Re: [PATCH] powerpc: Allow scheduling of IRQ with PREEMPT_RT

[Shrikanth Hegde <sshegde@linux.ibm.com>]

On 1/7/25 17:26, Sebastian Andrzej Siewior wrote:
> On 2024-12-26 11:12:38 [-0800], Robert Joslyn wrote:
>> This branch is needed to allow preemptible IRQs to be scheduled,
>> otherwise the kernel hangs during boot on my e300 based platforms.
>> This branch is needed for both CONFIG_PREEMPT and CONFIG_PREEMPT_RT,

Hi Robert.

Are you running with preempt=full or enabled preempt_rt ?
rt is yet to be supported on powerpc. are you having it enabled out of tree?

Also, I have sent out patch to enable dynamic preemption. It uses preempt_model_preemptible.
I think it should work for you as well. If you could give that a try and ensure it works, it would be great.
https://lore.kernel.org/all/20250106051919.55020-2-sshegde@linux.ibm.com/

>> so broaden the check to CONFIG_PREEMPTION, which is selected by both.
>>
>> This was previously set to CONFIG_PREEMPTION, but was reverted as part
>> of removing LAZY_PREEMPT support.
> 
> I remember I pointed this out while Shrikanth posted the preempt-lazy
> patches but didn't realize that I need this in RT tree, too.
> 
> The lazy-preempt patches for powerpc were merged upstream, I am going to
> pull them into the v6.6-series.
> I am going to Cc you, hopefully you can confirm that it is working.
> 
>> Fixes: b33022a48c79
>> Signed-off-by: Robert Joslyn <robert_joslyn@selinc.com>
>> ---
>>   arch/powerpc/kernel/interrupt.c | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/arch/powerpc/kernel/interrupt.c b/arch/powerpc/kernel/interrupt.c
>> index af62ec974b97..970039af8e92 100644
>> --- a/arch/powerpc/kernel/interrupt.c
>> +++ b/arch/powerpc/kernel/interrupt.c
>> @@ -396,7 +396,7 @@ notrace unsigned long interrupt_exit_kernel_prepare(struct pt_regs *regs)
>>   		/* Returning to a kernel context with local irqs enabled. */
>>   		WARN_ON_ONCE(!(regs->msr & MSR_EE));
>>   again:
>> -		if (IS_ENABLED(CONFIG_PREEMPT)) {
>> +		if (IS_ENABLED(CONFIG_PREEMPTION)) {
>>   			/* Return to preemptible kernel context */
>>   			if (unlikely(read_thread_flags() & _TIF_NEED_RESCHED)) {
>>   				if (preempt_count() == 0)
> 
> Sebastian