestedMatch, KernRe
17from kdoc.kdoc_item import KdocItem
18
19#
20# Regular expressions used to parse kernel-doc markups at KernelDoc class.
21#
22# Let's declare them in lowercase outside any class to make it easier to
23# convert from the Perl script.
24#
25# As those are evaluated at the beginning, no need to cache them
26#
27
28# Allow whitespace at end of comment start.
29doc_start = KernRe(r'^/\*\*\s*$', cache=False)
30
31doc_end = KernRe(r'\*/', cache=False)
32doc_com = KernRe(r'\s*\*\s*', cache=False)
33doc_com_body = KernRe(r'\s*\* ?', cache=False)
34doc_decl = doc_com + KernRe(r'(\w+)', cache=False)
35
36# @params and a strictly limited set of supported section names
37# Specifically:
38#   Match @word:
39#         @...:
40#         @{section-name}:
41# while trying to not match literal block starts like "example::"
42#
43known_section_names = 'description|context|returns?|notes?|examples?'
44known_sections = KernRe(known_section_names, flags = re.I)
45doc_sect = doc_com + \
46    KernRe(r'\s*(@[.\w]+|@\.\.\.|' + known_section_names + r')\s*:([^:].*)?$',
47           flags=re.I, cache=False)
48
49doc_content = doc_com_body + KernRe(r'(.*)', cache=False)
50doc_inline_start = KernRe(r'^\s*/\*\*\s*$', cache=False)
51doc_inline_sect = KernRe(r'\s*\*\s*(@\s*[\w][\w\.]*\s*):(.*)', cache=False)
52doc_inline_end = KernRe(r'^\s*\*/\s*$', cache=False)
53doc_inline_oneline = KernRe(r'^\s*/\*\*\s*(@\s*[\w][\w\.]*\s*):\s*(.*)\s*\*/\s*$', cache=False)
54
55export_symbol = KernRe(r'^\s*EXPORT_SYMBOL(_GPL)?\s*\(\s*(\w+)\s*\)\s*', cache=False)
56export_symbol_ns = KernRe(r'^\s*EXPORT_SYMBOL_NS(_GPL)?\s*\(\s*(\w+)\s*,\s*"\S+"\)\s*', cache=False)
57
58type_param = KernRe(r"@(\w*((\.\w+)|(->\w+))*(\.\.\.)?)", cache=False)
59
60#
61# Tests for the beginning of a kerneldoc block in its various forms.
62#
63doc_block = doc_com + KernRe(r'DOC:\s*(.*)?', cache=False)
64doc_begin_data = KernRe(r"^\s*\*?\s*(struct|union|enum|typedef|var)\b\s*(\w*)", cache = False)
65doc_begin_func = KernRe(str(doc_com) +			# initial " * '
66                        r"(?:\w+\s*\*\s*)?" + 		# type (not captured)
67                        r'(?:define\s+)?' + 		# possible "define" (not captured)
68                        r'(\w+)\s*(?:\(\w*\))?\s*' +	# name and optional "(...)"
69                        r'(?:[-:].*)?$',		# description (not captured)
70                        cache = False)
71
72#
73# Regexes here are guaranteed to have the end delimiter matching
74# the start delimiter. Yet, right now, only one replace group
75# is allowed.
76#
77struct_nested_prefixes = [
78    (re.compile(r'\bSTRUCT_GROUP\('), r'\1'),
79]
80
81#
82# Ancillary functions
83#
84
85multi_space = KernRe(r'\s\s+')
86def trim_whitespace(s):
87    """
88    A little helper to get rid of excess white space.
89    """
90    return multi_space.sub(' ', s.strip())
91
92def trim_private_members(text):
93    """
94    Remove ``struct``/``enum`` members that have been marked "private".
95    """
96    # First look for a "public:" block that ends a private region, then
97    # handle the "private until the end" case.
98    #
99    text = KernRe(r'/\*\s*private:.*?/\*\s*public:.*?\*/', flags=re.S).sub('', text)
100    text = KernRe(r'/\*\s*private:.*', flags=re.S).sub('', text)
101    #
102    # We needed the comments to do the above, but now we can take them out.
103    #
104    return KernRe(r'\s*/\*.*?\*/\s*', flags=re.S).sub('', text).strip()
105
106class state:
107    """
108    States used by the parser's state machine.
109    """
110
111    # Parser states
112    NORMAL        = 0        #: Normal code.
113    NAME          = 1        #: Looking for function name.
114    DECLARATION   = 2        #: We have seen a declaration which might not be done.
115    BODY          = 3        #: The body of the comment.
116    SPECIAL_SECTION = 4      #: Doc section ending with a blank line.
117    PROTO         = 5        #: Scanning prototype.
118    DOCBLOCK      = 6        #: Documentation block.
119    INLINE_NAME   = 7        #: Gathering doc outside main block.
120    INLINE_TEXT   = 8	     #: Reading the body of inline docs.
121
122    #: Names for each parser state.
123    name = [
124        "NORMAL",
125        "NAME",
126        "DECLARATION",
127        "BODY",
128        "SPECIAL_SECTION",
129        "PROTO",
130        "DOCBLOCK",
131        "INLINE_NAME",
132        "INLINE_TEXT",
133    ]
134
135
136SECTION_DEFAULT = "Description"  #: Default section.
137
138class KernelEntry:
139    """
140    Encapsulates a Kernel documentation entry.
141    """
142
143    def __init__(self, config, fname, ln):
144        self.config = config
145        self.fname = fname
146
147        self._contents = []
148        self.prototype = ""
149
150        self.warnings = []
151
152        self.parameterlist = []
153        self.parameterdescs = {}
154        self.parametertypes = {}
155        self.parameterdesc_start_lines = {}
156
157        self.section_start_lines = {}
158        self.sections = {}
159
160        self.anon_struct_union = False
161
162        self.leading_space = None
163
164        self.fname = fname
165
166        # State flags
167        self.brcount = 0
168        self.declaration_start_line = ln + 1
169
170    #
171    # Management of section contents
172    #
173    def add_text(self, text):
174        """Add a new text to the entry contents list."""
