Port-xen archive

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

Re: X11 on intel

>>>>> "Jean-Yves" == Jean-Yves Migeon <jeanyves.migeon%free.fr@localhost> writes:

    Jean-Yves> Le 2015-09-25 11:23, Cherry G. Mathew a écrit :
    >> On 25 September 2015 at 13:01, Robert Swindells <rjs%fdy2.co.uk@localhost> wrote:
    >>> There isn't currently a way of getting accelerated X11 in a Xen
    >>> DOM0 for Intel GPUs. There is PR #49330 basically on this.
    >> I'm working on this (see previous email) - my patches would work
    >> on a machine with RAM <= the dma maximum, but as Manuel pointed
    >> out, our uvm and drm infrastructure itself needs changes to do
    >> this properly.
    >> I really need testing on those patches (which will most certainly
    >> panic()/corrupt memory on a >4GB RAM modern machine), but I need
    >> to see multiple crashes to help chase a few dangling pointers I
    >> haven't got yet.

    Jean-Yves> Any reliable way to be sure that pages are allocated in
    Jean-Yves> +4GiB region?  

Actually no - that was Manuel's point - so short of pulling out your RAM
until you get 4GB (I don't know if there's a way to limit xen itself to
4GB), my patches won't work in any way.

    Jean-Yves> I have a 12GiB RAM machine here and can 
    Jean-Yves> start it 32 bits, but allocations seem to happen always 
    Jean-Yves> in the lower 4GiB region for PA.

    Jean-Yves> dom0_mem=8192 is enough?

It's too much :-)

Basically, I'm trying to hunt down one or more m<->p translations in the
drm code. outside of that, the issues in PR/49330 are still wip.


Home | Main Index | Thread Index | Old Index