From 1d27c062b62cfdb682f7b3dd5f5ac7bc777c8a06 Mon Sep 17 00:00:00 2001 From: jhb Date: Thu, 24 Dec 2009 14:32:11 +0000 Subject: Add a new tool which attempts to check for kernel configuration options that are missing from NOTES files. --- tools/tools/README | 1 + tools/tools/notescheck/Makefile | 5 + tools/tools/notescheck/notescheck.py | 361 +++++++++++++++++++++++++++++++++++ 3 files changed, 367 insertions(+) create mode 100644 tools/tools/notescheck/Makefile create mode 100644 tools/tools/notescheck/notescheck.py (limited to 'tools') diff --git a/tools/tools/README b/tools/tools/README index c052e7d..8a28d98 100644 --- a/tools/tools/README +++ b/tools/tools/README @@ -48,6 +48,7 @@ mfc Merge a directory from HEAD to a branch where it does not mid Create a Message-ID database for mailing lists. mwl Tools specific to the Marvell 88W8363 support ncpus Count the number of processors +notescheck Check for missing devices and options in NOTES files. npe Tools specific to the Intel IXP4XXX NPE device nxge A diagnostic tool for the nxge(4) driver pciid Generate src/share/misc/pci_vendors. diff --git a/tools/tools/notescheck/Makefile b/tools/tools/notescheck/Makefile new file mode 100644 index 0000000..8899a0a --- /dev/null +++ b/tools/tools/notescheck/Makefile @@ -0,0 +1,5 @@ +# $FreeBSD$ + +SCRIPTS= notescheck.py + +.include diff --git a/tools/tools/notescheck/notescheck.py b/tools/tools/notescheck/notescheck.py new file mode 100644 index 0000000..5aae296 --- /dev/null +++ b/tools/tools/notescheck/notescheck.py @@ -0,0 +1,361 @@ +#!/usr/local/bin/python +# +# This script analyzes sys/conf/files*, sys/conf/options*, +# sys/conf/NOTES, and sys/*/conf/NOTES and checks for inconsistencies +# such as options or devices that are not specified in any NOTES files +# or MI devices specified in MD NOTES files. +# +# $FreeBSD$ + +import glob +import os.path +import sys + +def usage(): + print >>sys.stderr, "notescheck " + print >>sys.stderr + print >>sys.stderr, "Where 'path' is a path to a kernel source tree." + +# These files are used to determine if a path is a valid kernel source tree. +requiredfiles = ['conf/files', 'conf/options', 'conf/NOTES'] + +# This special platform string is used for managing MI options. +global_platform = 'global' + +# This is a global string that represents the current file and line +# being parsed. +location = "" + +# Format the contents of a set into a sorted, comma-separated string +def format_set(set): + l = [] + for item in set: + l.append(item) + if len(l) == 0: + return "(empty)" + l.sort() + if len(l) == 2: + return "%s and %s" % (l[0], l[1]) + s = "%s" % (l[0]) + if len(l) == 1: + return s + for item in l[1:-1]: + s = "%s, %s" % (s, item) + s = "%s, and %s" % (s, l[-1]) + return s + +# This class actually covers both options and devices. For each named +# option we maintain two different lists. One is the list of +# platforms that the option was defined in via an options or files +# file. The other is the list of platforms that the option was tested +# in via a NOTES file. All options are stored as lowercase since +# config(8) treats the names as case-insensitive. +class Option: + def __init__(self, name): + self.name = name + self.type = None + self.defines = set() + self.tests = set() + + def set_type(self, type): + if self.type is None: + self.type = type + self.type_location = location + elif self.type != type: + print "WARN: Attempt to change type of %s from %s to %s%s" % \ + (self.name, self.type, type, location) + print " Previous type set%s" % (self.type_location) + + def add_define(self, platform): + self.defines.add(platform) + + def add_test(self, platform): + self.tests.add(platform) + + def title(self): + if self.type == 'option': + return 'option %s' % (self.name.upper()) + if self.type == None: + return self.name + return '%s %s' % (self.type, self.name) + + def warn(self): + # If the defined and tested sets are equal, then this option + # is ok. + if self.defines == self.tests: + return + + # If the tested set contains the global platform, then this + # option is ok. + if global_platform in self.tests: + return + + if global_platform in self.defines: + # If the device is defined globally ans is never tested, whine. + if len(self.tests) == 0: + print 'WARN: %s is defined globally but never tested' % \ + (self.title()) + return + + # If the device is defined globally and is tested on + # multiple MD platforms, then it is ok. This often occurs + # for drivers that are shared across multiple, but not + # all, platforms (e.g. acpi, agp). + if len(self.tests) > 1: + return + + # If a device is defined globally but is only tested on a + # single MD platform, then whine about this. + print 'WARN: %s is defined globally but only tested in %s NOTES' % \ + (self.title(), format_set(self.tests)) + return + + # If an option or device is never tested, whine. + if len(self.tests) == 0: + print 'WARN: %s is defined in %s but never tested' % \ + (self.title(), format_set(self.defines)) + return + + # The set of MD platforms where this option is defined, but not tested. + notest = self.defines - self.tests + if len(notest) != 0: + print 'WARN: %s is not tested in %s NOTES' % \ + (self.title(), format_set(notest)) + return + + print 'ERROR: bad state for %s: defined in %s, tested in %s' % \ + (self.title(), format_set(self.defines), format_set(self.tests)) + +# This class maintains a dictionary of options keyed by name. +class Options: + def __init__(self): + self.options = {} + + # Look up the object for a given option by name. If the option + # doesn't already exist, then add a new option. + def find(self, name): + name = name.lower() + if name in self.options: + return self.options[name] + option = Option(name) + self.options[name] = option + return option + + # Warn about inconsistencies + def warn(self): + keys = self.options.keys() + keys.sort() + for