]> git.proxmox.com Git - efi-boot-shim.git/log
efi-boot-shim.git
11 years agoPass the right arguments to EFI_PXE_BASE_CODE_TFTP_READ_FILE
Steve Langasek [Fri, 20 Sep 2013 16:29:23 +0000 (11:29 -0500)]
Pass the right arguments to EFI_PXE_BASE_CODE_TFTP_READ_FILE

A wrong pointer was being passed to EFI_PXE_BASE_CODE_TFTP_READ_FILE,
preventing us from getting the file size back from the tftp call, ensuring
that we don't have enough information to properly secureboot-validate the
retrieved image.

11 years agoAdd MokListRT option rom entry.
Peter Jones [Mon, 23 Sep 2013 17:24:32 +0000 (13:24 -0400)]
Add MokListRT option rom entry.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoUpdate TODO with missing description.
Peter Jones [Mon, 23 Sep 2013 15:04:48 +0000 (11:04 -0400)]
Update TODO with missing description.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoMade TODO represent the present.
Peter Jones [Mon, 23 Sep 2013 14:40:18 +0000 (10:40 -0400)]
Made TODO represent the present.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoMake SHIM_LOCK_GUID a first-class object with a symbol.
Peter Jones [Mon, 23 Sep 2013 14:22:48 +0000 (10:22 -0400)]
Make SHIM_LOCK_GUID a first-class object with a symbol.

Right now the CA is checking if shim builds expose a particular version
of the shim protocol.  To do this, they're looking for SHIM_LOCK_GUID's
value in the resulting binary.

Currently, with SHIM_LOCK_GUID as a macro that gets assigned to local
variables, that means they have to compensate for mov instructions mixed
in with the actual value.  This is completely absurd, so promote it to a
first-class object with a symbol to make it both easy to find and
continuous.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoDon't print things on the screen by default when everything works.
Peter Jones [Thu, 25 Jul 2013 14:51:05 +0000 (10:51 -0400)]
Don't print things on the screen by default when everything works.

There's no point to this text, and it generally confuses people.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoDon't print that fallback isn't found in should_use_fallback()
Peter Jones [Thu, 20 Jun 2013 16:41:14 +0000 (12:41 -0400)]
Don't print that fallback isn't found in should_use_fallback()

The call can simply fail if it isn't found - which will be the case on
removeable install media.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoFix some pointer casting issues.
Peter Jones [Tue, 11 Jun 2013 18:58:25 +0000 (14:58 -0400)]
Fix some pointer casting issues.

This also fixes the size of an empty vendor_cert or dbx_cert.

Signed-off-by: Peter Jones <shim-owner@fedoraproject.org>
11 years agoBump version to 0.4
Peter Jones [Mon, 10 Jun 2013 21:44:16 +0000 (17:44 -0400)]
Bump version to 0.4

Since I've finally merged in the "sections" branch, best to increment
the version number.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoMake DBX be included in build if the environment is set right.
Peter Jones [Mon, 10 Jun 2013 21:48:33 +0000 (17:48 -0400)]
Make DBX be included in build if the environment is set right.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoMake .vendor_cert get the right flags set.
Peter Jones [Mon, 10 Jun 2013 21:27:48 +0000 (17:27 -0400)]
Make .vendor_cert get the right flags set.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoadd a .gitignore
Peter Jones [Tue, 4 Jun 2013 13:22:11 +0000 (09:22 -0400)]
add a .gitignore

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoMove embedded certificates to their own section.
Peter Jones [Wed, 20 Mar 2013 17:18:34 +0000 (13:18 -0400)]
Move embedded certificates to their own section.

With this change, the embedded certificate and dbx lists (vendor_cert,
vendor_cert_size, vendor_dbx, and vendor_dbx_size) wind up being in a
section named .vendor_cert, and so will look something like:
------
fenchurch:~/devel/github.com/shim$ objdump -h shim.efi

shim.efi:     file format pei-x86-64

Sections:
Idx Name          Size      VMA               LMA               File off  Algn
  0 .eh_frame     000174a8  0000000000005000  0000000000005000  00000400  2**3
                  CONTENTS, ALLOC, LOAD, READONLY, DATA
  1 .text         000aa7e1  000000000001d000  000000000001d000  00017a00  2**4
                  CONTENTS, ALLOC, LOAD, READONLY, CODE
  2 .reloc        0000000a  00000000000c8000  00000000000c8000  000c2200  2**0
                  CONTENTS, ALLOC, LOAD, READONLY, DATA
  3 .data         00031228  00000000000c9000  00000000000c9000  000c2400  2**5
                  CONTENTS, ALLOC, LOAD, DATA
  4 .vendor_cert  00000375  00000000000fb000  00000000000fb000  000f3800  2**0
                  CONTENTS, READONLY
  5 .dynamic      000000f0  00000000000fc000  00000000000fc000  000f3c00  2**3
                  CONTENTS, ALLOC, LOAD, DATA
  6 .rela         0002afa8  00000000000fd000  00000000000fd000  000f3e00  2**3
                  CONTENTS, ALLOC, LOAD, READONLY, DATA
  7 .dynsym       0000f1f8  0000000000128000  0000000000128000  0011ee00  2**3
                  CONTENTS, ALLOC, LOAD, READONLY, DATA
------

This simplifies a security audit, because it means that different
versions of shim with substantially the same code with different keys
will be more easily comperable, and therefore logic differences may be
more easily identified.

This also means that if there's a trusted build you want to use, you can
remove the certificates, implant new ones, and have it signed, and the
code sections won't change.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agovendor_cert_size's size in the binary should be 4, not -4.
Peter Jones [Wed, 20 Mar 2013 17:16:08 +0000 (13:16 -0400)]
vendor_cert_size's size in the binary should be 4, not -4.

The thing about subtraction is that the minuend needs to be before the
subtrahend in the text.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoRemove FALLBACK_OBJS during clean as well.
Peter Jones [Mon, 10 Jun 2013 21:30:14 +0000 (17:30 -0400)]
Remove FALLBACK_OBJS during clean as well.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoMake sure all the Makefiles use the same arguments for mmx/sse/ms_abi.
Peter Jones [Mon, 10 Jun 2013 20:38:05 +0000 (16:38 -0400)]
Make sure all the Makefiles use the same arguments for mmx/sse/ms_abi.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoEFI_PXE_BASE_CODE_DHCPV6_PACKET is in gnu-efi-3.0t
Peter Jones [Thu, 30 May 2013 21:23:19 +0000 (17:23 -0400)]
EFI_PXE_BASE_CODE_DHCPV6_PACKET is in gnu-efi-3.0t

11 years agoDon't use MMX and SSE registers, they aren't initialized.
Peter Jones [Fri, 31 May 2013 19:22:37 +0000 (15:22 -0400)]
Don't use MMX and SSE registers, they aren't initialized.

GCC 4.8.0 will try to use these by default, and you'll wind up looping
across the (uninitialized!) trap handler for uninitialized instructions.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoBump version to 0.3
Peter Jones [Thu, 16 May 2013 15:03:25 +0000 (11:03 -0400)]
Bump version to 0.3

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoUse MS ABI instead of terrible wrappers.
Peter Jones [Thu, 16 May 2013 14:09:46 +0000 (10:09 -0400)]
Use MS ABI instead of terrible wrappers.

This means that we now require gnu-efi 3.0s

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoUse the correct define on Open.
Peter Jones [Wed, 15 May 2013 17:38:44 +0000 (13:38 -0400)]
Use the correct define on Open.

The value here doesn't actually change any, but we should still use the
right name.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoAdd some error messages when things don't work.
Peter Jones [Wed, 15 May 2013 17:38:27 +0000 (13:38 -0400)]
Add some error messages when things don't work.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoRemove some unnecessary code.
Peter Jones [Wed, 15 May 2013 17:38:00 +0000 (13:38 -0400)]
Remove some unnecessary code.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoFix some minor type errors.
Peter Jones [Wed, 15 May 2013 17:36:31 +0000 (13:36 -0400)]
Fix some minor type errors.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoPass parameters correctly when booting.
Peter Jones [Tue, 14 May 2013 17:10:34 +0000 (13:10 -0400)]
Pass parameters correctly when booting.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years ago[fallback] Try to execute the first new boot option.
Peter Jones [Tue, 30 Apr 2013 14:30:04 +0000 (10:30 -0400)]
[fallback] Try to execute the first new boot option.

I'm told rebooting is sometimes unreliable when called here, and we'll
get bootx64.efi loaded anyway.  I'll just assume that's true and try to
load the first option, since it's clearly what we'd prefer happens next.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoExplain byte order handling better.
Peter Jones [Tue, 30 Apr 2013 14:58:59 +0000 (10:58 -0400)]
Explain byte order handling better.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoDon't update BootOrder until all csv files are processed
Gary Ching-Pang Lin [Tue, 30 Apr 2013 09:04:16 +0000 (17:04 +0800)]
Don't update BootOrder until all csv files are processed

11 years agoReset the system after restoring the boot entries
Gary Ching-Pang Lin [Tue, 30 Apr 2013 06:16:11 +0000 (14:16 +0800)]
Reset the system after restoring the boot entries

11 years agoFix crash due to memory allocation
Gary Ching-Pang Lin [Tue, 30 Apr 2013 03:58:51 +0000 (11:58 +0800)]
Fix crash due to memory allocation

11 years agoGet rid of extra "continue".
Peter Jones [Fri, 26 Apr 2013 18:10:34 +0000 (14:10 -0400)]
Get rid of extra "continue".

It's confusing, and it doesn't actually accomplish anything when applied
to *either* loop.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoFix error checking on AllocateZeroPool() in update_boot_order()
Peter Jones [Fri, 26 Apr 2013 17:57:36 +0000 (13:57 -0400)]
Fix error checking on AllocateZeroPool() in update_boot_order()

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoMake shim use fallback when appropriate.
Peter Jones [Fri, 26 Apr 2013 16:14:51 +0000 (12:14 -0400)]
Make shim use fallback when appropriate.

If we're called as /BOOT/EFI/BOOT*.EFI, and /BOOT/EFI/FALLBACK.EFI exists,
give it a shot.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoAdd a fallback loader for when shim is invoked as BOOTX64.EFI
Peter Jones [Fri, 15 Mar 2013 16:40:47 +0000 (12:40 -0400)]
Add a fallback loader for when shim is invoked as BOOTX64.EFI

If shim is invoked as \EFI\BOOT\BOOT*.EFI and a file exists named
\EFI\BOOT\FALLBACK.EFI, try it instead of our second stage.  So don't
put fallback.efi on your install media in \EFI\BOOT, because that won't
do whatever it is you're hoping for, unless you're hoping not to start
the installer.

So here's the process for using this:
in /EFI/fedora/ (or whichever directory you happen to own), you put:
  shim.efi
  grub.efi
  boot.csv - format is: shim.efi,Nice Label,cmdline arguments,comments
           - filenames refer only to files in this directory, with no
     leading characters such as L"./" or L"/EFI/fedora/"
           - note that while this is CSV, the character encoding is
     UCS-2

and if /EFI/BOOT/BOOTX64.EFI doesn't already exist, then in /EFI/BOOT:
  shim.efi as BOOTX64.EFI
  fallback.efi

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoAdd StrCSpn()
Peter Jones [Fri, 26 Apr 2013 16:09:01 +0000 (12:09 -0400)]
Add StrCSpn()

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoAdopt the UEFI shell style LoadOptions
Gary Ching-Pang Lin [Mon, 14 Jan 2013 08:53:19 +0000 (16:53 +0800)]
Adopt the UEFI shell style LoadOptions

The previous commit, 14d4b8e, caused shim failed to parse the name
of the 2nd stage loader in UEFI shell. Amend parsing of the name the
2nd stage loader to be compatible with UEFI shell.

To create an boot entry for elilo.efi:

 # efibootmgr -c -L "shim elilo" -l "efi\\shim.efi" -u "shim.efi elilo.efi"

11 years agoGet rid of extra label.
Peter Jones [Fri, 26 Apr 2013 15:29:37 +0000 (11:29 -0400)]
Get rid of extra label.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoDon't put the directory in the file path twice.
Peter Jones [Fri, 26 Apr 2013 15:38:10 +0000 (11:38 -0400)]
Don't put the directory in the file path twice.

Sometimes when we're creating paths, the ImagePath can contain the
directory name already.  If that happens, don't add it in again.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoUEFI Shell sticks the UCS2 of li->FilePath in li->LoadOptions. Ignore it.
Peter Jones [Fri, 26 Apr 2013 15:31:40 +0000 (11:31 -0400)]
UEFI Shell sticks the UCS2 of li->FilePath in li->LoadOptions. Ignore it.

If li->LoadOptions tells us to execute our own binary, it's clearly not
what we want to do for the second stage.  So simply ignore that case.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoMake sure the menu shows when the callback fails
Gary Ching-Pang Lin [Wed, 26 Dec 2012 03:44:46 +0000 (11:44 +0800)]
Make sure the menu shows when the callback fails

Since Pause() doesn't clear the key from the input queue, the next
ReadKeyStroke reads the queued key instead of the new one. If the
user presses "Enter", MokManager exits directly without showing
the menu again.

11 years agoAdd support for deleting specific keys
Gary Ching-Pang Lin [Fri, 21 Dec 2012 09:59:31 +0000 (17:59 +0800)]
Add support for deleting specific keys

11 years agoMOK doesn't include the signature owner
Gary Ching-Pang Lin [Fri, 21 Dec 2012 07:02:09 +0000 (15:02 +0800)]
MOK doesn't include the signature owner

11 years agoAdd a general function for password matching
Gary Ching-Pang Lin [Thu, 20 Dec 2012 09:02:35 +0000 (17:02 +0800)]
Add a general function for password matching

11 years agoReboot the system after enrolling/erasing keys
Gary Ching-Pang Lin [Mon, 10 Dec 2012 10:24:45 +0000 (18:24 +0800)]
Reboot the system after enrolling/erasing keys

11 years agoClear the screen before erasing keys
Gary Ching-Pang Lin [Mon, 10 Dec 2012 09:54:05 +0000 (17:54 +0800)]
Clear the screen before erasing keys

11 years agoGet the second stage loader from the Load Options
Gary Ching-Pang Lin [Thu, 6 Dec 2012 09:47:26 +0000 (17:47 +0800)]
Get the second stage loader from the Load Options

This commit replaces the 2nd stage loader path with the first
argument in the Load Options and moves the rest arguments (if any)
to the Load Options for the 2nd stage loader.

For example, to make shim to load elilo.efi, just create a new
boot entry with efibootmgr:

 # efibootmgr -c -L "shim elilo" -l "efi\\shim.efi" -u "elilo.efi"

11 years agoRemove debug code
Matthew Garrett [Wed, 28 Nov 2012 04:52:27 +0000 (23:52 -0500)]
Remove debug code

secure_mode() was altered to always return true for debug purposes, and this
accidentally got committed to mainline. Fix that.

11 years agoSign MokManager with a locally-generated key
Matthew Garrett [Sat, 24 Nov 2012 05:07:11 +0000 (00:07 -0500)]
Sign MokManager with a locally-generated key

shim needs to verify that MokManager hasn't been modified, but we want to
be able to support configurations where shim is shipped without a vendor
certificate. This patch adds support for generating a certificate at build
time, incorporating the public half into shim and signing MokManager with
the private half. It uses pesign and nss, but still requires openssl for
key generation. Anyone using sbsign will need to figure this out for
themselves.

11 years agoDon't fail if there's no network devices
Matthew Garrett [Thu, 1 Nov 2012 20:03:24 +0000 (16:03 -0400)]
Don't fail if there's no network devices

findNetboot() would continue blindly even if no PXE-capable devices were
found. Fix that.

11 years agoBump version
Matthew Garrett [Thu, 1 Nov 2012 19:32:12 +0000 (15:32 -0400)]
Bump version

This seems pretty much functionally complete, so let's call it 0.2.

11 years agoMerge branch 'mok' into netboot
Matthew Garrett [Thu, 1 Nov 2012 19:29:23 +0000 (15:29 -0400)]
Merge branch 'mok' into netboot

Conflicts:
shim.c

11 years agoAdd comments
Matthew Garrett [Thu, 1 Nov 2012 19:26:36 +0000 (15:26 -0400)]
Add comments

11 years agoMerge branch 'mok' into netboot
Matthew Garrett [Thu, 1 Nov 2012 14:45:22 +0000 (10:45 -0400)]
Merge branch 'mok' into netboot

Conflicts:
Makefile
shim.c

11 years agoFix AuthenticodeVerify loop
Matthew Garrett [Thu, 1 Nov 2012 14:39:31 +0000 (10:39 -0400)]
Fix AuthenticodeVerify loop

Cert needs to be modified inside the Index loop, not outside it. This is unlikely to
ever trigger since there will typically only be one X509 certificate per
EFI_SIGNATURE_LIST, but fix it anyway.

