Comment 11 for bug 1853506

Revision history for this message
Rafael David Tinoco (rafaeldtinoco) wrote :

Yes Seth!

Unfortunately (or not, as picking up things like this in this phase is very good) there is an on-going issue with nvdimm regions boundaries. I'm documenting everything in another bug:

https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1855177

And the issue is well known upstream (and capable of happening to some arches/firmwares). Since there isn't a clear "spec" on how HW manufactores can force the region alignments, mitigating the issue, I'm keeping documentation in that bug not to block this MIR, as the upstream ndctl also faces the same issue and end-user can use some directions by reading that bug.

In our "test-scenario", for the qa-regressions tests, the mitigation will be to reduce from 2 emulated nvdimms to just 1, and report the alignment issue to QEMU. I'm using some of the most important tests from ndctl/test source code.

Will post the repo pushes soon.

Cheers