[RISCV] Allow llvm-objdump to disassemble objects with unrecognised versions of known extensions

[asb]

Summary

This Moves ELFObjectFile to using RISCVISAInfo::parseNormalizedArchString which is not an NFC, as the test changes show. D144353 transitioned LLD to using this function, which is specialised to parsing arch strings in the normalised format specified in the psABI rather than user-authored strings accepted in -march, which has greater flexibility.

parseNormalizedArchString does not ignore or produce an error for ISA extensions with a version that isn't recognised/supported by LLVM. As current GCC is marking its objects with a higher version of the A, F, and D extensions than LLVM (see extension versioning discussion this massively improves the usability of llvm-objdump with such binaries.

This patch also solves the issues addressed in D146070, though I've opted to post both because I suspect this patch may attract more discussion about questions such as if/when warning should be produced, while D146070 is hopefully more straightforward to land. One challenge with producing warnings for unrecognised versions is that LLVM lacks a good infrastructure for doing this from utility code like RISCVISAInfo.

[MaskRay]

The unknown_ext_version.s case is common (for GCC uses) and will get improved.

It seems that llvm-objdump -d noncanonicalized_arch.o can ideally report a warning and continue dumping. The case is rare and less important, so stopping dumping for the case is fine with me.

llvm/test/tools/llvm-objdump/ELF/RISCV/riscv-attributes.s

19

Use FileCheck -DFILE=noncanonicalized_arch.o

Add error: to check that this is an error.

Add NONCANON-NOT: {{.}} to assert that there is no dumping after the error message.

[asb]

Rebase and address review comments.

llvm/lib/Object/ELFObjectFile.cpp

Expected<SubtargetFeatures> ELFObjectFileBase::getRISCVFeatures() const {
   RISCVAttributeParser Attributes;
   if (Error E = getBuildAttributes(Attributes)) {
     return std::move(E);
   }
 
   std::optional<StringRef> Attr =
       Attributes.getAttributeString(RISCVAttrs::ARCH);
   if (Attr) {
-    // Suppress version checking for experimental extensions to prevent erroring
-    // when getting any unknown version of experimental extension.
-    auto ParseResult = RISCVISAInfo::parseArchString(
-        *Attr, /*EnableExperimentalExtension=*/true,
-        /*ExperimentalExtensionVersionCheck=*/false,
-        /*IgnoreUnknown=*/true);
+    auto ParseResult = RISCVISAInfo::parseNormalizedArchString(*Attr);
     if (!ParseResult)
       return ParseResult.takeError();
     auto &ISAInfo = *ParseResult;
 
     if (ISAInfo->getXLen() == 32)
       Features.AddFeature("64bit", false);
     else if (ISAInfo->getXLen() == 64)
       Features.AddFeature("64bit");

llvm/test/tools/llvm-objdump/ELF/RISCV/riscv-attributes.s

 # RUN: rm -rf %t && split-file %s %t && cd %t
 
 # RUN: llvm-mc -filetype=obj -triple=riscv64 -mattr=+m,+f,+d,+v noncanonicalized_arch.s -o noncanonicalized_arch.o
-# RUN: llvm-objdump -d noncanonicalized_arch.o | FileCheck %s --check-prefix=NONCANON
+# RUN: not llvm-objdump -d noncanonicalized_arch.o 2>&1 | FileCheck %s --check-prefix=NONCANON
 
 # RUN: llvm-mc -filetype=obj -triple=riscv64 invalid_arch.s -o invalid_arch.o
 # RUN: not llvm-objdump -d invalid_arch.o 2>&1 | FileCheck %s --check-prefix=INVALID
 
 # RUN: llvm-mc -filetype=obj -triple=riscv32 unknown_i_version.s -o unknown_i_version.o
 # RUN: llvm-objdump -d unknown_i_version.o 2>&1 | FileCheck %s --check-prefix=UNKNOWN-I-VERSION
 
 # RUN: llvm-mc -filetype=obj -triple=riscv32 -mattr=+zicbom unknown_ext_version.s -o unknown_ext_version.o
 # RUN: llvm-objdump -d unknown_ext_version.o 2>&1 | FileCheck %s --check-prefix=UNKNOWN-EXT-VERSION
 
 # RUN: llvm-mc -filetype=obj -triple=riscv64 unknown_ext_name.s -o unknown_ext_name.o
 # RUN: llvm-objdump -d unknown_ext_name.o 2>&1 | FileCheck %s --check-prefix=UNKNOWN-EXT-NAME
 
 #--- noncanonicalized_arch.s
-# NONCANON: vsetvli a3, a2, e8, m8, tu, mu
+# NONCANON: arch string must begin with valid base ISA
 vsetvli a3, a2, e8, m8, tu, mu
 
 .section .riscv.attributes,"",@0x70000003
 .byte 0x41
 .long .Lend-.riscv.attributes-1
 .asciz "riscv"  # vendor
 .Lbegin:
 .byte 1  # Tag_File
 .long .Lend-.Lbegin
 .byte 5  # Tag_RISCV_arch
 .asciz "rv64gcv"
 .Lend:
 
 #--- invalid_arch.s
-# INVALID: string must begin with rv32{i,e,g} or rv64{i,g} 
+# INVALID: arch string must begin with valid base ISA
 nop
 
 .section .riscv.attributes,"",@0x70000003
 .byte 0x41
 .long .Lend-.riscv.attributes-1
 .asciz "riscv"  # vendor
 .Lbegin:
 .byte 1  # Tag_File

 .Lbegin:
 .byte 1  # Tag_File
 .long .Lend-.Lbegin
 .byte 5  # Tag_RISCV_arch
 .asciz "rv32i99p99"
 .Lend:
 
 #--- unknown_ext_version.s
-# UNKNOWN-EXT-VERSION: <unknown>
+# UNKNOWN-EXT-VERSION: cbo.clean (t0)
 cbo.clean (t0)
 
 .section .riscv.attributes,"",@0x70000003
 .byte 0x41
 .long .Lend-.riscv.attributes-1
 .asciz "riscv"  # vendor
 .Lbegin:
 .byte 1  # Tag_File