xref: /linux/scripts/lib/kdoc/kdoc_parser.py (revision 260f6f4fda93c8485c8037865c941b42b9cba5d2)
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_block = doc_com + KernRe(r'DOC:\s*(.*)?', cache=False)
51doc_inline_start = KernRe(r'^\s*/\*\*\s*$', cache=False)
52doc_inline_sect = KernRe(r'\s*\*\s*(@\s*[\w][\w\.]*\s*):(.*)', cache=False)
53doc_inline_end = KernRe(r'^\s*\*/\s*$', cache=False)
54doc_inline_oneline = KernRe(r'^\s*/\*\*\s*(@[\w\s]+):\s*(.*)\s*\*/\s*$', cache=False)
55attribute = KernRe(r"__attribute__\s*\(\([a-z0-9,_\*\s\(\)]*\)\)",
56               flags=re.I | re.S, cache=False)
57
58export_symbol = KernRe(r'^\s*EXPORT_SYMBOL(_GPL)?\s*\(\s*(\w+)\s*\)\s*', cache=False)
59export_symbol_ns = KernRe(r'^\s*EXPORT_SYMBOL_NS(_GPL)?\s*\(\s*(\w+)\s*,\s*"\S+"\)\s*', cache=False)
60
61type_param = KernRe(r"\@(\w*((\.\w+)|(->\w+))*(\.\.\.)?)", cache=False)
62
63class state:
64    """
65    State machine enums
66    """
67
68    # Parser states
69    NORMAL        = 0        # normal code
70    NAME          = 1        # looking for function name
71    BODY_MAYBE    = 2        # body - or maybe more description
72    BODY          = 3        # the body of the comment
73    BODY_WITH_BLANK_LINE = 4 # the body which has a blank line
74    PROTO         = 5        # scanning prototype
75    DOCBLOCK      = 6        # documentation block
76    INLINE        = 7        # gathering doc outside main block
77
78    name = [
79        "NORMAL",
80        "NAME",
81        "BODY_MAYBE",
82        "BODY",
83        "BODY_WITH_BLANK_LINE",
84        "PROTO",
85        "DOCBLOCK",
86        "INLINE",
87    ]
88
89    # Inline documentation state
90    INLINE_NA     = 0 # not applicable ($state != INLINE)
91    INLINE_NAME   = 1 # looking for member name (@foo:)
92    INLINE_TEXT   = 2 # looking for member documentation
93    INLINE_END    = 3 # done
94    INLINE_ERROR  = 4 # error - Comment without header was found.
95                      # Spit a warning as it's not
96                      # proper kernel-doc and ignore the rest.
97
98    inline_name = [
99        "",
100        "_NAME",
101        "_TEXT",
102        "_END",
103        "_ERROR",
104    ]
105
106SECTION_DEFAULT = "Description"  # default section
107
108class KernelEntry:
109
110    def __init__(self, config, ln):
111        self.config = config
112
113        self.contents = ""
114        self.function = ""
115        self.sectcheck = ""
116        self.struct_actual = ""
117        self.prototype = ""
118
119        self.warnings = []
120
121        self.parameterlist = []
122        self.parameterdescs = {}
123        self.parametertypes = {}
124        self.parameterdesc_start_lines = {}
125
126        self.section_start_lines = {}
127        self.sectionlist = []
128        self.sections = {}
129
130        self.anon_struct_union = False
131
132        self.leading_space = None
133
134        # State flags
135        self.brcount = 0
136
137        self.in_doc_sect = False
138        self.declaration_start_line = ln + 1
139
140    # TODO: rename to emit_message after removal of kernel-doc.pl
141    def emit_msg(self, log_msg, warning=True):
142        """Emit a message"""
143
144        if not warning:
145            self.config.log.info(log_msg)
146            return
147
148        # Delegate warning output to output logic, as this way it
149        # will report warnings/info only for symbols that are output
150
151        self.warnings.append(log_msg)
152        return
153
154    def dump_section(self, start_new=True):
155        """
156        Dumps section contents to arrays/hashes intended for that purpose.
157        """
158
159        name = self.section
160        contents = self.contents
161
162        if type_param.match(name):
163            name = type_param.group(1)
164
165            self.parameterdescs[name] = contents
166            self.parameterdesc_start_lines[name] = self.new_start_line
167
168            self.sectcheck += name + " "
169            self.new_start_line = 0
170
171        elif name == "@...":
172            name = "..."
173            self.parameterdescs[name] = contents
174            self.sectcheck += name + " "
175            self.parameterdesc_start_lines[name] = self.new_start_line
176            self.new_start_line = 0
177
178        else:
179            if name in self.sections and self.sections[name] != "":
180                # Only warn on user-specified duplicate section names
181                if name != SECTION_DEFAULT:
182                    self.emit_msg(self.new_start_line,
183                                  f"duplicate section name '{name}'\n")
184                self.sections[name] += contents
185            else:
186                self.sections[name] = contents
187                self.sectionlist.append(name)
188                self.section_start_lines[name] = self.new_start_line
189                self.new_start_line = 0
190
191#        self.config.log.debug("Section: %s : %s", name, pformat(vars(self)))
192
193        if start_new:
194            self.section = SECTION_DEFAULT
195            self.contents = ""
196
197
198class KernelDoc:
199    """
200    Read a C language source or header FILE and extract embedded
201    documentation comments.
202    """
203
204    # Section names
205
206    section_intro = "Introduction"
207    section_context = "Context"
208    section_return = "Return"
209
210    undescribed = "-- undescribed --"
211
212    def __init__(self, config, fname):
213        """Initialize internal variables"""
214
215        self.fname = fname
216        self.config = config
217
218        # Initial state for the state machines
219        self.state = state.NORMAL
220        self.inline_doc_state = state.INLINE_NA
221
222        # Store entry currently being processed
223        self.entry = None
224
225        # Place all potential outputs into an array
226        self.entries = []
227
228    def emit_msg(self, ln, msg, warning=True):
229        """Emit a message"""
230
231        log_msg = f"{self.fname}:{ln} {msg}"
232
233        if self.entry:
234            self.entry.emit_msg(log_msg, warning)
235            return
236
237        if warning:
238            self.config.log.warning(log_msg)
239        else:
240            self.config.log.info(log_msg)
241
242    def dump_section(self, start_new=True):
243        """
244        Dumps section contents to arrays/hashes intended for that purpose.
245        """
246
247        if self.entry:
248            self.entry.dump_section(start_new)
249
250    # TODO: rename it to store_declaration after removal of kernel-doc.pl
251    def output_declaration(self, dtype, name, **args):
252        """
253        Stores the entry into an entry array.
254
255        The actual output and output filters will be handled elsewhere
256        """
257
258        # The implementation here is different than the original kernel-doc:
259        # instead of checking for output filters or actually output anything,
260        # it just stores the declaration content at self.entries, as the
261        # output will happen on a separate class.
262        #
263        # For now, we're keeping the same name of the function just to make
264        # easier to compare the source code of both scripts
265
266        args["declaration_start_line"] = self.entry.declaration_start_line
267        args["type"] = dtype
268        args["warnings"] = self.entry.warnings
269
270        # TODO: use colletions.OrderedDict to remove sectionlist
271
272        sections = args.get('sections', {})
273        sectionlist = args.get('sectionlist', [])
274
275        # Drop empty sections
276        # TODO: improve empty sections logic to emit warnings
277        for section in ["Description", "Return"]:
278            if section in sectionlist:
279                if not sections[section].rstrip():
280                    del sections[section]
281                    sectionlist.remove(section)
282
283        self.entries.append((name, args))
284
285        self.config.log.debug("Output: %s:%s = %s", dtype, name, pformat(args))
286
287    def reset_state(self, ln):
288        """
289        Ancillary routine to create a new entry. It initializes all
290        variables used by the state machine.
