[PATCH] rxrpc/proc: size address buffers for %pISpc output

[Anderson Nascimento]

On 4/6/26 3:10 AM, Pengpeng Hou wrote:
> Hi,
>
> Yes. My original changelog example was too loose, and your quick test is
> right for a fully expanded plain IPv6 form such as
>
>    [ffff:ffff:ffff:ffff:ffff:ffff:ffff:ffff]:65535
>
> That form is only 47 visible characters, so it fits in the current
> char[50] buffers.
>
> The reason I still think the bug is real is the current %pISpc
> implementation in lib/vsprintf.c.
>
> For AF_INET6, %pISpc goes through ip6_addr_string_sa(), and the compressed
> path uses ip6_compressed_string(). That helper switches to a dotted-quad
> tail not only for v4mapped addresses, but also for ISATAP addresses:
>
>    useIPv4 = ipv6_addr_v4mapped(&in6) || ipv6_addr_is_isatap(&in6);
>
> So a current-tree case such as
>
>    [ffff:ffff:ffff:ffff:0:5efe:255.255.255.255]:65535
>
> is possible. That string is 50 visible characters, i.e. 51 bytes
> including the trailing NUL, which does not fit in the existing char[50]
> buffers used by the rxrpc procfs helpers.
>
> So I agree the example in my changelog should be corrected, but I do not
> think the underlying bug goes away. The claim should be framed around the
> ISATAP case rather than the plain IPv6 or mapped-v4 examples I used
> originally.
>
> If that makes sense, I can resend with the changelog corrected to cite the
> actual maximum case explicitly.
>
> Thanks,
> Pengpeng

Thanks! I confirm the ISATAP case really triggers the issue. It ends up 
writing 51 bytes, including the NULL byte.

$ cat /proc/net/rxrpc/locals
Proto Local                                           Use Act RxQ
UDP   [ffff:ffff:ffff:ffff:0:5efe:255.255.255.255]:65   1   1   0
$

Regards,

Anderson Nascimento