Re: [GIT pull] irq updates for 4.13

[Linus Torvalds]

On Tue, Jul 11, 2017 at 7:41 AM, Thomas Gleixner <tglx@xxxxxxxxxxxxx> wrote:
>
> Ah. Now that makes sense.
>
> Unpatched the ordering is:
>
>           chip_bus_lock(desc);
>           irq_request_resources(desc);

I *looked* at that ordering and then went "Naah, that makes no sense".

But if that's the only issue, how about we just re-order those things
- we still don't need to move the irq_request_resources() into the
spinlock, we just move it to below the chip_bus_lock().

IOW, something like the (COMPLETELY UNTEESTED!) attached patch.

This assumes that the chip_bus_lock() thing is still ok for the RT
case, but it looks like it might be: the only other one I looked at
(apart from the gpio-omap one) used a mutex.

             Linus
 kernel/irq/manage.c | 11 +++++------
 1 file changed, 5 insertions(+), 6 deletions(-)

diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c
index 5624b2dd6b58..ea1b9404c041 100644
--- a/kernel/irq/manage.c
+++ b/kernel/irq/manage.c
@@ -1168,17 +1168,17 @@ __setup_irq(unsigned int irq, struct irq_desc *desc, struct irqaction *new)
 		new->flags &= ~IRQF_ONESHOT;
 
 	mutex_lock(&desc->request_mutex);
+	chip_bus_lock(desc);
+
 	if (!desc->action) {
 		ret = irq_request_resources(desc);
 		if (ret) {
 			pr_err("Failed to request resources for %s (irq %d) on irqchip %s\n",
 			       new->name, irq, desc->irq_data.chip->name);
-			goto out_mutex;
+			goto out_unlock_chip_bus;
 		}
 	}
 
-	chip_bus_lock(desc);
-
 	/*
 	 * The following block of code has to be executed atomically
 	 */
@@ -1385,12 +1385,11 @@ __setup_irq(unsigned int irq, struct irq_desc *desc, struct irqaction *new)
 out_unlock:
 	raw_spin_unlock_irqrestore(&desc->lock, flags);
 
-	chip_bus_sync_unlock(desc);
-
 	if (!desc->action)
 		irq_release_resources(desc);
 
-out_mutex:
+out_unlock_chip_bus:
+	chip_bus_sync_unlock(desc);
 	mutex_unlock(&desc->request_mutex);
 
 out_thread: