[MC] Rename temporary symbols of empty name to ".L0 " (#89693)

MaskRay · web-flow · commit 96c45a7fa126 · 2024-04-24T13:16:02.000-07:00
Temporary symbols generated for .eh_frame and .debug_line have an empty name, which appear in .symtab in the presence of RISC-V style linker relaxation and will not be discarded by ld/objcopy --discard-locals (-X). In contrast, GNU assembler's riscv port assigns a fake name ".L0 " (with a trailing space) to these symbols so that will be discarded by ld/objcopy --discard-locals. This patch matches the GNU behavior. Since Clang's RISC-V targets pass -X to ld, and GNU ld defaults to -X for RISC-V targets, these ".L0 " symbols will be discarded after linking by default, as expected by users. The llvm-symbolizer special case for RISC-V `SF_FormatSpecific` symbols https://reviews.llvm.org/D98669 needs to be adjusted. Note: `"":` in assembly currently crashes.
diff --git a/lld/test/ELF/mips-eh_frame-pic.s b/lld/test/ELF/mips-eh_frame-pic.s
@@ -36,8 +36,8 @@
 # RELOCS:            .rel{{a?}}.eh_frame {
 # ABS32-RELOCS-NEXT:   0x1C R_MIPS_32 .text
 # ABS64-RELOCS-NEXT:   0x1C R_MIPS_64/R_MIPS_NONE/R_MIPS_NONE .text
-# PIC64-RELOCS-NEXT:   0x1C R_MIPS_PC32/R_MIPS_NONE/R_MIPS_NONE <null>
-# PIC32-RELOCS-NEXT:   0x1C R_MIPS_PC32 <null>
+# PIC64-RELOCS-NEXT:   0x1C R_MIPS_PC32/R_MIPS_NONE/R_MIPS_NONE .L0
+# PIC32-RELOCS-NEXT:   0x1C R_MIPS_PC32 .L0
 # RELOCS-NEXT:       }
 
 # ABS64-EH-FRAME: Augmentation data: 0C
diff --git a/llvm/include/llvm/Object/ELFObjectFile.h b/llvm/include/llvm/Object/ELFObjectFile.h
@@ -801,9 +801,8 @@ Expected<uint32_t> ELFObjectFile<ELFT>::getSymbolFlags(DataRefImpl Sym) const {
   } else if (EF.getHeader().e_machine == ELF::EM_RISCV) {
     if (Expected<StringRef> NameOrErr = getSymbolName(Sym)) {
       StringRef Name = *NameOrErr;
-      // Mark empty name symbols (used for label differences) and mapping
-      // symbols.
-      if (Name.empty() || Name.starts_with("$d") || Name.starts_with("$x"))
+      // Mark fake labels (used for label differences) and mapping symbols.
+      if (Name == ".L0 " || Name.starts_with("$d") || Name.starts_with("$x"))
         Result |= SymbolRef::SF_FormatSpecific;
     } else {
       // TODO: Actually report errors helpfully.
diff --git a/llvm/lib/MC/ELFObjectWriter.cpp b/llvm/lib/MC/ELFObjectWriter.cpp
@@ -725,7 +725,13 @@ void ELFWriter::computeSymbolTable(
         HasLargeSectionIndex = true;
     }
 
+    // Temporary symbols generated for certain assembler features (.eh_frame,
+    // .debug_line) of an empty name may be referenced by relocations due to
+    // linker relaxation. Rename them to ".L0 " to match the gas fake label name
+    // and allow ld/objcopy --discard-locals to discard such symbols.
     StringRef Name = Symbol.getName();
+    if (Name.empty())
+      Name = ".L0 ";
 
     // Sections have their own string table
     if (Symbol.getType() != ELF::STT_SECTION) {
diff --git a/llvm/test/CodeGen/RISCV/fixups-diff.ll b/llvm/test/CodeGen/RISCV/fixups-diff.ll
@@ -27,7 +27,7 @@ entry:
 ; CHECK:      }
 
 ; CHECK:      Section {{.*}} .rela.eh_frame {
-; CHECK-NEXT:   0x1C R_RISCV_32_PCREL <null> 0x0
+; CHECK-NEXT:   0x1C R_RISCV_32_PCREL .L0  0x0
 ; CHECK-NEXT: }
 
 !llvm.dbg.cu = !{!0}
diff --git a/llvm/test/DebugInfo/LoongArch/dwarf-loongarch-relocs.ll b/llvm/test/DebugInfo/LoongArch/dwarf-loongarch-relocs.ll
@@ -18,21 +18,21 @@
 ; RELOCS-BOTH:         Section ({{.*}}) .rela.debug_frame {
 ; RELOCS-NORL-NEXT:      0x1C R_LARCH_32 .debug_frame 0x0
 ; RELOCS-NORL-NEXT:      0x20 R_LARCH_64 .text 0x0
-; RELOCS-ENRL-NEXT:      0x1C R_LARCH_32 <null> 0x0
-; RELOCS-ENRL-NEXT:      0x20 R_LARCH_64 <null> 0x0
-; RELOCS-ENRL-NEXT:      0x28 R_LARCH_ADD64 <null> 0x0
-; RELOCS-ENRL-NEXT:      0x28 R_LARCH_SUB64 <null> 0x0
-; RELOCS-ENRL-NEXT:      0x3F R_LARCH_ADD6 <null> 0x0
-; RELOCS-ENRL-NEXT:      0x3F R_LARCH_SUB6 <null> 0x0
+; RELOCS-ENRL-NEXT:      0x1C R_LARCH_32 .L0  0x0
+; RELOCS-ENRL-NEXT:      0x20 R_LARCH_64 .L0  0x0
+; RELOCS-ENRL-NEXT:      0x28 R_LARCH_ADD64 .L0  0x0
+; RELOCS-ENRL-NEXT:      0x28 R_LARCH_SUB64 .L0  0x0
+; RELOCS-ENRL-NEXT:      0x3F R_LARCH_ADD6 .L0  0x0
+; RELOCS-ENRL-NEXT:      0x3F R_LARCH_SUB6 .L0  0x0
 ; RELOCS-BOTH-NEXT:    }
 ; RELOCS-BOTH:         Section ({{.*}}) .rela.debug_line {
 ; RELOCS-BOTH-NEXT:      0x22 R_LARCH_32 .debug_line_str 0x0
 ; RELOCS-BOTH-NEXT:      0x31 R_LARCH_32 .debug_line_str 0x2
 ; RELOCS-BOTH-NEXT:      0x46 R_LARCH_32 .debug_line_str 0x1B
 ; RELOCS-NORL-NEXT:      0x4F R_LARCH_64 .text 0x0
-; RELOCS-ENRL-NEXT:      0x4F R_LARCH_64 <null> 0x0
-; RELOCS-ENRL-NEXT:      0x5F R_LARCH_ADD16 <null> 0x0
-; RELOCS-ENRL-NEXT:      0x5F R_LARCH_SUB16 <null> 0x0
+; RELOCS-ENRL-NEXT:      0x4F R_LARCH_64 .L0  0x0
+; RELOCS-ENRL-NEXT:      0x5F R_LARCH_ADD16 .L0  0x0
+; RELOCS-ENRL-NEXT:      0x5F R_LARCH_SUB16 .L0  0x0
 ; RELOCS-BOTH-NEXT:    }
 ; RELOCS-BOTH-NEXT:  ]
 
diff --git a/llvm/test/DebugInfo/RISCV/dwarf-riscv-relocs.ll b/llvm/test/DebugInfo/RISCV/dwarf-riscv-relocs.ll
@@ -6,14 +6,14 @@
 
 ; Check that we actually have relocations, otherwise this is kind of pointless.
 ; READOBJ-RELOCS:  Section ({{.*}}) .rela.debug_info {
-; READOBJ-RELOCS:    0x1B R_RISCV_ADD32 <null> 0x0
-; READOBJ-RELOCS-NEXT:    0x1B R_RISCV_SUB32 <null> 0x0
+; READOBJ-RELOCS:    0x1B R_RISCV_ADD32 .L0  0x0
+; READOBJ-RELOCS-NEXT:    0x1B R_RISCV_SUB32 .L0  0x0
 ; READOBJ-RELOCS:  Section ({{.*}}) .rela.debug_frame {
-; READOBJ-RELOCS:    0x20 R_RISCV_ADD32 <null> 0x0
-; READOBJ-RELOCS-NEXT:    0x20 R_RISCV_SUB32 <null> 0x0
+; READOBJ-RELOCS:    0x20 R_RISCV_ADD32 .L0  0x0
+; READOBJ-RELOCS-NEXT:    0x20 R_RISCV_SUB32 .L0  0x0
 ; READOBJ-RELOCS:  Section ({{.*}}) .rela.debug_line {
-; READOBJ-RELOCS:    0x5A R_RISCV_ADD16 <null> 0x0
-; READOBJ-RELOCS-NEXT:    0x5A R_RISCV_SUB16 <null> 0x0
+; READOBJ-RELOCS:    0x5A R_RISCV_ADD16 .L0  0x0
+; READOBJ-RELOCS-NEXT:    0x5A R_RISCV_SUB16 .L0  0x0
 
