[edk2-devel] [PATCH v2 0/9] Migrate Pointer from flash to permanent memory (CVE-2019-11098)
Laszlo Ersek
lersek at redhat.com
Fri Jul 3 14:06:41 UTC 2020
Hi,
I'm adding Michael Kubacki's new email address to the "To:" list, as
Michael has authored a significant portion of this code, plus he seems
to have contributed a large part of the design, in
<https://bugzilla.tianocore.org/show_bug.cgi?id=1614>.
Also CC'ing Ard.
Comments below.
On 07/02/20 07:15, Guomin Jiang wrote:
> The TOCTOU vulnerability allow that the physical present person to replace the code with the normal BootGuard check and PCR0 value.
> The issue occur when BootGuard measure IBB and access flash code after NEM disable.
> the reason why we access the flash code is that we have some pointer to flash.
> To avoid this vulnerability, we need to convert those pointers, the patch series do this work and make sure that no code will access flash address.
I've now read through the comments in
<https://bugzilla.tianocore.org/show_bug.cgi?id=1614>, and I've also
checked the slides at:
[1]
https://conference.hitb.org/hitbsecconf2019ams/sessions/now-you-see-it-toctou-attacks-against-secure-boot-and-bootguard/
https://conference.hitb.org/hitbsecconf2019ams/materials/D1T1%20-%20Toctou%20Attacks%20Against%20Secure%20Boot%20-%20Trammell%20Hudson%20&%20Peter%20Bosch.pdf
My understanding is that this vulnerability (and fix) do not apply to
virtualization, or even most other emulation platforms (such as
EmulatorPkg).
Therefore I'm requesting that the approach seen in this patch series be
reversed, as follows.
* The patch v3 #1 should introduce the new PCD called
"PcdMigrateTemporaryRamFirmwareVolumes". (Currently: patch v2 #7.)
The comments in the DEC file (and in the UNI file) should *very* clearly
explain what the PCD controls. It should provide a *concise* description
of the entire feature. The default value of the PCD should be TRUE.
* The next patches in the series (v3 patches #2, #3, #4) should set the
PCD to FALSE in at least the following platform DSC files:
- ArmVirtPkg [v3 #2]
- EmulatorPkg [v3 #3],
- OvmfPkg [v3 #4].
The commit messages on these patches should explain that the
vulnerability simply doesn't exist on those platforms, as BootGuard is
undefined on them in the first place. "CAR" (Cache-As-RAM) is also
undefined on them.
In particular, slide #8 in the presentation, titled "Chain of Trust
(simplified)", presents two "cascades" (sub-chains of trust). My
understanding is that the attack targets the left hand side cascade
("Fused OEM key", "Signed by Intel"). And that cascade doesn't seem to
exist at all on virtualized platforms.
Furthermore, the abstract at [1] writes,
"These protections are supposed to be secure against physical attacks on
the SPI flash"
foreshadowing that the attack is indeed a physical machine attack.
Doesn't apply to virtualization.
Therefore both the attack and the mitigation appear moot, on virtual
platforms. This should be stated in the commit messages of v3 patches
#2, #3, #4.
* Patch v3 #5 should be the current (v2) patch #1, namely:
MdeModulePkg/PeiCore: Enable T-RAM evacuation in PeiCore
This patch should *immediately* depend on
"PcdMigrateTemporaryRamFirmwareVolumes". That is, please let's not
introduce the feature first as unconditionally active, and then gate it
on "PcdMigrateTemporaryRamFirmwareVolumes" separately. The feature
should honor "PcdMigrateTemporaryRamFirmwareVolumes" right off the bat.
Technically, this more or less means squashing part of patch v2 #7 into
v2 #1.
The commit message should point out that EvacuateTempRam() is never
called if "PcdMigrateTemporaryRamFirmwareVolumes" is FALSE.
* I've made the rest of my comments under the individual patches.
