[PATCH] SecurityPkg: Initailize variable Status before it is consumed.

[PATCH] SecurityPkg: Initailize variable Status before it is consumed.

[Zhiguang Liu]

REF: https://bugzilla.tianocore.org/show_bug.cgi?id=2945

Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jian J Wang <jian.j.wang@intel.com>
Cc: Qi Zhang <qi1.zhang@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Signed-off-by: Zhiguang Liu <zhiguang.liu@intel.com>
---
 SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c b/SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c
index 0e770f4485..5e883f0cc5 100644
--- a/SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c
+++ b/SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c
@@ -449,6 +449,7 @@ HashLogExtendEvent (
   EFI_STATUS                        Status;
   TPML_DIGEST_VALUES                DigestList;
 
+  Status = EFI_SUCCESS;
   if (GetFirstGuidHob (&gTpmErrorHobGuid) != NULL) {
     return EFI_DEVICE_ERROR;
   }
-- 
2.25.1.windows.1

Re: [edk2-devel] [PATCH] SecurityPkg: Initailize variable Status before it is consumed.

[Laszlo Ersek]

On 08/31/20 10:15, Zhiguang Liu wrote:
> REF: https://bugzilla.tianocore.org/show_bug.cgi?id=2945
> 
> Cc: Jiewen Yao <jiewen.yao@intel.com>
> Cc: Jian J Wang <jian.j.wang@intel.com>
> Cc: Qi Zhang <qi1.zhang@intel.com>
> Cc: Rahul Kumar <rahul1.kumar@intel.com>
> Signed-off-by: Zhiguang Liu <zhiguang.liu@intel.com>
> ---
>  SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c b/SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c
> index 0e770f4485..5e883f0cc5 100644
> --- a/SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c
> +++ b/SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c
> @@ -449,6 +449,7 @@ HashLogExtendEvent (
>    EFI_STATUS                        Status;
>    TPML_DIGEST_VALUES                DigestList;
>  
> +  Status = EFI_SUCCESS;
>    if (GetFirstGuidHob (&gTpmErrorHobGuid) != NULL) {
>      return EFI_DEVICE_ERROR;
>    }
> 

I agree that there is a path in the code where Status is read without
having been set.

It seems that using EFI_SUCCESS as default value makes sense (assuming
the LogHashEvent() call is the important one). I'll let SecurityPkg
maintainers decide about this.

I think it would be nicer to set Status to EFI_SUCCESS either on the
(currently missing) "else" branch, or else just before the
EDKII_TCG_PRE_HASH check. In particular, setting Status so early that we
may still exit with EFI_DEVICE_ERROR is wasteful. So at least I'd move
the assignment past the "return EFI_DEVICE_ERROR" statement.

But... it's late. I agree that this patch qualifies for the stable tag.
So I'm not asking for a repost. I just wish more thought had been given
to the placement of the assignment.

Laszlo

Re: [edk2-devel] [PATCH] SecurityPkg: Initailize variable Status before it is consumed.

[Yao, Jiewen]

Agree with Laszlo.

I prefer to move "Status = EFI_SUCCESS;" before the EDKII_TCG_PRE_HASH check.

With that moving, reviewed-by: Jiewen Yao <jiewen.yao@intel.com>

Thank you
Yao Jiewen


> -----Original Message-----
> From: Laszlo Ersek <lersek@redhat.com>
> Sent: Monday, August 31, 2020 11:46 PM
> To: devel@edk2.groups.io; Liu, Zhiguang <zhiguang.liu@intel.com>
> Cc: Yao, Jiewen <jiewen.yao@intel.com>; Wang, Jian J <jian.j.wang@intel.com>;
> Zhang, Qi1 <qi1.zhang@intel.com>; Kumar, Rahul1 <rahul1.kumar@intel.com>
> Subject: Re: [edk2-devel] [PATCH] SecurityPkg: Initailize variable Status before it
> is consumed.
> 
> On 08/31/20 10:15, Zhiguang Liu wrote:
> > REF: https://bugzilla.tianocore.org/show_bug.cgi?id=2945
> >
> > Cc: Jiewen Yao <jiewen.yao@intel.com>
> > Cc: Jian J Wang <jian.j.wang@intel.com>
> > Cc: Qi Zhang <qi1.zhang@intel.com>
> > Cc: Rahul Kumar <rahul1.kumar@intel.com>
> > Signed-off-by: Zhiguang Liu <zhiguang.liu@intel.com>
> > ---
> >  SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c | 1 +
> >  1 file changed, 1 insertion(+)
> >
> > diff --git a/SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c
> b/SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c
> > index 0e770f4485..5e883f0cc5 100644
> > --- a/SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c
> > +++ b/SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c
> > @@ -449,6 +449,7 @@ HashLogExtendEvent (
> >    EFI_STATUS                        Status;
> >    TPML_DIGEST_VALUES                DigestList;
> >
> > +  Status = EFI_SUCCESS;
> >    if (GetFirstGuidHob (&gTpmErrorHobGuid) != NULL) {
> >      return EFI_DEVICE_ERROR;
> >    }
> >
> 
> I agree that there is a path in the code where Status is read without
> having been set.
> 
> It seems that using EFI_SUCCESS as default value makes sense (assuming
> the LogHashEvent() call is the important one). I'll let SecurityPkg
> maintainers decide about this.
> 
> I think it would be nicer to set Status to EFI_SUCCESS either on the
> (currently missing) "else" branch, or else just before the
> EDKII_TCG_PRE_HASH check. In particular, setting Status so early that we
> may still exit with EFI_DEVICE_ERROR is wasteful. So at least I'd move
> the assignment past the "return EFI_DEVICE_ERROR" statement.
> 
> But... it's late. I agree that this patch qualifies for the stable tag.
> So I'm not asking for a repost. I just wish more thought had been given
> to the placement of the assignment.
> 
> Laszlo

[PATCH] SecurityPkg: Initailize variable Status before it is consumed.

[Zhiguang Liu]

REF: https://bugzilla.tianocore.org/show_bug.cgi?id=2945

V2: Move "Status = EFI_SUCCESS;" before the EDKII_TCG_PRE_HASH check.

Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jian J Wang <jian.j.wang@intel.com>
Cc: Qi Zhang <qi1.zhang@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jiewen Yao <jiewen.yao@intel.com>
Signed-off-by: Zhiguang Liu <zhiguang.liu@intel.com>
---
 SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c b/SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c
index 0e770f4485..93a8803ff6 100644
--- a/SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c
+++ b/SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c
@@ -456,6 +456,7 @@ HashLogExtendEvent (
   if ((Flags & EDKII_TCG_PRE_HASH) != 0 || (Flags & EDKII_TCG_PRE_HASH_LOG_ONLY) != 0) {
     ZeroMem (&DigestList, sizeof(DigestList));
     CopyMem (&DigestList, HashData, sizeof(DigestList));
+    Status = EFI_SUCCESS;
     if ((Flags & EDKII_TCG_PRE_HASH) !=0 ) {
       Status = Tpm2PcrExtend (
                NewEventHdr->PCRIndex,
-- 
2.25.1.windows.1

Re: [PATCH] SecurityPkg: Initailize variable Status before it is consumed.

[Laszlo Ersek]

On 09/01/20 02:55, Zhiguang Liu wrote:
> REF: https://bugzilla.tianocore.org/show_bug.cgi?id=2945
> 
> V2: Move "Status = EFI_SUCCESS;" before the EDKII_TCG_PRE_HASH check.
> 
> Cc: Jiewen Yao <jiewen.yao@intel.com>
> Cc: Jian J Wang <jian.j.wang@intel.com>
> Cc: Qi Zhang <qi1.zhang@intel.com>
> Cc: Rahul Kumar <rahul1.kumar@intel.com>
> Cc: Laszlo Ersek <lersek@redhat.com>
> Reviewed-by: Jiewen Yao <jiewen.yao@intel.com>
> Signed-off-by: Zhiguang Liu <zhiguang.liu@intel.com>
> ---
>  SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c b/SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c
> index 0e770f4485..93a8803ff6 100644
> --- a/SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c
> +++ b/SecurityPkg/Tcg/Tcg2Pei/Tcg2Pei.c
> @@ -456,6 +456,7 @@ HashLogExtendEvent (
>    if ((Flags & EDKII_TCG_PRE_HASH) != 0 || (Flags & EDKII_TCG_PRE_HASH_LOG_ONLY) != 0) {
>      ZeroMem (&DigestList, sizeof(DigestList));
>      CopyMem (&DigestList, HashData, sizeof(DigestList));
> +    Status = EFI_SUCCESS;
>      if ((Flags & EDKII_TCG_PRE_HASH) !=0 ) {
>        Status = Tpm2PcrExtend (
>                 NewEventHdr->PCRIndex,
> 

Reviewed-by: Laszlo Ersek <lersek@redhat.com>

I'll let Jiewen or Jian merge this.

Please change the status of TianoCore#2945 to IN_PROGRESS, and link both
versions of this patch into the ticket as well (in a new comment).

Thanks
Laszlo