Re: [Xen-devel] [patch] Initialize xen_vcpu0 before initialize irq_ops

[Anthoine Bourgeois <anthoine.bourgeois@xxxxxxxxx>]

2011/11/24 Ian Campbell <Ian.Campbell@xxxxxxxxxx>:
> CC'ing Xen Liunx maintainers, please consult MAINTAINERS or
> use ./scripts/get_maintainer.pl.
>
> This local_irq_disable is interesting. Aren't IRQs supposed to already
> be disabled from entry to xen_start_kernel (really, since start of time)
> until at least this point?
>
> Enabling (or disabling) interrupts would require both xen_init_irq_ops()
> and xen_vcpu[0] to be setup, so it seems that either interrupts are not
> disabled at start of day (I'm fairly sure they are) or there is some
> magic code somewhere which does it directly without using the generic
> infrastructure (I can't find anything like that).
>
> So where do interrupts get enabled? Is before xen_init_irq_ops really
> early enough? I can't find anything between the old and new locations of
> this setup code which looks like it would enable them. It is possible
> that you just win the race on your slow systems now but that the window
> is still there.

Hum, you're right, there is something strange here.
I don't know why interrupts are enabled. I investigate and come back to
you later. Looks like my bug is elsewhere.
I'll CC Xen Liunx maintainers as you have advised me next time.
Thanks.

--
Anthoine

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