key in keys: + option = self.options[key] + option.warn() + +# Global map of options +options = Options() + +# Look for MD NOTES files to build our list of platforms. We ignore +# platforms that do not have a NOTES file. +def find_platforms(tree): + platforms = [] + for file in glob.glob(tree + '*/conf/NOTES'): + if not file.startswith(tree): + print >>sys.stderr, "Bad MD NOTES file %s" %(file) + sys.exit(1) + platforms.append(file[len(tree):].split('/')[0]) + if global_platform in platforms: + print >>sys.stderr, "Found MD NOTES file for global platform" + sys.exit(1) + return platforms + +# Parse a file that has escaped newlines. Any escaped newlines are +# coalesced and each logical line is passed to the callback function. +# This also skips blank lines and comments. +def parse_file(file, callback, *args): + global location + + f = open(file) + current = None + i = 0 + for line in f: + # Update parsing location + i = i + 1 + location = ' at %s:%d' % (file, i) + + # Trim the newline + line = line[:-1] + + # If the previous line had an escaped newline, append this + # line to that. + if current is not None: + line = current + line + current = None + + # If the line ends in a '\', set current to the line (minus + # the escape) and continue. + if len(line) > 0 and line[-1] == '\\': + current = line[:-1] + continue + + # Skip blank lines or lines with only whitespace + if len(line) == 0 or len(line.split()) == 0: + continue + + # Skip comment lines. Any line whose first non-space + # character is a '#' is considered a comment. + if line.split()[0][0] == '#': + continue + + # Invoke the callback on this line + callback(line, *args) + if current is not None: + callback(current, *args) + + location = "" + +# Split a line into words on whitespace with the exception that quoted +# strings are always treated as a single word. +def tokenize(line): + if len(line) == 0: + return [] + + # First, split the line on quote characters. + groups = line.split('"') + + # Ensure we have an even number of quotes. The 'groups' array + # will contain 'number of quotes' + 1 entries, so it should have + # an odd number of entries. + if len(groups) % 2 == 0: + print >>sys.stderr, "Failed to tokenize: %s%s" (line, location) + return [] + + # String split all the "odd" groups since they are not quoted strings. + quoted = False + words = [] + for group in groups: + if quoted: + words.append(group) + quoted = False + else: + for word in group.split(): + words.append(word) + quoted = True + return words + +# Parse a sys/conf/files* file adding defines for any options +# encountered. Note files does not differentiate between options and +# devices. +def parse_files_line(line, platform): + words = tokenize(line) + + # Skip include lines. + if words[0] == 'include': + return + + # Skip standard lines as they have no devices or options. + if words[1] == 'standard': + return + + # Remaining lines better be optional or mandatory lines. + if words[1] != 'optional' and words[1] != 'mandatory': + print >>sys.stderr, "Invalid files line: %s%s" % (line, location) + + # Drop the first two words and begin parsing keywords and devices. + skip = False + for word in words[2:]: + if skip: + skip = False + continue + + # Skip keywords + if word == 'no-obj' or word == 'no-implicit-rule' or \ + word == 'before-depend' or word == 'local' or \ + word == 'no-depend' or word == 'profiling-routine' or \ + word == 'nowerror': + continue + + # Skip keywords and their following argument + if word == 'dependency' or word == 'clean' or \ + word == 'compile-with' or word == 'warning': + skip = True + continue + + # Ignore pipes + if word == '|': + continue + + option = options.find(word) + option.add_define(platform) + +# Parse a sys/conf/options* file adding defines for any options +# encountered. Unlike a files file, options files only add options. +def parse_options_line(line, platform): + # The first word is the option name. + name = line.split()[0] + + # Ignore DEV_xxx options. These are magic options that are + # aliases for 'device xxx'. + if name.startswith('DEV_'): + return + + option = options.find(name) + option.add_define(platform) + option.set_type('option') + +# Parse a sys/conf/NOTES file adding tests for any options or devices +# encountered. +def parse_notes_line(line, platform): + words = line.split() + + # Skip lines with just whitespace + if len(words) == 0: + return + + if words[0] == 'device' or words[0] == 'devices': + option = options.find(words[1]) + option.add_test(platform) + option.set_type('device') + return + + if words[0] == 'option' or words[0] == 'options': + option = options.find(words[1].split('=')[0]) + option.add_test(platform) + option.set_type('option') + return + +def main(argv=None): + if argv is None: + argv = sys.argv + if len(sys.argv) != 2: + usage() + return 2 + + # Ensure the path has a trailing '/'. + tree = sys.argv[1] + if tree[-1] != '/': + tree = tree + '/' + for file in requiredfiles: + if not os.path.exists(tree + file): + print>> sys.stderr, "Kernel source tree missing %s" % (file) + return 1 + + platforms = find_platforms(tree) + + # First, parse global files. + parse_file(tree + 'conf/files', parse_files_line, global_platform) + parse_file(tree + 'conf/options', parse_options_line, global_platform) + parse_file(tree + 'conf/NOTES', parse_notes_line, global_platform) + + # Next, parse MD files. + for platform in platforms: + files_file = tree + 'conf/files.' + platform + if os.path.exists(files_file): + parse_file(files_file, parse_files_line, platform) + options_file = tree + 'conf/options.' + platform + if os.path.exists(options_file): + parse_file(options_file, parse_options_line, platform) + parse_file(tree + platform + '/conf/NOTES', parse_notes_line, platform) + + options.warn() + return 0 + +if __name__ == "__main__": + sys.exit(main()) -- cgit v1.1