291        """
292
293        self.entry = KernelEntry(self.config, ln)
294
295        # State flags
296        self.state = state.NORMAL
297        self.inline_doc_state = state.INLINE_NA
298
299    def push_parameter(self, ln, decl_type, param, dtype,
300                       org_arg, declaration_name):
301        """
302        Store parameters and their descriptions at self.entry.
303        """
304
305        if self.entry.anon_struct_union and dtype == "" and param == "}":
306            return  # Ignore the ending }; from anonymous struct/union
307
308        self.entry.anon_struct_union = False
309
310        param = KernRe(r'[\[\)].*').sub('', param, count=1)
311
312        if dtype == "" and param.endswith("..."):
313            if KernRe(r'\w\.\.\.$').search(param):
314                # For named variable parameters of the form `x...`,
315                # remove the dots
316                param = param[:-3]
317            else:
318                # Handles unnamed variable parameters
319                param = "..."
320
321            if param not in self.entry.parameterdescs or \
322                not self.entry.parameterdescs[param]:
323
324                self.entry.parameterdescs[param] = "variable arguments"
325
326        elif dtype == "" and (not param or param == "void"):
327            param = "void"
328            self.entry.parameterdescs[param] = "no arguments"
329
330        elif dtype == "" and param in ["struct", "union"]:
331            # Handle unnamed (anonymous) union or struct
332            dtype = param
333            param = "{unnamed_" + param + "}"
334            self.entry.parameterdescs[param] = "anonymous\n"
335            self.entry.anon_struct_union = True
336
337        # Handle cache group enforcing variables: they do not need
338        # to be described in header files
339        elif "__cacheline_group" in param:
340            # Ignore __cacheline_group_begin and __cacheline_group_end
341            return
342
343        # Warn if parameter has no description
344        # (but ignore ones starting with # as these are not parameters
345        # but inline preprocessor statements)
346        if param not in self.entry.parameterdescs and not param.startswith("#"):
347            self.entry.parameterdescs[param] = self.undescribed
348
349            if "." not in param:
350                if decl_type == 'function':
351                    dname = f"{decl_type} parameter"
352                else:
353                    dname = f"{decl_type} member"
354
355                self.emit_msg(ln,
356                              f"{dname} '{param}' not described in '{declaration_name}'")
357
358        # Strip spaces from param so that it is one continuous string on
359        # parameterlist. This fixes a problem where check_sections()
360        # cannot find a parameter like "addr[6 + 2]" because it actually
361        # appears as "addr[6", "+", "2]" on the parameter list.
362        # However, it's better to maintain the param string unchanged for
363        # output, so just weaken the string compare in check_sections()
364        # to ignore "[blah" in a parameter string.
365
366        self.entry.parameterlist.append(param)
367        org_arg = KernRe(r'\s\s+').sub(' ', org_arg)
368        self.entry.parametertypes[param] = org_arg
369
370    def save_struct_actual(self, actual):
371        """
372        Strip all spaces from the actual param so that it looks like
373        one string item.
374        """
375
376        actual = KernRe(r'\s*').sub("", actual, count=1)
377
378        self.entry.struct_actual += actual + " "
379
380    def create_parameter_list(self, ln, decl_type, args,
381                              splitter, declaration_name):
382        """
383        Creates a list of parameters, storing them at self.entry.
384        """
385
386        # temporarily replace all commas inside function pointer definition
387        arg_expr = KernRe(r'(\([^\),]+),')
388        while arg_expr.search(args):
389            args = arg_expr.sub(r"\1#", args)
390
391        for arg in args.split(splitter):
392            # Strip comments
393            arg = KernRe(r'\/\*.*\*\/').sub('', arg)
394
395            # Ignore argument attributes
396            arg = KernRe(r'\sPOS0?\s').sub(' ', arg)
397
398            # Strip leading/trailing spaces
399            arg = arg.strip()
400            arg = KernRe(r'\s+').sub(' ', arg, count=1)
401
402            if arg.startswith('#'):
403                # Treat preprocessor directive as a typeless variable just to fill
404                # corresponding data structures "correctly". Catch it later in
405                # output_* subs.
406
407                # Treat preprocessor directive as a typeless variable
408                self.push_parameter(ln, decl_type, arg, "",
409                                    "", declaration_name)
410
411            elif KernRe(r'\(.+\)\s*\(').search(arg):
412                # Pointer-to-function
413
414                arg = arg.replace('#', ',')
415
416                r = KernRe(r'[^\(]+\(\*?\s*([\w\[\]\.]*)\s*\)')
417                if r.match(arg):
418                    param = r.group(1)
419                else:
420                    self.emit_msg(ln, f"Invalid param: {arg}")
421                    param = arg
422
423                dtype = KernRe(r'([^\(]+\(\*?)\s*' + re.escape(param)).sub(r'\1', arg)
424                self.save_struct_actual(param)
425                self.push_parameter(ln, decl_type, param, dtype,
426                                    arg, declaration_name)
427
428            elif KernRe(r'\(.+\)\s*\[').search(arg):
429                # Array-of-pointers
430
431                arg = arg.replace('#', ',')
432                r = KernRe(r'[^\(]+\(\s*\*\s*([\w\[\]\.]*?)\s*(\s*\[\s*[\w]+\s*\]\s*)*\)')
433                if r.match(arg):
434                    param = r.group(1)
435                else:
436                    self.emit_msg(ln, f"Invalid param: {arg}")
437                    param = arg
438
439                dtype = KernRe(r'([^\(]+\(\*?)\s*' + re.escape(param)).sub(r'\1', arg)
440
441                self.save_struct_actual(param)
442                self.push_parameter(ln, decl_type, param, dtype,
443                                    arg, declaration_name)
444
445            elif arg:
446                arg = KernRe(r'\s*:\s*').sub(":", arg)
447                arg = KernRe(r'\s*\[').sub('[', arg)
448
449                args = KernRe(r'\s*,\s*').split(arg)
450                if args[0] and '*' in args[0]:
451                    args[0] = re.sub(r'(\*+)\s*', r' \1', args[0])
452
453                first_arg = []
454                r = KernRe(r'^(.*\s+)(.*?\[.*\].*)$')
455                if args[0] and r.match(args[0]):
456                    args.pop(0)
457                    first_arg.extend(r.group(1))
458                    first_arg.append(r.group(2))
459                else:
460                    first_arg = KernRe(r'\s+').split(args.pop(0))
461
462                args.insert(0, first_arg.pop())
463                dtype = ' '.join(first_arg)
464
465                for param in args:
466                    if KernRe(r'^(\*+)\s*(.*)').match(param):
467                        r = KernRe(r'^(\*+)\s*(.*)')
468                        if not r.match(param):
469                            self.emit_msg(ln, f"Invalid param: {param}")
470                            continue
471
472                        param = r.group(1)
473
474                        self.save_struct_actual(r.group(2))
475                        self.push_parameter(ln, decl_type, r.group(2),
476                                            f"{dtype} {r.group(1)}",
477                                            arg, declaration_name)
478
479                    elif KernRe(r'(.*?):(\w+)').search(param):
480                        r = KernRe(r'(.*?):(\w+)')
481                        if not r.match(param):
482                            self.emit_msg(ln, f"Invalid param: {param}")
483                            continue
484
485                        if dtype != "":  # Skip unnamed bit-fields
486                            self.save_struct_actual(r.group(1))
487                            self.push_parameter(ln, decl_type, r.group(1),
488                                                f"{dtype}:{r.group(2)}",
489                                                arg, declaration_name)
490                    else:
491                        self.save_struct_actual(param)
492                        self.push_parameter(ln, decl_type, param, dtype,
493                                            arg, declaration_name)
494
495    def check_sections(self, ln, decl_name, decl_type, sectcheck, prmscheck):
496        """
497        Check for errors inside sections, emitting warnings if not found
498        parameters are described.
