On 04/01/2023 16:45, Ulf Hansson wrote: > On Mon, 19 Dec 2022 at 16:15, Krzysztof Kozlowski > <krzysztof.kozlowski@xxxxxxxxxx> wrote: >> >> Realtime kernels with PREEMPT_RT must use raw_spinlock_t for domains >> which are invoked from CPU idle (thus from atomic section). Example is >> cpuidle PSCI domain driver which itself is PREEMPT_RT safe, but is being >> called as part of cpuidle. > > Just so I don't get this wrong, since the cpuidle-psci also calls > pm_runtime_* functions so it isn't PREEMPT_RT safe, at least not yet? You are correct. Patch 3 here addresses it by... just not doing runtime PM. This is a hacky workaround but: 1. I don't have any other idea, 2. It's not a big problem because RT systems are not supposed to have any CPU idle (one of first things during RT system tuning is to disable cpuidle). > >> >> Add a flag allowing a power domain provider to indicate it is RT safe. >> The flag is supposed to be used with existing GENPD_FLAG_IRQ_SAFE. >> >> Cc: Adrien Thierry <athierry@xxxxxxxxxx> >> Cc: Brian Masney <bmasney@xxxxxxxxxx> >> Cc: linux-rt-users@xxxxxxxxxxxxxxx >> Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@xxxxxxxxxx> > > For genpd, overall, I think this looks like an okay approach to me. > Although, let me check the whole series (I need some more time to do > that) before I give this my blessing. Sure, we are all have too many mails in inbox. :) Best regards, Krzysztof
