[PATCH v5 2/5] i2c: mux: add support for per channel bus frequency

[Marcus Folkesson]

Hi Peter!

On Fri, Feb 13, 2026 at 12:48:13PM +0100, Peter Rosin wrote:
> Hi!
> 

[...]

> > +	if (priv->adap.clock_hz && priv->adap.clock_hz != parent->clock_hz) {
> > +		mux_locked_ancestor = i2c_mux_topmost_mux_locked(adap);
> > +		root = i2c_root_adapter(&adap->dev);
> > +
> > +		/*
> > +		 * If there's a mux-locked mux in our ancestry, lock the parent
> > +		 * of the topmost one. Mux-locked muxes don't propagate locking
> > +		 * to their parents, so we must explicitly acquire the lock above
> > +		 * the highest mux-locked ancestor to reach the root adapter.
> > +		 */
> 
> As may be more apparent from my recent v4 response(?), this is not
> what you want. You want mux_locked_ancestor to be the nearest
> mux-locked ancestor.

Hrm, I think we are pointing at the same thing but naming it differently; M1 in this chain:
Root - P1 - M1 - M2 - P2 - D1

I'm thinking 'topmost' means the one closest to the root (ancestor most
far away), but I don't stick to that naming and I am happy to change it.

> 
> 
> > +		if (mux_locked_ancestor)
> > +			i2c_lock_bus(mux_locked_ancestor->parent, I2C_LOCK_ROOT_ADAPTER);
> > +
> > +		ret = i2c_adapter_set_clk_freq(root, priv->adap.clock_hz);
> > +
> > +		if (mux_locked_ancestor)
> > +			i2c_unlock_bus(mux_locked_ancestor->parent, I2C_LOCK_ROOT_ADAPTER);
> > +
> > +		if (ret < 0) {
> > +			dev_err(&adap->dev,
> > +				"Failed to set clock frequency %dHz on root adapter %s: %d\n",
> > +				priv->adap.clock_hz, root->name, ret);
> > +
> > +			return ret;
> > +		}
> > +	}
> > +
> 
> Here, you set the frequency /before/ ->select.
> 
> > +	return muxc->select(muxc, priv->chan_id);
> > +}
> > +
> > +static void i2c_mux_deselect_chan(struct i2c_adapter *adap, u32 chan_id)
> > +{
> > +	struct i2c_mux_priv *priv = adap->algo_data;
> > +	struct i2c_mux_core *muxc = priv->muxc;
> > +	struct i2c_adapter *parent = muxc->parent;
> > +	struct i2c_mux_core *mux_locked_ancestor = NULL;
> > +	struct i2c_adapter *root;
> > +	int ret;
> 
> And here, you restore the frequency, but /before/ ->deselect.
> 
> Thers's no symmetry in that...

Good point, I will change that.

> 
> Cheers,
> Peter

Best regards,
Marcus Folkesson
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20260213/10e1364e/attachment.sig>