xref: /linux/scripts/lib/kdoc/kdoc_parser.py (revision 636d4d9ec641025b98e8df4623a77ecc09026209)
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
19from kdoc_item import KdocItem
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*$', cache=False)
32
33doc_end = KernRe(r'\*/', cache=False)
34doc_com = KernRe(r'\s*\*\s*', cache=False)
35doc_com_body = KernRe(r'\s*\* ?', cache=False)
36doc_decl = doc_com + KernRe(r'(\w+)', cache=False)
37
38# @params and a strictly limited set of supported section names
39# Specifically:
40#   Match @word:
41#         @...:
42#         @{section-name}:
43# while trying to not match literal block starts like "example::"
44#
45known_section_names = 'description|context|returns?|notes?|examples?'
46known_sections = KernRe(known_section_names, flags = re.I)
47doc_sect = doc_com + \
48    KernRe(r'\s*(\@[.\w]+|\@\.\.\.|' + known_section_names + r')\s*:([^:].*)?$',
49           flags=re.I, cache=False)
50
51doc_content = doc_com_body + KernRe(r'(.*)', cache=False)
52doc_inline_start = KernRe(r'^\s*/\*\*\s*$', cache=False)
53doc_inline_sect = KernRe(r'\s*\*\s*(@\s*[\w][\w\.]*\s*):(.*)', cache=False)
54doc_inline_end = KernRe(r'^\s*\*/\s*$', cache=False)
55doc_inline_oneline = KernRe(r'^\s*/\*\*\s*(@[\w\s]+):\s*(.*)\s*\*/\s*$', cache=False)
56attribute = KernRe(r"__attribute__\s*\(\([a-z0-9,_\*\s\(\)]*\)\)",
57               flags=re.I | re.S, cache=False)
58
59export_symbol = KernRe(r'^\s*EXPORT_SYMBOL(_GPL)?\s*\(\s*(\w+)\s*\)\s*', cache=False)
60export_symbol_ns = KernRe(r'^\s*EXPORT_SYMBOL_NS(_GPL)?\s*\(\s*(\w+)\s*,\s*"\S+"\)\s*', cache=False)
61
62type_param = KernRe(r"\@(\w*((\.\w+)|(->\w+))*(\.\.\.)?)", cache=False)
63
64#
65# Tests for the beginning of a kerneldoc block in its various forms.
66#
67doc_block = doc_com + KernRe(r'DOC:\s*(.*)?', cache=False)
68doc_begin_data = KernRe(r"^\s*\*?\s*(struct|union|enum|typedef)\b\s*(\w*)", cache = False)
69doc_begin_func = KernRe(str(doc_com) +			# initial " * '
70                        r"(?:\w+\s*\*\s*)?" + 		# type (not captured)
71                        r'(?:define\s+)?' + 		# possible "define" (not captured)
72                        r'(\w+)\s*(?:\(\w*\))?\s*' +	# name and optional "(...)"
73                        r'(?:[-:].*)?$',		# description (not captured)
74                        cache = False)
75
76#
77# A little helper to get rid of excess white space
78#
79multi_space = KernRe(r'\s\s+')
80def trim_whitespace(s):
81    return multi_space.sub(' ', s.strip())
82
83class state:
84    """
85    State machine enums
86    """
87
88    # Parser states
89    NORMAL        = 0        # normal code
90    NAME          = 1        # looking for function name
91    DECLARATION   = 2        # We have seen a declaration which might not be done
92    BODY          = 3        # the body of the comment
93    SPECIAL_SECTION = 4      # doc section ending with a blank line
94    PROTO         = 5        # scanning prototype
95    DOCBLOCK      = 6        # documentation block
96    INLINE_NAME   = 7        # gathering doc outside main block
97    INLINE_TEXT   = 8	     # reading the body of inline docs
98
99    name = [
100        "NORMAL",
101        "NAME",
102        "DECLARATION",
103        "BODY",
104        "SPECIAL_SECTION",
105        "PROTO",
106        "DOCBLOCK",
107        "INLINE_NAME",
108        "INLINE_TEXT",
109    ]
110
111
112SECTION_DEFAULT = "Description"  # default section
113
114class KernelEntry:
115
116    def __init__(self, config, ln):
117        self.config = config
118
119        self._contents = []
120        self.prototype = ""
121
122        self.warnings = []
123
124        self.parameterlist = []
125        self.parameterdescs = {}
126        self.parametertypes = {}
127        self.parameterdesc_start_lines = {}
128
129        self.section_start_lines = {}
130        self.sections = {}
131
132        self.anon_struct_union = False
133
134        self.leading_space = None
135
136        # State flags
137        self.brcount = 0
138        self.declaration_start_line = ln + 1
139
140    #
141    # Management of section contents
142    #
143    def add_text(self, text):
144        self._contents.append(text)
145
146    def contents(self):
147        return '\n'.join(self._contents) + '\n'
148
149    # TODO: rename to emit_message after removal of kernel-doc.pl
150    def emit_msg(self, log_msg, warning=True):
151        """Emit a message"""
152
153        if not warning:
154            self.config.log.info(log_msg)
155            return
156
157        # Delegate warning output to output logic, as this way it
158        # will report warnings/info only for symbols that are output
159
160        self.warnings.append(log_msg)
161        return
162
163    #
164    # Begin a new section.
165    #
166    def begin_section(self, line_no, title = SECTION_DEFAULT, dump = False):
167        if dump:
168            self.dump_section(start_new = True)
169        self.section = title
170        self.new_start_line = line_no
171
172    def dump_section(self, start_new=True):
173        """
174        Dumps section contents to arrays/hashes intended for that purpose.
175        """
176        #
177        # If we have accumulated no contents in the default ("description")
178        # section, don't bother.
179        #
180        if self.section == SECTION_DEFAULT and not self._contents:
181            return
182        name = self.section
183        contents = self.contents()
184
185        if type_param.match(name):
186            name = type_param.group(1)
187
188            self.parameterdescs[name] = contents
189            self.parameterdesc_start_lines[name] = self.new_start_line
190
191            self.new_start_line = 0
192
193        else:
194            if name in self.sections and self.sections[name] != "":
195                # Only warn on user-specified duplicate section names
196                if name != SECTION_DEFAULT:
197                    self.emit_msg(self.new_start_line,
198                                  f"duplicate section name '{name}'\n")
199                # Treat as a new paragraph - add a blank line
200                self.sections[name] += '\n' + contents
201            else:
202                self.sections[name] = contents
203                self.section_start_lines[name] = self.new_start_line
204                self.new_start_line = 0
205
206#        self.config.log.debug("Section: %s : %s", name, pformat(vars(self)))
207
208        if start_new:
209            self.section = SECTION_DEFAULT
210            self._contents = []
211
212
213class KernelDoc:
214    """
215    Read a C language source or header FILE and extract embedded
216    documentation comments.
