Re: [PATCH v2 1/2] i3c: Remove the const qualifier from i2c_msg pointer in i2c_xfers API

[Alexandre Belloni <alexandre.belloni@bootlin.com>]

On 04/02/2025 17:50:14+0100, Thomas Weißschuh wrote:
> Hi Billy,
> 
> On 2025-02-04 10:09:31+0000, Billy Tsai wrote:
> > > On 2025-02-04 17:17:01+0800, Billy Tsai wrote:
> > > > The change is necessary to enable the use of the
> > > > `i2c_get_dma_safe_msg_buf()` API, which requires a non-const
> > > > `struct i2c_msg *` to operate. The `i2c_get_dma_safe_msg_buf()` function
> > > > ensures safe handling of I2C messages when using DMA, making it essential
> > > > for scenarios where DMA transfers are involved. By removing the `const`
> > > > qualifier, this patch allows drivers to prepare and manage DMA-safe
> > > > buffers directly.
> > 
> > > This is missing a changelog to v1 of the series.
> > 
> > > Also I asked before why it is not possible to change the signature of
> > > i2c_get_dma_safe_msg_buf() to accept 'const struct i2c_msg *' [0].
> > > That looks like the nicer solution to me.
> 
> > The i2c_get_dma_safe_msg_buf() function has existed for a long time,
> > but I do not know the original reason why it declares struct i2c_msg *
> > without const.
> 
> I would guess that nobody ever cared before.
> 
> > However, I believe this is because the
> > i2c_put_dma_safe_msg_buf() function modifies the buffer data, so for
> > consistency, it is declared without const.
> 
> IMO that's a weak argument.
> Maybe the maintainers have a preference?
> 

I guess Wolfram will be the one have the preference, I'm fine using the
existing API as-is.

-- 
Alexandre Belloni, co-owner and COO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com