Thread (1 message) 1 message, 1 author, 2016-09-09

Re: [PATCH v4 3/4] arm64: arch_timer: Work around QorIQ Erratum A-008585

From: Marc Zyngier <hidden>
Date: 2016-09-09 07:06:44
Also in: linux-arm-kernel

On Fri, 9 Sep 2016 13:20:46 +0800
Ding Tianhong [off-list ref] wrote:

Hi Ding,
quoted
I'm still worried that this series doesn't address Xen or KVM guests
that need to be made aware of the broken timers.
  
Hi Marc:

I think "ARCH_TIMER_REG_READ("cntv_tval_el0", arch_timer_get_vtval)"
could cover the tval changes in KVM guest, and I fould the guest
still use the get_cycles() to get the timer cycles, if I miss
something, please remind me, thanks.
You're missing the point that to tell the guest that the HW has this
erratum, we need the DT to be populated with the right property.
quoted
At the very least, I'd like a kernel command line option that'd let
the user reliably run its VMs. You can do something along the lines
of 46fd5c6b, and have a command line argument like
"clocksource.arm_arch_timer.fsl-a008585=1", which would enable the
workaround.
  
I don't think adding in command line is a good solution for this bug,
we should not told the OS user how to fix it by adding command line
option, and need to fix it by dts or ACPI.
Sure. I'm happy to review the patches that will:
- Add a new KVM kernel API describing properties for the timer so that
  we can expose the erratum from the kernel to userspace
- Add a similar API for Xen
- Have all the userspace (QEMU, kvmtool, the Xen tools) to be converted
  to use this new API so that they can emit the correct DT
- Point to a change in the ACPI spec so that we can encode errata there
- Add support for handling this erratum using ACPI in the kernel
- Allow QEMU and the Xen tools to expose this erratum in the guests
  ACPI tables

Unless you're in a position to write all this code (and
specifications), and post the patches *now*, I'll stand by my
recommendation to have a command-line option. Not exposing this issue
to the guest results in unreliable operations. You may not care, but I
do.

Thanks,

	M.
-- 
Without deviation from the norm, progress is not possible.
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help