[edk2-devel] [RESEND PATCH v2] BaseTools: Add support for RISCV GOT/PLT relocations

Pete Batard pete at akeo.ie
Wed Jun 16 12:35:27 UTC 2021


Sunil, Daniel, thanks for the patch.

I confirm that this addresses the 0x13 and 0x14 relocation issues that I 
was seeing.

However, this patch appears to introduces new R_RISCV_PCREL_LO12_S 
relocation errors that I was not seeing previously, so I still can't 
manage to get a successful compilation.

Especially the stream of 0x13 and 0x14 relocation errors I was getting 
at 
https://github.com/pbatard/ntfs-3g/runs/2278190652?check_suite_focus=true has 
now switched to (tested on up to date Ubuntu with latest EDK2):

-------------------------------------------------------------------------
"GenFw" -e UEFI_DRIVER -o 
/usr/src/ntfs-3g/Build/RELEASE_GCC5/RISCV64/uefi-driver/uefi_driver/OUTPUT/ntfs.efi 
/usr/src/ntfs-3g/Build/RELEASE_GCC5/RISCV64/uefi-driver/uefi_driver/DEBUG/ntfs.dll
GenFw: ERROR 3000: Invalid
   WriteSections64(): 
/usr/src/ntfs-3g/Build/RELEASE_GCC5/RISCV64/uefi-driver/uefi_driver/DEBUG/ntfs.dll 
unsupported ELF EM_RISCV64 relocation 0x19.
GenFw: ERROR 3000: Invalid
   WriteSections64(): 
/usr/src/ntfs-3g/Build/RELEASE_GCC5/RISCV64/uefi-driver/uefi_driver/DEBUG/ntfs.dll 
unsupported ELF EM_RISCV64 relocation 0x19.
GenFw: ERROR 3000: Invalid
   WriteSections64(): 
/usr/src/ntfs-3g/Build/RELEASE_GCC5/RISCV64/uefi-driver/uefi_driver/DEBUG/ntfs.dll 
unsupported ELF EM_RISCV64 relocation 0x19.
GenFw: ERROR 3000: Invalid
   WriteRelocations64(): 
/usr/src/ntfs-3g/Build/RELEASE_GCC5/RISCV64/uefi-driver/uefi_driver/DEBUG/ntfs.dll 
unsupported ELF EM_RISCV64 relocation 0x19.
GenFw: ERROR 3000: Invalid
   WriteRelocations64(): 
/usr/src/ntfs-3g/Build/RELEASE_GCC5/RISCV64/uefi-driver/uefi_driver/DEBUG/ntfs.dll 
unsupported ELF EM_RISCV64 relocation 0x19.
GenFw: ERROR 3000: Invalid
   WriteRelocations64(): 
/usr/src/ntfs-3g/Build/RELEASE_GCC5/RISCV64/uefi-driver/uefi_driver/DEBUG/ntfs.dll 
unsupported ELF EM_RISCV64 relocation 0x19.
make: *** [GNUmakefile:553: 
/usr/src/ntfs-3g/Build/RELEASE_GCC5/RISCV64/uefi-driver/uefi_driver/OUTPUT/ntfs.efi] 
Error 2
-------------------------------------------------------------------------

So, in effect, some of the earlier relocation errors appear to have 
morphed into 0x19/R_RISCV_PCREL_LO12_S ones...

I can open a new bug for this issue if you prefer.

