[v1,0/2] w1: add UART w1 bus driver

  Hello!

Krzysztof, thank your very much for your feedback!

This patch contains a driver for a 1-Wire bus over UART. The driver
utilizes the UART interface via the Serial Device Bus to create the
1-Wire timing patterns.

Version 1

- In v1, the driver requests a baud-rate (9600 for reset and 115200 for
write/read) and tries to adapt the transmitted byte according to the
actual baud-rate returned from serdev. Is this the correct direction or
should the baud-rate be specified in the device-tree? Alternatively,
it could make sense to specify the minimum and maximum times for the
1-Wire operations in the device-tree, instead of using hard-coded ones
similar as in "Figure 11. Configuration tab" of the linked document
"Using UART to Implement a 1-Wire Bus Master".

- In addition, the received byte is now protected with a mutex - instead
of the atomic, which I used before due to the concurrent store and load.

- Receiving more than one byte results in an error, since the w1-uart
driver is the only writer, it writes a single-byte and should receive
a single byte.

Changes:
- support different baud-rates
- fix variable names, errno-returns, wrong define CONFIG_OF
- fix log flooding
- fix locking problem for serdev-receive and w1-master reset/touch
- fix driver remove (error-path for rxtx-function)
- add documentation for dt-binding


It was tested on a "Raspberry Pi 3 Model B+" with a DS18B20 and on a
"Variscite DART-6UL" with a DS18S20 temperature sensor.

Content:
- Patch 1: device tree binding
- Patch 2: driver and documentation

The patch was created against the w1 subsytem tree (branch w1-next):
  Link: https://git.kernel.org/pub/scm/linux/kernel/git/krzk/linux-w1.git/

The checkpatch.pl script reported the following error - which I am not
sure how to fix:
  WARNING: added, moved or deleted file(s), does MAINTAINERS need
  updating?

The technical details for 1-Wire over UART are in the document:
  Link: https://www.analog.com/en/technical-articles/using-a-uart-to-implement-a-1wire-bus-master.html

  In short, the UART peripheral must support full-duplex and operate in
open-drain mode. The timing patterns are generated by a specific
combination of baud-rate and transmitted byte, which corresponds to a
1-Wire read bit, write bit or reset pulse.

For instance the timing pattern for a 1-Wire reset and presence detect
uses the baud-rate 9600, i.e. 104.2 us per bit. The transmitted byte
0xf0 over UART (least significant bit first, start-bit low) sets the
reset low time for 1-Wire to 521 us. A present 1-Wire device changes the
received byte by pulling the line low, which is used by the driver to
evaluate the result of the 1-Wire operation.

Similar for a 1-Wire read bit or write bit, which uses the baud-rate
115200, i.e. 8.7 us per bit. The transmitted byte 0x00 is used for a
Write-0 operation and the byte 0xff for Read-0, Read-1 and Write-1.

Hope the driver is helpful.

Thanks,
Christoph

Christoph Winklhofer (2):
  dt-bindings: w1: UART 1-wire bus
  w1: add UART w1 bus driver

 .../devicetree/bindings/w1/w1-uart.yaml       |  44 +++
 Documentation/w1/masters/index.rst            |   1 +
 Documentation/w1/masters/w1-uart.rst          |  53 +++
 drivers/w1/masters/Kconfig                    |  10 +
 drivers/w1/masters/Makefile                   |   1 +
 drivers/w1/masters/w1-uart.c                  | 307 ++++++++++++++++++
 6 files changed, 416 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/w1/w1-uart.yaml
 create mode 100644 Documentation/w1/masters/w1-uart.rst
 create mode 100644 drivers/w1/masters/w1-uart.c


base-commit: efc19c44aa442197ddcbb157c6ca54a56eba8c4e
  

On 21/12/2023 07:50, Christoph Winklhofer wrote:
> Hello!
> 
> Krzysztof, thank your very much for your feedback!
> 
> This patch contains a driver for a 1-Wire bus over UART. The driver
> utilizes the UART interface via the Serial Device Bus to create the
> 1-Wire timing patterns.
> 
> Version 1
> 

