[PATCH] net: ipv6: Add support for default gateway discovery.

[Vyacheslav V. Mitrofanov]

On Thu, 2023-03-02 at 08:58 -0800, emohandesi at linux.microsoft.com
wrote:
> 
> From: Ehsan Mohandesi <emohandesi at microsoft.com>
> 
> In IPv6, the default gateway and prefix length are determined by
> receiving
> a router advertisement as defined in -
> https://www.rfc-editor.org/rfc/rfc4861.
> 
> Add support for sending router solicitation (RS) and processing
> router
> advertisements (RA).
> 
> If the RA has prefix info option and following conditions are met,
> then
> gatewayip6 and net_prefix_length of ip6addr env variables are
> initialized.
> These are later consumed by IPv6 code for non-local destination IP.
> 
> - "Router Lifetime" != 0
> - Prefix is NOT link-local prefix (0xfe80::/10)
> - L flag is 1
> - "Valid Lifetime" != 0
> 
> Timing Parameters:
> - MAX_RTR_SOLICITATION_DELAY (0-1s)
> - RTR_SOLICITATION_INTERVAL (4s) (min retransmit delay)
> - MAX_RTR_SOLICITATIONS (3 RS transmissions)
> 
> The functionality is enabled by CONFIG_IPV6_ROUTER_DISCOVERY and
> invoked
> automatically from net_init_loop().
> 
> Signed-off-by: Ehsan Mohandesi <emohandesi at microsoft.com>
> 
> Conflicts:
>         cmd/Kconfig
>         include/net.h
>         net/net.c
> ---
>  cmd/Kconfig     |   7 ++
>  include/ndisc.h |  23 ++++++
>  include/net.h   |   2 +-
>  include/net6.h  |  40 ++++++++++
>  net/ndisc.c     | 243
> +++++++++++++++++++++++++++++++++++++++++++++++++++++---
>  net/net.c       |  23 +++++-
>  net/net6.c      |   1 +
>  7 files changed, 327 insertions(+), 12 deletions(-)
> 

I reviewed this patch and it looks good. I have no critical remarks,
only some small notes. 

I've tested it on SiFive Unmatched board.


> 
> +config IPV6_ROUTER_DISCOVERY
> +       bool "Do router discovery"
> +       depends on IPV6
> +       help
> +         Will automatically perform router solicitation on first
> IPv6
> +         network operation
> +
>  endif
> 
I think it is better to write sth like Do IPv6 router discovery because
IPv4 has also router discovery protocol and it could lead to
misunderstanding


> 
> net_set_timeout_handler(0, 0);
> 
Maybe net_set_timeout_handler(0, NULL); is better



> +/*
> + * validate_ra() - Validate the router advertisement message.
> + *
> + * @ip6:
> + * @len: Length of the router advertisement packet
> + *
> + * Check if the router advertisement message is valid. Conditions
> are
> + * according to RFC 4861 section 6.1.2. Validation of Router
> Advertisement
> + * Messages.
> + *
> + * Return: true if the message is valid and false if it is invalid.
> + */
> +static bool validate_ra(struct ip6_hdr *ip6, int len)
> +{
> +       struct icmp6hdr *icmp = (struct icmp6hdr *)(ip6 + 1);
> +
> +       /* ICMP length (derived from the IP length) should be 16 or
> more octets. */
> +       if (ip6->payload_len < 16)
> +               return false;
> +
> +       /* Source IP Address should be a valid link-local address. */
> +       if ((ntohs(ip6->saddr.s6_addr16[0]) & IPV6_LINK_LOCAL_MASK)
> !=
> +           IPV6_LINK_LOCAL_PREFIX)
> +               return false;
> +
> +       /*
> +        * The IP Hop Limit field should have a value of 255, i.e.,
> the packet
> +        * could not possibly have been forwarded by a router.
> +        */
> +       if (ip6->hop_limit != 255)
> +               return false;
> +
Unicast hop limit only?

> diff --git a/net/net.c b/net/net.c
> index c9a749f..39f0b81 100644
> --- a/net/net.c
> +++ b/net/net.c
> @@ -24,7 +24,7 @@
>   *                     - name of bootfile
>   *     Next step:      ARP
>   *
> - * LINK_LOCAL:
> + * LINKLOCAL:
> 
Maybe it is better to move to other patch?!


Reviewed-by: Viacheslav Mitrofanov <v.v.mitrofanov at yadro.com>
Tested-by: Viacheslav Mitrofanov <v.v.mitrofanov at yadro.com>