x86-64: Fix TLSDESC relaxation for x32

For x32, we must encode "lea x@TLSDESC(%rip), %reg" with a REX prefix
even if it isn't required.  Otherwise linker can’t safely perform
GDesc -> IE/LE optimization.  X32 TLSDESC sequences can be:

40 8d 05 00 00 00 00	rex lea	x@TLSDESC(%rip), %reg
...
67 ff 10		call	*x@TLSCALL(%eax)

or the same sequence as LP64:

48 8d 05 00 00 00 00	lea	foo@TLSDESC(%rip), %reg
...
ff 10			call	*foo@TLSCALL(%rax)

We need to support both sequences for x32.  For both GDesc -> IE/LE
transitions,

67 ff 10		call	*x@TLSCALL(%eax)

should relaxed to

0f 1f 00		nopl	(%rax)

For GDesc -> LE transition,

40 8d 05 00 00 00 00	rex lea	x@TLSDESC(%rip), %reg

should relaxed to

40 c7 c0 fc ff ff ff	rex movl $x@tpoff, %reg

For GDesc -> IE transition,

40 8d 05 00 00 00 00	rex lea	x@TLSDESC(%rip), %reg

should relaxed to

40 8b 05 00 00 00 00	rex movl x@gottpoff(%rip), %eax

bfd/

	PR ld/25416
	* elf64-x86-64.c (elf_x86_64_check_tls_transition): Support
	"rex leal x@tlsdesc(%rip), %reg" and "call *x@tlsdesc(%eax)" in
	X32 mode.
	(elf_x86_64_relocate_section): In x32 mode, for GDesc -> LE
	transition, relax "rex leal x@tlsdesc(%rip), %reg" to
	"rex movl $x@tpoff, %reg", for GDesc -> IE transition, relax
	"rex leal x@tlsdesc(%rip), %reg" to
	"rex movl x@gottpoff(%rip), %eax".  For both transitions, relax
	"call *(%eax)" to "nopl (%rax)".

gas/

	PR ld/25416
	* config/tc-i386.c (output_insn): Add a dummy REX_OPCODE prefix
	for lea with R_X86_64_GOTPC32_TLSDESC relocation when generating
	x32 object.
	* testsuite/gas/i386/ilp32/x32-tls.d: Updated.
	* testsuite/gas/i386/ilp32/x32-tls.s: Add tests for lea with
	R_X86_64_GOTPC32_TLSDESC relocation.

ld/

	PR ld/25416
	* testsuite/ld-x86-64/pr25416-1.s: New file
	* testsuite/ld-x86-64/pr25416-1a.d: Likewise.
	* testsuite/ld-x86-64/pr25416-1b.d: Likewise.
	* testsuite/ld-x86-64/pr25416-1.s: Likewise.
	* testsuite/ld-x86-64/pr25416-2.s: Likewise.
	* testsuite/ld-x86-64/pr25416-2a.d: Likewise.
	* testsuite/ld-x86-64/pr25416-2b.d: Likewise.
	* testsuite/ld-x86-64/pr25416-3.d: Likewise.
	* testsuite/ld-x86-64/pr25416-3.s: Likewise.
	* testsuite/ld-x86-64/pr25416-4.d: Likewise.
	* testsuite/ld-x86-64/pr25416-4.s: Likewise.
	* testsuite/ld-x86-64/pr25416-5a.c: Likewise.
	* testsuite/ld-x86-64/pr25416-5b.s: Likewise.
	* testsuite/ld-x86-64/pr25416-5c.s: Likewise.
	* testsuite/ld-x86-64/pr25416-5d.s: Likewise.
	* testsuite/ld-x86-64/pr25416-5e.s: Likewise.
	* testsuite/ld-x86-64/x86-64.exp: Run PR ld/25416 tests.

bfd/ChangeLog

@@ -1,3 +1,16 @@
+2020-01-20  H.J. Lu  <hongjiu.lu@intel.com>
+
+	PR ld/25416
+	* elf64-x86-64.c (elf_x86_64_check_tls_transition): Support
+	"rex leal x@tlsdesc(%rip), %reg" and "call *x@tlsdesc(%eax)" in
+	X32 mode.
+	(elf_x86_64_relocate_section): In x32 mode, for GDesc -> LE
+	transition, relax "rex leal x@tlsdesc(%rip), %reg" to
+	"rex movl $x@tpoff, %reg", for GDesc -> IE transition, relax
+	"rex leal x@tlsdesc(%rip), %reg" to
+	"rex movl x@gottpoff(%rip), %eax".  For both transitions, relax
+	"call *(%eax)" to "nopl (%rax)".
+
 2020-01-20  Alan Modra  <amodra@gmail.com>
 