217    """
218
219    # Section names
220
221    section_context = "Context"
222    section_return = "Return"
223
224    undescribed = "-- undescribed --"
225
226    def __init__(self, config, fname):
227        """Initialize internal variables"""
228
229        self.fname = fname
230        self.config = config
231
232        # Initial state for the state machines
233        self.state = state.NORMAL
234
235        # Store entry currently being processed
236        self.entry = None
237
238        # Place all potential outputs into an array
239        self.entries = []
240
241    def emit_msg(self, ln, msg, warning=True):
242        """Emit a message"""
243
244        log_msg = f"{self.fname}:{ln} {msg}"
245
246        if self.entry:
247            self.entry.emit_msg(log_msg, warning)
248            return
249
250        if warning:
251            self.config.log.warning(log_msg)
252        else:
253            self.config.log.info(log_msg)
254
255    def dump_section(self, start_new=True):
256        """
257        Dumps section contents to arrays/hashes intended for that purpose.
258        """
259
260        if self.entry:
261            self.entry.dump_section(start_new)
262
263    # TODO: rename it to store_declaration after removal of kernel-doc.pl
264    def output_declaration(self, dtype, name, **args):
265        """
266        Stores the entry into an entry array.
267
268        The actual output and output filters will be handled elsewhere
269        """
270
271        item = KdocItem(name, dtype, self.entry.declaration_start_line, **args)
272        item.warnings = self.entry.warnings
273
274        # Drop empty sections
275        # TODO: improve empty sections logic to emit warnings
276        sections = self.entry.sections
277        for section in ["Description", "Return"]:
278            if section in sections and not sections[section].rstrip():
279                del sections[section]
280        item.set_sections(sections, self.entry.section_start_lines)
281        item.set_params(self.entry.parameterlist, self.entry.parameterdescs,
282                        self.entry.parametertypes,
283                        self.entry.parameterdesc_start_lines)
284        self.entries.append(item)
285
286        self.config.log.debug("Output: %s:%s = %s", dtype, name, pformat(args))
287
288    def reset_state(self, ln):
289        """
290        Ancillary routine to create a new entry. It initializes all
291        variables used by the state machine.
292        """
293
294        self.entry = KernelEntry(self.config, ln)
295
296        # State flags
297        self.state = state.NORMAL
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
371    def create_parameter_list(self, ln, decl_type, args,
372                              splitter, declaration_name):
373        """
374        Creates a list of parameters, storing them at self.entry.
375        """
376
377        # temporarily replace all commas inside function pointer definition
378        arg_expr = KernRe(r'(\([^\),]+),')
379        while arg_expr.search(args):
380            args = arg_expr.sub(r"\1#", args)
381
382        for arg in args.split(splitter):
383            # Strip comments
384            arg = KernRe(r'\/\*.*\*\/').sub('', arg)
385
386            # Ignore argument attributes
387            arg = KernRe(r'\sPOS0?\s').sub(' ', arg)
388
389            # Strip leading/trailing spaces
390            arg = arg.strip()
391            arg = KernRe(r'\s+').sub(' ', arg, count=1)
392
393            if arg.startswith('#'):
394                # Treat preprocessor directive as a typeless variable just to fill
395                # corresponding data structures "correctly". Catch it later in
396                # output_* subs.
397
398                # Treat preprocessor directive as a typeless variable
399                self.push_parameter(ln, decl_type, arg, "",
400                                    "", declaration_name)
401
402            elif KernRe(r'\(.+\)\s*\(').search(arg):
403                # Pointer-to-function
404
405                arg = arg.replace('#', ',')
406
407                r = KernRe(r'[^\(]+\(\*?\s*([\w\[\]\.]*)\s*\)')
408                if r.match(arg):
409                    param = r.group(1)
410                else:
411                    self.emit_msg(ln, f"Invalid param: {arg}")
412                    param = arg
413
414                dtype = KernRe(r'([^\(]+\(\*?)\s*' + re.escape(param)).sub(r'\1', arg)
415                self.push_parameter(ln, decl_type, param, dtype,
416                                    arg, declaration_name)
417
418            elif KernRe(r'\(.+\)\s*\[').search(arg):
419                # Array-of-pointers
420
421                arg = arg.replace('#', ',')
422                r = KernRe(r'[^\(]+\(\s*\*\s*([\w\[\]\.]*?)\s*(\s*\[\s*[\w]+\s*\]\s*)*\)')
423                if r.match(arg):
424                    param = r.group(1)
425                else:
426                    self.emit_msg(ln, f"Invalid param: {arg}")
427                    param = arg
428
429                dtype = KernRe(r'([^\(]+\(\*?)\s*' + re.escape(param)).sub(r'\1', arg)
430
431                self.push_parameter(ln, decl_type, param, dtype,
432                                    arg, declaration_name)
433
434            elif arg:
435                arg = KernRe(r'\s*:\s*').sub(":", arg)
436                arg = KernRe(r'\s*\[').sub('[', arg)
437
438                args = KernRe(r'\s*,\s*').split(arg)
439                if args[0] and '*' in args[0]:
440                    args[0] = re.sub(r'(\*+)\s*', r' \1', args[0])
441
442                first_arg = []
443                r = KernRe(r'^(.*\s+)(.*?\[.*\].*)$')
444                if args[0] and r.match(args[0]):
445                    args.pop(0)
446                    first_arg.extend(r.group(1))
447                    first_arg.append(r.group(2))
448                else:
449                    first_arg = KernRe(r'\s+').split(args.pop(0))
450
451                args.insert(0, first_arg.pop())
452                dtype = ' '.join(first_arg)
453
454                for param in args:
455                    if KernRe(r'^(\*+)\s*(.*)').match(param):
456                        r = KernRe(r'^(\*+)\s*(.*)')
457                        if not r.match(param):
458                            self.emit_msg(ln, f"Invalid param: {param}")
459                            continue
460
461                        param = r.group(1)
462
463                        self.push_parameter(ln, decl_type, r.group(2),
464                                            f"{dtype} {r.group(1)}",
465                                            arg, declaration_name)
466
467                    elif KernRe(r'(.*?):(\w+)').search(param):
468                        r = KernRe(r'(.*?):(\w+)')
469                        if not r.match(param):
470                            self.emit_msg(ln, f"Invalid param: {param}")
471                            continue
472
473                        if dtype != "":  # Skip unnamed bit-fields
474                            self.push_parameter(ln, decl_type, r.group(1),
475                                                f"{dtype}:{r.group(2)}",
476                                                arg, declaration_name)
477                    else:
478                        self.push_parameter(ln, decl_type, param, dtype,
479                                            arg, declaration_name)
480
481    def check_sections(self, ln, decl_name, decl_type):
482        """
483        Check for errors inside sections, emitting warnings if not found
484        parameters are described.
485        """
486        for section in self.entry.sections:
487            if section not in self.entry.parameterlist and \
488               not known_sections.search(section):
489                if decl_type == 'function':
490                    dname = f"{decl_type} parameter"
491                else:
492                    dname = f"{decl_type} member"
493                self.emit_msg(ln,
494                              f"Excess {dname} '{section}' description in '{decl_name}'")
495
496    def check_return_section(self, ln, declaration_name, return_type):
497        """
498        If the function doesn't return void, warns about the lack of a
499        return description.
