On Mon, Jul 07, 2025 at 06:49:02PM +0000, Michael Kelley wrote:

From: Nam Cao <redacted> Sent: Monday, July 7, 2025 1:20 AM

quoted

Move away from the legacy MSI domain setup, switch to use
msi_create_parent_irq_domain().

While doing the conversion, I noticed that hv_compose_msi_msg() is doing
more than it is supposed to (composing message). This function also
allocates and populates struct tran_int_desc, which should be done in
hv_pcie_domain_alloc() instead. It works, but it is not the correct design.
However, I have no hardware to test such change, therefore I leave a TODO
note.

Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Thomas Gleixner <redacted>
Signed-off-by: Nam Cao <redacted>

[Adding linux-hyperv@vger.kernel.org so that the Linux on Hyper-V folks
have visibility.]

This all looks good to me now. Thanks for the additional explanation of
the TODO. I understand what you are suggesting. Moving the interaction
with the Hyper-V host into hv_pcie_domain_alloc() has additional appeal
because it should eliminate the need for the ugly polling for a VMBus
response. However, I'm unlikely to be the person implementing the
TODO. hv_compose_msi_msg() is a real beast of a function, and I lack
access to hardware to fully test the move, particularly a device that
does multi MSI. I don't think such a device is available in a VM in the
Azure public cloud.

I've tested this patch in an Azure VM that has a MANA NIC. The MANA
driver has updates in linux-next to use MSIX dynamic allocation, and
that dynamic allocation appears to work correctly with this patch. My
testing included unbind and rebind the driver several times so that
the full round-trip is tested.

Reviewed-by: Michael Kelley <redacted>
Tested-by: Michael Kelley <redacted>

Acked-by: Wei Liu <wei.liu@kernel.org>