Re: [Xen-devel] Issues with Intel HD graphics 4000

[Konrad Rzeszutek Wilk <konrad.wilk@xxxxxxxxxx>]

On Wed, Jun 19, 2013 at 01:36:56PM +0200, Dietmar Hahn wrote:
> Am Dienstag 18 Juni 2013, 13:37:08 schrieb Dietmar Hahn:
> > Am Montag 17 Juni 2013, 09:14:30 schrieb Konrad Rzeszutek Wilk:
> > 
> > > Hey Dietmar,
> > > 
> > > Did you ever test the extra change for the debug patch?
> > > 
> > > http://lists.xenproject.org/archives/html/xen-devel/2013-05/msg02051.html
> > > "
> > > I tried both but none helped. I think it couldn't be expected as the first
> > > > patch handles an error case and the line with the second patch,
> > > > the call of pci_dma_sync_single_for_device(), gets not reached.
> > > 
> > > OK, perhaps move the pci_dma_sync_single_for_device in the while loop?
> > > 
> > > " ?
> > 
> > I believe I did and I think it didn't work but I'am not sure and so I'll try
> > again. But this will take some time as the PC is used for my daily work!
> 
> OK, tried again today and as expected it didn't work. Following code:
> 
> linux-3.7.10-1.11/drivers/char/agp/intel-gtt.c:
> 
> void intel_gtt_clear_range(unsigned int first_entry, unsigned int num_entries)
> {
>       unsigned int i;
> 
>       intel_private.driver->write_entry(intel_private.base.scratch_page_dma,
>                                                 first_entry, 0);
>       printk(KERN_WARNING "HAHN-%s: sync the first page\n", __func__);
>       pci_dma_sync_single_for_device(intel_private.pcidev,
>                                      intel_private.base.scratch_page_dma,
>                                      PAGE_SIZE, PCI_DMA_BIDIRECTIONAL);
> 
>       for (i = first_entry + 1; i < (first_entry + num_entries); i++) {
> 
>               
> intel_private.driver->write_entry(intel_private.base.scratch_page_dma,
>                                                 i, 0);
>               pci_dma_sync_single_for_device(intel_private.pcidev,
>                                              
> intel_private.base.scratch_page_dma,
>                                              PAGE_SIZE, 
> PCI_DMA_BIDIRECTIONAL);
>       }
>       printk(KERN_WARNING "HAHN-%s: End of loop\n", __func__);
>       readl(intel_private.gtt+i-1);
> }
> 
> The second printk() with "End of loop" is not reached.

That is unfortunate. I've another idea (that is to allocate said region
to use pci-alloc-coherent) - but before I go that route - could you give me
your machine details. Please include the 'dmesg' as well (I am curious to where
the intel-agp tries to allocate the region), and how much memory you allocate
to dom0. And naturally 'lspci' to get an idea of which Intel card it is.

Thanks!
> 
> Dietmar.
> 
> -- 
> Company details: http://ts.fujitsu.com/imprint.html

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel