Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > linux.debian.kernel > #78251 > unrolled thread

Bug#910727: Overlays cannot be applied to dtbs - missing DTC_FLAGS='-@'

Started byNicolas Frattaroli <frattaroli.nicolas@gmail.com>
First post2023-02-12 19:00 +0100
Last post2025-04-25 12:30 +0200
Articles 5 — 5 participants

Back to article view | Back to linux.debian.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.


Contents

  Bug#910727: Overlays cannot be applied to dtbs - missing DTC_FLAGS='-@' Nicolas Frattaroli <frattaroli.nicolas@gmail.com> - 2023-02-12 19:00 +0100
    Bug#910727: Overlays cannot be applied to dtbs - missing DTC_FLAGS='-@' Diederik de Haas <didi.debian@cknow.org> - 2023-07-31 10:10 +0200
      Bug#910727: Overlays cannot be applied to dtbs - missing DTC_FLAGS='-@' Salvatore Bonaccorso <carnil@debian.org> - 2025-04-25 11:20 +0200
        Processed: Re: Bug#910727: Overlays cannot be applied to dtbs -  missing DTC_FLAGS='-@' "Debian Bug Tracking System" <owner@bugs.debian.org> - 2025-04-25 11:20 +0200
        Bug#910727: Overlays cannot be applied to dtbs - missing DTC_FLAGS='-@' "Diederik de Haas" <didi.debian@cknow.org> - 2025-04-25 12:30 +0200

#78251 — Bug#910727: Overlays cannot be applied to dtbs - missing DTC_FLAGS='-@'

FromNicolas Frattaroli <frattaroli.nicolas@gmail.com>
Date2023-02-12 19:00 +0100
SubjectBug#910727: Overlays cannot be applied to dtbs - missing DTC_FLAGS='-@'
Message-ID<FYoNP-5n83-3@gated-at.bofh.it>
Hello,

I realise this is quite an old bug, but it would still be of interest
to me to get this enabled. The -@ option will increase the size of the
compiled device tree blobs somewhat, but on the flipside, u-boot-menu's
device tree overlay functionality will actually be useful.

In my case, I'm interested in this for PINE64's Quartz64 lineup of ARM
devices. Since these are development boards that expose non-enumerating
protocols like I2C and SPI, the use of device tree overlays is pretty
much required to add additional modules to the board.

Kind regards,
Nicolas Frattaroli

[toc] | [next] | [standalone]


#79824

FromDiederik de Haas <didi.debian@cknow.org>
Date2023-07-31 10:10 +0200
Message-ID<GXw8x-40o0-1@gated-at.bofh.it>
In reply to#78251

[Multipart message — attachments visible in raw view] — view raw

On Monday, 31 July 2023 04:28:31 CEST MOESSBAUER, Felix wrote:
> On Sun, 12 Feb 2023 18:52:07 +0100 Nicolas Frattaroli
> <frattaroli.nicolas@gmail.com> wrote:
> > I realise this is quite an old bug, but it would still be of interest
> > to me to get this enabled. The -@ option will increase the size of
> > the compiled device tree blobs somewhat, but on the flipside, 
> > u-boot-menu's device tree overlay functionality will actually be useful.
> 
> The increase is around 30% in average. For the linux-image-6.1.0-10-
> arm64, we have around 30MB of dtb files (uncompressed). By that, the
> package would grow by 10MB (uncompressed) or even less when compressed.

I'm not a kernel maintainer, but I wouldn't expect it to be applied to 
Debian's Stable kernel, ie the 6.1 series.

> On many (if not most) arm based boards, these device tree overlays are
> very useful. There are also many kernel patches about enabling the
> symbol support on a per-device basis. However, kernel devs seem to have
> some mixed feelings about that. A proposed solution is to enable this
> at distro level [1].
> @Ben: Would it be an accepted solution to enable the DTC_FLAGS += -@
> for all armhf, armmp, arm64 and riscv devices?

AFAIUI, the upstream kernel devs were against such a blanket enablement.
(Consequently it sounds like a bad idea if Debian would do it)

https://lore.kernel.org/all/20220328000915.15041-1-ansuelsmth@gmail.com/ was a 
renewed attempt at restructuring the various .dts* into vendor directories and 
with 6.5 that got merged into the upstream kernel.
And (AFAIUI) that was seen as a prerequisite to start enabling the "-@" flag on 
a vendor/directory level, with optionally a subselection if it didn't make 
sense to enable it for all boards ... pretty much (exactly) as you proposed.

