xref: /linux/scripts/lib/kdoc/kdoc_parser.py (revision d1af2889682e83acc791e2a2191687958b548da1)
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"""
8kdoc_parser
9===========
10
11Read a C language source or header FILE and extract embedded
12documentation comments
13"""
14
15import re
16from pprint import pformat
17
18from kdoc_re import NestedMatch, KernRe
19
20
21#
22# Regular expressions used to parse kernel-doc markups at KernelDoc class.
23#
24# Let's declare them in lowercase outside any class to make easier to
25# convert from the python script.
26#
27# As those are evaluated at the beginning, no need to cache them
28#
29
30# Allow whitespace at end of comment start.
31doc_start = KernRe(r'^/\*\*\s*$', cache=False)
32
33doc_end = KernRe(r'\*/', cache=False)
34doc_com = KernRe(r'\s*\*\s*', cache=False)
35doc_com_body = KernRe(r'\s*\* ?', cache=False)
36doc_decl = doc_com + KernRe(r'(\w+)', cache=False)
37
38# @params and a strictly limited set of supported section names
39# Specifically:
40#   Match @word:
41#         @...:
42#         @{section-name}:
43# while trying to not match literal block starts like "example::"
44#
45doc_sect = doc_com + \
46            KernRe(r'\s*(\@[.\w]+|\@\.\.\.|description|context|returns?|notes?|examples?)\s*:([^:].*)?$',
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*(@[\w\s]+):\s*(.*)\s*\*/\s*$', cache=False)
54attribute = KernRe(r"__attribute__\s*\(\([a-z0-9,_\*\s\(\)]*\)\)",
55               flags=re.I | re.S, cache=False)
56
57export_symbol = KernRe(r'^\s*EXPORT_SYMBOL(_GPL)?\s*\(\s*(\w+)\s*\)\s*', cache=False)
58export_symbol_ns = KernRe(r'^\s*EXPORT_SYMBOL_NS(_GPL)?\s*\(\s*(\w+)\s*,\s*"\S+"\)\s*', cache=False)
59
60type_param = KernRe(r"\@(\w*((\.\w+)|(->\w+))*(\.\.\.)?)", cache=False)
61
62#
63# Tests for the beginning of a kerneldoc block in its various forms.
64#
65doc_block = doc_com + KernRe(r'DOC:\s*(.*)?', cache=False)
66doc_begin_data = KernRe(r"^\s*\*?\s*(struct|union|enum|typedef)\b\s*(\w*)", cache = False)
67doc_begin_func = KernRe(str(doc_com) +			# initial " * '
68                        r"(?:\w+\s*\*\s*)?" + 		# type (not captured)
69                        r'(?:define\s+)?' + 		# possible "define" (not captured)
70                        r'(\w+)\s*(?:\(\w*\))?\s*' +	# name and optional "(...)"
71                        r'(?:[-:].*)?$',		# description (not captured)
72                        cache = False)
73
74#
75# A little helper to get rid of excess white space
76#
77multi_space = KernRe(r'\s\s+')
78def trim_whitespace(s):
79    return multi_space.sub(' ', s.strip())
80
81class state:
82    """
83    State machine enums
84    """
85
86    # Parser states
87    NORMAL        = 0        # normal code
88    NAME          = 1        # looking for function name
89    DECLARATION   = 2        # We have seen a declaration which might not be done
90    BODY          = 3        # the body of the comment
91    SPECIAL_SECTION = 4      # doc section ending with a blank line
92    PROTO         = 5        # scanning prototype
93    DOCBLOCK      = 6        # documentation block
94    INLINE_NAME   = 7        # gathering doc outside main block
95    INLINE_TEXT   = 8	     # reading the body of inline docs
96
97    name = [
98        "NORMAL",
99        "NAME",
100        "DECLARATION",
101        "BODY",
102        "SPECIAL_SECTION",
103        "PROTO",
104        "DOCBLOCK",
105        "INLINE_NAME",
106        "INLINE_TEXT",
107    ]
108
109
110SECTION_DEFAULT = "Description"  # default section
111
112class KernelEntry:
113
114    def __init__(self, config, ln):
115        self.config = config
116
117        self._contents = []
118        self.sectcheck = ""
119        self.struct_actual = ""
120        self.prototype = ""
121
122        self.warnings = []
123
124        self.parameterlist = []
125        self.parameterdescs = {}
126        self.parametertypes = {}
127        self.parameterdesc_start_lines = {}
128
129        self.section_start_lines = {}
130        self.sectionlist = []
131        self.sections = {}
132
133        self.anon_struct_union = False
134
135        self.leading_space = None
136
137        # State flags
138        self.brcount = 0
139        self.declaration_start_line = ln + 1
140
141    #
142    # Management of section contents
143    #
144    def add_text(self, text):
145        self._contents.append(text)
146
147    def contents(self):
148        return '\n'.join(self._contents) + '\n'
149
150    # TODO: rename to emit_message after removal of kernel-doc.pl
151    def emit_msg(self, log_msg, warning=True):
152        """Emit a message"""
153
154        if not warning:
155            self.config.log.info(log_msg)
156            return
157
158        # Delegate warning output to output logic, as this way it
159        # will report warnings/info only for symbols that are output
160
161        self.warnings.append(log_msg)
162        return
163
164    #
165    # Begin a new section.
166    #
167    def begin_section(self, line_no, title = SECTION_DEFAULT, dump = False):
168        if dump:
169            self.dump_section(start_new = True)
170        self.section = title
171        self.new_start_line = line_no
172
173    def dump_section(self, start_new=True):
174        """
175        Dumps section contents to arrays/hashes intended for that purpose.
176        """
177        #
178        # If we have accumulated no contents in the default ("description")
179        # section, don't bother.
180        #
181        if self.section == SECTION_DEFAULT and not self._contents:
182            return
183        name = self.section
184        contents = self.contents()
185
186        if type_param.match(name):
187            name = type_param.group(1)
188
189            self.parameterdescs[name] = contents
190            self.parameterdesc_start_lines[name] = self.new_start_line
191
192            self.sectcheck += name + " "
193            self.new_start_line = 0
194
195        else:
196            if name in self.sections and self.sections[name] != "":
197                # Only warn on user-specified duplicate section names
198                if name != SECTION_DEFAULT:
199                    self.emit_msg(self.new_start_line,
200                                  f"duplicate section name '{name}'\n")
201                # Treat as a new paragraph - add a blank line
202                self.sections[name] += '\n' + contents
203            else:
204                self.sections[name] = contents
205                self.sectionlist.append(name)
206                self.section_start_lines[name] = self.new_start_line
207                self.new_start_line = 0
208
209#        self.config.log.debug("Section: %s : %s", name, pformat(vars(self)))
210
211        if start_new:
212            self.section = SECTION_DEFAULT
213            self._contents = []
214
215
216class KernelDoc:
217    """
218    Read a C language source or header FILE and extract embedded
219    documentation comments.
220    """
221
222    # Section names
223
224    section_context = "Context"
225    section_return = "Return"
226
227    undescribed = "-- undescribed --"
228
229    def __init__(self, config, fname):
230        """Initialize internal variables"""
231
232        self.fname = fname
233        self.config = config
234
235        # Initial state for the state machines
236        self.state = state.NORMAL
237
238        # Store entry currently being processed
239        self.entry = None
240
241        # Place all potential outputs into an array
242        self.entries = []
243
244    def emit_msg(self, ln, msg, warning=True):
245        """Emit a message"""
246
247        log_msg = f"{self.fname}:{ln} {msg}"
248
249        if self.entry:
250            self.entry.emit_msg(log_msg, warning)
251            return
252
253        if warning:
254            self.config.log.warning(log_msg)
255        else:
256            self.config.log.info(log_msg)
257
258    def dump_section(self, start_new=True):
259        """
260        Dumps section contents to arrays/hashes intended for that purpose.
261        """
262
263        if self.entry:
264            self.entry.dump_section(start_new)
265
266    # TODO: rename it to store_declaration after removal of kernel-doc.pl
267    def output_declaration(self, dtype, name, **args):
268        """
269        Stores the entry into an entry array.
270
271        The actual output and output filters will be handled elsewhere
272        """
273
274        # The implementation here is different than the original kernel-doc:
275        # instead of checking for output filters or actually output anything,
276        # it just stores the declaration content at self.entries, as the
277        # output will happen on a separate class.
