RE: [PATCH RESEND] i2c: tegra: Add SMBus block read function

[Akhil R]

> 12.02.2022 19:20, Dmitry Osipenko пишет:
> > 12.02.2022 19:15, Dmitry Osipenko пишет:
> >> 11.02.2022 12:11, Akhil R пишет:
> >>>> 10.02.2022 18:36, Akhil R пишет:
> >>>>> Emulate SMBus block read using ContinueXfer to read the length byte
> >>>>>
> >>>>> Signed-off-by: Akhil R <akhilrajeev@xxxxxxxxxx>
> >>>>> ---
> >>>>>  drivers/i2c/busses/i2c-tegra.c | 18 ++++++++++++++++--
> >>>>>  1 file changed, 16 insertions(+), 2 deletions(-)
> >>>>>
> >>>>> diff --git a/drivers/i2c/busses/i2c-tegra.c b/drivers/i2c/busses/i2c-tegra.c
> >>>>> index 03cea102ab76..2941e42aa6a0 100644
> >>>>> --- a/drivers/i2c/busses/i2c-tegra.c
> >>>>> +++ b/drivers/i2c/busses/i2c-tegra.c
> >>>>> @@ -1233,6 +1233,11 @@ static int tegra_i2c_xfer_msg(struct
> tegra_i2c_dev
> >>>> *i2c_dev,
> >>>>>               return err;
> >>>>>
> >>>>>       i2c_dev->msg_buf = msg->buf;
> >>>>> +
> >>>>> +     /* The condition true implies smbus block read and len is already read
> */
> >>>>> +     if (msg->flags & I2C_M_RECV_LEN && end_state !=
> >>>> MSG_END_CONTINUE)
> >>>>> +             i2c_dev->msg_buf = msg->buf + 1;
> >>>>> +
> >>>>>       i2c_dev->msg_buf_remaining = msg->len;
> >>>>>       i2c_dev->msg_err = I2C_ERR_NONE;
> >>>>>       i2c_dev->msg_read = !!(msg->flags & I2C_M_RD);
> >>>>> @@ -1389,6 +1394,15 @@ static int tegra_i2c_xfer(struct i2c_adapter
> *adap,
> >>>> struct i2c_msg msgs[],
> >>>>>                       else
> >>>>>                               end_type = MSG_END_REPEAT_START;
> >>>>>               }
> >>>>> +             /* If M_RECV_LEN use ContinueXfer to read the first byte */
> >>>>> +             if (msgs[i].flags & I2C_M_RECV_LEN) {
> >>>>> +                     ret = tegra_i2c_xfer_msg(i2c_dev, &msgs[i],
> >>>> MSG_END_CONTINUE);
> >>>>> +                     if (ret)
> >>>>> +                             break;
> >>>>> +                     /* Set the read byte as msg len */
> >>>>> +                     msgs[i].len = msgs[i].buf[0];
> >>>>> +                     dev_dbg(i2c_dev->dev, "reading %d bytes\n", msgs[i].len);
> >>>>> +             }
> >>>>>               ret = tegra_i2c_xfer_msg(i2c_dev, &msgs[i], end_type);
> >>>>>               if (ret)
> >>>>>                       break;
> >>>>> @@ -1416,10 +1430,10 @@ static u32 tegra_i2c_func(struct i2c_adapter
> >>>> *adap)
> >>>>>  {
> >>>>>       struct tegra_i2c_dev *i2c_dev = i2c_get_adapdata(adap);
> >>>>>       u32 ret = I2C_FUNC_I2C | (I2C_FUNC_SMBUS_EMUL &
> >>>> ~I2C_FUNC_SMBUS_QUICK) |
> >>>>> -               I2C_FUNC_10BIT_ADDR | I2C_FUNC_PROTOCOL_MANGLING;
> >>>>> +               I2C_FUNC_10BIT_ADDR | I2C_FUNC_PROTOCOL_MANGLING;
> >>>>>
> >>>>>       if (i2c_dev->hw->has_continue_xfer_support)
> >>>>> -             ret |= I2C_FUNC_NOSTART;
> >>>>> +             ret |= I2C_FUNC_NOSTART |
> I2C_FUNC_SMBUS_READ_BLOCK_DATA;
> >>>>>
> >>>>>       return ret;
> >>>>>  }
> >>>>
> >>>> Please describe how this was tested.
> >>> This is tested using an I2C EEPROM to emulate SMBus block read in which
> >>> we read the first byte as the length of bytes to read. This is an expected
> >>> feature for NVIDIA Grace chipset where there will be an actual SMBus
> device.
> >>
> >> We have several Tegra30+ tablets that have EC on SMBus. Svyatoslav tried
> >> your I2C patch + [1] on Asus TF201 and reported that it breaks EC. Any
> >> idea why it doesn't work?
> >>
> >> [1]
> >> https://github.com/grate-
> driver/linux/commit/aa8d71f5a960ef40503e5448c622d62d1c53a2c0
> >
> > Ah, I see now that I2C_FUNC_SMBUS_WRITE_BLOCK_DATA not supported,
> we
> > should check again without the write then.
> 
> We also missed that i2c_smbus_read_i2c_block_data() populates the first
> byte of the read data with the transfer size. So
> i2c_smbus_read_block_data() actually works properly.
> 
> It's unclear to me what's the point of emulating
> I2C_FUNC_SMBUS_READ_BLOCK_DATA within the driver if you could use
> i2c_smbus_read_i2c_block_data().

We are looking to support I2C_M_RECV_LEN where the length is read from the
first byte of data. I see that i2c_smbus_read_i2c_block_data() requires the length
to be passed from the client driver.

BTW, I2C_FUNC_SMBUS_WRITE_BLOCK_DATA is also expected to be supported.
It is included in I2C_FUNC_SMBUS_EMUL. I suppose, it doesn't require any additional
change in the driver. The client driver should populate the first byte as the length
of data to be transferred.

Thanks,
Akhil