Support mixed SSE/AVX audit and check AVX only once.

This patch fixes mixed SSE/AVX audit and checks AVX only once in
_dl_runtime_profile. When an AVX or SSE register value in pltenter is
modified, we have to make sure that the SSE part value is the same in both
lr_xmm and lr_vector fields so that pltexit will get the correct value
from either lr_xmm or lr_vector fields. AVX-enabled pltenter should
update both lr_xmm and lr_vector fields to support stacked AVX/SSE
pltenter functions.

11 files changed, 1062 insertions, 239 deletions

diff --git a/ChangeLog b/ChangeLog
index 072af87..12fa4cd 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,31 @@
+2009-08-01  H.J. Lu  <hongjiu.lu@intel.com>
+
+	* elf/Makefile (distribute): Add tst-audit6.c tst-auditmod6a.c
+	tst-auditmod6b.c tst-auditmod6c.c tst-audit7.c tst-auditmod7a.c
+	tst-auditmod7b.c.
+	(tests): Add tst-audit6 tst-audit7.
+	(modules-names): Add st-auditmod6a tst-auditmod6b tst-auditmod6c
+	tst-auditmod7a tst-auditmod7b.
+	($(objpfx)tst-audit6): New.
+	($(objpfx)tst-audit6.out): Likewise.
+	($(objpfx)tst-audit7): Likewise.
+	($(objpfx)tst-audit7.out): Likewise.
+	(tst-audit6-ENV): Likewise.
+	(tst-audit7-ENV): Likewise.
+	(CFLAGS-tst-auditmod6b.c): Likewise.
+	(CFLAGS-tst-auditmod6c.c): Likewise.
+	(CFLAGS-tst-auditmod7b.c): Likewise.
+	* elf/tst-audit6.c: New file.
+	* elf/tst-audit7.c: New file.
+	* elf/tst-auditmod6a.c: New file.
+	* elf/tst-auditmod6b.c: New file.
+	* elf/tst-auditmod6c.c: New file.
+	* elf/tst-auditmod7a.c: New file.
+	* elf/tst-auditmod7b.c: New file.
+	* sysdeps/x86_64/dl-trampoline.S (_dl_runtime_profile): Move
+	saving and restoring SSE/AVX registers to ...
+	* sysdeps/x86_64/dl-trampoline.h: This.  New file.
+
 2009-08-07  H.J. Lu  <hongjiu.lu@intel.com>
 
 	* sysdeps/i386/i686/multiarch/strcspn.S (STRCSPN): Use PIC
diff --git a/elf/Makefile b/elf/Makefile
index 3baad96..d57c7fe 100644
--- a/elf/Makefile
+++ b/elf/Makefile
@@ -93,6 +93,9 @@ distribute	:= rtld-Rules \
 		   tst-auditmod1.c tst-auditmod3a.c tst-auditmod3b.c \
 		   tst-auditmod4a.c tst-auditmod4b.c \
 		   tst-audit5.c tst-auditmod5a.c tst-auditmod5b.c \
+		   tst-audit6.c tst-auditmod6a.c tst-auditmod6b.c \
+		   tst-auditmod6c.c \
+		   tst-audit7.c tst-auditmod7a.c tst-auditmod7b.c \
 		   order2mod1.c order2mod2.c order2mod3.c order2mod4.c \
 		   tst-stackguard1.c tst-stackguard1-static.c \
 		   tst-array5.c tst-array5-static.c tst-array5dep.c \
@@ -200,7 +203,7 @@ tests += loadtest restest1 preloadtest loadfail multiload origtest resolvfail \
 test-srcs = tst-pathopt
 tests-execstack-yes = tst-execstack tst-execstack-needed tst-execstack-prog
 ifeq (x86_64,$(config-machine))
-tests += tst-audit3 tst-audit4 tst-audit5
+tests += tst-audit3 tst-audit4 tst-audit5 tst-audit6 tst-audit7
 endif
 endif
 ifeq (yesyes,$(have-fpie)$(build-shared))
@@ -255,7 +258,9 @@ endif
 ifeq (x86_64,$(config-machine))
 modules-names += tst-auditmod3a tst-auditmod3b \
 		tst-auditmod4a tst-auditmod4b \
-		tst-auditmod5a tst-auditmod5b
+		tst-auditmod5a tst-auditmod5b \
+		tst-auditmod6a tst-auditmod6b tst-auditmod6c \
+		tst-auditmod7a tst-auditmod7b
 endif
 modules-execstack-yes = tst-execstack-mod
 extra-test-objs += $(addsuffix .os,$(strip $(modules-names)))
@@ -987,6 +992,15 @@ $(objpfx)tst-audit5: $(objpfx)tst-auditmod5a.so
 $(objpfx)tst-audit5.out: $(objpfx)tst-auditmod5b.so
 tst-audit5-ENV = LD_AUDIT=$(objpfx)tst-auditmod5b.so
 
+$(objpfx)tst-audit6: $(objpfx)tst-auditmod6a.so
+$(objpfx)tst-audit6.out: $(objpfx)tst-auditmod6b.so \
+			 $(objpfx)tst-auditmod6c.so
+tst-audit6-ENV = LD_AUDIT=$(objpfx)tst-auditmod6b.so:$(objpfx)tst-auditmod6c.so
+
+$(objpfx)tst-audit7: $(objpfx)tst-auditmod7a.so
+$(objpfx)tst-audit7.out: $(objpfx)tst-auditmod7b.so
+tst-audit7-ENV = LD_AUDIT=$(objpfx)tst-auditmod7b.so
+
 $(objpfx)tst-global1: $(libdl)
 $(objpfx)tst-global1.out: $(objpfx)testobj6.so $(objpfx)testobj2.so
 
@@ -1134,4 +1148,7 @@ ifeq (yes,$(config-cflags-avx))
 CFLAGS-tst-audit4.c += -mavx
 CFLAGS-tst-auditmod4a.c += -mavx
 CFLAGS-tst-auditmod4b.c += -mavx
+CFLAGS-tst-auditmod6b.c += -mavx
+CFLAGS-tst-auditmod6c.c += -mavx
+CFLAGS-tst-auditmod7b.c += -mavx
 endif
diff --git a/elf/tst-audit6.c b/elf/tst-audit6.c
new file mode 100644
index 0000000..1f6dcb1
--- /dev/null
+++ b/elf/tst-audit6.c
@@ -0,0 +1,28 @@
+/* Test case for x86-64 preserved registers in dynamic linker.  */
+
+#include <stdlib.h>
+#include <string.h>
+#include <cpuid.h>
+#include <emmintrin.h>
+
+extern __m128i audit_test (__m128i, __m128i, __m128i, __m128i,
+			   __m128i, __m128i, __m128i, __m128i);
+
+int
+main (void)
+{
+  unsigned int eax, ebx, ecx, edx;
+
+  /* Run AVX test only if AVX is supported.  */
+  if (__get_cpuid (1, &eax, &ebx, &ecx, &edx)
+      && (ecx & bit_AVX))
+    {
+      __m128i xmm = _mm_setzero_si128 ();
+      __m128i ret = audit_test (xmm, xmm, xmm, xmm, xmm, xmm, xmm, xmm);
+
+      xmm = _mm_set1_epi32 (0x98abcdef);
+      if (memcmp (&xmm, &ret, sizeof (ret)))
+	abort ();
+    }
+  return 0;
+}
diff --git a/elf/tst-audit7.c b/elf/tst-audit7.c
new file mode 100644
index 0000000..1d2a7de
--- /dev/null
