bionic/libc/kernel/tools/kernel.py
Elliott Hughes 40596aa005 Make cpp.py less braindead.
The old code ignored operator precedence (!), despite having two tables
of operator precedence. The code's still pretty awful, but I've cleaned
it up enough to fix this, the most important bug.

This patch lets us correctly clean the uapi unistd.h, stat.h, and swab.h files,
and also fixes the mess we were already making of various old kernel
header files. I've added a bunch more tests, fixed the existing tests that
the existing script was already failing (!), and changed the script so that
the tests are run every time the script is run.

We can probably remove some of the old kernel header files that we were
parsing incorrectly, but we can worry about that later.

Bug: 11253477
Change-Id: Ie66c65b3a7ae13b4e98ed8038a6a534f06eae0e5
2013-11-06 12:04:48 -08:00

341 lines
12 KiB
Python

# this file contains definitions related to the Linux kernel itself
#
# list here the macros that you know are always defined/undefined when including
# the kernel headers
#
import sys, cpp, re, os.path, string, time
from defaults import *
verboseSearch = 0
verboseFind = 0
########################################################################
########################################################################
##### #####
##### H E A D E R S C A N N E R #####
##### #####
########################################################################
########################################################################
class HeaderScanner:
"""a class used to non-recursively detect which Linux kernel headers are
used by a given set of input source files"""
# to use the HeaderScanner, do the following:
#
# scanner = HeaderScanner()
# for path in <your list of files>:
# scanner.parseFile(path)
#
# # get the set of Linux headers included by your files
# headers = scanner.getHeaders()
#
# # get the set of of input files that do include Linux headers
# files = scanner.getFiles()
#
# note that the result of getHeaders() is a set of strings, each one
# corresponding to a non-bracketed path name, e.g.:
#
# set("linux/types","asm/types.h")
#
# the default algorithm is pretty smart and will analyze the input
# files with a custom C pre-processor in order to optimize out macros,
# get rid of comments, empty lines, etc..
#
# this avoids many annoying false positives... !!
#
# this regular expression is used to detect include paths that relate to
# the kernel, by default, it selects one of:
# <linux/*>
# <asm/*>
# <asm-generic/*>
# <mtd/*>
#
re_combined_str=\
r"^.*<((%s)/[\d\w_\+\.\-/]*)>.*$" % string.join(kernel_dirs,"|")
re_combined = re.compile(re_combined_str)
# some kernel files choose to include files with relative paths (x86 32/64
# dispatch for instance)
re_rel_dir = re.compile(r'^.*"([\d\w_\+\.\-/]+)".*$')
def __init__(self,config={}):
"""initialize a HeaderScanner"""
self.reset()
self.config = config
def reset(self,config={}):
self.files = set() # set of files being parsed for headers
self.headers = {} # maps headers to set of users
self.config = config
def checkInclude(self, line, from_file, kernel_root=None):
relative = False
m = HeaderScanner.re_combined.match(line)
if kernel_root and not m:
m = HeaderScanner.re_rel_dir.match(line)
relative = True
if not m: return
header = m.group(1)
if from_file:
self.files.add(from_file)
if kernel_root and relative:
hdr_dir = os.path.realpath(os.path.dirname(from_file))
hdr_dir = hdr_dir.replace("%s/" % os.path.realpath(kernel_root),
"")
if hdr_dir:
_prefix = "%s/" % hdr_dir
else:
_prefix = ""
header = "%s%s" % (_prefix, header)
if not header in self.headers:
self.headers[header] = set()
if from_file:
if verboseFind:
print "=== %s uses %s" % (from_file, header)
self.headers[header].add(from_file)
def parseFile(self, path, arch=None, kernel_root=None):
"""parse a given file for Linux headers"""
if not os.path.exists(path):
return
# since tokenizing the file is very slow, we first try a quick grep
# to see if this returns any meaningful results. only if this is true
# do we do the tokenization"""
try:
f = open(path, "rt")
except:
print "!!! can't read '%s'" % path
return
hasIncludes = False
for line in f:
if (HeaderScanner.re_combined.match(line) or
(kernel_root and HeaderScanner.re_rel_dir.match(line))):
hasIncludes = True
break
if not hasIncludes:
if verboseSearch: print "::: " + path
return
if verboseSearch: print "*** " + path
list = cpp.BlockParser().parseFile(path)
if list:
macros = kernel_known_macros.copy()
if kernel_root:
macros.update(self.config)
if arch and arch in kernel_default_arch_macros:
macros.update(kernel_default_arch_macros[arch])
list.optimizeMacros(macros)
list.optimizeIf01()
includes = list.findIncludes()
for inc in includes:
self.checkInclude(inc, path, kernel_root)
def getHeaders(self):
"""return the set of all needed kernel headers"""
return set(self.headers.keys())
def getHeaderUsers(self,header):
"""return the set of all users for a given header"""
