On Tue, Jan 05, 2021 at 04:39:38PM +0000, Michael Kelley wrote:

From: Wei Liu <wei.liu@kernel.org> Sent: Tuesday, January 5, 2021 5:04 AM

quoted

On Mon, Dec 21, 2020 at 10:55:41PM -0800, Dexuan Cui wrote:

quoted

Currently the kexec kernel can panic or hang due to 2 causes:

1) hv_cpu_die() is not called upon kexec, so the hypervisor corrupts the
old VP Assist Pages when the kexec kernel runs. The same issue is fixed
for hibernation in commit 421f090c819d ("x86/hyperv: Suspend/resume the
VP assist page for hibernation"). Now fix it for kexec.

2) hyperv_cleanup() is called too early. In the kexec path, the other CPUs
are stopped in hv_machine_shutdown() -> native_machine_shutdown(), so
between hv_kexec_handler() and native_machine_shutdown(), the other CPUs
can still try to access the hypercall page and cause panic. The workaround
"hv_hypercall_pg = NULL;" in hyperv_cleanup() is unreliabe. Move
hyperv_cleanup() to a better place.

Signed-off-by: Dexuan Cui <decui@microsoft.com>

The code looks a bit intrusive. On the other hand, this does sound like
something needs backporting for older stable kernels.

On a more practical note, I need to decide whether to take it via
hyperv-fixes or hyperv-next. What do you think?

I'd like to see this in hyperv-fixes and backported to older stable kernels.
In its current form, the kexec path in a Hyper-V guest has multiple problems
that make it unreliable, so the downside risk of taking these fixes is minimal
while the upside benefit is considerable.

Applied to hyperv-fixes.

Wei.