[LLD] Add memory ORIGIN and LENGTH expression support

[raidiun]

Summary

Adds support for the ORIGIN and MEMORY functions in expression parsing.
ORIGIN produces an absolute address based on the memory section it references.
LENGTH produces the length of the referenced memory section.

[ruiu]

ELF/LinkerScript.cpp

1871

Is this guaranteed that MemoryRegion[Name] exists?

1872

Fix indentation

1876

Ditto

test/ELF/linkerscript/symbol-memoryexpr.s

34

Please add a newline at end.

[raidiun]

Fixes indentation and newline.

[raidiun]

No, MemoryRegion[Name] is not guaranteed to exist.

Is a call to setError sufficient to generate the error?
Should probably also add a test case for this.

[raidiun]

Add parse-time checking for existence of MemoryRegion and test case for undefined region.
NB: Error is triggered if MEMORY definitions come after SECTION. Not sure how ld handles this.

[ruiu]

ELF/LinkerScript.cpp

1871

I believe that if a linker script file contains MEMORY commands, they must appear before any SECTIONS commands. Therefore, we can check the existence of MemoryRegions[Name] outside the lambdas. (And raising an error early is better than doing that in the lambdas.)

1873

Why return has braces?

1877

Please format this patch in the LLVM style.

test/ELF/linkerscript/symbol-memoryexpr.s

34

You added two newlines -- I wanted to make sure that last line ends with "\n" (which didn't). Now it ends with "\n\n".

[raidiun]

Better MemoryRegion existence check
Updates styling
Removes superfluous braces
Removes extra newline

[ruiu]

Almost fine, a few nits.

ELF/LinkerScript.cpp

1871

Do s/ScriptConfig/Opt/g

(Opt is an alias to ScriptConfig in this class.)

[raidiun]

OK to use Opt in ifs at 1871 and 1877.
Using Opt in lambdas results in segfault. Presumably instance is gone by the time the lambda expressions are called. Could evaluate outside lambda expressions if preferred.
Similar pattern is used in lambdas for other commands (e.g. line 1868) accessing ScriptBase (LinkerScript.h:314).
ScriptBase->Opt may improve consistency but is currently private.

[ruiu]

LGTM

[raidiun]

Refactoring has rendered patch irrelevant. Implemented in D32934 for refactored script parser.

ELF/LinkerScript.cpp

Expr ScriptParser::readPrimary() {
   if (Tok == "ALIGNOF") {
     StringRef Name = readParenLiteral();
     return [=](uint64_t Dot) {
       return ScriptBase->getOutputSection(Location, Name)->Addralign;
     };
   }
   if (Tok == "SIZEOF_HEADERS")
     return [=](uint64_t Dot) { return ScriptBase->getHeaderSize(); };
+  if (Tok == "ORIGIN") {
+    StringRef Name = readParenLiteral();
+    if (ScriptConfig->MemoryRegions.count(Name) == 0)
+      setError("memory region not defined: " + Name);
+    return [=](uint64_t Dot) { return ScriptConfig->MemoryRegions[Name].Origin; };
+  }
+  if (Tok == "LENGTH") {
+    StringRef Name = readParenLiteral();
+    if (ScriptConfig->MemoryRegions.count(Name) == 0)
+      setError("memory region not defined: " + Name);
+    return [=](uint64_t Dot) { return ScriptConfig->MemoryRegions[Name].Length; };
+  }
 
   // Tok is a literal number.
   uint64_t V;
   if (readInteger(Tok, V))
     return [=](uint64_t Dot) { return V; };
 
   // Tok is a symbol name.
   if (Tok != "." && !isValidCIdentifier(Tok))

test/ELF/linkerscript/symbol-memoryexpr.s

+# REQUIRES: x86
+# RUN: llvm-mc -filetype=obj -triple=x86_64-unknown-linux %s -o %t
+
+# RUN: echo "MEMORY { \
+# RUN:   ram (rwx)  : ORIGIN = 0x8000, LENGTH = 256K \
+# RUN: } \
+# RUN: SECTIONS { \
+# RUN:         origin = ORIGIN(ram); \
+# RUN:         length = LENGTH(ram); \
+# RUN:         end    = ORIGIN(ram) + LENGTH(ram); \
+# RUN:       }" > %t.script
+# RUN: ld.lld -o %t1 --script %t.script %t
+# RUN: llvm-objdump -t %t1 | FileCheck %s
+
+# CHECK:      SYMBOL TABLE:
+# CHECK-NEXT: 0000000000000000 *UND* 00000000
+# CHECK-NEXT: 0000000000008000 .text 00000000 _start
+# CHECK-NEXT: 0000000000008000 *ABS* 00000000 origin
+# CHECK-NEXT: 0000000000040000 *ABS* 00000000 length
+# CHECK-NEXT: 0000000000048000 *ABS* 00000000 end
+
+# RUN: echo "SECTIONS { \
+# RUN:         no_exist_origin = ORIGIN(ram); \
+# RUN:         no_exist_length = LENGTH(ram); \
+# RUN:       }" > %t2.script
+# RUN: not ld.lld -o %t2 --script %t2.script %t 2>&1 \
+# RUN:  | FileCheck -check-prefix=ERR %s
+# ERR: {{.*}}.script:1: memory region not defined: ram
+
+
+.global _start
+_start:
+ nop