xref: /linux/tools/lib/python/kdoc/kdoc_parser.py (revision caa642bf3b767c4b33a62c6e2c2708fafb88ea6c)
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.kdoc_re import NestedMatch, KernRe
20from kdoc.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)
57
58export_symbol = KernRe(r'^\s*EXPORT_SYMBOL(_GPL)?\s*\(\s*(\w+)\s*\)\s*', cache=False)
59export_symbol_ns = KernRe(r'^\s*EXPORT_SYMBOL_NS(_GPL)?\s*\(\s*(\w+)\s*,\s*"\S+"\)\s*', cache=False)
60
61type_param = KernRe(r"@(\w*((\.\w+)|(->\w+))*(\.\.\.)?)", cache=False)
62
63#
64# Tests for the beginning of a kerneldoc block in its various forms.
65#
66doc_block = doc_com + KernRe(r'DOC:\s*(.*)?', cache=False)
67doc_begin_data = KernRe(r"^\s*\*?\s*(struct|union|enum|typedef)\b\s*(\w*)", cache = False)
68doc_begin_func = KernRe(str(doc_com) +			# initial " * '
69                        r"(?:\w+\s*\*\s*)?" + 		# type (not captured)
70                        r'(?:define\s+)?' + 		# possible "define" (not captured)
71                        r'(\w+)\s*(?:\(\w*\))?\s*' +	# name and optional "(...)"
72                        r'(?:[-:].*)?$',		# description (not captured)
73                        cache = False)
74
75#
76# Here begins a long set of transformations to turn structure member prefixes
77# and macro invocations into something we can parse and generate kdoc for.
78#
79struct_args_pattern = r'([^,)]+)'
80
81struct_xforms = [
82    # Strip attributes
83    (KernRe(r"__attribute__\s*\(\([a-z0-9,_\*\s\(\)]*\)\)", flags=re.I | re.S, cache=False), ' '),
84    (KernRe(r'\s*__aligned\s*\([^;]*\)', re.S), ' '),
85    (KernRe(r'\s*__counted_by\s*\([^;]*\)', re.S), ' '),
86    (KernRe(r'\s*__counted_by_(le|be)\s*\([^;]*\)', re.S), ' '),
87    (KernRe(r'\s*__packed\s*', re.S), ' '),
88    (KernRe(r'\s*CRYPTO_MINALIGN_ATTR', re.S), ' '),
89    (KernRe(r'\s*____cacheline_aligned_in_smp', re.S), ' '),
90    (KernRe(r'\s*____cacheline_aligned', re.S), ' '),
91    (KernRe(r'\s*__cacheline_group_(begin|end)\([^\)]+\);'), ''),
92    #
93    # Unwrap struct_group macros based on this definition:
94    # __struct_group(TAG, NAME, ATTRS, MEMBERS...)
95    # which has variants like: struct_group(NAME, MEMBERS...)
96    # Only MEMBERS arguments require documentation.
97    #
98    # Parsing them happens on two steps:
99    #
100    # 1. drop struct group arguments that aren't at MEMBERS,
101    #    storing them as STRUCT_GROUP(MEMBERS)
102    #
103    # 2. remove STRUCT_GROUP() ancillary macro.
104    #
105    # The original logic used to remove STRUCT_GROUP() using an
106    # advanced regex:
107    #
108    #   \bSTRUCT_GROUP(\(((?:(?>[^)(]+)|(?1))*)\))[^;]*;
109    #
110    # with two patterns that are incompatible with
111    # Python re module, as it has:
112    #
113    #   - a recursive pattern: (?1)
114    #   - an atomic grouping: (?>...)
115    #
116    # I tried a simpler version: but it didn't work either:
117    #   \bSTRUCT_GROUP\(([^\)]+)\)[^;]*;
118    #
119    # As it doesn't properly match the end parenthesis on some cases.
120    #
121    # So, a better solution was crafted: there's now a NestedMatch
122    # class that ensures that delimiters after a search are properly
123    # matched. So, the implementation to drop STRUCT_GROUP() will be
124    # handled in separate.
125    #
126    (KernRe(r'\bstruct_group\s*\(([^,]*,)', re.S), r'STRUCT_GROUP('),
127    (KernRe(r'\bstruct_group_attr\s*\(([^,]*,){2}', re.S), r'STRUCT_GROUP('),
128    (KernRe(r'\bstruct_group_tagged\s*\(([^,]*),([^,]*),', re.S), r'struct \1 \2; STRUCT_GROUP('),
129    (KernRe(r'\b__struct_group\s*\(([^,]*,){3}', re.S), r'STRUCT_GROUP('),
130    #
131    # Replace macros
132    #
133    # TODO: use NestedMatch for FOO($1, $2, ...) matches
134    #
135    # it is better to also move those to the NestedMatch logic,
136    # to ensure that parenthesis will be properly matched.
137    #
138    (KernRe(r'__ETHTOOL_DECLARE_LINK_MODE_MASK\s*\(([^\)]+)\)', re.S),
139     r'DECLARE_BITMAP(\1, __ETHTOOL_LINK_MODE_MASK_NBITS)'),
140    (KernRe(r'DECLARE_PHY_INTERFACE_MASK\s*\(([^\)]+)\)', re.S),
141     r'DECLARE_BITMAP(\1, PHY_INTERFACE_MODE_MAX)'),
142    (KernRe(r'DECLARE_BITMAP\s*\(' + struct_args_pattern + r',\s*' + struct_args_pattern + r'\)',
143            re.S), r'unsigned long \1[BITS_TO_LONGS(\2)]'),
144    (KernRe(r'DECLARE_HASHTABLE\s*\(' + struct_args_pattern + r',\s*' + struct_args_pattern + r'\)',
145            re.S), r'unsigned long \1[1 << ((\2) - 1)]'),
146    (KernRe(r'DECLARE_KFIFO\s*\(' + struct_args_pattern + r',\s*' + struct_args_pattern +
147            r',\s*' + struct_args_pattern + r'\)', re.S), r'\2 *\1'),
148    (KernRe(r'DECLARE_KFIFO_PTR\s*\(' + struct_args_pattern + r',\s*' +
149            struct_args_pattern + r'\)', re.S), r'\2 *\1'),
150    (KernRe(r'(?:__)?DECLARE_FLEX_ARRAY\s*\(' + struct_args_pattern + r',\s*' +
151            struct_args_pattern + r'\)', re.S), r'\1 \2[]'),
152    (KernRe(r'DEFINE_DMA_UNMAP_ADDR\s*\(' + struct_args_pattern + r'\)', re.S), r'dma_addr_t \1'),
153    (KernRe(r'DEFINE_DMA_UNMAP_LEN\s*\(' + struct_args_pattern + r'\)', re.S), r'__u32 \1'),
154]
155#
156# Regexes here are guaranteed to have the end limiter 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"__deprecated +"), ""),
180    (KernRe(r"__flatten +"), ""),
181    (KernRe(r"__meminit +"), ""),
182    (KernRe(r"__must_check +"), ""),
183    (KernRe(r"__weak +"), ""),
184    (KernRe(r"__sched +"), ""),
185    (KernRe(r"_noprof"), ""),
186    (KernRe(r"__printf\s*\(\s*\d*\s*,\s*\d*\s*\) +"), ""),
187    (KernRe(r"__(?:re)?alloc_size\s*\(\s*\d+\s*(?:,\s*\d+\s*)?\) +"), ""),
188    (KernRe(r"__diagnose_as\s*\(\s*\S+\s*(?:,\s*\d+\s*)*\) +"), ""),
189    (KernRe(r"DECL_BUCKET_PARAMS\s*\(\s*(\S+)\s*,\s*(\S+)\s*\)"), r"\1, \2"),
190    (KernRe(r"__attribute_const__ +"), ""),
191    (KernRe(r"__attribute__\s*\(\((?:[\w\s]+(?:\([^)]*\))?\s*,?)+\)\)\s+"), ""),
192]
193
194#
195# Apply a set of transforms to a block of text.
196#
197def apply_transforms(xforms, text):
198    for search, subst in xforms:
199        text = search.sub(subst, text)
200    return text
201
202#
203# A little helper to get rid of excess white space
204#
205multi_space = KernRe(r'\s\s+')
206def trim_whitespace(s):
207    return multi_space.sub(' ', s.strip())
208
209#
210# Remove struct/enum members that have been marked "private".
211#
212def trim_private_members(text):
213    #
214    # First look for a "public:" block that ends a private region, then
215    # handle the "private until the end" case.
216    #
217    text = KernRe(r'/\*\s*private:.*?/\*\s*public:.*?\*/', flags=re.S).sub('', text)
218    text = KernRe(r'/\*\s*private:.*', flags=re.S).sub('', text)
219    #
220    # We needed the comments to do the above, but now we can take them out.
221    #
222    return KernRe(r'\s*/\*.*?\*/\s*', flags=re.S).sub('', text).strip()
223
224class state:
225    """
226    State machine enums
227    """
228
229    # Parser states
230    NORMAL        = 0        # normal code
231    NAME          = 1        # looking for function name
232    DECLARATION   = 2        # We have seen a declaration which might not be done
233    BODY          = 3        # the body of the comment
234    SPECIAL_SECTION = 4      # doc section ending with a blank line
235    PROTO         = 5        # scanning prototype
236    DOCBLOCK      = 6        # documentation block
237    INLINE_NAME   = 7        # gathering doc outside main block
238    INLINE_TEXT   = 8	     # reading the body of inline docs
239
240    name = [
241        "NORMAL",
242        "NAME",
243        "DECLARATION",
244        "BODY",
245        "SPECIAL_SECTION",
246        "PROTO",
247        "DOCBLOCK",
248        "INLINE_NAME",
249        "INLINE_TEXT",
250    ]
251
252
253SECTION_DEFAULT = "Description"  # default section
254
255class KernelEntry:
256
257    def __init__(self, config, fname, ln):
258        self.config = config
259        self.fname = fname
260
261        self._contents = []
262        self.prototype = ""
263
264        self.warnings = []
265
266        self.parameterlist = []
267        self.parameterdescs = {}
268        self.parametertypes = {}
269        self.parameterdesc_start_lines = {}
270
271        self.section_start_lines = {}
272        self.sections = {}
273
274        self.anon_struct_union = False
275
276        self.leading_space = None
277
278        self.fname = fname
279
280        # State flags
281        self.brcount = 0
282        self.declaration_start_line = ln + 1
283
284    #
285    # Management of section contents
286    #
287    def add_text(self, text):
288        self._contents.append(text)
289
290    def contents(self):
291        return '\n'.join(self._contents) + '\n'
292
293    # TODO: rename to emit_message after removal of kernel-doc.pl
294    def emit_msg(self, ln, msg, *, warning=True):
295        """Emit a message"""
296
297        log_msg = f"{self.fname}:{ln} {msg}"
298
299        if not warning:
300            self.config.log.info(log_msg)
301            return
302
303        # Delegate warning output to output logic, as this way it
304        # will report warnings/info only for symbols that are output
305
306        self.warnings.append(log_msg)
307        return
308
309    #
310    # Begin a new section.
311    #
312    def begin_section(self, line_no, title = SECTION_DEFAULT, dump = False):
313        if dump:
314            self.dump_section(start_new = True)
315        self.section = title
316        self.new_start_line = line_no
317
318    def dump_section(self, start_new=True):
319        """
320        Dumps section contents to arrays/hashes intended for that purpose.
321        """
322        #
323        # If we have accumulated no contents in the default ("description")
324        # section, don't bother.
325        #
326        if self.section == SECTION_DEFAULT and not self._contents:
327            return
328        name = self.section
329        contents = self.contents()
330
331        if type_param.match(name):
332            name = type_param.group(1)
333
334            self.parameterdescs[name] = contents
335            self.parameterdesc_start_lines[name] = self.new_start_line
336
337            self.new_start_line = 0
338
339        else:
340            if name in self.sections and self.sections[name] != "":
341                # Only warn on user-specified duplicate section names
342                if name != SECTION_DEFAULT:
343                    self.emit_msg(self.new_start_line,
344                                  f"duplicate section name '{name}'")
345                # Treat as a new paragraph - add a blank line
346                self.sections[name] += '\n' + contents
347            else:
348                self.sections[name] = contents
349                self.section_start_lines[name] = self.new_start_line
350                self.new_start_line = 0
351
352#        self.config.log.debug("Section: %s : %s", name, pformat(vars(self)))
353
354        if start_new:
355            self.section = SECTION_DEFAULT
356            self._contents = []
357
358python_warning = False
359
360class KernelDoc:
361    """
362    Read a C language source or header FILE and extract embedded
363    documentation comments.
364    """
365
366    # Section names
367
368    section_context = "Context"
369    section_return = "Return"
370
371    undescribed = "-- undescribed --"
372
373    def __init__(self, config, fname):
374        """Initialize internal variables"""
375
376        self.fname = fname
377        self.config = config
378
379        # Initial state for the state machines
380        self.state = state.NORMAL
381
382        # Store entry currently being processed
383        self.entry = None
384
385        # Place all potential outputs into an array
386        self.entries = []
387
388        #
389        # We need Python 3.7 for its "dicts remember the insertion
390        # order" guarantee
391        #
392        global python_warning
393        if (not python_warning and
394            sys.version_info.major == 3 and sys.version_info.minor < 7):
395
396            self.emit_msg(0,
397                          'Python 3.7 or later is required for correct results')
398            python_warning = True
399
400    def emit_msg(self, ln, msg, *, warning=True):
401        """Emit a message"""
402
403        if self.entry:
404            self.entry.emit_msg(ln, msg, warning=warning)
405            return
406
407        log_msg = f"{self.fname}:{ln} {msg}"
408
409        if warning:
410            self.config.log.warning(log_msg)
411        else:
412            self.config.log.info(log_msg)
413
414    def dump_section(self, start_new=True):
415        """
416        Dumps section contents to arrays/hashes intended for that purpose.
417        """
418
419        if self.entry:
420            self.entry.dump_section(start_new)
421
422    # TODO: rename it to store_declaration after removal of kernel-doc.pl
423    def output_declaration(self, dtype, name, **args):
424        """
425        Stores the entry into an entry array.
426
427        The actual output and output filters will be handled elsewhere
428        """
429
430        item = KdocItem(name, self.fname, dtype,
431                        self.entry.declaration_start_line, **args)
432        item.warnings = self.entry.warnings
433
434        # Drop empty sections
435        # TODO: improve empty sections logic to emit warnings
436        sections = self.entry.sections
437        for section in ["Description", "Return"]:
438            if section in sections and not sections[section].rstrip():
439                del sections[section]
440        item.set_sections(sections, self.entry.section_start_lines)
441        item.set_params(self.entry.parameterlist, self.entry.parameterdescs,
442                        self.entry.parametertypes,
443                        self.entry.parameterdesc_start_lines)
444        self.entries.append(item)
445
446        self.config.log.debug("Output: %s:%s = %s", dtype, name, pformat(args))
447
448    def reset_state(self, ln):
449        """
450        Ancillary routine to create a new entry. It initializes all
451        variables used by the state machine.
452        """
453
454        #
455        # Flush the warnings out before we proceed further
456        #
457        if self.entry and self.entry not in self.entries:
458            for log_msg in self.entry.warnings:
459                self.config.log.warning(log_msg)
460
461        self.entry = KernelEntry(self.config, self.fname, ln)
462
463        # State flags
464        self.state = state.NORMAL
465
466    def push_parameter(self, ln, decl_type, param, dtype,
467                       org_arg, declaration_name):
468        """
469        Store parameters and their descriptions at self.entry.
470        """
471
472        if self.entry.anon_struct_union and dtype == "" and param == "}":
473            return  # Ignore the ending }; from anonymous struct/union
474
475        self.entry.anon_struct_union = False
476
477        param = KernRe(r'[\[\)].*').sub('', param, count=1)
478
479        #
480        # Look at various "anonymous type" cases.
481        #
482        if dtype == '':
483            if param.endswith("..."):
484                if len(param) > 3: # there is a name provided, use that
485                    param = param[:-3]
486                if not self.entry.parameterdescs.get(param):
487                    self.entry.parameterdescs[param] = "variable arguments"
488
489            elif (not param) or param == "void":
490                param = "void"
491                self.entry.parameterdescs[param] = "no arguments"
492
493            elif param in ["struct", "union"]:
494                # Handle unnamed (anonymous) union or struct
495                dtype = param
496                param = "{unnamed_" + param + "}"
497                self.entry.parameterdescs[param] = "anonymous\n"
498                self.entry.anon_struct_union = True
499
500        # Warn if parameter has no description
501        # (but ignore ones starting with # as these are not parameters
502        # but inline preprocessor statements)
503        if param not in self.entry.parameterdescs and not param.startswith("#"):
504            self.entry.parameterdescs[param] = self.undescribed
505
506            if "." not in param:
507                if decl_type == 'function':
508                    dname = f"{decl_type} parameter"
509                else:
510                    dname = f"{decl_type} member"
511
512                self.emit_msg(ln,
513                              f"{dname} '{param}' not described in '{declaration_name}'")
514
515        # Strip spaces from param so that it is one continuous string on
516        # parameterlist. This fixes a problem where check_sections()
517        # cannot find a parameter like "addr[6 + 2]" because it actually
518        # appears as "addr[6", "+", "2]" on the parameter list.
519        # However, it's better to maintain the param string unchanged for
520        # output, so just weaken the string compare in check_sections()
521        # to ignore "[blah" in a parameter string.
522
523        self.entry.parameterlist.append(param)
524        org_arg = KernRe(r'\s\s+').sub(' ', org_arg)
525        self.entry.parametertypes[param] = org_arg
526
527
528    def create_parameter_list(self, ln, decl_type, args,
529                              splitter, declaration_name):
530        """
531        Creates a list of parameters, storing them at self.entry.
532        """
533
534        # temporarily replace all commas inside function pointer definition
535        arg_expr = KernRe(r'(\([^\),]+),')
536        while arg_expr.search(args):
537            args = arg_expr.sub(r"\1#", args)
538
539        for arg in args.split(splitter):
540            # Ignore argument attributes
541            arg = KernRe(r'\sPOS0?\s').sub(' ', arg)
542
543            # Strip leading/trailing spaces
544            arg = arg.strip()
545            arg = KernRe(r'\s+').sub(' ', arg, count=1)
546
547            if arg.startswith('#'):
548                # Treat preprocessor directive as a typeless variable just to fill
549                # corresponding data structures "correctly". Catch it later in
550                # output_* subs.
551
552                # Treat preprocessor directive as a typeless variable
553                self.push_parameter(ln, decl_type, arg, "",
554                                    "", declaration_name)
555            #
556            # The pointer-to-function case.
557            #
558            elif KernRe(r'\(.+\)\s*\(').search(arg):
559                arg = arg.replace('#', ',')
560                r = KernRe(r'[^\(]+\(\*?\s*'  # Everything up to "(*"
561                           r'([\w\[\].]*)'    # Capture the name and possible [array]
562                           r'\s*\)')	      # Make sure the trailing ")" is there
563                if r.match(arg):
564                    param = r.group(1)
565                else:
566                    self.emit_msg(ln, f"Invalid param: {arg}")
567                    param = arg
568                dtype = arg.replace(param, '')
569                self.push_parameter(ln, decl_type, param, dtype, arg, declaration_name)
570            #
571            # The array-of-pointers case.  Dig the parameter name out from the middle
572            # of the declaration.
573            #
574            elif KernRe(r'\(.+\)\s*\[').search(arg):
575                r = KernRe(r'[^\(]+\(\s*\*\s*'		# Up to "(" and maybe "*"
576                           r'([\w.]*?)'			# The actual pointer name
577                           r'\s*(\[\s*\w+\s*\]\s*)*\)') # The [array portion]
578                if r.match(arg):
579                    param = r.group(1)
580                else:
581                    self.emit_msg(ln, f"Invalid param: {arg}")
582                    param = arg
583                dtype = arg.replace(param, '')
584                self.push_parameter(ln, decl_type, param, dtype, arg, declaration_name)
585            elif arg:
586                #
587                # Clean up extraneous spaces and split the string at commas; the first
588                # element of the resulting list will also include the type information.
589                #
590                arg = KernRe(r'\s*:\s*').sub(":", arg)
591                arg = KernRe(r'\s*\[').sub('[', arg)
592                args = KernRe(r'\s*,\s*').split(arg)
593                args[0] = re.sub(r'(\*+)\s*', r' \1', args[0])
594                #
595                # args[0] has a string of "type a".  If "a" includes an [array]
596                # declaration, we want to not be fooled by any white space inside
597                # the brackets, so detect and handle that case specially.
598                #
599                r = KernRe(r'^([^[\]]*\s+)(.*)$')
600                if r.match(args[0]):
601                    args[0] = r.group(2)
602                    dtype = r.group(1)
603                else:
604                    # No space in args[0]; this seems wrong but preserves previous behavior
605                    dtype = ''
606
607                bitfield_re = KernRe(r'(.*?):(\w+)')
608                for param in args:
609                    #
610                    # For pointers, shift the star(s) from the variable name to the
611                    # type declaration.
612                    #
613                    r = KernRe(r'^(\*+)\s*(.*)')
614                    if r.match(param):
615                        self.push_parameter(ln, decl_type, r.group(2),
616                                            f"{dtype} {r.group(1)}",
617                                            arg, declaration_name)
618                    #
619                    # Perform a similar shift for bitfields.
620                    #
621                    elif bitfield_re.search(param):
622                        if dtype != "":  # Skip unnamed bit-fields
623                            self.push_parameter(ln, decl_type, bitfield_re.group(1),
624                                                f"{dtype}:{bitfield_re.group(2)}",
625                                                arg, declaration_name)
626                    else:
627                        self.push_parameter(ln, decl_type, param, dtype,
628                                            arg, declaration_name)
629
630    def check_sections(self, ln, decl_name, decl_type):
631        """
632        Check for errors inside sections, emitting warnings if not found
633        parameters are described.
