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

Re: [Xen-devel] [PATCH RFC 0/3] Move RTC interrupt injection back into the vpt code.



On 14/02/14 11:52, George Dunlap wrote:
> On Thu, Feb 13, 2014 at 5:39 PM, Andrew Cooper
> <andrew.cooper3@xxxxxxxxxx> wrote:
>> On 13/02/14 17:32, Tim Deegan wrote:
>>> Hi,
>>>
>>> This series implements the most recent idea I was proposing about
>>> reworking the RTC PF interrupt injection.
>>>
>>> Patch 1 switches handling the !PIE case to calculate the right answer
>>> for REG_C.PF on demand rather than running the timers.
>>> Patch 2 switches back to the old model of having the vpt code control
>>> the timer interrupt injection; this is the fix for the w2k3 hang.
>>> Patch 3 is just a minor cleanup, and not particularly necessary.
>>>
>>> N.B. In its current state it DOES NOT WORK.  I got distracted by
>>> other things today and didn't get a chance to finish working on it,
>>> but I wanted to send it out for feedback on the general approach.
>>> If it seems broadly acceptable then either I can pick it up again next
>>> week or maybe Andrew can look at fixing it.
>>>
>>> Cheers,
>>>
>>> Tim.
>>>
>> I should have time to look at the series tomorrow.
> The next question to ask is this:
>
> This is the last big disruptive bug / bugfix on my list.  We're
> planning on cutting an RC Monday probably, with a test day Tuesday.
> This bug was originally marked as "Not for 4.4".

It was originally for 4.4 when I proposed the very first fix, then not
for 4.4 when it was clear my first fix didn't work and I had to start
from scratch.  The efforts for fixing now is simply because of getting
some traction on the problem.

>
> So our options are:
> * Delay the release, waiting for this new series to be ready
> * Take the patch Andy posted last week for now, and backport Tim's fix
> when it's ready
> * Release without this bug being fixed
>
> As a reminder (for those who haven't been following the thread), the
> effect of this bug is that w2k3 guests sometimes hang during boot.
> I'm not sure exactly how often this is, but from talking to Andy it
> seems to be fairly low -- one percent maybe?  The code is very subtle
> and any change may risk causing similar hangs in other situations; in
> particular we would want to be able to test it pretty well.
>
> At the moment I'm leaning towards not delaying the release for it.
> That could either mean checking the patch we have to hand today (so it
> can make it into the RC Monday hopefully), or just going without it.
>
> Any thoughts?
>
>  -George

The bug has been present in Xen since the 4.3 dev cycle, meaning that
All Xen-4.3.x are susceptible.

I do not believe holding the release for a fix is sensible, especially
as the fix is still not considered ready yet

Also, I do not advise taking last weeks patch as an interim fix.  As
repeatedly proved, this is very tricky to get right, and it is less risk
to leave it as-is.


I will attempt to get the series working, and if it is well tested and
still within an acceptable timeframe for 4.4 then it should go in.  If
not, releasing with this bug is not the end of the world.

~Andrew

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel


 


Rackspace

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