> Date: Tue, 14 Nov 2006 17:38:16 +0000 (GMT) > From: Robert Watson <rwatson_at_FreeBSD.org> > Sender: owner-freebsd-current_at_freebsd.org > > On Tue, 14 Nov 2006, John-Mark Gurney wrote: > > > Robert Watson wrote this message on Tue, Nov 14, 2006 at 13:53 +0000: > >> I updated two boxes to recent kernels from a kernel around October 7 or so, > >> and they now both hang on boot if I have a Neterion 10gbps ethernet card in > >> the PCIe slot. Since I don't have the driver loaded at boot, it seems more > >> likely it's a kernel bug. Both identical machines now have the following > >> vpd warning during boot, which wasn't present previously, but may be > >> unrelated: > > > > It's very likely you are plagued w/ non-standard PCI cards... > > > > I assume you mean Nov 7th? If so, there was fix committed for more normal > > bad VPD data (v1.321 of sys/dev/pci/pci.c)... > > > > There is still an outstanding bug of a device that doesn't even properly > > handle VPD accesses and it hang waiting for a bit to clear... I need to > > inspect the patch closer before committing.. > > What I mean specifically is that the kernel dated October 7 works fine, and > any more recent kernel hangs solidly if I boot it. Obviously, this is > somewhat inconvenient. :-) > > The device in question is a PCI-X Neterion 10gbps card. The output from the > kernel when the device driver is loaded is: > > Copyright(c) 2002-2005 Neterion Inc. > xge0: <Neterion XframeII 10GbE Adapter, Revision 2, Driver v2.0.0.6765> mem > 0xd8300000-0xd8307fff,0xd8400000-0xd84fffff,0xd8308000-0xd83087ff irq 25 at > device 1.0 on pci129 > xge0: Device is on 64 bit PCIX(M1) 133MHz bus > > If there's more information I can provide I'm happy to do so, just let me know > what's needed. > > Is there a way I can disable vpd support at boot-time in some form -- i.e., > via a tunable? It would be very useful if these machines worked. The patch Ian Dowse submitted for this (which jmg is evaluating) has fixed the problem for at least two cases, my ancient Dell P3 and a newer ASUS A8V. It was posted back on 8 Nov. It simply adds a timeout for getting VPD info and lets the system proceed. Note that the last hunk is has been committed by jmg, so you only need the first two hunks. -- R. Kevin Oberman, Network Engineer Energy Sciences Network (ESnet) Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab) E-mail: oberman_at_es.net Phone: +1 510 486-8634 Key fingerprint:059B 2DDF 031C 9BA3 14A4 EADA 927D EBB3 987B 3751
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:02 UTC