634        """
635        for section in self.entry.sections:
636            if section not in self.entry.parameterlist and \
637               not known_sections.search(section):
638                if decl_type == 'function':
639                    dname = f"{decl_type} parameter"
640                else:
641                    dname = f"{decl_type} member"
642                self.emit_msg(ln,
643                              f"Excess {dname} '{section}' description in '{decl_name}'")
644
645    def check_return_section(self, ln, declaration_name, return_type):
646        """
647        If the function doesn't return void, warns about the lack of a
648        return description.
649        """
650
651        if not self.config.wreturn:
652            return
653
654        # Ignore an empty return type (It's a macro)
655        # Ignore functions with a "void" return type (but not "void *")
656        if not return_type or KernRe(r'void\s*\w*\s*$').search(return_type):
657            return
658
659        if not self.entry.sections.get("Return", None):
660            self.emit_msg(ln,
661                          f"No description found for return value of '{declaration_name}'")
662
663    #
664    # Split apart a structure prototype; returns (struct|union, name, members) or None
665    #
666    def split_struct_proto(self, proto):
667        type_pattern = r'(struct|union)'
668        qualifiers = [
669            "__attribute__",
670            "__packed",
671            "__aligned",
672            "____cacheline_aligned_in_smp",
673            "____cacheline_aligned",
674        ]
675        definition_body = r'\{(.*)\}\s*' + "(?:" + '|'.join(qualifiers) + ")?"
676
677        r = KernRe(type_pattern + r'\s+(\w+)\s*' + definition_body)
678        if r.search(proto):
679            return (r.group(1), r.group(2), r.group(3))
680        else:
681            r = KernRe(r'typedef\s+' + type_pattern + r'\s*' + definition_body + r'\s*(\w+)\s*;')
682            if r.search(proto):
683                return (r.group(1), r.group(3), r.group(2))
684        return None
685    #
686    # Rewrite the members of a structure or union for easier formatting later on.
687    # Among other things, this function will turn a member like:
688    #
689    #  struct { inner_members; } foo;
690    #
691    # into:
692    #
693    #  struct foo; inner_members;
694    #
695    def rewrite_struct_members(self, members):
696        #
697        # Process struct/union members from the most deeply nested outward.  The
698        # trick is in the ^{ below - it prevents a match of an outer struct/union
699        # until the inner one has been munged (removing the "{" in the process).
700        #
701        struct_members = KernRe(r'(struct|union)'   # 0: declaration type
702                                r'([^\{\};]+)' 	    # 1: possible name
703                                r'(\{)'
704                                r'([^\{\}]*)'       # 3: Contents of declaration
705                                r'(\})'
706                                r'([^\{\};]*)(;)')  # 5: Remaining stuff after declaration
707        tuples = struct_members.findall(members)
708        while tuples:
709            for t in tuples:
710                newmember = ""
711                oldmember = "".join(t) # Reconstruct the original formatting
712                dtype, name, lbr, content, rbr, rest, semi = t
713                #
714                # Pass through each field name, normalizing the form and formatting.
715                #
716                for s_id in rest.split(','):
717                    s_id = s_id.strip()
718                    newmember += f"{dtype} {s_id}; "
719                    #
720                    # Remove bitfield/array/pointer info, getting the bare name.
721                    #
722                    s_id = KernRe(r'[:\[].*').sub('', s_id)
723                    s_id = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', s_id)
724                    #
725                    # Pass through the members of this inner structure/union.
726                    #
727                    for arg in content.split(';'):
728                        arg = arg.strip()
729                        #
730                        # Look for (type)(*name)(args) - pointer to function
731                        #
732                        r = KernRe(r'^([^\(]+\(\*?\s*)([\w.]*)(\s*\).*)')
733                        if r.match(arg):
734                            dtype, name, extra = r.group(1), r.group(2), r.group(3)
735                            # Pointer-to-function
736                            if not s_id:
737                                # Anonymous struct/union
738                                newmember += f"{dtype}{name}{extra}; "
739                            else:
740                                newmember += f"{dtype}{s_id}.{name}{extra}; "
741                        #
742                        # Otherwise a non-function member.
743                        #
744                        else:
745                            #
746                            # Remove bitmap and array portions and spaces around commas
747                            #
748                            arg = KernRe(r':\s*\d+\s*').sub('', arg)
749                            arg = KernRe(r'\[.*\]').sub('', arg)
750                            arg = KernRe(r'\s*,\s*').sub(',', arg)
751                            #
752                            # Look for a normal decl - "type name[,name...]"
753                            #
754                            r = KernRe(r'(.*)\s+([\S+,]+)')
755                            if r.search(arg):
756                                for name in r.group(2).split(','):
757                                    name = KernRe(r'^\s*\**(\S+)\s*').sub(r'\1', name)
758                                    if not s_id:
759                                        # Anonymous struct/union
760                                        newmember += f"{r.group(1)} {name}; "
761                                    else:
762                                        newmember += f"{r.group(1)} {s_id}.{name}; "
763                            else:
764                                newmember += f"{arg}; "
765                #
766                # At the end of the s_id loop, replace the original declaration with
767                # the munged version.
768                #
769                members = members.replace(oldmember, newmember)
770            #
771            # End of the tuple loop - search again and see if there are outer members
772            # that now turn up.
773            #
774            tuples = struct_members.findall(members)
775        return members
776
777    #
778    # Format the struct declaration into a standard form for inclusion in the
779    # resulting docs.
780    #
781    def format_struct_decl(self, declaration):
782        #
783        # Insert newlines, get rid of extra spaces.
784        #
785        declaration = KernRe(r'([\{;])').sub(r'\1\n', declaration)
786        declaration = KernRe(r'\}\s+;').sub('};', declaration)
787        #
788        # Format inline enums with each member on its own line.
789        #
790        r = KernRe(r'(enum\s+\{[^\}]+),([^\n])')
791        while r.search(declaration):
792            declaration = r.sub(r'\1,\n\2', declaration)
793        #
794        # Now go through and supply the right number of tabs
795        # for each line.
796        #
797        def_args = declaration.split('\n')
798        level = 1
799        declaration = ""
800        for clause in def_args:
801            clause = KernRe(r'\s+').sub(' ', clause.strip(), count=1)
802            if clause:
803                if '}' in clause and level > 1:
804                    level -= 1
805                if not clause.startswith('#'):
806                    declaration += "\t" * level
807                declaration += "\t" + clause + "\n"
808                if "{" in clause and "}" not in clause:
809                    level += 1
810        return declaration
811
812
813    def dump_struct(self, ln, proto):
814        """
815        Store an entry for an struct or union
816        """
817        #
818        # Do the basic parse to get the pieces of the declaration.
819        #
820        struct_parts = self.split_struct_proto(proto)
821        if not struct_parts:
822            self.emit_msg(ln, f"{proto} error: Cannot parse struct or union!")
823            return
824        decl_type, declaration_name, members = struct_parts
825
826        if self.entry.identifier != declaration_name:
827            self.emit_msg(ln, f"expecting prototype for {decl_type} {self.entry.identifier}. "
828                          f"Prototype was for {decl_type} {declaration_name} instead\n")
829            return
830        #
831        # Go through the list of members applying all of our transformations.
832        #
833        members = trim_private_members(members)
834        members = apply_transforms(struct_xforms, members)
835
836        nested = NestedMatch()
837        for search, sub in struct_nested_prefixes:
838            members = nested.sub(search, sub, members)
839        #
840        # Deal with embedded struct and union members, and drop enums entirely.
841        #
842        declaration = members
843        members = self.rewrite_struct_members(members)
844        members = re.sub(r'(\{[^\{\}]*\})', '', members)
845        #
846        # Output the result and we are done.
847        #
848        self.create_parameter_list(ln, decl_type, members, ';',
849                                   declaration_name)
850        self.check_sections(ln, declaration_name, decl_type)
851        self.output_declaration(decl_type, declaration_name,
852                                definition=self.format_struct_decl(declaration),
853                                purpose=self.entry.declaration_purpose)
854
855    def dump_enum(self, ln, proto):
856        """
857        Stores an enum inside self.entries array.
858        """
859        #
860        # Strip preprocessor directives.  Note that this depends on the
861        # trailing semicolon we added in process_proto_type().
862        #
863        proto = KernRe(r'#\s*((define|ifdef|if)\s+|endif)[^;]*;', flags=re.S).sub('', proto)
864        #
865        # Parse out the name and members of the enum.  Typedef form first.
866        #
867        r = KernRe(r'typedef\s+enum\s*\{(.*)\}\s*(\w*)\s*;')
868        if r.search(proto):
869            declaration_name = r.group(2)
870            members = trim_private_members(r.group(1))
871        #
872        # Failing that, look for a straight enum
873        #
874        else:
875            r = KernRe(r'enum\s+(\w*)\s*\{(.*)\}')
876            if r.match(proto):
877                declaration_name = r.group(1)
878                members = trim_private_members(r.group(2))
879        #
880        # OK, this isn't going to work.
881        #
882            else:
883                self.emit_msg(ln, f"{proto}: error: Cannot parse enum!")
884                return
885        #
886        # Make sure we found what we were expecting.
887        #
888        if self.entry.identifier != declaration_name:
889            if self.entry.identifier == "":
890                self.emit_msg(ln,
891                              f"{proto}: wrong kernel-doc identifier on prototype")
892            else:
893                self.emit_msg(ln,
894                              f"expecting prototype for enum {self.entry.identifier}. "
895                              f"Prototype was for enum {declaration_name} instead")
896            return
897
898        if not declaration_name:
899            declaration_name = "(anonymous)"
900        #
901        # Parse out the name of each enum member, and verify that we
902        # have a description for it.
903        #
904        member_set = set()
905        members = KernRe(r'\([^;)]*\)').sub('', members)
906        for arg in members.split(','):
907            if not arg:
908                continue
909            arg = KernRe(r'^\s*(\w+).*').sub(r'\1', arg)
910            self.entry.parameterlist.append(arg)
911            if arg not in self.entry.parameterdescs:
912                self.entry.parameterdescs[arg] = self.undescribed
913                self.emit_msg(ln,
914                              f"Enum value '{arg}' not described in enum '{declaration_name}'")
915            member_set.add(arg)
916        #
917        # Ensure that every described member actually exists in the enum.
918        #
919        for k in self.entry.parameterdescs:
920            if k not in member_set:
921                self.emit_msg(ln,
922                              f"Excess enum value '%{k}' description in '{declaration_name}'")
923
924        self.output_declaration('enum', declaration_name,
925                                purpose=self.entry.declaration_purpose)
926
927    def dump_declaration(self, ln, prototype):
928        """
929        Stores a data declaration inside self.entries array.
930        """
931
932        if self.entry.decl_type == "enum":
933            self.dump_enum(ln, prototype)
934        elif self.entry.decl_type == "typedef":
935            self.dump_typedef(ln, prototype)
936        elif self.entry.decl_type in ["union", "struct"]:
937            self.dump_struct(ln, prototype)
938        else:
939            # This would be a bug
940            self.emit_message(ln, f'Unknown declaration type: {self.entry.decl_type}')
941
942    def dump_function(self, ln, prototype):
943        """
944        Stores a function of function macro inside self.entries array.
945        """
946
947        found = func_macro = False
948        return_type = ''
949        decl_type = 'function'
950        #
951        # Apply the initial transformations.
952        #
953        prototype = apply_transforms(function_xforms, prototype)
954        #
955        # If we have a macro, remove the "#define" at the front.
956        #
957        new_proto = KernRe(r"^#\s*define\s+").sub("", prototype)
958        if new_proto != prototype:
959            prototype = new_proto
960            #
961            # Dispense with the simple "#define A B" case here; the key
962            # is the space after the name of the symbol being defined.
963            # NOTE that the seemingly misnamed "func_macro" indicates a
964            # macro *without* arguments.
965            #
966            r = KernRe(r'^(\w+)\s+')
967            if r.search(prototype):
968                return_type = ''
969                declaration_name = r.group(1)
970                func_macro = True
971                found = True
972
973        # Yes, this truly is vile.  We are looking for:
974        # 1. Return type (may be nothing if we're looking at a macro)
975        # 2. Function name
976        # 3. Function parameters.
977        #
978        # All the while we have to watch out for function pointer parameters
979        # (which IIRC is what the two sections are for), C types (these
980        # regexps don't even start to express all the possibilities), and
981        # so on.
982        #
983        # If you mess with these regexps, it's a good idea to check that
984        # the following functions' documentation still comes out right:
985        # - parport_register_device (function pointer parameters)
986        # - atomic_set (macro)
987        # - pci_match_device, __copy_to_user (long return type)
988
989        name = r'\w+'
990        type1 = r'(?:[\w\s]+)?'
991        type2 = r'(?:[\w\s]+\*+)+'
992        #
993        # Attempt to match first on (args) with no internal parentheses; this
994        # lets us easily filter out __acquires() and other post-args stuff.  If
995        # that fails, just grab the rest of the line to the last closing
996        # parenthesis.
997        #
998        proto_args = r'\(([^\(]*|.*)\)'
999        #
1000        # (Except for the simple macro case) attempt to split up the prototype
1001        # in the various ways we understand.
1002        #
1003        if not found:
1004            patterns = [
1005                rf'^()({name})\s*{proto_args}',
1006                rf'^({type1})\s+({name})\s*{proto_args}',
1007                rf'^({type2})\s*({name})\s*{proto_args}',
1008            ]
1009
1010            for p in patterns:
1011                r = KernRe(p)
1012                if r.match(prototype):
1013                    return_type = r.group(1)
1014                    declaration_name = r.group(2)
1015                    args = r.group(3)
1016                    self.create_parameter_list(ln, decl_type, args, ',',
1017                                               declaration_name)
1018                    found = True
1019                    break
1020        #
1021        # Parsing done; make sure that things are as we expect.
1022        #
1023        if not found:
1024            self.emit_msg(ln,
1025                          f"cannot understand function prototype: '{prototype}'")
1026            return
1027        if self.entry.identifier != declaration_name:
1028            self.emit_msg(ln, f"expecting prototype for {self.entry.identifier}(). "
1029                          f"Prototype was for {declaration_name}() instead")
1030            return
1031        self.check_sections(ln, declaration_name, "function")
1032        self.check_return_section(ln, declaration_name, return_type)
1033        #
1034        # Store the result.
1035        #
1036        self.output_declaration(decl_type, declaration_name,
1037                                typedef=('typedef' in return_type),
1038                                functiontype=return_type,
1039                                purpose=self.entry.declaration_purpose,
1040                                func_macro=func_macro)
1041
1042
1043    def dump_typedef(self, ln, proto):
1044        """
1045        Stores a typedef inside self.entries array.
1046        """
1047        #
1048        # We start by looking for function typedefs.
1049        #
1050        typedef_type = r'typedef((?:\s+[\w*]+\b){0,7}\s+(?:\w+\b|\*+))\s*'
1051        typedef_ident = r'\*?\s*(\w\S+)\s*'
1052        typedef_args = r'\s*\((.*)\);'
1053
1054        typedef1 = KernRe(typedef_type + r'\(' + typedef_ident + r'\)' + typedef_args)
1055        typedef2 = KernRe(typedef_type + typedef_ident + typedef_args)
1056
1057        # Parse function typedef prototypes
1058        for r in [typedef1, typedef2]:
1059            if not r.match(proto):
1060                continue
1061
1062            return_type = r.group(1).strip()
1063            declaration_name = r.group(2)
1064            args = r.group(3)
1065
1066            if self.entry.identifier != declaration_name:
1067                self.emit_msg(ln,
1068                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1069                return
1070
1071            self.create_parameter_list(ln, 'function', args, ',', declaration_name)
1072
1073            self.output_declaration('function', declaration_name,
1074                                    typedef=True,
1075                                    functiontype=return_type,
1076                                    purpose=self.entry.declaration_purpose)
1077            return
1078        #
1079        # Not a function, try to parse a simple typedef.
1080        #
1081        r = KernRe(r'typedef.*\s+(\w+)\s*;')
1082        if r.match(proto):
1083            declaration_name = r.group(1)
1084
1085            if self.entry.identifier != declaration_name:
1086                self.emit_msg(ln,
1087                              f"expecting prototype for typedef {self.entry.identifier}. Prototype was for typedef {declaration_name} instead\n")
1088                return
1089
1090            self.output_declaration('typedef', declaration_name,
1091                                    purpose=self.entry.declaration_purpose)
1092            return
1093
1094        self.emit_msg(ln, "error: Cannot parse typedef!")
1095
1096    @staticmethod
1097    def process_export(function_set, line):
1098        """
1099        process EXPORT_SYMBOL* tags
1100
1101        This method doesn't use any variable from the class, so declare it
1102        with a staticmethod decorator.
1103        """
1104
1105        # We support documenting some exported symbols with different
1106        # names.  A horrible hack.
1107        suffixes = [ '_noprof' ]
1108
1109        # Note: it accepts only one EXPORT_SYMBOL* per line, as having
1110        # multiple export lines would violate Kernel coding style.
1111
1112        if export_symbol.search(line):
1113            symbol = export_symbol.group(2)
1114        elif export_symbol_ns.search(line):
1115            symbol = export_symbol_ns.group(2)
1116        else:
1117            return False
1118        #
1119        # Found an export, trim out any special suffixes
1120        #
1121        for suffix in suffixes:
1122            # Be backward compatible with Python < 3.9
1123            if symbol.endswith(suffix):
1124                symbol = symbol[:-len(suffix)]
1125        function_set.add(symbol)
1126        return True
1127
1128    def process_normal(self, ln, line):
1129        """
1130        STATE_NORMAL: looking for the /** to begin everything.
1131        """
1132
1133        if not doc_start.match(line):
1134            return
1135
1136        # start a new entry
1137        self.reset_state(ln)
1138
1139        # next line is always the function name
1140        self.state = state.NAME
1141
1142    def process_name(self, ln, line):
1143        """
1144        STATE_NAME: Looking for the "name - description" line
1145        """
1146        #
1147        # Check for a DOC: block and handle them specially.
1148        #
1149        if doc_block.search(line):
1150
1151            if not doc_block.group(1):
1152                self.entry.begin_section(ln, "Introduction")
1153            else:
1154                self.entry.begin_section(ln, doc_block.group(1))
1155
1156            self.entry.identifier = self.entry.section
1157            self.state = state.DOCBLOCK
1158        #
1159        # Otherwise we're looking for a normal kerneldoc declaration line.
1160        #
1161        elif doc_decl.search(line):
1162            self.entry.identifier = doc_decl.group(1)
1163
1164            # Test for data declaration
1165            if doc_begin_data.search(line):
1166                self.entry.decl_type = doc_begin_data.group(1)
1167                self.entry.identifier = doc_begin_data.group(2)
1168            #
1169            # Look for a function description
1170            #
1171            elif doc_begin_func.search(line):
1172                self.entry.identifier = doc_begin_func.group(1)
1173                self.entry.decl_type = "function"
1174            #
1175            # We struck out.
1176            #
1177            else:
1178                self.emit_msg(ln,
1179                              f"This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst\n{line}")
1180                self.state = state.NORMAL
1181                return
1182            #
1183            # OK, set up for a new kerneldoc entry.
1184            #
1185            self.state = state.BODY
1186            self.entry.identifier = self.entry.identifier.strip(" ")
1187            # if there's no @param blocks need to set up default section here
1188            self.entry.begin_section(ln + 1)
1189            #
1190            # Find the description portion, which *should* be there but
1191            # isn't always.
1192            # (We should be able to capture this from the previous parsing - someday)
1193            #
1194            r = KernRe("[-:](.*)")
1195            if r.search(line):
1196                self.entry.declaration_purpose = trim_whitespace(r.group(1))
1197                self.state = state.DECLARATION
1198            else:
1199                self.entry.declaration_purpose = ""
1200
1201            if not self.entry.declaration_purpose and self.config.wshort_desc:
1202                self.emit_msg(ln,
1203                              f"missing initial short description on line:\n{line}")
1204
1205            if not self.entry.identifier and self.entry.decl_type != "enum":
1206                self.emit_msg(ln,
1207                              f"wrong kernel-doc identifier on line:\n{line}")
1208                self.state = state.NORMAL
1209
1210            if self.config.verbose:
1211                self.emit_msg(ln,
1212                              f"Scanning doc for {self.entry.decl_type} {self.entry.identifier}",
1213                                  warning=False)
1214        #
1215        # Failed to find an identifier. Emit a warning
1216        #
1217        else:
1218            self.emit_msg(ln, f"Cannot find identifier on line:\n{line}")
1219
1220    #
1221    # Helper function to determine if a new section is being started.
1222    #
1223    def is_new_section(self, ln, line):
1224        if doc_sect.search(line):
1225            self.state = state.BODY
1226            #
1227            # Pick out the name of our new section, tweaking it if need be.
1228            #
1229            newsection = doc_sect.group(1)
1230            if newsection.lower() == 'description':
1231                newsection = 'Description'
1232            elif newsection.lower() == 'context':
1233                newsection = 'Context'
1234                self.state = state.SPECIAL_SECTION
1235            elif newsection.lower() in ["@return", "@returns",
1236                                        "return", "returns"]:
1237                newsection = "Return"
1238                self.state = state.SPECIAL_SECTION
1239            elif newsection[0] == '@':
1240                self.state = state.SPECIAL_SECTION
1241            #
1242            # Initialize the contents, and get the new section going.
1243            #
1244            newcontents = doc_sect.group(2)
1245            if not newcontents:
1246                newcontents = ""
1247            self.dump_section()
1248            self.entry.begin_section(ln, newsection)
1249            self.entry.leading_space = None
1250
1251            self.entry.add_text(newcontents.lstrip())
1252            return True
1253        return False
1254
1255    #
1256    # Helper function to detect (and effect) the end of a kerneldoc comment.
1257    #
1258    def is_comment_end(self, ln, line):
1259        if doc_end.search(line):
1260            self.dump_section()
1261
1262            # Look for doc_com + <text> + doc_end:
1263            r = KernRe(r'\s*\*\s*[a-zA-Z_0-9:.]+\*/')
1264            if r.match(line):
1265                self.emit_msg(ln, f"suspicious ending line: {line}")
1266
1267            self.entry.prototype = ""
1268            self.entry.new_start_line = ln + 1
1269
1270            self.state = state.PROTO
1271            return True
1272        return False
1273
1274
1275    def process_decl(self, ln, line):
1276        """
1277        STATE_DECLARATION: We've seen the beginning of a declaration
1278        """
1279        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1280            return
1281        #
1282        # Look for anything with the " * " line beginning.
1283        #
1284        if doc_content.search(line):
1285            cont = doc_content.group(1)
1286            #
1287            # A blank line means that we have moved out of the declaration
1288            # part of the comment (without any "special section" parameter
1289            # descriptions).
1290            #
1291            if cont == "":
1292                self.state = state.BODY
1293            #
1294            # Otherwise we have more of the declaration section to soak up.
1295            #
1296            else:
1297                self.entry.declaration_purpose = \
1298                    trim_whitespace(self.entry.declaration_purpose + ' ' + cont)
1299        else:
1300            # Unknown line, ignore
1301            self.emit_msg(ln, f"bad line: {line}")
1302
1303
1304    def process_special(self, ln, line):
1305        """
1306        STATE_SPECIAL_SECTION: a section ending with a blank line
1307        """
1308        #
1309        # If we have hit a blank line (only the " * " marker), then this
1310        # section is done.
1311        #
1312        if KernRe(r"\s*\*\s*$").match(line):
1313            self.entry.begin_section(ln, dump = True)
1314            self.state = state.BODY
1315            return
1316        #
1317        # Not a blank line, look for the other ways to end the section.
1318        #
1319        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1320            return
1321        #
1322        # OK, we should have a continuation of the text for this section.
1323        #
1324        if doc_content.search(line):
1325            cont = doc_content.group(1)
1326            #
1327            # If the lines of text after the first in a special section have
1328            # leading white space, we need to trim it out or Sphinx will get
1329            # confused.  For the second line (the None case), see what we
1330            # find there and remember it.
1331            #
1332            if self.entry.leading_space is None:
1333                r = KernRe(r'^(\s+)')
1334                if r.match(cont):
1335                    self.entry.leading_space = len(r.group(1))
1336                else:
1337                    self.entry.leading_space = 0
1338            #
1339            # Otherwise, before trimming any leading chars, be *sure*
1340            # that they are white space.  We should maybe warn if this
1341            # isn't the case.
1342            #
1343            for i in range(0, self.entry.leading_space):
1344                if cont[i] != " ":
1345                    self.entry.leading_space = i
1346                    break
1347            #
1348            # Add the trimmed result to the section and we're done.
1349            #
1350            self.entry.add_text(cont[self.entry.leading_space:])
1351        else:
1352            # Unknown line, ignore
1353            self.emit_msg(ln, f"bad line: {line}")
1354
1355    def process_body(self, ln, line):
1356        """
1357        STATE_BODY: the bulk of a kerneldoc comment.
1358        """
1359        if self.is_new_section(ln, line) or self.is_comment_end(ln, line):
1360            return
1361
1362        if doc_content.search(line):
1363            cont = doc_content.group(1)
1364            self.entry.add_text(cont)
1365        else:
1366            # Unknown line, ignore
1367            self.emit_msg(ln, f"bad line: {line}")
1368
1369    def process_inline_name(self, ln, line):
1370        """STATE_INLINE_NAME: beginning of docbook comments within a prototype."""
1371
1372        if doc_inline_sect.search(line):
1373            self.entry.begin_section(ln, doc_inline_sect.group(1))
1374            self.entry.add_text(doc_inline_sect.group(2).lstrip())
1375            self.state = state.INLINE_TEXT
1376        elif doc_inline_end.search(line):
1377            self.dump_section()
1378            self.state = state.PROTO
1379        elif doc_content.search(line):
1380            self.emit_msg(ln, f"Incorrect use of kernel-doc format: {line}")
1381            self.state = state.PROTO
1382        # else ... ??
1383
1384    def process_inline_text(self, ln, line):
1385        """STATE_INLINE_TEXT: docbook comments within a prototype."""
1386
1387        if doc_inline_end.search(line):
1388            self.dump_section()
1389            self.state = state.PROTO
1390        elif doc_content.search(line):
1391            self.entry.add_text(doc_content.group(1))
1392        # else ... ??
1393
1394    def syscall_munge(self, ln, proto):         # pylint: disable=W0613
1395        """
1396        Handle syscall definitions
1397        """
1398
1399        is_void = False
1400
1401        # Strip newlines/CR's
1402        proto = re.sub(r'[\r\n]+', ' ', proto)
1403
1404        # Check if it's a SYSCALL_DEFINE0
1405        if 'SYSCALL_DEFINE0' in proto:
1406            is_void = True
1407
1408        # Replace SYSCALL_DEFINE with correct return type & function name
1409        proto = KernRe(r'SYSCALL_DEFINE.*\(').sub('long sys_', proto)
1410
1411        r = KernRe(r'long\s+(sys_.*?),')
1412        if r.search(proto):
1413            proto = KernRe(',').sub('(', proto, count=1)
1414        elif is_void:
1415            proto = KernRe(r'\)').sub('(void)', proto, count=1)
1416
1417        # Now delete all of the odd-numbered commas in the proto
1418        # so that argument types & names don't have a comma between them
1419        count = 0
1420        length = len(proto)
1421
1422        if is_void:
1423            length = 0  # skip the loop if is_void
1424
1425        for ix in range(length):
1426            if proto[ix] == ',':
1427                count += 1
1428                if count % 2 == 1:
1429                    proto = proto[:ix] + ' ' + proto[ix + 1:]
1430
1431        return proto
1432
1433    def tracepoint_munge(self, ln, proto):
1434        """
1435        Handle tracepoint definitions
1436        """
1437
1438        tracepointname = None
1439        tracepointargs = None
1440
1441        # Match tracepoint name based on different patterns
1442        r = KernRe(r'TRACE_EVENT\((.*?),')
1443        if r.search(proto):
1444            tracepointname = r.group(1)
1445
1446        r = KernRe(r'DEFINE_SINGLE_EVENT\((.*?),')
1447        if r.search(proto):
1448            tracepointname = r.group(1)
1449
1450        r = KernRe(r'DEFINE_EVENT\((.*?),(.*?),')
1451        if r.search(proto):
1452            tracepointname = r.group(2)
1453
1454        if tracepointname:
1455            tracepointname = tracepointname.lstrip()
1456
1457        r = KernRe(r'TP_PROTO\((.*?)\)')
1458        if r.search(proto):
1459            tracepointargs = r.group(1)
1460
1461        if not tracepointname or not tracepointargs:
1462            self.emit_msg(ln,
1463                          f"Unrecognized tracepoint format:\n{proto}\n")
1464        else:
1465            proto = f"static inline void trace_{tracepointname}({tracepointargs})"
1466            self.entry.identifier = f"trace_{self.entry.identifier}"
1467
1468        return proto
1469
1470    def process_proto_function(self, ln, line):
1471        """Ancillary routine to process a function prototype"""
1472
1473        # strip C99-style comments to end of line
1474        line = KernRe(r"//.*$", re.S).sub('', line)
1475        #
1476        # Soak up the line's worth of prototype text, stopping at { or ; if present.
