[Code Trace] dom0 Nouveau accelerated mmap problem

[Code Trace] dom0 Nouveau accelerated mmap problem

[Arvind R]

Hi,
In continuation of my 'Nouveau on dom0' posting, I dug into the code ...
As Konrad had predicted, it is the discontinuous PFN in the mmapped
area that seems to the problem.

booted bare hardware
X: libdrm:OUT_RING: Enter: chan=0x8170a0, id=2, data=0x44000,
chan->cur=0x7fcdaff5e000
*(chan->cur++) = (data);
X: libdrm:OUT_RING: Exit
<some time later>
kernel: ttm_bo_vm_fault: bo=ffff88007ccfb200
kernel: ttm_bo_vm_fault: is_iomem=0, page_offset=0x0, page_last=0x10,
bo->num_pages=16
kernel: ttm_bo_vm_fault: vma=ffff88007b493870, address=0x7fcdaff5e000,
pfn=0x73f95, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88007b493870, address=0x7fcdaff5f000,
pfn=0x73f96, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88007b493870, address=0x7fcdaff60000,
pfn=0x73f97, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88007b493870, address=0x7fcdaff61000,
pfn=0x73f98, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88007b493870, address=0x7fcdaff62000,
pfn=0x73f99, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88007b493870, address=0x7fcdaff63000,
pfn=0x73f9a, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88007b493870, address=0x7fcdaff64000,
pfn=0x73f9b, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88007b493870, address=0x7fcdaff65000,
pfn=0x73f9c, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88007b493870, address=0x7fcdaff66000,
pfn=0x73f9d, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88007b493870, address=0x7fcdaff67000,
pfn=0x73f9e, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88007b493870, address=0x7fcdaff68000,
pfn=0x73f9f, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88007b493870, address=0x7fcdaff69000,
pfn=0x73fa0, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88007b493870, address=0x7fcdaff6a000,
pfn=0x73fa1, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88007b493870, address=0x7fcdaff6b000,
pfn=0x73fa2, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88007b493870, address=0x7fcdaff6c000,
pfn=0x73fa3, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88007b493870, address=0x7fcdaff6d000,
pfn=0x73fa4, ret=0

booted Xen
X: libdrm:OUT_RING: Enter: chan=0x8170a0, id=2, data=0x44000,
chan->cur=0x7faed9add000
<some time later, repeats never-ending till shutdown, OUT_RING never exits>
kernel: ttm_bo_vm_fault: bo=ffff88000e79ba00
kernel: ttm_bo_vm_fault: is_iomem=0, page_offset=0x0, page_last=0x10,
bo->num_pages=16
kernel: ttm_bo_vm_fault: vma=ffff88000e691918, address=0x7faed9add000,
pfn=0xe6db, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88000e691918, address=0x7faed9ade000,
pfn=0xe718, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88000e691918, address=0x7faed9adf000,
pfn=0xe772, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88000e691918, address=0x7faed9ae0000,
pfn=0xe778, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88000e691918, address=0x7faed9ae1000,
pfn=0xe6eb, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88000e691918, address=0x7faed9ae2000,
pfn=0xe738, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88000e691918, address=0x7faed9ae3000,
pfn=0xe6da, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88000e691918, address=0x7faed9ae4000,
pfn=0xe710, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88000e691918, address=0x7faed9ae5000,
pfn=0xe76e, ret=0
kernel: ttm_bo_vm_fault: vma=ffff88000e691918, address=0x7faed9ae6000,
pfn=0xe7ad, ret=0

ttm_bo_vm_fault() is in drivers/gpu/drm/ttm/ttm_bo_vm.c in the kernel tree.

HELP!
Arvind R.

Re: [Code Trace] dom0 Nouveau accelerated mmap problem

[Konrad Rzeszutek Wilk]

On Sun, Feb 28, 2010 at 01:01:16PM +0530, Arvind R wrote:
> Hi,
> In continuation of my 'Nouveau on dom0' posting, I dug into the code ...
> As Konrad had predicted, it is the discontinuous PFN in the mmapped
> area that seems to the problem.

<groan>
I had hoped it would not be...

So the is_iomem flag is set to zero, right? If that is the case, did
this patch:
http://lists.xensource.com/archives/html/xen-devel/2009-10/msg00986.html

help in any regard? And if it did, what did your ttm_bo_vm_fault start
spitting out?