[PATCH 0/2] i2c: core: prepare dropping support for I2C_COMPAT

[PATCH 0/2] i2c: core: prepare dropping support for I2C_COMPAT

[Heiner Kallweit]

I2C_COMPAT has been considered deprecated for 15 years now.
Therefore make it default n, before we remove support for it in the
near future. In addition warn any potential user of the old ABI.

Heiner Kallweit (2):
  i2c: core: Switch I2C_COMPAT to default n
  driver core: class: warn if a compatibility class is registered

 drivers/base/class.c | 3 +++
 drivers/i2c/Kconfig  | 4 ++--
 2 files changed, 5 insertions(+), 2 deletions(-)

-- 
2.46.0

[PATCH 1/2] i2c: core: Switch I2C_COMPAT to default n

[Heiner Kallweit]

I2C_COMPAT has been considered deprecated for 15 years now.
Therefore make it default n, before we remove support for it
in the near future.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
---
 drivers/i2c/Kconfig | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/i2c/Kconfig b/drivers/i2c/Kconfig
index 44710267d..e5721cebb 100644
--- a/drivers/i2c/Kconfig
+++ b/drivers/i2c/Kconfig
@@ -42,11 +42,11 @@ config I2C_BOARDINFO
 
 config I2C_COMPAT
 	bool "Enable compatibility bits for old user-space"
-	default y
+	default n
 	help
 	  Say Y here if you intend to run lm-sensors 3.1.1 or older, or any
 	  other user-space package which expects i2c adapters to be class
-	  devices. If you don't know, say Y.
+	  devices. If you don't know, say N.
 
 config I2C_CHARDEV
 	tristate "I2C device interface"
-- 
2.46.0

Re: [PATCH 1/2] i2c: core: Switch I2C_COMPAT to default n

[Greg Kroah-Hartman]

On Mon, Sep 02, 2024 at 09:01:18PM +0200, Heiner Kallweit wrote:
> I2C_COMPAT has been considered deprecated for 15 years now.
> Therefore make it default n, before we remove support for it
> in the near future.
> 
> Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
> ---
>  drivers/i2c/Kconfig | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/i2c/Kconfig b/drivers/i2c/Kconfig
> index 44710267d..e5721cebb 100644
> --- a/drivers/i2c/Kconfig
> +++ b/drivers/i2c/Kconfig
> @@ -42,11 +42,11 @@ config I2C_BOARDINFO
>  
>  config I2C_COMPAT
>  	bool "Enable compatibility bits for old user-space"
> -	default y
> +	default n

Just remove the default line, that way it will default to 'n'.

thanks,

greg k-h

[PATCH 2/2] driver core: class: warn if a compatibility class is registered

[Heiner Kallweit]

Kernel doc for this function states:
"Compatibility class are meant as a temporary user-space compatibility
workaround when converting a family of class devices to a bus devices."

Therefore remind any potential user of the old ABI that support for it
will be dropped soon.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
---
 drivers/base/class.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/base/class.c b/drivers/base/class.c
index 7b38fdf8e..f12a43736 100644
--- a/drivers/base/class.c
+++ b/drivers/base/class.c
@@ -556,6 +556,9 @@ struct class_compat *class_compat_register(const char *name)
 {
 	struct class_compat *cls;
 
+	pr_warn("Compatibility class %s will go away soon, please migrate userspace tools to use bus devices\n",
+		name);
+
 	cls = kmalloc(sizeof(struct class_compat), GFP_KERNEL);
 	if (!cls)
 		return NULL;
-- 
2.46.0

Re: [PATCH 2/2] driver core: class: warn if a compatibility class is registered

[Greg Kroah-Hartman]

On Mon, Sep 02, 2024 at 09:02:17PM +0200, Heiner Kallweit wrote:
> Kernel doc for this function states:
> "Compatibility class are meant as a temporary user-space compatibility
> workaround when converting a family of class devices to a bus devices."
> 
> Therefore remind any potential user of the old ABI that support for it
> will be dropped soon.
> 
> Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
> ---
>  drivers/base/class.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/drivers/base/class.c b/drivers/base/class.c
> index 7b38fdf8e..f12a43736 100644
> --- a/drivers/base/class.c
> +++ b/drivers/base/class.c
> @@ -556,6 +556,9 @@ struct class_compat *class_compat_register(const char *name)
>  {
>  	struct class_compat *cls;
>  
> +	pr_warn("Compatibility class %s will go away soon, please migrate userspace tools to use bus devices\n",
> +		name);

That's not going to do anything except annoy users who have no control
over this, sorry.  Please just fix up all of the kernel and then delete
this function.

thanks,

greg k-h

Re: [PATCH 2/2] driver core: class: warn if a compatibility class is registered

[Wolfram Sang]

[-- Attachment #1: Type: text/plain, Size: 778 bytes --]


> > +	pr_warn("Compatibility class %s will go away soon, please migrate userspace tools to use bus devices\n",
> > +		name);
> 
> That's not going to do anything except annoy users who have no control
> over this, sorry.  Please just fix up all of the kernel and then delete
> this function.

