[3.16.y-ckt stable] Patch "sparc64: Fix several bugs in memmove()." has been added to staging queue

Luis Henriques luis.henriques at canonical.com
Tue Mar 24 15:35:31 UTC 2015

This is a note to let you know that I have just added a patch titled

    sparc64: Fix several bugs in memmove().

to the linux-3.16.y-queue branch of the 3.16.y-ckt extended stable tree 
which can be found at:


This patch is scheduled to be released in version 3.16.7-ckt9.

If you, or anyone else, feels it should not be added to this tree, please 
reply to this email.

For more information about the 3.16.y-ckt tree, see



>From f27d9c417050dc67f1c9cd254f8a1b34783803ee Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem at davemloft.net>
Date: Mon, 23 Mar 2015 09:22:10 -0700
Subject: sparc64: Fix several bugs in memmove().

commit 2077cef4d5c29cf886192ec32066f783d6a80db8 upstream.

Firstly, handle zero length calls properly.  Believe it or not there
are a few of these happening during early boot.

Next, we can't just drop to a memcpy() call in the forward copy case
where dst <= src.  The reason is that the cache initializing stores
used in the Niagara memcpy() implementations can end up clearing out
cache lines before we've sourced their original contents completely.

For example, considering NG4memcpy, the main unrolled loop begins like

     load   src + 0x00
     load   src + 0x08
     load   src + 0x10
     load   src + 0x18
     load   src + 0x20
     store  dst + 0x00

Assume dst is 64 byte aligned and let's say that dst is src - 8 for
this memcpy() call.  That store at the end there is the one to the
first line in the cache line, thus clearing the whole line, which thus
clobbers "src + 0x28" before it even gets loaded.

To avoid this, just fall through to a simple copy only mildly
optimized for the case where src and dst are 8 byte aligned and the
length is a multiple of 8 as well.  We could get fancy and call
GENmemcpy() but this is good enough for how this thing is actually

Reported-by: David Ahern <david.ahern at oracle.com>
Reported-by: Bob Picco <bpicco at meloft.net>
Signed-off-by: David S. Miller <davem at davemloft.net>
Signed-off-by: Luis Henriques <luis.henriques at canonical.com>
 arch/sparc/lib/memmove.S | 35 ++++++++++++++++++++++++++++++++---
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/arch/sparc/lib/memmove.S b/arch/sparc/lib/memmove.S
index b7f6334e159f..857ad4f8905f 100644
--- a/arch/sparc/lib/memmove.S
+++ b/arch/sparc/lib/memmove.S
@@ -8,9 +8,11 @@

 ENTRY(memmove) /* o0=dst o1=src o2=len */
-	mov		%o0, %g1
+	brz,pn		%o2, 99f
+	 mov		%o0, %g1
 	cmp		%o0, %o1
-	bleu,pt		%xcc, memcpy
+	bleu,pt		%xcc, 2f
 	 add		%o1, %o2, %g7
 	cmp		%g7, %o0
 	bleu,pt		%xcc, memcpy
@@ -24,7 +26,34 @@ ENTRY(memmove) /* o0=dst o1=src o2=len */
 	stb		%g7, [%o0]
 	bne,pt		%icc, 1b
 	 sub		%o0, 1, %o0
 	 mov		%g1, %o0
+	/* We can't just call memcpy for these memmove cases.  On some
+	 * chips the memcpy uses cache initializing stores and when dst
+	 * and src are close enough, those can clobber the source data
+	 * before we've loaded it in.
+	 */
+2:	or		%o0, %o1, %g7
+	or		%o2, %g7, %g7
+	andcc		%g7, 0x7, %g0
+	bne,pn		%xcc, 4f
+	 nop
+3:	ldx		[%o1], %g7
+	add		%o1, 8, %o1
+	subcc		%o2, 8, %o2
+	add		%o0, 8, %o0
+	bne,pt		%icc, 3b
+	 stx		%g7, [%o0 - 0x8]
+	ba,a,pt		%xcc, 99b
+4:	ldub		[%o1], %g7
+	add		%o1, 1, %o1
+	subcc		%o2, 1, %o2
+	add		%o0, 1, %o0
+	bne,pt		%icc, 4b
+	 stb		%g7, [%o0 - 0x1]
+	ba,a,pt		%xcc, 99b

More information about the kernel-team mailing list