Re: of balloon drivers and such

["Cherry G. Mathew" <cherry.g.mathew%gmail.com@localhost>]

On 5 May 2010 19:31, Jean-Yves Migeon <jeanyves.migeon%free.fr@localhost> wrote:
>
> On Sat, 1 May 2010 15:46:49 +0530, "Cherry G. Mathew"
> <cherry.g.mathew%gmail.com@localhost> wrote:
>> Hello,
>>
>> ftp://ftp.netbsd.org/pub/NetBSD/misc/cherry/balloon/
>>
>> I spent a bit more time messing around with this. If people could test
>> it on real systems that'd be a real help.
>>
>> I have no idea how it performs under real life work loads.
>
> Very quick overview (PAE keeps me busy, did not test it, sorry):
> - x86_btop and XPMAP_OFFSET: use xpmap_ptom()/mtop() directly instead, for
> portability (most Xen drivers (balloon is one), will eventually move to
> dev/xen).
> - protect xpmap_phys_to_machine_mapping access with splvm(), preferably in
> "sane" places (before entering a loop, for example), as flaglock mtx is
> IPL_NONE and kthread is PRI_NONE.
>

Hi,

Thanks I've updated the code similar to arch/xen/x86/xen_bus_dma.c.

There's a panic in supervisor mode that hits the domU when any
COMPAT_LINUX app is run. Not sure if this is balloon related. Will
investigate further and report here.

Cheers,
-- 
~Cherry