[syslinux] extlinux 6.03 IPMI SOL (serial-over-lan) console not responding to input

[Ady]

> Hi,
> 
> On Mon, May 18, 2015 at 10:09:12PM +0300, Ady via Syslinux wrote:
> > First, you should add all the lib*.c32 files for BIOS (located under 
> > the BIOS subdir), so to avoid the error messages.
> 
> Thanks.  But the problem was demonstrated in any case, regardless
> of whether these files are present.  The input is not received
> by extlinux.
 
Sure. I just mentioned the library modules because menu.c32 needs them.
 
> 
> > Then, just in case, have you tried changing:
> > 
> >  serial 0 115200 0x003 
> > 
> > with alternative values?
> 
> I previously tried different flowcontrol bits, such as 0 or 0x00b.
> 
> > For instance:
> >  port is a number (0 =/dev/ttyS0 = COM1, etc.) or an I/O port address 
> > (e.g. 0x3F8).
> 
> The port is certainly right -- the serial output is displayed correctly.
> I cannot see how changing it to the wrong port would help.
 
My intention was to see whether any different behaviors could be seen, 
for example when using a different notation (instead of using a port 
number, using an equivalent I/O port).
 
> 
> Similarly, the speed is correct.  It works properly with the physical
> serial connection, and it displays properly over IPMI.
> 
> > Or trying a different speed, just for a test?
> 
> I have not tried this.  It does not seem plausible that it could matter.
> That being said, two of the three machines exhibiting the problem have the
> speed configured at 115,200, and the other one is at 57,600.  So I have
> effectively tried two different speeds.
 
I understand. As I said, my intention was to trigger some potential 
different behavior. Different (and correctly configured) speeds could 
potentially show such difference, if there is one.
 
> 
> > Or trying a different flowcontrol value for a test?
> 
> That was the first thing I tried.  It did not help.
> 
> > I am not saying it will certainly solve your situation (probably it 
> > won't), and a bug in Syslinux 6.03 is not at all discarded.
> > 
> > The other suggestion I can think of would be to search for a BIOS 
> > update.
> 
> I think I said this already, but for the sake of clarity: the exact
> same extlinux.conf setup was working properly on the same machine with
> the same motherboard and the same BIOS under Fedora 19 with extlinux 4.05.
> The only thing that changed was upgrading the O/S (and extlinux along
> with it).
 
We have seen v.4.xx behaving "nice" with some BIOS versions, and 
Syslinux v.6.xx _triggering_ some problem. The problem was not always 
in Syslinux v.6.xx, but in the BIOS; Syslinux was only the trigger. In 
some cases, users updated their BIOS versions and the problem was 
solved.
 
> 
> > I am aware that these are not the type of comments you would hope for.
> 
> I appreciate your responding.  But I feel that I have done enough testing at
> this point to demonstrate that there is a bug in extlinux 6.03's serial
> implementation.  Grub2 works fine, and older extlinux works fine.
> 
> Should I bother to create a bugzilla for this?  I'm willing to run more
> tests from which we have some hope of learning something.  But I respectfully
> submit that most of your suggestions above are exceedingly unlikely to
> reveal any additional useful information.  I've already done a lot of
> troubleshooting, and I think I've narrowed down the problem pretty well.
> The next step is to start looking at the code.  Is there a git repository
> for this that can show the difference in the serial handling between 4.05
> and 6.03?  I have been a maintainer on some open-source projects.  At
> a certain point, enough information has been collected, and it's time to
> dive into the code or find a way to get some debugging info.  I think
> we have exhausted what can be learned from running tests with various
> configurations.
> 
 
I am not at all surprised that the console in v.6.xx is showing some 
problems under certain situations. I am just trying to narrow down the 
source of the problem, and one way to do it is to trigger different 
behaviors with some changes (such as using official binaries).

Hopefully someone else might have some more-specific tests (such as 
testing specific versions so to narrow down the commits' range). I'll 
leave such suggestions to others.

Some potential additional tests could include using alternative keys, 
e.g. [Ctrl-E] instead of [End]. See:

 http://www.syslinux.org/wiki/index.php/Cli#Keystrokes 
 
We appreciate the feedback and the tests.

Thank you,
Ady.