RE: [PATCH 1/1] i2c: ocores: fix polling mode workaround on FU540-C000 SoC

[Sagar Kadam <sagar.kadam@openfive.com>]

Hello Peter,

> -----Original Message-----
> From: Peter Korsgaard <jacmet@gmail.com> On Behalf Of Peter Korsgaard
> Sent: Wednesday, October 7, 2020 5:10 PM
> To: Sagar Kadam <sagar.kadam@openfive.com>
> Cc: linux-kernel@vger.kernel.org; linux-riscv@lists.infradead.org; linux-
> i2c@vger.kernel.org; andrew@lunn.ch; Paul Walmsley ( Sifive)
> <paul.walmsley@sifive.com>; palmer@dabbelt.com;
> aou@eecs.berkeley.edu
> Subject: Re: [PATCH 1/1] i2c: ocores: fix polling mode workaround on FU540-
> C000 SoC
> 
> [External Email] Do not click links or attachments unless you recognize the
> sender and know the content is safe
> 
> >>>>> "Sagar" == Sagar Shrikant Kadam <sagar.kadam@sifive.com> writes:
> 
>  > The FU540-C000 has a broken IRQ and support was added earlier
>  > so that it will operate in polling mode, but seems to work only
>  > in case interrupts property is missing from the i2c0 dt-node.
>  > This should not be the case and the driver should handle polling
>  > mode with the interrupt property present in i2c0 node of the
>  > device tree.
>  > So check if it's the FU540-C000 soc and enable polling mode master
>  > xfers, as the IRQ for this chip is broken.
> 
>  > Fixes commit c45d4ba86731 ("i2c: ocores: add polling mode workaround
>  > for Sifive FU540-C000 SoC")
> 
>  > Signed-off-by: Sagar Shrikant Kadam <sagar.kadam@sifive.com>
>  > ---
>  >  drivers/i2c/busses/i2c-ocores.c | 22 +++++++++++++---------
>  >  1 file changed, 13 insertions(+), 9 deletions(-)
> 
>  > diff --git a/drivers/i2c/busses/i2c-ocores.c b/drivers/i2c/busses/i2c-
> ocores.c
>  > index f5fc75b..4405244 100644
>  > --- a/drivers/i2c/busses/i2c-ocores.c
>  > +++ b/drivers/i2c/busses/i2c-ocores.c
>  > @@ -686,17 +686,21 @@ static int ocores_i2c_probe(struct
> platform_device *pdev)
> 
>  >      init_waitqueue_head(&i2c->wait);
> 
>  > +    /*
>  > +     * Set OCORES_FLAG_BROKEN_IRQ to enable workaround for
>  > +     * FU540-C000 SoC in polling mode.
>  > +     * Since the SoC does have interrupt it's dt has the interrupt
>  > +     * defined but it should be bypassed in driver as this SoC has
>  > +     * a broken IRQ, hence update the master_xfer to use polling
>  > +     * transfers.
>  > +     */
>  > +    match = of_match_node(ocores_i2c_match, pdev->dev.of_node);
>  > +    if (match && (long)match->data == TYPE_SIFIVE_REV0)
>  > +            i2c->flags |= OCORES_FLAG_BROKEN_IRQ;
>  > +
>  >      irq = platform_get_irq(pdev, 0);
>  > -    if (irq == -ENXIO) {
>  > +    if (i2c->flags == OCORES_FLAG_BROKEN_IRQ || irq == -ENXIO) {
> 
> NIT: flags is a bitmask, so i2c->flags & OCORES_FLAG_BROKEN_IRQ would be
> better, even if there currently doesn't exist any other flags.
> 
Thanks for your suggestions. I will apply the bitmask, this will be better.

> TYPE_SIFIVE_REV0 is also set for two compatibles:
> 
>         {
>                 .compatible = "sifive,fu540-c000-i2c",
>                 .data = (void *)TYPE_SIFIVE_REV0,
>         },
>         {
>                 .compatible = "sifive,i2c0",
>                 .data = (void *)TYPE_SIFIVE_REV0,
>         },
> 
> Are both affected by this issue? if not, we will need to extend the code
> to handle them differently.
> 

No, this issue is present in FU540-C000 SoC i.e SoC- specific, and so I can check 
for the soc-compatible string as follows:

-       match = of_match_node(ocores_i2c_match, pdev->dev.of_node);
-       if (match && (long)match->data == TYPE_SIFIVE_REV0)
+       if (of_device_is_compatible(pdev->dev.of_node,
+                                       "sifive,fu540-c000-i2c"))

Please let me know if this is okay.

Thanks & BR,
Sagar

> Other than that, it looks OK to me.
> 
> --
> Bye, Peter Korsgaard