User Details
- User Since
- Jan 20 2016, 6:04 AM (126 w, 1 d)
Tue, Jun 12
Fri, May 25
Mar 16 2018
Mar 15 2018
Mar 14 2018
Mar 1 2018
Feb 24 2018
We'll handle the BE part.
Feb 23 2018
Looks OK, we'll test it on our hardware.
Feb 22 2018
I think we talked out all issues here.
I'll split this to separate patches and commit to the HEAD. Please let me know if you have any objections.
Feb 21 2018
Feb 20 2018
Feb 16 2018
Actually, this is the only way to release resources by idling CPUs within the same SMT group - Linux does the same.
Sleep mode used here has the shortest wakeup delay, so the only performance hit is from saving and restoring context. However, it's still less complex than the one in cpu_switch.
Feb 14 2018
Feb 8 2018
Jan 31 2018
Jan 29 2018
It SHOULD work on PowerNV as the code for cpuid swap is there, however I did not run much of testing, so I'd rather consider this as experimental/not-working. The CHRP platform has no such code, so there the BSP must be at CPU0.
Jan 26 2018
Jan 25 2018
Any objections for commiting this?
Sounds reasonable. I've removed all references to PIR and left the previous PCPU-assignment mechanism intact.
Jan 24 2018
Jan 23 2018
Jan 22 2018
Jan 18 2018
Jan 17 2018
Yes, the recent toolchain works well.
Jan 16 2018
It's already outdated
Merged as r302292
Jan 15 2018
Jan 12 2018
Please let me know if there are any objections to this patch. I'd like to have it commited on Monday.
Jan 11 2018
Jan 9 2018
Just tested on arm64 (LE), works fine.
Nov 16 2017
Closing this one as I'm going to change this function much more.
Nov 15 2017
Yes, sysctl shows all doorbells just fine
Nov 14 2017
Nov 2 2017
After some testing of all this CPU startup thing, I think it would be best to panic if any of AP fails to start.
The rest of the system is not prepared to handle the scenario when the cpumask has "holes" inside. I then can see hangs in various places: for example the smp_rendezvouz, returning from kdb_enter, but there still might be more such things which I couldn't find.