278        #
279        # For now, we're keeping the same name of the function just to make
280        # easier to compare the source code of both scripts
281
282        args["declaration_start_line"] = self.entry.declaration_start_line
283        args["type"] = dtype
284        args["warnings"] = self.entry.warnings
285
286        # TODO: use colletions.OrderedDict to remove sectionlist
287
288        sections = args.get('sections', {})
289        sectionlist = args.get('sectionlist', [])
290
291        # Drop empty sections
292        # TODO: improve empty sections logic to emit warnings
293        for section in ["Description", "Return"]:
294            if section in sectionlist:
295                if not sections[section].rstrip():
296                    del sections[section]
297                    sectionlist.remove(section)
298
299        self.entries.append((name, args))
300
301        self.config.log.debug("Output: %s:%s = %s", dtype, name, pformat(args))
302
303    def reset_state(self, ln):
304        """
305        Ancillary routine to create a new entry. It initializes all
306        variables used by the state machine.
307        """
308
309        self.entry = KernelEntry(self.config, ln)
310
311        # State flags
312        self.state = state.NORMAL
313
314    def push_parameter(self, ln, decl_type, param, dtype,
315                       org_arg, declaration_name):
316        """
317        Store parameters and their descriptions at self.entry.
318        """
319
320        if self.entry.anon_struct_union and dtype == "" and param == "}":
321            return  # Ignore the ending }; from anonymous struct/union
322
323        self.entry.anon_struct_union = False
324
325        param = KernRe(r'[\[\)].*').sub('', param, count=1)
326
327        if dtype == "" and param.endswith("..."):
328            if KernRe(r'\w\.\.\.$').search(param):
329                # For named variable parameters of the form `x...`,
330                # remove the dots
331                param = param[:-3]
332            else:
333                # Handles unnamed variable parameters
334                param = "..."
335
336            if param not in self.entry.parameterdescs or \
337                not self.entry.parameterdescs[param]:
338
339                self.entry.parameterdescs[param] = "variable arguments"
340
341        elif dtype == "" and (not param or param == "void"):
342            param = "void"
343            self.entry.parameterdescs[param] = "no arguments"
344
345        elif dtype == "" and param in ["struct", "union"]:
346            # Handle unnamed (anonymous) union or struct
347            dtype = param
348            param = "{unnamed_" + param + "}"
349            self.entry.parameterdescs[param] = "anonymous\n"
350            self.entry.anon_struct_union = True
351
352        # Handle cache group enforcing variables: they do not need
353        # to be described in header files
354        elif "__cacheline_group" in param:
355            # Ignore __cacheline_group_begin and __cacheline_group_end
356            return
357
358        # Warn if parameter has no description
359        # (but ignore ones starting with # as these are not parameters
360        # but inline preprocessor statements)
361        if param not in self.entry.parameterdescs and not param.startswith("#"):
362            self.entry.parameterdescs[param] = self.undescribed
363
364            if "." not in param:
365                if decl_type == 'function':
366                    dname = f"{decl_type} parameter"
367                else:
368                    dname = f"{decl_type} member"
369
370                self.emit_msg(ln,
371                              f"{dname} '{param}' not described in '{declaration_name}'")
372
373        # Strip spaces from param so that it is one continuous string on
374        # parameterlist. This fixes a problem where check_sections()
375        # cannot find a parameter like "addr[6 + 2]" because it actually
376        # appears as "addr[6", "+", "2]" on the parameter list.
377        # However, it's better to maintain the param string unchanged for
378        # output, so just weaken the string compare in check_sections()
379        # to ignore "[blah" in a parameter string.
380
381        self.entry.parameterlist.append(param)
382        org_arg = KernRe(r'\s\s+').sub(' ', org_arg)
383        self.entry.parametertypes[param] = org_arg
384
385    def save_struct_actual(self, actual):
386        """
387        Strip all spaces from the actual param so that it looks like
388        one string item.
389        """
390
391        actual = KernRe(r'\s*').sub("", actual, count=1)
392
393        self.entry.struct_actual += actual + " "
394
395    def create_parameter_list(self, ln, decl_type, args,
396                              splitter, declaration_name):
397        """
398        Creates a list of parameters, storing them at self.entry.
399        """
400
401        # temporarily replace all commas inside function pointer definition
402        arg_expr = KernRe(r'(\([^\),]+),')
403        while arg_expr.search(args):
404            args = arg_expr.sub(r"\1#", args)
405
406        for arg in args.split(splitter):
407            # Strip comments
408            arg = KernRe(r'\/\*.*\*\/').sub('', arg)
409
410            # Ignore argument attributes
411            arg = KernRe(r'\sPOS0?\s').sub(' ', arg)
412
413            # Strip leading/trailing spaces
414            arg = arg.strip()
415            arg = KernRe(r'\s+').sub(' ', arg, count=1)
416
417            if arg.startswith('#'):
418                # Treat preprocessor directive as a typeless variable just to fill
419                # corresponding data structures "correctly". Catch it later in
420                # output_* subs.
421
422                # Treat preprocessor directive as a typeless variable
423                self.push_parameter(ln, decl_type, arg, "",
424                                    "", declaration_name)
425
426            elif KernRe(r'\(.+\)\s*\(').search(arg):
427                # Pointer-to-function
428
429                arg = arg.replace('#', ',')
430
431                r = KernRe(r'[^\(]+\(\*?\s*([\w\[\]\.]*)\s*\)')
432                if r.match(arg):
433                    param = r.group(1)
434                else:
435                    self.emit_msg(ln, f"Invalid param: {arg}")
436                    param = arg
437
438                dtype = KernRe(r'([^\(]+\(\*?)\s*' + re.escape(param)).sub(r'\1', arg)
439                self.save_struct_actual(param)
440                self.push_parameter(ln, decl_type, param, dtype,
441                                    arg, declaration_name)
442
443            elif KernRe(r'\(.+\)\s*\[').search(arg):
444                # Array-of-pointers
445
446                arg = arg.replace('#', ',')
447                r = KernRe(r'[^\(]+\(\s*\*\s*([\w\[\]\.]*?)\s*(\s*\[\s*[\w]+\s*\]\s*)*\)')
448                if r.match(arg):
449                    param = r.group(1)
450                else:
451                    self.emit_msg(ln, f"Invalid param: {arg}")
452                    param = arg
453
454                dtype = KernRe(r'([^\(]+\(\*?)\s*' + re.escape(param)).sub(r'\1', arg)
455
456                self.save_struct_actual(param)
457                self.push_parameter(ln, decl_type, param, dtype,
458                                    arg, declaration_name)
459
460            elif arg:
461                arg = KernRe(r'\s*:\s*').sub(":", arg)
462                arg = KernRe(r'\s*\[').sub('[', arg)
463
464                args = KernRe(r'\s*,\s*').split(arg)
465                if args[0] and '*' in args[0]:
466                    args[0] = re.sub(r'(\*+)\s*', r' \1', args[0])
467
468                first_arg = []
469                r = KernRe(r'^(.*\s+)(.*?\[.*\].*)$')
470                if args[0] and r.match(args[0]):
471                    args.pop(0)
472                    first_arg.extend(r.group(1))
473                    first_arg.append(r.group(2))
474                else:
475                    first_arg = KernRe(r'\s+').split(args.pop(0))
476
477                args.insert(0, first_arg.pop())
478                dtype = ' '.join(first_arg)
479
480                for param in args:
481                    if KernRe(r'^(\*+)\s*(.*)').match(param):
482                        r = KernRe(r'^(\*+)\s*(.*)')
483                        if not r.match(param):
484                            self.emit_msg(ln, f"Invalid param: {param}")
485                            continue
486
487                        param = r.group(1)
488
489                        self.save_struct_actual(r.group(2))
490                        self.push_parameter(ln, decl_type, r.group(2),
491                                            f"{dtype} {r.group(1)}",
492                                            arg, declaration_name)
493
494                    elif KernRe(r'(.*?):(\w+)').search(param):
495                        r = KernRe(r'(.*?):(\w+)')
496                        if not r.match(param):
497                            self.emit_msg(ln, f"Invalid param: {param}")
498                            continue
499
500                        if dtype != "":  # Skip unnamed bit-fields
501                            self.save_struct_actual(r.group(1))
502                            self.push_parameter(ln, decl_type, r.group(1),
503                                                f"{dtype}:{r.group(2)}",
504                                                arg, declaration_name)
505                    else:
506                        self.save_struct_actual(param)
507                        self.push_parameter(ln, decl_type, param, dtype,
508                                            arg, declaration_name)
509
510    def check_sections(self, ln, decl_name, decl_type, sectcheck, prmscheck):
511        """
512        Check for errors inside sections, emitting warnings if not found
513        parameters are described.
