[PATCH net] powerpc: use big endian to hash len and proto in csum_ipv6_magic

The function csum_ipv6_magic doesn't convert len and proto to big
endian before doing ipv6 csum hash, which is not consistent with
RFC and other arches.

Jianlin found it when ICMPv6 packets from other hosts were dropped
in the powerpc64 system.

This patch is to fix it by using instruction 'lwbrx' to do this
conversion in powerpc32/64 csum_ipv6_magic.

Fixes: e9c4943a107b ("powerpc: Implement csum_ipv6_magic in assembly")
Reported-by: Jianlin Shi <redacted>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
---
 arch/powerpc/lib/checksum_32.S | 4 ++++
 arch/powerpc/lib/checksum_64.S | 4 ++++
 2 files changed, 8 insertions(+)

diff --git a/arch/powerpc/lib/checksum_32.S b/arch/powerpc/lib/checksum_32.S
index aa22406..7d3446e 100644
--- a/arch/powerpc/lib/checksum_32.S
+++ b/arch/powerpc/lib/checksum_32.S

@@ -325,6 +325,10 @@ _GLOBAL(csum_ipv6_magic)
 	adde	r0, r0, r9
 	lwz	r11, 12(r4)
 	adde	r0, r0, r10
+	STWX_BE	r5, 0, r1
+	lwz	r5, 0(r1)
+	STWX_BE	r6, 0, r1
+	lwz	r6, 0(r1)
 	add	r5, r5, r6	/* assumption: len + proto doesn't carry */
 	adde	r0, r0, r11
 	adde	r0, r0, r5

diff --git a/arch/powerpc/lib/checksum_64.S b/arch/powerpc/lib/checksum_64.S
index 886ed94..302e732 100644
--- a/arch/powerpc/lib/checksum_64.S
+++ b/arch/powerpc/lib/checksum_64.S

@@ -439,6 +439,10 @@ EXPORT_SYMBOL(csum_partial_copy_generic)
 _GLOBAL(csum_ipv6_magic)
 	ld	r8, 0(r3)
 	ld	r9, 8(r3)
+	STWX_BE	r5, 0, r1
+	lwz	r5, 0(r1)
+	STWX_BE	r6, 0, r1
+	lwz	r6, 0(r1)
 	add	r5, r5, r6
 	addc	r0, r8, r9
 	ld	r10, 0(r4)

-- 
2.1.0

Xin Long [off-list ref] a =C3=A9crit=C2=A0:

2.S

PPC32 doesn't support little endian, so nothing to do here.

4.S

This is overkill. For LE it should be enough to rotate r5 by 8 bits=20=20
after=20the sum. Best place to do it would be after ld r11 I think.

Christophe

 	addc	r0, r8, r9
 	ld	r10, 0(r4)
--
2.1.0

From: Xin Long <lucien.xin@gmail.com>
Date: Sat,  8 Sep 2018 18:15:12 +0800

The function csum_ipv6_magic doesn't convert len and proto to big
endian before doing ipv6 csum hash, which is not consistent with
RFC and other arches.

Jianlin found it when ICMPv6 packets from other hosts were dropped
in the powerpc64 system.

This patch is to fix it by using instruction 'lwbrx' to do this
conversion in powerpc32/64 csum_ipv6_magic.

Fixes: e9c4943a107b ("powerpc: Implement csum_ipv6_magic in assembly")
Reported-by: Jianlin Shi <redacted>
Signed-off-by: Xin Long <lucien.xin@gmail.com>

Xin, please address the feedback you were given.

Thank you.

Le 12/09/2018 à 08:01, David Miller a écrit :

From: Xin Long <lucien.xin@gmail.com>
Date: Sat,  8 Sep 2018 18:15:12 +0800

quoted

The function csum_ipv6_magic doesn't convert len and proto to big
endian before doing ipv6 csum hash, which is not consistent with
RFC and other arches.

Jianlin found it when ICMPv6 packets from other hosts were dropped
in the powerpc64 system.

This patch is to fix it by using instruction 'lwbrx' to do this
conversion in powerpc32/64 csum_ipv6_magic.

Fixes: e9c4943a107b ("powerpc: Implement csum_ipv6_magic in assembly")
Reported-by: Jianlin Shi <redacted>
Signed-off-by: Xin Long <lucien.xin@gmail.com>

Xin, please address the feedback you were given.

I submitted an alternative fix, and Lucien Xin gave its Tested-by:

See https://patchwork.ozlabs.org/patch/967868/

Christophe

On Wed, Sep 12, 2018 at 2:01 PM David Miller [off-list ref] wrote:

From: Xin Long <lucien.xin@gmail.com>
Date: Sat,  8 Sep 2018 18:15:12 +0800

quoted

The function csum_ipv6_magic doesn't convert len and proto to big
endian before doing ipv6 csum hash, which is not consistent with
RFC and other arches.

Jianlin found it when ICMPv6 packets from other hosts were dropped
in the powerpc64 system.

This patch is to fix it by using instruction 'lwbrx' to do this
conversion in powerpc32/64 csum_ipv6_magic.

Fixes: e9c4943a107b ("powerpc: Implement csum_ipv6_magic in assembly")
Reported-by: Jianlin Shi <redacted>
Signed-off-by: Xin Long <lucien.xin@gmail.com>

Xin, please address the feedback you were given.

Christophe posted another one,
https://lore.kernel.org/patchwork/patch/983905/

Sorry, I didn't notice netdev wasn't in its CC-list.