Re: [RFC V1 2/7] drivers/base: Introduce callbacks for IMS interrupt domain
From: Jason Gunthorpe
Date: Fri Sep 13 2019 - 10:50:10 EST
On Thu, Sep 12, 2019 at 06:32:03PM -0700, Megha Dey wrote:
> This patch serves as a preparatory patch to introduce a new IMS
> (Interrupt Message Store) domain. It consists of APIs which would
> be used as callbacks to the IRQ chip associated with the IMS domain.
> The APIs introduced in this patch are:
> dev_ims_mask_irq - Generic irq chip callback to mask IMS interrupts
> dev_ims_unmask_irq - Generic irq chip callback to unmask IMS interrupts
> dev_ims_domain_write_msg - Helper to write MSI message to Device IMS
> It also introduces IMS specific structures namely:
> dev_ims_ops - Callbacks for IMS domain ops
> dev_ims_desc - Device specific IMS msi descriptor data
> dev_ims_priv_data - Internal data structure containing a unique devid
> and a pointer to the IMS domain ops
> Lastly, it adds a new config option MSI_IMS which must be enabled by
> any driver who would want to use the IMS infrastructure.
> Since IMS is not PCI compliant (like platform-msi), most of the code is
> similar to platform-msi.c.
It is very unclear what driver facing API is being proposed here
without a driver code example.
> TODO: Conclude if ims-msi.c and platform-msi.c can be merged.
Wow, the commit message for platform-msi.c explains that it is doing
*exactly* the same thing as ims - at least as far as the driver is
Clearly the ims idea already exists and the driver facing API should
be harmonized and enhanced to support both PCI and platform devices.
It would also make sense if the arch facing API was harmonized between
ARM and x86 as well. Would updating x86 to use the new generic
msi_domains ARM is using be needed?