Re: [Xen-devel] [Xen-unstable] regression in pci passthrough to HVM guests due to commit 568da4f8c43d2e5b614964c6aefd768de3e3af14 "pt-irq fixes and improvements".

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 05.08.14 at 12:02, <linux@xxxxxxxxxxxxxx> wrote:
> But 2 questions:
> 
> 1) the:
>    (XEN) [2014-08-05 09:25:03.463] io.c:320: d17: unbind: m_gsi=87 g_gsi=16 
> dev=00:00.0 intx=0
>    (XEN) [2014-08-05 09:25:03.463] io.c:404: d17 final unmap: m_irq=87 
> dev=00:00.0 intx=0 
> 
>    That's not the device that is passed through, is that the unbind message 
> you 
>    where expecting due to the "pci_msitranslate=1" you were referring to 
> before ?

The message is bogus for the MSI case altogether - it prints data
from the PCI part of the union which isn't required to be filled in
at all. I'll have to further modify the patch.

> 2) And the second is about the logging:
> 
>    - It doesn't seem to print the binding of the msi irq's of pt-devices, i 
> only get 2 
>      messages of bindings which seem to be for the emulated xen pci device:
>      (XEN) [2014-08-05 09:46:05.355] io.c:280: d20: bind: m_gsi=37 g_gsi=36 
> dev=00.00.5 intx=0
>      (XEN) [2014-08-05 09:46:06.433] io.c:280: d20: bind: m_gsi=47 g_gsi=40 
> dev=00.00.6 intx=0
>      but in the guest i don't see any devices with those BDF's, looks like it 
> is 
>      printing some wrong values ?  

The absence of messages for binding is expected with current code;
bindings for other than passed through devices shouldn't make it here
at all, however.

>      at least it's not symmetric with what it is saying to unbind:
>     (XEN) [2014-08-05 09:47:51.658] io.c:320: d20: unbind: m_gsi=87 g_gsi=16 
> dev=00:00.0 intx=0
>     (XEN) [2014-08-05 09:47:51.658] io.c:404: d20 final unmap: m_irq=87 
> dev=00:00.0 intx=0
>     (XEN) [2014-08-05 09:47:51.658] io.c:320: d20: unbind: m_gsi=86 g_gsi=45 
> dev=00:02.0 intx=59
>     (XEN) [2014-08-05 09:47:51.658] io.c:404: d20 final unmap: m_irq=86 
> dev=00:02.0 intx=59
>     (XEN) [2014-08-05 09:47:51.658] io.c:320: d20: unbind: m_gsi=85 g_gsi=45 
> dev=00:02.0 intx=59
>     (XEN) [2014-08-05 09:47:51.658] io.c:404: d20 final unmap: m_irq=85 
> dev=00:02.0 intx=59
>     (XEN) [2014-08-05 09:47:51.658] io.c:320: d20: unbind: m_gsi=84 g_gsi=45 
> dev=00:02.0 intx=59
>     (XEN) [2014-08-05 09:47:51.658] io.c:404: d20 final unmap: m_irq=84 
> dev=00:02.0 intx=59
>     (XEN) [2014-08-05 09:47:51.658] io.c:320: d20: unbind: m_gsi=83 g_gsi=45 
> dev=00:02.0 intx=59
>     (XEN) [2014-08-05 09:47:51.658] io.c:404: d20 final unmap: m_irq=83 
> dev=00:02.0 intx=59
> 
>    - There are no mapping/unmapping messages for non-msi irq's ?

There should be for PCI ones, but there's not even implementation
for ISA ones.

Jan


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