xref: /linux/scripts/lib/kdoc/kdoc_parser.py (revision 8d9d122915492ea6984f32e5df30cef5c582f062)
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.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.sectcheck += name + " "
192            self.new_start_line = 0
193
194        else:
195            if name in self.sections and self.sections[name] != "":
196                # Only warn on user-specified duplicate section names
197                if name != SECTION_DEFAULT:
198                    self.emit_msg(self.new_start_line,
199                                  f"duplicate section name '{name}'\n")
200                # Treat as a new paragraph - add a blank line
201                self.sections[name] += '\n' + contents
202            else:
203                self.sections[name] = contents
204                self.section_start_lines[name] = self.new_start_line
205                self.new_start_line = 0
206
207#        self.config.log.debug("Section: %s : %s", name, pformat(vars(self)))
208
209        if start_new:
210            self.section = SECTION_DEFAULT
211            self._contents = []
212
213
214class KernelDoc:
215    """
216    Read a C language source or header FILE and extract embedded
217    documentation comments.
218    """
219
220    # Section names
221
222    section_context = "Context"
223    section_return = "Return"
224
225    undescribed = "-- undescribed --"
226
227    def __init__(self, config, fname):
228        """Initialize internal variables"""
229
230        self.fname = fname
231        self.config = config
232
233        # Initial state for the state machines
234        self.state = state.NORMAL
235
236        # Store entry currently being processed
237        self.entry = None
238
239        # Place all potential outputs into an array
240        self.entries = []
241
242    def emit_msg(self, ln, msg, warning=True):
243        """Emit a message"""
244
245        log_msg = f"{self.fname}:{ln} {msg}"
246
247        if self.entry:
248            self.entry.emit_msg(log_msg, warning)
249            return
250
251        if warning:
252            self.config.log.warning(log_msg)
253        else:
254            self.config.log.info(log_msg)
255
256    def dump_section(self, start_new=True):
257        """
258        Dumps section contents to arrays/hashes intended for that purpose.
259        """
260
261        if self.entry:
262            self.entry.dump_section(start_new)
263
264    # TODO: rename it to store_declaration after removal of kernel-doc.pl
265    def output_declaration(self, dtype, name, **args):
266        """
267        Stores the entry into an entry array.
268
269        The actual output and output filters will be handled elsewhere
270        """
271
272        item = KdocItem(name, dtype, self.entry.declaration_start_line, **args)
273        item.warnings = self.entry.warnings
274
275        sections = item.get('sections', {})
276
277        # Drop empty sections
278        # TODO: improve empty sections logic to emit warnings
279        for section in ["Description", "Return"]:
280            if section in sections and not sections[section].rstrip():
281                del sections[section]
282
283        self.entries.append(item)
284
285        self.config.log.debug("Output: %s:%s = %s", dtype, name, pformat(args))
286
287    def reset_state(self, ln):
288        """
289        Ancillary routine to create a new entry. It initializes all
290        variables used by the state machine.
291        """
292
293        self.entry = KernelEntry(self.config, ln)
294
295        # State flags
296        self.state = state.NORMAL
297
298    def push_parameter(self, ln, decl_type, param, dtype,
299                       org_arg, declaration_name):
300        """
301        Store parameters and their descriptions at self.entry.
302        """
303
304        if self.entry.anon_struct_union and dtype == "" and param == "}":
305            return  # Ignore the ending }; from anonymous struct/union
306
307        self.entry.anon_struct_union = False
308
309        param = KernRe(r'[\[\)].*').sub('', param, count=1)
310
311        if dtype == "" and param.endswith("..."):
312            if KernRe(r'\w\.\.\.$').search(param):
313                # For named variable parameters of the form `x...`,
314                # remove the dots
315                param = param[:-3]
316            else:
317                # Handles unnamed variable parameters
318                param = "..."
319
320            if param not in self.entry.parameterdescs or \
321                not self.entry.parameterdescs[param]:
322
323                self.entry.parameterdescs[param] = "variable arguments"
324
325        elif dtype == "" and (not param or param == "void"):
326            param = "void"
327            self.entry.parameterdescs[param] = "no arguments"
328
329        elif dtype == "" and param in ["struct", "union"]:
330            # Handle unnamed (anonymous) union or struct
331            dtype = param
332            param = "{unnamed_" + param + "}"
333            self.entry.parameterdescs[param] = "anonymous\n"
334            self.entry.anon_struct_union = True
335
336        # Handle cache group enforcing variables: they do not need
337        # to be described in header files
338        elif "__cacheline_group" in param:
339            # Ignore __cacheline_group_begin and __cacheline_group_end
340            return
341
342        # Warn if parameter has no description
343        # (but ignore ones starting with # as these are not parameters
344        # but inline preprocessor statements)
345        if param not in self.entry.parameterdescs and not param.startswith("#"):
346            self.entry.parameterdescs[param] = self.undescribed
347
348            if "." not in param:
349                if decl_type == 'function':
350                    dname = f"{decl_type} parameter"
351                else:
352                    dname = f"{decl_type} member"
353
354                self.emit_msg(ln,
355                              f"{dname} '{param}' not described in '{declaration_name}'")
356
357        # Strip spaces from param so that it is one continuous string on
358        # parameterlist. This fixes a problem where check_sections()
359        # cannot find a parameter like "addr[6 + 2]" because it actually
360        # appears as "addr[6", "+", "2]" on the parameter list.
361        # However, it's better to maintain the param string unchanged for
362        # output, so just weaken the string compare in check_sections()
363        # to ignore "[blah" in a parameter string.
364
365        self.entry.parameterlist.append(param)
366        org_arg = KernRe(r'\s\s+').sub(' ', org_arg)
367        self.entry.parametertypes[param] = org_arg
368
369    def save_struct_actual(self, actual):
370        """
371        Strip all spaces from the actual param so that it looks like
372        one string item.
373        """
374
375        actual = KernRe(r'\s*').sub("", actual, count=1)
376
377        self.entry.struct_actual += actual + " "
378
379    def create_parameter_list(self, ln, decl_type, args,
380                              splitter, declaration_name):
381        """
382        Creates a list of parameters, storing them at self.entry.
383        """
384
385        # temporarily replace all commas inside function pointer definition
386        arg_expr = KernRe(r'(\([^\),]+),')
387        while arg_expr.search(args):
388            args = arg_expr.sub(r"\1#", args)
389
390        for arg in args.split(splitter):
391            # Strip comments
392            arg = KernRe(r'\/\*.*\*\/').sub('', arg)
393
394            # Ignore argument attributes
395            arg = KernRe(r'\sPOS0?\s').sub(' ', arg)
396
397            # Strip leading/trailing spaces
398            arg = arg.strip()
399            arg = KernRe(r'\s+').sub(' ', arg, count=1)
400
401            if arg.startswith('#'):
402                # Treat preprocessor directive as a typeless variable just to fill
403                # corresponding data structures "correctly". Catch it later in
404                # output_* subs.
