Re: [RFC PATCH v5 2/6] i2c: add helpers to ease DMA handling

[Jonathan Cameron <Jonathan.Cameron@huawei.com>]

On Thu, 21 Sep 2017 16:15:28 +0200
Wolfram Sang <wsa@the-dreams.de> wrote:

> > > > +/**
> > > > + * i2c_release_dma_safe_msg_buf - release DMA safe buffer and sync with i2c_msg
> > > > + * @msg: the message to be synced with
> > > > + * @buf: the buffer obtained from i2c_get_dma_safe_msg_buf(). May be NULL.
> > > > + */
> > > > +void i2c_release_dma_safe_msg_buf(struct i2c_msg *msg, u8 *buf)
> > > > +{
> > > > +	if (!buf || buf == msg->buf)
> > > > +		return;
> > > > +
> > > > +	if (msg->flags & I2C_M_RD)
> > > > +		memcpy(msg->buf, buf, msg->len);
> > > > +
> > > > +	kfree(buf);  
> > 
> > Only free when you actually allocated it.  Seems to me like you need
> > to check if (!(msg->flags & I2C_M_DMA_SAFE)) before kfree.
> > 
> > Otherwise the logic to do this will be needed in every driver
> > which will get irritating fast.  
> 
> Well, I return early if (buf == msg->buf) which is only true for
> I2C_M_DMA_SAFE. If not, I allocated the buffer. Am I missing something?
> It would be very strange to call this function if the caller allocated
> the buffer manually.
> 
> Thanks for the review!

Doh missed that check and my comment was bonkers even if it hadn't been there.
I come back to the claim of insufficient caffeine.

You are quite correct.  Please ignore previous comment - the code is
fine as is. 

Jonathan
> 
>