On 12/18/19 11:01 AM, Niek Linnenbank wrote: > Hello Richard, > > On Tue, Dec 17, 2019 at 5:41 PM Richard Henderson > <[email protected] > <mailto:[email protected]>> wrote: > > On 12/17/19 6:12 AM, Peter Maydell wrote: > > Cc'ing Richard : this is one for you I think... (surely we > > need to rebuild the hflags from scratch when we power up > > a CPU anyway?) > > We do compute hflags from scratch in reset. > > It has also turned out that there were a few board models that poked at > the > contents of the cpu and needed special help. Some of that I would imagine > would be fixed properly with the multi-phase reset patches, where we could > rebuild hflags when *leaving* reset. > > In arm_set_cpu_on_async_work, we start by resetting the cpu and then start > poking at the contents of some system registers. So, yes, we do need to > rebuild after doing that. Also, I'm not sure how this function should > fit into > the multi-phase reset future. > > > Great, thanks a lot for confirming and clarifying this! > You mention the multi-phase reset feature, is that going to replace the > arm_set_cpu_on() functionality?
I don't think so, but I'm not sure. As I said above, I don't immediately see how arm_set_cpu_on() will integrate. In any case, multi-phase reset is still pending, though I believe it is high on Peter's priority queue for the 5.0 development cycle. r~
