IVTV and Xen

IVTV and Xen

[Jason]

Has anyone had any luck getting IVTV .61 working in xen (Im on 3.0 unstable changeset  Fri Mar 31
00:26:07 2006 +0100 9494:8a8361994904 because of some unrelated driver issues with another device).
I keep getting a DMA issue. If this is fixed in a later rev, I will upgrade but that will be a
pretty major undertaking on this particular machine, so I thought I would ask for others experience
first.   This is what I get when the module inserts, this ia DomU kernel BTW.

Apr 29 09:15:47 localhost kernel: ------------[ cut here ]------------
Apr 29 09:15:47 localhost kernel: kernel BUG at arch/i386/kernel/pci-dma-xen.c:298!
Apr 29 09:15:47 localhost kernel: invalid opcode: 0000 [#1]
Apr 29 09:15:47 localhost kernel: Modules linked in: ivtv
Apr 29 09:15:47 localhost kernel: CPU:    0
Apr 29 09:15:47 localhost kernel: EIP:    0061:[<c010aa9c>]    Not tainted VLI
Apr 29 09:15:47 localhost kernel: EFLAGS: 00010286   (2.6.16-xenU #12) 
Apr 29 09:15:47 localhost kernel: EIP is at dma_map_single+0x10c/0x170
Apr 29 09:15:47 localhost kernel: eax: 0000002f   ebx: 638fa000   ecx: 00000000   edx: fbfef000
Apr 29 09:15:47 localhost kernel: esi: 00000000   edi: 00008000   ebp: d6865cfc   esp: d6865cdc
Apr 29 09:15:47 localhost kernel: ds: 007b   es: 007b   ss: 0069
Apr 29 09:15:47 localhost kernel: Process modprobe (pid: 960, threadinfo=d6864000 task=d68115b0)
Apr 29 09:15:47 localhost kernel: Stack: <0>c0393d3c d6940000 d6865cfc c015531a 638fa000 d78b84c0
00000002 d6940000 
Apr 29 09:15:47 localhost kernel:        d6865d20 d9065043 c139a048 d6ab8000 00008000 00000002
00000000 00000000 
Apr 29 09:15:47 localhost kernel:        00008000 d6865d3c d9065152 d6940000 d68e4800 00000000
d6940000 00008000 
Apr 29 09:15:47 localhost kernel: Call Trace:
Apr 29 09:15:47 localhost kernel:  [<c01050fa>] show_stack_log_lvl+0xaa/0xe0
Apr 29 09:15:47 localhost kernel:  [<c010530e>] show_registers+0x18e/0x210
Apr 29 09:15:47 localhost kernel:  [<c0105509>] die+0xd9/0x180
Apr 29 09:15:47 localhost kernel:  [<c0105639>] do_trap+0x89/0xd0
Apr 29 09:15:47 localhost kernel:  [<c01059b8>] do_invalid_op+0xb8/0xd0
Apr 29 09:15:47 localhost kernel:  [<c0104d1f>] error_code+0x2b/0x30
Apr 29 09:15:47 localhost kernel:  [<d9065043>] ivtv_init_buffer+0xd3/0x1b0 [ivtv]
Apr 29 09:15:47 localhost kernel:  [<d9065152>] ivtv_init_queue_nolock+0x32/0xa0 [ivtv]
Apr 29 09:15:47 localhost kernel:  [<d9065215>] ivtv_init_queues_nolock+0x55/0x80 [ivtv]
Apr 29 09:15:47 localhost kernel:  [<d9066797>] ivtv_stream_alloc+0xd7/0x2b0 [ivtv]
Apr 29 09:15:47 localhost kernel:  [<d905f5d7>] ivtv_streams_setup+0x107/0x1c0 [ivtv]
Apr 29 09:15:47 localhost kernel:  [<d9059c33>] ivtv_probe+0x3b3/0xae0 [ivtv]
Apr 29 09:15:47 localhost kernel:  [<c0215879>] pci_call_probe+0x19/0x20
Apr 29 09:15:47 localhost kernel:  [<c02158d9>] __pci_device_probe+0x59/0x70
Apr 29 09:15:47 localhost kernel:  [<c021591f>] pci_device_probe+0x2f/0x60
Apr 29 09:15:47 localhost kernel:  [<c0236e99>] driver_probe_device+0x99/0xf0
Apr 29 09:15:47 localhost kernel:  [<c0236fef>] __driver_attach+0x6f/0x80
Apr 29 09:15:47 localhost kernel:  [<c0236234>] bus_for_each_dev+0x54/0x80
Apr 29 09:15:47 localhost kernel:  [<c0237027>] driver_attach+0x27/0x30
Apr 29 09:15:47 localhost kernel:  [<c0236810>] bus_add_driver+0x80/0xd0
Apr 29 09:15:47 localhost kernel:  [<c0237515>] driver_register+0x65/0xa0
Apr 29 09:15:47 localhost kernel:  [<c0215ba6>] __pci_register_driver+0x56/0x80
Apr 29 09:15:47 localhost kernel:  [<d905a986>] module_start+0xa6/0x120 [ivtv]
Apr 29 09:15:47 localhost kernel:  [<c0134c05>] sys_init_module+0x135/0x1c0
Apr 29 09:15:47 localhost kernel:  [<c0104ba9>] syscall_call+0x7/0xb
Apr 29 09:15:47 localhost kernel: Code: 5d f4 8b 45 f0 8b 75 f8 8b 7d fc 89 ec 5d c3 0f 0b 23 01 1c
3d 39 c0 e9 24 ff ff ff 8d 76 00 c7 04 24 3c 3d 39 c0 e8 e4 f4 00 00
  <0f> 0b 2a 01 1c 3d 39 c0 e9 69 ff ff ff c7 04 24 88 2c 39 c0 b8

