I have tested this patch on my 2Big Network v2 board.
Sometimes I get a weird "No such device" error while reading or writing
the hwmon sysfs attributes. For example:
# cat pwm1
cat: read error: No such device
# cat pwm1
0
Weird, I cannot reproduce that at all on my ReadyNAS Duo v2.
Note, that if the same command is executed again (just after the faulty
one), then it will succeed most of the time. The errors seems to happen
quite randomly and not very often. I'd say one time over ten tries.
I suppose this error is now visible due to the enhanced error path in
g762_update_client. It looks like we learn something by checking the
return value for i2c_smbus_read_byte_data :)
Except that, all is working as expected. This new version fixes
effectively open-loop driving when the set_cnt register is set to 0xff.
Also I am planning (hopefully this Wednesday) to test your patch on my
5Big Network v2 board. This last also embed a g762 controller but with a
three wires fan instead. Then, I should be able to test the closed-loop
mode behaviour.
On Wed, Jun 19, 2013 at 10:15:56PM +0200, Arnaud Ebalard wrote:
Hi Simon,
Simon Guinot [off-list ref] writes:
quoted
I have tested this patch on my 2Big Network v2 board.
Sometimes I get a weird "No such device" error while reading or writing
the hwmon sysfs attributes. For example:
# cat pwm1
cat: read error: No such device
# cat pwm1
0
This is odd, since ENODEV should not be returned as error code from i2c read
operations.
What is the i2c controller ?
Thanks,
Guenter
Weird, I cannot reproduce that at all on my ReadyNAS Duo v2.
quoted
Note, that if the same command is executed again (just after the faulty
one), then it will succeed most of the time. The errors seems to happen
quite randomly and not very often. I'd say one time over ten tries.
I suppose this error is now visible due to the enhanced error path in
g762_update_client. It looks like we learn something by checking the
return value for i2c_smbus_read_byte_data :)
Except that, all is working as expected. This new version fixes
effectively open-loop driving when the set_cnt register is set to 0xff.
Also I am planning (hopefully this Wednesday) to test your patch on my
5Big Network v2 board. This last also embed a g762 controller but with a
three wires fan instead. Then, I should be able to test the closed-loop
mode behaviour.
From: Simon Guinot <hidden> Date: 2013-06-19 21:26:33
On Wed, Jun 19, 2013 at 01:53:21PM -0700, Guenter Roeck wrote:
On Wed, Jun 19, 2013 at 10:15:56PM +0200, Arnaud Ebalard wrote:
quoted
Hi Simon,
Simon Guinot [off-list ref] writes:
quoted
I have tested this patch on my 2Big Network v2 board.
Sometimes I get a weird "No such device" error while reading or writing
the hwmon sysfs attributes. For example:
# cat pwm1
cat: read error: No such device
# cat pwm1
0
This is odd, since ENODEV should not be returned as error code from i2c read
operations.
What is the i2c controller ?
The controller is the very same as for the ReadyNAS Duo v2. It is built
in the Marvell Kirkwood SoC and the driver is i2c-mv64xxx.
I will try an another board to see if I reproduce the error.
Regards,
Simon
Thanks,
Guenter
quoted
Weird, I cannot reproduce that at all on my ReadyNAS Duo v2.
quoted
Note, that if the same command is executed again (just after the faulty
one), then it will succeed most of the time. The errors seems to happen
quite randomly and not very often. I'd say one time over ten tries.
I suppose this error is now visible due to the enhanced error path in
g762_update_client. It looks like we learn something by checking the
return value for i2c_smbus_read_byte_data :)
Except that, all is working as expected. This new version fixes
effectively open-loop driving when the set_cnt register is set to 0xff.
Also I am planning (hopefully this Wednesday) to test your patch on my
5Big Network v2 board. This last also embed a g762 controller but with a
three wires fan instead. Then, I should be able to test the closed-loop
mode behaviour.
On Wed, Jun 19, 2013 at 11:26:33PM +0200, Simon Guinot wrote:
On Wed, Jun 19, 2013 at 01:53:21PM -0700, Guenter Roeck wrote:
quoted
On Wed, Jun 19, 2013 at 10:15:56PM +0200, Arnaud Ebalard wrote:
quoted
Hi Simon,
Simon Guinot [off-list ref] writes:
quoted
I have tested this patch on my 2Big Network v2 board.
Sometimes I get a weird "No such device" error while reading or writing
the hwmon sysfs attributes. For example:
# cat pwm1
cat: read error: No such device
# cat pwm1
0
This is odd, since ENODEV should not be returned as error code from i2c read
operations.
What is the i2c controller ?
The controller is the very same as for the ReadyNAS Duo v2. It is built
in the Marvell Kirkwood SoC and the driver is i2c-mv64xxx.
I will try an another board to see if I reproduce the error.
Might be a flaky i2c connection or a problematic board. The i2c-mv64xxx driver
does return ENODEV in error cases. I submitted a patch to change it to ENXIO;
we'll see if it gets accepted.
Guenter
Regards,
Simon
quoted
Thanks,
Guenter
quoted
Weird, I cannot reproduce that at all on my ReadyNAS Duo v2.
quoted
Note, that if the same command is executed again (just after the faulty
one), then it will succeed most of the time. The errors seems to happen
quite randomly and not very often. I'd say one time over ten tries.
I suppose this error is now visible due to the enhanced error path in
g762_update_client. It looks like we learn something by checking the
return value for i2c_smbus_read_byte_data :)
Except that, all is working as expected. This new version fixes
effectively open-loop driving when the set_cnt register is set to 0xff.
Also I am planning (hopefully this Wednesday) to test your patch on my
5Big Network v2 board. This last also embed a g762 controller but with a
three wires fan instead. Then, I should be able to test the closed-loop
mode behaviour.