xref: /linux/scripts/lib/kdoc/kdoc_parser.py (revision 823d6f956605cb2f009f75de138622fcd7e03817)
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'(?:[-:].*)?$',		# description (not captured)
72                        cache = False)
73
74#
75# A little helper to get rid of excess white space
76#
77multi_space = KernRe(r'\s\s+')
78def trim_whitespace(s):
79    return multi_space.sub(' ', s.strip())
80
81class state:
82    """
83    State machine enums
84    """
85
86    # Parser states
87    NORMAL        = 0        # normal code
88    NAME          = 1        # looking for function name
89    BODY_MAYBE    = 2        # body - or maybe more description
90    BODY          = 3        # the body of the comment
91    BODY_WITH_BLANK_LINE = 4 # the body which has a blank line
92    PROTO         = 5        # scanning prototype
93    DOCBLOCK      = 6        # documentation block
94    INLINE        = 7        # gathering doc outside main block
95
96    name = [
97        "NORMAL",
98        "NAME",
99        "BODY_MAYBE",
100        "BODY",
101        "BODY_WITH_BLANK_LINE",
102        "PROTO",
103        "DOCBLOCK",
104        "INLINE",
105    ]
106
107    # Inline documentation state
108    INLINE_NA     = 0 # not applicable ($state != INLINE)
109    INLINE_NAME   = 1 # looking for member name (@foo:)
110    INLINE_TEXT   = 2 # looking for member documentation
111    INLINE_END    = 3 # done
112    INLINE_ERROR  = 4 # error - Comment without header was found.
113                      # Spit a warning as it's not
114                      # proper kernel-doc and ignore the rest.
115
116    inline_name = [
117        "",
118        "_NAME",
119        "_TEXT",
120        "_END",
121        "_ERROR",
122    ]
123
124SECTION_DEFAULT = "Description"  # default section
125
126class KernelEntry:
127
128    def __init__(self, config, ln):
129        self.config = config
130
131        self.contents = ""
132        self.function = ""
133        self.sectcheck = ""
134        self.struct_actual = ""
135        self.prototype = ""
136
137        self.warnings = []
138
139        self.parameterlist = []
140        self.parameterdescs = {}
141        self.parametertypes = {}
142        self.parameterdesc_start_lines = {}
143
144        self.section_start_lines = {}
145        self.sectionlist = []
146        self.sections = {}
147
148        self.anon_struct_union = False
149
150        self.leading_space = None
151
152        # State flags
153        self.brcount = 0
154
155        self.in_doc_sect = False
156        self.declaration_start_line = ln + 1
157
158    # TODO: rename to emit_message after removal of kernel-doc.pl
159    def emit_msg(self, log_msg, warning=True):
160        """Emit a message"""
161
162        if not warning:
163            self.config.log.info(log_msg)
164            return
165
166        # Delegate warning output to output logic, as this way it
167        # will report warnings/info only for symbols that are output
168
169        self.warnings.append(log_msg)
170        return
171
172    def dump_section(self, start_new=True):
173        """
174        Dumps section contents to arrays/hashes intended for that purpose.
175        """
176
177        name = self.section
178        contents = self.contents
179
180        if type_param.match(name):
181            name = type_param.group(1)
182
183            self.parameterdescs[name] = contents
184            self.parameterdesc_start_lines[name] = self.new_start_line
185
186            self.sectcheck += name + " "
187            self.new_start_line = 0
188
189        elif name == "@...":
190            name = "..."
191            self.parameterdescs[name] = contents
192            self.sectcheck += name + " "
193            self.parameterdesc_start_lines[name] = self.new_start_line
194            self.new_start_line = 0
195
196        else:
197            if name in self.sections and self.sections[name] != "":
198                # Only warn on user-specified duplicate section names
199                if name != SECTION_DEFAULT:
200                    self.emit_msg(self.new_start_line,
201                                  f"duplicate section name '{name}'\n")
202                self.sections[name] += 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        self.inline_doc_state = state.INLINE_NA
238
239        # Store entry currently being processed
240        self.entry = None
241
242        # Place all potential outputs into an array
243        self.entries = []
244
245    def emit_msg(self, ln, msg, warning=True):
246        """Emit a message"""
247
248        log_msg = f"{self.fname}:{ln} {msg}"
249
250        if self.entry:
251            self.entry.emit_msg(log_msg, warning)
252            return
253
254        if warning:
255            self.config.log.warning(log_msg)
256        else:
257            self.config.log.info(log_msg)
258
259    def dump_section(self, start_new=True):
260        """
261        Dumps section contents to arrays/hashes intended for that purpose.
262        """
263
264        if self.entry:
265            self.entry.dump_section(start_new)
266
267    # TODO: rename it to store_declaration after removal of kernel-doc.pl
268    def output_declaration(self, dtype, name, **args):
269        """
270        Stores the entry into an entry array.
271
272        The actual output and output filters will be handled elsewhere
273        """
274
275        # The implementation here is different than the original kernel-doc:
276        # instead of checking for output filters or actually output anything,
277        # it just stores the declaration content at self.entries, as the
278        # output will happen on a separate class.
279        #
280        # For now, we're keeping the same name of the function just to make
281        # easier to compare the source code of both scripts
282
283        args["declaration_start_line"] = self.entry.declaration_start_line
284        args["type"] = dtype
285        args["warnings"] = self.entry.warnings
286
287        # TODO: use colletions.OrderedDict to remove sectionlist
288
289        sections = args.get('sections', {})
290        sectionlist = args.get('sectionlist', [])
291
292        # Drop empty sections
293        # TODO: improve empty sections logic to emit warnings
294        for section in ["Description", "Return"]:
295            if section in sectionlist:
296                if not sections[section].rstrip():
297                    del sections[section]
298                    sectionlist.remove(section)
299
300        self.entries.append((name, args))
301
302        self.config.log.debug("Output: %s:%s = %s", dtype, name, pformat(args))
303
304    def reset_state(self, ln):
305        """
306        Ancillary routine to create a new entry. It initializes all
307        variables used by the state machine.
308        """
309
310        self.entry = KernelEntry(self.config, ln)
311
312        # State flags
313        self.state = state.NORMAL
314        self.inline_doc_state = state.INLINE_NA
315
316    def push_parameter(self, ln, decl_type, param, dtype,
317                       org_arg, declaration_name):
318        """
319        Store parameters and their descriptions at self.entry.
320        """
321
322        if self.entry.anon_struct_union and dtype == "" and param == "}":
323            return  # Ignore the ending }; from anonymous struct/union
324
325        self.entry.anon_struct_union = False
326
327        param = KernRe(r'[\[\)].*').sub('', param, count=1)
328
329        if dtype == "" and param.endswith("..."):
330            if KernRe(r'\w\.\.\.$').search(param):
331                # For named variable parameters of the form `x...`,
332                # remove the dots
333                param = param[:-3]
334            else:
335                # Handles unnamed variable parameters
336                param = "..."