return set(self.headers.get(header))
def getAllUsers(self):
"""return a dictionary mapping heaaders to their user set"""
return self.headers.copy()
def getFiles(self):
"""returns the set of files that do include kernel headers"""
return self.files.copy()
##########################################################################
##########################################################################
##### #####
##### H E A D E R F I N D E R #####
##### #####
##########################################################################
##########################################################################
class KernelHeaderFinder:
"""a class used to scan the kernel headers themselves."""
# this is different
# from a HeaderScanner because we need to translate the path returned by
# HeaderScanner.getHeaders() into possibly architecture-specific ones.
#
# for example, <asm/XXXX.h> needs to be translated in <asm-ARCH/XXXX.h>
# where ARCH is appropriately chosen
# here's how to use this:
#
# scanner = HeaderScanner()
# for path in <your list of user sources>:
# scanner.parseFile(path)
#
# used_headers = scanner.getHeaders()
# finder = KernelHeaderFinder(used_headers, [ "arm", "x86" ],
# "<kernel_include_path>")
# all_headers = finder.scanForAllArchs()
#
# not that the result of scanForAllArchs() is a list of relative
# header paths that are not bracketed
#
def __init__(self,headers,archs,kernel_root,kernel_config):
"""init a KernelHeaderScanner,
'headers' is a list or set of headers,
'archs' is a list of architectures
'kernel_root' is the path to the 'include' directory
of your original kernel sources
"""
if len(kernel_root) > 0 and kernel_root[-1] != "/":
kernel_root += "/"
#print "using kernel_root %s" % kernel_root
self.archs = archs
self.searched = set(headers)
self.kernel_root = kernel_root
self.kernel_config = kernel_config
self.needed = {}
self.setArch(arch=None)
def setArch(self,arch=None):
self.curr_arch = arch
self.arch_headers = set()
if arch:
self.prefix = "asm-%s/" % arch
else:
self.prefix = None
def pathFromHeader(self,header):
path = header
if self.prefix and path.startswith("asm/"):
path = "%s%s" % (self.prefix, path[4:])
return path
def pathToHeader(self,path):
if self.prefix and path.startswith(self.prefix):
path = "asm/%s" % path[len(self.prefix):]
return "%s" % path
def setSearchedHeaders(self,headers):
self.searched = set(headers)
def scanForArch(self):
fparser = HeaderScanner(config=self.kernel_config)
workqueue = []
needed = {}
for h in self.searched:
path = self.pathFromHeader(h)
if not path in needed:
needed[path] = set()
workqueue.append(path)
i = 0
while i < len(workqueue):
path = workqueue[i]
i += 1
fparser.parseFile(self.kernel_root + path,
arch=self.curr_arch, kernel_root=self.kernel_root)
for used in fparser.getHeaders():
path = self.pathFromHeader(used)
if not path in needed:
needed[path] = set()
workqueue.append(path)
for user in fparser.getHeaderUsers(used):
needed[path].add(user)
# now copy the arch-specific headers into the global list
for header in needed.keys():
users = needed[header]
if not header in self.needed:
self.needed[header] = set()
for user in users:
self.needed[header].add(user)
def scanForAllArchs(self):
"""scan for all architectures and return the set of all needed kernel headers"""
for arch in self.archs:
self.setArch(arch)
self.scanForArch()
return set(self.needed.keys())
def getHeaderUsers(self,header):
"""return the set of all users for a given header"""
return set(self.needed[header])
def getArchHeaders(self,arch):
"""return the set of all <asm/...> headers required by a given architecture"""
return set() # XXX: TODO
#####################################################################################
#####################################################################################
##### #####
##### C O N F I G P A R S E R #####
##### #####
#####################################################################################
#####################################################################################
class ConfigParser:
"""a class used to parse the Linux kernel .config file"""
re_CONFIG_ = re.compile(r"^(CONFIG_\w+)=(.*)$")
def __init__(self):
self.items = {}
self.duplicates = False
def parseLine(self,line):
line = string.strip(line)
# skip empty and comment lines
if len(line) == 0 or line[0] == "#":
return
m = ConfigParser.re_CONFIG_.match(line)
if not m: return
name = m.group(1)
value = m.group(2)
if name in self.items: # aarg, duplicate value
self.duplicates = True
self.items[name] = value
def parseFile(self,path):
f = file(path, "r")
for line in f:
if len(line) > 0:
if line[-1] == "\n":
line = line[:-1]
if len(line) > 0 and line[-1] == "\r":
line = line[:-1]
self.parseLine(line)
f.close()
def getDefinitions(self):
"""retrieve a dictionary containing definitions for CONFIG_XXX"""
return self.items.copy()
def __repr__(self):
return repr(self.items)
def __str__(self):
return str(self.items)