499        """
500
501        sects = sectcheck.split()
502        prms = prmscheck.split()
503        err = False
504
505        for sx in range(len(sects)):                  # pylint: disable=C0200
506            err = True
507            for px in range(len(prms)):               # pylint: disable=C0200
508                prm_clean = prms[px]
509                prm_clean = KernRe(r'\[.*\]').sub('', prm_clean)
510                prm_clean = attribute.sub('', prm_clean)
511
512                # ignore array size in a parameter string;
513                # however, the original param string may contain
514                # spaces, e.g.:  addr[6 + 2]
515                # and this appears in @prms as "addr[6" since the
516                # parameter list is split at spaces;
517                # hence just ignore "[..." for the sections check;
518                prm_clean = KernRe(r'\[.*').sub('', prm_clean)
519
520                if prm_clean == sects[sx]:
521                    err = False
522                    break
523
524            if err:
525                if decl_type == 'function':
526                    dname = f"{decl_type} parameter"
527                else:
528                    dname = f"{decl_type} member"
529
530                self.emit_msg(ln,
531                              f"Excess {dname} '{sects[sx]}' description in '{decl_name}'")
532
533    def check_return_section(self, ln, declaration_name, return_type):
534        """
535        If the function doesn't return void, warns about the lack of a
536        return description.
537        """
538
539        if not self.config.wreturn:
540            return
541
542        # Ignore an empty return type (It's a macro)
543        # Ignore functions with a "void" return type (but not "void *")
544        if not return_type or KernRe(r'void\s*\w*\s*$').search(return_type):
545            return
546
547        if not self.entry.sections.get("Return", None):
548            self.emit_msg(ln,
549                          f"No description found for return value of '{declaration_name}'")
550
551    def dump_struct(self, ln, proto):
552        """
553        Store an entry for an struct or union
554        """
555
556        type_pattern = r'(struct|union)'
557
558        qualifiers = [
559            "__attribute__",
560            "__packed",
561            "__aligned",
562            "____cacheline_aligned_in_smp",
563            "____cacheline_aligned",
564        ]
565
566        definition_body = r'\{(.*)\}\s*' + "(?:" + '|'.join(qualifiers) + ")?"
567        struct_members = KernRe(type_pattern + r'([^\{\};]+)(\{)([^\{\}]*)(\})([^\{\}\;]*)(\;)')
568
569        # Extract struct/union definition
570        members = None
571        declaration_name = None
572        decl_type = None
573
574        r = KernRe(type_pattern + r'\s+(\w+)\s*' + definition_body)
575        if r.search(proto):
576            decl_type = r.group(1)
577            declaration_name = r.group(2)
578            members = r.group(3)
579        else:
580            r = KernRe(r'typedef\s+' + type_pattern + r'\s*' + definition_body + r'\s*(\w+)\s*;')
581
582            if r.search(proto):
583                decl_type = r.group(1)
584                declaration_name = r.group(3)
585                members = r.group(2)
586
587        if not members:
588            self.emit_msg(ln, f"{proto} error: Cannot parse struct or union!")
589            return
590
591        if self.entry.identifier != declaration_name:
592            self.emit_msg(ln,
593                          f"expecting prototype for {decl_type} {self.entry.identifier}. Prototype was for {decl_type} {declaration_name} instead\n")
594            return
595
596        args_pattern = r'([^,)]+)'
597
598        sub_prefixes = [
599            (KernRe(r'\/\*\s*private:.*?\/\*\s*public:.*?\*\/', re.S | re.I), ''),
600            (KernRe(r'\/\*\s*private:.*', re.S | re.I), ''),
601
602            # Strip comments
603            (KernRe(r'\/\*.*?\*\/', re.S), ''),
604
605            # Strip attributes
606            (attribute, ' '),
607            (KernRe(r'\s*__aligned\s*\([^;]*\)', re.S), ' '),
608            (KernRe(r'\s*__counted_by\s*\([^;]*\)', re.S), ' '),
609            (KernRe(r'\s*__counted_by_(le|be)\s*\([^;]*\)', re.S), ' '),
610            (KernRe(r'\s*__packed\s*', re.S), ' '),
611            (KernRe(r'\s*CRYPTO_MINALIGN_ATTR', re.S), ' '),
612            (KernRe(r'\s*____cacheline_aligned_in_smp', re.S), ' '),
613            (KernRe(r'\s*____cacheline_aligned', re.S), ' '),
614
615            # Unwrap struct_group macros based on this definition:
616            # __struct_group(TAG, NAME, ATTRS, MEMBERS...)
617            # which has variants like: struct_group(NAME, MEMBERS...)
618            # Only MEMBERS arguments require documentation.
619            #
620            # Parsing them happens on two steps:
621            #
622            # 1. drop struct group arguments that aren't at MEMBERS,
623            #    storing them as STRUCT_GROUP(MEMBERS)
624            #
625            # 2. remove STRUCT_GROUP() ancillary macro.
626            #
627            # The original logic used to remove STRUCT_GROUP() using an
628            # advanced regex:
629            #
630            #   \bSTRUCT_GROUP(\(((?:(?>[^)(]+)|(?1))*)\))[^;]*;
631            #
632            # with two patterns that are incompatible with
633            # Python re module, as it has:
634            #
635            #   - a recursive pattern: (?1)
636            #   - an atomic grouping: (?>...)
637            #
638            # I tried a simpler version: but it didn't work either:
639            #   \bSTRUCT_GROUP\(([^\)]+)\)[^;]*;
640            #
641            # As it doesn't properly match the end parenthesis on some cases.
642            #
643            # So, a better solution was crafted: there's now a NestedMatch
644            # class that ensures that delimiters after a search are properly
645            # matched. So, the implementation to drop STRUCT_GROUP() will be
646            # handled in separate.
647
648            (KernRe(r'\bstruct_group\s*\(([^,]*,)', re.S), r'STRUCT_GROUP('),
649            (KernRe(r'\bstruct_group_attr\s*\(([^,]*,){2}', re.S), r'STRUCT_GROUP('),
650            (KernRe(r'\bstruct_group_tagged\s*\(([^,]*),([^,]*),', re.S), r'struct \1 \2; STRUCT_GROUP('),
651            (KernRe(r'\b__struct_group\s*\(([^,]*,){3}', re.S), r'STRUCT_GROUP('),
652
653            # Replace macros
654            #
655            # TODO: use NestedMatch for FOO($1, $2, ...) matches
656            #
657            # it is better to also move those to the NestedMatch logic,
658            # to ensure that parenthesis will be properly matched.
659
660            (KernRe(r'__ETHTOOL_DECLARE_LINK_MODE_MASK\s*\(([^\)]+)\)', re.S), r'DECLARE_BITMAP(\1, __ETHTOOL_LINK_MODE_MASK_NBITS)'),
661            (KernRe(r'DECLARE_PHY_INTERFACE_MASK\s*\(([^\)]+)\)', re.S), r'DECLARE_BITMAP(\1, PHY_INTERFACE_MODE_MAX)'),
662            (KernRe(r'DECLARE_BITMAP\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'unsigned long \1[BITS_TO_LONGS(\2)]'),
663            (KernRe(r'DECLARE_HASHTABLE\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'unsigned long \1[1 << ((\2) - 1)]'),
664            (KernRe(r'DECLARE_KFIFO\s*\(' + args_pattern + r',\s*' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\2 *\1'),
665            (KernRe(r'DECLARE_KFIFO_PTR\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\2 *\1'),
666            (KernRe(r'(?:__)?DECLARE_FLEX_ARRAY\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\1 \2[]'),
667            (KernRe(r'DEFINE_DMA_UNMAP_ADDR\s*\(' + args_pattern + r'\)', re.S), r'dma_addr_t \1'),
668            (KernRe(r'DEFINE_DMA_UNMAP_LEN\s*\(' + args_pattern + r'\)', re.S), r'__u32 \1'),
669            (KernRe(r'VIRTIO_DECLARE_FEATURES\s*\(' + args_pattern + r'\)', re.S), r'u64 \1; u64 \1_array[VIRTIO_FEATURES_DWORDS]'),
670        ]
671
672        # Regexes here are guaranteed to have the end limiter matching
673        # the start delimiter. Yet, right now, only one replace group
674        # is allowed.
