On Mar 1, 2020, at 3:41 PM, Michael D Kinney <michael.d.kinney@intel.com> wrote:

Andrew,

Thanks.  I did not realize this was a struct with 
a single flexible array member.  I agree that does
not make sense.

The top level struct here is a HOB that must be in
contiguous memory.  The HOB can vary in size based on
MicrocodeCount and StorageType.  We are trying to make
it easy to write the C code that produces and consumes
this HOB.  Using flexible array members helps improve
the readability when structures have this attribute.

For the storage type defined, the StorageContext is a 
single UINT64 value.  Other storage type GUIDs may have
more complex StorageContext structures.  This array
of structures is guaranteed to start on an 8-byte
boundary, so the use case is to cast the start of
the StorageContext to the defined type and access
the array if structures and their fields.

Do you have any suggestions on how to make this
better?


Mike,

The flexible array member can not be in a struct by its self or a member of a union (I looked into that, but clang warns on that too). It has to be a the end of a structure with size. 

If we can't build a compound structure with the preamble data as that data varies too much the best I can think of is this:

Add the known size element to enforce alignment, and make the flexible size member legal C. Add a macro that converts the pointer to raw data to the FLEXIBLE_ARRAY. Basically you need to subtract the fixed size element to get the pointer to the flexible array aligned and indexable. 

typedef struct {
  UINT64  ForceAlign;
  UINT8   Array[];
}  FLEXIBLE_ARRAY;


FLEXIBLE_ARRAY *gFlex = FLEXIBLE_ARRAY_ALIGN (RawData);

Now FLEXIBLE_ARRAY.Array[] is an index into the RawData, and the definition of the FLEXIBLE_ARRAY_ALIGN can explain the trick and why it is required or standard C. 

Then from the C code it looks more like code that requires aligning a buffer? 

Sorry that is the best I've got.  But at the end of the day a pointer is also a flexible array in C so that is the other option. 

Thanks,
 
Andrew Fish


Mike

-----Original Message-----
From: afish@apple.com <afish@apple.com>
Sent: Thursday, February 27, 2020 7:43 AM
To: devel@edk2.groups.io; Kinney, Michael D
<michael.d.kinney@intel.com>
Cc: Gao, Liming <liming.gao@intel.com>; Fu, Siyuan
<siyuan.fu@intel.com>; Ni, Ray <ray.ni@intel.com>;
Chaganty, Rangasai V <rangasai.v.chaganty@intel.com>
Subject: Re: [edk2-devel] [Patch]
IntelSiliconPkg/ShadowMicrocodePei: Fix GCC build
error.

Mike,

Flexible array members must be the last element of a
struct  but they can not be the only element.

This is non standard behavior from the compilers that
are not throwing the error.

Why not just use a pointer?



On Feb 26, 2020, at 10:03 PM, Michael D Kinney
<michael.d.kinney@intel.com> wrote:

Liming,

This does not make sense.  Those compilers
should support C99 flexible array members.

Structured PCDs also require use of flexible
array members.

We need to make sure the compiler flags for
those tool chain are correct to support flexible
array members.

Mike

-----Original Message-----
From: Gao, Liming <liming.gao@intel.com>
Sent: Wednesday, February 26, 2020 9:58 PM
To: devel@edk2.groups.io; Kinney, Michael D
<michael.d.kinney@intel.com>; Fu, Siyuan
<siyuan.fu@intel.com>
Cc: Ni, Ray <ray.ni@intel.com>; Chaganty, Rangasai V
<rangasai.v.chaganty@intel.com>
Subject: RE: [edk2-devel] [Patch]
IntelSiliconPkg/ShadowMicrocodePei: Fix GCC build
error.

Mike:
I find this issue on GCC5 tool chain tag with GCC
5.5
and CLANGPDB tool chain tag with CLANG 9.0.0

Thanks
Liming
-----Original Message-----
From: devel@edk2.groups.io <devel@edk2.groups.io>
On
Behalf Of Michael D Kinney
Sent: Thursday, February 27, 2020 1:54 PM
To: Gao, Liming <liming.gao@intel.com>;
devel@edk2.groups.io; Fu, Siyuan
<siyuan.fu@intel.com>;
Kinney, Michael D
<michael.d.kinney@intel.com>
Cc: Ni, Ray <ray.ni@intel.com>; Chaganty, Rangasai
V
<rangasai.v.chaganty@intel.com>
Subject: Re: [edk2-devel] [Patch]
IntelSiliconPkg/ShadowMicrocodePei: Fix GCC build
error.

Which GCC and CLANG tool chain tags?

Flexible array member is a standard C feature
documented in C99.

