xref: /linux/scripts/lib/kdoc/kdoc_parser.py (revision 3e443d167327b10966166c1953631936547b03d0)
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        ]
670
671        # Regexes here are guaranteed to have the end limiter matching
672        # the start delimiter. Yet, right now, only one replace group
673        # is allowed.
674
675        sub_nested_prefixes = [
676            (re.compile(r'\bSTRUCT_GROUP\('), r'\1'),
677        ]
678
679        for search, sub in sub_prefixes:
680            members = search.sub(sub, members)
681
682        nested = NestedMatch()
683
684        for search, sub in sub_nested_prefixes:
685            members = nested.sub(search, sub, members)
686
687        # Keeps the original declaration as-is
688        declaration = members
689
690        # Split nested struct/union elements
691        #
692        # This loop was simpler at the original kernel-doc perl version, as
693        #   while ($members =~ m/$struct_members/) { ... }
694        # reads 'members' string on each interaction.
695        #
696        # Python behavior is different: it parses 'members' only once,
697        # creating a list of tuples from the first interaction.
698        #
699        # On other words, this won't get nested structs.
700        #
701        # So, we need to have an extra loop on Python to override such
702        # re limitation.
703
704        while True:
705            tuples = struct_members.findall(members)
706            if not tuples:
707                break
708
709            for t in tuples:
710                newmember = ""
711                maintype = t[0]
712                s_ids = t[5]
713                content = t[3]
714
715                oldmember = "".join(t)
716
717                for s_id in s_ids.split(','):
718                    s_id = s_id.strip()
719
720                    newmember += f"{maintype} {s_id}; "
721                    s_id = KernRe(r'[:\[].*').sub('', s_id)
722                    s_id = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', s_id)
723
724                    for arg in content.split(';'):
725                        arg = arg.strip()
726
727                        if not arg:
728                            continue
729
730                        r = KernRe(r'^([^\(]+\(\*?\s*)([\w\.]*)(\s*\).*)')
731                        if r.match(arg):
732                            # Pointer-to-function
733                            dtype = r.group(1)
734                            name = r.group(2)
735                            extra = r.group(3)
736
737                            if not name:
738                                continue
739
740                            if not s_id:
741                                # Anonymous struct/union
742                                newmember += f"{dtype}{name}{extra}; "
743                            else:
744                                newmember += f"{dtype}{s_id}.{name}{extra}; "
745
746                        else:
747                            arg = arg.strip()
748                            # Handle bitmaps
749                            arg = KernRe(r':\s*\d+\s*').sub('', arg)
750
751                            # Handle arrays
752                            arg = KernRe(r'\[.*\]').sub('', arg)
753
754                            # Handle multiple IDs
755                            arg = KernRe(r'\s*,\s*').sub(',', arg)
756
757                            r = KernRe(r'(.*)\s+([\S+,]+)')
758
759                            if r.search(arg):
760                                dtype = r.group(1)
761                                names = r.group(2)
762                            else:
763                                newmember += f"{arg}; "
764                                continue
765
766                            for name in names.split(','):
767                                name = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', name).strip()
768
769                                if not name:
770                                    continue
771
772                                if not s_id:
773                                    # Anonymous struct/union
774                                    newmember += f"{dtype} {name}; "
775                                else:
776                                    newmember += f"{dtype} {s_id}.{name}; "
777
778                members = members.replace(oldmember, newmember)
779
780        # Ignore other nested elements, like enums
781        members = re.sub(r'(\{[^\{\}]*\})', '', members)
782
783        self.create_parameter_list(ln, decl_type, members, ';',
784                                   declaration_name)
785        self.check_sections(ln, declaration_name, decl_type,
786                            self.entry.sectcheck, self.entry.struct_actual)
787
788        # Adjust declaration for better display
789        declaration = KernRe(r'([\{;])').sub(r'\1\n', declaration)
790        declaration = KernRe(r'\}\s+;').sub('};', declaration)
791
792        # Better handle inlined enums
793        while True:
794            r = KernRe(r'(enum\s+\{[^\}]+),([^\n])')
795            if not r.search(declaration):
796                break
797
798            declaration = r.sub(r'\1,\n\2', declaration)
799
800        def_args = declaration.split('\n')
801        level = 1
802        declaration = ""
803        for clause in def_args:
804
805            clause = clause.strip()
806            clause = KernRe(r'\s+').sub(' ', clause, count=1)
807
808            if not clause:
809                continue
810
811            if '}' in clause and level > 1:
812                level -= 1
813
814            if not KernRe(r'^\s*#').match(clause):
815                declaration += "\t" * level
816
817            declaration += "\t" + clause + "\n"
818            if "{" in clause and "}" not in clause:
819                level += 1
820
821        self.output_declaration(decl_type, declaration_name,
822                                struct=declaration_name,
823                                definition=declaration,
824                                parameterlist=self.entry.parameterlist,
825                                parameterdescs=self.entry.parameterdescs,
826                                parametertypes=self.entry.parametertypes,
827                                parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
828                                sectionlist=self.entry.sectionlist,
829                                sections=self.entry.sections,
830                                section_start_lines=self.entry.section_start_lines,
831                                purpose=self.entry.declaration_purpose)
832
833    def dump_enum(self, ln, proto):
834        """
835        Stores an enum inside self.entries array.
836        """
837
838        # Ignore members marked private
839        proto = KernRe(r'\/\*\s*private:.*?\/\*\s*public:.*?\*\/', flags=re.S).sub('', proto)
840        proto = KernRe(r'\/\*\s*private:.*}', flags=re.S).sub('}', proto)
841
842        # Strip comments
843        proto = KernRe(r'\/\*.*?\*\/', flags=re.S).sub('', proto)
844
845        # Strip #define macros inside enums
846        proto = KernRe(r'#\s*((define|ifdef|if)\s+|endif)[^;]*;', flags=re.S).sub('', proto)
847
848        members = None
849        declaration_name = None
850
851        r = KernRe(r'typedef\s+enum\s*\{(.*)\}\s*(\w*)\s*;')
852        if r.search(proto):
853            declaration_name = r.group(2)
854            members = r.group(1).rstrip()
855        else:
856            r = KernRe(r'enum\s+(\w*)\s*\{(.*)\}')
857            if r.match(proto):
858                declaration_name = r.group(1)
859                members = r.group(2).rstrip()
860
861        if not members:
862            self.emit_msg(ln, f"{proto}: error: Cannot parse enum!")
863            return
864
865        if self.entry.identifier != declaration_name:
866            if self.entry.identifier == "":
867                self.emit_msg(ln,
868                              f"{proto}: wrong kernel-doc identifier on prototype")
869            else:
870                self.emit_msg(ln,
871                              f"expecting prototype for enum {self.entry.identifier}. Prototype was for enum {declaration_name} instead")
872            return
873
874        if not declaration_name:
875            declaration_name = "(anonymous)"
876
877        member_set = set()
878
879        members = KernRe(r'\([^;]*?[\)]').sub('', members)
880
881        for arg in members.split(','):
882            if not arg:
883                continue
884            arg = KernRe(r'^\s*(\w+).*').sub(r'\1', arg)
885            self.entry.parameterlist.append(arg)
886            if arg not in self.entry.parameterdescs:
887                self.entry.parameterdescs[arg] = self.undescribed
888                self.emit_msg(ln,
889                              f"Enum value '{arg}' not described in enum '{declaration_name}'")
890            member_set.add(arg)
891
892        for k in self.entry.parameterdescs:
893            if k not in member_set:
894                self.emit_msg(ln,
895                              f"Excess enum value '%{k}' description in '{declaration_name}'")
896
897        self.output_declaration('enum', declaration_name,
898                                enum=declaration_name,
899                                parameterlist=self.entry.parameterlist,
900                                parameterdescs=self.entry.parameterdescs,
901                                parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
902                                sectionlist=self.entry.sectionlist,
903                                sections=self.entry.sections,
904                                section_start_lines=self.entry.section_start_lines,
905                                purpose=self.entry.declaration_purpose)
906
907    def dump_declaration(self, ln, prototype):
908        """
909        Stores a data declaration inside self.entries array.
910        """
911
912        if self.entry.decl_type == "enum":
913            self.dump_enum(ln, prototype)
914            return
915
916        if self.entry.decl_type == "typedef":
917            self.dump_typedef(ln, prototype)
918            return
919
920        if self.entry.decl_type in ["union", "struct"]:
921            self.dump_struct(ln, prototype)
922            return
923
924        self.output_declaration(self.entry.decl_type, prototype,
925                                entry=self.entry)
926
927    def dump_function(self, ln, prototype):
928        """
