Fix check for AVX enablement

The AVX bit is set if the CPU supports AVX.  But this doesn't mean the
kernel does.  Add checks according to Intel's documentation.

5 files changed, 58 insertions, 16 deletions

diff --git a/ChangeLog b/ChangeLog
index 593dddd8d8..f47300f92d 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,6 +1,13 @@
 2011-07-20  Ulrich Drepper  <drepper@gmail.com>
 
-	* sysdeps/x86_64/bits/link.h (La_x86_64_ymm): Force 16-byt alignment.
+	[BZ #13007]
+	* sysdeps/x86_64/dl-trampoline.S (_dl_runtime_profile): More complete
+	check for AVX enablement so that we don't crash with old kernels and
+	new hardware.
+	* elf/tst-audit4.c: Add same checks here.
+	* elf/tst-audit6.c: Likewise.
+
+	* sysdeps/x86_64/bits/link.h (La_x86_64_ymm): Force 16-byte alignment.
 
 2011-07-09  Andreas Schwab  <schwab@linux-m68k.org>
 
diff --git a/NEWS b/NEWS
index eeef7398fc..2dd7bea2ef 100644
--- a/NEWS
+++ b/NEWS
@@ -1,4 +1,4 @@
-GNU C Library NEWS -- history of user-visible changes.  2011-7-19
+GNU C Library NEWS -- history of user-visible changes.  2011-7-20
 Copyright (C) 1992-2009, 2010, 2011 Free Software Foundation, Inc.
 See the end for copying conditions.
 
@@ -9,7 +9,7 @@ Version 2.15
 
 * The following bugs are resolved with this release:
 
-  9696, 12868, 12874, 12885, 12907, 12922, 12935
+  9696, 12868, 12874, 12885, 12907, 12922, 12935, 13007
 
 * New program pldd to list loaded object of a process
   Implemented by Ulrich Drepper.
diff --git a/elf/tst-audit4.c b/elf/tst-audit4.c
index b17d4a61a7..c4f1d5bdb9 100644
--- a/elf/tst-audit4.c
+++ b/elf/tst-audit4.c
@@ -6,16 +6,30 @@
 #include <cpuid.h>
 #include <immintrin.h>
 
+
+static int
+avx_enabled (void)
+{
+  unsigned int eax, ebx, ecx, edx;
+
+  if (__get_cpuid (1, &eax, &ebx, &ecx, &edx) == 0
+      || (ecx & (bit_AVX | bit_OSXSAVE)) != (bit_AVX | bit_OSXSAVE))
+    return 0;
+
+  /* Check the OS has AVX and SSE saving enabled.  */
+  asm ("xgetbv" : "=a" (eax), "=d" (edx) : "c" (0));
+
+  return (eax & 6) == 6;
+}
+
+
 extern __m256i audit_test (__m256i, __m256i, __m256i, __m256i,
 			   __m256i, __m256i, __m256i, __m256i);
 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))
+  if (avx_enabled ())
     {
       __m256i ymm = _mm256_setzero_si256 ();
       __m256i ret = audit_test (ymm, ymm, ymm, ymm, ymm, ymm, ymm, ymm);
diff --git a/elf/tst-audit6.c b/elf/tst-audit6.c
index 1f6dcb16e9..64209a152e 100644
--- a/elf/tst-audit6.c
+++ b/elf/tst-audit6.c
@@ -8,14 +8,28 @@
 extern __m128i audit_test (__m128i, __m128i, __m128i, __m128i,
 			   __m128i, __m128i, __m128i, __m128i);
 
-int
-main (void)
+
+static int
+avx_enabled (void)
 {
   unsigned int eax, ebx, ecx, edx;
 
+  if (__get_cpuid (1, &eax, &ebx, &ecx, &edx) == 0
+      || (ecx & (bit_AVX | bit_OSXSAVE)) != (bit_AVX | bit_OSXSAVE))
+    return 0;
+
+  /* Check the OS has AVX and SSE saving enabled.  */
+  asm ("xgetbv" : "=a" (eax), "=d" (edx) : "c" (0));
+
+  return (eax & 6) == 6;
+}
+
+
+int
+main (void)
+{
   /* Run AVX test only if AVX is supported.  */
-  if (__get_cpuid (1, &eax, &ebx, &ecx, &edx)
-      && (ecx & bit_AVX))
+  if (avx_enabled ())
     {
       __m128i xmm = _mm_setzero_si128 ();
       __m128i ret = audit_test (xmm, xmm, xmm, xmm, xmm, xmm, xmm, xmm);
diff --git a/sysdeps/x86_64/dl-trampoline.S b/sysdeps/x86_64/dl-trampoline.S
index 5564a11af2..1b97929aaa 100644
--- a/sysdeps/x86_64/dl-trampoline.S
+++ b/sysdeps/x86_64/dl-trampoline.S
@@ -1,5 +1,5 @@
 /* PLT trampolines.  x86-64 version.
-   Copyright (C) 2004, 2005, 2007, 2009 Free Software Foundation, Inc.
+   Copyright (C) 2004, 2005, 2007, 2009, 2011 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
@@ -139,10 +139,17 @@ L(have_avx):
 	movl	$1, %eax
 	cpuid
 	movq	%r11,%rbx		# Restore rbx
-	movl	$1, %eax
-	testl	$(1 << 28), %ecx
-	jne	2f
-	negl	%eax
+	xorl	%eax, %eax
+	// AVX and XSAVE supported?
+	testl	$((1 << 28) | (1 << 27)), %ecx
+	je	2f
+	xorl	%ecx, %ecx
+	// Get XFEATURE_ENABLED_MASK
+	xgetbv
+	andl	$0x6, %eax
+	cmpl	$0x6, %eax
+	// Nonzero if SSE and AVX state saving is enabled.
+	sete	%al
 2:	movl	%eax, L(have_avx)(%rip)
 	cmpl	$0, %eax