+++ b/elf/tst-audit7.c
@@ -0,0 +1 @@
+#include "tst-audit6.c"
diff --git a/elf/tst-auditmod6a.c b/elf/tst-auditmod6a.c
new file mode 100644
index 0000000..c3a850c
--- /dev/null
+++ b/elf/tst-auditmod6a.c
@@ -0,0 +1,46 @@
+/* Test case for x86-64 preserved registers in dynamic linker.  */
+
+#include <stdlib.h>
+#include <string.h>
+#include <emmintrin.h>
+
+__m128i
+audit_test (__m128i x0, __m128i x1, __m128i x2, __m128i x3,
+	    __m128i x4, __m128i x5, __m128i x6, __m128i x7)
+{
+  __m128i xmm;
+
+  xmm =  _mm_set1_epi32 (0x100);
+  if (memcmp (&xmm, &x0, sizeof (xmm)))
+    abort ();
+
+  xmm =  _mm_set1_epi32 (0x101);
+  if (memcmp (&xmm, &x1, sizeof (xmm)))
+    abort ();
+
+  xmm =  _mm_set1_epi32 (0x102);
+  if (memcmp (&xmm, &x2, sizeof (xmm)))
+    abort ();
+
+  xmm =  _mm_set1_epi32 (0x103);
+  if (memcmp (&xmm, &x3, sizeof (xmm)))
+    abort ();
+
+  xmm =  _mm_set1_epi32 (0x104);
+  if (memcmp (&xmm, &x4, sizeof (xmm)))
+    abort ();
+
+  xmm =  _mm_set1_epi32 (0x105);
+  if (memcmp (&xmm, &x5, sizeof (xmm)))
+    abort ();
+
+  xmm =  _mm_set1_epi32 (0x106);
+  if (memcmp (&xmm, &x6, sizeof (xmm)))
+    abort ();
+
+  xmm =  _mm_set1_epi32 (0x107);
+  if (memcmp (&xmm, &x7, sizeof (xmm)))
+    abort ();
+
+  return _mm_setzero_si128 ();
+}
diff --git a/elf/tst-auditmod6b.c b/elf/tst-auditmod6b.c
new file mode 100644
index 0000000..f756b50
--- /dev/null
+++ b/elf/tst-auditmod6b.c
@@ -0,0 +1,220 @@
+/* Verify that changing AVX registers in audit library won't affect
+   function parameter passing/return.  */
+
+#include <dlfcn.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <bits/wordsize.h>
+#include <gnu/lib-names.h>
+
+unsigned int
+la_version (unsigned int v)
+{
+  setlinebuf (stdout);
+
+  printf ("version: %u\n", v);
+
+  char buf[20];
+  sprintf (buf, "%u", v);
+
+  return v;
+}
+
+void
+la_activity (uintptr_t *cookie, unsigned int flag)
+{
+  if (flag == LA_ACT_CONSISTENT)
+    printf ("activity: consistent\n");
+  else if (flag == LA_ACT_ADD)
+    printf ("activity: add\n");
+  else if (flag == LA_ACT_DELETE)
+    printf ("activity: delete\n");
+  else
+    printf ("activity: unknown activity %u\n", flag);
+}
+
+char *
+la_objsearch (const char *name, uintptr_t *cookie, unsigned int flag)
+{
+  char buf[100];
+  const char *flagstr;
+  if (flag == LA_SER_ORIG)
+    flagstr = "LA_SET_ORIG";
+  else if (flag == LA_SER_LIBPATH)
+    flagstr = "LA_SER_LIBPATH";
+  else if (flag == LA_SER_RUNPATH)
+    flagstr = "LA_SER_RUNPATH";
+  else if (flag == LA_SER_CONFIG)
+    flagstr = "LA_SER_CONFIG";
+  else if (flag == LA_SER_DEFAULT)
+    flagstr = "LA_SER_DEFAULT";
+  else if (flag == LA_SER_SECURE)
+    flagstr = "LA_SER_SECURE";
+  else
+    {
+       sprintf (buf, "unknown flag %d", flag);
+       flagstr = buf;
+    }
+  printf ("objsearch: %s, %s\n", name, flagstr);
+
+  return (char *) name;
+}
+
+unsigned int
+la_objopen (struct link_map *l, Lmid_t lmid, uintptr_t *cookie)
+{
+  printf ("objopen: %ld, %s\n", lmid, l->l_name);
+
+  return 3;
+}
+
+void
+la_preinit (uintptr_t *cookie)
+{
+  printf ("preinit\n");
+}
+
+unsigned int
+la_objclose  (uintptr_t *cookie)
+{
+  printf ("objclose\n");
+  return 0;
+}
+
+uintptr_t
+la_symbind64 (Elf64_Sym *sym, unsigned int ndx, uintptr_t *refcook,
+	      uintptr_t *defcook, unsigned int *flags, const char *symname)
+{
+  printf ("symbind64: symname=%s, st_value=%#lx, ndx=%u, flags=%u\n",
+	  symname, (long int) sym->st_value, ndx, *flags);
+
+  return sym->st_value;
+}
+
+#define pltenter la_x86_64_gnu_pltenter
+#define pltexit la_x86_64_gnu_pltexit
+#define La_regs La_x86_64_regs
+#define La_retval La_x86_64_retval
+#define int_retval lrv_rax
+
+#include <tst-audit.h>
+
+#ifdef __AVX__
+#include <immintrin.h>
+#include <cpuid.h>
+
+static int avx = -1;
+
+static int
+__attribute ((always_inline))
+check_avx (void)
+{
+  if (avx == -1)
+    {
+      unsigned int eax, ebx, ecx, edx;
+
+      if (__get_cpuid (1, &eax, &ebx, &ecx, &edx)
+	  && (ecx & bit_AVX))
+	avx = 1;
+      else
+	avx = 0;
+    }
+  return avx;
+}
+#else
+#include <emmintrin.h>
+#endif
+
+ElfW(Addr)
+pltenter (ElfW(Sym) *sym, unsigned int ndx, uintptr_t *refcook,
+	  uintptr_t *defcook, La_regs *regs, unsigned int *flags,
+	  const char *symname, long int *framesizep)
+{
+  printf ("pltenter: symname=%s, st_value=%#lx, ndx=%u, flags=%u\n",
+	  symname, (long int) sym->st_value, ndx, *flags);
+
+#ifdef __AVX__
+  if (check_avx () && strcmp (symname, "audit_test") == 0)
+    {
+      int i;
+
+      __m128i xmm = _mm_setzero_si128 ();
+      for (i = 0; i < 8; i++)
+	if (memcmp (&regs->lr_xmm[i], &xmm, sizeof (xmm))
+	    || memcmp (&regs->lr_vector[i], &xmm, sizeof (xmm)))
+	  abort ();
+
+      for (i = 0; i < 8; i += 2)
+	{
+	  regs->lr_xmm[i] = (La_x86_64_xmm) _mm_set1_epi32 (i + 1);
+	  regs->lr_vector[i].xmm[0] = regs->lr_xmm[i];
+	  regs->lr_vector[i + 1].ymm[0]
+	    = (La_x86_64_ymm) _mm256_set1_epi32 (i + 2);
+	  regs->lr_xmm[i + 1] = regs->lr_vector[i + 1].xmm[0];
+	}
+
+      __m256i ymm = _mm256_set1_epi32 (-1);
+      asm volatile ("vmovdqa %0, %%ymm0" : : "x" (ymm) : "xmm0" );
