-
Notifications
You must be signed in to change notification settings - Fork 762
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[csrng] Coverage of stalling behaviour at interfaces #15744
Comments
CSRNG pushback on EDN. EDN then pushback on CSRNG. Etc. Both would then keep hanging |
We need coverage of the handshake between EDN and CSRNG as well as CSRNG and entropy source. |
We discussed today that this has also high priority. @ctopal has been looking into that. It would be great to get some update tomorrow. :-) |
Overall findings: In the DV environment we have a transaction level model FIFO for EDN side of the communication named As we are not configuring this push_pull_agent to have a FIFO and we use
As I explained above, we don't have complicated behaviour in communication between EDN and CSRNG. Entropy source and CSRNG connection is even simpler considering it is a basic req/ack handshake rather than ready/valid one. The common So with all this looking around and such, I'm still struggling to understand what is needed to be done here. I checked the discussion for EDN hanging when CSRNG ready drops (in #15561). Looks like from CSRNG block level perspective a coverpoint for this specific deal might not be needed. I'd really appreciate if someone can help me identify what specifically would be needed for this one to be considered covered. |
Thanks for summarizing your findings, @ctopal. The functional coverage that comes with However, as those interfaces are connected to FIFOs and arbiters controlled by state machines within CSRNG, I think it's worth adding functional coverage on at least the FIFOs in In a second step, I would add a cross between the full signal of the FIFO and the |
I've now reviewed the covergroups added by @ctopal as well as the resulting coverage based on the last nightly regression. PR #16494 adds some ignore bins for cross points we're not expecting to hit at all (e.g. those covered by SVAs that we disable when doing FI testing) or not to hit for every possible command stage FIFO (where we have combined alerts for the FIFO errors). In addition, a new cover point is added for the read side of the command FIFOs to cover also the case when the FIFOs are not immediately read as proposed by @andreaskurth . From a quick coverage run it seems that all the points are being hit, except for some points that we expect to be hit by the disable/re-enable testing. I've added comments for clarification for these. So, once #16494 is being merged, we can close this issue. |
Coverage of stalling behaviour at interfaces? Probably should include this as we’ve seen EDN issues around this
original estimate 4
The text was updated successfully, but these errors were encountered: