questions about kvm-unit-tests results

questions about kvm-unit-tests results

[Conny Seidel]

[-- Attachment #1: Type: text/plain, Size: 966 bytes --]

Hi,

I have some questions about the kvm-unit-tests results.

Is there any advice how the results of tsc and kvmclock_test are to
interpret?

They both only print out some numbers and not, if the test has passed or
failed.

Does it make sense, that I prepare some patches, to unify the output of
all kvm unit tests? 

regards,
Conny

##
##################################################################
# Email : conny.seidel@amd.com            GnuPG-Key : 0xA6AB055D #
# Fingerprint: 17C4 5DB2 7C4C C1C7 1452 8148 F139 7C09 A6AB 055D #
##################################################################
# Advanced Micro Devices GmbH Einsteinring 24 85609 Dornach      #
# General Managers: Alberto Bozzo, Andrew Bowd                   #
# Registration: Dornach, Landkr. Muenchen; Registerger. Muenchen #
#               HRB Nr. 43632                                    #
##################################################################

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: questions about kvm-unit-tests results

[Jason Wang]

----- "Conny Seidel" <conny.seidel@amd.com> wrote:

> Hi,
> 
> I have some questions about the kvm-unit-tests results.
> 
> Is there any advice how the results of tsc and kvmclock_test are to
> interpret?

For tsc, it first shows the latency of rdtsc, and then the result of
rdtsc after giving two different values to the TSC. Currently, TSC
test always succeed.

For kvmclock, the tests contains three parts:
First part is a test of wallclock which check whether the wallclock
get from host match the value get from kvmclock. The test would show
the value get both from guest and host and their offset. The test
would fail if the offset were too large.

Second part is a monotonic cycle growth test, it first tries to detect
whether the cycle provided by the hyperviosr growth monotonically and
shows the number of loops, warps and stalls. This detection just serve
as a refrence and does not impact the result of test. Then it would
try to let the kvmclock driver in guest to produce a monotonical
growth cycle and check the result. The test would fail if warps were
detected and the result were shown.

The last part is just a performance test which measure the performance
degredation when trying to correct the cycle supplied by
hypervisor. It first measure the performance by just returning the
cycle supplied by hypervisor and show the TSC cycle spent on this. And
then measure the performance when driver tries to correct and produce
a correct cylce. These two tests do not have impact on the result of 
test and are always considered success.

I must admit the result of kvmclock_test may be not user-friendly. And
seems Amos have post a patch to give a summary of kvmclock_test, does
that satisfied your requirement?

> 
> They both only print out some numbers and not, if the test has passed
> or
> failed.

As a common rule, a zero value returned by qemu-kvm indicates a
success of test otherwise a failure.

> 
> Does it make sense, that I prepare some patches, to unify the output
> of
> all kvm unit tests? 
> 
> regards,
> Conny
> 
> ##
> ##################################################################
> # Email : conny.seidel@amd.com            GnuPG-Key : 0xA6AB055D #
> # Fingerprint: 17C4 5DB2 7C4C C1C7 1452 8148 F139 7C09 A6AB 055D #
> ##################################################################
> # Advanced Micro Devices GmbH Einsteinring 24 85609 Dornach      #
> # General Managers: Alberto Bozzo, Andrew Bowd                   #
> # Registration: Dornach, Landkr. Muenchen; Registerger. Muenchen #
> #               HRB Nr. 43632                                    #
> ##################################################################