514        """
515
516        sects = sectcheck.split()
517        prms = prmscheck.split()
518        err = False
519
520        for sx in range(len(sects)):                  # pylint: disable=C0200
521            err = True
522            for px in range(len(prms)):               # pylint: disable=C0200
523                prm_clean = prms[px]
524                prm_clean = KernRe(r'\[.*\]').sub('', prm_clean)
525                prm_clean = attribute.sub('', prm_clean)
526
527                # ignore array size in a parameter string;
528                # however, the original param string may contain
529                # spaces, e.g.:  addr[6 + 2]
530                # and this appears in @prms as "addr[6" since the
531                # parameter list is split at spaces;
532                # hence just ignore "[..." for the sections check;
533                prm_clean = KernRe(r'\[.*').sub('', prm_clean)
534
535                if prm_clean == sects[sx]:
536                    err = False
537                    break
538
539            if err:
540                if decl_type == 'function':
541                    dname = f"{decl_type} parameter"
542                else:
543                    dname = f"{decl_type} member"
544
545                self.emit_msg(ln,
546                              f"Excess {dname} '{sects[sx]}' description in '{decl_name}'")
547
548    def check_return_section(self, ln, declaration_name, return_type):
549        """
550        If the function doesn't return void, warns about the lack of a
551        return description.
552        """
553
554        if not self.config.wreturn:
555            return
556
557        # Ignore an empty return type (It's a macro)
558        # Ignore functions with a "void" return type (but not "void *")
559        if not return_type or KernRe(r'void\s*\w*\s*$').search(return_type):
560            return
561
562        if not self.entry.sections.get("Return", None):
563            self.emit_msg(ln,
564                          f"No description found for return value of '{declaration_name}'")
565
566    def dump_struct(self, ln, proto):
567        """
568        Store an entry for an struct or union
569        """
570
571        type_pattern = r'(struct|union)'
572
573        qualifiers = [
574            "__attribute__",
575            "__packed",
576            "__aligned",
577            "____cacheline_aligned_in_smp",
578            "____cacheline_aligned",
579        ]
580
581        definition_body = r'\{(.*)\}\s*' + "(?:" + '|'.join(qualifiers) + ")?"
582        struct_members = KernRe(type_pattern + r'([^\{\};]+)(\{)([^\{\}]*)(\})([^\{\}\;]*)(\;)')
583
584        # Extract struct/union definition
585        members = None
586        declaration_name = None
587        decl_type = None
588
589        r = KernRe(type_pattern + r'\s+(\w+)\s*' + definition_body)
590        if r.search(proto):
591            decl_type = r.group(1)
592            declaration_name = r.group(2)
593            members = r.group(3)
594        else:
595            r = KernRe(r'typedef\s+' + type_pattern + r'\s*' + definition_body + r'\s*(\w+)\s*;')
596
597            if r.search(proto):
598                decl_type = r.group(1)
599                declaration_name = r.group(3)
600                members = r.group(2)
601
602        if not members:
603            self.emit_msg(ln, f"{proto} error: Cannot parse struct or union!")
604            return
605
606        if self.entry.identifier != declaration_name:
607            self.emit_msg(ln,
608                          f"expecting prototype for {decl_type} {self.entry.identifier}. Prototype was for {decl_type} {declaration_name} instead\n")
609            return
610
611        args_pattern = r'([^,)]+)'
612
613        sub_prefixes = [
614            (KernRe(r'\/\*\s*private:.*?\/\*\s*public:.*?\*\/', re.S | re.I), ''),
615            (KernRe(r'\/\*\s*private:.*', re.S | re.I), ''),
616
617            # Strip comments
618            (KernRe(r'\/\*.*?\*\/', re.S), ''),
619
620            # Strip attributes
621            (attribute, ' '),
622            (KernRe(r'\s*__aligned\s*\([^;]*\)', re.S), ' '),
623            (KernRe(r'\s*__counted_by\s*\([^;]*\)', re.S), ' '),
624            (KernRe(r'\s*__counted_by_(le|be)\s*\([^;]*\)', re.S), ' '),
625            (KernRe(r'\s*__packed\s*', re.S), ' '),
626            (KernRe(r'\s*CRYPTO_MINALIGN_ATTR', re.S), ' '),
627            (KernRe(r'\s*____cacheline_aligned_in_smp', re.S), ' '),
628            (KernRe(r'\s*____cacheline_aligned', re.S), ' '),
629
630            # Unwrap struct_group macros based on this definition:
631            # __struct_group(TAG, NAME, ATTRS, MEMBERS...)
632            # which has variants like: struct_group(NAME, MEMBERS...)
633            # Only MEMBERS arguments require documentation.
634            #
635            # Parsing them happens on two steps:
636            #
637            # 1. drop struct group arguments that aren't at MEMBERS,
638            #    storing them as STRUCT_GROUP(MEMBERS)
639            #
640            # 2. remove STRUCT_GROUP() ancillary macro.
641            #
642            # The original logic used to remove STRUCT_GROUP() using an
643            # advanced regex:
644            #
645            #   \bSTRUCT_GROUP(\(((?:(?>[^)(]+)|(?1))*)\))[^;]*;
646            #
647            # with two patterns that are incompatible with
648            # Python re module, as it has:
649            #
650            #   - a recursive pattern: (?1)
651            #   - an atomic grouping: (?>...)
652            #
653            # I tried a simpler version: but it didn't work either:
654            #   \bSTRUCT_GROUP\(([^\)]+)\)[^;]*;
655            #
656            # As it doesn't properly match the end parenthesis on some cases.
657            #
658            # So, a better solution was crafted: there's now a NestedMatch
659            # class that ensures that delimiters after a search are properly
660            # matched. So, the implementation to drop STRUCT_GROUP() will be
661            # handled in separate.
662
663            (KernRe(r'\bstruct_group\s*\(([^,]*,)', re.S), r'STRUCT_GROUP('),
664            (KernRe(r'\bstruct_group_attr\s*\(([^,]*,){2}', re.S), r'STRUCT_GROUP('),
665            (KernRe(r'\bstruct_group_tagged\s*\(([^,]*),([^,]*),', re.S), r'struct \1 \2; STRUCT_GROUP('),
666            (KernRe(r'\b__struct_group\s*\(([^,]*,){3}', re.S), r'STRUCT_GROUP('),
667
668            # Replace macros
669            #
670            # TODO: use NestedMatch for FOO($1, $2, ...) matches
671            #
672            # it is better to also move those to the NestedMatch logic,
673            # to ensure that parenthesis will be properly matched.
674
675            (KernRe(r'__ETHTOOL_DECLARE_LINK_MODE_MASK\s*\(([^\)]+)\)', re.S), r'DECLARE_BITMAP(\1, __ETHTOOL_LINK_MODE_MASK_NBITS)'),
676            (KernRe(r'DECLARE_PHY_INTERFACE_MASK\s*\(([^\)]+)\)', re.S), r'DECLARE_BITMAP(\1, PHY_INTERFACE_MODE_MAX)'),
677            (KernRe(r'DECLARE_BITMAP\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'unsigned long \1[BITS_TO_LONGS(\2)]'),
678            (KernRe(r'DECLARE_HASHTABLE\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'unsigned long \1[1 << ((\2) - 1)]'),
679            (KernRe(r'DECLARE_KFIFO\s*\(' + args_pattern + r',\s*' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\2 *\1'),
680            (KernRe(r'DECLARE_KFIFO_PTR\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\2 *\1'),
681            (KernRe(r'(?:__)?DECLARE_FLEX_ARRAY\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\1 \2[]'),
682            (KernRe(r'DEFINE_DMA_UNMAP_ADDR\s*\(' + args_pattern + r'\)', re.S), r'dma_addr_t \1'),
683            (KernRe(r'DEFINE_DMA_UNMAP_LEN\s*\(' + args_pattern + r'\)', re.S), r'__u32 \1'),
684        ]
685
686        # Regexes here are guaranteed to have the end limiter matching
687        # the start delimiter. Yet, right now, only one replace group
688        # is allowed.
689
690        sub_nested_prefixes = [
691            (re.compile(r'\bSTRUCT_GROUP\('), r'\1'),
692        ]
693
694        for search, sub in sub_prefixes:
695            members = search.sub(sub, members)
696
697        nested = NestedMatch()
698
699        for search, sub in sub_nested_prefixes:
700            members = nested.sub(search, sub, members)
701
702        # Keeps the original declaration as-is
703        declaration = members
704
705        # Split nested struct/union elements
706        #
707        # This loop was simpler at the original kernel-doc perl version, as
708        #   while ($members =~ m/$struct_members/) { ... }
709        # reads 'members' string on each interaction.
710        #
711        # Python behavior is different: it parses 'members' only once,
712        # creating a list of tuples from the first interaction.
713        #
714        # On other words, this won't get nested structs.
715        #
716        # So, we need to have an extra loop on Python to override such
717        # re limitation.
718
719        while True:
720            tuples = struct_members.findall(members)
721            if not tuples:
722                break
723
724            for t in tuples:
725                newmember = ""
726                maintype = t[0]
727                s_ids = t[5]
728                content = t[3]
729
730                oldmember = "".join(t)
731
732                for s_id in s_ids.split(','):
733                    s_id = s_id.strip()
734
735                    newmember += f"{maintype} {s_id}; "
736                    s_id = KernRe(r'[:\[].*').sub('', s_id)
737                    s_id = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', s_id)
738
739                    for arg in content.split(';'):
740                        arg = arg.strip()
741
742                        if not arg:
743                            continue
744
745                        r = KernRe(r'^([^\(]+\(\*?\s*)([\w\.]*)(\s*\).*)')
746                        if r.match(arg):
747                            # Pointer-to-function
748                            dtype = r.group(1)
749                            name = r.group(2)
750                            extra = r.group(3)
751
752                            if not name:
753                                continue
754
755                            if not s_id:
756                                # Anonymous struct/union
757                                newmember += f"{dtype}{name}{extra}; "
758                            else:
759                                newmember += f"{dtype}{s_id}.{name}{extra}; "
760
761                        else:
762                            arg = arg.strip()
763                            # Handle bitmaps
764                            arg = KernRe(r':\s*\d+\s*').sub('', arg)
765
766                            # Handle arrays
767                            arg = KernRe(r'\[.*\]').sub('', arg)
768
769                            # Handle multiple IDs
770                            arg = KernRe(r'\s*,\s*').sub(',', arg)
771
772                            r = KernRe(r'(.*)\s+([\S+,]+)')
773
774                            if r.search(arg):
775                                dtype = r.group(1)
776                                names = r.group(2)
777                            else:
778                                newmember += f"{arg}; "
779                                continue
780
781                            for name in names.split(','):
782                                name = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', name).strip()
783
784                                if not name:
785                                    continue
786
787                                if not s_id:
788                                    # Anonymous struct/union
789                                    newmember += f"{dtype} {name}; "
790                                else:
791                                    newmember += f"{dtype} {s_id}.{name}; "
792
793                members = members.replace(oldmember, newmember)
794
795        # Ignore other nested elements, like enums
796        members = re.sub(r'(\{[^\{\}]*\})', '', members)
797
798        self.create_parameter_list(ln, decl_type, members, ';',
799                                   declaration_name)
800        self.check_sections(ln, declaration_name, decl_type,
801                            self.entry.sectcheck, self.entry.struct_actual)
802
803        # Adjust declaration for better display
804        declaration = KernRe(r'([\{;])').sub(r'\1\n', declaration)
805        declaration = KernRe(r'\}\s+;').sub('};', declaration)
806
807        # Better handle inlined enums
808        while True:
809            r = KernRe(r'(enum\s+\{[^\}]+),([^\n])')
810            if not r.search(declaration):
811                break
812
813            declaration = r.sub(r'\1,\n\2', declaration)
814
815        def_args = declaration.split('\n')
816        level = 1
817        declaration = ""
818        for clause in def_args:
819
820            clause = clause.strip()
821            clause = KernRe(r'\s+').sub(' ', clause, count=1)
822
823            if not clause:
824                continue
825
826            if '}' in clause and level > 1:
827                level -= 1
828
829            if not KernRe(r'^\s*#').match(clause):
830                declaration += "\t" * level
831
832            declaration += "\t" + clause + "\n"
833            if "{" in clause and "}" not in clause:
834                level += 1
835
836        self.output_declaration(decl_type, declaration_name,
837                                struct=declaration_name,
838                                definition=declaration,
839                                parameterlist=self.entry.parameterlist,
840                                parameterdescs=self.entry.parameterdescs,
841                                parametertypes=self.entry.parametertypes,
842                                parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
843                                sectionlist=self.entry.sectionlist,
844                                sections=self.entry.sections,
845                                section_start_lines=self.entry.section_start_lines,
846                                purpose=self.entry.declaration_purpose)
847
848    def dump_enum(self, ln, proto):
849        """
850        Stores an enum inside self.entries array.
851        """
852
853        # Ignore members marked private
854        proto = KernRe(r'\/\*\s*private:.*?\/\*\s*public:.*?\*\/', flags=re.S).sub('', proto)
855        proto = KernRe(r'\/\*\s*private:.*}', flags=re.S).sub('}', proto)
856
857        # Strip comments
858        proto = KernRe(r'\/\*.*?\*\/', flags=re.S).sub('', proto)
859
860        # Strip #define macros inside enums
861        proto = KernRe(r'#\s*((define|ifdef|if)\s+|endif)[^;]*;', flags=re.S).sub('', proto)
862
863        #
864        # Parse out the name and members of the enum.  Typedef form first.
865        #
866        r = KernRe(r'typedef\s+enum\s*\{(.*)\}\s*(\w*)\s*;')
867        if r.search(proto):
868            declaration_name = r.group(2)
869            members = r.group(1).rstrip()
870        #
871        # Failing that, look for a straight enum
872        #
873        else:
874            r = KernRe(r'enum\s+(\w*)\s*\{(.*)\}')
875            if r.match(proto):
876                declaration_name = r.group(1)
877                members = r.group(2).rstrip()
878        #
879        # OK, this isn't going to work.
880        #
881            else:
882                self.emit_msg(ln, f"{proto}: error: Cannot parse enum!")
883                return
884        #
885        # Make sure we found what we were expecting.
