On Fri, 2024-12-13 at 18:05 +0100, Thomas Gleixner wrote: > > > Agreed. The hacky proof of concept I posted earlier invoking > > machine_kexec() instead of suspend_ops->enter() works fine. I'll look > > at cleaning it up and making it not invoke all the ACPI hooks for > > *actual* suspend to RAM, etc. > > Something like the below? It survived an hour of loop testing. If I read that correctly, it's still invoking the standard platform (e.g. ACPI) hooks for suspend-to-RAM, when it probably shouldn't? I suspect it wants its *own* set of platform_suspend_ops, which are mostly empty apart from the ->enter() ? I started looking at that, but now my eyes are currently bleeding after seeing the existing platform_suspend_ops vs. platform_s2idle_ops structures, which are kind of similar but not the same. And the set of helper functions which invoke one or the other, from the barely tolerable platform_resume_end()... static void platform_resume_end(suspend_state_t state) { if (state == PM_SUSPEND_TO_IDLE && s2idle_ops && s2idle_ops->end) s2idle_ops->end(); else if (suspend_ops && suspend_ops->end) suspend_ops->end(); } ... to the extra-special platform_resume_noirq() which is similar except that it needs three *different* names (_resume_noirq vs. restore_early vs. wake): static void platform_resume_noirq(suspend_state_t state) { if (state == PM_SUSPEND_TO_IDLE) { if (s2idle_ops && s2idle_ops->restore_early) s2idle_ops->restore_early(); } else if (suspend_ops->wake) { suspend_ops->wake(); } } I wonder if we end up wanting a *third* set there, for the kjump_ops? Except can we unify the structure definitions and then just *use* the appropriate one of the three, which is either passed down or selected using the 'state'?
Attachment:
smime.p7s
Description: S/MIME cryptographic signature