[llvm-objdump][ARM] support --symbolize-operands for ARM/ELF

[covanam]

Summary

Support --symbolize-operands on ARM/Thumb architecture.
Only support ELF object files for now.

[hoy]

llvm/tools/llvm-objdump/llvm-objdump.cpp

1169–1171

Is this always true for linked binary?

[Esme]

Please include as much context as possible with your diff. https://llvm.org/docs/Phabricator.html#requesting-a-review-via-the-web-interface

llvm/tools/llvm-objdump/llvm-objdump.cpp

1145

Please capitalize the first letter of variable names.

1172

ditto.

[covanam]

In D143725#4127923, @Esme wrote:

Please include as much context as possible with your diff. https://llvm.org/docs/Phabricator.html#requesting-a-review-via-the-web-interface

Thank you. Will read that and will add more context.

llvm/tools/llvm-objdump/llvm-objdump.cpp

1145

Sorry I am newcomer. clang-format didn't say anything, so I thought this is okay.
Will be fixed. Same for the other one.

1169–1171

Actually I don't know, from my practical experience, it seems to be true. However, I checked ARM's documentation on ELF files and they do not say anything about this.

I will think about this and will come back later.

[jhenderson]

llvm/tools/llvm-objdump/llvm-objdump.cpp

1145

clang-format is only for formatting. It won't do things like variable name styles. You may want to look into clang-tidy which does a lot more. Also, make sure you've fully read and digested the LLVM coding standards.

FWIW, here you can't simply do triple -> Triple, since Triple is a type. You probably can use TargetTriple or something along those lines.

llvm/docs/CommandGuide/llvm-objdump.rst

 
   When a bb-address-map section is present (i.e., the object file is built with ``-fbasic-block-sections=labels``), labels are retrieved from that section instead.
 
-  Only works with PowerPC objects or X86 linked images.
+  Only works with PowerPC objects, ARM/Thumb objects or X86 linked images.
 
   Example:
     A non-symbolized branch instruction with a local target and pc-relative memory access like

llvm/test/tools/llvm-objdump/ELF/ARM/symbolize-operand-arm.s

+@ RUN: llvm-mc %s -triple=armv8 -filetype=obj | \
+@ RUN:   llvm-objdump --symbolize-operands --no-show-raw-insn -d --triple=armv8 - | \
+@ RUN:   FileCheck %s
+
+.text
+foo:
+@ CHECK:      00000000 <foo>:
+
+@@ Check symbolize local branch
+  b .L0
+.L4:
+  beq .L1
+.L0:
+  b .L2
+  ble .L3
+.L2:
+  b .L3
+.L1:
+  b .L0
+.L3:
+  b .L4
+@ CHECK-NEXT:  0:   b   0x8  <L0>       @ imm = #0x0
+@ CHECK-NEXT: <L4>:
+@ CHECK-NEXT:  4:   beq 0x14 <L1>       @ imm = #0x8
+@ CHECK-NEXT: <L0>:
+@ CHECK-NEXT:  8:   b    0x10 <L2>       @ imm = #0x0
+@ CHECK-NEXT:  c:   ble  0x18 <L3>       @ imm = #0x4
+@ CHECK-NEXT: <L2>:
+@ CHECK-NEXT: 10:   b   0x18 <L3>       @ imm = #0x0
+@ CHECK-NEXT: <L1>:
+@ CHECK-NEXT: 14:   b   0x8  <L0>       @ imm = #-0x14
+@ CHECK-NEXT: <L3>:
+@ CHECK-NEXT: 18:   b   0x4  <L4>       @ imm = #-0x1c
+
+
+@@ Check function calls are not symbolized
+  bl bar
+@ CHECK-NEXT: 1c:   bl  0x1c <foo+0x1c> @ imm = #-0x8

llvm/test/tools/llvm-objdump/ELF/ARM/symbolize-operand-thumb.s

+@ RUN: llvm-mc %s -triple=thumbv7m -filetype=obj | \
+@ RUN:   llvm-objdump --symbolize-operands --no-show-raw-insn -d --triple=thumbv7m - | \
+@ RUN:   FileCheck %s
+
+.text
+foo:
+@ CHECK:      00000000 <foo>:
+
+@@ Check symbolize local branch
+  b .L0
+  beq.w .L1
+  b .L2
+.L0:
+  b .L2
+  bne .L1
+  b.w .L0
+.L2:
+  b.w .L1
+  b .L2
+.L1:
+  b.w .L0
+@ CHECK-NEXT:  0:   b      0x8 <L0>     @ imm = #0x4
+@ CHECK-NEXT:  2:   beq.w 0x16 <L1>    @ imm = #0x10
+@ CHECK-NEXT:  6:   b     0x10 <L2>    @ imm = #0x6
+@ CHECK-NEXT: <L0>:
+@ CHECK-NEXT:  8:   b     0x10 <L2>    @ imm = #0x4
+@ CHECK-NEXT:  a:   bne   0x16 <L1>    @ imm = #0x8
+@ CHECK-NEXT:  c:   b.w    0x8 <L0>    @ imm = #-0x8
+@ CHECK-NEXT: <L2>:
+@ CHECK-NEXT: 10:   b.w   0x16 <L1>    @ imm = #0x2
+@ CHECK-NEXT: 14:   b     0x10 <L2>    @ imm = #-0x8
+@ CHECK-NEXT: <L1>:
+@ CHECK-NEXT: 16:   b.w    0x8 <L0>    @ imm = #-0x12
+
+@@ Check function calls are not symbolized
+  bl bar
+@ CHECK-NEXT: 1a:   bl 0x1a <foo+0x1a> @ imm = #-0x4

llvm/tools/llvm-objdump/llvm-objdump.cpp

     ArrayRef<uint8_t> Bytes, const MCInstrAnalysis *MIA, MCDisassembler *DisAsm,
     MCInstPrinter *IP, const MCSubtargetInfo *STI, uint64_t SectionAddr,
     uint64_t Start, uint64_t End, std::unordered_map<uint64_t, std::string> &Labels) {
-  // So far only supports PowerPC and X86.
-  if (!STI->getTargetTriple().isPPC() && !STI->getTargetTriple().isX86())
+  // So far only supports ARM/Thumb, PowerPC and X86.
+  Triple triple = STI->getTargetTriple();
+  if (!triple.isPPC() && !triple.isX86() && !triple.isARM() &&
+      !triple.isThumb())
     return;
 
   Labels.clear();

     if (Disassembled && MIA) {
       uint64_t Target;
       bool TargetKnown = MIA->evaluateBranch(Inst, Index, Size, Target);
-      // On PowerPC, if the address of a branch is the same as the target, it
-      // means that it's a function call. Do not mark the label for this case.
+      // On PowerPC and ARM/Thumb, if the address of a branch is the same as
+      // the target, it means that it's a function call. Do not mark the label
+      // for this case.
+      bool isFunctionCall =
+          Target == Index &&
+          (triple.isPPC() || triple.isARM() || triple.isThumb());
       if (TargetKnown && (Target >= Start && Target < End) &&
-          !Labels.count(Target) &&
-          !(STI->getTargetTriple().isPPC() && Target == Index))
+          !Labels.count(Target) && !isFunctionCall)
         Labels[Target] = ("L" + Twine(LabelCount++)).str();
     }
     Index += Size;