Hi Geert-san,
> From: Yoshihiro Shimoda, Sent: Thursday, May 31, 2018 6:12 PM
>
> Hi Geert-san, Wolfram-san,
>
> > From: Geert Uytterhoeven, Sent: Thursday, May 31, 2018 5:45 PM
> >
> > Hi Shimoda-san,
> >
> > On Wed, May 30, 2018 at 10:35 AM, Yoshihiro Shimoda
> > <yoshihiro.shimoda.uh@xxxxxxxxxxx> wrote:
> > >> From: Wolfram Sang, Sent: Tuesday, May 29, 2018 7:59 PM
> > >> Subject: [RFC PATCH 1/1] i2c: rcar: handle RXDMA HW bug on Gen3
> > >
> > > If possible, I'd like to replace "bug" with "specification" or other words :)
> > >
> > > <snip>
> > >> @@ -743,6 +753,16 @@ static int rcar_i2c_master_xfer(struct i2c_adapter *adap,
> > >>
> > >> pm_runtime_get_sync(dev);
> > >>
> > >> + /* Gen3 has a HW bug which needs a reset before allowing RX DMA once */
> > >> + if (priv->devtype == I2C_RCAR_GEN3) {
> > >> + priv->flags |= ID_P_NO_RXDMA;
> > >> + if (!IS_ERR(priv->rstc)) {
> > >> + ret = reset_control_reset(priv->rstc);
> > >
> > > According to the datasheet Rev.1.00 page 57-69, we should do:
> > > reset_control_assert();
> > > udelay(1);
> > > reset_control_deassert();
> > > while (reset_control_status())
> > > ;
> > > instead of reset_control_reset(), I think.
> >
> > The i2c-specific procedure documented at page 57-69 thus differs from
> > the generic one at page 8A-58, which is what cpg_mssr_reset() implements.
> >
> > The latter waits 35µs instead of 1µs, so that should be safe.
> > But it doesn't check the status bit. Is the longer delay sufficient, or should
> > a check polling the status bit be added to cpg_mssr_reset()?
>
> Thank you for the pointed out.
> I agree we should wait 35us for safe.
> But, anyway I'll ask HW team about this contradiction and really need polling the status.
I got information about this topic.
< In CPG / MSSR point of view >
- This needs 35 usec wait while asserting.
- After deasserted a reset, no wait needs.
- In other words, there is each hardware IP dependence.
< In I2C point of view >
- After deasserted the reset, this nees SRCR register polling.
So, I don't think cpg_mssr_reset() checks the status bit after deasserted a reset.
But, what do you think?
Best regards,
Yoshihiro Shimoda
