Re: [Xen-devel] [PATCH v3] xen/balloon: don't online new memory initially

[Juergen Gross]

On 03/10/17 23:33, Boris Ostrovsky wrote:
> On 10/02/2017 05:37 PM, HW42 wrote:
>> Juergen Gross:
>>> When setting up the Xenstore watch for the memory target size the new
>>> watch will fire at once. Don't try to reach the configured target size
>>> by onlining new memory in this case, as the current memory size will
>>> be smaller in almost all cases due to e.g. BIOS reserved pages.
>>>
>>> Onlining new memory will lead to more problems e.g. undesired conflicts
>>> with NVMe devices meant to be operated as block devices.
>>>
>>> Instead remember the difference between target size and current size
>>> when the watch fires for the first time and apply it to any further
>>> size changes, too.
>>>
>>> In order to avoid races between balloon.c and xen-balloon.c init calls
>>> do the xen-balloon.c initialization from balloon.c.
>>>
>>> Signed-off-by: Juergen Gross <jgross@xxxxxxxx>
>> This patch seems to introduce a regression. If I boot a HVM or PVH
>> domain with memory != maxmem then the kernel inside the domain reports
>> that it has maxmem available even though Xen reports only what is set as
>> memory. Sooner or later Xen logs "out of PoD memory!" and kills the
>> domain. If I revert the corresponding commit (96edd61d) then everything
>> works as expected.
>>
>> Tested this with Xen 4.9.0 and Linux 4.13.4.
>>
> 
> 
> Yes, this indeed doesn't look like it's doing the right thing (although
> I haven't seen the "out of memory" error).

You need to use enough memory (e.g. via memhog).

> I wonder whether target_diff should be computed against xenstore's
> "static-max" and not "target" and the memory should be ballooned down
> immediately and not on a subsequent watch firing.

Right. And we need to keep target_diff = 0 for PV domains.

Patch coming soon.


Juergen