Re: [PATCH v4 2/2] dhcpv6: Inject custom options into DHCPv6 replies

[Stefano Brivio <sbrivio@redhat.com>]

On Tue,  6 Oct 2026 14:34:53 +0530
Anshu Kumari <anskuma@redhat.com> wrote:

> Append user-specified options from --dhcpv6-opt to DHCPv6 reply
> messages.  Options that exceed available buffer space are skipped
> with a warning.
> 
> Link: https://bugs.passt.top/show_bug.cgi?id=192
> Signed-off-by: Anshu Kumari <anskuma@redhat.com>
> ---
> v4:
>   - warn() instead of debug() when options don't fit in the reply.
>   - doc comment for dhcpv6_user_opts_fill() updated.
> 
> v3:
>   - rename dhcpv6_custom_opts_fill() to dhcpv6_user_opts_fill().
>   - Iterate directly over the pre-parsed binary dhcpv6_user_opts[]
>     array instead of re-parsing string values at reply time.
> 
> v2:
>   - Updated dhcpv6_custom_opts_fill() to parse str at reply time
>     using dhcpv6_opt_parse() instead of copying cached val/len.
> ---
>  dhcpv6.c | 36 ++++++++++++++++++++++++++++++++++++
>  1 file changed, 36 insertions(+)
> 
> diff --git a/dhcpv6.c b/dhcpv6.c
> index abc45813..a0545406 100644
> --- a/dhcpv6.c
> +++ b/dhcpv6.c
> @@ -879,6 +879,41 @@ static size_t dhcpv6_client_fqdn_fill(const struct iov_tail *data,
>  	return offset + sizeof(struct opt_hdr) + opt_len;
>  }
>  
> +/**
> + * dhcpv6_user_opts_fill() - Append user-specified options to reply
> + * @buf:	Response message buffer
> + * @offset:	Current offset in buffer
> + *
> + * Return: offset after appended user options
> + */
> +static size_t dhcpv6_user_opts_fill(char *buf, size_t offset)
> +{
> +	unsigned int i;
> +
> +	for (i = 0; i < ARRAY_SIZE(dhcpv6_user_opts); i++) {
> +		int vlen = dhcpv6_user_opts[i].len;
> +		struct opt_hdr *hdr;
> +
> +		if (!vlen)
> +			continue;
> +
> +		if (offset + sizeof(struct opt_hdr) + vlen > OPT_MAX_SIZE) {
> +			warn("DHCPv6: option %u doesn't fit, skipping", i);

I've seen Laurent's comment in:

  https://archives.passt.top/passt-dev/6fcca48c-a992-436f-a308-061ae0017d4f@redhat.com/

but the issue with a warn() call here is that the guest could try to
flood logs that way.  You could rate-limit it by passing the timestamp
to here via dhcpv6(), it's just one caller to modify, it doesn't look
that bad.

> +			continue;
> +		}
> +
> +		hdr = (struct opt_hdr *)(buf + offset);
> +		hdr->t = htons(i);
> +		hdr->l = htons(vlen);
> +		offset += sizeof(struct opt_hdr);
> +
> +		memcpy(buf + offset, dhcpv6_user_opts[i].val, vlen);
> +		offset += vlen;
> +	}
> +
> +	return offset;
> +}
> +
>  /**
>   * dhcpv6() - Check if this is a DHCPv6 message, reply as needed
>   * @c:		Execution context
> @@ -1017,6 +1052,7 @@ int dhcpv6(struct ctx *c, struct iov_tail *data,
>  	    sizeof(struct opt_hdr) + ntohs(client_id->l);
>  	n = dhcpv6_dns_fill(c, (char *)&resp, n);
>  	n = dhcpv6_client_fqdn_fill(data, c, (char *)&resp, n);
> +	n = dhcpv6_user_opts_fill((char *)&resp, n);
>  
>  	resp.hdr.xid = mh->xid;
>  

-- 
Stefano