xref: /linux/scripts/lib/kdoc/kdoc_parser.py (revision b1cce98493a095925fb51be045ccf6e08edb4aa0)
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 sys
16import re
17from pprint import pformat
18
19from kdoc_re import NestedMatch, KernRe
20from kdoc_item import KdocItem
21
22#
23# Regular expressions used to parse kernel-doc markups at KernelDoc class.
24#
25# Let's declare them in lowercase outside any class to make easier to
26# convert from the python script.
27#
28# As those are evaluated at the beginning, no need to cache them
29#
30
31# Allow whitespace at end of comment start.
32doc_start = KernRe(r'^/\*\*\s*$', cache=False)
33
34doc_end = KernRe(r'\*/', cache=False)
35doc_com = KernRe(r'\s*\*\s*', cache=False)
36doc_com_body = KernRe(r'\s*\* ?', cache=False)
37doc_decl = doc_com + KernRe(r'(\w+)', cache=False)
38
39# @params and a strictly limited set of supported section names
40# Specifically:
41#   Match @word:
42#         @...:
43#         @{section-name}:
44# while trying to not match literal block starts like "example::"
45#
46known_section_names = 'description|context|returns?|notes?|examples?'
47known_sections = KernRe(known_section_names, flags = re.I)
48doc_sect = doc_com + \
49    KernRe(r'\s*(\@[.\w]+|\@\.\.\.|' + known_section_names + r')\s*:([^:].*)?$',
50           flags=re.I, cache=False)
51
52doc_content = doc_com_body + KernRe(r'(.*)', cache=False)
53doc_inline_start = KernRe(r'^\s*/\*\*\s*$', cache=False)
54doc_inline_sect = KernRe(r'\s*\*\s*(@\s*[\w][\w\.]*\s*):(.*)', cache=False)
55doc_inline_end = KernRe(r'^\s*\*/\s*$', cache=False)
56doc_inline_oneline = KernRe(r'^\s*/\*\*\s*(@[\w\s]+):\s*(.*)\s*\*/\s*$', cache=False)
57attribute = KernRe(r"__attribute__\s*\(\([a-z0-9,_\*\s\(\)]*\)\)",
58               flags=re.I | re.S, cache=False)
59
60export_symbol = KernRe(r'^\s*EXPORT_SYMBOL(_GPL)?\s*\(\s*(\w+)\s*\)\s*', cache=False)
61export_symbol_ns = KernRe(r'^\s*EXPORT_SYMBOL_NS(_GPL)?\s*\(\s*(\w+)\s*,\s*"\S+"\)\s*', cache=False)
62
63type_param = KernRe(r"\@(\w*((\.\w+)|(->\w+))*(\.\.\.)?)", cache=False)
64
65#
66# Tests for the beginning of a kerneldoc block in its various forms.
67#
68doc_block = doc_com + KernRe(r'DOC:\s*(.*)?', cache=False)
69doc_begin_data = KernRe(r"^\s*\*?\s*(struct|union|enum|typedef)\b\s*(\w*)", cache = False)
70doc_begin_func = KernRe(str(doc_com) +			# initial " * '
71                        r"(?:\w+\s*\*\s*)?" + 		# type (not captured)
72                        r'(?:define\s+)?' + 		# possible "define" (not captured)
73                        r'(\w+)\s*(?:\(\w*\))?\s*' +	# name and optional "(...)"
74                        r'(?:[-:].*)?$',		# description (not captured)
75                        cache = False)
76
77#
78# A little helper to get rid of excess white space
79#
80multi_space = KernRe(r'\s\s+')
81def trim_whitespace(s):
82    return multi_space.sub(' ', s.strip())
83
84class state:
85    """
86    State machine enums
87    """
88
89    # Parser states
90    NORMAL        = 0        # normal code
91    NAME          = 1        # looking for function name
92    DECLARATION   = 2        # We have seen a declaration which might not be done
93    BODY          = 3        # the body of the comment
94    SPECIAL_SECTION = 4      # doc section ending with a blank line
95    PROTO         = 5        # scanning prototype
96    DOCBLOCK      = 6        # documentation block
97    INLINE_NAME   = 7        # gathering doc outside main block
98    INLINE_TEXT   = 8	     # reading the body of inline docs
99
100    name = [
101        "NORMAL",
102        "NAME",
103        "DECLARATION",
104        "BODY",
105        "SPECIAL_SECTION",
106        "PROTO",
107        "DOCBLOCK",
108        "INLINE_NAME",
109        "INLINE_TEXT",
110    ]
111
112
113SECTION_DEFAULT = "Description"  # default section
114
115class KernelEntry:
116
117    def __init__(self, config, ln):
118        self.config = config
119
120        self._contents = []
121        self.prototype = ""
122
123        self.warnings = []
124
125        self.parameterlist = []
126        self.parameterdescs = {}
127        self.parametertypes = {}
128        self.parameterdesc_start_lines = {}
129
130        self.section_start_lines = {}
131        self.sections = {}
132
133        self.anon_struct_union = False
134
135        self.leading_space = None
136
137        # State flags
138        self.brcount = 0
139        self.declaration_start_line = ln + 1
140
141    #
142    # Management of section contents
143    #
144    def add_text(self, text):
145        self._contents.append(text)
146
147    def contents(self):
148        return '\n'.join(self._contents) + '\n'
149
150    # TODO: rename to emit_message after removal of kernel-doc.pl
151    def emit_msg(self, log_msg, warning=True):
152        """Emit a message"""
153
154        if not warning:
155            self.config.log.info(log_msg)
156            return
157
158        # Delegate warning output to output logic, as this way it
159        # will report warnings/info only for symbols that are output
160
161        self.warnings.append(log_msg)
162        return
163
164    #
165    # Begin a new section.
166    #
167    def begin_section(self, line_no, title = SECTION_DEFAULT, dump = False):
168        if dump:
169            self.dump_section(start_new = True)
170        self.section = title
171        self.new_start_line = line_no
172
173    def dump_section(self, start_new=True):
174        """
175        Dumps section contents to arrays/hashes intended for that purpose.
176        """
177        #
178        # If we have accumulated no contents in the default ("description")
179        # section, don't bother.
180        #
181        if self.section == SECTION_DEFAULT and not self._contents:
182            return
183        name = self.section
184        contents = self.contents()
185
186        if type_param.match(name):
187            name = type_param.group(1)
188
189            self.parameterdescs[name] = contents
190            self.parameterdesc_start_lines[name] = self.new_start_line
191
192            self.new_start_line = 0
193
194        else:
195            if name in self.sections and self.sections[name] != "":
196                # Only warn on user-specified duplicate section names
197                if name != SECTION_DEFAULT:
198                    self.emit_msg(self.new_start_line,
199                                  f"duplicate section name '{name}'\n")
200                # Treat as a new paragraph - add a blank line
201                self.sections[name] += '\n' + contents
202            else:
203                self.sections[name] = contents
204                self.section_start_lines[name] = self.new_start_line
205                self.new_start_line = 0
206
207#        self.config.log.debug("Section: %s : %s", name, pformat(vars(self)))
208
209        if start_new:
210            self.section = SECTION_DEFAULT
211            self._contents = []
212
213
214class KernelDoc:
215    """
216    Read a C language source or header FILE and extract embedded
217    documentation comments.
218    """
219
220    # Section names
221
222    section_context = "Context"
223    section_return = "Return"
224
225    undescribed = "-- undescribed --"
226
227    def __init__(self, config, fname):
228        """Initialize internal variables"""
229
230        self.fname = fname
231        self.config = config
232
233        # Initial state for the state machines
234        self.state = state.NORMAL
235
236        # Store entry currently being processed
237        self.entry = None
238
239        # Place all potential outputs into an array
240        self.entries = []
241
242        #
243        # We need Python 3.7 for its "dicts remember the insertion
244        # order" guarantee
245        #
246        if sys.version_info.major == 3 and sys.version_info.minor < 7:
247            self.emit_msg(0,
248                          'Python 3.7 or later is required for correct results')
249
250    def emit_msg(self, ln, msg, warning=True):
251        """Emit a message"""
252
253        log_msg = f"{self.fname}:{ln} {msg}"
254
255        if self.entry:
256            self.entry.emit_msg(log_msg, warning)
257            return
258
259        if warning:
260            self.config.log.warning(log_msg)
261        else:
262            self.config.log.info(log_msg)
263
264    def dump_section(self, start_new=True):
265        """
266        Dumps section contents to arrays/hashes intended for that purpose.
267        """
268
269        if self.entry:
270            self.entry.dump_section(start_new)
271
272    # TODO: rename it to store_declaration after removal of kernel-doc.pl
273    def output_declaration(self, dtype, name, **args):
274        """
275        Stores the entry into an entry array.
276
277        The actual output and output filters will be handled elsewhere
278        """
279
280        item = KdocItem(name, dtype, self.entry.declaration_start_line, **args)
281        item.warnings = self.entry.warnings
282
283        # Drop empty sections
284        # TODO: improve empty sections logic to emit warnings
285        sections = self.entry.sections
286        for section in ["Description", "Return"]:
287            if section in sections and not sections[section].rstrip():
288                del sections[section]
289        item.set_sections(sections, self.entry.section_start_lines)
290        item.set_params(self.entry.parameterlist, self.entry.parameterdescs,
291                        self.entry.parametertypes,
292                        self.entry.parameterdesc_start_lines)
293        self.entries.append(item)
294
295        self.config.log.debug("Output: %s:%s = %s", dtype, name, pformat(args))
296
297    def reset_state(self, ln):
298        """
299        Ancillary routine to create a new entry. It initializes all
300        variables used by the state machine.
301        """
302
303        self.entry = KernelEntry(self.config, ln)
304
305        # State flags
306        self.state = state.NORMAL
307
308    def push_parameter(self, ln, decl_type, param, dtype,
309                       org_arg, declaration_name):
310        """
311        Store parameters and their descriptions at self.entry.
312        """
313
314        if self.entry.anon_struct_union and dtype == "" and param == "}":
315            return  # Ignore the ending }; from anonymous struct/union
316
317        self.entry.anon_struct_union = False
318
319        param = KernRe(r'[\[\)].*').sub('', param, count=1)
320
321        if dtype == "" and param.endswith("..."):
322            if KernRe(r'\w\.\.\.$').search(param):
323                # For named variable parameters of the form `x...`,
324                # remove the dots
325                param = param[:-3]
326            else:
327                # Handles unnamed variable parameters
328                param = "..."
329
330            if param not in self.entry.parameterdescs or \
331                not self.entry.parameterdescs[param]:
332
333                self.entry.parameterdescs[param] = "variable arguments"
334
335        elif dtype == "" and (not param or param == "void"):
336            param = "void"
337            self.entry.parameterdescs[param] = "no arguments"
338
339        elif dtype == "" and param in ["struct", "union"]:
340            # Handle unnamed (anonymous) union or struct
341            dtype = param
342            param = "{unnamed_" + param + "}"
343            self.entry.parameterdescs[param] = "anonymous\n"
344            self.entry.anon_struct_union = True
345
346        # Handle cache group enforcing variables: they do not need
347        # to be described in header files
348        elif "__cacheline_group" in param:
349            # Ignore __cacheline_group_begin and __cacheline_group_end
350            return
351
352        # Warn if parameter has no description
353        # (but ignore ones starting with # as these are not parameters
354        # but inline preprocessor statements)
355        if param not in self.entry.parameterdescs and not param.startswith("#"):
356            self.entry.parameterdescs[param] = self.undescribed
357
358            if "." not in param:
359                if decl_type == 'function':
360                    dname = f"{decl_type} parameter"
361                else:
362                    dname = f"{decl_type} member"
363
364                self.emit_msg(ln,
365                              f"{dname} '{param}' not described in '{declaration_name}'")
366
367        # Strip spaces from param so that it is one continuous string on
368        # parameterlist. This fixes a problem where check_sections()
369        # cannot find a parameter like "addr[6 + 2]" because it actually
370        # appears as "addr[6", "+", "2]" on the parameter list.
371        # However, it's better to maintain the param string unchanged for
372        # output, so just weaken the string compare in check_sections()
373        # to ignore "[blah" in a parameter string.
374
375        self.entry.parameterlist.append(param)
376        org_arg = KernRe(r'\s\s+').sub(' ', org_arg)
377        self.entry.parametertypes[param] = org_arg
378
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.push_parameter(ln, decl_type, param, dtype,
425                                    arg, declaration_name)
426
427            elif KernRe(r'\(.+\)\s*\[').search(arg):
428                # Array-of-pointers
429
430                arg = arg.replace('#', ',')
431                r = KernRe(r'[^\(]+\(\s*\*\s*([\w\[\]\.]*?)\s*(\s*\[\s*[\w]+\s*\]\s*)*\)')
432                if r.match(arg):
433                    param = r.group(1)
434                else:
435                    self.emit_msg(ln, f"Invalid param: {arg}")
436                    param = arg
437
438                dtype = KernRe(r'([^\(]+\(\*?)\s*' + re.escape(param)).sub(r'\1', arg)
439
440                self.push_parameter(ln, decl_type, param, dtype,
441                                    arg, declaration_name)
442
443            elif arg:
444                arg = KernRe(r'\s*:\s*').sub(":", arg)
445                arg = KernRe(r'\s*\[').sub('[', arg)
446
447                args = KernRe(r'\s*,\s*').split(arg)
448                if args[0] and '*' in args[0]:
449                    args[0] = re.sub(r'(\*+)\s*', r' \1', args[0])
450
451                first_arg = []
452                r = KernRe(r'^(.*\s+)(.*?\[.*\].*)$')
453                if args[0] and r.match(args[0]):
454                    args.pop(0)
455                    first_arg.extend(r.group(1))
456                    first_arg.append(r.group(2))
457                else:
458                    first_arg = KernRe(r'\s+').split(args.pop(0))
459
460                args.insert(0, first_arg.pop())
461                dtype = ' '.join(first_arg)
462
463                for param in args:
464                    if KernRe(r'^(\*+)\s*(.*)').match(param):
465                        r = KernRe(r'^(\*+)\s*(.*)')
466                        if not r.match(param):
467                            self.emit_msg(ln, f"Invalid param: {param}")
468                            continue
469
470                        param = r.group(1)
471
472                        self.push_parameter(ln, decl_type, r.group(2),
473                                            f"{dtype} {r.group(1)}",
474                                            arg, declaration_name)
475
476                    elif KernRe(r'(.*?):(\w+)').search(param):
477                        r = KernRe(r'(.*?):(\w+)')
478                        if not r.match(param):
479                            self.emit_msg(ln, f"Invalid param: {param}")
480                            continue
481
482                        if dtype != "":  # Skip unnamed bit-fields
483                            self.push_parameter(ln, decl_type, r.group(1),
484                                                f"{dtype}:{r.group(2)}",
485                                                arg, declaration_name)
486                    else:
487                        self.push_parameter(ln, decl_type, param, dtype,
488                                            arg, declaration_name)
489
490    def check_sections(self, ln, decl_name, decl_type):
491        """
492        Check for errors inside sections, emitting warnings if not found
493        parameters are described.
494        """
495        for section in self.entry.sections:
496            if section not in self.entry.parameterlist and \
497               not known_sections.search(section):
498                if decl_type == 'function':
499                    dname = f"{decl_type} parameter"
500                else:
501                    dname = f"{decl_type} member"
502                self.emit_msg(ln,
503                              f"Excess {dname} '{section}' description in '{decl_name}'")
504
505    def check_return_section(self, ln, declaration_name, return_type):
506        """
507        If the function doesn't return void, warns about the lack of a
508        return description.
509        """
510
511        if not self.config.wreturn:
512            return
513
514        # Ignore an empty return type (It's a macro)
515        # Ignore functions with a "void" return type (but not "void *")
516        if not return_type or KernRe(r'void\s*\w*\s*$').search(return_type):
517            return
518
519        if not self.entry.sections.get("Return", None):
520            self.emit_msg(ln,
521                          f"No description found for return value of '{declaration_name}'")
522
523    def dump_struct(self, ln, proto):
524        """
525        Store an entry for an struct or union
526        """
527
528        type_pattern = r'(struct|union)'
529
530        qualifiers = [
531            "__attribute__",
532            "__packed",
533            "__aligned",
534            "____cacheline_aligned_in_smp",
535            "____cacheline_aligned",
536        ]
537
538        definition_body = r'\{(.*)\}\s*' + "(?:" + '|'.join(qualifiers) + ")?"
539        struct_members = KernRe(type_pattern + r'([^\{\};]+)(\{)([^\{\}]*)(\})([^\{\}\;]*)(\;)')
540
541        # Extract struct/union definition
542        members = None
543        declaration_name = None
544        decl_type = None
545
546        r = KernRe(type_pattern + r'\s+(\w+)\s*' + definition_body)
547        if r.search(proto):
548            decl_type = r.group(1)
549            declaration_name = r.group(2)
550            members = r.group(3)
551        else:
552            r = KernRe(r'typedef\s+' + type_pattern + r'\s*' + definition_body + r'\s*(\w+)\s*;')
553
554            if r.search(proto):
555                decl_type = r.group(1)
556                declaration_name = r.group(3)
557                members = r.group(2)
558
559        if not members:
560            self.emit_msg(ln, f"{proto} error: Cannot parse struct or union!")
561            return
562
563        if self.entry.identifier != declaration_name:
564            self.emit_msg(ln,
565                          f"expecting prototype for {decl_type} {self.entry.identifier}. Prototype was for {decl_type} {declaration_name} instead\n")
566            return
567
568        args_pattern = r'([^,)]+)'
569
570        sub_prefixes = [
571            (KernRe(r'\/\*\s*private:.*?\/\*\s*public:.*?\*\/', re.S | re.I), ''),
572            (KernRe(r'\/\*\s*private:.*', re.S | re.I), ''),
573
574            # Strip comments
575            (KernRe(r'\/\*.*?\*\/', re.S), ''),
576
577            # Strip attributes
578            (attribute, ' '),
579            (KernRe(r'\s*__aligned\s*\([^;]*\)', re.S), ' '),
580            (KernRe(r'\s*__counted_by\s*\([^;]*\)', re.S), ' '),
581            (KernRe(r'\s*__counted_by_(le|be)\s*\([^;]*\)', re.S), ' '),
582            (KernRe(r'\s*__packed\s*', re.S), ' '),
583            (KernRe(r'\s*CRYPTO_MINALIGN_ATTR', re.S), ' '),
584            (KernRe(r'\s*____cacheline_aligned_in_smp', re.S), ' '),
585            (KernRe(r'\s*____cacheline_aligned', re.S), ' '),
586
587            # Unwrap struct_group macros based on this definition:
588            # __struct_group(TAG, NAME, ATTRS, MEMBERS...)
589            # which has variants like: struct_group(NAME, MEMBERS...)
590            # Only MEMBERS arguments require documentation.
591            #
592            # Parsing them happens on two steps:
593            #
594            # 1. drop struct group arguments that aren't at MEMBERS,
595            #    storing them as STRUCT_GROUP(MEMBERS)
596            #
597            # 2. remove STRUCT_GROUP() ancillary macro.
598            #
599            # The original logic used to remove STRUCT_GROUP() using an
600            # advanced regex:
601            #
602            #   \bSTRUCT_GROUP(\(((?:(?>[^)(]+)|(?1))*)\))[^;]*;
603            #
604            # with two patterns that are incompatible with
605            # Python re module, as it has:
606            #
607            #   - a recursive pattern: (?1)
608            #   - an atomic grouping: (?>...)
609            #
610            # I tried a simpler version: but it didn't work either:
611            #   \bSTRUCT_GROUP\(([^\)]+)\)[^;]*;
612            #
613            # As it doesn't properly match the end parenthesis on some cases.
614            #
615            # So, a better solution was crafted: there's now a NestedMatch
616            # class that ensures that delimiters after a search are properly
617            # matched. So, the implementation to drop STRUCT_GROUP() will be
618            # handled in separate.
619
620            (KernRe(r'\bstruct_group\s*\(([^,]*,)', re.S), r'STRUCT_GROUP('),
621            (KernRe(r'\bstruct_group_attr\s*\(([^,]*,){2}', re.S), r'STRUCT_GROUP('),
622            (KernRe(r'\bstruct_group_tagged\s*\(([^,]*),([^,]*),', re.S), r'struct \1 \2; STRUCT_GROUP('),
623            (KernRe(r'\b__struct_group\s*\(([^,]*,){3}', re.S), r'STRUCT_GROUP('),
624
625            # Replace macros
626            #
627            # TODO: use NestedMatch for FOO($1, $2, ...) matches
628            #
629            # it is better to also move those to the NestedMatch logic,
630            # to ensure that parenthesis will be properly matched.
631
632            (KernRe(r'__ETHTOOL_DECLARE_LINK_MODE_MASK\s*\(([^\)]+)\)', re.S), r'DECLARE_BITMAP(\1, __ETHTOOL_LINK_MODE_MASK_NBITS)'),
633            (KernRe(r'DECLARE_PHY_INTERFACE_MASK\s*\(([^\)]+)\)', re.S), r'DECLARE_BITMAP(\1, PHY_INTERFACE_MODE_MAX)'),
634            (KernRe(r'DECLARE_BITMAP\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'unsigned long \1[BITS_TO_LONGS(\2)]'),
635            (KernRe(r'DECLARE_HASHTABLE\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'unsigned long \1[1 << ((\2) - 1)]'),
636            (KernRe(r'DECLARE_KFIFO\s*\(' + args_pattern + r',\s*' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\2 *\1'),
637            (KernRe(r'DECLARE_KFIFO_PTR\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\2 *\1'),
638            (KernRe(r'(?:__)?DECLARE_FLEX_ARRAY\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\1 \2[]'),
639            (KernRe(r'DEFINE_DMA_UNMAP_ADDR\s*\(' + args_pattern + r'\)', re.S), r'dma_addr_t \1'),
640            (KernRe(r'DEFINE_DMA_UNMAP_LEN\s*\(' + args_pattern + r'\)', re.S), r'__u32 \1'),
641            (KernRe(r'VIRTIO_DECLARE_FEATURES\s*\(' + args_pattern + r'\)', re.S), r'u64 \1; u64 \1_array[VIRTIO_FEATURES_DWORDS]'),
642        ]
643
644        # Regexes here are guaranteed to have the end limiter matching
645        # the start delimiter. Yet, right now, only one replace group
646        # is allowed.
647
648        sub_nested_prefixes = [
649            (re.compile(r'\bSTRUCT_GROUP\('), r'\1'),
650        ]
651
652        for search, sub in sub_prefixes:
653            members = search.sub(sub, members)
654
655        nested = NestedMatch()
656
657        for search, sub in sub_nested_prefixes:
658            members = nested.sub(search, sub, members)
659
660        # Keeps the original declaration as-is
661        declaration = members
662
663        # Split nested struct/union elements
664        #
665        # This loop was simpler at the original kernel-doc perl version, as
666        #   while ($members =~ m/$struct_members/) { ... }
667        # reads 'members' string on each interaction.
668        #
669        # Python behavior is different: it parses 'members' only once,
670        # creating a list of tuples from the first interaction.
671        #
672        # On other words, this won't get nested structs.
673        #
674        # So, we need to have an extra loop on Python to override such
675        # re limitation.
676
677        while True:
678            tuples = struct_members.findall(members)
679            if not tuples:
680                break
681
682            for t in tuples:
683                newmember = ""
684                maintype = t[0]
685                s_ids = t[5]
686                content = t[3]
687
688                oldmember = "".join(t)
689
690                for s_id in s_ids.split(','):
691                    s_id = s_id.strip()
692
693                    newmember += f"{maintype} {s_id}; "
694                    s_id = KernRe(r'[:\[].*').sub('', s_id)
695                    s_id = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', s_id)
696
697                    for arg in content.split(';'):
698                        arg = arg.strip()
699
700                        if not arg:
701                            continue
702
703                        r = KernRe(r'^([^\(]+\(\*?\s*)([\w\.]*)(\s*\).*)')
704                        if r.match(arg):
705                            # Pointer-to-function
706                            dtype = r.group(1)
707                            name = r.group(2)
708                            extra = r.group(3)
709
710                            if not name:
711                                continue
712
713                            if not s_id:
714                                # Anonymous struct/union
715                                newmember += f"{dtype}{name}{extra}; "
716                            else:
717                                newmember += f"{dtype}{s_id}.{name}{extra}; "
718
719                        else:
720                            arg = arg.strip()
721                            # Handle bitmaps
722                            arg = KernRe(r':\s*\d+\s*').sub('', arg)
723
724                            # Handle arrays
725                            arg = KernRe(r'\[.*\]').sub('', arg)
726
727                            # Handle multiple IDs
728                            arg = KernRe(r'\s*,\s*').sub(',', arg)
729
730                            r = KernRe(r'(.*)\s+([\S+,]+)')
731
732                            if r.search(arg):
733                                dtype = r.group(1)
734                                names = r.group(2)
735                            else:
736                                newmember += f"{arg}; "
737                                continue
738
739                            for name in names.split(','):
740                                name = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', name).strip()
741
742                                if not name:
743                                    continue
744
745                                if not s_id:
746                                    # Anonymous struct/union
747                                    newmember += f"{dtype} {name}; "
748                                else:
749                                    newmember += f"{dtype} {s_id}.{name}; "
750
751                members = members.replace(oldmember, newmember)
752
753        # Ignore other nested elements, like enums
754        members = re.sub(r'(\{[^\{\}]*\})', '', members)
755
756        self.create_parameter_list(ln, decl_type, members, ';',
757                                   declaration_name)
758        self.check_sections(ln, declaration_name, decl_type)
759
760        # Adjust declaration for better display
761        declaration = KernRe(r'([\{;])').sub(r'\1\n', declaration)
762        declaration = KernRe(r'\}\s+;').sub('};', declaration)
763
764        # Better handle inlined enums
765        while True:
766            r = KernRe(r'(enum\s+\{[^\}]+),([^\n])')
767            if not r.search(declaration):
768                break
769
770            declaration = r.sub(r'\1,\n\2', declaration)
771
772        def_args = declaration.split('\n')
773        level = 1
774        declaration = ""
775        for clause in def_args:
776
777            clause = clause.strip()
778            clause = KernRe(r'\s+').sub(' ', clause, count=1)
779
780            if not clause:
781                continue
782
783            if '}' in clause and level > 1:
784                level -= 1
785
786            if not KernRe(r'^\s*#').match(clause):
787                declaration += "\t" * level
788
789            declaration += "\t" + clause + "\n"
790            if "{" in clause and "}" not in clause:
791                level += 1
792
793        self.output_declaration(decl_type, declaration_name,
794                                definition=declaration,
795                                purpose=self.entry.declaration_purpose)
796
797    def dump_enum(self, ln, proto):
798        """
799        Stores an enum inside self.entries array.
800        """
801
802        # Ignore members marked private
803        proto = KernRe(r'\/\*\s*private:.*?\/\*\s*public:.*?\*\/', flags=re.S).sub('', proto)
804        proto = KernRe(r'\/\*\s*private:.*}', flags=re.S).sub('}', proto)
805
806        # Strip comments
807        proto = KernRe(r'\/\*.*?\*\/', flags=re.S).sub('', proto)
808
809        # Strip #define macros inside enums
810        proto = KernRe(r'#\s*((define|ifdef|if)\s+|endif)[^;]*;', flags=re.S).sub('', proto)
811
812        #
813        # Parse out the name and members of the enum.  Typedef form first.
814        #
815        r = KernRe(r'typedef\s+enum\s*\{(.*)\}\s*(\w*)\s*;')
816        if r.search(proto):
817            declaration_name = r.group(2)
818            members = r.group(1).rstrip()
819        #
820        # Failing that, look for a straight enum
821        #
822        else:
823            r = KernRe(r'enum\s+(\w*)\s*\{(.*)\}')
824            if r.match(proto):
825                declaration_name = r.group(1)
826                members = r.group(2).rstrip()
827        #
828        # OK, this isn't going to work.
829        #
830            else:
831                self.emit_msg(ln, f"{proto}: error: Cannot parse enum!")
832                return
833        #
834        # Make sure we found what we were expecting.
835        #
836        if self.entry.identifier != declaration_name:
837            if self.entry.identifier == "":
838                self.emit_msg(ln,
839                              f"{proto}: wrong kernel-doc identifier on prototype")
840            else:
841                self.emit_msg(ln,
842                              f"expecting prototype for enum {self.entry.identifier}. "
843                              f"Prototype was for enum {declaration_name} instead")
844            return
845
846        if not declaration_name:
847            declaration_name = "(anonymous)"
848        #
849        # Parse out the name of each enum member, and verify that we
850        # have a description for it.
851        #
852        member_set = set()
853        members = KernRe(r'\([^;)]*\)').sub('', members)
854        for arg in members.split(','):
855            if not arg:
856                continue
857            arg = KernRe(r'^\s*(\w+).*').sub(r'\1', arg)
858            self.entry.parameterlist.append(arg)
859            if arg not in self.entry.parameterdescs:
860                self.entry.parameterdescs[arg] = self.undescribed
861                self.emit_msg(ln,
862                              f"Enum value '{arg}' not described in enum '{declaration_name}'")
863            member_set.add(arg)
864        #
865        # Ensure that every described member actually exists in the enum.
866        #
867        for k in self.entry.parameterdescs:
868            if k not in member_set:
869                self.emit_msg(ln,
870                              f"Excess enum value '%{k}' description in '{declaration_name}'")
871
872        self.output_declaration('enum', declaration_name,
873                                purpose=self.entry.declaration_purpose)
874
875    def dump_declaration(self, ln, prototype):
876        """
877        Stores a data declaration inside self.entries array.
878        """
879
880        if self.entry.decl_type == "enum":
881            self.dump_enum(ln, prototype)
882        elif self.entry.decl_type == "typedef":
883            self.dump_typedef(ln, prototype)
884        elif self.entry.decl_type in ["union", "struct"]:
885            self.dump_struct(ln, prototype)
886        else:
887            # This would be a bug
888            self.emit_message(ln, f'Unknown declaration type: {self.entry.decl_type}')
889
890    def dump_function(self, ln, prototype):
891        """
892        Stores a function of function macro inside self.entries array.
893        """
894
895        func_macro = False
896        return_type = ''
897        decl_type = 'function'
898
899        # Prefixes that would be removed
900        sub_prefixes = [
901            (r"^static +", "", 0),
902            (r"^extern +", "", 0),
903            (r"^asmlinkage +", "", 0),
904            (r"^inline +", "", 0),
905            (r"^__inline__ +", "", 0),
906            (r"^__inline +", "", 0),
907            (r"^__always_inline +", "", 0),
908            (r"^noinline +", "", 0),
909            (r"^__FORTIFY_INLINE +", "", 0),
910            (r"__init +", "", 0),
911            (r"__init_or_module +", "", 0),
912            (r"__deprecated +", "", 0),
913            (r"__flatten +", "", 0),
914            (r"__meminit +", "", 0),
915            (r"__must_check +", "", 0),
916            (r"__weak +", "", 0),
917            (r"__sched +", "", 0),
918            (r"_noprof", "", 0),
919            (r"__printf\s*\(\s*\d*\s*,\s*\d*\s*\) +", "", 0),
920            (r"__(?:re)?alloc_size\s*\(\s*\d+\s*(?:,\s*\d+\s*)?\) +", "", 0),
921            (r"__diagnose_as\s*\(\s*\S+\s*(?:,\s*\d+\s*)*\) +", "", 0),
922            (r"DECL_BUCKET_PARAMS\s*\(\s*(\S+)\s*,\s*(\S+)\s*\)", r"\1, \2", 0),
923            (r"__attribute_const__ +", "", 0),
924
925            # It seems that Python support for re.X is broken:
926            # At least for me (Python 3.13), this didn't work
927#            (r"""
928#              __attribute__\s*\(\(
929#                (?:
930#                    [\w\s]+          # attribute name
931#                    (?:\([^)]*\))?   # attribute arguments
932#                    \s*,?            # optional comma at the end
933#                )+
934#              \)\)\s+
935#             """, "", re.X),
936
937            # So, remove whitespaces and comments from it
938            (r"__attribute__\s*\(\((?:[\w\s]+(?:\([^)]*\))?\s*,?)+\)\)\s+", "", 0),
939        ]
940
941        for search, sub, flags in sub_prefixes:
942            prototype = KernRe(search, flags).sub(sub, prototype)
943
944        # Macros are a special case, as they change the prototype format
945        new_proto = KernRe(r"^#\s*define\s+").sub("", prototype)
946        if new_proto != prototype:
947            is_define_proto = True
948            prototype = new_proto
949        else:
950            is_define_proto = False
951
952        # Yes, this truly is vile.  We are looking for:
953        # 1. Return type (may be nothing if we're looking at a macro)
954        # 2. Function name
955        # 3. Function parameters.
956        #
957        # All the while we have to watch out for function pointer parameters
958        # (which IIRC is what the two sections are for), C types (these
959        # regexps don't even start to express all the possibilities), and
960        # so on.
961        #
962        # If you mess with these regexps, it's a good idea to check that
963        # the following functions' documentation still comes out right:
964        # - parport_register_device (function pointer parameters)
965        # - atomic_set (macro)
966        # - pci_match_device, __copy_to_user (long return type)
967
968        name = r'[a-zA-Z0-9_~:]+'
969        prototype_end1 = r'[^\(]*'
970        prototype_end2 = r'[^\{]*'
971        prototype_end = fr'\(({prototype_end1}|{prototype_end2})\)'
972
973        # Besides compiling, Perl qr{[\w\s]+} works as a non-capturing group.
974        # So, this needs to be mapped in Python with (?:...)? or (?:...)+
975
976        type1 = r'(?:[\w\s]+)?'
977        type2 = r'(?:[\w\s]+\*+)+'
978
979        found = False
980
981        if is_define_proto:
982            r = KernRe(r'^()(' + name + r')\s+')
983
984            if r.search(prototype):
985                return_type = ''
986                declaration_name = r.group(2)
987                func_macro = True
988
989                found = True
990
991        if not found:
992            patterns = [
993                rf'^()({name})\s*{prototype_end}',
994                rf'^({type1})\s+({name})\s*{prototype_end}',
995                rf'^({type2})\s*({name})\s*{prototype_end}',
996            ]
997
998            for p in patterns:
999                r = KernRe(p)
1000
1001                if r.match(prototype):
1002
1003                    return_type = r.group(1)
1004                    declaration_name = r.group(2)
1005                    args = r.group(3)
1006
1007                    self.create_parameter_list(ln, decl_type, args, ',',
1008                                               declaration_name)
1009
1010                    found = True
1011                    break
1012        if not found:
1013            self.emit_msg(ln,
1014                          f"cannot understand function prototype: '{prototype}'")
1015            return
1016
1017        if self.entry.identifier != declaration_name:
1018            self.emit_msg(ln,
1019                          f"expecting prototype for {self.entry.identifier}(). Prototype was for {declaration_name}() instead")
1020            return
1021
1022        self.check_sections(ln, declaration_name, "function")
1023
1024        self.check_return_section(ln, declaration_name, return_type)
1025
1026        if 'typedef' in return_type:
1027            self.output_declaration(decl_type, declaration_name,
1028                                    typedef=True,
1029                                    functiontype=return_type,
1030                                    purpose=self.entry.declaration_purpose,
1031                                    func_macro=func_macro)
1032        else:
1033            self.output_declaration(decl_type, declaration_name,
1034                                    typedef=False,
1035                                    functiontype=return_type,
1036                                    purpose=self.entry.declaration_purpose,
1037                                    func_macro=func_macro)
1038
1039    def dump_typedef(self, ln, proto):
1040        """
1041        Stores a typedef inside self.entries array.
1042        """
1043
1044        typedef_type = r'((?:\s+[\w\*]+\b){0,7}\s+(?:\w+\b|\*+))\s*'
1045        typedef_ident = r'\*?\s*(\w\S+)\s*'
1046        typedef_args = r'\s*\((.*)\);'
1047
1048        typedef1 = KernRe(r'typedef' + typedef_type + r'\(' + typedef_ident + r'\)' + typedef_args)
1049        typedef2 = KernRe(r'typedef' + typedef_type + typedef_ident + typedef_args)
1050
1051        # Strip comments
1052        proto = KernRe(r'/\*.*?\*/', flags=re.S).sub('', proto)
1053
1054        # Parse function typedef prototypes
1055        for r in [typedef1, typedef2]:
1056            if not r.match(proto):
1057                continue
1058
1059            return_type = r.group(1).strip()
1060            declaration_name = r.group(2)
1061            args = r.group(3)
1062
1063            if self.entry.identifier != declaration_name:
1064                self.emit_msg(ln,
1065                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1066                return
1067
1068            decl_type = 'function'
1069            self.create_parameter_list(ln, decl_type, args, ',', declaration_name)
1070
1071            self.output_declaration(decl_type, declaration_name,
1072                                    typedef=True,
1073                                    functiontype=return_type,
1074                                    purpose=self.entry.declaration_purpose)
1075            return
1076
1077        # Handle nested parentheses or brackets
1078        r = KernRe(r'(\(*.\)\s*|\[*.\]\s*);$')
1079        while r.search(proto):
1080            proto = r.sub('', proto)
1081
1082        # Parse simple typedefs
1083        r = KernRe(r'typedef.*\s+(\w+)\s*;')
1084        if r.match(proto):
1085            declaration_name = r.group(1)
1086
1087            if self.entry.identifier != declaration_name:
1088                self.emit_msg(ln,
1089                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1090                return
1091
1092            self.output_declaration('typedef', declaration_name,
1093                                    purpose=self.entry.declaration_purpose)
1094            return
1095
1096        self.emit_msg(ln, "error: Cannot parse typedef!")
1097
1098    @staticmethod
1099    def process_export(function_set, line):
1100        """
1101        process EXPORT_SYMBOL* tags
1102
1103        This method doesn't use any variable from the class, so declare it
1104        with a staticmethod decorator.
1105        """
1106
1107        # We support documenting some exported symbols with different
1108        # names.  A horrible hack.
1109        suffixes = [ '_noprof' ]
1110
1111        # Note: it accepts only one EXPORT_SYMBOL* per line, as having
1112        # multiple export lines would violate Kernel coding style.
1113
1114        if export_symbol.search(line):
1115            symbol = export_symbol.group(2)
1116        elif export_symbol_ns.search(line):
1117            symbol = export_symbol_ns.group(2)
1118        else:
1119            return False
1120        #
1121        # Found an export, trim out any special suffixes
1122        #
1123        for suffix in suffixes:
1124            # Be backward compatible with Python < 3.9
1125            if symbol.endswith(suffix):
1126                symbol = symbol[:-len(suffix)]
1127        function_set.add(symbol)
1128        return True
1129
1130    def process_normal(self, ln, line):
1131        """
1132        STATE_NORMAL: looking for the /** to begin everything.
1133        """
1134
1135        if not doc_start.match(line):
1136            return
1137
1138        # start a new entry
1139        self.reset_state(ln)
1140
1141        # next line is always the function name
1142        self.state = state.NAME
1143
1144    def process_name(self, ln, line):
1145        """
1146        STATE_NAME: Looking for the "name - description" line
1147        """
1148        #
1149        # Check for a DOC: block and handle them specially.
1150        #
1151        if doc_block.search(line):
1152
1153            if not doc_block.group(1):
1154                self.entry.begin_section(ln, "Introduction")
1155            else:
1156                self.entry.begin_section(ln, doc_block.group(1))
1157
1158            self.entry.identifier = self.entry.section
1159            self.state = state.DOCBLOCK
1160        #
1161        # Otherwise we're looking for a normal kerneldoc declaration line.
1162        #
1163        elif doc_decl.search(line):
1164            self.entry.identifier = doc_decl.group(1)
1165
1166            # Test for data declaration
1167            if doc_begin_data.search(line):
1168                self.entry.decl_type = doc_begin_data.group(1)
1169                self.entry.identifier = doc_begin_data.group(2)
1170            #
1171            # Look for a function description
1172            #
1173            elif doc_begin_func.search(line):
1174                self.entry.identifier = doc_begin_func.group(1)
1175                self.entry.decl_type = "function"
1176            #
1177            # We struck out.
1178            #
1179            else:
1180                self.emit_msg(ln,
1181                              f"This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst\n{line}")
1182                self.state = state.NORMAL
1183                return
1184            #
1185            # OK, set up for a new kerneldoc entry.
1186            #
1187            self.state = state.BODY
1188            self.entry.identifier = self.entry.identifier.strip(" ")
1189            # if there's no @param blocks need to set up default section here
1190            self.entry.begin_section(ln + 1)
1191            #
1192            # Find the description portion, which *should* be there but
1193            # isn't always.
1194            # (We should be able to capture this from the previous parsing - someday)
1195            #
1196            r = KernRe("[-:](.*)")
1197            if r.search(line):
1198                self.entry.declaration_purpose = trim_whitespace(r.group(1))
1199                self.state = state.DECLARATION
1200            else:
1201                self.entry.declaration_purpose = ""
1202
1203            if not self.entry.declaration_purpose and self.config.wshort_desc:
1204                self.emit_msg(ln,
1205                              f"missing initial short description on line:\n{line}")
1206
1207            if not self.entry.identifier and self.entry.decl_type != "enum":
1208                self.emit_msg(ln,
1209                              f"wrong kernel-doc identifier on line:\n{line}")
1210                self.state = state.NORMAL
1211
1212            if self.config.verbose:
1213                self.emit_msg(ln,
1214                              f"Scanning doc for {self.entry.decl_type} {self.entry.identifier}",
1215                                  warning=False)
1216        #
1217        # Failed to find an identifier. Emit a warning
1218        #
1219        else:
1220            self.emit_msg(ln, f"Cannot find identifier on line:\n{line}")
1221
1222    #
1223    # Helper function to determine if a new section is being started.
1224    #
1225    def is_new_section(self, ln, line):
1226        if doc_sect.search(line):
1227            self.state = state.BODY
1228            #
1229            # Pick out the name of our new section, tweaking it if need be.
1230            #
1231            newsection = doc_sect.group(1)
1232            if newsection.lower() == 'description':
1233                newsection = 'Description'
1234            elif newsection.lower() == 'context':
1235                newsection = 'Context'
1236                self.state = state.SPECIAL_SECTION
1237            elif newsection.lower() in ["@return", "@returns",
1238                                        "return", "returns"]:
1239                newsection = "Return"
1240                self.state = state.SPECIAL_SECTION
1241            elif newsection[0] == '@':
1242                self.state = state.SPECIAL_SECTION
1243            #
1244            # Initialize the contents, and get the new section going.
1245            #
1246            newcontents = doc_sect.group(2)
1247            if not newcontents:
1248                newcontents = ""
1249            self.dump_section()
1250            self.entry.begin_section(ln, newsection)
1251            self.entry.leading_space = None
1252
1253            self.entry.add_text(newcontents.lstrip())
1254            return True
1255        return False
1256
1257    #
1258    # Helper function to detect (and effect) the end of a kerneldoc comment.
1259    #
1260    def is_comment_end(self, ln, line):
1261        if doc_end.search(line):
1262            self.dump_section()
1263
1264            # Look for doc_com + <text> + doc_end:
1265            r = KernRe(r'\s*\*\s*[a-zA-Z_0-9:\.]+\*/')
1266            if r.match(line):
1267                self.emit_msg(ln, f"suspicious ending line: {line}")
1268
1269            self.entry.prototype = ""
1270            self.entry.new_start_line = ln + 1
1271
1272            self.state = state.PROTO
1273            return True
1274        return False
1275
1276
1277    def process_decl(self, ln, line):
1278        """
1279        STATE_DECLARATION: We've seen the beginning of a declaration
1280        """
1281        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1282            return
1283        #
1284        # Look for anything with the " * " line beginning.
1285        #
1286        if doc_content.search(line):
1287            cont = doc_content.group(1)
1288            #
1289            # A blank line means that we have moved out of the declaration
1290            # part of the comment (without any "special section" parameter
1291            # descriptions).
1292            #
1293            if cont == "":
1294                self.state = state.BODY
1295            #
1296            # Otherwise we have more of the declaration section to soak up.
1297            #
1298            else:
1299                self.entry.declaration_purpose = \
1300                    trim_whitespace(self.entry.declaration_purpose + ' ' + cont)
1301        else:
1302            # Unknown line, ignore
1303            self.emit_msg(ln, f"bad line: {line}")
1304
1305
1306    def process_special(self, ln, line):
1307        """
1308        STATE_SPECIAL_SECTION: a section ending with a blank line
1309        """
1310        #
1311        # If we have hit a blank line (only the " * " marker), then this
1312        # section is done.
1313        #
1314        if KernRe(r"\s*\*\s*$").match(line):
1315            self.entry.begin_section(ln, dump = True)
1316            self.state = state.BODY
1317            return
1318        #
1319        # Not a blank line, look for the other ways to end the section.
1320        #
1321        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1322            return
1323        #
1324        # OK, we should have a continuation of the text for this section.
1325        #
1326        if doc_content.search(line):
1327            cont = doc_content.group(1)
1328            #
1329            # If the lines of text after the first in a special section have
1330            # leading white space, we need to trim it out or Sphinx will get
1331            # confused.  For the second line (the None case), see what we
1332            # find there and remember it.
1333            #
1334            if self.entry.leading_space is None:
1335                r = KernRe(r'^(\s+)')
1336                if r.match(cont):
1337                    self.entry.leading_space = len(r.group(1))
1338                else:
1339                    self.entry.leading_space = 0
1340            #
1341            # Otherwise, before trimming any leading chars, be *sure*
1342            # that they are white space.  We should maybe warn if this
1343            # isn't the case.
1344            #
1345            for i in range(0, self.entry.leading_space):
1346                if cont[i] != " ":
1347                    self.entry.leading_space = i
1348                    break
1349            #
1350            # Add the trimmed result to the section and we're done.
1351            #
1352            self.entry.add_text(cont[self.entry.leading_space:])
1353        else:
1354            # Unknown line, ignore
1355            self.emit_msg(ln, f"bad line: {line}")
1356
1357    def process_body(self, ln, line):
1358        """
1359        STATE_BODY: the bulk of a kerneldoc comment.
1360        """
1361        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1362            return
1363
1364        if doc_content.search(line):
1365            cont = doc_content.group(1)
1366            self.entry.add_text(cont)
1367        else:
1368            # Unknown line, ignore
1369            self.emit_msg(ln, f"bad line: {line}")
1370
1371    def process_inline_name(self, ln, line):
1372        """STATE_INLINE_NAME: beginning of docbook comments within a prototype."""
1373
1374        if doc_inline_sect.search(line):
1375            self.entry.begin_section(ln, doc_inline_sect.group(1))
1376            self.entry.add_text(doc_inline_sect.group(2).lstrip())
1377            self.state = state.INLINE_TEXT
1378        elif doc_inline_end.search(line):
1379            self.dump_section()
1380            self.state = state.PROTO
1381        elif doc_content.search(line):
1382            self.emit_msg(ln, f"Incorrect use of kernel-doc format: {line}")
1383            self.state = state.PROTO
1384        # else ... ??
1385
1386    def process_inline_text(self, ln, line):
1387        """STATE_INLINE_TEXT: docbook comments within a prototype."""
1388
1389        if doc_inline_end.search(line):
1390            self.dump_section()
1391            self.state = state.PROTO
1392        elif doc_content.search(line):
1393            self.entry.add_text(doc_content.group(1))
1394        # else ... ??
1395
1396    def syscall_munge(self, ln, proto):         # pylint: disable=W0613
1397        """
1398        Handle syscall definitions
1399        """
1400
1401        is_void = False
1402
1403        # Strip newlines/CR's
1404        proto = re.sub(r'[\r\n]+', ' ', proto)
1405
1406        # Check if it's a SYSCALL_DEFINE0
1407        if 'SYSCALL_DEFINE0' in proto:
1408            is_void = True
1409
1410        # Replace SYSCALL_DEFINE with correct return type & function name
1411        proto = KernRe(r'SYSCALL_DEFINE.*\(').sub('long sys_', proto)
1412
1413        r = KernRe(r'long\s+(sys_.*?),')
1414        if r.search(proto):
1415            proto = KernRe(',').sub('(', proto, count=1)
1416        elif is_void:
1417            proto = KernRe(r'\)').sub('(void)', proto, count=1)
1418
1419        # Now delete all of the odd-numbered commas in the proto
1420        # so that argument types & names don't have a comma between them
1421        count = 0
1422        length = len(proto)
1423
1424        if is_void:
1425            length = 0  # skip the loop if is_void
1426
1427        for ix in range(length):
1428            if proto[ix] == ',':
1429                count += 1
1430                if count % 2 == 1:
1431                    proto = proto[:ix] + ' ' + proto[ix + 1:]
1432
1433        return proto
1434
1435    def tracepoint_munge(self, ln, proto):
1436        """
1437        Handle tracepoint definitions
1438        """
1439
1440        tracepointname = None
1441        tracepointargs = None
1442
1443        # Match tracepoint name based on different patterns
1444        r = KernRe(r'TRACE_EVENT\((.*?),')
1445        if r.search(proto):
1446            tracepointname = r.group(1)
1447
1448        r = KernRe(r'DEFINE_SINGLE_EVENT\((.*?),')
1449        if r.search(proto):
1450            tracepointname = r.group(1)
1451
1452        r = KernRe(r'DEFINE_EVENT\((.*?),(.*?),')
1453        if r.search(proto):
1454            tracepointname = r.group(2)
1455
1456        if tracepointname:
1457            tracepointname = tracepointname.lstrip()
1458
1459        r = KernRe(r'TP_PROTO\((.*?)\)')
1460        if r.search(proto):
1461            tracepointargs = r.group(1)
1462
1463        if not tracepointname or not tracepointargs:
1464            self.emit_msg(ln,
1465                          f"Unrecognized tracepoint format:\n{proto}\n")
1466        else:
1467            proto = f"static inline void trace_{tracepointname}({tracepointargs})"
1468            self.entry.identifier = f"trace_{self.entry.identifier}"
1469
1470        return proto
1471
1472    def process_proto_function(self, ln, line):
1473        """Ancillary routine to process a function prototype"""
1474
1475        # strip C99-style comments to end of line
1476        line = KernRe(r"\/\/.*$", re.S).sub('', line)
1477        #
1478        # Soak up the line's worth of prototype text, stopping at { or ; if present.
1479        #
1480        if KernRe(r'\s*#\s*define').match(line):
1481            self.entry.prototype = line
1482        elif not line.startswith('#'):   # skip other preprocessor stuff
1483            r = KernRe(r'([^\{]*)')
1484            if r.match(line):
1485                self.entry.prototype += r.group(1) + " "
1486        #
1487        # If we now have the whole prototype, clean it up and declare victory.
1488        #
1489        if '{' in line or ';' in line or KernRe(r'\s*#\s*define').match(line):
1490            # strip comments and surrounding spaces
1491            self.entry.prototype = KernRe(r'/\*.*\*/').sub('', self.entry.prototype).strip()
1492            #
1493            # Handle self.entry.prototypes for function pointers like:
1494            #       int (*pcs_config)(struct foo)
1495            # by turning it into
1496            #	    int pcs_config(struct foo)
1497            #
1498            r = KernRe(r'^(\S+\s+)\(\s*\*(\S+)\)')
1499            self.entry.prototype = r.sub(r'\1\2', self.entry.prototype)
1500            #
1501            # Handle special declaration syntaxes
1502            #
1503            if 'SYSCALL_DEFINE' in self.entry.prototype:
1504                self.entry.prototype = self.syscall_munge(ln,
1505                                                          self.entry.prototype)
1506            else:
1507                r = KernRe(r'TRACE_EVENT|DEFINE_EVENT|DEFINE_SINGLE_EVENT')
1508                if r.search(self.entry.prototype):
1509                    self.entry.prototype = self.tracepoint_munge(ln,
1510                                                                 self.entry.prototype)
1511            #
1512            # ... and we're done
1513            #
1514            self.dump_function(ln, self.entry.prototype)
1515            self.reset_state(ln)
1516
1517    def process_proto_type(self, ln, line):
1518        """Ancillary routine to process a type"""
1519
1520        # Strip C99-style comments and surrounding whitespace
1521        line = KernRe(r"//.*$", re.S).sub('', line).strip()
1522        if not line:
1523            return # nothing to see here
1524
1525        # To distinguish preprocessor directive from regular declaration later.
1526        if line.startswith('#'):
1527            line += ";"
1528        #
1529        # Split the declaration on any of { } or ;, and accumulate pieces
1530        # until we hit a semicolon while not inside {brackets}
1531        #
1532        r = KernRe(r'(.*?)([{};])')
1533        for chunk in r.split(line):
1534            if chunk:  # Ignore empty matches
1535                self.entry.prototype += chunk
1536                #
1537                # This cries out for a match statement ... someday after we can
1538                # drop Python 3.9 ...
1539                #
1540                if chunk == '{':
1541                    self.entry.brcount += 1
1542                elif chunk == '}':
1543                    self.entry.brcount -= 1
1544                elif chunk == ';' and self.entry.brcount <= 0:
1545                    self.dump_declaration(ln, self.entry.prototype)
1546                    self.reset_state(ln)
1547                    return
1548        #
1549        # We hit the end of the line while still in the declaration; put
1550        # in a space to represent the newline.
1551        #
1552        self.entry.prototype += ' '
1553
1554    def process_proto(self, ln, line):
1555        """STATE_PROTO: reading a function/whatever prototype."""
1556
1557        if doc_inline_oneline.search(line):
1558            self.entry.begin_section(ln, doc_inline_oneline.group(1))
1559            self.entry.add_text(doc_inline_oneline.group(2))
1560            self.dump_section()
1561
1562        elif doc_inline_start.search(line):
1563            self.state = state.INLINE_NAME
1564
1565        elif self.entry.decl_type == 'function':
1566            self.process_proto_function(ln, line)
1567
1568        else:
1569            self.process_proto_type(ln, line)
1570
1571    def process_docblock(self, ln, line):
1572        """STATE_DOCBLOCK: within a DOC: block."""
1573
1574        if doc_end.search(line):
1575            self.dump_section()
1576            self.output_declaration("doc", self.entry.identifier)
1577            self.reset_state(ln)
1578
1579        elif doc_content.search(line):
1580            self.entry.add_text(doc_content.group(1))
1581
1582    def parse_export(self):
1583        """
1584        Parses EXPORT_SYMBOL* macros from a single Kernel source file.
1585        """
1586
1587        export_table = set()
1588
1589        try:
1590            with open(self.fname, "r", encoding="utf8",
1591                      errors="backslashreplace") as fp:
1592
1593                for line in fp:
1594                    self.process_export(export_table, line)
1595
1596        except IOError:
1597            return None
1598
1599        return export_table
1600
1601    #
1602    # The state/action table telling us which function to invoke in
1603    # each state.
1604    #
1605    state_actions = {
1606        state.NORMAL:			process_normal,
1607        state.NAME:			process_name,
1608        state.BODY:			process_body,
1609        state.DECLARATION:		process_decl,
1610        state.SPECIAL_SECTION:		process_special,
1611        state.INLINE_NAME:		process_inline_name,
1612        state.INLINE_TEXT:		process_inline_text,
1613        state.PROTO:			process_proto,
1614        state.DOCBLOCK:			process_docblock,
1615        }
1616
1617    def parse_kdoc(self):
1618        """
1619        Open and process each line of a C source file.
1620        The parsing is controlled via a state machine, and the line is passed
1621        to a different process function depending on the state. The process
1622        function may update the state as needed.
1623
1624        Besides parsing kernel-doc tags, it also parses export symbols.
1625        """
1626
1627        prev = ""
1628        prev_ln = None
1629        export_table = set()
1630
1631        try:
1632            with open(self.fname, "r", encoding="utf8",
1633                      errors="backslashreplace") as fp:
1634                for ln, line in enumerate(fp):
1635
1636                    line = line.expandtabs().strip("\n")
1637
1638                    # Group continuation lines on prototypes
1639                    if self.state == state.PROTO:
1640                        if line.endswith("\\"):
1641                            prev += line.rstrip("\\")
1642                            if not prev_ln:
1643                                prev_ln = ln
1644                            continue
1645
1646                        if prev:
1647                            ln = prev_ln
1648                            line = prev + line
1649                            prev = ""
1650                            prev_ln = None
1651
1652                    self.config.log.debug("%d %s: %s",
1653                                          ln, state.name[self.state],
1654                                          line)
1655
1656                    # This is an optimization over the original script.
1657                    # There, when export_file was used for the same file,
1658                    # it was read twice. Here, we use the already-existing
1659                    # loop to parse exported symbols as well.
1660                    #
1661                    if (self.state != state.NORMAL) or \
1662                       not self.process_export(export_table, line):
1663                        # Hand this line to the appropriate state handler
1664                        self.state_actions[self.state](self, ln, line)
1665
1666        except OSError:
1667            self.config.log.error(f"Error: Cannot open file {self.fname}")
1668
1669        return export_table, self.entries
1670