[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [patch RFC 26/38] x86/xen: Wrap XEN MSI management into irqdomain
- To: Jürgen Groß <jgross@xxxxxxxx>, LKML <linux-kernel@xxxxxxxxxxxxxxx>
- From: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
- Date: Tue, 25 Aug 2020 09:57:44 +0200
- Cc: x86@xxxxxxxxxx, linux-pci@xxxxxxxxxxxxxxx, xen-devel@xxxxxxxxxxxxxxxxxxxx, Joerg Roedel <joro@xxxxxxxxxx>, iommu@xxxxxxxxxxxxxxxxxxxxxxxxxx, linux-hyperv@xxxxxxxxxxxxxxx, Haiyang Zhang <haiyangz@xxxxxxxxxxxxx>, Jon Derrick <jonathan.derrick@xxxxxxxxx>, Lu Baolu <baolu.lu@xxxxxxxxxxxxxxx>, Wei Liu <wei.liu@xxxxxxxxxx>, "K. Y. Srinivasan" <kys@xxxxxxxxxxxxx>, Stephen Hemminger <sthemmin@xxxxxxxxxxxxx>, Steve Wahl <steve.wahl@xxxxxxx>, Dimitri Sivanich <sivanich@xxxxxxx>, Russ Anderson <rja@xxxxxxx>, Bjorn Helgaas <bhelgaas@xxxxxxxxxx>, Lorenzo Pieralisi <lorenzo.pieralisi@xxxxxxx>, Konrad Rzeszutek Wilk <konrad.wilk@xxxxxxxxxx>, Boris Ostrovsky <boris.ostrovsky@xxxxxxxxxx>, Stefano Stabellini <sstabellini@xxxxxxxxxx>, Marc Zyngier <maz@xxxxxxxxxx>, Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>, "Rafael J. Wysocki" <rafael@xxxxxxxxxx>, Megha Dey <megha.dey@xxxxxxxxx>, Jason Gunthorpe <jgg@xxxxxxxxxxxx>, Dave Jiang <dave.jiang@xxxxxxxxx>, Alex Williamson <alex.williamson@xxxxxxxxxx>, Jacob Pan <jacob.jun.pan@xxxxxxxxx>, Baolu Lu <baolu.lu@xxxxxxxxx>, Kevin Tian <kevin.tian@xxxxxxxxx>, Dan Williams <dan.j.williams@xxxxxxxxx>
- Delivery-date: Tue, 25 Aug 2020 07:57:56 +0000
- List-id: Xen developer discussion <xen-devel.lists.xenproject.org>
On Mon, Aug 24 2020 at 08:21, Jürgen Groß wrote:
> On 21.08.20 02:24, Thomas Gleixner wrote:
>
> Looking into https://www.kernel.org/doc/Documentation/IRQ-domain.txt (is
> this still valid?)
It's halfways correct. Emphasis on halfways.
> I believe Xen should be able to use the "No Map" approach, as Xen only
> ever uses software IRQs (at least those are the only ones visible to
> any driver). The (virtualized) hardware interrupts are Xen events
> after all.
>
> So maybe morphing Xen into supporting irqdomains in a sane way isn't
> that complicated. Maybe I'm missing the main complexities, though.
The wrapper domain I did is pretty much that, but with the extra
functionality required by hierarchical irq domains. So, yes it's
functionally correct, but it's only utilizing the alloc/free interface
and not any of the other mechanisms provided by irqdomains. The latter
should make the overall code simpler but that obviously needs some
thought.
Thanks,
tglx
|
