xref: /linux/tools/lib/python/kdoc/kdoc_parser.py (revision 97d4e70bc2c6f75911a9a5e1a75f2de13fde9b6b)
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"""
8Classes and functions related to reading a C language source or header FILE
9and extract embedded documentation comments from it.
10"""
11
12import sys
13import re
14from pprint import pformat
15
16from kdoc.kdoc_re import NestedMatch, KernRe
17from kdoc.kdoc_item import KdocItem
18
19#
20# Regular expressions used to parse kernel-doc markups at KernelDoc class.
21#
22# Let's declare them in lowercase outside any class to make it easier to
23# convert from the Perl script.
24#
25# As those are evaluated at the beginning, no need to cache them
26#
27
28# Allow whitespace at end of comment start.
29doc_start = KernRe(r'^/\*\*\s*$', cache=False)
30
31doc_end = KernRe(r'\*/', cache=False)
32doc_com = KernRe(r'\s*\*\s*', cache=False)
33doc_com_body = KernRe(r'\s*\* ?', cache=False)
34doc_decl = doc_com + KernRe(r'(\w+)', cache=False)
35
36# @params and a strictly limited set of supported section names
37# Specifically:
38#   Match @word:
39#         @...:
40#         @{section-name}:
41# while trying to not match literal block starts like "example::"
42#
43known_section_names = 'description|context|returns?|notes?|examples?'
44known_sections = KernRe(known_section_names, flags = re.I)
45doc_sect = doc_com + \
46    KernRe(r'\s*(@[.\w]+|@\.\.\.|' + known_section_names + r')\s*:([^:].*)?$',
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*(@\s*[\w][\w\.]*\s*):\s*(.*)\s*\*/\s*$', cache=False)
54
55export_symbol = KernRe(r'^\s*EXPORT_SYMBOL(_GPL)?\s*\(\s*(\w+)\s*\)\s*', cache=False)
56export_symbol_ns = KernRe(r'^\s*EXPORT_SYMBOL_NS(_GPL)?\s*\(\s*(\w+)\s*,\s*"\S+"\)\s*', cache=False)
57
58type_param = KernRe(r"@(\w*((\.\w+)|(->\w+))*(\.\.\.)?)", cache=False)
59
60#
61# Tests for the beginning of a kerneldoc block in its various forms.
62#
63doc_block = doc_com + KernRe(r'DOC:\s*(.*)?', cache=False)
64doc_begin_data = KernRe(r"^\s*\*?\s*(struct|union|enum|typedef|var)\b\s*(\w*)", cache = False)
65doc_begin_func = KernRe(str(doc_com) +			# initial " * '
66                        r"(?:\w+\s*\*\s*)?" + 		# type (not captured)
67                        r'(?:define\s+)?' + 		# possible "define" (not captured)
68                        r'(\w+)\s*(?:\(\w*\))?\s*' +	# name and optional "(...)"
69                        r'(?:[-:].*)?$',		# description (not captured)
70                        cache = False)
71
72#
73# Here begins a long set of transformations to turn structure member prefixes
74# and macro invocations into something we can parse and generate kdoc for.
75#
76struct_args_pattern = r'([^,)]+)'
77
78struct_xforms = [
79    # Strip attributes
80    (KernRe(r"__attribute__\s*\(\([a-z0-9,_\*\s\(\)]*\)\)", flags=re.I | re.S, cache=False), ' '),
81    (KernRe(r'\s*__aligned\s*\([^;]*\)', re.S), ' '),
82    (KernRe(r'\s*__counted_by\s*\([^;]*\)', re.S), ' '),
83    (KernRe(r'\s*__counted_by_(le|be)\s*\([^;]*\)', re.S), ' '),
84    (KernRe(r'\s*__packed\s*', re.S), ' '),
85    (KernRe(r'\s*CRYPTO_MINALIGN_ATTR', re.S), ' '),
86    (KernRe(r'\s*__private', re.S), ' '),
87    (KernRe(r'\s*__rcu', re.S), ' '),
88    (KernRe(r'\s*____cacheline_aligned_in_smp', re.S), ' '),
89    (KernRe(r'\s*____cacheline_aligned', re.S), ' '),
90    (KernRe(r'\s*__cacheline_group_(begin|end)\([^\)]+\);'), ''),
91    #
92    # Unwrap struct_group macros based on this definition:
93    # __struct_group(TAG, NAME, ATTRS, MEMBERS...)
94    # which has variants like: struct_group(NAME, MEMBERS...)
95    # Only MEMBERS arguments require documentation.
96    #
97    # Parsing them happens on two steps:
98    #
99    # 1. drop struct group arguments that aren't at MEMBERS,
100    #    storing them as STRUCT_GROUP(MEMBERS)
101    #
102    # 2. remove STRUCT_GROUP() ancillary macro.
103    #
104    # The original logic used to remove STRUCT_GROUP() using an
105    # advanced regex:
106    #
107    #   \bSTRUCT_GROUP(\(((?:(?>[^)(]+)|(?1))*)\))[^;]*;
108    #
109    # with two patterns that are incompatible with
110    # Python re module, as it has:
111    #
112    #   - a recursive pattern: (?1)
113    #   - an atomic grouping: (?>...)
114    #
115    # I tried a simpler version: but it didn't work either:
116    #   \bSTRUCT_GROUP\(([^\)]+)\)[^;]*;
117    #
118    # As it doesn't properly match the end parenthesis on some cases.
119    #
120    # So, a better solution was crafted: there's now a NestedMatch
121    # class that ensures that delimiters after a search are properly
122    # matched. So, the implementation to drop STRUCT_GROUP() will be
123    # handled in separate.
124    #
125    (KernRe(r'\bstruct_group\s*\(([^,]*,)', re.S), r'STRUCT_GROUP('),
126    (KernRe(r'\bstruct_group_attr\s*\(([^,]*,){2}', re.S), r'STRUCT_GROUP('),
127    (KernRe(r'\bstruct_group_tagged\s*\(([^,]*),([^,]*),', re.S), r'struct \1 \2; STRUCT_GROUP('),
128    (KernRe(r'\b__struct_group\s*\(([^,]*,){3}', re.S), r'STRUCT_GROUP('),
129    #
130    # Replace macros
131    #
132    # TODO: use NestedMatch for FOO($1, $2, ...) matches
133    #
134    # it is better to also move those to the NestedMatch logic,
135    # to ensure that parentheses will be properly matched.
136    #
137    (KernRe(r'__ETHTOOL_DECLARE_LINK_MODE_MASK\s*\(([^\)]+)\)', re.S),
138     r'DECLARE_BITMAP(\1, __ETHTOOL_LINK_MODE_MASK_NBITS)'),
139    (KernRe(r'DECLARE_PHY_INTERFACE_MASK\s*\(([^\)]+)\)', re.S),
140     r'DECLARE_BITMAP(\1, PHY_INTERFACE_MODE_MAX)'),
141    (KernRe(r'DECLARE_BITMAP\s*\(' + struct_args_pattern + r',\s*' + struct_args_pattern + r'\)',
142            re.S), r'unsigned long \1[BITS_TO_LONGS(\2)]'),
143    (KernRe(r'DECLARE_HASHTABLE\s*\(' + struct_args_pattern + r',\s*' + struct_args_pattern + r'\)',
144            re.S), r'unsigned long \1[1 << ((\2) - 1)]'),
145    (KernRe(r'DECLARE_KFIFO\s*\(' + struct_args_pattern + r',\s*' + struct_args_pattern +
146            r',\s*' + struct_args_pattern + r'\)', re.S), r'\2 *\1'),
147    (KernRe(r'DECLARE_KFIFO_PTR\s*\(' + struct_args_pattern + r',\s*' +
148            struct_args_pattern + r'\)', re.S), r'\2 *\1'),
149    (KernRe(r'(?:__)?DECLARE_FLEX_ARRAY\s*\(' + struct_args_pattern + r',\s*' +
150            struct_args_pattern + r'\)', re.S), r'\1 \2[]'),
151    (KernRe(r'DEFINE_DMA_UNMAP_ADDR\s*\(' + struct_args_pattern + r'\)', re.S), r'dma_addr_t \1'),
152    (KernRe(r'DEFINE_DMA_UNMAP_LEN\s*\(' + struct_args_pattern + r'\)', re.S), r'__u32 \1'),
153    (KernRe(r'VIRTIO_DECLARE_FEATURES\(([\w_]+)\)'), r'union { u64 \1; u64 \1_array[VIRTIO_FEATURES_U64S]; }'),
154]
155#
156# Regexes here are guaranteed to have the end delimiter matching
157# the start delimiter. Yet, right now, only one replace group
158# is allowed.
159#
160struct_nested_prefixes = [
161    (re.compile(r'\bSTRUCT_GROUP\('), r'\1'),
162]
163
164#
165# Transforms for function prototypes
166#
167function_xforms = [
168    (KernRe(r"^static +"), ""),
169    (KernRe(r"^extern +"), ""),
170    (KernRe(r"^asmlinkage +"), ""),
171    (KernRe(r"^inline +"), ""),
172    (KernRe(r"^__inline__ +"), ""),
173    (KernRe(r"^__inline +"), ""),
174    (KernRe(r"^__always_inline +"), ""),
175    (KernRe(r"^noinline +"), ""),
176    (KernRe(r"^__FORTIFY_INLINE +"), ""),
177    (KernRe(r"__init +"), ""),
178    (KernRe(r"__init_or_module +"), ""),
179    (KernRe(r"__exit +"), ""),
180    (KernRe(r"__deprecated +"), ""),
181    (KernRe(r"__flatten +"), ""),
182    (KernRe(r"__meminit +"), ""),
183    (KernRe(r"__must_check +"), ""),
184    (KernRe(r"__weak +"), ""),
185    (KernRe(r"__sched +"), ""),
186    (KernRe(r"_noprof"), ""),
187    (KernRe(r"__always_unused *"), ""),
188    (KernRe(r"__printf\s*\(\s*\d*\s*,\s*\d*\s*\) +"), ""),
189    (KernRe(r"__(?:re)?alloc_size\s*\(\s*\d+\s*(?:,\s*\d+\s*)?\) +"), ""),
190    (KernRe(r"__diagnose_as\s*\(\s*\S+\s*(?:,\s*\d+\s*)*\) +"), ""),
191    (KernRe(r"DECL_BUCKET_PARAMS\s*\(\s*(\S+)\s*,\s*(\S+)\s*\)"), r"\1, \2"),
192    (KernRe(r"__attribute_const__ +"), ""),
193    (KernRe(r"__attribute__\s*\(\((?:[\w\s]+(?:\([^)]*\))?\s*,?)+\)\)\s+"), ""),
194]
195
196#
197# Transforms for variable prototypes
198#
199var_xforms = [
200    (KernRe(r"__read_mostly"), ""),
201    (KernRe(r"__ro_after_init"), ""),
202    (KernRe(r"LIST_HEAD\(([\w_]+)\)"), r"struct list_head \1"),
203    (KernRe(r"(?://.*)$"), ""),
204    (KernRe(r"(?:/\*.*\*/)"), ""),
205    (KernRe(r";$"), ""),
206]
207
208#
209# Ancillary functions
210#
211
212def apply_transforms(xforms, text):
213    """
214    Apply a set of transforms to a block of text.
215    """
216    for search, subst in xforms:
217        text = search.sub(subst, text)
218    return text
219
220multi_space = KernRe(r'\s\s+')
221def trim_whitespace(s):
222    """
223    A little helper to get rid of excess white space.
224    """
225    return multi_space.sub(' ', s.strip())
226
227def trim_private_members(text):
228    """
229    Remove ``struct``/``enum`` members that have been marked "private".
230    """
231    # First look for a "public:" block that ends a private region, then
232    # handle the "private until the end" case.
233    #
234    text = KernRe(r'/\*\s*private:.*?/\*\s*public:.*?\*/', flags=re.S).sub('', text)
235    text = KernRe(r'/\*\s*private:.*', flags=re.S).sub('', text)
236    #
237    # We needed the comments to do the above, but now we can take them out.
238    #
239    return KernRe(r'\s*/\*.*?\*/\s*', flags=re.S).sub('', text).strip()
240
241class state:
242    """
243    States used by the parser's state machine.
244    """
245
246    # Parser states
247    NORMAL        = 0        #: Normal code.
248    NAME          = 1        #: Looking for function name.
249    DECLARATION   = 2        #: We have seen a declaration which might not be done.
250    BODY          = 3        #: The body of the comment.
251    SPECIAL_SECTION = 4      #: Doc section ending with a blank line.
252    PROTO         = 5        #: Scanning prototype.
253    DOCBLOCK      = 6        #: Documentation block.
254    INLINE_NAME   = 7        #: Gathering doc outside main block.
255    INLINE_TEXT   = 8	     #: Reading the body of inline docs.
256
257    #: Names for each parser state.
258    name = [
259        "NORMAL",
260        "NAME",
261        "DECLARATION",
262        "BODY",
263        "SPECIAL_SECTION",
264        "PROTO",
265        "DOCBLOCK",
266        "INLINE_NAME",
267        "INLINE_TEXT",
268    ]
269
270
271SECTION_DEFAULT = "Description"  #: Default section.
272
273class KernelEntry:
274    """
275    Encapsulates a Kernel documentation entry.
276    """
277
278    def __init__(self, config, fname, ln):
279        self.config = config
280        self.fname = fname
281
282        self._contents = []
283        self.prototype = ""
284
285        self.warnings = []
286
287        self.parameterlist = []
288        self.parameterdescs = {}
289        self.parametertypes = {}
290        self.parameterdesc_start_lines = {}
291
292        self.section_start_lines = {}
293        self.sections = {}
294
295        self.anon_struct_union = False
296
297        self.leading_space = None
298
299        self.fname = fname
300
301        # State flags
302        self.brcount = 0
303        self.declaration_start_line = ln + 1
304
305    #
306    # Management of section contents
307    #
308    def add_text(self, text):
309        """Add a new text to the entry contents list."""
310        self._contents.append(text)
311
312    def contents(self):
313        """Returns a string with all content texts that were added."""
314        return '\n'.join(self._contents) + '\n'
315
316    # TODO: rename to emit_message after removal of kernel-doc.pl
317    def emit_msg(self, ln, msg, *, warning=True):
318        """Emit a message."""
319
320        log_msg = f"{self.fname}:{ln} {msg}"
321
322        if not warning:
323            self.config.log.info(log_msg)
324            return
325
326        # Delegate warning output to output logic, as this way it
327        # will report warnings/info only for symbols that are output
328
329        self.warnings.append(log_msg)
330        return
331
332    def begin_section(self, line_no, title = SECTION_DEFAULT, dump = False):
333        """
334        Begin a new section.
335        """
336        if dump:
337            self.dump_section(start_new = True)
338        self.section = title
339        self.new_start_line = line_no
340
341    def dump_section(self, start_new=True):
342        """
343        Dumps section contents to arrays/hashes intended for that purpose.
344        """
345        #
346        # If we have accumulated no contents in the default ("description")
347        # section, don't bother.
348        #
349        if self.section == SECTION_DEFAULT and not self._contents:
350            return
351        name = self.section
352        contents = self.contents()
353
354        if type_param.match(name):
355            name = type_param.group(1)
356
357            self.parameterdescs[name] = contents
358            self.parameterdesc_start_lines[name] = self.new_start_line
359
360            self.new_start_line = 0
361
362        else:
363            if name in self.sections and self.sections[name] != "":
364                # Only warn on user-specified duplicate section names
365                if name != SECTION_DEFAULT:
366                    self.emit_msg(self.new_start_line,
367                                  f"duplicate section name '{name}'")
368                # Treat as a new paragraph - add a blank line
369                self.sections[name] += '\n' + contents
370            else:
371                self.sections[name] = contents
372                self.section_start_lines[name] = self.new_start_line
373                self.new_start_line = 0
374
375#        self.config.log.debug("Section: %s : %s", name, pformat(vars(self)))
376
377        if start_new:
378            self.section = SECTION_DEFAULT
379            self._contents = []
380
381python_warning = False
382
383class KernelDoc:
384    """
385    Read a C language source or header FILE and extract embedded
386    documentation comments.
387    """
388
389    #: Name of context section.
