Re: [PATCH] xen: do not clear and mask evtchns in __xen_evtchn_do_upcall

[Stefano Stabellini <stefano.stabellini@eu.citrix.com>]

On Mon, 4 Apr 2011, Konrad Rzeszutek Wilk wrote:
> > I reworded this part of the commit message, see below updated patch.
> > 
> [looking]
> .. snip..
> > > Can you expand? As in for the GSI? Or for the MSI/MSI-X?
> > 
> > Linux on native would use handle_edge_irq for edge irqs and msis, and
> > handle_fasteoi_irq for everything else.
> 
> What about per_cpu one?
> 

That is the simplest case, we are already using the same semantic as in
the native code.


> > > OK. You need a big comment about this in the code. Explain
> > > why this is happening. B/c if you look at this from code
> > > it seems like wrong thing to do for gsi's (as in you would
> > > think handle_level_irq would the right choice).
> > 
> > Except handle_level_irq is not used anymore anywhere in the kernel, give
> > a look at arch/x86/kernel/apic/io_apic.c:ioapic_register_intr.
> 
> 'make_8259A_irq' ? But yeah, I don't think we will hit machines with
> that architecture anymore.

Actually I was looking at:

setup_IO_APIC_irqs -> __io_apic_setup_irqs -> io_apic_setup_irq_pin



> > Updated patch, rebased on 2.6.39-rc1 follows:
> > 
> > ---
> > 
> > 
> > commit 6978531913b45abf3aff048475a2174a2cdaf288
> > Author: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
> > Date:   Tue Mar 29 14:15:06 2011 +0000
> > 
> >     xen: do not clear and mask evtchns in __xen_evtchn_do_upcall
> >     
> >     Change the irq handler of virqs and pirqs that don't need EOI (pirqs
> >     that correspond to physical edge interrupts) to handle_edge_irq.
> >     
> >     Use handle_fasteoi_irq for pirqs that need eoi (they generally
> >     correspond to level triggered irqs), no risk in loosing interrupts
> >     because we have to EOI the irq anyway.
> >     
> >     This change has the following benefits:
> >     
> >     - it uses the very same handlers that Linux would use on native for the
> >     same irqs (handle_edge_irq for edge irqs and msis, and
>                                                    ^^^-'MSIs/MSI-Xs'
> >     handle_fasteoi_irq for everything else);
> >     
> >     - it uses these handlers in the same way Linux would use them: it let
>                                                                            ^- 's'
> 
> >     Linux mask\unmask and ack the irq when Linux want to mask\unmask and ack
>                                                        ^- 's'
> >     the irq;
> >     
> >     See genericirq in the kernel docbook docs for more informations.
> 
> Say 'Documentation/DocBook/genericirq.tmpl'
> 
> [edit: The patch looks OK to me, but let me think about this some more over this
> week and sketch out the flow].

done