1477        #
1478        if KernRe(r'\s*#\s*define').match(line):
1479            self.entry.prototype = line
1480        elif not line.startswith('#'):   # skip other preprocessor stuff
1481            r = KernRe(r'([^\{]*)')
1482            if r.match(line):
1483                self.entry.prototype += r.group(1) + " "
1484        #
1485        # If we now have the whole prototype, clean it up and declare victory.
1486        #
1487        if '{' in line or ';' in line or KernRe(r'\s*#\s*define').match(line):
1488            # strip comments and surrounding spaces
1489            self.entry.prototype = KernRe(r'/\*.*\*/').sub('', self.entry.prototype).strip()
1490            #
1491            # Handle self.entry.prototypes for function pointers like:
1492            #       int (*pcs_config)(struct foo)
1493            # by turning it into
1494            #	    int pcs_config(struct foo)
1495            #
1496            r = KernRe(r'^(\S+\s+)\(\s*\*(\S+)\)')
1497            self.entry.prototype = r.sub(r'\1\2', self.entry.prototype)
1498            #
1499            # Handle special declaration syntaxes
1500            #
1501            if 'SYSCALL_DEFINE' in self.entry.prototype:
1502                self.entry.prototype = self.syscall_munge(ln,
1503                                                          self.entry.prototype)
1504            else:
1505                r = KernRe(r'TRACE_EVENT|DEFINE_EVENT|DEFINE_SINGLE_EVENT')
1506                if r.search(self.entry.prototype):
1507                    self.entry.prototype = self.tracepoint_munge(ln,
1508                                                                 self.entry.prototype)
1509            #
1510            # ... and we're done
1511            #
1512            self.dump_function(ln, self.entry.prototype)
1513            self.reset_state(ln)
1514
1515    def process_proto_type(self, ln, line):
1516        """Ancillary routine to process a type"""
1517
1518        # Strip C99-style comments and surrounding whitespace
1519        line = KernRe(r"//.*$", re.S).sub('', line).strip()
1520        if not line:
1521            return # nothing to see here
1522
1523        # To distinguish preprocessor directive from regular declaration later.
1524        if line.startswith('#'):
1525            line += ";"
1526        #
1527        # Split the declaration on any of { } or ;, and accumulate pieces
1528        # until we hit a semicolon while not inside {brackets}
1529        #
1530        r = KernRe(r'(.*?)([{};])')
1531        for chunk in r.split(line):
1532            if chunk:  # Ignore empty matches
1533                self.entry.prototype += chunk
1534                #
1535                # This cries out for a match statement ... someday after we can
1536                # drop Python 3.9 ...
1537                #
1538                if chunk == '{':
1539                    self.entry.brcount += 1
1540                elif chunk == '}':
1541                    self.entry.brcount -= 1
1542                elif chunk == ';' and self.entry.brcount <= 0:
1543                    self.dump_declaration(ln, self.entry.prototype)
1544                    self.reset_state(ln)
1545                    return
1546        #
1547        # We hit the end of the line while still in the declaration; put
1548        # in a space to represent the newline.
1549        #
1550        self.entry.prototype += ' '
1551
1552    def process_proto(self, ln, line):
1553        """STATE_PROTO: reading a function/whatever prototype."""
1554
1555        if doc_inline_oneline.search(line):
1556            self.entry.begin_section(ln, doc_inline_oneline.group(1))
1557            self.entry.add_text(doc_inline_oneline.group(2))
1558            self.dump_section()
1559
1560        elif doc_inline_start.search(line):
1561            self.state = state.INLINE_NAME
1562
1563        elif self.entry.decl_type == 'function':
1564            self.process_proto_function(ln, line)
1565
1566        else:
1567            self.process_proto_type(ln, line)
1568
1569    def process_docblock(self, ln, line):
1570        """STATE_DOCBLOCK: within a DOC: block."""
1571
1572        if doc_end.search(line):
1573            self.dump_section()
1574            self.output_declaration("doc", self.entry.identifier)
1575            self.reset_state(ln)
1576
1577        elif doc_content.search(line):
1578            self.entry.add_text(doc_content.group(1))
1579
1580    def parse_export(self):
1581        """
1582        Parses EXPORT_SYMBOL* macros from a single Kernel source file.
1583        """
1584
1585        export_table = set()
1586
1587        try:
1588            with open(self.fname, "r", encoding="utf8",
1589                      errors="backslashreplace") as fp:
1590
1591                for line in fp:
1592                    self.process_export(export_table, line)
1593
1594        except IOError:
1595            return None
1596
1597        return export_table
1598
1599    #
1600    # The state/action table telling us which function to invoke in
1601    # each state.
1602    #
1603    state_actions = {
1604        state.NORMAL:			process_normal,
1605        state.NAME:			process_name,
1606        state.BODY:			process_body,
1607        state.DECLARATION:		process_decl,
1608        state.SPECIAL_SECTION:		process_special,
1609        state.INLINE_NAME:		process_inline_name,
1610        state.INLINE_TEXT:		process_inline_text,
1611        state.PROTO:			process_proto,
1612        state.DOCBLOCK:			process_docblock,
1613        }
1614
1615    def parse_kdoc(self):
1616        """
1617        Open and process each line of a C source file.
1618        The parsing is controlled via a state machine, and the line is passed
1619        to a different process function depending on the state. The process
1620        function may update the state as needed.
1621
1622        Besides parsing kernel-doc tags, it also parses export symbols.
1623        """
1624
1625        prev = ""
1626        prev_ln = None
1627        export_table = set()
1628
1629        try:
1630            with open(self.fname, "r", encoding="utf8",
1631                      errors="backslashreplace") as fp:
1632                for ln, line in enumerate(fp):
1633
1634                    line = line.expandtabs().strip("\n")
1635
1636                    # Group continuation lines on prototypes
1637                    if self.state == state.PROTO:
1638                        if line.endswith("\\"):
1639                            prev += line.rstrip("\\")
1640                            if not prev_ln:
1641                                prev_ln = ln
1642                            continue
1643
1644                        if prev:
1645                            ln = prev_ln
1646                            line = prev + line
1647                            prev = ""
1648                            prev_ln = None
1649
1650                    self.config.log.debug("%d %s: %s",
1651                                          ln, state.name[self.state],
1652                                          line)
1653
1654                    # This is an optimization over the original script.
1655                    # There, when export_file was used for the same file,
1656                    # it was read twice. Here, we use the already-existing
1657                    # loop to parse exported symbols as well.
1658                    #
1659                    if (self.state != state.NORMAL) or \
1660                       not self.process_export(export_table, line):
1661                        # Hand this line to the appropriate state handler
1662                        self.state_actions[self.state](self, ln, line)
1663
1664        except OSError:
1665            self.config.log.error(f"Error: Cannot open file {self.fname}")
1666
1667        return export_table, self.entries
1668