> + /* If we should retrieve the length from the buffer, make sure */
> + /* to read enough bytes to avoid sending the */
> + /* STOP bit after the read if the first byte */
Please stick to Kernel commenting style.
> + /* length=0 should not be allowed, but is indeed possible.
> + * To avoid locking the chip, we keep reading at least 2 bytes
> + */
Ditto.
> + drv_data->effective_length = data+1;
> + drv_data->bytes_left = data+1;
> + drv_data->msg->len = data+1;
Also, checkpatch complains about this if you run it with "--strict'
CHECKPATCH
CHECK: spaces preferred around that '+' (ctx:VxV)
#81: FILE: drivers/i2c/busses/i2c-mv64xxx.c:384:
+ drv_data->effective_length = data+1;
^
CHECK: spaces preferred around that '+' (ctx:VxV)
#82: FILE: drivers/i2c/busses/i2c-mv64xxx.c:385:
+ drv_data->bytes_left = data+1;
^
CHECK: spaces preferred around that '+' (ctx:VxV)
#83: FILE: drivers/i2c/busses/i2c-mv64xxx.c:386:
+ drv_data->msg->len = data+1;
^
I think the complaints make sense.
> - return I2C_FUNC_I2C | I2C_FUNC_10BIT_ADDR | I2C_FUNC_SMBUS_EMUL;
> + return I2C_FUNC_I2C | I2C_FUNC_10BIT_ADDR |
> + I2C_FUNC_SMBUS_READ_BLOCK_DATA | I2C_FUNC_SMBUS_EMUL;
Likely I2C_FUNC_SMBUS_BLOCK_PROC_CALL will work as well, but I assume
you can't test it?
Otherwise, looks good.
Attachment:
signature.asc
Description: PGP signature