+      asm volatile ("vmovdqa %0, %%ymm1" : : "x" (ymm) : "xmm1" );
+      asm volatile ("vmovdqa %0, %%ymm2" : : "x" (ymm) : "xmm2" );
+      asm volatile ("vmovdqa %0, %%ymm3" : : "x" (ymm) : "xmm3" );
+      asm volatile ("vmovdqa %0, %%ymm4" : : "x" (ymm) : "xmm4" );
+      asm volatile ("vmovdqa %0, %%ymm5" : : "x" (ymm) : "xmm5" );
+      asm volatile ("vmovdqa %0, %%ymm6" : : "x" (ymm) : "xmm6" );
+      asm volatile ("vmovdqa %0, %%ymm7" : : "x" (ymm) : "xmm7" );
+
+      *framesizep = 1024;
+    }
+#endif
+
+  return sym->st_value;
+}
+
+unsigned int
+pltexit (ElfW(Sym) *sym, unsigned int ndx, uintptr_t *refcook,
+	 uintptr_t *defcook, const La_regs *inregs, La_retval *outregs,
+	 const char *symname)
+{
+  printf ("pltexit: symname=%s, st_value=%#lx, ndx=%u, retval=%tu\n",
+	  symname, (long int) sym->st_value, ndx, outregs->int_retval);
+
+#ifdef __AVX__
+  if (check_avx () && strcmp (symname, "audit_test") == 0)
+    {
+      int i;
+
+      __m128i xmm = _mm_setzero_si128 ();
+      if (memcmp (&outregs->lrv_xmm0, &xmm, sizeof (xmm))
+	  || memcmp (&outregs->lrv_vector0, &xmm, sizeof (xmm)))
+	abort ();
+
+      __m256i ymm;
+
+      for (i = 0; i < 8; i += 2)
+	{
+	  xmm = _mm_set1_epi32 (i + 0x100);
+	  if (memcmp (&inregs->lr_xmm[i], &xmm, sizeof (xmm))
+	      || memcmp (&inregs->lr_vector[i], &xmm, sizeof (xmm)))
+	    abort ();
+
+	  ymm = _mm256_set1_epi32 (i + 0x101);
+	  if (memcmp (&inregs->lr_xmm[i + 1],
+		      &inregs->lr_vector[i + 1].xmm[0], sizeof (xmm))
+	      || memcmp (&inregs->lr_vector[i + 1], &ymm, sizeof (ymm)))
+	    abort ();
+	}
+
+      outregs->lrv_vector0.ymm[0]
+	= (La_x86_64_ymm) _mm256_set1_epi32 (0x12349876);
+
+      ymm = _mm256_set1_epi32 (-1);
+      asm volatile ("vmovdqa %0, %%ymm0" : : "x" (ymm) : "xmm0" );
+      asm volatile ("vmovdqa %0, %%ymm1" : : "x" (ymm) : "xmm1" );
+    }
+#endif
+
+  return 0;
+}
diff --git a/elf/tst-auditmod6c.c b/elf/tst-auditmod6c.c
new file mode 100644
index 0000000..49cbf05
--- /dev/null
+++ b/elf/tst-auditmod6c.c
@@ -0,0 +1,225 @@
+/* Verify that changing AVX registers in audit library won't affect
+   function parameter passing/return.  */
+
+#include <dlfcn.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <bits/wordsize.h>
+#include <gnu/lib-names.h>
+
+unsigned int
+la_version (unsigned int v)
+{
+  setlinebuf (stdout);
+
+  printf ("version: %u\n", v);
+
+  char buf[20];
+  sprintf (buf, "%u", v);
+
+  return v;
+}
+
+void
+la_activity (uintptr_t *cookie, unsigned int flag)
+{
+  if (flag == LA_ACT_CONSISTENT)
+    printf ("activity: consistent\n");
+  else if (flag == LA_ACT_ADD)
+    printf ("activity: add\n");
+  else if (flag == LA_ACT_DELETE)
+    printf ("activity: delete\n");
+  else
+    printf ("activity: unknown activity %u\n", flag);
+}
+
+char *
+la_objsearch (const char *name, uintptr_t *cookie, unsigned int flag)
+{
+  char buf[100];
+  const char *flagstr;
+  if (flag == LA_SER_ORIG)
+    flagstr = "LA_SET_ORIG";
+  else if (flag == LA_SER_LIBPATH)
+    flagstr = "LA_SER_LIBPATH";
+  else if (flag == LA_SER_RUNPATH)
+    flagstr = "LA_SER_RUNPATH";
+  else if (flag == LA_SER_CONFIG)
+    flagstr = "LA_SER_CONFIG";
+  else if (flag == LA_SER_DEFAULT)
+    flagstr = "LA_SER_DEFAULT";
+  else if (flag == LA_SER_SECURE)
+    flagstr = "LA_SER_SECURE";
+  else
+    {
+       sprintf (buf, "unknown flag %d", flag);
+       flagstr = buf;
+    }
+  printf ("objsearch: %s, %s\n", name, flagstr);
+
+  return (char *) name;
+}
+
+unsigned int
+la_objopen (struct link_map *l, Lmid_t lmid, uintptr_t *cookie)
+{
+  printf ("objopen: %ld, %s\n", lmid, l->l_name);
+
+  return 3;
+}
+
+void
+la_preinit (uintptr_t *cookie)
+{
+  printf ("preinit\n");
+}
+
+unsigned int
+la_objclose  (uintptr_t *cookie)
+{
+  printf ("objclose\n");
+  return 0;
+}
+
+uintptr_t
+la_symbind64 (Elf64_Sym *sym, unsigned int ndx, uintptr_t *refcook,
+	      uintptr_t *defcook, unsigned int *flags, const char *symname)
+{
+  printf ("symbind64: symname=%s, st_value=%#lx, ndx=%u, flags=%u\n",
+	  symname, (long int) sym->st_value, ndx, *flags);
+
+  return sym->st_value;
+}
+
+#define pltenter la_x86_64_gnu_pltenter
+#define pltexit la_x86_64_gnu_pltexit
+#define La_regs La_x86_64_regs
+#define La_retval La_x86_64_retval
+#define int_retval lrv_rax
+
+#include <tst-audit.h>
+
+#ifdef __AVX__
+#include <immintrin.h>
+#include <cpuid.h>
+
+static int avx = -1;
+
+static int
+__attribute ((always_inline))
+check_avx (void)
+{
+  if (avx == -1)
+    {
+      unsigned int eax, ebx, ecx, edx;
+
+      if (__get_cpuid (1, &eax, &ebx, &ecx, &edx)
+	  && (ecx & bit_AVX))
+	avx = 1;
+      else
+	avx = 0;
+    }
+  return avx;
+}
+#else
+#include <emmintrin.h>
+#endif
+
+ElfW(Addr)
+pltenter (ElfW(Sym) *sym, unsigned int ndx, uintptr_t *refcook,
+	  uintptr_t *defcook, La_regs *regs, unsigned int *flags,
+	  const char *symname, long int *framesizep)
+{
+  printf ("pltenter: symname=%s, st_value=%#lx, ndx=%u, flags=%u\n",
+	  symname, (long int) sym->st_value, ndx, *flags);
+
+#ifdef __AVX__
+  if (check_avx () && strcmp (symname, "audit_test") == 0)
+    {
+      int i;
+      __m128i xmm;
+      __m256i ymm;
+
+      for (i = 0; i < 8; i += 2)
+	{
+	  xmm = _mm_set1_epi32 (i + 1);
+	  if (memcmp (&regs->lr_xmm[i], &xmm, sizeof (xmm))
+	      || memcmp (&regs->lr_vector[i], &xmm, sizeof (xmm)))
+	    abort ();
+	  regs->lr_xmm[i] = (La_x86_64_xmm) _mm_set1_epi32 (i + 0x100);
+	  regs->lr_vector[i].xmm[0] = regs->lr_xmm[i];
+
+	  ymm = _mm256_set1_epi32 (i + 2);
+	  if (memcmp (&regs->lr_xmm[i + 1],
+		      &regs->lr_vector[i + 1].xmm[0], sizeof (xmm))
+	      || memcmp (&regs->lr_vector[i + 1], &ymm, sizeof (ymm)))
+	    abort ();
+	  regs->lr_vector[i + 1].ymm[0]
+	    = (La_x86_64_ymm) _mm256_set1_epi32 (i + 0x101);
+	  regs->lr_xmm[i + 1] = regs->lr_vector[i + 1].xmm[0];
+	}
+
+      ymm = _mm256_set1_epi32 (-1);
+      asm volatile ("vmovdqa %0, %%ymm0" : : "x" (ymm) : "xmm0" );
+      asm volatile ("vmovdqa %0, %%ymm1" : : "x" (ymm) : "xmm1" );
+      asm volatile ("vmovdqa %0, %%ymm2" : : "x" (ymm) : "xmm2" );
+      asm volatile ("vmovdqa %0, %%ymm3" : : "x" (ymm) : "xmm3" );
+      asm volatile ("vmovdqa %0, %%ymm4" : : "x" (ymm) : "xmm4" );
+      asm volatile ("vmovdqa %0, %%ymm5" : : "x" (ymm) : "xmm5" );
+      asm volatile ("vmovdqa %0, %%ymm6" : : "x" (ymm) : "xmm6" );
+      asm volatile ("vmovdqa %0, %%ymm7" : : "x" (ymm) : "xmm7" );
+
+      *framesizep = 1024;
+    }
+#endif
+
+  return sym->st_value;
+}
+
+unsigned int
+pltexit (ElfW(Sym) *sym, unsigned int ndx, uintptr_t *refcook,
+	 uintptr_t *defcook, const La_regs *inregs, La_retval *outregs,
+	 const char *symname)
+{
+  printf ("pltexit: symname=%s, st_value=%#lx, ndx=%u, retval=%tu\n",
+	  symname, (long int) sym->st_value, ndx, outregs->int_retval);
+
+#ifdef __AVX__
+  if (check_avx () && strcmp (symname, "audit_test") == 0)
+    {
+      int i;
+
+      __m256i ymm = _mm256_set1_epi32 (0x12349876);;
+      if (memcmp (&outregs->lrv_vector0, &ymm, sizeof (ymm)))
+	abort ();
+
+      __m128i xmm;
+
+      for (i = 0; i < 8; i += 2)
+	{
+	  xmm = _mm_set1_epi32 (i + 0x100);
+	  if (memcmp (&inregs->lr_xmm[i], &xmm, sizeof (xmm))
+	      || memcmp (&inregs->lr_vector[i], &xmm, sizeof (xmm)))
+	    abort ();
+
+	  ymm = _mm256_set1_epi32 (i + 0x101);
+	  if (memcmp (&inregs->lr_xmm[i + 1],
+		      &inregs->lr_vector[i + 1].xmm[0], sizeof (xmm))
+	      || memcmp (&inregs->lr_vector[i + 1], &ymm, sizeof (ymm)))
+	    abort ();
+	}
+
+      outregs->lrv_vector0.ymm[0]
+	= (La_x86_64_ymm) _mm256_set1_epi32 (0x98abcdef);
+
+      ymm = _mm256_set1_epi32 (-1);
+      asm volatile ("vmovdqa %0, %%ymm0" : : "x" (ymm) : "xmm0" );
+      asm volatile ("vmovdqa %0, %%ymm1" : : "x" (ymm) : "xmm1" );
+    }
+#endif
+
+  return 0;
+}
diff --git a/elf/tst-auditmod7a.c b/elf/tst-auditmod7a.c
new file mode 100644
index 0000000..b379df7
--- /dev/null
+++ b/elf/tst-auditmod7a.c
@@ -0,0 +1 @@
+#include "tst-auditmod6a.c"
diff --git a/elf/tst-auditmod7b.c b/elf/tst-auditmod7b.c
new file mode 100644
index 0000000..eb23758
--- /dev/null
+++ b/elf/tst-auditmod7b.c
@@ -0,0 +1,218 @@
+/* Verify that changing AVX registers in audit library won't affect
+   function parameter passing/return.  */
+
+#include <dlfcn.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <bits/wordsize.h>
+#include <gnu/lib-names.h>
+
+unsigned int
+la_version (unsigned int v)
+{
+  setlinebuf (stdout);
+
+  printf ("version: %u\n", v);
+
+  char buf[20];
+  sprintf (buf, "%u", v);
+
+  return v;
+}
+
+void
+la_activity (uintptr_t *cookie, unsigned int flag)
+{
+  if (flag == LA_ACT_CONSISTENT)
+    printf ("activity: consistent\n");
+  else if (flag == LA_ACT_ADD)
+    printf ("activity: add\n");
+  else if (flag == LA_ACT_DELETE)
+    printf ("activity: delete\n");
+  else
+    printf ("activity: unknown activity %u\n", flag);
+}
+
+char *
+la_objsearch (const char *name, uintptr_t *cookie, unsigned int flag)
+{
+  char buf[100];
+  const char *flagstr;
+  if (flag == LA_SER_ORIG)
+    flagstr = "LA_SET_ORIG";
+  else if (flag == LA_SER_LIBPATH)
+    flagstr = "LA_SER_LIBPATH";
+  else if (flag == LA_SER_RUNPATH)
+    flagstr = "LA_SER_RUNPATH";
+  else if (flag == LA_SER_CONFIG)
+    flagstr = "LA_SER_CONFIG";
+  else if (flag == LA_SER_DEFAULT)
+    flagstr = "LA_SER_DEFAULT";
+  else if (flag == LA_SER_SECURE)
+    flagstr = "LA_SER_SECURE";
+  else
+    {
+       sprintf (buf, "unknown flag %d", flag);
+       flagstr = buf;
+    }
+  printf ("objsearch: %s, %s\n", name, flagstr);
+
+  return (char *) name;
+}
+
+unsigned int
+la_objopen (struct link_map *l, Lmid_t lmid, uintptr_t *cookie)
+{
+  printf ("objopen: %ld, %s\n", lmid, l->l_name);
+
+  return 3;
+}
+
+void
+la_preinit (uintptr_t *cookie)
+{
+  printf ("preinit\n");
+}
+
+unsigned int
+la_objclose  (uintptr_t *cookie)
+{
+  printf ("objclose\n");
+  return 0;
+}
+
+uintptr_t
+la_symbind64 (Elf64_Sym *sym, unsigned int ndx, uintptr_t *refcook,
+	      uintptr_t *defcook, unsigned int *flags, const char *symname)
+{
+  printf ("symbind64: symname=%s, st_value=%#lx, ndx=%u, flags=%u\n",
+	  symname, (long int) sym->st_value, ndx, *flags);
+
+  return sym->st_value;
+}
+
+#define pltenter la_x86_64_gnu_pltenter
+#define pltexit la_x86_64_gnu_pltexit
+#define La_regs La_x86_64_regs
+#define La_retval La_x86_64_retval
+#define int_retval lrv_rax
+
+#include <tst-audit.h>
+
+#ifdef __AVX__
+#include <immintrin.h>
+#include <cpuid.h>
+
+static int avx = -1;
+
+static int
+__attribute ((always_inline))
+check_avx (void)
+{
+  if (avx == -1)
+    {
+      unsigned int eax, ebx, ecx, edx;
+
+      if (__get_cpuid (1, &eax, &ebx, &ecx, &edx)
+	  && (ecx & bit_AVX))
+	avx = 1;
+      else
+	avx = 0;
+    }
+  return avx;
+}
+#else
+#include <emmintrin.h>
+#endif
+
+ElfW(Addr)
+pltenter (ElfW(Sym) *sym, unsigned int ndx, uintptr_t *refcook,
+	  uintptr_t *defcook, La_regs *regs, unsigned int *flags,
+	  const char *symname, long int *framesizep)
+{
+  printf ("pltenter: symname=%s, st_value=%#lx, ndx=%u, flags=%u\n",
+	  symname, (long int) sym->st_value, ndx, *flags);
+
+#ifdef __AVX__
+  if (check_avx () && strcmp (symname, "audit_test") == 0)
+    {
+      int i;
+
+      __m128i xmm = _mm_setzero_si128 ();
+      for (i = 0; i < 8; i++)
+	if (memcmp (&regs->lr_xmm[i], &xmm, sizeof (xmm))
+	    || memcmp (&regs->lr_vector[i], &xmm, sizeof (xmm)))
+	  abort ();
+
+      for (i = 0; i < 8; i += 2)
+	{
+	  regs->lr_xmm[i] = (La_x86_64_xmm) _mm_set1_epi32 (i + 0x100);
+	  regs->lr_vector[i + 1].ymm[0]
+	    = (La_x86_64_ymm) _mm256_set1_epi32 (i + 0x101);
+	}
+
+      __m256i ymm = _mm256_set1_epi32 (-1);
+      asm volatile ("vmovdqa %0, %%ymm0" : : "x" (ymm) : "xmm0" );
+      asm volatile ("vmovdqa %0, %%ymm1" : : "x" (ymm) : "xmm1" );
+      asm volatile ("vmovdqa %0, %%ymm2" : : "x" (ymm) : "xmm2" );
+      asm volatile ("vmovdqa %0, %%ymm3" : : "x" (ymm) : "xmm3" );
+      asm volatile ("vmovdqa %0, %%ymm4" : : "x" (ymm) : "xmm4" );
+      asm volatile ("vmovdqa %0, %%ymm5" : : "x" (ymm) : "xmm5" );
+      asm volatile ("vmovdqa %0, %%ymm6" : : "x" (ymm) : "xmm6" );
+      asm volatile ("vmovdqa %0, %%ymm7" : : "x" (ymm) : "xmm7" );
+
+      *framesizep = 1024;
+    }
+#endif
+
+  return sym->st_value;
+}
+
+unsigned int
+pltexit (ElfW(Sym) *sym, unsigned int ndx, uintptr_t *refcook,
+	 uintptr_t *defcook, const La_regs *inregs, La_retval *outregs,
+	 const char *symname)
+{
+  printf ("pltexit: symname=%s, st_value=%#lx, ndx=%u, retval=%tu\n",
+	  symname, (long int) sym->st_value, ndx, outregs->int_retval);
+
+#ifdef __AVX__
+  if (check_avx () && strcmp (symname, "audit_test") == 0)
+    {
+      int i;
+
+      __m128i xmm = _mm_setzero_si128 ();
+      if (memcmp (&outregs->lrv_xmm0, &xmm, sizeof (xmm))
+	  || memcmp (&outregs->lrv_vector0, &xmm, sizeof (xmm)))
+	abort ();
+
+      __m256i ymm;
+
+      for (i = 0; i < 8; i += 2)
+	{
+	  xmm = _mm_set1_epi32 (i + 0x100);
+	  if (memcmp (&inregs->lr_xmm[i], &xmm, sizeof (xmm))
+	      || memcmp (&inregs->lr_vector[i], &xmm, sizeof (xmm)))
+	    abort ();
+
+	  ymm = _mm256_set1_epi32 (i + 0x101);
+	  if (memcmp (&inregs->lr_xmm[i + 1],
+		      &inregs->lr_vector[i + 1].xmm[0], sizeof (xmm))
+	      || memcmp (&inregs->lr_vector[i + 1], &ymm, sizeof (ymm)))
+	    abort ();
+	}
+
+      outregs->lrv_vector0.ymm[0]
+	= (La_x86_64_ymm) _mm256_set1_epi32 (0x98abcdef);
+
+      ymm = _mm256_set1_epi32 (-1);
+      asm volatile ("vmovdqa %0, %%ymm0" : : "x" (ymm) : "xmm0" );
+      asm volatile ("vmovdqa %0, %%ymm1" : : "x" (ymm) : "xmm1" );
+    }
+#endif
+
+  return 0;
+}
diff --git a/sysdeps/x86_64/dl-trampoline.S b/sysdeps/x86_64/dl-trampoline.S
index 20da695..f9c60ad 100644
--- a/sysdeps/x86_64/dl-trampoline.S
+++ b/sysdeps/x86_64/dl-trampoline.S
@@ -146,247 +146,17 @@ L(have_avx):
 2:	movl	%eax, L(have_avx)(%rip)
 	cmpl	$0, %eax
 
-1:	js	L(no_avx1)
+1:	js	L(no_avx)
 
-	/* This is to support AVX audit modules.  */
-	vmovdqu %ymm0,		      (LR_VECTOR_OFFSET)(%rsp)
-	vmovdqu %ymm1, (LR_VECTOR_OFFSET +   VECTOR_SIZE)(%rsp)
-	vmovdqu %ymm2, (LR_VECTOR_OFFSET + VECTOR_SIZE*2)(%rsp)
-	vmovdqu %ymm3, (LR_VECTOR_OFFSET + VECTOR_SIZE*3)(%rsp)
-	vmovdqu %ymm4, (LR_VECTOR_OFFSET + VECTOR_SIZE*4)(%rsp)
-	vmovdqu %ymm5, (LR_VECTOR_OFFSET + VECTOR_SIZE*5)(%rsp)
-	vmovdqu %ymm6, (LR_VECTOR_OFFSET + VECTOR_SIZE*6)(%rsp)
-	vmovdqu %ymm7, (LR_VECTOR_OFFSET + VECTOR_SIZE*7)(%rsp)
+#  define RESTORE_AVX
+#  include "dl-trampoline.h"
 
-	/* Save xmm0-xmm7 registers to detect if any of them are
-	   changed by audit module.  */
-	vmovdqa %xmm0,		    (LR_SIZE)(%rsp)
-	vmovdqa %xmm1, (LR_SIZE +   XMM_SIZE)(%rsp)
-	vmovdqa %xmm2, (LR_SIZE + XMM_SIZE*2)(%rsp)
-	vmovdqa %xmm3, (LR_SIZE + XMM_SIZE*3)(%rsp)
-	vmovdqa %xmm4, (LR_SIZE + XMM_SIZE*4)(%rsp)
-	vmovdqa %xmm5, (LR_SIZE + XMM_SIZE*5)(%rsp)
-	vmovdqa %xmm6, (LR_SIZE + XMM_SIZE*6)(%rsp)
-	vmovdqa %xmm7, (LR_SIZE + XMM_SIZE*7)(%rsp)
-
-L(no_avx1):
-# endif
-
-	movq %rsp, %rcx		# La_x86_64_regs pointer to %rcx.
-	movq 48(%rbx), %rdx	# Load return address if needed.
-	movq 40(%rbx), %rsi	# Copy args pushed by PLT in register.
-	movq 32(%rbx), %rdi	# %rdi: link_map, %rsi: reloc_index
-	leaq 16(%rbx), %r8
-	call _dl_profile_fixup	# Call resolver.
-
-	movq %rax, %r11		# Save return value.
-
-	movq 8(%rbx), %rax	# Get back register content.
