Skip to contentSkip to conversation

gpio: rework locking and object life-time control

PATCH Jiawen Wu 3 participants 4 messages Apr 19, 2024 – Apr 22, 2024 GPIO

lore
Share & tools
lore

Jiawen Wu

author
#0 lore raw

Hi Bartosz Golaszewski,

I ran into a kernel crash problem when I pull the latest net-next.git, and finally it was found that is caused by this patch series merged.

The kernel crashed because I got gpio=0 when I called irq_find_mapping() and then struct irq_data *d=null, as my driver describes:

	int gpio = irq_find_mapping(gc->irq.domain, hwirq);
	struct irq_data *d = irq_get_irq_data(gpio);

	txgbe_gpio_irq_ack(d);

The deeper positioning is this line in __irq_resolve_mapping().

	data = rcu_dereference(domain->revmap[hwirq]);

So, is it the addition of SRCU infrastructure that causes this issue?

Thanks,
Jiawen

#1 lore raw
Quoted 4 lines · Jiawen Wu

Hi Bartosz Golaszewski,

I ran into a kernel crash problem when I pull the latest net-next.git, and finally it was found that is caused by this patch series merged.

Can you bisect further, i.e. which patch (now a commit message) is the culprit?

Quoted 8 lines

The kernel crashed because I got gpio=0 when I called irq_find_mapping() and then struct irq_data *d=null, as my driver describes:

	int gpio = irq_find_mapping(gc->irq.domain, hwirq);
	struct irq_data *d = irq_get_irq_data(gpio);

	txgbe_gpio_irq_ack(d);
Quoted 5 lines

The deeper positioning is this line in __irq_resolve_mapping().

	data = rcu_dereference(domain->revmap[hwirq]);

So, is it the addition of SRCU infrastructure that causes this issue?

Do you have a full traceback / Oops message to share?

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#2 lore raw
In reply to Jiawen Wu
Quoted 14 lines · Jiawen Wu

Hi Bartosz Golaszewski,

I ran into a kernel crash problem when I pull the latest net-next.git, and finally it was found that is caused by this patch series merged.

The kernel crashed because I got gpio=0 when I called irq_find_mapping() and then struct irq_data *d=null, as my driver describes:

        int gpio = irq_find_mapping(gc->irq.domain, hwirq);
        struct irq_data *d = irq_get_irq_data(gpio);

        txgbe_gpio_irq_ack(d);
Quoted 6 lines

The deeper positioning is this line in __irq_resolve_mapping().

        data = rcu_dereference(domain->revmap[hwirq]);

So, is it the addition of SRCU infrastructure that causes this issue?

This is irq-specific RCU that I did not add in the GPIO series. Please provide us with more information. Bisect to the exact commit causing the issue and post the kernel log (we don't know what kind of crash you trigger and what the stack trace is).

Bart

author
#3 lore raw
Quoted 22 lines · Bartosz Golaszewski
Quoted 14 lines · Jiawen Wu

Hi Bartosz Golaszewski,

I ran into a kernel crash problem when I pull the latest net-next.git, and finally it was found that is caused by this patch series merged.

The kernel crashed because I got gpio=0 when I called irq_find_mapping() and then struct irq_data *d=null, as my driver describes:

        int gpio = irq_find_mapping(gc->irq.domain, hwirq);
        struct irq_data *d = irq_get_irq_data(gpio);

        txgbe_gpio_irq_ack(d);
Quoted 6 lines

The deeper positioning is this line in __irq_resolve_mapping().

        data = rcu_dereference(domain->revmap[hwirq]);

So, is it the addition of SRCU infrastructure that causes this issue?

Quoted 7 lines

This is irq-specific RCU that I did not add in the GPIO series. Please provide us with more information. Bisect to the exact commit causing the issue and post the kernel log (we don't know what kind of crash you trigger and what the stack trace is).

Bart

Hi Bartosz & Andy,

Thanks for your replies.

I'm sorry for the misunderstanding, and glad this patch doesn't cause any problems. I thought the issue was in this patch because of my mistake. It's actually caused by other patches. :)