Problem with a hih6130 sensor in a OMAP I2C bus

Problem with a hih6130 sensor in a OMAP I2C bus

[José Miguel Gonçalves]

Hi,

While testing an HIH6130 humidity and temperature sensor with I2C 
interface in a BeagleBone board I've found that I was unable to read it 
because the driver always returned EINVAL. With some debugging I've 
found that the error was due to a test on omap_i2c_xfer_msg() in OMAP 
I2C driver that invalidates zero length writes. The hwmon hih6130 driver 
issues such kind of request in hih6130_update_measurements() to issue a 
measurement request to the sensor.

I was able to get measurements from the sensor by hacking the hih6130 
driver replacing the following line in hih6130_update_measurements();

         ret = i2c_master_send(client, tmp, 0);

by

         tmp[0] = 0;
         ret = i2c_master_send(client, tmp, 1);

Is this the correct way to fix this issue, or should the fix be in the 
I2C OMAP driver to accept zero length transfers?

Best regards,
José Gonçalves

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: Problem with a hih6130 sensor in a OMAP I2C bus

[Jean Delvare]

Hi José,

On Mon, 09 Dec 2013 11:42:13 +0000, José Miguel Gonçalves wrote:
> While testing an HIH6130 humidity and temperature sensor with I2C 
> interface in a BeagleBone board I've found that I was unable to read it 
> because the driver always returned EINVAL. With some debugging I've 
> found that the error was due to a test on omap_i2c_xfer_msg() in OMAP 
> I2C driver that invalidates zero length writes. The hwmon hih6130 driver 
> issues such kind of request in hih6130_update_measurements() to issue a 
> measurement request to the sensor.
> 
> I was able to get measurements from the sensor by hacking the hih6130 
> driver replacing the following line in hih6130_update_measurements();
> 
>          ret = i2c_master_send(client, tmp, 0);
> 
> by
> 
>          tmp[0] = 0;
>          ret = i2c_master_send(client, tmp, 1);
> 
> Is this the correct way to fix this issue, or should the fix be in the 
> I2C OMAP driver to accept zero length transfers?

Ideally it should be fixed in the i2c-omap bus driver. However, if
zero-length transfers are explicitly prohibited, it may mean that the
hardware itself can't do it (in which case adding a comment saying so
in the source code would be great.) You'll have to check. The driver
supports several generations of OMAP chipsets so you'll have to check
them all, maybe some can handle zero-length messages and some do not.

If fixing it in the i2c-omap driver is not possible, then working
around it in the hih6130 driver would be acceptable. You'll have to
make sure there is no side effect to sending a single byte to the chip.
If performance is a concern, you may have to enable the workaround only
when zero-byte messages aren't supported. You can check with:

i2c_get_functionality(i2c_adapter) & I2C_FUNC_SMBUS_QUICK

-- 
Jean Delvare

Re: Problem with a hih6130 sensor in a OMAP I2C bus

[José Miguel Gonçalves]

Hi Jean,

On 09-12-2013 12:22, Jean Delvare wrote:
> Hi José,
>
> On Mon, 09 Dec 2013 11:42:13 +0000, José Miguel Gonçalves wrote:
>> While testing an HIH6130 humidity and temperature sensor with I2C
>> interface in a BeagleBone board I've found that I was unable to read it
>> because the driver always returned EINVAL. With some debugging I've
>> found that the error was due to a test on omap_i2c_xfer_msg() in OMAP
>> I2C driver that invalidates zero length writes. The hwmon hih6130 driver
>> issues such kind of request in hih6130_update_measurements() to issue a
>> measurement request to the sensor.
>>
>> I was able to get measurements from the sensor by hacking the hih6130
>> driver replacing the following line in hih6130_update_measurements();
>>
>>           ret = i2c_master_send(client, tmp, 0);
>>
>> by
>>
>>           tmp[0] = 0;
>>           ret = i2c_master_send(client, tmp, 1);
>>
>> Is this the correct way to fix this issue, or should the fix be in the
>> I2C OMAP driver to accept zero length transfers?
> Ideally it should be fixed in the i2c-omap bus driver. However, if
> zero-length transfers are explicitly prohibited, it may mean that the
> hardware itself can't do it (in which case adding a comment saying so
> in the source code would be great.) You'll have to check. The driver
> supports several generations of OMAP chipsets so you'll have to check
> them all, maybe some can handle zero-length messages and some do not.
>
> If fixing it in the i2c-omap driver is not possible, then working
> around it in the hih6130 driver would be acceptable. You'll have to
> make sure there is no side effect to sending a single byte to the chip.
> If performance is a concern, you may have to enable the workaround only
> when zero-byte messages aren't supported. You can check with:
>
> i2c_get_functionality(i2c_adapter) & I2C_FUNC_SMBUS_QUICK
>

I am by no means an expert on OMAP platforms and/or I2C drivers so, if 
anyone else with more knowledge of them has a better suggestion, I will 
follow your suggestion and submit my workaround patch for the hih6130 
driver with the added check for support of zero-byte messages by the 
underlying I2C driver. From my understanding of the technical note for 
I2C communication on this chip;

http://www.mouser.com/pdfdocs/I2CCommunications.pdf

I see no problems on writing that dummy byte to it and my tests on a 
real hardware also show that.

Best regards,
José Gonçalves