337
338            if param not in self.entry.parameterdescs or \
339                not self.entry.parameterdescs[param]:
340
341                self.entry.parameterdescs[param] = "variable arguments"
342
343        elif dtype == "" and (not param or param == "void"):
344            param = "void"
345            self.entry.parameterdescs[param] = "no arguments"
346
347        elif dtype == "" and param in ["struct", "union"]:
348            # Handle unnamed (anonymous) union or struct
349            dtype = param
350            param = "{unnamed_" + param + "}"
351            self.entry.parameterdescs[param] = "anonymous\n"
352            self.entry.anon_struct_union = True
353
354        # Handle cache group enforcing variables: they do not need
355        # to be described in header files
356        elif "__cacheline_group" in param:
357            # Ignore __cacheline_group_begin and __cacheline_group_end
358            return
359
360        # Warn if parameter has no description
361        # (but ignore ones starting with # as these are not parameters
362        # but inline preprocessor statements)
363        if param not in self.entry.parameterdescs and not param.startswith("#"):
364            self.entry.parameterdescs[param] = self.undescribed
365
366            if "." not in param:
367                if decl_type == 'function':
368                    dname = f"{decl_type} parameter"
369                else:
370                    dname = f"{decl_type} member"
371
372                self.emit_msg(ln,
373                              f"{dname} '{param}' not described in '{declaration_name}'")
374
375        # Strip spaces from param so that it is one continuous string on
376        # parameterlist. This fixes a problem where check_sections()
377        # cannot find a parameter like "addr[6 + 2]" because it actually
378        # appears as "addr[6", "+", "2]" on the parameter list.
379        # However, it's better to maintain the param string unchanged for
380        # output, so just weaken the string compare in check_sections()
381        # to ignore "[blah" in a parameter string.
382
383        self.entry.parameterlist.append(param)
384        org_arg = KernRe(r'\s\s+').sub(' ', org_arg)
385        self.entry.parametertypes[param] = org_arg
386
387    def save_struct_actual(self, actual):
388        """
389        Strip all spaces from the actual param so that it looks like
390        one string item.
391        """
392
393        actual = KernRe(r'\s*').sub("", actual, count=1)
394
395        self.entry.struct_actual += actual + " "
396
397    def create_parameter_list(self, ln, decl_type, args,
398                              splitter, declaration_name):
399        """
400        Creates a list of parameters, storing them at self.entry.
401        """
402
403        # temporarily replace all commas inside function pointer definition
404        arg_expr = KernRe(r'(\([^\),]+),')
405        while arg_expr.search(args):
406            args = arg_expr.sub(r"\1#", args)
407
408        for arg in args.split(splitter):
409            # Strip comments
410            arg = KernRe(r'\/\*.*\*\/').sub('', arg)
411
412            # Ignore argument attributes
413            arg = KernRe(r'\sPOS0?\s').sub(' ', arg)
414
415            # Strip leading/trailing spaces
416            arg = arg.strip()
417            arg = KernRe(r'\s+').sub(' ', arg, count=1)
418
419            if arg.startswith('#'):
420                # Treat preprocessor directive as a typeless variable just to fill
421                # corresponding data structures "correctly". Catch it later in
422                # output_* subs.
423
424                # Treat preprocessor directive as a typeless variable
425                self.push_parameter(ln, decl_type, arg, "",
426                                    "", declaration_name)
427
428            elif KernRe(r'\(.+\)\s*\(').search(arg):
429                # Pointer-to-function
430
431                arg = arg.replace('#', ',')
432
433                r = KernRe(r'[^\(]+\(\*?\s*([\w\[\]\.]*)\s*\)')
434                if r.match(arg):
435                    param = r.group(1)
436                else:
437                    self.emit_msg(ln, f"Invalid param: {arg}")
438                    param = arg
439
440                dtype = KernRe(r'([^\(]+\(\*?)\s*' + re.escape(param)).sub(r'\1', arg)
441                self.save_struct_actual(param)
442                self.push_parameter(ln, decl_type, param, dtype,
443                                    arg, declaration_name)
444
445            elif KernRe(r'\(.+\)\s*\[').search(arg):
446                # Array-of-pointers
447
448                arg = arg.replace('#', ',')
449                r = KernRe(r'[^\(]+\(\s*\*\s*([\w\[\]\.]*?)\s*(\s*\[\s*[\w]+\s*\]\s*)*\)')
450                if r.match(arg):
451                    param = r.group(1)
452                else:
453                    self.emit_msg(ln, f"Invalid param: {arg}")
454                    param = arg
455
456                dtype = KernRe(r'([^\(]+\(\*?)\s*' + re.escape(param)).sub(r'\1', arg)
457
458                self.save_struct_actual(param)
459                self.push_parameter(ln, decl_type, param, dtype,
460                                    arg, declaration_name)
461
462            elif arg:
463                arg = KernRe(r'\s*:\s*').sub(":", arg)
464                arg = KernRe(r'\s*\[').sub('[', arg)
465
466                args = KernRe(r'\s*,\s*').split(arg)
467                if args[0] and '*' in args[0]:
468                    args[0] = re.sub(r'(\*+)\s*', r' \1', args[0])
469
470                first_arg = []
471                r = KernRe(r'^(.*\s+)(.*?\[.*\].*)$')
472                if args[0] and r.match(args[0]):
473                    args.pop(0)
474                    first_arg.extend(r.group(1))
475                    first_arg.append(r.group(2))
476                else:
477                    first_arg = KernRe(r'\s+').split(args.pop(0))
478
479                args.insert(0, first_arg.pop())
480                dtype = ' '.join(first_arg)
481
482                for param in args:
483                    if KernRe(r'^(\*+)\s*(.*)').match(param):
484                        r = KernRe(r'^(\*+)\s*(.*)')
485                        if not r.match(param):
486                            self.emit_msg(ln, f"Invalid param: {param}")
487                            continue
488
489                        param = r.group(1)
490
491                        self.save_struct_actual(r.group(2))
492                        self.push_parameter(ln, decl_type, r.group(2),
493                                            f"{dtype} {r.group(1)}",
494                                            arg, declaration_name)
495
496                    elif KernRe(r'(.*?):(\w+)').search(param):
497                        r = KernRe(r'(.*?):(\w+)')
498                        if not r.match(param):
499                            self.emit_msg(ln, f"Invalid param: {param}")
500                            continue
501
502                        if dtype != "":  # Skip unnamed bit-fields
503                            self.save_struct_actual(r.group(1))
504                            self.push_parameter(ln, decl_type, r.group(1),
505                                                f"{dtype}:{r.group(2)}",
506                                                arg, declaration_name)
507                    else:
508                        self.save_struct_actual(param)
509                        self.push_parameter(ln, decl_type, param, dtype,
510                                            arg, declaration_name)
511
512    def check_sections(self, ln, decl_name, decl_type, sectcheck, prmscheck):
513        """
514        Check for errors inside sections, emitting warnings if not found
515        parameters are described.
516        """
517
518        sects = sectcheck.split()
519        prms = prmscheck.split()
520        err = False
521
522        for sx in range(len(sects)):                  # pylint: disable=C0200
523            err = True
524            for px in range(len(prms)):               # pylint: disable=C0200
525                prm_clean = prms[px]
526                prm_clean = KernRe(r'\[.*\]').sub('', prm_clean)
527                prm_clean = attribute.sub('', prm_clean)
528
529                # ignore array size in a parameter string;
530                # however, the original param string may contain
531                # spaces, e.g.:  addr[6 + 2]
532                # and this appears in @prms as "addr[6" since the
533                # parameter list is split at spaces;
534                # hence just ignore "[..." for the sections check;
535                prm_clean = KernRe(r'\[.*').sub('', prm_clean)
536
537                if prm_clean == sects[sx]:
538                    err = False
539                    break
540
541            if err:
542                if decl_type == 'function':
543                    dname = f"{decl_type} parameter"
544                else:
545                    dname = f"{decl_type} member"
546
547                self.emit_msg(ln,
548                              f"Excess {dname} '{sects[sx]}' description in '{decl_name}'")
549
550    def check_return_section(self, ln, declaration_name, return_type):
551        """
552        If the function doesn't return void, warns about the lack of a
553        return description.
554        """
555
556        if not self.config.wreturn:
557            return
558
559        # Ignore an empty return type (It's a macro)
560        # Ignore functions with a "void" return type (but not "void *")
561        if not return_type or KernRe(r'void\s*\w*\s*$').search(return_type):
562            return
563
564        if not self.entry.sections.get("Return", None):
565            self.emit_msg(ln,
566                          f"No description found for return value of '{declaration_name}'")
567
568    def dump_struct(self, ln, proto):
569        """
570        Store an entry for an struct or union
571        """
572
573        type_pattern = r'(struct|union)'
574
575        qualifiers = [
576            "__attribute__",
577            "__packed",
578            "__aligned",
579            "____cacheline_aligned_in_smp",
580            "____cacheline_aligned",
581        ]
582
583        definition_body = r'\{(.*)\}\s*' + "(?:" + '|'.join(qualifiers) + ")?"
584        struct_members = KernRe(type_pattern + r'([^\{\};]+)(\{)([^\{\}]*)(\})([^\{\}\;]*)(\;)')
585
586        # Extract struct/union definition
587        members = None
588        declaration_name = None
589        decl_type = None
590
591        r = KernRe(type_pattern + r'\s+(\w+)\s*' + definition_body)
592        if r.search(proto):
593            decl_type = r.group(1)
594            declaration_name = r.group(2)
595            members = r.group(3)
596        else:
597            r = KernRe(r'typedef\s+' + type_pattern + r'\s*' + definition_body + r'\s*(\w+)\s*;')
598
599            if r.search(proto):
600                decl_type = r.group(1)
601                declaration_name = r.group(3)
602                members = r.group(2)
603
604        if not members:
605            self.emit_msg(ln, f"{proto} error: Cannot parse struct or union!")
606            return
607
608        if self.entry.identifier != declaration_name:
609            self.emit_msg(ln,
610                          f"expecting prototype for {decl_type} {self.entry.identifier}. Prototype was for {decl_type} {declaration_name} instead\n")
611            return
612
613        args_pattern = r'([^,)]+)'
614
615        sub_prefixes = [
616            (KernRe(r'\/\*\s*private:.*?\/\*\s*public:.*?\*\/', re.S | re.I), ''),
617            (KernRe(r'\/\*\s*private:.*', re.S | re.I), ''),
618
619            # Strip comments
620            (KernRe(r'\/\*.*?\*\/', re.S), ''),
621
622            # Strip attributes
623            (attribute, ' '),
624            (KernRe(r'\s*__aligned\s*\([^;]*\)', re.S), ' '),
625            (KernRe(r'\s*__counted_by\s*\([^;]*\)', re.S), ' '),
626            (KernRe(r'\s*__counted_by_(le|be)\s*\([^;]*\)', re.S), ' '),
627            (KernRe(r'\s*__packed\s*', re.S), ' '),
628            (KernRe(r'\s*CRYPTO_MINALIGN_ATTR', re.S), ' '),
629            (KernRe(r'\s*____cacheline_aligned_in_smp', re.S), ' '),
630            (KernRe(r'\s*____cacheline_aligned', re.S), ' '),
631
632            # Unwrap struct_group macros based on this definition:
633            # __struct_group(TAG, NAME, ATTRS, MEMBERS...)
634            # which has variants like: struct_group(NAME, MEMBERS...)
635            # Only MEMBERS arguments require documentation.
636            #
637            # Parsing them happens on two steps:
638            #
639            # 1. drop struct group arguments that aren't at MEMBERS,
640            #    storing them as STRUCT_GROUP(MEMBERS)
641            #
642            # 2. remove STRUCT_GROUP() ancillary macro.
643            #
644            # The original logic used to remove STRUCT_GROUP() using an
645            # advanced regex:
646            #
647            #   \bSTRUCT_GROUP(\(((?:(?>[^)(]+)|(?1))*)\))[^;]*;
648            #
649            # with two patterns that are incompatible with
650            # Python re module, as it has:
651            #
652            #   - a recursive pattern: (?1)
653            #   - an atomic grouping: (?>...)
654            #
655            # I tried a simpler version: but it didn't work either:
656            #   \bSTRUCT_GROUP\(([^\)]+)\)[^;]*;
657            #
658            # As it doesn't properly match the end parenthesis on some cases.
659            #
660            # So, a better solution was crafted: there's now a NestedMatch
661            # class that ensures that delimiters after a search are properly
662            # matched. So, the implementation to drop STRUCT_GROUP() will be
663            # handled in separate.
664
665            (KernRe(r'\bstruct_group\s*\(([^,]*,)', re.S), r'STRUCT_GROUP('),
666            (KernRe(r'\bstruct_group_attr\s*\(([^,]*,){2}', re.S), r'STRUCT_GROUP('),
667            (KernRe(r'\bstruct_group_tagged\s*\(([^,]*),([^,]*),', re.S), r'struct \1 \2; STRUCT_GROUP('),
668            (KernRe(r'\b__struct_group\s*\(([^,]*,){3}', re.S), r'STRUCT_GROUP('),
669
670            # Replace macros
671            #
672            # TODO: use NestedMatch for FOO($1, $2, ...) matches
673            #
674            # it is better to also move those to the NestedMatch logic,
675            # to ensure that parenthesis will be properly matched.
676
677            (KernRe(r'__ETHTOOL_DECLARE_LINK_MODE_MASK\s*\(([^\)]+)\)', re.S), r'DECLARE_BITMAP(\1, __ETHTOOL_LINK_MODE_MASK_NBITS)'),
678            (KernRe(r'DECLARE_PHY_INTERFACE_MASK\s*\(([^\)]+)\)', re.S), r'DECLARE_BITMAP(\1, PHY_INTERFACE_MODE_MAX)'),
679            (KernRe(r'DECLARE_BITMAP\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'unsigned long \1[BITS_TO_LONGS(\2)]'),
680            (KernRe(r'DECLARE_HASHTABLE\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'unsigned long \1[1 << ((\2) - 1)]'),
681            (KernRe(r'DECLARE_KFIFO\s*\(' + args_pattern + r',\s*' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\2 *\1'),
682            (KernRe(r'DECLARE_KFIFO_PTR\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\2 *\1'),
683            (KernRe(r'(?:__)?DECLARE_FLEX_ARRAY\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\1 \2[]'),
684            (KernRe(r'DEFINE_DMA_UNMAP_ADDR\s*\(' + args_pattern + r'\)', re.S), r'dma_addr_t \1'),
685            (KernRe(r'DEFINE_DMA_UNMAP_LEN\s*\(' + args_pattern + r'\)', re.S), r'__u32 \1'),
686        ]
687
688        # Regexes here are guaranteed to have the end limiter matching
689        # the start delimiter. Yet, right now, only one replace group
690        # is allowed.
691
692        sub_nested_prefixes = [
693            (re.compile(r'\bSTRUCT_GROUP\('), r'\1'),
694        ]
695
696        for search, sub in sub_prefixes:
697            members = search.sub(sub, members)
698
699        nested = NestedMatch()
700
701        for search, sub in sub_nested_prefixes:
702            members = nested.sub(search, sub, members)
703
704        # Keeps the original declaration as-is
705        declaration = members
706
707        # Split nested struct/union elements
708        #
709        # This loop was simpler at the original kernel-doc perl version, as
710        #   while ($members =~ m/$struct_members/) { ... }
711        # reads 'members' string on each interaction.