11 years agoFix signature checking
Matthew Garrett [Thu, 1 Nov 2012 14:31:14 +0000 (10:31 -0400)]
Fix signature checking

We could potentially find a valid signature and then fail to validate it
due to not breaking out of the outer while loop.

11 years agoFix double free
Matthew Garrett [Thu, 1 Nov 2012 14:12:20 +0000 (10:12 -0400)]
Fix double free

load_image() didn't allocate PathName, don't have it free it.

11 years agoFix up some types
Matthew Garrett [Thu, 1 Nov 2012 13:27:38 +0000 (09:27 -0400)]
Fix up some types

Type-checking the UEFI calls picked up a couple of problems. Fix them up.

11 years agoAdd documentation of the Mok variables
Matthew Garrett [Tue, 30 Oct 2012 20:14:02 +0000 (16:14 -0400)]
Add documentation of the Mok variables

Brief overview of the function and format of the various variables used
by Shim and MokManager.

11 years agoMerge branch 'mok' of github.com:mjg59/shim into mok
Matthew Garrett [Tue, 30 Oct 2012 19:59:40 +0000 (15:59 -0400)]
Merge branch 'mok' of github.com:mjg59/shim into mok

11 years agoCheck the vendor blacklist correctly
Gary Ching-Pang Lin [Tue, 30 Oct 2012 02:32:35 +0000 (10:32 +0800)]
Check the vendor blacklist correctly

11 years agoInitialize the size of vendor dbx as 0
Gary Ching-Pang Lin [Tue, 30 Oct 2012 02:32:34 +0000 (10:32 +0800)]
Initialize the size of vendor dbx as 0

The size of vendor dbx must be 0 if there is no vendor dbx provided
or the functions of db check will crash.

11 years agoClean up password setting
Matthew Garrett [Wed, 24 Oct 2012 05:14:50 +0000 (01:14 -0400)]
Clean up password setting

Permit clearing of the password, and avoid a case where choosing not to set
a password would result in an error message on exit. Fix the same problem
with MokSB.

11 years agoImprove signature validation enable/disable
Matthew Garrett [Wed, 24 Oct 2012 05:05:45 +0000 (01:05 -0400)]
Improve signature validation enable/disable

The logic used in checking the signature validation password was a bit
ugly. Improve that so it behaves rather more as expected.

11 years agoBoot unsigned binaries if we're not in secure mode
Matthew Garrett [Wed, 24 Oct 2012 04:10:29 +0000 (00:10 -0400)]
Boot unsigned binaries if we're not in secure mode

read_header would fail if the binary was unsigned, even if we weren't then
going to verify the signature. Move that check to the verify function
instead.

11 years agoMiscellaneous small fixups
Matthew Garrett [Wed, 24 Oct 2012 04:09:08 +0000 (00:09 -0400)]
Miscellaneous small fixups

Fixes for some small bugs noticed during review

11 years agoAdd another missing screen clearing
Matthew Garrett [Wed, 24 Oct 2012 03:46:44 +0000 (23:46 -0400)]
Add another missing screen clearing

Another case where we were drawing text over existing text.

11 years agoMerge branch 'mok' of github.com:mjg59/shim into mok
Matthew Garrett [Tue, 23 Oct 2012 19:45:04 +0000 (15:45 -0400)]
Merge branch 'mok' of github.com:mjg59/shim into mok

11 years agoFix password hash calculation
Matthew Garrett [Tue, 23 Oct 2012 19:42:31 +0000 (15:42 -0400)]
Fix password hash calculation

This was hardcoded, rather than being based on the actual password length,
resulting in incorrect hashes being generated.

11 years agoUpdate image validation enable/disable
Matthew Garrett [Tue, 23 Oct 2012 19:06:59 +0000 (15:06 -0400)]
Update image validation enable/disable

Update this to match the new mokutil behaviour

11 years agoDelete MokList properly
Matthew Garrett [Tue, 23 Oct 2012 15:57:22 +0000 (11:57 -0400)]
Delete MokList properly

A cut and paste error meant that attempts to delete MokList were instead
appending a zero-length addition.

