xref: /linux/scripts/lib/kdoc/kdoc_parser.py (revision 60016e0116b8d33f95e797b011799e717766ec13)
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*$', cache=False)
32
33doc_end = KernRe(r'\*/', cache=False)
34doc_com = KernRe(r'\s*\*\s*', cache=False)
35doc_com_body = KernRe(r'\s*\* ?', cache=False)
36doc_decl = doc_com + KernRe(r'(\w+)', cache=False)
37
38# @params and a strictly limited set of supported section names
39# Specifically:
40#   Match @word:
41#         @...:
42#         @{section-name}:
43# while trying to not match literal block starts like "example::"
44#
45doc_sect = doc_com + \
46            KernRe(r'\s*(\@[.\w]+|\@\.\.\.|description|context|returns?|notes?|examples?)\s*:([^:].*)?$',
47                flags=re.I, cache=False)
48
49doc_content = doc_com_body + KernRe(r'(.*)', cache=False)
50doc_inline_start = KernRe(r'^\s*/\*\*\s*$', cache=False)
51doc_inline_sect = KernRe(r'\s*\*\s*(@\s*[\w][\w\.]*\s*):(.*)', cache=False)
52doc_inline_end = KernRe(r'^\s*\*/\s*$', cache=False)
53doc_inline_oneline = KernRe(r'^\s*/\*\*\s*(@[\w\s]+):\s*(.*)\s*\*/\s*$', cache=False)
54attribute = KernRe(r"__attribute__\s*\(\([a-z0-9,_\*\s\(\)]*\)\)",
55               flags=re.I | re.S, cache=False)
56
57export_symbol = KernRe(r'^\s*EXPORT_SYMBOL(_GPL)?\s*\(\s*(\w+)\s*\)\s*', cache=False)
58export_symbol_ns = KernRe(r'^\s*EXPORT_SYMBOL_NS(_GPL)?\s*\(\s*(\w+)\s*,\s*"\S+"\)\s*', cache=False)
59
60type_param = KernRe(r"\@(\w*((\.\w+)|(->\w+))*(\.\.\.)?)", cache=False)
61
62#
63# Tests for the beginning of a kerneldoc block in its various forms.
64#
65doc_block = doc_com + KernRe(r'DOC:\s*(.*)?', cache=False)
66doc_begin_data = KernRe(r"^\s*\*?\s*(struct|union|enum|typedef)\b\s*(\w*)", cache = False)
67doc_begin_func = KernRe(str(doc_com) +			# initial " * '
68                        r"(?:\w+\s*\*\s*)?" + 		# type (not captured)
69                        r'(?:define\s+)?' + 		# possible "define" (not captured)
70                        r'(\w+)\s*(?:\(\w*\))?\s*' +	# name and optional "(...)"
71                        r'(?:[-:].*)?$',		# description (not captured)
72                        cache = False)
73
74#
75# A little helper to get rid of excess white space
76#
77multi_space = KernRe(r'\s\s+')
78def trim_whitespace(s):
79    return multi_space.sub(' ', s.strip())
80
81class state:
82    """
83    State machine enums
84    """
85
86    # Parser states
87    NORMAL        = 0        # normal code
88    NAME          = 1        # looking for function name
89    DECLARATION   = 2        # We have seen a declaration which might not be done
90    BODY          = 3        # the body of the comment
91    SPECIAL_SECTION = 4      # doc section ending with a blank line
92    PROTO         = 5        # scanning prototype
93    DOCBLOCK      = 6        # documentation block
94    INLINE_NAME   = 7        # gathering doc outside main block
95    INLINE_TEXT   = 8	     # reading the body of inline docs
96
97    name = [
98        "NORMAL",
99        "NAME",
100        "DECLARATION",
101        "BODY",
102        "SPECIAL_SECTION",
103        "PROTO",
104        "DOCBLOCK",
105        "INLINE_NAME",
106        "INLINE_TEXT",
107    ]
108
109
110SECTION_DEFAULT = "Description"  # default section
111
112class KernelEntry:
113
114    def __init__(self, config, ln):
115        self.config = config
116
117        self._contents = []
118        self.sectcheck = ""
119        self.struct_actual = ""
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.sectionlist = []
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.sectcheck += name + " "
193            self.new_start_line = 0
194
195        else:
196            if name in self.sections and self.sections[name] != "":
197                # Only warn on user-specified duplicate section names
198                if name != SECTION_DEFAULT:
199                    self.emit_msg(self.new_start_line,
200                                  f"duplicate section name '{name}'\n")
201                # Treat as a new paragraph - add a blank line
202                self.sections[name] += '\n' + contents
203            else:
204                self.sections[name] = contents
205                self.sectionlist.append(name)
206                self.section_start_lines[name] = self.new_start_line
207                self.new_start_line = 0
208
209#        self.config.log.debug("Section: %s : %s", name, pformat(vars(self)))
210
211        if start_new:
212            self.section = SECTION_DEFAULT
213            self._contents = []
214
215
216class KernelDoc:
217    """
218    Read a C language source or header FILE and extract embedded
219    documentation comments.
220    """
221
222    # Section names
223
224    section_context = "Context"
225    section_return = "Return"
226
227    undescribed = "-- undescribed --"
228
229    def __init__(self, config, fname):
230        """Initialize internal variables"""
231
232        self.fname = fname
233        self.config = config
234
235        # Initial state for the state machines
236        self.state = state.NORMAL
237
238        # Store entry currently being processed
239        self.entry = None
240
241        # Place all potential outputs into an array
242        self.entries = []
243
244    def emit_msg(self, ln, msg, warning=True):
245        """Emit a message"""
246
247        log_msg = f"{self.fname}:{ln} {msg}"
248
249        if self.entry:
250            self.entry.emit_msg(log_msg, warning)
251            return
252
253        if warning:
254            self.config.log.warning(log_msg)
255        else:
256            self.config.log.info(log_msg)
257
258    def dump_section(self, start_new=True):
259        """
260        Dumps section contents to arrays/hashes intended for that purpose.
261        """
262
263        if self.entry:
264            self.entry.dump_section(start_new)
265
266    # TODO: rename it to store_declaration after removal of kernel-doc.pl
267    def output_declaration(self, dtype, name, **args):
268        """
269        Stores the entry into an entry array.
270
271        The actual output and output filters will be handled elsewhere
272        """
273
274        item = KdocItem(name, dtype, self.entry.declaration_start_line, **args)
275        item.warnings = self.entry.warnings
276
277        sections = item.get('sections', {})
278        sectionlist = item.get('sectionlist', [])
279
280        # Drop empty sections
281        # TODO: improve empty sections logic to emit warnings
282        for section in ["Description", "Return"]:
283            if section in sectionlist and not sections[section].rstrip():
284                del sections[section]
285                sectionlist.remove(section)
286
287        self.entries.append((name, item))
288
289        self.config.log.debug("Output: %s:%s = %s", dtype, name, pformat(args))
290
291    def reset_state(self, ln):
292        """
293        Ancillary routine to create a new entry. It initializes all
294        variables used by the state machine.
295        """
296
297        self.entry = KernelEntry(self.config, ln)
298
299        # State flags
300        self.state = state.NORMAL
301
302    def push_parameter(self, ln, decl_type, param, dtype,
303                       org_arg, declaration_name):
304        """
305        Store parameters and their descriptions at self.entry.
306        """
307
308        if self.entry.anon_struct_union and dtype == "" and param == "}":
309            return  # Ignore the ending }; from anonymous struct/union
310
311        self.entry.anon_struct_union = False
312
313        param = KernRe(r'[\[\)].*').sub('', param, count=1)
314
315        if dtype == "" and param.endswith("..."):
316            if KernRe(r'\w\.\.\.$').search(param):
317                # For named variable parameters of the form `x...`,
318                # remove the dots
319                param = param[:-3]
320            else:
321                # Handles unnamed variable parameters
322                param = "..."
