mbox series

[v7,0/7] rsa: extend rsa_verify() for UEFI secure boot

Message ID 20200221061301.19660-1-takahiro.akashi@linaro.org
Headers show
Series rsa: extend rsa_verify() for UEFI secure boot | expand

Message

AKASHI Takahiro Feb. 21, 2020, 6:12 a.m. UTC
# This patch set is a prerequisite for UEFI secure boot.

The current rsa_verify() requires five parameters for a RSA public key
for efficiency while RSA, in theory, requires only two. In addition,
those parameters are expected to come from FIT image.

So this function won't fit very well when we want to use it for the purpose
of implementing UEFI secure boot, in particular, image authentication
as well as variable authentication, where the essential two parameters
are set to be retrieved from one of X509 certificates in signature
database.

So, in this patch, additional three parameters will be calculated
on the fly when rsa_verify() is called without fdt which should contain
parameters above.

This calculation heavily relies on "big-number (or multi-precision)
library." Therefore some routines from BearSSL[1] under MIT license are
imported in this implementation. See Patch#4.
# Please let me know if this is not appropriate.

Prerequisite:
* public key parser in my "import x509/pkcs7 parser" patch[2]

# Checkpatch will complain with lots of warnings/errors, but
# I intentionally don't fix them for maximum maintainability.

  [1] https://bearssl.org/
  [2] https://lists.denx.de/pipermail/u-boot/2019-November/390127.html

Changes in v7 (Feb 21, 2020)
* rebased to v2020.04-rc2
* no functional change

Changes in v6 (Jan 27, 2020)
* rebased to v2020.01
* change CONFIG_UT_LIB_RSA dependencies after Tom's suggestion (patch#6)
* add patch#7 to enable RSA library test on sandbox

Changes in v5 (Dec 17, 2019)
* modify RSA_VERIFY-related configuration to fix size growth problem
  on some platforms (T1042RDB_PI_NAND_SECURE_BOOT and else), adding
  IMAGE_SIGN_INFO config (patch #1, #2, #6)
* simplify kconfig dependencies;
  - RSA_VERIFY_WITH_PKEY *selects* RSA_VERIFY instead of "depends on"
    (patch#2)
  - remove some implicit dependencies (patch#6)

Changes in v4 (Nov 21, 2019)
* rebased to v2020.01-rc3
* change a function prototype of rsa_gen_key_prop() to return an error
  code (patch#4,#5)
* re-order include files in alphabetical order (patch#6)
* add some comments per Simon's review comments

Changes in v3 (Nov 13, 2019)
* remove RSA_VERIFY_WITH_PKEY, which is to be added in patch#2 (patch#1)
* modify unit test Kconfg due to removal of test/lib/Kconfig (patch#6)

Changes in v2 (Oct 29, 2019)
* fix build errors at Travis CI
* not include linux/kconfig.h (patch#1)
* add a separate patch for adding CONFIG_RSA_VERIFY_WITH_PKEY (patch#2)
* take a prerequisite patch from my "secure boot patch" (patch#3)
* add a dependency on RSA_PUBLIC_KEY_PARSER (patch#4)
* remove "inline" directives (patch#4)
* add function descriptions, which mostly come from BearSSL's src/inner.h
  (patch#4)
* improve Kconfig help text after Simon's comment (patch#5)
* add function description of rsa_verify_with_pkey() (patch#5)
* modify rsa_verify() to use "if (CONFIG_IS_ENABLED(...) " style
  rather than "#ifdef CONFIG_..." (patch#5)
* add function tests (patch#6)

Changes in v1 (Oct 9, 2019)
* fix a build error on pine64-lts_defconfig (reported by Heinrich)
  by defining FIT_IMAGE_ENABLE_VERIFY flag and adding
  SPL_RSA_VERIFY config (patch#1)
* remove FIT-specific code from image-sig.c and put them to new
  image-fit-sig.c to allow us to disable CONFIG_FIT_SIGNATURE (patch#1)
* compile rsa-keyprop.c only if necessary (i.e. if
  CONFIG_RSA_VERIFY_WITH_PKEY) (patch#2)
* add SPDX license identifier in rsa-keyprop.c (patch#2)
* include <common.h> instead of <stdio.h> (patch#2)
* use U-Boot's byteorder helper functions instead of BearSSL's (patch#2)

AKASHI Takahiro (7):
  lib: rsa: decouple rsa from FIT image verification
  rsa: add CONFIG_RSA_VERIFY_WITH_PKEY config
  include: image.h: add key info to image_sign_info
  lib: rsa: generate additional parameters for public key
  lib: rsa: add rsa_verify_with_pkey()
  test: add rsa_verify() unit test
  test: enable RSA library test on sandbox

 Kconfig                            |   4 +
 common/Kconfig                     |   7 +
 common/Makefile                    |   3 +-
 common/image-fit-sig.c             | 417 +++++++++++++++++
 common/image-fit.c                 |   6 +-
 common/image-sig.c                 | 396 ----------------
 configs/sandbox64_defconfig        |   1 +
 configs/sandbox_defconfig          |   1 +
 configs/sandbox_flattree_defconfig |   1 +
 configs/sandbox_spl_defconfig      |   1 +
 include/image.h                    |  21 +-
 include/u-boot/rsa-mod-exp.h       |  23 +
 lib/rsa/Kconfig                    |  27 ++
 lib/rsa/Makefile                   |   3 +-
 lib/rsa/rsa-keyprop.c              | 725 +++++++++++++++++++++++++++++
 lib/rsa/rsa-verify.c               | 137 ++++--
 test/Kconfig                       |  10 +
 test/lib/Makefile                  |   1 +
 test/lib/rsa.c                     | 206 ++++++++
 tools/Makefile                     |   2 +-
 20 files changed, 1556 insertions(+), 436 deletions(-)
 create mode 100644 common/image-fit-sig.c
 create mode 100644 lib/rsa/rsa-keyprop.c
 create mode 100644 test/lib/rsa.c

Comments

Tom Rini Feb. 21, 2020, 5:18 p.m. UTC | #1
On Fri, Feb 21, 2020 at 03:12:54PM +0900, AKASHI Takahiro wrote:

> # This patch set is a prerequisite for UEFI secure boot.
> 
> The current rsa_verify() requires five parameters for a RSA public key
> for efficiency while RSA, in theory, requires only two. In addition,
> those parameters are expected to come from FIT image.
> 
> So this function won't fit very well when we want to use it for the purpose
> of implementing UEFI secure boot, in particular, image authentication
> as well as variable authentication, where the essential two parameters
> are set to be retrieved from one of X509 certificates in signature
> database.
> 
> So, in this patch, additional three parameters will be calculated
> on the fly when rsa_verify() is called without fdt which should contain
> parameters above.
> 
> This calculation heavily relies on "big-number (or multi-precision)
> library." Therefore some routines from BearSSL[1] under MIT license are
> imported in this implementation. See Patch#4.
> # Please let me know if this is not appropriate.
> 
> Prerequisite:
> * public key parser in my "import x509/pkcs7 parser" patch[2]

This has been applied a long while back.

And for the record, without http://patchwork.ozlabs.org/patch/1239098/
applied sandbox fails to build.  I had said I would take care of that
specific issue, so I'm just noting it here.  I'm kicking off a larger
test now.
AKASHI Takahiro Feb. 25, 2020, 4:55 a.m. UTC | #2
On Fri, Feb 21, 2020 at 12:18:41PM -0500, Tom Rini wrote:
> On Fri, Feb 21, 2020 at 03:12:54PM +0900, AKASHI Takahiro wrote:
> 
> > # This patch set is a prerequisite for UEFI secure boot.
> > 
> > The current rsa_verify() requires five parameters for a RSA public key
> > for efficiency while RSA, in theory, requires only two. In addition,
> > those parameters are expected to come from FIT image.
> > 
> > So this function won't fit very well when we want to use it for the purpose
> > of implementing UEFI secure boot, in particular, image authentication
> > as well as variable authentication, where the essential two parameters
> > are set to be retrieved from one of X509 certificates in signature
> > database.
> > 
> > So, in this patch, additional three parameters will be calculated
> > on the fly when rsa_verify() is called without fdt which should contain
> > parameters above.
> > 
> > This calculation heavily relies on "big-number (or multi-precision)
> > library." Therefore some routines from BearSSL[1] under MIT license are
> > imported in this implementation. See Patch#4.
> > # Please let me know if this is not appropriate.
> > 
> > Prerequisite:
> > * public key parser in my "import x509/pkcs7 parser" patch[2]
> 
> This has been applied a long while back.

Yes, I forgot to remove this line.

> And for the record, without http://patchwork.ozlabs.org/patch/1239098/
> applied sandbox fails to build.  I had said I would take care of that
> specific issue, so I'm just noting it here.  I'm kicking off a larger
> test now.

Thank you!
-Takahiro Akashi

> -- 
> Tom