Hi,

 

If the source code below was linkable for all UEFI  targets PEI+DXE+SMM+Shell,

then the most urgent integer/memset()/memcpy() issues would be gone.

NOTE: The program can not run anywhere.

 

> That means I believe GCCINTRINx86-32.lib, CLANGINTRINx86-32.lib and XCODEINTRINx86-32.lib could be simply created
> using these files her:
> https://opensource.apple.com/source/clang/clang-600.0.57/src/projects/compiler-rt/lib/i386/
> or
> https://github.com/llvm/llvm-project/tree/main/compiler-rt/lib/builtins/i386
> Than we have the space saving assembler implementation for all GCC derivates.

 

 

For Microsoft tool chain the intrinsic libraries shipped with the build environment (VS2015 etc.) should be used.

For memset(), memcpy() that should be manually added by compiling from source since the original Microsoft

functions depends on other functions or were not optimal for BIOS usage.

 

Best regards,

Kilian

 

 

typedef unsigned long long uint64_t;                // allow a C99-Standard definition to save ink

typedef          long long  int64_t;                // allow another C99-Standard definition to save ink

typedef struct _BUFFER {

    char buffer[5869];

}BUFFER, *PBUFFER;

 

int main(int argc, char** argv)

{

    //

    // memset()  http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1256.pdf#page=345

    //

    BUFFER buffer = { argc };                       // sometimes invokes automatically a function called memset()

    //

    // memcpy()  http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1256.pdf#page=337

    //

    PBUFFER pBuf = (void*)argv[1];

    *pBuf = buffer;                                 // sometimes invokes automatically a function called memcpy()

    //

    // shift operators  http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1256.pdf#page=96

    //

    uint64_t ullshl = ((uint64_t)argc) <<  3ULL;    // <<

    uint64_t ullshr = ((uint64_t)argc) >>  5ULL;    // >>

    int64_t  llshl  = (( int64_t)argc) <<  7LL;     // <<

    int64_t  llshr  = (( int64_t)argc) >> 11LL;     // >>

    //

    // multiplicative operators  http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1256.pdf#page=94

    //

    uint64_t ullmul = ((uint64_t)argc) * 13ULL;     // *

    uint64_t ulldiv = ((uint64_t)argc) / 17ULL;     // /

    uint64_t ullrem = ((uint64_t)argc) % 19ULL;     // %

    int64_t  llmul  = (( int64_t)argc) * 23LL;      // *

    int64_t  lldiv  = (( int64_t)argc) / 29LL;      // /

    int64_t  llrem  = (( int64_t)argc) % 31LL;      // %

    ////

    //// floating point intrinsics

    ////

    //volatile double dblmul = ((double)argc) * 37ULL;

    //volatile double dbldiv = ((double)argc) / 37ULL;

   

    //__builtin_trap();

   

    //__debugbreak();

 

    return (int)(

          ullshl

        + ullshr

        + ullmul

        + ulldiv

        + ullrem

        +  llshl

        +  llshr

        +  llmul

        +  lldiv

        +  llrem

        + (uint64_t)pBuf);

}

 

Sent from Mail for Windows

 

From: kraxel@redhat.com
Sent: Tuesday, February 1, 2022 10:55 AM
Subject: Re: [edk2-devel] [PATCH 00/24] CryptoPkg/openssl: update openssl submodule to v3.0

 

  Hi,

> That means I believe GCCINTRINx86-32.lib, CLANGINTRINx86-32.lib and XCODEINTRINx86-32.lib could be simply created
> using these files her:
> https://opensource.apple.com/source/clang/clang-600.0.57/src/projects/compiler-rt/lib/i386/
> or
> https://github.com/llvm/llvm-project/tree/main/compiler-rt/lib/builtins/i386
> Than we have the space saving assembler implementation for all GCC derivates.

Yes, using the source and compiling our own probably works better
because we don't run into abi problems that way.

> BTW: Microsoft source is leaked here:
> https://github.com/microsoft/Ironclad/blob/main/ironclad-apps/src/Checked/BootLoader/SingLdrPc/x86/blcrtasm.asm

I doubt using leaked sources is a good idea from a legal/licensing point
of view ...

> There is also a library INT64.LIB in the current Windows 10 SDK, and also in the old WinDDK 7600.
> C:\WinDDK\7600.16385.1\lib\wxp\i386\int64.lib

... but if there is a library with only the 64bit math intrinsics and
nothing else it is maybe easy enough to use the library as-is when
building with the microsoft compiler.

take care,
  Gerd

 

From: Kilian Kegel
Sent: Sunday, January 30, 2022 09:18 PM
To: devel@edk2.groups.io; afish@apple.com
Cc: kraxel@redhat.com; Mike Kinney; Yao, Jiewen; Sean Brogan; Bret Barkelew; Wang, Jian J; Jiang, Guomin; Pawel Polawski; Lu, XiaoyuX; Pedro Falcato
Subject: Re: [edk2-devel] [PATCH 00/24] CryptoPkg/openssl: update openssl submodule to v3.0

 

Hi Andrew, hi Mike, hi Gerd,

 

>Very cool idea but …

The good news:

A common solution for the / % << >>  C operator issue for all platforms is in sight for 64Bit integers.

 

I will discuss my findings along the non-Microsoft udivdi3() function, that is invoked for unsigned long long division

 

Regrettably binary isolation of the required .OBJ/.o during EDKSETUP is not possible for XCODE.

Regrettably these intrinsic functions, are available as C- and Assembler-implementations

and may consume more flash space for opcode, in case of C-versions.

 

Surprisingly both, the assembler and the C implementation found here uses __cdel calling convention.

 

>1) We don’t always use the systems native compiler and sometimes we use

>     a cross compiler so making assumptions about system libs is not always valid.

You’re fully right. Maybe the extraction/isolation during startup can not be used reliable.

 

 

>For bonus point i386 has been obsoleted in Xcode.

The code generator creates a function call to __udivdi3().

The latest Apple-Clang source code for that particular function can be found here:

https://opensource.apple.com/source/clang/clang-600.0.57/src/projects/compiler-rt/lib/i386/udivdi3.S.auto.html

and here for original LLVM that is still maintained:

https://github.com/llvm/llvm-project/blob/main/compiler-rt/lib/builtins/i386/udivdi3.S

No significant difference for both.

 

>2) Are these system libs architecturally defined to be free standing?

>    Can they make assumptions about the runtime?

>    If so seems like they could do random bad things to make them not work in

>    EFI like a system call or make other assumptions like writing to a magic address to generate a trap.

>    Just inspecting them tells us the implementation, not the architecture of the lib.

You are fully right.

Inspecting is necessary in any case.

But extracting from the original library or reimplementing and providing as an

additionally the same for all the other supported architectures…

is much easier than again and again remind the pitfalls of the UEFI guidelines, get build break, wait for rebuild...

for all BIOS developers, world wide for ever.

 

>3) Are the paths to these libs architectural or are they arbitrary implementation that is

>    normal abstracted by how the compiler is released?

>    Could some packaging change break the magic paths to libs?

Truly, I am not absolutely sure anymore, but (except for XCODE) a fully installed build machine 32Bit/64Bit

must contain such libraries by definition.

e.g.:

 

 

When I disassemble the main function of the WSL Ubuntu build of i386.c it looks like

Since parameters and return value passed in __cdecl calling convention, Ubuntu-__udivdi3()

would run in XCODE UEFI too, if the .ELF object format is the same. I have appended the udivdi3.o to test this in XCODE.

 

> This crazy is an example about how assumptions that are not EFI centric can leak into the generic libs produced for the compiler.

 

That means I believe GCCINTRINx86-32.lib, CLANGINTRINx86-32.lib and XCODEINTRINx86-32.lib could be simply created

using these files her:

https://opensource.apple.com/source/clang/clang-600.0.57/src/projects/compiler-rt/lib/i386/

or

https://github.com/llvm/llvm-project/tree/main/compiler-rt/lib/builtins/i386

Than we have the space saving assembler implementation for all GCC derivates.

 

BTW: Microsoft source is leaked here:

https://github.com/microsoft/Ironclad/blob/main/ironclad-apps/src/Checked/BootLoader/SingLdrPc/x86/blcrtasm.asm

 

There is also a library INT64.LIB in the current Windows 10 SDK, and also in the old WinDDK 7600.

C:\WinDDK\7600.16385.1\lib\wxp\i386\int64.lib

C:\Program Files (x86)\Windows Kits\10\Lib\10.0.22000.0\um\x86\int64.lib

 

Best regards,

Kilian

 

From: Andrew Fish via groups.io
Sent: Friday, January 28, 2022 01:55 AM
To: edk2-devel-groups-io; KILIAN_KEGEL@outlook.com
Cc: kraxel@redhat.com; Mike Kinney; Yao, Jiewen; Sean Brogan; Bret Barkelew; Wang, Jian J; Jiang, Guomin; Pawel Polawski; Lu, XiaoyuX; Pedro Falcato
Subject: Re: [edk2-devel] [PATCH 00/24] CryptoPkg/openssl: update openssl submodule to v3.0

 

Very cool idea but …..

 

1) We don’t always use the systems native compiler and sometimes we use a cross compiler so making assumptions about system libs is not always valid.

 

On a Mac with Xcode clang I’ve got full SysV ABI libs (not supper helpful for EFI), but not EFI/MSFT x86_64 ABI. For bonus point i386 has been obsoleted in Xcode.

 

~/work/Compiler/Math>cat hello.c                                

#include <stdio.h>

 

int main(int argc, char **argv)

{

  unsigned long long ulldiv = ((unsigned long long) argc)/3;

 

  // prevent the optimizer to calculate result at build time

  printf("ulldiv %lld\n", ulldiv );

}

~/work/Compiler/Math>clang hello.c

Thus I can link Sys V ABI with the wrong calling convention for EFI.

 

But for the EFI ABIs not so much….

~/work/Compiler/Math>clang -target x86_64-pc-win32-macho hello.c

clang: warning: unable to find a Visual Studio installation; try running Clang from a developer command prompt [-Wmsvc-not-found]

hello.c:1:10: fatal error: 'stdio.h' file not found

#include <stdio.h>

         ^~~~~~~~~

1 error generated.

~/work/Compiler/Math>clang -arch i386 hello.c                   

ld: warning: The i386 architecture is deprecated for macOS (remove from the Xcode build setting: ARCHS)

ld: warning: ignoring file /Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX12.0.sdk/usr/lib/libSystem.tbd, missing required architecture i386 in file /Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX12.0.sdk/usr/lib/libSystem.tbd (3 slices)

Undefined symbols for architecture i386:

  "_printf", referenced from:

      _main in hello-380c0a.o

ld: symbol(s) not found for architecture i386

clang: error: linker command failed with exit code 1 (use -v to see invocation)

 

 

2) Are these system libs architecturally defined to be free standing? Can they make assumptions about the runtime? If so seems like they could do random bad things to make them not work in EFI like a system call or make other assumptions like writing to a magic address to generate a trap. Just inspecting them tells us the implementation, not the architecture of the lib.

 

3) Are the paths to these libs architectural or are they arbitrary implementation that is normal abstracted by how the compiler is released? Could some packaging change break the magic paths to libs?

 

4) I asked the Xcode/clang team a long time ago what to do for free standing match and they pointed me at some open source implementation of these math libs that had been implemented in C. They did not want us using their libs that shipped with macOS. 

 

 

For the GCC/clang tools seems like we are better off just providing the code. 

 

We have magic for compiler specific inline assembly 

https://github.com/tianocore/edk2/blob/master/MdePkg/Library/BaseLib/Ia32/GccInline.c

 

We have magic to abstract some compiler intrinsics today:

https://github.com/tianocore/edk2/blob/master/MdePkg/Library/BaseIoLibIntrinsic/IoLibMsc.c

https://github.com/tianocore/edk2/blob/master/MdePkg/Library/BaseIoLibIntrinsic/IoLibGcc.c

 

While we try to avoid it when at all possible the build system does support doing things differently for different compilers if we have to

https://github.com/tianocore/edk2/blob/master/MdePkg/Library/BaseIoLibIntrinsic/BaseIoLibIntrinsic.inf#L38

[Sources.IA32]

IoLibGcc.c | GCC

IoLibMsc.c | MSFT

IoLib.c

 

 

Thanks,

 

Andrew Fish

 

PS The compiler still works like you think we just don’t have the libs you might need. 

 

~/work/Compiler/Math>cat i386.c             

typedef unsigned long long UINT64;

 

UINT64

main2(int argc, char **argv)

{

  UINT64 ulldiv = ((UINT64) argc)/3;

 

  // prevent the optimizer to calculate result at build time

  return  ulldiv;