 ; Check that we can print the source, even with relocations.
 ; OBJDUMP-SOURCE: Disassembly of section .text:
diff --git a/llvm/test/DebugInfo/RISCV/relax-debug-frame.ll b/llvm/test/DebugInfo/RISCV/relax-debug-frame.ll
@@ -4,11 +4,11 @@
 ; RUN:     | FileCheck -check-prefix=RELAX-DWARFDUMP %s
 ;
 ; RELAX:      Section ({{.*}}) .rela.eh_frame {
-; RELAX-NEXT:   0x1C R_RISCV_32_PCREL <null> 0x0
-; RELAX-NEXT:   0x30 R_RISCV_32_PCREL <null> 0x0
-; RELAX-NEXT:   0x44 R_RISCV_32_PCREL <null> 0x0
-; RELAX-NEXT:   0x48 R_RISCV_ADD32 <null> 0x0
-; RELAX-NEXT:   0x48 R_RISCV_SUB32 <null> 0x0
+; RELAX-NEXT:   0x1C R_RISCV_32_PCREL .L0  0x0
+; RELAX-NEXT:   0x30 R_RISCV_32_PCREL .L0  0x0
+; RELAX-NEXT:   0x44 R_RISCV_32_PCREL .L0  0x0
+; RELAX-NEXT:   0x48 R_RISCV_ADD32 .L0  0x0
+; RELAX-NEXT:   0x48 R_RISCV_SUB32 .L0  0x0
 ; RELAX-NEXT:  }
 
 ; RELAX-DWARFDUMP-NOT: error: failed to compute relocation
diff --git a/llvm/test/DebugInfo/Symbolize/ELF/riscv-temporary-symbol.s b/llvm/test/DebugInfo/Symbolize/ELF/riscv-temporary-symbol.s
@@ -1,10 +1,11 @@
 # REQUIRES: riscv-registered-target
-## Ignore empty name symbols.
+## Ignore .L0 symbols that are generated by LLVM integrated assembler and GNU
+## assembler for .debug_line/.eh_frame related assembler directives.
 
 # RUN: llvm-mc -filetype=obj -triple=riscv64 %s -o %t
 # RUN: llvm-readelf -s %t | FileCheck %s --check-prefix=SYM
 
-# SYM: 0000000000000004  0 NOTYPE LOCAL  DEFAULT [[#]] {{$}}
+# SYM: 0000000000000004  0 NOTYPE LOCAL  DEFAULT [[#]] .L0 {{$}}
 # SYM: 0000000000000000  0 NOTYPE GLOBAL DEFAULT [[#]] foo
 
 ## Make sure we test at an address larger than or equal to an empty name symbol.
diff --git a/llvm/test/ExecutionEngine/JITLink/RISCV/anonymous_symbol.s b/llvm/test/ExecutionEngine/JITLink/RISCV/anonymous_symbol.s
@@ -7,7 +7,7 @@
 # the section start and section end. So that by relocating these symbol, the section length
 # can be calculated.
 #
-# CHECK: Creating defined graph symbol for ELF symbol ""
+# CHECK: Creating defined graph symbol for ELF symbol ".L0 "
 # CHECK: Creating defined graph symbol for ELF symbol "main"
         .text
         .globl main
diff --git a/llvm/test/MC/ELF/RISCV/gen-dwarf.s b/llvm/test/MC/ELF/RISCV/gen-dwarf.s
@@ -40,28 +40,28 @@
 # CHECK-NEXT: 0x00000020: [DW_RLE_end_of_list ]
 
