A question about generated phandle during dtb=>dts conversion

A question about generated phandle during dtb=>dts conversion

[Chan Kim]

Hello, all
I extracted a dtb file from qemu execution (arm virt machine) and converted
the dtb file to dts file using dtc.
But for some node that is referenced by other nodes, the phandle is
generated like “phandle = <0x8000>” as shown below.

apb-pclk {
        phandle = <0x8000>;
        clock-output-names = "clk24mhz";
        clock-frequency = <0x16e3600>;
        #clock-cells = <0x00>;
        compatible = "fixed-clock";
    };

And another node is using this phandle like this. (the clocks is the phandle
for the apb-pclk above.)

pl061@9030000 {
        phandle = <0x8003>;
        clock-names = "apb_pclk";
        clocks = <0x8000>;
        interrupts = <0x00 0x07 0x04>;
        gpio-controller;
        #gpio-cells = <0x02>;
        compatible = "arm,pl061\0arm,primecell";
        reg = <0x00 0x9030000 0x00 0x1000>;
    };

When I convert the dts back to the dtb, dtc doesn’t understand the phandle
and complains that it is not a phandle referenece.

dts.dts:276.3-21: Warning (clocks_property): /pl061@9030000:clocks: cell 0
is not a phandle reference

I’ve checked event with this warning, the phandle reference is being
preserved through out the procedure.
So, if I want to extract a dtb and convert it to dts, edit some and convert
it back to dtb, how can I prevent this kind of phandle warning? 
And why doesn't dtc add labels to the referenced nodes and let other nodes
use that label as phandle?

Thank you!
Regards,

Chan Kim

Re: A question about generated phandle during dtb=>dts conversion

[David Gibson]

[-- Attachment #1: Type: text/plain, Size: 3012 bytes --]

On Tue, Nov 02, 2021 at 06:03:27PM +0900, Chan Kim wrote:
> Hello, all
> I extracted a dtb file from qemu execution (arm virt machine) and converted
> the dtb file to dts file using dtc.
> But for some node that is referenced by other nodes, the phandle is
> generated like “phandle = <0x8000>” as shown below.

Note that the value is generated by qemu and is in the dtb.  It's not
generated by dtc during decompilation.

> apb-pclk {
>         phandle = <0x8000>;
>         clock-output-names = "clk24mhz";
>         clock-frequency = <0x16e3600>;
>         #clock-cells = <0x00>;
>         compatible = "fixed-clock";
>     };
> 
> And another node is using this phandle like this. (the clocks is the phandle
> for the apb-pclk above.)
> 
> pl061@9030000 {
>         phandle = <0x8003>;
>         clock-names = "apb_pclk";
>         clocks = <0x8000>;
>         interrupts = <0x00 0x07 0x04>;
>         gpio-controller;
>         #gpio-cells = <0x02>;
>         compatible = "arm,pl061\0arm,primecell";
>         reg = <0x00 0x9030000 0x00 0x1000>;
>     };
> 
> When I convert the dts back to the dtb, dtc doesn’t understand the phandle
> and complains that it is not a phandle referenece.
> 
> dts.dts:276.3-21: Warning (clocks_property): /pl061@9030000:clocks: cell 0
> is not a phandle reference
> 
> I’ve checked event with this warning, the phandle reference is being
> preserved through out the procedure.
> So, if I want to extract a dtb and convert it to dts, edit some and convert
> it back to dtb, how can I prevent this kind of phandle warning? 
> And why doesn't dtc add labels to the referenced nodes and let other nodes
> use that label as phandle?

The short answer is because dtc doesn't know it's a phandle.  dtb is
not a self-describing format - the properties are just bytestrings,
and you need to know what you're expecting in there in order to decode
them (that's what dt bindings give you).

So the only way that dtc could know that those 4 bytes in the 'clocks'
property are actually a phandle is by specifically knowing about the
"clocks" property.  dtc has *some* knowledge of specific properties in
the checks, but it's far from complete (and complete knowledge of the
bindings is pretty much out of scope for dtc).

But.. for the normal case of writing a dts file directly, it's useful
to get a warning if you don't use a reference somewhere you probably
should (anywhere that's a phandle), so dtc does check that.