675
676        sub_nested_prefixes = [
677            (re.compile(r'\bSTRUCT_GROUP\('), r'\1'),
678        ]
679
680        for search, sub in sub_prefixes:
681            members = search.sub(sub, members)
682
683        nested = NestedMatch()
684
685        for search, sub in sub_nested_prefixes:
686            members = nested.sub(search, sub, members)
687
688        # Keeps the original declaration as-is
689        declaration = members
690
691        # Split nested struct/union elements
692        #
693        # This loop was simpler at the original kernel-doc perl version, as
694        #   while ($members =~ m/$struct_members/) { ... }
695        # reads 'members' string on each interaction.
696        #
697        # Python behavior is different: it parses 'members' only once,
698        # creating a list of tuples from the first interaction.
699        #
700        # On other words, this won't get nested structs.
701        #
702        # So, we need to have an extra loop on Python to override such
703        # re limitation.
704
705        while True:
706            tuples = struct_members.findall(members)
707            if not tuples:
708                break
709
710            for t in tuples:
711                newmember = ""
712                maintype = t[0]
713                s_ids = t[5]
714                content = t[3]
715
716                oldmember = "".join(t)
717
718                for s_id in s_ids.split(','):
719                    s_id = s_id.strip()
720
721                    newmember += f"{maintype} {s_id}; "
722                    s_id = KernRe(r'[:\[].*').sub('', s_id)
723                    s_id = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', s_id)
724
725                    for arg in content.split(';'):
726                        arg = arg.strip()
727
728                        if not arg:
729                            continue
730
731                        r = KernRe(r'^([^\(]+\(\*?\s*)([\w\.]*)(\s*\).*)')
732                        if r.match(arg):
733                            # Pointer-to-function
734                            dtype = r.group(1)
735                            name = r.group(2)
736                            extra = r.group(3)
737
738                            if not name:
739                                continue
740
741                            if not s_id:
742                                # Anonymous struct/union
743                                newmember += f"{dtype}{name}{extra}; "
744                            else:
745                                newmember += f"{dtype}{s_id}.{name}{extra}; "
746
747                        else:
748                            arg = arg.strip()
749                            # Handle bitmaps
750                            arg = KernRe(r':\s*\d+\s*').sub('', arg)
751
752                            # Handle arrays
753                            arg = KernRe(r'\[.*\]').sub('', arg)
754
755                            # Handle multiple IDs
756                            arg = KernRe(r'\s*,\s*').sub(',', arg)
757
758                            r = KernRe(r'(.*)\s+([\S+,]+)')
759
760                            if r.search(arg):
761                                dtype = r.group(1)
762                                names = r.group(2)
763                            else:
764                                newmember += f"{arg}; "
765                                continue
766
767                            for name in names.split(','):
768                                name = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', name).strip()
769
770                                if not name:
771                                    continue
772
773                                if not s_id:
774                                    # Anonymous struct/union
775                                    newmember += f"{dtype} {name}; "
776                                else:
777                                    newmember += f"{dtype} {s_id}.{name}; "
778
779                members = members.replace(oldmember, newmember)
780
781        # Ignore other nested elements, like enums
782        members = re.sub(r'(\{[^\{\}]*\})', '', members)
783
784        self.create_parameter_list(ln, decl_type, members, ';',
785                                   declaration_name)
786        self.check_sections(ln, declaration_name, decl_type,
787                            self.entry.sectcheck, self.entry.struct_actual)
788
789        # Adjust declaration for better display
790        declaration = KernRe(r'([\{;])').sub(r'\1\n', declaration)
791        declaration = KernRe(r'\}\s+;').sub('};', declaration)
792
793        # Better handle inlined enums
794        while True:
795            r = KernRe(r'(enum\s+\{[^\}]+),([^\n])')
796            if not r.search(declaration):
797                break
798
799            declaration = r.sub(r'\1,\n\2', declaration)
800
801        def_args = declaration.split('\n')
802        level = 1
803        declaration = ""
804        for clause in def_args:
805
806            clause = clause.strip()
807            clause = KernRe(r'\s+').sub(' ', clause, count=1)
808
809            if not clause:
810                continue
811
812            if '}' in clause and level > 1:
813                level -= 1
814
815            if not KernRe(r'^\s*#').match(clause):
816                declaration += "\t" * level
817
818            declaration += "\t" + clause + "\n"
819            if "{" in clause and "}" not in clause:
820                level += 1
821
822        self.output_declaration(decl_type, declaration_name,
823                                struct=declaration_name,
824                                definition=declaration,
825                                parameterlist=self.entry.parameterlist,
826                                parameterdescs=self.entry.parameterdescs,
827                                parametertypes=self.entry.parametertypes,
828                                parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
829                                sectionlist=self.entry.sectionlist,
830                                sections=self.entry.sections,
831                                section_start_lines=self.entry.section_start_lines,
832                                purpose=self.entry.declaration_purpose)
833
834    def dump_enum(self, ln, proto):
835        """
836        Stores an enum inside self.entries array.
837        """
838
839        # Ignore members marked private
840        proto = KernRe(r'\/\*\s*private:.*?\/\*\s*public:.*?\*\/', flags=re.S).sub('', proto)
841        proto = KernRe(r'\/\*\s*private:.*}', flags=re.S).sub('}', proto)
842
843        # Strip comments
844        proto = KernRe(r'\/\*.*?\*\/', flags=re.S).sub('', proto)
845
846        # Strip #define macros inside enums
847        proto = KernRe(r'#\s*((define|ifdef|if)\s+|endif)[^;]*;', flags=re.S).sub('', proto)
848
849        members = None
850        declaration_name = None
851
852        r = KernRe(r'typedef\s+enum\s*\{(.*)\}\s*(\w*)\s*;')
853        if r.search(proto):
854            declaration_name = r.group(2)
855            members = r.group(1).rstrip()
856        else:
857            r = KernRe(r'enum\s+(\w*)\s*\{(.*)\}')
858            if r.match(proto):
859                declaration_name = r.group(1)
860                members = r.group(2).rstrip()
861
862        if not members:
863            self.emit_msg(ln, f"{proto}: error: Cannot parse enum!")
864            return
865
866        if self.entry.identifier != declaration_name:
867            if self.entry.identifier == "":
868                self.emit_msg(ln,
869                              f"{proto}: wrong kernel-doc identifier on prototype")
870            else:
871                self.emit_msg(ln,
872                              f"expecting prototype for enum {self.entry.identifier}. Prototype was for enum {declaration_name} instead")
873            return
874
875        if not declaration_name:
876            declaration_name = "(anonymous)"
877
878        member_set = set()
879
880        members = KernRe(r'\([^;]*?[\)]').sub('', members)
881
882        for arg in members.split(','):
883            if not arg:
884                continue
885            arg = KernRe(r'^\s*(\w+).*').sub(r'\1', arg)
886            self.entry.parameterlist.append(arg)
887            if arg not in self.entry.parameterdescs:
888                self.entry.parameterdescs[arg] = self.undescribed
889                self.emit_msg(ln,
890                              f"Enum value '{arg}' not described in enum '{declaration_name}'")
891            member_set.add(arg)
892
893        for k in self.entry.parameterdescs:
894            if k not in member_set:
895                self.emit_msg(ln,
896                              f"Excess enum value '%{k}' description in '{declaration_name}'")
897
898        self.output_declaration('enum', declaration_name,
899                                enum=declaration_name,
900                                parameterlist=self.entry.parameterlist,
901                                parameterdescs=self.entry.parameterdescs,
902                                parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
903                                sectionlist=self.entry.sectionlist,
904                                sections=self.entry.sections,
905                                section_start_lines=self.entry.section_start_lines,
906                                purpose=self.entry.declaration_purpose)
907
908    def dump_declaration(self, ln, prototype):
909        """
910        Stores a data declaration inside self.entries array.
911        """
912
913        if self.entry.decl_type == "enum":
914            self.dump_enum(ln, prototype)
915            return
916
917        if self.entry.decl_type == "typedef":
918            self.dump_typedef(ln, prototype)
919            return
920
921        if self.entry.decl_type in ["union", "struct"]:
922            self.dump_struct(ln, prototype)
923            return
924
925        self.output_declaration(self.entry.decl_type, prototype,
926                                entry=self.entry)
927
928    def dump_function(self, ln, prototype):
929        """
930        Stores a function of function macro inside self.entries array.
931        """
932
933        func_macro = False
934        return_type = ''
935        decl_type = 'function'
936
937        # Prefixes that would be removed
938        sub_prefixes = [
939            (r"^static +", "", 0),
940            (r"^extern +", "", 0),
941            (r"^asmlinkage +", "", 0),
942            (r"^inline +", "", 0),
943            (r"^__inline__ +", "", 0),
944            (r"^__inline +", "", 0),
945            (r"^__always_inline +", "", 0),
946            (r"^noinline +", "", 0),
947            (r"^__FORTIFY_INLINE +", "", 0),
948            (r"__init +", "", 0),
949            (r"__init_or_module +", "", 0),
950            (r"__deprecated +", "", 0),
951            (r"__flatten +", "", 0),
952            (r"__meminit +", "", 0),
953            (r"__must_check +", "", 0),
954            (r"__weak +", "", 0),
955            (r"__sched +", "", 0),
956            (r"_noprof", "", 0),
957            (r"__printf\s*\(\s*\d*\s*,\s*\d*\s*\) +", "", 0),
958            (r"__(?:re)?alloc_size\s*\(\s*\d+\s*(?:,\s*\d+\s*)?\) +", "", 0),
959            (r"__diagnose_as\s*\(\s*\S+\s*(?:,\s*\d+\s*)*\) +", "", 0),
960            (r"DECL_BUCKET_PARAMS\s*\(\s*(\S+)\s*,\s*(\S+)\s*\)", r"\1, \2", 0),
961            (r"__attribute_const__ +", "", 0),
962
963            # It seems that Python support for re.X is broken:
964            # At least for me (Python 3.13), this didn't work
965#            (r"""
966#              __attribute__\s*\(\(
967#                (?:
968#                    [\w\s]+          # attribute name
969#                    (?:\([^)]*\))?   # attribute arguments
970#                    \s*,?            # optional comma at the end
971#                )+
972#              \)\)\s+
973#             """, "", re.X),
974
975            # So, remove whitespaces and comments from it
976            (r"__attribute__\s*\(\((?:[\w\s]+(?:\([^)]*\))?\s*,?)+\)\)\s+", "", 0),
977        ]
978
979        for search, sub, flags in sub_prefixes:
980            prototype = KernRe(search, flags).sub(sub, prototype)
981
982        # Macros are a special case, as they change the prototype format
983        new_proto = KernRe(r"^#\s*define\s+").sub("", prototype)
984        if new_proto != prototype:
985            is_define_proto = True
986            prototype = new_proto
987        else:
988            is_define_proto = False
989
990        # Yes, this truly is vile.  We are looking for:
991        # 1. Return type (may be nothing if we're looking at a macro)
992        # 2. Function name
993        # 3. Function parameters.
994        #
995        # All the while we have to watch out for function pointer parameters
996        # (which IIRC is what the two sections are for), C types (these
997        # regexps don't even start to express all the possibilities), and
998        # so on.
999        #
1000        # If you mess with these regexps, it's a good idea to check that
1001        # the following functions' documentation still comes out right:
1002        # - parport_register_device (function pointer parameters)
1003        # - atomic_set (macro)
1004        # - pci_match_device, __copy_to_user (long return type)
1005
1006        name = r'[a-zA-Z0-9_~:]+'
1007        prototype_end1 = r'[^\(]*'
1008        prototype_end2 = r'[^\{]*'
1009        prototype_end = fr'\(({prototype_end1}|{prototype_end2})\)'
1010
1011        # Besides compiling, Perl qr{[\w\s]+} works as a non-capturing group.
1012        # So, this needs to be mapped in Python with (?:...)? or (?:...)+
1013
1014        type1 = r'(?:[\w\s]+)?'
1015        type2 = r'(?:[\w\s]+\*+)+'
1016
1017        found = False
1018
1019        if is_define_proto:
1020            r = KernRe(r'^()(' + name + r')\s+')
1021
1022            if r.search(prototype):
1023                return_type = ''
1024                declaration_name = r.group(2)
1025                func_macro = True
1026
1027                found = True
1028
1029        if not found:
1030            patterns = [
1031                rf'^()({name})\s*{prototype_end}',
1032                rf'^({type1})\s+({name})\s*{prototype_end}',
1033                rf'^({type2})\s*({name})\s*{prototype_end}',
1034            ]
1035
1036            for p in patterns:
1037                r = KernRe(p)
1038
1039                if r.match(prototype):
1040
1041                    return_type = r.group(1)
1042                    declaration_name = r.group(2)
1043                    args = r.group(3)
1044
1045                    self.create_parameter_list(ln, decl_type, args, ',',
1046                                               declaration_name)
1047
1048                    found = True
1049                    break
1050        if not found:
1051            self.emit_msg(ln,
1052                          f"cannot understand function prototype: '{prototype}'")
1053            return
1054
1055        if self.entry.identifier != declaration_name:
1056            self.emit_msg(ln,
1057                          f"expecting prototype for {self.entry.identifier}(). Prototype was for {declaration_name}() instead")
1058            return
1059
1060        prms = " ".join(self.entry.parameterlist)
1061        self.check_sections(ln, declaration_name, "function",
1062                            self.entry.sectcheck, prms)
1063
1064        self.check_return_section(ln, declaration_name, return_type)
1065
1066        if 'typedef' in return_type:
1067            self.output_declaration(decl_type, declaration_name,
1068                                    function=declaration_name,
1069                                    typedef=True,
1070                                    functiontype=return_type,
1071                                    parameterlist=self.entry.parameterlist,
1072                                    parameterdescs=self.entry.parameterdescs,
1073                                    parametertypes=self.entry.parametertypes,
1074                                    parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
1075                                    sectionlist=self.entry.sectionlist,
1076                                    sections=self.entry.sections,
1077                                    section_start_lines=self.entry.section_start_lines,
1078                                    purpose=self.entry.declaration_purpose,
1079                                    func_macro=func_macro)
1080        else:
1081            self.output_declaration(decl_type, declaration_name,
1082                                    function=declaration_name,
1083                                    typedef=False,
1084                                    functiontype=return_type,
1085                                    parameterlist=self.entry.parameterlist,
1086                                    parameterdescs=self.entry.parameterdescs,
1087                                    parametertypes=self.entry.parametertypes,
1088                                    parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
1089                                    sectionlist=self.entry.sectionlist,
1090                                    sections=self.entry.sections,
1091                                    section_start_lines=self.entry.section_start_lines,
1092                                    purpose=self.entry.declaration_purpose,
1093                                    func_macro=func_macro)
1094
1095    def dump_typedef(self, ln, proto):
1096        """
1097        Stores a typedef inside self.entries array.
1098        """
1099
1100        typedef_type = r'((?:\s+[\w\*]+\b){0,7}\s+(?:\w+\b|\*+))\s*'
1101        typedef_ident = r'\*?\s*(\w\S+)\s*'
1102        typedef_args = r'\s*\((.*)\);'
1103
1104        typedef1 = KernRe(r'typedef' + typedef_type + r'\(' + typedef_ident + r'\)' + typedef_args)
1105        typedef2 = KernRe(r'typedef' + typedef_type + typedef_ident + typedef_args)
1106
1107        # Strip comments
1108        proto = KernRe(r'/\*.*?\*/', flags=re.S).sub('', proto)
1109
1110        # Parse function typedef prototypes
1111        for r in [typedef1, typedef2]:
1112            if not r.match(proto):
1113                continue
1114
1115            return_type = r.group(1).strip()
1116            declaration_name = r.group(2)
1117            args = r.group(3)
1118
1119            if self.entry.identifier != declaration_name:
1120                self.emit_msg(ln,
1121                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1122                return
1123
1124            decl_type = 'function'
1125            self.create_parameter_list(ln, decl_type, args, ',', declaration_name)
1126
1127            self.output_declaration(decl_type, declaration_name,
1128                                    function=declaration_name,
1129                                    typedef=True,
1130                                    functiontype=return_type,
1131                                    parameterlist=self.entry.parameterlist,
1132                                    parameterdescs=self.entry.parameterdescs,
1133                                    parametertypes=self.entry.parametertypes,
1134                                    parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
1135                                    sectionlist=self.entry.sectionlist,
1136                                    sections=self.entry.sections,
1137                                    section_start_lines=self.entry.section_start_lines,
1138                                    purpose=self.entry.declaration_purpose)
1139            return
1140
1141        # Handle nested parentheses or brackets
1142        r = KernRe(r'(\(*.\)\s*|\[*.\]\s*);$')
1143        while r.search(proto):
1144            proto = r.sub('', proto)
1145
1146        # Parse simple typedefs
1147        r = KernRe(r'typedef.*\s+(\w+)\s*;')
1148        if r.match(proto):
1149            declaration_name = r.group(1)
1150
1151            if self.entry.identifier != declaration_name:
1152                self.emit_msg(ln,
1153                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1154                return
1155
1156            self.output_declaration('typedef', declaration_name,
1157                                    typedef=declaration_name,
1158                                    sectionlist=self.entry.sectionlist,
1159                                    sections=self.entry.sections,
1160                                    section_start_lines=self.entry.section_start_lines,
1161                                    purpose=self.entry.declaration_purpose)
1162            return
1163
1164        self.emit_msg(ln, "error: Cannot parse typedef!")
1165
1166    @staticmethod
1167    def process_export(function_set, line):
1168        """
1169        process EXPORT_SYMBOL* tags
1170
1171        This method doesn't use any variable from the class, so declare it
1172        with a staticmethod decorator.
1173        """
1174
1175        # Note: it accepts only one EXPORT_SYMBOL* per line, as having
1176        # multiple export lines would violate Kernel coding style.
1177
1178        if export_symbol.search(line):
1179            symbol = export_symbol.group(2)
1180            function_set.add(symbol)
1181            return
1182
1183        if export_symbol_ns.search(line):
1184            symbol = export_symbol_ns.group(2)
1185            function_set.add(symbol)
1186
1187    def process_normal(self, ln, line):
1188        """
1189        STATE_NORMAL: looking for the /** to begin everything.
1190        """
1191
1192        if not doc_start.match(line):
1193            return
1194
1195        # start a new entry
1196        self.reset_state(ln)
1197        self.entry.in_doc_sect = False
1198
1199        # next line is always the function name
1200        self.state = state.NAME
1201
1202    def process_name(self, ln, line):
1203        """
1204        STATE_NAME: Looking for the "name - description" line
1205        """
1206
1207        if doc_block.search(line):
1208            self.entry.new_start_line = ln
1209
1210            if not doc_block.group(1):
1211                self.entry.section = self.section_intro
1212            else:
1213                self.entry.section = doc_block.group(1)
1214
1215            self.entry.identifier = self.entry.section
1216            self.state = state.DOCBLOCK
1217            return
1218
1219        if doc_decl.search(line):
1220            self.entry.identifier = doc_decl.group(1)
1221            self.entry.is_kernel_comment = False
1222
1223            decl_start = str(doc_com)       # comment block asterisk
1224            fn_type = r"(?:\w+\s*\*\s*)?"  # type (for non-functions)
1225            parenthesis = r"(?:\(\w*\))?"   # optional parenthesis on function
1226            decl_end = r"(?:[-:].*)"         # end of the name part
1227
1228            # test for pointer declaration type, foo * bar() - desc
1229            r = KernRe(fr"^{decl_start}([\w\s]+?){parenthesis}?\s*{decl_end}?$")
1230            if r.search(line):
1231                self.entry.identifier = r.group(1)
1232
1233            # Test for data declaration
1234            r = KernRe(r"^\s*\*?\s*(struct|union|enum|typedef)\b\s*(\w*)")
1235            if r.search(line):
1236                self.entry.decl_type = r.group(1)
1237                self.entry.identifier = r.group(2)
1238                self.entry.is_kernel_comment = True
1239            else:
1240                # Look for foo() or static void foo() - description;
1241                # or misspelt identifier
1242
1243                r1 = KernRe(fr"^{decl_start}{fn_type}(\w+)\s*{parenthesis}\s*{decl_end}?$")
1244                r2 = KernRe(fr"^{decl_start}{fn_type}(\w+[^-:]*){parenthesis}\s*{decl_end}$")
1245
1246                for r in [r1, r2]:
1247                    if r.search(line):
1248                        self.entry.identifier = r.group(1)
1249                        self.entry.decl_type = "function"
1250
1251                        r = KernRe(r"define\s+")
1252                        self.entry.identifier = r.sub("", self.entry.identifier)
1253                        self.entry.is_kernel_comment = True
1254                        break
1255
1256            self.entry.identifier = self.entry.identifier.strip(" ")
1257
1258            self.state = state.BODY
1259
1260            # if there's no @param blocks need to set up default section here
1261            self.entry.section = SECTION_DEFAULT
1262            self.entry.new_start_line = ln + 1
1263
1264            r = KernRe("[-:](.*)")
1265            if r.search(line):
1266                # strip leading/trailing/multiple spaces
1267                self.entry.descr = r.group(1).strip(" ")
1268
1269                r = KernRe(r"\s+")
1270                self.entry.descr = r.sub(" ", self.entry.descr)
1271                self.entry.declaration_purpose = self.entry.descr
1272                self.state = state.BODY_MAYBE
1273            else:
1274                self.entry.declaration_purpose = ""
1275
1276            if not self.entry.is_kernel_comment:
1277                self.emit_msg(ln,
1278                              f"This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst\n{line}")
1279                self.state = state.NORMAL
1280
1281            if not self.entry.declaration_purpose and self.config.wshort_desc:
1282                self.emit_msg(ln,
1283                              f"missing initial short description on line:\n{line}")
1284
1285            if not self.entry.identifier and self.entry.decl_type != "enum":
1286                self.emit_msg(ln,
1287                              f"wrong kernel-doc identifier on line:\n{line}")
1288                self.state = state.NORMAL
1289
1290            if self.config.verbose:
1291                self.emit_msg(ln,
1292                              f"Scanning doc for {self.entry.decl_type} {self.entry.identifier}",
1293                                  warning=False)
1294
1295            return
1296
1297        # Failed to find an identifier. Emit a warning
1298        self.emit_msg(ln, f"Cannot find identifier on line:\n{line}")
1299
1300    def process_body(self, ln, line):
1301        """
1302        STATE_BODY and STATE_BODY_MAYBE: the bulk of a kerneldoc comment.
1303        """
1304
1305        if self.state == state.BODY_WITH_BLANK_LINE:
1306            r = KernRe(r"\s*\*\s?\S")
1307            if r.match(line):
1308                self.dump_section()
1309                self.entry.section = SECTION_DEFAULT
1310                self.entry.new_start_line = ln
1311                self.entry.contents = ""
1312
1313        if doc_sect.search(line):
1314            self.entry.in_doc_sect = True
1315            newsection = doc_sect.group(1)
1316
1317            if newsection.lower() in ["description", "context"]:
1318                newsection = newsection.title()
1319
1320            # Special case: @return is a section, not a param description
1321            if newsection.lower() in ["@return", "@returns",
1322                                      "return", "returns"]:
1323                newsection = "Return"
1324
1325            # Perl kernel-doc has a check here for contents before sections.
1326            # the logic there is always false, as in_doc_sect variable is
1327            # always true. So, just don't implement Wcontents_before_sections
1328
1329            # .title()
1330            newcontents = doc_sect.group(2)
1331            if not newcontents:
1332                newcontents = ""
1333
1334            if self.entry.contents.strip("\n"):
1335                self.dump_section()
1336
1337            self.entry.new_start_line = ln
1338            self.entry.section = newsection
1339            self.entry.leading_space = None
1340
1341            self.entry.contents = newcontents.lstrip()
1342            if self.entry.contents:
1343                self.entry.contents += "\n"
1344
1345            self.state = state.BODY
1346            return
1347
1348        if doc_end.search(line):
1349            self.dump_section()
1350
1351            # Look for doc_com + <text> + doc_end:
1352            r = KernRe(r'\s*\*\s*[a-zA-Z_0-9:\.]+\*/')
1353            if r.match(line):
1354                self.emit_msg(ln, f"suspicious ending line: {line}")
1355
1356            self.entry.prototype = ""
1357            self.entry.new_start_line = ln + 1
1358
1359            self.state = state.PROTO
1360            return
1361
1362        if doc_content.search(line):
1363            cont = doc_content.group(1)
1364
1365            if cont == "":
1366                if self.entry.section == self.section_context:
1367                    self.dump_section()
1368
1369                    self.entry.new_start_line = ln
1370                    self.state = state.BODY
1371                else:
1372                    if self.entry.section != SECTION_DEFAULT:
1373                        self.state = state.BODY_WITH_BLANK_LINE
1374                    else:
1375                        self.state = state.BODY
1376
1377                    self.entry.contents += "\n"
1378
1379            elif self.state == state.BODY_MAYBE:
1380
1381                # Continued declaration purpose
1382                self.entry.declaration_purpose = self.entry.declaration_purpose.rstrip()
1383                self.entry.declaration_purpose += " " + cont
1384
1385                r = KernRe(r"\s+")
1386                self.entry.declaration_purpose = r.sub(' ',
1387                                                       self.entry.declaration_purpose)
1388
1389            else:
1390                if self.entry.section.startswith('@') or        \
1391                   self.entry.section == self.section_context:
1392                    if self.entry.leading_space is None:
1393                        r = KernRe(r'^(\s+)')
1394                        if r.match(cont):
1395                            self.entry.leading_space = len(r.group(1))
1396                        else:
1397                            self.entry.leading_space = 0
1398
1399                    # Double-check if leading space are realy spaces
1400                    pos = 0
1401                    for i in range(0, self.entry.leading_space):
1402                        if cont[i] != " ":
1403                            break
1404                        pos += 1
1405
1406                    cont = cont[pos:]
1407
1408                    # NEW LOGIC:
1409                    # In case it is different, update it
1410                    if self.entry.leading_space != pos:
1411                        self.entry.leading_space = pos
1412
1413                self.entry.contents += cont + "\n"
1414            return
1415
1416        # Unknown line, ignore
1417        self.emit_msg(ln, f"bad line: {line}")
1418
1419    def process_inline(self, ln, line):
1420        """STATE_INLINE: docbook comments within a prototype."""
1421
1422        if self.inline_doc_state == state.INLINE_NAME and \
1423           doc_inline_sect.search(line):
1424            self.entry.section = doc_inline_sect.group(1)
1425            self.entry.new_start_line = ln
1426
1427            self.entry.contents = doc_inline_sect.group(2).lstrip()
1428            if self.entry.contents != "":
1429                self.entry.contents += "\n"
1430
1431            self.inline_doc_state = state.INLINE_TEXT
1432            # Documentation block end */
1433            return
1434
1435        if doc_inline_end.search(line):
1436            if self.entry.contents not in ["", "\n"]:
1437                self.dump_section()
1438
1439            self.state = state.PROTO
1440            self.inline_doc_state = state.INLINE_NA
1441            return
1442
1443        if doc_content.search(line):
1444            if self.inline_doc_state == state.INLINE_TEXT:
1445                self.entry.contents += doc_content.group(1) + "\n"
1446                if not self.entry.contents.strip(" ").rstrip("\n"):
1447                    self.entry.contents = ""
1448
1449            elif self.inline_doc_state == state.INLINE_NAME:
1450                self.emit_msg(ln,
1451                              f"Incorrect use of kernel-doc format: {line}")
1452
1453                self.inline_doc_state = state.INLINE_ERROR
1454
1455    def syscall_munge(self, ln, proto):         # pylint: disable=W0613
1456        """
1457        Handle syscall definitions
1458        """
1459
1460        is_void = False
1461
1462        # Strip newlines/CR's
1463        proto = re.sub(r'[\r\n]+', ' ', proto)
1464
1465        # Check if it's a SYSCALL_DEFINE0
1466        if 'SYSCALL_DEFINE0' in proto:
1467            is_void = True
1468
1469        # Replace SYSCALL_DEFINE with correct return type & function name
1470        proto = KernRe(r'SYSCALL_DEFINE.*\(').sub('long sys_', proto)
1471
1472        r = KernRe(r'long\s+(sys_.*?),')
1473        if r.search(proto):
1474            proto = KernRe(',').sub('(', proto, count=1)
1475        elif is_void:
1476            proto = KernRe(r'\)').sub('(void)', proto, count=1)
1477
1478        # Now delete all of the odd-numbered commas in the proto
1479        # so that argument types & names don't have a comma between them
1480        count = 0
1481        length = len(proto)
1482
1483        if is_void:
1484            length = 0  # skip the loop if is_void
1485
1486        for ix in range(length):
1487            if proto[ix] == ',':
1488                count += 1
1489                if count % 2 == 1:
1490                    proto = proto[:ix] + ' ' + proto[ix + 1:]
1491
1492        return proto
1493
1494    def tracepoint_munge(self, ln, proto):
1495        """
1496        Handle tracepoint definitions
1497        """
1498
1499        tracepointname = None
1500        tracepointargs = None
1501
1502        # Match tracepoint name based on different patterns
1503        r = KernRe(r'TRACE_EVENT\((.*?),')
1504        if r.search(proto):
1505            tracepointname = r.group(1)
1506
1507        r = KernRe(r'DEFINE_SINGLE_EVENT\((.*?),')
1508        if r.search(proto):
1509            tracepointname = r.group(1)
1510
1511        r = KernRe(r'DEFINE_EVENT\((.*?),(.*?),')
1512        if r.search(proto):
1513            tracepointname = r.group(2)
1514
1515        if tracepointname:
1516            tracepointname = tracepointname.lstrip()
1517
1518        r = KernRe(r'TP_PROTO\((.*?)\)')
1519        if r.search(proto):
1520            tracepointargs = r.group(1)
1521
1522        if not tracepointname or not tracepointargs:
1523            self.emit_msg(ln,
1524                          f"Unrecognized tracepoint format:\n{proto}\n")
1525        else:
1526            proto = f"static inline void trace_{tracepointname}({tracepointargs})"
1527            self.entry.identifier = f"trace_{self.entry.identifier}"
1528
1529        return proto
1530
1531    def process_proto_function(self, ln, line):
1532        """Ancillary routine to process a function prototype"""
1533
1534        # strip C99-style comments to end of line
1535        r = KernRe(r"\/\/.*$", re.S)
1536        line = r.sub('', line)
1537
1538        if KernRe(r'\s*#\s*define').match(line):
1539            self.entry.prototype = line
1540        elif line.startswith('#'):
1541            # Strip other macros like #ifdef/#ifndef/#endif/...
1542            pass
1543        else:
1544            r = KernRe(r'([^\{]*)')
1545            if r.match(line):
1546                self.entry.prototype += r.group(1) + " "
1547
1548        if '{' in line or ';' in line or KernRe(r'\s*#\s*define').match(line):
1549            # strip comments
1550            r = KernRe(r'/\*.*?\*/')
1551            self.entry.prototype = r.sub('', self.entry.prototype)
1552
1553            # strip newlines/cr's
1554            r = KernRe(r'[\r\n]+')
1555            self.entry.prototype = r.sub(' ', self.entry.prototype)
1556
1557            # strip leading spaces
1558            r = KernRe(r'^\s+')
1559            self.entry.prototype = r.sub('', self.entry.prototype)
1560
1561            # Handle self.entry.prototypes for function pointers like:
1562            #       int (*pcs_config)(struct foo)
1563
1564            r = KernRe(r'^(\S+\s+)\(\s*\*(\S+)\)')
1565            self.entry.prototype = r.sub(r'\1\2', self.entry.prototype)
1566
1567            if 'SYSCALL_DEFINE' in self.entry.prototype:
1568                self.entry.prototype = self.syscall_munge(ln,
1569                                                          self.entry.prototype)
1570
1571            r = KernRe(r'TRACE_EVENT|DEFINE_EVENT|DEFINE_SINGLE_EVENT')
1572            if r.search(self.entry.prototype):
1573                self.entry.prototype = self.tracepoint_munge(ln,
1574                                                             self.entry.prototype)
1575
1576            self.dump_function(ln, self.entry.prototype)
1577            self.reset_state(ln)
1578
1579    def process_proto_type(self, ln, line):
1580        """Ancillary routine to process a type"""
1581
1582        # Strip newlines/cr's.
1583        line = KernRe(r'[\r\n]+', re.S).sub(' ', line)
1584
1585        # Strip leading spaces
1586        line = KernRe(r'^\s+', re.S).sub('', line)
1587
1588        # Strip trailing spaces
1589        line = KernRe(r'\s+$', re.S).sub('', line)
1590
1591        # Strip C99-style comments to the end of the line
1592        line = KernRe(r"\/\/.*$", re.S).sub('', line)
1593
1594        # To distinguish preprocessor directive from regular declaration later.
1595        if line.startswith('#'):
1596            line += ";"
1597
1598        r = KernRe(r'([^\{\};]*)([\{\};])(.*)')
1599        while True:
1600            if r.search(line):
1601                if self.entry.prototype:
1602                    self.entry.prototype += " "
1603                self.entry.prototype += r.group(1) + r.group(2)
1604
1605                self.entry.brcount += r.group(2).count('{')
1606                self.entry.brcount -= r.group(2).count('}')
1607
1608                self.entry.brcount = max(self.entry.brcount, 0)
1609
1610                if r.group(2) == ';' and self.entry.brcount == 0:
1611                    self.dump_declaration(ln, self.entry.prototype)
1612                    self.reset_state(ln)
1613                    break
1614
1615                line = r.group(3)
1616            else:
1617                self.entry.prototype += line
1618                break
1619
1620    def process_proto(self, ln, line):
1621        """STATE_PROTO: reading a function/whatever prototype."""
1622
1623        if doc_inline_oneline.search(line):
1624            self.entry.section = doc_inline_oneline.group(1)
1625            self.entry.contents = doc_inline_oneline.group(2)
1626
1627            if self.entry.contents != "":
1628                self.entry.contents += "\n"
1629                self.dump_section(start_new=False)
1630
1631        elif doc_inline_start.search(line):
1632            self.state = state.INLINE
1633            self.inline_doc_state = state.INLINE_NAME
1634
1635        elif self.entry.decl_type == 'function':
1636            self.process_proto_function(ln, line)
1637
1638        else:
1639            self.process_proto_type(ln, line)
1640
1641    def process_docblock(self, ln, line):
1642        """STATE_DOCBLOCK: within a DOC: block."""
1643
1644        if doc_end.search(line):
1645            self.dump_section()
1646            self.output_declaration("doc", self.entry.identifier,
1647                                    sectionlist=self.entry.sectionlist,
1648                                    sections=self.entry.sections,
1649                                    section_start_lines=self.entry.section_start_lines)
1650            self.reset_state(ln)
1651
1652        elif doc_content.search(line):
1653            self.entry.contents += doc_content.group(1) + "\n"
1654
1655    def parse_export(self):
1656        """
1657        Parses EXPORT_SYMBOL* macros from a single Kernel source file.
1658        """
1659
1660        export_table = set()
1661
1662        try:
1663            with open(self.fname, "r", encoding="utf8",
1664                      errors="backslashreplace") as fp:
1665
1666                for line in fp:
1667                    self.process_export(export_table, line)
1668
1669        except IOError:
1670            return None
1671
1672        return export_table
1673
1674    def parse_kdoc(self):
1675        """
1676        Open and process each line of a C source file.
1677        The parsing is controlled via a state machine, and the line is passed
1678        to a different process function depending on the state. The process
1679        function may update the state as needed.
1680
1681        Besides parsing kernel-doc tags, it also parses export symbols.
1682        """
1683
1684        cont = False
1685        prev = ""
1686        prev_ln = None
1687        export_table = set()
1688
1689        try:
1690            with open(self.fname, "r", encoding="utf8",
1691                      errors="backslashreplace") as fp:
1692                for ln, line in enumerate(fp):
1693
1694                    line = line.expandtabs().strip("\n")
1695
1696                    # Group continuation lines on prototypes
1697                    if self.state == state.PROTO:
1698                        if line.endswith("\\"):
1699                            prev += line.rstrip("\\")
1700                            cont = True
1701
1702                            if not prev_ln:
1703                                prev_ln = ln
1704
1705                            continue
1706
1707                        if cont:
1708                            ln = prev_ln
1709                            line = prev + line
1710                            prev = ""
1711                            cont = False
1712                            prev_ln = None
1713
1714                    self.config.log.debug("%d %s%s: %s",
1715                                          ln, state.name[self.state],
1716                                          state.inline_name[self.inline_doc_state],
1717                                          line)
1718
1719                    # This is an optimization over the original script.
1720                    # There, when export_file was used for the same file,
1721                    # it was read twice. Here, we use the already-existing
1722                    # loop to parse exported symbols as well.
1723                    #
1724                    # TODO: It should be noticed that not all states are
1725                    # needed here. On a future cleanup, process export only
1726                    # at the states that aren't handling comment markups.
1727                    self.process_export(export_table, line)
1728
1729                    # Hand this line to the appropriate state handler
1730                    if self.state == state.NORMAL:
1731                        self.process_normal(ln, line)
1732                    elif self.state == state.NAME:
1733                        self.process_name(ln, line)
1734                    elif self.state in [state.BODY, state.BODY_MAYBE,
1735                                        state.BODY_WITH_BLANK_LINE]:
1736                        self.process_body(ln, line)
1737                    elif self.state == state.INLINE:  # scanning for inline parameters
1738                        self.process_inline(ln, line)
1739                    elif self.state == state.PROTO:
1740                        self.process_proto(ln, line)
1741                    elif self.state == state.DOCBLOCK:
1742                        self.process_docblock(ln, line)
1743        except OSError:
1744            self.config.log.error(f"Error: Cannot open file {self.fname}")
1745
1746        return export_table, self.entries
1747