[OpenWrt-Devel] [PATCH] ath79: ar7100: remove IRQ code from PCI driver

[Dmitry Tunin]

> This is true for ar7100. On other chips the entire
> RST_MISC_INTERRUPT_STATUS block is marked as Read/Write-to-Clear.
Do you have the ar7240 datasheet with those specs?

> > It makes no sense to have a separate ack done this way. It slow things down.
> You can use an implementation without ack for PCI but you should keep
> the one used for MISC_INTC unchanged.

It is not "without ack". Implementations can be with separate "mask"
and "ack", or with "mask_ack" that does both.
The second way is faster and recommended.

For this simple controller there is no reason of using separate "mask"
and "ack", especially doing "ack" by writhing to the status register.
What is the point of doing it? What is the goal?

It looks like someone when writing the code didn't know what to define
for "ack" and wrote that.
Then someone else did a better job for 7100, by doing a correct way.
A third person merged these two ways in one driver without having a
clue why it is done this way.

Regarding the "real" miscintc controller it doesn't make much
difference because it serves slow devices like ohci.
But for PCI it is not a good idea.

Anyway in my current patch the compatible should be changed to
"qca,ar7100-misc-intc".

I can suggest other changes for discussion and testing later on.

_______________________________________________
openwrt-devel mailing list
openwrt-devel at lists.openwrt.org
https://lists.openwrt.org/mailman/listinfo/openwrt-devel