Re: [PATCH v4 3/4] dhcp: Add --dhcp-opt with option table and value parser

[Stefano Brivio <sbrivio@redhat.com>]

On Fri, 19 Jun 2026 12:59:10 +0530
Anshu Kumari <anskuma@redhat.com> wrote:

> On Fri, Jun 19, 2026 at 9:24 AM David Gibson <david@gibson.dropbear.id.au>
> wrote:
> 
> > On Wed, Jun 17, 2026 at 06:52:37PM +0530, Anshu Kumari wrote:  
> > > Introduce the --dhcp-opt flag that allows setting arbitrary DHCP
> > > options from command-line in the form [--dhcp-opt CODE,VALUE].
> > >
> > > Add a type lookup table mapping option codes to RFC 2132 value types
> > > (IPv4, IPv4 list, integer, string) and dhcp_opt_parse() to convert
> > > CLI strings to binary wire format.  Parsed options are stored in
> > > struct ctx and injected into DHCP replies.  If the same option code
> > > is given more than once, the last value wins.
> > >
> > > Link: https://bugs.passt.top/show_bug.cgi?id=192
> > > Signed-off-by: Anshu Kumari <anskuma@redhat.com>
> > > ---
> > > v4:
> > >   - Renamed custom_opts to dhcp_opts, 256 entries indexed by option
> > >     code, removed MAX_CUSTOM_DHCP_OPTS and count field.
> > >   - Changed str buffer from 256 to 255 bytes.
> > >   - Moved function to conf.c as static conf_dhcp_option(), renamed
> > >     from dhcp_add_option().
> > >   - Made dhcp_opt_parse() non-static, declared in dhcp.h
> > >   - Dropped val/len from ctx struct; conf_dhcp_option() validates
> > >     with temp buffer, dhcp() parses str directly into opts[] at
> > >     reply time.  
> >
> > Hmm.  So each option is parsed twice.  What prevents you from parsing
> > directly into the opts[] array at conf() time?
> 
> The first parse acts as a validation step to check that the user has
> entered the
> correct value format for the option. Without it, if the user passes
> something
> like *--dhcp-opt 3,notanip*, the error would surface only when the first
> DHCP client connects, not at startup.
> 
> I think it's better to fail during startup if correct value format is not
> entered in command-line rather than failing at later
> stage during reply time.

Sure, I would say we have to.

But I think David was suggesting (that would be my suggestion as well)
that you actually validate values right away, and store them directly
into opts[], instead of having two steps and keeping a c->dhcp_opts[]
array in parallel.

The downside of a single parsing step, without duplicating values, is
that you would need to render values back from opts[] to print them in
conf_print(). But it should be relatively trivial. It might actually be
less code than it is now.

-- 
Stefan