x86-64: Add ENDBR64 to the TLSDESC PLT entry

The TLSDESC entry in a lazy procedure linkage table is called indirectly
with "callq *(%rax)".  This patch adds an ENDBR64 to support indirect
branch tracking in Intel CET.  The TLSDESC PLT entry now looks like:

0xf3, 0x0f, 0x1e, 0xfa,  /* endbr64             */
0xff, 0x35, 8, 0, 0, 0,  /* pushq GOT+8(%rip)   */
0xff, 0x25, 16, 0, 0, 0  /* jmpq *GOT+TDG(%rip)  */

The BND prefix isn't needed since MPX isn't used for TLSDESC.

bfd/

	PR ld/23000
	* elf64-x86-64.c (elf_x86_64_finish_dynamic_sections): Add
	ENDBR64 to the TLSDESC PLT entry.

ld/

	PR ld/23000
	* testsuite/ld-x86-64/tlsdesc.pd: Updated.

diff --git a/bfd/ChangeLog b/bfd/ChangeLog
index 7534b192f693671f431bd170e0c578c2cbec4d02..426561a00eba41c5535ea2d8d021418297e26fcf 100644 (file)
--- a/bfd/ChangeLog
+++ b/bfd/ChangeLog
@@ -1,3 +1,9 @@
+2018-03-25  H.J. Lu  <hongjiu.lu@intel.com>
+
+       PR ld/23000
+       * elf64-x86-64.c (elf_x86_64_finish_dynamic_sections): Add
+       ENDBR64 to the TLSDESC PLT entry.
+
 2018-03-21  Alan Modra  <amodra@gmail.com>
 
        * elf64-ppc.c (ppc64_elf_get_synthetic_symtab): Trim uninteresting

diff --git a/bfd/elf64-x86-64.c b/bfd/elf64-x86-64.c
index a964316e44f778bfaea2ad6f89ca483a08833cae..fcc82363735c0125f34614ed8ca80f0e3533e52b 100644 (file)
--- a/bfd/elf64-x86-64.c
+++ b/bfd/elf64-x86-64.c
@@ -4411,15 +4411,23 @@ elf_x86_64_finish_dynamic_sections (bfd *output_bfd,
 
       if (htab->tlsdesc_plt)
        {
+         /* The TLSDESC entry in a lazy procedure linkage table.  */
+         static const bfd_byte tlsdesc_plt_entry[LAZY_PLT_ENTRY_SIZE] =
+           {
+             0xf3, 0x0f, 0x1e, 0xfa,   /* endbr64              */
+             0xff, 0x35, 8, 0, 0, 0,   /* pushq GOT+8(%rip)    */
+             0xff, 0x25, 16, 0, 0, 0   /* jmpq *GOT+TDG(%rip)  */
+           };
+
          bfd_put_64 (output_bfd, (bfd_vma) 0,
                      htab->elf.sgot->contents + htab->tlsdesc_got);
 
          memcpy (htab->elf.splt->contents + htab->tlsdesc_plt,
-                 htab->lazy_plt->plt0_entry,
-                 htab->lazy_plt->plt0_entry_size);
+                 tlsdesc_plt_entry, LAZY_PLT_ENTRY_SIZE);
 
-         /* Add offset for pushq GOT+8(%rip), since the
-            instruction uses 6 bytes subtract this value.  */
+         /* Add offset for pushq GOT+8(%rip), since ENDBR64 uses 4
+            bytes and the instruction uses 6 bytes, subtract these
+            values.  */
          bfd_put_32 (output_bfd,
                      (htab->elf.sgotplt->output_section->vma
                       + htab->elf.sgotplt->output_offset
@@ -4427,14 +4435,13 @@ elf_x86_64_finish_dynamic_sections (bfd *output_bfd,
                       - htab->elf.splt->output_section->vma
                       - htab->elf.splt->output_offset
                       - htab->tlsdesc_plt
-                      - 6),
+                      - 4 - 6),
                      (htab->elf.splt->contents
                       + htab->tlsdesc_plt
-                      + htab->lazy_plt->plt0_got1_offset));
-         /* Add offset for the PC-relative instruction accessing
-            GOT+TDG, where TDG stands for htab->tlsdesc_got,
-            subtracting the offset to the end of that
-            instruction.  */
+                      + 4 + 2));
+         /* Add offset for indirect branch via GOT+TDG, where TDG
+            stands for htab->tlsdesc_got, subtracting the offset
+            to the end of that instruction.  */
          bfd_put_32 (output_bfd,
                      (htab->elf.sgot->output_section->vma
                       + htab->elf.sgot->output_offset
@@ -4442,10 +4449,9 @@ elf_x86_64_finish_dynamic_sections (bfd *output_bfd,
                       - htab->elf.splt->output_section->vma
                       - htab->elf.splt->output_offset
                       - htab->tlsdesc_plt
-                      - htab->lazy_plt->plt0_got2_insn_end),
+                      - 4 - 6 - 6),
                      (htab->elf.splt->contents
-                      + htab->tlsdesc_plt
-                      + htab->lazy_plt->plt0_got2_offset));
+                      + htab->tlsdesc_plt + 4 + 6 + 2));
        }
     }
 

diff --git a/ld/ChangeLog b/ld/ChangeLog
index 330998cb42a9af7a26e3f95b50530c3469179095..dfdc648fc65457501d64c950934c0bc1247967f5 100644 (file)
--- a/ld/ChangeLog
+++ b/ld/ChangeLog
@@ -1,3 +1,8 @@
+2018-03-25  H.J. Lu  <hongjiu.lu@intel.com>
+
+       PR ld/23000
+       * testsuite/ld-x86-64/tlsdesc.pd: Updated.
+
 2018-03-23  Nick Clifton  <nickc@redhat.com>
 
        PR 22948

diff --git a/ld/testsuite/ld-x86-64/tlsdesc.pd b/ld/testsuite/ld-x86-64/tlsdesc.pd
index 0fa36f327390172a298267ba6b2452286c9a2d75..08b4fa3aef9f84c011a9e02c724e6f04c378683c 100644 (file)
--- a/ld/testsuite/ld-x86-64/tlsdesc.pd
+++ b/ld/testsuite/ld-x86-64/tlsdesc.pd
@@ -13,7 +13,7 @@ Disassembly of section .plt:
  [0-9a-f]+:    ff 35 .. .. 20 00       pushq  .*\(%rip\)        # 201358 <_GLOBAL_OFFSET_TABLE_\+0x8>
  [0-9a-f]+:    ff 25 .. .. 20 00       jmpq   \*.*\(%rip\)        # 201360 <_GLOBAL_OFFSET_TABLE_\+0x10>
  [0-9a-f]+:    0f 1f 40 00             nopl   0x0\(%rax\)
+ [0-9a-f]+:    f3 0f 1e fa             endbr64 
  [0-9a-f]+:    ff 35 .. .. 20 00       pushq  .*\(%rip\)        # 201358 <_GLOBAL_OFFSET_TABLE_\+0x8>
  [0-9a-f]+:    ff 25 .. .. 20 00       jmpq   \*.*\(%rip\)        # 201348 <.*>
- [0-9a-f]+:    0f 1f 40 00             nopl   0x0\(%rax\)