Re: [PATCH] tpm: increase timeout for kselftests

[Jarkko Sakkinen]

On Fri, May 20, 2022 at 03:28:42PM +0200, Stefan Mahnke-Hartmann wrote:
> 
> 
> On 11.05.22 17:14, Jarkko Sakkinen wrote:
> > On Tue, May 10, 2022 at 01:16:08PM +0200, Johannes Holland wrote:
> >> Due to CreatePrimary commands which need to create RSA keys of
> >> increasing size, the timeout value need to be raised, as well.
> >> Default is 300s.
> >>
> >> Signed-off-by: Johannes Holland <johannes.holland@xxxxxxxxxxxx>
> >> ---
> >> A timeout of anything below 600s still lead to occasional timeouts for
> >> me. Therefore, I propose 600s as a new value. 
> >>
> >>  tools/testing/selftests/tpm2/settings | 2 ++
> >>  1 file changed, 2 insertions(+)
> >>  create mode 100644 tools/testing/selftests/tpm2/settings
> >>
> >> diff --git a/tools/testing/selftests/tpm2/settings b/tools/testing/selftests/tpm2/settings
> >> new file mode 100644
> >> index 000000000000..919bc3803f03
> >> --- /dev/null
> >> +++ b/tools/testing/selftests/tpm2/settings
> >> @@ -0,0 +1,2 @@
> >> +timeout=600
> >> +
> >> -- 
> >> 2.34.1
> >>
> > Could cope but I did not get why it needs to be raised.
> 
> The TPM2 SpaceTest testsuite currently creates 8 primary RSA 2k keys + 1 for
> setup. Generating a RSA2k key can take up to ~1-2 minutes on some of our TPMs.
> => 2x9 = 18 minutes. In the kernel we even define a duration timeout for rsa2k
> keygen of 5 min per key! (TPM2_DURATION_LONG_LONG = 300000) => up to 45 minutes
> would be "acceptable".
> 
> However since the average key generation time is much faster, a value of ~10
> minutes should be fine enough.
> 
> The reason why you did not experience this yet on your test system is either
> because PTT TPMs are faster in that respect and/or some TPMs implement pre-gen
> mechanisms for key generation, (or just plain luck)
> 
> BR, Stefan
> 
> > BR, Jarkko

I think this is ok.


Acked-by: Jarkko Sakkinen <jarkko@xxxxxxxxxx>

BR, Jarkko