batman-adv: generalize batman-adv icmp packet handling

  Instead of handling icmp packets only up to length of icmp_packet_rr,
the code should handle any icmp length size. Therefore the length
truncating is moved to when the packet is actually sent to userspace
(this does not support lengths longer than icmp_packet_rr yet). Longer
packets are forwarded without truncating.

This patch also cleans up some parts where the icmp header struct could
be used instead of other icmp_packet(_rr) structs to make the code more
readable.

Signed-off-by: Simon Wunderlich <siwu@hrz.tu-chemnitz.de>
---
 icmp_socket.c |   38 ++++++++++++++++++++-------
 icmp_socket.h |    2 +-
 routing.c     |   80 +++++++++++++++++++++++++++++++++++----------------------
 3 files changed, 79 insertions(+), 41 deletions(-)
  

On Friday 13 September 2013 18:08:10 Simon Wunderlich wrote:
> +/**
> + * batadv_socket_receive_packet - schedule an icmp packet to be sent to
> userspace + *  on an icmp socket.
> + * @socket_client: the socket this packet belongs to
> + * @icmph: pointer to the header of the icmp packet
> + * @icmp_len: total length of the icmp packet
> + */
>  static void batadv_socket_add_packet(struct batadv_socket_client
> *socket_client, -				     struct batadv_icmp_packet_rr 
*icmp_packet,
> +				     struct batadv_icmp_header *icmph,
>  				     size_t icmp_len)
>  {
>  	struct batadv_socket_packet *socket_packet;
> +	size_t len;
> 
>  	socket_packet = kmalloc(sizeof(*socket_packet), GFP_ATOMIC);
> 
>  	if (!socket_packet)
>  		return;
> 
> +	len = icmp_len;
> +	/* check the maximum length before filling the buffer */
> +	if (len > sizeof(socket_packet->icmp_packet))
> +		len = sizeof(socket_packet->icmp_packet);
> +
>  	INIT_LIST_HEAD(&socket_packet->list);
> -	memcpy(&socket_packet->icmp_packet, icmp_packet, icmp_len);
> +	memcpy(&socket_packet->icmp_packet, icmph, icmp_len);

Shouldn't "len" be used here ?

Besides, if we make everything generic batadv_socket_packet->icmp_packet 
should not be hard-coded to batadv_icmp_packet_rr but the largest available 
ICMP packet type ?


> +/**
> + * batadv_recv_my_icmp_packet - receive an icmp packet locally
> + * @bat_priv: the bat priv with all the soft interface information
> + * @skb: icmp packet to process
> + *
> + * Returns NET_RX_SUCCESS if the packet has been consumed or NET_RX_DROP
> + * otherwise.
> + */
>  static int batadv_recv_my_icmp_packet(struct batadv_priv *bat_priv,
> -				      struct sk_buff *skb, size_t icmp_len)
> +				      struct sk_buff *skb)
>  {
>  	struct batadv_hard_iface *primary_if = NULL;
>  	struct batadv_orig_node *orig_node = NULL;
> -	struct batadv_icmp_packet_rr *icmp_packet;
> +	struct batadv_icmp_header *icmph;
>  	int ret = NET_RX_DROP;
> 
> -	icmp_packet = (struct batadv_icmp_packet_rr *)skb->data;
> +	icmph = (struct batadv_icmp_header *)skb->data;
> 
>  	/* add data to device queue */
> -	if (icmp_packet->icmph.msg_type != BATADV_ECHO_REQUEST) {
> -		batadv_socket_receive_packet(icmp_packet, icmp_len);
> +	if (icmph->msg_type != BATADV_ECHO_REQUEST) {
> +		if (skb_linearize(skb) < 0)
> +			goto out;
> +
> +		batadv_socket_receive_packet(icmph, skb->len);
>  		goto out;
>  	}

Wouldn't it be better to dump unkown icmp types for us instead of copying 
everything to user space ?

Same is true for batadv_socket_write(). We should use the icmp header and not 
assume icmp echo.

Cheers,
Marek
  

On Tue, Oct 15, 2013 at 04:04:31PM +0800, Marek Lindner wrote:
> On Friday 13 September 2013 18:08:10 Simon Wunderlich wrote:
> > +/**
> > + * batadv_socket_receive_packet - schedule an icmp packet to be sent to
> > userspace + *  on an icmp socket.
> > + * @socket_client: the socket this packet belongs to
> > + * @icmph: pointer to the header of the icmp packet
> > + * @icmp_len: total length of the icmp packet
> > + */
> >  static void batadv_socket_add_packet(struct batadv_socket_client
> > *socket_client, -				     struct batadv_icmp_packet_rr 
> *icmp_packet,
> > +				     struct batadv_icmp_header *icmph,
> >  				     size_t icmp_len)
> >  {
> >  	struct batadv_socket_packet *socket_packet;
> > +	size_t len;
> > 
> >  	socket_packet = kmalloc(sizeof(*socket_packet), GFP_ATOMIC);
> > 
> >  	if (!socket_packet)
> >  		return;
> > 
> > +	len = icmp_len;
> > +	/* check the maximum length before filling the buffer */
> > +	if (len > sizeof(socket_packet->icmp_packet))
> > +		len = sizeof(socket_packet->icmp_packet);
> > +
> >  	INIT_LIST_HEAD(&socket_packet->list);
> > -	memcpy(&socket_packet->icmp_packet, icmp_packet, icmp_len);
> > +	memcpy(&socket_packet->icmp_packet, icmph, icmp_len);
> 
> Shouldn't "len" be used here ?
> 
> Besides, if we make everything generic batadv_socket_packet->icmp_packet 
> should not be hard-coded to batadv_icmp_packet_rr but the largest available 
> ICMP packet type ?


or we dynamically allocate a buffer of size 'len'? In this way we don't need to
change icmp_packet each time (hopefully not so many but still..)  the
"largest available ICMP packet type" changes.


> 
> 
> > +/**
> > + * batadv_recv_my_icmp_packet - receive an icmp packet locally
> > + * @bat_priv: the bat priv with all the soft interface information
> > + * @skb: icmp packet to process
> > + *
> > + * Returns NET_RX_SUCCESS if the packet has been consumed or NET_RX_DROP
> > + * otherwise.
> > + */
> >  static int batadv_recv_my_icmp_packet(struct batadv_priv *bat_priv,
> > -				      struct sk_buff *skb, size_t icmp_len)
> > +				      struct sk_buff *skb)
> >  {
> >  	struct batadv_hard_iface *primary_if = NULL;
> >  	struct batadv_orig_node *orig_node = NULL;
> > -	struct batadv_icmp_packet_rr *icmp_packet;
> > +	struct batadv_icmp_header *icmph;
> >  	int ret = NET_RX_DROP;
> > 
> > -	icmp_packet = (struct batadv_icmp_packet_rr *)skb->data;
> > +	icmph = (struct batadv_icmp_header *)skb->data;
> > 
> >  	/* add data to device queue */
> > -	if (icmp_packet->icmph.msg_type != BATADV_ECHO_REQUEST) {
> > -		batadv_socket_receive_packet(icmp_packet, icmp_len);
> > +	if (icmph->msg_type != BATADV_ECHO_REQUEST) {
> > +		if (skb_linearize(skb) < 0)
> > +			goto out;
> > +
> > +		batadv_socket_receive_packet(icmph, skb->len);
> >  		goto out;
> >  	}
> 
> Wouldn't it be better to dump unkown icmp types for us instead of copying 
> everything to user space ?

dump == drop ? in that case I agree. Delivering unknown packets to batctl may
also be dangerous.

> 
> Same is true for batadv_socket_write(). We should use the icmp header and not 
> assume icmp echo.

do you mean using the ICMP header to understand what packet it is and then
behave accordingly? Also in this case I agree with you :)


Moving to a packet type based "check" was also part of the original idea of this
generalisation (if I remember correctly), but not really needed when looking at
avoiding further compatibility breakage (this should be the reason why this
"feature" is not part of this patch).



I am not replying on Simon's behalf, I was just eager to share my 2 cents.

Cheers,
  

On Tuesday 15 October 2013 21:20:23 Antonio Quartulli wrote:
> > Besides, if we make everything generic batadv_socket_packet->icmp_packet
> > should not be hard-coded to batadv_icmp_packet_rr but the largest
> > available
> > ICMP packet type ?
> 
> or we dynamically allocate a buffer of size 'len'? In this way we don't need
> to change icmp_packet each time (hopefully not so many but still..)  the
> "largest available ICMP packet type" changes.

Allocating a dynamic buffer does not solve the underlying issue. At some point 
we would want to check the packet size - either through sizeof(socket_packet-
>icmp_packet) or a macro or whatever. 
Take a look at batadv_max_header_len() for an idea how to address the matter.


> > Wouldn't it be better to dump unkown icmp types for us instead of copying
> > everything to user space ?
> 
> dump == drop ? in that case I agree. Delivering unknown packets to batctl
> may also be dangerous.

Yes, that is what I was talking about.


> > Same is true for batadv_socket_write(). We should use the icmp header and
> > not assume icmp echo.
> 
> do you mean using the ICMP header to understand what packet it is and then
> behave accordingly? Also in this case I agree with you :)

Yap.


> Moving to a packet type based "check" was also part of the original idea of
> this generalisation (if I remember correctly), but not really needed when
> looking at avoiding further compatibility breakage (this should be the
> reason why this "feature" is not part of this patch).

I agree - it is not strictly needed but would fit since the rest of the patch 
works into the same direction.

Cheers,
Marek
  

On Wed, Oct 16, 2013 at 01:59:18PM +0800, Marek Lindner wrote:
> On Tuesday 15 October 2013 21:20:23 Antonio Quartulli wrote:
> > > Besides, if we make everything generic batadv_socket_packet->icmp_packet
> > > should not be hard-coded to batadv_icmp_packet_rr but the largest
> > > available
> > > ICMP packet type ?
> > 
> > or we dynamically allocate a buffer of size 'len'? In this way we don't need
> > to change icmp_packet each time (hopefully not so many but still..)  the
> > "largest available ICMP packet type" changes.
> 
> Allocating a dynamic buffer does not solve the underlying issue. At some point 
> we would want to check the packet size - either through sizeof(socket_packet-
> >icmp_packet) or a macro or whatever. 
> Take a look at batadv_max_header_len() for an idea how to address the matter.
> 

Right. Thanks for the clarification


Cheers,