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

Bug#273106: cmd64x driver doesn't load properly on ia64



Package: kernel-source-2.4.27

This bug appears in the kernel-image-2.4.27-1-itanium_2.4.27-3_ia64.deb
package although the .udeb derived from this package does work.

The CDROM drive, using cmd64x does not properly install.  I have traced
the problem on an HP rx2600 dual Itanium server to the following:

1. ide-cd installs followed by ide-detect.  This causes ide_scan_pcidev 
  to start its scan.  The CMD649 device (0x1095:0x649) is noted in
  the scan but the cmd64x probe function (cmd64x_init_one) is not
  called.

2. cmd64x is loaded and its init function is called which does call
  the probe function (pre_init == 0).  The console reports the IDE
  controller at PCI slot ... thru the ide0 and ide1 lines with the
  dma and BIOS settings (hd[a-d]:pio).  There are no messages
  reporting the CDROM drive itself.

/proc/modules shows all of the modules loaded but /proc/ide only
has the files drivers and cmd64x.  The subdirs ide[01] and the
symlinks hd[a-d] are not present.  Access of the device (like mount)
returns a "not a valid device" error.

The udeb works on the install kernel because the cmd64x driver
seems to get loaded first.  I also reproduced this in a running
system by loading re-compiled modules (with added printk's...) in that
order.  Others have reported correct loading on their running
systems as well.  Once I rebooted my system the same error
behavior I saw in the stage 2 booted kernel appeared.

In looking at the changelog for kernel-source, I note that patch
070_ide_scan_pcibus.diff was a missing patch on an earlier
rev.  Might there be more to this patch?  Given the two slightly
different paths to and thru the init_one function, this seems to be
an issue in the pci layer somewhere although it may be triggered
by the cmd64x.

Note that this is not a problem with either the 2.4.25 or 2.6.8-1 
kernels.  Also, although I could unload all of the ide modules in
the earlier (-1) kernel, I can't unload cmd64x (busy) and ide-core
(also busy because of cmd64x).

I've looked at stock kernels for 2.4.27 and 2.4.28 and don't see these
patches so I don't know their origin.

Regards,

Jim
-- 
*************
Jim Lieb			Wild Open Source Inc.
lieb@wildopensource.com		Cell: 831.295.9317
Office: 831.421.0883            Fax:  831.421.0885




Reply to: