On Thu, 17 Mar 2022 23:52:35 -0500 Samuel Holland <sam...@sholland.org> wrote:
Hi, > Chips attached to the RSB bus require an initialization command before > they can be used. (Specifically, this command programs the chip's > runtime address.) The driver does this in its .probe_chip hook, under > the assumption that .probe_chip is called during child probe. This is > not the case; .probe_chip is only called by dm_i2c_probe, which is > intended for use by board-level code, not for chips with OF nodes. > > Since this initialization command must be run before a child chip can be > used, do it before probing each child. > > Signed-off-by: Samuel Holland <sam...@sholland.org> Same as for patch 1/4: currently that code path is not called by any sunxi code, so without looking too deep or having the ability to test this, it looks alright to me: Acked-by: Andre Przywara <andre.przyw...@arm.com> Cheers, Andre > --- > > drivers/i2c/sun8i_rsb.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/drivers/i2c/sun8i_rsb.c b/drivers/i2c/sun8i_rsb.c > index 716b245a00..0681e398f7 100644 > --- a/drivers/i2c/sun8i_rsb.c > +++ b/drivers/i2c/sun8i_rsb.c > @@ -252,11 +252,12 @@ static int sun8i_rsb_probe(struct udevice *bus) > static int sun8i_rsb_child_pre_probe(struct udevice *child) > { > struct dm_i2c_chip *chip = dev_get_parent_plat(child); > + struct udevice *bus = child->parent; > > /* Ensure each transfer is for a single register. */ > chip->flags |= DM_I2C_CHIP_RD_ADDRESS | DM_I2C_CHIP_WR_ADDRESS; > > - return 0; > + return sun8i_rsb_probe_chip(bus, chip->chip_addr, 0); > } > > static const struct dm_i2c_ops sun8i_rsb_ops = {