On 11/02/2015 09:38 PM, Thomas Gleixner wrote:

On Mon, 2 Nov 2015, Grygorii Strashko wrote:

quoted

Now in kernel below code pattern is used by many drivers:
static irqreturn_t driver_xx_hw_irq_handler(int irq, void *arg)
{
	<read IRQ status register>
   	<perform HW specific operations>

	for (<each set bit in IRQ status register>) {
		<get Linux IRQ number>
		generic_handle_irq(<Linux IRQ number>);
		|- handle_simple_irq()
		|-or- handle_level_irq()
		|-or- handle_edge_irq()
		   |-handle_irq_event()
		     |-handle_irq_event_percpu()
===
"WARNING: CPU: 1 PID: 82 at kernel/irq/handle.c:150 handle_irq_event_percpu+0x14c/0x174()
  irq 460 handler irq_default_primary_handler+0x0/0x14 enabled interrupts"
===
	}
}

On -RT above code will generate warnings, because driver_xx_hw_irq_handler()
will be forced threaded (by default) and, as result, generic_handle_irq()
will be called with IRQs enabled. To W/A this issue generic_handle_irq() can
be surrounded by raw_spin_lock_irqsave/irqrestore(wa_lock).

Instead of spreading this W/A directly in many drivers this patch
introduces -RT specific version of generic_handle_irq() API -
generic_handle_irq_rt_wa(). This new generic_handle_irq_rt_wa() just calls
generic_handle_irq() surrounded by raw_spin_lock_irqsave/irqrestore().
If -RT is disabled It will fallback to generic_handle_irq().

Why aren't you simply marking these demultiplex handlers with IRQ_NO_THREAD?

In general, it's possible. But, in this case, worst scenario will look like:
dra7xx_pcie_msi_irq_handler()
-> dw_handle_msi_irq()
   [code simplified]
   -> for (i = 0; i < MAX_MSI_IRQS; i++) {
	...
	generic_handle_irq(Y(i));
	...
   }
where MAX_MSI_IRQS = 32 now, but potentially can be increased up to 256.

Thanks.
-- 
regards,
-grygorii