500        """
501
502        if not self.config.wreturn:
503            return
504
505        # Ignore an empty return type (It's a macro)
506        # Ignore functions with a "void" return type (but not "void *")
507        if not return_type or KernRe(r'void\s*\w*\s*$').search(return_type):
508            return
509
510        if not self.entry.sections.get("Return", None):
511            self.emit_msg(ln,
512                          f"No description found for return value of '{declaration_name}'")
513
514    def dump_struct(self, ln, proto):
515        """
516        Store an entry for an struct or union
517        """
518
519        type_pattern = r'(struct|union)'
520
521        qualifiers = [
522            "__attribute__",
523            "__packed",
524            "__aligned",
525            "____cacheline_aligned_in_smp",
526            "____cacheline_aligned",
527        ]
528
529        definition_body = r'\{(.*)\}\s*' + "(?:" + '|'.join(qualifiers) + ")?"
530        struct_members = KernRe(type_pattern + r'([^\{\};]+)(\{)([^\{\}]*)(\})([^\{\}\;]*)(\;)')
531
532        # Extract struct/union definition
533        members = None
534        declaration_name = None
535        decl_type = None
536
537        r = KernRe(type_pattern + r'\s+(\w+)\s*' + definition_body)
538        if r.search(proto):
539            decl_type = r.group(1)
540            declaration_name = r.group(2)
541            members = r.group(3)
542        else:
543            r = KernRe(r'typedef\s+' + type_pattern + r'\s*' + definition_body + r'\s*(\w+)\s*;')
544
545            if r.search(proto):
546                decl_type = r.group(1)
547                declaration_name = r.group(3)
548                members = r.group(2)
549
550        if not members:
551            self.emit_msg(ln, f"{proto} error: Cannot parse struct or union!")
552            return
553
554        if self.entry.identifier != declaration_name:
555            self.emit_msg(ln,
556                          f"expecting prototype for {decl_type} {self.entry.identifier}. Prototype was for {decl_type} {declaration_name} instead\n")
557            return
558
559        args_pattern = r'([^,)]+)'
560
561        sub_prefixes = [
562            (KernRe(r'\/\*\s*private:.*?\/\*\s*public:.*?\*\/', re.S | re.I), ''),
563            (KernRe(r'\/\*\s*private:.*', re.S | re.I), ''),
564
565            # Strip comments
566            (KernRe(r'\/\*.*?\*\/', re.S), ''),
567
568            # Strip attributes
569            (attribute, ' '),
570            (KernRe(r'\s*__aligned\s*\([^;]*\)', re.S), ' '),
571            (KernRe(r'\s*__counted_by\s*\([^;]*\)', re.S), ' '),
572            (KernRe(r'\s*__counted_by_(le|be)\s*\([^;]*\)', re.S), ' '),
573            (KernRe(r'\s*__packed\s*', re.S), ' '),
574            (KernRe(r'\s*CRYPTO_MINALIGN_ATTR', re.S), ' '),
575            (KernRe(r'\s*____cacheline_aligned_in_smp', re.S), ' '),
576            (KernRe(r'\s*____cacheline_aligned', re.S), ' '),
577
578            # Unwrap struct_group macros based on this definition:
579            # __struct_group(TAG, NAME, ATTRS, MEMBERS...)
580            # which has variants like: struct_group(NAME, MEMBERS...)
581            # Only MEMBERS arguments require documentation.
582            #
583            # Parsing them happens on two steps:
584            #
585            # 1. drop struct group arguments that aren't at MEMBERS,
586            #    storing them as STRUCT_GROUP(MEMBERS)
587            #
588            # 2. remove STRUCT_GROUP() ancillary macro.
589            #
590            # The original logic used to remove STRUCT_GROUP() using an
591            # advanced regex:
592            #
593            #   \bSTRUCT_GROUP(\(((?:(?>[^)(]+)|(?1))*)\))[^;]*;
594            #
595            # with two patterns that are incompatible with
596            # Python re module, as it has:
597            #
598            #   - a recursive pattern: (?1)
599            #   - an atomic grouping: (?>...)
600            #
601            # I tried a simpler version: but it didn't work either:
602            #   \bSTRUCT_GROUP\(([^\)]+)\)[^;]*;
603            #
604            # As it doesn't properly match the end parenthesis on some cases.
605            #
606            # So, a better solution was crafted: there's now a NestedMatch
607            # class that ensures that delimiters after a search are properly
608            # matched. So, the implementation to drop STRUCT_GROUP() will be
609            # handled in separate.
610
611            (KernRe(r'\bstruct_group\s*\(([^,]*,)', re.S), r'STRUCT_GROUP('),
612            (KernRe(r'\bstruct_group_attr\s*\(([^,]*,){2}', re.S), r'STRUCT_GROUP('),
613            (KernRe(r'\bstruct_group_tagged\s*\(([^,]*),([^,]*),', re.S), r'struct \1 \2; STRUCT_GROUP('),
614            (KernRe(r'\b__struct_group\s*\(([^,]*,){3}', re.S), r'STRUCT_GROUP('),
615
616            # Replace macros
617            #
618            # TODO: use NestedMatch for FOO($1, $2, ...) matches
619            #
620            # it is better to also move those to the NestedMatch logic,
621            # to ensure that parenthesis will be properly matched.
622
623            (KernRe(r'__ETHTOOL_DECLARE_LINK_MODE_MASK\s*\(([^\)]+)\)', re.S), r'DECLARE_BITMAP(\1, __ETHTOOL_LINK_MODE_MASK_NBITS)'),
624            (KernRe(r'DECLARE_PHY_INTERFACE_MASK\s*\(([^\)]+)\)', re.S), r'DECLARE_BITMAP(\1, PHY_INTERFACE_MODE_MAX)'),
625            (KernRe(r'DECLARE_BITMAP\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'unsigned long \1[BITS_TO_LONGS(\2)]'),
626            (KernRe(r'DECLARE_HASHTABLE\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'unsigned long \1[1 << ((\2) - 1)]'),
627            (KernRe(r'DECLARE_KFIFO\s*\(' + args_pattern + r',\s*' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\2 *\1'),
628            (KernRe(r'DECLARE_KFIFO_PTR\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\2 *\1'),
629            (KernRe(r'(?:__)?DECLARE_FLEX_ARRAY\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\1 \2[]'),
630            (KernRe(r'DEFINE_DMA_UNMAP_ADDR\s*\(' + args_pattern + r'\)', re.S), r'dma_addr_t \1'),
631            (KernRe(r'DEFINE_DMA_UNMAP_LEN\s*\(' + args_pattern + r'\)', re.S), r'__u32 \1'),
632        ]
633
634        # Regexes here are guaranteed to have the end limiter matching
635        # the start delimiter. Yet, right now, only one replace group
636        # is allowed.
637
638        sub_nested_prefixes = [
639            (re.compile(r'\bSTRUCT_GROUP\('), r'\1'),
640        ]
641
642        for search, sub in sub_prefixes:
643            members = search.sub(sub, members)
644
645        nested = NestedMatch()
646
647        for search, sub in sub_nested_prefixes:
648            members = nested.sub(search, sub, members)
649
650        # Keeps the original declaration as-is
651        declaration = members
652
653        # Split nested struct/union elements
654        #
655        # This loop was simpler at the original kernel-doc perl version, as
656        #   while ($members =~ m/$struct_members/) { ... }
657        # reads 'members' string on each interaction.
658        #
659        # Python behavior is different: it parses 'members' only once,
660        # creating a list of tuples from the first interaction.
661        #
662        # On other words, this won't get nested structs.
663        #
664        # So, we need to have an extra loop on Python to override such
665        # re limitation.
666
667        while True:
668            tuples = struct_members.findall(members)
669            if not tuples:
670                break
671
672            for t in tuples:
673                newmember = ""
674                maintype = t[0]
675                s_ids = t[5]
676                content = t[3]
677
678                oldmember = "".join(t)
679
680                for s_id in s_ids.split(','):
681                    s_id = s_id.strip()
682
683                    newmember += f"{maintype} {s_id}; "
684                    s_id = KernRe(r'[:\[].*').sub('', s_id)
685                    s_id = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', s_id)
686
687                    for arg in content.split(';'):
688                        arg = arg.strip()
689
690                        if not arg:
691                            continue
692
693                        r = KernRe(r'^([^\(]+\(\*?\s*)([\w\.]*)(\s*\).*)')
694                        if r.match(arg):
695                            # Pointer-to-function
696                            dtype = r.group(1)
697                            name = r.group(2)
698                            extra = r.group(3)
699
700                            if not name:
701                                continue
702
703                            if not s_id:
704                                # Anonymous struct/union
705                                newmember += f"{dtype}{name}{extra}; "
706                            else:
707                                newmember += f"{dtype}{s_id}.{name}{extra}; "
708
709                        else:
710                            arg = arg.strip()
711                            # Handle bitmaps
712                            arg = KernRe(r':\s*\d+\s*').sub('', arg)
713
714                            # Handle arrays
715                            arg = KernRe(r'\[.*\]').sub('', arg)
716
717                            # Handle multiple IDs
718                            arg = KernRe(r'\s*,\s*').sub(',', arg)
719
720                            r = KernRe(r'(.*)\s+([\S+,]+)')
721
722                            if r.search(arg):
723                                dtype = r.group(1)
724                                names = r.group(2)
725                            else:
726                                newmember += f"{arg}; "
727                                continue
728
729                            for name in names.split(','):
730                                name = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', name).strip()
731
732                                if not name:
733                                    continue
734
735                                if not s_id:
736                                    # Anonymous struct/union
737                                    newmember += f"{dtype} {name}; "
738                                else:
739                                    newmember += f"{dtype} {s_id}.{name}; "
740
741                members = members.replace(oldmember, newmember)
742
743        # Ignore other nested elements, like enums
744        members = re.sub(r'(\{[^\{\}]*\})', '', members)
745
746        self.create_parameter_list(ln, decl_type, members, ';',
747                                   declaration_name)
748        self.check_sections(ln, declaration_name, decl_type)
749
750        # Adjust declaration for better display
751        declaration = KernRe(r'([\{;])').sub(r'\1\n', declaration)
752        declaration = KernRe(r'\}\s+;').sub('};', declaration)
753
754        # Better handle inlined enums
755        while True:
756            r = KernRe(r'(enum\s+\{[^\}]+),([^\n])')
757            if not r.search(declaration):
758                break
759
760            declaration = r.sub(r'\1,\n\2', declaration)
761
762        def_args = declaration.split('\n')
763        level = 1
764        declaration = ""
765        for clause in def_args:
766
767            clause = clause.strip()
768            clause = KernRe(r'\s+').sub(' ', clause, count=1)
769
770            if not clause:
771                continue
772
773            if '}' in clause and level > 1:
774                level -= 1
775
776            if not KernRe(r'^\s*#').match(clause):
777                declaration += "\t" * level
778
779            declaration += "\t" + clause + "\n"
780            if "{" in clause and "}" not in clause:
781                level += 1
782
783        self.output_declaration(decl_type, declaration_name,
784                                definition=declaration,
785                                purpose=self.entry.declaration_purpose)
786
787    def dump_enum(self, ln, proto):
788        """
789        Stores an enum inside self.entries array.
790        """
791
792        # Ignore members marked private
793        proto = KernRe(r'\/\*\s*private:.*?\/\*\s*public:.*?\*\/', flags=re.S).sub('', proto)
794        proto = KernRe(r'\/\*\s*private:.*}', flags=re.S).sub('}', proto)
795
796        # Strip comments
797        proto = KernRe(r'\/\*.*?\*\/', flags=re.S).sub('', proto)
798
799        # Strip #define macros inside enums
800        proto = KernRe(r'#\s*((define|ifdef|if)\s+|endif)[^;]*;', flags=re.S).sub('', proto)
801
802        #
803        # Parse out the name and members of the enum.  Typedef form first.
804        #
805        r = KernRe(r'typedef\s+enum\s*\{(.*)\}\s*(\w*)\s*;')
806        if r.search(proto):
807            declaration_name = r.group(2)
808            members = r.group(1).rstrip()
809        #
810        # Failing that, look for a straight enum
811        #
812        else:
813            r = KernRe(r'enum\s+(\w*)\s*\{(.*)\}')
814            if r.match(proto):
815                declaration_name = r.group(1)
816                members = r.group(2).rstrip()
817        #
818        # OK, this isn't going to work.
819        #
820            else:
821                self.emit_msg(ln, f"{proto}: error: Cannot parse enum!")
822                return
823        #
824        # Make sure we found what we were expecting.
825        #
826        if self.entry.identifier != declaration_name:
827            if self.entry.identifier == "":
828                self.emit_msg(ln,
829                              f"{proto}: wrong kernel-doc identifier on prototype")
830            else:
831                self.emit_msg(ln,
832                              f"expecting prototype for enum {self.entry.identifier}. "
833                              f"Prototype was for enum {declaration_name} instead")
834            return
835
836        if not declaration_name:
837            declaration_name = "(anonymous)"
838        #
839        # Parse out the name of each enum member, and verify that we
840        # have a description for it.
841        #
842        member_set = set()
843        members = KernRe(r'\([^;)]*\)').sub('', members)
844        for arg in members.split(','):
845            if not arg:
846                continue
847            arg = KernRe(r'^\s*(\w+).*').sub(r'\1', arg)
848            self.entry.parameterlist.append(arg)
849            if arg not in self.entry.parameterdescs:
850                self.entry.parameterdescs[arg] = self.undescribed
851                self.emit_msg(ln,
852                              f"Enum value '{arg}' not described in enum '{declaration_name}'")
853            member_set.add(arg)
854        #
855        # Ensure that every described member actually exists in the enum.
