Re: [Qemu-devel] Re: [PATCH] spice: add qxl device

["Michael S. Tsirkin" <mst@redhat.com>]

On Thu, Nov 18, 2010 at 01:55:29PM +0200, Gleb Natapov wrote:
> On Thu, Nov 18, 2010 at 01:33:08PM +0200, Michael S. Tsirkin wrote:
> > On Thu, Nov 18, 2010 at 11:57:51AM +0200, Gleb Natapov wrote:
> > > On Thu, Nov 18, 2010 at 11:30:27AM +0200, Michael S. Tsirkin wrote:
> > > > On Thu, Nov 18, 2010 at 11:11:49AM +0200, Gleb Natapov wrote:
> > > > > On Thu, Nov 18, 2010 at 11:03:21AM +0200, Michael S. Tsirkin wrote:
> > > > > > On Thu, Nov 18, 2010 at 10:09:35AM +0200, Gleb Natapov wrote:
> > > > > > > On Wed, Nov 17, 2010 at 08:00:08PM +0200, Michael S. Tsirkin wrote:
> > > > > > > > > >>  If so: does qemu
> > > > > > > > > >>emulate this correctly?
> > > > > > > > > >
> > > > > > > > > >It mostly does.
> > > > > > > > > 
> > > > > > > > > I doubt it actually enables/disables the legacy vga ports.
> > > > > > > > 
> > > > > > > > I'll check when I have the time. We can fix it if it doesn't,
> > > > > > > > 
> > > > > > > So many guests (all of them?) just assume that vga ports and
> > > > > > > framebuffer is there.
> > > > > > 
> > > > > > Why do you think they disable io memory then?
> > > > > > 
> > > > > Who and how and when disables io memory?
> > > > 
> > > > I think guest will do this if you disable the device through the device
> > > > manager. This might need a reboot to become effective.
> > > > 
> > > Try to do it with primary VGA adapter and tell us what happens :)
> > > 
> > > > > Some guests are designed to run
> > > > > even on old ISA machines that have no way to disable anything. The
> > > > > device is just there.
> > > > > 
> > > > > This is the same with IDE ports. BIOS "knows" legacy ISA ports and just
> > > > > program them into PCI IO bars to be nice.
> > > > 
> > > > HAven't checked IDE, for VGA AFAIK BIOS does not program legacy ports in
> > > > the card, they are hardwired there. However, the card must not claim any
> > > > io transactions if IO memory is disabled in command register.
> > > > 
> > > Is this correct also for legacy ports?
> > 
> > Yes. The spec is quite explicit on this point:
> > 
> > A function that supports a PC legacy function (IDE, VGA, etc.) is
> > allowed to claim those addresses associated with the specific function
> > when the I/O Space (see Figure 6-2) enable bit is set.  These addresses
> > are not requested using a Base Address register but are assigned by
> > initialization software. If a device identifies itself as a legacy
> > function (class code), the initialization software grants the device
> > permission to claim the I/O legacy addresses by setting the device’s I/O
> > Space enable bit.
> > 
> > 
> What do they mean by "initialization software".

BIOS or OS.

> How addresses is
> assigned by initialization software without use of Base Address register? 

As it says:
" If a device identifies itself as a legacy
" function (class code), the initialization software grants the device
" permission to claim the I/O legacy addresses by setting the device’s
" I/O
" Space enable bit.

So you look at the class code and know which addresses will be claimed.
The relevant table is in the appendix D, take a look there.

> Looks like "initialization software" is something internal to HW.

Not really. Seabios does this simply by enabling io unconditionally.
It could easily detect multiple VGA cards and only enable one.

> And what spec says about legacy mmio?

What do you want to know?

> > > This wouldn't be backwards
> > > compatible to ISA machines, so old software my not run properly back in
> > > the days when transaction from ISA to PCI happened.
> > 
> > initialization software could be the BIOS.
> > So maybe BIOS update was needed in the transition.
> > 
> That is possible.
> 
> > > So my guess is that
> > > old ISA ports works in backwards compatible way.
> > 
> > The spec seems to contradict this.
> > 
> > > > When qemu is started, it works correctly: the io memory is disabled and card does
> > > > not claim any io. Then BIOS comes along and enables io. At this point
> > > > map callback is invoked and maps io memory, card starts claiming io.
> > > Looking at the code I see that cirrus claims all IO ports and
> > > framebuffer memory during init function unconditionally.
> > 
> > So that may be OK for ISA, but not for PCI.
> > 
> The code does it for both.

Yep. So it's a bug.

> > > > 
> > > > What is broken is that if BIOS/guest then disables IO memory,
> > > > (I think - even if guest is rebooted!) we will keep claiming IO transactions.
> > > > That our emulation does this seems to be a clear spec violation, we are
> > > > just lucky that BIOS/guest does not do this at the moment.
> > > > 
> > > > 
> > > > 
> > > > > > > So what "fixing" this will buy us?
> > > > > > 
> > > > > > Besides spec compliancy, you mean?  Ability to support multiple VGA
> > > > > > cards. That's how it works I think: BIOS enables IO on the primary
> > > > > > VGA device only.
> > > > > > 
> > > > > What spec defines hot-plug for primary VGA adapter?
> > > > 
> > > > No idea about hotplug. I am talking about multiple VGA cards,
> > > > enabling/disabling them dynamically should be possible.
> > > Of course. With properly designed VGA card you should be able to have
> > > more then one,
> > 
> > And, for that to have a chance to work when all cards are identical, you
> > don't claim IO when IO is disabled.
> > 
> But then only one card will be able to use IO since enabling IO on more
> the one cards will cause conflict.

Sure. That's life for legacy io though.

> > > but one of them will provide legacy functionality
> > > and is not removable.
> > 
> > The guest might not support hotplug. But there's no way
> > it can prevent surprise removal. qemu should not crash
> > when this happens.
> Qemu can prevent any removal, surprise or not. Qemu can just
> disallow device removal.

Yes, but that won't emulate real hardware faithfully.
On real hardware with a hotplug supporting slot
(and without an EM lock :) ) you can yank the card out
and the guest can do nothing about it.

> --
> 			Gleb.