-- 
Jason
The place where you made your stand never mattered,
only that you were there... and still on your feet

Re: IVTV and Xen

[Andy Clayton]

Jason wrote:
> Has anyone had any luck getting IVTV .61 working in xen (Im on 3.0 
> unstable changeset  Fri Mar 31
> 00:26:07 2006 +0100 9494:8a8361994904 because of some unrelated driver 
> issues with another device).
> I keep getting a DMA issue. If this is fixed in a later rev, I will 
> upgrade but that will be a
> pretty major undertaking on this particular machine, so I thought I 
> would ask for others experience
> first.   This is what I get when the module inserts, this ia DomU 
> kernel BTW.
>
It has been a while, but I think I had a similar error with ivtv that 
could be fixed by passing "swiotlb=force" to the kernel (extra in the 
configs).  This was with an unstable changeset in between 3.0.1 and 
3.0.2.  When I changed over to 3.0.2 I believe it then started to panic 
with the setting, but booted fine without.  Since I've been busy trying 
to get a FreeBSD domU to work and have it not panic every two seconds, I 
have yet to look into it further.  Additionally when using that setting 
I could capture video initially, but after some time it would lock up.  
I don't know if that issue was related.   Anyways it would seem as if 
something affecting the issue was changed in or around 3.0.2--assuming 
my memory is even partially correct.  That setting may still be worth a 
try, though, if you do not want to try upgrading.

Sorry for the continuous stream of guessing and I thinks---Hopefully it 
helps,
Andy Clayton

Re: IVTV and Xen

[Keir Fraser]

On 1 May 2006, at 06:20, Andy Clayton wrote:

>> Has anyone had any luck getting IVTV .61 working in xen (Im on 3.0 
>> unstable changeset  Fri Mar 31
>> 00:26:07 2006 +0100 9494:8a8361994904 because of some unrelated 
>> driver issues with another device).
>> I keep getting a DMA issue. If this is fixed in a later rev, I will 
>> upgrade but that will be a
>> pretty major undertaking on this particular machine, so I thought I 
>> would ask for others experience
>> first.   This is what I get when the module inserts, this ia DomU 
>> kernel BTW.
>>
> It has been a while, but I think I had a similar error with ivtv that 
> could be fixed by passing "swiotlb=force" to the kernel (extra in the 
> configs).  This was with an unstable changeset in between 3.0.1 and 
> 3.0.2.

We've never enabled swiotlb by default for domUs, even when they're 
driver domains. So you will definitely need 'swiotlb=force'. You could 
try 'swiotlb=16,force' for a 16MB swiotlb (the default is 64MB which 
may burn an unnecessary amount of RAM).