175        self._contents.append(text)
176
177    def contents(self):
178        """Returns a string with all content texts that were added."""
179        return '\n'.join(self._contents) + '\n'
180
181    # TODO: rename to emit_message after removal of kernel-doc.pl
182    def emit_msg(self, ln, msg, *, warning=True):
183        """Emit a message."""
184
185        log_msg = f"{self.fname}:{ln} {msg}"
186
187        if not warning:
188            self.config.log.info(log_msg)
189            return
190
191        # Delegate warning output to output logic, as this way it
192        # will report warnings/info only for symbols that are output
193
194        self.warnings.append(log_msg)
195        return
196
197    def begin_section(self, line_no, title = SECTION_DEFAULT, dump = False):
198        """
199        Begin a new section.
200        """
201        if dump:
202            self.dump_section(start_new = True)
203        self.section = title
204        self.new_start_line = line_no
205
206    def dump_section(self, start_new=True):
207        """
208        Dumps section contents to arrays/hashes intended for that purpose.
209        """
210        #
211        # If we have accumulated no contents in the default ("description")
212        # section, don't bother.
213        #
214        if self.section == SECTION_DEFAULT and not self._contents:
215            return
216        name = self.section
217        contents = self.contents()
218
219        if type_param.match(name):
220            name = type_param.group(1)
221
222            self.parameterdescs[name] = contents
223            self.parameterdesc_start_lines[name] = self.new_start_line
224
225            self.new_start_line = 0
226
227        else:
228            if name in self.sections and self.sections[name] != "":
229                # Only warn on user-specified duplicate section names
230                if name != SECTION_DEFAULT:
231                    self.emit_msg(self.new_start_line,
232                                  f"duplicate section name '{name}'")
233                # Treat as a new paragraph - add a blank line
234                self.sections[name] += '\n' + contents
235            else:
236                self.sections[name] = contents
237                self.section_start_lines[name] = self.new_start_line
238                self.new_start_line = 0
239
240#        self.config.log.debug("Section: %s : %s", name, pformat(vars(self)))
241
242        if start_new:
243            self.section = SECTION_DEFAULT
244            self._contents = []
245
246python_warning = False
247
248class KernelDoc:
249    """
250    Read a C language source or header FILE and extract embedded
251    documentation comments.
252    """
253
254    #: Name of context section.
255    section_context = "Context"
256
257    #: Name of return section.
258    section_return = "Return"
259
260    #: String to write when a parameter is not described.
261    undescribed = "-- undescribed --"
262
263    def __init__(self, config, fname, xforms):
264        """Initialize internal variables"""
265
266        self.fname = fname
267        self.config = config
268        self.xforms = xforms
269
270        # Initial state for the state machines
271        self.state = state.NORMAL
272
273        # Store entry currently being processed
274        self.entry = None
275
276        # Place all potential outputs into an array
277        self.entries = []
278
279        #
280        # We need Python 3.7 for its "dicts remember the insertion
281        # order" guarantee
282        #
283        global python_warning
284        if (not python_warning and
285            sys.version_info.major == 3 and sys.version_info.minor < 7):
286
287            self.emit_msg(0,
288                          'Python 3.7 or later is required for correct results')
289            python_warning = True
290
291    def emit_msg(self, ln, msg, *, warning=True):
292        """Emit a message"""
293
294        if self.entry:
295            self.entry.emit_msg(ln, msg, warning=warning)
296            return
297
298        log_msg = f"{self.fname}:{ln} {msg}"
299
300        if warning:
301            self.config.log.warning(log_msg)
302        else:
303            self.config.log.info(log_msg)
304
305    def dump_section(self, start_new=True):
306        """
307        Dump section contents to arrays/hashes intended for that purpose.
308        """
309
310        if self.entry:
311            self.entry.dump_section(start_new)
312
313    # TODO: rename it to store_declaration after removal of kernel-doc.pl
314    def output_declaration(self, dtype, name, **args):
315        """
316        Store the entry into an entry array.
317
318        The actual output and output filters will be handled elsewhere.
319        """
320
321        item = KdocItem(name, self.fname, dtype,
322                        self.entry.declaration_start_line, **args)
323        item.warnings = self.entry.warnings
324
325        # Drop empty sections
326        # TODO: improve empty sections logic to emit warnings
327        sections = self.entry.sections
328        for section in ["Description", "Return"]:
329            if section in sections and not sections[section].rstrip():
330                del sections[section]
331        item.set_sections(sections, self.entry.section_start_lines)
332        item.set_params(self.entry.parameterlist, self.entry.parameterdescs,
333                        self.entry.parametertypes,
334                        self.entry.parameterdesc_start_lines)
335        self.entries.append(item)
336
337        self.config.log.debug("Output: %s:%s = %s", dtype, name, pformat(args))
338
339    def emit_unused_warnings(self):
340        """
341        When the parser fails to produce a valid entry, it places some
342        warnings under `entry.warnings` that will be discarded when resetting
343        the state.
344
345        Ensure that those warnings are not lost.
346
347        .. note::
348
349              Because we are calling `config.warning()` here, those
350              warnings are not filtered by the `-W` parameters: they will all
351              be produced even when `-Wreturn`, `-Wshort-desc`, and/or
352              `-Wcontents-before-sections` are used.
353
354              Allowing those warnings to be filtered is complex, because it
355              would require storing them in a buffer and then filtering them
356              during the output step of the code, depending on the
357              selected symbols.
358        """
359        if self.entry and self.entry not in self.entries:
360            for log_msg in self.entry.warnings:
361                self.config.warning(log_msg)
362
363    def reset_state(self, ln):
364        """
365        Ancillary routine to create a new entry. It initializes all
366        variables used by the state machine.
367        """
368
369        self.emit_unused_warnings()
370
371        self.entry = KernelEntry(self.config, self.fname, ln)
372
373        # State flags
374        self.state = state.NORMAL
375
376    def push_parameter(self, ln, decl_type, param, dtype,
377                       org_arg, declaration_name):
378        """
379        Store parameters and their descriptions at self.entry.
380        """
381
382        if self.entry.anon_struct_union and dtype == "" and param == "}":
383            return  # Ignore the ending }; from anonymous struct/union
384
385        self.entry.anon_struct_union = False
386
387        param = KernRe(r'[\[\)].*').sub('', param, count=1)
388
389        #
390        # Look at various "anonymous type" cases.
391        #
392        if dtype == '':
393            if param.endswith("..."):
394                if len(param) > 3: # there is a name provided, use that
395                    param = param[:-3]
396                if not self.entry.parameterdescs.get(param):
397                    self.entry.parameterdescs[param] = "variable arguments"
398
399            elif (not param) or param == "void":
400                param = "void"
401                self.entry.parameterdescs[param] = "no arguments"
402
403            elif param in ["struct", "union"]:
404                # Handle unnamed (anonymous) union or struct
405                dtype = param
406                param = "{unnamed_" + param + "}"
407                self.entry.parameterdescs[param] = "anonymous\n"
408                self.entry.anon_struct_union = True
409
410        # Warn if parameter has no description
411        # (but ignore ones starting with # as these are not parameters
412        # but inline preprocessor statements)
413        if param not in self.entry.parameterdescs and not param.startswith("#"):
414            self.entry.parameterdescs[param] = self.undescribed
415
416            if "." not in param:
417                if decl_type == 'function':
418                    dname = f"{decl_type} parameter"
419                else:
420                    dname = f"{decl_type} member"
421
422                self.emit_msg(ln,
423                              f"{dname} '{param}' not described in '{declaration_name}'")
424
425        # Strip spaces from param so that it is one continuous string on
426        # parameterlist. This fixes a problem where check_sections()
427        # cannot find a parameter like "addr[6 + 2]" because it actually
428        # appears as "addr[6", "+", "2]" on the parameter list.
429        # However, it's better to maintain the param string unchanged for
430        # output, so just weaken the string compare in check_sections()
431        # to ignore "[blah" in a parameter string.
432
433        self.entry.parameterlist.append(param)
434        org_arg = KernRe(r'\s\s+').sub(' ', org_arg)
435        self.entry.parametertypes[param] = org_arg
436
437
438    def create_parameter_list(self, ln, decl_type, args,
439                              splitter, declaration_name):
440        """
441        Creates a list of parameters, storing them at self.entry.
442        """
443
444        # temporarily replace all commas inside function pointer definition
445        arg_expr = KernRe(r'(\([^\),]+),')
446        while arg_expr.search(args):
447            args = arg_expr.sub(r"\1#", args)
448
449        for arg in args.split(splitter):
450            # Ignore argument attributes
451            arg = KernRe(r'\sPOS0?\s').sub(' ', arg)
452
453            # Strip leading/trailing spaces
454            arg = arg.strip()
455            arg = KernRe(r'\s+').sub(' ', arg, count=1)
456
457            if arg.startswith('#'):
458                # Treat preprocessor directive as a typeless variable just to fill
459                # corresponding data structures "correctly". Catch it later in
460                # output_* subs.
461
462                # Treat preprocessor directive as a typeless variable
463                self.push_parameter(ln, decl_type, arg, "",
464                                    "", declaration_name)
465            #
466            # The pointer-to-function case.
467            #
468            elif KernRe(r'\(.+\)\s*\(').search(arg):
469                arg = arg.replace('#', ',')
470                r = KernRe(r'[^\(]+\(\*?\s*'  # Everything up to "(*"
471                           r'([\w\[\].]*)'    # Capture the name and possible [array]
472                           r'\s*\)')	      # Make sure the trailing ")" is there
473                if r.match(arg):
474                    param = r.group(1)
475                else:
476                    self.emit_msg(ln, f"Invalid param: {arg}")
477                    param = arg
478                dtype = arg.replace(param, '')
479                self.push_parameter(ln, decl_type, param, dtype, arg, declaration_name)
480            #
481            # The array-of-pointers case.  Dig the parameter name out from the middle
482            # of the declaration.
483            #
484            elif KernRe(r'\(.+\)\s*\[').search(arg):
485                r = KernRe(r'[^\(]+\(\s*\*\s*'		# Up to "(" and maybe "*"
486                           r'([\w.]*?)'			# The actual pointer name
487                           r'\s*(\[\s*\w+\s*\]\s*)*\)') # The [array portion]
488                if r.match(arg):
489                    param = r.group(1)
490                else:
491                    self.emit_msg(ln, f"Invalid param: {arg}")
492                    param = arg
493                dtype = arg.replace(param, '')
494                self.push_parameter(ln, decl_type, param, dtype, arg, declaration_name)
495            elif arg:
496                #
497                # Clean up extraneous spaces and split the string at commas; the first
498                # element of the resulting list will also include the type information.
499                #
500                arg = KernRe(r'\s*:\s*').sub(":", arg)
501                arg = KernRe(r'\s*\[').sub('[', arg)
502                args = KernRe(r'\s*,\s*').split(arg)
503                args[0] = re.sub(r'(\*+)\s*', r' \1', args[0])
504                #
505                # args[0] has a string of "type a".  If "a" includes an [array]
506                # declaration, we want to not be fooled by any white space inside
507                # the brackets, so detect and handle that case specially.
508                #
509                r = KernRe(r'^([^[\]]*\s+)(.*)$')
510                if r.match(args[0]):
511                    args[0] = r.group(2)
512                    dtype = r.group(1)
513                else:
514                    # No space in args[0]; this seems wrong but preserves previous behavior
515                    dtype = ''
516
517                bitfield_re = KernRe(r'(.*?):(\w+)')
518                for param in args:
519                    #
520                    # For pointers, shift the star(s) from the variable name to the
521                    # type declaration.
522                    #
523                    r = KernRe(r'^(\*+)\s*(.*)')
524                    if r.match(param):
525                        self.push_parameter(ln, decl_type, r.group(2),
526                                            f"{dtype} {r.group(1)}",
527                                            arg, declaration_name)
528                    #
529                    # Perform a similar shift for bitfields.
530                    #
531                    elif bitfield_re.search(param):
532                        if dtype != "":  # Skip unnamed bit-fields
533                            self.push_parameter(ln, decl_type, bitfield_re.group(1),
534                                                f"{dtype}:{bitfield_re.group(2)}",
535                                                arg, declaration_name)
536                    else:
537                        self.push_parameter(ln, decl_type, param, dtype,
538                                            arg, declaration_name)
539
540    def check_sections(self, ln, decl_name, decl_type):
541        """
542        Check for errors inside sections, emitting warnings if not found
543        parameters are described.
544        """
545        for section in self.entry.sections:
546            if section not in self.entry.parameterlist and \
547               not known_sections.search(section):
548                if decl_type == 'function':
549                    dname = f"{decl_type} parameter"
550                else:
551                    dname = f"{decl_type} member"
552                self.emit_msg(ln,
553                              f"Excess {dname} '{section}' description in '{decl_name}'")
554
555    def check_return_section(self, ln, declaration_name, return_type):
556        """
557        If the function doesn't return void, warns about the lack of a
558        return description.
559        """
560
561        if not self.config.wreturn:
562            return
563
564        # Ignore an empty return type (It's a macro)
565        # Ignore functions with a "void" return type (but not "void *")
566        if not return_type or KernRe(r'void\s*\w*\s*$').search(return_type):
567            return
568
569        if not self.entry.sections.get("Return", None):
570            self.emit_msg(ln,
571                          f"No description found for return value of '{declaration_name}'")
572
573    def split_struct_proto(self, proto):
574        """
575        Split apart a structure prototype; returns (struct|union, name,
576        members) or ``None``.
577        """
578
579        type_pattern = r'(struct|union)'
580        qualifiers = [
581            "__attribute__",
582            "__packed",
583            "__aligned",
584            "____cacheline_aligned_in_smp",
585            "____cacheline_aligned",
586        ]
587        definition_body = r'\{(.*)\}\s*' + "(?:" + '|'.join(qualifiers) + ")?"
588
589        r = KernRe(type_pattern + r'\s+(\w+)\s*' + definition_body)
590        if r.search(proto):
591            return (r.group(1), r.group(2), r.group(3))
592        else:
593            r = KernRe(r'typedef\s+' + type_pattern + r'\s*' + definition_body + r'\s*(\w+)\s*;')
594            if r.search(proto):
595                return (r.group(1), r.group(3), r.group(2))
596        return None
597
598    def rewrite_struct_members(self, members):
599        """
600        Process ``struct``/``union`` members from the most deeply nested
601        outward.
602
603        Rewrite the members of a ``struct`` or ``union`` for easier formatting
604        later on. Among other things, this function will turn a member like::
605
606          struct { inner_members; } foo;
607
608        into::
609
610          struct foo; inner_members;
611        """
612
613        #
614        # The trick is in the ``^{`` below - it prevents a match of an outer
615        # ``struct``/``union`` until the inner one has been munged
616        # (removing the ``{`` in the process).
617        #
618        struct_members = KernRe(r'(struct|union)'   # 0: declaration type
619                                r'([^\{\};]+)' 	    # 1: possible name
620                                r'(\{)'
621                                r'([^\{\}]*)'       # 3: Contents of declaration
622                                r'(\})'
623                                r'([^\{\};]*)(;)')  # 5: Remaining stuff after declaration
624        tuples = struct_members.findall(members)
625        while tuples:
626            for t in tuples:
627                newmember = ""
628                oldmember = "".join(t) # Reconstruct the original formatting
629                dtype, name, lbr, content, rbr, rest, semi = t
630                #
631                # Pass through each field name, normalizing the form and formatting.
632                #
633                for s_id in rest.split(','):
634                    s_id = s_id.strip()
635                    newmember += f"{dtype} {s_id}; "
636                    #
637                    # Remove bitfield/array/pointer info, getting the bare name.
638                    #
639                    s_id = KernRe(r'[:\[].*').sub('', s_id)
640                    s_id = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', s_id)
641                    #
642                    # Pass through the members of this inner structure/union.
643                    #
644                    for arg in content.split(';'):
645                        arg = arg.strip()
646                        #
647                        # Look for (type)(*name)(args) - pointer to function
648                        #
649                        r = KernRe(r'^([^\(]+\(\*?\s*)([\w.]*)(\s*\).*)')
650                        if r.match(arg):
651                            dtype, name, extra = r.group(1), r.group(2), r.group(3)
652                            # Pointer-to-function
653                            if not s_id:
654                                # Anonymous struct/union
655                                newmember += f"{dtype}{name}{extra}; "
656                            else:
657                                newmember += f"{dtype}{s_id}.{name}{extra}; "
658                        #
659                        # Otherwise a non-function member.
660                        #
661                        else:
662                            #
663                            # Remove bitmap and array portions and spaces around commas
664                            #
665                            arg = KernRe(r':\s*\d+\s*').sub('', arg)
666                            arg = KernRe(r'\[.*\]').sub('', arg)
667                            arg = KernRe(r'\s*,\s*').sub(',', arg)
668                            #
669                            # Look for a normal decl - "type name[,name...]"
670                            #
671                            r = KernRe(r'(.*)\s+([\S+,]+)')
672                            if r.search(arg):
673                                for name in r.group(2).split(','):
674                                    name = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', name)
675                                    if not s_id:
676                                        # Anonymous struct/union
677                                        newmember += f"{r.group(1)} {name}; "
678                                    else:
679                                        newmember += f"{r.group(1)} {s_id}.{name}; "
680                            else:
681                                newmember += f"{arg}; "
682                #
683                # At the end of the s_id loop, replace the original declaration with
684                # the munged version.
685                #
686                members = members.replace(oldmember, newmember)
687            #
688            # End of the tuple loop - search again and see if there are outer members
689            # that now turn up.
690            #
691            tuples = struct_members.findall(members)
692        return members
693
694    def format_struct_decl(self, declaration):
695        """
696        Format the ``struct`` declaration into a standard form for inclusion
697        in the resulting docs.
698        """
699
700        #
701        # Insert newlines, get rid of extra spaces.
702        #
703        declaration = KernRe(r'([\{;])').sub(r'\1\n', declaration)
704        declaration = KernRe(r'\}\s+;').sub('};', declaration)
705        #
706        # Format inline enums with each member on its own line.
707        #
708        r = KernRe(r'(enum\s+\{[^\}]+),([^\n])')
709        while r.search(declaration):
710            declaration = r.sub(r'\1,\n\2', declaration)
711        #
712        # Now go through and supply the right number of tabs
713        # for each line.
714        #
715        def_args = declaration.split('\n')
716        level = 1
717        declaration = ""
718        for clause in def_args:
719            clause = KernRe(r'\s+').sub(' ', clause.strip(), count=1)
720            if clause:
721                if '}' in clause and level > 1:
722                    level -= 1
723                if not clause.startswith('#'):
724                    declaration += "\t" * level
725                declaration += "\t" + clause + "\n"
726                if "{" in clause and "}" not in clause:
727                    level += 1
728        return declaration
729
730
731    def dump_struct(self, ln, proto):
732        """
733        Store an entry for a ``struct`` or ``union``
734        """
735        #
736        # Do the basic parse to get the pieces of the declaration.
737        #
738        struct_parts = self.split_struct_proto(proto)
739        if not struct_parts:
740            self.emit_msg(ln, f"{proto} error: Cannot parse struct or union!")
741            return
742        decl_type, declaration_name, members = struct_parts
743
744        if self.entry.identifier != declaration_name:
745            self.emit_msg(ln, f"expecting prototype for {decl_type} {self.entry.identifier}. "
746                          f"Prototype was for {decl_type} {declaration_name} instead\n")
747            return
748        #
749        # Go through the list of members applying all of our transformations.
750        #
751        members = trim_private_members(members)
752        members = self.xforms.apply("struct", members)
753
754        nested = NestedMatch()
755        for search, sub in struct_nested_prefixes:
756            members = nested.sub(search, sub, members)
757        #
758        # Deal with embedded struct and union members, and drop enums entirely.
759        #
760        declaration = members
761        members = self.rewrite_struct_members(members)
762        members = re.sub(r'(\{[^\{\}]*\})', '', members)
763        #
764        # Output the result and we are done.
765        #
766        self.create_parameter_list(ln, decl_type, members, ';',
767                                   declaration_name)
768        self.check_sections(ln, declaration_name, decl_type)
769        self.output_declaration(decl_type, declaration_name,
770                                definition=self.format_struct_decl(declaration),
771                                purpose=self.entry.declaration_purpose)
772
773    def dump_enum(self, ln, proto):
774        """
775        Store an ``enum`` inside self.entries array.
776        """
777        #
778        # Strip preprocessor directives.  Note that this depends on the
779        # trailing semicolon we added in process_proto_type().
780        #
781        proto = KernRe(r'#\s*((define|ifdef|if)\s+|endif)[^;]*;', flags=re.S).sub('', proto)
782        #
783        # Parse out the name and members of the enum.  Typedef form first.
784        #
785        r = KernRe(r'typedef\s+enum\s*\{(.*)\}\s*(\w*)\s*;')
786        if r.search(proto):
787            declaration_name = r.group(2)
788            members = trim_private_members(r.group(1))
789        #
790        # Failing that, look for a straight enum
791        #
792        else:
793            r = KernRe(r'enum\s+(\w*)\s*\{(.*)\}')
794            if r.match(proto):
795                declaration_name = r.group(1)
796                members = trim_private_members(r.group(2))
797        #
798        # OK, this isn't going to work.
799        #
800            else:
801                self.emit_msg(ln, f"{proto}: error: Cannot parse enum!")
802                return
803        #
804        # Make sure we found what we were expecting.
805        #
806        if self.entry.identifier != declaration_name:
807            if self.entry.identifier == "":
808                self.emit_msg(ln,
809                              f"{proto}: wrong kernel-doc identifier on prototype")
810            else:
811                self.emit_msg(ln,
812                              f"expecting prototype for enum {self.entry.identifier}. "
813                              f"Prototype was for enum {declaration_name} instead")
814            return
815
816        if not declaration_name:
817            declaration_name = "(anonymous)"
818        #
819        # Parse out the name of each enum member, and verify that we
820        # have a description for it.
821        #
822        member_set = set()
823        members = KernRe(r'\([^;)]*\)').sub('', members)
824        for arg in members.split(','):
825            if not arg:
826                continue
827            arg = KernRe(r'^\s*(\w+).*').sub(r'\1', arg)
828            self.entry.parameterlist.append(arg)
829            if arg not in self.entry.parameterdescs:
830                self.entry.parameterdescs[arg] = self.undescribed
831                self.emit_msg(ln,
832                              f"Enum value '{arg}' not described in enum '{declaration_name}'")
833            member_set.add(arg)
834        #
835        # Ensure that every described member actually exists in the enum.
