Problem with the serial port in Fedora Core 6
craigwhite at azapple.com
Sun Nov 12 03:43:22 UTC 2006
On Sat, 2006-11-11 at 22:17 -0500, Gene Heskett wrote:
> On Saturday 11 November 2006 19:42, Craig White wrote:
> >> >https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=204825
> >> But, I'll repeat myself AGAIN, I don't think udev has anything to
> >> with THIS problem. One clue from this bug report may be that xen is
> >> hogging the serial ports somehow, and from the looks of the grub.conf,
> >> it has first dibs and locks everybody else out since with xen running
> >> (AIUI) the kernel itself is a client and xen prevents the discovery of
> >> something xen (the serial ports in this case) is useing. In any
> >> event, its appears that getting rid of xen solves the problem. So
> >> will you please quit harping on whether or not this udev nuwbie has a
> >> clue? Its not a udev problem according to what my detective work has
> >> found.
> >you're detective work notwithstanding, Bill Nottingham's suggestion to
> >put an entry into 'udev' rules was the suggested recommendation for the
> >problem but since you didn't try that with the xen kernel, didn't post
> >your results on this bugzilla entry after trying the recommended entry,
> >I guess none of us will ever be certain.
> And what sort of an entry might that be?
My reference was to Bill Nottingham's comments - I believe that they are
> It does create 4 devices now, as
> its written to do, but the devices are worthless when the onboard
> hardware hasn't been detected.
that was what the OP posted on that bugzilla too.
> >I guess I am glad that the non-xen kernel gives you what you wanted
> >after all (the serial port interfaces for your X11). I just assumed that
> >you chose the xen kernel with the intent to run xen virtualization, but
> >I am guessing that your motherboard doesn't support xen which is
> >probably the cause of the problem in the first place.
> Either way (I didn't choose it, anaconda did), and I have NDI if this
> biostar mobo supports xen or not, there has been precious little
> discussions about it on the 50+ mailing lists I lurk on.
I have scripted all my FC-6 installs now with kickstart so I don't
recall the choices I was given the one time I used it but I thought I
recalled a Xen option.
> >Now I agree with you that getting rid of the xen kernel solves this
> >problem. I don't agree with you a bit that it wasn't a udev problem
> >because if you again read Comment #5 on the above referenced bugzilla
> >entry, it is obvious that Bill Nottingham believes it is a udev related
> I still don't see what, if anything, udev could do about it when the
> hardware has not been detected. Does it have some sort of a magic
> twanger for that case that we common mortals don't know about? I'm now
> looking at that #5 comment, but have no clue because theres no preamble
> describing where to put those commands, which I assume would go
> in /.etc/udev/rules.d/50-udev.rules, but it contains no similar rules.
> Those in the know will I assume know where in that file, but to me thats
> not a usable fix without the whole story. But with some use of a
> figurative shovel to dig a little deeper it seems to indicate it should
> be added into /etc/udev/rules.d/05-udev-early-rules. It would be more at
> home there by far.
> Is this correct?
this is where I always refer to when thinking about udev changes
this suggests your own rule in /etc/udev/rules.d/10-whavever
05-udev-early-rules might be OK
> There is a second reason I didn't pursue that angle, any attempt to
> modprobe anything 8250 related failed due to undefined labels in the
> module(s). Which brings me back to square one, what can udev do about
> it when the kernel can't find the hardware in the first place?
I honestly don't know for sure - that was Bill Nottingham's suggestion
and he is obviously in a much better position to know these things than
I. I am no expert in udev myself.
> right now, with them running, an lspci -vv doesn't show them. Whether
> they should show I don't recall. I just checked on my firewall box, and
> they don't appear there either, so perhaps I'm using the wrong tool.
I don't recall ever seeing a serial port appearing in an lspci command
> Sorry I'm so dense in your opinion. If it will make you feel better, I'll
> plead oldtimers since I had a birthday last month, having made it to 72.
You are who you are. I didn't make any characterization whatsoever. I
helped you fix your serial port and you don't say thanks...ok...that's
fine. You then stated by your 'detective work' that the problem couldn't
have been udev but I suggested that you didn't do the testing to know
that - and even if you had done the testing as recommended in the
referred to bugzilla, you might not have proved anything except a
particular problem between your motherboard and xen kernel.
I see the point of bugzilla is to contribute the data set in the hands
of the developers so that they can be better informed. Your problem
though is solved and you don't need to do anything further and that's
fine. You haven't proven to yourself or anyone else that the problem
wasn't udev though - that was my point.
More information about the users