}

~/work/Compiler/Math>clang -arch i386 i386.c -S

~/work/Compiler/Math>cat i386.S                

.section __TEXT,__text,regular,pure_instructions

.build_version macos, 12, 0 sdk_version 12, 0

.globl _main2                          ## -- Begin function main2

.p2align 4, 0x90

_main2:                                 ## @main2

.cfi_startproc

## %bb.0:

pushl %ebp

.cfi_def_cfa_offset 8

.cfi_offset %ebp, -8

movl %esp, %ebp

.cfi_def_cfa_register %ebp

subl $24, %esp

movl 12(%ebp), %eax

movl 8(%ebp), %eax

movl 8(%ebp), %ecx

movl %ecx, %edx

sarl $31, %edx

movl %esp, %eax

movl %edx, 4(%eax)

movl %ecx, (%eax)

movl $0, 12(%eax)

movl $3, 8(%eax)

calll ___udivdi3

movl %edx, -4(%ebp)

movl %eax, -8(%ebp)

movl -8(%ebp), %eax

movl -4(%ebp), %edx

addl $24, %esp

popl %ebp

retl

.cfi_endproc

                                        ## -- End function

.subsections_via_symbols

 

PPS Fun story about ABI differences as the macOS i386 ABI requires 16-byte aligned stack accesses and that is more strict than EFI. Luckily it does not break EFI, but it means you can NOT call macOS code from EFI code. To make the emulator work on macOS I had to write assembly gaskets to align the stacks to make calls between the worlds possible. Not my finest hour but it works….

https://github.com/tianocore/edk2/blob/master/EmulatorPkg/Unix/Host/Ia32/Gasket.S

 

This crazy is an example about how assumptions that are not EFI centric can leak into the generic libs produced for the compiler.

 

On Jan 27, 2022, at 2:26 PM, Kilian Kegel <KILIAN_KEGEL@OUTLOOK.COM> wrote:

 

Hi Gerd,

 

>* On my system the gcc intrinsics are only available as shared library,
>   so the "just unpack the lib and use the object files" idea is not
>   going to work.

<D175366159184B98AC9B192EC485505B.png>

 

This little C program makes an unsigned 64Bit division on PC compilers.

Running a 32Bit code generator, it usually invokes an intrinsic function.

 

On my 32Bit Ubuntu standard installation I ran

  1. cc - Xlinker -Map=static.map hello.c -static
  2. cc  -Xlinker -Map=shared.map hello.c

 

The first .OBJ file mentioned in the .MAP file is in both cases:

/usr/lib/gcc/i686-linux-gnu/6/libgcc.a(_udivdi3.o)

<DC03DFFFFFF14308B3A615804A1BF474.png>

 

<377AC53F424C47F794809BA1A5953904.png>

 

Then for each a.out I did:

 

In both cases the intrinsic function is fully linked into the .ELF executable.

<B5A273DA2C3A4898BAEB0A354C667FE5.png>

 

>so the "just unpack the lib and use the object files" idea is not
>going to work.

It seems to me that GNU holds the intrinsic functions in a separate library

that can be used without any change, and is always correct by definition.

 

For Microsoft that is only true when a SDK is installed (INT64.LIB).

Without SDK the intrinsic functions were included in LIBCMT.LIB and

must be isolated manually.

 

Gerd, can you please doublecheck in your GCC build, if that works:

  1. add a 64Bit div to an x86 PEI module like:

<500D8F2283CD4FAE9B0E45647823894A.png>

 

  1. add libgcc.a as a search library, adjust the conf\tools_def.txt like:

DEBUG_GCCxx_IA32_DLINK_FLAGS   = …predefined parameter … /usr/lib/gcc/i686-linux-gnu/6/libgcc.a

to match your build system

  1. build the BIOS 
  2. if the build gets ready, check the .MAP file whether it contains  __udivdi3() or not

 

>* I have my doubts that compiler's builtin libraries are optimized for
>   size, so I'd suspect we would see a noticeable size grow from that.

Please check the size of __udivdi3() and whether the tianocore reimplementation is smaller or not

 

If this works for all build platforms, independently of using the tianocore reimplementation or

using the original compiler intrinsics, this is correct location to place the address of the intrinsic library.

For all optimization modes, operation modes (64Bit/32Bit) the intrinsic library is available for the compiler.

 

GNU lists the intrinsics:

https://gcc.gnu.org/onlinedocs/gccint/Libgcc.html#Libgcc

 

The intrinsic library belongs to the compiler not to the build system.

If the build system changes from EDK2 to VS2022/MSBUILD, the compiler

expects the same intrinsic library.

 

Leave the intrinsic restrictions behind and just provide all required intrinsics the compiler needs

to fulfil the C-Standard!

