Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1431051 > unrolled thread

[RFT PATCH] regmap-i2c: Only use I2C_FUNC_I2C if val_bits == 8

Back to article view | Back to linux.kernel

Contents

  [RFT PATCH] regmap-i2c: Only use I2C_FUNC_I2C if val_bits == 8 Guenter Roeck <linux@roeck-us.net> - 2016-06-25 05:00 +0200
    Re: [RFT PATCH] regmap-i2c: Only use I2C_FUNC_I2C if val_bits == 8 Mark Brown <broonie@kernel.org> - 2016-06-25 13:40 +0200
      Re: [RFT PATCH] regmap-i2c: Only use I2C_FUNC_I2C if val_bits == 8 Guenter Roeck <linux@roeck-us.net> - 2016-06-25 17:10 +0200

#1431051 — [RFT PATCH] regmap-i2c: Only use I2C_FUNC_I2C if val_bits == 8

On chips with a register value size of 16 bits, I2C block reads will
typically not return the expected values. Instead, returned values
are most likely undefined if an operation crosses a register boundary.

This can be observed, for example, with a TMP102 connected to an omap
i2c controller. With this chip, the initial regmap read operation to fill
the register cache looks as follows (debug log from modified kernel).

omap_i2c 48070000.i2c: MSG[0]: add=0x0048, len: 1, flags: 0x0
omap_i2c 48070000.i2c: [0] 0x00
omap_i2c 48070000.i2c: MSG[1]: add=0x0048, len: 8, flags: 0x1
omap_i2c 48070000.i2c: [0] 0x23
omap_i2c 48070000.i2c: [1] 0xa0
omap_i2c 48070000.i2c: [2] 0xff
omap_i2c 48070000.i2c: [3] 0xff
omap_i2c 48070000.i2c: [4] 0xff
omap_i2c 48070000.i2c: [5] 0xff
omap_i2c 48070000.i2c: [6] 0xff
omap_i2c 48070000.i2c: [7] 0xff

Reported-by: Nishanth Menon <nm@ti.com>
Cc: Nishanth Menon <nm@ti.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
---
Nishanth,

it would be great if you can test this patch, to see if it solves the problem
you had observed. Please note that https://patchwork.kernel.org/patch/9191185/
needs to be applied first (it is available in next-20160624).

 drivers/base/regmap/regmap-i2c.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/base/regmap/regmap-i2c.c b/drivers/base/regmap/regmap-i2c.c
index 4735318f4268..f319e6829816 100644
--- a/drivers/base/regmap/regmap-i2c.c
+++ b/drivers/base/regmap/regmap-i2c.c
@@ -257,7 +257,8 @@ static struct regmap_bus regmap_i2c_smbus_i2c_block = {
 static const struct regmap_bus *regmap_get_i2c_bus(struct i2c_client *i2c,
 					const struct regmap_config *config)
 {
-	if (i2c_check_functionality(i2c->adapter, I2C_FUNC_I2C))
+	if (config->val_bits == 8 &&
+	    i2c_check_functionality(i2c->adapter, I2C_FUNC_I2C))
 		return &regmap_i2c;
 	else if (config->val_bits == 8 && config->reg_bits == 8 &&
 		 i2c_check_functionality(i2c->adapter,
-- 
2.5.0

[toc] | [next] | [standalone]

#1431119

[Multipart message — attachments visible in raw view] — view raw

On Fri, Jun 24, 2016 at 07:53:53PM -0700, Guenter Roeck wrote:

> On chips with a register value size of 16 bits, I2C block reads will
> typically not return the expected values. Instead, returned values
> are most likely undefined if an operation crosses a register boundary.

No, this just isn't true in general - if a device has trouble with bulk
reads it needs to set use_sigle_rw in the config to disable them.
Device limitations with bulk reads are a per device thing and are
orthogonal to the value size, being more to do with the age and quality
of the register map implementation than anything else.

[toc] | [prev] | [next] | [standalone]

#1431140

On 06/25/2016 04:34 AM, Mark Brown wrote:
> On Fri, Jun 24, 2016 at 07:53:53PM -0700, Guenter Roeck wrote:
>
>> On chips with a register value size of 16 bits, I2C block reads will
>> typically not return the expected values. Instead, returned values
>> are most likely undefined if an operation crosses a register boundary.
>
> No, this just isn't true in general - if a device has trouble with bulk
> reads it needs to set use_sigle_rw in the config to disable them.
> Device limitations with bulk reads are a per device thing and are
> orthogonal to the value size, being more to do with the age and quality
> of the register map implementation than anything else.
>

Ok, thanks for letting me know. I'll try that.

Guenter

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web