929        Stores a function of function macro inside self.entries array.
930        """
931
932        func_macro = False
933        return_type = ''
934        decl_type = 'function'
935
936        # Prefixes that would be removed
937        sub_prefixes = [
938            (r"^static +", "", 0),
939            (r"^extern +", "", 0),
940            (r"^asmlinkage +", "", 0),
941            (r"^inline +", "", 0),
942            (r"^__inline__ +", "", 0),
943            (r"^__inline +", "", 0),
944            (r"^__always_inline +", "", 0),
945            (r"^noinline +", "", 0),
946            (r"^__FORTIFY_INLINE +", "", 0),
947            (r"__init +", "", 0),
948            (r"__init_or_module +", "", 0),
949            (r"__deprecated +", "", 0),
950            (r"__flatten +", "", 0),
951            (r"__meminit +", "", 0),
952            (r"__must_check +", "", 0),
953            (r"__weak +", "", 0),
954            (r"__sched +", "", 0),
955            (r"_noprof", "", 0),
956            (r"__printf\s*\(\s*\d*\s*,\s*\d*\s*\) +", "", 0),
957            (r"__(?:re)?alloc_size\s*\(\s*\d+\s*(?:,\s*\d+\s*)?\) +", "", 0),
958            (r"__diagnose_as\s*\(\s*\S+\s*(?:,\s*\d+\s*)*\) +", "", 0),
959            (r"DECL_BUCKET_PARAMS\s*\(\s*(\S+)\s*,\s*(\S+)\s*\)", r"\1, \2", 0),
960            (r"__attribute_const__ +", "", 0),
961
962            # It seems that Python support for re.X is broken:
963            # At least for me (Python 3.13), this didn't work
964#            (r"""
965#              __attribute__\s*\(\(
966#                (?:
967#                    [\w\s]+          # attribute name
968#                    (?:\([^)]*\))?   # attribute arguments
969#                    \s*,?            # optional comma at the end
970#                )+
971#              \)\)\s+
972#             """, "", re.X),
973
974            # So, remove whitespaces and comments from it
975            (r"__attribute__\s*\(\((?:[\w\s]+(?:\([^)]*\))?\s*,?)+\)\)\s+", "", 0),
976        ]
977
978        for search, sub, flags in sub_prefixes:
979            prototype = KernRe(search, flags).sub(sub, prototype)
980
981        # Macros are a special case, as they change the prototype format
982        new_proto = KernRe(r"^#\s*define\s+").sub("", prototype)
983        if new_proto != prototype:
984            is_define_proto = True
985            prototype = new_proto
986        else:
987            is_define_proto = False
988
989        # Yes, this truly is vile.  We are looking for:
990        # 1. Return type (may be nothing if we're looking at a macro)
991        # 2. Function name
992        # 3. Function parameters.
993        #
994        # All the while we have to watch out for function pointer parameters
995        # (which IIRC is what the two sections are for), C types (these
996        # regexps don't even start to express all the possibilities), and
997        # so on.
998        #
999        # If you mess with these regexps, it's a good idea to check that
1000        # the following functions' documentation still comes out right:
1001        # - parport_register_device (function pointer parameters)
1002        # - atomic_set (macro)
1003        # - pci_match_device, __copy_to_user (long return type)
1004
1005        name = r'[a-zA-Z0-9_~:]+'
1006        prototype_end1 = r'[^\(]*'
1007        prototype_end2 = r'[^\{]*'
1008        prototype_end = fr'\(({prototype_end1}|{prototype_end2})\)'
1009
1010        # Besides compiling, Perl qr{[\w\s]+} works as a non-capturing group.
1011        # So, this needs to be mapped in Python with (?:...)? or (?:...)+
1012
1013        type1 = r'(?:[\w\s]+)?'
1014        type2 = r'(?:[\w\s]+\*+)+'
1015
1016        found = False
1017
1018        if is_define_proto:
1019            r = KernRe(r'^()(' + name + r')\s+')
1020
1021            if r.search(prototype):
1022                return_type = ''
1023                declaration_name = r.group(2)
1024                func_macro = True
1025
1026                found = True
1027
1028        if not found:
1029            patterns = [
1030                rf'^()({name})\s*{prototype_end}',
1031                rf'^({type1})\s+({name})\s*{prototype_end}',
1032                rf'^({type2})\s*({name})\s*{prototype_end}',
1033            ]
1034
1035            for p in patterns:
1036                r = KernRe(p)
1037
1038                if r.match(prototype):
1039
1040                    return_type = r.group(1)
1041                    declaration_name = r.group(2)
1042                    args = r.group(3)
1043
1044                    self.create_parameter_list(ln, decl_type, args, ',',
1045                                               declaration_name)
1046
1047                    found = True
1048                    break
1049        if not found:
1050            self.emit_msg(ln,
1051                          f"cannot understand function prototype: '{prototype}'")
1052            return
1053
1054        if self.entry.identifier != declaration_name:
1055            self.emit_msg(ln,
1056                          f"expecting prototype for {self.entry.identifier}(). Prototype was for {declaration_name}() instead")
1057            return
1058
1059        prms = " ".join(self.entry.parameterlist)
1060        self.check_sections(ln, declaration_name, "function",
1061                            self.entry.sectcheck, prms)
1062
1063        self.check_return_section(ln, declaration_name, return_type)
1064
1065        if 'typedef' in return_type:
1066            self.output_declaration(decl_type, declaration_name,
1067                                    function=declaration_name,
1068                                    typedef=True,
1069                                    functiontype=return_type,
1070                                    parameterlist=self.entry.parameterlist,
1071                                    parameterdescs=self.entry.parameterdescs,
1072                                    parametertypes=self.entry.parametertypes,
1073                                    parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
1074                                    sectionlist=self.entry.sectionlist,
1075                                    sections=self.entry.sections,
1076                                    section_start_lines=self.entry.section_start_lines,
1077                                    purpose=self.entry.declaration_purpose,
1078                                    func_macro=func_macro)
1079        else:
1080            self.output_declaration(decl_type, declaration_name,
1081                                    function=declaration_name,
1082                                    typedef=False,
1083                                    functiontype=return_type,
1084                                    parameterlist=self.entry.parameterlist,
1085                                    parameterdescs=self.entry.parameterdescs,
1086                                    parametertypes=self.entry.parametertypes,
1087                                    parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
1088                                    sectionlist=self.entry.sectionlist,
1089                                    sections=self.entry.sections,
1090                                    section_start_lines=self.entry.section_start_lines,
1091                                    purpose=self.entry.declaration_purpose,
1092                                    func_macro=func_macro)
1093
1094    def dump_typedef(self, ln, proto):
1095        """
1096        Stores a typedef inside self.entries array.
1097        """
1098
1099        typedef_type = r'((?:\s+[\w\*]+\b){0,7}\s+(?:\w+\b|\*+))\s*'
1100        typedef_ident = r'\*?\s*(\w\S+)\s*'
1101        typedef_args = r'\s*\((.*)\);'
1102
1103        typedef1 = KernRe(r'typedef' + typedef_type + r'\(' + typedef_ident + r'\)' + typedef_args)
1104        typedef2 = KernRe(r'typedef' + typedef_type + typedef_ident + typedef_args)
1105
1106        # Strip comments
1107        proto = KernRe(r'/\*.*?\*/', flags=re.S).sub('', proto)
1108
1109        # Parse function typedef prototypes
1110        for r in [typedef1, typedef2]:
1111            if not r.match(proto):
1112                continue
1113
1114            return_type = r.group(1).strip()
1115            declaration_name = r.group(2)
1116            args = r.group(3)
1117
1118            if self.entry.identifier != declaration_name:
1119                self.emit_msg(ln,
1120                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1121                return
1122
1123            decl_type = 'function'
1124            self.create_parameter_list(ln, decl_type, args, ',', declaration_name)
1125
1126            self.output_declaration(decl_type, declaration_name,
1127                                    function=declaration_name,
1128                                    typedef=True,
1129                                    functiontype=return_type,
1130                                    parameterlist=self.entry.parameterlist,
1131                                    parameterdescs=self.entry.parameterdescs,
1132                                    parametertypes=self.entry.parametertypes,
1133                                    parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
1134                                    sectionlist=self.entry.sectionlist,
1135                                    sections=self.entry.sections,
1136                                    section_start_lines=self.entry.section_start_lines,
1137                                    purpose=self.entry.declaration_purpose)
1138            return
1139
1140        # Handle nested parentheses or brackets
1141        r = KernRe(r'(\(*.\)\s*|\[*.\]\s*);$')
1142        while r.search(proto):
1143            proto = r.sub('', proto)
1144
1145        # Parse simple typedefs
1146        r = KernRe(r'typedef.*\s+(\w+)\s*;')
1147        if r.match(proto):
1148            declaration_name = r.group(1)
1149
1150            if self.entry.identifier != declaration_name:
1151                self.emit_msg(ln,
1152                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1153                return
1154
1155            self.output_declaration('typedef', declaration_name,
1156                                    typedef=declaration_name,
1157                                    sectionlist=self.entry.sectionlist,
1158                                    sections=self.entry.sections,
1159                                    section_start_lines=self.entry.section_start_lines,
1160                                    purpose=self.entry.declaration_purpose)
1161            return
1162
1163        self.emit_msg(ln, "error: Cannot parse typedef!")
1164
1165    @staticmethod
1166    def process_export(function_set, line):
1167        """
1168        process EXPORT_SYMBOL* tags
1169
1170        This method doesn't use any variable from the class, so declare it
1171        with a staticmethod decorator.
1172        """
1173
1174        # Note: it accepts only one EXPORT_SYMBOL* per line, as having
1175        # multiple export lines would violate Kernel coding style.
1176
1177        if export_symbol.search(line):
1178            symbol = export_symbol.group(2)
1179            function_set.add(symbol)
1180            return
1181
1182        if export_symbol_ns.search(line):
1183            symbol = export_symbol_ns.group(2)
1184            function_set.add(symbol)
1185
1186    def process_normal(self, ln, line):
1187        """
1188        STATE_NORMAL: looking for the /** to begin everything.
1189        """
1190
1191        if not doc_start.match(line):
1192            return
1193
1194        # start a new entry
1195        self.reset_state(ln)
1196        self.entry.in_doc_sect = False
1197
1198        # next line is always the function name
1199        self.state = state.NAME
1200
1201    def process_name(self, ln, line):
1202        """
1203        STATE_NAME: Looking for the "name - description" line
1204        """
1205
1206        if doc_block.search(line):
1207            self.entry.new_start_line = ln
1208
1209            if not doc_block.group(1):
1210                self.entry.section = self.section_intro
1211            else:
1212                self.entry.section = doc_block.group(1)
1213
1214            self.entry.identifier = self.entry.section
1215            self.state = state.DOCBLOCK
1216            return
1217
1218        if doc_decl.search(line):
1219            self.entry.identifier = doc_decl.group(1)
1220            self.entry.is_kernel_comment = False
1221
1222            decl_start = str(doc_com)       # comment block asterisk
1223            fn_type = r"(?:\w+\s*\*\s*)?"  # type (for non-functions)
1224            parenthesis = r"(?:\(\w*\))?"   # optional parenthesis on function
1225            decl_end = r"(?:[-:].*)"         # end of the name part
1226
1227            # test for pointer declaration type, foo * bar() - desc
1228            r = KernRe(fr"^{decl_start}([\w\s]+?){parenthesis}?\s*{decl_end}?$")
1229            if r.search(line):
1230                self.entry.identifier = r.group(1)
1231
1232            # Test for data declaration
1233            r = KernRe(r"^\s*\*?\s*(struct|union|enum|typedef)\b\s*(\w*)")
1234            if r.search(line):
1235                self.entry.decl_type = r.group(1)
1236                self.entry.identifier = r.group(2)
1237                self.entry.is_kernel_comment = True
1238            else:
1239                # Look for foo() or static void foo() - description;
1240                # or misspelt identifier
1241
1242                r1 = KernRe(fr"^{decl_start}{fn_type}(\w+)\s*{parenthesis}\s*{decl_end}?$")
1243                r2 = KernRe(fr"^{decl_start}{fn_type}(\w+[^-:]*){parenthesis}\s*{decl_end}$")
1244
1245                for r in [r1, r2]:
1246                    if r.search(line):
1247                        self.entry.identifier = r.group(1)
1248                        self.entry.decl_type = "function"
1249
1250                        r = KernRe(r"define\s+")
1251                        self.entry.identifier = r.sub("", self.entry.identifier)
1252                        self.entry.is_kernel_comment = True
1253                        break
1254
1255            self.entry.identifier = self.entry.identifier.strip(" ")
1256
1257            self.state = state.BODY
1258
1259            # if there's no @param blocks need to set up default section here
1260            self.entry.section = SECTION_DEFAULT
1261            self.entry.new_start_line = ln + 1
1262
1263            r = KernRe("[-:](.*)")
1264            if r.search(line):
1265                # strip leading/trailing/multiple spaces
1266                self.entry.descr = r.group(1).strip(" ")
1267
1268                r = KernRe(r"\s+")
1269                self.entry.descr = r.sub(" ", self.entry.descr)
1270                self.entry.declaration_purpose = self.entry.descr
1271                self.state = state.BODY_MAYBE
1272            else:
1273                self.entry.declaration_purpose = ""
1274
1275            if not self.entry.is_kernel_comment:
1276                self.emit_msg(ln,
1277                              f"This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst\n{line}")
1278                self.state = state.NORMAL
1279
1280            if not self.entry.declaration_purpose and self.config.wshort_desc:
1281                self.emit_msg(ln,
1282                              f"missing initial short description on line:\n{line}")
1283
1284            if not self.entry.identifier and self.entry.decl_type != "enum":
1285                self.emit_msg(ln,
1286                              f"wrong kernel-doc identifier on line:\n{line}")
1287                self.state = state.NORMAL
1288
1289            if self.config.verbose:
1290                self.emit_msg(ln,
1291                              f"Scanning doc for {self.entry.decl_type} {self.entry.identifier}",
1292                                  warning=False)
1293
1294            return
1295
1296        # Failed to find an identifier. Emit a warning
1297        self.emit_msg(ln, f"Cannot find identifier on line:\n{line}")
1298
1299    def process_body(self, ln, line):
1300        """
1301        STATE_BODY and STATE_BODY_MAYBE: the bulk of a kerneldoc comment.
1302        """
1303
1304        if self.state == state.BODY_WITH_BLANK_LINE:
1305            r = KernRe(r"\s*\*\s?\S")
1306            if r.match(line):
1307                self.dump_section()
1308                self.entry.section = SECTION_DEFAULT
1309                self.entry.new_start_line = ln
1310                self.entry.contents = ""
1311
1312        if doc_sect.search(line):
1313            self.entry.in_doc_sect = True
1314            newsection = doc_sect.group(1)
1315
1316            if newsection.lower() in ["description", "context"]:
1317                newsection = newsection.title()
1318
1319            # Special case: @return is a section, not a param description
1320            if newsection.lower() in ["@return", "@returns",
1321                                      "return", "returns"]:
1322                newsection = "Return"
1323
1324            # Perl kernel-doc has a check here for contents before sections.