You already sent v1, so this is v2:

b4 diff '<20231221065049.30703-1-cj.winklhofer@gmail.com>'
Grabbing thread from
lore.kernel.org/all/20231221065049.30703-1-cj.winklhofer@gmail.com/t.mbox.gz
---
Analyzing 4 messages in the thread
ERROR: Could not auto-find previous revision
       Run "b4 am -T" manually, then "b4 diff -m mbx1 mbx2"

I still cannot find the changelog. Does it mean nothing improved?


> - In v1, the driver requests a baud-rate (9600 for reset and 115200 for
> write/read) and tries to adapt the transmitted byte according to the
> actual baud-rate returned from serdev. Is this the correct direction or
> should the baud-rate be specified in the device-tree? Alternatively,
> it could make sense to specify the minimum and maximum times for the
> 1-Wire operations in the device-tree, instead of using hard-coded ones
> similar as in "Figure 11. Configuration tab" of the linked document
> "Using UART to Implement a 1-Wire Bus Master".

Depends, are these hardware properties? Are these runtime? What do they
depend on?

> 
> - In addition, the received byte is now protected with a mutex - instead
> of the atomic, which I used before due to the concurrent store and load.
> 
> - Receiving more than one byte results in an error, since the w1-uart
> driver is the only writer, it writes a single-byte and should receive
> a single byte.
> 
> Changes:
> - support different baud-rates
> - fix variable names, errno-returns, wrong define CONFIG_OF
> - fix log flooding
> - fix locking problem for serdev-receive and w1-master reset/touch
> - fix driver remove (error-path for rxtx-function)
> - add documentation for dt-binding

So this looks like changelog. Please make it explicit - move it to the
beginning of cover letter and say "changes in v2".


Best regards,
Krzysztof
  

On Thu, Dec 21, 2023 at 10:13:01PM +0100, Krzysztof Kozlowski wrote:
> On 21/12/2023 07:50, Christoph Winklhofer wrote:
> > Hello!
> > 
> > Krzysztof, thank your very much for your feedback!
> > 
> > This patch contains a driver for a 1-Wire bus over UART. The driver
> > utilizes the UART interface via the Serial Device Bus to create the
> > 1-Wire timing patterns.
> > 
> > Version 1
> > 
> 
> You already sent v1, so this is v2:
> 
> b4 diff '<20231221065049.30703-1-cj.winklhofer@gmail.com>'
> Grabbing thread from
> lore.kernel.org/all/20231221065049.30703-1-cj.winklhofer@gmail.com/t.mbox.gz
> ---
> Analyzing 4 messages in the thread
> ERROR: Could not auto-find previous revision
>        Run "b4 am -T" manually, then "b4 diff -m mbx1 mbx2"
> 
> I still cannot find the changelog. Does it mean nothing improved?
> 
> 

Sorry, I will fix the patch and resend it.

> > - In v1, the driver requests a baud-rate (9600 for reset and 115200 for
> > write/read) and tries to adapt the transmitted byte according to the
> > actual baud-rate returned from serdev. Is this the correct direction or
> > should the baud-rate be specified in the device-tree? Alternatively,
> > it could make sense to specify the minimum and maximum times for the
> > 1-Wire operations in the device-tree, instead of using hard-coded ones
> > similar as in "Figure 11. Configuration tab" of the linked document
> > "Using UART to Implement a 1-Wire Bus Master".
> 
> Depends, are these hardware properties? Are these runtime? What do they
> depend on?
> 

Ok, the timing constraints came from the 1-Wire protocol, so DT makes no
sense. Probably it would be nice to tweak them for different 1-Wire slaves
via parameter to the driver - however, I will left them hardcoded for now.

Thanks!
Christoph