-	movq LR_RDX_OFFSET(%rsp), %rdx
-	movq  LR_R8_OFFSET(%rsp), %r8
-	movq  LR_R9_OFFSET(%rsp), %r9
-
-	movaps		    (LR_XMM_OFFSET)(%rsp), %xmm0
-	movaps	 (LR_XMM_OFFSET + XMM_SIZE)(%rsp), %xmm1
-	movaps (LR_XMM_OFFSET + XMM_SIZE*2)(%rsp), %xmm2
-	movaps (LR_XMM_OFFSET + XMM_SIZE*3)(%rsp), %xmm3
-	movaps (LR_XMM_OFFSET + XMM_SIZE*4)(%rsp), %xmm4
-	movaps (LR_XMM_OFFSET + XMM_SIZE*5)(%rsp), %xmm5
-	movaps (LR_XMM_OFFSET + XMM_SIZE*6)(%rsp), %xmm6
-	movaps (LR_XMM_OFFSET + XMM_SIZE*7)(%rsp), %xmm7
-
-# ifdef HAVE_AVX_SUPPORT
-	cmpl	$0, L(have_avx)(%rip)
-	js	L(no_avx2)
-
-	/* Check if any xmm0-xmm7 registers are changed by audit
-	   module.  */
-	vpcmpeqq (LR_SIZE)(%rsp), %xmm0, %xmm8
-	vpmovmskb %xmm8, %esi
-	cmpl $0xffff, %esi
-	jne 1f
-	vmovdqu			(LR_VECTOR_OFFSET)(%rsp), %ymm0
-
-1:	vpcmpeqq (LR_SIZE + XMM_SIZE)(%rsp), %xmm1, %xmm8
-	vpmovmskb %xmm8, %esi
-	cmpl $0xffff, %esi
-	jne 1f
-	vmovdqu	  (LR_VECTOR_OFFSET + VECTOR_SIZE)(%rsp), %ymm1
-
-1:	vpcmpeqq (LR_SIZE + XMM_SIZE*2)(%rsp), %xmm2, %xmm8
-	vpmovmskb %xmm8, %esi
-	cmpl $0xffff, %esi
-	jne 1f
-	vmovdqu (LR_VECTOR_OFFSET + VECTOR_SIZE*2)(%rsp), %ymm2
-
-1:	vpcmpeqq (LR_SIZE + XMM_SIZE*3)(%rsp), %xmm3, %xmm8
-	vpmovmskb %xmm8, %esi
-	cmpl $0xffff, %esi
-	jne 1f
-	vmovdqu (LR_VECTOR_OFFSET + VECTOR_SIZE*3)(%rsp), %ymm3
-
-1:	vpcmpeqq (LR_SIZE + XMM_SIZE*4)(%rsp), %xmm4, %xmm8
-	vpmovmskb %xmm8, %esi
-	cmpl $0xffff, %esi
-	jne 1f
-	vmovdqu (LR_VECTOR_OFFSET + VECTOR_SIZE*4)(%rsp), %ymm4
-
-1:	vpcmpeqq (LR_SIZE + XMM_SIZE*5)(%rsp), %xmm5, %xmm8
-	vpmovmskb %xmm8, %esi
-	cmpl $0xffff, %esi
-	jne 1f
-	vmovdqu (LR_VECTOR_OFFSET + VECTOR_SIZE*5)(%rsp), %ymm5
-
-1:	vpcmpeqq (LR_SIZE + XMM_SIZE*6)(%rsp), %xmm6, %xmm8
-	vpmovmskb %xmm8, %esi
-	cmpl $0xffff, %esi
-	jne 1f
-	vmovdqu (LR_VECTOR_OFFSET + VECTOR_SIZE*6)(%rsp), %ymm6
-
-1:	vpcmpeqq (LR_SIZE + XMM_SIZE*7)(%rsp), %xmm7, %xmm8
-	vpmovmskb %xmm8, %esi
-	cmpl $0xffff, %esi
-	jne 1f
-	vmovdqu (LR_VECTOR_OFFSET + VECTOR_SIZE*7)(%rsp), %ymm7
-
-L(no_avx2):
-1:
-# endif
-	movq 16(%rbx), %r10	# Anything in framesize?
-	testq %r10, %r10
-	jns 3f
-
-	/* There's nothing in the frame size, so there
-	   will be no call to the _dl_call_pltexit. */
-
-	/* Get back registers content.  */
-	movq LR_RCX_OFFSET(%rsp), %rcx
-	movq LR_RSI_OFFSET(%rsp), %rsi
-	movq LR_RDI_OFFSET(%rsp), %rdi
-
-	movq %rbx, %rsp
-	movq (%rsp), %rbx
-	cfi_restore(rbx)
-	cfi_def_cfa_register(%rsp)
-
-	addq $48, %rsp		# Adjust the stack to the return value
-				# (eats the reloc index and link_map)
-	cfi_adjust_cfa_offset(-48)
-	jmp *%r11		# Jump to function address.
-
-3:
-	cfi_adjust_cfa_offset(48)
-	cfi_rel_offset(%rbx, 0)
-	cfi_def_cfa_register(%rbx)
-
-	/* At this point we need to prepare new stack for the function
-	   which has to be called.  We copy the original stack to a
-	   temporary buffer of the size specified by the 'framesize'
-	   returned from _dl_profile_fixup */
-
-	leaq LR_RSP_OFFSET(%rbx), %rsi	# stack
-	addq $8, %r10
-	andq $0xfffffffffffffff0, %r10
-	movq %r10, %rcx
-	subq %r10, %rsp
-	movq %rsp, %rdi
-	shrq $3, %rcx
-	rep
-	movsq
-
-	movq 24(%rdi), %rcx	# Get back register content.
-	movq 32(%rdi), %rsi
-	movq 40(%rdi), %rdi
-
-	call *%r11
-
-	mov 24(%rbx), %rsp	# Drop the copied stack content
-
-	/* Now we have to prepare the La_x86_64_retval structure for the
-	   _dl_call_pltexit.  The La_x86_64_regs is being pointed by rsp now,
-	   so we just need to allocate the sizeof(La_x86_64_retval) space on
-	   the stack, since the alignment has already been taken care of. */
-# ifdef HAVE_AVX_SUPPORT
-	/* sizeof(La_x86_64_retval).  Need extra space for 2 SSE
-	   registers to detect if xmm0/xmm1 registers are changed
-	   by audit module.  */
-	subq $(LRV_SIZE + XMM_SIZE*2), %rsp
-# else
-	subq $LRV_SIZE, %rsp	# sizeof(La_x86_64_retval)
-# endif
-	movq %rsp, %rcx		# La_x86_64_retval argument to %rcx.
-
-	/* Fill in the La_x86_64_retval structure.  */
-	movq %rax, LRV_RAX_OFFSET(%rcx)
-	movq %rdx, LRV_RDX_OFFSET(%rcx)
-
-	movaps %xmm0, LRV_XMM0_OFFSET(%rcx)
-	movaps %xmm1, LRV_XMM1_OFFSET(%rcx)
-
-# ifdef HAVE_AVX_SUPPORT
-	cmpl	$0, L(have_avx)(%rip)
-	js	L(no_avx3)
-
-	/* This is to support AVX audit modules.  */
-	vmovdqu %ymm0, LRV_VECTOR0_OFFSET(%rcx)
-	vmovdqu %ymm1, LRV_VECTOR1_OFFSET(%rcx)
-
-	/* Save xmm0/xmm1 registers to detect if they are changed
-	   by audit module.  */
-	vmovdqa %xmm0,		  (LRV_SIZE)(%rcx)
-	vmovdqa %xmm1, (LRV_SIZE + XMM_SIZE)(%rcx)
-
-L(no_avx3):
-# endif
-
-	fstpt LRV_ST0_OFFSET(%rcx)
-	fstpt LRV_ST1_OFFSET(%rcx)
-
-	movq 24(%rbx), %rdx	# La_x86_64_regs argument to %rdx.
-	movq 40(%rbx), %rsi	# Copy args pushed by PLT in register.
-        movq 32(%rbx), %rdi	# %rdi: link_map, %rsi: reloc_index
-	call _dl_call_pltexit
-
-	/* Restore return registers.  */
-	movq LRV_RAX_OFFSET(%rsp), %rax
-	movq LRV_RDX_OFFSET(%rsp), %rdx
-
-	movaps LRV_XMM0_OFFSET(%rsp), %xmm0
-	movaps LRV_XMM1_OFFSET(%rsp), %xmm1
-
-# ifdef HAVE_AVX_SUPPORT
-	cmpl	$0, L(have_avx)(%rip)
-	js	L(no_avx4)
-
-	/* Check if xmm0/xmm1 registers are changed by audit module.  */
-	vpcmpeqq (LRV_SIZE)(%rsp), %xmm0, %xmm2
-	vpmovmskb %xmm2, %esi
-	cmpl $0xffff, %esi
-	jne 1f
-	vmovdqu LRV_VECTOR0_OFFSET(%rsp), %ymm0
-
-1:	vpcmpeqq (LRV_SIZE + XMM_SIZE)(%rsp), %xmm1, %xmm2
-	vpmovmskb %xmm2, %esi
-	cmpl $0xffff, %esi
-	jne 1f
-	vmovdqu LRV_VECTOR1_OFFSET(%rsp), %ymm1
-
-L(no_avx4):
-1:
+	.align 16
+L(no_avx):
 # endif
 
-	fldt LRV_ST1_OFFSET(%rsp)
-	fldt LRV_ST0_OFFSET(%rsp)
-
-	movq %rbx, %rsp
-	movq (%rsp), %rbx
-	cfi_restore(rbx)
-	cfi_def_cfa_register(%rsp)
-
-	addq $48, %rsp		# Adjust the stack to the return value
-				# (eats the reloc index and link_map)
-	cfi_adjust_cfa_offset(-48)
-	retq
+#  undef RESTORE_AVX
+#  include "dl-trampoline.h"
 
 	cfi_endproc
 	.size _dl_runtime_profile, .-_dl_runtime_profile
diff --git a/sysdeps/x86_64/dl-trampoline.h b/sysdeps/x86_64/dl-trampoline.h
new file mode 100644
index 0000000..5d49ed4
--- /dev/null
+++ b/sysdeps/x86_64/dl-trampoline.h
@@ -0,0 +1,269 @@
+/* Partial PLT profile trampoline to save and restore x86-64 vector
+   registers.
+   Copyright (C) 2009 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#ifdef RESTORE_AVX
+	/* This is to support AVX audit modules.  */
+	vmovdqu %ymm0,		      (LR_VECTOR_OFFSET)(%rsp)
+	vmovdqu %ymm1, (LR_VECTOR_OFFSET +   VECTOR_SIZE)(%rsp)
+	vmovdqu %ymm2, (LR_VECTOR_OFFSET + VECTOR_SIZE*2)(%rsp)
+	vmovdqu %ymm3, (LR_VECTOR_OFFSET + VECTOR_SIZE*3)(%rsp)
+	vmovdqu %ymm4, (LR_VECTOR_OFFSET + VECTOR_SIZE*4)(%rsp)
+	vmovdqu %ymm5, (LR_VECTOR_OFFSET + VECTOR_SIZE*5)(%rsp)
+	vmovdqu %ymm6, (LR_VECTOR_OFFSET + VECTOR_SIZE*6)(%rsp)
+	vmovdqu %ymm7, (LR_VECTOR_OFFSET + VECTOR_SIZE*7)(%rsp)
+
+	/* Save xmm0-xmm7 registers to detect if any of them are
+	   changed by audit module.  */
+	vmovdqa %xmm0,		    (LR_SIZE)(%rsp)
+	vmovdqa %xmm1, (LR_SIZE +   XMM_SIZE)(%rsp)
+	vmovdqa %xmm2, (LR_SIZE + XMM_SIZE*2)(%rsp)
+	vmovdqa %xmm3, (LR_SIZE + XMM_SIZE*3)(%rsp)
+	vmovdqa %xmm4, (LR_SIZE + XMM_SIZE*4)(%rsp)
+	vmovdqa %xmm5, (LR_SIZE + XMM_SIZE*5)(%rsp)
+	vmovdqa %xmm6, (LR_SIZE + XMM_SIZE*6)(%rsp)
+	vmovdqa %xmm7, (LR_SIZE + XMM_SIZE*7)(%rsp)
+#endif
+
+	movq %rsp, %rcx		# La_x86_64_regs pointer to %rcx.
+	movq 48(%rbx), %rdx	# Load return address if needed.
+	movq 40(%rbx), %rsi	# Copy args pushed by PLT in register.
+	movq 32(%rbx), %rdi	# %rdi: link_map, %rsi: reloc_index
+	leaq 16(%rbx), %r8
+	call _dl_profile_fixup	# Call resolver.
+
+	movq %rax, %r11		# Save return value.
+
+	movq 8(%rbx), %rax	# Get back register content.
+	movq LR_RDX_OFFSET(%rsp), %rdx
+	movq  LR_R8_OFFSET(%rsp), %r8
+	movq  LR_R9_OFFSET(%rsp), %r9
+
+	movaps		    (LR_XMM_OFFSET)(%rsp), %xmm0
+	movaps	 (LR_XMM_OFFSET + XMM_SIZE)(%rsp), %xmm1
+	movaps (LR_XMM_OFFSET + XMM_SIZE*2)(%rsp), %xmm2
+	movaps (LR_XMM_OFFSET + XMM_SIZE*3)(%rsp), %xmm3
+	movaps (LR_XMM_OFFSET + XMM_SIZE*4)(%rsp), %xmm4
+	movaps (LR_XMM_OFFSET + XMM_SIZE*5)(%rsp), %xmm5
+	movaps (LR_XMM_OFFSET + XMM_SIZE*6)(%rsp), %xmm6
+	movaps (LR_XMM_OFFSET + XMM_SIZE*7)(%rsp), %xmm7
+
+#ifdef RESTORE_AVX
+	/* Check if any xmm0-xmm7 registers are changed by audit
+	   module.  */
+	vpcmpeqq (LR_SIZE)(%rsp), %xmm0, %xmm8
+	vpmovmskb %xmm8, %esi
+	cmpl $0xffff, %esi
+	je 2f
+	vmovdqa	%xmm0, (LR_VECTOR_OFFSET)(%rsp)
+	jmp 1f
+2:	vmovdqu	(LR_VECTOR_OFFSET)(%rsp), %ymm0
+	vmovdqa	%xmm0, (LR_XMM_OFFSET)(%rsp)
+
+1:	vpcmpeqq (LR_SIZE + XMM_SIZE)(%rsp), %xmm1, %xmm8
+	vpmovmskb %xmm8, %esi
+	cmpl $0xffff, %esi
+	je 2f
+	vmovdqa	%xmm1, (LR_VECTOR_OFFSET + VECTOR_SIZE)(%rsp)
+	jmp 1f
+2:	vmovdqu	(LR_VECTOR_OFFSET + VECTOR_SIZE)(%rsp), %ymm1
+	vmovdqa	%xmm1, (LR_XMM_OFFSET + XMM_SIZE)(%rsp)
+
+1:	vpcmpeqq (LR_SIZE + XMM_SIZE*2)(%rsp), %xmm2, %xmm8
+	vpmovmskb %xmm8, %esi
+	cmpl $0xffff, %esi
+	je 2f
+	vmovdqa	%xmm2, (LR_VECTOR_OFFSET + VECTOR_SIZE*2)(%rsp)
+	jmp 1f
+2:	vmovdqu (LR_VECTOR_OFFSET + VECTOR_SIZE*2)(%rsp), %ymm2
+	vmovdqa	%xmm2, (LR_XMM_OFFSET + XMM_SIZE*2)(%rsp)
+
+1:	vpcmpeqq (LR_SIZE + XMM_SIZE*3)(%rsp), %xmm3, %xmm8
+	vpmovmskb %xmm8, %esi
+	cmpl $0xffff, %esi
+	je 2f
+	vmovdqa	%xmm3, (LR_VECTOR_OFFSET + VECTOR_SIZE*3)(%rsp)
+	jmp 1f
+2:	vmovdqu (LR_VECTOR_OFFSET + VECTOR_SIZE*3)(%rsp), %ymm3
+	vmovdqa	%xmm3, (LR_XMM_OFFSET + XMM_SIZE*3)(%rsp)
+
+1:	vpcmpeqq (LR_SIZE + XMM_SIZE*4)(%rsp), %xmm4, %xmm8
+	vpmovmskb %xmm8, %esi
+	cmpl $0xffff, %esi
+	je 2f
+	vmovdqa	%xmm4, (LR_VECTOR_OFFSET + VECTOR_SIZE*4)(%rsp)
+	jmp 1f
+2:	vmovdqu (LR_VECTOR_OFFSET + VECTOR_SIZE*4)(%rsp), %ymm4
+	vmovdqa	%xmm4, (LR_XMM_OFFSET + XMM_SIZE*4)(%rsp)
+
+1:	vpcmpeqq (LR_SIZE + XMM_SIZE*5)(%rsp), %xmm5, %xmm8
+	vpmovmskb %xmm8, %esi
+	cmpl $0xffff, %esi
+	je 2f
+	vmovdqa	%xmm5, (LR_VECTOR_OFFSET + VECTOR_SIZE*5)(%rsp)
+	jmp 1f
+2:	vmovdqu (LR_VECTOR_OFFSET + VECTOR_SIZE*5)(%rsp), %ymm5
+	vmovdqa	%xmm5, (LR_XMM_OFFSET + XMM_SIZE*5)(%rsp)
+
+1:	vpcmpeqq (LR_SIZE + XMM_SIZE*6)(%rsp), %xmm6, %xmm8
+	vpmovmskb %xmm8, %esi
+	cmpl $0xffff, %esi
+	je 2f
+	vmovdqa	%xmm6, (LR_VECTOR_OFFSET + VECTOR_SIZE*6)(%rsp)
+	jmp 1f
+2:	vmovdqu (LR_VECTOR_OFFSET + VECTOR_SIZE*6)(%rsp), %ymm6
+	vmovdqa	%xmm6, (LR_XMM_OFFSET + XMM_SIZE*6)(%rsp)
+
+1:	vpcmpeqq (LR_SIZE + XMM_SIZE*7)(%rsp), %xmm7, %xmm8
+	vpmovmskb %xmm8, %esi
+	cmpl $0xffff, %esi
+	je 2f
+	vmovdqa	%xmm7, (LR_VECTOR_OFFSET + VECTOR_SIZE*7)(%rsp)
+	jmp 1f
+2:	vmovdqu (LR_VECTOR_OFFSET + VECTOR_SIZE*7)(%rsp), %ymm7
+	vmovdqa	%xmm7, (LR_XMM_OFFSET + XMM_SIZE*7)(%rsp)
+
+1:
+#endif
+	movq 16(%rbx), %r10	# Anything in framesize?
+	testq %r10, %r10
+	jns 3f
+
+	/* There's nothing in the frame size, so there
+	   will be no call to the _dl_call_pltexit. */
+
+	/* Get back registers content.  */
+	movq LR_RCX_OFFSET(%rsp), %rcx
+	movq LR_RSI_OFFSET(%rsp), %rsi
+	movq LR_RDI_OFFSET(%rsp), %rdi
+
+	movq %rbx, %rsp
+	movq (%rsp), %rbx
+	cfi_restore(rbx)
+	cfi_def_cfa_register(%rsp)
+
+	addq $48, %rsp		# Adjust the stack to the return value
+				# (eats the reloc index and link_map)
+	cfi_adjust_cfa_offset(-48)
+	jmp *%r11		# Jump to function address.
+
+3:
+	cfi_adjust_cfa_offset(48)
+	cfi_rel_offset(%rbx, 0)
+	cfi_def_cfa_register(%rbx)
+
+	/* At this point we need to prepare new stack for the function
+	   which has to be called.  We copy the original stack to a
+	   temporary buffer of the size specified by the 'framesize'
+	   returned from _dl_profile_fixup */
+
+	leaq LR_RSP_OFFSET(%rbx), %rsi	# stack
+	addq $8, %r10
+	andq $0xfffffffffffffff0, %r10
+	movq %r10, %rcx
+	subq %r10, %rsp
+	movq %rsp, %rdi
+	shrq $3, %rcx
+	rep
+	movsq
+
+	movq 24(%rdi), %rcx	# Get back register content.
+	movq 32(%rdi), %rsi
+	movq 40(%rdi), %rdi
+
+	call *%r11
+
+	mov 24(%rbx), %rsp	# Drop the copied stack content
+
+	/* Now we have to prepare the La_x86_64_retval structure for the
+	   _dl_call_pltexit.  The La_x86_64_regs is being pointed by rsp now,
+	   so we just need to allocate the sizeof(La_x86_64_retval) space on
+	   the stack, since the alignment has already been taken care of. */
+# ifdef RESTORE_AVX
+	/* sizeof(La_x86_64_retval).  Need extra space for 2 SSE
+	   registers to detect if xmm0/xmm1 registers are changed
+	   by audit module.  */
+	subq $(LRV_SIZE + XMM_SIZE*2), %rsp
+# else
+	subq $LRV_SIZE, %rsp	# sizeof(La_x86_64_retval)
+# endif
+	movq %rsp, %rcx		# La_x86_64_retval argument to %rcx.
+
+	/* Fill in the La_x86_64_retval structure.  */
+	movq %rax, LRV_RAX_OFFSET(%rcx)
+	movq %rdx, LRV_RDX_OFFSET(%rcx)
+
+	movaps %xmm0, LRV_XMM0_OFFSET(%rcx)
+	movaps %xmm1, LRV_XMM1_OFFSET(%rcx)
+
+# ifdef RESTORE_AVX
+	/* This is to support AVX audit modules.  */
+	vmovdqu %ymm0, LRV_VECTOR0_OFFSET(%rcx)
+	vmovdqu %ymm1, LRV_VECTOR1_OFFSET(%rcx)
+
+	/* Save xmm0/xmm1 registers to detect if they are changed
+	   by audit module.  */
+	vmovdqa %xmm0,		  (LRV_SIZE)(%rcx)
+	vmovdqa %xmm1, (LRV_SIZE + XMM_SIZE)(%rcx)
+# endif
+
+	fstpt LRV_ST0_OFFSET(%rcx)
+	fstpt LRV_ST1_OFFSET(%rcx)
+
+	movq 24(%rbx), %rdx	# La_x86_64_regs argument to %rdx.
+	movq 40(%rbx), %rsi	# Copy args pushed by PLT in register.
+        movq 32(%rbx), %rdi	# %rdi: link_map, %rsi: reloc_index
+	call _dl_call_pltexit
+
+	/* Restore return registers.  */
+	movq LRV_RAX_OFFSET(%rsp), %rax
+	movq LRV_RDX_OFFSET(%rsp), %rdx
+
+	movaps LRV_XMM0_OFFSET(%rsp), %xmm0
+	movaps LRV_XMM1_OFFSET(%rsp), %xmm1
+
+# ifdef RESTORE_AVX
+	/* Check if xmm0/xmm1 registers are changed by audit module.  */
+	vpcmpeqq (LRV_SIZE)(%rsp), %xmm0, %xmm2
+	vpmovmskb %xmm2, %esi
+	cmpl $0xffff, %esi
+	jne 1f
+	vmovdqu LRV_VECTOR0_OFFSET(%rsp), %ymm0
+
+1:	vpcmpeqq (LRV_SIZE + XMM_SIZE)(%rsp), %xmm1, %xmm2
+	vpmovmskb %xmm2, %esi
+	cmpl $0xffff, %esi
+	jne 1f
+	vmovdqu LRV_VECTOR1_OFFSET(%rsp), %ymm1
+
+1:
+# endif
+
+	fldt LRV_ST1_OFFSET(%rsp)
+	fldt LRV_ST0_OFFSET(%rsp)
+
+	movq %rbx, %rsp
+	movq (%rsp), %rbx
+	cfi_restore(rbx)
+	cfi_def_cfa_register(%rsp)
+
+	addq $48, %rsp		# Adjust the stack to the return value
+				# (eats the reloc index and link_map)
+	cfi_adjust_cfa_offset(-48)
+	retq