xref: /linux/scripts/lib/kdoc/kdoc_parser.py (revision f587722aa5c9e620fd8692054f0159da60052697)
1#!/usr/bin/env python3
2# SPDX-License-Identifier: GPL-2.0
3# Copyright(c) 2025: Mauro Carvalho Chehab <mchehab@kernel.org>.
4#
5# pylint: disable=C0301,C0302,R0904,R0912,R0913,R0914,R0915,R0917,R1702
6
7"""
8kdoc_parser
9===========
10
11Read a C language source or header FILE and extract embedded
12documentation comments
13"""
14
15import sys
16import re
17from pprint import pformat
18
19from kdoc_re import NestedMatch, KernRe
20from kdoc_item import KdocItem
21
22#
23# Regular expressions used to parse kernel-doc markups at KernelDoc class.
24#
25# Let's declare them in lowercase outside any class to make easier to
26# convert from the python script.
27#
28# As those are evaluated at the beginning, no need to cache them
29#
30
31# Allow whitespace at end of comment start.
32doc_start = KernRe(r'^/\*\*\s*$', cache=False)
33
34doc_end = KernRe(r'\*/', cache=False)
35doc_com = KernRe(r'\s*\*\s*', cache=False)
36doc_com_body = KernRe(r'\s*\* ?', cache=False)
37doc_decl = doc_com + KernRe(r'(\w+)', cache=False)
38
39# @params and a strictly limited set of supported section names
40# Specifically:
41#   Match @word:
42#         @...:
43#         @{section-name}:
44# while trying to not match literal block starts like "example::"
45#
46known_section_names = 'description|context|returns?|notes?|examples?'
47known_sections = KernRe(known_section_names, flags = re.I)
48doc_sect = doc_com + \
49    KernRe(r'\s*(\@[.\w]+|\@\.\.\.|' + known_section_names + r')\s*:([^:].*)?$',
50           flags=re.I, cache=False)
51
52doc_content = doc_com_body + KernRe(r'(.*)', cache=False)
53doc_inline_start = KernRe(r'^\s*/\*\*\s*$', cache=False)
54doc_inline_sect = KernRe(r'\s*\*\s*(@\s*[\w][\w\.]*\s*):(.*)', cache=False)
55doc_inline_end = KernRe(r'^\s*\*/\s*$', cache=False)
56doc_inline_oneline = KernRe(r'^\s*/\*\*\s*(@[\w\s]+):\s*(.*)\s*\*/\s*$', cache=False)
57attribute = KernRe(r"__attribute__\s*\(\([a-z0-9,_\*\s\(\)]*\)\)",
58               flags=re.I | re.S, cache=False)
59
60export_symbol = KernRe(r'^\s*EXPORT_SYMBOL(_GPL)?\s*\(\s*(\w+)\s*\)\s*', cache=False)
61export_symbol_ns = KernRe(r'^\s*EXPORT_SYMBOL_NS(_GPL)?\s*\(\s*(\w+)\s*,\s*"\S+"\)\s*', cache=False)
62
63type_param = KernRe(r"\@(\w*((\.\w+)|(->\w+))*(\.\.\.)?)", cache=False)
64
65#
66# Tests for the beginning of a kerneldoc block in its various forms.
67#
68doc_block = doc_com + KernRe(r'DOC:\s*(.*)?', cache=False)
69doc_begin_data = KernRe(r"^\s*\*?\s*(struct|union|enum|typedef)\b\s*(\w*)", cache = False)
70doc_begin_func = KernRe(str(doc_com) +			# initial " * '
71                        r"(?:\w+\s*\*\s*)?" + 		# type (not captured)
72                        r'(?:define\s+)?' + 		# possible "define" (not captured)
73                        r'(\w+)\s*(?:\(\w*\))?\s*' +	# name and optional "(...)"
74                        r'(?:[-:].*)?$',		# description (not captured)
75                        cache = False)
76
77#
78# A little helper to get rid of excess white space
79#
80multi_space = KernRe(r'\s\s+')
81def trim_whitespace(s):
82    return multi_space.sub(' ', s.strip())
83
84class state:
85    """
86    State machine enums
87    """
88
89    # Parser states
90    NORMAL        = 0        # normal code
91    NAME          = 1        # looking for function name
92    DECLARATION   = 2        # We have seen a declaration which might not be done
93    BODY          = 3        # the body of the comment
94    SPECIAL_SECTION = 4      # doc section ending with a blank line
95    PROTO         = 5        # scanning prototype
96    DOCBLOCK      = 6        # documentation block
97    INLINE_NAME   = 7        # gathering doc outside main block
98    INLINE_TEXT   = 8	     # reading the body of inline docs
99
100    name = [
101        "NORMAL",
102        "NAME",
103        "DECLARATION",
104        "BODY",
105        "SPECIAL_SECTION",
106        "PROTO",
107        "DOCBLOCK",
108        "INLINE_NAME",
109        "INLINE_TEXT",
110    ]
111
112
113SECTION_DEFAULT = "Description"  # default section
114
115class KernelEntry:
116
117    def __init__(self, config, ln):
118        self.config = config
119
120        self._contents = []
121        self.prototype = ""
122
123        self.warnings = []
124
125        self.parameterlist = []
126        self.parameterdescs = {}
127        self.parametertypes = {}
128        self.parameterdesc_start_lines = {}
129
130        self.section_start_lines = {}
131        self.sections = {}
132
133        self.anon_struct_union = False
134
135        self.leading_space = None
136
137        # State flags
138        self.brcount = 0
139        self.declaration_start_line = ln + 1
140
141    #
142    # Management of section contents
143    #
144    def add_text(self, text):
145        self._contents.append(text)
146
147    def contents(self):
148        return '\n'.join(self._contents) + '\n'
149
150    # TODO: rename to emit_message after removal of kernel-doc.pl
151    def emit_msg(self, log_msg, warning=True):
152        """Emit a message"""
153
154        if not warning:
155            self.config.log.info(log_msg)
156            return
157
158        # Delegate warning output to output logic, as this way it
159        # will report warnings/info only for symbols that are output
160
161        self.warnings.append(log_msg)
162        return
163
164    #
165    # Begin a new section.
166    #
167    def begin_section(self, line_no, title = SECTION_DEFAULT, dump = False):
168        if dump:
169            self.dump_section(start_new = True)
170        self.section = title
171        self.new_start_line = line_no
172
173    def dump_section(self, start_new=True):
174        """
175        Dumps section contents to arrays/hashes intended for that purpose.
176        """
177        #
178        # If we have accumulated no contents in the default ("description")
179        # section, don't bother.
180        #
181        if self.section == SECTION_DEFAULT and not self._contents:
182            return
183        name = self.section
184        contents = self.contents()
185
186        if type_param.match(name):
187            name = type_param.group(1)
188
189            self.parameterdescs[name] = contents
190            self.parameterdesc_start_lines[name] = self.new_start_line
191
192            self.new_start_line = 0
193
194        else:
195            if name in self.sections and self.sections[name] != "":
196                # Only warn on user-specified duplicate section names
197                if name != SECTION_DEFAULT:
198                    self.emit_msg(self.new_start_line,
199                                  f"duplicate section name '{name}'\n")
200                # Treat as a new paragraph - add a blank line
201                self.sections[name] += '\n' + contents
202            else:
203                self.sections[name] = contents
204                self.section_start_lines[name] = self.new_start_line
205                self.new_start_line = 0
206
207#        self.config.log.debug("Section: %s : %s", name, pformat(vars(self)))
208
209        if start_new:
210            self.section = SECTION_DEFAULT
211            self._contents = []
212
213
214class KernelDoc:
215    """
216    Read a C language source or header FILE and extract embedded
217    documentation comments.
218    """
219
220    # Section names
221
222    section_context = "Context"
223    section_return = "Return"
224
225    undescribed = "-- undescribed --"
226
227    def __init__(self, config, fname):
228        """Initialize internal variables"""
229
230        self.fname = fname
231        self.config = config
232
233        # Initial state for the state machines
234        self.state = state.NORMAL
235
236        # Store entry currently being processed
237        self.entry = None
238
239        # Place all potential outputs into an array
240        self.entries = []
241
242        #
243        # We need Python 3.7 for its "dicts remember the insertion
244        # order" guarantee
245        #
246        if sys.version_info.major == 3 and sys.version_info.minor < 7:
247            self.emit_msg(0,
248                          'Python 3.7 or later is required for correct results')
249
250    def emit_msg(self, ln, msg, warning=True):
251        """Emit a message"""
252
253        log_msg = f"{self.fname}:{ln} {msg}"
254
255        if self.entry:
256            self.entry.emit_msg(log_msg, warning)
257            return
258
259        if warning:
260            self.config.log.warning(log_msg)
261        else:
262            self.config.log.info(log_msg)
263
264    def dump_section(self, start_new=True):
265        """
266        Dumps section contents to arrays/hashes intended for that purpose.
267        """
268
269        if self.entry:
270            self.entry.dump_section(start_new)
271
272    # TODO: rename it to store_declaration after removal of kernel-doc.pl
273    def output_declaration(self, dtype, name, **args):
274        """
275        Stores the entry into an entry array.
276
277        The actual output and output filters will be handled elsewhere
278        """
279
280        item = KdocItem(name, dtype, self.entry.declaration_start_line, **args)
281        item.warnings = self.entry.warnings
282
283        # Drop empty sections
284        # TODO: improve empty sections logic to emit warnings
285        sections = self.entry.sections
286        for section in ["Description", "Return"]:
287            if section in sections and not sections[section].rstrip():
288                del sections[section]
289        item.set_sections(sections, self.entry.section_start_lines)
290        item.set_params(self.entry.parameterlist, self.entry.parameterdescs,
291                        self.entry.parametertypes,
292                        self.entry.parameterdesc_start_lines)
293        self.entries.append(item)
294
295        self.config.log.debug("Output: %s:%s = %s", dtype, name, pformat(args))
296
297    def reset_state(self, ln):
298        """
299        Ancillary routine to create a new entry. It initializes all
300        variables used by the state machine.
301        """
302
303        self.entry = KernelEntry(self.config, ln)
304
305        # State flags
306        self.state = state.NORMAL
307
308    def push_parameter(self, ln, decl_type, param, dtype,
309                       org_arg, declaration_name):
310        """
311        Store parameters and their descriptions at self.entry.
312        """
313
314        if self.entry.anon_struct_union and dtype == "" and param == "}":
315            return  # Ignore the ending }; from anonymous struct/union
316
317        self.entry.anon_struct_union = False
318
319        param = KernRe(r'[\[\)].*').sub('', param, count=1)
320
321        if dtype == "" and param.endswith("..."):
322            if KernRe(r'\w\.\.\.$').search(param):
323                # For named variable parameters of the form `x...`,
324                # remove the dots
325                param = param[:-3]
326            else:
327                # Handles unnamed variable parameters
328                param = "..."
329
330            if param not in self.entry.parameterdescs or \
331                not self.entry.parameterdescs[param]:
332
333                self.entry.parameterdescs[param] = "variable arguments"
334
335        elif dtype == "" and (not param or param == "void"):
336            param = "void"
337            self.entry.parameterdescs[param] = "no arguments"
338
339        elif dtype == "" and param in ["struct", "union"]:
340            # Handle unnamed (anonymous) union or struct
341            dtype = param
342            param = "{unnamed_" + param + "}"
343            self.entry.parameterdescs[param] = "anonymous\n"
344            self.entry.anon_struct_union = True
345
346        # Handle cache group enforcing variables: they do not need
347        # to be described in header files
348        elif "__cacheline_group" in param:
349            # Ignore __cacheline_group_begin and __cacheline_group_end
350            return
351
352        # Warn if parameter has no description
353        # (but ignore ones starting with # as these are not parameters
354        # but inline preprocessor statements)
355        if param not in self.entry.parameterdescs and not param.startswith("#"):
356            self.entry.parameterdescs[param] = self.undescribed
357
358            if "." not in param:
359                if decl_type == 'function':
360                    dname = f"{decl_type} parameter"
361                else:
362                    dname = f"{decl_type} member"
363
364                self.emit_msg(ln,
365                              f"{dname} '{param}' not described in '{declaration_name}'")
366
367        # Strip spaces from param so that it is one continuous string on
368        # parameterlist. This fixes a problem where check_sections()
369        # cannot find a parameter like "addr[6 + 2]" because it actually
370        # appears as "addr[6", "+", "2]" on the parameter list.
371        # However, it's better to maintain the param string unchanged for
372        # output, so just weaken the string compare in check_sections()
373        # to ignore "[blah" in a parameter string.
374
375        self.entry.parameterlist.append(param)
376        org_arg = KernRe(r'\s\s+').sub(' ', org_arg)
377        self.entry.parametertypes[param] = org_arg
378
379
380    def create_parameter_list(self, ln, decl_type, args,
381                              splitter, declaration_name):
382        """
383        Creates a list of parameters, storing them at self.entry.
384        """
385
386        # temporarily replace all commas inside function pointer definition
387        arg_expr = KernRe(r'(\([^\),]+),')
388        while arg_expr.search(args):
389            args = arg_expr.sub(r"\1#", args)
390
391        for arg in args.split(splitter):
392            # Strip comments
393            arg = KernRe(r'\/\*.*\*\/').sub('', arg)
394
395            # Ignore argument attributes
396            arg = KernRe(r'\sPOS0?\s').sub(' ', arg)
397
398            # Strip leading/trailing spaces
399            arg = arg.strip()
400            arg = KernRe(r'\s+').sub(' ', arg, count=1)
401
402            if arg.startswith('#'):
403                # Treat preprocessor directive as a typeless variable just to fill
404                # corresponding data structures "correctly". Catch it later in
405                # output_* subs.
