[syslinux] [PATCH] Updated udp.c to use real client ip and subnetmask values if on local subnet

[Gene Cumm]

On Tue, Sep 15, 2015 at 11:08 AM,  <jeff_sloan at selinc.com> wrote:
>
>
> Jeff Sloan
> Software Engineer - Computing Systems
> Schweitzer Engineering Laboratories, Inc.
> (509) 336-2098
> jeff_sloan at selinc.com
>
>
>
> From:        Gene Cumm <gene.cumm at gmail.com>
> To:        Patrick Masotta <masottaus at yahoo.com>, jeff_sloan at selinc.com,
> Cc:        For discussion of Syslinux and tftp-hpa <syslinux at zytor.com>
> Date:        09/12/2015 07:35 AM
> Subject:        Re: [PATCH] Updated udp.c to use real client ip and
> subnetmask values if on local subnet
> ________________________________
>
>
>
>>On Sat, Aug 15, 2015 at 9:50 AM, Gene Cumm <gene.cumm at gmail.com> wrote:
>>> On Jul 27, 2015 12:30 PM, "Patrick Masotta" <masottaus at yahoo.com> wrote:
>>>>
>>>> >>>
>>>>  > I think these changes would solve the thing.
>>>>  >
>>>>  > ...
>>>>  > -EFI_SERVICE_BINDING *sbp;
>>>>  > +EFI_SERVICE_BINDING *sbp =NULL;
>>>>  > ...
>>>>  >
>>>>  > -    if (status != EFI_SUCCESS)
>>>>  > +    if (status != EFI_SUCCESS || sbp ==
>>>>  NULL)
>>>>  >        goto free_binding;
>>>>  > ...
>>>>
>>>>  The presumption was that we'd see an error rather than
>>>>  non-error from LibLocateHandle() when nr_handles is 0 and be guaranteed
>>>> to find a
>>>>  matching handle.  Your additional measures seem the best choice.
>>>>  --
>>>>  -Gene
>>>> <<<
>>>>
>>>> OK I have adopted this code approach; testing it since last
>>>> week, so far so good.
>>>> Presumably there are other people testing this; feedback appreciated.
>>>>
>>>> Best,
>>>> Patrick
>>>>
>>>
>>> Jeff Sloan, have you been able to test the latest commit, e466d249 ?
>>
>>> Should have replied on this thread.
>>
>>--
>>-Gene
>
> I downloaded the latest commit, 02f3609, and tested on two different
> environments. The first is my test network which has dhcp server on one
> subnet and client, proxy and tftp server on another.
>
> Second network is a setup that simulates our customer environment of three
> subnets: dhcp on one, client on a second and proxy/tftp on a third.
>
> Both environments fail with: core_udp_sendto: udp->configure unsuccessful
> (?)
>
> Let me know what else I can do to help.

Excellent testing.  I presume these two machines are identical except
for which network they're in and obviously unique attributes like
UUID/MAC?

I think the better approach that may help your machines and some
others that have been having issues is to build a list of conditions
under which we declare the default address invalid THEN forcibly
insert the correct addresses.

 - udp->configure() never seems to succeed in a reasonable time (it
should already have been working)
 - If we're working on retrieving ldlinux.* and we have have yet to
receive a packet back.

-- 
-Gene