 	* elf64-ppc.c (LD_R11_0R3, CMPDI_R11_0, STD_R11_0R1, LD_R11_0R1),

bfd/elf64-x86-64.c

@@ -1223,7 +1223,8 @@ elf_x86_64_check_tls_transition (bfd *abfd,
 
     case R_X86_64_GOTPC32_TLSDESC:
       /* Check transition from GDesc access model:
-		leaq x@tlsdesc(%rip), %rax
+		leaq x@tlsdesc(%rip), %rax <--- LP64 mode.
+		rex leal x@tlsdesc(%rip), %eax <--- X32 mode.
 
 	 Make sure it's a leaq adding rip to a 32-bit offset
 	 into any register, although it's probably almost always
@@ -1233,7 +1234,8 @@ elf_x86_64_check_tls_transition (bfd *abfd,
 	return FALSE;
 
       val = bfd_get_8 (abfd, contents + offset - 3);
-      if ((val & 0xfb) != 0x48)
+      val &= 0xfb;
+      if (val != 0x48 && (ABI_64_P (abfd) || val != 0x40))
 	return FALSE;
 
       if (bfd_get_8 (abfd, contents + offset - 2) != 0x8d)
@@ -1244,13 +1246,26 @@ elf_x86_64_check_tls_transition (bfd *abfd,
 
     case R_X86_64_TLSDESC_CALL:
       /* Check transition from GDesc access model:
-		call *x@tlsdesc(%rax)
+		call *x@tlsdesc(%rax) <--- LP64 mode.
+		call *x@tlsdesc(%eax) <--- X32 mode.
        */
       if (offset + 2 <= sec->size)
 	{
-	  /* Make sure that it's a call *x@tlsdesc(%rax).  */
+	  unsigned int prefix;
 	  call = contents + offset;
-	  return call[0] == 0xff && call[1] == 0x10;
+	  prefix = 0;
+	  if (!ABI_64_P (abfd))
+	    {
+	      /* Check for call *x@tlsdesc(%eax).  */
+	      if (call[0] == 0x67)
+		{
+		  prefix = 1;
+		  if (offset + 3 > sec->size)
+		    return FALSE;
+		}
+	    }
+	  /* Make sure that it's a call *x@tlsdesc(%rax).  */
+	  return call[prefix] == 0xff && call[1 + prefix] == 0x10;
 	}
 
       return FALSE;
@@ -3401,10 +3416,13 @@ corrupt_input:
 		{
 		  /* GDesc -> LE transition.
 		     It's originally something like:
-		     leaq x@tlsdesc(%rip), %rax
+		     leaq x@tlsdesc(%rip), %rax <--- LP64 mode.
+		     rex leal x@tlsdesc(%rip), %eax <--- X32 mode.
 
 		     Change it to:
-		     movl $x@tpoff, %rax.  */
+		     movq $x@tpoff, %rax <--- LP64 mode.
+		     rex movl $x@tpoff, %eax <--- X32 mode.
+		   */
 
 		  unsigned int val, type;
 
@@ -3412,7 +3430,8 @@ corrupt_input:
 		    goto corrupt_input;
 		  type = bfd_get_8 (input_bfd, contents + roff - 3);
 		  val = bfd_get_8 (input_bfd, contents + roff - 1);
-		  bfd_put_8 (output_bfd, 0x48 | ((type >> 2) & 1),
+		  bfd_put_8 (output_bfd,
+			     (type & 0x48) | ((type >> 2) & 1),
 			     contents + roff - 3);
 		  bfd_put_8 (output_bfd, 0xc7, contents + roff - 2);
 		  bfd_put_8 (output_bfd, 0xc0 | ((val >> 3) & 7),
@@ -3426,11 +3445,30 @@ corrupt_input:
 		{
 		  /* GDesc -> LE transition.
 		     It's originally:
-		     call *(%rax)
+		     call *(%rax) <--- LP64 mode.
+		     call *(%eax) <--- X32 mode.
 		     Turn it into:
-		     xchg %ax,%ax.  */
+		     xchg %ax,%ax <-- LP64 mode.
+		     nopl (%rax)  <-- X32 mode.
+		   */
+		  unsigned int prefix = 0;
+		  if (!ABI_64_P (input_bfd))
+		    {
+		      /* Check for call *x@tlsdesc(%eax).  */
+		      if (contents[roff] == 0x67)
+			prefix = 1;
+		    }
+		  if (prefix)
+		    {
+		      bfd_put_8 (output_bfd, 0x0f, contents + roff);
+		      bfd_put_8 (output_bfd, 0x1f, contents + roff + 1);
+		      bfd_put_8 (output_bfd, 0x00, contents + roff + 2);
+		    }
+		  else
+		    {
 		      bfd_put_8 (output_bfd, 0x66, contents + roff);
 		      bfd_put_8 (output_bfd, 0x90, contents + roff + 1);
+		    }
 		  continue;
 		}
 	      else if (r_type == R_X86_64_GOTTPOFF)
@@ -3741,13 +3779,18 @@ corrupt_input:
 		{
 		  /* GDesc -> IE transition.
 		     It's originally something like:
-		     leaq x@tlsdesc(%rip), %rax
+		     leaq x@tlsdesc(%rip), %rax <--- LP64 mode.
+		     rex leal x@tlsdesc(%rip), %eax <--- X32 mode.
 