406
407                # Treat preprocessor directive as a typeless variable
408                self.push_parameter(ln, decl_type, arg, "",
409                                    "", declaration_name)
410
411            elif KernRe(r'\(.+\)\s*\(').search(arg):
412                # Pointer-to-function
413
414                arg = arg.replace('#', ',')
415
416                r = KernRe(r'[^\(]+\(\*?\s*([\w\[\]\.]*)\s*\)')
417                if r.match(arg):
418                    param = r.group(1)
419                else:
420                    self.emit_msg(ln, f"Invalid param: {arg}")
421                    param = arg
422
423                dtype = KernRe(r'([^\(]+\(\*?)\s*' + re.escape(param)).sub(r'\1', arg)
424                self.push_parameter(ln, decl_type, param, dtype,
425                                    arg, declaration_name)
426
427            elif KernRe(r'\(.+\)\s*\[').search(arg):
428                # Array-of-pointers
429
430                arg = arg.replace('#', ',')
431                r = KernRe(r'[^\(]+\(\s*\*\s*([\w\[\]\.]*?)\s*(\s*\[\s*[\w]+\s*\]\s*)*\)')
432                if r.match(arg):
433                    param = r.group(1)
434                else:
435                    self.emit_msg(ln, f"Invalid param: {arg}")
436                    param = arg
437
438                dtype = KernRe(r'([^\(]+\(\*?)\s*' + re.escape(param)).sub(r'\1', arg)
439
440                self.push_parameter(ln, decl_type, param, dtype,
441                                    arg, declaration_name)
442
443            elif arg:
444                arg = KernRe(r'\s*:\s*').sub(":", arg)
445                arg = KernRe(r'\s*\[').sub('[', arg)
446
447                args = KernRe(r'\s*,\s*').split(arg)
448                if args[0] and '*' in args[0]:
449                    args[0] = re.sub(r'(\*+)\s*', r' \1', args[0])
450
451                first_arg = []
452                r = KernRe(r'^(.*\s+)(.*?\[.*\].*)$')
453                if args[0] and r.match(args[0]):
454                    args.pop(0)
455                    first_arg.extend(r.group(1))
456                    first_arg.append(r.group(2))
457                else:
458                    first_arg = KernRe(r'\s+').split(args.pop(0))
459
460                args.insert(0, first_arg.pop())
461                dtype = ' '.join(first_arg)
462
463                for param in args:
464                    if KernRe(r'^(\*+)\s*(.*)').match(param):
465                        r = KernRe(r'^(\*+)\s*(.*)')
466                        if not r.match(param):
467                            self.emit_msg(ln, f"Invalid param: {param}")
468                            continue
469
470                        param = r.group(1)
471
472                        self.push_parameter(ln, decl_type, r.group(2),
473                                            f"{dtype} {r.group(1)}",
474                                            arg, declaration_name)
475
476                    elif KernRe(r'(.*?):(\w+)').search(param):
477                        r = KernRe(r'(.*?):(\w+)')
478                        if not r.match(param):
479                            self.emit_msg(ln, f"Invalid param: {param}")
480                            continue
481
482                        if dtype != "":  # Skip unnamed bit-fields
483                            self.push_parameter(ln, decl_type, r.group(1),
484                                                f"{dtype}:{r.group(2)}",
485                                                arg, declaration_name)
486                    else:
487                        self.push_parameter(ln, decl_type, param, dtype,
488                                            arg, declaration_name)
489
490    def check_sections(self, ln, decl_name, decl_type):
491        """
492        Check for errors inside sections, emitting warnings if not found
493        parameters are described.
494        """
495        for section in self.entry.sections:
496            if section not in self.entry.parameterlist and \
497               not known_sections.search(section):
498                if decl_type == 'function':
499                    dname = f"{decl_type} parameter"
500                else:
501                    dname = f"{decl_type} member"
502                self.emit_msg(ln,
503                              f"Excess {dname} '{section}' description in '{decl_name}'")
504
505    def check_return_section(self, ln, declaration_name, return_type):
506        """
507        If the function doesn't return void, warns about the lack of a
508        return description.
509        """
510
511        if not self.config.wreturn:
512            return
513
514        # Ignore an empty return type (It's a macro)
515        # Ignore functions with a "void" return type (but not "void *")
516        if not return_type or KernRe(r'void\s*\w*\s*$').search(return_type):
517            return
518
519        if not self.entry.sections.get("Return", None):
520            self.emit_msg(ln,
521                          f"No description found for return value of '{declaration_name}'")
522
523    def dump_struct(self, ln, proto):
524        """
525        Store an entry for an struct or union
526        """
527
528        type_pattern = r'(struct|union)'
529
530        qualifiers = [
531            "__attribute__",
532            "__packed",
533            "__aligned",
534            "____cacheline_aligned_in_smp",
535            "____cacheline_aligned",
536        ]
537
538        definition_body = r'\{(.*)\}\s*' + "(?:" + '|'.join(qualifiers) + ")?"
539        struct_members = KernRe(type_pattern + r'([^\{\};]+)(\{)([^\{\}]*)(\})([^\{\}\;]*)(\;)')
540
541        # Extract struct/union definition
542        members = None
543        declaration_name = None
544        decl_type = None
545
546        r = KernRe(type_pattern + r'\s+(\w+)\s*' + definition_body)
547        if r.search(proto):
548            decl_type = r.group(1)
549            declaration_name = r.group(2)
550            members = r.group(3)
551        else:
552            r = KernRe(r'typedef\s+' + type_pattern + r'\s*' + definition_body + r'\s*(\w+)\s*;')
553
554            if r.search(proto):
555                decl_type = r.group(1)
556                declaration_name = r.group(3)
557                members = r.group(2)
558
559        if not members:
560            self.emit_msg(ln, f"{proto} error: Cannot parse struct or union!")
561            return
562
563        if self.entry.identifier != declaration_name:
564            self.emit_msg(ln,
565                          f"expecting prototype for {decl_type} {self.entry.identifier}. Prototype was for {decl_type} {declaration_name} instead\n")
566            return
567
568        args_pattern = r'([^,)]+)'
569
570        sub_prefixes = [
571            (KernRe(r'\/\*\s*private:.*?\/\*\s*public:.*?\*\/', re.S | re.I), ''),
572            (KernRe(r'\/\*\s*private:.*', re.S | re.I), ''),
573
574            # Strip comments
575            (KernRe(r'\/\*.*?\*\/', re.S), ''),
576
577            # Strip attributes
578            (attribute, ' '),
579            (KernRe(r'\s*__aligned\s*\([^;]*\)', re.S), ' '),
580            (KernRe(r'\s*__counted_by\s*\([^;]*\)', re.S), ' '),
581            (KernRe(r'\s*__counted_by_(le|be)\s*\([^;]*\)', re.S), ' '),
582            (KernRe(r'\s*__packed\s*', re.S), ' '),
583            (KernRe(r'\s*CRYPTO_MINALIGN_ATTR', re.S), ' '),
584            (KernRe(r'\s*____cacheline_aligned_in_smp', re.S), ' '),
585            (KernRe(r'\s*____cacheline_aligned', re.S), ' '),
586
587            # Unwrap struct_group macros based on this definition:
588            # __struct_group(TAG, NAME, ATTRS, MEMBERS...)
589            # which has variants like: struct_group(NAME, MEMBERS...)
590            # Only MEMBERS arguments require documentation.
591            #
592            # Parsing them happens on two steps:
593            #
594            # 1. drop struct group arguments that aren't at MEMBERS,
595            #    storing them as STRUCT_GROUP(MEMBERS)
596            #
597            # 2. remove STRUCT_GROUP() ancillary macro.
598            #
599            # The original logic used to remove STRUCT_GROUP() using an
600            # advanced regex:
601            #
602            #   \bSTRUCT_GROUP(\(((?:(?>[^)(]+)|(?1))*)\))[^;]*;
603            #
604            # with two patterns that are incompatible with
605            # Python re module, as it has:
606            #
607            #   - a recursive pattern: (?1)
608            #   - an atomic grouping: (?>...)
609            #
610            # I tried a simpler version: but it didn't work either:
611            #   \bSTRUCT_GROUP\(([^\)]+)\)[^;]*;
612            #
613            # As it doesn't properly match the end parenthesis on some cases.
614            #
615            # So, a better solution was crafted: there's now a NestedMatch
616            # class that ensures that delimiters after a search are properly
617            # matched. So, the implementation to drop STRUCT_GROUP() will be
618            # handled in separate.
619
620            (KernRe(r'\bstruct_group\s*\(([^,]*,)', re.S), r'STRUCT_GROUP('),
621            (KernRe(r'\bstruct_group_attr\s*\(([^,]*,){2}', re.S), r'STRUCT_GROUP('),
622            (KernRe(r'\bstruct_group_tagged\s*\(([^,]*),([^,]*),', re.S), r'struct \1 \2; STRUCT_GROUP('),
623            (KernRe(r'\b__struct_group\s*\(([^,]*,){3}', re.S), r'STRUCT_GROUP('),
624
625            # Replace macros
626            #
627            # TODO: use NestedMatch for FOO($1, $2, ...) matches
628            #
629            # it is better to also move those to the NestedMatch logic,
630            # to ensure that parenthesis will be properly matched.
631
632            (KernRe(r'__ETHTOOL_DECLARE_LINK_MODE_MASK\s*\(([^\)]+)\)', re.S), r'DECLARE_BITMAP(\1, __ETHTOOL_LINK_MODE_MASK_NBITS)'),
633            (KernRe(r'DECLARE_PHY_INTERFACE_MASK\s*\(([^\)]+)\)', re.S), r'DECLARE_BITMAP(\1, PHY_INTERFACE_MODE_MAX)'),
634            (KernRe(r'DECLARE_BITMAP\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'unsigned long \1[BITS_TO_LONGS(\2)]'),
635            (KernRe(r'DECLARE_HASHTABLE\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'unsigned long \1[1 << ((\2) - 1)]'),
636            (KernRe(r'DECLARE_KFIFO\s*\(' + args_pattern + r',\s*' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\2 *\1'),
637            (KernRe(r'DECLARE_KFIFO_PTR\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\2 *\1'),
638            (KernRe(r'(?:__)?DECLARE_FLEX_ARRAY\s*\(' + args_pattern + r',\s*' + args_pattern + r'\)', re.S), r'\1 \2[]'),
639            (KernRe(r'DEFINE_DMA_UNMAP_ADDR\s*\(' + args_pattern + r'\)', re.S), r'dma_addr_t \1'),
640            (KernRe(r'DEFINE_DMA_UNMAP_LEN\s*\(' + args_pattern + r'\)', re.S), r'__u32 \1'),
641        ]
642
643        # Regexes here are guaranteed to have the end limiter matching
644        # the start delimiter. Yet, right now, only one replace group
645        # is allowed.
646
647        sub_nested_prefixes = [
648            (re.compile(r'\bSTRUCT_GROUP\('), r'\1'),
649        ]
650
651        for search, sub in sub_prefixes:
652            members = search.sub(sub, members)
653
654        nested = NestedMatch()
655
656        for search, sub in sub_nested_prefixes:
657            members = nested.sub(search, sub, members)
658
659        # Keeps the original declaration as-is
660        declaration = members
661
662        # Split nested struct/union elements
663        #
664        # This loop was simpler at the original kernel-doc perl version, as
665        #   while ($members =~ m/$struct_members/) { ... }
666        # reads 'members' string on each interaction.
667        #
668        # Python behavior is different: it parses 'members' only once,
669        # creating a list of tuples from the first interaction.
670        #
671        # On other words, this won't get nested structs.
672        #
673        # So, we need to have an extra loop on Python to override such
674        # re limitation.
675
676        while True:
677            tuples = struct_members.findall(members)
678            if not tuples:
679                break
680
681            for t in tuples:
682                newmember = ""
683                maintype = t[0]
684                s_ids = t[5]
685                content = t[3]
686
687                oldmember = "".join(t)
688
689                for s_id in s_ids.split(','):
690                    s_id = s_id.strip()
691
692                    newmember += f"{maintype} {s_id}; "
693                    s_id = KernRe(r'[:\[].*').sub('', s_id)
694                    s_id = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', s_id)
695
696                    for arg in content.split(';'):
697                        arg = arg.strip()
698
699                        if not arg:
700                            continue
701
702                        r = KernRe(r'^([^\(]+\(\*?\s*)([\w\.]*)(\s*\).*)')
703                        if r.match(arg):
704                            # Pointer-to-function
705                            dtype = r.group(1)
706                            name = r.group(2)
707                            extra = r.group(3)
708
709                            if not name:
710                                continue
711
712                            if not s_id:
713                                # Anonymous struct/union
714                                newmember += f"{dtype}{name}{extra}; "
715                            else:
716                                newmember += f"{dtype}{s_id}.{name}{extra}; "
717
718                        else:
719                            arg = arg.strip()
720                            # Handle bitmaps
721                            arg = KernRe(r':\s*\d+\s*').sub('', arg)
722
723                            # Handle arrays
724                            arg = KernRe(r'\[.*\]').sub('', arg)
725
726                            # Handle multiple IDs
727                            arg = KernRe(r'\s*,\s*').sub(',', arg)
728
729                            r = KernRe(r'(.*)\s+([\S+,]+)')
730
731                            if r.search(arg):
732                                dtype = r.group(1)
733                                names = r.group(2)
734                            else:
735                                newmember += f"{arg}; "
736                                continue
737
738                            for name in names.split(','):
739                                name = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', name).strip()
740
741                                if not name:
742                                    continue
743
744                                if not s_id:
745                                    # Anonymous struct/union
746                                    newmember += f"{dtype} {name}; "
747                                else:
748                                    newmember += f"{dtype} {s_id}.{name}; "
749
750                members = members.replace(oldmember, newmember)
751
752        # Ignore other nested elements, like enums
753        members = re.sub(r'(\{[^\{\}]*\})', '', members)
754
755        self.create_parameter_list(ln, decl_type, members, ';',
756                                   declaration_name)
757        self.check_sections(ln, declaration_name, decl_type)
758
759        # Adjust declaration for better display
760        declaration = KernRe(r'([\{;])').sub(r'\1\n', declaration)
761        declaration = KernRe(r'\}\s+;').sub('};', declaration)
762
763        # Better handle inlined enums
764        while True:
765            r = KernRe(r'(enum\s+\{[^\}]+),([^\n])')
766            if not r.search(declaration):
767                break
768
769            declaration = r.sub(r'\1,\n\2', declaration)
770
771        def_args = declaration.split('\n')
772        level = 1
773        declaration = ""
774        for clause in def_args:
775
776            clause = clause.strip()
777            clause = KernRe(r'\s+').sub(' ', clause, count=1)
778
779            if not clause:
780                continue
781
782            if '}' in clause and level > 1:
783                level -= 1
784
785            if not KernRe(r'^\s*#').match(clause):
786                declaration += "\t" * level
787
788            declaration += "\t" + clause + "\n"
789            if "{" in clause and "}" not in clause:
790                level += 1
791
792        self.output_declaration(decl_type, declaration_name,
793                                definition=declaration,
794                                purpose=self.entry.declaration_purpose)
795
796    def dump_enum(self, ln, proto):
797        """
798        Stores an enum inside self.entries array.
799        """
800
801        # Ignore members marked private
802        proto = KernRe(r'\/\*\s*private:.*?\/\*\s*public:.*?\*\/', flags=re.S).sub('', proto)
803        proto = KernRe(r'\/\*\s*private:.*}', flags=re.S).sub('}', proto)
804
805        # Strip comments
806        proto = KernRe(r'\/\*.*?\*\/', flags=re.S).sub('', proto)
807
808        # Strip #define macros inside enums
809        proto = KernRe(r'#\s*((define|ifdef|if)\s+|endif)[^;]*;', flags=re.S).sub('', proto)
810
811        #
812        # Parse out the name and members of the enum.  Typedef form first.
813        #
814        r = KernRe(r'typedef\s+enum\s*\{(.*)\}\s*(\w*)\s*;')
815        if r.search(proto):
816            declaration_name = r.group(2)
817            members = r.group(1).rstrip()
818        #
819        # Failing that, look for a straight enum
820        #
821        else:
822            r = KernRe(r'enum\s+(\w*)\s*\{(.*)\}')
823            if r.match(proto):
824                declaration_name = r.group(1)
825                members = r.group(2).rstrip()
826        #
827        # OK, this isn't going to work.
828        #
829            else:
830                self.emit_msg(ln, f"{proto}: error: Cannot parse enum!")
831                return
832        #
833        # Make sure we found what we were expecting.
834        #
835        if self.entry.identifier != declaration_name:
836            if self.entry.identifier == "":
837                self.emit_msg(ln,
838                              f"{proto}: wrong kernel-doc identifier on prototype")
839            else:
840                self.emit_msg(ln,
841                              f"expecting prototype for enum {self.entry.identifier}. "
842                              f"Prototype was for enum {declaration_name} instead")
843            return
844
845        if not declaration_name:
846            declaration_name = "(anonymous)"
847        #
848        # Parse out the name of each enum member, and verify that we
849        # have a description for it.
850        #
851        member_set = set()
852        members = KernRe(r'\([^;)]*\)').sub('', members)
853        for arg in members.split(','):
854            if not arg:
855                continue
856            arg = KernRe(r'^\s*(\w+).*').sub(r'\1', arg)
857            self.entry.parameterlist.append(arg)
858            if arg not in self.entry.parameterdescs:
859                self.entry.parameterdescs[arg] = self.undescribed
860                self.emit_msg(ln,
861                              f"Enum value '{arg}' not described in enum '{declaration_name}'")
862            member_set.add(arg)
863        #
864        # Ensure that every described member actually exists in the enum.
865        #
866        for k in self.entry.parameterdescs:
867            if k not in member_set:
868                self.emit_msg(ln,
869                              f"Excess enum value '%{k}' description in '{declaration_name}'")
870
871        self.output_declaration('enum', declaration_name,
872                                purpose=self.entry.declaration_purpose)
873
874    def dump_declaration(self, ln, prototype):
875        """
876        Stores a data declaration inside self.entries array.
877        """
878
879        if self.entry.decl_type == "enum":
880            self.dump_enum(ln, prototype)
881        elif self.entry.decl_type == "typedef":
882            self.dump_typedef(ln, prototype)
883        elif self.entry.decl_type in ["union", "struct"]:
884            self.dump_struct(ln, prototype)
885        else:
886            # This would be a bug
887            self.emit_message(ln, f'Unknown declaration type: {self.entry.decl_type}')
888
889    def dump_function(self, ln, prototype):
890        """
891        Stores a function of function macro inside self.entries array.
892        """
893
894        func_macro = False
895        return_type = ''
896        decl_type = 'function'
897
898        # Prefixes that would be removed
899        sub_prefixes = [
900            (r"^static +", "", 0),
901            (r"^extern +", "", 0),
902            (r"^asmlinkage +", "", 0),
903            (r"^inline +", "", 0),
904            (r"^__inline__ +", "", 0),
905            (r"^__inline +", "", 0),
906            (r"^__always_inline +", "", 0),
907            (r"^noinline +", "", 0),
908            (r"^__FORTIFY_INLINE +", "", 0),
909            (r"__init +", "", 0),
910            (r"__init_or_module +", "", 0),
911            (r"__deprecated +", "", 0),
912            (r"__flatten +", "", 0),
913            (r"__meminit +", "", 0),
914            (r"__must_check +", "", 0),
915            (r"__weak +", "", 0),
916            (r"__sched +", "", 0),
917            (r"_noprof", "", 0),
918            (r"__printf\s*\(\s*\d*\s*,\s*\d*\s*\) +", "", 0),
919            (r"__(?:re)?alloc_size\s*\(\s*\d+\s*(?:,\s*\d+\s*)?\) +", "", 0),
920            (r"__diagnose_as\s*\(\s*\S+\s*(?:,\s*\d+\s*)*\) +", "", 0),
921            (r"DECL_BUCKET_PARAMS\s*\(\s*(\S+)\s*,\s*(\S+)\s*\)", r"\1, \2", 0),
922            (r"__attribute_const__ +", "", 0),
923
924            # It seems that Python support for re.X is broken:
925            # At least for me (Python 3.13), this didn't work
926#            (r"""
927#              __attribute__\s*\(\(
928#                (?:
929#                    [\w\s]+          # attribute name
930#                    (?:\([^)]*\))?   # attribute arguments
931#                    \s*,?            # optional comma at the end
932#                )+
933#              \)\)\s+
934#             """, "", re.X),
935
936            # So, remove whitespaces and comments from it
937            (r"__attribute__\s*\(\((?:[\w\s]+(?:\([^)]*\))?\s*,?)+\)\)\s+", "", 0),
938        ]
939
940        for search, sub, flags in sub_prefixes:
941            prototype = KernRe(search, flags).sub(sub, prototype)
942
943        # Macros are a special case, as they change the prototype format
944        new_proto = KernRe(r"^#\s*define\s+").sub("", prototype)
945        if new_proto != prototype:
946            is_define_proto = True
947            prototype = new_proto
948        else:
949            is_define_proto = False
950
951        # Yes, this truly is vile.  We are looking for:
952        # 1. Return type (may be nothing if we're looking at a macro)
953        # 2. Function name
954        # 3. Function parameters.
955        #
956        # All the while we have to watch out for function pointer parameters
957        # (which IIRC is what the two sections are for), C types (these
958        # regexps don't even start to express all the possibilities), and
959        # so on.
960        #
961        # If you mess with these regexps, it's a good idea to check that
962        # the following functions' documentation still comes out right:
963        # - parport_register_device (function pointer parameters)
964        # - atomic_set (macro)
965        # - pci_match_device, __copy_to_user (long return type)
966
967        name = r'[a-zA-Z0-9_~:]+'
968        prototype_end1 = r'[^\(]*'
969        prototype_end2 = r'[^\{]*'
970        prototype_end = fr'\(({prototype_end1}|{prototype_end2})\)'
971
972        # Besides compiling, Perl qr{[\w\s]+} works as a non-capturing group.
973        # So, this needs to be mapped in Python with (?:...)? or (?:...)+
974
975        type1 = r'(?:[\w\s]+)?'
976        type2 = r'(?:[\w\s]+\*+)+'
977
978        found = False
979
980        if is_define_proto:
981            r = KernRe(r'^()(' + name + r')\s+')
982
983            if r.search(prototype):
984                return_type = ''
985                declaration_name = r.group(2)
986                func_macro = True
987
988                found = True
989
990        if not found:
991            patterns = [
992                rf'^()({name})\s*{prototype_end}',
993                rf'^({type1})\s+({name})\s*{prototype_end}',
994                rf'^({type2})\s*({name})\s*{prototype_end}',
995            ]
996
997            for p in patterns:
998                r = KernRe(p)
999
1000                if r.match(prototype):
1001
1002                    return_type = r.group(1)
1003                    declaration_name = r.group(2)
1004                    args = r.group(3)
1005
1006                    self.create_parameter_list(ln, decl_type, args, ',',
1007                                               declaration_name)
1008
1009                    found = True
1010                    break
1011        if not found:
1012            self.emit_msg(ln,
1013                          f"cannot understand function prototype: '{prototype}'")
1014            return
1015
1016        if self.entry.identifier != declaration_name:
1017            self.emit_msg(ln,
1018                          f"expecting prototype for {self.entry.identifier}(). Prototype was for {declaration_name}() instead")
1019            return
1020
1021        self.check_sections(ln, declaration_name, "function")
1022
1023        self.check_return_section(ln, declaration_name, return_type)
1024
1025        if 'typedef' in return_type:
1026            self.output_declaration(decl_type, declaration_name,
1027                                    typedef=True,
1028                                    functiontype=return_type,
1029                                    purpose=self.entry.declaration_purpose,
1030                                    func_macro=func_macro)
1031        else:
1032            self.output_declaration(decl_type, declaration_name,
1033                                    typedef=False,
1034                                    functiontype=return_type,
1035                                    purpose=self.entry.declaration_purpose,
1036                                    func_macro=func_macro)
1037
1038    def dump_typedef(self, ln, proto):
1039        """
1040        Stores a typedef inside self.entries array.
1041        """
1042
1043        typedef_type = r'((?:\s+[\w\*]+\b){0,7}\s+(?:\w+\b|\*+))\s*'
1044        typedef_ident = r'\*?\s*(\w\S+)\s*'
1045        typedef_args = r'\s*\((.*)\);'
1046
1047        typedef1 = KernRe(r'typedef' + typedef_type + r'\(' + typedef_ident + r'\)' + typedef_args)
1048        typedef2 = KernRe(r'typedef' + typedef_type + typedef_ident + typedef_args)
1049
1050        # Strip comments
1051        proto = KernRe(r'/\*.*?\*/', flags=re.S).sub('', proto)
1052
1053        # Parse function typedef prototypes
1054        for r in [typedef1, typedef2]:
1055            if not r.match(proto):
1056                continue
1057
1058            return_type = r.group(1).strip()
1059            declaration_name = r.group(2)
1060            args = r.group(3)
1061
1062            if self.entry.identifier != declaration_name:
1063                self.emit_msg(ln,
1064                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1065                return
1066
1067            decl_type = 'function'
1068            self.create_parameter_list(ln, decl_type, args, ',', declaration_name)
1069
1070            self.output_declaration(decl_type, declaration_name,
1071                                    typedef=True,
1072                                    functiontype=return_type,
1073                                    purpose=self.entry.declaration_purpose)
1074            return
1075
1076        # Handle nested parentheses or brackets
1077        r = KernRe(r'(\(*.\)\s*|\[*.\]\s*);$')
1078        while r.search(proto):
1079            proto = r.sub('', proto)
1080
1081        # Parse simple typedefs
1082        r = KernRe(r'typedef.*\s+(\w+)\s*;')
1083        if r.match(proto):
1084            declaration_name = r.group(1)
1085
1086            if self.entry.identifier != declaration_name:
1087                self.emit_msg(ln,
1088                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1089                return
1090
1091            self.output_declaration('typedef', declaration_name,
1092                                    purpose=self.entry.declaration_purpose)
1093            return
1094
1095        self.emit_msg(ln, "error: Cannot parse typedef!")
1096
1097    @staticmethod
1098    def process_export(function_set, line):
1099        """
1100        process EXPORT_SYMBOL* tags
1101
1102        This method doesn't use any variable from the class, so declare it
1103        with a staticmethod decorator.
1104        """
1105
1106        # We support documenting some exported symbols with different
1107        # names.  A horrible hack.
1108        suffixes = [ '_noprof' ]
1109
1110        # Note: it accepts only one EXPORT_SYMBOL* per line, as having
1111        # multiple export lines would violate Kernel coding style.
1112
1113        if export_symbol.search(line):
1114            symbol = export_symbol.group(2)
1115        elif export_symbol_ns.search(line):
1116            symbol = export_symbol_ns.group(2)
1117        else:
1118            return False
1119        #
1120        # Found an export, trim out any special suffixes
1121        #
1122        for suffix in suffixes:
1123            symbol = symbol.removesuffix(suffix)
1124        function_set.add(symbol)
1125        return True
1126
1127    def process_normal(self, ln, line):
1128        """
1129        STATE_NORMAL: looking for the /** to begin everything.
1130        """
1131
1132        if not doc_start.match(line):
1133            return
1134
1135        # start a new entry
1136        self.reset_state(ln)
1137
1138        # next line is always the function name
1139        self.state = state.NAME
1140
1141    def process_name(self, ln, line):
1142        """
1143        STATE_NAME: Looking for the "name - description" line
1144        """
1145        #
1146        # Check for a DOC: block and handle them specially.
1147        #
1148        if doc_block.search(line):
1149
1150            if not doc_block.group(1):
1151                self.entry.begin_section(ln, "Introduction")
1152            else:
1153                self.entry.begin_section(ln, doc_block.group(1))
1154
1155            self.entry.identifier = self.entry.section
1156            self.state = state.DOCBLOCK
1157        #
1158        # Otherwise we're looking for a normal kerneldoc declaration line.
1159        #
1160        elif doc_decl.search(line):
1161            self.entry.identifier = doc_decl.group(1)
1162
1163            # Test for data declaration
1164            if doc_begin_data.search(line):
1165                self.entry.decl_type = doc_begin_data.group(1)
1166                self.entry.identifier = doc_begin_data.group(2)
1167            #
1168            # Look for a function description
1169            #
1170            elif doc_begin_func.search(line):
1171                self.entry.identifier = doc_begin_func.group(1)
1172                self.entry.decl_type = "function"
1173            #
1174            # We struck out.
1175            #
1176            else:
1177                self.emit_msg(ln,
1178                              f"This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst\n{line}")
1179                self.state = state.NORMAL
1180                return
1181            #
1182            # OK, set up for a new kerneldoc entry.
1183            #
1184            self.state = state.BODY
1185            self.entry.identifier = self.entry.identifier.strip(" ")
1186            # if there's no @param blocks need to set up default section here
1187            self.entry.begin_section(ln + 1)
1188            #
1189            # Find the description portion, which *should* be there but
1190            # isn't always.
1191            # (We should be able to capture this from the previous parsing - someday)
1192            #
1193            r = KernRe("[-:](.*)")
1194            if r.search(line):
1195                self.entry.declaration_purpose = trim_whitespace(r.group(1))
1196                self.state = state.DECLARATION
1197            else:
1198                self.entry.declaration_purpose = ""
1199
1200            if not self.entry.declaration_purpose and self.config.wshort_desc:
1201                self.emit_msg(ln,
1202                              f"missing initial short description on line:\n{line}")
1203
1204            if not self.entry.identifier and self.entry.decl_type != "enum":
1205                self.emit_msg(ln,
1206                              f"wrong kernel-doc identifier on line:\n{line}")
1207                self.state = state.NORMAL
1208
1209            if self.config.verbose:
1210                self.emit_msg(ln,
1211                              f"Scanning doc for {self.entry.decl_type} {self.entry.identifier}",
1212                                  warning=False)
1213        #
1214        # Failed to find an identifier. Emit a warning
1215        #
1216        else:
1217            self.emit_msg(ln, f"Cannot find identifier on line:\n{line}")
1218
1219    #
1220    # Helper function to determine if a new section is being started.
1221    #
1222    def is_new_section(self, ln, line):
1223        if doc_sect.search(line):
1224            self.state = state.BODY
1225            #
1226            # Pick out the name of our new section, tweaking it if need be.
1227            #
1228            newsection = doc_sect.group(1)
1229            if newsection.lower() == 'description':
1230                newsection = 'Description'
1231            elif newsection.lower() == 'context':
1232                newsection = 'Context'
1233                self.state = state.SPECIAL_SECTION
1234            elif newsection.lower() in ["@return", "@returns",
1235                                        "return", "returns"]:
1236                newsection = "Return"
1237                self.state = state.SPECIAL_SECTION
1238            elif newsection[0] == '@':
1239                self.state = state.SPECIAL_SECTION
1240            #
1241            # Initialize the contents, and get the new section going.
