Re: [PATCH] rose_dev: fix memcpy-bug in rose_set_mac_address

Re: [PATCH] rose_dev: fix memcpy-bug in rose_set_mac_address

[Daniel Borkmann]

David Miller <davem@davemloft.net> wrote:

>> This patch fixes a small bug in rose_set_mac_address. If the current and new
>> MAC addresses match, then nothing needs to be done. However memcpy was used
>> instead of memcmp for comparison.
>>
>> The patch is against the latest net-tree.
>>
>> Signed-off-by: Daniel Borkmann <daniel.borkmann@tik.ee.ethz.ch>
>
> You're breaking this code, not fixing it.
>
> If you don't keep the memcpy, then the calls right below which setup
> the loopback node will use the previous device address not the new
> one being configured.

David, thanks for your feedback! So here's a small rework of the patch:

If both addresses equal, nothing needs to be done. If the device is down,
then we simply copy the new address to dev->dev_addr. If the device is up,
then we add another loopback device with the new address, and if that does
not fail, we remove the loopback device with the old address. And only
then, we update the dev->dev_addr.

Signed-off-by: Daniel Borkmann <daniel.borkmann@tik.ee.ethz.ch>
---
 net/rose/rose_dev.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/net/rose/rose_dev.c b/net/rose/rose_dev.c
index 178ff4f..2679507 100644
--- a/net/rose/rose_dev.c
+++ b/net/rose/rose_dev.c
@@ -96,11 +96,11 @@ static int rose_set_mac_address(struct net_device *dev, void *addr)
 	struct sockaddr *sa = addr;
 	int err;
 
-	if (!memcpy(dev->dev_addr, sa->sa_data, dev->addr_len))
+	if (!memcmp(dev->dev_addr, sa->sa_data, dev->addr_len))
 		return 0;
 
 	if (dev->flags & IFF_UP) {
-		err = rose_add_loopback_node((rose_address *)dev->dev_addr);
+		err = rose_add_loopback_node((rose_address *)sa->sa_data);
 		if (err)
 			return err;
 

Re: [PATCH] rose_dev: fix memcpy-bug in rose_set_mac_address

[David Miller]

From: Daniel Borkmann <danborkmann@iogearbox.net>
Date: Wed, 28 Mar 2012 10:47:43 +0200

> If both addresses equal, nothing needs to be done. If the device is down,
> then we simply copy the new address to dev->dev_addr. If the device is up,
> then we add another loopback device with the new address, and if that does
> not fail, we remove the loopback device with the old address. And only
> then, we update the dev->dev_addr.
> 
> Signed-off-by: Daniel Borkmann <daniel.borkmann@tik.ee.ethz.ch>

Applied and queued up for -stable.

Re: [PATCH] rose_dev: fix memcpy-bug in rose_set_mac_address

[David Miller]

From: Daniel Borkmann <danborkmann@iogearbox.net>
Date: Tue, 27 Mar 2012 15:27:56 +0200

> This patch fixes a small bug in rose_set_mac_address. If the current and new
> MAC addresses match, then nothing needs to be done. However memcpy was used
> instead of memcmp for comparison.
> 
> The patch is against the latest net-tree.
> 
> Signed-off-by: Daniel Borkmann <daniel.borkmann@tik.ee.ethz.ch>

You're breaking this code, not fixing it.

If you don't keep the memcpy, then the calls right below which setup
the loopback node will use the previous device address not the new
one being configured.

[PATCH] rose_dev: fix memcpy-bug in rose_set_mac_address

[Daniel Borkmann]

This patch fixes a small bug in rose_set_mac_address. If the current and new
MAC addresses match, then nothing needs to be done. However memcpy was used
instead of memcmp for comparison.

The patch is against the latest net-tree.

Signed-off-by: Daniel Borkmann <daniel.borkmann@tik.ee.ethz.ch>
Cc: Ralf Baechle <ralf@linux-mips.org>
---
 net/rose/rose_dev.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/net/rose/rose_dev.c b/net/rose/rose_dev.c
index 178ff4f..3c5ce5e 100644
--- a/net/rose/rose_dev.c
+++ b/net/rose/rose_dev.c
@@ -96,7 +96,7 @@ static int rose_set_mac_address(struct net_device *dev, void *addr)
 	struct sockaddr *sa = addr;
 	int err;
 
-	if (!memcpy(dev->dev_addr, sa->sa_data, dev->addr_len))
+	if (!memcmp(dev->dev_addr, sa->sa_data, dev->addr_len))
 		return 0;
 
 	if (dev->flags & IFF_UP) {