323
324            if param not in self.entry.parameterdescs or \
325                not self.entry.parameterdescs[param]:
326
327                self.entry.parameterdescs[param] = "variable arguments"
328
329        elif dtype == "" and (not param or param == "void"):
330            param = "void"
331            self.entry.parameterdescs[param] = "no arguments"
332
333        elif dtype == "" and param in ["struct", "union"]:
334            # Handle unnamed (anonymous) union or struct
335            dtype = param
336            param = "{unnamed_" + param + "}"
337            self.entry.parameterdescs[param] = "anonymous\n"
338            self.entry.anon_struct_union = True
339
340        # Handle cache group enforcing variables: they do not need
341        # to be described in header files
342        elif "__cacheline_group" in param:
343            # Ignore __cacheline_group_begin and __cacheline_group_end
344            return
345
346        # Warn if parameter has no description
347        # (but ignore ones starting with # as these are not parameters
348        # but inline preprocessor statements)
349        if param not in self.entry.parameterdescs and not param.startswith("#"):
350            self.entry.parameterdescs[param] = self.undescribed
351
352            if "." not in param:
353                if decl_type == 'function':
354                    dname = f"{decl_type} parameter"
355                else:
356                    dname = f"{decl_type} member"
357
358                self.emit_msg(ln,
359                              f"{dname} '{param}' not described in '{declaration_name}'")
360
361        # Strip spaces from param so that it is one continuous string on
362        # parameterlist. This fixes a problem where check_sections()
363        # cannot find a parameter like "addr[6 + 2]" because it actually
364        # appears as "addr[6", "+", "2]" on the parameter list.
365        # However, it's better to maintain the param string unchanged for
366        # output, so just weaken the string compare in check_sections()
367        # to ignore "[blah" in a parameter string.
368
369        self.entry.parameterlist.append(param)
370        org_arg = KernRe(r'\s\s+').sub(' ', org_arg)
371        self.entry.parametertypes[param] = org_arg
372
373    def save_struct_actual(self, actual):
374        """
375        Strip all spaces from the actual param so that it looks like
376        one string item.
377        """
378
379        actual = KernRe(r'\s*').sub("", actual, count=1)
380
381        self.entry.struct_actual += actual + " "
382
383    def create_parameter_list(self, ln, decl_type, args,
384                              splitter, declaration_name):
385        """
386        Creates a list of parameters, storing them at self.entry.
387        """
388
389        # temporarily replace all commas inside function pointer definition
390        arg_expr = KernRe(r'(\([^\),]+),')
391        while arg_expr.search(args):
392            args = arg_expr.sub(r"\1#", args)
393
394        for arg in args.split(splitter):
395            # Strip comments
396            arg = KernRe(r'\/\*.*\*\/').sub('', arg)
397
398            # Ignore argument attributes
399            arg = KernRe(r'\sPOS0?\s').sub(' ', arg)
400
401            # Strip leading/trailing spaces
402            arg = arg.strip()
403            arg = KernRe(r'\s+').sub(' ', arg, count=1)
404
405            if arg.startswith('#'):
406                # Treat preprocessor directive as a typeless variable just to fill
407                # corresponding data structures "correctly". Catch it later in
408                # output_* subs.
409
410                # Treat preprocessor directive as a typeless variable
411                self.push_parameter(ln, decl_type, arg, "",
412                                    "", declaration_name)
413
414            elif KernRe(r'\(.+\)\s*\(').search(arg):
415                # Pointer-to-function
416
417                arg = arg.replace('#', ',')
418
419                r = KernRe(r'[^\(]+\(\*?\s*([\w\[\]\.]*)\s*\)')
420                if r.match(arg):
421                    param = r.group(1)
422                else:
423                    self.emit_msg(ln, f"Invalid param: {arg}")
424                    param = arg
425
426                dtype = KernRe(r'([^\(]+\(\*?)\s*' + re.escape(param)).sub(r'\1', arg)
427                self.save_struct_actual(param)
428                self.push_parameter(ln, decl_type, param, dtype,
429                                    arg, declaration_name)
430
431            elif KernRe(r'\(.+\)\s*\[').search(arg):
432                # Array-of-pointers
433
434                arg = arg.replace('#', ',')
435                r = KernRe(r'[^\(]+\(\s*\*\s*([\w\[\]\.]*?)\s*(\s*\[\s*[\w]+\s*\]\s*)*\)')
436                if r.match(arg):
437                    param = r.group(1)
438                else:
439                    self.emit_msg(ln, f"Invalid param: {arg}")
440                    param = arg
441
442                dtype = KernRe(r'([^\(]+\(\*?)\s*' + re.escape(param)).sub(r'\1', arg)
443
444                self.save_struct_actual(param)
445                self.push_parameter(ln, decl_type, param, dtype,
446                                    arg, declaration_name)
447
448            elif arg:
449                arg = KernRe(r'\s*:\s*').sub(":", arg)
450                arg = KernRe(r'\s*\[').sub('[', arg)
451
452                args = KernRe(r'\s*,\s*').split(arg)
453                if args[0] and '*' in args[0]:
454                    args[0] = re.sub(r'(\*+)\s*', r' \1', args[0])
455
456                first_arg = []
457                r = KernRe(r'^(.*\s+)(.*?\[.*\].*)$')
458                if args[0] and r.match(args[0]):
459                    args.pop(0)
460                    first_arg.extend(r.group(1))
461                    first_arg.append(r.group(2))
462                else:
463                    first_arg = KernRe(r'\s+').split(args.pop(0))
464
465                args.insert(0, first_arg.pop())
466                dtype = ' '.join(first_arg)
467
468                for param in args:
469                    if KernRe(r'^(\*+)\s*(.*)').match(param):
470                        r = KernRe(r'^(\*+)\s*(.*)')
471                        if not r.match(param):
472                            self.emit_msg(ln, f"Invalid param: {param}")
473                            continue
474
475                        param = r.group(1)
476
477                        self.save_struct_actual(r.group(2))
478                        self.push_parameter(ln, decl_type, r.group(2),
479                                            f"{dtype} {r.group(1)}",
480                                            arg, declaration_name)
481
482                    elif KernRe(r'(.*?):(\w+)').search(param):
483                        r = KernRe(r'(.*?):(\w+)')
484                        if not r.match(param):
485                            self.emit_msg(ln, f"Invalid param: {param}")
486                            continue
487
488                        if dtype != "":  # Skip unnamed bit-fields
489                            self.save_struct_actual(r.group(1))
490                            self.push_parameter(ln, decl_type, r.group(1),
491                                                f"{dtype}:{r.group(2)}",
492                                                arg, declaration_name)
493                    else:
494                        self.save_struct_actual(param)
495                        self.push_parameter(ln, decl_type, param, dtype,
496                                            arg, declaration_name)
497
498    def check_sections(self, ln, decl_name, decl_type, sectcheck, prmscheck):
499        """
500        Check for errors inside sections, emitting warnings if not found
501        parameters are described.
502        """
503
504        sects = sectcheck.split()
505        prms = prmscheck.split()
506        err = False
507
508        for sx in range(len(sects)):                  # pylint: disable=C0200
509            err = True
510            for px in range(len(prms)):               # pylint: disable=C0200
511                prm_clean = prms[px]
512                prm_clean = KernRe(r'\[.*\]').sub('', prm_clean)
513                prm_clean = attribute.sub('', prm_clean)
514
515                # ignore array size in a parameter string;
516                # however, the original param string may contain
517                # spaces, e.g.:  addr[6 + 2]
518                # and this appears in @prms as "addr[6" since the
519                # parameter list is split at spaces;
520                # hence just ignore "[..." for the sections check;
521                prm_clean = KernRe(r'\[.*').sub('', prm_clean)
522
523                if prm_clean == sects[sx]:
524                    err = False
525                    break
526
527            if err:
528                if decl_type == 'function':
529                    dname = f"{decl_type} parameter"
530                else:
531                    dname = f"{decl_type} member"
532
533                self.emit_msg(ln,
534                              f"Excess {dname} '{sects[sx]}' description in '{decl_name}'")
535
536    def check_return_section(self, ln, declaration_name, return_type):
537        """
538        If the function doesn't return void, warns about the lack of a
539        return description.
540        """
541
542        if not self.config.wreturn:
543            return
544
545        # Ignore an empty return type (It's a macro)
546        # Ignore functions with a "void" return type (but not "void *")
547        if not return_type or KernRe(r'void\s*\w*\s*$').search(return_type):
548            return
549
550        if not self.entry.sections.get("Return", None):
551            self.emit_msg(ln,
552                          f"No description found for return value of '{declaration_name}'")
553
554    def dump_struct(self, ln, proto):
555        """
556        Store an entry for an struct or union
557        """
558
559        type_pattern = r'(struct|union)'
560
561        qualifiers = [
562            "__attribute__",
563            "__packed",
564            "__aligned",
565            "____cacheline_aligned_in_smp",
566            "____cacheline_aligned",
567        ]
568
569        definition_body = r'\{(.*)\}\s*' + "(?:" + '|'.join(qualifiers) + ")?"
570        struct_members = KernRe(type_pattern + r'([^\{\};]+)(\{)([^\{\}]*)(\})([^\{\}\;]*)(\;)')
571
572        # Extract struct/union definition
573        members = None
574        declaration_name = None
575        decl_type = None
576
577        r = KernRe(type_pattern + r'\s+(\w+)\s*' + definition_body)
578        if r.search(proto):
579            decl_type = r.group(1)
580            declaration_name = r.group(2)
581            members = r.group(3)
582        else:
583            r = KernRe(r'typedef\s+' + type_pattern + r'\s*' + definition_body + r'\s*(\w+)\s*;')
584
585            if r.search(proto):
586                decl_type = r.group(1)
587                declaration_name = r.group(3)
588                members = r.group(2)
589
590        if not members:
591            self.emit_msg(ln, f"{proto} error: Cannot parse struct or union!")
592            return
593
594        if self.entry.identifier != declaration_name:
595            self.emit_msg(ln,
596                          f"expecting prototype for {decl_type} {self.entry.identifier}. Prototype was for {decl_type} {declaration_name} instead\n")
597            return
598
599        args_pattern = r'([^,)]+)'
600
601        sub_prefixes = [
602            (KernRe(r'\/\*\s*private:.*?\/\*\s*public:.*?\*\/', re.S | re.I), ''),
603            (KernRe(r'\/\*\s*private:.*', re.S | re.I), ''),
604
605            # Strip comments
606            (KernRe(r'\/\*.*?\*\/', re.S), ''),
607
608            # Strip attributes
609            (attribute, ' '),
610            (KernRe(r'\s*__aligned\s*\([^;]*\)', re.S), ' '),
611            (KernRe(r'\s*__counted_by\s*\([^;]*\)', re.S), ' '),
612            (KernRe(r'\s*__counted_by_(le|be)\s*\([^;]*\)', re.S), ' '),
613            (KernRe(r'\s*__packed\s*', re.S), ' '),
614            (KernRe(r'\s*CRYPTO_MINALIGN_ATTR', re.S), ' '),
615            (KernRe(r'\s*____cacheline_aligned_in_smp', re.S), ' '),
616            (KernRe(r'\s*____cacheline_aligned', re.S), ' '),
617
618            # Unwrap struct_group macros based on this definition:
619            # __struct_group(TAG, NAME, ATTRS, MEMBERS...)
620            # which has variants like: struct_group(NAME, MEMBERS...)
621            # Only MEMBERS arguments require documentation.
622            #
623            # Parsing them happens on two steps:
624            #
625            # 1. drop struct group arguments that aren't at MEMBERS,
626            #    storing them as STRUCT_GROUP(MEMBERS)
627            #
628            # 2. remove STRUCT_GROUP() ancillary macro.
629            #
630            # The original logic used to remove STRUCT_GROUP() using an
631            # advanced regex:
632            #
633            #   \bSTRUCT_GROUP(\(((?:(?>[^)(]+)|(?1))*)\))[^;]*;
634            #
635            # with two patterns that are incompatible with
636            # Python re module, as it has:
637            #
638            #   - a recursive pattern: (?1)
639            #   - an atomic grouping: (?>...)
640            #
641            # I tried a simpler version: but it didn't work either:
642            #   \bSTRUCT_GROUP\(([^\)]+)\)[^;]*;
643            #
644            # As it doesn't properly match the end parenthesis on some cases.
645            #
646            # So, a better solution was crafted: there's now a NestedMatch
647            # class that ensures that delimiters after a search are properly
648            # matched. So, the implementation to drop STRUCT_GROUP() will be
649            # handled in separate.
650
651            (KernRe(r'\bstruct_group\s*\(([^,]*,)', re.S), r'STRUCT_GROUP('),
652            (KernRe(r'\bstruct_group_attr\s*\(([^,]*,){2}', re.S), r'STRUCT_GROUP('),
653            (KernRe(r'\bstruct_group_tagged\s*\(([^,]*),([^,]*),', re.S), r'struct \1 \2; STRUCT_GROUP('),
654            (KernRe(r'\b__struct_group\s*\(([^,]*,){3}', re.S), r'STRUCT_GROUP('),
655
656            # Replace macros
657            #
658            # TODO: use NestedMatch for FOO($1, $2, ...) matches
659            #
660            # it is better to also move those to the NestedMatch logic,
661            # to ensure that parenthesis will be properly matched.
662
663            (KernRe(r'__ETHTOOL_DECLARE_LINK_MODE_MASK\s*\(([^\)]+)\)', re.S), r'DECLARE_BITMAP(\1, __ETHTOOL_LINK_MODE_MASK_NBITS)'),
664            (KernRe(r'DECLARE_PHY_INTERFACE_MASK\s*\(([^\)]+)\)', re.S), r'DECLARE_BITMAP(\1, PHY_INTERFACE_MODE_MAX)'),
665            (KernRe(r'DECLARE_BITMAP\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'unsigned long \1[BITS_TO_LONGS(\2)]'),
666            (KernRe(r'DECLARE_HASHTABLE\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'unsigned long \1[1 << ((\2) - 1)]'),
667            (KernRe(r'DECLARE_KFIFO\s*\(' + args_pattern + r',\s*' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\2 *\1'),
668            (KernRe(r'DECLARE_KFIFO_PTR\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\2 *\1'),
669            (KernRe(r'(?:__)?DECLARE_FLEX_ARRAY\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\1 \2[]'),
670            (KernRe(r'DEFINE_DMA_UNMAP_ADDR\s*\(' + args_pattern + r'\)', re.S), r'dma_addr_t \1'),
671            (KernRe(r'DEFINE_DMA_UNMAP_LEN\s*\(' + args_pattern + r'\)', re.S), r'__u32 \1'),
672        ]
673
674        # Regexes here are guaranteed to have the end limiter matching
675        # the start delimiter. Yet, right now, only one replace group
676        # is allowed.
677
678        sub_nested_prefixes = [
679            (re.compile(r'\bSTRUCT_GROUP\('), r'\1'),
680        ]
681
682        for search, sub in sub_prefixes:
683            members = search.sub(sub, members)
684
685        nested = NestedMatch()
686
687        for search, sub in sub_nested_prefixes:
688            members = nested.sub(search, sub, members)
689
690        # Keeps the original declaration as-is
691        declaration = members
692
693        # Split nested struct/union elements
694        #
695        # This loop was simpler at the original kernel-doc perl version, as
696        #   while ($members =~ m/$struct_members/) { ... }
697        # reads 'members' string on each interaction.
698        #
699        # Python behavior is different: it parses 'members' only once,
700        # creating a list of tuples from the first interaction.
701        #
702        # On other words, this won't get nested structs.
703        #
704        # So, we need to have an extra loop on Python to override such
705        # re limitation.
706
707        while True:
708            tuples = struct_members.findall(members)
709            if not tuples:
710                break
711
712            for t in tuples:
713                newmember = ""
714                maintype = t[0]
715                s_ids = t[5]
716                content = t[3]
717
718                oldmember = "".join(t)
719
720                for s_id in s_ids.split(','):
721                    s_id = s_id.strip()
722
723                    newmember += f"{maintype} {s_id}; "
724                    s_id = KernRe(r'[:\[].*').sub('', s_id)
725                    s_id = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', s_id)
726
727                    for arg in content.split(';'):
728                        arg = arg.strip()
729
730                        if not arg:
731                            continue
732
733                        r = KernRe(r'^([^\(]+\(\*?\s*)([\w\.]*)(\s*\).*)')
734                        if r.match(arg):
735                            # Pointer-to-function
736                            dtype = r.group(1)
737                            name = r.group(2)
738                            extra = r.group(3)
739
740                            if not name:
741                                continue
742
743                            if not s_id:
744                                # Anonymous struct/union
745                                newmember += f"{dtype}{name}{extra}; "
746                            else:
747                                newmember += f"{dtype}{s_id}.{name}{extra}; "
748
749                        else:
750                            arg = arg.strip()
751                            # Handle bitmaps
752                            arg = KernRe(r':\s*\d+\s*').sub('', arg)
753
754                            # Handle arrays
755                            arg = KernRe(r'\[.*\]').sub('', arg)
756
757                            # Handle multiple IDs
758                            arg = KernRe(r'\s*,\s*').sub(',', arg)
759
760                            r = KernRe(r'(.*)\s+([\S+,]+)')
761
762                            if r.search(arg):
763                                dtype = r.group(1)
764                                names = r.group(2)
765                            else:
766                                newmember += f"{arg}; "
767                                continue
768
769                            for name in names.split(','):
770                                name = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', name).strip()
771
772                                if not name:
773                                    continue
774
775                                if not s_id:
776                                    # Anonymous struct/union
777                                    newmember += f"{dtype} {name}; "
778                                else:
779                                    newmember += f"{dtype} {s_id}.{name}; "
780
781                members = members.replace(oldmember, newmember)
782
783        # Ignore other nested elements, like enums
784        members = re.sub(r'(\{[^\{\}]*\})', '', members)
785
786        self.create_parameter_list(ln, decl_type, members, ';',
787                                   declaration_name)
788        self.check_sections(ln, declaration_name, decl_type,
789                            self.entry.sectcheck, self.entry.struct_actual)
790
791        # Adjust declaration for better display
792        declaration = KernRe(r'([\{;])').sub(r'\1\n', declaration)
793        declaration = KernRe(r'\}\s+;').sub('};', declaration)
794
795        # Better handle inlined enums
796        while True:
797            r = KernRe(r'(enum\s+\{[^\}]+),([^\n])')
798            if not r.search(declaration):
799                break
800
801            declaration = r.sub(r'\1,\n\2', declaration)
802
803        def_args = declaration.split('\n')
804        level = 1
805        declaration = ""
806        for clause in def_args:
807
808            clause = clause.strip()
809            clause = KernRe(r'\s+').sub(' ', clause, count=1)
810
811            if not clause:
812                continue
813
814            if '}' in clause and level > 1:
815                level -= 1
816
817            if not KernRe(r'^\s*#').match(clause):
818                declaration += "\t" * level
819
820            declaration += "\t" + clause + "\n"
821            if "{" in clause and "}" not in clause:
822                level += 1
823
824        self.output_declaration(decl_type, declaration_name,
825                                struct=declaration_name,
826                                definition=declaration,
827                                parameterlist=self.entry.parameterlist,
828                                parameterdescs=self.entry.parameterdescs,
829                                parametertypes=self.entry.parametertypes,
830                                parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
831                                sectionlist=self.entry.sectionlist,
832                                sections=self.entry.sections,
833                                section_start_lines=self.entry.section_start_lines,
834                                purpose=self.entry.declaration_purpose)
835
836    def dump_enum(self, ln, proto):
837        """
838        Stores an enum inside self.entries array.
839        """
840
841        # Ignore members marked private
842        proto = KernRe(r'\/\*\s*private:.*?\/\*\s*public:.*?\*\/', flags=re.S).sub('', proto)
843        proto = KernRe(r'\/\*\s*private:.*}', flags=re.S).sub('}', proto)
844
845        # Strip comments
846        proto = KernRe(r'\/\*.*?\*\/', flags=re.S).sub('', proto)
847
848        # Strip #define macros inside enums
849        proto = KernRe(r'#\s*((define|ifdef|if)\s+|endif)[^;]*;', flags=re.S).sub('', proto)
850
851        #
852        # Parse out the name and members of the enum.  Typedef form first.
853        #
854        r = KernRe(r'typedef\s+enum\s*\{(.*)\}\s*(\w*)\s*;')
855        if r.search(proto):
856            declaration_name = r.group(2)
857            members = r.group(1).rstrip()
858        #
859        # Failing that, look for a straight enum
860        #
861        else:
862            r = KernRe(r'enum\s+(\w*)\s*\{(.*)\}')
863            if r.match(proto):
864                declaration_name = r.group(1)
865                members = r.group(2).rstrip()
866        #
867        # OK, this isn't going to work.
868        #
869            else:
870                self.emit_msg(ln, f"{proto}: error: Cannot parse enum!")
871                return
872        #
873        # Make sure we found what we were expecting.
874        #
875        if self.entry.identifier != declaration_name:
876            if self.entry.identifier == "":
877                self.emit_msg(ln,
878                              f"{proto}: wrong kernel-doc identifier on prototype")
879            else:
880                self.emit_msg(ln,
881                              f"expecting prototype for enum {self.entry.identifier}. "
882                              f"Prototype was for enum {declaration_name} instead")
883            return
884
885        if not declaration_name:
886            declaration_name = "(anonymous)"
887        #
888        # Parse out the name of each enum member, and verify that we
889        # have a description for it.
890        #
891        member_set = set()
892        members = KernRe(r'\([^;)]*\)').sub('', members)
893        for arg in members.split(','):
894            if not arg:
895                continue
896            arg = KernRe(r'^\s*(\w+).*').sub(r'\1', arg)
897            self.entry.parameterlist.append(arg)
898            if arg not in self.entry.parameterdescs:
899                self.entry.parameterdescs[arg] = self.undescribed
900                self.emit_msg(ln,
901                              f"Enum value '{arg}' not described in enum '{declaration_name}'")
902            member_set.add(arg)
903        #
904        # Ensure that every described member actually exists in the enum.
905        #
906        for k in self.entry.parameterdescs:
907            if k not in member_set:
908                self.emit_msg(ln,
909                              f"Excess enum value '%{k}' description in '{declaration_name}'")
910
911        self.output_declaration('enum', declaration_name,
912                                enum=declaration_name,
913                                parameterlist=self.entry.parameterlist,
914                                parameterdescs=self.entry.parameterdescs,
915                                parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
916                                sectionlist=self.entry.sectionlist,
917                                sections=self.entry.sections,
918                                section_start_lines=self.entry.section_start_lines,
919                                purpose=self.entry.declaration_purpose)
920
921    def dump_declaration(self, ln, prototype):
922        """
923        Stores a data declaration inside self.entries array.
924        """
925
926        if self.entry.decl_type == "enum":
927            self.dump_enum(ln, prototype)
928            return
929
930        if self.entry.decl_type == "typedef":
931            self.dump_typedef(ln, prototype)
932            return
933
934        if self.entry.decl_type in ["union", "struct"]:
935            self.dump_struct(ln, prototype)
936            return
937
938        self.output_declaration(self.entry.decl_type, prototype,
939                                entry=self.entry)
940
941    def dump_function(self, ln, prototype):
942        """
943        Stores a function of function macro inside self.entries array.
944        """
945
946        func_macro = False
947        return_type = ''
948        decl_type = 'function'
949
950        # Prefixes that would be removed
951        sub_prefixes = [
952            (r"^static +", "", 0),
953            (r"^extern +", "", 0),
954            (r"^asmlinkage +", "", 0),
955            (r"^inline +", "", 0),
956            (r"^__inline__ +", "", 0),
957            (r"^__inline +", "", 0),
958            (r"^__always_inline +", "", 0),
959            (r"^noinline +", "", 0),
960            (r"^__FORTIFY_INLINE +", "", 0),
961            (r"__init +", "", 0),
962            (r"__init_or_module +", "", 0),
963            (r"__deprecated +", "", 0),
964            (r"__flatten +", "", 0),
965            (r"__meminit +", "", 0),
966            (r"__must_check +", "", 0),
967            (r"__weak +", "", 0),
968            (r"__sched +", "", 0),
969            (r"_noprof", "", 0),
970            (r"__printf\s*\(\s*\d*\s*,\s*\d*\s*\) +", "", 0),
971            (r"__(?:re)?alloc_size\s*\(\s*\d+\s*(?:,\s*\d+\s*)?\) +", "", 0),
972            (r"__diagnose_as\s*\(\s*\S+\s*(?:,\s*\d+\s*)*\) +", "", 0),
973            (r"DECL_BUCKET_PARAMS\s*\(\s*(\S+)\s*,\s*(\S+)\s*\)", r"\1, \2", 0),
974            (r"__attribute_const__ +", "", 0),
975
976            # It seems that Python support for re.X is broken:
977            # At least for me (Python 3.13), this didn't work
978#            (r"""
979#              __attribute__\s*\(\(
980#                (?:
981#                    [\w\s]+          # attribute name
982#                    (?:\([^)]*\))?   # attribute arguments
983#                    \s*,?            # optional comma at the end
984#                )+
985#              \)\)\s+
986#             """, "", re.X),
987
988            # So, remove whitespaces and comments from it
989            (r"__attribute__\s*\(\((?:[\w\s]+(?:\([^)]*\))?\s*,?)+\)\)\s+", "", 0),
990        ]
991
992        for search, sub, flags in sub_prefixes:
993            prototype = KernRe(search, flags).sub(sub, prototype)
994
995        # Macros are a special case, as they change the prototype format
996        new_proto = KernRe(r"^#\s*define\s+").sub("", prototype)
997        if new_proto != prototype:
998            is_define_proto = True
999            prototype = new_proto
1000        else:
1001            is_define_proto = False
1002
1003        # Yes, this truly is vile.  We are looking for:
1004        # 1. Return type (may be nothing if we're looking at a macro)
1005        # 2. Function name
1006        # 3. Function parameters.
1007        #
1008        # All the while we have to watch out for function pointer parameters
1009        # (which IIRC is what the two sections are for), C types (these
1010        # regexps don't even start to express all the possibilities), and
1011        # so on.
1012        #
1013        # If you mess with these regexps, it's a good idea to check that
1014        # the following functions' documentation still comes out right:
1015        # - parport_register_device (function pointer parameters)
1016        # - atomic_set (macro)
1017        # - pci_match_device, __copy_to_user (long return type)
1018
1019        name = r'[a-zA-Z0-9_~:]+'
1020        prototype_end1 = r'[^\(]*'
1021        prototype_end2 = r'[^\{]*'
1022        prototype_end = fr'\(({prototype_end1}|{prototype_end2})\)'
1023
1024        # Besides compiling, Perl qr{[\w\s]+} works as a non-capturing group.
1025        # So, this needs to be mapped in Python with (?:...)? or (?:...)+
1026
1027        type1 = r'(?:[\w\s]+)?'
1028        type2 = r'(?:[\w\s]+\*+)+'
1029
1030        found = False
1031
1032        if is_define_proto:
1033            r = KernRe(r'^()(' + name + r')\s+')
1034
1035            if r.search(prototype):
1036                return_type = ''
1037                declaration_name = r.group(2)
1038                func_macro = True
1039
1040                found = True
1041
1042        if not found:
1043            patterns = [
1044                rf'^()({name})\s*{prototype_end}',
1045                rf'^({type1})\s+({name})\s*{prototype_end}',
1046                rf'^({type2})\s*({name})\s*{prototype_end}',
1047            ]
1048
1049            for p in patterns:
1050                r = KernRe(p)
1051
1052                if r.match(prototype):
1053
1054                    return_type = r.group(1)
1055                    declaration_name = r.group(2)
1056                    args = r.group(3)
1057
1058                    self.create_parameter_list(ln, decl_type, args, ',',
1059                                               declaration_name)
1060
1061                    found = True
1062                    break
1063        if not found:
1064            self.emit_msg(ln,
1065                          f"cannot understand function prototype: '{prototype}'")
1066            return
1067
1068        if self.entry.identifier != declaration_name:
1069            self.emit_msg(ln,
1070                          f"expecting prototype for {self.entry.identifier}(). Prototype was for {declaration_name}() instead")
1071            return
1072
1073        prms = " ".join(self.entry.parameterlist)
1074        self.check_sections(ln, declaration_name, "function",
1075                            self.entry.sectcheck, prms)
1076
1077        self.check_return_section(ln, declaration_name, return_type)
1078
1079        if 'typedef' in return_type:
1080            self.output_declaration(decl_type, declaration_name,
1081                                    function=declaration_name,
1082                                    typedef=True,
1083                                    functiontype=return_type,
1084                                    parameterlist=self.entry.parameterlist,
1085                                    parameterdescs=self.entry.parameterdescs,
1086                                    parametertypes=self.entry.parametertypes,
1087                                    parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
1088                                    sectionlist=self.entry.sectionlist,
1089                                    sections=self.entry.sections,
1090                                    section_start_lines=self.entry.section_start_lines,
1091                                    purpose=self.entry.declaration_purpose,
1092                                    func_macro=func_macro)
1093        else:
1094            self.output_declaration(decl_type, declaration_name,
1095                                    function=declaration_name,
1096                                    typedef=False,
1097                                    functiontype=return_type,
1098                                    parameterlist=self.entry.parameterlist,
1099                                    parameterdescs=self.entry.parameterdescs,
1100                                    parametertypes=self.entry.parametertypes,
1101                                    parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
1102                                    sectionlist=self.entry.sectionlist,
1103                                    sections=self.entry.sections,
1104                                    section_start_lines=self.entry.section_start_lines,
1105                                    purpose=self.entry.declaration_purpose,
1106                                    func_macro=func_macro)
1107
1108    def dump_typedef(self, ln, proto):
1109        """
