APPLIED: [LP 1851709] Cannot launch KVM guests if EPT is disabled
t.lamprecht at proxmox.com
Wed Nov 13 06:35:36 UTC 2019
On 11/12/19 7:37 PM, Thadeu Lima de Souza Cascardo wrote:
> On Tue, Nov 12, 2019 at 02:00:45PM +0100, Thomas Lamprecht wrote:
>> On 11/12/19 11:32 AM, Stefan Bader wrote:
>>> On 12.11.19 10:16, Thadeu Lima de Souza Cascardo wrote:
>>>> It has been reported as failure to launch 64-bit linux VMs when kvm-intel
>>>> ept module parameter is off. It is off by default on old Intel CPUs.
>>>> The bionic revert has been tested as works fine. kvm-unit-tests has been
>>>> run there as well and no regressions found.
>>> Applied to disco+bionic/master-next. Thanks.
>> Why did you go the revert route and not the "backport the missing
>> pieces" one? For Disco cherry-picking upstream commit
>> bf03d4f9334728bf7c8ffc7de787df48abd6340e is enough (tested here). For
>> disco the backport for 4.14 should be a working (did not tested yet).
>> The patch which made problem show up was not a issue itself, and a valid
>> fix, older kernel just missed some supporting fix, i.e., . So, FWIWW,
>> IMO it would be better to backport that instead of reverting.
>> : https://firstname.lastname@example.org/T/
> Reverting was considered the safest option at the time we decided to fix
> this regression in time for releases today.
> In fact, the backport had not been sent by Sean yet at the time, and,
> though we have already noticed the message where it was explained what
> needed to be backported, we still considered it safer to revert.
> The revert of the revert followed by the fix backport will probably follow
> in a future release.
> Hope that explains the situation.
Yes it does, and sorry I was not aware of that timing difference then :)
FWIW, It's the same route we went with anyway (first revert, then backport).
More information about the kernel-team