RFE: Make ANDROID_BINDER_IPC tristate

RFE: Make ANDROID_BINDER_IPC tristate

[Diederik de Haas]

[-- Attachment #1: Type: text/plain, Size: 646 bytes --]

Hi,

Programs like Waydroid allows people to run Android in a Linux container
but it needs ANDROID_BINDER_IPC to do that.
It would be really helpful if ANDROID_BINDER_IPC could be dynamically
loaded when someone wants to use Waydroid (and unload when no longer
needed). But currently ANDROID_BINDER_IPC is a 'bool' so either you have
it at all times or you will never have it.

So hereby the request to make ANDROID_BINDER_IPC tristate so people that
want to use it can load it when needed.

Debian has a patch to make it tristate since 2018, but it would be so
much more useful to have it tristate upstream.

Cheers,
  Diederik

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

Re: RFE: Make ANDROID_BINDER_IPC tristate

[Greg Kroah-Hartman]

On Sat, Aug 16, 2025 at 02:11:24PM +0200, Diederik de Haas wrote:
> Hi,
> 
> Programs like Waydroid allows people to run Android in a Linux container
> but it needs ANDROID_BINDER_IPC to do that.
> It would be really helpful if ANDROID_BINDER_IPC could be dynamically
> loaded when someone wants to use Waydroid (and unload when no longer
> needed). But currently ANDROID_BINDER_IPC is a 'bool' so either you have
> it at all times or you will never have it.
> 
> So hereby the request to make ANDROID_BINDER_IPC tristate so people that
> want to use it can load it when needed.
> 
> Debian has a patch to make it tristate since 2018, but it would be so
> much more useful to have it tristate upstream.

Please submit a patch to do so!

Re: RFE: Make ANDROID_BINDER_IPC tristate

[Diederik de Haas]

[-- Attachment #1: Type: text/plain, Size: 634 bytes --]

On Sat Aug 16, 2025 at 2:47 PM CEST, Greg Kroah-Hartman wrote:
> On Sat, Aug 16, 2025 at 02:11:24PM +0200, Diederik de Haas wrote:
>> Programs like Waydroid allows people to run Android in a Linux container
>> but it needs ANDROID_BINDER_IPC to do that.
>> 
>> So hereby the request to make ANDROID_BINDER_IPC tristate so people that
>> want to use it can load it when needed.
>> 
>> Debian has a patch to make it tristate since 2018, but it would be so
>> much more useful to have it tristate upstream.
>
> Please submit a patch to do so!

I would have if I thought I had the knowledge/skills to do that,
but I don't.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

Re: RFE: Make ANDROID_BINDER_IPC tristate

[Diederik de Haas]

[-- Attachment #1: Type: text/plain, Size: 1591 bytes --]

On Sat Aug 16, 2025 at 6:37 PM CEST, Todd Kjos wrote:
> This is not feasible since binder relies on many kernel internal functions
> and data that cannot be exported for loadable module. Patches for this have
> been attempted in the past, but soundly rejected. You can see some of that
> discussion at
> https://lore.kernel.org/lkml/20180730143710.14413-1-christian@brauner.io/.

That seems to be the upstreaming attempt (or at least trying to start a
discussion about it) of the initial patch.
A lot could've happened in these 7 years, but apparently not enough for
a different outcome. Good to know, thanks :-)

FWIW these are the current ones:
https://salsa.debian.org/kernel-team/linux/-/blob/debian/6.16.1-1_exp1/debian/patches/debian/export-symbols-needed-by-binder.patch
https://salsa.debian.org/kernel-team/linux/-/blob/debian/6.16.1-1_exp1/debian/patches/debian/android-enable-building-binder-as-module.patch

