Re: swiotlb/caamjr regression (Was: [GIT PULL] (swiotlb) stable/for-linus-5.12)

From: Konrad Rzeszutek Wilk
Date: Mon Jun 21 2021 - 09:16:54 EST


On Mon, Jun 21, 2021 at 01:14:48PM +0900, 'Dominique MARTINET' wrote:
> Chanho Park wrote on Mon, Jun 21, 2021 at 11:55:22AM +0900:
> > Sure. No problem. But, the patch was already stacked on Konrad's tree
> > and linux-next as well.
> >
> > https://git.kernel.org/pub/scm/linux/kernel/git/konrad/swiotlb.git/commit/?h=devel/for-linus-5.14&id=33d1641f38f0c327bc3e5c21de585c77a6512bc6
>
> That patch is slightly different, it's a rewrite Konrad did that mixes
> in Linus' suggestion[1], which breaks things for the NVMe usecase
> Jianxiong Gao has.
>
> [1] offset = (tlb_addr - mem->start) & (IO_TLB_SIZE - 1)
>
>
> Konrad is aware so I think it shouldn't be submitted :)

The beaty of 'devel' and 'linux-next' is that they can be reshuffled and
mangled. I pushed them original patch from Bumyong there and will let
it sit for a day and then create a stable branch and give it to Linus.

Then I need to expand the test-regression bucket so that this does not
happen again. Dominique, how easy would it be to purchase one of those
devices?

I was originally thinking to create a crypto device in QEMU to simulate
this but that may take longer to write than just getting the real thing.

Or I could create some fake devices with weird offsets and write a driver
for it to exercise this.. like this one I had done some time ago that
needs some brushing off.