On 10/23/2013 11:23 AM, Felipe Balbi wrote:
Hi again,
On Wed, Oct 23, 2013 at 01:21:43PM -0500, Felipe Balbi wrote:
On Wed, Oct 23, 2013 at 11:10:48AM -0700, David Cohen wrote:
My idea is always use threaded irq and passing flags into request.
Like as:
unsigned long flags = res->flags & IORESOURCE_BITS;
...
request_threaded_irq(max->irq, serial_m3110_irq, IRQF_ONESHOT | flags, "max3110", max);
Oh, maybe we were talking about different things afterall :)
The reason this struct plat_max3110 was created is to allow platform
code (located under arch/x86/platform/intel-mid/device_libs/) to define
the irq edge type.
When I saw your comment I though you were referring to struct resource
(which has IORESOURCE_IRQ_* flags). But unlike platform_device,
spi_device has no struct resource * to replace the need of struct
plat_max3110.
OTOH your suggestion can replace this piece of code:
@@ -68,6 +69,7 @@ struct uart_max3110 {
u8 clock;
u8 parity, word_7bits;
u16 irq;
+ u16 irq_edge_triggered;
max3110 is already edge triggered:
495 ret = request_irq(max->irq, serial_m3110_irq,
496 IRQ_TYPE_EDGE_FALLING, "max3110", max);
it would be nice a threaded IRQ instead of using a singlethread
workqueue, though.
making it clearer, you only converted to threaded IRQ when
edge_triggered isn't set, why don't you *always* use threaded IRQs ?
This is a very difficult question to answer, considering the
non-threaded irq is used just to wake up a wq :)
I wish this driver was better maintained.
I'll rework this whole IRQ scheme here.
Br, David Cohen
--
To unsubscribe from this list: send the line "unsubscribe linux-serial" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