1325            # the logic there is always false, as in_doc_sect variable is
1326            # always true. So, just don't implement Wcontents_before_sections
1327
1328            # .title()
1329            newcontents = doc_sect.group(2)
1330            if not newcontents:
1331                newcontents = ""
1332
1333            if self.entry.contents.strip("\n"):
1334                self.dump_section()
1335
1336            self.entry.new_start_line = ln
1337            self.entry.section = newsection
1338            self.entry.leading_space = None
1339
1340            self.entry.contents = newcontents.lstrip()
1341            if self.entry.contents:
1342                self.entry.contents += "\n"
1343
1344            self.state = state.BODY
1345            return
1346
1347        if doc_end.search(line):
1348            self.dump_section()
1349
1350            # Look for doc_com + <text> + doc_end:
1351            r = KernRe(r'\s*\*\s*[a-zA-Z_0-9:\.]+\*/')
1352            if r.match(line):
1353                self.emit_msg(ln, f"suspicious ending line: {line}")
1354
1355            self.entry.prototype = ""
1356            self.entry.new_start_line = ln + 1
1357
1358            self.state = state.PROTO
1359            return
1360
1361        if doc_content.search(line):
1362            cont = doc_content.group(1)
1363
1364            if cont == "":
1365                if self.entry.section == self.section_context:
1366                    self.dump_section()
1367
1368                    self.entry.new_start_line = ln
1369                    self.state = state.BODY
1370                else:
1371                    if self.entry.section != SECTION_DEFAULT:
1372                        self.state = state.BODY_WITH_BLANK_LINE
1373                    else:
1374                        self.state = state.BODY
1375
1376                    self.entry.contents += "\n"
1377
1378            elif self.state == state.BODY_MAYBE:
1379
1380                # Continued declaration purpose
1381                self.entry.declaration_purpose = self.entry.declaration_purpose.rstrip()
1382                self.entry.declaration_purpose += " " + cont
1383
1384                r = KernRe(r"\s+")
1385                self.entry.declaration_purpose = r.sub(' ',
1386                                                       self.entry.declaration_purpose)
1387
1388            else:
1389                if self.entry.section.startswith('@') or        \
1390                   self.entry.section == self.section_context:
1391                    if self.entry.leading_space is None:
1392                        r = KernRe(r'^(\s+)')
1393                        if r.match(cont):
1394                            self.entry.leading_space = len(r.group(1))
1395                        else:
1396                            self.entry.leading_space = 0
1397
1398                    # Double-check if leading space are realy spaces
1399                    pos = 0
1400                    for i in range(0, self.entry.leading_space):
1401                        if cont[i] != " ":
1402                            break
1403                        pos += 1
1404
1405                    cont = cont[pos:]
1406
1407                    # NEW LOGIC:
1408                    # In case it is different, update it
1409                    if self.entry.leading_space != pos:
1410                        self.entry.leading_space = pos
1411
1412                self.entry.contents += cont + "\n"
1413            return
1414
1415        # Unknown line, ignore
1416        self.emit_msg(ln, f"bad line: {line}")
1417
1418    def process_inline(self, ln, line):
1419        """STATE_INLINE: docbook comments within a prototype."""
1420
1421        if self.inline_doc_state == state.INLINE_NAME and \
1422           doc_inline_sect.search(line):
1423            self.entry.section = doc_inline_sect.group(1)
1424            self.entry.new_start_line = ln
1425
1426            self.entry.contents = doc_inline_sect.group(2).lstrip()
1427            if self.entry.contents != "":
1428                self.entry.contents += "\n"
1429
1430            self.inline_doc_state = state.INLINE_TEXT
1431            # Documentation block end */
1432            return
1433
1434        if doc_inline_end.search(line):
1435            if self.entry.contents not in ["", "\n"]:
1436                self.dump_section()
1437
1438            self.state = state.PROTO
1439            self.inline_doc_state = state.INLINE_NA
1440            return
1441
1442        if doc_content.search(line):
1443            if self.inline_doc_state == state.INLINE_TEXT:
1444                self.entry.contents += doc_content.group(1) + "\n"
1445                if not self.entry.contents.strip(" ").rstrip("\n"):
1446                    self.entry.contents = ""
1447
1448            elif self.inline_doc_state == state.INLINE_NAME:
1449                self.emit_msg(ln,
1450                              f"Incorrect use of kernel-doc format: {line}")
1451
1452                self.inline_doc_state = state.INLINE_ERROR
1453
1454    def syscall_munge(self, ln, proto):         # pylint: disable=W0613
1455        """
