[SLOF] [PATCH 1/2] ping: use gateway address for routing

[Nikunj A Dadhania]

Thomas Huth <thuth at redhat.com> writes:

> On 22.04.2016 09:39, Nikunj A Dadhania wrote:
>> Thomas Huth <thuth at redhat.com> writes:
> ...
>>> Sorry again for not using the right terms ... I meant the specifying the
>>> prefix length (and thus the netmask) with a CIDR prefix notation, see
>>> https://en.wikipedia.org/wiki/Classless_Inter-Domain_Routing#CIDR_notation
>>>
>>> That means, it might maybe be useful to support specifying the client
>>> address with a prefix length like this:
>>>
>>>  load net:192.168.3.2,filename,192.168.1.10/24
>>>
>>> What do you think?
>> 
>> So that will be client-ip/nn, we need to take care we dont break old
>> cases. how about
>> 
>> Currently, its this
>> 
>> [bootp,ipv6,dhcp,]siaddr,filename,ciaddr,giaddr,bootp-retries,tftp-retries
>> 
>> Change that to:
>> 
>> [bootp,ipv6,dhcp,]siaddr,filename,ciaddr,giaddr,bootp-retries,tftp-retries,netmask
>> 
>> One more argument at the end.
>
> That would of course work, too. But actually, I already always have a
> hard time to remember the current order of the arguments here ... adding
> some more does not sound too appealing to me.
>
> There is one more thing to consider: IPv6. Currently, the IPv6 code
> silently assumes a prefix length of 64 for all addresses (I think),
> which works in most cases, but is technically also wrong, of course. So
> we might need a way to specify the prefix length / netmask for IPv6,
> too, one day. And for IPv6, you certainly don't want to type the whole
> netmask but use the /prefixlen notation instead. So I'd recommend to use
> that for IPv4 here, too, to keep both as close as possible.

If we could handle this without breaking the current parsing logic, that
should be fine.

Regards
Nikunj