On Mon, Jun 18, 2018 at 12:49:37PM -0600, Rob Herring wrote: > +devicetree-spec > > On Sun, Jun 17, 2018 at 5:47 AM, David Gibson > <david@xxxxxxxxxxxxxxxxxxxxx> wrote: > > On Fri, Jun 15, 2018 at 03:37:16PM -0600, Rob Herring wrote: > >> According to the DT Spec (and ePAPR), node and property names should be > >> 1-31 characters. > >> > >> Signed-off-by: Rob Herring <robh@xxxxxxxxxx> > > > > I don't think this is a good idea. > > > >> --- > >> Surprisingly, there aren't just a ton of warnings introduced with this. > >> I'm not really sure where the 31 character limit came from. Maybe it > >> should be increased rather than trying to fix any long names. Either > >> way, we should have a check if there's a spec'ed limit. > > > > The 31 character limit comes from IEEE 1275. But.. even long before > > flattened trees, device trees from both IBM and Apple frequently > > violated it. I think we should probably just ditch this limitation > > from the spec. > > Having no limit means names are only limited by variable constraints > of the cpu or OS/client and there's no way for the OS to know what the > max is. In reality, we don't need *anywhere* close to what would be a > maximum string length of a client. Do we really want it undefined in > the spec? Don't know about IBM and Apple DTs, but just doubling the > limit would eliminate all the warnings of the dts files in the kernel. > > Even if it is not a spec limit, I think there should be a style or > best practice limit. We already have that type of check for the > character set, so why not the length? Good points. Yes, we should probably still have a limit, just a larger one than current. I'd be inclined to set it to 255 - it's plenty long enough but still fits in 8-bits, which might be useful to someone. -- David Gibson | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_ | _way_ _around_! http://www.ozlabs.org/~dgibson
Attachment:
signature.asc
Description: PGP signature