 # RELOC:      Section ([[#]]) .rela.eh_frame {
-# RELOC-NEXT:   0x1C R_RISCV_32_PCREL <null> 0x0
-# RELOC-NEXT:   0x20 R_RISCV_ADD32 <null> 0x0
-# RELOC-NEXT:   0x20 R_RISCV_SUB32 <null> 0x0
-# RELOC-NEXT:   0x25 R_RISCV_SET6 <null> 0x0
-# RELOC-NEXT:   0x25 R_RISCV_SUB6 <null> 0x0
-# RELOC-NEXT:   0x34 R_RISCV_32_PCREL <null> 0x0
+# RELOC-NEXT:   0x1C R_RISCV_32_PCREL .L0  0x0
+# RELOC-NEXT:   0x20 R_RISCV_ADD32 .L0  0x0
+# RELOC-NEXT:   0x20 R_RISCV_SUB32 .L0  0x0
+# RELOC-NEXT:   0x25 R_RISCV_SET6 .L0  0x0
+# RELOC-NEXT:   0x25 R_RISCV_SUB6 .L0  0x0
+# RELOC-NEXT:   0x34 R_RISCV_32_PCREL .L0  0x0
 # RELOC-NEXT: }
 
 # RELOC:      Section ([[#]]) .rela.debug_rnglists {
 # RELOC-NEXT:   0xD R_RISCV_64 .text.foo 0x0
-# RELOC-NEXT:   0x15 R_RISCV_SET_ULEB128 <null> 0x0
+# RELOC-NEXT:   0x15 R_RISCV_SET_ULEB128 .L0  0x0
 # RELOC-NEXT:   0x15 R_RISCV_SUB_ULEB128 .text.foo 0x0
 # RELOC-NEXT:   0x17 R_RISCV_64 .text.bar 0x0
 # RELOC-NEXT: }
 
 # RELOC:      Section ([[#]]) .rela.debug_line {
-# RELOC:        R_RISCV_ADD16 <null> 0x0
-# RELOC-NEXT:   R_RISCV_SUB16 <null> 0x0
-# RELOC-NEXT:   R_RISCV_ADD16 <null> 0x0
-# RELOC-NEXT:   R_RISCV_SUB16 <null> 0x0
-# RELOC-NEXT:   R_RISCV_ADD16 <null> 0x0
-# RELOC-NEXT:   R_RISCV_SUB16 <null> 0x0
+# RELOC:        R_RISCV_ADD16 .L0  0x0
+# RELOC-NEXT:   R_RISCV_SUB16 .L0  0x0
+# RELOC-NEXT:   R_RISCV_ADD16 .L0  0x0
+# RELOC-NEXT:   R_RISCV_SUB16 .L0  0x0
+# RELOC-NEXT:   R_RISCV_ADD16 .L0  0x0
+# RELOC-NEXT:   R_RISCV_SUB16 .L0  0x0
 # RELOC:      }
 
 # RELOC:      Hex dump of section '.eh_frame':
diff --git a/llvm/test/MC/RISCV/cfi-advance.s b/llvm/test/MC/RISCV/cfi-advance.s
@@ -1,13 +1,27 @@
 # RUN: llvm-mc -filetype=obj -triple riscv32 %s -o %t.o  
-# RUN: llvm-readobj -r %t.o | FileCheck -check-prefix=CHECK %s
+# RUN: llvm-readelf -sr %t.o | FileCheck %s
 # RUN: llvm-dwarfdump --debug-frame %t.o 2>&1 \
 # RUN:     | FileCheck -check-prefix=CHECK-DWARFDUMP %s
 
