[llvm-symbolizer] Add switch to adjust addresses by fixed offset

[jhenderson]

Summary

If a stack trace or similar has a list of addresses from a program or DSO loaded at a variable address (e.g. due to ASLR), the addresses will not directly correspond to the addresses stored in the object file. If a user wishes to use llvm-symbolizer, they have to subtract the load address from every address. This is somewhat inconvenient, especially as the output of --print-address will result in the adjusted address being listed, rather than the address coming from the stack trace, making it harder to map results between the two.

This change adds a new switch to llvm-symbolizer --adjust-vma which takes an offset, which is then used to automatically do this calculation. The printed address remains the input address (allowing for easy mapping), whilst the specified offset is applied to the addresses when performing the lookup.

The switch is conceptually similar to llvm-objdump's new switch of the same name (see D57051), which in turn mirrors a GNU switch. There is no equivalent switch in addr2line.

[jhenderson]

Add documentation.

[grimar]

I think this is a good one. Interesting what others think.

test/tools/llvm-symbolizer/adjust-vma.s

18

I would probably suggest to do:
llvm-symbolizer 0xa 0xb --print-address --adjust-vma 0x100

That way it would be easy to match this invocation with one used in the first test and see a difference in result.

I have no strong preference, just was a bit confused with 0xb at first, because it was already used as
an address in the previous test and my brain tried to match it with it :)

[jhenderson]

test/tools/llvm-symbolizer/adjust-vma.s

18

I can understand the confusion. That's what I did originally, but a colleague suggested that I change the offset to 0xa to test the case where the calculation only just overflows. How about the following:

llvm-symbolizer 0xa 0xb --print-address --adjust-vma 0xc

[grimar]

test/tools/llvm-symbolizer/adjust-vma.s

18

Yeah, that will also work, thanks!

[jhenderson]

Tweak test to reduce risk of confusion.

[grimar]

LGTM

docs/CommandGuide/llvm-symbolizer.rst

 
  Print human readable output. If ``-inlining`` is specified, enclosing scope is
  prefixed by (inlined by). Refer to listed examples.
 
 .. option:: -basenames, -s
 
  Strip directories when printing the file path.
 
+.. option:: -adjust-vma=<offset>
+
+ Add the specified offset to object file addresses when performing lookups. This
+ can be used to simplify lookups when the object is not loaded at a dynamically
+ relocated address.
+
 EXIT STATUS
 -----------
 
 :program:`llvm-symbolizer` returns 0. Other exit codes imply internal program error.

test/tools/llvm-symbolizer/adjust-vma.s

+# REQUIRES: x86-registered-target
+
+.type foo,@function
+.size foo,12
+foo:
+    .space 10
+    nop
+    nop
+
+# RUN: llvm-mc -filetype=obj -triple=x86_64-pc-linux %s -o %t.o -g
+
+# RUN: llvm-symbolizer 0xa 0xb --print-address --obj=%t.o \
+# RUN:   | FileCheck %s --check-prefix=NORMAL
+# RUN: llvm-symbolizer 0x10a 0x10b --print-address --adjust-vma 0x100 --obj=%t.o \
+# RUN:   | FileCheck %s --check-prefix=ADJUST
+
+# Show that we can handle addresses less than the offset.
+# RUN: llvm-symbolizer 0xa --print-address --adjust-vma 0xb --obj=%t.o \
+# RUN:   | FileCheck %s --check-prefix=OVERFLOW
+
+# NORMAL:      0xa
+# NORMAL-NEXT: foo
+# NORMAL-NEXT: adjust-vma.s:7:0
+# NORMAL-EMPTY:
+# NORMAL-NEXT: 0xb
+# NORMAL-NEXT: foo
+# NORMAL-NEXT: adjust-vma.s:8:0
+
+# ADJUST:      0x10a
+# ADJUST-NEXT: foo
+# ADJUST-NEXT: adjust-vma.s:7:0
+# ADJUST-EMPTY:
+# ADJUST-NEXT: 0x10b
+# ADJUST-NEXT: foo
+# ADJUST-NEXT: adjust-vma.s:8:0
+
+# OVERFLOW:      0xa
+# OVERFLOW-NEXT: ??
+# OVERFLOW-NEXT: ??

tools/llvm-symbolizer/llvm-symbolizer.cpp

 
 static cl::opt<int> ClPrintSourceContextLines(
     "print-source-context-lines", cl::init(0),
     cl::desc("Print N number of source file context"));
 
 static cl::opt<bool> ClVerbose("verbose", cl::init(false),
                                cl::desc("Print verbose line info"));
 
+// -adjust-vma
+static cl::opt<unsigned long long>
+    ClAdjustVMA("adjust-vma", cl::init(0), cl::value_desc("offset"),
+                cl::desc("Add specified offset to object file addresses"));
+
 static cl::list<std::string> ClInputAddresses(cl::Positional,
                                               cl::desc("<input addresses>..."),
                                               cl::ZeroOrMore);
 
 template<typename T>
 static bool error(Expected<T> &ResOrErr) {
   if (ResOrErr)
     return false;
static void symbolizeInput(StringRef InputString, LLVMSymbolizer &Symbolizer,
   }
 
   if (ClPrintAddress) {
     outs() << "0x";
     outs().write_hex(ModuleOffset);
     StringRef Delimiter = ClPrettyPrint ? ": " : "\n";
     outs() << Delimiter;
   }
+  ModuleOffset -= ClAdjustVMA;
   if (IsData) {
     auto ResOrErr = Symbolizer.symbolizeData(ModuleName, ModuleOffset);
     Printer << (error(ResOrErr) ? DIGlobal() : ResOrErr.get());
   } else if (ClPrintInlining) {
     auto ResOrErr =
         Symbolizer.symbolizeInlinedCode(ModuleName, ModuleOffset, ClDwpName);
     Printer << (error(ResOrErr) ? DIInliningInfo() : ResOrErr.get());
   } else {