So, we deprecated this sysfs-class 15 years ago and hid it with a
Kconfig symbol. However, we never pursued this further, so e.g. Debian
has the Kconfig symbol still enabled. Can we really remove this from one
release to the next without another transition period? I am not afraid
of tools like lm-sensors which were converted long ago. But custom code
might rely on sysfs-paths created by this class. It was even advertised
in IPMI docs until last week (fixed now).


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

Re: [PATCH 2/2] driver core: class: warn if a compatibility class is registered

[Wolfram Sang]

[-- Attachment #1: Type: text/plain, Size: 1069 bytes --]

On Tue, Sep 03, 2024 at 11:04:44AM +0200, Wolfram Sang wrote:
> 
> > > +	pr_warn("Compatibility class %s will go away soon, please migrate userspace tools to use bus devices\n",
> > > +		name);
> > 
> > That's not going to do anything except annoy users who have no control
> > over this, sorry.  Please just fix up all of the kernel and then delete
> > this function.
> 
> So, we deprecated this sysfs-class 15 years ago and hid it with a
> Kconfig symbol. However, we never pursued this further, so e.g. Debian
> has the Kconfig symbol still enabled. Can we really remove this from one
> release to the next without another transition period? I am not afraid
> of tools like lm-sensors which were converted long ago. But custom code
> might rely on sysfs-paths created by this class. It was even advertised
> in IPMI docs until last week (fixed now).

I missed that Heiner was changing the driver core, not I2C core. So, to
give more details, I am talking about I2C_COMPAT and the "i2c-adapter"
class. The main question from above still stands.


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

Re: [PATCH 2/2] driver core: class: warn if a compatibility class is registered

[Greg Kroah-Hartman]

On Tue, Sep 03, 2024 at 11:06:44AM +0200, Wolfram Sang wrote:
> On Tue, Sep 03, 2024 at 11:04:44AM +0200, Wolfram Sang wrote:
> > 
> > > > +	pr_warn("Compatibility class %s will go away soon, please migrate userspace tools to use bus devices\n",
> > > > +		name);
> > > 
> > > That's not going to do anything except annoy users who have no control
> > > over this, sorry.  Please just fix up all of the kernel and then delete
> > > this function.
> > 
> > So, we deprecated this sysfs-class 15 years ago and hid it with a
> > Kconfig symbol. However, we never pursued this further, so e.g. Debian
> > has the Kconfig symbol still enabled. Can we really remove this from one
> > release to the next without another transition period? I am not afraid
> > of tools like lm-sensors which were converted long ago. But custom code
> > might rely on sysfs-paths created by this class. It was even advertised
> > in IPMI docs until last week (fixed now).
> 
> I missed that Heiner was changing the driver core, not I2C core. So, to
> give more details, I am talking about I2C_COMPAT and the "i2c-adapter"
> class. The main question from above still stands.

Delete the code and see if anyone notices?  :)

greg k-h

Re: [PATCH 2/2] driver core: class: warn if a compatibility class is registered

[Wolfram Sang]

[-- Attachment #1: Type: text/plain, Size: 327 bytes --]


> Delete the code and see if anyone notices?  :)

"Never ever break userspace, at least until Greg says so" :D

Seriously, Heiner initially sent a patch simply removing the code in
question [1]. May I interpret your above statement as "Acked-by"?

[1] https://lore.kernel.org/r/80c4a898-5867-4162-ac85-bdf7c7c68746@gmail.com


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

Re: [PATCH 2/2] driver core: class: warn if a compatibility class is registered

[Greg Kroah-Hartman]

On Tue, Sep 03, 2024 at 12:43:52PM +0200, Wolfram Sang wrote:
> 
> > Delete the code and see if anyone notices?  :)
> 
> "Never ever break userspace, at least until Greg says so" :D

"Never break userspace in a way that anyone notices" is the real rule we
have :)

> Seriously, Heiner initially sent a patch simply removing the code in
> question [1]. May I interpret your above statement as "Acked-by"?
> 
> [1] https://lore.kernel.org/r/80c4a898-5867-4162-ac85-bdf7c7c68746@gmail.com

Ack now sent there, thanks!

greg k-h

Re: [PATCH 2/2] driver core: class: warn if a compatibility class is registered

[Wolfram Sang]

[-- Attachment #1: Type: text/plain, Size: 385 bytes --]


> "Never break userspace in a way that anyone notices" is the real rule we
> have :)

Noted!

> > Seriously, Heiner initially sent a patch simply removing the code in
> > question [1]. May I interpret your above statement as "Acked-by"?
> > 
> > [1] https://lore.kernel.org/r/80c4a898-5867-4162-ac85-bdf7c7c68746@gmail.com
> 
> Ack now sent there, thanks!

Thank you!


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