 		     Change it to:
-		     movq x@gottpoff(%rip), %rax # before xchg %ax,%ax.  */
+		     # before xchg %ax,%ax in LP64 mode.
+		     movq x@gottpoff(%rip), %rax
+		     # before nopl (%rax) in X32 mode.
+		     rex movl x@gottpoff(%rip), %eax
+		  */
 
 		  /* Now modify the instruction as appropriate. To
-		     turn a leaq into a movq in the form we use it, it
+		     turn a lea into a mov in the form we use it, it
 		     suffices to change the second byte from 0x8d to
 		     0x8b.  */
 		  if (roff < 2)
@@ -3768,13 +3811,32 @@ corrupt_input:
 		{
 		  /* GDesc -> IE transition.
 		     It's originally:
-		     call *(%rax)
+		     call *(%rax) <--- LP64 mode.
+		     call *(%eax) <--- X32 mode.
 
 		     Change it to:
-		     xchg %ax, %ax.  */
+		     xchg %ax, %ax <-- LP64 mode.
+		     nopl (%rax)  <-- X32 mode.
+		   */
 
+		  unsigned int prefix = 0;
+		  if (!ABI_64_P (input_bfd))
+		    {
+		      /* Check for call *x@tlsdesc(%eax).  */
+		      if (contents[roff] == 0x67)
+			prefix = 1;
+		    }
+		  if (prefix)
+		    {
+		      bfd_put_8 (output_bfd, 0x0f, contents + roff);
+		      bfd_put_8 (output_bfd, 0x1f, contents + roff + 1);
+		      bfd_put_8 (output_bfd, 0x00, contents + roff + 2);
+		    }
+		  else
+		    {
 		      bfd_put_8 (output_bfd, 0x66, contents + roff);
 		      bfd_put_8 (output_bfd, 0x90, contents + roff + 1);
+		    }
 		  continue;
 		}
 	      else

gas/ChangeLog

@@ -1,3 +1,13 @@
+2020-01-20  H.J. Lu  <hongjiu.lu@intel.com>
+
+	PR ld/25416
+	* config/tc-i386.c (output_insn): Add a dummy REX_OPCODE prefix
+	for lea with R_X86_64_GOTPC32_TLSDESC relocation when generating
+	x32 object.
+	* testsuite/gas/i386/ilp32/x32-tls.d: Updated.
+	* testsuite/gas/i386/ilp32/x32-tls.s: Add tests for lea with
+	R_X86_64_GOTPC32_TLSDESC relocation.
+
 2020-01-18  Nick Clifton  <nickc@redhat.com>
 
 	* configure: Regenerate.

gas/config/tc-i386.c

@@ -8713,10 +8713,13 @@ output_insn (void)
 #if defined (OBJ_MAYBE_ELF) || defined (OBJ_ELF)
 	  /* For x32, add a dummy REX_OPCODE prefix for mov/add with
 	     R_X86_64_GOTTPOFF relocation so that linker can safely
-	     perform IE->LE optimization.  */
+	     perform IE->LE optimization.  A dummy REX_OPCODE prefix
+	     is also needed for lea with R_X86_64_GOTPC32_TLSDESC
+	     relocation for GDesc -> IE/LE optimization.  */
 	  if (x86_elf_abi == X86_64_X32_ABI
 	      && i.operands == 2
-	      && i.reloc[0] == BFD_RELOC_X86_64_GOTTPOFF
+	      && (i.reloc[0] == BFD_RELOC_X86_64_GOTTPOFF
+		  || i.reloc[0] == BFD_RELOC_X86_64_GOTPC32_TLSDESC)
 	      && i.prefix[REX_PREFIX] == 0)
 	    add_prefix (REX_OPCODE);
 #endif

gas/testsuite/gas/i386/ilp32/x32-tls.d

@@ -10,4 +10,6 @@ Disassembly of section .text:
 [ 	]*[a-f0-9]+:	4c 8b 25 00 00 00 00 	mov    0x0\(%rip\),%r12        # e <_start\+0xe>
 [ 	]*[a-f0-9]+:	40 03 05 00 00 00 00 	rex add 0x0\(%rip\),%eax        # 15 <_start\+0x15>
 [ 	]*[a-f0-9]+:	44 03 25 00 00 00 00 	add    0x0\(%rip\),%r12d        # 1c <_start\+0x1c>
+[ 	]*[a-f0-9]+:	40 8d 05 00 00 00 00 	rex lea 0x0\(%rip\),%eax        # 23 <_start\+0x23>
+[ 	]*[a-f0-9]+:	44 8d 25 00 00 00 00 	lea    0x0\(%rip\),%r12d        # 2a <_start\+0x2a>
 #pass

gas/testsuite/gas/i386/ilp32/x32-tls.s

@@ -4,6 +4,8 @@ _start:
 	mov	foo@gottpoff(%rip), %r12
 	add	foo@gottpoff(%rip), %eax
 	add	foo@gottpoff(%rip), %r12d
+	lea	foo@tlsdesc(%rip), %eax
+	lea	foo@tlsdesc(%rip), %r12d
 	.globl	foo
 	.section	.tdata,"awT",@progbits
 	.align 4

ld/ChangeLog

@@ -1,3 +1,24 @@
+2020-01-20  H.J. Lu  <hongjiu.lu@intel.com>
+
+	PR ld/25416
+	* testsuite/ld-x86-64/pr25416-1.s: New file
+	* testsuite/ld-x86-64/pr25416-1a.d: Likewise.
+	* testsuite/ld-x86-64/pr25416-1b.d: Likewise.
+	* testsuite/ld-x86-64/pr25416-1.s: Likewise.
+	* testsuite/ld-x86-64/pr25416-2.s: Likewise.
+	* testsuite/ld-x86-64/pr25416-2a.d: Likewise.
+	* testsuite/ld-x86-64/pr25416-2b.d: Likewise.
+	* testsuite/ld-x86-64/pr25416-3.d: Likewise.
+	* testsuite/ld-x86-64/pr25416-3.s: Likewise.
+	* testsuite/ld-x86-64/pr25416-4.d: Likewise.
+	* testsuite/ld-x86-64/pr25416-4.s: Likewise.
+	* testsuite/ld-x86-64/pr25416-5a.c: Likewise.
+	* testsuite/ld-x86-64/pr25416-5b.s: Likewise.
+	* testsuite/ld-x86-64/pr25416-5c.s: Likewise.
+	* testsuite/ld-x86-64/pr25416-5d.s: Likewise.
+	* testsuite/ld-x86-64/pr25416-5e.s: Likewise.
+	* testsuite/ld-x86-64/x86-64.exp: Run PR ld/25416 tests.
+
 2020-01-20  Alan Modra  <amodra@gmail.com>
 
 	* testsuite/ld-powerpc/tlsexe.d: Match new __tls_get_addr stub.

ld/testsuite/ld-x86-64/pr25416-1.s

@@ -0,0 +1,13 @@
+	.text
+	.globl	_start
+	.type	_start, @function
+_start:
+	lea	foo@TLSDESC(%rip), %eax
+	call	*foo@TLSCALL(%eax)
+	.section	.tdata,"awT",@progbits
+	.align 4
+	.type	foo, @object
+	.size	foo, 4
+foo:
+	.long	30
+	.section	.note.GNU-stack,"",@progbits

ld/testsuite/ld-x86-64/pr25416-1a.d

@@ -0,0 +1,16 @@
+#name: X32 GDesc 1
+#source: pr25416-1.s
+#as: --x32
+#ld: -melf32_x86_64 -shared
+#objdump: -dw
+
+.*: +file format .*
+
+
+#...
+Disassembly of section .text:
+
+[a-f0-9]+ <_start>:
+ +[a-f0-9]+:	40 8d 05 ([0-9a-f]{2} ){4}[ \t]+rex lea 0x[a-f0-9]+\(%rip\),%eax[ \t]+# [a-f0-9]+ <_GLOBAL_OFFSET_TABLE_\+0x[a-f0-9]+>
+ +[a-f0-9]+:	67 ff 10             	callq  \*\(%eax\)
+#pass

ld/testsuite/ld-x86-64/pr25416-1b.d

@@ -0,0 +1,15 @@
+#name: X32 GDesc -> LE 1
+#source: pr25416-1.s
+#as: --x32
+#ld: -melf32_x86_64
+#objdump: -dw
+
+.*: +file format .*
+
+
+Disassembly of section .text:
+
+[a-f0-9]+ <_start>:
+ +[a-f0-9]+:	40 c7 c0 ([0-9a-f]{2} ){4}[ \t]+rex mov \$0x[a-f0-9]+,%eax
+ +[a-f0-9]+:	0f 1f 00             	nopl   \(%rax\)
+#pass

ld/testsuite/ld-x86-64/pr25416-2.s

@@ -0,0 +1,13 @@
+	.text
+	.globl	_start
+	.type	_start, @function
+_start:
+	lea	foo@TLSDESC(%rip), %rax
+	call	*foo@TLSCALL(%rax)
+	.section	.tdata,"awT",@progbits
+	.align 4
+	.type	foo, @object
+	.size	foo, 4
+foo:
+	.long	30
+	.section	.note.GNU-stack,"",@progbits

ld/testsuite/ld-x86-64/pr25416-2a.d

@@ -0,0 +1,16 @@
+#name: X32 GDesc 2
+#source: pr25416-2.s
+#as: --x32
+#ld: -melf32_x86_64 -shared
+#objdump: -dw
+
+.*: +file format .*
+
+
+#...
+Disassembly of section .text:
+
+[a-f0-9]+ <_start>:
+ +[a-f0-9]+:	48 8d 05 ([0-9a-f]{2} ){4}[ \t]+lea    0x[a-f0-9]+\(%rip\),%rax[ \t]+# [a-f0-9]+ <_GLOBAL_OFFSET_TABLE_\+0x[a-f0-9]+>
+ +[a-f0-9]+:	ff 10                	callq  \*\(%rax\)
+#pass

ld/testsuite/ld-x86-64/pr25416-2b.d

@@ -0,0 +1,15 @@
+#name: X32 GDesc -> LE 2
+#source: pr25416-2.s
+#as: --x32
+#ld: -melf32_x86_64
+#objdump: -dw
+
+.*: +file format .*
+
+
+Disassembly of section .text:
+
+[a-f0-9]+ <_start>:
+ +[a-f0-9]+:	48 c7 c0 ([0-9a-f]{2} ){4}[ \t]+mov    \$0x[a-f0-9]+,%rax
+ +[a-f0-9]+:	66 90                	xchg   %ax,%ax
+#pass

ld/testsuite/ld-x86-64/pr25416-3.d

@@ -0,0 +1,16 @@
+#name: X32 GDesc -> IE 1
+#as: --x32
+#ld: -melf32_x86_64 -shared
+#objdump: -dw
+
+.*: +file format .*
+
+
+Disassembly of section .text:
+
+[a-f0-9]+ <_start>:
+ +[a-f0-9]+:	40 8b 05 ([0-9a-f]{2} ){4}[ \t]+rex mov 0x[a-f0-9]+\(%rip\),%eax[ \t]+# [a-f0-9]+ <.got>
+ +[a-f0-9]+:	0f 1f 00             	nopl   \(%rax\)
+ +[a-f0-9]+:	64 8b 0c 25 00 00 00 00 	mov    %fs:0x0,%ecx
+ +[a-f0-9]+:	40 03 0d ([0-9a-f]{2} ){4}[ \t]+rex add 0x[a-f0-9]+\(%rip\),%ecx[ \t]+# [a-f0-9]+ <.got>
+#pass

ld/testsuite/ld-x86-64/pr25416-3.s

@@ -0,0 +1,15 @@
+	.text
+	.globl	_start
+	.type	_start, @function
+_start:
+	lea	foo@TLSDESC(%rip), %eax
+	call	*foo@TLSCALL(%eax)
+	movl	%fs:0, %ecx
+	addl	foo@gottpoff(%rip), %ecx
+	.section	.tdata,"awT",@progbits
+	.align 4
+	.type	foo, @object
+	.size	foo, 4
+foo:
+	.long	30
+	.section	.note.GNU-stack,"",@progbits

ld/testsuite/ld-x86-64/pr25416-4.d

@@ -0,0 +1,16 @@
+#name: X32 GDesc -> IE 2
+#as: --x32
+#ld: -melf32_x86_64 -shared
+#objdump: -dw
+
+.*: +file format .*
+
+
+Disassembly of section .text:
+
+[a-f0-9]+ <_start>:
+ +[a-f0-9]+:	48 8b 05 ([0-9a-f]{2} ){4}[ \t]+mov    0x[a-f0-9]+\(%rip\),%rax[ \t]+# [a-f0-9]+ <.got>
+ +[a-f0-9]+:	66 90                	xchg   %ax,%ax
+ +[a-f0-9]+:	64 8b 0c 25 00 00 00 00 	mov    %fs:0x0,%ecx
+ +[a-f0-9]+:	40 03 0d ([0-9a-f]{2} ){4}[ \t]+rex add 0x[a-f0-9]+\(%rip\),%ecx[ \t]+# [a-f0-9]+ <.got>
+#pass

ld/testsuite/ld-x86-64/pr25416-4.s

@@ -0,0 +1,15 @@
+	.text
+	.globl	_start
+	.type	_start, @function
+_start:
+	lea	foo@TLSDESC(%rip), %rax
+	call	*foo@TLSCALL(%rax)
+	movl	%fs:0, %ecx
+	addl	foo@gottpoff(%rip), %ecx
+	.section	.tdata,"awT",@progbits
+	.align 4
+	.type	foo, @object
+	.size	foo, 4
+foo:
+	.long	30
+	.section	.note.GNU-stack,"",@progbits

ld/testsuite/ld-x86-64/pr25416-5a.c

@@ -0,0 +1,33 @@
+#include <stdlib.h>
+#include <stdio.h>
+
+__thread int bar = 301;
+
+extern int *test1 (int);
+extern int *test2 (int);
+extern int *test3 (int);
+
+int
+main ()
+{
+  int *p;
+  p = test1 (30);
+  if (*p != 30)
+    abort ();
+  *p = 40;
+  test1 (40);
+  p = test2 (301);
+  if (*p != 301)
+    abort ();
+  if (p != &bar)
+    abort ();
+  *p = 40;
+  test2 (40);
+  p = test3 (40);
+  if (*p != 40)
+    abort ();
+  *p = 50;
+  test3 (50);
+  puts ("PASS");
+  return 0;
+}

ld/testsuite/ld-x86-64/pr25416-5b.s

@@ -0,0 +1,52 @@
+	.text
+	.p2align 4
+	.globl	test1
+	.type	test1, @function
+test1:
+	.cfi_startproc
+	subl	$8, %esp
+	.cfi_def_cfa_offset 16
+	lea	foo@TLSDESC(%rip), %eax
+	call	*foo@TLSCALL(%eax)
+	addl	%fs:0, %eax
+	cmpl	%edi, (%eax)
+	jne	.L5
+	addl	$8, %esp
+	.cfi_remember_state
+	.cfi_def_cfa_offset 8
+	ret
+.L5:
+	.cfi_restore_state
+	call	abort@PLT
+	.cfi_endproc
+	.size	test1, .-test1
+	.p2align 4
+	.globl	test2
+	.type	test2, @function
+test2:
+	.cfi_startproc
+	subl	$8, %esp
+	.cfi_def_cfa_offset 16
+	lea	bar@TLSDESC(%rip), %eax
+	call	*bar@TLSCALL(%eax)
+	addl	%fs:0, %eax
+	cmpl	%edi, (%eax)
+	jne	.L9
+	addl	$8, %esp
+	.cfi_remember_state
+	.cfi_def_cfa_offset 8
+	ret
+.L9:
+	.cfi_restore_state
+	call	abort@PLT
+	.cfi_endproc
+	.size	test2, .-test2
+	.section	.tdata,"awT",@progbits
+	.align 4
+	.hidden foo
+	.globl foo
+	.type	foo, @object
+	.size	foo, 4
+foo:
+	.long	30
+	.section	.note.GNU-stack,"",@progbits