856        #
857        for k in self.entry.parameterdescs:
858            if k not in member_set:
859                self.emit_msg(ln,
860                              f"Excess enum value '%{k}' description in '{declaration_name}'")
861
862        self.output_declaration('enum', declaration_name,
863                                purpose=self.entry.declaration_purpose)
864
865    def dump_declaration(self, ln, prototype):
866        """
867        Stores a data declaration inside self.entries array.
868        """
869
870        if self.entry.decl_type == "enum":
871            self.dump_enum(ln, prototype)
872        elif self.entry.decl_type == "typedef":
873            self.dump_typedef(ln, prototype)
874        elif self.entry.decl_type in ["union", "struct"]:
875            self.dump_struct(ln, prototype)
876        else:
877            # This would be a bug
878            self.emit_message(ln, f'Unknown declaration type: {self.entry.decl_type}')
879
880    def dump_function(self, ln, prototype):
881        """
882        Stores a function of function macro inside self.entries array.
883        """
884
885        func_macro = False
886        return_type = ''
887        decl_type = 'function'
888
889        # Prefixes that would be removed
890        sub_prefixes = [
891            (r"^static +", "", 0),
892            (r"^extern +", "", 0),
893            (r"^asmlinkage +", "", 0),
894            (r"^inline +", "", 0),
895            (r"^__inline__ +", "", 0),
896            (r"^__inline +", "", 0),
897            (r"^__always_inline +", "", 0),
898            (r"^noinline +", "", 0),
899            (r"^__FORTIFY_INLINE +", "", 0),
900            (r"__init +", "", 0),
901            (r"__init_or_module +", "", 0),
902            (r"__deprecated +", "", 0),
903            (r"__flatten +", "", 0),
904            (r"__meminit +", "", 0),
905            (r"__must_check +", "", 0),
906            (r"__weak +", "", 0),
907            (r"__sched +", "", 0),
908            (r"_noprof", "", 0),
909            (r"__printf\s*\(\s*\d*\s*,\s*\d*\s*\) +", "", 0),
910            (r"__(?:re)?alloc_size\s*\(\s*\d+\s*(?:,\s*\d+\s*)?\) +", "", 0),
911            (r"__diagnose_as\s*\(\s*\S+\s*(?:,\s*\d+\s*)*\) +", "", 0),
912            (r"DECL_BUCKET_PARAMS\s*\(\s*(\S+)\s*,\s*(\S+)\s*\)", r"\1, \2", 0),
913            (r"__attribute_const__ +", "", 0),
914
915            # It seems that Python support for re.X is broken:
916            # At least for me (Python 3.13), this didn't work
917#            (r"""
918#              __attribute__\s*\(\(
919#                (?:
920#                    [\w\s]+          # attribute name
921#                    (?:\([^)]*\))?   # attribute arguments
922#                    \s*,?            # optional comma at the end
923#                )+
924#              \)\)\s+
925#             """, "", re.X),
926
927            # So, remove whitespaces and comments from it
928            (r"__attribute__\s*\(\((?:[\w\s]+(?:\([^)]*\))?\s*,?)+\)\)\s+", "", 0),
929        ]
930
931        for search, sub, flags in sub_prefixes:
932            prototype = KernRe(search, flags).sub(sub, prototype)
933
934        # Macros are a special case, as they change the prototype format
935        new_proto = KernRe(r"^#\s*define\s+").sub("", prototype)
936        if new_proto != prototype:
937            is_define_proto = True
938            prototype = new_proto
939        else:
940            is_define_proto = False
941
942        # Yes, this truly is vile.  We are looking for:
943        # 1. Return type (may be nothing if we're looking at a macro)
944        # 2. Function name
945        # 3. Function parameters.
946        #
947        # All the while we have to watch out for function pointer parameters
948        # (which IIRC is what the two sections are for), C types (these
949        # regexps don't even start to express all the possibilities), and
950        # so on.
951        #
952        # If you mess with these regexps, it's a good idea to check that
953        # the following functions' documentation still comes out right:
954        # - parport_register_device (function pointer parameters)
955        # - atomic_set (macro)
956        # - pci_match_device, __copy_to_user (long return type)
957
958        name = r'[a-zA-Z0-9_~:]+'
959        prototype_end1 = r'[^\(]*'
960        prototype_end2 = r'[^\{]*'
961        prototype_end = fr'\(({prototype_end1}|{prototype_end2})\)'
962
963        # Besides compiling, Perl qr{[\w\s]+} works as a non-capturing group.
964        # So, this needs to be mapped in Python with (?:...)? or (?:...)+
965
966        type1 = r'(?:[\w\s]+)?'
967        type2 = r'(?:[\w\s]+\*+)+'
968
969        found = False
970
971        if is_define_proto:
972            r = KernRe(r'^()(' + name + r')\s+')
973
974            if r.search(prototype):
975                return_type = ''
976                declaration_name = r.group(2)
977                func_macro = True
978
979                found = True
980
981        if not found:
982            patterns = [
983                rf'^()({name})\s*{prototype_end}',
984                rf'^({type1})\s+({name})\s*{prototype_end}',
985                rf'^({type2})\s*({name})\s*{prototype_end}',
986            ]
987
988            for p in patterns:
989                r = KernRe(p)
990
991                if r.match(prototype):
992
993                    return_type = r.group(1)
994                    declaration_name = r.group(2)
995                    args = r.group(3)
996
997                    self.create_parameter_list(ln, decl_type, args, ',',
998                                               declaration_name)
999
1000                    found = True
1001                    break
1002        if not found:
1003            self.emit_msg(ln,
1004                          f"cannot understand function prototype: '{prototype}'")
1005            return
1006
1007        if self.entry.identifier != declaration_name:
1008            self.emit_msg(ln,
1009                          f"expecting prototype for {self.entry.identifier}(). Prototype was for {declaration_name}() instead")
1010            return
1011
1012        self.check_sections(ln, declaration_name, "function")
1013
1014        self.check_return_section(ln, declaration_name, return_type)
1015
1016        if 'typedef' in return_type:
1017            self.output_declaration(decl_type, declaration_name,
1018                                    typedef=True,
1019                                    functiontype=return_type,
1020                                    purpose=self.entry.declaration_purpose,
1021                                    func_macro=func_macro)
1022        else:
1023            self.output_declaration(decl_type, declaration_name,
1024                                    typedef=False,
1025                                    functiontype=return_type,
1026                                    purpose=self.entry.declaration_purpose,
1027                                    func_macro=func_macro)
1028
1029    def dump_typedef(self, ln, proto):
1030        """
1031        Stores a typedef inside self.entries array.
1032        """
1033
1034        typedef_type = r'((?:\s+[\w\*]+\b){0,7}\s+(?:\w+\b|\*+))\s*'
1035        typedef_ident = r'\*?\s*(\w\S+)\s*'
1036        typedef_args = r'\s*\((.*)\);'
1037
1038        typedef1 = KernRe(r'typedef' + typedef_type + r'\(' + typedef_ident + r'\)' + typedef_args)
1039        typedef2 = KernRe(r'typedef' + typedef_type + typedef_ident + typedef_args)
1040
1041        # Strip comments
1042        proto = KernRe(r'/\*.*?\*/', flags=re.S).sub('', proto)
1043
1044        # Parse function typedef prototypes
1045        for r in [typedef1, typedef2]:
1046            if not r.match(proto):
1047                continue
1048
1049            return_type = r.group(1).strip()
1050            declaration_name = r.group(2)
1051            args = r.group(3)
1052
1053            if self.entry.identifier != declaration_name:
1054                self.emit_msg(ln,
1055                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1056                return
1057
1058            decl_type = 'function'
1059            self.create_parameter_list(ln, decl_type, args, ',', declaration_name)
1060
1061            self.output_declaration(decl_type, declaration_name,
1062                                    typedef=True,
1063                                    functiontype=return_type,
1064                                    purpose=self.entry.declaration_purpose)
1065            return
1066
1067        # Handle nested parentheses or brackets
1068        r = KernRe(r'(\(*.\)\s*|\[*.\]\s*);$')
1069        while r.search(proto):
1070            proto = r.sub('', proto)
1071
1072        # Parse simple typedefs
1073        r = KernRe(r'typedef.*\s+(\w+)\s*;')
1074        if r.match(proto):
1075            declaration_name = r.group(1)
1076
1077            if self.entry.identifier != declaration_name:
1078                self.emit_msg(ln,
1079                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1080                return
1081
1082            self.output_declaration('typedef', declaration_name,
1083                                    purpose=self.entry.declaration_purpose)
1084            return
1085
1086        self.emit_msg(ln, "error: Cannot parse typedef!")
1087
1088    @staticmethod
1089    def process_export(function_set, line):
1090        """
1091        process EXPORT_SYMBOL* tags
1092
1093        This method doesn't use any variable from the class, so declare it
1094        with a staticmethod decorator.
1095        """
1096
1097        # We support documenting some exported symbols with different
1098        # names.  A horrible hack.
1099        suffixes = [ '_noprof' ]
1100
1101        # Note: it accepts only one EXPORT_SYMBOL* per line, as having
1102        # multiple export lines would violate Kernel coding style.
1103
1104        if export_symbol.search(line):
1105            symbol = export_symbol.group(2)
1106        elif export_symbol_ns.search(line):
1107            symbol = export_symbol_ns.group(2)
1108        else:
1109            return False
1110        #
1111        # Found an export, trim out any special suffixes
1112        #
1113        for suffix in suffixes:
1114            symbol = symbol.removesuffix(suffix)
1115        function_set.add(symbol)
1116        return True
1117
1118    def process_normal(self, ln, line):
1119        """
1120        STATE_NORMAL: looking for the /** to begin everything.
1121        """
1122
1123        if not doc_start.match(line):
1124            return
1125
1126        # start a new entry
1127        self.reset_state(ln)
1128
1129        # next line is always the function name
1130        self.state = state.NAME
1131
1132    def process_name(self, ln, line):
1133        """
1134        STATE_NAME: Looking for the "name - description" line
1135        """
1136        #
1137        # Check for a DOC: block and handle them specially.
1138        #
1139        if doc_block.search(line):
1140
1141            if not doc_block.group(1):
1142                self.entry.begin_section(ln, "Introduction")
1143            else:
1144                self.entry.begin_section(ln, doc_block.group(1))
1145
1146            self.entry.identifier = self.entry.section
1147            self.state = state.DOCBLOCK
1148        #
1149        # Otherwise we're looking for a normal kerneldoc declaration line.
1150        #
1151        elif doc_decl.search(line):
1152            self.entry.identifier = doc_decl.group(1)
1153
1154            # Test for data declaration
1155            if doc_begin_data.search(line):
1156                self.entry.decl_type = doc_begin_data.group(1)
1157                self.entry.identifier = doc_begin_data.group(2)
1158            #
1159            # Look for a function description
1160            #
1161            elif doc_begin_func.search(line):
1162                self.entry.identifier = doc_begin_func.group(1)
1163                self.entry.decl_type = "function"
1164            #
1165            # We struck out.
1166            #
1167            else:
1168                self.emit_msg(ln,
1169                              f"This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst\n{line}")
1170                self.state = state.NORMAL
1171                return
1172            #
1173            # OK, set up for a new kerneldoc entry.
1174            #
1175            self.state = state.BODY
1176            self.entry.identifier = self.entry.identifier.strip(" ")
1177            # if there's no @param blocks need to set up default section here
1178            self.entry.begin_section(ln + 1)
1179            #
1180            # Find the description portion, which *should* be there but
1181            # isn't always.
1182            # (We should be able to capture this from the previous parsing - someday)
1183            #
1184            r = KernRe("[-:](.*)")
1185            if r.search(line):
1186                self.entry.declaration_purpose = trim_whitespace(r.group(1))
1187                self.state = state.DECLARATION
1188            else:
1189                self.entry.declaration_purpose = ""
1190
1191            if not self.entry.declaration_purpose and self.config.wshort_desc:
1192                self.emit_msg(ln,
1193                              f"missing initial short description on line:\n{line}")
1194
1195            if not self.entry.identifier and self.entry.decl_type != "enum":
1196                self.emit_msg(ln,
1197                              f"wrong kernel-doc identifier on line:\n{line}")
1198                self.state = state.NORMAL
1199
1200            if self.config.verbose:
1201                self.emit_msg(ln,
1202                              f"Scanning doc for {self.entry.decl_type} {self.entry.identifier}",
1203                                  warning=False)
1204        #
1205        # Failed to find an identifier. Emit a warning
1206        #
1207        else:
1208            self.emit_msg(ln, f"Cannot find identifier on line:\n{line}")
1209
1210    #
1211    # Helper function to determine if a new section is being started.
1212    #
1213    def is_new_section(self, ln, line):
1214        if doc_sect.search(line):
1215            self.state = state.BODY
1216            #
1217            # Pick out the name of our new section, tweaking it if need be.
1218            #
1219            newsection = doc_sect.group(1)
1220            if newsection.lower() == 'description':
1221                newsection = 'Description'
1222            elif newsection.lower() == 'context':
1223                newsection = 'Context'
1224                self.state = state.SPECIAL_SECTION
1225            elif newsection.lower() in ["@return", "@returns",
1226                                        "return", "returns"]:
1227                newsection = "Return"
1228                self.state = state.SPECIAL_SECTION
1229            elif newsection[0] == '@':
1230                self.state = state.SPECIAL_SECTION
1231            #
1232            # Initialize the contents, and get the new section going.
1233            #
1234            newcontents = doc_sect.group(2)
1235            if not newcontents:
1236                newcontents = ""
1237            self.dump_section()
1238            self.entry.begin_section(ln, newsection)
1239            self.entry.leading_space = None
1240
1241            self.entry.add_text(newcontents.lstrip())
1242            return True
1243        return False
1244
1245    #
1246    # Helper function to detect (and effect) the end of a kerneldoc comment.