-# CHECK:      .rela.eh_frame {
-# CHECK-NEXT:   0x1C R_RISCV_32_PCREL <null> 0x0
-# CHECK-NEXT:   0x35 R_RISCV_SET6 <null> 0x0
-# CHECK-NEXT:   0x35 R_RISCV_SUB6 <null> 0x0
-# CHECK-NEXT: }
+
+# CHECK:      Relocation section '.rela.text1' at offset {{.*}} contains 1 entries:
+# CHECK-NEXT:  Offset     Info    Type                Sym. Value  Symbol's Name + Addend
+# CHECK-NEXT: 00000000  00000313 R_RISCV_CALL_PLT       00000004   .L0 + 0
+# CHECK-EMPTY:
+# CHECK-NEXT: Relocation section '.rela.eh_frame' at offset {{.*}} contains 3 entries:
+# CHECK:       Offset     Info    Type                Sym. Value  Symbol's Name + Addend
+# CHECK-NEXT: 0000001c  00000139 R_RISCV_32_PCREL       00000000   .L0 + 0
+# CHECK-NEXT: 00000035  00000b35 R_RISCV_SET6           00010178   .L0 + 0
+# CHECK-NEXT: 00000035  00000934 R_RISCV_SUB6           0001016e   .L0 + 0
+# CHECK-EMPTY:
+# CHECK:      Symbol table '.symtab' contains 15 entries:
+# CHECK-NEXT:    Num:    Value  Size Type    Bind   Vis       Ndx Name
+# CHECK-NEXT:      0: 00000000     0 NOTYPE  LOCAL  DEFAULT   UND
+# CHECK-NEXT:      1: 00000000     0 NOTYPE  LOCAL  DEFAULT     2 .L0 {{$}}
+# CHECK:           3: 00000004     0 NOTYPE  LOCAL  DEFAULT     2 .L0{{$}}
+# CHECK:           9: 0001016e     0 NOTYPE  LOCAL  DEFAULT     2 .L0 {{$}}
+# CHECK:          11: 00010178     0 NOTYPE  LOCAL  DEFAULT     2 .L0 {{$}}
+
 # CHECK-DWARFDUMP: DW_CFA_advance_loc1: 104
 # CHECK-DWARFDUMP-NEXT: DW_CFA_def_cfa_offset: +8
 # CHECK-DWARFDUMP-NEXT: DW_CFA_advance_loc2: 259
@@ -23,6 +37,9 @@
 test:
         .cfi_startproc
         nop
+## This looks similar to fake label names ".L0 ". Even if this is ".L0 ",
+## the assembler will not conflate it with fake labels.
+.L0:
         .zero 100, 0x90
         .cfi_def_cfa_offset 8
         nop
@@ -36,3 +53,6 @@ test:
         .cfi_def_cfa_offset 8
         nop
         .cfi_endproc
+
+.section .text1,"ax"
+call .L0
diff --git a/llvm/test/MC/RISCV/fde-reloc.s b/llvm/test/MC/RISCV/fde-reloc.s
@@ -12,7 +12,7 @@ func:
 	.cfi_endproc
 
 # CHECK:   Section (4) .rela.eh_frame {
-# CHECK-NEXT:   0x1C R_RISCV_32_PCREL <null> 0x0
+# CHECK-NEXT:   0x1C R_RISCV_32_PCREL .L0  0x0
 # CHECK-NEXT: }
 # CHECK:      Hex dump of section '.eh_frame':
 # CHECK-NEXT: 0x00000000 10000000 00000000 017a5200 017c0101
diff --git a/llvm/test/MC/RISCV/scoped-relaxation.s b/llvm/test/MC/RISCV/scoped-relaxation.s
@@ -9,7 +9,7 @@
 .dword function - .
 
 # CHECK: 0x0 R_RISCV_ADD64 function 0x0
-# CHECK-NEXT: 0x0 R_RISCV_SUB64 <null> 0x0
+# CHECK-NEXT: 0x0 R_RISCV_SUB64 .L0  0x0
 
 # Relaxed reference, this will resolve to a pair of `RISCV_ADD64` and
 # `RISCV_SUB64` relocation.
@@ -19,7 +19,7 @@
 .option pop
 
 # CHECK: 0x8 R_RISCV_ADD64 function 0x0
-# CHECK-NEXT: 0x8 R_RISCV_SUB64 <null> 0x0
+# CHECK-NEXT: 0x8 R_RISCV_SUB64 .L0  0x0
 
 # Unrelaxed reference, this will resolve to a pair of `RISCV_ADD64` and
 # `RISCV_SUB64` relocation due to relaxation being sticky to the file.
@@ -29,6 +29,6 @@
 .option pop
 
 # CHECK: 0x10 R_RISCV_ADD64 function 0x0
-# CHECK-NEXT: 0x10 R_RISCV_SUB64 <null> 0x0
+# CHECK-NEXT: 0x10 R_RISCV_SUB64 .L0  0x0
 
 # CHECK: }

Original file line number	Diff line number	Diff line change
`@@ -7,7 +7,7 @@`
`7`	`7`	`# the section start and section end. So that by relocating these symbol, the section length`
`8`	`8`	`# can be calculated.`
`9`	`9`	`#`
`10`		`-# CHECK: Creating defined graph symbol for ELF symbol ""`
	`10`	`+# CHECK: Creating defined graph symbol for ELF symbol ".L0 "`
`11`	`11`	`# CHECK: Creating defined graph symbol for ELF symbol "main"`
`12`	`12`	`.text`
`13`	`13`	`.globl main`