886        #
887        if self.entry.identifier != declaration_name:
888            if self.entry.identifier == "":
889                self.emit_msg(ln,
890                              f"{proto}: wrong kernel-doc identifier on prototype")
891            else:
892                self.emit_msg(ln,
893                              f"expecting prototype for enum {self.entry.identifier}. "
894                              f"Prototype was for enum {declaration_name} instead")
895            return
896
897        if not declaration_name:
898            declaration_name = "(anonymous)"
899        #
900        # Parse out the name of each enum member, and verify that we
901        # have a description for it.
902        #
903        member_set = set()
904        members = KernRe(r'\([^;)]*\)').sub('', members)
905        for arg in members.split(','):
906            if not arg:
907                continue
908            arg = KernRe(r'^\s*(\w+).*').sub(r'\1', arg)
909            self.entry.parameterlist.append(arg)
910            if arg not in self.entry.parameterdescs:
911                self.entry.parameterdescs[arg] = self.undescribed
912                self.emit_msg(ln,
913                              f"Enum value '{arg}' not described in enum '{declaration_name}'")
914            member_set.add(arg)
915        #
916        # Ensure that every described member actually exists in the enum.
917        #
918        for k in self.entry.parameterdescs:
919            if k not in member_set:
920                self.emit_msg(ln,
921                              f"Excess enum value '%{k}' description in '{declaration_name}'")
922
923        self.output_declaration('enum', declaration_name,
924                                enum=declaration_name,
925                                parameterlist=self.entry.parameterlist,
926                                parameterdescs=self.entry.parameterdescs,
927                                parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
928                                sectionlist=self.entry.sectionlist,
929                                sections=self.entry.sections,
930                                section_start_lines=self.entry.section_start_lines,
931                                purpose=self.entry.declaration_purpose)
932
933    def dump_declaration(self, ln, prototype):
934        """
935        Stores a data declaration inside self.entries array.
936        """
937
938        if self.entry.decl_type == "enum":
939            self.dump_enum(ln, prototype)
940            return
941
942        if self.entry.decl_type == "typedef":
943            self.dump_typedef(ln, prototype)
944            return
945
946        if self.entry.decl_type in ["union", "struct"]:
947            self.dump_struct(ln, prototype)
948            return
949
950        self.output_declaration(self.entry.decl_type, prototype,
951                                entry=self.entry)
952
953    def dump_function(self, ln, prototype):
954        """
955        Stores a function of function macro inside self.entries array.
956        """
957
958        func_macro = False
959        return_type = ''
960        decl_type = 'function'
961
962        # Prefixes that would be removed
963        sub_prefixes = [
964            (r"^static +", "", 0),
965            (r"^extern +", "", 0),
966            (r"^asmlinkage +", "", 0),
967            (r"^inline +", "", 0),
968            (r"^__inline__ +", "", 0),
969            (r"^__inline +", "", 0),
970            (r"^__always_inline +", "", 0),
971            (r"^noinline +", "", 0),
972            (r"^__FORTIFY_INLINE +", "", 0),
973            (r"__init +", "", 0),
974            (r"__init_or_module +", "", 0),
975            (r"__deprecated +", "", 0),
976            (r"__flatten +", "", 0),
977            (r"__meminit +", "", 0),
978            (r"__must_check +", "", 0),
979            (r"__weak +", "", 0),
980            (r"__sched +", "", 0),
981            (r"_noprof", "", 0),
982            (r"__printf\s*\(\s*\d*\s*,\s*\d*\s*\) +", "", 0),
983            (r"__(?:re)?alloc_size\s*\(\s*\d+\s*(?:,\s*\d+\s*)?\) +", "", 0),
984            (r"__diagnose_as\s*\(\s*\S+\s*(?:,\s*\d+\s*)*\) +", "", 0),
985            (r"DECL_BUCKET_PARAMS\s*\(\s*(\S+)\s*,\s*(\S+)\s*\)", r"\1, \2", 0),
986            (r"__attribute_const__ +", "", 0),
987
988            # It seems that Python support for re.X is broken:
989            # At least for me (Python 3.13), this didn't work
990#            (r"""
991#              __attribute__\s*\(\(
992#                (?:
993#                    [\w\s]+          # attribute name
994#                    (?:\([^)]*\))?   # attribute arguments
995#                    \s*,?            # optional comma at the end
996#                )+
997#              \)\)\s+
998#             """, "", re.X),
999
1000            # So, remove whitespaces and comments from it
1001            (r"__attribute__\s*\(\((?:[\w\s]+(?:\([^)]*\))?\s*,?)+\)\)\s+", "", 0),
1002        ]
1003
1004        for search, sub, flags in sub_prefixes:
1005            prototype = KernRe(search, flags).sub(sub, prototype)
1006
1007        # Macros are a special case, as they change the prototype format
1008        new_proto = KernRe(r"^#\s*define\s+").sub("", prototype)
1009        if new_proto != prototype:
1010            is_define_proto = True
1011            prototype = new_proto
1012        else:
1013            is_define_proto = False
1014
1015        # Yes, this truly is vile.  We are looking for:
1016        # 1. Return type (may be nothing if we're looking at a macro)
1017        # 2. Function name
1018        # 3. Function parameters.
1019        #
1020        # All the while we have to watch out for function pointer parameters
1021        # (which IIRC is what the two sections are for), C types (these
1022        # regexps don't even start to express all the possibilities), and
1023        # so on.
1024        #
1025        # If you mess with these regexps, it's a good idea to check that
1026        # the following functions' documentation still comes out right:
1027        # - parport_register_device (function pointer parameters)
1028        # - atomic_set (macro)
1029        # - pci_match_device, __copy_to_user (long return type)
1030
1031        name = r'[a-zA-Z0-9_~:]+'
1032        prototype_end1 = r'[^\(]*'
1033        prototype_end2 = r'[^\{]*'
1034        prototype_end = fr'\(({prototype_end1}|{prototype_end2})\)'
1035
1036        # Besides compiling, Perl qr{[\w\s]+} works as a non-capturing group.
1037        # So, this needs to be mapped in Python with (?:...)? or (?:...)+
1038
1039        type1 = r'(?:[\w\s]+)?'
1040        type2 = r'(?:[\w\s]+\*+)+'
1041
1042        found = False
1043
1044        if is_define_proto:
1045            r = KernRe(r'^()(' + name + r')\s+')
1046
1047            if r.search(prototype):
1048                return_type = ''
1049                declaration_name = r.group(2)
1050                func_macro = True
1051
1052                found = True
1053
1054        if not found:
1055            patterns = [
1056                rf'^()({name})\s*{prototype_end}',
1057                rf'^({type1})\s+({name})\s*{prototype_end}',
1058                rf'^({type2})\s*({name})\s*{prototype_end}',
1059            ]
1060
1061            for p in patterns:
1062                r = KernRe(p)
1063
1064                if r.match(prototype):
1065
1066                    return_type = r.group(1)
1067                    declaration_name = r.group(2)
1068                    args = r.group(3)
1069
1070                    self.create_parameter_list(ln, decl_type, args, ',',
1071                                               declaration_name)
1072
1073                    found = True
1074                    break
1075        if not found:
1076            self.emit_msg(ln,
1077                          f"cannot understand function prototype: '{prototype}'")
1078            return
1079
1080        if self.entry.identifier != declaration_name:
1081            self.emit_msg(ln,
1082                          f"expecting prototype for {self.entry.identifier}(). Prototype was for {declaration_name}() instead")
1083            return
1084
1085        prms = " ".join(self.entry.parameterlist)
1086        self.check_sections(ln, declaration_name, "function",
1087                            self.entry.sectcheck, prms)
1088
1089        self.check_return_section(ln, declaration_name, return_type)
1090
1091        if 'typedef' in return_type:
1092            self.output_declaration(decl_type, declaration_name,
1093                                    function=declaration_name,
1094                                    typedef=True,
1095                                    functiontype=return_type,
1096                                    parameterlist=self.entry.parameterlist,
1097                                    parameterdescs=self.entry.parameterdescs,
1098                                    parametertypes=self.entry.parametertypes,
1099                                    parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
1100                                    sectionlist=self.entry.sectionlist,
1101                                    sections=self.entry.sections,
1102                                    section_start_lines=self.entry.section_start_lines,
1103                                    purpose=self.entry.declaration_purpose,
1104                                    func_macro=func_macro)
1105        else:
1106            self.output_declaration(decl_type, declaration_name,
1107                                    function=declaration_name,
1108                                    typedef=False,
1109                                    functiontype=return_type,
1110                                    parameterlist=self.entry.parameterlist,
1111                                    parameterdescs=self.entry.parameterdescs,
1112                                    parametertypes=self.entry.parametertypes,
1113                                    parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
1114                                    sectionlist=self.entry.sectionlist,
1115                                    sections=self.entry.sections,
1116                                    section_start_lines=self.entry.section_start_lines,
1117                                    purpose=self.entry.declaration_purpose,
1118                                    func_macro=func_macro)
1119
1120    def dump_typedef(self, ln, proto):
1121        """
1122        Stores a typedef inside self.entries array.
1123        """
1124
1125        typedef_type = r'((?:\s+[\w\*]+\b){0,7}\s+(?:\w+\b|\*+))\s*'
1126        typedef_ident = r'\*?\s*(\w\S+)\s*'
1127        typedef_args = r'\s*\((.*)\);'
1128
1129        typedef1 = KernRe(r'typedef' + typedef_type + r'\(' + typedef_ident + r'\)' + typedef_args)
1130        typedef2 = KernRe(r'typedef' + typedef_type + typedef_ident + typedef_args)
1131
1132        # Strip comments
1133        proto = KernRe(r'/\*.*?\*/', flags=re.S).sub('', proto)
1134
1135        # Parse function typedef prototypes
1136        for r in [typedef1, typedef2]:
1137            if not r.match(proto):
1138                continue
1139
1140            return_type = r.group(1).strip()
1141            declaration_name = r.group(2)
1142            args = r.group(3)
1143
1144            if self.entry.identifier != declaration_name:
1145                self.emit_msg(ln,
1146                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1147                return
1148
1149            decl_type = 'function'
1150            self.create_parameter_list(ln, decl_type, args, ',', declaration_name)
1151
1152            self.output_declaration(decl_type, declaration_name,
1153                                    function=declaration_name,
1154                                    typedef=True,
1155                                    functiontype=return_type,
1156                                    parameterlist=self.entry.parameterlist,
1157                                    parameterdescs=self.entry.parameterdescs,
1158                                    parametertypes=self.entry.parametertypes,
1159                                    parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
1160                                    sectionlist=self.entry.sectionlist,
1161                                    sections=self.entry.sections,
1162                                    section_start_lines=self.entry.section_start_lines,
1163                                    purpose=self.entry.declaration_purpose)
1164            return
1165
1166        # Handle nested parentheses or brackets
1167        r = KernRe(r'(\(*.\)\s*|\[*.\]\s*);$')
1168        while r.search(proto):
1169            proto = r.sub('', proto)
1170
1171        # Parse simple typedefs
1172        r = KernRe(r'typedef.*\s+(\w+)\s*;')
1173        if r.match(proto):
1174            declaration_name = r.group(1)
1175
1176            if self.entry.identifier != declaration_name:
1177                self.emit_msg(ln,
1178                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1179                return
1180
1181            self.output_declaration('typedef', declaration_name,
1182                                    typedef=declaration_name,
1183                                    sectionlist=self.entry.sectionlist,
1184                                    sections=self.entry.sections,
1185                                    section_start_lines=self.entry.section_start_lines,
1186                                    purpose=self.entry.declaration_purpose)
1187            return
1188
1189        self.emit_msg(ln, "error: Cannot parse typedef!")
1190
1191    @staticmethod
1192    def process_export(function_set, line):
1193        """
1194        process EXPORT_SYMBOL* tags
1195
1196        This method doesn't use any variable from the class, so declare it
1197        with a staticmethod decorator.
1198        """
1199
1200        # We support documenting some exported symbols with different
1201        # names.  A horrible hack.
1202        suffixes = [ '_noprof' ]
1203
1204        # Note: it accepts only one EXPORT_SYMBOL* per line, as having
1205        # multiple export lines would violate Kernel coding style.
1206
1207        if export_symbol.search(line):
1208            symbol = export_symbol.group(2)
1209        elif export_symbol_ns.search(line):
1210            symbol = export_symbol_ns.group(2)
1211        else:
1212            return False
1213        #
1214        # Found an export, trim out any special suffixes
1215        #
1216        for suffix in suffixes:
1217            symbol = symbol.removesuffix(suffix)
1218        function_set.add(symbol)
1219        return True
1220
1221    def process_normal(self, ln, line):
1222        """
1223        STATE_NORMAL: looking for the /** to begin everything.
1224        """
1225
1226        if not doc_start.match(line):
1227            return
1228
1229        # start a new entry
1230        self.reset_state(ln)
1231
1232        # next line is always the function name
1233        self.state = state.NAME
1234
1235    def process_name(self, ln, line):
1236        """
1237        STATE_NAME: Looking for the "name - description" line
1238        """
1239        #
1240        # Check for a DOC: block and handle them specially.
1241        #
1242        if doc_block.search(line):
1243
1244            if not doc_block.group(1):
1245                self.entry.begin_section(ln, "Introduction")
1246            else:
1247                self.entry.begin_section(ln, doc_block.group(1))
1248
1249            self.entry.identifier = self.entry.section
1250            self.state = state.DOCBLOCK
1251        #
1252        # Otherwise we're looking for a normal kerneldoc declaration line.
1253        #
1254        elif doc_decl.search(line):
1255            self.entry.identifier = doc_decl.group(1)
1256
1257            # Test for data declaration
1258            if doc_begin_data.search(line):
1259                self.entry.decl_type = doc_begin_data.group(1)
1260                self.entry.identifier = doc_begin_data.group(2)
1261            #
1262            # Look for a function description
1263            #
1264            elif doc_begin_func.search(line):
1265                self.entry.identifier = doc_begin_func.group(1)
1266                self.entry.decl_type = "function"
1267            #
1268            # We struck out.
1269            #
1270            else:
1271                self.emit_msg(ln,
1272                              f"This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst\n{line}")
1273                self.state = state.NORMAL
1274                return
1275            #
1276            # OK, set up for a new kerneldoc entry.
1277            #
1278            self.state = state.BODY
1279            self.entry.identifier = self.entry.identifier.strip(" ")
1280            # if there's no @param blocks need to set up default section here
1281            self.entry.begin_section(ln + 1)
1282            #
1283            # Find the description portion, which *should* be there but
1284            # isn't always.
1285            # (We should be able to capture this from the previous parsing - someday)
1286            #
1287            r = KernRe("[-:](.*)")
1288            if r.search(line):
1289                self.entry.declaration_purpose = trim_whitespace(r.group(1))
1290                self.state = state.DECLARATION
1291            else:
1292                self.entry.declaration_purpose = ""
1293
1294            if not self.entry.declaration_purpose and self.config.wshort_desc:
1295                self.emit_msg(ln,
1296                              f"missing initial short description on line:\n{line}")
1297
1298            if not self.entry.identifier and self.entry.decl_type != "enum":
1299                self.emit_msg(ln,
1300                              f"wrong kernel-doc identifier on line:\n{line}")
1301                self.state = state.NORMAL
1302
1303            if self.config.verbose:
1304                self.emit_msg(ln,
1305                              f"Scanning doc for {self.entry.decl_type} {self.entry.identifier}",
1306                                  warning=False)
1307        #
1308        # Failed to find an identifier. Emit a warning
1309        #
1310        else:
1311            self.emit_msg(ln, f"Cannot find identifier on line:\n{line}")
1312
1313    #
1314    # Helper function to determine if a new section is being started.