712        #
713        # Python behavior is different: it parses 'members' only once,
714        # creating a list of tuples from the first interaction.
715        #
716        # On other words, this won't get nested structs.
717        #
718        # So, we need to have an extra loop on Python to override such
719        # re limitation.
720
721        while True:
722            tuples = struct_members.findall(members)
723            if not tuples:
724                break
725
726            for t in tuples:
727                newmember = ""
728                maintype = t[0]
729                s_ids = t[5]
730                content = t[3]
731
732                oldmember = "".join(t)
733
734                for s_id in s_ids.split(','):
735                    s_id = s_id.strip()
736
737                    newmember += f"{maintype} {s_id}; "
738                    s_id = KernRe(r'[:\[].*').sub('', s_id)
739                    s_id = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', s_id)
740
741                    for arg in content.split(';'):
742                        arg = arg.strip()
743
744                        if not arg:
745                            continue
746
747                        r = KernRe(r'^([^\(]+\(\*?\s*)([\w\.]*)(\s*\).*)')
748                        if r.match(arg):
749                            # Pointer-to-function
750                            dtype = r.group(1)
751                            name = r.group(2)
752                            extra = r.group(3)
753
754                            if not name:
755                                continue
756
757                            if not s_id:
758                                # Anonymous struct/union
759                                newmember += f"{dtype}{name}{extra}; "
760                            else:
761                                newmember += f"{dtype}{s_id}.{name}{extra}; "
762
763                        else:
764                            arg = arg.strip()
765                            # Handle bitmaps
766                            arg = KernRe(r':\s*\d+\s*').sub('', arg)
767
768                            # Handle arrays
769                            arg = KernRe(r'\[.*\]').sub('', arg)
770
771                            # Handle multiple IDs
772                            arg = KernRe(r'\s*,\s*').sub(',', arg)
773
774                            r = KernRe(r'(.*)\s+([\S+,]+)')
775
776                            if r.search(arg):
777                                dtype = r.group(1)
778                                names = r.group(2)
779                            else:
780                                newmember += f"{arg}; "
781                                continue
782
783                            for name in names.split(','):
784                                name = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', name).strip()
785
786                                if not name:
787                                    continue
788
789                                if not s_id:
790                                    # Anonymous struct/union
791                                    newmember += f"{dtype} {name}; "
792                                else:
793                                    newmember += f"{dtype} {s_id}.{name}; "
794
795                members = members.replace(oldmember, newmember)
796
797        # Ignore other nested elements, like enums
798        members = re.sub(r'(\{[^\{\}]*\})', '', members)
799
800        self.create_parameter_list(ln, decl_type, members, ';',
801                                   declaration_name)
802        self.check_sections(ln, declaration_name, decl_type,
803                            self.entry.sectcheck, self.entry.struct_actual)
804
805        # Adjust declaration for better display
806        declaration = KernRe(r'([\{;])').sub(r'\1\n', declaration)
807        declaration = KernRe(r'\}\s+;').sub('};', declaration)
808
809        # Better handle inlined enums
810        while True:
811            r = KernRe(r'(enum\s+\{[^\}]+),([^\n])')
812            if not r.search(declaration):
813                break
814
815            declaration = r.sub(r'\1,\n\2', declaration)
816
817        def_args = declaration.split('\n')
818        level = 1
819        declaration = ""
820        for clause in def_args:
821
822            clause = clause.strip()
823            clause = KernRe(r'\s+').sub(' ', clause, count=1)
824
825            if not clause:
826                continue
827
828            if '}' in clause and level > 1:
829                level -= 1
830
831            if not KernRe(r'^\s*#').match(clause):
832                declaration += "\t" * level
833
834            declaration += "\t" + clause + "\n"
835            if "{" in clause and "}" not in clause:
836                level += 1
837
838        self.output_declaration(decl_type, declaration_name,
839                                struct=declaration_name,
840                                definition=declaration,
841                                parameterlist=self.entry.parameterlist,
842                                parameterdescs=self.entry.parameterdescs,
843                                parametertypes=self.entry.parametertypes,
844                                parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
845                                sectionlist=self.entry.sectionlist,
846                                sections=self.entry.sections,
847                                section_start_lines=self.entry.section_start_lines,
848                                purpose=self.entry.declaration_purpose)
849
850    def dump_enum(self, ln, proto):
851        """
852        Stores an enum inside self.entries array.
853        """
854
855        # Ignore members marked private
856        proto = KernRe(r'\/\*\s*private:.*?\/\*\s*public:.*?\*\/', flags=re.S).sub('', proto)
857        proto = KernRe(r'\/\*\s*private:.*}', flags=re.S).sub('}', proto)
858
859        # Strip comments
860        proto = KernRe(r'\/\*.*?\*\/', flags=re.S).sub('', proto)
861
862        # Strip #define macros inside enums
863        proto = KernRe(r'#\s*((define|ifdef|if)\s+|endif)[^;]*;', flags=re.S).sub('', proto)
864
865        members = None
866        declaration_name = None
867
868        r = KernRe(r'typedef\s+enum\s*\{(.*)\}\s*(\w*)\s*;')
869        if r.search(proto):
870            declaration_name = r.group(2)
871            members = r.group(1).rstrip()
872        else:
873            r = KernRe(r'enum\s+(\w*)\s*\{(.*)\}')
874            if r.match(proto):
875                declaration_name = r.group(1)
876                members = r.group(2).rstrip()
877
878        if not members:
879            self.emit_msg(ln, f"{proto}: error: Cannot parse enum!")
880            return
881
882        if self.entry.identifier != declaration_name:
883            if self.entry.identifier == "":
884                self.emit_msg(ln,
885                              f"{proto}: wrong kernel-doc identifier on prototype")
886            else:
887                self.emit_msg(ln,
888                              f"expecting prototype for enum {self.entry.identifier}. Prototype was for enum {declaration_name} instead")
889            return
890
891        if not declaration_name:
892            declaration_name = "(anonymous)"
893
894        member_set = set()
895
896        members = KernRe(r'\([^;]*?[\)]').sub('', members)
897
898        for arg in members.split(','):
899            if not arg:
900                continue
901            arg = KernRe(r'^\s*(\w+).*').sub(r'\1', arg)
902            self.entry.parameterlist.append(arg)
903            if arg not in self.entry.parameterdescs:
904                self.entry.parameterdescs[arg] = self.undescribed
905                self.emit_msg(ln,
906                              f"Enum value '{arg}' not described in enum '{declaration_name}'")
907            member_set.add(arg)
908
909        for k in self.entry.parameterdescs:
910            if k not in member_set:
911                self.emit_msg(ln,
912                              f"Excess enum value '%{k}' description in '{declaration_name}'")
913
914        self.output_declaration('enum', declaration_name,
915                                enum=declaration_name,
916                                parameterlist=self.entry.parameterlist,
917                                parameterdescs=self.entry.parameterdescs,
918                                parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
919                                sectionlist=self.entry.sectionlist,
920                                sections=self.entry.sections,
921                                section_start_lines=self.entry.section_start_lines,
922                                purpose=self.entry.declaration_purpose)
923
924    def dump_declaration(self, ln, prototype):
925        """
926        Stores a data declaration inside self.entries array.
927        """
928
929        if self.entry.decl_type == "enum":
930            self.dump_enum(ln, prototype)
931            return
932
933        if self.entry.decl_type == "typedef":
934            self.dump_typedef(ln, prototype)
935            return
936
937        if self.entry.decl_type in ["union", "struct"]:
938            self.dump_struct(ln, prototype)
939            return
940
941        self.output_declaration(self.entry.decl_type, prototype,
942                                entry=self.entry)
943
944    def dump_function(self, ln, prototype):
945        """
946        Stores a function of function macro inside self.entries array.
947        """
948
949        func_macro = False
950        return_type = ''
951        decl_type = 'function'
952
953        # Prefixes that would be removed
954        sub_prefixes = [
955            (r"^static +", "", 0),
956            (r"^extern +", "", 0),
957            (r"^asmlinkage +", "", 0),
958            (r"^inline +", "", 0),
959            (r"^__inline__ +", "", 0),
960            (r"^__inline +", "", 0),
961            (r"^__always_inline +", "", 0),
962            (r"^noinline +", "", 0),
963            (r"^__FORTIFY_INLINE +", "", 0),
964            (r"__init +", "", 0),
965            (r"__init_or_module +", "", 0),
966            (r"__deprecated +", "", 0),
967            (r"__flatten +", "", 0),
968            (r"__meminit +", "", 0),
969            (r"__must_check +", "", 0),
970            (r"__weak +", "", 0),
971            (r"__sched +", "", 0),
972            (r"_noprof", "", 0),
973            (r"__printf\s*\(\s*\d*\s*,\s*\d*\s*\) +", "", 0),
974            (r"__(?:re)?alloc_size\s*\(\s*\d+\s*(?:,\s*\d+\s*)?\) +", "", 0),
975            (r"__diagnose_as\s*\(\s*\S+\s*(?:,\s*\d+\s*)*\) +", "", 0),
976            (r"DECL_BUCKET_PARAMS\s*\(\s*(\S+)\s*,\s*(\S+)\s*\)", r"\1, \2", 0),
977            (r"__attribute_const__ +", "", 0),
978
979            # It seems that Python support for re.X is broken:
980            # At least for me (Python 3.13), this didn't work
981#            (r"""
982#              __attribute__\s*\(\(
983#                (?:
984#                    [\w\s]+          # attribute name
985#                    (?:\([^)]*\))?   # attribute arguments
986#                    \s*,?            # optional comma at the end
987#                )+
988#              \)\)\s+
989#             """, "", re.X),
990
991            # So, remove whitespaces and comments from it
992            (r"__attribute__\s*\(\((?:[\w\s]+(?:\([^)]*\))?\s*,?)+\)\)\s+", "", 0),
993        ]
994
995        for search, sub, flags in sub_prefixes:
996            prototype = KernRe(search, flags).sub(sub, prototype)
997
998        # Macros are a special case, as they change the prototype format
999        new_proto = KernRe(r"^#\s*define\s+").sub("", prototype)
1000        if new_proto != prototype:
1001            is_define_proto = True
1002            prototype = new_proto
1003        else:
1004            is_define_proto = False
1005
1006        # Yes, this truly is vile.  We are looking for:
1007        # 1. Return type (may be nothing if we're looking at a macro)
1008        # 2. Function name
1009        # 3. Function parameters.
1010        #
1011        # All the while we have to watch out for function pointer parameters
1012        # (which IIRC is what the two sections are for), C types (these
1013        # regexps don't even start to express all the possibilities), and
1014        # so on.
1015        #
1016        # If you mess with these regexps, it's a good idea to check that
1017        # the following functions' documentation still comes out right:
1018        # - parport_register_device (function pointer parameters)
1019        # - atomic_set (macro)
1020        # - pci_match_device, __copy_to_user (long return type)
1021
1022        name = r'[a-zA-Z0-9_~:]+'
1023        prototype_end1 = r'[^\(]*'
1024        prototype_end2 = r'[^\{]*'
1025        prototype_end = fr'\(({prototype_end1}|{prototype_end2})\)'
1026
1027        # Besides compiling, Perl qr{[\w\s]+} works as a non-capturing group.
1028        # So, this needs to be mapped in Python with (?:...)? or (?:...)+
1029
1030        type1 = r'(?:[\w\s]+)?'
1031        type2 = r'(?:[\w\s]+\*+)+'
1032
1033        found = False
1034
1035        if is_define_proto:
1036            r = KernRe(r'^()(' + name + r')\s+')
1037
1038            if r.search(prototype):
1039                return_type = ''
1040                declaration_name = r.group(2)
1041                func_macro = True
1042
1043                found = True
1044
1045        if not found:
1046            patterns = [
1047                rf'^()({name})\s*{prototype_end}',
1048                rf'^({type1})\s+({name})\s*{prototype_end}',
1049                rf'^({type2})\s*({name})\s*{prototype_end}',
1050            ]
1051
1052            for p in patterns:
1053                r = KernRe(p)
1054
1055                if r.match(prototype):
1056
1057                    return_type = r.group(1)
1058                    declaration_name = r.group(2)
1059                    args = r.group(3)
1060
1061                    self.create_parameter_list(ln, decl_type, args, ',',
1062                                               declaration_name)
1063
1064                    found = True
1065                    break
1066        if not found:
1067            self.emit_msg(ln,
1068                          f"cannot understand function prototype: '{prototype}'")
1069            return
1070
1071        if self.entry.identifier != declaration_name:
1072            self.emit_msg(ln,
1073                          f"expecting prototype for {self.entry.identifier}(). Prototype was for {declaration_name}() instead")
1074            return
1075
1076        prms = " ".join(self.entry.parameterlist)
1077        self.check_sections(ln, declaration_name, "function",
1078                            self.entry.sectcheck, prms)
1079
1080        self.check_return_section(ln, declaration_name, return_type)
1081
1082        if 'typedef' in return_type:
1083            self.output_declaration(decl_type, declaration_name,
1084                                    function=declaration_name,
1085                                    typedef=True,
1086                                    functiontype=return_type,
1087                                    parameterlist=self.entry.parameterlist,
1088                                    parameterdescs=self.entry.parameterdescs,
1089                                    parametertypes=self.entry.parametertypes,
1090                                    parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
1091                                    sectionlist=self.entry.sectionlist,
1092                                    sections=self.entry.sections,
1093                                    section_start_lines=self.entry.section_start_lines,
1094                                    purpose=self.entry.declaration_purpose,
1095                                    func_macro=func_macro)
1096        else:
1097            self.output_declaration(decl_type, declaration_name,
1098                                    function=declaration_name,
1099                                    typedef=False,
1100                                    functiontype=return_type,
1101                                    parameterlist=self.entry.parameterlist,
1102                                    parameterdescs=self.entry.parameterdescs,
1103                                    parametertypes=self.entry.parametertypes,
1104                                    parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
1105                                    sectionlist=self.entry.sectionlist,
1106                                    sections=self.entry.sections,
1107                                    section_start_lines=self.entry.section_start_lines,
1108                                    purpose=self.entry.declaration_purpose,
1109                                    func_macro=func_macro)
1110
1111    def dump_typedef(self, ln, proto):
1112        """
1113        Stores a typedef inside self.entries array.
1114        """
1115
1116        typedef_type = r'((?:\s+[\w\*]+\b){0,7}\s+(?:\w+\b|\*+))\s*'
1117        typedef_ident = r'\*?\s*(\w\S+)\s*'
1118        typedef_args = r'\s*\((.*)\);'
1119
1120        typedef1 = KernRe(r'typedef' + typedef_type + r'\(' + typedef_ident + r'\)' + typedef_args)
1121        typedef2 = KernRe(r'typedef' + typedef_type + typedef_ident + typedef_args)
1122
1123        # Strip comments
1124        proto = KernRe(r'/\*.*?\*/', flags=re.S).sub('', proto)
1125
1126        # Parse function typedef prototypes
1127        for r in [typedef1, typedef2]:
1128            if not r.match(proto):
1129                continue
1130
1131            return_type = r.group(1).strip()
1132            declaration_name = r.group(2)
1133            args = r.group(3)
1134
1135            if self.entry.identifier != declaration_name:
1136                self.emit_msg(ln,
1137                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1138                return
1139
1140            decl_type = 'function'
1141            self.create_parameter_list(ln, decl_type, args, ',', declaration_name)
1142
1143            self.output_declaration(decl_type, declaration_name,
1144                                    function=declaration_name,
1145                                    typedef=True,
1146                                    functiontype=return_type,
1147                                    parameterlist=self.entry.parameterlist,
1148                                    parameterdescs=self.entry.parameterdescs,
1149                                    parametertypes=self.entry.parametertypes,
1150                                    parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
1151                                    sectionlist=self.entry.sectionlist,
1152                                    sections=self.entry.sections,
1153                                    section_start_lines=self.entry.section_start_lines,
1154                                    purpose=self.entry.declaration_purpose)
1155            return
1156
1157        # Handle nested parentheses or brackets
1158        r = KernRe(r'(\(*.\)\s*|\[*.\]\s*);$')
1159        while r.search(proto):
1160            proto = r.sub('', proto)
1161
1162        # Parse simple typedefs
1163        r = KernRe(r'typedef.*\s+(\w+)\s*;')
1164        if r.match(proto):
1165            declaration_name = r.group(1)
1166
1167            if self.entry.identifier != declaration_name:
1168                self.emit_msg(ln,
1169                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1170                return
1171
1172            self.output_declaration('typedef', declaration_name,
1173                                    typedef=declaration_name,
1174                                    sectionlist=self.entry.sectionlist,
1175                                    sections=self.entry.sections,
1176                                    section_start_lines=self.entry.section_start_lines,
1177                                    purpose=self.entry.declaration_purpose)
1178            return
1179
1180        self.emit_msg(ln, "error: Cannot parse typedef!")
1181
1182    @staticmethod
1183    def process_export(function_set, line):
1184        """
1185        process EXPORT_SYMBOL* tags
1186
1187        This method doesn't use any variable from the class, so declare it
1188        with a staticmethod decorator.
1189        """
1190
1191        # We support documenting some exported symbols with different
1192        # names.  A horrible hack.
1193        suffixes = [ '_noprof' ]
1194
1195        # Note: it accepts only one EXPORT_SYMBOL* per line, as having
1196        # multiple export lines would violate Kernel coding style.
1197
1198        if export_symbol.search(line):
1199            symbol = export_symbol.group(2)
1200            for suffix in suffixes:
1201                symbol = symbol.removesuffix(suffix)
1202            function_set.add(symbol)
1203            return
1204
1205        if export_symbol_ns.search(line):
1206            symbol = export_symbol_ns.group(2)
1207            for suffix in suffixes:
1208                symbol = symbol.removesuffix(suffix)
1209            function_set.add(symbol)
1210
1211    def process_normal(self, ln, line):
1212        """
1213        STATE_NORMAL: looking for the /** to begin everything.
1214        """
1215
1216        if not doc_start.match(line):
1217            return
1218
1219        # start a new entry
1220        self.reset_state(ln)
1221        self.entry.in_doc_sect = False
1222
1223        # next line is always the function name
1224        self.state = state.NAME
1225
1226    def process_name(self, ln, line):
1227        """
1228        STATE_NAME: Looking for the "name - description" line
1229        """
1230        #
1231        # Check for a DOC: block and handle them specially.
1232        #
1233        if doc_block.search(line):
1234            self.entry.new_start_line = ln
1235
1236            if not doc_block.group(1):
1237                self.entry.section = "Introduction"
1238            else:
1239                self.entry.section = doc_block.group(1)
1240
1241            self.entry.identifier = self.entry.section
1242            self.state = state.DOCBLOCK
1243        #
1244        # Otherwise we're looking for a normal kerneldoc declaration line.
1245        #
1246        elif doc_decl.search(line):
1247            self.entry.identifier = doc_decl.group(1)
1248
1249            # Test for data declaration
1250            if doc_begin_data.search(line):
1251                self.entry.decl_type = doc_begin_data.group(1)
1252                self.entry.identifier = doc_begin_data.group(2)
1253            #
1254            # Look for a function description
1255            #
1256            elif doc_begin_func.search(line):
1257                self.entry.identifier = doc_begin_func.group(1)
1258                self.entry.decl_type = "function"
1259            #
1260            # We struck out.
1261            #
1262            else:
1263                self.emit_msg(ln,
1264                              f"This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst\n{line}")
1265                self.state = state.NORMAL
1266                return
1267            #
1268            # OK, set up for a new kerneldoc entry.
1269            #
1270            self.state = state.BODY
1271            self.entry.identifier = self.entry.identifier.strip(" ")
1272            # if there's no @param blocks need to set up default section here
1273            self.entry.section = SECTION_DEFAULT
1274            self.entry.new_start_line = ln + 1
1275            #
1276            # Find the description portion, which *should* be there but
1277            # isn't always.
1278            # (We should be able to capture this from the previous parsing - someday)
1279            #
1280            r = KernRe("[-:](.*)")
1281            if r.search(line):
1282                self.entry.declaration_purpose = trim_whitespace(r.group(1))
1283                self.state = state.BODY_MAYBE
1284            else:
1285                self.entry.declaration_purpose = ""
1286
1287            if not self.entry.declaration_purpose and self.config.wshort_desc:
1288                self.emit_msg(ln,
1289                              f"missing initial short description on line:\n{line}")
1290
1291            if not self.entry.identifier and self.entry.decl_type != "enum":
1292                self.emit_msg(ln,
1293                              f"wrong kernel-doc identifier on line:\n{line}")
1294                self.state = state.NORMAL
1295
1296            if self.config.verbose:
1297                self.emit_msg(ln,
1298                              f"Scanning doc for {self.entry.decl_type} {self.entry.identifier}",
1299                                  warning=False)
1300        #
1301        # Failed to find an identifier. Emit a warning
1302        #
1303        else:
1304            self.emit_msg(ln, f"Cannot find identifier on line:\n{line}")
1305
1306    def process_body(self, ln, line):
1307        """
1308        STATE_BODY and STATE_BODY_MAYBE: the bulk of a kerneldoc comment.
1309        """
1310
1311        if self.state == state.BODY_WITH_BLANK_LINE:
1312            r = KernRe(r"\s*\*\s*\S")
1313            if r.match(line):
1314                self.dump_section()
1315                self.entry.section = SECTION_DEFAULT
1316                self.entry.new_start_line = ln
1317                self.entry.contents = ""
1318
1319        if doc_sect.search(line):
1320            self.entry.in_doc_sect = True
1321            newsection = doc_sect.group(1)
1322
1323            if newsection.lower() in ["description", "context"]:
1324                newsection = newsection.title()
1325
1326            # Special case: @return is a section, not a param description
1327            if newsection.lower() in ["@return", "@returns",
1328                                      "return", "returns"]:
1329                newsection = "Return"
1330
1331            # Perl kernel-doc has a check here for contents before sections.
1332            # the logic there is always false, as in_doc_sect variable is
1333            # always true. So, just don't implement Wcontents_before_sections
1334
1335            # .title()
1336            newcontents = doc_sect.group(2)
1337            if not newcontents:
1338                newcontents = ""
1339
1340            if self.entry.contents.strip("\n"):
1341                self.dump_section()
1342
1343            self.entry.new_start_line = ln
1344            self.entry.section = newsection
1345            self.entry.leading_space = None
1346
1347            self.entry.contents = newcontents.lstrip()
1348            if self.entry.contents:
1349                self.entry.contents += "\n"
1350
1351            self.state = state.BODY
1352            return
1353
1354        if doc_end.search(line):
1355            self.dump_section()
1356
1357            # Look for doc_com + <text> + doc_end:
1358            r = KernRe(r'\s*\*\s*[a-zA-Z_0-9:\.]+\*/')
1359            if r.match(line):
1360                self.emit_msg(ln, f"suspicious ending line: {line}")
1361
1362            self.entry.prototype = ""
1363            self.entry.new_start_line = ln + 1
1364
1365            self.state = state.PROTO
1366            return
1367
1368        if doc_content.search(line):
1369            cont = doc_content.group(1)
1370
1371            if cont == "":
1372                if self.entry.section == self.section_context:
1373                    self.dump_section()
1374
1375                    self.entry.new_start_line = ln
1376                    self.state = state.BODY
1377                else:
1378                    if self.entry.section != SECTION_DEFAULT:
1379                        self.state = state.BODY_WITH_BLANK_LINE
1380                    else:
1381                        self.state = state.BODY
1382
1383                    self.entry.contents += "\n"
1384
1385            elif self.state == state.BODY_MAYBE:
1386
1387                # Continued declaration purpose
1388                self.entry.declaration_purpose = self.entry.declaration_purpose.rstrip()
1389                self.entry.declaration_purpose += " " + cont
1390
1391                r = KernRe(r"\s+")
1392                self.entry.declaration_purpose = r.sub(' ',
1393                                                       self.entry.declaration_purpose)
1394
1395            else:
1396                if self.entry.section.startswith('@') or        \
1397                   self.entry.section == self.section_context:
1398                    if self.entry.leading_space is None:
1399                        r = KernRe(r'^(\s+)')
1400                        if r.match(cont):
1401                            self.entry.leading_space = len(r.group(1))
1402                        else:
1403                            self.entry.leading_space = 0
1404
1405                    # Double-check if leading space are realy spaces
1406                    pos = 0
1407                    for i in range(0, self.entry.leading_space):
1408                        if cont[i] != " ":
1409                            break
1410                        pos += 1
1411
1412                    cont = cont[pos:]
1413
1414                    # NEW LOGIC:
1415                    # In case it is different, update it
1416                    if self.entry.leading_space != pos:
1417                        self.entry.leading_space = pos
1418
1419                self.entry.contents += cont + "\n"
1420            return
1421
1422        # Unknown line, ignore
1423        self.emit_msg(ln, f"bad line: {line}")
1424
1425    def process_inline(self, ln, line):
1426        """STATE_INLINE: docbook comments within a prototype."""
1427
1428        if self.inline_doc_state == state.INLINE_NAME and \
1429           doc_inline_sect.search(line):
1430            self.entry.section = doc_inline_sect.group(1)
1431            self.entry.new_start_line = ln
1432
1433            self.entry.contents = doc_inline_sect.group(2).lstrip()
1434            if self.entry.contents != "":
1435                self.entry.contents += "\n"
1436
1437            self.inline_doc_state = state.INLINE_TEXT
1438            # Documentation block end */
1439            return
1440
1441        if doc_inline_end.search(line):
1442            if self.entry.contents not in ["", "\n"]:
1443                self.dump_section()
1444
1445            self.state = state.PROTO
1446            self.inline_doc_state = state.INLINE_NA
1447            return
1448
1449        if doc_content.search(line):
1450            if self.inline_doc_state == state.INLINE_TEXT:
1451                self.entry.contents += doc_content.group(1) + "\n"
1452                if not self.entry.contents.strip(" ").rstrip("\n"):
1453                    self.entry.contents = ""
1454
1455            elif self.inline_doc_state == state.INLINE_NAME:
1456                self.emit_msg(ln,
1457                              f"Incorrect use of kernel-doc format: {line}")
1458
1459                self.inline_doc_state = state.INLINE_ERROR
1460
1461    def syscall_munge(self, ln, proto):         # pylint: disable=W0613
1462        """
1463        Handle syscall definitions
1464        """
1465
1466        is_void = False
1467
1468        # Strip newlines/CR's
1469        proto = re.sub(r'[\r\n]+', ' ', proto)
1470
1471        # Check if it's a SYSCALL_DEFINE0
1472        if 'SYSCALL_DEFINE0' in proto:
1473            is_void = True
1474
1475        # Replace SYSCALL_DEFINE with correct return type & function name
1476        proto = KernRe(r'SYSCALL_DEFINE.*\(').sub('long sys_', proto)
1477
1478        r = KernRe(r'long\s+(sys_.*?),')
1479        if r.search(proto):
1480            proto = KernRe(',').sub('(', proto, count=1)
1481        elif is_void:
1482            proto = KernRe(r'\)').sub('(void)', proto, count=1)
1483
1484        # Now delete all of the odd-numbered commas in the proto
1485        # so that argument types & names don't have a comma between them
1486        count = 0
1487        length = len(proto)
1488
1489        if is_void:
1490            length = 0  # skip the loop if is_void
1491
1492        for ix in range(length):
1493            if proto[ix] == ',':
1494                count += 1
1495                if count % 2 == 1:
1496                    proto = proto[:ix] + ' ' + proto[ix + 1:]
1497
1498        return proto
1499
1500    def tracepoint_munge(self, ln, proto):
1501        """
1502        Handle tracepoint definitions
1503        """
1504
1505        tracepointname = None
1506        tracepointargs = None
1507
1508        # Match tracepoint name based on different patterns
1509        r = KernRe(r'TRACE_EVENT\((.*?),')
1510        if r.search(proto):
1511            tracepointname = r.group(1)
1512
1513        r = KernRe(r'DEFINE_SINGLE_EVENT\((.*?),')
1514        if r.search(proto):
1515            tracepointname = r.group(1)
1516
1517        r = KernRe(r'DEFINE_EVENT\((.*?),(.*?),')
1518        if r.search(proto):
1519            tracepointname = r.group(2)
1520
1521        if tracepointname:
1522            tracepointname = tracepointname.lstrip()
1523
1524        r = KernRe(r'TP_PROTO\((.*?)\)')
1525        if r.search(proto):
1526            tracepointargs = r.group(1)
1527
1528        if not tracepointname or not tracepointargs:
1529            self.emit_msg(ln,
1530                          f"Unrecognized tracepoint format:\n{proto}\n")
1531        else:
1532            proto = f"static inline void trace_{tracepointname}({tracepointargs})"
1533            self.entry.identifier = f"trace_{self.entry.identifier}"
1534
1535        return proto
1536
1537    def process_proto_function(self, ln, line):
1538        """Ancillary routine to process a function prototype"""
1539
1540        # strip C99-style comments to end of line
1541        r = KernRe(r"\/\/.*$", re.S)
1542        line = r.sub('', line)
1543
1544        if KernRe(r'\s*#\s*define').match(line):
1545            self.entry.prototype = line
1546        elif line.startswith('#'):
1547            # Strip other macros like #ifdef/#ifndef/#endif/...
1548            pass
1549        else:
1550            r = KernRe(r'([^\{]*)')
1551            if r.match(line):
1552                self.entry.prototype += r.group(1) + " "
1553
1554        if '{' in line or ';' in line or KernRe(r'\s*#\s*define').match(line):
1555            # strip comments
1556            r = KernRe(r'/\*.*?\*/')
1557            self.entry.prototype = r.sub('', self.entry.prototype)
1558
1559            # strip newlines/cr's
1560            r = KernRe(r'[\r\n]+')
1561            self.entry.prototype = r.sub(' ', self.entry.prototype)
1562
1563            # strip leading spaces
1564            r = KernRe(r'^\s+')
1565            self.entry.prototype = r.sub('', self.entry.prototype)
1566
1567            # Handle self.entry.prototypes for function pointers like:
1568            #       int (*pcs_config)(struct foo)
1569
1570            r = KernRe(r'^(\S+\s+)\(\s*\*(\S+)\)')
1571            self.entry.prototype = r.sub(r'\1\2', self.entry.prototype)
1572
1573            if 'SYSCALL_DEFINE' in self.entry.prototype:
1574                self.entry.prototype = self.syscall_munge(ln,
1575                                                          self.entry.prototype)
1576
1577            r = KernRe(r'TRACE_EVENT|DEFINE_EVENT|DEFINE_SINGLE_EVENT')
1578            if r.search(self.entry.prototype):
1579                self.entry.prototype = self.tracepoint_munge(ln,
1580                                                             self.entry.prototype)
1581
1582            self.dump_function(ln, self.entry.prototype)
1583            self.reset_state(ln)
1584
1585    def process_proto_type(self, ln, line):
1586        """Ancillary routine to process a type"""
1587
1588        # Strip newlines/cr's.
1589        line = KernRe(r'[\r\n]+', re.S).sub(' ', line)
1590
1591        # Strip leading spaces
1592        line = KernRe(r'^\s+', re.S).sub('', line)
1593
1594        # Strip trailing spaces
1595        line = KernRe(r'\s+$', re.S).sub('', line)
1596
1597        # Strip C99-style comments to the end of the line
1598        line = KernRe(r"\/\/.*$", re.S).sub('', line)
1599
1600        # To distinguish preprocessor directive from regular declaration later.
1601        if line.startswith('#'):
1602            line += ";"
1603
1604        r = KernRe(r'([^\{\};]*)([\{\};])(.*)')
1605        while True:
1606            if r.search(line):
1607                if self.entry.prototype:
1608                    self.entry.prototype += " "
1609                self.entry.prototype += r.group(1) + r.group(2)
1610
1611                self.entry.brcount += r.group(2).count('{')
1612                self.entry.brcount -= r.group(2).count('}')
1613
1614                self.entry.brcount = max(self.entry.brcount, 0)
1615
1616                if r.group(2) == ';' and self.entry.brcount == 0:
1617                    self.dump_declaration(ln, self.entry.prototype)
1618                    self.reset_state(ln)
1619                    break
1620
1621                line = r.group(3)
1622            else:
1623                self.entry.prototype += line
1624                break
1625
1626    def process_proto(self, ln, line):
1627        """STATE_PROTO: reading a function/whatever prototype."""
1628
1629        if doc_inline_oneline.search(line):
1630            self.entry.section = doc_inline_oneline.group(1)
1631            self.entry.contents = doc_inline_oneline.group(2)
1632
1633            if self.entry.contents != "":
1634                self.entry.contents += "\n"
1635                self.dump_section(start_new=False)
1636
1637        elif doc_inline_start.search(line):
1638            self.state = state.INLINE
1639            self.inline_doc_state = state.INLINE_NAME
1640
1641        elif self.entry.decl_type == 'function':
1642            self.process_proto_function(ln, line)
1643
1644        else:
1645            self.process_proto_type(ln, line)
1646
1647    def process_docblock(self, ln, line):
1648        """STATE_DOCBLOCK: within a DOC: block."""
1649
1650        if doc_end.search(line):
1651            self.dump_section()
1652            self.output_declaration("doc", self.entry.identifier,
1653                                    sectionlist=self.entry.sectionlist,
1654                                    sections=self.entry.sections,
1655                                    section_start_lines=self.entry.section_start_lines)
1656            self.reset_state(ln)
1657
1658        elif doc_content.search(line):
1659            self.entry.contents += doc_content.group(1) + "\n"
1660
1661    def parse_export(self):
1662        """
1663        Parses EXPORT_SYMBOL* macros from a single Kernel source file.
1664        """
1665
1666        export_table = set()
1667
1668        try:
1669            with open(self.fname, "r", encoding="utf8",
1670                      errors="backslashreplace") as fp:
1671
1672                for line in fp:
1673                    self.process_export(export_table, line)
1674
1675        except IOError:
1676            return None
1677
1678        return export_table
1679
1680    #
1681    # The state/action table telling us which function to invoke in
1682    # each state.
1683    #
1684    state_actions = {
1685        state.NORMAL:			process_normal,
1686        state.NAME:			process_name,
1687        state.BODY:			process_body,
1688        state.BODY_MAYBE:		process_body,
1689        state.BODY_WITH_BLANK_LINE:	process_body,
1690        state.INLINE:			process_inline,
1691        state.PROTO:			process_proto,
1692        state.DOCBLOCK:			process_docblock,
1693        }
1694
1695    def parse_kdoc(self):
1696        """
1697        Open and process each line of a C source file.
1698        The parsing is controlled via a state machine, and the line is passed
1699        to a different process function depending on the state. The process
1700        function may update the state as needed.
1701
1702        Besides parsing kernel-doc tags, it also parses export symbols.
1703        """
1704
1705        prev = ""
1706        prev_ln = None
1707        export_table = set()
1708
1709        try:
1710            with open(self.fname, "r", encoding="utf8",
1711                      errors="backslashreplace") as fp:
1712                for ln, line in enumerate(fp):
1713
1714                    line = line.expandtabs().strip("\n")
1715
1716                    # Group continuation lines on prototypes
1717                    if self.state == state.PROTO:
1718                        if line.endswith("\\"):
1719                            prev += line.rstrip("\\")
1720                            if not prev_ln:
1721                                prev_ln = ln
1722                            continue
1723
1724                        if prev:
1725                            ln = prev_ln
1726                            line = prev + line
1727                            prev = ""
1728                            prev_ln = None
1729
1730                    self.config.log.debug("%d %s%s: %s",
1731                                          ln, state.name[self.state],
1732                                          state.inline_name[self.inline_doc_state],
1733                                          line)
1734
1735                    # This is an optimization over the original script.
1736                    # There, when export_file was used for the same file,
1737                    # it was read twice. Here, we use the already-existing
1738                    # loop to parse exported symbols as well.
1739                    #
1740                    # TODO: It should be noticed that not all states are
1741                    # needed here. On a future cleanup, process export only
1742                    # at the states that aren't handling comment markups.
1743                    self.process_export(export_table, line)
1744
1745                    # Hand this line to the appropriate state handler
1746                    self.state_actions[self.state](self, ln, line)
1747
1748        except OSError:
1749            self.config.log.error(f"Error: Cannot open file {self.fname}")
1750
1751        return export_table, self.entries
1752