Re: [PATCH v4] drm/vmwgfx: Treat zero SVGA_REG_CURSOR_MAX_DIMENSION as unset

[fausten]

From: fausten <yunfeng.li.nb@xxxxxxxxx>

On Tue, Sep 2, 2026 at 8:44 PM, Maaz Mombasawala
<maaz.mombasawala@xxxxxxxxxxxx> wrote:
> This patch bricked my testing VM, investigating..

Hi Maaz, hi Zack,

Just a gentle ping on this one. Maaz, did the investigation into your
bricked testing VM turn up anything in the meantime?

Meanwhile I've done more testing on my side. I applied this fix on
top of mainline 7.3 and the cursor works correctly again:

  Host:   VMware Fusion 26.0.1 on macOS 26.7.1 (x86_64)
  Guest:  Kylin Linux Advanced Server V10 (Halberd)
  Kernel: 7.3 + this patch

Before the patch, every cursor update was rejected with -EINVAL
because SVGA_REG_CURSOR_MAX_DIMENSION reads back 0 on this setup;
with the patch applied, cursor updates go through normally again.
3D acceleration is disabled in the VM settings.

So far I haven't been able to reproduce the failure you reported.
If you can share your test environment (hypervisor product and
version, host OS, guest kernel, 3D acceleration on/off) and the
failure symptoms (boot hang vs. cursor path only), I'd like to
narrow down what's different between our setups.

And if the conclusion is that treating 0 as "unset" with the current
fallback isn't the right fix, I'm happy to rework it as a v5.

Thanks,
fausten