Re: [video4linux-cvs] [hg:v4l-dvb] Add support for Opera S1- DVB-USB

[abraham.manu at gmail.com (Manu Abraham)]

Trent Piepho wrote:
> On Thu, 19 Apr 2007, Michael Krufky wrote:
>> Was this patch ever posted to the linux-dvb mailing list?  There are some
>> (minor) concerns that I have with this changeset.  Please see my comments inline
>> below:
> 
> Yes, it posted.  I made the same comment about the I2C addresses.  This
> driver does something funny with i2c and sometimes uses 16-bit addresses.
> 
> There is an addresses width extension to i2c, but I don't think that's
> the case here.  I looks like the i2c function code is based on reverse
> engineering from usb logs.  It is probably not 16-bit i2c addresses but
> something else that is going on.

I2C addresses are generally 7 bit only. There is a 10 bit extension
albeit, that's not the case here. What's to be noted is that the host
itself is 8 bit, so definitely it cannot be a 10 bit address.

Now, there are devices that do addressing in 32 bits etc, but those are
internal to the device, nothing external to be seen like this, those
would be used to access the internal pages on the specific host.

The STB0899 uses such an addressing scheme, where they are called "offsets"

>>> +static int opera1_xilinx_rw(struct usb_device *dev, u8 request, u16 value,
>>> +			    u8 * data, u16 len, int flags)
>>> +{
>>> +	int ret;
>>> +	u8 r;
>>> +	u8 u8buf[len];
> 
> Do you know if C99 variable length arrays like this are allowed in kernel
> code?


variable length arrays are a bit frowned upon, i think


Manu