On 13.02.2016, at 01:47, Eric Anholt [off-list ref] wrote:
Martin Sperl [off-list ref] writes:
quoted
Problem is that we need the VC4 bus-address (0x7e203000),
which is the actual <reg> value from the device tree without any
mapping.
Not the ARM MMU visible address mappings that mem->start provides
(typically 0x20203000 or 0x3f203000 for bcm2836)
Nor the mapped address (base) available in the kernel (typically
0xdc......).
Now that I've noticed the BCM2835_VCMMU_SHIFT removal, this makes sense,
but when you put unrelated changes like this together you end up slowing
down the review process on your patches. Please separate it out into a
separate commit.
OK - I agree: I really should have separated those, but as the
driver was broken for both reasons and the MMU_SHIFT was the earliest
fix I forgot about it when I got the final version for submission.
Anyway: Mark has already applied this patch on January 15th into for-next:
https://git.kernel.org/cgit/linux/kernel/git/broonie/sound.git/commit/sound/soc/bcm/bcm2835-i2s.c?h=for-next&id=517e7a1537ae4663268be5d0c0ec62c563b9fc99
so it should go into the next release...
So unless Mark wants to revert it and have two separate patches
instead, I guess we need to leave it like it is.
Martin