[ELF] - Linkerscript: implement NOLOAD section type.

[grimar]

Summary

This is PR32351

Each output section may have a type. The type is a keyword in parentheses. The following types are defined:
NOLOAD
The section should be marked as not loadable, so that it will not be loaded into memory when the program is run.
(https://sourceware.org/binutils/docs/ld/Output-Section-Type.html#Output-Section-Type)

Parch makes such sections to be SHT_NOBITS. That looks consistent with what BFD do. Gold do something different now,
what looks like a wrong implementation simply.

Use case is next (taken from https://mcuoneclipse.com/2014/04/19/gnu-linker-can-you-not-initialize-my-variable/):

1. We want to place variable at absolute address and use next code:

static unsigned char Var[4096] __attribute__((section (".data_noload_a")));

2. Our script looks like:

.data_noload_a 0x2000000 : { *(.data_noload_a) }

Problem is that Var will stop be SHT_NOBITS and became SHT_PROGBITS.
(if attribute used with section, the variable is treated like an initialized variable, see article above)
We may want to save space, and do not write 4096 of zeroes to output file. NOLOAD is a solution.

NOLOAD forces output section to be SHT_NOBITS.

[grimar]

• Addressed review comments.

[ruiu]

ELF/LinkerScript.h

134

Noload is probably a better name.

ELF/ScriptParser.cpp

575

Read a section type. Currently, only NOLOAD is supported.

577

You can move this above if (peek() != ":"), no?

[grimar]

• Addressed review comments.

ELF/LinkerScript.h

134

Done.

ELF/ScriptParser.cpp

575

Done.

577

Do you mean something like this (new diff) ?

[ruiu]

ELF/ScriptParser.cpp

576

Why can't you do peek() != "("?

[grimar]

• Addressed review comments.

ELF/ScriptParser.cpp

576

My mistype/mistake =\ Sure we can do that.

[ruiu]

LGTM

ELF/LinkerScript.cpp

431–432

Move this before SecToCmd[Sec] = Cmd; so that assignments to Sec->{SectionIndex,Type} are next to each other in the source code.

ELF/LinkerScript.h

struct OutputSectionCommand : BaseCommand {
   OutputSectionCommand(StringRef Name)
       : BaseCommand(OutputSectionKind), Name(Name) {}
 
   static bool classof(const BaseCommand *C);
 
   OutputSection *Sec = nullptr;
   MemoryRegion *MemRegion = nullptr;
   StringRef Name;
+  StringRef Type;
   Expr AddrExpr;
   Expr AlignExpr;
   Expr LMAExpr;
   Expr SubalignExpr;
   std::vector<BaseCommand *> Commands;
   std::vector<StringRef> Phdrs;
   llvm::Optional<uint32_t> Filler;
   ConstraintKind Constraint = ConstraintKind::NoConstraint;
   std::string Location;
   std::string MemoryRegionName;
 
   template <class ELFT> void writeTo(uint8_t *Buf);
   uint32_t getFiller();
 };
 
 // This struct represents one section match pattern in SECTIONS() command.
 // It can optionally have negative match pattern for EXCLUDED_FILE command.
 // Also it may be surrounded with SORT() command, so contains sorting rules.
 struct SectionPattern {

ELF/LinkerScript.cpp

if (auto *Cmd = dyn_cast<OutputSectionCommand>(Opt.Commands[I])) {
 
       // Add input sections to an output section.
       for (InputSectionBase *S : V)
         Factory.addInputSec(S, Cmd->Name, Cmd->Sec);
       if (OutputSection *Sec = Cmd->Sec) {
         assert(Sec->SectionIndex == INT_MAX);
         Sec->SectionIndex = I;
         SecToCommand[Sec] = Cmd;
+
+        if (!Cmd->Type.empty())
+          Sec->Type = SHT_NOBITS;
       }
     }
   }
   CurOutSec = nullptr;
 }
 
 void LinkerScript::fabricateDefaultCommands() {
   std::vector<BaseCommand *> Commands;
template <class ELFT> void OutputSectionCommand::writeTo(uint8_t *Buf) {
   // -compress-debug-sections option. Write it together with header.
   if (!Sec->CompressedData.empty()) {
     memcpy(Buf, Sec->ZDebugHeader.data(), Sec->ZDebugHeader.size());
     memcpy(Buf + Sec->ZDebugHeader.size(), Sec->CompressedData.data(),
            Sec->CompressedData.size());
     return;
   }
 
+  if (Sec->Type == SHT_NOBITS)
+    return;
+
   // Write leading padding.
   ArrayRef<InputSection *> Sections = Sec->Sections;
   uint32_t Filler = getFiller();
   if (Filler)
     fill(Buf, Sections.empty() ? Sec->Size : Sections[0]->OutSecOff, Filler);
 
   parallelForEachN(0, Sections.size(), [=](size_t I) {
     InputSection *IS = Sections[I];

ELF/ScriptParser.cpp

 
 OutputSectionCommand *
 ScriptParser::readOutputSectionDescription(StringRef OutSec) {
   OutputSectionCommand *Cmd = make<OutputSectionCommand>(OutSec);
   Cmd->Location = getCurrentLocation();
 
   // Read an address expression.
   // https://sourceware.org/binutils/docs/ld/Output-Section-Address.html
-  if (peek() != ":")
+  if (peek() != ":" && !peek().startswith("("))
     Cmd->AddrExpr = readExpr();
 
+  // Read section type. Only NOLOAD is supported currently.
+  // https://sourceware.org/binutils/docs/ld/Output-Section-Type.html
+  if (peek() != ":") {
+    expect("(");
+    Cmd->Type = peek();
+    expect("NOLOAD");
+    expect(")");
+  }
+
   expect(":");
 
   if (consume("AT"))
     Cmd->LMAExpr = readParenExpr();
   if (consume("ALIGN"))
     Cmd->AlignExpr = readParenExpr();
   if (consume("SUBALIGN"))
     Cmd->SubalignExpr = readParenExpr();

test/ELF/linkerscript/noload.s

+# REQUIRES: x86
+# RUN: llvm-mc -filetype=obj -triple=x86_64-unknown-linux %s -o %t.o
+# RUN: echo "SECTIONS { \
+# RUN:        .data_noload_a (NOLOAD) : { *(.data_noload_a) } \
+# RUN:        .data_noload_b 0x10000 (NOLOAD) : { *(.data_noload_b) } };" > %t.script
+# RUN: ld.lld -o %t --script %t.script %t.o
+# RUN: llvm-readobj --symbols -sections %t
+
+# CHECK:      Section {
+# CHECK-NEXT:   Index: 2
+# CHECK-NEXT:   Name: .data_noload_a
+# CHECK-NEXT:   Type: SHT_NOBITS
+# CHECK-NEXT:   Flags [
+# CHECK-NEXT:     SHF_ALLOC
+# CHECK-NEXT:     SHF_WRITE
+# CHECK-NEXT:   ]
+# CHECK-NEXT:   Address: 0x0
+# CHECK-NEXT:   Offset: 0x1000
+# CHECK-NEXT:   Size: 4096
+# CHECK-NEXT:   Link: 0
+# CHECK-NEXT:   Info: 0
+# CHECK-NEXT:   AddressAlignment: 1
+# CHECK-NEXT:   EntrySize: 0
+# CHECK-NEXT: }
+# CHECK-NEXT: Section {
+# CHECK-NEXT:   Index: 3
+# CHECK-NEXT:   Name: .data_noload_b
+# CHECK-NEXT:   Type: SHT_NOBITS
+# CHECK-NEXT:   Flags [
+# CHECK-NEXT:     SHF_ALLOC
+# CHECK-NEXT:     SHF_WRITE
+# CHECK-NEXT:   ]
+# CHECK-NEXT:   Address: 0x10000
+# CHECK-NEXT:   Offset: 0x1000
+# CHECK-NEXT:   Size: 4096
+# CHECK-NEXT:   Link: 0
+# CHECK-NEXT:   Info: 0
+# CHECK-NEXT:   AddressAlignment: 1
+# CHECK-NEXT:   EntrySize: 0
+# CHECK-NEXT: }
+
+.section .data_noload_a,"aw",@progbits
+.zero 4096
+
+.section .data_noload_b,"aw",@progbits
+.zero 4096