Re: [PATCH] dt-bindings: Convert vendor prefixes to json-schema

[Maxime Ripard]

On Mon, May 20, 2019 at 11:35:01AM -0500, Rob Herring wrote:
> On Mon, May 20, 2019 at 8:18 AM Maxime Ripard <maxime.ripard@xxxxxxxxxxx> wrote:
> >
> > Hi Rob,
> >
> > On Fri, May 10, 2019 at 02:40:18PM -0500, Rob Herring wrote:
> > > Convert the vendor prefix registry to a schema. This will enable checking
> > > that new vendor prefixes are added (in addition to the less than perfect
> > > checkpatch.pl check) and will also check against adding other prefixes
> > > which are not vendors.
> > >
> > > Converted vendor-prefixes.txt using the following sed script:
> > >
> > > sed -e 's/\([a-zA-Z0-9\-]*\)[[:space:]]*\([a-zA-Z0-9].*\)/  "^\1,\.\*\":\n    description: \2/'
> > >
> > > Signed-off-by: Rob Herring <robh@xxxxxxxxxx>
> > > ---
> > > As vendor prefix updates come in via multiple trees, I plan to merge
> > > this before -rc1 to avoid cross tree conflicts.
> >
> > I just tried this with the 5.2-rc1 release, and this very
> > significantly slows down the validation.
> >
> > With a dtbs_check run on (arm's) sunxi_defconfig, on my core-i5 with 4
> > threads, I go from 1.30 minutes to more than 12.
>
> Indeed. 6 min to 45 min for allmodconfig. However, it's only 5 min to
> run checks with only this file. I'd expect a more linear hit. Maybe
> we're exceeding some cache size and thrashing.
>
> > Should we improve the dt-validate tool before merging this patch?
>
> How? I've looked at optimizing things some and implemented areas I
> found (primarily, saving the fixed-up schema and not printing line
> numbers (of the yaml encoded DT)).

I guess we can always try to profile the dt-validate tool to see where
the bottleneck is.

Another area of improvement might be to use a faster language. Rust's
regex engine is known to be blazing fast for example. However, it's a
pretty exotic language, and it would probably hurt the
maintainability. Maybe go would be a good in-between?

> I've been wanting to have some way to categorize checks, so we can
> split rules from pedantic guidance. Maybe we can add a level keyword
> in select or something.

That would be a good mitigation indeed.

> Short term, I'm fine with just disabling this one by default with
> 'select: false'.

Ok. Do you want to send it or should I do it?

Thanks!
Maxime

--
Maxime Ripard, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

Attachment: signature.asc
Description: PGP signature