[U-Boot] [PATCH v3 0/1] x86: Support microcode in a single place

As part of the effort to support Atom processors it became apparent the they can only support microcode updates before RAM is available. In this early environment the device tree is not available.
This series solves this problem by updating ifdtool to provide very early microcode access on these platforms.
Changes in v3: - Drop patches that have already been applied - Fix pylint problems - Add date string to microcode - Automatically create microcode directory if needed - Move microcode program into a function
Changes in v2: - Drop patches that have already been applied - Add function comments - Fix problem with specifying the full microcode filename - Allow 'list' command to display a list of matching microcode chunks - Print a message when generating microcode - Expand the help a little - Print an error when an ambiguous microcode model is given
Simon Glass (1): x86: Add a script to process Intel microcode files
tools/microcode-tool | 1 + tools/microcode-tool.py | 253 ++++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 254 insertions(+) create mode 120000 tools/microcode-tool create mode 100755 tools/microcode-tool.py

Intel delivers microcode updates in a microcode.dat file which must be split up into individual files for each CPU. Add a tool which performs this task. It can list available microcode updates for each model and produce a new microcode update in U-Boot's .dtsi format.
Signed-off-by: Simon Glass sjg@chromium.org ---
Changes in v3: - Drop patches that have already been applied - Fix pylint problems - Add date string to microcode - Automatically create microcode directory if needed - Move microcode program into a function
Changes in v2: - Drop patches that have already been applied - Add function comments - Fix problem with specifying the full microcode filename - Allow 'list' command to display a list of matching microcode chunks - Print a message when generating microcode - Expand the help a little - Print an error when an ambiguous microcode model is given
tools/microcode-tool | 1 + tools/microcode-tool.py | 253 ++++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 254 insertions(+) create mode 120000 tools/microcode-tool create mode 100755 tools/microcode-tool.py
diff --git a/tools/microcode-tool b/tools/microcode-tool new file mode 120000 index 0000000..8be8507 --- /dev/null +++ b/tools/microcode-tool @@ -0,0 +1 @@ +microcode-tool.py \ No newline at end of file diff --git a/tools/microcode-tool.py b/tools/microcode-tool.py new file mode 100755 index 0000000..003716d --- /dev/null +++ b/tools/microcode-tool.py @@ -0,0 +1,253 @@ +#!/usr/bin/env python +# +# Copyright (c) 2014 Google, Inc +# +# SPDX-License-Identifier: GPL-2.0+ +# +# Intel microcode update tool + +from optparse import OptionParser +import os +import re +import struct +import sys + +MICROCODE_DIR = 'arch/x86/dts/microcode' + +class Microcode: + """Holds information about the microcode for a particular model of CPU. + + Attributes: + name: Name of the CPU this microcode is for, including any version + information (e.g. 'm12206a7_00000029') + model: Model code string (this is cpuid(1).eax, e.g. '206a7') + words: List of hex words containing the microcode. The first 16 words + are the public header. + """ + def __init__(self, name, data): + self.name = name + # Convert data into a list of hex words + self.words = [] + for value in ''.join(data).split(','): + hexval = value.strip() + if hexval: + self.words.append(int(hexval, 0)) + + # The model is in the 4rd hex word + self.model = '%x' % self.words[3] + +def ParseFile(fname): + """Parse a micrcode.dat file and return the component parts + + Args: + fname: Filename to parse + Returns: + 3-Tuple: + date: String containing date from the file's header + license_text: List of text lines for the license file + microcodes: List of Microcode objects from the file + """ + re_date = re.compile('/* *(.* [0-9]{4}) **/$') + re_license = re.compile('/[^-*+] *(.*)$') + re_name = re.compile('/* *(.*).inc **/', re.IGNORECASE) + microcodes = {} + license_text = [] + date = '' + data = [] + name = None + with open(fname) as fd: + for line in fd: + line = line.rstrip() + m_date = re_date.match(line) + m_license = re_license.match(line) + m_name = re_name.match(line) + if m_name: + if name: + microcodes[name] = Microcode(name, data) + name = m_name.group(1).lower() + data = [] + elif m_license: + license_text.append(m_license.group(1)) + elif m_date: + date = m_date.group(1) + else: + data.append(line) + if name: + microcodes[name] = Microcode(name, data) + return date, license_text, microcodes + +def List(date, microcodes, model): + """List the available microcode chunks + + Args: + date: Date of the microcode file + microcodes: Dict of Microcode objects indexed by name + model: Model string to search for, or None + """ + print 'Date: %s' % date + if model: + mcode_list, tried = FindMicrocode(microcodes, model.lower()) + print 'Matching models %s:' % (', '.join(tried)) + else: + print 'All models:' + mcode_list = [microcodes[m] for m in microcodes.keys()] + for mcode in mcode_list: + print '%-20s: model %s' % (mcode.name, mcode.model) + +def FindMicrocode(microcodes, model): + """Find all the microcode chunks which match the given model. + + This model is something like 306a9 (the value returned in eax from + cpuid(1) when running on Intel CPUs). But we allow a partial match, + omitting the last 1 or two characters to allow many families to have the + same microcode. + + If the model name is ambiguous we return a list of matches. + + Args: + microcodes: Dict of Microcode objects indexed by name + model: String containing model name to find + Returns: + Tuple: + List of matching Microcode objects + List of abbreviations we tried + """ + # Allow a full name to be used + mcode = microcodes.get(model) + if mcode: + return [mcode], [] + + tried = [] + found = [] + for i in range(3): + abbrev = model[:-i] if i else model + tried.append(abbrev) + for mcode in microcodes.values(): + if mcode.model.startswith(abbrev): + found.append(mcode) + if found: + break + return found, tried + +def CreateFile(date, license_text, mcode, outfile): + """Create a microcode file in U-Boot's .dtsi format + + Args: + date: String containing date of original microcode file + license: List of text lines for the license file + mcode: Microcode object to write + outfile: Filename to write to ('-' for stdout) + """ + out = '''/*%s + * --- + * This is a device tree fragment. Use #include to add these properties to a + * node. + * + * Date: %s + */ + +compatible = "intel,microcode"; +intel,header-version = <%d>; +intel,update-revision = <%#x>; +intel,date-code = <%#x>; +intel,processor-signature = <%#x>; +intel,checksum = <%#x>; +intel,loader-revision = <%d>; +intel,processor-flags = <%#x>; + +/* The first 48-bytes are the public header which repeats the above data */ +data = <%s +\t>;''' + words = '' + for i in range(len(mcode.words)): + if not (i & 3): + words += '\n' + val = mcode.words[i] + # Change each word so it will be little-endian in the FDT + # This data is needed before RAM is available on some platforms so we + # cannot do an endianness swap on boot. + val = struct.unpack("<I", struct.pack(">I", val))[0] + words += '\t%#010x' % val + + # Take care to avoid adding a space before a tab + text = '' + for line in license_text: + if line[0] == '\t': + text += '\n *' + line + else: + text += '\n * ' + line + args = [text, date] + args += [mcode.words[i] for i in range(7)] + args.append(words) + if outfile == '-': + print out % tuple(args) + else: + if not outfile: + if not os.path.exists(MICROCODE_DIR): + print >> sys.stderr, "Creating directory '%s'" % MICROCODE_DIR + os.makedirs(MICROCODE_DIR) + outfile = os.path.join(MICROCODE_DIR, mcode.name + '.dtsi') + print >> sys.stderr, "Writing microcode for '%s' to '%s'" % ( + mcode.name, outfile) + with open(outfile, 'w') as fd: + print >> fd, out % tuple(args) + +def MicrocodeTool(): + """Run the microcode tool""" + commands = 'create,license,list'.split(',') + parser = OptionParser() + parser.add_option('-d', '--mcfile', type='string', action='store', + help='Name of microcode.dat file') + parser.add_option('-m', '--model', type='string', action='store', + help='Model name to extract') + parser.add_option('-o', '--outfile', type='string', action='store', + help='Filename to use for output (- for stdout), default is' + ' %s/<name>.dtsi' % MICROCODE_DIR) + parser.usage += """ command + + Process an Intel microcode file (use -h for help). Commands: + + create Create microcode .dtsi file for a model + list List available models in microcode file + license Print the license + + Typical usage: + + ./tools/microcode-tool -d microcode.dat -m 306a create + + This will find the appropriate file and write it to %s.""" % MICROCODE_DIR + + (options, args) = parser.parse_args() + if not args: + parser.error('Please specify a command') + cmd = args[0] + if cmd not in commands: + parser.error("Unknown command '%s'" % cmd) + + if not options.mcfile: + parser.error('You must specify a microcode file') + date, license_text, microcodes = ParseFile(options.mcfile) + + if cmd == 'list': + List(date, microcodes, options.model) + elif cmd == 'license': + print '\n'.join(license_text) + elif cmd == 'create': + if not options.model: + parser.error('You must specify a model to create') + model = options.model.lower() + mcode_list, tried = FindMicrocode(microcodes, model) + if not mcode_list: + parser.error("Unknown model '%s' (%s) - try 'list' to list" % + (model, ', '.join(tried))) + if len(mcode_list) > 1: + parser.error("Ambiguous model '%s' (%s) matched %s - try 'list' " + "to list or specify a particular file" % + (model, ', '.join(tried), + ', '.join([m.name for m in mcode_list]))) + CreateFile(date, license_text, mcode_list[0], options.outfile) + else: + parser.error("Unknown command '%s'" % cmd) + +if __name__ == "__main__": + MicrocodeTool()

On Wed, Dec 17, 2014 at 11:21 AM, Simon Glass sjg@chromium.org wrote:
Intel delivers microcode updates in a microcode.dat file which must be split up into individual files for each CPU. Add a tool which performs this task. It can list available microcode updates for each model and produce a new microcode update in U-Boot's .dtsi format.