1315    #
1316    def is_new_section(self, ln, line):
1317        if doc_sect.search(line):
1318            self.state = state.BODY
1319            #
1320            # Pick out the name of our new section, tweaking it if need be.
1321            #
1322            newsection = doc_sect.group(1)
1323            if newsection.lower() == 'description':
1324                newsection = 'Description'
1325            elif newsection.lower() == 'context':
1326                newsection = 'Context'
1327                self.state = state.SPECIAL_SECTION
1328            elif newsection.lower() in ["@return", "@returns",
1329                                        "return", "returns"]:
1330                newsection = "Return"
1331                self.state = state.SPECIAL_SECTION
1332            elif newsection[0] == '@':
1333                self.state = state.SPECIAL_SECTION
1334            #
1335            # Initialize the contents, and get the new section going.
1336            #
1337            newcontents = doc_sect.group(2)
1338            if not newcontents:
1339                newcontents = ""
1340            self.dump_section()
1341            self.entry.begin_section(ln, newsection)
1342            self.entry.leading_space = None
1343
1344            self.entry.add_text(newcontents.lstrip())
1345            return True
1346        return False
1347
1348    #
1349    # Helper function to detect (and effect) the end of a kerneldoc comment.
1350    #
1351    def is_comment_end(self, ln, line):
1352        if doc_end.search(line):
1353            self.dump_section()
1354
1355            # Look for doc_com + <text> + doc_end:
1356            r = KernRe(r'\s*\*\s*[a-zA-Z_0-9:\.]+\*/')
1357            if r.match(line):
1358                self.emit_msg(ln, f"suspicious ending line: {line}")
1359
1360            self.entry.prototype = ""
1361            self.entry.new_start_line = ln + 1
1362
1363            self.state = state.PROTO
1364            return True
1365        return False
1366
1367
1368    def process_decl(self, ln, line):
1369        """
1370        STATE_DECLARATION: We've seen the beginning of a declaration
1371        """
1372        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1373            return
1374        #
1375        # Look for anything with the " * " line beginning.
1376        #
1377        if doc_content.search(line):
1378            cont = doc_content.group(1)
1379            #
1380            # A blank line means that we have moved out of the declaration
1381            # part of the comment (without any "special section" parameter
1382            # descriptions).
1383            #
1384            if cont == "":
1385                self.state = state.BODY
1386            #
1387            # Otherwise we have more of the declaration section to soak up.
1388            #
1389            else:
1390                self.entry.declaration_purpose = \
1391                    trim_whitespace(self.entry.declaration_purpose + ' ' + cont)
1392        else:
1393            # Unknown line, ignore
1394            self.emit_msg(ln, f"bad line: {line}")
1395
1396
1397    def process_special(self, ln, line):
1398        """
1399        STATE_SPECIAL_SECTION: a section ending with a blank line
1400        """
1401        #
1402        # If we have hit a blank line (only the " * " marker), then this
1403        # section is done.
1404        #
1405        if KernRe(r"\s*\*\s*$").match(line):
1406            self.entry.begin_section(ln, dump = True)
1407            self.state = state.BODY
1408            return
1409        #
1410        # Not a blank line, look for the other ways to end the section.
1411        #
1412        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1413            return
1414        #
1415        # OK, we should have a continuation of the text for this section.
1416        #
1417        if doc_content.search(line):
1418            cont = doc_content.group(1)
1419            #
1420            # If the lines of text after the first in a special section have
1421            # leading white space, we need to trim it out or Sphinx will get
1422            # confused.  For the second line (the None case), see what we
1423            # find there and remember it.
1424            #
1425            if self.entry.leading_space is None:
1426                r = KernRe(r'^(\s+)')
1427                if r.match(cont):
1428                    self.entry.leading_space = len(r.group(1))
1429                else:
1430                    self.entry.leading_space = 0
1431            #
1432            # Otherwise, before trimming any leading chars, be *sure*
1433            # that they are white space.  We should maybe warn if this
1434            # isn't the case.
1435            #
1436            for i in range(0, self.entry.leading_space):
1437                if cont[i] != " ":
1438                    self.entry.leading_space = i
1439                    break
1440            #
1441            # Add the trimmed result to the section and we're done.
1442            #
1443            self.entry.add_text(cont[self.entry.leading_space:])
1444        else:
1445            # Unknown line, ignore
1446            self.emit_msg(ln, f"bad line: {line}")
1447
1448    def process_body(self, ln, line):
1449        """
1450        STATE_BODY: the bulk of a kerneldoc comment.
1451        """
1452        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1453            return
1454
1455        if doc_content.search(line):
1456            cont = doc_content.group(1)
1457            self.entry.add_text(cont)
1458        else:
1459            # Unknown line, ignore
1460            self.emit_msg(ln, f"bad line: {line}")
1461
1462    def process_inline_name(self, ln, line):
1463        """STATE_INLINE_NAME: beginning of docbook comments within a prototype."""
1464
1465        if doc_inline_sect.search(line):
1466            self.entry.begin_section(ln, doc_inline_sect.group(1))
1467            self.entry.add_text(doc_inline_sect.group(2).lstrip())
1468            self.state = state.INLINE_TEXT
1469        elif doc_inline_end.search(line):
1470            self.dump_section()
1471            self.state = state.PROTO
1472        elif doc_content.search(line):
1473            self.emit_msg(ln, f"Incorrect use of kernel-doc format: {line}")
1474            self.state = state.PROTO
1475        # else ... ??
1476
1477    def process_inline_text(self, ln, line):
1478        """STATE_INLINE_TEXT: docbook comments within a prototype."""
1479
1480        if doc_inline_end.search(line):
1481            self.dump_section()
1482            self.state = state.PROTO
1483        elif doc_content.search(line):
1484            self.entry.add_text(doc_content.group(1))
1485        # else ... ??
1486
1487    def syscall_munge(self, ln, proto):         # pylint: disable=W0613
1488        """
1489        Handle syscall definitions
1490        """
1491
1492        is_void = False
1493
1494        # Strip newlines/CR's
1495        proto = re.sub(r'[\r\n]+', ' ', proto)
1496
1497        # Check if it's a SYSCALL_DEFINE0
1498        if 'SYSCALL_DEFINE0' in proto:
1499            is_void = True
1500
1501        # Replace SYSCALL_DEFINE with correct return type & function name
1502        proto = KernRe(r'SYSCALL_DEFINE.*\(').sub('long sys_', proto)
1503
1504        r = KernRe(r'long\s+(sys_.*?),')
1505        if r.search(proto):
1506            proto = KernRe(',').sub('(', proto, count=1)
1507        elif is_void:
1508            proto = KernRe(r'\)').sub('(void)', proto, count=1)
1509
1510        # Now delete all of the odd-numbered commas in the proto
1511        # so that argument types & names don't have a comma between them
1512        count = 0
1513        length = len(proto)
1514
1515        if is_void:
1516            length = 0  # skip the loop if is_void
1517
1518        for ix in range(length):
1519            if proto[ix] == ',':
1520                count += 1
1521                if count % 2 == 1:
1522                    proto = proto[:ix] + ' ' + proto[ix + 1:]
1523
1524        return proto
1525
1526    def tracepoint_munge(self, ln, proto):
1527        """
1528        Handle tracepoint definitions
1529        """
1530
1531        tracepointname = None
1532        tracepointargs = None
1533
1534        # Match tracepoint name based on different patterns
1535        r = KernRe(r'TRACE_EVENT\((.*?),')
1536        if r.search(proto):
1537            tracepointname = r.group(1)
1538
1539        r = KernRe(r'DEFINE_SINGLE_EVENT\((.*?),')
1540        if r.search(proto):
1541            tracepointname = r.group(1)
1542
1543        r = KernRe(r'DEFINE_EVENT\((.*?),(.*?),')
1544        if r.search(proto):
1545            tracepointname = r.group(2)
1546
1547        if tracepointname:
1548            tracepointname = tracepointname.lstrip()
1549
1550        r = KernRe(r'TP_PROTO\((.*?)\)')
1551        if r.search(proto):
1552            tracepointargs = r.group(1)
1553
1554        if not tracepointname or not tracepointargs:
1555            self.emit_msg(ln,
1556                          f"Unrecognized tracepoint format:\n{proto}\n")
1557        else:
1558            proto = f"static inline void trace_{tracepointname}({tracepointargs})"
1559            self.entry.identifier = f"trace_{self.entry.identifier}"
1560
1561        return proto
1562
1563    def process_proto_function(self, ln, line):
1564        """Ancillary routine to process a function prototype"""
1565
1566        # strip C99-style comments to end of line
1567        line = KernRe(r"\/\/.*$", re.S).sub('', line)
1568        #
1569        # Soak up the line's worth of prototype text, stopping at { or ; if present.
