[LLD][ELF] - Do not forget to use ch_addralign field after decompressing the sections.

[grimar]

Summary

LLD did not use ELF::Chdr::ch_addralign for decompressed sections.
This resulted in a broken output.

Fixes https://bugs.llvm.org/show_bug.cgi?id=40482.

[ruiu]

ELF/InputSection.cpp

256

What is the definition of ch_addralign if it's 0? I wonder if we have to handle 0 as 1.

[grimar]

ELF/InputSection.cpp

256

ch_addralign description says:
(https://docs.oracle.com/cd/E53394_01/html/E54813/section_compression.html)
Required alignment for the uncompressed data. See sh_addralign.

And sh_addralign description says
(http://www.sco.com/developers/gabi/latest/ch4.sheader.html):
Some sections have address alignment constraints. For example, if a section holds a doubleword, the system must ensure doubleword alignment for the entire section. The value of sh_addr must be congruent to 0, modulo the value of sh_addralign. Currently, only 0 and positive integral powers of two are allowed. Values 0 and 1 mean the section has no alignment constraints.

So looks in according to the standard we want to handle 0 as 1.

[grimar]

• Add support for ch_addralign == 0 and a test case.

[ruiu]

LGTM

ELF/InputSection.cpp

if (Config->Is64) {
 
     auto *Hdr = reinterpret_cast<const Chdr64 *>(RawData.data());
     if (Hdr->ch_type != ELFCOMPRESS_ZLIB) {
       error(toString(this) + ": unsupported compression type");
       return;
     }
 
     UncompressedSize = Hdr->ch_size;
+    Alignment = Hdr->ch_addralign;
     RawData = RawData.slice(sizeof(*Hdr));
     return;
   }
 
   // New-style 32-bit header
   if (RawData.size() < sizeof(Chdr32)) {
     error(toString(this) + ": corrupted compressed section");
     return;
   }
 
   auto *Hdr = reinterpret_cast<const Chdr32 *>(RawData.data());
   if (Hdr->ch_type != ELFCOMPRESS_ZLIB) {
     error(toString(this) + ": unsupported compression type");
     return;
   }
 
   UncompressedSize = Hdr->ch_size;
+  Alignment = Hdr->ch_addralign;
   RawData = RawData.slice(sizeof(*Hdr));
 }
 
 InputSection *InputSectionBase::getLinkOrderDep() const {
   assert(Link);
   assert(Flags & SHF_LINK_ORDER);
   return cast<InputSection>(File->getSections()[Link]);
 }

test/ELF/compressed-input-alignment.test

+# RUN: yaml2obj %s -o %t.o
+# RUN: ld.lld %t.o %t.o -o %t2
+# RUN: llvm-readobj -sections -section-data %t2 | FileCheck %s
+
+# CHECK:        Name: .debug_info
+# CHECK-NEXT:   Type: SHT_PROGBITS
+# CHECK-NEXT:   Flags [
+# CHECK-NEXT:   ]
+# CHECK-NEXT:   Address: 0x0
+# CHECK-NEXT:   Offset: 0xE8
+# CHECK-NEXT:   Size: 108
+# CHECK-NEXT:   Link: 0
+# CHECK-NEXT:   Info: 0
+# CHECK-NEXT:   AddressAlignment: 1
+# CHECK-NEXT:   EntrySize: 0
+# CHECK-NEXT:   SectionData (
+# CHECK-NEXT:     0000: {{.*}} |ABCDEFGHIJKLMNOP|
+# CHECK-NEXT:     0010: {{.*}} |QRSTUVWXYZ.ABCDE|
+# CHECK-NEXT:     0020: {{.*}} |FGHIJKLMNOPQRSTU|
+# CHECK-NEXT:     0030: {{.*}} |VWXYZ.ABCDEFGHIJ|
+# CHECK-NEXT:     0040: {{.*}} |KLMNOPQRSTUVWXYZ|
+# CHECK-NEXT:     0050: {{.*}} |.ABCDEFGHIJKLMNO|
+# CHECK-NEXT:     0060: {{.*}} |PQRSTUVWXYZ.|
+# CHECK-NEXT:   )
+# CHECK-NEXT: }
+
+## YAML below is produced from the following code. AddressAlign of .debug_info is 8,
+## while compressed header has ch_addralign = 1. LLD had a bug and did not use the
+## value of ch_addralign at all.
+## TODO: This test can avoid using YAML, but at the moment of 22 april 2019 llvm-mc
+## is broken and its sets address align to 1 for all GNU compressed sections.
+##
+## .section .debug_info,"",@progbits
+##  .string "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
+##  .string "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
+
+--- !ELF
+FileHeader:      
+  Class:           ELFCLASS64
+  Data:            ELFDATA2LSB
+  Type:            ET_REL
+  Machine:         EM_X86_64
+Sections:
+  - Name:            .debug_info
+    Type:            SHT_PROGBITS
+    Flags:           [ SHF_COMPRESSED ]
+    AddressAlign:    0x0000000000000008
+    Content:         010000000000000036000000000000000100000000000000789C73747276717573F7F0F4F2F6F1F5F30F080C0A0E090D0B8F888C6270C42D0500ADA00FBF