[lld] 2fc704a - [ELF] --emit-relocs: fix st_value of STT_SECTION in the presence of a gap before the first input section

[Fangrui Song via llvm-commits]

Author: Fangrui Song
Date: 2020-11-02T08:37:15-08:00
New Revision: 2fc704a0a529dd7eba7566a293f981a86bfa5c3e

URL: https://github.com/llvm/llvm-project/commit/2fc704a0a529dd7eba7566a293f981a86bfa5c3e
DIFF: https://github.com/llvm/llvm-project/commit/2fc704a0a529dd7eba7566a293f981a86bfa5c3e.diff

LOG: [ELF] --emit-relocs: fix st_value of STT_SECTION in the presence of a gap before the first input section

In the presence of a gap, the st_value field of a STT_SECTION symbol is the
address of the first input section (incorrect if there is a gap). Set it to the
output section address instead.

In -r mode, this bug can cause an incorrect non-zero st_value of a STT_SECTION
symbol (while output sections have zero addresses, input sections may have
non-zero outSecOff).  The non-zero st_value can cause the final link to have
incorrect relocation computation (both GNU ld and LLD add st_value of the
STT_SECTION symbol to the output section address).

Reviewed By: grimar

Differential Revision: https://reviews.llvm.org/D90520

Added: 
    lld/test/ELF/section-symbol-gap.s

Modified: 
    lld/ELF/SyntheticSections.cpp
    lld/ELF/Writer.cpp

Removed: 
    


################################################################################
diff  --git a/lld/ELF/SyntheticSections.cpp b/lld/ELF/SyntheticSections.cpp
index 0ffd6bfa81dd..eccd3ef1795e 100644
--- a/lld/ELF/SyntheticSections.cpp
+++ b/lld/ELF/SyntheticSections.cpp
@@ -2198,9 +2198,8 @@ template <class ELFT> void SymbolTableSection<ELFT>::writeTo(uint8_t *buf) {
     else
       eSym->st_size = sym->getSize();
 
-    // st_value is usually an address of a symbol, but that has a
-    // special meaning for uninstantiated common symbols (this can
-    // occur if -r is given).
+    // st_value is usually an address of a symbol, but that has a special
+    // meaning for uninstantiated common symbols (--no-define-common).
     if (BssSection *commonSec = getCommonSec(ent.sym))
       eSym->st_value = commonSec->alignment;
     else if (isDefinedHere)

diff  --git a/lld/ELF/Writer.cpp b/lld/ELF/Writer.cpp
index 78d6cd018d76..a02b08204d80 100644
--- a/lld/ELF/Writer.cpp
+++ b/lld/ELF/Writer.cpp
@@ -811,9 +811,12 @@ template <class ELFT> void Writer<ELFT>::addSectionSymbols() {
     if (isa<SyntheticSection>(isec) && !(isec->flags & SHF_MERGE))
       continue;
 
+    // Set the symbol to be relative to the output section so that its st_value
+    // equals the output section address. Note, there may be a gap between the
+    // start of the output section and isec.
     auto *sym =
         make<Defined>(isec->file, "", STB_LOCAL, /*stOther=*/0, STT_SECTION,
-                      /*value=*/0, /*size=*/0, isec);
+                      /*value=*/0, /*size=*/0, isec->getOutputSection());
     in.symTab->addSymbol(sym);
   }
 }

diff  --git a/lld/test/ELF/section-symbol-gap.s b/lld/test/ELF/section-symbol-gap.s
new file mode 100644
index 000000000000..75a207d6d6f7
--- /dev/null
+++ b/lld/test/ELF/section-symbol-gap.s
@@ -0,0 +1,51 @@
+# REQUIRES: x86
+## Test st_value of the STT_SECTION symbol equals the output section address,
+## instead of the first input section address.
+
+# RUN: split-file %s %t
+# RUN: llvm-mc -filetype=obj -triple=x86_64 %t/asm -o %t.o
+
+# RUN: ld.lld --emit-relocs -T %t/lds %t.o -o %t.out
+# RUN: llvm-readelf -S -r -s %t.out | FileCheck %s --check-prefix=EXE
+
+## In -r mode, section addresses are zeros, hence the st_value fields of
+## STT_SECTION are zeros.
+# RUN: ld.lld -r -T %t/lds %t.o -o %t.ro
+# RUN: llvm-readelf -S -r -s %t.ro | FileCheck %s --check-prefix=RO
+
+# EXE:      [Nr] Name  Type      Address
+# EXE-NEXT: [ 0]
+# EXE-NEXT: [ 1] .text PROGBITS  0000000000000000
+# EXE-NEXT: [ 2] .bss  NOBITS    000000000000000a
+
+# EXE:      R_X86_64_64 {{.*}} .bss + 1
+
+# EXE:      Symbol table '.symtab' contains 4 entries:
+# EXE-NEXT: Num:    Value          Size Type    Bind   Vis       Ndx Name
+# EXE-NEXT:   0: 0000000000000000     0 NOTYPE  LOCAL  DEFAULT   UND
+# EXE-NEXT:   1: 000000000000000a     0 SECTION LOCAL  DEFAULT     2 .bss
+# EXE-NEXT:   2: 0000000000000000     0 SECTION LOCAL  DEFAULT     1 .text
+# EXE-NEXT:   3: 0000000000000000     0 SECTION LOCAL  DEFAULT     4 .comment
+
+# RO:       [Nr] Name  Type      Address
+# RO-NEXT:  [ 0]
+# RO-NEXT:  [ 1] .bss  NOBITS    0000000000000000
+
+# RO:       R_X86_64_64 {{.*}} .bss + 1
+
+# RO:      Symbol table '.symtab' contains 3 entries:
+# RO-NEXT: Num:    Value          Size Type    Bind   Vis       Ndx Name
+# RO-NEXT:   0: 0000000000000000     0 NOTYPE  LOCAL  DEFAULT   UND
+# RO-NEXT:   1: 0000000000000000     0 SECTION LOCAL  DEFAULT     1 .bss
+# RO-NEXT:   2: 0000000000000000     0 SECTION LOCAL  DEFAULT     2 .text
+
+#--- asm
+movabsq .bss, %rax
+
+.bss
+.byte 0
+
+#--- lds
+SECTIONS {
+  .bss : { BYTE(0) *(.bss) }
+}