[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Xen-users] One question about vcpu-wake up in credit scheduler


  • To: xen-users@xxxxxxxxxxxxxxxxxxx
  • From: you_bear zhou <youbear.zhou@xxxxxxxxx>
  • Date: Mon, 11 May 2009 13:09:49 -0400
  • Delivery-date: Mon, 11 May 2009 10:10:29 -0700
  • Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:date:message-id:subject:from:to:content-type :content-transfer-encoding; b=M9XU6XNC9P+K1uXhZF1BLhQVZRkbmZoRF0M5tP5ilcr0iDHFDTy+iT883V1+j/exUJ FJTpQSym50J64Ej5eMJxoZKMY353AaAQ0LFnGb5gcEnKMTyI9SC5CYmRpOlwGdbRPiqK n/RcVe0KM7kbCTdkrn9BG2YMD3yLYbS+IXP4w=
  • List-id: Xen user discussion <xen-users.lists.xensource.com>

Hei guys,

I'm now trying to do an attacking to credit scheduler ( Xen 3.2.0).

As far as I know, the credit scheduler ticks the current runnable vcpu
each 10ms, and a vcpu can wake up with BOOST priority if it was in
UNDER state before waking up.

What I'm trying to do is modifying the cpu_idle loop in linux kernel,
let the idle process first spins for 9ms and then invokes halt(). I
expect the vcpu be blocked right before the tick bound and wakes up
right after that, so it can keep using CPU without paying credits.

However, I found the vcpu seldom woke up. Then I look into the
cpu_idle(), it checks if the cpu needs to reschedule, if so, then
stops the idle loop.

when I change the spin interval to 25ms, it did wake up frequently. So
I think the interrupts should come from the scheduler.

But why the vcpu didn't wake up with 9ms spin interval??

Can anyone explain this to me ?


Best,
Tina

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.