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

Re: [Xen-devel] backporting considerations (Re: [PATCH v9 0/9] xen/x86: various XPTI speedups)

On Wed, May 16, 2018 at 5:01 PM, Jan Beulich <JBeulich@xxxxxxxx> wrote:
>>>> On 16.05.18 at 16:53, <dunlapg@xxxxxxxxx> wrote:
>> On Wed, May 16, 2018 at 3:01 PM, Jan Beulich <JBeulich@xxxxxxxx> wrote:
>>>>>> On 16.05.18 at 15:18, <dunlapg@xxxxxxxxx> wrote:
>>>> If the latter, I think the same argument applies: turning on XPTI is a
>>>> requirement for many people, and thus represents a pretty hefty
>>>> performance regression.  While we don't need to backport normal fixes
>>>> to security-only releases, we should certainly try to avoid
>>>> regressions.
>>> I don't think we would have addressed non-security fallout (or other
>>> than really severe regressions) from other security patches in the
>>> past on security only branches. People caring about performance
>>> should upgrade.
>> If a security patch, when backported to 4.6, broke some fairly
>> critical bit of functionality (say,  openvswitch support), you would
>> oppose a subsequent patch which would fix that regression?
>> That doesn't seem very reasonable to me.  Users shouldn't have to
>> choose between being vulnerable to a security issue and losing
>> functionality which was working at the last release.  Otherwise,
>> what's the point of having "security supported" releases?
> Note how I did say "or other than really severe regressions". I think
> your "fairly critical bit of functionality" falls into exactly that area.

Right, so we agree on the basic principles, but disagree about whether
XPTI's performance hit counts as a "really severe regression".  At
least one of my CentOS users was seriously considering applying
juergen's XPTI improvement patch before it even hit staging, because,
"We're struggling with slowdowns from xpti within the dom0."  (I
advised her not to at that point, because XenRT had flagged up some
potential issues.)  If a user describes herself as "struggling", I
think that counts as a regression.  This was for Xen 4.8, but I think
the basic principle applies.  I'll definitely be backporting those to
the 4.8 CentOS packages now that they're in staging; and I'll probably
try backporting them to the 4.6 packages too (since they're currently
the default, although hopefully not for long).

Like I said, I don't expect you personally to do something you don't
think is worth your time.  But I think it would in general be good if
the XenProject fixed this performance regression for releases which
are still under security support.


Xen-devel mailing list



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