On Fri, Feb 16, 2024 at 9:03 PM Thomas Gleixner <tglx@xxxxxxxxxxxxx> wrote:
>
> On Sat, Jan 27 2024 at 21:47, Anup Patel wrote:
> > + priv->irqdomain = irq_domain_create_linear(dev->fwnode, nr_irqs + 1,
> > + &plic_irqdomain_ops, priv);
> > + if (WARN_ON(!priv->irqdomain))
> > + return -ENOMEM;
>
> While some of the stuff is cleaned up by devm, the error handling in
> this code looks pretty fragile as it leaves initialized contexts,
> hardware state, chained handlers etc. around.
Sure, let me try to improve the error handling.
>
> The question is whether the system can actually boot or work at all if
> any of this fails.
On platforms with PLIC, the PLIC only manages wired interrupts
whereas IPIs are provided through SBI (firmware interface) so a
system can actually continue and boot further without PLIC.
In fact, we do have a synthetic platform (namely QEMU spike)
where there is no PLIC instance and Linux boots using SBI based
polling console.
>
> > +
> > /*
> > * We can have multiple PLIC instances so setup cpuhp state
> > - * and register syscore operations only when context handler
> > - * for current/boot CPU is present.
> > + * and register syscore operations only after context handlers
> > + * of all online CPUs are initialized.
> > */
> > - handler = this_cpu_ptr(&plic_handlers);
> > - if (handler->present && !plic_cpuhp_setup_done) {
> > + cpuhp_setup = true;
> > + for_each_online_cpu(cpu) {
> > + handler = per_cpu_ptr(&plic_handlers, cpu);
> > + if (!handler->present) {
> > + cpuhp_setup = false;
> > + break;
> > + }
> > + }
> > + if (cpuhp_setup) {
> > cpuhp_setup_state(CPUHP_AP_IRQ_SIFIVE_PLIC_STARTING,
> > "irqchip/sifive/plic:starting",
> > plic_starting_cpu, plic_dying_cpu);
> > register_syscore_ops(&plic_irq_syscore_ops);
> > - plic_cpuhp_setup_done = true;
>
> I don't think that removing the setup protection is correct.
>
> Assume you have maxcpus=N on the kernel command line, then the above
> for_each_online_cpu() loop would result in cpuhp_setup == true when the
> instances for the not onlined CPUs are set up, no?
A platform can have multiple PLIC instances where each PLIC
instance targets a subset of HARTs (or CPUs).
Previously (before this patch), we were probing PLIC very early so on
a platform with multiple PLIC instances, we need to ensure that cpuhp
setup is done only after PLIC context associated with boot CPU is
initialized hence the plic_cpuhp_setup_done check.
This patch converts PLIC driver into a platform driver so now PLIC
instances are probed after all available CPUs are brought-up. In this
case, the cpuhp setup must be done only after PLIC context of all
available CPUs are initialized otherwise some of the CPUs crash
in plic_starting_cpu() due to lack of PLIC context initialization.
Regards,
Anup
