Re: current not very current (vs curr_vcpu)

[Mukesh Rathor <mukesh.rathor@oracle.com>]

On Fri, 19 Feb 2010 08:12:18 +0000
Keir Fraser <keir.fraser@eu.citrix.com> wrote:

> On 19/02/2010 04:21, "Mukesh Rathor" <mukesh.rathor@oracle.com> wrote:
> 
> > I noticed while debugging something that current is not pointing to
> > the current vcpu upon serial interrupt. The regs->SP clearly shows
> > 64bit dom0 stack, guest_mode(regs) returns 1, but current is
> > pointing to idle vcpu. I'm not able to figure how this is possible.
> > I can come up with scenario where dom0.vcpu yields cpu to idle vcpu
> > in which case curr_vcpu will point to dom0.vcpu and current to idle
> > vcpu. But in that case guest_mode(regs) will be false, and regs.SP
> > will show hyp stack. Correct?
> > 
> > Am I correct that if guest_mode() then current should always point
> > to guest vcpu? If yes, then I will debug this further.
> 
> The behaviour you see is expected. Guest_mode() is meaningless when
> running an idle vcpu. This is because the guest regs at the bottom of
> the stack are junk for an idle vcpu (and also we do lazy state
> synchronisation, so they may be the valid active regs for the last
> scheduled non-idle vcpu).
> 
>  -- Keir
> 

Yes, but my point is it doesn't appear to be running idle vcpu as
indicated by regs->rsp and regs->rip. They both point to dom0 context.
This from printk in ns16550_interrupt().

To rephrase the question, if regs->rip and regs->rsp show guest context
in do_IRQ(), then current must always point to guest vcpu, correct?

thanks,
Mukesh