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

Re: [Xen-devel] [Help] Trigger Watchdog when adding an IPI in vcpu_wake

On Sat, 2016-09-24 at 11:39 +0800, Wei Yang wrote:
> On Thu, Sep 22, 2016 at 11:12:15AM +0200, Dario Faggioli wrote:
> > _Almost_ correct. However, the problem is more that vcpu_wake() can
> > happen in response to an IRQ, and when you grab a spinlock in IRQ
> > context, you need to disable IRQs.
> > 
> Ok, now I have a question to this sentence.
> I have checked some posts which says in the interrupt handling, irq
> would be
> disabled. Well, this really depends on the implementation.
> This means in Xen, irq is still enabled during interrupt handling?
> Because of
> this, we have to disable IRQ when we need to grab the spin lock?
So, I don't think I'm getting all you're saying and asking.

The situation is like this: currently, vcpu_wake() can be called both
from inside or outside IRQ context, and both with IRQs enabled or

Given this situation, to be safe, we need to disable IRQs when taking
whatever spinlock vcpu_wake() (and the functions it calls, of course)
calls. Since it takes the scheduler's runq lock, this means we need to
take the lock with IRQ disabled.

And because of that, _everyone_ that takes the schedulr's lock must do
that with IRQs disabled. If we manage to *not* take the scheduler's
lock with IRQ disabled in vcpu_wake(), we then can do the same
everywhere else. But we can't do that with the current structure of the
code, and that's why we're thinking to defer the actual wakeup --i.e.,
the actual call to vcpu_wake()-- to a moment when:
 - IRQs are enabled,
 - we don't need to disable them.

This is it. About the following...

> Looks I almost get every thing. Let me do a summary to see whether I
> have got
> your words.
>    (spinlock usage convention in Xen)  &  (vcpu_wake() may happen in
> an IRQ)
This is ok.

> =>
>    (irq all disabled || irq all enabled) & (disable irq on grabbing
> spinlock)
This, I don't get it.

> =>
>    should grab schedule lock with IRQ disabled
Yes, sounds right.

<<This happens because I choose it to happen!>> (Raistlin Majere)
Dario Faggioli, Ph.D, http://about.me/dario.faggioli
Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)

Attachment: signature.asc
Description: This is a digitally signed message part

Xen-devel mailing list



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