From: Marc-André Lureau <marcandre.lur...@redhat.com>

The ZeroMem() call goes beyond the HashInterfaceHob structure, causing
HOB list corruption. The intent was to clear all but the Identifier,
that is starting from HashInterfaceCount.

Quoting Laszlo Ersek:
  Therefore I think the *first* approach to actually fixing this bug would
  be to simply add the "Count" suffix:

  > diff --git 
a/SecurityPkg/Library/HashLibBaseCryptoRouter/HashLibBaseCryptoRouterPei.c 
b/SecurityPkg/Library/HashLibBaseCryptoRouter/HashLibBaseCryptoRouterPei.c
  > index dbee0f2531bc..a869fffae3fe 100644
  > --- 
a/SecurityPkg/Library/HashLibBaseCryptoRouter/HashLibBaseCryptoRouterPei.c
  > +++ 
b/SecurityPkg/Library/HashLibBaseCryptoRouter/HashLibBaseCryptoRouterPei.c
  > @@ -424,7 +424,10 @@ HashLibBaseCryptoRouterPeiConstructor (
  >      // This is the second execution of this module, clear the hash 
interface
  >      // information registered at its first execution.
  >      //
  > -    ZeroMem (&HashInterfaceHob->HashInterface, sizeof (*HashInterfaceHob) 
- sizeof (EFI_GUID));
  > +    ZeroMem (
  > +      &HashInterfaceHob->HashInterfaceCount,
  > +      sizeof (*HashInterfaceHob) - sizeof (EFI_GUID)
  > +      );
  >    }
  >
  >    //

  On the other hand, I don't think such a fix would be great. First of
  all, the HASH_INTERFACE_HOB struct definition is not wrapped in

  #pragma pack (1)
  ...
  #pragma pack ()

  in other words, HASH_INTERFACE_HOB is not packed, hence there could be
  an unspecified amount of padding between "Identifier" and
  "HashInterfaceCount". That would lead to the same kind of buffer
  overflow, because "Length" includes the padding, but the "Buffer"
  argument doesn't.

  Second, even if there were no padding (and thus the byte count would be
  a perfect match for the full HOB structure), in the "Buffer" argument
  we take the address of the "HashInterfaceCount" field, and with the
  "Length" field, we still overflow that *individual* field. Although we
  wouldn't overflow the containing HOB structure, this is still (arguably)
  undefined behavior -- I seem to remember that some compilers actually
  blow up on this when you use the standard ISO C memcpy() or memset()
  functions in such contexts.

This patch takes Laszlo suggestion to fix the issue.

Cc: Jiewen Yao <jiewen....@intel.com>
Cc: Chao Zhang <chao.b.zh...@intel.com>
Cc: Star Zeng <star.z...@intel.com>
Cc: Laszlo Ersek <ler...@redhat.com>
Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Marc-André Lureau <marcandre.lur...@redhat.com>
---
 .../HashLibBaseCryptoRouter/HashLibBaseCryptoRouterPei.c   | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git 
a/SecurityPkg/Library/HashLibBaseCryptoRouter/HashLibBaseCryptoRouterPei.c 
b/SecurityPkg/Library/HashLibBaseCryptoRouter/HashLibBaseCryptoRouterPei.c
index dbee0f2531bc..84c1aaae70eb 100644
--- a/SecurityPkg/Library/HashLibBaseCryptoRouter/HashLibBaseCryptoRouterPei.c
+++ b/SecurityPkg/Library/HashLibBaseCryptoRouter/HashLibBaseCryptoRouterPei.c
@@ -397,6 +397,7 @@ HashLibBaseCryptoRouterPeiConstructor (
 {
   EFI_STATUS            Status;
   HASH_INTERFACE_HOB    *HashInterfaceHob;
+  UINTN                 ClearStartOffset;
 
   HashInterfaceHob = InternalGetHashInterfaceHob (&gZeroGuid);
   if (HashInterfaceHob == NULL) {
@@ -424,7 +425,11 @@ HashLibBaseCryptoRouterPeiConstructor (
     // This is the second execution of this module, clear the hash interface
     // information registered at its first execution.
     //
-    ZeroMem (&HashInterfaceHob->HashInterface, sizeof (*HashInterfaceHob) - 
sizeof (EFI_GUID));
+    ClearStartOffset = OFFSET_OF (HASH_INTERFACE_HOB, HashInterfaceCount);
+    ZeroMem (
+      (UINT8 *)HashInterfaceHob + ClearStartOffset,
+      sizeof (*HashInterfaceHob) - ClearStartOffset
+      );
   }
 
   //
-- 
2.16.2.346.g9779355e34

_______________________________________________
edk2-devel mailing list
edk2-devel@lists.01.org
https://lists.01.org/mailman/listinfo/edk2-devel

Reply via email to