cmd64x not happy about being hotplugged, 2.6.17

cmd64x not happy about being hotplugged, 2.6.17

[Josh Litherland]

I have a thinkpad dock 2 which includes a Silicon Image CMD648 IDE 
controller (1095:0648 (rev 01)) for the UltraBay2000 slot.  The dock now 
works as an acpiphp PCI hotplug device, and the PCI devs get detected 
safely and removed more or less safely when undocking.  The cmd64x 
doesn't seem to handle the hotplugging as well.  When it is attached, it 
doesn't enumerate the CDRW device that is connected to it (although the 
drive works if it is inserted at boot time.).  When the laptop is 
undocked, cmd64x seems not to notice at all.  Any further access to it, 
such as "cat /proc/ide/cmd64x", will cause a segfault as seen here:

http://downloads.emperorlinux.com/support/misc/bug.txt

Let me know if there's any further information I can provide.  Thanks!

-- 
Josh Litherland (josh@temp123.org)

Re: cmd64x not happy about being hotplugged, 2.6.17

[Alan Cox]

Ar Gwe, 2006-06-23 am 11:35 -0400, ysgrifennodd Josh Litherland:
> I have a thinkpad dock 2 which includes a Silicon Image CMD648 IDE 
> controller (1095:0648 (rev 01)) for the UltraBay2000 slot.  The dock now 
> works as an acpiphp PCI hotplug device, and the PCI devs get detected 

drivers/ide does not support hotplug except in 2.4-ac and sort-of (if
the device is closed entirely) in Red Hat Enterprise Linux 4. You'll
have to ask Bartlomiej if he has any plans to fix that for 2.6 proper.

The new libata layer has hotplug work afoot which should deal with this,
although the CMD648 driver for it is very experimental as the 648 is
pretty rare nowdays

Alan