Cheers,
  Diederik

> On Sat, Aug 16, 2025 at 5:52 AM Diederik de Haas <didi.debian@cknow.org>
> wrote:
>
>> On Sat Aug 16, 2025 at 2:47 PM CEST, Greg Kroah-Hartman wrote:
>> > On Sat, Aug 16, 2025 at 02:11:24PM +0200, Diederik de Haas wrote:
>> >> Programs like Waydroid allows people to run Android in a Linux container
>> >> but it needs ANDROID_BINDER_IPC to do that.
>> >>
>> >> So hereby the request to make ANDROID_BINDER_IPC tristate so people that
>> >> want to use it can load it when needed.
>> >>
>> >> Debian has a patch to make it tristate since 2018, but it would be so
>> >> much more useful to have it tristate upstream.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

Re: RFE: Make ANDROID_BINDER_IPC tristate

[Greg Kroah-Hartman]

On Sat, Aug 16, 2025 at 09:37:35AM -0700, Todd Kjos wrote:
> This is not feasible since binder relies on many kernel internal functions
> and data that cannot be exported for loadable module. Patches for this have
> been attempted in the past, but soundly rejected. You can see some of that
> discussion at
> https://lore.kernel.org/lkml/20180730143710.14413-1-christian@brauner.io/.

Then how is this working in the Debian kernel?

confused,

greg k-h

Re: RFE: Make ANDROID_BINDER_IPC tristate

[Greg Kroah-Hartman]

On Sat, Aug 16, 2025 at 07:26:16PM +0200, Diederik de Haas wrote:
> On Sat Aug 16, 2025 at 6:37 PM CEST, Todd Kjos wrote:
> > This is not feasible since binder relies on many kernel internal functions
> > and data that cannot be exported for loadable module. Patches for this have
> > been attempted in the past, but soundly rejected. You can see some of that
> > discussion at
> > https://lore.kernel.org/lkml/20180730143710.14413-1-christian@brauner.io/.
> 
> That seems to be the upstreaming attempt (or at least trying to start a
> discussion about it) of the initial patch.
> A lot could've happened in these 7 years, but apparently not enough for
> a different outcome. Good to know, thanks :-)
> 
> FWIW these are the current ones:
> https://salsa.debian.org/kernel-team/linux/-/blob/debian/6.16.1-1_exp1/debian/patches/debian/export-symbols-needed-by-binder.patch

Ah, that makes more sense, that patch is not going to be accepted
upsteam sorry.

greg k-h

Re: RFE: Make ANDROID_BINDER_IPC tristate

[Diederik de Haas]

[-- Attachment #1: Type: text/plain, Size: 1659 bytes --]

On Sat Aug 16, 2025 at 7:32 PM CEST, Greg Kroah-Hartman wrote:
> On Sat, Aug 16, 2025 at 07:26:16PM +0200, Diederik de Haas wrote:
>> On Sat Aug 16, 2025 at 6:37 PM CEST, Todd Kjos wrote:
>> > This is not feasible since binder relies on many kernel internal functions
>> > and data that cannot be exported for loadable module. Patches for this have
>> > been attempted in the past, but soundly rejected. You can see some of that
>> > discussion at
>> > https://lore.kernel.org/lkml/20180730143710.14413-1-christian@brauner.io/.
>> 
>> That seems to be the upstreaming attempt (or at least trying to start a
>> discussion about it) of the initial patch.
>> A lot could've happened in these 7 years, but apparently not enough for
>> a different outcome. Good to know, thanks :-)
>> 
>> FWIW these are the current ones:
>> https://salsa.debian.org/kernel-team/linux/-/blob/debian/6.16.1-1_exp1/debian/patches/debian/export-symbols-needed-by-binder.patch
>
> Ah, that makes more sense, that patch is not going to be accepted
> upsteam sorry.

I deliberately didn't link to those patches in my OP as I wanted to know
if it was possible/feasible at all (perhaps with some rework). I was
pretty sure if I had, the discussion would've only been about that
specific implementation.

The functionality (ie having the possibility to run Waydroid) seems
appealing (even ideal) to me, but (for myself) I didn't want to make it
built-in and having it available as a loadable module seems like a
great way to have both options in my (and other's) kernel.
And a functionality which should be available upstream if at all.

Cheers,
  Diederik

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]