1110        Stores a typedef inside self.entries array.
1111        """
1112
1113        typedef_type = r'((?:\s+[\w\*]+\b){0,7}\s+(?:\w+\b|\*+))\s*'
1114        typedef_ident = r'\*?\s*(\w\S+)\s*'
1115        typedef_args = r'\s*\((.*)\);'
1116
1117        typedef1 = KernRe(r'typedef' + typedef_type + r'\(' + typedef_ident + r'\)' + typedef_args)
1118        typedef2 = KernRe(r'typedef' + typedef_type + typedef_ident + typedef_args)
1119
1120        # Strip comments
1121        proto = KernRe(r'/\*.*?\*/', flags=re.S).sub('', proto)
1122
1123        # Parse function typedef prototypes
1124        for r in [typedef1, typedef2]:
1125            if not r.match(proto):
1126                continue
1127
1128            return_type = r.group(1).strip()
1129            declaration_name = r.group(2)
1130            args = r.group(3)
1131
1132            if self.entry.identifier != declaration_name:
1133                self.emit_msg(ln,
1134                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1135                return
1136
1137            decl_type = 'function'
1138            self.create_parameter_list(ln, decl_type, args, ',', declaration_name)
1139
1140            self.output_declaration(decl_type, declaration_name,
1141                                    function=declaration_name,
1142                                    typedef=True,
1143                                    functiontype=return_type,
1144                                    parameterlist=self.entry.parameterlist,
1145                                    parameterdescs=self.entry.parameterdescs,
1146                                    parametertypes=self.entry.parametertypes,
1147                                    parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
1148                                    sectionlist=self.entry.sectionlist,
1149                                    sections=self.entry.sections,
1150                                    section_start_lines=self.entry.section_start_lines,
1151                                    purpose=self.entry.declaration_purpose)
1152            return
1153
1154        # Handle nested parentheses or brackets
1155        r = KernRe(r'(\(*.\)\s*|\[*.\]\s*);$')
1156        while r.search(proto):
1157            proto = r.sub('', proto)
1158
1159        # Parse simple typedefs
1160        r = KernRe(r'typedef.*\s+(\w+)\s*;')
1161        if r.match(proto):
1162            declaration_name = r.group(1)
1163
1164            if self.entry.identifier != declaration_name:
1165                self.emit_msg(ln,
1166                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1167                return
1168
1169            self.output_declaration('typedef', declaration_name,
1170                                    typedef=declaration_name,
1171                                    sectionlist=self.entry.sectionlist,
1172                                    sections=self.entry.sections,
1173                                    section_start_lines=self.entry.section_start_lines,
1174                                    purpose=self.entry.declaration_purpose)
1175            return
1176
1177        self.emit_msg(ln, "error: Cannot parse typedef!")
1178
1179    @staticmethod
1180    def process_export(function_set, line):
1181        """
1182        process EXPORT_SYMBOL* tags
1183
1184        This method doesn't use any variable from the class, so declare it
1185        with a staticmethod decorator.
1186        """
1187
1188        # We support documenting some exported symbols with different
1189        # names.  A horrible hack.
1190        suffixes = [ '_noprof' ]
1191
1192        # Note: it accepts only one EXPORT_SYMBOL* per line, as having
1193        # multiple export lines would violate Kernel coding style.
1194
1195        if export_symbol.search(line):
1196            symbol = export_symbol.group(2)
1197        elif export_symbol_ns.search(line):
1198            symbol = export_symbol_ns.group(2)
1199        else:
1200            return False
1201        #
1202        # Found an export, trim out any special suffixes
1203        #
1204        for suffix in suffixes:
1205            symbol = symbol.removesuffix(suffix)
1206        function_set.add(symbol)
1207        return True
1208
1209    def process_normal(self, ln, line):
1210        """
1211        STATE_NORMAL: looking for the /** to begin everything.
1212        """
1213
1214        if not doc_start.match(line):
1215            return
1216
1217        # start a new entry
1218        self.reset_state(ln)
1219
1220        # next line is always the function name
1221        self.state = state.NAME
1222
1223    def process_name(self, ln, line):
1224        """
1225        STATE_NAME: Looking for the "name - description" line
1226        """
1227        #
1228        # Check for a DOC: block and handle them specially.
1229        #
1230        if doc_block.search(line):
1231
1232            if not doc_block.group(1):
1233                self.entry.begin_section(ln, "Introduction")
1234            else:
1235                self.entry.begin_section(ln, doc_block.group(1))
1236
1237            self.entry.identifier = self.entry.section
1238            self.state = state.DOCBLOCK
1239        #
1240        # Otherwise we're looking for a normal kerneldoc declaration line.
1241        #
1242        elif doc_decl.search(line):
1243            self.entry.identifier = doc_decl.group(1)
1244
1245            # Test for data declaration
1246            if doc_begin_data.search(line):
1247                self.entry.decl_type = doc_begin_data.group(1)
1248                self.entry.identifier = doc_begin_data.group(2)
1249            #
1250            # Look for a function description
1251            #
1252            elif doc_begin_func.search(line):
1253                self.entry.identifier = doc_begin_func.group(1)
1254                self.entry.decl_type = "function"
1255            #
1256            # We struck out.
1257            #
1258            else:
1259                self.emit_msg(ln,
1260                              f"This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst\n{line}")
1261                self.state = state.NORMAL
1262                return
1263            #
1264            # OK, set up for a new kerneldoc entry.
1265            #
1266            self.state = state.BODY
1267            self.entry.identifier = self.entry.identifier.strip(" ")
1268            # if there's no @param blocks need to set up default section here
1269            self.entry.begin_section(ln + 1)
1270            #
1271            # Find the description portion, which *should* be there but
1272            # isn't always.
1273            # (We should be able to capture this from the previous parsing - someday)
1274            #
1275            r = KernRe("[-:](.*)")
1276            if r.search(line):
1277                self.entry.declaration_purpose = trim_whitespace(r.group(1))
1278                self.state = state.DECLARATION
1279            else:
1280                self.entry.declaration_purpose = ""
1281
1282            if not self.entry.declaration_purpose and self.config.wshort_desc:
1283                self.emit_msg(ln,
1284                              f"missing initial short description on line:\n{line}")
1285
1286            if not self.entry.identifier and self.entry.decl_type != "enum":
1287                self.emit_msg(ln,
1288                              f"wrong kernel-doc identifier on line:\n{line}")
1289                self.state = state.NORMAL
1290
1291            if self.config.verbose:
1292                self.emit_msg(ln,
1293                              f"Scanning doc for {self.entry.decl_type} {self.entry.identifier}",
1294                                  warning=False)
1295        #
1296        # Failed to find an identifier. Emit a warning
1297        #
1298        else:
1299            self.emit_msg(ln, f"Cannot find identifier on line:\n{line}")
1300
1301    #
1302    # Helper function to determine if a new section is being started.
1303    #
1304    def is_new_section(self, ln, line):
1305        if doc_sect.search(line):
1306            self.state = state.BODY
1307            #
1308            # Pick out the name of our new section, tweaking it if need be.
1309            #
1310            newsection = doc_sect.group(1)
1311            if newsection.lower() == 'description':
1312                newsection = 'Description'
1313            elif newsection.lower() == 'context':
1314                newsection = 'Context'
1315                self.state = state.SPECIAL_SECTION
1316            elif newsection.lower() in ["@return", "@returns",
1317                                        "return", "returns"]:
1318                newsection = "Return"
1319                self.state = state.SPECIAL_SECTION
1320            elif newsection[0] == '@':
1321                self.state = state.SPECIAL_SECTION
1322            #
1323            # Initialize the contents, and get the new section going.
1324            #
1325            newcontents = doc_sect.group(2)
1326            if not newcontents:
1327                newcontents = ""
1328            self.dump_section()
1329            self.entry.begin_section(ln, newsection)
1330            self.entry.leading_space = None
1331
1332            self.entry.add_text(newcontents.lstrip())
1333            return True
1334        return False
1335
1336    #
1337    # Helper function to detect (and effect) the end of a kerneldoc comment.
1338    #
1339    def is_comment_end(self, ln, line):
1340        if doc_end.search(line):
1341            self.dump_section()
1342
1343            # Look for doc_com + <text> + doc_end:
1344            r = KernRe(r'\s*\*\s*[a-zA-Z_0-9:\.]+\*/')
1345            if r.match(line):
1346                self.emit_msg(ln, f"suspicious ending line: {line}")
1347
1348            self.entry.prototype = ""
1349            self.entry.new_start_line = ln + 1
1350
1351            self.state = state.PROTO
1352            return True
1353        return False
1354
1355
1356    def process_decl(self, ln, line):
1357        """
1358        STATE_DECLARATION: We've seen the beginning of a declaration
1359        """
1360        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1361            return
1362        #
1363        # Look for anything with the " * " line beginning.
1364        #
1365        if doc_content.search(line):
1366            cont = doc_content.group(1)
1367            #
1368            # A blank line means that we have moved out of the declaration
1369            # part of the comment (without any "special section" parameter
1370            # descriptions).
1371            #
1372            if cont == "":
1373                self.state = state.BODY
1374            #
1375            # Otherwise we have more of the declaration section to soak up.
1376            #
1377            else:
1378                self.entry.declaration_purpose = \
1379                    trim_whitespace(self.entry.declaration_purpose + ' ' + cont)
1380        else:
1381            # Unknown line, ignore
1382            self.emit_msg(ln, f"bad line: {line}")
1383
1384
1385    def process_special(self, ln, line):
1386        """
1387        STATE_SPECIAL_SECTION: a section ending with a blank line
1388        """
1389        #
1390        # If we have hit a blank line (only the " * " marker), then this
1391        # section is done.
1392        #
1393        if KernRe(r"\s*\*\s*$").match(line):
1394            self.entry.begin_section(ln, dump = True)
1395            self.state = state.BODY
1396            return
1397        #
1398        # Not a blank line, look for the other ways to end the section.
1399        #
1400        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1401            return
1402        #
1403        # OK, we should have a continuation of the text for this section.
1404        #
1405        if doc_content.search(line):
1406            cont = doc_content.group(1)
1407            #
1408            # If the lines of text after the first in a special section have
1409            # leading white space, we need to trim it out or Sphinx will get
1410            # confused.  For the second line (the None case), see what we
1411            # find there and remember it.
1412            #
1413            if self.entry.leading_space is None:
1414                r = KernRe(r'^(\s+)')
1415                if r.match(cont):
1416                    self.entry.leading_space = len(r.group(1))
1417                else:
1418                    self.entry.leading_space = 0
1419            #
1420            # Otherwise, before trimming any leading chars, be *sure*
1421            # that they are white space.  We should maybe warn if this
1422            # isn't the case.
1423            #
1424            for i in range(0, self.entry.leading_space):
1425                if cont[i] != " ":
1426                    self.entry.leading_space = i
1427                    break
1428            #
1429            # Add the trimmed result to the section and we're done.
1430            #
1431            self.entry.add_text(cont[self.entry.leading_space:])
1432        else:
1433            # Unknown line, ignore
1434            self.emit_msg(ln, f"bad line: {line}")
1435
1436    def process_body(self, ln, line):
1437        """
1438        STATE_BODY: the bulk of a kerneldoc comment.
1439        """
1440        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1441            return
1442
1443        if doc_content.search(line):
1444            cont = doc_content.group(1)
1445            self.entry.add_text(cont)
1446        else:
1447            # Unknown line, ignore
1448            self.emit_msg(ln, f"bad line: {line}")
1449
1450    def process_inline_name(self, ln, line):
1451        """STATE_INLINE_NAME: beginning of docbook comments within a prototype."""
1452
1453        if doc_inline_sect.search(line):
1454            self.entry.begin_section(ln, doc_inline_sect.group(1))
1455            self.entry.add_text(doc_inline_sect.group(2).lstrip())
1456            self.state = state.INLINE_TEXT
1457        elif doc_inline_end.search(line):
1458            self.dump_section()
1459            self.state = state.PROTO
1460        elif doc_content.search(line):
1461            self.emit_msg(ln, f"Incorrect use of kernel-doc format: {line}")
1462            self.state = state.PROTO
1463        # else ... ??
1464
1465    def process_inline_text(self, ln, line):
1466        """STATE_INLINE_TEXT: docbook comments within a prototype."""
1467
1468        if doc_inline_end.search(line):
1469            self.dump_section()
1470            self.state = state.PROTO
1471        elif doc_content.search(line):
1472            self.entry.add_text(doc_content.group(1))
1473        # else ... ??
1474
1475    def syscall_munge(self, ln, proto):         # pylint: disable=W0613
1476        """
1477        Handle syscall definitions
1478        """
1479
1480        is_void = False
1481
1482        # Strip newlines/CR's
1483        proto = re.sub(r'[\r\n]+', ' ', proto)
1484
1485        # Check if it's a SYSCALL_DEFINE0
1486        if 'SYSCALL_DEFINE0' in proto:
1487            is_void = True
1488
1489        # Replace SYSCALL_DEFINE with correct return type & function name
1490        proto = KernRe(r'SYSCALL_DEFINE.*\(').sub('long sys_', proto)
1491
1492        r = KernRe(r'long\s+(sys_.*?),')
1493        if r.search(proto):
1494            proto = KernRe(',').sub('(', proto, count=1)
1495        elif is_void:
1496            proto = KernRe(r'\)').sub('(void)', proto, count=1)
1497
1498        # Now delete all of the odd-numbered commas in the proto
1499        # so that argument types & names don't have a comma between them
1500        count = 0
1501        length = len(proto)
1502
1503        if is_void:
1504            length = 0  # skip the loop if is_void
1505
1506        for ix in range(length):
1507            if proto[ix] == ',':
1508                count += 1
1509                if count % 2 == 1:
1510                    proto = proto[:ix] + ' ' + proto[ix + 1:]
1511
1512        return proto
1513
1514    def tracepoint_munge(self, ln, proto):
1515        """
1516        Handle tracepoint definitions
1517        """
1518
1519        tracepointname = None
1520        tracepointargs = None
1521
1522        # Match tracepoint name based on different patterns
1523        r = KernRe(r'TRACE_EVENT\((.*?),')
1524        if r.search(proto):
1525            tracepointname = r.group(1)
1526
1527        r = KernRe(r'DEFINE_SINGLE_EVENT\((.*?),')
1528        if r.search(proto):
1529            tracepointname = r.group(1)
1530
1531        r = KernRe(r'DEFINE_EVENT\((.*?),(.*?),')
1532        if r.search(proto):
1533            tracepointname = r.group(2)
1534
1535        if tracepointname:
1536            tracepointname = tracepointname.lstrip()
1537
1538        r = KernRe(r'TP_PROTO\((.*?)\)')
1539        if r.search(proto):
1540            tracepointargs = r.group(1)
1541
1542        if not tracepointname or not tracepointargs:
1543            self.emit_msg(ln,
1544                          f"Unrecognized tracepoint format:\n{proto}\n")
1545        else:
1546            proto = f"static inline void trace_{tracepointname}({tracepointargs})"
1547            self.entry.identifier = f"trace_{self.entry.identifier}"
1548
1549        return proto
1550
1551    def process_proto_function(self, ln, line):
1552        """Ancillary routine to process a function prototype"""
1553
1554        # strip C99-style comments to end of line
1555        line = KernRe(r"\/\/.*$", re.S).sub('', line)
1556        #
1557        # Soak up the line's worth of prototype text, stopping at { or ; if present.
