[Xenomai] bcm2835 ipipe question

[Xenomai] bcm2835 ipipe question

[Greg Gallagher]

I'm getting the raspberry pi zero images ready and I see the following
in the dmesg:

[    0.000000] WARNING: CPU: 0 PID: 0 at
/home/ggallagher/devel/emb_linux/ipipe-arm/kernel/irq/chip.c:57
irq_set_chip+0x7c/0xac
[    0.000000] irqchip ARMCTRL-level is not pipeline-safe!
[    0.000000] Modules linked in:
[    0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 4.14.4-ipipe+ #2
[    0.000000] Hardware name: BCM2835
[    0.000000] I-pipe domain: Linux

it looks like cobalt isn't running either, I'm trying to find the best
way to debug this or if this is a bug?

Any help is appreciated

Greg

Re: [Xenomai] bcm2835 ipipe question

[Philippe Gerum]

On 02/11/2018 02:57 AM, Greg Gallagher wrote:
> I'm getting the raspberry pi zero images ready and I see the following
> in the dmesg:
> 
> [    0.000000] WARNING: CPU: 0 PID: 0 at
> /home/ggallagher/devel/emb_linux/ipipe-arm/kernel/irq/chip.c:57
> irq_set_chip+0x7c/0xac
> [    0.000000] irqchip ARMCTRL-level is not pipeline-safe!
> [    0.000000] Modules linked in:
> [    0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 4.14.4-ipipe+ #2
> [    0.000000] Hardware name: BCM2835
> [    0.000000] I-pipe domain: Linux
> 
> it looks like cobalt isn't running either, I'm trying to find the best
> way to debug this or if this is a bug?

This is a heads up I have recently added telling us about an active IC
driver which has not been fixed up for interrupt pipelining yet, which
would certainly cause breakage with CONFIG_IPIPE on.

In that case, that would be:
drivers/irqchip/irq-bcm2835.c:	.name = "ARMCTRL-level",

Looking at the code, it has been fixed up properly, but it is missing
the new flag introduced for tagging such condition, i.e.:

diff --git a/drivers/irqchip/irq-bcm2835.c b/drivers/irqchip/irq-bcm2835.c
index bf8c9d4dc00d..ea2c31b58073 100644
--- a/drivers/irqchip/irq-bcm2835.c
+++ b/drivers/irqchip/irq-bcm2835.c
@@ -115,6 +115,7 @@ static struct irq_chip armctrl_chip = {
 	.irq_hold = armctrl_mask_irq,
 	.irq_release = armctrl_unmask_irq,
 #endif
+	.flags	= IRQCHIP_PIPELINE_SAFE,
 };

-- 
Philippe.

Re: [Xenomai] bcm2835 ipipe question

[Greg Gallagher]

Thanks Philippe, that worked.  Working through a couple more issues
and the bcm2835 and bcm2836 images should be ready to share.

-Greg

On Sun, Feb 11, 2018 at 4:44 AM, Philippe Gerum <rpm@xenomai.org> wrote:
> On 02/11/2018 02:57 AM, Greg Gallagher wrote:
>> I'm getting the raspberry pi zero images ready and I see the following
>> in the dmesg:
>>
>> [    0.000000] WARNING: CPU: 0 PID: 0 at
>> /home/ggallagher/devel/emb_linux/ipipe-arm/kernel/irq/chip.c:57
>> irq_set_chip+0x7c/0xac
>> [    0.000000] irqchip ARMCTRL-level is not pipeline-safe!
>> [    0.000000] Modules linked in:
>> [    0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 4.14.4-ipipe+ #2
>> [    0.000000] Hardware name: BCM2835
>> [    0.000000] I-pipe domain: Linux
>>
>> it looks like cobalt isn't running either, I'm trying to find the best
>> way to debug this or if this is a bug?
>
> This is a heads up I have recently added telling us about an active IC
> driver which has not been fixed up for interrupt pipelining yet, which
> would certainly cause breakage with CONFIG_IPIPE on.
>
> In that case, that would be:
> drivers/irqchip/irq-bcm2835.c:  .name = "ARMCTRL-level",
>
> Looking at the code, it has been fixed up properly, but it is missing
> the new flag introduced for tagging such condition, i.e.:
>
> diff --git a/drivers/irqchip/irq-bcm2835.c b/drivers/irqchip/irq-bcm2835.c
> index bf8c9d4dc00d..ea2c31b58073 100644
> --- a/drivers/irqchip/irq-bcm2835.c
> +++ b/drivers/irqchip/irq-bcm2835.c
> @@ -115,6 +115,7 @@ static struct irq_chip armctrl_chip = {
>         .irq_hold = armctrl_mask_irq,
>         .irq_release = armctrl_unmask_irq,
>  #endif
> +       .flags  = IRQCHIP_PIPELINE_SAFE,
>  };
>
> --
> Philippe.