On 14.05.2014 16:18, jordan wrote:

Hey Kimmo,

quoted

quoted

you can get the patch here; http://pastebin.com/VG7X5WjK ... it
applies over 3.14.3-rt5 [after you've already applied the -rt patch,
you apply it]. It's yielded working AMD Phenom systems for me
[although, i don't have the exact same H/W as you]. it might be worth
a try, just to see if it works.

I can confirm that applying the patch in http://pastebin.com/VG7X5WjK to
3.14.3-rt5 makes my system bootable even with AMD C1E enabled.
To clarify my previous email, those other tricks "processor.max_cstate=0",
"processor.nocst", etc... did not help.

Well, it looks like you are another person who's system is unstable
when those 3 upstream -rt patches are applied. Reverting [as you have
done now] should result in a usable/stable system for you [its worked
for myself and others for a few releases now. Maybe you can post back
on how that goes for you?]. Although, depending on config; ie: if you
are using no_hz_idle/full you may see "NOHZ: softirq pending 40"
[which one of the patches we reverted fixes, IIRC]. Myself, I don't
use NO_HZ_* and use the 'periodic timers' in the kernel, instead.

I use also the "periodic timers". There are quite many more who have the 
system effected by this problem. It's the feedback from Daphile users. 
Daphile is my "distro" using the RT kernel (www.daphile.com).

Actually I'm not able give you feedback on the longterm stability of the 
reverted version because the Phenom PC is my home server and I can use 
it only for quick testing. I'm hoping too that Paul or someone else 
finds a better solutions for this :-).

Kimmo

anyway, I am not sure how to fix this, beyond reverting / using that
patch. I know that some of the devs have had a hard time reproducing
the problem - maybe Paul will have better luck on another AMD system
[here is hoping]. But at least for now, you should be able to use that
patch.

jordan

quoted

Kimmo

quoted

quoted

OK, so I never got back to finding a system that could reproduce this,
but it sounds like it is still worthwhile; thanks for the reminder and
I'll go see if I can reproduce this on a quad since the six core was OK.

Hey Paul, I also forgot to get back to you ;) ... I had been carrying
over my .config - so i tried using defconfig too [just to see] and it
was no different. I was astonished that yours booted up just fine,
because i know of other AMD users [aside from Kimmo] who can't boot on
vainlla-rt... Again, 3.14.x-rtX boot and work just fine for me, by
reverting those patches. I had thought maybe -rt5 would boot [it
looked to have some substantial changes], but alas -rt5 also required
the patch for a working system.

Jordan

quoted

P.
--

quoted

Cheers
Kimmo

quoted

--
To unsubscribe from this list: send the line "unsubscribe
linux-rt-users" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html