[U-Boot] U-Boot IPv6 state

[U-Boot] U-Boot IPv6 state

[Chris Packham]

On 18/01/2017 5:56 AM, "Luka Kova?i?" <luka.kovacica@gmail.com> wrote:

Hi,
My name is Luka Kova?i? and I am just getting interest to U-Boot. I was
really confused, because IPv6 is all over U-Boot on the mailing lists, but
it actually isn't available in the mainline tree. What is the current state
of it? Does it work?
Kind regards,
Luka


Wow 2 inquiries in 2 days. Something is going on in the ether.

The last series that was posted is functional but now a little out of date
w.r.t. other u-boot changes. I've also got a github repo with the changes
applied https://github.com/cpackham/u-boot/tree/ipv6

Tom has applied 2 of the trivial clean up patches so I plan to rebase the
series on top of that soon. After that I'll update to use kconfig. Then
I'll finally get round to doing what Joe actually asked for and see about
adding some tests. I could use some help with that last bit.

[U-Boot] U-Boot IPv6 state

[Teddy Reed]

On 01/17/2017 09:49 PM, Luka Kova?i? wrote:
> That's great. I could help you to test it. Link-local and maybe
> auto-configuration on the home router.
>
> V tor., 17. jan. 2017 23:19 je oseba Chris Packham
> <judge.packham at gmail.com <mailto:judge.packham@gmail.com>> napisala:
>
>
>
>     On 18/01/2017 5:56 AM, "Luka Kova?i?" <luka.kovacica@gmail.com
>     <mailto:luka.kovacica@gmail.com>> wrote:
>
>         Hi,
>         My name is Luka Kova?i? and I am just getting interest to
>         U-Boot. I was really confused, because IPv6 is all over U-Boot
>         on the mailing lists, but it actually isn't available in the
>         mainline tree. What is the current state of it? Does it work?
>         Kind regards,
>         Luka
>
>
>     Wow 2 inquiries in 2 days. Something is going on in the ether.
>
>     The last series that was posted is functional but now a little out
>     of date w.r.t. other u-boot changes. I've also got a github repo
>     with the changes applied https://github.com/cpackham/u-boot/tree/ipv6
>

Thanks!

I like the test-command approach.

For some context, I'm testing in QEMU with the added switches:
   -net tap,vlan=0,name=br0,id=net0,script=ifup.sh \
   -net nic,name=br0,vlan=0,model=ftgmac100

Following Redhat's how-to: http://www.linux-kvm.org/page/Networking

And applying a quick patch to net/net.c (2016.07 tag):

                 }
                 goto common;
  #endif
-#ifdef CONFIG_CMD_PING6
+#if defined(CONFIG_CMD_PING6)
         case PING6:
                 if (ip6_is_unspecified_addr(&net_ping_ip6)) {
                         puts("*** ERROR: ping address not given\n");
                         return 1;
                 }
-               goto common;
+               break;
  #endif
  #if defined(CONFIG_CMD_SNTP)

I wanted to avoid the check for `ipaddr`.

 From there I can set CONFIG_PREBOOT commands to ping a pre-configured 
IPv6 address on the selected interface. The NS packets work perfectly, 
but my NIC model is not buffering the entire NA ICMPv6 packet. I'll test 
with hardware tomorrow and report back but I don't suspect any issues 
with the V6 code at all.

Once I apply some tweaks my test case will be similar to the U-Boot 
sandbox tests [1]; but instead a QEMU model is executed to run a 
ping6/tftpboot6. Then an expect script verifies the output and halts the 
emulator. These are nasty integration tests but the most useful if 
you're like me and most of your work is configuring boards rather than 
developing.

>     Tom has applied 2 of the trivial clean up patches so I plan to
>     rebase the series on top of that soon. After that I'll update to use
>     kconfig. Then I'll finally get round to doing what Joe actually
>     asked for and see about adding some tests. I could use some help
>     with that last bit.

[1] https://github.com/trini/u-boot/blob/master/test/py/tests/test_net.py

>

[U-Boot] U-Boot IPv6 state

[Luka Kovačič]

OK. What can I test? I could do it on real hardware like odroid c2 or qemu.

V sre., 18. jan. 2017 09:44 je oseba Teddy Reed <reed@fb.com> napisala:



On 01/17/2017 09:49 PM, Luka Kova?i? wrote:
> That's great. I could help you to test it. Link-local and maybe
> auto-configuration on the home router.
>
> V tor., 17. jan. 2017 23:19 je oseba Chris Packham
> <judge.packham at gmail.com <mailto:judge.packham@gmail.com>> napisala:
>
>
>
>     On 18/01/2017 5:56 AM, "Luka Kova?i?" <luka.kovacica@gmail.com
>     <mailto:luka.kovacica@gmail.com>> wrote:
>
>         Hi,
>         My name is Luka Kova?i? and I am just getting interest to
>         U-Boot. I was really confused, because IPv6 is all over U-Boot
>         on the mailing lists, but it actually isn't available in the
>         mainline tree. What is the current state of it? Does it work?
>         Kind regards,
>         Luka
>
>
>     Wow 2 inquiries in 2 days. Something is going on in the ether.
>
>     The last series that was posted is functional but now a little out
>     of date w.r.t. other u-boot changes. I've also got a github repo
>     with the changes applied https://github.com/cpackham/u-boot/tree/ipv6
>

Thanks!

I like the test-command approach.

For some context, I'm testing in QEMU with the added switches:
   -net tap,vlan=0,name=br0,id=net0,script=ifup.sh \
   -net nic,name=br0,vlan=0,model=ftgmac100

Following Redhat's how-to: http://www.linux-kvm.org/page/Networking

And applying a quick patch to net/net.c (2016.07 tag):

                 }
                 goto common;
  #endif
-#ifdef CONFIG_CMD_PING6
+#if defined(CONFIG_CMD_PING6)
         case PING6:
                 if (ip6_is_unspecified_addr(&net_ping_ip6)) {
                         puts("*** ERROR: ping address not given\n");
                         return 1;
                 }
-               goto common;
+               break;
  #endif
  #if defined(CONFIG_CMD_SNTP)

I wanted to avoid the check for `ipaddr`.

 From there I can set CONFIG_PREBOOT commands to ping a pre-configured
IPv6 address on the selected interface. The NS packets work perfectly,
but my NIC model is not buffering the entire NA ICMPv6 packet. I'll test
with hardware tomorrow and report back but I don't suspect any issues
with the V6 code at all.

Once I apply some tweaks my test case will be similar to the U-Boot
sandbox tests [1]; but instead a QEMU model is executed to run a
ping6/tftpboot6. Then an expect script verifies the output and halts the
emulator. These are nasty integration tests but the most useful if
you're like me and most of your work is configuring boards rather than
developing.

>     Tom has applied 2 of the trivial clean up patches so I plan to
>     rebase the series on top of that soon. After that I'll update to use
>     kconfig. Then I'll finally get round to doing what Joe actually
>     asked for and see about adding some tests. I could use some help
>     with that last bit.

[1] https://github.com/trini/u-boot/blob/master/test/py/tests/test_net.py

>

[U-Boot] U-Boot IPv6 state

[Joe Hershberger]

Hi Chris,

On Tue, Jan 17, 2017 at 4:19 PM, Chris Packham <judge.packham@gmail.com> wrote:
> On 18/01/2017 5:56 AM, "Luka Kova?i?" <luka.kovacica@gmail.com> wrote:
>
> Hi,
> My name is Luka Kova?i? and I am just getting interest to U-Boot. I was
> really confused, because IPv6 is all over U-Boot on the mailing lists, but
> it actually isn't available in the mainline tree. What is the current state
> of it? Does it work?
> Kind regards,
> Luka
>
>
> Wow 2 inquiries in 2 days. Something is going on in the ether.

Glad to see activity on this again. I too would like to see it in the mainline.

> The last series that was posted is functional but now a little out of date
> w.r.t. other u-boot changes. I've also got a github repo with the changes
> applied https://github.com/cpackham/u-boot/tree/ipv6
>
> Tom has applied 2 of the trivial clean up patches so I plan to rebase the
> series on top of that soon. After that I'll update to use kconfig. Then
> I'll finally get round to doing what Joe actually asked for and see about
> adding some tests. I could use some help with that last bit.

Cool. I agree that unit tests are gonna be critical for keeping it
functional since at least for now it's not likely to see a ton of use
immediately. Also just really good practice.

Another thing I'd like to see happen is a Travis CI test.

Cheers,
-Joe