On 2017-06-26 18:26, Gustavo Lima Chaves wrote:
> * Henning Schild <[email protected]> [2017-06-26 09:55:59 +0200]:
> 
>> Am Fri, 23 Jun 2017 14:56:58 -0700
>> schrieb Gustavo Lima Chaves <[email protected]>:
>>
>>> * Jan Kiszka <[email protected]> [2017-06-23 22:36:35 +0200]:
>>>
>>>> On 2017-06-23 19:33, Gustavo Lima Chaves wrote:  
>>>>> [...]
>>>>>   
>>>>>>>
>>>>>>> When I artificially block this interrupt after a while, I see
>>>>>>> the real time jitter behaving almost perfectly.
>>>>>>>
>>>>>>>
>>>>>>> Can you see how interrupt SPI storm ,though to the root cell,
>>>>>>> might damage RT jitter on another cell? 
>>>>>>
>>>>>> The interrupt storm may be just one symptom of load on shared
>>>>>> resources. Maybe the GPU or its driver are also issuing a high
>>>>>> load of memory accesses or are otherwise stressing the shared
>>>>>> interconnects.
>>>>>>
>>>>>> Looking at the hypervisor, there should be no interference
>>>>>> between two cells if one receives lots of interrupts.
>>>>>> Specifically, there is only one shared lock between both in
>>>>>> irqchip.c (dist_lock), but it's taken only for a very short
>>>>>> read-modify-write access.
>>>>>>
>>>>>> Jan  
>>>>>
>>>>> What about Linux inmate cells on Intel? Has anybody else seem
>>>>> horrid timer resolutions (like ".resolution: 1000000 nsecs") as
>>>>> well? The aforementioned ARM patch makes no sense for the x86's
>>>>> version of time.c, I guess. Should we be providing HPET address
>>>>> to the inmate the same way it's done for PM timer address? Any
>>>>> other ideas?  
>>>>
>>>> What's your setup exactly?  
>>>
>>> Maybe it's something in my inmate Linux config? I've tried with two
>>> codebases there: RT Linux and upstream (both with Jailhouse patches,
>>> naturally). I'll attach one of the configs.
>>>
>>>>
>>>> There is no need to use the slow HPET on x86 anymore, with or
>>>> without Jailhouse. We now have the local APIC as reliable
>>>> high-resolution and high-performance (because it's core-local)
>>>> timer.  
>>>
>>> OK, fair. Still wondering what's giving that final bad resolution. The
>>> rest of the setup is a Fedora 25 system on QEMU, kernel 4.8, with
>>> pristine qemu-vm and linux-x86-demo cells running.
>>
>> So you are running Jailhouse and the rt-preempt non-root cell on qemu?
>> Getting solid rt-performance there might be tricky, that should not be
>> your starting point when looking at realtime!
> 
> Yeah, I know QEMU is the place to get RT performance, but I have my
> kernel already in place for bare metal testing. Going forward with
> those timer resolutions would be a no-go, though, thus why I wanted to
> sort that out in the first place. Thanks for the reminder anyway!

If you tested inside QEMU, maybe you didn't provide the guest all CPU
features it desired. -cpu kvm64,...+arat?

Jan

-- 
Siemens AG, Corporate Technology, CT RDA ITP SES-DE
Corporate Competence Center Embedded Linux

-- 
You received this message because you are subscribed to the Google Groups 
"Jailhouse" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to