Signed-off-by: Simon Glass sjg@chromium.org
Changes in v3:
- Drop patches that have already been applied
- Fix pylint problems
- Add date string to microcode
- Automatically create microcode directory if needed
- Move microcode program into a function
Changes in v2:
- Drop patches that have already been applied
- Add function comments
- Fix problem with specifying the full microcode filename
- Allow 'list' command to display a list of matching microcode chunks
- Print a message when generating microcode
- Expand the help a little
- Print an error when an ambiguous microcode model is given
tools/microcode-tool | 1 + tools/microcode-tool.py | 253 ++++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 254 insertions(+) create mode 120000 tools/microcode-tool create mode 100755 tools/microcode-tool.py
diff --git a/tools/microcode-tool b/tools/microcode-tool new file mode 120000 index 0000000..8be8507 --- /dev/null +++ b/tools/microcode-tool @@ -0,0 +1 @@ +microcode-tool.py \ No newline at end of file diff --git a/tools/microcode-tool.py b/tools/microcode-tool.py new file mode 100755 index 0000000..003716d --- /dev/null +++ b/tools/microcode-tool.py @@ -0,0 +1,253 @@ +#!/usr/bin/env python +# +# Copyright (c) 2014 Google, Inc +# +# SPDX-License-Identifier: GPL-2.0+ +# +# Intel microcode update tool
+from optparse import OptionParser +import os +import re +import struct +import sys
+MICROCODE_DIR = 'arch/x86/dts/microcode'
+class Microcode:
- """Holds information about the microcode for a particular model of CPU.
- Attributes:
name: Name of the CPU this microcode is for, including any version
information (e.g. 'm12206a7_00000029')
model: Model code string (this is cpuid(1).eax, e.g. '206a7')
words: List of hex words containing the microcode. The first 16 words
are the public header.
- """
- def __init__(self, name, data):
self.name = name
# Convert data into a list of hex words
self.words = []
for value in ''.join(data).split(','):
hexval = value.strip()
if hexval:
self.words.append(int(hexval, 0))
# The model is in the 4rd hex word
self.model = '%x' % self.words[3]
+def ParseFile(fname):
- """Parse a micrcode.dat file and return the component parts
- Args:
fname: Filename to parse
- Returns:
3-Tuple:
date: String containing date from the file's header
license_text: List of text lines for the license file
microcodes: List of Microcode objects from the file
- """
- re_date = re.compile('/* *(.* [0-9]{4}) **/$')
- re_license = re.compile('/[^-*+] *(.*)$')
- re_name = re.compile('/* *(.*).inc **/', re.IGNORECASE)
- microcodes = {}
- license_text = []
- date = ''
- data = []
- name = None
- with open(fname) as fd:
for line in fd:
line = line.rstrip()
m_date = re_date.match(line)
m_license = re_license.match(line)
m_name = re_name.match(line)
if m_name:
if name:
microcodes[name] = Microcode(name, data)
name = m_name.group(1).lower()
data = []
elif m_license:
license_text.append(m_license.group(1))
elif m_date:
date = m_date.group(1)
else:
data.append(line)
- if name:
microcodes[name] = Microcode(name, data)
- return date, license_text, microcodes
+def List(date, microcodes, model):
- """List the available microcode chunks
- Args:
date: Date of the microcode file
microcodes: Dict of Microcode objects indexed by name
model: Model string to search for, or None
- """
- print 'Date: %s' % date
- if model:
mcode_list, tried = FindMicrocode(microcodes, model.lower())
print 'Matching models %s:' % (', '.join(tried))
- else:
print 'All models:'
mcode_list = [microcodes[m] for m in microcodes.keys()]
- for mcode in mcode_list:
print '%-20s: model %s' % (mcode.name, mcode.model)
+def FindMicrocode(microcodes, model):
- """Find all the microcode chunks which match the given model.
- This model is something like 306a9 (the value returned in eax from
- cpuid(1) when running on Intel CPUs). But we allow a partial match,
- omitting the last 1 or two characters to allow many families to have the
- same microcode.
- If the model name is ambiguous we return a list of matches.
- Args:
microcodes: Dict of Microcode objects indexed by name
model: String containing model name to find
- Returns:
Tuple:
List of matching Microcode objects
List of abbreviations we tried
- """
- # Allow a full name to be used
- mcode = microcodes.get(model)
- if mcode:
return [mcode], []
- tried = []
- found = []
- for i in range(3):
abbrev = model[:-i] if i else model
tried.append(abbrev)
for mcode in microcodes.values():
if mcode.model.startswith(abbrev):
found.append(mcode)
if found:
break
- return found, tried
+def CreateFile(date, license_text, mcode, outfile):
- """Create a microcode file in U-Boot's .dtsi format
- Args:
date: String containing date of original microcode file
license: List of text lines for the license file
mcode: Microcode object to write
outfile: Filename to write to ('-' for stdout)
- """
- out = '''/*%s
- This is a device tree fragment. Use #include to add these properties to a
- node.
- Date: %s
- */
+compatible = "intel,microcode"; +intel,header-version = <%d>; +intel,update-revision = <%#x>; +intel,date-code = <%#x>; +intel,processor-signature = <%#x>; +intel,checksum = <%#x>; +intel,loader-revision = <%d>; +intel,processor-flags = <%#x>;
+/* The first 48-bytes are the public header which repeats the above data */ +data = <%s +\t>;'''
- words = ''
- for i in range(len(mcode.words)):
if not (i & 3):
words += '\n'
val = mcode.words[i]
# Change each word so it will be little-endian in the FDT
# This data is needed before RAM is available on some platforms so we
# cannot do an endianness swap on boot.
val = struct.unpack("<I", struct.pack(">I", val))[0]
words += '\t%#010x' % val
- # Take care to avoid adding a space before a tab
- text = ''
- for line in license_text:
if line[0] == '\t':
text += '\n *' + line
else:
text += '\n * ' + line
- args = [text, date]
- args += [mcode.words[i] for i in range(7)]
- args.append(words)
- if outfile == '-':
print out % tuple(args)
- else:
if not outfile:
if not os.path.exists(MICROCODE_DIR):
print >> sys.stderr, "Creating directory '%s'" % MICROCODE_DIR
os.makedirs(MICROCODE_DIR)
outfile = os.path.join(MICROCODE_DIR, mcode.name + '.dtsi')
print >> sys.stderr, "Writing microcode for '%s' to '%s'" % (
mcode.name, outfile)
with open(outfile, 'w') as fd:
print >> fd, out % tuple(args)
+def MicrocodeTool():
- """Run the microcode tool"""
- commands = 'create,license,list'.split(',')
- parser = OptionParser()
- parser.add_option('-d', '--mcfile', type='string', action='store',
help='Name of microcode.dat file')
- parser.add_option('-m', '--model', type='string', action='store',
help='Model name to extract')
- parser.add_option('-o', '--outfile', type='string', action='store',
help='Filename to use for output (- for stdout), default is'
' %s/<name>.dtsi' % MICROCODE_DIR)
- parser.usage += """ command
- Process an Intel microcode file (use -h for help). Commands:
create Create microcode .dtsi file for a model
list List available models in microcode file
license Print the license
- Typical usage:
./tools/microcode-tool -d microcode.dat -m 306a create
- This will find the appropriate file and write it to %s.""" % MICROCODE_DIR
- (options, args) = parser.parse_args()
- if not args:
parser.error('Please specify a command')
- cmd = args[0]
- if cmd not in commands:
parser.error("Unknown command '%s'" % cmd)
- if not options.mcfile:
parser.error('You must specify a microcode file')
- date, license_text, microcodes = ParseFile(options.mcfile)
- if cmd == 'list':
List(date, microcodes, options.model)
- elif cmd == 'license':
print '\n'.join(license_text)
- elif cmd == 'create':
if not options.model:
parser.error('You must specify a model to create')
model = options.model.lower()
mcode_list, tried = FindMicrocode(microcodes, model)
if not mcode_list:
parser.error("Unknown model '%s' (%s) - try 'list' to list" %
(model, ', '.join(tried)))
if len(mcode_list) > 1:
parser.error("Ambiguous model '%s' (%s) matched %s - try 'list' "
"to list or specify a particular file" %
(model, ', '.join(tried),
', '.join([m.name for m in mcode_list])))
CreateFile(date, license_text, mcode_list[0], options.outfile)
- else:
parser.error("Unknown command '%s'" % cmd)
+if __name__ == "__main__":
- MicrocodeTool()
--
Tested-by: Bin Meng bmeng.cn@gmail.com
participants (2)
-
Bin Meng
-
Simon Glass