390    section_context = "Context"
391
392    #: Name of return section.
393    section_return = "Return"
394
395    #: String to write when a parameter is not described.
396    undescribed = "-- undescribed --"
397
398    def __init__(self, config, fname):
399        """Initialize internal variables"""
400
401        self.fname = fname
402        self.config = config
403
404        # Initial state for the state machines
405        self.state = state.NORMAL
406
407        # Store entry currently being processed
408        self.entry = None
409
410        # Place all potential outputs into an array
411        self.entries = []
412
413        #
414        # We need Python 3.7 for its "dicts remember the insertion
415        # order" guarantee
416        #
417        global python_warning
418        if (not python_warning and
419            sys.version_info.major == 3 and sys.version_info.minor < 7):
420
421            self.emit_msg(0,
422                          'Python 3.7 or later is required for correct results')
423            python_warning = True
424
425    def emit_msg(self, ln, msg, *, warning=True):
426        """Emit a message"""
427
428        if self.entry:
429            self.entry.emit_msg(ln, msg, warning=warning)
430            return
431
432        log_msg = f"{self.fname}:{ln} {msg}"
433
434        if warning:
435            self.config.log.warning(log_msg)
436        else:
437            self.config.log.info(log_msg)
438
439    def dump_section(self, start_new=True):
440        """
441        Dump section contents to arrays/hashes intended for that purpose.
442        """
443
444        if self.entry:
445            self.entry.dump_section(start_new)
446
447    # TODO: rename it to store_declaration after removal of kernel-doc.pl
448    def output_declaration(self, dtype, name, **args):
449        """
450        Store the entry into an entry array.
451
452        The actual output and output filters will be handled elsewhere.
453        """
454
455        item = KdocItem(name, self.fname, dtype,
456                        self.entry.declaration_start_line, **args)
457        item.warnings = self.entry.warnings
458
459        # Drop empty sections
460        # TODO: improve empty sections logic to emit warnings
461        sections = self.entry.sections
462        for section in ["Description", "Return"]:
463            if section in sections and not sections[section].rstrip():
464                del sections[section]
465        item.set_sections(sections, self.entry.section_start_lines)
466        item.set_params(self.entry.parameterlist, self.entry.parameterdescs,
467                        self.entry.parametertypes,
468                        self.entry.parameterdesc_start_lines)
469        self.entries.append(item)
470
471        self.config.log.debug("Output: %s:%s = %s", dtype, name, pformat(args))
472
473    def emit_unused_warnings(self):
474        """
475        When the parser fails to produce a valid entry, it places some
476        warnings under `entry.warnings` that will be discarded when resetting
477        the state.
478
479        Ensure that those warnings are not lost.
480
481        .. note::
482
483              Because we are calling `config.warning()` here, those
484              warnings are not filtered by the `-W` parameters: they will all
485              be produced even when `-Wreturn`, `-Wshort-desc`, and/or
486              `-Wcontents-before-sections` are used.
487
488              Allowing those warnings to be filtered is complex, because it
489              would require storing them in a buffer and then filtering them
490              during the output step of the code, depending on the
491              selected symbols.
492        """
493        if self.entry and self.entry not in self.entries:
494            for log_msg in self.entry.warnings:
495                self.config.warning(log_msg)
496
497    def reset_state(self, ln):
498        """
499        Ancillary routine to create a new entry. It initializes all
500        variables used by the state machine.
501        """
502
503        self.emit_unused_warnings()
504
505        self.entry = KernelEntry(self.config, self.fname, ln)
506
507        # State flags
508        self.state = state.NORMAL
509
510    def push_parameter(self, ln, decl_type, param, dtype,
511                       org_arg, declaration_name):
512        """
513        Store parameters and their descriptions at self.entry.
514        """
515
516        if self.entry.anon_struct_union and dtype == "" and param == "}":
517            return  # Ignore the ending }; from anonymous struct/union
518
519        self.entry.anon_struct_union = False
520
521        param = KernRe(r'[\[\)].*').sub('', param, count=1)
522
523        #
524        # Look at various "anonymous type" cases.
525        #
526        if dtype == '':
527            if param.endswith("..."):
528                if len(param) > 3: # there is a name provided, use that
529                    param = param[:-3]
530                if not self.entry.parameterdescs.get(param):
531                    self.entry.parameterdescs[param] = "variable arguments"
532
533            elif (not param) or param == "void":
534                param = "void"
535                self.entry.parameterdescs[param] = "no arguments"
536
537            elif param in ["struct", "union"]:
538                # Handle unnamed (anonymous) union or struct
539                dtype = param
540                param = "{unnamed_" + param + "}"
541                self.entry.parameterdescs[param] = "anonymous\n"
542                self.entry.anon_struct_union = True
543
544        # Warn if parameter has no description
545        # (but ignore ones starting with # as these are not parameters
546        # but inline preprocessor statements)
547        if param not in self.entry.parameterdescs and not param.startswith("#"):
548            self.entry.parameterdescs[param] = self.undescribed
549
550            if "." not in param:
551                if decl_type == 'function':
552                    dname = f"{decl_type} parameter"
553                else:
554                    dname = f"{decl_type} member"
555
556                self.emit_msg(ln,
557                              f"{dname} '{param}' not described in '{declaration_name}'")
558
559        # Strip spaces from param so that it is one continuous string on
560        # parameterlist. This fixes a problem where check_sections()
561        # cannot find a parameter like "addr[6 + 2]" because it actually
562        # appears as "addr[6", "+", "2]" on the parameter list.
563        # However, it's better to maintain the param string unchanged for
564        # output, so just weaken the string compare in check_sections()
565        # to ignore "[blah" in a parameter string.
566
567        self.entry.parameterlist.append(param)
568        org_arg = KernRe(r'\s\s+').sub(' ', org_arg)
569        self.entry.parametertypes[param] = org_arg
570
571
572    def create_parameter_list(self, ln, decl_type, args,
573                              splitter, declaration_name):
574        """
575        Creates a list of parameters, storing them at self.entry.
576        """
577
578        # temporarily replace all commas inside function pointer definition
579        arg_expr = KernRe(r'(\([^\),]+),')
580        while arg_expr.search(args):
581            args = arg_expr.sub(r"\1#", args)
582
583        for arg in args.split(splitter):
584            # Ignore argument attributes
585            arg = KernRe(r'\sPOS0?\s').sub(' ', arg)
586
587            # Strip leading/trailing spaces
588            arg = arg.strip()
589            arg = KernRe(r'\s+').sub(' ', arg, count=1)
590
591            if arg.startswith('#'):
592                # Treat preprocessor directive as a typeless variable just to fill
593                # corresponding data structures "correctly". Catch it later in
594                # output_* subs.
595
596                # Treat preprocessor directive as a typeless variable
597                self.push_parameter(ln, decl_type, arg, "",
598                                    "", declaration_name)
599            #
600            # The pointer-to-function case.
601            #
602            elif KernRe(r'\(.+\)\s*\(').search(arg):
603                arg = arg.replace('#', ',')
604                r = KernRe(r'[^\(]+\(\*?\s*'  # Everything up to "(*"
605                           r'([\w\[\].]*)'    # Capture the name and possible [array]
606                           r'\s*\)')	      # Make sure the trailing ")" is there
607                if r.match(arg):
608                    param = r.group(1)
609                else:
610                    self.emit_msg(ln, f"Invalid param: {arg}")
611                    param = arg
612                dtype = arg.replace(param, '')
613                self.push_parameter(ln, decl_type, param, dtype, arg, declaration_name)
614            #
615            # The array-of-pointers case.  Dig the parameter name out from the middle
616            # of the declaration.
617            #
618            elif KernRe(r'\(.+\)\s*\[').search(arg):
619                r = KernRe(r'[^\(]+\(\s*\*\s*'		# Up to "(" and maybe "*"
620                           r'([\w.]*?)'			# The actual pointer name
621                           r'\s*(\[\s*\w+\s*\]\s*)*\)') # The [array portion]
622                if r.match(arg):
623                    param = r.group(1)
624                else:
625                    self.emit_msg(ln, f"Invalid param: {arg}")
626                    param = arg
627                dtype = arg.replace(param, '')
628                self.push_parameter(ln, decl_type, param, dtype, arg, declaration_name)
629            elif arg:
630                #
631                # Clean up extraneous spaces and split the string at commas; the first
632                # element of the resulting list will also include the type information.
