Re: [PATCH 1/2] dma-mapping: add a dma_ops_bypass flag to struct device

On 07/04/2020 20:12, Alexey Kardashevskiy wrote:

On 07/04/2020 03:17, Christoph Hellwig wrote:

quoted

On Mon, Apr 06, 2020 at 11:25:09PM +1000, Alexey Kardashevskiy wrote:

quoted

quoted

quoted

Do you see any serious problem with this approach? Thanks!

Do you have a link to the whole branch?  The github UI is unfortunately
unusable for that (or I'm missing something).

The UI shows the branch but since I rebased and forcepushed it, it does
not. Here is the current one with:

https://github.com/aik/linux/commits/dma-bypass.3

Ok, so we use the core bypass without persistent memory, and then
have another bypass mode on top.  Not great, but I can't think
of anything better.  Note that your checks for the map_sg case
aren't very efficient - for one it would make sense to calculate
the limit only once, 

Good points, I'll post revised version when you post your v3 of this.

Any plans on posting v3 of this? Thanks,

quoted

but also it would make sense to reuse the
calculted diecect mapping addresses instead of doing another pass
later on in the dma-direct code.

Probably but I wonder what kind of hardware we need to see the
difference. I might try, just need to ride to the office to plug the
cable in my 100GBit eth machines :) Thanks,

-- 
Alexey