1558        #
1559        if KernRe(r'\s*#\s*define').match(line):
1560            self.entry.prototype = line
1561        elif not line.startswith('#'):   # skip other preprocessor stuff
1562            r = KernRe(r'([^\{]*)')
1563            if r.match(line):
1564                self.entry.prototype += r.group(1) + " "
1565        #
1566        # If we now have the whole prototype, clean it up and declare victory.
1567        #
1568        if '{' in line or ';' in line or KernRe(r'\s*#\s*define').match(line):
1569            # strip comments and surrounding spaces
1570            self.entry.prototype = KernRe(r'/\*.*\*/').sub('', self.entry.prototype).strip()
1571            #
1572            # Handle self.entry.prototypes for function pointers like:
1573            #       int (*pcs_config)(struct foo)
1574            # by turning it into
1575            #	    int pcs_config(struct foo)
1576            #
1577            r = KernRe(r'^(\S+\s+)\(\s*\*(\S+)\)')
1578            self.entry.prototype = r.sub(r'\1\2', self.entry.prototype)
1579            #
1580            # Handle special declaration syntaxes
1581            #
1582            if 'SYSCALL_DEFINE' in self.entry.prototype:
1583                self.entry.prototype = self.syscall_munge(ln,
1584                                                          self.entry.prototype)
1585            else:
1586                r = KernRe(r'TRACE_EVENT|DEFINE_EVENT|DEFINE_SINGLE_EVENT')
1587                if r.search(self.entry.prototype):
1588                    self.entry.prototype = self.tracepoint_munge(ln,
1589                                                                 self.entry.prototype)
1590            #
1591            # ... and we're done
1592            #
1593            self.dump_function(ln, self.entry.prototype)
1594            self.reset_state(ln)
1595
1596    def process_proto_type(self, ln, line):
1597        """Ancillary routine to process a type"""
1598
1599        # Strip C99-style comments and surrounding whitespace
1600        line = KernRe(r"//.*$", re.S).sub('', line).strip()
1601        if not line:
1602            return # nothing to see here
1603
1604        # To distinguish preprocessor directive from regular declaration later.
1605        if line.startswith('#'):
1606            line += ";"
1607        #
1608        # Split the declaration on any of { } or ;, and accumulate pieces
1609        # until we hit a semicolon while not inside {brackets}
1610        #
1611        r = KernRe(r'(.*?)([{};])')
1612        for chunk in r.split(line):
1613            if chunk:  # Ignore empty matches
1614                self.entry.prototype += chunk
1615                #
1616                # This cries out for a match statement ... someday after we can
1617                # drop Python 3.9 ...
1618                #
1619                if chunk == '{':
1620                    self.entry.brcount += 1
1621                elif chunk == '}':
1622                    self.entry.brcount -= 1
1623                elif chunk == ';' and self.entry.brcount <= 0:
1624                    self.dump_declaration(ln, self.entry.prototype)
1625                    self.reset_state(ln)
1626                    return
1627        #
1628        # We hit the end of the line while still in the declaration; put
1629        # in a space to represent the newline.
1630        #
1631        self.entry.prototype += ' '
1632
1633    def process_proto(self, ln, line):
1634        """STATE_PROTO: reading a function/whatever prototype."""
1635
1636        if doc_inline_oneline.search(line):
1637            self.entry.begin_section(ln, doc_inline_oneline.group(1))
1638            self.entry.add_text(doc_inline_oneline.group(2))
1639            self.dump_section()
1640
1641        elif doc_inline_start.search(line):
1642            self.state = state.INLINE_NAME
1643
1644        elif self.entry.decl_type == 'function':
1645            self.process_proto_function(ln, line)
1646
1647        else:
1648            self.process_proto_type(ln, line)
1649
1650    def process_docblock(self, ln, line):
1651        """STATE_DOCBLOCK: within a DOC: block."""
1652
1653        if doc_end.search(line):
1654            self.dump_section()
1655            self.output_declaration("doc", self.entry.identifier,
1656                                    sectionlist=self.entry.sectionlist,
1657                                    sections=self.entry.sections,
1658                                    section_start_lines=self.entry.section_start_lines)
1659            self.reset_state(ln)
1660
1661        elif doc_content.search(line):
1662            self.entry.add_text(doc_content.group(1))
1663
1664    def parse_export(self):
1665        """
1666        Parses EXPORT_SYMBOL* macros from a single Kernel source file.
1667        """
1668
1669        export_table = set()
1670
1671        try:
1672            with open(self.fname, "r", encoding="utf8",
1673                      errors="backslashreplace") as fp:
1674
1675                for line in fp:
1676                    self.process_export(export_table, line)
1677
1678        except IOError:
1679            return None
1680
1681        return export_table
1682
1683    #
1684    # The state/action table telling us which function to invoke in
1685    # each state.
1686    #
1687    state_actions = {
1688        state.NORMAL:			process_normal,
1689        state.NAME:			process_name,
1690        state.BODY:			process_body,
1691        state.DECLARATION:		process_decl,
1692        state.SPECIAL_SECTION:		process_special,
1693        state.INLINE_NAME:		process_inline_name,
1694        state.INLINE_TEXT:		process_inline_text,
1695        state.PROTO:			process_proto,
1696        state.DOCBLOCK:			process_docblock,
1697        }
1698
1699    def parse_kdoc(self):
1700        """
1701        Open and process each line of a C source file.
1702        The parsing is controlled via a state machine, and the line is passed
1703        to a different process function depending on the state. The process
1704        function may update the state as needed.
1705
1706        Besides parsing kernel-doc tags, it also parses export symbols.
1707        """
1708
1709        prev = ""
1710        prev_ln = None
1711        export_table = set()
1712
1713        try:
1714            with open(self.fname, "r", encoding="utf8",
1715                      errors="backslashreplace") as fp:
1716                for ln, line in enumerate(fp):
1717
1718                    line = line.expandtabs().strip("\n")
1719
1720                    # Group continuation lines on prototypes
1721                    if self.state == state.PROTO:
1722                        if line.endswith("\\"):
1723                            prev += line.rstrip("\\")
1724                            if not prev_ln:
1725                                prev_ln = ln
1726                            continue
1727
1728                        if prev:
1729                            ln = prev_ln
1730                            line = prev + line
1731                            prev = ""
1732                            prev_ln = None
1733
1734                    self.config.log.debug("%d %s: %s",
1735                                          ln, state.name[self.state],
1736                                          line)
1737
1738                    # This is an optimization over the original script.
1739                    # There, when export_file was used for the same file,
1740                    # it was read twice. Here, we use the already-existing
1741                    # loop to parse exported symbols as well.
1742                    #
1743                    if (self.state != state.NORMAL) or \
1744                       not self.process_export(export_table, line):
1745                        # Hand this line to the appropriate state handler
1746                        self.state_actions[self.state](self, ln, line)
1747
1748        except OSError:
1749            self.config.log.error(f"Error: Cannot open file {self.fname}")
1750
1751        return export_table, self.entries
1752