1570        #
1571        if KernRe(r'\s*#\s*define').match(line):
1572            self.entry.prototype = line
1573        elif not line.startswith('#'):   # skip other preprocessor stuff
1574            r = KernRe(r'([^\{]*)')
1575            if r.match(line):
1576                self.entry.prototype += r.group(1) + " "
1577        #
1578        # If we now have the whole prototype, clean it up and declare victory.
1579        #
1580        if '{' in line or ';' in line or KernRe(r'\s*#\s*define').match(line):
1581            # strip comments and surrounding spaces
1582            self.entry.prototype = KernRe(r'/\*.*\*/').sub('', self.entry.prototype).strip()
1583            #
1584            # Handle self.entry.prototypes for function pointers like:
1585            #       int (*pcs_config)(struct foo)
1586            # by turning it into
1587            #	    int pcs_config(struct foo)
1588            #
1589            r = KernRe(r'^(\S+\s+)\(\s*\*(\S+)\)')
1590            self.entry.prototype = r.sub(r'\1\2', self.entry.prototype)
1591            #
1592            # Handle special declaration syntaxes
1593            #
1594            if 'SYSCALL_DEFINE' in self.entry.prototype:
1595                self.entry.prototype = self.syscall_munge(ln,
1596                                                          self.entry.prototype)
1597            else:
1598                r = KernRe(r'TRACE_EVENT|DEFINE_EVENT|DEFINE_SINGLE_EVENT')
1599                if r.search(self.entry.prototype):
1600                    self.entry.prototype = self.tracepoint_munge(ln,
1601                                                                 self.entry.prototype)
1602            #
1603            # ... and we're done
1604            #
1605            self.dump_function(ln, self.entry.prototype)
1606            self.reset_state(ln)
1607
1608    def process_proto_type(self, ln, line):
1609        """Ancillary routine to process a type"""
1610
1611        # Strip C99-style comments and surrounding whitespace
1612        line = KernRe(r"//.*$", re.S).sub('', line).strip()
1613        if not line:
1614            return # nothing to see here
1615
1616        # To distinguish preprocessor directive from regular declaration later.
1617        if line.startswith('#'):
1618            line += ";"
1619        #
1620        # Split the declaration on any of { } or ;, and accumulate pieces
1621        # until we hit a semicolon while not inside {brackets}
1622        #
1623        r = KernRe(r'(.*?)([{};])')
1624        for chunk in r.split(line):
1625            if chunk:  # Ignore empty matches
1626                self.entry.prototype += chunk
1627                #
1628                # This cries out for a match statement ... someday after we can
1629                # drop Python 3.9 ...
1630                #
1631                if chunk == '{':
1632                    self.entry.brcount += 1
1633                elif chunk == '}':
1634                    self.entry.brcount -= 1
1635                elif chunk == ';' and self.entry.brcount <= 0:
1636                    self.dump_declaration(ln, self.entry.prototype)
1637                    self.reset_state(ln)
1638                    return
1639        #
1640        # We hit the end of the line while still in the declaration; put
1641        # in a space to represent the newline.
1642        #
1643        self.entry.prototype += ' '
1644
1645    def process_proto(self, ln, line):
1646        """STATE_PROTO: reading a function/whatever prototype."""
1647
1648        if doc_inline_oneline.search(line):
1649            self.entry.begin_section(ln, doc_inline_oneline.group(1))
1650            self.entry.add_text(doc_inline_oneline.group(2))
1651            self.dump_section()
1652
1653        elif doc_inline_start.search(line):
1654            self.state = state.INLINE_NAME
1655
1656        elif self.entry.decl_type == 'function':
1657            self.process_proto_function(ln, line)
1658
1659        else:
1660            self.process_proto_type(ln, line)
1661
1662    def process_docblock(self, ln, line):
1663        """STATE_DOCBLOCK: within a DOC: block."""
1664
1665        if doc_end.search(line):
1666            self.dump_section()
1667            self.output_declaration("doc", self.entry.identifier,
1668                                    sectionlist=self.entry.sectionlist,
1669                                    sections=self.entry.sections,
1670                                    section_start_lines=self.entry.section_start_lines)
1671            self.reset_state(ln)
1672
1673        elif doc_content.search(line):
1674            self.entry.add_text(doc_content.group(1))
1675
1676    def parse_export(self):
1677        """
1678        Parses EXPORT_SYMBOL* macros from a single Kernel source file.
1679        """
1680
1681        export_table = set()
1682
1683        try:
1684            with open(self.fname, "r", encoding="utf8",
1685                      errors="backslashreplace") as fp:
1686
1687                for line in fp:
1688                    self.process_export(export_table, line)
1689
1690        except IOError:
1691            return None
1692
1693        return export_table
1694
1695    #
1696    # The state/action table telling us which function to invoke in
1697    # each state.
1698    #
1699    state_actions = {
1700        state.NORMAL:			process_normal,
1701        state.NAME:			process_name,
1702        state.BODY:			process_body,
1703        state.DECLARATION:		process_decl,
1704        state.SPECIAL_SECTION:		process_special,
1705        state.INLINE_NAME:		process_inline_name,
1706        state.INLINE_TEXT:		process_inline_text,
1707        state.PROTO:			process_proto,
1708        state.DOCBLOCK:			process_docblock,
1709        }
1710
1711    def parse_kdoc(self):
1712        """
1713        Open and process each line of a C source file.
1714        The parsing is controlled via a state machine, and the line is passed
1715        to a different process function depending on the state. The process
1716        function may update the state as needed.
1717
1718        Besides parsing kernel-doc tags, it also parses export symbols.
1719        """
1720
1721        prev = ""
1722        prev_ln = None
1723        export_table = set()
1724
1725        try:
1726            with open(self.fname, "r", encoding="utf8",
1727                      errors="backslashreplace") as fp:
1728                for ln, line in enumerate(fp):
1729
1730                    line = line.expandtabs().strip("\n")
1731
1732                    # Group continuation lines on prototypes
1733                    if self.state == state.PROTO:
1734                        if line.endswith("\\"):
1735                            prev += line.rstrip("\\")
1736                            if not prev_ln:
1737                                prev_ln = ln
1738                            continue
1739
1740                        if prev:
1741                            ln = prev_ln
1742                            line = prev + line
1743                            prev = ""
1744                            prev_ln = None
1745
1746                    self.config.log.debug("%d %s: %s",
1747                                          ln, state.name[self.state],
1748                                          line)
1749
1750                    # This is an optimization over the original script.
1751                    # There, when export_file was used for the same file,
1752                    # it was read twice. Here, we use the already-existing
1753                    # loop to parse exported symbols as well.
1754                    #
1755                    if (self.state != state.NORMAL) or \
1756                       not self.process_export(export_table, line):
1757                        # Hand this line to the appropriate state handler
1758                        self.state_actions[self.state](self, ln, line)
1759
1760        except OSError:
1761            self.config.log.error(f"Error: Cannot open file {self.fname}")
1762
1763        return export_table, self.entries
1764