Remap, SPI and parity bit

Remap, SPI and parity bit

[christopher tyerman]

hello

I'm not sure if this is the correct place to be asking this, if there is 
a better place please redirect me.

I'm currently working on a kernel driver for the Texas Instruments 
tps92518-Q1 and tps92518HV-Q1.

This device uses 16-bit SPI Frames (7-bit control, 9-bit data).

is "Command Frame"

|1-bit command bit 5-bit Address 1-bit parity 9-bit data|

requires a parity bit

|PARITY = XNOR(CMD, A4..A0, D8..D0)|

I'm trying to use regmaps where possible and I can define a regmap, e.g.

|static struct regmap_config tps92518_regmap_config = { .reg_bits = 6, 
.pad_bits = 1, // parity bit .val_bits = 9, .max_register = 
tps92518_RESET, }; |

however so far i cannot currently see how to automatically calculate it 
without replacing large chunks of code

e.g replacing chunks of /drivers/base/regmap/regmap-spi.c with a new 
regmap_bus

some where i need to be able to calculate and insert a parity bit

is there anywhere i can setup a callback that preprocesses the spi 
frames before sending?


I'm working up the chain an i can see a potential patch to regmaps that 
would  fix this if nothing better already exists

with regmap_config exists /@read_flag_mask/ and /@write_flag_mask/, 
along the same line there could be added a "@read_parity_mask" and 
"@write_parity_mask"
and an optional "@reg_parity" callback that calculates the parity 
from/reg/ and /val/ similar/@reg_read/ and /@reg_write/ but without 
replacing the entire read/write operations

I may even be able to write this but that is outside the spi-subsystem, 
and i'm currently not sure what list regmaps fall under



i apologise if this is the wrong list and, thankyou for any responses

kind regards

christopher tyerman

Re: Remap, SPI and parity bit

[Andy Shevchenko]

On Sun, Nov 14, 2021 at 8:41 PM christopher tyerman
<linuxkernel@fbautosys.co.uk> wrote:

> I'm not sure if this is the correct place to be asking this, if there is
> a better place please redirect me.

I guess it's induced by your (?) posts on SO?
Anyway, it's the correct place to ask, yes. I Cc'ed this to regulator,
LED, and MFD subsystem maintainers to gather their opinions.

> I'm currently working on a kernel driver for the Texas Instruments
> tps92518-Q1 and tps92518HV-Q1.
>
> This device uses 16-bit SPI Frames (7-bit control, 9-bit data).
>
> is "Command Frame"
>
> |1-bit command bit 5-bit Address 1-bit parity 9-bit data|
>
> requires a parity bit
>
> |PARITY = XNOR(CMD, A4..A0, D8..D0)|
>
> I'm trying to use regmaps where possible and I can define a regmap, e.g.
>
> |static struct regmap_config tps92518_regmap_config = { .reg_bits = 6,
> .pad_bits = 1, // parity bit .val_bits = 9, .max_register =
> tps92518_RESET, }; |
>
> however so far i cannot currently see how to automatically calculate it
> without replacing large chunks of code
>
> e.g replacing chunks of /drivers/base/regmap/regmap-spi.c with a new
> regmap_bus
>
> some where i need to be able to calculate and insert a parity bit
>
> is there anywhere i can setup a callback that preprocesses the spi
> frames before sending?
>
> I'm working up the chain an i can see a potential patch to regmaps that
> would  fix this if nothing better already exists
>
> with regmap_config exists /@read_flag_mask/ and /@write_flag_mask/,
> along the same line there could be added a "@read_parity_mask" and
> "@write_parity_mask"
> and an optional "@reg_parity" callback that calculates the parity
> from/reg/ and /val/ similar/@reg_read/ and /@reg_write/ but without
> replacing the entire read/write operations
>
> I may even be able to write this but that is outside the spi-subsystem,
> and i'm currently not sure what list regmaps fall under

-- 
With Best Regards,
Andy Shevchenko

Re: Remap, SPI and parity bit

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 651 bytes --]

On Mon, Nov 15, 2021 at 05:14:52PM +0200, Andy Shevchenko wrote:
> On Sun, Nov 14, 2021 at 8:41 PM christopher tyerman
> 
> > I'm trying to use regmaps where possible and I can define a regmap, e.g.

...

> > however so far i cannot currently see how to automatically calculate it
> > without replacing large chunks of code

This seems like a sign that you should be extending the core so that it
supports having a parity bit, it looks like there's at least some other
parts with this functionality. The pattern seems to be a bit that that
it's a bit at some specified offset which is the parity of all other
non-pad bits in the write.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]