On 9/20/21 3:45 PM, Fabio Estevam wrote:

quoted

quoted

IMHO the current code is correct as-is.

   Not quite... I don't want to leave a bad example for the future driver authors. What should
I change in the patch description for the patch to become acceptable for you?

Please see how the PCI subsystem has converted the handling of
platform_get_irq():
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?h=v5.15-rc2&id=0584bff09629666eea97c7ac428e55b00df211f5

   Thanks for the link -- that's what I've been doing for the drivers outside PCI in the
past few months. :-)

 Why does drivers/ata/ need to handle platform_get_irq() differently?

   Because ata_host_activate() treats irq0 as polling indicater and complains about
the 'handler' being non-NULL.
 

I still don't see the need for changing drivers/ata/pata_imx.c in this aspect.

   And now?

MBR, Sergei