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

Re: [Xen-devel] IPI sending difference between x86 and ARM



On Wed, 2014-09-10 at 11:27 +0100, Jan Beulich wrote:
> >>> On 10.09.14 at 12:19, <Ian.Campbell@xxxxxxxxxxxxx> wrote:
> > On Wed, 2014-09-10 at 11:10 +0100, Jan Beulich wrote:
> >> >>> On 10.09.14 at 12:01, <Ian.Campbell@xxxxxxxxxxxxx> wrote:
> >> > Anyway, if smp_send_event_check_mask is supposed to exclude the calling
> >> > CPU then we should fix that on ARM in the name of consistency/sanity. In
> >> > the event it causes other issues then we should fix those as they arise.
> >> 
> >> Actually I was instead considering to filter the local CPU in
> >> cpumask_raise_softirq(), just like cpu_raise_softirq() does. That
> >> would make changes to smp_send_event_check_mask()
> >> unnecessary.
> > 
> > Sure, that would be fine and good too.
> > 
> > But independent of that I think smp_send_event_check_mask() should
> > behave the same on x86 as ARM wrt including the current CPU.
> 
> But the dropping of the local CPU doesn't happen in
> smp_send_event_check_mask() on x86 - it's further down the
> call chain where this gets done.

I don't think this is relevant. smp_send_event_check_mask is the arch
agnostic interface and it should have some specified behaviour (which is
de facto to remove the local CPU from mask).

Whether the arch chooses to implement this at the top or the bottom of
the call chain from there on is up to the arch code.

>  And by fixing the caller we make
> it so that the specific arch's behavior doesn't matter anymore.

Except to any other callers, present and future.

Having generic arch interfaces which differ between architectures is
just unhelpful to everyone IMHO.

Ian.


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel


 


Rackspace

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