633                #
634                arg = KernRe(r'\s*:\s*').sub(":", arg)
635                arg = KernRe(r'\s*\[').sub('[', arg)
636                args = KernRe(r'\s*,\s*').split(arg)
637                args[0] = re.sub(r'(\*+)\s*', r' \1', args[0])
638                #
639                # args[0] has a string of "type a".  If "a" includes an [array]
640                # declaration, we want to not be fooled by any white space inside
641                # the brackets, so detect and handle that case specially.
642                #
643                r = KernRe(r'^([^[\]]*\s+)(.*)$')
644                if r.match(args[0]):
645                    args[0] = r.group(2)
646                    dtype = r.group(1)
647                else:
648                    # No space in args[0]; this seems wrong but preserves previous behavior
649                    dtype = ''
650
651                bitfield_re = KernRe(r'(.*?):(\w+)')
652                for param in args:
653                    #
654                    # For pointers, shift the star(s) from the variable name to the
655                    # type declaration.
656                    #
657                    r = KernRe(r'^(\*+)\s*(.*)')
658                    if r.match(param):
659                        self.push_parameter(ln, decl_type, r.group(2),
660                                            f"{dtype} {r.group(1)}",
661                                            arg, declaration_name)
662                    #
663                    # Perform a similar shift for bitfields.
664                    #
665                    elif bitfield_re.search(param):
666                        if dtype != "":  # Skip unnamed bit-fields
667                            self.push_parameter(ln, decl_type, bitfield_re.group(1),
668                                                f"{dtype}:{bitfield_re.group(2)}",
669                                                arg, declaration_name)
670                    else:
671                        self.push_parameter(ln, decl_type, param, dtype,
672                                            arg, declaration_name)
673
674    def check_sections(self, ln, decl_name, decl_type):
675        """
676        Check for errors inside sections, emitting warnings if not found
677        parameters are described.
678        """
679        for section in self.entry.sections:
680            if section not in self.entry.parameterlist and \
681               not known_sections.search(section):
682                if decl_type == 'function':
683                    dname = f"{decl_type} parameter"
684                else:
685                    dname = f"{decl_type} member"
686                self.emit_msg(ln,
687                              f"Excess {dname} '{section}' description in '{decl_name}'")
688
689    def check_return_section(self, ln, declaration_name, return_type):
690        """
691        If the function doesn't return void, warns about the lack of a
692        return description.
693        """
694
695        if not self.config.wreturn:
696            return
697
698        # Ignore an empty return type (It's a macro)
699        # Ignore functions with a "void" return type (but not "void *")
700        if not return_type or KernRe(r'void\s*\w*\s*$').search(return_type):
701            return
702
703        if not self.entry.sections.get("Return", None):
704            self.emit_msg(ln,
705                          f"No description found for return value of '{declaration_name}'")
706
707    def split_struct_proto(self, proto):
708        """
709        Split apart a structure prototype; returns (struct|union, name,
710        members) or ``None``.
711        """
712
713        type_pattern = r'(struct|union)'
714        qualifiers = [
715            "__attribute__",
716            "__packed",
717            "__aligned",
718            "____cacheline_aligned_in_smp",
719            "____cacheline_aligned",
720        ]
721        definition_body = r'\{(.*)\}\s*' + "(?:" + '|'.join(qualifiers) + ")?"
722
723        r = KernRe(type_pattern + r'\s+(\w+)\s*' + definition_body)
724        if r.search(proto):
725            return (r.group(1), r.group(2), r.group(3))
726        else:
727            r = KernRe(r'typedef\s+' + type_pattern + r'\s*' + definition_body + r'\s*(\w+)\s*;')
728            if r.search(proto):
729                return (r.group(1), r.group(3), r.group(2))
730        return None
731
732    def rewrite_struct_members(self, members):
733        """
734        Process ``struct``/``union`` members from the most deeply nested
735        outward.
736
737        Rewrite the members of a ``struct`` or ``union`` for easier formatting
738        later on. Among other things, this function will turn a member like::
739
740          struct { inner_members; } foo;
741
742        into::
743
744          struct foo; inner_members;
745        """
746
747        #
748        # The trick is in the ``^{`` below - it prevents a match of an outer
749        # ``struct``/``union`` until the inner one has been munged
750        # (removing the ``{`` in the process).
751        #
752        struct_members = KernRe(r'(struct|union)'   # 0: declaration type
753                                r'([^\{\};]+)' 	    # 1: possible name
754                                r'(\{)'
755                                r'([^\{\}]*)'       # 3: Contents of declaration
756                                r'(\})'
757                                r'([^\{\};]*)(;)')  # 5: Remaining stuff after declaration
758        tuples = struct_members.findall(members)
759        while tuples:
760            for t in tuples:
761                newmember = ""
762                oldmember = "".join(t) # Reconstruct the original formatting
763                dtype, name, lbr, content, rbr, rest, semi = t
764                #
765                # Pass through each field name, normalizing the form and formatting.
766                #
767                for s_id in rest.split(','):
768                    s_id = s_id.strip()
769                    newmember += f"{dtype} {s_id}; "
770                    #
771                    # Remove bitfield/array/pointer info, getting the bare name.
772                    #
773                    s_id = KernRe(r'[:\[].*').sub('', s_id)
774                    s_id = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', s_id)
775                    #
776                    # Pass through the members of this inner structure/union.
777                    #
778                    for arg in content.split(';'):
779                        arg = arg.strip()
780                        #
781                        # Look for (type)(*name)(args) - pointer to function
782                        #
783                        r = KernRe(r'^([^\(]+\(\*?\s*)([\w.]*)(\s*\).*)')
784                        if r.match(arg):
785                            dtype, name, extra = r.group(1), r.group(2), r.group(3)
786                            # Pointer-to-function
787                            if not s_id:
788                                # Anonymous struct/union
789                                newmember += f"{dtype}{name}{extra}; "
790                            else:
791                                newmember += f"{dtype}{s_id}.{name}{extra}; "
792                        #
793                        # Otherwise a non-function member.
794                        #
795                        else:
796                            #
797                            # Remove bitmap and array portions and spaces around commas
798                            #
799                            arg = KernRe(r':\s*\d+\s*').sub('', arg)
800                            arg = KernRe(r'\[.*\]').sub('', arg)
801                            arg = KernRe(r'\s*,\s*').sub(',', arg)
802                            #
803                            # Look for a normal decl - "type name[,name...]"
804                            #
805                            r = KernRe(r'(.*)\s+([\S+,]+)')
806                            if r.search(arg):
807                                for name in r.group(2).split(','):
808                                    name = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', name)
809                                    if not s_id:
810                                        # Anonymous struct/union
811                                        newmember += f"{r.group(1)} {name}; "
812                                    else:
813                                        newmember += f"{r.group(1)} {s_id}.{name}; "
814                            else:
815                                newmember += f"{arg}; "
816                #
817                # At the end of the s_id loop, replace the original declaration with
818                # the munged version.
819                #
820                members = members.replace(oldmember, newmember)
821            #
822            # End of the tuple loop - search again and see if there are outer members
823            # that now turn up.
824            #
825            tuples = struct_members.findall(members)
826        return members
827
828    def format_struct_decl(self, declaration):
829        """
830        Format the ``struct`` declaration into a standard form for inclusion
831        in the resulting docs.
832        """
833
834        #
835        # Insert newlines, get rid of extra spaces.
836        #
837        declaration = KernRe(r'([\{;])').sub(r'\1\n', declaration)
838        declaration = KernRe(r'\}\s+;').sub('};', declaration)
839        #
840        # Format inline enums with each member on its own line.
841        #
842        r = KernRe(r'(enum\s+\{[^\}]+),([^\n])')
843        while r.search(declaration):
844            declaration = r.sub(r'\1,\n\2', declaration)
845        #
846        # Now go through and supply the right number of tabs
847        # for each line.
848        #
849        def_args = declaration.split('\n')
850        level = 1
851        declaration = ""
852        for clause in def_args:
853            clause = KernRe(r'\s+').sub(' ', clause.strip(), count=1)
854            if clause:
855                if '}' in clause and level > 1:
856                    level -= 1
857                if not clause.startswith('#'):
858                    declaration += "\t" * level
859                declaration += "\t" + clause + "\n"
860                if "{" in clause and "}" not in clause:
861                    level += 1
862        return declaration
863
864
865    def dump_struct(self, ln, proto):
866        """
867        Store an entry for a ``struct`` or ``union``
868        """
869        #
870        # Do the basic parse to get the pieces of the declaration.
871        #
872        struct_parts = self.split_struct_proto(proto)
873        if not struct_parts:
874            self.emit_msg(ln, f"{proto} error: Cannot parse struct or union!")
875            return
876        decl_type, declaration_name, members = struct_parts
877
878        if self.entry.identifier != declaration_name:
879            self.emit_msg(ln, f"expecting prototype for {decl_type} {self.entry.identifier}. "
880                          f"Prototype was for {decl_type} {declaration_name} instead\n")
881            return
882        #
883        # Go through the list of members applying all of our transformations.
884        #
885        members = trim_private_members(members)
886        members = apply_transforms(struct_xforms, members)
887
888        nested = NestedMatch()
889        for search, sub in struct_nested_prefixes:
890            members = nested.sub(search, sub, members)
891        #
892        # Deal with embedded struct and union members, and drop enums entirely.
893        #
894        declaration = members
895        members = self.rewrite_struct_members(members)
896        members = re.sub(r'(\{[^\{\}]*\})', '', members)
897        #
898        # Output the result and we are done.
899        #
900        self.create_parameter_list(ln, decl_type, members, ';',
901                                   declaration_name)
902        self.check_sections(ln, declaration_name, decl_type)
903        self.output_declaration(decl_type, declaration_name,
904                                definition=self.format_struct_decl(declaration),
905                                purpose=self.entry.declaration_purpose)
906
907    def dump_enum(self, ln, proto):
908        """
909        Store an ``enum`` inside self.entries array.
910        """
911        #
912        # Strip preprocessor directives.  Note that this depends on the
913        # trailing semicolon we added in process_proto_type().
914        #
915        proto = KernRe(r'#\s*((define|ifdef|if)\s+|endif)[^;]*;', flags=re.S).sub('', proto)
916        #
917        # Parse out the name and members of the enum.  Typedef form first.
918        #
919        r = KernRe(r'typedef\s+enum\s*\{(.*)\}\s*(\w*)\s*;')
920        if r.search(proto):
921            declaration_name = r.group(2)
922            members = trim_private_members(r.group(1))
923        #
924        # Failing that, look for a straight enum
925        #
926        else:
927            r = KernRe(r'enum\s+(\w*)\s*\{(.*)\}')
928            if r.match(proto):
929                declaration_name = r.group(1)
930                members = trim_private_members(r.group(2))
931        #
932        # OK, this isn't going to work.
933        #
934            else:
935                self.emit_msg(ln, f"{proto}: error: Cannot parse enum!")
936                return
937        #
938        # Make sure we found what we were expecting.
939        #
940        if self.entry.identifier != declaration_name:
941            if self.entry.identifier == "":
942                self.emit_msg(ln,
943                              f"{proto}: wrong kernel-doc identifier on prototype")
944            else:
945                self.emit_msg(ln,
946                              f"expecting prototype for enum {self.entry.identifier}. "
947                              f"Prototype was for enum {declaration_name} instead")
948            return
949
950        if not declaration_name:
951            declaration_name = "(anonymous)"
952        #
953        # Parse out the name of each enum member, and verify that we
954        # have a description for it.
955        #
956        member_set = set()
957        members = KernRe(r'\([^;)]*\)').sub('', members)
958        for arg in members.split(','):
959            if not arg:
960                continue
961            arg = KernRe(r'^\s*(\w+).*').sub(r'\1', arg)
962            self.entry.parameterlist.append(arg)
963            if arg not in self.entry.parameterdescs:
964                self.entry.parameterdescs[arg] = self.undescribed
965                self.emit_msg(ln,
966                              f"Enum value '{arg}' not described in enum '{declaration_name}'")
967            member_set.add(arg)
968        #
969        # Ensure that every described member actually exists in the enum.
970        #
971        for k in self.entry.parameterdescs:
972            if k not in member_set:
973                self.emit_msg(ln,
974                              f"Excess enum value '@{k}' description in '{declaration_name}'")
975
976        self.output_declaration('enum', declaration_name,
977                                purpose=self.entry.declaration_purpose)
978
979    def dump_var(self, ln, proto):
980        """
981        Store variables that are part of kAPI.
982        """
983        VAR_ATTRIBS = [
984            "extern",
985            "const",
986        ]
987        OPTIONAL_VAR_ATTR = r"^(?:\b(?:" +"|".join(VAR_ATTRIBS) +r")\b\s*)*"
988
989        #
990        # Store the full prototype before modifying it
991        #
992        full_proto = proto
993        declaration_name = None
994
995        #
996        # Handle macro definitions
997        #
998        macro_prefixes = [
999            KernRe(r"DEFINE_[\w_]+\s*\(([\w_]+)\)"),
1000        ]
1001
1002        for r in macro_prefixes:
1003            match = r.search(proto)
1004            if match:
1005                declaration_name = match.group(1)
1006                break
1007
1008        #
1009        # Drop comments and macros to have a pure C prototype
1010        #
1011        if not declaration_name:
1012            for r, sub in var_xforms:
1013                proto = r.sub(sub, proto)
1014
1015        proto = proto.rstrip()
1016
1017        #
1018        # Variable name is at the end of the declaration
1019        #
1020
1021        default_val = None
1022
1023        r= KernRe(OPTIONAL_VAR_ATTR + r"\s*[\w_\s]*\s+(?:\*+)?([\w_]+)\s*[\d\]\[]*\s*(=.*)?")
1024        if r.match(proto):
1025            if not declaration_name:
1026                declaration_name = r.group(1)
1027
1028            default_val = r.group(2)
1029        else:
1030            r= KernRe(OPTIONAL_VAR_ATTR + r"(?:[\w_\s]*)?\s+(?:\*+)?(?:[\w_]+)\s*[\d\]\[]*\s*(=.*)?")
1031
1032            if r.match(proto):
1033                default_val = r.group(1)
1034        if not declaration_name:
1035           self.emit_msg(ln,f"{proto}: can't parse variable")
1036           return
1037
1038        if default_val:
1039            default_val = default_val.lstrip("=").strip()
1040
1041        self.output_declaration("var", declaration_name,
1042                                full_proto=full_proto,
1043                                default_val=default_val,
1044                                purpose=self.entry.declaration_purpose)
1045
1046    def dump_declaration(self, ln, prototype):
1047        """
1048        Store a data declaration inside self.entries array.
1049        """
1050
1051        if self.entry.decl_type == "enum":
1052            self.dump_enum(ln, prototype)
1053        elif self.entry.decl_type == "typedef":
1054            self.dump_typedef(ln, prototype)
1055        elif self.entry.decl_type in ["union", "struct"]:
1056            self.dump_struct(ln, prototype)
1057        elif self.entry.decl_type == "var":
1058            self.dump_var(ln, prototype)
1059        else:
1060            # This would be a bug
1061            self.emit_message(ln, f'Unknown declaration type: {self.entry.decl_type}')
1062
1063    def dump_function(self, ln, prototype):
1064        """
1065        Store a function or function macro inside self.entries array.
1066        """
1067
1068        found = func_macro = False
1069        return_type = ''
1070        decl_type = 'function'
1071
1072        #
1073        # If we have a macro, remove the "#define" at the front.
1074        #
1075        new_proto = KernRe(r"^#\s*define\s+").sub("", prototype)
1076        if new_proto != prototype:
1077            prototype = new_proto
1078            #
1079            # Dispense with the simple "#define A B" case here; the key
1080            # is the space after the name of the symbol being defined.
1081            # NOTE that the seemingly misnamed "func_macro" indicates a
1082            # macro *without* arguments.
1083            #
1084            r = KernRe(r'^(\w+)\s+')
1085            if r.search(prototype):
1086                return_type = ''
1087                declaration_name = r.group(1)
1088                func_macro = True
1089                found = True
1090        else:
1091            #
1092            # Apply the initial transformations.