So, I think basically in just the same way that you might expect
generated or decompiled code to be rather messy and generate compiler
warnings, so a decompiled dts might have messy things and generate
warnings.  You can switch off the warnings explicitly if you need
(-Wno-clocks_property in this case).

-- 
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 #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

RE: A question about generated phandle during dtb=>dts conversion

[Chan Kim]

Hi, David Gibson,
Thank you for the answer. I can understand it now.
Best regards,
Chan Kim
> -----Original Message-----
> From: David Gibson <david-xT8FGy+AXnRB3Ne2BGzF6laj5H9X9Tb+@public.gmane.org>
> Sent: Wednesday, November 3, 2021 11:34 AM
> To: Chan Kim <ckim-mINagvnPBMnWF1p7bDn8Kw@public.gmane.org>
> Cc: devicetree-compiler-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> Subject: Re: A question about generated phandle during dtb=>dts conversion
> 
> On Tue, Nov 02, 2021 at 06:03:27PM +0900, Chan Kim wrote:
> > Hello, all
> > I extracted a dtb file from qemu execution (arm virt machine) and
> > converted the dtb file to dts file using dtc.
> > But for some node that is referenced by other nodes, the phandle is
> > generated like “phandle = <0x8000>” as shown below.
> 
> Note that the value is generated by qemu and is in the dtb.  It's not
> generated by dtc during decompilation.
> 
> > apb-pclk {
> >         phandle = <0x8000>;
> >         clock-output-names = "clk24mhz";
> >         clock-frequency = <0x16e3600>;
> >         #clock-cells = <0x00>;
> >         compatible = "fixed-clock";
> >     };
> >
> > And another node is using this phandle like this. (the clocks is the
> > phandle for the apb-pclk above.)
> >
> > pl061@9030000 {
> >         phandle = <0x8003>;
> >         clock-names = "apb_pclk";
> >         clocks = <0x8000>;
> >         interrupts = <0x00 0x07 0x04>;
> >         gpio-controller;
> >         #gpio-cells = <0x02>;
> >         compatible = "arm,pl061\0arm,primecell";
> >         reg = <0x00 0x9030000 0x00 0x1000>;
> >     };
> >
> > When I convert the dts back to the dtb, dtc doesn’t understand the
> > phandle and complains that it is not a phandle referenece.
> >
> > dts.dts:276.3-21: Warning (clocks_property): /pl061@9030000:clocks:
> > cell 0 is not a phandle reference
> >
> > I’ve checked event with this warning, the phandle reference is being
> > preserved through out the procedure.
> > So, if I want to extract a dtb and convert it to dts, edit some and
> > convert it back to dtb, how can I prevent this kind of phandle warning?
> > And why doesn't dtc add labels to the referenced nodes and let other
> > nodes use that label as phandle?
> 
> The short answer is because dtc doesn't know it's a phandle.  dtb is not a
> self-describing format - the properties are just bytestrings, and you need
> to know what you're expecting in there in order to decode them (that's
> what dt bindings give you).
> 
> So the only way that dtc could know that those 4 bytes in the 'clocks'
> property are actually a phandle is by specifically knowing about the
> "clocks" property.  dtc has *some* knowledge of specific properties in the
> checks, but it's far from complete (and complete knowledge of the bindings
> is pretty much out of scope for dtc).
> 
> But.. for the normal case of writing a dts file directly, it's useful to
> get a warning if you don't use a reference somewhere you probably should
> (anywhere that's a phandle), so dtc does check that.
> 
> So, I think basically in just the same way that you might expect generated
> or decompiled code to be rather messy and generate compiler warnings, so a
> decompiled dts might have messy things and generate warnings.  You can
> switch off the warnings explicitly if you need (-Wno-clocks_property in
> this case).
> 
> --
> 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

Re: A question about generated phandle during dtb=>dts conversion

['David Gibson']

[-- Attachment #1: Type: text/plain, Size: 4372 bytes --]

On Wed, Nov 03, 2021 at 06:06:47PM +0900, Chan Kim wrote:
> 
> Hi, David Gibson,
> Thank you for the answer. I can understand it now.
> Best regards,
> Chan Kim
> > -----Original Message-----
> > From: David Gibson <david-xT8FGy+AXnRB3Ne2BGzF6laj5H9X9Tb+@public.gmane.org>
> > Sent: Wednesday, November 3, 2021 11:34 AM
> > To: Chan Kim <ckim-mINagvnPBMnWF1p7bDn8Kw@public.gmane.org>
> > Cc: devicetree-compiler-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> > Subject: Re: A question about generated phandle during dtb=>dts conversion
> > 
> > On Tue, Nov 02, 2021 at 06:03:27PM +0900, Chan Kim wrote:
> > > Hello, all
> > > I extracted a dtb file from qemu execution (arm virt machine) and
> > > converted the dtb file to dts file using dtc.
> > > But for some node that is referenced by other nodes, the phandle is
> > > generated like “phandle = <0x8000>” as shown below.
> > 
> > Note that the value is generated by qemu and is in the dtb.  It's not
> > generated by dtc during decompilation.
> > 
> > > apb-pclk {
> > >         phandle = <0x8000>;
> > >         clock-output-names = "clk24mhz";
> > >         clock-frequency = <0x16e3600>;
> > >         #clock-cells = <0x00>;
> > >         compatible = "fixed-clock";
> > >     };
> > >
> > > And another node is using this phandle like this. (the clocks is the
> > > phandle for the apb-pclk above.)
> > >
> > > pl061@9030000 {
> > >         phandle = <0x8003>;
> > >         clock-names = "apb_pclk";
> > >         clocks = <0x8000>;
> > >         interrupts = <0x00 0x07 0x04>;
> > >         gpio-controller;
> > >         #gpio-cells = <0x02>;
> > >         compatible = "arm,pl061\0arm,primecell";
> > >         reg = <0x00 0x9030000 0x00 0x1000>;
> > >     };
> > >
> > > When I convert the dts back to the dtb, dtc doesn’t understand the
> > > phandle and complains that it is not a phandle referenece.
> > >
> > > dts.dts:276.3-21: Warning (clocks_property): /pl061@9030000:clocks:
> > > cell 0 is not a phandle reference
> > >
> > > I’ve checked event with this warning, the phandle reference is being
> > > preserved through out the procedure.
> > > So, if I want to extract a dtb and convert it to dts, edit some and
> > > convert it back to dtb, how can I prevent this kind of phandle warning?
> > > And why doesn't dtc add labels to the referenced nodes and let other
> > > nodes use that label as phandle?
> > 
> > The short answer is because dtc doesn't know it's a phandle.  dtb is not a
> > self-describing format - the properties are just bytestrings, and you need
> > to know what you're expecting in there in order to decode them (that's
> > what dt bindings give you).
> > 
> > So the only way that dtc could know that those 4 bytes in the 'clocks'
> > property are actually a phandle is by specifically knowing about the
> > "clocks" property.  dtc has *some* knowledge of specific properties in the
> > checks, but it's far from complete (and complete knowledge of the bindings
> > is pretty much out of scope for dtc).
> > 
> > But.. for the normal case of writing a dts file directly, it's useful to
> > get a warning if you don't use a reference somewhere you probably should
> > (anywhere that's a phandle), so dtc does check that.
> > 
> > So, I think basically in just the same way that you might expect generated
> > or decompiled code to be rather messy and generate compiler warnings, so a
> > decompiled dts might have messy things and generate warnings.  You can
> > switch off the warnings explicitly if you need (-Wno-clocks_property in
> > this case).

Actually, a nicer way of handling this just occurred to me.  It would
probably be a good idea to add to the dts syntax a new tag, maybe
"/strict/".  Without this tag, checks which rely on information that's
only in the source, not the dtb (types, phandles etc.) would be
disabled by default.

The tag itself obviously wouldn't be encoded into the dtb, so if you
compiled then decompiled the tag would be lost, meaning that the
output dtb wouldn't trigger checks.

Unfortunately, I'm very unlikely to have time to implement this any
time soon.

-- 
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 #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]