1. Patchew
  2. linux
  3. [PATCH v7 0/6] imu: st_lsm6dsx: Add support for rotation sensor
  4. View patch

[PATCH v7 4/6] iio: ABI: Add support for floating-point numbers in buffer scan elements

Francesco Lavra posted 6 patches 1 month, 1 week ago
There is a newer version of this series
[PATCH v7 4/6] iio: ABI: Add support for floating-point numbers in buffer scan elements
Posted by Francesco Lavra 1 month, 1 week ago 
In the data storage description of a scan element, the first character
after the colon can have the values 's' and 'u' to specify signed and
unsigned integers, respectively.
Add 'f' as an allowed value to specify floating-point numbers formatted
according to the IEEE 754 standard.

Signed-off-by: Francesco Lavra <flavra@baylibre.com>
---
 Documentation/ABI/testing/sysfs-bus-iio  | 33 +++++++++++++-----------
 Documentation/driver-api/iio/buffers.rst |  3 ++-
 Documentation/iio/iio_devbuf.rst         |  3 ++-
 3 files changed, 22 insertions(+), 17 deletions(-)

diff --git a/Documentation/ABI/testing/sysfs-bus-iio b/Documentation/ABI/testing/sysfs-bus-iio
index 5f87dcee78f7..bd6c3305dd2b 100644
--- a/Documentation/ABI/testing/sysfs-bus-iio
+++ b/Documentation/ABI/testing/sysfs-bus-iio
@@ -1510,21 +1510,24 @@ Contact:	linux-iio@vger.kernel.org
 Description:
 		Description of the scan element data storage within the buffer
 		and hence the form in which it is read from user-space.
-		Form is [be|le]:[s|u]bits/storagebits[>>shift].
-		be or le specifies big or little endian. s or u specifies if
-		signed (2's complement) or unsigned. bits is the number of bits
-		of data and storagebits is the space (after padding) that it
-		occupies in the buffer. shift if specified, is the shift that
-		needs to be applied prior to masking out unused bits. Some
-		devices put their data in the middle of the transferred elements
-		with additional information on both sides.  Note that some
-		devices will have additional information in the unused bits
-		so to get a clean value, the bits value must be used to mask
-		the buffer output value appropriately.  The storagebits value
-		also specifies the data alignment.  So s48/64>>2 will be a
-		signed 48 bit integer stored in a 64 bit location aligned to
-		a 64 bit boundary. To obtain the clean value, shift right 2
-		and apply a mask to zero the top 16 bits of the result.
+		Form is [be|le]:[f|s|u]bits/storagebits[>>shift].
+		be or le specifies big or little endian. f means floating-point
+		(IEEE 754 binary format), s means signed (2's complement), u means
+		unsigned. bits is the number of bits of data and storagebits is the
+		space (after padding) that it occupies in the buffer; when using a
+		floating-point format, bits must be one of the width values defined
+		in the IEEE 754 standard for binary interchange formats (e.g. 16
+		indicates the binary16 format for half-precision numbers). shift,
+		if specified, is the shift that needs to be applied prior to
+		masking out unused bits. Some devices put their data in the middle
+		of the transferred elements with additional information on both
+		sides. Note that some devices will have additional information in
+		the unused bits, so to get a clean value the bits value must be
+		used to mask the buffer output value appropriately. The storagebits
+		value also specifies the data alignment. So s48/64>>2 will be a
+		signed 48 bit integer stored in a 64 bit location aligned to a 64
+		bit boundary. To obtain the clean value, shift right 2 and apply a
+		mask to zero the top 16 bits of the result.
 		For other storage combinations this attribute will be extended
 		appropriately.
 
diff --git a/Documentation/driver-api/iio/buffers.rst b/Documentation/driver-api/iio/buffers.rst
index f36e6d00173f..2fc9c2951a9d 100644
--- a/Documentation/driver-api/iio/buffers.rst
+++ b/Documentation/driver-api/iio/buffers.rst
@@ -37,9 +37,10 @@ directory contains attributes of the following form:
 * :file:`index`, the scan_index of the channel.
 * :file:`type`, description of the scan element data storage within the buffer
   and hence the form in which it is read from user space.
-  Format is [be|le]:[s|u]bits/storagebits[Xrepeat][>>shift] .
+  Format is [be|le]:[f|s|u]bits/storagebits[Xrepeat][>>shift] .
 
   * *be* or *le*, specifies big or little endian.
+  * *f*, specifies if floating-point.
   * *s* or *u*, specifies if signed (2's complement) or unsigned.
   * *bits*, is the number of valid data bits.
   * *storagebits*, is the number of bits (after padding) that it occupies in the
diff --git a/Documentation/iio/iio_devbuf.rst b/Documentation/iio/iio_devbuf.rst
index dca1f0200b0d..e91730fa3cea 100644
--- a/Documentation/iio/iio_devbuf.rst
+++ b/Documentation/iio/iio_devbuf.rst
@@ -83,9 +83,10 @@ and the relevant _type attributes to establish the data storage format.
 
 Read-only attribute containing the description of the scan element data storage
 within the buffer and hence the form in which it is read from userspace. Format
-is [be|le]:[s|u]bits/storagebits[Xrepeat][>>shift], where:
+is [be|le]:[f|s|u]bits/storagebits[Xrepeat][>>shift], where:
 
 - **be** or **le** specifies big or little-endian.
+- **f** specifies if floating-point.
 - **s** or **u** specifies if signed (2's complement) or unsigned.
 - **bits** is the number of valid data bits.
 - **storagebits** is the number of bits (after padding) that it occupies in the
-- 
2.39.5
Re: [PATCH v7 4/6] iio: ABI: Add support for floating-point numbers in buffer scan elements
Posted by David Lechner 1 month, 1 week ago 
On 3/4/26 2:06 AM, Francesco Lavra wrote:
> In the data storage description of a scan element, the first character
> after the colon can have the values 's' and 'u' to specify signed and
> unsigned integers, respectively.
> Add 'f' as an allowed value to specify floating-point numbers formatted
> according to the IEEE 754 standard.
> 
...

> diff --git a/Documentation/driver-api/iio/buffers.rst b/Documentation/driver-api/iio/buffers.rst
> index f36e6d00173f..2fc9c2951a9d 100644
> --- a/Documentation/driver-api/iio/buffers.rst
> +++ b/Documentation/driver-api/iio/buffers.rst
> @@ -37,9 +37,10 @@ directory contains attributes of the following form:
>  * :file:`index`, the scan_index of the channel.
>  * :file:`type`, description of the scan element data storage within the buffer
>    and hence the form in which it is read from user space.
> -  Format is [be|le]:[s|u]bits/storagebits[Xrepeat][>>shift] .
> +  Format is [be|le]:[f|s|u]bits/storagebits[Xrepeat][>>shift] .
>  
>    * *be* or *le*, specifies big or little endian.
> +  * *f*, specifies if floating-point.
>    * *s* or *u*, specifies if signed (2's complement) or unsigned.

I would keep all of the format options on one bullet point.

>    * *bits*, is the number of valid data bits.
>    * *storagebits*, is the number of bits (after padding) that it occupies in the
> diff --git a/Documentation/iio/iio_devbuf.rst b/Documentation/iio/iio_devbuf.rst
> index dca1f0200b0d..e91730fa3cea 100644
> --- a/Documentation/iio/iio_devbuf.rst
> +++ b/Documentation/iio/iio_devbuf.rst
> @@ -83,9 +83,10 @@ and the relevant _type attributes to establish the data storage format.
>  
>  Read-only attribute containing the description of the scan element data storage
>  within the buffer and hence the form in which it is read from userspace. Format
> -is [be|le]:[s|u]bits/storagebits[Xrepeat][>>shift], where:
> +is [be|le]:[f|s|u]bits/storagebits[Xrepeat][>>shift], where:
>  
>  - **be** or **le** specifies big or little-endian.
> +- **f** specifies if floating-point.
>  - **s** or **u** specifies if signed (2's complement) or unsigned.

same here

>  - **bits** is the number of valid data bits.
>  - **storagebits** is the number of bits (after padding) that it occupies in the
Re: [PATCH v7 4/6] iio: ABI: Add support for floating-point numbers in buffer scan elements
Posted by Francesco Lavra 1 month, 1 week ago 
On Wed, 2026-03-04 at 16:45 -0600, David Lechner wrote:
> On 3/4/26 2:06 AM, Francesco Lavra wrote:
> > In the data storage description of a scan element, the first character
> > after the colon can have the values 's' and 'u' to specify signed and
> > unsigned integers, respectively.
> > Add 'f' as an allowed value to specify floating-point numbers formatted
> > according to the IEEE 754 standard.
> > 
> ...
> 
> > diff --git a/Documentation/driver-api/iio/buffers.rst
> > b/Documentation/driver-api/iio/buffers.rst
> > index f36e6d00173f..2fc9c2951a9d 100644
> > --- a/Documentation/driver-api/iio/buffers.rst
> > +++ b/Documentation/driver-api/iio/buffers.rst
> > @@ -37,9 +37,10 @@ directory contains attributes of the following form:
> >  * :file:`index`, the scan_index of the channel.
> >  * :file:`type`, description of the scan element data storage within
> > the buffer
> >    and hence the form in which it is read from user space.
> > -  Format is [be|le]:[s|u]bits/storagebits[Xrepeat][>>shift] .
> > +  Format is [be|le]:[f|s|u]bits/storagebits[Xrepeat][>>shift] .
> >  
> >    * *be* or *le*, specifies big or little endian.
> > +  * *f*, specifies if floating-point.
> >    * *s* or *u*, specifies if signed (2's complement) or unsigned.
> 
> I would keep all of the format options on one bullet point.

That's what I did initially, but Andy suggested doing differently [1].


> >    * *bits*, is the number of valid data bits.
> >    * *storagebits*, is the number of bits (after padding) that it
> > occupies in the
> > diff --git a/Documentation/iio/iio_devbuf.rst
> > b/Documentation/iio/iio_devbuf.rst
> > index dca1f0200b0d..e91730fa3cea 100644
> > --- a/Documentation/iio/iio_devbuf.rst
> > +++ b/Documentation/iio/iio_devbuf.rst
> > @@ -83,9 +83,10 @@ and the relevant _type attributes to establish the
> > data storage format.
> >  
> >  Read-only attribute containing the description of the scan element
> > data storage
> >  within the buffer and hence the form in which it is read from
> > userspace. Format
> > -is [be|le]:[s|u]bits/storagebits[Xrepeat][>>shift], where:
> > +is [be|le]:[f|s|u]bits/storagebits[Xrepeat][>>shift], where:
> >  
> >  - **be** or **le** specifies big or little-endian.
> > +- **f** specifies if floating-point.
> >  - **s** or **u** specifies if signed (2's complement) or unsigned.
> 
> same here

[1] https://lore.kernel.org/linux-iio/aZ7dCdLs5xcJ4UGW@smile.fi.intel.com/
>
Re: [PATCH v7 4/6] iio: ABI: Add support for floating-point numbers in buffer scan elements
Posted by Andy Shevchenko 1 month, 1 week ago 
On Thu, Mar 5, 2026 at 11:09 AM Francesco Lavra <flavra@baylibre.com> wrote:
> On Wed, 2026-03-04 at 16:45 -0600, David Lechner wrote:
> > On 3/4/26 2:06 AM, Francesco Lavra wrote:
> > > In the data storage description of a scan element, the first character
> > > after the colon can have the values 's' and 'u' to specify signed and
> > > unsigned integers, respectively.
> > > Add 'f' as an allowed value to specify floating-point numbers formatted
> > > according to the IEEE 754 standard.

...

> > > -  Format is [be|le]:[s|u]bits/storagebits[Xrepeat][>>shift] .
> > > +  Format is [be|le]:[f|s|u]bits/storagebits[Xrepeat][>>shift] .
> > >
> > >    * *be* or *le*, specifies big or little endian.
> > > +  * *f*, specifies if floating-point.
> > >    * *s* or *u*, specifies if signed (2's complement) or unsigned.
> >
> > I would keep all of the format options on one bullet point.
>
> That's what I did initially, but Andy suggested doing differently [1].

And still I think it's better to not mix them. The floating in the
same sentence is confusing (along with 2's complement mention and
sign).

...

> > > -is [be|le]:[s|u]bits/storagebits[Xrepeat][>>shift], where:
> > > +is [be|le]:[f|s|u]bits/storagebits[Xrepeat][>>shift], where:
> > >
> > >  - **be** or **le** specifies big or little-endian.
> > > +- **f** specifies if floating-point.
> > >  - **s** or **u** specifies if signed (2's complement) or unsigned.
> >
> > same here
>
> [1] https://lore.kernel.org/linux-iio/aZ7dCdLs5xcJ4UGW@smile.fi.intel.com/

Same here.

-- 
With Best Regards,
Andy Shevchenko
Re: [PATCH v7 4/6] iio: ABI: Add support for floating-point numbers in buffer scan elements
Posted by David Lechner 1 month, 1 week ago 
On 3/5/26 3:23 AM, Andy Shevchenko wrote:
> On Thu, Mar 5, 2026 at 11:09 AM Francesco Lavra <flavra@baylibre.com> wrote:
>> On Wed, 2026-03-04 at 16:45 -0600, David Lechner wrote:
>>> On 3/4/26 2:06 AM, Francesco Lavra wrote:
>>>> In the data storage description of a scan element, the first character
>>>> after the colon can have the values 's' and 'u' to specify signed and
>>>> unsigned integers, respectively.
>>>> Add 'f' as an allowed value to specify floating-point numbers formatted
>>>> according to the IEEE 754 standard.
> 
> ...
> 
>>>> -  Format is [be|le]:[s|u]bits/storagebits[Xrepeat][>>shift] .
>>>> +  Format is [be|le]:[f|s|u]bits/storagebits[Xrepeat][>>shift] .
>>>>
>>>>    * *be* or *le*, specifies big or little endian.
>>>> +  * *f*, specifies if floating-point.
>>>>    * *s* or *u*, specifies if signed (2's complement) or unsigned.
>>>
>>> I would keep all of the format options on one bullet point.
>>
>> That's what I did initially, but Andy suggested doing differently [1].
> 
> And still I think it's better to not mix them. The floating in the
> same sentence is confusing (along with 2's complement mention and
> sign).

Then I would split up all 3. It is strange to mix some and not
all.

> 
> ...
> 
>>>> -is [be|le]:[s|u]bits/storagebits[Xrepeat][>>shift], where:
>>>> +is [be|le]:[f|s|u]bits/storagebits[Xrepeat][>>shift], where:
>>>>
>>>>  - **be** or **le** specifies big or little-endian.
>>>> +- **f** specifies if floating-point.
>>>>  - **s** or **u** specifies if signed (2's complement) or unsigned.
>>>
>>> same here
>>
>> [1] https://lore.kernel.org/linux-iio/aZ7dCdLs5xcJ4UGW@smile.fi.intel.com/
> 
> Same here.
>
Re: [PATCH v7 4/6] iio: ABI: Add support for floating-point numbers in buffer scan elements
Posted by Andy Shevchenko 1 month, 1 week ago 
On Thu, Mar 05, 2026 at 08:37:48AM -0600, David Lechner wrote:
> On 3/5/26 3:23 AM, Andy Shevchenko wrote:
> > On Thu, Mar 5, 2026 at 11:09 AM Francesco Lavra <flavra@baylibre.com> wrote:
> >> On Wed, 2026-03-04 at 16:45 -0600, David Lechner wrote:
> >>> On 3/4/26 2:06 AM, Francesco Lavra wrote:
> >>>> In the data storage description of a scan element, the first character
> >>>> after the colon can have the values 's' and 'u' to specify signed and
> >>>> unsigned integers, respectively.
> >>>> Add 'f' as an allowed value to specify floating-point numbers formatted
> >>>> according to the IEEE 754 standard.

...

> >>>> -  Format is [be|le]:[s|u]bits/storagebits[Xrepeat][>>shift] .
> >>>> +  Format is [be|le]:[f|s|u]bits/storagebits[Xrepeat][>>shift] .
> >>>>
> >>>>    * *be* or *le*, specifies big or little endian.
> >>>> +  * *f*, specifies if floating-point.
> >>>>    * *s* or *u*, specifies if signed (2's complement) or unsigned.
> >>>
> >>> I would keep all of the format options on one bullet point.
> >>
> >> That's what I did initially, but Andy suggested doing differently [1].
> > 
> > And still I think it's better to not mix them. The floating in the
> > same sentence is confusing (along with 2's complement mention and
> > sign).
> 
> Then I would split up all 3. It is strange to mix some and not
> all.

I don't find it 'strange'. The integer are grouped together, floats do not
belong to that group.

...

> >>>> -is [be|le]:[s|u]bits/storagebits[Xrepeat][>>shift], where:
> >>>> +is [be|le]:[f|s|u]bits/storagebits[Xrepeat][>>shift], where:
> >>>>
> >>>>  - **be** or **le** specifies big or little-endian.
> >>>> +- **f** specifies if floating-point.
> >>>>  - **s** or **u** specifies if signed (2's complement) or unsigned.
> >>>
> >>> same here
> >>
> >> [1] https://lore.kernel.org/linux-iio/aZ7dCdLs5xcJ4UGW@smile.fi.intel.com/
> > 
> > Same here.

-- 
With Best Regards,
Andy Shevchenko
Re: [PATCH v7 4/6] iio: ABI: Add support for floating-point numbers in buffer scan elements
Posted by Jonathan Cameron 1 month, 1 week ago 
On Fri, 6 Mar 2026 14:09:51 +0200
Andy Shevchenko <andriy.shevchenko@intel.com> wrote:

> On Thu, Mar 05, 2026 at 08:37:48AM -0600, David Lechner wrote:
> > On 3/5/26 3:23 AM, Andy Shevchenko wrote:  
> > > On Thu, Mar 5, 2026 at 11:09 AM Francesco Lavra <flavra@baylibre.com> wrote:  
> > >> On Wed, 2026-03-04 at 16:45 -0600, David Lechner wrote:  
> > >>> On 3/4/26 2:06 AM, Francesco Lavra wrote:  
> > >>>> In the data storage description of a scan element, the first character
> > >>>> after the colon can have the values 's' and 'u' to specify signed and
> > >>>> unsigned integers, respectively.
> > >>>> Add 'f' as an allowed value to specify floating-point numbers formatted
> > >>>> according to the IEEE 754 standard.  
> 
> ...
> 
> > >>>> -  Format is [be|le]:[s|u]bits/storagebits[Xrepeat][>>shift] .
> > >>>> +  Format is [be|le]:[f|s|u]bits/storagebits[Xrepeat][>>shift] .
> > >>>>
> > >>>>    * *be* or *le*, specifies big or little endian.
> > >>>> +  * *f*, specifies if floating-point.
> > >>>>    * *s* or *u*, specifies if signed (2's complement) or unsigned.  
> > >>>
> > >>> I would keep all of the format options on one bullet point.  
> > >>
> > >> That's what I did initially, but Andy suggested doing differently [1].  
> > > 
> > > And still I think it's better to not mix them. The floating in the
> > > same sentence is confusing (along with 2's complement mention and
> > > sign).  
> > 
> > Then I would split up all 3. It is strange to mix some and not
> > all.  
> 
> I don't find it 'strange'. The integer are grouped together, floats do not
> belong to that group.
Maybe two paragaraphs in one bullet point?
	* *f*, specifies if floating-point.
	  *s* or *u*, specifies if signed (2's complement) or unsigned.  

Though then we'll definitely need to check it didn't break the formatting
in the docs generated from these files.

For me any of the above are fine.