1247    #
1248    def is_comment_end(self, ln, line):
1249        if doc_end.search(line):
1250            self.dump_section()
1251
1252            # Look for doc_com + <text> + doc_end:
1253            r = KernRe(r'\s*\*\s*[a-zA-Z_0-9:\.]+\*/')
1254            if r.match(line):
1255                self.emit_msg(ln, f"suspicious ending line: {line}")
1256
1257            self.entry.prototype = ""
1258            self.entry.new_start_line = ln + 1
1259
1260            self.state = state.PROTO
1261            return True
1262        return False
1263
1264
1265    def process_decl(self, ln, line):
1266        """
1267        STATE_DECLARATION: We've seen the beginning of a declaration
1268        """
1269        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1270            return
1271        #
1272        # Look for anything with the " * " line beginning.
1273        #
1274        if doc_content.search(line):
1275            cont = doc_content.group(1)
1276            #
1277            # A blank line means that we have moved out of the declaration
1278            # part of the comment (without any "special section" parameter
1279            # descriptions).
1280            #
1281            if cont == "":
1282                self.state = state.BODY
1283            #
1284            # Otherwise we have more of the declaration section to soak up.
1285            #
1286            else:
1287                self.entry.declaration_purpose = \
1288                    trim_whitespace(self.entry.declaration_purpose + ' ' + cont)
1289        else:
1290            # Unknown line, ignore
1291            self.emit_msg(ln, f"bad line: {line}")
1292
1293
1294    def process_special(self, ln, line):
1295        """
1296        STATE_SPECIAL_SECTION: a section ending with a blank line
1297        """
1298        #
1299        # If we have hit a blank line (only the " * " marker), then this
1300        # section is done.
1301        #
1302        if KernRe(r"\s*\*\s*$").match(line):
1303            self.entry.begin_section(ln, dump = True)
1304            self.state = state.BODY
1305            return
1306        #
1307        # Not a blank line, look for the other ways to end the section.
1308        #
1309        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1310            return
1311        #
1312        # OK, we should have a continuation of the text for this section.
1313        #
1314        if doc_content.search(line):
1315            cont = doc_content.group(1)
1316            #
1317            # If the lines of text after the first in a special section have
1318            # leading white space, we need to trim it out or Sphinx will get
1319            # confused.  For the second line (the None case), see what we
1320            # find there and remember it.
1321            #
1322            if self.entry.leading_space is None:
1323                r = KernRe(r'^(\s+)')
1324                if r.match(cont):
1325                    self.entry.leading_space = len(r.group(1))
1326                else:
1327                    self.entry.leading_space = 0
1328            #
1329            # Otherwise, before trimming any leading chars, be *sure*
1330            # that they are white space.  We should maybe warn if this
1331            # isn't the case.
1332            #
1333            for i in range(0, self.entry.leading_space):
1334                if cont[i] != " ":
1335                    self.entry.leading_space = i
1336                    break
1337            #
1338            # Add the trimmed result to the section and we're done.
1339            #
1340            self.entry.add_text(cont[self.entry.leading_space:])
1341        else:
1342            # Unknown line, ignore
1343            self.emit_msg(ln, f"bad line: {line}")
1344
1345    def process_body(self, ln, line):
1346        """
1347        STATE_BODY: the bulk of a kerneldoc comment.
1348        """
1349        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1350            return
1351
1352        if doc_content.search(line):
1353            cont = doc_content.group(1)
1354            self.entry.add_text(cont)
1355        else:
1356            # Unknown line, ignore
1357            self.emit_msg(ln, f"bad line: {line}")
1358
1359    def process_inline_name(self, ln, line):
1360        """STATE_INLINE_NAME: beginning of docbook comments within a prototype."""
1361
1362        if doc_inline_sect.search(line):
1363            self.entry.begin_section(ln, doc_inline_sect.group(1))
1364            self.entry.add_text(doc_inline_sect.group(2).lstrip())
1365            self.state = state.INLINE_TEXT
1366        elif doc_inline_end.search(line):
1367            self.dump_section()
1368            self.state = state.PROTO
1369        elif doc_content.search(line):
1370            self.emit_msg(ln, f"Incorrect use of kernel-doc format: {line}")
1371            self.state = state.PROTO
1372        # else ... ??
1373
1374    def process_inline_text(self, ln, line):
1375        """STATE_INLINE_TEXT: docbook comments within a prototype."""
1376
1377        if doc_inline_end.search(line):
1378            self.dump_section()
1379            self.state = state.PROTO
1380        elif doc_content.search(line):
1381            self.entry.add_text(doc_content.group(1))
1382        # else ... ??
1383
1384    def syscall_munge(self, ln, proto):         # pylint: disable=W0613
1385        """
1386        Handle syscall definitions
1387        """
1388
1389        is_void = False
1390
1391        # Strip newlines/CR's
1392        proto = re.sub(r'[\r\n]+', ' ', proto)
1393
1394        # Check if it's a SYSCALL_DEFINE0
1395        if 'SYSCALL_DEFINE0' in proto:
1396            is_void = True
1397
1398        # Replace SYSCALL_DEFINE with correct return type & function name
1399        proto = KernRe(r'SYSCALL_DEFINE.*\(').sub('long sys_', proto)
1400
1401        r = KernRe(r'long\s+(sys_.*?),')
1402        if r.search(proto):
1403            proto = KernRe(',').sub('(', proto, count=1)
1404        elif is_void:
1405            proto = KernRe(r'\)').sub('(void)', proto, count=1)
1406
1407        # Now delete all of the odd-numbered commas in the proto
1408        # so that argument types & names don't have a comma between them
1409        count = 0
1410        length = len(proto)
1411
1412        if is_void:
1413            length = 0  # skip the loop if is_void
1414
1415        for ix in range(length):
1416            if proto[ix] == ',':
1417                count += 1
1418                if count % 2 == 1:
1419                    proto = proto[:ix] + ' ' + proto[ix + 1:]
1420
1421        return proto
1422
1423    def tracepoint_munge(self, ln, proto):
1424        """
1425        Handle tracepoint definitions
1426        """
1427
1428        tracepointname = None
1429        tracepointargs = None
1430
1431        # Match tracepoint name based on different patterns
1432        r = KernRe(r'TRACE_EVENT\((.*?),')
1433        if r.search(proto):
1434            tracepointname = r.group(1)
1435
1436        r = KernRe(r'DEFINE_SINGLE_EVENT\((.*?),')
1437        if r.search(proto):
1438            tracepointname = r.group(1)
1439
1440        r = KernRe(r'DEFINE_EVENT\((.*?),(.*?),')
1441        if r.search(proto):
1442            tracepointname = r.group(2)
1443
1444        if tracepointname:
1445            tracepointname = tracepointname.lstrip()
1446
1447        r = KernRe(r'TP_PROTO\((.*?)\)')
1448        if r.search(proto):
1449            tracepointargs = r.group(1)
1450
1451        if not tracepointname or not tracepointargs:
1452            self.emit_msg(ln,
1453                          f"Unrecognized tracepoint format:\n{proto}\n")
1454        else:
1455            proto = f"static inline void trace_{tracepointname}({tracepointargs})"
1456            self.entry.identifier = f"trace_{self.entry.identifier}"
1457
1458        return proto
1459
1460    def process_proto_function(self, ln, line):
1461        """Ancillary routine to process a function prototype"""
1462
1463        # strip C99-style comments to end of line
1464        line = KernRe(r"\/\/.*$", re.S).sub('', line)
1465        #
1466        # Soak up the line's worth of prototype text, stopping at { or ; if present.