Mike

-----Original Message-----
From: Gao, Liming <liming.gao@intel.com>
Sent: Wednesday, February 26, 2020 9:38 PM
To: Kinney, Michael D
<michael.d.kinney@intel.com>;
devel@edk2.groups.io; Fu, Siyuan
<siyuan.fu@intel.com>
Cc: Ni, Ray <ray.ni@intel.com>; Chaganty, Rangasai
V
<rangasai.v.chaganty@intel.com>
Subject: RE: [edk2-devel] [Patch]
IntelSiliconPkg/ShadowMicrocodePei: Fix GCC build
error.

Mike:
I find GCC and CLANG will report the error for
the
empty struct.

d:\allpkg\edk2-


platforms\Silicon\Intel\IntelSiliconPkg\Include\Guid/Mi
crocodeShadowInfoHob.h:61:11: error: flexible
array
member 'MicrocodeAddressInFlash' not allowed in
otherwise empty struct
UINT64  MicrocodeAddressInFlash[];
        ^
1 error generated.

Thanks
Liming
-----Original Message-----
From: Kinney, Michael D
<michael.d.kinney@intel.com>
Sent: Thursday, February 27, 2020 1:25 PM
To: devel@edk2.groups.io; Fu, Siyuan
<siyuan.fu@intel.com>; Kinney, Michael D
<michael.d.kinney@intel.com>
Cc: Ni, Ray <ray.ni@intel.com>; Chaganty,
Rangasai V
<rangasai.v.chaganty@intel.com>; Gao, Liming
<liming.gao@intel.com>
Subject: RE: [edk2-devel] [Patch]
IntelSiliconPkg/ShadowMicrocodePei: Fix GCC build
error.

What compiler does not like the flexible array
member syntax [].

Mike

-----Original Message-----
From: devel@edk2.groups.io
<devel@edk2.groups.io>
On
Behalf Of Siyuan, Fu
Sent: Wednesday, February 26, 2020 5:58 PM
To: devel@edk2.groups.io
Cc: Ni, Ray <ray.ni@intel.com>; Chaganty,
Rangasai
V
<rangasai.v.chaganty@intel.com>; Gao, Liming
<liming.gao@intel.com>
Subject: [edk2-devel] [Patch]
IntelSiliconPkg/ShadowMicrocodePei: Fix GCC
build
error.

This patch fixes compiler error introduced by
commit
b0099a39bd.

BZ:



https://tianocore.acgmultimedia.com/show_bug.cgi?id=244
9
Cc: Ray Ni <ray.ni@intel.com>
Cc: Rangasai V Chaganty
<rangasai.v.chaganty@intel.com>
Cc: Liming Gao <liming.gao@intel.com>
Signed-off-by: Siyuan Fu <siyuan.fu@intel.com>
---

.../Feature/ShadowMicrocode/ShadowMicrocodePei.c
| 2 +-




.../Intel/IntelSiliconPkg/Include/Guid/MicrocodeShadowI
nfoHob.h | 2 +-
2 files changed, 2 insertions(+), 2
deletions(-)

diff --git



a/Silicon/Intel/IntelSiliconPkg/Feature/ShadowMicrocode
/ShadowMicrocodePei.c



b/Silicon/Intel/IntelSiliconPkg/Feature/ShadowMicrocode
/ShadowMicrocodePei.c
index 7e4084247e..8d6574f667 100644
---



a/Silicon/Intel/IntelSiliconPkg/Feature/ShadowMicrocode
/ShadowMicrocodePei.c
+++



b/Silicon/Intel/IntelSiliconPkg/Feature/ShadowMicrocode
/ShadowMicrocodePei.c
@@ -247,7 +247,7 @@ ShadowMicrocodePatchWorker
(
     (VOID *) Patches[Index].Address,
     Patches[Index].Size
     );
-    MicrocodeAddressInMemory[Index] = (UINT64)
Walker;
+    MicrocodeAddressInMemory[Index] = (UINT64)
(UINTN)
Walker;
   Flashcontext-
MicrocodeAddressInFlash[Index]
=
(UINT64) Patches[Index].Address;
   Walker += Patches[Index].Size;
 }
diff --git



a/Silicon/Intel/IntelSiliconPkg/Include/Guid/MicrocodeS
hadowInfoHob.h



b/Silicon/Intel/IntelSiliconPkg/Include/Guid/MicrocodeS
hadowInfoHob.h
index d887b39123..1daae1234a 100644
---



a/Silicon/Intel/IntelSiliconPkg/Include/Guid/MicrocodeS
hadowInfoHob.h
+++



b/Silicon/Intel/IntelSiliconPkg/Include/Guid/MicrocodeS
hadowInfoHob.h
@@ -58,7 +58,7 @@ typedef struct {
 // microcode patch address on flash. The
address
is
placed in same
 // order as the microcode patches in
MicrocodeAddrInMemory.
 //
-  UINT64  MicrocodeAddressInFlash[];
+  UINT64  MicrocodeAddressInFlash[0];
} EFI_MICROCODE_STORAGE_TYPE_FLASH_CONTEXT;

#endif
--
2.19.1.windows.1