Doing it at the distro level was rejected (at least for Debian) with the 
argument that it should be done upstream.
As there is some (?) coordination/synchronization wrt DeviceTrees between the 
upstream kernel and u-boot, that also seems required for the functionality 
quoted at the top of this email.

You mentioned you proposed it for RISC-V boards and some DDs proposed it for 
RPi boards and both got accepted. I think/expected that we would see (many?) 
more such patches getting accepted going forward now that the above mentioned 
dts directory restructuring has taken place.

IOW: I agree with the reasoning you brought up in your patch submission thread 
and keep targeting inclusion in the upstream kernel.

Cheers,
  Diederik

[toc] | [prev] | [next] | [standalone]


#87043

FromSalvatore Bonaccorso <carnil@debian.org>
Date2025-04-25 11:20 +0200
Message-ID<KFnnY-g7aW-15@gated-at.bofh.it>
In reply to#79824
Control: tags -1 + moreinfo

On Mon, Jul 31, 2023 at 10:03:52AM +0200, Diederik de Haas wrote:
> On Monday, 31 July 2023 04:28:31 CEST MOESSBAUER, Felix wrote:
> > On Sun, 12 Feb 2023 18:52:07 +0100 Nicolas Frattaroli
> > <frattaroli.nicolas@gmail.com> wrote:
> > > I realise this is quite an old bug, but it would still be of interest
> > > to me to get this enabled. The -@ option will increase the size of
> > > the compiled device tree blobs somewhat, but on the flipside, 
> > > u-boot-menu's device tree overlay functionality will actually be useful.
> > 
> > The increase is around 30% in average. For the linux-image-6.1.0-10-
> > arm64, we have around 30MB of dtb files (uncompressed). By that, the
> > package would grow by 10MB (uncompressed) or even less when compressed.
> 
> I'm not a kernel maintainer, but I wouldn't expect it to be applied to 
> Debian's Stable kernel, ie the 6.1 series.
> 
> > On many (if not most) arm based boards, these device tree overlays are
> > very useful. There are also many kernel patches about enabling the
> > symbol support on a per-device basis. However, kernel devs seem to have
> > some mixed feelings about that. A proposed solution is to enable this
> > at distro level [1].
> > @Ben: Would it be an accepted solution to enable the DTC_FLAGS += -@
> > for all armhf, armmp, arm64 and riscv devices?
> 
> AFAIUI, the upstream kernel devs were against such a blanket enablement.
> (Consequently it sounds like a bad idea if Debian would do it)
> 
> https://lore.kernel.org/all/20220328000915.15041-1-ansuelsmth@gmail.com/ was a 
> renewed attempt at restructuring the various .dts* into vendor directories and 
> with 6.5 that got merged into the upstream kernel.
> And (AFAIUI) that was seen as a prerequisite to start enabling the "-@" flag on 
> a vendor/directory level, with optionally a subselection if it didn't make 
> sense to enable it for all boards ... pretty much (exactly) as you proposed.
> 
> Doing it at the distro level was rejected (at least for Debian) with the 
> argument that it should be done upstream.
> As there is some (?) coordination/synchronization wrt DeviceTrees between the 
> upstream kernel and u-boot, that also seems required for the functionality 
> quoted at the top of this email.
> 
> You mentioned you proposed it for RISC-V boards and some DDs proposed it for 
> RPi boards and both got accepted. I think/expected that we would see (many?) 
> more such patches getting accepted going forward now that the above mentioned 
> dts directory restructuring has taken place.
> 
> IOW: I agree with the reasoning you brought up in your patch submission thread 
> and keep targeting inclusion in the upstream kernel.

Was there progress here, or what should we further do? If it's not
something which is going to land upstream then I would suggest that we
close the bug.

Regards,
Salvatore

[toc] | [prev] | [next] | [standalone]


#87045 — Processed: Re: Bug#910727: Overlays cannot be applied to dtbs - missing DTC_FLAGS='-@'

From"Debian Bug Tracking System" <owner@bugs.debian.org>
Date2025-04-25 11:20 +0200
SubjectProcessed: Re: Bug#910727: Overlays cannot be applied to dtbs - missing DTC_FLAGS='-@'
Message-ID<KFnnY-g7aW-21@gated-at.bofh.it>
In reply to#87043
Processing control commands:

> tags -1 + moreinfo
Bug #910727 [src:linux] Overlays cannot be applied to dtbs - missing DTC_FLAGS='-@'
Added tag(s) moreinfo.