1093            #
1094            prototype = apply_transforms(function_xforms, prototype)
1095
1096        # Yes, this truly is vile.  We are looking for:
1097        # 1. Return type (may be nothing if we're looking at a macro)
1098        # 2. Function name
1099        # 3. Function parameters.
1100        #
1101        # All the while we have to watch out for function pointer parameters
1102        # (which IIRC is what the two sections are for), C types (these
1103        # regexps don't even start to express all the possibilities), and
1104        # so on.
1105        #
1106        # If you mess with these regexps, it's a good idea to check that
1107        # the following functions' documentation still comes out right:
1108        # - parport_register_device (function pointer parameters)
1109        # - atomic_set (macro)
1110        # - pci_match_device, __copy_to_user (long return type)
1111
1112        name = r'\w+'
1113        type1 = r'(?:[\w\s]+)?'
1114        type2 = r'(?:[\w\s]+\*+)+'
1115        #
1116        # Attempt to match first on (args) with no internal parentheses; this
1117        # lets us easily filter out __acquires() and other post-args stuff.  If
1118        # that fails, just grab the rest of the line to the last closing
1119        # parenthesis.
1120        #
1121        proto_args = r'\(([^\(]*|.*)\)'
1122        #
1123        # (Except for the simple macro case) attempt to split up the prototype
1124        # in the various ways we understand.
1125        #
1126        if not found:
1127            patterns = [
1128                rf'^()({name})\s*{proto_args}',
1129                rf'^({type1})\s+({name})\s*{proto_args}',
1130                rf'^({type2})\s*({name})\s*{proto_args}',
1131            ]
1132
1133            for p in patterns:
1134                r = KernRe(p)
1135                if r.match(prototype):
1136                    return_type = r.group(1)
1137                    declaration_name = r.group(2)
1138                    args = r.group(3)
1139                    self.create_parameter_list(ln, decl_type, args, ',',
1140                                               declaration_name)
1141                    found = True
1142                    break
1143        #
1144        # Parsing done; make sure that things are as we expect.
1145        #
1146        if not found:
1147            self.emit_msg(ln,
1148                          f"cannot understand function prototype: '{prototype}'")
1149            return
1150        if self.entry.identifier != declaration_name:
1151            self.emit_msg(ln, f"expecting prototype for {self.entry.identifier}(). "
1152                          f"Prototype was for {declaration_name}() instead")
1153            return
1154        self.check_sections(ln, declaration_name, "function")
1155        self.check_return_section(ln, declaration_name, return_type)
1156        #
1157        # Store the result.
1158        #
1159        self.output_declaration(decl_type, declaration_name,
1160                                typedef=('typedef' in return_type),
1161                                functiontype=return_type,
1162                                purpose=self.entry.declaration_purpose,
1163                                func_macro=func_macro)
1164
1165
1166    def dump_typedef(self, ln, proto):
1167        """
1168        Store a ``typedef`` inside self.entries array.
1169        """
1170        #
1171        # We start by looking for function typedefs.
1172        #
1173        typedef_type = r'typedef((?:\s+[\w*]+\b){0,7}\s+(?:\w+\b|\*+))\s*'
1174        typedef_ident = r'\*?\s*(\w\S+)\s*'
1175        typedef_args = r'\s*\((.*)\);'
1176
1177        typedef1 = KernRe(typedef_type + r'\(' + typedef_ident + r'\)' + typedef_args)
1178        typedef2 = KernRe(typedef_type + typedef_ident + typedef_args)
1179
1180        # Parse function typedef prototypes
1181        for r in [typedef1, typedef2]:
1182            if not r.match(proto):
1183                continue
1184
1185            return_type = r.group(1).strip()
1186            declaration_name = r.group(2)
1187            args = r.group(3)
1188
1189            if self.entry.identifier != declaration_name:
1190                self.emit_msg(ln,
1191                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1192                return
1193
1194            self.create_parameter_list(ln, 'function', args, ',', declaration_name)
1195
1196            self.output_declaration('function', declaration_name,
1197                                    typedef=True,
1198                                    functiontype=return_type,
1199                                    purpose=self.entry.declaration_purpose)
1200            return
1201        #
1202        # Not a function, try to parse a simple typedef.
1203        #
1204        r = KernRe(r'typedef.*\s+(\w+)\s*;')
1205        if r.match(proto):
1206            declaration_name = r.group(1)
1207
1208            if self.entry.identifier != declaration_name:
1209                self.emit_msg(ln,
1210                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1211                return
1212
1213            self.output_declaration('typedef', declaration_name,
1214                                    purpose=self.entry.declaration_purpose)
1215            return
1216
1217        self.emit_msg(ln, "error: Cannot parse typedef!")
1218
1219    @staticmethod
1220    def process_export(function_set, line):
1221        """
1222        process ``EXPORT_SYMBOL*`` tags
1223
1224        This method doesn't use any variable from the class, so declare it
1225        with a staticmethod decorator.
1226        """
1227
1228        # We support documenting some exported symbols with different
1229        # names.  A horrible hack.
1230        suffixes = [ '_noprof' ]
1231
1232        # Note: it accepts only one EXPORT_SYMBOL* per line, as having
1233        # multiple export lines would violate Kernel coding style.
1234
1235        if export_symbol.search(line):
1236            symbol = export_symbol.group(2)
1237        elif export_symbol_ns.search(line):
1238            symbol = export_symbol_ns.group(2)
1239        else:
1240            return False
1241        #
1242        # Found an export, trim out any special suffixes
1243        #
1244        for suffix in suffixes:
1245            # Be backward compatible with Python < 3.9
1246            if symbol.endswith(suffix):
1247                symbol = symbol[:-len(suffix)]
1248        function_set.add(symbol)
1249        return True
1250
1251    def process_normal(self, ln, line):
1252        """
1253        STATE_NORMAL: looking for the ``/**`` to begin everything.
1254        """
1255
1256        if not doc_start.match(line):
1257            return
1258
1259        # start a new entry
1260        self.reset_state(ln)
1261
1262        # next line is always the function name
1263        self.state = state.NAME
1264
1265    def process_name(self, ln, line):
1266        """
1267        STATE_NAME: Looking for the "name - description" line
1268        """
1269        #
1270        # Check for a DOC: block and handle them specially.
1271        #
1272        if doc_block.search(line):
1273
1274            if not doc_block.group(1):
1275                self.entry.begin_section(ln, "Introduction")
1276            else:
1277                self.entry.begin_section(ln, doc_block.group(1))
1278
1279            self.entry.identifier = self.entry.section
1280            self.state = state.DOCBLOCK
1281        #
1282        # Otherwise we're looking for a normal kerneldoc declaration line.
1283        #
1284        elif doc_decl.search(line):
1285            self.entry.identifier = doc_decl.group(1)
1286
1287            # Test for data declaration
1288            if doc_begin_data.search(line):
1289                self.entry.decl_type = doc_begin_data.group(1)
1290                self.entry.identifier = doc_begin_data.group(2)
1291            #
1292            # Look for a function description
1293            #
1294            elif doc_begin_func.search(line):
1295                self.entry.identifier = doc_begin_func.group(1)
1296                self.entry.decl_type = "function"
1297            #
1298            # We struck out.
1299            #
1300            else:
1301                self.emit_msg(ln,
1302                              f"This comment starts with '/**', but isn't a kernel-doc comment. Refer to Documentation/doc-guide/kernel-doc.rst\n{line}")
1303                self.state = state.NORMAL
1304                return
1305            #
1306            # OK, set up for a new kerneldoc entry.
1307            #
1308            self.state = state.BODY
1309            self.entry.identifier = self.entry.identifier.strip(" ")
1310            # if there's no @param blocks need to set up default section here
1311            self.entry.begin_section(ln + 1)
1312            #
1313            # Find the description portion, which *should* be there but
1314            # isn't always.
