[LLD] [COFF] Order .debug_* sections at the end, to avoid leaving gaps if stripped

[mstorsjo]

Summary

So far, we sort all discardable sections at the end, with only some
extra logic to make sure that the .reloc section is at the start
of that group of sections. But if there are other discardable
sections, other than .reloc, they must also be ordered before
.debug_* sections, to avoid leaving gaps if the executable is
stripped.

(Stripping executables doesn't remove all discardable sections,
only the ones named .debug_*).

Rust binaries seem to include a .rmeta section, which is marked
discardable. This fixes stripping such binaries if built with
dwarf debug info included.

This fixes issues observed in MSYS2 in
https://github.com/msys2/MINGW-packages/pull/10555.

[mstorsjo]

Other earlier changes in the same area can be found here:

https://github.com/llvm/llvm-project/commit/98ff9f845ded3df23b5c0a6386e3bba96d9bb557 / D49351 - "Sort .reloc before all other discardable sections"
https://github.com/llvm/llvm-project/commit/13ac40ea6ef9621838068a70cbef86d544ee9700 / D11655 - "Sort output sections which start with .debug to the end of the file"

(Despite the commit subject in the latter of those, it only looked at the IMAGE_SCN_MEM_DISCARDABLE flag.)

[mstorsjo]

Also, see https://github.com/llvm/llvm-project/blob/main/llvm/lib/ObjCopy/COFF/COFFObjcopy.cpp#L142-L147 (plus line 31) as reference to what stripping does here - llvm-objcopy matches GNU objcopy in this aspect.

[rnk]

lgtm

[mati865]

LGTM

Rebuilt Rust with this patch applied to host LLVM and it still works after stripping binaries and libraries.

[mstorsjo]

In D120805#3355837, @mati865 wrote:

LGTM

Rebuilt Rust with this patch applied to host LLVM and it still works after stripping binaries and libraries.

Thanks for testing!

lld/COFF/Writer.cpp

for (auto it : partialSections) {
     sec->addContributingPartialSection(pSec);
   }
 
   // Finally, move some output sections to the end.
   auto sectionOrder = [&](const OutputSection *s) {
     // Move DISCARDABLE (or non-memory-mapped) sections to the end of file
     // because the loader cannot handle holes. Stripping can remove other
     // discardable ones than .reloc, which is first of them (created early).
-    if (s->header.Characteristics & IMAGE_SCN_MEM_DISCARDABLE)
+    if (s->header.Characteristics & IMAGE_SCN_MEM_DISCARDABLE) {
+      // Move discardable sections named .debug_ to the end, after other
+      // discardable sections. Stripping only removes the sections named
+      // .debug_* - thus try to avoid leaving holes after stripping.
+      if (s->name.startswith(".debug_"))
+        return 3;
       return 2;
+    }
     // .rsrc should come at the end of the non-discardable sections because its
     // size may change by the Win32 UpdateResources() function, causing
     // subsequent sections to move (see https://crbug.com/827082).
     if (s == rsrcSec)
       return 1;
     return 0;
   };
   llvm::stable_sort(ctx.outputSections,

lld/test/COFF/sort-debug.test

 # RUN: yaml2obj %s -o %t.obj
 # RUN: lld-link /debug /out:%t.exe /entry:main %t.obj
 # RUN: llvm-readobj --sections %t.exe | FileCheck %s
 # RUN: lld-link /debug:dwarf /out:%t.exe /entry:main %t.obj
 # RUN: llvm-readobj --sections %t.exe | FileCheck %s
 # RUN: lld-link /out:%t.exe /entry:main %t.obj
 # RUN: llvm-readobj --sections %t.exe | FileCheck -check-prefix=NODEBUG %s
 # RUN: lld-link /debug:symtab /out:%t.exe /entry:main %t.obj
 # RUN: llvm-readobj --sections %t.exe | FileCheck -check-prefix=NODEBUG %s
 
 # CHECK: Name: .text
 # CHECK: Name: .reloc
+# CHECK: Name: .rmeta
 # CHECK: Name: .debug_abbrev
 # CHECK: Name: .debug_info
 # CHECK: Name: .debug_line
 # CHECK: Name: .debug_pubnames
 # CHECK: Name: .debug_pubtypes
 
 # NODEBUG: Name: .text
 # NODEBUG: Name: .reloc
+# NODEBUG: Name: .rmeta
 # NODEBUG-NOT: Name: .debug_abbrev
 # NODEBUG-NOT: Name: .debug_info
 # NODEBUG-NOT: Name: .debug_line
 # NODEBUG-NOT: Name: .debug_pubnames
 # NODEBUG-NOT: Name: .debug_pubtypes
 
 --- !COFF
 header:
sections:
   - Name:            .debug_line
     Characteristics: [ IMAGE_SCN_CNT_INITIALIZED_DATA, IMAGE_SCN_MEM_DISCARDABLE, IMAGE_SCN_MEM_READ ]
     Alignment:       1
     SectionData:     3300000002001E0000000101FB0E0D000101010100000001000001003C737464696E3E000000000000050200000000010AD60202000101
     Relocations:
       - VirtualAddress:  43
         SymbolName:      .text
         Type:            IMAGE_REL_I386_DIR32
+  - Name:            .rmeta
+    Characteristics: [ IMAGE_SCN_CNT_INITIALIZED_DATA, IMAGE_SCN_MEM_DISCARDABLE, IMAGE_SCN_MEM_READ ]
+    Alignment:       1
+    SectionData:     00112233
 symbols:
   - Name:            .text
     Value:           0
     SectionNumber:   1
     SimpleType:      IMAGE_SYM_TYPE_NULL
     ComplexType:     IMAGE_SYM_DTYPE_NULL
     StorageClass:    IMAGE_SYM_CLASS_STATIC
     SectionDefinition: