render / rpms / edk2

Forked from rpms/edk2 3 months ago
Clone

Blame SOURCES/edk2-MdeModulePkg-Variable-Fix-Timestamp-zeroing-issue-on.patch

cc9195
From 3b8ff18ad4ac1af740a979ad27fb83dbbdca70ef Mon Sep 17 00:00:00 2001
cc9195
From: Laszlo Ersek <lersek@redhat.com>
cc9195
Date: Wed, 24 Oct 2018 21:03:42 +0200
cc9195
Subject: [PATCH 1/4] MdeModulePkg Variable: Fix Timestamp zeroing issue on
cc9195
 APPEND_WRITE
cc9195
cc9195
Message-id: <20181024190345.15288-2-lersek@redhat.com>
cc9195
Patchwork-id: 82887
cc9195
O-Subject:  [RHEL8 edk2 PATCH 1/4] MdeModulePkg Variable: Fix Timestamp zeroing
cc9195
	issue on APPEND_WRITE
cc9195
Bugzilla: 1641436
cc9195
Acked-by: Vitaly Kuznetsov <vkuznets@redhat.com>
cc9195
Acked-by: Thomas Huth <thuth@redhat.com>
cc9195
cc9195
From: Star Zeng <star.zeng@intel.com>
cc9195
cc9195
--v-- RHEL8 note start --v--
cc9195
cc9195
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1641436
cc9195
cc9195
This patch fixes CVE-2018-3613. Unfortunately, the upstream subject line
cc9195
does not include the CVE number. I've decided to stick with the upstream
cc9195
subject verbatim in the backport, so we can more easily drop this patch at
cc9195
the next rebase. On the upstream list, I did complain loudly, so there's
cc9195
hope the next CVE fix will advertise the CVE number in the subject.
cc9195
cc9195
In practice, the vulnerability is difficult to exploit. Please refer to
cc9195
the following messages in the upstream discussion:
cc9195
cc9195
  https://lists.01.org/pipermail/edk2-devel/2018-October/031103.html
cc9195
  https://lists.01.org/pipermail/edk2-devel/2018-October/031140.html
cc9195
cc9195
--^-- RHEL8 note end --^--
cc9195
cc9195
REF: https://bugzilla.tianocore.org/show_bug.cgi?id=415
cc9195
cc9195
When SetVariable() to a time based auth variable with APPEND_WRITE
cc9195
attribute, and if the EFI_VARIABLE_AUTHENTICATION_2.TimeStamp in
cc9195
the input Data is earlier than current value, it will cause timestamp
cc9195
zeroing.
cc9195
cc9195
This issue may bring time based auth variable downgrade problem.
cc9195
For example:
cc9195
A vendor released three certs at 2014, 2015, and 2016, and system
cc9195
integrated the 2016 cert. User can SetVariable() with 2015 cert and
cc9195
APPEND_WRITE attribute to cause timestamp zeroing first, then
cc9195
SetVariable() with 2014 cert to downgrade the cert.
cc9195
cc9195
This patch fixes this issue.
cc9195
cc9195
Cc: Jiewen Yao <jiewen.yao@intel.com>
cc9195
Cc: Chao Zhang <chao.b.zhang@intel.com>
cc9195
Cc: Jian J Wang <jian.j.wang@intel.com>
cc9195
Contributed-under: TianoCore Contribution Agreement 1.1
cc9195
Signed-off-by: Star Zeng <star.zeng@intel.com>
cc9195
Reviewed-by: Jiewen Yao <jiewen.yao@intel.com>
cc9195
(cherry picked from commit b7dc8888f31402f410c53242839271ba3b94b619)
cc9195
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
cc9195
---
cc9195
 MdeModulePkg/Universal/Variable/RuntimeDxe/Variable.c | 2 ++
cc9195
 1 file changed, 2 insertions(+)
cc9195
cc9195
diff --git a/MdeModulePkg/Universal/Variable/RuntimeDxe/Variable.c b/MdeModulePkg/Universal/Variable/RuntimeDxe/Variable.c
cc9195
index 6caf603..60439b5 100644
cc9195
--- a/MdeModulePkg/Universal/Variable/RuntimeDxe/Variable.c
cc9195
+++ b/MdeModulePkg/Universal/Variable/RuntimeDxe/Variable.c
cc9195
@@ -2460,6 +2460,8 @@ UpdateVariable (
cc9195
         if (Variable->CurrPtr != NULL) {
cc9195
           if (VariableCompareTimeStampInternal (&(((AUTHENTICATED_VARIABLE_HEADER *) CacheVariable->CurrPtr)->TimeStamp), TimeStamp)) {
cc9195
             CopyMem (&AuthVariable->TimeStamp, TimeStamp, sizeof (EFI_TIME));
cc9195
+          } else {
cc9195
+            CopyMem (&AuthVariable->TimeStamp, &(((AUTHENTICATED_VARIABLE_HEADER *) CacheVariable->CurrPtr)->TimeStamp), sizeof (EFI_TIME));
cc9195
           }
cc9195
         }
cc9195
       }
cc9195
-- 
cc9195
1.8.3.1
cc9195