I2C_RDRW_IOCTL_MAX_MSGS

I2C_RDRW_IOCTL_MAX_MSGS

[Srinivasan, Nageswari]

Hi,

Is there any specific reason to limit no of i2c messages 
to 42 in i2c-dev.c file

/* Put an arbitrary limit on the number of messages that can
         * be sent at once */
        if (rdwr_arg.nmsgs > I2C_RDRW_IOCTL_MAX_MSGS)
                return -EINVAL;

Regards,
Nageswari

Re: I2C_RDRW_IOCTL_MAX_MSGS

[Jean Delvare]

Hi Nageswari,

On Tue, 17 Jan 2012 10:56:43 +0000, Srinivasan, Nageswari wrote:
> Is there any specific reason to limit no of i2c messages 
> to 42 in i2c-dev.c file
> 
> /* Put an arbitrary limit on the number of messages that can
>          * be sent at once */
>         if (rdwr_arg.nmsgs > I2C_RDRW_IOCTL_MAX_MSGS)
>                 return -EINVAL;

No, as the comment says there is no good reason, it's purely arbitrary.
If this limitation is a problem for you it can be raised.

But you should also keep in mind that pushing many messages at once
increases the I2C bus latency so other devices on the same bus might
become unresponsive.

As a matter of fact the limitation has been there for over 10 years and
nobody ever complained about it, so you must be doing something quite
unusual. I invite you to explain your case in details if you need
I2C_RDRW_IOCTL_MAX_MSGS to be increased.

-- 
Jean Delvare

RE: I2C_RDRW_IOCTL_MAX_MSGS

[Srinivasan, Nageswari]

Jean,

On Tue, Jan 17, 2012 at 18:39:54, Jean Delvare wrote:
> Hi Nageswari,
> 
>
[...] 
> No, as the comment says there is no good reason, it's purely arbitrary.
> If this limitation is a problem for you it can be raised.
> 
> But you should also keep in mind that pushing many messages at once
> increases the I2C bus latency so other devices on the same bus might
> become unresponsive.
> 
> As a matter of fact the limitation has been there for over 10 years and
> nobody ever complained about it, so you must be doing something quite
> unusual. I invite you to explain your case in details if you need
> I2C_RDRW_IOCTL_MAX_MSGS to be increased.
> 
> -- 

We have a slave device which requires more than 500 registers
to be configured. (kind of bulk write) So, looking for an option
of configuring all of them in single command.

> Jean Delvare
> 

Nageswari

Re: I2C_RDRW_IOCTL_MAX_MSGS

[Jean Delvare]

On Tue, 17 Jan 2012 14:49:01 +0000, Srinivasan, Nageswari wrote:
> We have a slave device which requires more than 500 registers
> to be configured. (kind of bulk write) So, looking for an option
> of configuring all of them in single command.

How long will it take? You should ask yourself if it really OK to block
the bus for so long.

What happens if you configure the chip in several goes? The time
penalty shouldn't be that big and I can't believe a chip can't be
configured one register after the other (or at least group by group.)

If you really have to do it all in one go (and even if not) you should
check if the chip supports register address pointer autoincrement
(or block writes.) This would let you write several register values in
one message, so you no longer have to send 500 messages, plus this
would be faster than your current strategy.

-- 
Jean Delvare

RE: I2C_RDRW_IOCTL_MAX_MSGS

[Srinivasan, Nageswari]

Jean,

On Tue, Jan 17, 2012 at 20:36:55, Jean Delvare wrote:
> On Tue, 17 Jan 2012 14:49:01 +0000, Srinivasan, Nageswari wrote:
> > We have a slave device which requires more than 500 registers
> > to be configured. (kind of bulk write) So, looking for an option
> > of configuring all of them in single command.
> 
> How long will it take? You should ask yourself if it really OK to block
> the bus for so long.
> 
> What happens if you configure the chip in several goes? The time
> penalty shouldn't be that big and I can't believe a chip can't be
> configured one register after the other (or at least group by group.)
> 
> If you really have to do it all in one go (and even if not) you should
> check if the chip supports register address pointer autoincrement
> (or block writes.) This would let you write several register values in
> one message, so you no longer have to send 500 messages, plus this
> would be faster than your current strategy.
> 

Thanks. Will try these options.

> -- 


> Jean Delvare
>

Nageswari