1456        Handle syscall definitions
1457        """
1458
1459        is_void = False
1460
1461        # Strip newlines/CR's
1462        proto = re.sub(r'[\r\n]+', ' ', proto)
1463
1464        # Check if it's a SYSCALL_DEFINE0
1465        if 'SYSCALL_DEFINE0' in proto:
1466            is_void = True
1467
1468        # Replace SYSCALL_DEFINE with correct return type & function name
1469        proto = KernRe(r'SYSCALL_DEFINE.*\(').sub('long sys_', proto)
1470
1471        r = KernRe(r'long\s+(sys_.*?),')
1472        if r.search(proto):
1473            proto = KernRe(',').sub('(', proto, count=1)
1474        elif is_void:
1475            proto = KernRe(r'\)').sub('(void)', proto, count=1)
1476
1477        # Now delete all of the odd-numbered commas in the proto
1478        # so that argument types & names don't have a comma between them
1479        count = 0
1480        length = len(proto)
1481
1482        if is_void:
1483            length = 0  # skip the loop if is_void
1484
1485        for ix in range(length):
1486            if proto[ix] == ',':
1487                count += 1
1488                if count % 2 == 1:
1489                    proto = proto[:ix] + ' ' + proto[ix + 1:]
1490
1491        return proto
1492
1493    def tracepoint_munge(self, ln, proto):
1494        """
1495        Handle tracepoint definitions
1496        """
1497
1498        tracepointname = None
1499        tracepointargs = None
1500
1501        # Match tracepoint name based on different patterns
1502        r = KernRe(r'TRACE_EVENT\((.*?),')
1503        if r.search(proto):
1504            tracepointname = r.group(1)
1505
1506        r = KernRe(r'DEFINE_SINGLE_EVENT\((.*?),')
1507        if r.search(proto):
1508            tracepointname = r.group(1)
1509
1510        r = KernRe(r'DEFINE_EVENT\((.*?),(.*?),')
1511        if r.search(proto):
1512            tracepointname = r.group(2)
1513
1514        if tracepointname:
1515            tracepointname = tracepointname.lstrip()
1516
1517        r = KernRe(r'TP_PROTO\((.*?)\)')
1518        if r.search(proto):
1519            tracepointargs = r.group(1)
1520
1521        if not tracepointname or not tracepointargs:
1522            self.emit_msg(ln,
1523                          f"Unrecognized tracepoint format:\n{proto}\n")
1524        else:
1525            proto = f"static inline void trace_{tracepointname}({tracepointargs})"
1526            self.entry.identifier = f"trace_{self.entry.identifier}"
1527
1528        return proto
1529
1530    def process_proto_function(self, ln, line):
1531        """Ancillary routine to process a function prototype"""
1532
1533        # strip C99-style comments to end of line
1534        r = KernRe(r"\/\/.*$", re.S)
1535        line = r.sub('', line)
1536
1537        if KernRe(r'\s*#\s*define').match(line):
1538            self.entry.prototype = line
1539        elif line.startswith('#'):
1540            # Strip other macros like #ifdef/#ifndef/#endif/...
1541            pass
1542        else:
1543            r = KernRe(r'([^\{]*)')
1544            if r.match(line):
1545                self.entry.prototype += r.group(1) + " "
1546
1547        if '{' in line or ';' in line or KernRe(r'\s*#\s*define').match(line):
1548            # strip comments
1549            r = KernRe(r'/\*.*?\*/')
1550            self.entry.prototype = r.sub('', self.entry.prototype)
1551
1552            # strip newlines/cr's
1553            r = KernRe(r'[\r\n]+')
1554            self.entry.prototype = r.sub(' ', self.entry.prototype)
1555
1556            # strip leading spaces
1557            r = KernRe(r'^\s+')
1558            self.entry.prototype = r.sub('', self.entry.prototype)
1559
1560            # Handle self.entry.prototypes for function pointers like:
1561            #       int (*pcs_config)(struct foo)
1562
1563            r = KernRe(r'^(\S+\s+)\(\s*\*(\S+)\)')
1564            self.entry.prototype = r.sub(r'\1\2', self.entry.prototype)
1565
1566            if 'SYSCALL_DEFINE' in self.entry.prototype:
1567                self.entry.prototype = self.syscall_munge(ln,
1568                                                          self.entry.prototype)
1569
1570            r = KernRe(r'TRACE_EVENT|DEFINE_EVENT|DEFINE_SINGLE_EVENT')
1571            if r.search(self.entry.prototype):
1572                self.entry.prototype = self.tracepoint_munge(ln,
1573                                                             self.entry.prototype)
1574
1575            self.dump_function(ln, self.entry.prototype)
1576            self.reset_state(ln)
1577
1578    def process_proto_type(self, ln, line):
1579        """Ancillary routine to process a type"""
1580
1581        # Strip newlines/cr's.
1582        line = KernRe(r'[\r\n]+', re.S).sub(' ', line)
1583
1584        # Strip leading spaces
1585        line = KernRe(r'^\s+', re.S).sub('', line)
1586
1587        # Strip trailing spaces
1588        line = KernRe(r'\s+$', re.S).sub('', line)
1589
1590        # Strip C99-style comments to the end of the line
1591        line = KernRe(r"\/\/.*$", re.S).sub('', line)
1592
1593        # To distinguish preprocessor directive from regular declaration later.