 

Thanks a lot,

Kilian

https://github.com/tianocore/edk2-staging/tree/CdePkg#cdepkgblog

 

From: kraxel@redhat.com
Sent: Wednesday, January 26, 2022 12:02 PM
To: Pedro Falcato
Cc: edk2-devel-groups-io; Kinney, Michael D; Kilian Kegel; Yao, Jiewen; Sean Brogan; Bret Barkelew; Wang, Jian J; Jiang, Guomin; Pawel Polawski; Lu, XiaoyuX
Subject: Re: [edk2-devel] [PATCH 00/24] CryptoPkg/openssl: update openssl submodule to v3.0

 

  Hi,

> I think adding intrinsic libraries is a mixed bag, for the following
> reasons:
> 
> 1) The intrinsic libraries are completely internal to the compilers.
> Breaking down that toolchain/code barrier is not a good idea if we want the
> project to compile using a good variety of compilers. The API is unstable
> (as it's internal to the compiler + version) and sometimes undocumented;
> while in reality the ABI doesn't really change (at least in the LLVM/GCC
> world, not sure about the other compilers), it's something to consider.

Yes.  But apparently there is no way around them.  We have them for arm.
We have them for openssl.  IntelUndiPkg in edk2-staging has some too
(see IntelUndiPkg/LibC).  And I wouldn't be surprised if there are more
cases ...

Having a policy to outlaw Intrinsics, but then hand out exceptions left
and right doesn't look like a good idea to me.  I think we should revisit
that and accept that there simply is no way around Intrinsics in some
cases.

I think it makes sense to consolidate all the Intrinsics we have, i.e.
move them over to MdePkg, make everybody use that, so we have only a
single version to maintain.

I think it also makes sense to restrict Intrinsics to the cases where we
have no other option, to keep them as small as possible and also make it
as easy as possible to maintain them.

> 2) Linking the compiler's builtin libraries would fix our issues, except
> that it doesn't work in cases where object files are tagged with ABI (such
> as hard FP vs soft FP).

Also:

 * On my system the gcc intrinsics are only available as shared library,
   so the "just unpack the lib and use the object files" idea is not
   going to work.

 * I have my doubts that compiler's builtin libraries are optimized for
   size, so I'd suspect we would see a noticeable size grow from that.

 * I'd very much prefer to continue with the current approach to have
   source code for the Intrinsics we need.  In case we run into trouble
   things tend to be much easier to fix when you have the source code at
   hand.  That's actually part of the open source success story.

take care,
  Gerd

 

Sent from Mail for Windows

 

From: Kilian Kegel
Sent: Tuesday, January 25, 2022 09:06 PM
To: Kinney, Michael D; devel@edk2.groups.io; kraxel@redhat.com; Yao, Jiewen; Sean Brogan; Bret Barkelew
Cc: devel@edk2.groups.io; Wang, Jian J; Jiang, Guomin; Pawel Polawski; Lu, XiaoyuX
Subject: Re: [edk2-devel] [PATCH 00/24] CryptoPkg/openssl: update openssl submodule to v3.0

 

Hi Mike,

 

thank you for your explanation. I understand all the technical aspects.

But let me go into the details of my approach, that skips step 2) to 5)

and adds step 1.5)

 

>I think in practice, the intrinsic APIs we are seeing from use

>of C code from submodules is a very limited set that do not

>change across compiler releases,

I totally agree. E.g INT64.LIB is the same since 2004 (WinXP DDK).

 

But my perspective is different anyway:

  1. an intrinsic library belongs to a particular compiler/linker couple       
  2. an intrinsic library does not belong to a UEFI tianocore or commercial BIOS feature set and should be managed

outside from any EDK2 specific build description (.INF, .DSC)

 

Let’s assume there is a C  build environment fully installed, e.g. Microsoft VS2022, on an EDK2 build machine

and all legal aspects were fulfilled.

 

In that case the advanced developer is able to locate the library, that holds the intrinsic functions

(intrinsic .OBJ modules) we needed to extract. (simply by checking the .MAP of a 32bit executable

that pulls in the particular intrinsics) 

This is step 1)

 

>One of the challenges is that compilers are allowed to add/remove/modify intrinsic APIs

>across compiler releases.  We would need to define a solution that will work if there are

>these types of changes, which would potentially mean a different instance of the intrinsic

>library for each tool chain tag.  

 

Here comes step 1.5):

In case of Microsoft build it is LIBCMT.LIB that can be found when walking through the

LIB environment string.

 

It is easy to extend EDKSETUP.BAT to generate MSFTINTRINx86-32.LIB each time:

  1. locate LIBCMT.LIB
  2. extract the identified .OBJ modules from step 1: “LIB.EXE /extract:full_path_name.obj /out:name.obj LIBCMT.LIB”
  3. bind extracted .OBJ to the MSFTINTRINx86-32.LIB: “LIB.EXE *.obj /out:%CONF_PATH%\MSFTINTRINx86-32.lib”

 

Now MSFTINTRINx86-32.LIB is located in  the conf directory.

 

Adjust the DLINK_FLAGS in tools_def.txt to hold MSFTINTRINx86-32.LIB as a search library:

 

  DEBUG_VS2015_IA32_DLINK_FLAGS   = /NOLOGO /NODEFAULTLIB /IGNORE:4001 /OPT:REF /OPT:ICF=10 /MAP /ALIGN:32 /SECTION:.xdata,D /SECTION:.pdata,D /MACHINE:X86 /LTCG /DLL /ENTRY:$(IMAGE_ENTRY_POINT) /SUBSYSTEM:EFI_BOOT_SERVICE_DRIVER /SAFESEH:NO /BASE:0 /DRIVER /DEBUG %CONF_PATH%\MSFTINTRINx86-32.LIB

 

RELEASE_VS2015_IA32_DLINK_FLAGS   = /NOLOGO /NODEFAULTLIB /IGNORE:4001 /IGNORE:4254 /OPT:REF /OPT:ICF=10 /MAP /ALIGN:32 /SECTION:.xdata,D /SECTION:.pdata,D /MACHINE:X86 /LTCG /DLL /ENTRY:$(IMAGE_ENTRY_POINT) /SUBSYSTEM:EFI_BOOT_SERVICE_DRIVER /SAFESEH:NO /BASE:0 /DRIVER /MERGE:.rdata=.data %CONF_PATH%\MSFTINTRINx86-32.LIB

 

From now on the intrinsics are available for all 32Bit components.

 

With that procedure it is guaranteed by design, that the intrinsics are always available and match a particular compiler/linker release.

 

The script below is a demonstration of the above arguments. It additionally adds memcpy() and memcmp() to MSFTINTRINx86-32.LIB,

that the compiler sometimes needs, depending on optimization style, array size, instruction set, whatsoever …

<A1B2261BD5014BBF968AF0AA25EF9349.png>

 

I have checked some more .OBJ from LIBCMT.LIB and some of them (ftol3.obj to get __ltod3() long to double needed for difftime()) 

seem not to be space optimized for BIOS usage, because the ftol3.obj holds multiple functions

(and so violates the ODR one definition rule).

 

But also such cases could be handled automatically by a script (I wrote a C program < 200 lines)

that disassembles (using Microsoft DUMPBIN.EXE) the .OBJ, splits by simple text processing into

single-function-.ASM-files that could be assembled back to multiple .OBJ of smaller code size.

 

For this approach compiler, library manager, disassembler (DUMPBIN, OBJDUMP) were needed,

that are available on all build machines by definition.

 

Best regards

Kilian

 

 

From: Kinney, Michael D
Sent: Monday, January 24, 2022 06:28 PM
To: Kilian Kegel; devel@edk2.groups.io; kraxel@redhat.com; Yao, Jiewen; Sean Brogan; Bret Barkelew; Kinney, Michael D
Cc: devel@edk2.groups.io; Wang, Jian J; Jiang, Guomin; Pawel Polawski; Lu, XiaoyuX
Subject: RE: [edk2-devel] [PATCH 00/24] CryptoPkg/openssl: update openssl submodule to v3.0

 

Hi Kilian,

 

I am in favor of an intrinsic lib to improve the EDK II development environment.

 

This has already been done for ARM compilers.  The solution should mirror that approach.

 

It would be best if we had source code (either in the edk2 repo or through a submodule) for

the required intrinsic APIs.  If source code is not possible and we have to use a binary, then

that must be accessed through a submodule.  The edk2 repo does not host binaries.  We use

repos such as edk2-non-osi for binaries.

 

We also have to provide a solution that works with supported compilers (VS, GCC, CLANG, XCODE).

 

One of the challenges is that compilers are allowed to add/remove/modify intrinsic APIs

across compiler releases.  We would need to define a solution that will work if there are

these types of changes, which would potentially mean a different instance of the intrinsic

library for each tool chain tag.  I think in practice, the intrinsic APIs we are seeing from use

of C code from submodules is a very limited set that do not change across compiler releases,

so the maintenance of these intrinsic libs would be manageable.