1315            # (We should be able to capture this from the previous parsing - someday)
1316            #
1317            r = KernRe("[-:](.*)")
1318            if r.search(line):
1319                self.entry.declaration_purpose = trim_whitespace(r.group(1))
1320                self.state = state.DECLARATION
1321            else:
1322                self.entry.declaration_purpose = ""
1323
1324            if not self.entry.declaration_purpose and self.config.wshort_desc:
1325                self.emit_msg(ln,
1326                              f"missing initial short description on line:\n{line}")
1327
1328            if not self.entry.identifier and self.entry.decl_type != "enum":
1329                self.emit_msg(ln,
1330                              f"wrong kernel-doc identifier on line:\n{line}")
1331                self.state = state.NORMAL
1332
1333            if self.config.verbose:
1334                self.emit_msg(ln,
1335                              f"Scanning doc for {self.entry.decl_type} {self.entry.identifier}",
1336                                  warning=False)
1337        #
1338        # Failed to find an identifier. Emit a warning
1339        #
1340        else:
1341            self.emit_msg(ln, f"Cannot find identifier on line:\n{line}")
1342
1343    def is_new_section(self, ln, line):
1344        """
1345        Helper function to determine if a new section is being started.
1346        """
1347        if doc_sect.search(line):
1348            self.state = state.BODY
1349            #
1350            # Pick out the name of our new section, tweaking it if need be.
1351            #
1352            newsection = doc_sect.group(1)
1353            if newsection.lower() == 'description':
1354                newsection = 'Description'
1355            elif newsection.lower() == 'context':
1356                newsection = 'Context'
1357                self.state = state.SPECIAL_SECTION
1358            elif newsection.lower() in ["@return", "@returns",
1359                                        "return", "returns"]:
1360                newsection = "Return"
1361                self.state = state.SPECIAL_SECTION
1362            elif newsection[0] == '@':
1363                self.state = state.SPECIAL_SECTION
1364            #
1365            # Initialize the contents, and get the new section going.
1366            #
1367            newcontents = doc_sect.group(2)
1368            if not newcontents:
1369                newcontents = ""
1370            self.dump_section()
1371            self.entry.begin_section(ln, newsection)
1372            self.entry.leading_space = None
1373
1374            self.entry.add_text(newcontents.lstrip())
1375            return True
1376        return False
1377
1378    def is_comment_end(self, ln, line):
1379        """
1380        Helper function to detect (and effect) the end of a kerneldoc comment.
1381        """
1382        if doc_end.search(line):
1383            self.dump_section()
1384
1385            # Look for doc_com + <text> + doc_end:
1386            r = KernRe(r'\s*\*\s*[a-zA-Z_0-9:.]+\*/')
1387            if r.match(line):
1388                self.emit_msg(ln, f"suspicious ending line: {line}")
1389
1390            self.entry.prototype = ""
1391            self.entry.new_start_line = ln + 1
1392
1393            self.state = state.PROTO
1394            return True
1395        return False
1396
1397
1398    def process_decl(self, ln, line):
1399        """
1400        STATE_DECLARATION: We've seen the beginning of a declaration.
1401        """
1402        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1403            return
1404        #
1405        # Look for anything with the " * " line beginning.
1406        #
1407        if doc_content.search(line):
1408            cont = doc_content.group(1)
1409            #
1410            # A blank line means that we have moved out of the declaration
1411            # part of the comment (without any "special section" parameter
1412            # descriptions).
1413            #
1414            if cont == "":
1415                self.state = state.BODY
1416            #
1417            # Otherwise we have more of the declaration section to soak up.
1418            #
1419            else:
1420                self.entry.declaration_purpose = \
1421                    trim_whitespace(self.entry.declaration_purpose + ' ' + cont)
1422        else:
1423            # Unknown line, ignore
1424            self.emit_msg(ln, f"bad line: {line}")
1425
1426
1427    def process_special(self, ln, line):
1428        """
1429        STATE_SPECIAL_SECTION: a section ending with a blank line.
1430        """
1431        #
1432        # If we have hit a blank line (only the " * " marker), then this
1433        # section is done.
1434        #
1435        if KernRe(r"\s*\*\s*$").match(line):
1436            self.entry.begin_section(ln, dump = True)
1437            self.state = state.BODY
1438            return
1439        #
1440        # Not a blank line, look for the other ways to end the section.
1441        #
1442        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1443            return
1444        #
1445        # OK, we should have a continuation of the text for this section.
1446        #
1447        if doc_content.search(line):
1448            cont = doc_content.group(1)
1449            #
1450            # If the lines of text after the first in a special section have
1451            # leading white space, we need to trim it out or Sphinx will get
1452            # confused.  For the second line (the None case), see what we
1453            # find there and remember it.
1454            #
1455            if self.entry.leading_space is None:
1456                r = KernRe(r'^(\s+)')
1457                if r.match(cont):
1458                    self.entry.leading_space = len(r.group(1))
1459                else:
1460                    self.entry.leading_space = 0
1461            #
1462            # Otherwise, before trimming any leading chars, be *sure*
1463            # that they are white space.  We should maybe warn if this
1464            # isn't the case.
1465            #
1466            for i in range(0, self.entry.leading_space):
1467                if cont[i] != " ":
1468                    self.entry.leading_space = i
1469                    break
1470            #
1471            # Add the trimmed result to the section and we're done.
1472            #
1473            self.entry.add_text(cont[self.entry.leading_space:])
1474        else:
1475            # Unknown line, ignore
1476            self.emit_msg(ln, f"bad line: {line}")
1477
1478    def process_body(self, ln, line):
1479        """
1480        STATE_BODY: the bulk of a kerneldoc comment.
1481        """
1482        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1483            return
1484
1485        if doc_content.search(line):
1486            cont = doc_content.group(1)
1487            self.entry.add_text(cont)
1488        else:
1489            # Unknown line, ignore
1490            self.emit_msg(ln, f"bad line: {line}")
1491
1492    def process_inline_name(self, ln, line):
1493        """STATE_INLINE_NAME: beginning of docbook comments within a prototype."""
1494
1495        if doc_inline_sect.search(line):
1496            self.entry.begin_section(ln, doc_inline_sect.group(1))
1497            self.entry.add_text(doc_inline_sect.group(2).lstrip())
1498            self.state = state.INLINE_TEXT
1499        elif doc_inline_end.search(line):
1500            self.dump_section()
1501            self.state = state.PROTO
1502        elif doc_content.search(line):
1503            self.emit_msg(ln, f"Incorrect use of kernel-doc format: {line}")
1504            self.state = state.PROTO
1505        # else ... ??
1506
1507    def process_inline_text(self, ln, line):
1508        """STATE_INLINE_TEXT: docbook comments within a prototype."""
1509
1510        if doc_inline_end.search(line):
1511            self.dump_section()
1512            self.state = state.PROTO
1513        elif doc_content.search(line):
1514            self.entry.add_text(doc_content.group(1))
1515        # else ... ??
1516
1517    def syscall_munge(self, ln, proto):         # pylint: disable=W0613
1518        """
1519        Handle syscall definitions.
1520        """
1521
1522        is_void = False
1523
1524        # Strip newlines/CR's
1525        proto = re.sub(r'[\r\n]+', ' ', proto)
1526
1527        # Check if it's a SYSCALL_DEFINE0
1528        if 'SYSCALL_DEFINE0' in proto:
1529            is_void = True
1530
1531        # Replace SYSCALL_DEFINE with correct return type & function name
1532        proto = KernRe(r'SYSCALL_DEFINE.*\(').sub('long sys_', proto)
1533
1534        r = KernRe(r'long\s+(sys_.*?),')
1535        if r.search(proto):
1536            proto = KernRe(',').sub('(', proto, count=1)
1537        elif is_void:
1538            proto = KernRe(r'\)').sub('(void)', proto, count=1)
1539
1540        # Now delete all of the odd-numbered commas in the proto
1541        # so that argument types & names don't have a comma between them
1542        count = 0
1543        length = len(proto)
1544
1545        if is_void:
1546            length = 0  # skip the loop if is_void
1547
1548        for ix in range(length):
1549            if proto[ix] == ',':
1550                count += 1
1551                if count % 2 == 1:
1552                    proto = proto[:ix] + ' ' + proto[ix + 1:]
1553
1554        return proto
1555
1556    def tracepoint_munge(self, ln, proto):
1557        """
1558        Handle tracepoint definitions.
1559        """
1560
1561        tracepointname = None
1562        tracepointargs = None
1563
1564        # Match tracepoint name based on different patterns
1565        r = KernRe(r'TRACE_EVENT\((.*?),')
1566        if r.search(proto):
1567            tracepointname = r.group(1)
1568
1569        r = KernRe(r'DEFINE_SINGLE_EVENT\((.*?),')
1570        if r.search(proto):
1571            tracepointname = r.group(1)
1572
1573        r = KernRe(r'DEFINE_EVENT\((.*?),(.*?),')
1574        if r.search(proto):
1575            tracepointname = r.group(2)
1576
1577        if tracepointname:
1578            tracepointname = tracepointname.lstrip()
1579
1580        r = KernRe(r'TP_PROTO\((.*?)\)')
1581        if r.search(proto):
1582            tracepointargs = r.group(1)
1583
1584        if not tracepointname or not tracepointargs:
1585            self.emit_msg(ln,
1586                          f"Unrecognized tracepoint format:\n{proto}\n")
1587        else:
1588            proto = f"static inline void trace_{tracepointname}({tracepointargs})"
1589            self.entry.identifier = f"trace_{self.entry.identifier}"
1590
1591        return proto
1592
1593    def process_proto_function(self, ln, line):
1594        """Ancillary routine to process a function prototype."""
1595
1596        # strip C99-style comments to end of line
1597        line = KernRe(r"//.*$", re.S).sub('', line)
1598        #
1599        # Soak up the line's worth of prototype text, stopping at { or ; if present.
1600        #
1601        if KernRe(r'\s*#\s*define').match(line):
1602            self.entry.prototype = line
1603        elif not line.startswith('#'):   # skip other preprocessor stuff
1604            r = KernRe(r'([^\{]*)')
1605            if r.match(line):
1606                self.entry.prototype += r.group(1) + " "
1607        #
1608        # If we now have the whole prototype, clean it up and declare victory.
1609        #
1610        if '{' in line or ';' in line or KernRe(r'\s*#\s*define').match(line):
1611            # strip comments and surrounding spaces
1612            self.entry.prototype = KernRe(r'/\*.*\*/').sub('', self.entry.prototype).strip()
1613            #
1614            # Handle self.entry.prototypes for function pointers like:
1615            #       int (*pcs_config)(struct foo)
1616            # by turning it into
1617            #	    int pcs_config(struct foo)
1618            #
1619            r = KernRe(r'^(\S+\s+)\(\s*\*(\S+)\)')
1620            self.entry.prototype = r.sub(r'\1\2', self.entry.prototype)
1621            #
1622            # Handle special declaration syntaxes
1623            #
1624            if 'SYSCALL_DEFINE' in self.entry.prototype:
1625                self.entry.prototype = self.syscall_munge(ln,
1626                                                          self.entry.prototype)
1627            else:
1628                r = KernRe(r'TRACE_EVENT|DEFINE_EVENT|DEFINE_SINGLE_EVENT')
1629                if r.search(self.entry.prototype):
1630                    self.entry.prototype = self.tracepoint_munge(ln,
1631                                                                 self.entry.prototype)
1632            #
1633            # ... and we're done
1634            #
1635            self.dump_function(ln, self.entry.prototype)
1636            self.reset_state(ln)
1637
1638    def process_proto_type(self, ln, line):
1639        """
1640        Ancillary routine to process a type.
1641        """
1642
1643        # Strip C99-style comments and surrounding whitespace
1644        line = KernRe(r"//.*$", re.S).sub('', line).strip()
1645        if not line:
1646            return # nothing to see here
1647
1648        # To distinguish preprocessor directive from regular declaration later.
1649        if line.startswith('#'):
1650            line += ";"
1651        #
1652        # Split the declaration on any of { } or ;, and accumulate pieces
1653        # until we hit a semicolon while not inside {brackets}
1654        #
1655        r = KernRe(r'(.*?)([{};])')
1656        for chunk in r.split(line):
1657            if chunk:  # Ignore empty matches
1658                self.entry.prototype += chunk
1659                #
1660                # This cries out for a match statement ... someday after we can
1661                # drop Python 3.9 ...
1662                #
1663                if chunk == '{':
1664                    self.entry.brcount += 1
1665                elif chunk == '}':
1666                    self.entry.brcount -= 1
1667                elif chunk == ';' and self.entry.brcount <= 0:
1668                    self.dump_declaration(ln, self.entry.prototype)
1669                    self.reset_state(ln)
1670                    return
1671        #
1672        # We hit the end of the line while still in the declaration; put
1673        # in a space to represent the newline.
1674        #
1675        self.entry.prototype += ' '
1676
1677    def process_proto(self, ln, line):
1678        """STATE_PROTO: reading a function/whatever prototype."""
1679
1680        if doc_inline_oneline.search(line):
1681            self.entry.begin_section(ln, doc_inline_oneline.group(1))
1682            self.entry.add_text(doc_inline_oneline.group(2))
1683            self.dump_section()
1684
1685        elif doc_inline_start.search(line):
1686            self.state = state.INLINE_NAME
1687
1688        elif self.entry.decl_type == 'function':
1689            self.process_proto_function(ln, line)
1690
1691        else:
1692            self.process_proto_type(ln, line)
1693
1694    def process_docblock(self, ln, line):
1695        """STATE_DOCBLOCK: within a ``DOC:`` block."""
1696
1697        if doc_end.search(line):
1698            self.dump_section()
1699            self.output_declaration("doc", self.entry.identifier)
1700            self.reset_state(ln)
1701
1702        elif doc_content.search(line):
1703            self.entry.add_text(doc_content.group(1))
1704
1705    def parse_export(self):
1706        """
1707        Parses ``EXPORT_SYMBOL*`` macros from a single Kernel source file.
1708        """
1709
1710        export_table = set()
1711
1712        try:
1713            with open(self.fname, "r", encoding="utf8",
1714                      errors="backslashreplace") as fp:
1715
1716                for line in fp:
1717                    self.process_export(export_table, line)
1718
1719        except IOError:
1720            return None
1721
1722        return export_table
1723
1724    #: The state/action table telling us which function to invoke in each state.
1725    state_actions = {
1726        state.NORMAL:			process_normal,
1727        state.NAME:			process_name,
1728        state.BODY:			process_body,
1729        state.DECLARATION:		process_decl,
1730        state.SPECIAL_SECTION:		process_special,
1731        state.INLINE_NAME:		process_inline_name,
1732        state.INLINE_TEXT:		process_inline_text,
1733        state.PROTO:			process_proto,
1734        state.DOCBLOCK:			process_docblock,
1735        }
1736
1737    def parse_kdoc(self):
1738        """
1739        Open and process each line of a C source file.
1740        The parsing is controlled via a state machine, and the line is passed
1741        to a different process function depending on the state. The process
1742        function may update the state as needed.
1743
1744        Besides parsing kernel-doc tags, it also parses export symbols.
1745        """
1746
1747        prev = ""
1748        prev_ln = None
1749        export_table = set()
1750
1751        try:
1752            with open(self.fname, "r", encoding="utf8",
1753                      errors="backslashreplace") as fp:
1754                for ln, line in enumerate(fp):
1755
1756                    line = line.expandtabs().strip("\n")
1757
1758                    # Group continuation lines on prototypes
1759                    if self.state == state.PROTO:
1760                        if line.endswith("\\"):
1761                            prev += line.rstrip("\\")
1762                            if not prev_ln:
1763                                prev_ln = ln
1764                            continue
1765
1766                        if prev:
1767                            ln = prev_ln
1768                            line = prev + line
1769                            prev = ""
1770                            prev_ln = None
1771
1772                    self.config.log.debug("%d %s: %s",
1773                                          ln, state.name[self.state],
1774                                          line)
1775
1776                    # This is an optimization over the original script.
1777                    # There, when export_file was used for the same file,
1778                    # it was read twice. Here, we use the already-existing
1779                    # loop to parse exported symbols as well.
1780                    #
1781                    if (self.state != state.NORMAL) or \
1782                       not self.process_export(export_table, line):
1783                        # Hand this line to the appropriate state handler
1784                        self.state_actions[self.state](self, ln, line)
1785
1786            self.emit_unused_warnings()
1787
1788        except OSError:
1789            self.config.log.error(f"Error: Cannot open file {self.fname}")
1790
1791        return export_table, self.entries
1792