Preserve SSE registers in runtime relocations on x86-64.

SSE registers are used for passing parameters and must be preserved
in runtime relocations.  This is inside ld.so enforced through the
tests in tst-xmmymm.sh.  But the malloc routines used after startup
come from libc.so and can be arbitrarily complex.  It's overkill
to save the SSE registers all the time because of that.  These calls
are rare.  Instead we save them on demand.  The new infrastructure
put in place in this patch makes this possible and efficient.

2 files changed, 86 insertions, 3 deletions

diff --git a/sysdeps/x86_64/dl-trampoline.S b/sysdeps/x86_64/dl-trampoline.S
index 49d239f075..7ecf1b0c64 100644
--- a/sysdeps/x86_64/dl-trampoline.S
+++ b/sysdeps/x86_64/dl-trampoline.S
@@ -390,3 +390,85 @@ L(no_avx4):
 	cfi_endproc
 	.size _dl_runtime_profile, .-_dl_runtime_profile
 #endif
+
+
+#ifdef SHARED
+	.globl _dl_x86_64_save_sse
+	.type _dl_x86_64_save_sse, @function
+	.align 16
+	cfi_startproc
+_dl_x86_64_save_sse:
+# ifdef HAVE_AVX_SUPPORT
+	cmpl	$0, L(have_avx)(%rip)
+	jne	1f
+	movq	%rbx, %r11		# Save rbx
+	movl	$1, %eax
+	cpuid
+	movq	%r11,%rbx		# Restore rbx
+	movl	$1, %eax
+	testl	$(1 << 28), %ecx
+	jne	2f
+	negl	%eax
+2:	movl	%eax, L(have_avx)(%rip)
+	cmpl	$0, %eax
+
+1:	js	L(no_avx5)
+
+#  define YMM_SIZE 32
+	vmovdqa %ymm0, %fs:RTLD_SAVESPACE_SSE+0*YMM_SIZE
+	vmovdqa %ymm1, %fs:RTLD_SAVESPACE_SSE+1*YMM_SIZE
+	vmovdqa %ymm2, %fs:RTLD_SAVESPACE_SSE+2*YMM_SIZE
+	vmovdqa %ymm3, %fs:RTLD_SAVESPACE_SSE+3*YMM_SIZE
+	vmovdqa %ymm4, %fs:RTLD_SAVESPACE_SSE+4*YMM_SIZE
+	vmovdqa %ymm5, %fs:RTLD_SAVESPACE_SSE+5*YMM_SIZE
+	vmovdqa %ymm6, %fs:RTLD_SAVESPACE_SSE+6*YMM_SIZE
+	vmovdqa %ymm7, %fs:RTLD_SAVESPACE_SSE+7*YMM_SIZE
+	ret
+L(no_avx5):
+# endif
+# define YMM_SIZE 16
+	movdqa	%xmm0, %fs:RTLD_SAVESPACE_SSE+0*XMM_SIZE
+	movdqa	%xmm1, %fs:RTLD_SAVESPACE_SSE+1*XMM_SIZE
+	movdqa	%xmm2, %fs:RTLD_SAVESPACE_SSE+2*XMM_SIZE
+	movdqa	%xmm3, %fs:RTLD_SAVESPACE_SSE+3*XMM_SIZE
+	movdqa	%xmm4, %fs:RTLD_SAVESPACE_SSE+4*XMM_SIZE
+	movdqa	%xmm5, %fs:RTLD_SAVESPACE_SSE+5*XMM_SIZE
+	movdqa	%xmm6, %fs:RTLD_SAVESPACE_SSE+6*XMM_SIZE
+	movdqa	%xmm7, %fs:RTLD_SAVESPACE_SSE+7*XMM_SIZE
+	ret
+	cfi_endproc
+	.size _dl_x86_64_save_sse, .-_dl_x86_64_save_sse
+
+
+	.globl _dl_x86_64_restore_sse
+	.type _dl_x86_64_restore_sse, @function
+	.align 16
+	cfi_startproc
+_dl_x86_64_restore_sse:
+# ifdef HAVE_AVX_SUPPORT
+	cmpl	$0, L(have_avx)(%rip)
+	js	L(no_avx6)
+
+	vmovdqa %fs:RTLD_SAVESPACE_SSE+0*YMM_SIZE, %ymm0
+	vmovdqa %fs:RTLD_SAVESPACE_SSE+1*YMM_SIZE, %ymm1
+	vmovdqa %fs:RTLD_SAVESPACE_SSE+2*YMM_SIZE, %ymm2
+	vmovdqa %fs:RTLD_SAVESPACE_SSE+3*YMM_SIZE, %ymm3
+	vmovdqa %fs:RTLD_SAVESPACE_SSE+4*YMM_SIZE, %ymm4
+	vmovdqa %fs:RTLD_SAVESPACE_SSE+5*YMM_SIZE, %ymm5
+	vmovdqa %fs:RTLD_SAVESPACE_SSE+6*YMM_SIZE, %ymm6
+	vmovdqa %fs:RTLD_SAVESPACE_SSE+7*YMM_SIZE, %ymm7
+	ret
+L(no_avx6):
+# endif
+	movdqa	%fs:RTLD_SAVESPACE_SSE+0*XMM_SIZE, %xmm0
+	movdqa	%fs:RTLD_SAVESPACE_SSE+1*XMM_SIZE, %xmm1
+	movdqa	%fs:RTLD_SAVESPACE_SSE+2*XMM_SIZE, %xmm2
+	movdqa	%fs:RTLD_SAVESPACE_SSE+3*XMM_SIZE, %xmm3
+	movdqa	%fs:RTLD_SAVESPACE_SSE+4*XMM_SIZE, %xmm4
+	movdqa	%fs:RTLD_SAVESPACE_SSE+5*XMM_SIZE, %xmm5
+	movdqa	%fs:RTLD_SAVESPACE_SSE+6*XMM_SIZE, %xmm6
+	movdqa	%fs:RTLD_SAVESPACE_SSE+7*XMM_SIZE, %xmm7
+	ret
+	cfi_endproc
+	.size _dl_x86_64_restore_sse, .-_dl_x86_64_restore_sse
+#endif
diff --git a/sysdeps/x86_64/tst-xmmymm.sh b/sysdeps/x86_64/tst-xmmymm.sh
index a576e7da0d..da8af7e686 100755
--- a/sysdeps/x86_64/tst-xmmymm.sh
+++ b/sysdeps/x86_64/tst-xmmymm.sh
@@ -59,10 +59,11 @@ for f in $tocheck; do
   objdump -d "$objpfx"../*/"$f" |
   awk 'BEGIN { last="" } /^[[:xdigit:]]* <[_[:alnum:]]*>:$/ { fct=substr($2, 2, length($2)-3) } /,%[xy]mm[[:digit:]]*$/ { if (last != fct) { print fct; last=fct} }' |
   while read fct; do
-    if test "$fct" != "_dl_runtime_profile"; then
-      echo "function $fct in $f modifies xmm/ymm" >> "$tmp"
-      result=1
+    if test "$fct" = "_dl_runtime_profile" -o "$fct" = "_dl_x86_64_restore_sse"; then
+      continue;
     fi
+    echo "function $fct in $f modifies xmm/ymm" >> "$tmp"
+    result=1
   done
 done