[RISCV] Support Parsing Nonstandard Relocations

[lenary]

This allows nonstandard relocation names to be used in .reloc assembly directives (giving the correct relocation number).

No translation is done by the assembler into R_RISCV_CUSTOM<n> names, and the assembler does not automatically add the relevant R_RISCV_VENDOR relocation with the vendor symbol. If we want, we can have a different directive that does this later.

The first batch of relocations to be added are from Qualcomm's RISC-V psABI extensions.

[llvmbot]

@llvm/pr-subscribers-mc
@llvm/pr-subscribers-backend-risc-v

@llvm/pr-subscribers-llvm-binary-utilities

Author: Sam Elliott (lenary)

Changes

This allows nonstandard relocation names to be used in .reloc assembly directives (giving the correct relocation number).

No translation is done by the assembler into R_RISCV_CUSTOM<n> names, and the assembler does not automatically add the relevant R_RISCV_VENDOR relocation with the vendor symbol. If we want, we can have a different directive that does this later.

The first batch of relocations to be added are from Qualcomm's RISC-V psABI extensions.

---

Full diff: https://github.com/llvm/llvm-project/pull/119909.diff

3 Files Affected:

• (added) llvm/include/llvm/BinaryFormat/ELFRelocs/RISCV_nonstandard.def (+30)
• (modified) llvm/lib/Target/RISCV/MCTargetDesc/RISCVAsmBackend.cpp (+4-1)
• (modified) llvm/test/MC/RISCV/custom_reloc.s (+17)

diff --git a/llvm/include/llvm/BinaryFormat/ELFRelocs/RISCV_nonstandard.def b/llvm/include/llvm/BinaryFormat/ELFRelocs/RISCV_nonstandard.def
new file mode 100644
index 00000000000000..8a9f856f196b5f
--- /dev/null
+++ b/llvm/include/llvm/BinaryFormat/ELFRelocs/RISCV_nonstandard.def
@@ -0,0 +1,30 @@
+//===--- RISC-V Nonstandard Relocation List ---------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef ELF_RISCV_NONSTANDARD_RELOC
+#error "ELF_RISCV_NONSTANDARD_RELOC must be defined"
+#endif
+
+/*
+  ELF_RISCV_NONSTANDARD_RELOC(VENDOR, NAME, ID) defines information about
+  nonstandard relocation codes. This can be used when parsing relocations, or
+  when printing them, to provide better information.
+
+  VENDOR should be the symbol name expected in the associated `R_RISCV_VENDOR`
+  relocation. NAME and ID work like `ELF_RELOC` but the mapping is not expected
+  to be 1:1.
+
+  The mapping in RISCV.def is 1:1, and should be used when the only information
+  available is the relocation enum value.
+*/
+
+/* Qualcomm Nonstandard Relocations */
+ELF_RISCV_NONSTANDARD_RELOC(QUALCOMM, R_RISCV_QC_LO20_U,     192)
+ELF_RISCV_NONSTANDARD_RELOC(QUALCOMM, R_RISCV_QC_E_BRANCH,   193)
+ELF_RISCV_NONSTANDARD_RELOC(QUALCOMM, R_RISCV_QC_E_32,       194)
+ELF_RISCV_NONSTANDARD_RELOC(QUALCOMM, R_RISCV_QC_E_JUMP_PLT, 195)
diff --git a/llvm/lib/Target/RISCV/MCTargetDesc/RISCVAsmBackend.cpp b/llvm/lib/Target/RISCV/MCTargetDesc/RISCVAsmBackend.cpp
index eab4a5e77d96e5..0cb1ef0a66b60b 100644
--- a/llvm/lib/Target/RISCV/MCTargetDesc/RISCVAsmBackend.cpp
+++ b/llvm/lib/Target/RISCV/MCTargetDesc/RISCVAsmBackend.cpp
@@ -38,9 +38,12 @@ std::optional<MCFixupKind> RISCVAsmBackend::getFixupKind(StringRef Name) const {
   if (STI.getTargetTriple().isOSBinFormatELF()) {
     unsigned Type;
     Type = llvm::StringSwitch<unsigned>(Name)
-#define ELF_RELOC(X, Y) .Case(#X, Y)
+#define ELF_RELOC(NAME, ID) .Case(#NAME, ID)
 #include "llvm/BinaryFormat/ELFRelocs/RISCV.def"
 #undef ELF_RELOC
+#define ELF_RISCV_NONSTANDARD_RELOC(_VENDOR, NAME, ID) .Case(#NAME, ID)
+#include "llvm/BinaryFormat/ELFRelocs/RISCV_nonstandard.def"
+#undef ELF_RISCV_NONSTANDARD_RELOC
                .Case("BFD_RELOC_NONE", ELF::R_RISCV_NONE)
                .Case("BFD_RELOC_32", ELF::R_RISCV_32)
                .Case("BFD_RELOC_64", ELF::R_RISCV_64)
diff --git a/llvm/test/MC/RISCV/custom_reloc.s b/llvm/test/MC/RISCV/custom_reloc.s
index d6fea8fac36713..feb6bd371e65a8 100644
--- a/llvm/test/MC/RISCV/custom_reloc.s
+++ b/llvm/test/MC/RISCV/custom_reloc.s
@@ -33,3 +33,20 @@
   nop
   # CHECK-ASM: nop
   # CHECK-OBJ: addi zero, zero, 0x0
+
+  .reloc ., R_RISCV_VENDOR,    QUALCOMM
+  .reloc ., R_RISCV_QC_LO20_U, my_bar + 2
+  addi a1, a1, 0
+  # CHECK-ASM: [[L3:.L[^:]+]]:
+  # CHECK-ASM-NEXT: .reloc [[L3]], R_RISCV_VENDOR, QUALCOMM
+  # CHECK-ASM-NEXT: [[L4:.L[^:]+]]:
+  # CHECK-ASM-NEXT: .reloc [[L4]], R_RISCV_QC_LO20_U, my_bar+2
+  # CHECK-ASM-NEXT: mv a1, a1
+
+  # CHECK-OBJ: addi a1, a1, 0
+  # CHECK-OBJ-NEXT: R_RISCV_VENDOR    QUALCOMM
+  # CHECK-OBJ-NEXT: R_RISCV_CUSTOM192 my_bar+0x2
+
+  nop
+  # CHECK-ASM: nop
+  # CHECK-OBJ: addi zero, zero, 0x0

[lenary]

The assembly streamer here just re-emits the string which it was passed from the directive parser in the call. I don't know if we want to do anything to normalise this into R_RISCV_CUSTOM<n> - it's a possibility.

[lenary]

Updated with the new relocation name from quic/riscv-elf-psabi-quic-extensions@50f54da and included these names in the RISC-V elf relocation code enum so future code can be a bit clearer.

[svs-quic]

LGTM but please wait for someone else to review before merging.

[topperc]

LGTM

[MaskRay]

Prefer //

[lenary]

Done

[MaskRay]

//

[lenary]

Done

[MaskRay]

The ASM: line, shorter than the next line by 5, should be indented to align with the following lines

ditto below

[lenary]

Done