1467        #
1468        if KernRe(r'\s*#\s*define').match(line):
1469            self.entry.prototype = line
1470        elif not line.startswith('#'):   # skip other preprocessor stuff
1471            r = KernRe(r'([^\{]*)')
1472            if r.match(line):
1473                self.entry.prototype += r.group(1) + " "
1474        #
1475        # If we now have the whole prototype, clean it up and declare victory.
1476        #
1477        if '{' in line or ';' in line or KernRe(r'\s*#\s*define').match(line):
1478            # strip comments and surrounding spaces
1479            self.entry.prototype = KernRe(r'/\*.*\*/').sub('', self.entry.prototype).strip()
1480            #
1481            # Handle self.entry.prototypes for function pointers like:
1482            #       int (*pcs_config)(struct foo)
1483            # by turning it into
1484            #	    int pcs_config(struct foo)
1485            #
1486            r = KernRe(r'^(\S+\s+)\(\s*\*(\S+)\)')
1487            self.entry.prototype = r.sub(r'\1\2', self.entry.prototype)
1488            #
1489            # Handle special declaration syntaxes
1490            #
1491            if 'SYSCALL_DEFINE' in self.entry.prototype:
1492                self.entry.prototype = self.syscall_munge(ln,
1493                                                          self.entry.prototype)
1494            else:
1495                r = KernRe(r'TRACE_EVENT|DEFINE_EVENT|DEFINE_SINGLE_EVENT')
1496                if r.search(self.entry.prototype):
1497                    self.entry.prototype = self.tracepoint_munge(ln,
1498                                                                 self.entry.prototype)
1499            #
1500            # ... and we're done
1501            #
1502            self.dump_function(ln, self.entry.prototype)
1503            self.reset_state(ln)
1504
1505    def process_proto_type(self, ln, line):
1506        """Ancillary routine to process a type"""
1507
1508        # Strip C99-style comments and surrounding whitespace
1509        line = KernRe(r"//.*$", re.S).sub('', line).strip()
1510        if not line:
1511            return # nothing to see here
1512
1513        # To distinguish preprocessor directive from regular declaration later.
1514        if line.startswith('#'):
1515            line += ";"
1516        #
1517        # Split the declaration on any of { } or ;, and accumulate pieces
1518        # until we hit a semicolon while not inside {brackets}
1519        #
1520        r = KernRe(r'(.*?)([{};])')
1521        for chunk in r.split(line):
1522            if chunk:  # Ignore empty matches
1523                self.entry.prototype += chunk
1524                #
1525                # This cries out for a match statement ... someday after we can
1526                # drop Python 3.9 ...
1527                #
1528                if chunk == '{':
1529                    self.entry.brcount += 1
1530                elif chunk == '}':
1531                    self.entry.brcount -= 1
1532                elif chunk == ';' and self.entry.brcount <= 0:
1533                    self.dump_declaration(ln, self.entry.prototype)
1534                    self.reset_state(ln)
1535                    return
1536        #
1537        # We hit the end of the line while still in the declaration; put
1538        # in a space to represent the newline.
1539        #
1540        self.entry.prototype += ' '
1541
1542    def process_proto(self, ln, line):
1543        """STATE_PROTO: reading a function/whatever prototype."""
1544
1545        if doc_inline_oneline.search(line):
1546            self.entry.begin_section(ln, doc_inline_oneline.group(1))
1547            self.entry.add_text(doc_inline_oneline.group(2))
1548            self.dump_section()
1549
1550        elif doc_inline_start.search(line):
1551            self.state = state.INLINE_NAME
1552
1553        elif self.entry.decl_type == 'function':
1554            self.process_proto_function(ln, line)
1555
1556        else:
1557            self.process_proto_type(ln, line)
1558
1559    def process_docblock(self, ln, line):
1560        """STATE_DOCBLOCK: within a DOC: block."""
1561
1562        if doc_end.search(line):
1563            self.dump_section()
1564            self.output_declaration("doc", self.entry.identifier)
1565            self.reset_state(ln)
1566
1567        elif doc_content.search(line):
1568            self.entry.add_text(doc_content.group(1))
1569
1570    def parse_export(self):
1571        """
1572        Parses EXPORT_SYMBOL* macros from a single Kernel source file.
1573        """
1574
1575        export_table = set()
1576
1577        try:
1578            with open(self.fname, "r", encoding="utf8",
1579                      errors="backslashreplace") as fp:
1580
1581                for line in fp:
1582                    self.process_export(export_table, line)
1583
1584        except IOError:
1585            return None
1586
1587        return export_table
1588
1589    #
1590    # The state/action table telling us which function to invoke in
1591    # each state.
1592    #
1593    state_actions = {
1594        state.NORMAL:			process_normal,
1595        state.NAME:			process_name,
1596        state.BODY:			process_body,
1597        state.DECLARATION:		process_decl,
1598        state.SPECIAL_SECTION:		process_special,
1599        state.INLINE_NAME:		process_inline_name,
1600        state.INLINE_TEXT:		process_inline_text,
1601        state.PROTO:			process_proto,
1602        state.DOCBLOCK:			process_docblock,
1603        }
1604
1605    def parse_kdoc(self):
1606        """
1607        Open and process each line of a C source file.
1608        The parsing is controlled via a state machine, and the line is passed
1609        to a different process function depending on the state. The process
1610        function may update the state as needed.
1611
1612        Besides parsing kernel-doc tags, it also parses export symbols.
1613        """
1614
1615        prev = ""
1616        prev_ln = None
1617        export_table = set()
1618
1619        try:
1620            with open(self.fname, "r", encoding="utf8",
1621                      errors="backslashreplace") as fp:
1622                for ln, line in enumerate(fp):
1623
1624                    line = line.expandtabs().strip("\n")
1625
1626                    # Group continuation lines on prototypes
1627                    if self.state == state.PROTO:
1628                        if line.endswith("\\"):
1629                            prev += line.rstrip("\\")
1630                            if not prev_ln:
1631                                prev_ln = ln
1632                            continue
1633
1634                        if prev:
1635                            ln = prev_ln
1636                            line = prev + line
1637                            prev = ""
1638                            prev_ln = None
1639
1640                    self.config.log.debug("%d %s: %s",
1641                                          ln, state.name[self.state],
1642                                          line)
1643
1644                    # This is an optimization over the original script.
1645                    # There, when export_file was used for the same file,
1646                    # it was read twice. Here, we use the already-existing
1647                    # loop to parse exported symbols as well.
1648                    #
1649                    if (self.state != state.NORMAL) or \
1650                       not self.process_export(export_table, line):
1651                        # Hand this line to the appropriate state handler
1652                        self.state_actions[self.state](self, ln, line)
1653
1654        except OSError:
1655            self.config.log.error(f"Error: Cannot open file {self.fname}")
1656
1657        return export_table, self.entries
1658