[Xen-devel] Re: Xen, ustable and VGA passthrough

[Liwei <xieliwei@xxxxxxxxx>]

(I'm still unsure which list to post to, so I'm cross posting again.
Prehaps someone can recommend me which list to stick to?)

Hey,
    Just to tell you I've read your email and will keep you updated
when I get down to trying again. Currently the computer I'm trying to
passthrough on is my work computer, so it takes some motivation for me
to reboot it into xen. =)
    Meanwhile, you can check out these two email snippets I've
collected over the week (I've lost their sources, but they're part of
the threads pointed to in the VGAPassthrough wiki page).

----------------------------------------
(This one shows how to properly reserve the memory ranges)
----------------------------------------
I forgot to mention that you should change memory addresses in
xen-vBAR-pBAR.patch according to your MMIO BARs of your assigned gfx
card.

For example, the output of 'lspci -s 01:00.0 -v':

        01:00.0 VGA compatible controller: nVidia Corporation Unknown
device 05ff (rev a1) (prog-if 00 [VGA controller])
        Subsystem: nVidia Corporation Unknown device 0661
        Flags: bus master, fast devsel, latency 0, IRQ 16
        Memory at c2000000 (32-bit, non-prefetchable) [size=16M]
        Memory at e0000000 (64-bit, prefetchable) [size=256M]
        Memory at c0000000 (64-bit, non-prefetchable) [size=32M]
        I/O ports at 9c00 [size=128]
        Expansion ROM at bff00000 [disabled] [size=512K]
        Capabilities: [60] Power Management version 3
        Capabilities: [68] Message Signalled Interrupts: 64bit+
Queue=0/0 Enable+
        Capabilities: [78] Express Endpoint IRQ 0

you can see I reserve above memories in dsdt.asl in
xen-vBAR-pBAR.patch. you should replace them with memories of your
gfx.
----------------------------------------

----------------------------------------
(I'm having a problem similar to this one. My VGA ROM is actually
bigger than what xen allocated. I recall the first VGA passthrough
patch manually modified the range, so I'd try that. You might want to
check if your ROM size is larger than xen can handle too.)
----------------------------------------
I suspect your vga bios file is incorrect, its size is very small. I found
following VGA bios in xm_dmesg.log: ... (XEN) HVM1: BIOS map: (XEN)
HVM1: c0000-c07ff: VGA BIOS ...

You can check if the VGA bios is really loaded. You can use 'xxd vgabios-pt.bin'
to see the VGA bios content, of course you can see the size. After you apply the
patches I posted and mak the code, you can see vgabios_pt[] in
xen-unstable.hg/tools/firmware/hvmloader/roms.h. content of vgabios_pt[] should
be the same with output of 'xxd vgabios-pt.bin'.
----------------------------------------

    Do keep everyone updated on the Xen list.

Liwei


On 21 May 2011 23:58, Gennady Marchenko <gennady.marchenko@xxxxxxxxx> wrote:
> Hello again Liwei!
> Thanks for your patch it applied good :)
> With your patch I can run my ATI 5450 as primary GPU (I have no secondary at
> all) as basic VGA:
> Windows XP can start and show the desktop if it has no installed drivers. It
> is very big progress for me :)
> After I installed drivers it can't init the GPU. I think when you tries to
> install Win7 you got same - because it already have buildin drivers and
> tries to init extended display mode.
> My logs is:
> root@xen:/var/log/xen# tail -f /var/log/xen/qemu-dm-winxp.log
> (here domU started up)
> pt_iomem_map: e_phys=d0000000 maddr=d0000000 type=8 len=268435456 index=0
> first_map=1
> pt_bar_reg_read: first read BARs of gfx
> pt_bar_reg_read: first read BARs of gfx
> pt_iomem_map: e_phys=febc0000 maddr=febc0000 type=0 len=131072 index=2
> first_map=1
> pt_bar_reg_read: first read BARs of gfx
> pt_bar_reg_read: first read BARs of gfx
> pt_ioport_map: e_phys=e000 pio_base=e000 len=256 index=4 first_map=1
> pt_iomem_map: e_phys=f1000000 maddr=febfc000 type=0 len=16384 index=0
> first_map=1
> platform_fixed_ioport: changed ro/rw state of ROM memory area. now is rw
> state.
> platform_fixed_ioport: changed ro/rw state of ROM memory area. now is ro
> state.
>
>
>
> (here I see windows xp 'load' progress with logo, mayb it tries to init vga)
> pt_iomem_map: e_phys=ffffffff maddr=d0000000 type=8 len=268435456 index=0
> first_map=0
> pt_iomem_map: e_phys=ffffffff maddr=febc0000 type=0 len=131072 index=2
> first_map=0
> pt_ioport_map: e_phys=ffff pio_base=e000 len=256 index=4 first_map=0
> pt_iomem_map: e_phys=d0000000 maddr=d0000000 type=8 len=268435456 index=0
> first_map=0
> pt_iomem_map: e_phys=febc0000 maddr=febc0000 type=0 len=131072 index=2
> first_map=0
> pt_ioport_map: e_phys=e000 pio_base=e000 len=256 index=4 first_map=0
> pt_iomem_map: e_phys=ffffffff maddr=febfc000 type=0 len=16384 index=0
> first_map=0
> pt_pci_write_config: Warning: Guest attempt to set address to unused Base
> Address Register. [00:06.0][Offset:30h][Length:4]
> pt_iomem_map: e_phys=f1000000 maddr=febfc000 type=0 len=16384 index=0
> first_map=0
>
> (here it must start to show the desktop)
> pt_iomem_map: e_phys=ffffffff maddr=d0000000 type=8 len=268435456 index=0
> first_map=0
> pt_iomem_map: e_phys=ffffffff maddr=febc0000 type=0 len=131072 index=2
> first_map=0
> pt_ioport_map: e_phys=ffff pio_base=e000 len=256 index=4 first_map=0
> pt_iomem_map: e_phys=e0000000 maddr=d0000000 type=8 len=268435456 index=0
> first_map=0
> pt_iomem_map: e_phys=fbfe0000 maddr=febc0000 type=0 len=131072 index=2
> first_map=0
> pt_ioport_map: e_phys=e000 pio_base=e000 len=256 index=4 first_map=0
> (but something goes wrong :))
> reset requested in cpu_handle_ioreq.
> Issued domain 6 reboot
> It may be problem with memory map in dsdt.asl but I don't know how it may
> fix. :(
> Thanks a lot for your patch.
> Gennady.
>
> On Fri, May 20, 2011 at 10:08 AM, Liwei <xieliwei@xxxxxxxxx> wrote:
>>
>> Hello Gennady!
>>    Wasn't planning on posting the patch until it works. Also not sure
>> if it'll end up destroying stuff so, do take care! Not my fault if
>> unfortunate stuff happens!
>>    The patch was hastily created on a separate Ubuntu machine with
>> the source partition mounted, so I couldn't do a compile test nor am I
>> sure if all the changes I made are in the patch. Tell me if it doesn't
>> compile after applying the patch, I probably missed a file or two.
>>    Patch is based on the original VGA passthrough patches posted by
>> Weidong, modified to account for the ROMBIOS code being moved to its
>> own file. It includes the claim cycle patch for secondary graphics
>> card passthrough as well.
>>    Remember to place your graphics card's firmware into
>> tools/firmware/vgabios/ after each "make clean".
>>
>> Regards,
>> Liwei
>>
>>
>> On 20 May 2011 08:11, Gennady Marchenko <gennady.marchenko@xxxxxxxxx>
>> wrote:
>> > Hi Liwei!
>> > Some days ago you post a report to xen-devel about your tries to run
>> > your
>> > vga in pt mode on xen unstable.So could you share the last patched
>> > sources
>> > are you tried to work?
>> > Because I can't apply several patch to current unstable could you share
>> > it
>> > please?
>> > I'm already have done about IGD but can't PT the primary ati 5450 at all
>> > :(
>> > Best wishes,
>> > Gennady.
>
>

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