 

If we go down the source code path, we can break it up into the following tasks:

  1. Identify the specific subset of intrinsic APIs from each compiler that is required for the edk2 use cases. 
  2. Obtain the function prototype and full documentation for each intrinsic API to support implementation and unit tests.
  3. Implement the APIs for all compilers.
  4. Implement unit tests for all APIs for all compilers using UnitTestFrameworkPkg unit tests.
  5. Update MdeLibs.dsc.inc with the NULL instances for the intrinsic libs
  6. Remove intrinsic APIs from EDK II modules that currently maintain their own implementations of intrinsic APIs.

 

Best regards,

 

Mike

 

From: Kilian Kegel <kilian_kegel@outlook.com> 
Sent: Monday, January 24, 2022 8:25 AM
To: devel@edk2.groups.io; Kinney, Michael D <michael.d.kinney@intel.com>; kraxel@redhat.com; Yao, Jiewen <jiewen.yao@intel.com>; Sean Brogan <sean.brogan@microsoft.com>; Bret Barkelew <Bret.Barkelew@microsoft.com>
Cc: devel@edk2.groups.io; Wang, Jian J <jian.j.wang@intel.com>; Jiang, Guomin <guomin.jiang@intel.com>; Pawel Polawski <ppolawsk@redhat.com>; Lu, XiaoyuX <xiaoyux.lu@intel.com>
Subject: RE: [edk2-devel] [PATCH 00/24] CryptoPkg/openssl: update openssl submodule to v3.0

 

The 64-bit integer math intrinsics and other intrinsic

problems could be solved easily for ever:

 

  1. Putting all .OBJ files together from LIBCMT.H or INT64.LIB (for ll*.obj and ull*.obj only)

ltod3.obj

ftol2.obj

lldiv.obj

lldvrm.obj

llmul.obj

llrem.obj

llshl.obj

llshr.obj

ulldiv.obj

ulldvrm.obj

ullrem.obj

ullshr.obj

memcmp.obj

memcpycpy.obj

                and adjust for usability in EDK2 (remove / solve further internal dependencies or rewrite “*cpy” and “*cmp” functions)

This is already done in IntrinsicLib.lib for some of the above functions, just complete this task!

  1. Put all the .OBJ into a e.g. edk2\Conf \“MSFTINTRINx86-32<compilerversion>.lib”
  2. Update the MSFT_DEF.txt tool chain definition path

DEBUG_*_IA32_DLINK_FLAGS     = %CONF_PATH%\ MSFTINTRINx86-32<compilerversion>.lib

RELEASE_*_IA32_DLINK_FLAGS   = %CONF_PATH%\ MSFTINTRINx86-32<compilerversion>.lib

  1. Resolve build conflicts with other existing intrinsic libraries from CryptoPkg, RedfishPkg… – remove these libraries

 

From now on all existing 32Bit components have access to their own compiler intrinsics without

touching any .INF file and the problem is instantly gone.

 

Please do the same for 

 

Leave the intrinsic restrictions behind and just provide all required intrinsics the compiler needs

to fulfil the C-Standard!

 

UEFI shall conform the execution environment described in the C Specification

http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1256.pdf#page=23

and shall not try to create a new restricted “UEFI execution environment”

that currently prohibits some “expressions” (shift << >> , divide / % ) on some “data types” (64bit “long long”)

but maybe in the future will prohibit some more “expressions” (logical AND &&, relational-expression < >) on

still speculative “data types” (e.g. a 128bit “extended long”) or just because a new compiler

(version) with some new optimization(ultra slow)/security(specdown/meltre) capabilities introduces

some new intrinsic functions.

Who knows…

 

In contrast to:

“I think we shouldn't add any intrinsics unless we are absolutely forced

to. I do agree however that, for those intrinsics that we cannot at all

avoid reimplementing, we should at least collect them in a common

library.

(In theory, I can also imagine reimplementing all possible intrinsics

*if* the edk2 coding style spec / requirements are updated in parallel,

permitting all new code to universally rely on the intrinsics, rather

than the BaseLib / BaseMemoryLib functions.)”

https://bugzilla.tianocore.org/show_bug.cgi?id=1516#c2

 

This mindset violates edk2 coding style spec too: 

https://edk2-docs.gitbook.io/edk-ii-c-coding-standards-specification/2_guiding_principles

 

Have fun,

Kilian

 

From: Michael D Kinney
Sent: Friday, January 21, 2022 05:39 PM
To: kraxel@redhat.com; Yao, Jiewen; Sean Brogan; Bret Barkelew; Kinney, Michael D
Cc: devel@edk2.groups.io; Wang, Jian J; Jiang, Guomin; Pawel Polawski; Lu, XiaoyuX
Subject: Re: [edk2-devel] [PATCH 00/24] CryptoPkg/openssl: update openssl submodule to v3.0

 

Comments below.

Mike

> -----Original Message-----
> From: kraxel@redhat.com <kraxel@redhat.com>
> Sent: Friday, January 21, 2022 12:31 AM
> To: Yao, Jiewen <jiewen.yao@intel.com>
> Cc: devel@edk2.groups.io; Kinney, Michael D <michael.d.kinney@intel.com>; Wang, Jian J <jian.j.wang@intel.com>; Jiang, Guomin
> <guomin.jiang@intel.com>; Pawel Polawski <ppolawsk@redhat.com>; Lu, XiaoyuX <xiaoyux.lu@intel.com>
> Subject: Re: [edk2-devel] [PATCH 00/24] CryptoPkg/openssl: update openssl submodule to v3.0
> 
> > > No changes in SEC and PEI.
> > [Jiewen] Do you mean the Crypto consumer in PEI has no size difference? Such as
> > https://github.com/tianocore/edk2/tree/master/SecurityPkg/Tcg/Tcg2Pei ,
> > https://github.com/tianocore/edk2/tree/master/SecurityPkg/FvReportPei ,
> > https://github.com/tianocore/edk2/tree/master/SignedCapsulePkg/Universal/RecoveryModuleLoadPei linking
> https://github.com/tianocore/edk2/tree/master/SecurityPkg/Library/FmpAuthenticationLibRsa2048Sha256.
> 
> PEI has this (OvmfIa32X64Pkg build):
> 
>     7062 TpmMmioSevDecryptPei
>     7830 StatusCodeHandlerPei
>     7902 ReportStatusCodeRouterPei
>     8470 FaultTolerantWritePei
>     9734 SmmAccessPei
>    11206 Tcg2ConfigPei
>    11842 PeiVariable
>    14730 Tcg2PlatformPei
>    17274 TcgPei
>    18438 S3Resume2Pei
>    18682 DxeIpl
>    18938 PcdPeim
>    38014 CpuMpPei
>    39554 PlatformPei
>    45050 PeiCore
>    49274 Tcg2Pei
> 
> No size change for Tcg2Pei.
> 
> The other modules are not there.  Seems they are related to firmware
> updates.  We don't have that on ovmf as we can simply update the
> firmware image files on the host machine ...
> 
> Is there some target I could use to test-build those modules?
> 
> > > INFO - OpensslLibCrypto.lib(rsa_lib.obj) : error LNK2001: unresolved external
> > > symbol __allmul
> > > INFO - OpensslLibCrypto.lib(rsa_lib.obj) : error LNK2001: unresolved external
> > > symbol __aulldiv
> > > INFO - OpensslLibCrypto.lib(bio_print.obj) : error LNK2001: unresolved external
> > > symbol __aulldvrm
> > > INFO - OpensslLibCrypto.lib(bio_print.obj) : error LNK2001: unresolved external
> > > symbol __ftol2_sse
> > >
> > > Those symbols look like they reference helper functions to do 64bit math
> > > on 32bit architecture.  Any hints how to fix that?
> > [Jiewen] Please add them to https://github.com/tianocore/edk2/tree/master/CryptoPkg/Library/IntrinsicLib
> 
> Any hints where I could get them?  Given this happens on windows builds
> it's probably somewhere in the microsoft standard C library?  Is that
> available as open source somewhere?

Sean and Bret may be able to help with these.

There is also a BZ on this topic.

https://bugzilla.tianocore.org/show_bug.cgi?id=1516

> 
> > > (3) Some NOOPT builds are failing due to the size growing ...
> > [Jiewen] Size becomes big challenge...
> > Have you tried to use https://github.com/tianocore/edk2/tree/master/CryptoPkg/Driver solution?
> 
> Seems the idea is to have only one openssl copy in the dxe image by
> calling a protocol instead of linking a lib.  Makes sense.
> 
> Is this documented somewhere?  Is there some easy way to use that as
> drop-in replacement?  Or do we have to change all crypto users to call
> the driver instead of linking the lib?
> 
> take care,
>   Gerd


 

 

 

<shared.dis><shared.map><static.dis><static.map>