Re: u3g for XEN3_DOM0 ?

[Frank Kardel <kardel%netbsd.org@localhost>]

As of now there still seem issues with XEN3_DOM0 and MSIX.
When hitting that issue NO_PCI_MSI_MSIX was introduced.
PR kern/55667 (regression: XEN3_DOM0 fails to boot on) was created.
Later on on a different machine the MSIX issues seem to come up again
see PR kern/56291 (XEN3_DOM0 nvme abysmal performance - very slow boot)

So far I have not seen updated to those PRs so I did not retest after 
9.99.84.

The wm network interface issue may also be linked the the MSIX issue
suspected in the above PRs.

Frank

On 09/11/21 02:49, Emmanuel Dreyfus wrote:
> On Mon, Sep 06, 2021 at 11:53:08AM -0400, Greg Troxel wrote:
> > I didn't mean that.  If some driver doesn't work on xen, that's not a
> > big deal.  I just meant that when I was messing with the config that I
> > felt like i needed to actually do a build, and then install the kernel
> > on a xen box and actually boot it and make sure it still could run xen,
> > so that people would be no worse off.
>
> I tried to move XEN3_DOM0 as close as possible to GENERIC. I left the
> following differences
> - maxusers 32 vs 64
> - xen specific options and device attachments
> - things that do not build: acpicpu, vga and framebuffer, sysbeep
> - AGP, which is known to crash
> - VirtIO ent Hyper-V, which I assume are not relevant.
>
> The kernel boot, but my wm network interface gets deaf. I had to
> asdd options NO_PCI_MSI_MSIX to fix that, but perhaps it is not
> related to my change, I need to test a vanilla -current XEN3_DOM0
> to check.
>
> Attached is the patch to XEN3_DOM0. Once applied, you can
> diff XEN3_DOM0 GENERIC to check what differences are left.