Re: [PATCH v5 00/40] x86/resctrl: Move the resctrl filesystem code to /fs/resctrl

From: Tony Luck
Date: Tue Oct 08 2024 - 19:24:30 EST


On Fri, Oct 04, 2024 at 06:03:07PM +0000, James Morse wrote:
> Changes since v4?:
> * Dropped the percentage/mbps distinction, this can be future cleanup as I
> think the difference matters to user-space. These are both treated as a
> 'range'.
> * Picked a pre-requisite cleanup patch from Christophe to make merging
> easier.
> * More of the __init/__exit stuff has consolodated in the patch that removes
> them from filesystem code.
> Regardless, changes are noted on each patch.
>
> ~
>
> This is the final series that allows other architectures to implement resctrl.
> The final patch to move the code has been omited, but can be generated using
> the python script at the end of the series.
> The final move is a bit of a monster. I don't expect that to get merged as part
> of this series - we should wait for it to make less impact on other series.

I've been playing around with this series (after running the python
script to get the full effect of where the code is headed).

Things seem to look pretty good. I haven't noticed anything failing
(but I haven't done extensive testing).

I've skimmed over the patches and posted some nitpicks on a couple
of them. I'll try to find time to make a proper review pass.

Just to see whether I still understand the new layout I took my
patches for mba_MBps[1] and applied them on top of all of this.
The process was pretty simple, most functions that were changed
just moved to fs/resctrl. Just a few places where I needed to
tweak things to fit on top of the changes (like using the
new resctrl_arch_get_resource() to lookup rdt_resources
instead of indexing into the rdt_resources_all[] array,
of function names changed to add "_arch").

Overall it looks pretty good.

-Tony

[1] https://lore.kernel.org/all/20241003191228.67541-1-tony.luck@xxxxxxxxx/