836        #
837        for k in self.entry.parameterdescs:
838            if k not in member_set:
839                self.emit_msg(ln,
840                              f"Excess enum value '@{k}' description in '{declaration_name}'")
841
842        self.output_declaration('enum', declaration_name,
843                                purpose=self.entry.declaration_purpose)
844
845    def dump_var(self, ln, proto):
846        """
847        Store variables that are part of kAPI.
848        """
849        VAR_ATTRIBS = [
850            "extern",
851            "const",
852        ]
853        OPTIONAL_VAR_ATTR = r"^(?:\b(?:" +"|".join(VAR_ATTRIBS) +r")\b\s*)*"
854
855        #
856        # Store the full prototype before modifying it
857        #
858        full_proto = proto
859        declaration_name = None
860
861        #
862        # Handle macro definitions
863        #
864        macro_prefixes = [
865            KernRe(r"DEFINE_[\w_]+\s*\(([\w_]+)\)"),
866        ]
867
868        for r in macro_prefixes:
869            match = r.search(proto)
870            if match:
871                declaration_name = match.group(1)
872                break
873
874        #
875        # Drop comments and macros to have a pure C prototype
876        #
877        if not declaration_name:
878            proto = self.xforms.apply("var", proto)
879
880        proto = proto.rstrip()
881
882        #
883        # Variable name is at the end of the declaration
884        #
885
886        default_val = None
887
888        r= KernRe(OPTIONAL_VAR_ATTR + r"\s*[\w_\s]*\s+(?:\*+)?([\w_]+)\s*[\d\]\[]*\s*(=.*)?")
889        if r.match(proto):
890            if not declaration_name:
891                declaration_name = r.group(1)
892
893            default_val = r.group(2)
894        else:
895            r= KernRe(OPTIONAL_VAR_ATTR + r"(?:[\w_\s]*)?\s+(?:\*+)?(?:[\w_]+)\s*[\d\]\[]*\s*(=.*)?")
896
897            if r.match(proto):
898                default_val = r.group(1)
899        if not declaration_name:
900           self.emit_msg(ln,f"{proto}: can't parse variable")
901           return
902
903        if default_val:
904            default_val = default_val.lstrip("=").strip()
905
906        self.output_declaration("var", declaration_name,
907                                full_proto=full_proto,
908                                default_val=default_val,
909                                purpose=self.entry.declaration_purpose)
910
911    def dump_declaration(self, ln, prototype):
912        """
913        Store a data declaration inside self.entries array.
914        """
915
916        if self.entry.decl_type == "enum":
917            self.dump_enum(ln, prototype)
918        elif self.entry.decl_type == "typedef":
919            self.dump_typedef(ln, prototype)
920        elif self.entry.decl_type in ["union", "struct"]:
921            self.dump_struct(ln, prototype)
922        elif self.entry.decl_type == "var":
923            self.dump_var(ln, prototype)
924        else:
925            # This would be a bug
926            self.emit_message(ln, f'Unknown declaration type: {self.entry.decl_type}')
927
928    def dump_function(self, ln, prototype):
929        """
930        Store a function or function macro inside self.entries array.
931        """
932
933        found = func_macro = False
934        return_type = ''
935        decl_type = 'function'
936
937        #
938        # If we have a macro, remove the "#define" at the front.
939        #
940        new_proto = KernRe(r"^#\s*define\s+").sub("", prototype)
941        if new_proto != prototype:
942            prototype = new_proto
943            #
944            # Dispense with the simple "#define A B" case here; the key
945            # is the space after the name of the symbol being defined.
946            # NOTE that the seemingly misnamed "func_macro" indicates a
947            # macro *without* arguments.
948            #
949            r = KernRe(r'^(\w+)\s+')
950            if r.search(prototype):
951                return_type = ''
952                declaration_name = r.group(1)
953                func_macro = True
954                found = True
955        else:
956            #
957            # Apply the initial transformations.
958            #
959            prototype = self.xforms.apply("func", prototype)
960
961        # Yes, this truly is vile.  We are looking for:
962        # 1. Return type (may be nothing if we're looking at a macro)
963        # 2. Function name
964        # 3. Function parameters.
965        #
966        # All the while we have to watch out for function pointer parameters
967        # (which IIRC is what the two sections are for), C types (these
968        # regexps don't even start to express all the possibilities), and
969        # so on.
970        #
971        # If you mess with these regexps, it's a good idea to check that
972        # the following functions' documentation still comes out right:
973        # - parport_register_device (function pointer parameters)
974        # - atomic_set (macro)
975        # - pci_match_device, __copy_to_user (long return type)
976
977        name = r'\w+'
978        type1 = r'(?:[\w\s]+)?'
979        type2 = r'(?:[\w\s]+\*+)+'
980        #
981        # Attempt to match first on (args) with no internal parentheses; this
982        # lets us easily filter out __acquires() and other post-args stuff.  If
983        # that fails, just grab the rest of the line to the last closing
984        # parenthesis.
985        #
986        proto_args = r'\(([^\(]*|.*)\)'
987        #
988        # (Except for the simple macro case) attempt to split up the prototype
989        # in the various ways we understand.
990        #
991        if not found:
992            patterns = [
993                rf'^()({name})\s*{proto_args}',
994                rf'^({type1})\s+({name})\s*{proto_args}',
995                rf'^({type2})\s*({name})\s*{proto_args}',
996            ]
997
998            for p in patterns:
999                r = KernRe(p)
1000                if r.match(prototype):
1001                    return_type = r.group(1)
1002                    declaration_name = r.group(2)
1003                    args = r.group(3)
1004                    self.create_parameter_list(ln, decl_type, args, ',',
1005                                               declaration_name)
1006                    found = True
1007                    break
1008        #
1009        # Parsing done; make sure that things are as we expect.
1010        #
1011        if not found:
1012            self.emit_msg(ln,
1013                          f"cannot understand function prototype: '{prototype}'")
1014            return
1015        if self.entry.identifier != declaration_name:
1016            self.emit_msg(ln, f"expecting prototype for {self.entry.identifier}(). "
1017                          f"Prototype was for {declaration_name}() instead")
1018            return
1019        self.check_sections(ln, declaration_name, "function")
1020        self.check_return_section(ln, declaration_name, return_type)
1021        #
1022        # Store the result.
1023        #
1024        self.output_declaration(decl_type, declaration_name,
1025                                typedef=('typedef' in return_type),
1026                                functiontype=return_type,
1027                                purpose=self.entry.declaration_purpose,
1028                                func_macro=func_macro)
1029
1030
1031    def dump_typedef(self, ln, proto):
1032        """
1033        Store a ``typedef`` inside self.entries array.
1034        """
1035        #
1036        # We start by looking for function typedefs.
1037        #
1038        typedef_type = r'typedef((?:\s+[\w*]+\b){0,7}\s+(?:\w+\b|\*+))\s*'
1039        typedef_ident = r'\*?\s*(\w\S+)\s*'
1040        typedef_args = r'\s*\((.*)\);'
1041
1042        typedef1 = KernRe(typedef_type + r'\(' + typedef_ident + r'\)' + typedef_args)
1043        typedef2 = KernRe(typedef_type + typedef_ident + typedef_args)
1044
1045        # Parse function typedef prototypes
1046        for r in [typedef1, typedef2]:
1047            if not r.match(proto):
1048                continue
1049
1050            return_type = r.group(1).strip()
1051            declaration_name = r.group(2)
1052            args = r.group(3)
1053
1054            if self.entry.identifier != declaration_name:
1055                self.emit_msg(ln,
1056                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1057                return
1058
1059            self.create_parameter_list(ln, 'function', args, ',', declaration_name)
1060
1061            self.output_declaration('function', declaration_name,
1062                                    typedef=True,
1063                                    functiontype=return_type,
1064                                    purpose=self.entry.declaration_purpose)
1065            return
1066        #
1067        # Not a function, try to parse a simple typedef.
1068        #
1069        r = KernRe(r'typedef.*\s+(\w+)\s*;')
1070        if r.match(proto):
1071            declaration_name = r.group(1)
1072
1073            if self.entry.identifier != declaration_name:
1074                self.emit_msg(ln,
1075                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1076                return
1077
1078            self.output_declaration('typedef', declaration_name,
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):
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):
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):
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):
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):
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):
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        # else ... ??
1371
1372    def process_inline_text(self, ln, line):
1373        """STATE_INLINE_TEXT: docbook comments within a prototype."""
1374
1375        if doc_inline_end.search(line):
1376            self.dump_section()
1377            self.state = state.PROTO
1378        elif doc_content.search(line):
1379            self.entry.add_text(doc_content.group(1))
1380        # else ... ??
1381
1382    def syscall_munge(self, ln, proto):         # pylint: disable=W0613
1383        """
1384        Handle syscall definitions.
1385        """
1386
1387        is_void = False
1388
1389        # Strip newlines/CR's
1390        proto = re.sub(r'[\r\n]+', ' ', proto)
1391
1392        # Check if it's a SYSCALL_DEFINE0
1393        if 'SYSCALL_DEFINE0' in proto:
1394            is_void = True
1395
1396        # Replace SYSCALL_DEFINE with correct return type & function name
1397        proto = KernRe(r'SYSCALL_DEFINE.*\(').sub('long sys_', proto)
1398
1399        r = KernRe(r'long\s+(sys_.*?),')
1400        if r.search(proto):
1401            proto = KernRe(',').sub('(', proto, count=1)
1402        elif is_void:
1403            proto = KernRe(r'\)').sub('(void)', proto, count=1)
1404
1405        # Now delete all of the odd-numbered commas in the proto
1406        # so that argument types & names don't have a comma between them
1407        count = 0
1408        length = len(proto)
1409
1410        if is_void:
1411            length = 0  # skip the loop if is_void
1412
1413        for ix in range(length):
1414            if proto[ix] == ',':
1415                count += 1
1416                if count % 2 == 1:
1417                    proto = proto[:ix] + ' ' + proto[ix + 1:]
1418
1419        return proto
1420
1421    def tracepoint_munge(self, ln, proto):
1422        """
1423        Handle tracepoint definitions.
1424        """
1425
1426        tracepointname = None
1427        tracepointargs = None
1428
1429        # Match tracepoint name based on different patterns
1430        r = KernRe(r'TRACE_EVENT\((.*?),')
1431        if r.search(proto):
1432            tracepointname = r.group(1)
1433
1434        r = KernRe(r'DEFINE_SINGLE_EVENT\((.*?),')
1435        if r.search(proto):
1436            tracepointname = r.group(1)
1437
1438        r = KernRe(r'DEFINE_EVENT\((.*?),(.*?),')
1439        if r.search(proto):
1440            tracepointname = r.group(2)
1441
1442        if tracepointname:
1443            tracepointname = tracepointname.lstrip()
1444
1445        r = KernRe(r'TP_PROTO\((.*?)\)')
1446        if r.search(proto):
1447            tracepointargs = r.group(1)
1448
1449        if not tracepointname or not tracepointargs:
1450            self.emit_msg(ln,
1451                          f"Unrecognized tracepoint format:\n{proto}\n")
1452        else:
1453            proto = f"static inline void trace_{tracepointname}({tracepointargs})"
1454            self.entry.identifier = f"trace_{self.entry.identifier}"
1455
1456        return proto
1457
1458    def process_proto_function(self, ln, line):
1459        """Ancillary routine to process a function prototype."""
1460
1461        # strip C99-style comments to end of line
1462        line = KernRe(r"//.*$", re.S).sub('', line)
1463        #
1464        # Soak up the line's worth of prototype text, stopping at { or ; if present.
1465        #
1466        if KernRe(r'\s*#\s*define').match(line):
1467            self.entry.prototype = line
1468        elif not line.startswith('#'):   # skip other preprocessor stuff
1469            r = KernRe(r'([^\{]*)')
1470            if r.match(line):
1471                self.entry.prototype += r.group(1) + " "
1472        #
1473        # If we now have the whole prototype, clean it up and declare victory.