11 years agoClean up checks for MokManager entry
Matthew Garrett [Tue, 23 Oct 2012 15:36:02 +0000 (11:36 -0400)]
Clean up checks for MokManager entry

Add a helper function and tidy up the calls for getting into MokManager

11 years agoFix key database parsing
Matthew Garrett [Tue, 23 Oct 2012 14:04:02 +0000 (10:04 -0400)]
Fix key database parsing

The pointer to the certificate needs to be incremented by the size of the
entire certificate, not just the certificate data.

11 years agoSupport a vendor-specific DBX list.
Peter Jones [Tue, 23 Oct 2012 15:06:28 +0000 (11:06 -0400)]
Support a vendor-specific DBX list.

In some rare corner cases, it's useful to add a blacklist of things that
were allowed by a copy of shim that was never signed by the UEFI signing
service.  In these cases it's okay for them to go into a local dbx,
rather than taking up precious flash.

Signed-off-by: Peter Jones <pjones@redhat.com>
11 years agoClear screen before prompting
Matthew Garrett [Thu, 18 Oct 2012 21:34:14 +0000 (17:34 -0400)]
Clear screen before prompting

We were drawing prompts on top of existing text, which was less than
ideal.

11 years agoDon't print SHA1 sum when calculating file fingerprints
Matthew Garrett [Thu, 18 Oct 2012 21:29:01 +0000 (17:29 -0400)]
Don't print SHA1 sum when calculating file fingerprints

There's no point in printing the SHA1 of a SHA256...

11 years agoClean up timeout counter handling
Matthew Garrett [Thu, 18 Oct 2012 21:24:52 +0000 (17:24 -0400)]
Clean up timeout counter handling

Reduce menu redrawing by only redrawing the invalidated section of the menu
during the timeout countdown.

11 years agoAdd MOK password auth
Matthew Garrett [Thu, 18 Oct 2012 01:40:05 +0000 (21:40 -0400)]
Add MOK password auth

Add support for setting an MOK password. The OS passes down a password hash.
MokManager then presents an option for setting a password. Selecting it
prompts the user for the same password again. If they match, the hash is
enrolled into a boot services variable and MokManager will prompt for the
password whenever it's started.

11 years agoPause on callback failures
Matthew Garrett [Wed, 17 Oct 2012 22:23:35 +0000 (18:23 -0400)]
Pause on callback failures

If a callback returns any kind of failure, wait for a keypress in order to
give the user an opportunity to read any failure messages.

11 years agoSkip signature checking if insecure
Matthew Garrett [Wed, 17 Oct 2012 22:36:07 +0000 (18:36 -0400)]
Skip signature checking if insecure

If we're configured to run untrusted code, print a message and skip the
validation checks.

11 years agoAdd support for disabling signature verification
Matthew Garrett [Wed, 17 Oct 2012 22:22:11 +0000 (18:22 -0400)]
Add support for disabling signature verification

Provide a mechanism for a physically present end user to disable signature
verification. This is handled by the OS passing down a variable that
contains a UINT32 and a SHA256 hash. If this variable is present, MokManager
prompts the user to choose whether to enable or disable signature validation
(depending on the value of the UINT32). They are then asked to type the
passphrase that matches the hash. This then saves a boot services variable
which is checked by shim, and if set will skip verification of signatures.

11 years agoAdd section headers
Matthew Garrett [Sat, 13 Oct 2012 05:07:43 +0000 (01:07 -0400)]
Add section headers

Provide a little more contextual information when people are in shim
menus.

11 years agoAdd draft version of Neil's netboot code
Matthew Garrett [Sat, 13 Oct 2012 00:13:46 +0000 (20:13 -0400)]
Add draft version of Neil's netboot code

11 years agoReallocate the DevPath space for the volume label
Gary Ching-Pang Lin [Fri, 12 Oct 2012 09:55:09 +0000 (17:55 +0800)]
Reallocate the DevPath space for the volume label

The size of the DevPath string array was not sufficient to append
the volume label. This patch extends the size for the label and
re-enables the menu freeing.

11 years agoRemove LoadImage/StartImage support
Matthew Garrett [Fri, 12 Oct 2012 01:05:01 +0000 (21:05 -0400)]
Remove LoadImage/StartImage support

Some systems will show an error dialog if LoadImage() returned
EFI_ACCESS_DENIED, which then requires physical user interaction to skip.
Let's just remove the LoadImage/StartImage code, since the built-in code
is theoretically equivalent.

11 years agoSwitch to using db format for MokList and MokNew
Matthew Garrett [Thu, 11 Oct 2012 19:54:36 +0000 (15:54 -0400)]
Switch to using db format for MokList and MokNew

Using the same format as the UEFI key databases makes it easier for the
kernel to parse and extract keys from MOK, and also permits MOK to contain
multiple key or hash types. Additionally, add support for enrolling hashes.

11 years agoSplit out hashing
Matthew Garrett [Thu, 11 Oct 2012 16:24:36 +0000 (12:24 -0400)]
Split out hashing

We want to be able to generate hashes, so split out the hash generation
function from the verification function

11 years agoAdd missing header define
Matthew Garrett [Thu, 11 Oct 2012 16:13:18 +0000 (12:13 -0400)]
Add missing header define

11 years agoAdd SHA1 support
Matthew Garrett [Thu, 11 Oct 2012 15:30:41 +0000 (11:30 -0400)]
Add SHA1 support

In theory vendors could blacklist binaries with SHA1, so make sure we
calculate and check that hash as well.

11 years agoAdd menu header
Matthew Garrett [Thu, 11 Oct 2012 13:55:14 +0000 (09:55 -0400)]
Add menu header

Add a basic header to the menu to make it clearer what's going on.
Define SHIM_VENDOR in order to override the default.

11 years agoClear screen on exit
Matthew Garrett [Thu, 11 Oct 2012 13:45:14 +0000 (09:45 -0400)]
Clear screen on exit

11 years agoAdd timeout support
Matthew Garrett [Thu, 11 Oct 2012 13:44:27 +0000 (09:44 -0400)]
Add timeout support

We should time out if there aren't any keypresses at the top level menu
within a reasonable timeframe.

11 years agoFix whitespace
Matthew Garrett [Thu, 11 Oct 2012 13:26:16 +0000 (09:26 -0400)]
Fix whitespace

11 years agoRemove unnecessary newline
Gary Ching-Pang Lin [Thu, 11 Oct 2012 13:20:30 +0000 (09:20 -0400)]
Remove unnecessary newline

The newline character shouldn't be present in the menu item string

11 years agoPass the size of MokNew to the key enrollment function
Gary Ching-Pang Lin [Thu, 11 Oct 2012 13:19:32 +0000 (09:19 -0400)]
Pass the size of MokNew to the key enrollment function

The size of MokNew was missing and it caused crash when enrolling new MOKs
with mokutil

11 years agoCallbacks should return INTN, not UINTN
Matthew Garrett [Tue, 9 Oct 2012 13:44:55 +0000 (09:44 -0400)]
Callbacks should return INTN, not UINTN

11 years agoRemove unused variable
Matthew Garrett [Tue, 9 Oct 2012 13:43:37 +0000 (09:43 -0400)]
Remove unused variable

11 years agoDon't lose the last file in top-level directories
Matthew Garrett [Tue, 9 Oct 2012 13:39:36 +0000 (09:39 -0400)]
Don't lose the last file in top-level directories

The filesystem callback was failing to account for the additional menu
item to return to the filesystem list, and so the last file entry in the
root directory would be missing from the list.

11 years agoFix filesystem enrollment
Matthew Garrett [Mon, 8 Oct 2012 19:17:49 +0000 (15:17 -0400)]
Fix filesystem enrollment

We would always enrol a single key, deleting any existing keys in the
database. Fix that up.

11 years agoAdd an auth argument to store_keys()
Matthew Garrett [Tue, 9 Oct 2012 01:45:38 +0000 (21:45 -0400)]
Add an auth argument to store_keys()

If the user is manually installing keys from a filesystem then we don't need
to ask for the key password.

11 years agoAdd __attribute__ ((packed)) to MokListNode definition
Matthew Garrett [Tue, 9 Oct 2012 01:44:43 +0000 (21:44 -0400)]
Add __attribute__ ((packed)) to MokListNode definition

Packing this lets us use MokListNode with the binary MokList representation.