Regards,

/Pete

On 2021.06.15 03:26, Daniel Schaefer wrote:
> Great commit message, thanks Sunil!
> Maintainers, please take a look and let us know if there's any other 
> concern.
> This patch lets us build the RISC-V platforms using modern toolchains 
> that are provided directly by the distributions, rather than building 
> your own from source.
> 
> Thanks,
> Daniel
> ------------------------------------------------------------------------
> *From:* Sunil V L <sunilvl at ventanamicro.com>
> *Sent:* Friday, June 11, 2021 22:08
> *To:* devel at edk2.groups.io <devel at edk2.groups.io>
> *Cc:* Chang, Abner (HPS SW/FW Technologist) <abner.chang at hpe.com>; 
> Schaefer, Daniel <daniel.schaefer at hpe.com>; Bob Feng 
> <bob.c.feng at intel.com>; Liming Gao <gaoliming at byosoft.com.cn>; Yuwei 
> Chen <yuwei.chen at intel.com>; Heinrich Schuchardt <xypron.glpk at gmx.de>
> *Subject:* Re: [RESEND PATCH v2] BaseTools: Add support for RISCV 
> GOT/PLT relocations
> Hi,
>      I just edited the commit message to indicate the module and CC the
>      maintainers. Could I get the feedback please?
> Thanks
> Sunil
> 
> On Fri, Jun 11, 2021 at 07:35:03PM +0530, Sunil V L wrote:
>> Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=3096 
> <https://bugzilla.tianocore.org/show_bug.cgi?id=3096>
>> 
>> This patch adds support for R_RISCV_CALL_PLT and R_RISCV_GOT_HI20
>> relocations generated by PIE enabled compiler. This also needed
>> changes to R_RISCV_32 and R_RISCV_64 relocations as explained in
>> https://github.com/riscv/riscv-gnu-toolchain/issues/905#issuecomment-846682710 
> <https://github.com/riscv/riscv-gnu-toolchain/issues/905#issuecomment-846682710>
>> 
>> Changes in v2:
>>   - Addressed Daniel's comment on formatting
>> 
>> Testing:
>> 1) Debian GCC 8.3.0 and booted sifive_u and QMEU virt models.
>> 2) Debian 10.2.0 and booted QEMU virt model.
>> 3) riscv-gnu-tool chain 9.2 and booted QEMU virt model.
>> 
>> Signed-off-by: Sunil V L <sunilvl at ventanamicro.com>
>> 
>> Acked-by: Abner Chang <abner.chang at hpe.com>
>> Reviewed-by: Daniel Schaefer <daniel.schaefer at hpe.com>
>> Tested-by: <daniel.schaefer at hpe.com>
>> 
>> Cc: Bob Feng <bob.c.feng at intel.com>
>> Cc: Liming Gao <gaoliming at byosoft.com.cn>
>> Cc: Yuwei Chen <yuwei.chen at intel.com>
>> Cc: Heinrich Schuchardt <xypron.glpk at gmx.de>
>> ---
>>  BaseTools/Source/C/GenFw/Elf64Convert.c | 44 +++++++++++++++++++++----
>>  1 file changed, 38 insertions(+), 6 deletions(-)
>> 
>> diff --git a/BaseTools/Source/C/GenFw/Elf64Convert.c b/BaseTools/Source/C/GenFw/Elf64Convert.c
>> index d097db8632..d684318269 100644
>> --- a/BaseTools/Source/C/GenFw/Elf64Convert.c
>> +++ b/BaseTools/Source/C/GenFw/Elf64Convert.c
>> @@ -129,6 +129,8 @@ STATIC UINT32 mDebugOffset;
>>  STATIC UINT8       *mRiscVPass1Targ = NULL;
>>  STATIC Elf_Shdr    *mRiscVPass1Sym = NULL;
>>  STATIC Elf64_Half  mRiscVPass1SymSecIndex = 0;
>> +STATIC INT32       mRiscVPass1Offset;
>> +STATIC INT32       mRiscVPass1GotFixup;
>>  
>>  //
>>  // Initialization Function
>> @@ -479,11 +481,11 @@ WriteSectionRiscV64 (
>>      break;
>>  
>>    case R_RISCV_32:
>> -    *(UINT32 *)Targ = (UINT32)((UINT64)(*(UINT32 *)Targ) - SymShdr->sh_addr + mCoffSectionsOffset[Sym->st_shndx]);
>> +    *(UINT64 *)Targ = Sym->st_value + Rel->r_addend;
>>      break;
>>  
>>    case R_RISCV_64:
>> -    *(UINT64 *)Targ = *(UINT64 *)Targ - SymShdr->sh_addr + mCoffSectionsOffset[Sym->st_shndx];
>> +    *(UINT64 *)Targ = Sym->st_value + Rel->r_addend;
>>      break;
>>  
>>    case R_RISCV_HI20:
>> @@ -533,6 +535,18 @@ WriteSectionRiscV64 (
>>      mRiscVPass1SymSecIndex = 0;
>>      break;
>>  
>> +  case R_RISCV_GOT_HI20:
>> +    Value = (Sym->st_value - Rel->r_offset);
>> +    mRiscVPass1Offset = RV_X(Value, 0, 12);
>> +    Value = RV_X(Value, 12, 20);
>> +    *(UINT32 *)Targ = (Value << 12) | (RV_X(*(UINT32*)Targ, 0, 12));
>> +
>> +    mRiscVPass1Targ = Targ;
>> +    mRiscVPass1Sym = SymShdr;
>> +    mRiscVPass1SymSecIndex = Sym->st_shndx;
>> +    mRiscVPass1GotFixup = 1;
>> +    break;
>> +
>>    case R_RISCV_PCREL_HI20:
>>      mRiscVPass1Targ = Targ;
>>      mRiscVPass1Sym = SymShdr;
>> @@ -545,11 +559,17 @@ WriteSectionRiscV64 (
>>      if (mRiscVPass1Targ != NULL && mRiscVPass1Sym != NULL && mRiscVPass1SymSecIndex != 0) {
>>        int i;
>>        Value2 = (UINT32)(RV_X(*(UINT32 *)mRiscVPass1Targ, 12, 20));
>> -      Value = (UINT32)(RV_X(*(UINT32 *)Targ, 20, 12));
>> -      if(Value & (RISCV_IMM_REACH/2)) {
>> -        Value |= ~(RISCV_IMM_REACH-1);
>> +
>> +      if(mRiscVPass1GotFixup) {
>> +        Value = (UINT32)(mRiscVPass1Offset);
>> +      } else {
>> +        Value = (UINT32)(RV_X(*(UINT32 *)Targ, 20, 12));
>> +        if(Value & (RISCV_IMM_REACH/2)) {
>> +          Value |= ~(RISCV_IMM_REACH-1);
>> +        }
>>        }
>>        Value = Value - (UINT32)mRiscVPass1Sym->sh_addr + mCoffSectionsOffset[mRiscVPass1SymSecIndex];
>> +
>>        if(-2048 > (INT32)Value) {
>>          i = (((INT32)Value * -1) / 4096);
>>          Value2 -= i;
>> @@ -569,12 +589,21 @@ WriteSectionRiscV64 (
>>          }
>>        }
>>  
>> -      *(UINT32 *)Targ = (RV_X(Value, 0, 12) << 20) | (RV_X(*(UINT32*)Targ, 0, 20));
>> +      if(mRiscVPass1GotFixup) {
>> +        *(UINT32 *)Targ = (RV_X((UINT32)Value, 0, 12) << 20)
>> +                            | (RV_X(*(UINT32*)Targ, 0, 20));
>> +        /* Convert LD instruction to ADDI */
>> +        *(UINT32 *)Targ = ((*(UINT32 *)Targ & ~0x707f) | 0x13);
>> +      } else {
>> +        *(UINT32 *)Targ = (RV_X(Value, 0, 12) << 20) | (RV_X(*(UINT32*)Targ, 0, 20));
>> +      }
>>        *(UINT32 *)mRiscVPass1Targ = (RV_X(Value2, 0, 20)<<12) | (RV_X(*(UINT32 *)mRiscVPass1Targ, 0, 12));
>>      }
>>      mRiscVPass1Sym = NULL;
>>      mRiscVPass1Targ = NULL;
>>      mRiscVPass1SymSecIndex = 0;
>> +    mRiscVPass1Offset = 0;
>> +    mRiscVPass1GotFixup = 0;
>>      break;
>>  
>>    case R_RISCV_ADD64:
>> @@ -586,6 +615,7 @@ WriteSectionRiscV64 (
>>    case R_RISCV_GPREL_I:
>>    case R_RISCV_GPREL_S:
>>    case R_RISCV_CALL:
>> +  case R_RISCV_CALL_PLT:
>>    case R_RISCV_RVC_BRANCH:
>>    case R_RISCV_RVC_JUMP:
>>    case R_RISCV_RELAX:
>> @@ -1528,6 +1558,7 @@ WriteRelocations64 (
>>              case R_RISCV_GPREL_I:
>>              case R_RISCV_GPREL_S:
>>              case R_RISCV_CALL:
>> +            case R_RISCV_CALL_PLT:
>>              case R_RISCV_RVC_BRANCH:
>>              case R_RISCV_RVC_JUMP:
>>              case R_RISCV_RELAX:
>> @@ -1537,6 +1568,7 @@ WriteRelocations64 (
>>              case R_RISCV_SET16:
>>              case R_RISCV_SET32:
>>              case R_RISCV_PCREL_HI20:
>> +            case R_RISCV_GOT_HI20:
>>              case R_RISCV_PCREL_LO12_I:
>>                break;
>>  
>> -- 
>> 2.25.1
>> 
> 



-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#76591): https://edk2.groups.io/g/devel/message/76591
Mute This Topic: https://groups.io/mt/83468967/1813853
Group Owner: devel+owner at edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [edk2-devel-archive at redhat.com]
-=-=-=-=-=-=-=-=-=-=-=-





More information about the edk2-devel-archive mailing list