1474        #
1475        if '{' in line or ';' in line or KernRe(r'\s*#\s*define').match(line):
1476            # strip comments and surrounding spaces
1477            self.entry.prototype = KernRe(r'/\*.*\*/').sub('', self.entry.prototype).strip()
1478            #
1479            # Handle self.entry.prototypes for function pointers like:
1480            #       int (*pcs_config)(struct foo)
1481            # by turning it into
1482            #	    int pcs_config(struct foo)
1483            #
1484            r = KernRe(r'^(\S+\s+)\(\s*\*(\S+)\)')
1485            self.entry.prototype = r.sub(r'\1\2', self.entry.prototype)
1486            #
1487            # Handle special declaration syntaxes
1488            #
1489            if 'SYSCALL_DEFINE' in self.entry.prototype:
1490                self.entry.prototype = self.syscall_munge(ln,
1491                                                          self.entry.prototype)
1492            else:
1493                r = KernRe(r'TRACE_EVENT|DEFINE_EVENT|DEFINE_SINGLE_EVENT')
1494                if r.search(self.entry.prototype):
1495                    self.entry.prototype = self.tracepoint_munge(ln,
1496                                                                 self.entry.prototype)
1497            #
1498            # ... and we're done
1499            #
1500            self.dump_function(ln, self.entry.prototype)
1501            self.reset_state(ln)
1502
1503    def process_proto_type(self, ln, line):
1504        """
1505        Ancillary routine to process a type.
1506        """
1507
1508        # Strip C99-style comments and surrounding whitespace
1509        line = KernRe(r"//.*$", re.S).sub('', line).strip()
1510        if not line:
1511            return # nothing to see here
1512
1513        # To distinguish preprocessor directive from regular declaration later.
1514        if line.startswith('#'):
1515            line += ";"
1516        #
1517        # Split the declaration on any of { } or ;, and accumulate pieces
1518        # until we hit a semicolon while not inside {brackets}
1519        #
1520        r = KernRe(r'(.*?)([{};])')
1521        for chunk in r.split(line):
1522            if chunk:  # Ignore empty matches
1523                self.entry.prototype += chunk
1524                #
1525                # This cries out for a match statement ... someday after we can
1526                # drop Python 3.9 ...
1527                #
1528                if chunk == '{':
1529                    self.entry.brcount += 1
1530                elif chunk == '}':
1531                    self.entry.brcount -= 1
1532                elif chunk == ';' and self.entry.brcount <= 0:
1533                    self.dump_declaration(ln, self.entry.prototype)
1534                    self.reset_state(ln)
1535                    return
1536        #
1537        # We hit the end of the line while still in the declaration; put
1538        # in a space to represent the newline.
1539        #
1540        self.entry.prototype += ' '
1541
1542    def process_proto(self, ln, line):
1543        """STATE_PROTO: reading a function/whatever prototype."""
1544
1545        if doc_inline_oneline.search(line):
1546            self.entry.begin_section(ln, doc_inline_oneline.group(1))
1547            self.entry.add_text(doc_inline_oneline.group(2))
1548            self.dump_section()
1549
1550        elif doc_inline_start.search(line):
1551            self.state = state.INLINE_NAME
1552
1553        elif self.entry.decl_type == 'function':
1554            self.process_proto_function(ln, line)
1555
1556        else:
1557            self.process_proto_type(ln, line)
1558
1559    def process_docblock(self, ln, line):
1560        """STATE_DOCBLOCK: within a ``DOC:`` block."""
1561
1562        if doc_end.search(line):
1563            self.dump_section()
1564            self.output_declaration("doc", self.entry.identifier)
1565            self.reset_state(ln)
1566
1567        elif doc_content.search(line):
1568            self.entry.add_text(doc_content.group(1))
1569
1570    def parse_export(self):
1571        """
1572        Parses ``EXPORT_SYMBOL*`` macros from a single Kernel source file.
1573        """
1574
1575        export_table = set()
1576
1577        try:
1578            with open(self.fname, "r", encoding="utf8",
1579                      errors="backslashreplace") as fp:
1580
1581                for line in fp:
1582                    self.process_export(export_table, line)
1583
1584        except IOError:
1585            return None
1586
1587        return export_table
1588
1589    #: The state/action table telling us which function to invoke in each state.
1590    state_actions = {
1591        state.NORMAL:			process_normal,
1592        state.NAME:			process_name,
1593        state.BODY:			process_body,
1594        state.DECLARATION:		process_decl,
1595        state.SPECIAL_SECTION:		process_special,
1596        state.INLINE_NAME:		process_inline_name,
1597        state.INLINE_TEXT:		process_inline_text,
1598        state.PROTO:			process_proto,
1599        state.DOCBLOCK:			process_docblock,
1600        }
1601
1602    def parse_kdoc(self):
1603        """
1604        Open and process each line of a C source file.
1605        The parsing is controlled via a state machine, and the line is passed
1606        to a different process function depending on the state. The process
1607        function may update the state as needed.
1608
1609        Besides parsing kernel-doc tags, it also parses export symbols.
1610        """
1611
1612        prev = ""
1613        prev_ln = None
1614        export_table = set()
1615
1616        try:
1617            with open(self.fname, "r", encoding="utf8",
1618                      errors="backslashreplace") as fp:
1619                for ln, line in enumerate(fp):
1620
1621                    line = line.expandtabs().strip("\n")
1622
1623                    # Group continuation lines on prototypes
1624                    if self.state == state.PROTO:
1625                        if line.endswith("\\"):
1626                            prev += line.rstrip("\\")
1627                            if not prev_ln:
1628                                prev_ln = ln
1629                            continue
1630
1631                        if prev:
1632                            ln = prev_ln
1633                            line = prev + line
1634                            prev = ""
1635                            prev_ln = None
1636
1637                    self.config.log.debug("%d %s: %s",
1638                                          ln, state.name[self.state],
1639                                          line)
1640
1641                    # This is an optimization over the original script.
1642                    # There, when export_file was used for the same file,
1643                    # it was read twice. Here, we use the already-existing
1644                    # loop to parse exported symbols as well.
1645                    #
1646                    if (self.state != state.NORMAL) or \
1647                       not self.process_export(export_table, line):
1648                        # Hand this line to the appropriate state handler
1649                        self.state_actions[self.state](self, ln, line)
1650
1651            self.emit_unused_warnings()
1652
1653        except OSError:
1654            self.config.log.error(f"Error: Cannot open file {self.fname}")
1655
1656        return export_table, self.entries
1657