Mark Williamson
2005-May-06 18:07 UTC
Re: [Xen-devel] PCI interrupt setup failure: gateway
Hi there, On Friday 06 May 2005 19:28, Jonathan S. Shapiro wrote:> I''m seeing what appears to be an interrupt setup failure. It occurs both > in the current pre-FC4 xen RPM and also in the xen-unstable tree from > last night. I''m unable to pull from the bitkeeper repository -- Larry > McVoy denied us a license because we work on OpenCM and that competes > with Bitkeeper.Have you tried SourcePuller? http://sourceforge.net/projects/sourcepuller/ It''s the notorious Tridge-designed BK client. It doesn''t have the SCM functionality but should perhaps be sufficient to get the latest source out of BK? Or there''s the BK open source client http://www.bitmover.com/bk-client.shar - this was originally under the "No Whiners License" but I think it''s actually BSD licensed, or something similar. Anyhow, I stuck a tarball of the latest and greatest BK tree at: http://www.cl.cam.ac.uk/~maw48/xen-devel-20050506-maw.bk.tar.bz2 It should contain the latest patches - maybe it will help.> Since I cannot get to the bitkeeper repository, I attempted to build and > run the latest xen-unstable tarball -- partly because that has some of > the new ACPI patches. This didn''t help at all -- I suspect the problem > is that cardbus support may not exist in Xen yet.Xen doesn''t let Linux see behind Cardbus bridges, so that wouldn''t work. The new PCI / ACPI patches may support it, although I haven''t tried. Cheers, Mark> However, I *did* notice that xen-unstable exhibits significantly reduced > function relative to the current FC4-devel RPM. X windows crashes, one > of the early setup tools in the boot process also crashes. Possibly > Jeremy''s patches are not migrating back into the tree promptly? > > At the moment, I''ld appreciate some guidance on where to look. > > shap > > On Thu, 2005-05-05 at 02:35 +0100, Ian Pratt wrote: > > > I now have xen up from the FC4-- repository on two machines. > > > One works fine, the other is a very new amd64, and xen is > > > having trouble parsing the bios tables (therefore unable to > > > find the interrupt line for the enet controller). > > > > The ACPI patch went in to unstable.bk yesterday -- it would be very > > interesting to know whether it helped or made things worse. > > > > > I''m happy to instrument/debug, but who''s the right contact > > > person? Can join xen-developers list if needed, but the > > > traffic is heavy... > > > > I''d prefer I you could post to xen-devel, even if you don''t subscribe. > > > > Thanks, > > Ian > > > > > shap > > > > > > On Sun, 2005-05-01 at 05:26 +0100, Ian Pratt wrote: > > > > > > BTW: big ACPI support patch going in next week. Mightn''t be > > > > > > > > > > too hard > > > > > > > > > > > to get power management working on top of it... > > > > > > > > > > Great. Now if only you would dump bitkeeper. One of the > > > > > > reasons that > > > > > > > > I have been unable to track the Xen project more closely > > > > > > is that I > > > > > > > > cannot install the bitkeeper client. Linus isn''t the only > > > > > > person for > > > > > > > > whom Larry has made legal problems. > > > > > > > > We''ll be putting up a CVS server of the head of the repository. > > > > > > > > Ian > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xensource.com > http://lists.xensource.com/xen-devel_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Jonathan S. Shapiro
2005-May-06 18:28 UTC
[Xen-devel] PCI interrupt setup failure: gateway
I''m seeing what appears to be an interrupt setup failure. It occurs
both
in the current pre-FC4 xen RPM and also in the xen-unstable tree from
last night. I''m unable to pull from the bitkeeper repository -- Larry
McVoy denied us a license because we work on OpenCM and that competes
with Bitkeeper.
In the pre-FC4 version of xen that is currently in the Fedora
development tree, I get the following diagnostic (from Xen) about midway
through the linux boot:
PCI: No IRQ known for interrupt pin A of device 00:0a.0.
Probably buggy MP table.
The PCI address in question is a cardbus bridge. The lspci info is
(retyping):
00:0a.o CardBus bridge: ENE Technology Inc CB1410 Cardbus Controller
(rev 01)
Subsystem: Rioworks: Unknown device 2032
Flags: bus master, medium devsel, latency 168
Memory at ffe7f000 (32-bit, non-prefetchable) [size=4K]
Bus: primary=00, secondary=02, subordinate=05, sec-latency=176
Memory window 1: 3ec00000-3efff000
I/O window 0: 00004000-000040ff
I/O window 1: 00004400-000044ff
16-bit legacy interface ports at 0001
This is on a fairly new machine: Gateway 7426gx, which has an AMD64
mobile CPU.
Curiously, the system is also unable to locate the rhine ethernet device
at PCI 00:12.0, which the xen kernel claims is not present. This
suggests that other interrupt setup isn''t going quite right either.
Since I cannot get to the bitkeeper repository, I attempted to build and
run the latest xen-unstable tarball -- partly because that has some of
the new ACPI patches. This didn''t help at all -- I suspect the problem
is that cardbus support may not exist in Xen yet.
However, I *did* notice that xen-unstable exhibits significantly reduced
function relative to the current FC4-devel RPM. X windows crashes, one
of the early setup tools in the boot process also crashes. Possibly
Jeremy''s patches are not migrating back into the tree promptly?
At the moment, I''ld appreciate some guidance on where to look.
shap
On Thu, 2005-05-05 at 02:35 +0100, Ian Pratt wrote: >
> > I now have xen up from the FC4-- repository on two machines.
> > One works fine, the other is a very new amd64, and xen is
> > having trouble parsing the bios tables (therefore unable to
> > find the interrupt line for the enet controller).
>
> The ACPI patch went in to unstable.bk yesterday -- it would be very
> interesting to know whether it helped or made things worse.
>
> > I''m happy to instrument/debug, but who''s the right
contact
> > person? Can join xen-developers list if needed, but the
> > traffic is heavy...
>
> I''d prefer I you could post to xen-devel, even if you
don''t subscribe.
>
> Thanks,
> Ian
>
> > shap
> >
> > On Sun, 2005-05-01 at 05:26 +0100, Ian Pratt wrote:
> > >
> > > > > BTW: big ACPI support patch going in next week.
Mightn''t be
> > > > too hard
> > > > > to get power management working on top of it...
> > > >
> > > > Great. Now if only you would dump bitkeeper. One of the
> > reasons that
> > > > I have been unable to track the Xen project more closely
> > is that I
> > > > cannot install the bitkeeper client. Linus isn''t
the only
> > person for
> > > > whom Larry has made legal problems.
> > >
> > > We''ll be putting up a CVS server of the head of the
repository.
> > >
> > > Ian
> >
> >
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
On 6 May 2005, at 19:28, Jonathan S. Shapiro wrote:> Since I cannot get to the bitkeeper repository, I attempted to build > and > run the latest xen-unstable tarball -- partly because that has some of > the new ACPI patches. This didn''t help at all -- I suspect the problem > is that cardbus support may not exist in Xen yet.I''ll be very surprised if these problems still show up in the latest snapshot. Things were in a state of flux after checking in Intel''s ACPI/irq-routing patch, and there were still some pretty serious problems until Friday (there may still be more lurking but I think the worst are gone).> However, I *did* notice that xen-unstable exhibits significantly > reduced > function relative to the current FC4-devel RPM. X windows crashes, one > of the early setup tools in the boot process also crashes. Possibly > Jeremy''s patches are not migrating back into the tree promptly?Probably due to bugs added after Redhat last took a snapshot, rather than fixes that haven''t been fed back. I fixed a nasty process destruction bug this morning (a domain would blow up when destroying a process that has an LDT). It''s possible this will sort out your problems -- if not then please open a bug at bugzilla.xensource.com. If you can reproduce with a non-X (text console only) application then it will be easier for us to reproduce and track down. -- Keir _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Jonathan S. Shapiro
2005-May-08 18:57 UTC
Re: [Xen-devel] PCI interrupt setup failure: gateway
Mark: Before attempting it on the amd64 laptop, I tested you pull of xen- devel-20050506 on my thinkpad T-40. It''s a net reduction in functionality. The version of xen currently in the FC4 devel tree doesn''t have ACPI support, but it configures interrupts correctly for this machine. The version from your pull does not configure interrupts correctly to the extent that the e1000 device (which worked fine in Xen from the FC4 devel tree) cannot be found at all with the 05/06 pull. It''s not a cardbus issue. shap _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel