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

Re: [Xen-devel] Xen 4.5 development update (September update). Feature freeze slip by two weeks.

Hi Konrad,

On 25 Sep 2014, at 22:26, Konrad Rzeszutek Wilk <konrad.wilk@xxxxxxxxxx> wrote:

> On Thu, Sep 25, 2014 at 10:25:42AM +0100, George Dunlap wrote:
>> On 09/25/2014 10:22 AM, Dave Scott wrote:
>>> Hi Konrad,
>>> On 10 Sep 2014, at 18:05, <konrad.wilk@xxxxxxxxxx> <konrad.wilk@xxxxxxxxxx> 
>>> wrote:
>>>> *  extend the xenstore ring with a 'closing' signal (fair)
>>>>  RFC patch posted
>>>> -  David Scott
>>> I’d like to propose an exception for this one. The title is perhaps 
>>> underselling it: although it is a protocol extension (and hence a new 
>>> feature) the main motivation is to fix a bug in hvmloader which causes HVM 
>>> guests to hang during booting on a busy machine. The bug is fairly easy to 
>>> reproduce with ~200 VMs: you’ll probably find at least one failed to boot.
>>> The patch set has got to a v4 but needs a little bit of work to clarify the 
>>> relationship between the closing signal and the existing RESET_WATCHES 
>>> xenstore protocol request.
> It looked (from a brief look) as it has OCaml and I have no experience
> with that. Is there somebody who can review it?
> I hadn't dug in it to give it yet an opinion (sorry).

I’ve managed to gather an Acked-by from IanJ (NB without reviewing the OCaml 


On 26 Sep 2014, at 15:00, Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx> wrote:

> Acked-by: Ian Jackson <ian.jackson@xxxxxxxxxxxxx>
> I'm happy to take this from you without really reviewing the ocaml
> code.

and I’ve got a positive review from Jon Ludlam (an OCaml expert)[added to cc:] 
just now on the OCaml parts:


On 7 Oct 2014, at 12:26, Jon Ludlam <jonathan.ludlam@xxxxxxxxxxxxx> wrote:

> Handling the exception higher up looks a fair bit nicer, and my previous
> comment has been addressed, so it looks fine to me.
> Reviewed-by: Jon Ludlam <jonathan.ludlam@xxxxxxxxxx>

FWIW I think this is still worth taking for 4.5 because

1. it fixes a real bug (HVM VMs occasionally failing to boot because of bad 
hvmloader/xenstored interaction)

2. the oxenstored changes are quite unobtrusive: the flag is sampled at the 
bottom, an exception bubbles all the way (almost) to the top and is handled 
there. If something were to go wrong, that exception would be caught at the 
main loop allowing the process to continue.

What do you think?


Xen-devel mailing list



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