  -- Keir

Re: IVTV and Xen

[Jason]

It looks like swiotlb=16,force works, but just swiotlb=force causes my domu to panic.

-- 
Jason
The place where you made your stand never mattered,
only that you were there... and still on your feet

On Mon, 1 May 2006, Keir Fraser wrote:

>
> On 1 May 2006, at 06:20, Andy Clayton wrote:
>
>>> Has anyone had any luck getting IVTV .61 working in xen (Im on 3.0 
>>> unstable changeset  Fri Mar 31
>>> 00:26:07 2006 +0100 9494:8a8361994904 because of some unrelated driver 
>>> issues with another device).
>>> I keep getting a DMA issue. If this is fixed in a later rev, I will 
>>> upgrade but that will be a
>>> pretty major undertaking on this particular machine, so I thought I would 
>>> ask for others experience
>>> first.   This is what I get when the module inserts, this ia DomU kernel 
>>> BTW.
>>> 
>> It has been a while, but I think I had a similar error with ivtv that could 
>> be fixed by passing "swiotlb=force" to the kernel (extra in the configs). 
>> This was with an unstable changeset in between 3.0.1 and 3.0.2.
>
> We've never enabled swiotlb by default for domUs, even when they're driver 
> domains. So you will definitely need 'swiotlb=force'. You could try 
> 'swiotlb=16,force' for a 16MB swiotlb (the default is 64MB which may burn an 
> unnecessary amount of RAM).
>
> -- Keir
>
>

Re: IVTV and Xen

[Muli Ben-Yehuda]

On Mon, May 01, 2006 at 08:14:25PM -0500, Jason wrote:

> It looks like swiotlb=16,force works, but just swiotlb=force causes my domu 
> to panic.

How much memory are you giving it?

Cheers,
Muli

Re: IVTV and Xen

[Jason]

512MB, it runs pretty consistently around 300MB.

-- 
Jason
The place where you made your stand never mattered,
only that you were there... and still on your feet

On Tue, 2 May 2006, Muli Ben-Yehuda wrote:

> On Mon, May 01, 2006 at 08:14:25PM -0500, Jason wrote:
>
>> It looks like swiotlb=16,force works, but just swiotlb=force causes my domu
>> to panic.
>
> How much memory are you giving it?
>
> Cheers,
> Muli
>
>
>

Re: IVTV and Xen

[Keir Fraser]

On 2 May 2006, at 02:14, Jason wrote:

> It looks like swiotlb=16,force works, but just swiotlb=force causes my 
> domu to panic.
>

Makes sense if it's a small domU.

  -- Keir

Re: IVTV and Xen

[jason]

The swiotlb issue does cause a panic on my rev. I think I must be on that
same middle ground that you were on.


--- Andy Clayton <xen.nospam@q3aiml.net> wrote:
> Jason wrote:
> > Has anyone had any luck getting IVTV .61 working in xen (Im on 3.0 
> > unstable changeset  Fri Mar 31
> > 00:26:07 2006 +0100 9494:8a8361994904 because of some unrelated driver 
> > issues with another device).
> > I keep getting a DMA issue. If this is fixed in a later rev, I will 
> > upgrade but that will be a
> > pretty major undertaking on this particular machine, so I thought I 
> > would ask for others experience
> > first.   This is what I get when the module inserts, this ia DomU 
> > kernel BTW.
> >
> It has been a while, but I think I had a similar error with ivtv that 
> could be fixed by passing "swiotlb=force" to the kernel (extra in the 
> configs).  This was with an unstable changeset in between 3.0.1 and 
> 3.0.2.  When I changed over to 3.0.2 I believe it then started to panic 
> with the setting, but booted fine without.  Since I've been busy trying 
> to get a FreeBSD domU to work and have it not panic every two seconds, I 
> have yet to look into it further.  Additionally when using that setting 
> I could capture video initially, but after some time it would lock up.  
> I don't know if that issue was related.   Anyways it would seem as if 
> something affecting the issue was changed in or around 3.0.2--assuming 
> my memory is even partially correct.  That setting may still be worth a 
> try, though, if you do not want to try upgrading.
> 
> Sorry for the continuous stream of guessing and I thinks---Hopefully it 
> helps,
> Andy Clayton
>