Attention is currently required from: Bora Guvendik, Furquan Shaikh, Selma Bensaid, Tim Wawrzynczak, Subrata Banik, Angel Pons, Aaron Durbin. Julius Werner has posted comments on this change. ( https://review.coreboot.org/c/coreboot/+/51445 )
Change subject: timestamp: Add new helper functions ......................................................................
Patch Set 8:
(1 comment)
Commit Message:
https://review.coreboot.org/c/coreboot/+/51445/comment/d892396b_b8c88b80 PS3, Line 7: Add helper fucntions
I guess we have other options as well to create 2 cbmem tables
No, we definitely don't want two separate timestamp tables, it should all be withing one, internally consistent table. The "rewind" method definitely works for that, but I think my suggestion (just using negative timestamps) would also work and be simpler.
Additionally, unlike some other SoC platform, where #1 can be collected as early after CPU reset, is not likely the case with IA platform. Rather when we are able to collect #1 (typically, at early romstage state or late might be), we already have #2 available.
Right, but that shouldn't make a difference. You can add timestamps whenever you want. Note that timestamps do not need to appear in chronological order in the timestamp table... it's perfectly legal for the CSE timestamps to be added to the table after the bootblock timestamps, the cbmem utility will just sort them before printing.
I kind of get what your are suggesting but I would let Bora to share a sample timestamp for #1 and #2 above to illustrate the scenario. Want to make sure we have exact timestamp value appearing at base rather delta which may not be the exact while doing (Tn - T(n-1)) = negative number.
I don't really understand what you mean here, unfortunately. Note that timestamps in the table are always stored relative (i.e. as a delta) to the base timestamp. For the CSE timestamps, in the approach I'm suggesting, that delta would simply be negative.
-
Julius Werner (Code Review)