Xen users - Oct 2011 - Using pciback to dynamically bind/unbind primary display driver: lost console, help

On Debain 3.1.0-rc7, Radeon 6870 and Xen 4.1 (no patches)

I used

echo -n 0000:01:00.0 > /sys/bus/pci/drivers/radeon/unbind
echo -n 0000:01:00.0 > /sys/bus/pci/drivers/pciback/new_slot
echo -n 0000:01:00.0 > /sys/bus/pci/drivers/pciback/bind

to move my primary display adapter (01:00.0) to pciback for domU use.  The
display, as expected, turned off.  I had enabled ssh to do this, and have
access to the system still.  I then used

echo -n 0000:01:00.0 > /sys/bus/pci/drivers/pciback/unbind
echo -n 0000:01:00.0 > /sys/bus/pci/drivers/radeon/bind

to pass it back to dom0 at which point the display turned back on.  Thinking
all was well, I made a script to do the above and executed it (with a sleep
of 10 seconds after the first bind) at which point I lost the display, as
expected, but it didn''t come back.  I then logged in through ssh and
find
that the radeon driver is bound to 01:00.0, but still no display.

I think I need to tell the console to output to the radeon driver, but I
have no idea how and a google search is proving fruitless.  Any help? 
I''ve
also noticed that unbinding/binding the radeon driver is now (it wasn''t
initially) very slow to respond.

Thanks!


_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

Turns out I passed "dynpm" to radeon/power_profile instead of
radeon/power_method and it crashed atom bios (kept locking, timing out in
the error log).  Works fine now.  How do I control, on the fly, where kernel
output goes though?

On Fri, Oct 28, 2011 at 9:35 PM, Ben Himberg <bhimberg@gmail.com> wrote:
> On Debain 3.1.0-rc7, Radeon 6870 and Xen 4.1 (no patches)
>
> I used
>
> echo -n 0000:01:00.0 > /sys/bus/pci/drivers/radeon/unbind
> echo -n 0000:01:00.0 > /sys/bus/pci/drivers/pciback/new_slot
> echo -n 0000:01:00.0 > /sys/bus/pci/drivers/pciback/bind
>
> to move my primary display adapter (01:00.0) to pciback for domU use.  The
> display, as expected, turned off.  I had enabled ssh to do this, and have
> access to the system still.  I then used
>
> echo -n 0000:01:00.0 > /sys/bus/pci/drivers/pciback/unbind
> echo -n 0000:01:00.0 > /sys/bus/pci/drivers/radeon/bind
>
> to pass it back to dom0 at which point the display turned back on.
> Thinking all was well, I made a script to do the above and executed it
(with
> a sleep of 10 seconds after the first bind) at which point I lost the
> display, as expected, but it didn''t come back.  I then logged in
through ssh
> and find that the radeon driver is bound to 01:00.0, but still no display.
>
> I think I need to tell the console to output to the radeon driver, but I
> have no idea how and a google search is proving fruitless.  Any help? 
I''ve
> also noticed that unbinding/binding the radeon driver is now (it
wasn''t
> initially) very slow to respond.
>
> Thanks!
>

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

err console output... I''m new to this

On Fri, Oct 28, 2011 at 10:14 PM, Ben Himberg <bhimberg@gmail.com> wrote:
> Turns out I passed "dynpm" to radeon/power_profile instead of
> radeon/power_method and it crashed atom bios (kept locking, timing out in
> the error log).  Works fine now.  How do I control, on the fly, where
kernel
> output goes though?
>
>
> On Fri, Oct 28, 2011 at 9:35 PM, Ben Himberg <bhimberg@gmail.com>
wrote:
>
>> On Debain 3.1.0-rc7, Radeon 6870 and Xen 4.1 (no patches)
>>
>> I used
>>
>> echo -n 0000:01:00.0 > /sys/bus/pci/drivers/radeon/unbind
>> echo -n 0000:01:00.0 > /sys/bus/pci/drivers/pciback/new_slot
>> echo -n 0000:01:00.0 > /sys/bus/pci/drivers/pciback/bind
>>
>> to move my primary display adapter (01:00.0) to pciback for domU use. 
The
>> display, as expected, turned off.  I had enabled ssh to do this, and
have
>> access to the system still.  I then used
>>
>> echo -n 0000:01:00.0 > /sys/bus/pci/drivers/pciback/unbind
>> echo -n 0000:01:00.0 > /sys/bus/pci/drivers/radeon/bind
>>
>> to pass it back to dom0 at which point the display turned back on.
>> Thinking all was well, I made a script to do the above and executed it
(with
>> a sleep of 10 seconds after the first bind) at which point I lost the
>> display, as expected, but it didn''t come back.  I then logged
in through ssh
>> and find that the radeon driver is bound to 01:00.0, but still no
display.
>>
>> I think I need to tell the console to output to the radeon driver, but
I
>> have no idea how and a google search is proving fruitless.  Any help? 
I''ve
>> also noticed that unbinding/binding the radeon driver is now (it
wasn''t
>> initially) very slow to respond.
>>
>> Thanks!
>>
>
>

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

2011/10/29 Ben Himberg <bhimberg@gmail.com>:
> Turns out I passed "dynpm" to radeon/power_profile instead of
> radeon/power_method and it crashed atom bios (kept locking, timing out in
> the error log).  Works fine now.  How do I control, on the fly, where
kernel
> output goes though?
I think Linux is not able to re-wire the system console to another
device on the fly.
Not 100% sure, but so far I have not seen it done.
For your kind of experiments, it would be easiest to run over a serial
console, or use 2 graphics cards.

Florian

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users