Impact on KPI between smack and Non-smack build

[shreeharsha_ind]

Hi Team,

We are seeing a Increase in KPI in SMACK enabled build when compared to non-SMACK build. After doing  further trails we understood that system call execution in smack build is consuming more time compared to non-SMACK build. 

Below is the trial data from a sample test app which calculates the time taken for 1000 calls of read and write of data (1024) operation to a file.

Smack build:
root@tas600-eda-t255:~# ./test
Time taken: 449.494 milliseconds
root@tas600-eda-t255:~# ./test
Time taken: 468.849 milliseconds
root@tas600-eda-t255:~# ./test
Time taken: 452.376 milliseconds
root@tas600-eda-t255:~# ./test
Time taken: 446.216 milliseconds
root@tas600-eda-t255:~# ./test
Time taken: 448.251 milliseconds

Average ~ 453 milliseconds

Non-smack build:
root@tas600-eda-t255:~# ./test
Time taken: 430.927 milliseconds
root@tas600-eda-t255:~# ./test
Time taken: 435.279 milliseconds
root@tas600-eda-t255:~# ./test
Time taken: 431.131 milliseconds
root@tas600-eda-t255:~# ./test
Time taken: 441.297 milliseconds
root@tas600-eda-t255:~# ./test
Time taken: 447.747 milliseconds

Average ~ 437 milliseconds

Is there any mechanism to enable caching feature for SMACK and thus reduce the overhead for systemcall? 

[Bio_TICFSL]

Hello,

Are you working with the AGL demo for i.MX8MP? You can find the readme here:

https://docs.automotivelinux.org/en/koi/#0_Getting_Started/2_Building_AGL_Image/3_Initializing_Your_...

The AGL demo is provided “as is”, as a proof of concept, and it was tested only on the i.MX6Q so no support is provided for this demo and it hasn’t been tested on the i.MX8.

To enable kernel extensions and other options you may use menuconfig with the following command, although we do not have documentation on enabling SMACK step by step.

$ bitbake linux-imx -c menuconfig

My apologies for the inconvenience.

Regards