Help needed with 3c90x driver in RHL 6.2 kickstart

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



	Hi,
	I am in need of help, and I am hoping someone here will be able
to help me with some information -- I am probably missin gthe obvious.

	We have a RHL 6.2 kickstart setup; it has worked fine so far,
but the new batch of systems we got have a network device which
identifies itself as 3Com 905C (PCI ID 9200), does not work with the
3c59x driver. Our older systems' network device also identifies itself
as PCI device ID 9200, but it works fine with 3c59x driver.

	Even though both devices offer the same PCI ID, the newer
systems' network device does not work with 3c59x driver; it does work
with the 3c90x driver, but the latter isn't included on the standard
bootnet floppy. No biggie, I figure -- I unzipped and mounted the
INITRD, uncompressed modules.cgz, added 3c90x.o file from
kernel-BOOT-2.2.14-5.0.i386.rpm, removed some older modules (3c501,
3c503 & 3c505) edited module-info and pcitable, and rolled everything
back into INITRD and onto the floppy.

	Now a system (any system, whether the one that worked with the
old kickstart's 3c59x driver or one that requires 3c90x driver) starts
booting, gets past the DHCP stage, and bails when trying to get the NFS
volume where the kickstart file resides (we distribute the kickstart
config file over the network). The same volume can be mounted normally
when the system is up in Linux (booted from a SuperRescue 2.0 CD, or
booted from HD installation in the case if the older systems).

	Both older systems and newer ones, when booted with the modified
kickstart floppy (the one that ties PCI device 9200 to 3c90x driver
instead of 3c59x), produce the same output on the error (3rd) console
(with appropriate names blotted out to protect the guilty parties):

* probing buses
* finished bus probing
* found suggestion of 3c90x
* found 3c90x device
* found devices justProbe is 0
* going to insmod 3c90x.o (path is NULL)
* sending DHCP request through device eth0
* nodns is 0
* reverse name lookup failed
* ks server: ###.###.###.###:/xxxxx/xxxxx/xxxxx file xxxxx.cfg
* going to insmod sunrpc.o (path is NULL)
* going to insmod lockd.o (path is NULL)
* going to insmod nfs.o (path is NULL)
* failed to mount ###.###.###.###:/xxxxx/xxxxx/xxxxx

	At this point, there is a message box on the primary console
indicating that the kickstart config file couldn't be found; if I then
proceed with standard netwoprk install and try to get the IP via DHCP,
the following lines are added to the error console:

* no install method specified for kickstart
* 65 keymaps are available
* loaded 9 keymap tables
* pump told us: SIOCSIFFLAGS: No such device

	Now I double-checked, and the kernel on my kickstart floppy is
identical to the kernel in the kernel-BOOT-2.2.14-5.0.i386.rpm;
correspondingly, the 3c90x.o file I put onto the floppy is indeed the
one I yanked from that same RPM. It should work, to the best of my
knowledge, but it does not. The 3c90x driver obviously gets the DHCP
info; the content of the syslog (4th) console indicate that it does see
the network, at least temporarily -- it gets the correct info about our
DNS and NTP servers over the net, so there is definitely at least some
IP negotiations going on. However, something in the network driver just
as obviously doesn't work, and I have no idea what it is (the last error
entry, about 'SIOCSIFFLAGS', is really weird).

	I am of course still able to kickstart the older machines from
the unmodified kickstart floppy. The difference on the old machines'
error log between the modified (3c90x) kickstart and the old, working
one, starts after the "nodns" line, when the following line is logged:

* reverse name lookup worked

	After which the kickstart (using 3c59x driver) proceeds normally.

	So the reverse DNS lookup works with 3c59x driver and fails with
the 3c90x driver, but I have no idea why; I do suspect that it's just a
symptom of a deeper problem, though. Furthermore, that same older
system, when kickstarted normally (with 3c59x driver) and then
configured to use 3c90x instead, works without any problems, so the
3c90x driver itself doesn't seem to be the culprit...

	Can anyone help me? I really have no idea what the problem here
is... Many thanks in advance.

-- 
|  Victor  Danilchenko  +----------------------------------------------+
| danilche@xxxxxxxxxxxx | Does Emacs have the buddha nature?           |
|   CSCF   |   5-4231   | Why not, it has bloody well everything else! |





[Index of Archives]     [Red Hat General]     [CentOS Users]     [Fedora Users]     [Fedora Maintainers]     [Fedora Desktop]     [Fedora SELinux]     [Big List of Linux Books]     [Yosemite News]     [KDE Users]

  Powered by Linux