Re: [ 06/79] net: Swap ver and type in pppoe_hdr

[Ben Hutchings]

On Fri, 2013-07-26 at 13:46 -0700, Greg Kroah-Hartman wrote:
> 3.10-stable review patch.  If anyone has any objections, please let me know.
> 
> ------------------
> 
> From: Changli Gao <xiaosuo@xxxxxxxxx>
> 
> [ Upstream commit b1a5a34bd0b8767ea689e68f8ea513e9710b671e ]
> 
> Ver and type in pppoe_hdr should be swapped as defined by RFC2516
> section-4.
> 
> Signed-off-by: David S. Miller <davem@xxxxxxxxxxxxx>
> Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>

I'm not so sure this is suitable for stable.  It doesn't seem to be
helpful to either userland or kernel code.

In the kernel, we were checking that ver == 1 and type == 1 thus this
wasn't hurting us.

If userland does anything more with these fields then it's probably
already working around the swapped fields.  It can work with both old
and new headers by doing:

#if LINUX_VERSION_CODE >= KERNEL_VERSION(3,11,0)
	/* use pppoe_hdr fields normally */
#else
	/* definitions were wrong before, work around it */
#endif

If this is backported to stable, there is no usable version test.
Userland will have to bypass the bitfields entirely:

	type = *(uint8_t *)hdr & 0xf;
	ver = *(uint8_t *)hdr >> 4;

Ben.

> ---
>  include/uapi/linux/if_pppox.h |    4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> --- a/include/uapi/linux/if_pppox.h
> +++ b/include/uapi/linux/if_pppox.h
> @@ -135,11 +135,11 @@ struct pppoe_tag {
>  
>  struct pppoe_hdr {
>  #if defined(__LITTLE_ENDIAN_BITFIELD)
> -	__u8 ver : 4;
>  	__u8 type : 4;
> +	__u8 ver : 4;
>  #elif defined(__BIG_ENDIAN_BITFIELD)
> -	__u8 type : 4;
>  	__u8 ver : 4;
> +	__u8 type : 4;
>  #else
>  #error	"Please fix <asm/byteorder.h>"
>  #endif


-- 
Ben Hutchings
Once a job is fouled up, anything done to improve it makes it worse.

Attachment: signature.asc
Description: This is a digitally signed message part