Bug 1265744

Summary: multiple guests run cyclictest tools cause guest CPU stall
Product: Red Hat Enterprise Linux 7 Reporter: FuXiangChun <xfu>
Component: kernel-rtAssignee: Clark Williams <williams>
Status: CLOSED NOTABUG QA Contact: Jiri Kastner <jkastner>
Severity: high Docs Contact:
Priority: unspecified    
Version: 7.2CC: bhu, chayang, juzhang, lcapitulino, michen, qzhang
Target Milestone: rc   
Target Release: ---   
Hardware: x86_64   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2015-09-24 16:33:49 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:

Comment 4 Luiz Capitulino 2015-09-23 19:42:28 UTC
If you're starting two guests with real-time vCPUs pinned to the same cores in the host, that will deadlock.

Generally speaking, each host core isolated for real-time should run only one real-time vCPU.

Comment 5 FuXiangChun 2015-09-24 16:10:48 UTC
(In reply to Luiz Capitulino from comment #4)
> If you're starting two guests with real-time vCPUs pinned to the same cores
> in the host, that will deadlock.
> 
> Generally speaking, each host core isolated for real-time should run only
> one real-time vCPU.

I see what you mean. I tested this scenario. Both guests work. 

In addition.
 1. Customer also possible to start multiple guests with real-time vCPUs pinned to the same cores in the host. Do we have a relevant way to prevent customer? or let they know the right way. 

 2. Can I use cores do real-time test are not isolated in host?
 
 e.g
cores 1 3 5 7(belong to node1) are isolated in host. vCPUs pinned to these cores. guest1 use them to do real-time test.

cores 0 2 4 6(belong to node0) are not isolated in host. vCPUs pinned to these cores. guest2 use them do do real-time test inside guest.

Comment 6 Luiz Capitulino 2015-09-24 16:33:49 UTC
(In reply to FuXiangChun from comment #5)

> In addition.
>  1. Customer also possible to start multiple guests with real-time vCPUs
> pinned to the same cores in the host. Do we have a relevant way to prevent
> customer? or let they know the right way. 

No, there's no way to prevent this in the real-time kernel. We can write documentation warning about this, but there's no way to prevent it.

>  2. Can I use cores do real-time test are not isolated in host?

You may have higher latency due to other tasks re-activating the tick and/or pushing work to workqueues.

I'm closing this as NOTABUG just to avoid BZs that are not real issues hanging around.