[yaml2obj] - Add a way to override the sh_addralign field of a section.

[grimar]

Summary

Imagine the following declaration of a section:

Sections:
  - Name:         .dynsym
    Type:         SHT_DYNSYM
    AddressAlign: 0x1111111111111111

The aligment is large and yaml2obj reports an error currently:
"the desired output size is greater than permitted. Use the --max-size option to change the limit"

This patch implements the "ShAddrAlign" key, which is similar to other "Sh*" keys we have.
With it it is possible to override the sh_addralign field, ignoring the writing of alignment bytes.

[jhenderson]

llvm/test/tools/yaml2obj/ELF/override-shaddralign.yaml

5–10

I'm not sure we need to do this. Showing that the output offset matches what you'd expect due to the AddressAlign field and not the ShAddrAlign field would be better.

40–42

I don't understand why we need multiple different kinds of sections. This is generic behaviour, so doesn't need testing with everything, right?

[grimar]

llvm/test/tools/yaml2obj/ELF/override-shaddralign.yaml

40–42

Yes, though we do it for ShName , ShOffset and ShSize. I mostly wanted to be consistent.

[grimar]

• Addressed review comments.

[MaskRay]

llvm/test/tools/yaml2obj/ELF/override-shaddralign.yaml

6

use of "ShAddrAlign" key -> ShAddrAlign or the "ShAddrAlign" key

[jhenderson]

One nit, otherwise LGTM.

llvm/test/tools/yaml2obj/ELF/override-shaddralign.yaml

26

I'd explicitly set this to be something larger than the size of the ELF header, e.g. 0x80, to demonstrate that alignment has actually occurred here.

llvm/include/llvm/ObjectYAML/ELFYAML.h

struct Section : public Chunk {
   // validation.
   virtual std::vector<std::pair<StringRef, bool>> getEntries() const {
     return {};
   };
 
   // The following members are used to override section fields which is
   // useful for creating invalid objects.
 
+  // This can be used to override the sh_addralign field.
+  Optional<llvm::yaml::Hex64> ShAddrAlign;
+
   // This can be used to override the offset stored in the sh_name field.
   // It does not affect the name stored in the string table.
   Optional<llvm::yaml::Hex64> ShName;
 
   // This can be used to override the sh_offset field. It does not place the
   // section data at the offset specified.
   Optional<llvm::yaml::Hex64> ShOffset;
 

llvm/lib/ObjectYAML/ELFEmitter.cpp

unsigned ELFState<ELFT>::toSymbolIndex(StringRef S, StringRef LocSec,
   }
   return Index;
 }
 
 template <class ELFT>
 static void overrideFields(ELFYAML::Section *From, typename ELFT::Shdr &To) {
   if (!From)
     return;
+  if (From->ShAddrAlign)
+    To.sh_addralign = *From->ShAddrAlign;
   if (From->ShFlags)
     To.sh_flags = *From->ShFlags;
   if (From->ShName)
     To.sh_name = *From->ShName;
   if (From->ShOffset)
     To.sh_offset = *From->ShOffset;
   if (From->ShSize)
     To.sh_size = *From->ShSize;

llvm/lib/ObjectYAML/ELFYAML.cpp

static void commonSectionMapping(IO &IO, ELFYAML::Section &Section) {
 
   IO.mapOptional("Content", Section.Content);
   IO.mapOptional("Size", Section.Size);
 
   // obj2yaml does not dump these fields. They are expected to be empty when we
   // are producing YAML, because yaml2obj sets appropriate values for them
   // automatically when they are not explicitly defined.
   assert(!IO.outputting() ||
-         (!Section.ShOffset.hasValue() && !Section.ShSize.hasValue() &&
-          !Section.ShName.hasValue() && !Section.ShFlags.hasValue() &&
-          !Section.ShType.hasValue()));
+         (!Section.ShOffset && !Section.ShSize && !Section.ShName &&
+          !Section.ShFlags && !Section.ShType && !Section.ShAddrAlign));
+  IO.mapOptional("ShAddrAlign", Section.ShAddrAlign);
   IO.mapOptional("ShName", Section.ShName);
   IO.mapOptional("ShOffset", Section.ShOffset);
   IO.mapOptional("ShSize", Section.ShSize);
   IO.mapOptional("ShFlags", Section.ShFlags);
   IO.mapOptional("ShType", Section.ShType);
 }
 
 static void sectionMapping(IO &IO, ELFYAML::DynamicSection &Section) {

llvm/test/tools/yaml2obj/ELF/override-shaddralign.yaml

+## Check we are able to set a custom sh_addralign field for different sections
+## and that doing this does not affect the output size.
+
+## Test that we are able to override the sh_addralign section
+## field with use of the "ShAddrAlign" key.
+
+# RUN: yaml2obj %s -o %t
+# RUN: llvm-readelf --sections %t | FileCheck %s --check-prefix=CHECK
+
+# CHECK:      Section Headers:
+# CHECK-NEXT:   [Nr] Name    {{.*}} Off    Size   ES Flg Lk Inf Al
+# CHECK-NEXT:   [ 0]         {{.*}} 000000 000000 00     0   0  0
+# CHECK-NEXT:   [ 1] .foo    {{.*}} 000080 000000 00     0   0 1229782938247303441
+# CHECK-NEXT:   [ 2] .bar    {{.*}} 000100 000000 00     0   0 2459565876494606882
+# CHECK-NEXT:   [ 3] .strtab {{.*}} 000100 000001 00     0   0  1
+
+--- !ELF
+FileHeader:
+  Class: ELFCLASS64
+  Data:  ELFDATA2LSB
+  Type:  ET_REL
+Sections:
+  - Name:         .foo
+    Type:         SHT_PROGBITS
+    AddressAlign: 0x80
+    ShAddrAlign:  0x1111111111111111
+  - Name:         .bar
+    Type:         SHT_PROGBITS
+    AddressAlign: 0x100
+    ShAddrAlign:  0x2222222222222222