ld/testsuite/ld-x86-64/pr25416-5c.s

@@ -0,0 +1,56 @@
+	.text
+	.p2align 4
+	.globl	test1
+	.type	test1, @function
+test1:
+	.cfi_startproc
+	subq	$8, %rsp
+	.cfi_def_cfa_offset 16
+	lea	foo@TLSDESC(%rip), %rax
+	call	*foo@TLSCALL(%rax)
+	addl	%fs:0, %eax
+	cmpl	%edi, (%eax)
+	jne	.L5
+	movl	%eax, %r8d
+	addq	$8, %rsp
+	.cfi_remember_state
+	.cfi_def_cfa_offset 8
+	movq	%r8, %rax
+	ret
+.L5:
+	.cfi_restore_state
+	call	abort@PLT
+	.cfi_endproc
+	.size	test1, .-test1
+	.p2align 4
+	.globl	test2
+	.type	test2, @function
+test2:
+	.cfi_startproc
+	subq	$8, %rsp
+	.cfi_def_cfa_offset 16
+	lea	bar@TLSDESC(%rip), %rax
+	call	*bar@TLSCALL(%rax)
+	addl	%fs:0, %eax
+	cmpl	%edi, (%eax)
+	jne	.L9
+	movl	%eax, %r8d
+	addq	$8, %rsp
+	.cfi_remember_state
+	.cfi_def_cfa_offset 8
+	movq	%r8, %rax
+	ret
+.L9:
+	.cfi_restore_state
+	call	abort@PLT
+	.cfi_endproc
+	.size	test2, .-test2
+	.section	.tdata,"awT",@progbits
+	.align 4
+	.hidden foo
+	.globl foo
+	.type	foo, @object
+	.size	foo, 4
+foo:
+	.long	30
+	.section	.note.GNU-stack,"",@progbits

ld/testsuite/ld-x86-64/pr25416-5d.s

@@ -0,0 +1,19 @@
+	.text
+	.p2align 4
+	.globl	test3
+	.type	test3, @function
+test3:
+	.cfi_startproc
+	movl	%fs:0, %eax
+	addq	foo@gottpoff(%rip), %rax
+	cmpl	%edi, (%eax)
+	jne	.L7
+	movl	%eax, %eax
+	ret
+.L7:
+	pushq	%rax
+	.cfi_def_cfa_offset 16
+	call	abort@PLT
+	.cfi_endproc
+	.size	test3, .-test3
+	.section	.note.GNU-stack,"",@progbits

ld/testsuite/ld-x86-64/pr25416-5e.s

@@ -0,0 +1,23 @@
+	.text
+	.p2align 4
+	.globl	test3
+	.type	test3, @function
+test3:
+	.cfi_startproc
+	subl	$8, %esp
+	.cfi_def_cfa_offset 16
+	lea	foo@TLSDESC(%rip), %eax
+	call	*foo@TLSCALL(%eax)
+	addl	%fs:0, %eax
+	cmpl	%edi, (%eax)
+	jne	.L5
+	addl	$8, %esp
+	.cfi_remember_state
+	.cfi_def_cfa_offset 8
+	ret
+.L5:
+	.cfi_restore_state
+	call	abort@PLT
+	.cfi_endproc
+	.size	test3, .-test3
+	.section	.note.GNU-stack,"",@progbits

ld/testsuite/ld-x86-64/x86-64.exp

@@ -461,6 +461,12 @@ run_dump_test "pr24721-x32"
 run_dump_test "pr24905"
 run_dump_test "pr24905-x32"
 run_dump_test "align-branch-1"
+run_dump_test "pr25416-1a"
+run_dump_test "pr25416-2b"
+run_dump_test "pr25416-2a"
+run_dump_test "pr25416-2b"
+run_dump_test "pr25416-3"
+run_dump_test "pr25416-4"
 
 if { ![istarget "x86_64-*-linux*"] && ![istarget "x86_64-*-nacl*"]} {
     return
@@ -1302,6 +1308,51 @@ if { [isnative] && [check_compiler_available] } {
     ]
 