Thanks,
Laszlo
>
> Cc: Jian J Wang <jian.j.wang at intel.com>
> Cc: Hao A Wu <hao.a.wu at intel.com>
> Cc: Dandan Bi <dandan.bi at intel.com>
> Cc: Liming Gao <liming.gao at intel.com>
> Cc: Debkumar De <debkumar.de at intel.com>
> Cc: Harry Han <harry.han at intel.com>
> Cc: Catharine West <catharine.west at intel.com>
> Cc: Eric Dong <eric.dong at intel.com>
> Cc: Ray Ni <ray.ni at intel.com>
> Cc: Laszlo Ersek <lersek at redhat.com>
> Cc: Rahul Kumar <rahul1.kumar at intel.com>
> Cc: Jiewen Yao <jiewen.yao at intel.com>
> Cc: Chao Zhang <chao.b.zhang at intel.com>
> Cc: Qi Zhang <qi1.zhang at intel.com>
>
> Guomin Jiang (5):
> MdeModulePkg/Core: Create Migrated FV Info Hob for calculating hash
> (CVE-2019-11098)
> SecurityPkg/Tcg2Pei: Use Migrated FV Info Hob for calculating hash
> (CVE-2019-11098)
> MdeModulePkg/Core: Add switch to enable or disable TOCTOU feature
> (CVE-2019-11098)
> UefiCpuPkg/SecMigrationPei: Add switch to control if produce PPI
> (CVE-2019-11098)
> UefiCpuPkg/CpuMpPei: Enable paging and set NP flag to avoid TOCTOU
> (CVE-2019-11098)
>
> Jian J Wang (1):
> MdeModulePkg/DxeIplPeim: Register for shadow on S3 shadowed boot
> (CVE-2019-11098)
>
> Michael Kubacki (3):
> MdeModulePkg/PeiCore: Enable T-RAM evacuation in PeiCore
> (CVE-2019-11098)
> UefiCpuPkg/CpuMpPei: Add GDT and IDT migration support
> (CVE-2019-11098)
> UefiCpuPkg/SecMigrationPei: Add initial PEIM (CVE-2019-11098)
>
> MdeModulePkg/Core/DxeIplPeim/DxeIpl.inf | 3 +
> MdeModulePkg/Core/DxeIplPeim/DxeLoad.c | 2 +-
> MdeModulePkg/Core/Pei/Dispatcher/Dispatcher.c | 417 ++++++++++++++++++
> MdeModulePkg/Core/Pei/Image/Image.c | 115 +++++
> MdeModulePkg/Core/Pei/Memory/MemoryServices.c | 82 ++++
> MdeModulePkg/Core/Pei/PeiMain.h | 169 +++++++
> MdeModulePkg/Core/Pei/PeiMain.inf | 3 +
> MdeModulePkg/Core/Pei/PeiMain/PeiMain.c | 17 +
> MdeModulePkg/Core/Pei/Ppi/Ppi.c | 287 ++++++++++++
> MdeModulePkg/Include/Guid/MigratedFvInfo.h | 22 +
> MdeModulePkg/MdeModulePkg.dec | 8 +
> SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c | 31 +-
> SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.inf | 1 +
> UefiCpuPkg/CpuMpPei/CpuMpPei.c | 40 +-
> UefiCpuPkg/CpuMpPei/CpuMpPei.h | 13 +
> UefiCpuPkg/CpuMpPei/CpuMpPei.inf | 3 +
> UefiCpuPkg/CpuMpPei/CpuPaging.c | 31 +-
> UefiCpuPkg/Include/Ppi/RepublishSecPpi.h | 54 +++
> .../Ia32/ArchExceptionHandler.c | 4 +-
> .../SecPeiCpuException.c | 2 +-
> UefiCpuPkg/SecCore/SecCore.inf | 2 +
> UefiCpuPkg/SecCore/SecMain.c | 26 +-
> UefiCpuPkg/SecCore/SecMain.h | 1 +
> UefiCpuPkg/SecMigrationPei/SecMigrationPei.c | 374 ++++++++++++++++
> UefiCpuPkg/SecMigrationPei/SecMigrationPei.h | 170 +++++++
> .../SecMigrationPei/SecMigrationPei.inf | 68 +++
> .../SecMigrationPei/SecMigrationPei.uni | 13 +
> UefiCpuPkg/UefiCpuPkg.dec | 4 +
> UefiCpuPkg/UefiCpuPkg.dsc | 1 +
> 29 files changed, 1947 insertions(+), 16 deletions(-)
> create mode 100644 MdeModulePkg/Include/Guid/MigratedFvInfo.h
> create mode 100644 UefiCpuPkg/Include/Ppi/RepublishSecPpi.h
> create mode 100644 UefiCpuPkg/SecMigrationPei/SecMigrationPei.c
> create mode 100644 UefiCpuPkg/SecMigrationPei/SecMigrationPei.h
> create mode 100644 UefiCpuPkg/SecMigrationPei/SecMigrationPei.inf
> create mode 100644 UefiCpuPkg/SecMigrationPei/SecMigrationPei.uni
>
-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#62037): https://edk2.groups.io/g/devel/message/62037
Mute This Topic: https://groups.io/mt/75252659/1813853
Group Owner: devel+owner at edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [edk2-devel-archive at redhat.com]
-=-=-=-=-=-=-=-=-=-=-=-
More information about the edk2-devel-archive
mailing list