1242            #
1243            newcontents = doc_sect.group(2)
1244            if not newcontents:
1245                newcontents = ""
1246            self.dump_section()
1247            self.entry.begin_section(ln, newsection)
1248            self.entry.leading_space = None
1249
1250            self.entry.add_text(newcontents.lstrip())
1251            return True
1252        return False
1253
1254    #
1255    # Helper function to detect (and effect) the end of a kerneldoc comment.
1256    #
1257    def is_comment_end(self, ln, line):
1258        if doc_end.search(line):
1259            self.dump_section()
1260
1261            # Look for doc_com + <text> + doc_end:
1262            r = KernRe(r'\s*\*\s*[a-zA-Z_0-9:\.]+\*/')
1263            if r.match(line):
1264                self.emit_msg(ln, f"suspicious ending line: {line}")
1265
1266            self.entry.prototype = ""
1267            self.entry.new_start_line = ln + 1
1268
1269            self.state = state.PROTO
1270            return True
1271        return False
1272
1273
1274    def process_decl(self, ln, line):
1275        """
1276        STATE_DECLARATION: We've seen the beginning of a declaration
1277        """
1278        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1279            return
1280        #
1281        # Look for anything with the " * " line beginning.
1282        #
1283        if doc_content.search(line):
1284            cont = doc_content.group(1)
1285            #
1286            # A blank line means that we have moved out of the declaration
1287            # part of the comment (without any "special section" parameter
1288            # descriptions).
1289            #
1290            if cont == "":
1291                self.state = state.BODY
1292            #
1293            # Otherwise we have more of the declaration section to soak up.
1294            #
1295            else:
1296                self.entry.declaration_purpose = \
1297                    trim_whitespace(self.entry.declaration_purpose + ' ' + cont)
1298        else:
1299            # Unknown line, ignore
1300            self.emit_msg(ln, f"bad line: {line}")
1301
1302
1303    def process_special(self, ln, line):
1304        """
1305        STATE_SPECIAL_SECTION: a section ending with a blank line
1306        """
1307        #
1308        # If we have hit a blank line (only the " * " marker), then this
1309        # section is done.
1310        #
1311        if KernRe(r"\s*\*\s*$").match(line):
1312            self.entry.begin_section(ln, dump = True)
1313            self.state = state.BODY
1314            return
1315        #
1316        # Not a blank line, look for the other ways to end the section.
1317        #
1318        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1319            return
1320        #
1321        # OK, we should have a continuation of the text for this section.
1322        #
1323        if doc_content.search(line):
1324            cont = doc_content.group(1)
1325            #
1326            # If the lines of text after the first in a special section have
1327            # leading white space, we need to trim it out or Sphinx will get
1328            # confused.  For the second line (the None case), see what we
1329            # find there and remember it.
1330            #
1331            if self.entry.leading_space is None:
1332                r = KernRe(r'^(\s+)')
1333                if r.match(cont):
1334                    self.entry.leading_space = len(r.group(1))
1335                else:
1336                    self.entry.leading_space = 0
1337            #
1338            # Otherwise, before trimming any leading chars, be *sure*
1339            # that they are white space.  We should maybe warn if this
1340            # isn't the case.
1341            #
1342            for i in range(0, self.entry.leading_space):
1343                if cont[i] != " ":
1344                    self.entry.leading_space = i
1345                    break
1346            #
1347            # Add the trimmed result to the section and we're done.
1348            #
1349            self.entry.add_text(cont[self.entry.leading_space:])
1350        else:
1351            # Unknown line, ignore
1352            self.emit_msg(ln, f"bad line: {line}")
1353
1354    def process_body(self, ln, line):
1355        """
1356        STATE_BODY: the bulk of a kerneldoc comment.
1357        """
1358        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1359            return
1360
1361        if doc_content.search(line):
1362            cont = doc_content.group(1)
1363            self.entry.add_text(cont)
1364        else:
1365            # Unknown line, ignore
1366            self.emit_msg(ln, f"bad line: {line}")
1367
1368    def process_inline_name(self, ln, line):
1369        """STATE_INLINE_NAME: beginning of docbook comments within a prototype."""
1370
1371        if doc_inline_sect.search(line):
1372            self.entry.begin_section(ln, doc_inline_sect.group(1))
1373            self.entry.add_text(doc_inline_sect.group(2).lstrip())
1374            self.state = state.INLINE_TEXT
1375        elif doc_inline_end.search(line):
1376            self.dump_section()
1377            self.state = state.PROTO
1378        elif doc_content.search(line):
1379            self.emit_msg(ln, f"Incorrect use of kernel-doc format: {line}")
1380            self.state = state.PROTO
1381        # else ... ??
1382
1383    def process_inline_text(self, ln, line):
1384        """STATE_INLINE_TEXT: docbook comments within a prototype."""
1385
1386        if doc_inline_end.search(line):
1387            self.dump_section()
1388            self.state = state.PROTO
1389        elif doc_content.search(line):
1390            self.entry.add_text(doc_content.group(1))
1391        # else ... ??
1392
1393    def syscall_munge(self, ln, proto):         # pylint: disable=W0613
1394        """
1395        Handle syscall definitions
1396        """
1397
1398        is_void = False
1399
1400        # Strip newlines/CR's
1401        proto = re.sub(r'[\r\n]+', ' ', proto)
1402
1403        # Check if it's a SYSCALL_DEFINE0
1404        if 'SYSCALL_DEFINE0' in proto:
1405            is_void = True
1406
1407        # Replace SYSCALL_DEFINE with correct return type & function name
1408        proto = KernRe(r'SYSCALL_DEFINE.*\(').sub('long sys_', proto)
1409
1410        r = KernRe(r'long\s+(sys_.*?),')
1411        if r.search(proto):
1412            proto = KernRe(',').sub('(', proto, count=1)
1413        elif is_void:
1414            proto = KernRe(r'\)').sub('(void)', proto, count=1)
1415
1416        # Now delete all of the odd-numbered commas in the proto
1417        # so that argument types & names don't have a comma between them
1418        count = 0
1419        length = len(proto)
1420
1421        if is_void:
1422            length = 0  # skip the loop if is_void
1423
1424        for ix in range(length):
1425            if proto[ix] == ',':
1426                count += 1
1427                if count % 2 == 1:
1428                    proto = proto[:ix] + ' ' + proto[ix + 1:]
1429
1430        return proto
1431
1432    def tracepoint_munge(self, ln, proto):
1433        """
1434        Handle tracepoint definitions
1435        """
1436
1437        tracepointname = None
1438        tracepointargs = None
1439
1440        # Match tracepoint name based on different patterns
1441        r = KernRe(r'TRACE_EVENT\((.*?),')
1442        if r.search(proto):
1443            tracepointname = r.group(1)
1444
1445        r = KernRe(r'DEFINE_SINGLE_EVENT\((.*?),')
1446        if r.search(proto):
1447            tracepointname = r.group(1)
1448
1449        r = KernRe(r'DEFINE_EVENT\((.*?),(.*?),')
1450        if r.search(proto):
1451            tracepointname = r.group(2)
1452
1453        if tracepointname:
1454            tracepointname = tracepointname.lstrip()
1455
1456        r = KernRe(r'TP_PROTO\((.*?)\)')
1457        if r.search(proto):
1458            tracepointargs = r.group(1)
1459
1460        if not tracepointname or not tracepointargs:
1461            self.emit_msg(ln,
1462                          f"Unrecognized tracepoint format:\n{proto}\n")
1463        else:
1464            proto = f"static inline void trace_{tracepointname}({tracepointargs})"
1465            self.entry.identifier = f"trace_{self.entry.identifier}"
1466
1467        return proto
1468
1469    def process_proto_function(self, ln, line):
1470        """Ancillary routine to process a function prototype"""
1471
1472        # strip C99-style comments to end of line
1473        line = KernRe(r"\/\/.*$", re.S).sub('', line)
1474        #
1475        # Soak up the line's worth of prototype text, stopping at { or ; if present.
1476        #
1477        if KernRe(r'\s*#\s*define').match(line):
1478            self.entry.prototype = line
1479        elif not line.startswith('#'):   # skip other preprocessor stuff
1480            r = KernRe(r'([^\{]*)')
1481            if r.match(line):
1482                self.entry.prototype += r.group(1) + " "
1483        #
1484        # If we now have the whole prototype, clean it up and declare victory.
1485        #
1486        if '{' in line or ';' in line or KernRe(r'\s*#\s*define').match(line):
1487            # strip comments and surrounding spaces
1488            self.entry.prototype = KernRe(r'/\*.*\*/').sub('', self.entry.prototype).strip()
1489            #
1490            # Handle self.entry.prototypes for function pointers like:
1491            #       int (*pcs_config)(struct foo)
1492            # by turning it into
1493            #	    int pcs_config(struct foo)
1494            #
1495            r = KernRe(r'^(\S+\s+)\(\s*\*(\S+)\)')
1496            self.entry.prototype = r.sub(r'\1\2', self.entry.prototype)
1497            #
1498            # Handle special declaration syntaxes
1499            #
1500            if 'SYSCALL_DEFINE' in self.entry.prototype:
1501                self.entry.prototype = self.syscall_munge(ln,
1502                                                          self.entry.prototype)
1503            else:
1504                r = KernRe(r'TRACE_EVENT|DEFINE_EVENT|DEFINE_SINGLE_EVENT')
1505                if r.search(self.entry.prototype):
1506                    self.entry.prototype = self.tracepoint_munge(ln,
1507                                                                 self.entry.prototype)
1508            #
1509            # ... and we're done
1510            #
1511            self.dump_function(ln, self.entry.prototype)
1512            self.reset_state(ln)
1513
1514    def process_proto_type(self, ln, line):
1515        """Ancillary routine to process a type"""
1516
1517        # Strip C99-style comments and surrounding whitespace
1518        line = KernRe(r"//.*$", re.S).sub('', line).strip()
1519        if not line:
1520            return # nothing to see here
1521
1522        # To distinguish preprocessor directive from regular declaration later.
1523        if line.startswith('#'):
1524            line += ";"
1525        #
1526        # Split the declaration on any of { } or ;, and accumulate pieces
1527        # until we hit a semicolon while not inside {brackets}
1528        #
1529        r = KernRe(r'(.*?)([{};])')
1530        for chunk in r.split(line):
1531            if chunk:  # Ignore empty matches
1532                self.entry.prototype += chunk
1533                #
1534                # This cries out for a match statement ... someday after we can
1535                # drop Python 3.9 ...
1536                #
1537                if chunk == '{':
1538                    self.entry.brcount += 1
1539                elif chunk == '}':
1540                    self.entry.brcount -= 1
1541                elif chunk == ';' and self.entry.brcount <= 0:
1542                    self.dump_declaration(ln, self.entry.prototype)
1543                    self.reset_state(ln)
1544                    return
1545        #
1546        # We hit the end of the line while still in the declaration; put
1547        # in a space to represent the newline.
1548        #
1549        self.entry.prototype += ' '
1550
1551    def process_proto(self, ln, line):
1552        """STATE_PROTO: reading a function/whatever prototype."""
1553
1554        if doc_inline_oneline.search(line):
1555            self.entry.begin_section(ln, doc_inline_oneline.group(1))
1556            self.entry.add_text(doc_inline_oneline.group(2))
1557            self.dump_section()
1558
1559        elif doc_inline_start.search(line):
1560            self.state = state.INLINE_NAME
1561
1562        elif self.entry.decl_type == 'function':
1563            self.process_proto_function(ln, line)
1564
1565        else:
1566            self.process_proto_type(ln, line)
1567
1568    def process_docblock(self, ln, line):
1569        """STATE_DOCBLOCK: within a DOC: block."""
1570
1571        if doc_end.search(line):
1572            self.dump_section()
1573            self.output_declaration("doc", self.entry.identifier)
1574            self.reset_state(ln)
1575
1576        elif doc_content.search(line):
1577            self.entry.add_text(doc_content.group(1))
1578
1579    def parse_export(self):
1580        """
1581        Parses EXPORT_SYMBOL* macros from a single Kernel source file.
1582        """
1583
1584        export_table = set()
1585
1586        try:
1587            with open(self.fname, "r", encoding="utf8",
1588                      errors="backslashreplace") as fp:
1589
1590                for line in fp:
1591                    self.process_export(export_table, line)
1592
1593        except IOError:
1594            return None
1595
1596        return export_table
1597
1598    #
1599    # The state/action table telling us which function to invoke in
1600    # each state.
1601    #
1602    state_actions = {
1603        state.NORMAL:			process_normal,
1604        state.NAME:			process_name,
1605        state.BODY:			process_body,
1606        state.DECLARATION:		process_decl,
1607        state.SPECIAL_SECTION:		process_special,
1608        state.INLINE_NAME:		process_inline_name,
1609        state.INLINE_TEXT:		process_inline_text,
1610        state.PROTO:			process_proto,
1611        state.DOCBLOCK:			process_docblock,
1612        }
1613
1614    def parse_kdoc(self):
1615        """
1616        Open and process each line of a C source file.
1617        The parsing is controlled via a state machine, and the line is passed
1618        to a different process function depending on the state. The process
1619        function may update the state as needed.
1620
1621        Besides parsing kernel-doc tags, it also parses export symbols.
1622        """
1623
1624        prev = ""
1625        prev_ln = None
1626        export_table = set()
1627
1628        try:
1629            with open(self.fname, "r", encoding="utf8",
1630                      errors="backslashreplace") as fp:
1631                for ln, line in enumerate(fp):
1632
1633                    line = line.expandtabs().strip("\n")
1634
1635                    # Group continuation lines on prototypes
1636                    if self.state == state.PROTO:
1637                        if line.endswith("\\"):
1638                            prev += line.rstrip("\\")
1639                            if not prev_ln:
1640                                prev_ln = ln
1641                            continue
1642
1643                        if prev:
1644                            ln = prev_ln
1645                            line = prev + line
1646                            prev = ""
1647                            prev_ln = None
1648
1649                    self.config.log.debug("%d %s: %s",
1650                                          ln, state.name[self.state],
1651                                          line)
1652
1653                    # This is an optimization over the original script.
1654                    # There, when export_file was used for the same file,
1655                    # it was read twice. Here, we use the already-existing
1656                    # loop to parse exported symbols as well.
1657                    #
1658                    if (self.state != state.NORMAL) or \
1659                       not self.process_export(export_table, line):
1660                        # Hand this line to the appropriate state handler
1661                        self.state_actions[self.state](self, ln, line)
1662
1663        except OSError:
1664            self.config.log.error(f"Error: Cannot open file {self.fname}")
1665
1666        return export_table, self.entries
1667