     if  {[istarget "x86_64-*-linux*-gnux32"]} {
+	run_cc_link_tests [list \
+	    [list \
+		"Build pr25416-5b.o (GDesc -maddress-mode=short)" \
+		"" \
+		"-Wa,-mx86-used-note=yes" \
+		{ pr25416-5b.s } \
+	    ] \
+	    [list \
+		"Build pr25416-5b.so (GDesc -> IE -maddress-mode=short)" \
+		"-shared" \
+		"-fPIC -Wa,-mx86-used-note=yes" \
+		{ pr25416-5b.s pr25416-5d.s } \
+		{} \
+		"pr25416-5b.so" \
+	    ] \
+	    [list \
+		"Build pr25416-5c.o (GDesc -maddress-mode=long)" \
+		"" \
+		"-Wa,-mx86-used-note=yes" \
+		{ pr25416-5c.s } \
+	    ] \
+	    [list \
+		"Build pr25416-5c.so (GDesc -> IE -maddress-mode=long)" \
+		"-shared" \
+		"-fPIC -Wa,-mx86-used-note=yes" \
+		{ pr25416-5c.s pr25416-5d.s } \
+		{} \
+		"pr25416-5c.so" \
+	    ] \
+	    [list \
+		"Build pr25416-5d.so (GDesc -maddress-mode=short)" \
+		"-shared" \
+		"-fPIC -Wa,-mx86-used-note=yes" \
+		{ pr25416-5b.s pr25416-5e.s } \
+		{} \
+		"pr25416-5d.so" \
+	    ] \
+	    [list \
+		"Build pr25416-5d.o (IE -maddress-mode=short)" \
+		"" \
+		"-Wa,-mx86-used-note=yes" \
+		{ pr25416-5d.s } \
+	    ] \
+	]
+
 	run_ld_link_exec_tests [list \
 	    [list \
 		"Run pr22001-1b" \
@@ -1321,6 +1372,67 @@ if { [isnative] && [check_compiler_available] } {
 		"pass.out" \
 		"$NOPIE_CFLAGS" \
 	    ] \
+	    [list \
+		"Run pr25416-5a (GDesc -> IE -maddress-mode=short)" \
+		"$NOPIE_LDFLAGS -Wl,--no-as-needed tmpdir/pr25416-5b.so" \
+		"-Wa,-mx86-used-note=yes" \
+		{ pr25416-5a.c } \
+		"pr25416-5a" \
+		"pass.out" \
+		"$NOPIE_CFLAGS" \
+	    ] \
+	    [list \
+		"Run pr25416-5b (GDesc -> LE -maddress-mode=short" \
+		"$NOPIE_LDFLAGS tmpdir/pr25416-5b.o tmpdir/pr25416-5d.o" \
+		"-Wa,-mx86-used-note=yes" \
+		{ pr25416-5a.c } \
+		"pr25416-5b" \
+		"pass.out" \
+		"$NOPIE_CFLAGS" \
+	    ] \
+	    [list \
+		"Run pr25416-5c (GDesc -> IE -maddress-mode=long)" \
+		"$NOPIE_LDFLAGS -Wl,--no-as-needed tmpdir/pr25416-5c.so" \
+		"-Wa,-mx86-used-note=yes" \
+		{ pr25416-5a.c } \
+		"pr25416-5c" \
+		"pass.out" \
+		"$NOPIE_CFLAGS" \
+	    ] \
+	    [list \
+		"Run pr25416-5d (GDesc -> LE -maddress-mode=long)" \
+		"$NOPIE_LDFLAGS tmpdir/pr25416-5c.o tmpdir/pr25416-5d.o" \
+		"-Wa,-mx86-used-note=yes" \
+		{ pr25416-5a.c } \
+		"pr25416-5d" \
+		"pass.out" \
+		"$NOPIE_CFLAGS" \
+	    ] \
+	    [list \
+		"Run pr25416-5e (GDesc -maddress-mode=short)" \
+		"$NOPIE_LDFLAGS -Wl,--no-as-needed tmpdir/pr25416-5d.so" \
+		"-Wa,-mx86-used-note=yes" \
+		{ pr25416-5a.c } \
+		"pr25416-5e" \
+		"pass.out" \
+		"$NOPIE_CFLAGS" \
+	    ] \
+	    [list \
+		"Run pr25416-5f (PIE GDesc -> LE -maddress-mode=short)" \
+		"-pie tmpdir/pr25416-5b.o tmpdir/pr25416-5d.o" \
+		"-Wa,-mx86-used-note=yes" \
+		{ pr25416-5a.c } \
+		"pr25416-5f" \
+		"pass.out" \
+	    ] \
+	    [list \
+		"Run pr25416-5g (PIE GDesc -> LE -maddress-mode=long)" \
+		"-pie tmpdir/pr25416-5c.o tmpdir/pr25416-5d.o" \
+		"-Wa,-mx86-used-note=yes" \
+		{ pr25416-5a.c } \
+		"pr25416-5g" \
+		"pass.out" \
+	    ] \
 	]
     } else {
 	run_cc_link_tests [list \