405
406                # Treat preprocessor directive as a typeless variable
407                self.push_parameter(ln, decl_type, arg, "",
408                                    "", declaration_name)
409
410            elif KernRe(r'\(.+\)\s*\(').search(arg):
411                # Pointer-to-function
412
413                arg = arg.replace('#', ',')
414
415                r = KernRe(r'[^\(]+\(\*?\s*([\w\[\]\.]*)\s*\)')
416                if r.match(arg):
417                    param = r.group(1)
418                else:
419                    self.emit_msg(ln, f"Invalid param: {arg}")
420                    param = arg
421
422                dtype = KernRe(r'([^\(]+\(\*?)\s*' + re.escape(param)).sub(r'\1', arg)
423                self.save_struct_actual(param)
424                self.push_parameter(ln, decl_type, param, dtype,
425                                    arg, declaration_name)
426
427            elif KernRe(r'\(.+\)\s*\[').search(arg):
428                # Array-of-pointers
429
430                arg = arg.replace('#', ',')
431                r = KernRe(r'[^\(]+\(\s*\*\s*([\w\[\]\.]*?)\s*(\s*\[\s*[\w]+\s*\]\s*)*\)')
432                if r.match(arg):
433                    param = r.group(1)
434                else:
435                    self.emit_msg(ln, f"Invalid param: {arg}")
436                    param = arg
437
438                dtype = KernRe(r'([^\(]+\(\*?)\s*' + re.escape(param)).sub(r'\1', arg)
439
440                self.save_struct_actual(param)
441                self.push_parameter(ln, decl_type, param, dtype,
442                                    arg, declaration_name)
443
444            elif arg:
445                arg = KernRe(r'\s*:\s*').sub(":", arg)
446                arg = KernRe(r'\s*\[').sub('[', arg)
447
448                args = KernRe(r'\s*,\s*').split(arg)
449                if args[0] and '*' in args[0]:
450                    args[0] = re.sub(r'(\*+)\s*', r' \1', args[0])
451
452                first_arg = []
453                r = KernRe(r'^(.*\s+)(.*?\[.*\].*)$')
454                if args[0] and r.match(args[0]):
455                    args.pop(0)
456                    first_arg.extend(r.group(1))
457                    first_arg.append(r.group(2))
458                else:
459                    first_arg = KernRe(r'\s+').split(args.pop(0))
460
461                args.insert(0, first_arg.pop())
462                dtype = ' '.join(first_arg)
463
464                for param in args:
465                    if KernRe(r'^(\*+)\s*(.*)').match(param):
466                        r = KernRe(r'^(\*+)\s*(.*)')
467                        if not r.match(param):
468                            self.emit_msg(ln, f"Invalid param: {param}")
469                            continue
470
471                        param = r.group(1)
472
473                        self.save_struct_actual(r.group(2))
474                        self.push_parameter(ln, decl_type, r.group(2),
475                                            f"{dtype} {r.group(1)}",
476                                            arg, declaration_name)
477
478                    elif KernRe(r'(.*?):(\w+)').search(param):
479                        r = KernRe(r'(.*?):(\w+)')
480                        if not r.match(param):
481                            self.emit_msg(ln, f"Invalid param: {param}")
482                            continue
483
484                        if dtype != "":  # Skip unnamed bit-fields
485                            self.save_struct_actual(r.group(1))
486                            self.push_parameter(ln, decl_type, r.group(1),
487                                                f"{dtype}:{r.group(2)}",
488                                                arg, declaration_name)
489                    else:
490                        self.save_struct_actual(param)
491                        self.push_parameter(ln, decl_type, param, dtype,
492                                            arg, declaration_name)
493
494    def check_sections(self, ln, decl_name, decl_type, sectcheck, prmscheck):
495        """
496        Check for errors inside sections, emitting warnings if not found
497        parameters are described.
498        """
499
500        sects = sectcheck.split()
501        prms = prmscheck.split()
502        err = False
503
504        for sx in range(len(sects)):                  # pylint: disable=C0200
505            err = True
506            for px in range(len(prms)):               # pylint: disable=C0200
507                prm_clean = prms[px]
508                prm_clean = KernRe(r'\[.*\]').sub('', prm_clean)
509                prm_clean = attribute.sub('', prm_clean)
510
511                # ignore array size in a parameter string;
512                # however, the original param string may contain
513                # spaces, e.g.:  addr[6 + 2]
514                # and this appears in @prms as "addr[6" since the
515                # parameter list is split at spaces;
516                # hence just ignore "[..." for the sections check;
517                prm_clean = KernRe(r'\[.*').sub('', prm_clean)
518
519                if prm_clean == sects[sx]:
520                    err = False
521                    break
522
523            if err:
524                if decl_type == 'function':
525                    dname = f"{decl_type} parameter"
526                else:
527                    dname = f"{decl_type} member"
528
529                self.emit_msg(ln,
530                              f"Excess {dname} '{sects[sx]}' description in '{decl_name}'")
531
532    def check_return_section(self, ln, declaration_name, return_type):
533        """
534        If the function doesn't return void, warns about the lack of a
535        return description.
536        """
537
538        if not self.config.wreturn:
539            return
540
541        # Ignore an empty return type (It's a macro)
542        # Ignore functions with a "void" return type (but not "void *")
543        if not return_type or KernRe(r'void\s*\w*\s*$').search(return_type):
544            return
545
546        if not self.entry.sections.get("Return", None):
547            self.emit_msg(ln,
548                          f"No description found for return value of '{declaration_name}'")
549
550    def dump_struct(self, ln, proto):
551        """
552        Store an entry for an struct or union
553        """
554
555        type_pattern = r'(struct|union)'
556
557        qualifiers = [
558            "__attribute__",
559            "__packed",
560            "__aligned",
561            "____cacheline_aligned_in_smp",
562            "____cacheline_aligned",
563        ]
564
565        definition_body = r'\{(.*)\}\s*' + "(?:" + '|'.join(qualifiers) + ")?"
566        struct_members = KernRe(type_pattern + r'([^\{\};]+)(\{)([^\{\}]*)(\})([^\{\}\;]*)(\;)')
567
568        # Extract struct/union definition
569        members = None
570        declaration_name = None
571        decl_type = None
572
573        r = KernRe(type_pattern + r'\s+(\w+)\s*' + definition_body)
574        if r.search(proto):
575            decl_type = r.group(1)
576            declaration_name = r.group(2)
577            members = r.group(3)
578        else:
579            r = KernRe(r'typedef\s+' + type_pattern + r'\s*' + definition_body + r'\s*(\w+)\s*;')
580
581            if r.search(proto):
582                decl_type = r.group(1)
583                declaration_name = r.group(3)
584                members = r.group(2)
585
586        if not members:
587            self.emit_msg(ln, f"{proto} error: Cannot parse struct or union!")
588            return
589
590        if self.entry.identifier != declaration_name:
591            self.emit_msg(ln,
592                          f"expecting prototype for {decl_type} {self.entry.identifier}. Prototype was for {decl_type} {declaration_name} instead\n")
593            return
594
595        args_pattern = r'([^,)]+)'
596
597        sub_prefixes = [
598            (KernRe(r'\/\*\s*private:.*?\/\*\s*public:.*?\*\/', re.S | re.I), ''),
599            (KernRe(r'\/\*\s*private:.*', re.S | re.I), ''),
600
601            # Strip comments
602            (KernRe(r'\/\*.*?\*\/', re.S), ''),
603
604            # Strip attributes
605            (attribute, ' '),
606            (KernRe(r'\s*__aligned\s*\([^;]*\)', re.S), ' '),
607            (KernRe(r'\s*__counted_by\s*\([^;]*\)', re.S), ' '),
608            (KernRe(r'\s*__counted_by_(le|be)\s*\([^;]*\)', re.S), ' '),
609            (KernRe(r'\s*__packed\s*', re.S), ' '),
610            (KernRe(r'\s*CRYPTO_MINALIGN_ATTR', re.S), ' '),
611            (KernRe(r'\s*____cacheline_aligned_in_smp', re.S), ' '),
612            (KernRe(r'\s*____cacheline_aligned', re.S), ' '),
613
614            # Unwrap struct_group macros based on this definition:
615            # __struct_group(TAG, NAME, ATTRS, MEMBERS...)
616            # which has variants like: struct_group(NAME, MEMBERS...)
617            # Only MEMBERS arguments require documentation.
618            #
619            # Parsing them happens on two steps:
620            #
621            # 1. drop struct group arguments that aren't at MEMBERS,
622            #    storing them as STRUCT_GROUP(MEMBERS)
623            #
624            # 2. remove STRUCT_GROUP() ancillary macro.
625            #
626            # The original logic used to remove STRUCT_GROUP() using an
627            # advanced regex:
628            #
629            #   \bSTRUCT_GROUP(\(((?:(?>[^)(]+)|(?1))*)\))[^;]*;
630            #
631            # with two patterns that are incompatible with
632            # Python re module, as it has:
633            #
634            #   - a recursive pattern: (?1)
635            #   - an atomic grouping: (?>...)
636            #
637            # I tried a simpler version: but it didn't work either:
638            #   \bSTRUCT_GROUP\(([^\)]+)\)[^;]*;
639            #
640            # As it doesn't properly match the end parenthesis on some cases.
641            #
642            # So, a better solution was crafted: there's now a NestedMatch
643            # class that ensures that delimiters after a search are properly
644            # matched. So, the implementation to drop STRUCT_GROUP() will be
645            # handled in separate.
646
647            (KernRe(r'\bstruct_group\s*\(([^,]*,)', re.S), r'STRUCT_GROUP('),
648            (KernRe(r'\bstruct_group_attr\s*\(([^,]*,){2}', re.S), r'STRUCT_GROUP('),
649            (KernRe(r'\bstruct_group_tagged\s*\(([^,]*),([^,]*),', re.S), r'struct \1 \2; STRUCT_GROUP('),
650            (KernRe(r'\b__struct_group\s*\(([^,]*,){3}', re.S), r'STRUCT_GROUP('),
651
652            # Replace macros
653            #
654            # TODO: use NestedMatch for FOO($1, $2, ...) matches
655            #
656            # it is better to also move those to the NestedMatch logic,
657            # to ensure that parenthesis will be properly matched.
658
659            (KernRe(r'__ETHTOOL_DECLARE_LINK_MODE_MASK\s*\(([^\)]+)\)', re.S), r'DECLARE_BITMAP(\1, __ETHTOOL_LINK_MODE_MASK_NBITS)'),
660            (KernRe(r'DECLARE_PHY_INTERFACE_MASK\s*\(([^\)]+)\)', re.S), r'DECLARE_BITMAP(\1, PHY_INTERFACE_MODE_MAX)'),
661            (KernRe(r'DECLARE_BITMAP\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'unsigned long \1[BITS_TO_LONGS(\2)]'),
662            (KernRe(r'DECLARE_HASHTABLE\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'unsigned long \1[1 << ((\2) - 1)]'),
663            (KernRe(r'DECLARE_KFIFO\s*\(' + args_pattern + r',\s*' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\2 *\1'),
664            (KernRe(r'DECLARE_KFIFO_PTR\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\2 *\1'),
665            (KernRe(r'(?:__)?DECLARE_FLEX_ARRAY\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\1 \2[]'),
666            (KernRe(r'DEFINE_DMA_UNMAP_ADDR\s*\(' + args_pattern + r'\)', re.S), r'dma_addr_t \1'),
667            (KernRe(r'DEFINE_DMA_UNMAP_LEN\s*\(' + args_pattern + r'\)', re.S), r'__u32 \1'),
668        ]
669
670        # Regexes here are guaranteed to have the end limiter matching
671        # the start delimiter. Yet, right now, only one replace group
672        # is allowed.
673
674        sub_nested_prefixes = [
675            (re.compile(r'\bSTRUCT_GROUP\('), r'\1'),
676        ]
677
678        for search, sub in sub_prefixes:
679            members = search.sub(sub, members)
680
681        nested = NestedMatch()
682
683        for search, sub in sub_nested_prefixes:
684            members = nested.sub(search, sub, members)
685
686        # Keeps the original declaration as-is
687        declaration = members
688
689        # Split nested struct/union elements
690        #
691        # This loop was simpler at the original kernel-doc perl version, as
692        #   while ($members =~ m/$struct_members/) { ... }
693        # reads 'members' string on each interaction.
694        #
695        # Python behavior is different: it parses 'members' only once,
696        # creating a list of tuples from the first interaction.
697        #
698        # On other words, this won't get nested structs.
699        #
700        # So, we need to have an extra loop on Python to override such
701        # re limitation.