-- 
910727: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=910727
Debian Bug Tracking System
Contact owner@bugs.debian.org with problems

[toc] | [prev] | [next] | [standalone]


#87055

From"Diederik de Haas" <didi.debian@cknow.org>
Date2025-04-25 12:30 +0200
Message-ID<KFotH-g7Oa-1@gated-at.bofh.it>
In reply to#87043

[Multipart message — attachments visible in raw view] — view raw

On Fri Apr 25, 2025 at 11:10 AM CEST, Salvatore Bonaccorso wrote:
> Control: tags -1 + moreinfo
>
> On Mon, Jul 31, 2023 at 10:03:52AM +0200, Diederik de Haas wrote:
>> On Monday, 31 July 2023 04:28:31 CEST MOESSBAUER, Felix wrote:
>> > On Sun, 12 Feb 2023 18:52:07 +0100 Nicolas Frattaroli
>> > <frattaroli.nicolas@gmail.com> wrote:
>> > > I realise this is quite an old bug, but it would still be of interest
>> > > to me to get this enabled. The -@ option will increase the size of
>> > > the compiled device tree blobs somewhat, but on the flipside, 
>> > > u-boot-menu's device tree overlay functionality will actually be useful.
>> > 
>> > The increase is around 30% in average. For the linux-image-6.1.0-10-
>> > arm64, we have around 30MB of dtb files (uncompressed). By that, the
>> > package would grow by 10MB (uncompressed) or even less when compressed.
>> > 
>> > On many (if not most) arm based boards, these device tree overlays are
>> > very useful. There are also many kernel patches about enabling the
>> > symbol support on a per-device basis. However, kernel devs seem to have
>> > some mixed feelings about that. A proposed solution is to enable this
>> > at distro level [1].
>> > @Ben: Would it be an accepted solution to enable the DTC_FLAGS += -@
>> > for all armhf, armmp, arm64 and riscv devices?
>> 
>> AFAIUI, the upstream kernel devs were against such a blanket enablement.
>> (Consequently it sounds like a bad idea if Debian would do it)
>> 
>> https://lore.kernel.org/all/20220328000915.15041-1-ansuelsmth@gmail.com/ was a 
>> renewed attempt at restructuring the various .dts* into vendor directories and 
>> with 6.5 that got merged into the upstream kernel.
>> And (AFAIUI) that was seen as a prerequisite to start enabling the "-@" flag on 
>> a vendor/directory level, with optionally a subselection if it didn't make 
>> sense to enable it for all boards ... pretty much (exactly) as you proposed.
>> 
>> You mentioned you proposed it for RISC-V boards and some DDs proposed it for 
>> RPi boards and both got accepted. I think/expected that we would see (many?) 
>> more such patches getting accepted going forward now that the above mentioned 
>> dts directory restructuring has taken place.
>> 
>> IOW: I agree with the reasoning you brought up in your patch submission thread 
>> and keep targeting inclusion in the upstream kernel.
>
> Was there progress here, or what should we further do? If it's not
> something which is going to land upstream then I would suggest that we
> close the bug.

It seems there was _some_ progress:
me@pc:~/dev/kernel.org/linux$ grep -rn "DTC_FLAGS" arch/arm*/boot/dts/ | grep "\-@"

But AFA*I*CT, it has not been enabled for the Odroid C2 or for PINE64's
Quartz64 lineup of ARM devices, which were the explicit and targeted
requests that were made.

But I'm not an expert and I don't even fully understand how it
can/should work. 

``arch/arm64/boot/dts/rockchip/Makefile`` has this:
# This will make the <base>.dtb have symbols (like when DTC_FLAGS has -@ passed)
# and generate a new DTB (<overlay-application-test>.dtb) which is the
# result of the application of <overlay-1>.dtbo and other listed overlays on top
# of <base>.dtb.

which seems to imply that it can also be enabled in a different way
(then ``-@``) and IIRC it was 'inspired' by what some other subsystem
did, but I don't recall which that was.
So while the above mentioned ``grep`` statement had 87 results, it *may*
be available on more boards.

What the Debian kernel team does, is up to them.
This and my previous reply to this bug was to provide more info to the
best of *my* knowledge.

Cheers,
  Diederik

[toc] | [prev] | [standalone]


Back to top | Article view | linux.debian.kernel


csiph-web