Hi Thomas,
On 8/11/2020 2:53 AM, Thomas Gleixner wrote:
Thomas Gleixner <tglx@xxxxxxxxxxxxx> writes:
CC+: XEN folks
Thomas Gleixner <tglx@xxxxxxxxxxxxx> writes:
The infrastructure itself is not more than a thin wrapper around the
existing msi domain infrastructure and might even share code with
platform-msi.
And the annoying fact that you need XEN support which opens another can
of worms...
hmm I am not sure why we need Xen support... are you referring to idxd
using xen?
which needs some real cleanup first.
x86 still does not associate the irq domain to devices at device
discovery time, i.e. the device::msi_domain pointer is never populated.
So to support this new fangled device MSI stuff we'd need yet more
x86/xen specific arch_*msi_irqs() indirection and hackery, which is not
going to happen.
The right thing to do is to convert XEN MSI support over to proper irq
domains. This allows to populate device::msi_domain which makes a lot of
things simpler and also more consistent.
do you think this cleanup is to be a precursor to my patches? I could
look into it but I am not familiar with the background of Xen
and this stuff. Can you please provide further guidance on where to look?
Thanks,
tglx
