[i2c] [PATCH 8/12] drivers: PMC MSP71xx TWI driver

[Marc St-Jean]

Marc St-Jean wrote:
> 
> 
> Jean Delvare wrote:
>  > Hi Marc,
>  >
>  > On Thu, 28 Jun 2007 18:01:11 -0600, Marc St-Jean wrote:
>  >  > [PATCH 8/12] drivers: PMC MSP71xx TWI driver
>  >  >
>  >  > Patch to add TWI driver for the PMC-Sierra MSP71xx devices.
>  >  >
>  >  > Thanks,
>  >  > Marc
>  >  >
>  >  > Signed-off-by: Marc St-Jean <Marc_St-Jean at pmc-sierra.com>
>  >  > ---
>  >  > Changes since last post:
>  >  > - Changed config item to no longer depend on I2C and updated help 
> text.
>  >  > - Converted to platform device driver and rewrote probe/remove
>  > functions.
>  >  > - Changed xfer_command to fix up be64 handling in .
>  >  > - Changed master_xfer function to eliminate support for multiple
>  > commands
>  >  > other that simple write-read. Fixed probing workaround to not perform
>  > a read.
>  >  > - Updated IRQ flags.
>  >  > - Returned more meaningful error codes.
>  >  > - Removed unnecessary headers and definitions.
>  >  > - Removed get_clock_config function as not currently used.
>  >  > - Removed algo_data function pointers and made direct calls.
>  >  > - Removed extra casts and backets.
>  >  > - Removed redundant debugging output.
>  >  > - Other cleanups as requested by feedback.
>  >  >
>  >  >  Kconfig      |    6
>  >  >  Makefile     |    1
>  >  >  i2c-pmcmsp.c |  674
>  > +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>  >  >  3 files changed, 681 insertions(+)
>  >
...

>  >
>  >  > +
>  >  > +     if (msg->len == 0) {
>  >  > +             if (msg->flags & I2C_M_RD) {
>  >  > +                     dev_dbg(&adap->dev,
>  >  > +                             "Read of 0 bytes unsupported\n");
>  >  > +                     return -EINVAL;
>  >  > +             } else {
>  >  > +                     dev_dbg(&adap->dev, "Probing for slave at
>  > 0x%02x\n",
>  >  > +                             msg->addr & 0xff);
>  >  > +                     probe = true;
>  >  > +
>  >  > +                     /* Probe is a special write of 1 byte */
>  >  > +                     cmd.write_len = 1;
>  >  > +                     cmd.write_data = probe_buf;
>  >  > +             }
>  >  > +     }
>  >
>  > This should go away. As explained in my other post, new-style i2c
>  > drivers don't need probing (and if you really want probing, it no
>  > longer requires zero-byte messages.)
> 
> I understand but is that not dependent on a the chip drivers?
> We can't assume our customers only use new style drivers to get
> at devices on their boards.
> 
>  >  > +
>  >  > +     cmd.addr = msg->addr;
>  >  > +
>  >  > +     if (probe || (msg->flags & I2C_M_TEN)) {
>  >  > +             pmcmsptwi_get_twi_config(&newcfg, data);
>  >  > +             memcpy(&oldcfg, &newcfg, sizeof(oldcfg));
>  >  > +
>  >  > +             /* For probes, we don't want any retries */
>  >  > +             if (probe)
>  >  > +                     newcfg.nak = 0;
>  >  > +
>  >  > +             /* Set the special 10-bit address flag, if required */
>  >  > +             if (msg->flags & I2C_M_TEN)
>  >  > +                     newcfg.add10 = 1;
>  >  > +
>  >  > +             pmcmsptwi_set_twi_config(&newcfg, data);
>  >  > +     }
>  >  > +
>  >  > +     /* Execute the command */
>  >  > +     ret = pmcmsptwi_xfer_cmd(&cmd, data);
>  >  > +
>  >  > +     if (probe || (msg->flags & I2C_M_TEN))
>  >  > +             pmcmsptwi_set_twi_config(&oldcfg, data);
>  >  > +
>  >  > +     dev_dbg(&adap->dev, "I2C %s of %d bytes ",
>  >  > +             (msg->flags & I2C_M_RD) ? "read" : "write", msg->len);
>  >  > +     if (ret != MSP_TWI_XFER_OK) {
>  >  > +             /*
>  >  > +              * TODO: We could potentially loop and retry in the 
> case
>  >  > +              * of MSP_TWI_XFER_TIMEOUT.
>  >  > +              */
>  >  > +             dev_dbg(&adap->dev, "failed\n");
>  >  > +             return -1;
>  >  > +     }
>  >  > +
>  >  > +     dev_dbg(&adap->dev, "succeeded\n");
>  >  > +     return 0;
>  >  > +}
>  >  > +
>  >  > +static u32 pmcmsptwi_i2c_func(struct i2c_adapter *adapter)
>  >  > +{
>  >  > +     return I2C_FUNC_I2C | I2C_FUNC_10BIT_ADDR | 
> I2C_FUNC_SMBUS_EMUL;
>  >  > +}
>  >
>  > You'll have to change this. I2C_FUNC_SMBUS_EMUL includes
>  > I2C_FUNC_SMBUS_QUICK and it's important that you don't pretend to
>  > support that when you don't.
> 
> OK, changed to:
>         return I2C_FUNC_I2C | | I2C_FUNC_10BIT_ADDR |
>                 I2C_FUNC_SMBUS_BYTE | I2C_FUNC_SMBUS_BYTE_DATA |
>                 I2C_FUNC_SMBUS_WORD_DATA | I2C_FUNC_SMBUS_PROC_CALLL;
> 

Hi Jean,

I guess what this function returns is dependent on the answer to the first item 
above.

I just found out that removing I2C_FUNC_SMBUS_QUICK will prevent probes from 
functioning, so I've added it for now.

The error returned is:
i2c /dev entries driver
pmcmsptwi: Registering MSP71xx I2C adapter
PMC MSP TWI-LED Driver
i2c_adapter i2c-0: SMBus Quick command not supported, can't probe for chips

...

Marc