Dne 27.11.2012 08:07, Mike napsal(a):
在 2012-11-27二的 04:58 -0200,Lucas Meneghel Rodrigues写道:
On Tue, Nov 27, 2012 at 3:32 AM, Mike <[email protected]> wrote:
Hi all

Now I try to test the KSM case in my machine.
the path: virt/kvm/tests/ksm_overcommit.py

It has two cases , one is in serial mode other one is parallel mode

for parallel mode it just create one VM and test KSM.
this shouldn't be a fault????
 From the source code:

409    max_vms = int(params.get("max_vms", 2))
...
563    # Creating other guest systems
564    for i in range(1, vmsc):

So I'm not sure why you thought the test only used a single vm.
if (params['ksm_mode'] == "parallel"):
439         vmsc = 1
440         overcommit = 1
441         mem = host_mem
442         # 32bit system adjustment

Yes you right, but it has changed in the next code, if
params['ksm_mode'] == "parallel" the vmsc will change to 1.
Hi Mike,

ksm test was written before we start really documenting. I'll try to take a look at it this week and modify/add comments.

I haven't seen this test for a while, but from what I remember the "parallel" mode is there because we needed to test KSM and KVM. KSM requires a program to set the memory mergable. KVM does that. We also wanted to test that VM will actually survive this. The basic idea ways to start VM, run a simple python script inside the guest and fill it's memory with constant pages. This results in mergable pages on host (lots of pages with the same value). When we then change the value of each page, the pages should split (on host). Guest shouldn't know about this and shouldn't prevent it.

Please let me know if you have any other questions, I'll try to explain it in docstrings this or the following week. I need to take a deeper look to avoid some miss-thoughts.
Regards,
Lukáš


KSM should work for more than one VM,
Yes, and that's precisely what the test does. You might want to get
back to the source code. One of the mistakes we made on this one was
to not put a docstring explaining the steps taken by the test. Copying
Lukas.

Lukas, could you send us a patch with a docstring description of the test?

Cheers,



_______________________________________________
Autotest-kernel mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/autotest-kernel

Reply via email to