1594        if line.startswith('#'):
1595            line += ";"
1596
1597        r = KernRe(r'([^\{\};]*)([\{\};])(.*)')
1598        while True:
1599            if r.search(line):
1600                if self.entry.prototype:
1601                    self.entry.prototype += " "
1602                self.entry.prototype += r.group(1) + r.group(2)
1603
1604                self.entry.brcount += r.group(2).count('{')
1605                self.entry.brcount -= r.group(2).count('}')
1606
1607                self.entry.brcount = max(self.entry.brcount, 0)
1608
1609                if r.group(2) == ';' and self.entry.brcount == 0:
1610                    self.dump_declaration(ln, self.entry.prototype)
1611                    self.reset_state(ln)
1612                    break
1613
1614                line = r.group(3)
1615            else:
1616                self.entry.prototype += line
1617                break
1618
1619    def process_proto(self, ln, line):
1620        """STATE_PROTO: reading a function/whatever prototype."""
1621
1622        if doc_inline_oneline.search(line):
1623            self.entry.section = doc_inline_oneline.group(1)
1624            self.entry.contents = doc_inline_oneline.group(2)
1625
1626            if self.entry.contents != "":
1627                self.entry.contents += "\n"
1628                self.dump_section(start_new=False)
1629
1630        elif doc_inline_start.search(line):
1631            self.state = state.INLINE
1632            self.inline_doc_state = state.INLINE_NAME
1633
1634        elif self.entry.decl_type == 'function':
1635            self.process_proto_function(ln, line)
1636
1637        else:
1638            self.process_proto_type(ln, line)
1639
1640    def process_docblock(self, ln, line):
1641        """STATE_DOCBLOCK: within a DOC: block."""
1642
1643        if doc_end.search(line):
1644            self.dump_section()
1645            self.output_declaration("doc", self.entry.identifier,
1646                                    sectionlist=self.entry.sectionlist,
1647                                    sections=self.entry.sections,
1648                                    section_start_lines=self.entry.section_start_lines)
1649            self.reset_state(ln)
1650
1651        elif doc_content.search(line):
1652            self.entry.contents += doc_content.group(1) + "\n"
1653
1654    def parse_export(self):
1655        """
1656        Parses EXPORT_SYMBOL* macros from a single Kernel source file.
1657        """
1658
1659        export_table = set()
1660
1661        try:
1662            with open(self.fname, "r", encoding="utf8",
1663                      errors="backslashreplace") as fp:
1664
1665                for line in fp:
1666                    self.process_export(export_table, line)
1667
1668        except IOError:
1669            return None
1670
1671        return export_table
1672
1673    def parse_kdoc(self):
1674        """
1675        Open and process each line of a C source file.
1676        The parsing is controlled via a state machine, and the line is passed
1677        to a different process function depending on the state. The process
1678        function may update the state as needed.
1679
1680        Besides parsing kernel-doc tags, it also parses export symbols.
1681        """
1682
1683        cont = False
1684        prev = ""
1685        prev_ln = None
1686        export_table = set()
1687
1688        try:
1689            with open(self.fname, "r", encoding="utf8",
1690                      errors="backslashreplace") as fp:
1691                for ln, line in enumerate(fp):
1692
1693                    line = line.expandtabs().strip("\n")
1694
1695                    # Group continuation lines on prototypes
1696                    if self.state == state.PROTO:
1697                        if line.endswith("\\"):
1698                            prev += line.rstrip("\\")
1699                            cont = True
1700
1701                            if not prev_ln:
1702                                prev_ln = ln
1703
1704                            continue
1705
1706                        if cont:
1707                            ln = prev_ln
1708                            line = prev + line
1709                            prev = ""
1710                            cont = False
1711                            prev_ln = None
1712
1713                    self.config.log.debug("%d %s%s: %s",
1714                                          ln, state.name[self.state],
1715                                          state.inline_name[self.inline_doc_state],
1716                                          line)
1717
1718                    # This is an optimization over the original script.
1719                    # There, when export_file was used for the same file,
1720                    # it was read twice. Here, we use the already-existing
1721                    # loop to parse exported symbols as well.
1722                    #
1723                    # TODO: It should be noticed that not all states are
1724                    # needed here. On a future cleanup, process export only
1725                    # at the states that aren't handling comment markups.
1726                    self.process_export(export_table, line)
1727
1728                    # Hand this line to the appropriate state handler
1729                    if self.state == state.NORMAL:
1730                        self.process_normal(ln, line)
1731                    elif self.state == state.NAME:
1732                        self.process_name(ln, line)
1733                    elif self.state in [state.BODY, state.BODY_MAYBE,
1734                                        state.BODY_WITH_BLANK_LINE]:
1735                        self.process_body(ln, line)
1736                    elif self.state == state.INLINE:  # scanning for inline parameters
1737                        self.process_inline(ln, line)
1738                    elif self.state == state.PROTO:
1739                        self.process_proto(ln, line)
1740                    elif self.state == state.DOCBLOCK:
1741                        self.process_docblock(ln, line)
1742        except OSError:
1743            self.config.log.error(f"Error: Cannot open file {self.fname}")
1744
1745        return export_table, self.entries
1746