[RFC PATCH 0/5] LWIP stack integration

[Ilias Apalodimas]

Hi Maxim

On Mon, 22 May 2023 at 12:01, Maxim Uvarov <maxim.uvarov at linaro.org> wrote:
>
> My measurements for binary after LTO looks like:
>
> U-boot WGET |  LWIP WGET + ping |  LWIP WGET| diff bytes| diff %
> 870728            |  915000                    | 912560          | 41832    | 4.8


I think you'll need to analyze that a bit more.  First of all I don't
think the '+ping' tab is useful.  What is is trying to achieve?
- How was LWIP compiled?
- Was ipv6 supported?
- Can we strip it down even further?

 In general please give as much information as you can with what we
gain in functionality from LWIP with those extra bytes of code.

Thanks
/Ilias

> BR,
> Maxim.
>
> On Fri, 19 May 2023 at 09:52, Tom Rini <trini at konsulko.com> wrote:
>>
>> On Fri, May 19, 2023 at 04:17:06PM +0300, Ilias Apalodimas wrote:
>> > Hi Tom,
>> >
>> > Apologies for being late to the party
>> >
>> > > On Thu, May 11, 2023 at 09:52:04AM -0400, Tom Rini wrote:
>> > > On Fri, May 05, 2023 at 10:25:24AM +0000, Maxim Uvarov wrote:
>> > >
>> > > > Greetings,
>> > > >
>> > > > This RFC patchset is an attempt to try to use an already existing IP network stack inside U-boot.
>> > > > U-Boot recently got basic TCP/IP support, implementing wget, but in order to get a full IP stack
>> > > > with new features (e.g ipv6), it would be preferable to use an established embedded ip library,
>> > > > instead of rewriting the code from scratch.
>> > > >
>> > > > For this experiment LWIP network stack was selected:
>> > > > https://savannah.nongnu.org/git/?group=lwip
>> > > >
>> > > > LWIP main features include:
>> > > > - Protocols: IP, IPv6, ICMP, ND, MLD, UDP, TCP, IGMP, ARP, PPPoS, PPPoE
>> > > > - DHCP client, DNS client (incl. mDNS hostname resolver), AutoIP/APIPA (Zeroconf),
>> > > >   SNMP agent (v1, v2c, v3, private MIB support & MIB compiler)
>> > > > - APIs: specialized APIs for enhanced performance, optional Berkeley-alike socket API
>> > > > - Extended features: IP forwarding over multiple network interfaces, TCP congestion control,
>> > > >   RTT estimation and fast recovery/fast retransmit
>> > > > - Addon applications: HTTP(S) server, SNTP client, SMTP(S) client, ping, NetBIOS nameserver,
>> > > >   mDNS responder, MQTT client, TFTP server.
>> > > >
>> > > > This RFC work is a demo to enable lwIP (lightweight IP) which is a widely used open-source
>> > > > TCP/IP stack designed for embedded systems for U-boot. That will allow using already
>> > > > written network applications for microcontrollers.
>> > > >
>> > > > lwIP is licensed under a BSD-style license: http://lwip.wikia.com/wiki/License.
>> > > > Which should be compatible with u-boot.
>> > > >
>> > > > In the current RFC I tried to use minimal changes to better see how LWIP code can be embedded into
>> > > > U-boot. Patches implement ping and wget commands work. Both commands are currently copy pasting and
>> > > > reusing lwIP examples.  Whether we want to add the final application in U-Boot or lwIP is up to
>> > > > discussion, but the current approach was the easiest one for an RFC.
>> > >
>> > > I'm honestly not sure this is the most useful way of doing an RFC.  The
>> > > long term goal would be that we replace our existing net/ with lwIP,
>> > > yes? So what I'd see as more valuable is what it looks like to limit
>> > > yourself to either sandbox or some QEMU target, disable the current
>> > > network stack, and instead use lwIP to support just cmd/net.c so that
>> > > the scope of the conversion is visible.  Then the size comparison you do
>> > > should be between platform + net + cmd/net.c (and the rest of networking
>> > > turned off) and platform + lwip + cmd/net.c converted.
>> >
>> > This might be a bit too much imho.  How about replacing the TCP stack which
>> > is new an under heavy devel as well.  If we do that we could replace lwip
>> > with the version Maxim proposes and check the difference between
>> > U-boot + homegrown TCP + wget
>> > U-Boot + LWIP (for tcp only) + new wget
>> >
>> > That would give us an idea before trying to replace the UDP portion which
>> > is way bigger
>>
>> I guess we can try that as a starting point and see what things look
>> like.  My gut feeling however is that's not going to look like a win.
>>
>> --
>> Tom