[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH v4 4/6] microcode: don't call apply_microcode() in cpu_request_microcode()

>>> On 30.11.18 at 09:57, <chao.gao@xxxxxxxxx> wrote:
> On Thu, Nov 29, 2018 at 10:46:05AM +0100, Roger Pau Monné wrote:
>>On Thu, Nov 29, 2018 at 12:28:46PM +0800, Chao Gao wrote:
>>> It is better that all CPUs have the same microcode revision. 
>>> Linux kernel rejects late microcode update if finding some CPUs
>>> offlined. I may port this patch to Xen too in a separate patch.
>>What happens with hotplug CPUs?
>>Even if you disable late loading when there are offlined CPUs you
>>still need to handle hotplug CPUs, which IMO should share the same
>>path with offlined CPUs: the microcode update should be done ASAP
>>after bringing the CPU up.
> In linux, CPU's being offline is just a logical offline. It may participate
> actions like MCE. It would lead to a situation that some cpus are using old
> microcode while some are using the new one, which introduces instability.
> CPU hotplug doesn't have such issue.

But please be careful not to disallow ucode loading in cases where
doing so is fine. I'm in particular thinking about "smt=no", where
secondary hyperthreads get parked. Since ucode updates on one
hyperthread of a core affect all hyperthreads, updating is fine in
this particular case.

Furthermore, if you look at my patch series "x86: more power-
efficient CPU parking", it is also within realm of possibility to
briefly wake up parked CPUs, just to get their ucode updated.


Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.