702
703        while True:
704            tuples = struct_members.findall(members)
705            if not tuples:
706                break
707
708            for t in tuples:
709                newmember = ""
710                maintype = t[0]
711                s_ids = t[5]
712                content = t[3]
713
714                oldmember = "".join(t)
715
716                for s_id in s_ids.split(','):
717                    s_id = s_id.strip()
718
719                    newmember += f"{maintype} {s_id}; "
720                    s_id = KernRe(r'[:\[].*').sub('', s_id)
721                    s_id = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', s_id)
722
723                    for arg in content.split(';'):
724                        arg = arg.strip()
725
726                        if not arg:
727                            continue
728
729                        r = KernRe(r'^([^\(]+\(\*?\s*)([\w\.]*)(\s*\).*)')
730                        if r.match(arg):
731                            # Pointer-to-function
732                            dtype = r.group(1)
733                            name = r.group(2)
734                            extra = r.group(3)
735
736                            if not name:
737                                continue
738
739                            if not s_id:
740                                # Anonymous struct/union
741                                newmember += f"{dtype}{name}{extra}; "
742                            else:
743                                newmember += f"{dtype}{s_id}.{name}{extra}; "
744
745                        else:
746                            arg = arg.strip()
747                            # Handle bitmaps
748                            arg = KernRe(r':\s*\d+\s*').sub('', arg)
749
750                            # Handle arrays
751                            arg = KernRe(r'\[.*\]').sub('', arg)
752
753                            # Handle multiple IDs
754                            arg = KernRe(r'\s*,\s*').sub(',', arg)
755
756                            r = KernRe(r'(.*)\s+([\S+,]+)')
757
758                            if r.search(arg):
759                                dtype = r.group(1)
760                                names = r.group(2)
761                            else:
762                                newmember += f"{arg}; "
763                                continue
764
765                            for name in names.split(','):
766                                name = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', name).strip()
767
768                                if not name:
769                                    continue
770
771                                if not s_id:
772                                    # Anonymous struct/union
773                                    newmember += f"{dtype} {name}; "
774                                else:
775                                    newmember += f"{dtype} {s_id}.{name}; "
776
777                members = members.replace(oldmember, newmember)
778
779        # Ignore other nested elements, like enums
780        members = re.sub(r'(\{[^\{\}]*\})', '', members)
781
782        self.create_parameter_list(ln, decl_type, members, ';',
783                                   declaration_name)
784        self.check_sections(ln, declaration_name, decl_type,
785                            self.entry.sectcheck, self.entry.struct_actual)
786
787        # Adjust declaration for better display
788        declaration = KernRe(r'([\{;])').sub(r'\1\n', declaration)
789        declaration = KernRe(r'\}\s+;').sub('};', declaration)
790
791        # Better handle inlined enums
792        while True:
793            r = KernRe(r'(enum\s+\{[^\}]+),([^\n])')
794            if not r.search(declaration):
795                break
796
797            declaration = r.sub(r'\1,\n\2', declaration)
798
799        def_args = declaration.split('\n')
800        level = 1
801        declaration = ""
802        for clause in def_args:
803
804            clause = clause.strip()
805            clause = KernRe(r'\s+').sub(' ', clause, count=1)
806
807            if not clause:
808                continue
809
810            if '}' in clause and level > 1:
811                level -= 1
812
813            if not KernRe(r'^\s*#').match(clause):
814                declaration += "\t" * level
815
816            declaration += "\t" + clause + "\n"
817            if "{" in clause and "}" not in clause:
818                level += 1
819
820        self.output_declaration(decl_type, declaration_name,
821                                struct=declaration_name,
822                                definition=declaration,
823                                parameterlist=self.entry.parameterlist,
824                                parameterdescs=self.entry.parameterdescs,
825                                parametertypes=self.entry.parametertypes,
826                                parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
827                                sections=self.entry.sections,
828                                section_start_lines=self.entry.section_start_lines,
829                                purpose=self.entry.declaration_purpose)
830
831    def dump_enum(self, ln, proto):
832        """
833        Stores an enum inside self.entries array.
834        """
835
836        # Ignore members marked private
837        proto = KernRe(r'\/\*\s*private:.*?\/\*\s*public:.*?\*\/', flags=re.S).sub('', proto)
838        proto = KernRe(r'\/\*\s*private:.*}', flags=re.S).sub('}', proto)
839
840        # Strip comments
841        proto = KernRe(r'\/\*.*?\*\/', flags=re.S).sub('', proto)
842
843        # Strip #define macros inside enums
844        proto = KernRe(r'#\s*((define|ifdef|if)\s+|endif)[^;]*;', flags=re.S).sub('', proto)
845
846        #
847        # Parse out the name and members of the enum.  Typedef form first.
848        #
849        r = KernRe(r'typedef\s+enum\s*\{(.*)\}\s*(\w*)\s*;')
850        if r.search(proto):
851            declaration_name = r.group(2)
852            members = r.group(1).rstrip()
853        #
854        # Failing that, look for a straight enum
855        #
856        else:
857            r = KernRe(r'enum\s+(\w*)\s*\{(.*)\}')
858            if r.match(proto):
859                declaration_name = r.group(1)
860                members = r.group(2).rstrip()
861        #
862        # OK, this isn't going to work.
863        #
864            else:
865                self.emit_msg(ln, f"{proto}: error: Cannot parse enum!")
866                return
867        #
868        # Make sure we found what we were expecting.
869        #
870        if self.entry.identifier != declaration_name:
871            if self.entry.identifier == "":
872                self.emit_msg(ln,
873                              f"{proto}: wrong kernel-doc identifier on prototype")
874            else:
875                self.emit_msg(ln,
876                              f"expecting prototype for enum {self.entry.identifier}. "
877                              f"Prototype was for enum {declaration_name} instead")
878            return
879
880        if not declaration_name:
881            declaration_name = "(anonymous)"
882        #
883        # Parse out the name of each enum member, and verify that we
884        # have a description for it.
885        #
886        member_set = set()
887        members = KernRe(r'\([^;)]*\)').sub('', members)
888        for arg in members.split(','):
889            if not arg:
890                continue
891            arg = KernRe(r'^\s*(\w+).*').sub(r'\1', arg)
892            self.entry.parameterlist.append(arg)
893            if arg not in self.entry.parameterdescs:
894                self.entry.parameterdescs[arg] = self.undescribed
895                self.emit_msg(ln,
896                              f"Enum value '{arg}' not described in enum '{declaration_name}'")
897            member_set.add(arg)
898        #
899        # Ensure that every described member actually exists in the enum.
900        #
901        for k in self.entry.parameterdescs:
902            if k not in member_set:
903                self.emit_msg(ln,
904                              f"Excess enum value '%{k}' description in '{declaration_name}'")
905
906        self.output_declaration('enum', declaration_name,
907                                enum=declaration_name,
908                                parameterlist=self.entry.parameterlist,
909                                parameterdescs=self.entry.parameterdescs,
910                                parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
911                                sections=self.entry.sections,
912                                section_start_lines=self.entry.section_start_lines,
913                                purpose=self.entry.declaration_purpose)
914
915    def dump_declaration(self, ln, prototype):
916        """
917        Stores a data declaration inside self.entries array.
918        """
919
920        if self.entry.decl_type == "enum":
921            self.dump_enum(ln, prototype)
922            return
923
924        if self.entry.decl_type == "typedef":
925            self.dump_typedef(ln, prototype)
926            return
927
928        if self.entry.decl_type in ["union", "struct"]:
929            self.dump_struct(ln, prototype)
930            return
931
932        self.output_declaration(self.entry.decl_type, prototype,
933                                entry=self.entry)
934
935    def dump_function(self, ln, prototype):
936        """
937        Stores a function of function macro inside self.entries array.
938        """
939
940        func_macro = False
941        return_type = ''
942        decl_type = 'function'
943
944        # Prefixes that would be removed
945        sub_prefixes = [
946            (r"^static +", "", 0),
947            (r"^extern +", "", 0),
948            (r"^asmlinkage +", "", 0),
949            (r"^inline +", "", 0),
950            (r"^__inline__ +", "", 0),
951            (r"^__inline +", "", 0),
952            (r"^__always_inline +", "", 0),
953            (r"^noinline +", "", 0),
954            (r"^__FORTIFY_INLINE +", "", 0),
955            (r"__init +", "", 0),
956            (r"__init_or_module +", "", 0),
957            (r"__deprecated +", "", 0),
958            (r"__flatten +", "", 0),
959            (r"__meminit +", "", 0),
960            (r"__must_check +", "", 0),
961            (r"__weak +", "", 0),
962            (r"__sched +", "", 0),
963            (r"_noprof", "", 0),
964            (r"__printf\s*\(\s*\d*\s*,\s*\d*\s*\) +", "", 0),
965            (r"__(?:re)?alloc_size\s*\(\s*\d+\s*(?:,\s*\d+\s*)?\) +", "", 0),
966            (r"__diagnose_as\s*\(\s*\S+\s*(?:,\s*\d+\s*)*\) +", "", 0),
967            (r"DECL_BUCKET_PARAMS\s*\(\s*(\S+)\s*,\s*(\S+)\s*\)", r"\1, \2", 0),
968            (r"__attribute_const__ +", "", 0),
969
970            # It seems that Python support for re.X is broken:
971            # At least for me (Python 3.13), this didn't work
972#            (r"""
973#              __attribute__\s*\(\(
974#                (?:
975#                    [\w\s]+          # attribute name
976#                    (?:\([^)]*\))?   # attribute arguments
977#                    \s*,?            # optional comma at the end
978#                )+
979#              \)\)\s+
980#             """, "", re.X),
981
982            # So, remove whitespaces and comments from it
983            (r"__attribute__\s*\(\((?:[\w\s]+(?:\([^)]*\))?\s*,?)+\)\)\s+", "", 0),
984        ]
985
986        for search, sub, flags in sub_prefixes:
987            prototype = KernRe(search, flags).sub(sub, prototype)
988
989        # Macros are a special case, as they change the prototype format
990        new_proto = KernRe(r"^#\s*define\s+").sub("", prototype)
991        if new_proto != prototype:
992            is_define_proto = True
993            prototype = new_proto
994        else:
995            is_define_proto = False
996
997        # Yes, this truly is vile.  We are looking for:
998        # 1. Return type (may be nothing if we're looking at a macro)
999        # 2. Function name
1000        # 3. Function parameters.
1001        #
1002        # All the while we have to watch out for function pointer parameters
1003        # (which IIRC is what the two sections are for), C types (these
1004        # regexps don't even start to express all the possibilities), and
1005        # so on.
1006        #
1007        # If you mess with these regexps, it's a good idea to check that
1008        # the following functions' documentation still comes out right:
1009        # - parport_register_device (function pointer parameters)
1010        # - atomic_set (macro)
1011        # - pci_match_device, __copy_to_user (long return type)
1012
1013        name = r'[a-zA-Z0-9_~:]+'
1014        prototype_end1 = r'[^\(]*'
1015        prototype_end2 = r'[^\{]*'
1016        prototype_end = fr'\(({prototype_end1}|{prototype_end2})\)'
1017
1018        # Besides compiling, Perl qr{[\w\s]+} works as a non-capturing group.
1019        # So, this needs to be mapped in Python with (?:...)? or (?:...)+
1020
1021        type1 = r'(?:[\w\s]+)?'
1022        type2 = r'(?:[\w\s]+\*+)+'
1023
1024        found = False
1025
1026        if is_define_proto:
1027            r = KernRe(r'^()(' + name + r')\s+')
1028
1029            if r.search(prototype):
1030                return_type = ''
1031                declaration_name = r.group(2)
1032                func_macro = True
1033
1034                found = True
1035
1036        if not found:
1037            patterns = [
1038                rf'^()({name})\s*{prototype_end}',
1039                rf'^({type1})\s+({name})\s*{prototype_end}',
1040                rf'^({type2})\s*({name})\s*{prototype_end}',
1041            ]
1042
1043            for p in patterns:
1044                r = KernRe(p)
1045
1046                if r.match(prototype):
1047
1048                    return_type = r.group(1)
1049                    declaration_name = r.group(2)
1050                    args = r.group(3)
1051
1052                    self.create_parameter_list(ln, decl_type, args, ',',
1053                                               declaration_name)
1054
1055                    found = True
1056                    break
1057        if not found:
1058            self.emit_msg(ln,
1059                          f"cannot understand function prototype: '{prototype}'")
1060            return
1061
1062        if self.entry.identifier != declaration_name:
1063            self.emit_msg(ln,
1064                          f"expecting prototype for {self.entry.identifier}(). Prototype was for {declaration_name}() instead")
1065            return
1066
1067        prms = " ".join(self.entry.parameterlist)
1068        self.check_sections(ln, declaration_name, "function",
1069                            self.entry.sectcheck, prms)
1070
1071        self.check_return_section(ln, declaration_name, return_type)
1072
1073        if 'typedef' in return_type:
1074            self.output_declaration(decl_type, declaration_name,
1075                                    function=declaration_name,
1076                                    typedef=True,
1077                                    functiontype=return_type,
1078                                    parameterlist=self.entry.parameterlist,
1079                                    parameterdescs=self.entry.parameterdescs,
1080                                    parametertypes=self.entry.parametertypes,
1081                                    parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
1082                                    sections=self.entry.sections,
1083                                    section_start_lines=self.entry.section_start_lines,
1084                                    purpose=self.entry.declaration_purpose,
1085                                    func_macro=func_macro)
1086        else:
1087            self.output_declaration(decl_type, declaration_name,
1088                                    function=declaration_name,
1089                                    typedef=False,
1090                                    functiontype=return_type,
1091                                    parameterlist=self.entry.parameterlist,
1092                                    parameterdescs=self.entry.parameterdescs,
1093                                    parametertypes=self.entry.parametertypes,
1094                                    parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
1095                                    sections=self.entry.sections,
1096                                    section_start_lines=self.entry.section_start_lines,
1097                                    purpose=self.entry.declaration_purpose,
1098                                    func_macro=func_macro)
1099
1100    def dump_typedef(self, ln, proto):
1101        """
1102        Stores a typedef inside self.entries array.
1103        """
1104
1105        typedef_type = r'((?:\s+[\w\*]+\b){0,7}\s+(?:\w+\b|\*+))\s*'
1106        typedef_ident = r'\*?\s*(\w\S+)\s*'
1107        typedef_args = r'\s*\((.*)\);'
1108
1109        typedef1 = KernRe(r'typedef' + typedef_type + r'\(' + typedef_ident + r'\)' + typedef_args)
1110        typedef2 = KernRe(r'typedef' + typedef_type + typedef_ident + typedef_args)
1111
1112        # Strip comments
1113        proto = KernRe(r'/\*.*?\*/', flags=re.S).sub('', proto)
1114
1115        # Parse function typedef prototypes
1116        for r in [typedef1, typedef2]:
1117            if not r.match(proto):
1118                continue
1119
1120            return_type = r.group(1).strip()
1121            declaration_name = r.group(2)
1122            args = r.group(3)
1123
1124            if self.entry.identifier != declaration_name:
1125                self.emit_msg(ln,
1126                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1127                return
1128
1129            decl_type = 'function'
1130            self.create_parameter_list(ln, decl_type, args, ',', declaration_name)
1131
1132            self.output_declaration(decl_type, declaration_name,
1133                                    function=declaration_name,
1134                                    typedef=True,
1135                                    functiontype=return_type,
1136                                    parameterlist=self.entry.parameterlist,
1137                                    parameterdescs=self.entry.parameterdescs,
1138                                    parametertypes=self.entry.parametertypes,
1139                                    parameterdesc_start_lines=self.entry.parameterdesc_start_lines,
1140                                    sections=self.entry.sections,
1141                                    section_start_lines=self.entry.section_start_lines,
1142                                    purpose=self.entry.declaration_purpose)
1143            return
1144
1145        # Handle nested parentheses or brackets
1146        r = KernRe(r'(\(*.\)\s*|\[*.\]\s*);$')
1147        while r.search(proto):
1148            proto = r.sub('', proto)
1149
1150        # Parse simple typedefs
1151        r = KernRe(r'typedef.*\s+(\w+)\s*;')
1152        if r.match(proto):
1153            declaration_name = r.group(1)
1154
1155            if self.entry.identifier != declaration_name:
1156                self.emit_msg(ln,
1157                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1158                return
1159
1160            self.output_declaration('typedef', declaration_name,
1161                                    typedef=declaration_name,
1162                                    sections=self.entry.sections,
1163                                    section_start_lines=self.entry.section_start_lines,
1164                                    purpose=self.entry.declaration_purpose)
1165            return
1166
1167        self.emit_msg(ln, "error: Cannot parse typedef!")
1168
1169    @staticmethod
1170    def process_export(function_set, line):
1171        """
1172        process EXPORT_SYMBOL* tags
1173
1174        This method doesn't use any variable from the class, so declare it
1175        with a staticmethod decorator.
1176        """
1177
1178        # We support documenting some exported symbols with different
1179        # names.  A horrible hack.
1180        suffixes = [ '_noprof' ]
1181
1182        # Note: it accepts only one EXPORT_SYMBOL* per line, as having
1183        # multiple export lines would violate Kernel coding style.
1184
1185        if export_symbol.search(line):
1186            symbol = export_symbol.group(2)
1187        elif export_symbol_ns.search(line):
1188            symbol = export_symbol_ns.group(2)
1189        else:
1190            return False
1191        #
1192        # Found an export, trim out any special suffixes
1193        #
1194        for suffix in suffixes:
1195            symbol = symbol.removesuffix(suffix)
1196        function_set.add(symbol)
1197        return True
1198
1199    def process_normal(self, ln, line):
1200        """
1201        STATE_NORMAL: looking for the /** to begin everything.
1202        """
1203
1204        if not doc_start.match(line):
1205            return
1206
1207        # start a new entry
1208        self.reset_state(ln)
1209
1210        # next line is always the function name
1211        self.state = state.NAME
1212
1213    def process_name(self, ln, line):
1214        """
1215        STATE_NAME: Looking for the "name - description" line
1216        """
1217        #
1218        # Check for a DOC: block and handle them specially.
1219        #
1220        if doc_block.search(line):
1221
1222            if not doc_block.group(1):
1223                self.entry.begin_section(ln, "Introduction")
1224            else:
1225                self.entry.begin_section(ln, doc_block.group(1))
1226
1227            self.entry.identifier = self.entry.section
1228            self.state = state.DOCBLOCK
1229        #
1230        # Otherwise we're looking for a normal kerneldoc declaration line.
1231        #
1232        elif doc_decl.search(line):
1233            self.entry.identifier = doc_decl.group(1)
1234
1235            # Test for data declaration
1236            if doc_begin_data.search(line):
1237                self.entry.decl_type = doc_begin_data.group(1)
1238                self.entry.identifier = doc_begin_data.group(2)
1239            #
1240            # Look for a function description
1241            #
1242            elif doc_begin_func.search(line):
1243                self.entry.identifier = doc_begin_func.group(1)
1244                self.entry.decl_type = "function"
1245            #
1246            # We struck out.
1247            #
1248            else:
1249                self.emit_msg(ln,
1250                              f"This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst\n{line}")
1251                self.state = state.NORMAL
1252                return
1253            #
1254            # OK, set up for a new kerneldoc entry.
1255            #
1256            self.state = state.BODY
1257            self.entry.identifier = self.entry.identifier.strip(" ")
1258            # if there's no @param blocks need to set up default section here
1259            self.entry.begin_section(ln + 1)
1260            #
1261            # Find the description portion, which *should* be there but
1262            # isn't always.
1263            # (We should be able to capture this from the previous parsing - someday)
1264            #
1265            r = KernRe("[-:](.*)")
1266            if r.search(line):
1267                self.entry.declaration_purpose = trim_whitespace(r.group(1))
1268                self.state = state.DECLARATION
1269            else:
1270                self.entry.declaration_purpose = ""
1271
1272            if not self.entry.declaration_purpose and self.config.wshort_desc:
1273                self.emit_msg(ln,
1274                              f"missing initial short description on line:\n{line}")
1275
1276            if not self.entry.identifier and self.entry.decl_type != "enum":
1277                self.emit_msg(ln,
1278                              f"wrong kernel-doc identifier on line:\n{line}")
1279                self.state = state.NORMAL
1280
1281            if self.config.verbose:
1282                self.emit_msg(ln,
1283                              f"Scanning doc for {self.entry.decl_type} {self.entry.identifier}",
1284                                  warning=False)
1285        #
1286        # Failed to find an identifier. Emit a warning
1287        #
1288        else:
1289            self.emit_msg(ln, f"Cannot find identifier on line:\n{line}")
1290
1291    #
1292    # Helper function to determine if a new section is being started.
1293    #
1294    def is_new_section(self, ln, line):
1295        if doc_sect.search(line):
1296            self.state = state.BODY
1297            #
1298            # Pick out the name of our new section, tweaking it if need be.
1299            #
1300            newsection = doc_sect.group(1)
1301            if newsection.lower() == 'description':
1302                newsection = 'Description'
1303            elif newsection.lower() == 'context':
1304                newsection = 'Context'
1305                self.state = state.SPECIAL_SECTION
1306            elif newsection.lower() in ["@return", "@returns",
1307                                        "return", "returns"]:
1308                newsection = "Return"
1309                self.state = state.SPECIAL_SECTION
1310            elif newsection[0] == '@':
1311                self.state = state.SPECIAL_SECTION
1312            #
1313            # Initialize the contents, and get the new section going.
1314            #
1315            newcontents = doc_sect.group(2)
1316            if not newcontents:
1317                newcontents = ""
1318            self.dump_section()
1319            self.entry.begin_section(ln, newsection)
1320            self.entry.leading_space = None
1321
1322            self.entry.add_text(newcontents.lstrip())
1323            return True
1324        return False
1325
1326    #
1327    # Helper function to detect (and effect) the end of a kerneldoc comment.
1328    #
1329    def is_comment_end(self, ln, line):
1330        if doc_end.search(line):
1331            self.dump_section()
1332
1333            # Look for doc_com + <text> + doc_end:
1334            r = KernRe(r'\s*\*\s*[a-zA-Z_0-9:\.]+\*/')
1335            if r.match(line):
1336                self.emit_msg(ln, f"suspicious ending line: {line}")
1337
1338            self.entry.prototype = ""
1339            self.entry.new_start_line = ln + 1
1340
1341            self.state = state.PROTO
1342            return True
1343        return False
1344
1345
1346    def process_decl(self, ln, line):
1347        """
1348        STATE_DECLARATION: We've seen the beginning of a declaration
1349        """
1350        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1351            return
1352        #
1353        # Look for anything with the " * " line beginning.
1354        #
1355        if doc_content.search(line):
1356            cont = doc_content.group(1)
1357            #
1358            # A blank line means that we have moved out of the declaration
1359            # part of the comment (without any "special section" parameter
1360            # descriptions).
1361            #
1362            if cont == "":
1363                self.state = state.BODY
1364            #
1365            # Otherwise we have more of the declaration section to soak up.
1366            #
1367            else:
1368                self.entry.declaration_purpose = \
1369                    trim_whitespace(self.entry.declaration_purpose + ' ' + cont)
1370        else:
1371            # Unknown line, ignore
1372            self.emit_msg(ln, f"bad line: {line}")
1373
1374
1375    def process_special(self, ln, line):
1376        """
1377        STATE_SPECIAL_SECTION: a section ending with a blank line
1378        """
1379        #
1380        # If we have hit a blank line (only the " * " marker), then this
1381        # section is done.
1382        #
1383        if KernRe(r"\s*\*\s*$").match(line):
1384            self.entry.begin_section(ln, dump = True)
1385            self.state = state.BODY
1386            return
1387        #
1388        # Not a blank line, look for the other ways to end the section.
1389        #
1390        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1391            return
1392        #
1393        # OK, we should have a continuation of the text for this section.
1394        #
1395        if doc_content.search(line):
1396            cont = doc_content.group(1)
1397            #
1398            # If the lines of text after the first in a special section have
1399            # leading white space, we need to trim it out or Sphinx will get
1400            # confused.  For the second line (the None case), see what we
1401            # find there and remember it.
1402            #
1403            if self.entry.leading_space is None:
1404                r = KernRe(r'^(\s+)')
1405                if r.match(cont):
1406                    self.entry.leading_space = len(r.group(1))
1407                else:
1408                    self.entry.leading_space = 0
1409            #
1410            # Otherwise, before trimming any leading chars, be *sure*
1411            # that they are white space.  We should maybe warn if this
1412            # isn't the case.
1413            #
1414            for i in range(0, self.entry.leading_space):
1415                if cont[i] != " ":
1416                    self.entry.leading_space = i
1417                    break
1418            #
1419            # Add the trimmed result to the section and we're done.
1420            #
1421            self.entry.add_text(cont[self.entry.leading_space:])
1422        else:
1423            # Unknown line, ignore
1424            self.emit_msg(ln, f"bad line: {line}")
1425
1426    def process_body(self, ln, line):
1427        """
1428        STATE_BODY: the bulk of a kerneldoc comment.
1429        """
1430        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1431            return
1432
1433        if doc_content.search(line):
1434            cont = doc_content.group(1)
1435            self.entry.add_text(cont)
1436        else:
1437            # Unknown line, ignore
1438            self.emit_msg(ln, f"bad line: {line}")
1439
1440    def process_inline_name(self, ln, line):
1441        """STATE_INLINE_NAME: beginning of docbook comments within a prototype."""
1442
1443        if doc_inline_sect.search(line):
1444            self.entry.begin_section(ln, doc_inline_sect.group(1))
1445            self.entry.add_text(doc_inline_sect.group(2).lstrip())
1446            self.state = state.INLINE_TEXT
1447        elif doc_inline_end.search(line):
1448            self.dump_section()
1449            self.state = state.PROTO
1450        elif doc_content.search(line):
1451            self.emit_msg(ln, f"Incorrect use of kernel-doc format: {line}")
1452            self.state = state.PROTO
1453        # else ... ??
1454
1455    def process_inline_text(self, ln, line):
1456        """STATE_INLINE_TEXT: docbook comments within a prototype."""
1457
1458        if doc_inline_end.search(line):
1459            self.dump_section()
1460            self.state = state.PROTO
1461        elif doc_content.search(line):
1462            self.entry.add_text(doc_content.group(1))
1463        # else ... ??
1464
1465    def syscall_munge(self, ln, proto):         # pylint: disable=W0613
1466        """
1467        Handle syscall definitions
1468        """
1469
1470        is_void = False
1471
1472        # Strip newlines/CR's
1473        proto = re.sub(r'[\r\n]+', ' ', proto)
1474
1475        # Check if it's a SYSCALL_DEFINE0
1476        if 'SYSCALL_DEFINE0' in proto:
1477            is_void = True
1478
1479        # Replace SYSCALL_DEFINE with correct return type & function name
1480        proto = KernRe(r'SYSCALL_DEFINE.*\(').sub('long sys_', proto)
1481
1482        r = KernRe(r'long\s+(sys_.*?),')
1483        if r.search(proto):
1484            proto = KernRe(',').sub('(', proto, count=1)
1485        elif is_void:
1486            proto = KernRe(r'\)').sub('(void)', proto, count=1)
1487
1488        # Now delete all of the odd-numbered commas in the proto
1489        # so that argument types & names don't have a comma between them
1490        count = 0
1491        length = len(proto)
1492
1493        if is_void:
1494            length = 0  # skip the loop if is_void
1495
1496        for ix in range(length):
1497            if proto[ix] == ',':
1498                count += 1
1499                if count % 2 == 1:
1500                    proto = proto[:ix] + ' ' + proto[ix + 1:]
1501
1502        return proto
1503
1504    def tracepoint_munge(self, ln, proto):
1505        """
1506        Handle tracepoint definitions
1507        """
1508
1509        tracepointname = None
1510        tracepointargs = None
1511
1512        # Match tracepoint name based on different patterns
1513        r = KernRe(r'TRACE_EVENT\((.*?),')
1514        if r.search(proto):
1515            tracepointname = r.group(1)
1516
1517        r = KernRe(r'DEFINE_SINGLE_EVENT\((.*?),')
1518        if r.search(proto):
1519            tracepointname = r.group(1)
1520
1521        r = KernRe(r'DEFINE_EVENT\((.*?),(.*?),')
1522        if r.search(proto):
1523            tracepointname = r.group(2)
1524
1525        if tracepointname:
1526            tracepointname = tracepointname.lstrip()
1527
1528        r = KernRe(r'TP_PROTO\((.*?)\)')
1529        if r.search(proto):
1530            tracepointargs = r.group(1)
1531
1532        if not tracepointname or not tracepointargs:
1533            self.emit_msg(ln,
1534                          f"Unrecognized tracepoint format:\n{proto}\n")
1535        else:
1536            proto = f"static inline void trace_{tracepointname}({tracepointargs})"
1537            self.entry.identifier = f"trace_{self.entry.identifier}"
1538
1539        return proto
1540
1541    def process_proto_function(self, ln, line):
1542        """Ancillary routine to process a function prototype"""
1543
1544        # strip C99-style comments to end of line
1545        line = KernRe(r"\/\/.*$", re.S).sub('', line)
1546        #
1547        # Soak up the line's worth of prototype text, stopping at { or ; if present.
1548        #
1549        if KernRe(r'\s*#\s*define').match(line):
1550            self.entry.prototype = line
1551        elif not line.startswith('#'):   # skip other preprocessor stuff
1552            r = KernRe(r'([^\{]*)')
1553            if r.match(line):
1554                self.entry.prototype += r.group(1) + " "
1555        #
1556        # If we now have the whole prototype, clean it up and declare victory.
1557        #
1558        if '{' in line or ';' in line or KernRe(r'\s*#\s*define').match(line):
1559            # strip comments and surrounding spaces
1560            self.entry.prototype = KernRe(r'/\*.*\*/').sub('', self.entry.prototype).strip()
1561            #
1562            # Handle self.entry.prototypes for function pointers like:
1563            #       int (*pcs_config)(struct foo)
1564            # by turning it into
1565            #	    int pcs_config(struct foo)
1566            #
1567            r = KernRe(r'^(\S+\s+)\(\s*\*(\S+)\)')
1568            self.entry.prototype = r.sub(r'\1\2', self.entry.prototype)
1569            #
1570            # Handle special declaration syntaxes
1571            #
1572            if 'SYSCALL_DEFINE' in self.entry.prototype:
1573                self.entry.prototype = self.syscall_munge(ln,
1574                                                          self.entry.prototype)
1575            else:
1576                r = KernRe(r'TRACE_EVENT|DEFINE_EVENT|DEFINE_SINGLE_EVENT')
1577                if r.search(self.entry.prototype):
1578                    self.entry.prototype = self.tracepoint_munge(ln,
1579                                                                 self.entry.prototype)
1580            #
1581            # ... and we're done
1582            #
1583            self.dump_function(ln, self.entry.prototype)
1584            self.reset_state(ln)
1585
1586    def process_proto_type(self, ln, line):
1587        """Ancillary routine to process a type"""
1588
1589        # Strip C99-style comments and surrounding whitespace
1590        line = KernRe(r"//.*$", re.S).sub('', line).strip()
1591        if not line:
1592            return # nothing to see here
1593
1594        # To distinguish preprocessor directive from regular declaration later.
1595        if line.startswith('#'):
1596            line += ";"
1597        #
1598        # Split the declaration on any of { } or ;, and accumulate pieces
1599        # until we hit a semicolon while not inside {brackets}
1600        #
1601        r = KernRe(r'(.*?)([{};])')
1602        for chunk in r.split(line):
1603            if chunk:  # Ignore empty matches
1604                self.entry.prototype += chunk
1605                #
1606                # This cries out for a match statement ... someday after we can
1607                # drop Python 3.9 ...
1608                #
1609                if chunk == '{':
1610                    self.entry.brcount += 1
1611                elif chunk == '}':
1612                    self.entry.brcount -= 1
1613                elif chunk == ';' and self.entry.brcount <= 0:
1614                    self.dump_declaration(ln, self.entry.prototype)
1615                    self.reset_state(ln)
1616                    return
1617        #
1618        # We hit the end of the line while still in the declaration; put
1619        # in a space to represent the newline.
1620        #
1621        self.entry.prototype += ' '
1622
1623    def process_proto(self, ln, line):
1624        """STATE_PROTO: reading a function/whatever prototype."""
1625
1626        if doc_inline_oneline.search(line):
1627            self.entry.begin_section(ln, doc_inline_oneline.group(1))
1628            self.entry.add_text(doc_inline_oneline.group(2))
1629            self.dump_section()
1630
1631        elif doc_inline_start.search(line):
1632            self.state = state.INLINE_NAME
1633
1634        elif self.entry.decl_type == 'function':
1635            self.process_proto_function(ln, line)
1636
1637        else:
1638            self.process_proto_type(ln, line)
1639
1640    def process_docblock(self, ln, line):
1641        """STATE_DOCBLOCK: within a DOC: block."""
1642
1643        if doc_end.search(line):
1644            self.dump_section()
1645            self.output_declaration("doc", self.entry.identifier,
1646                                    sections=self.entry.sections,
1647                                    section_start_lines=self.entry.section_start_lines)
1648            self.reset_state(ln)
1649
1650        elif doc_content.search(line):
1651            self.entry.add_text(doc_content.group(1))
1652
1653    def parse_export(self):
1654        """
1655        Parses EXPORT_SYMBOL* macros from a single Kernel source file.
1656        """
1657
1658        export_table = set()
1659
1660        try:
1661            with open(self.fname, "r", encoding="utf8",
1662                      errors="backslashreplace") as fp:
1663
1664                for line in fp:
1665                    self.process_export(export_table, line)
1666
1667        except IOError:
1668            return None
1669
1670        return export_table
1671
1672    #
1673    # The state/action table telling us which function to invoke in
1674    # each state.
1675    #
1676    state_actions = {
1677        state.NORMAL:			process_normal,
1678        state.NAME:			process_name,
1679        state.BODY:			process_body,
1680        state.DECLARATION:		process_decl,
1681        state.SPECIAL_SECTION:		process_special,
1682        state.INLINE_NAME:		process_inline_name,
1683        state.INLINE_TEXT:		process_inline_text,
1684        state.PROTO:			process_proto,
1685        state.DOCBLOCK:			process_docblock,
1686        }
1687
1688    def parse_kdoc(self):
1689        """
1690        Open and process each line of a C source file.
1691        The parsing is controlled via a state machine, and the line is passed
1692        to a different process function depending on the state. The process
1693        function may update the state as needed.
1694
1695        Besides parsing kernel-doc tags, it also parses export symbols.
1696        """
1697
1698        prev = ""
1699        prev_ln = None
1700        export_table = set()
1701
1702        try:
1703            with open(self.fname, "r", encoding="utf8",
1704                      errors="backslashreplace") as fp:
1705                for ln, line in enumerate(fp):
1706
1707                    line = line.expandtabs().strip("\n")
1708
1709                    # Group continuation lines on prototypes
1710                    if self.state == state.PROTO:
1711                        if line.endswith("\\"):
1712                            prev += line.rstrip("\\")
1713                            if not prev_ln:
1714                                prev_ln = ln
1715                            continue
1716
1717                        if prev:
1718                            ln = prev_ln
1719                            line = prev + line
1720                            prev = ""
1721                            prev_ln = None
1722
1723                    self.config.log.debug("%d %s: %s",
1724                                          ln, state.name[self.state],
1725                                          line)
1726
1727                    # This is an optimization over the original script.
1728                    # There, when export_file was used for the same file,
1729                    # it was read twice. Here, we use the already-existing
1730                    # loop to parse exported symbols as well.
1731                    #
1732                    if (self.state != state.NORMAL) or \
1733                       not self.process_export(export_table, line):
1734                        # Hand this line to the appropriate state handler
1735                        self.state_actions[self.state](self, ln, line)
1736
1737        except OSError:
1738            self.config.log.error(f"Error: Cannot open file {self.fname}")
1739
1740        return export_table, self.entries
1741