	
> 
> ...
> 
> > >>>> -is [be|le]:[s|u]bits/storagebits[Xrepeat][>>shift], where:
> > >>>> +is [be|le]:[f|s|u]bits/storagebits[Xrepeat][>>shift], where:
> > >>>>
> > >>>>  - **be** or **le** specifies big or little-endian.
> > >>>> +- **f** specifies if floating-point.
> > >>>>  - **s** or **u** specifies if signed (2's complement) or unsigned.  
> > >>>
> > >>> same here  
> > >>
> > >> [1] https://lore.kernel.org/linux-iio/aZ7dCdLs5xcJ4UGW@smile.fi.intel.com/  
> > > 
> > > Same here.  
>
Re: [PATCH v7 4/6] iio: ABI: Add support for floating-point numbers in buffer scan elements
Posted by Francesco Lavra 3 weeks, 6 days ago 
On Sat, 2026-03-07 at 12:51 +0000, Jonathan Cameron wrote:
> On Fri, 6 Mar 2026 14:09:51 +0200
> Andy Shevchenko <andriy.shevchenko@intel.com> wrote:
> 
> > On Thu, Mar 05, 2026 at 08:37:48AM -0600, David Lechner wrote:
> > > On 3/5/26 3:23 AM, Andy Shevchenko wrote:  
> > > > On Thu, Mar 5, 2026 at 11:09 AM Francesco Lavra
> > > > <flavra@baylibre.com> wrote:  
> > > > > On Wed, 2026-03-04 at 16:45 -0600, David Lechner wrote:  
> > > > > > On 3/4/26 2:06 AM, Francesco Lavra wrote:  
> > > > > > > In the data storage description of a scan element, the first
> > > > > > > character
> > > > > > > after the colon can have the values 's' and 'u' to specify
> > > > > > > signed and
> > > > > > > unsigned integers, respectively.
> > > > > > > Add 'f' as an allowed value to specify floating-point numbers
> > > > > > > formatted
> > > > > > > according to the IEEE 754 standard.  
> > 
> > ...
> > 
> > > > > > > -  Format is [be|le]:[s|u]bits/storagebits[Xrepeat][>>shift]
> > > > > > > .
> > > > > > > +  Format is
> > > > > > > [be|le]:[f|s|u]bits/storagebits[Xrepeat][>>shift] .
> > > > > > > 
> > > > > > >    * *be* or *le*, specifies big or little endian.
> > > > > > > +  * *f*, specifies if floating-point.
> > > > > > >    * *s* or *u*, specifies if signed (2's complement) or
> > > > > > > unsigned.  
> > > > > > 
> > > > > > I would keep all of the format options on one bullet point.  
> > > > > 
> > > > > That's what I did initially, but Andy suggested doing differently
> > > > > [1].  
> > > > 
> > > > And still I think it's better to not mix them. The floating in the
> > > > same sentence is confusing (along with 2's complement mention and
> > > > sign).  
> > > 
> > > Then I would split up all 3. It is strange to mix some and not
> > > all.  
> > 
> > I don't find it 'strange'. The integer are grouped together, floats do
> > not
> > belong to that group.
> Maybe two paragaraphs in one bullet point?
>         * *f*, specifies if floating-point.
>           *s* or *u*, specifies if signed (2's complement) or unsigned.  
> 
> Though then we'll definitely need to check it didn't break the formatting
> in the docs generated from these files.

This does break the formatting in the HTML docs, where the two paragraphs
end up in the same line.

> For me any of the above are fine.

I will keep it as is.

> > ...
> > 
> > > > > > > -is [be|le]:[s|u]bits/storagebits[Xrepeat][>>shift], where:
> > > > > > > +is [be|le]:[f|s|u]bits/storagebits[Xrepeat][>>shift], where:
> > > > > > > 
> > > > > > >  - **be** or **le** specifies big or little-endian.
> > > > > > > +- **f** specifies if floating-point.
> > > > > > >  - **s** or **u** specifies if signed (2's complement) or
> > > > > > > unsigned.  
> > > > > > 
> > > > > > same here  
> > > > > 
> > > > > [1]
> > > > > https://lore.kernel.org/linux-iio/aZ7dCdLs5xcJ4UGW@smile.fi.intel.com/
> > > > >  
> > > > 
> > > > Same here.

Same here.

Patchew 2.1-devel-ea86937

© 2016 - 2026 Red Hat, Inc.