Michael Zanetti wrote:
> Hi Simen, Rey,

Hi Michael, Rey,

I've been able to confirm on my own that this issue is not limited to 
SMP-kernels;
My 2.6.17.6 kernel is susceptible in both SMP and UP modes. My CentOS 2.6.9 
kernel is not usable in SMP mode (unrelated timer/motherboard problems), and 
  on the 2.6.9 UP kernel I had to disable the cpuspeed service (load-based 
clock throttling) as loading the single-CPU powernow-modules dies rather 
horribly on my X2 kernel. Still, i did /not/ see the problem with the 2.6.9 
kernel, and I feel the test was running for a sufficient length.

I saw Kristo (on the list) mentioning that this had been tied to CPU 
throttling, so I'm currently back to the 2.6.17 SMP-kernel, unmodified 0.7.0 
ivtv-source but cpuspeed disabled. As it was with this setup that I first 
noticed the errors, I think I will do so again. I've set up a regimen to 
grab and encode a lot more data tv-time than I usually do, so I hope the 
error will show up quickly.

Michael, Rey, can you confirm that you are running with a speed management 
utility, or (possibly) that your kernels are compiled with support for speed 
management?

This should show as the presence of the
/sys/devices/system/cpu/cpu0/cpufreq/
subtree (I'm not sure if there is a similar /proc) interface.

If so, perhaps you too could give running without speed managment a try?

I'll report back when I have run a while with this, or have seen errors.

Yours,
-S




> Am 05.08.2006 um 12:46 schrieb Simen Thoresen:
> 
>> Hi Michael, Rey,
>>
>> I have not seen any new replies to this, nor anything from any 
>> developers. As I wrote in my problem report, I've been an ivtv-user 
>> for a while, but only recently subscribed to the list - thus I don't 
>> know what to expect from developer interest and such.
>>
>> Still - assuming the developers are reading the list and and are 
>> interested in the problem, I thought I'd try to summarise our systems 
>> so that we can get similarities or differences sorted out.
>>
>> My system (as previously written) is an Asus A8N-SLI Deluxe 
>> motherboard, with an Athlon64 X2 chip, running in i386 mode, with ivtv 
>> 0.7.0 on a 2.6.17.6 kernel.
> 
> My system:
> ASUS A8V (VIA-chipset).
> tried kernel 2.6.14 and 2.6.15 (I cannot use >=2.6.16 because of missing 
> support from dmraid)
> tried ivtv-0.4.x
> 
> The best results I got with 0.4.0. The system stays up for about 2 days. 
> The worst case comes with 0.4.5. ivtv crashes after about 10 minutes.
> 
> 
>>
>> This means
>> -NForce4 (I would not thing that mobo brand or chipset features matter 
>> too much)
>> -SMP (This may be important. I've run my card with ivtv-0.4.x 
>> 'forever' in UP-mode on older kernels without issues)
>> -kernel 2.6.17 + ivtv 0.7.0
>>
>>
>> Michael;
>> From your mail, I understand you are running an A8V board (That would 
>> be a ViA K8T800pro-board, right?). Are you running an X2 CPU in SMP on 
>> this? In i386 and not x86-64 mode?
> 
> Yes, it is the A8V with the mentioned chipset. I do not have an X2 CPU 
> and tried both, SMP and not SMP. I tried a fresh x86 install and a 
> x86-64 install. I still have both of them on my drive so I can test them 
> both.
> 2 weeks ago I put my old HD from my PIII mythbox into this PC and tested 
> with this. For two days I could run the system without one error. 
> Unfortunately I had a disk headcrash and the working system is now away 
> (I know I'm too lazy with backups).
> 
>> Also, you're writing that ivtv 'crashes' - are we talking about the 
>> same DMA error;
>> warning: ENC: (0) DMA Error 0x0000000b
>> reported some or many times during a capture?
> 
>> The actual error number may be important, altho I don't know what it 
>> signifies.
>>
> 
> Right. It is the same Error and also the number matches. The message 
> appears just as the System load goes up. For example when running one 
> recording and mythcommflag or running 2 recordings and watching one at 
> the same time. I couldn't find a certain point of CPU load causing the 
> problems. Probably the DMA-access from the hard disk and the ivtv-cards 
> at the same time are disturbing each other.
> 
>> Altho my errors were a lot less frequent that I have seen others 
>> report, they never caused ivtv to stop working, nor impaired system 
>> stability.
>>
>>
> My system throws out some DMA errors and then the one:
> 
> ivtv: DMA still pending while stopping capture.
> 
> After this message the card is unusable and I have to reboot the system. 
> Unloading and reloading the ivtv module doesn't help.
> 
>>
>> Rey;
>> You write that you have the same setup as I have. Could you please 
>> expand on that?
>> Did the patch work for you as well as it does for me?
> 
> Of course the patch works well since DMA for the ivtv-cards is totally 
> switched off. Unfortunately its not a solution for me since I cannot 
> watch any recordings when both cards are capturing because of the CPU load.
>>
>>
>>
>> So - assuming Michael is experiencing the same problem I am, it would 
>> seem that his is /not/ a chipset issue (or, if it is, both NForce4 and 
>> K8T800pro chipsets are affected), but the common line is that it 
>> applies to Athlon64 CPUs, possibly SMP setups.
> 
> I agree with the chipsets but its not the SMP setup.
> 
>>
>> From my own experience from my own setup, I feel pretty confident that 
>> the the main change I introduced was the SMP CPU. I see others have 
>> reported this issue with other kernels and ivtv-versions, but good 
>> information is typically missing.
>>
>> With a little spare time, I will build an UP kernel and give that a 
>> spin with an unmodified ivtv-driver. Given that my errors never were 
>> very frequent, testing will take time to be conclusive.
> 
> 
> I did some more testing but couldn't find out any setups working fine.
> 
> As mentioned above I think it could have to do something with the DMA 
> transfers of the hard disks. I am using a SATA I -Raid 0 setup. Are you 
> using SATA and/or RAID?
> 
> Hope this helps. Let me know if you need some more information. 
> Currently I'm not at home so I cannot do any testing for the next week.
> 
> Michael
> 

-- 
Simen Thoresen, Dolphin ICS

_______________________________________________
ivtv-users mailing list
[email protected]
http://ivtvdriver.org/mailman/listinfo/ivtv-users

Reply via email to