Re: [PATCH v5 2/2] drm/bridge: parade-ps8640: Add support for AUX channel

[Philip Chen]

Hi

On Tue, Sep 21, 2021 at 9:02 AM Doug Anderson <dianders@xxxxxxxxxxxx> wrote:
>
> Hi,
>
> On Sat, Sep 18, 2021 at 10:21 AM Philip Chen <philipchen@xxxxxxxxxxxx> wrote:
> >
> > +static ssize_t ps8640_aux_transfer(struct drm_dp_aux *aux,
> > +                                  struct drm_dp_aux_msg *msg)
> > +{
> > +       struct ps8640 *ps_bridge = aux_to_ps8640(aux);
> > +       struct regmap *map = ps_bridge->regmap[PAGE0_DP_CNTL];
> > +       struct device *dev = &ps_bridge->page[PAGE0_DP_CNTL]->dev;
> > +
> > +       unsigned int len = msg->size;
>
> nit: usually no blank lines in the variable definition section.
Fixed in v6.
PTAL.

>
>
> > +       base = PAGE0_SWAUX_ADDR_7_0;
> > +       addr_len[PAGE0_SWAUX_ADDR_7_0 - base] = msg->address;
> > +       addr_len[PAGE0_SWAUX_ADDR_15_8 - base] = msg->address >> 8;
> > +       addr_len[PAGE0_SWAUX_ADDR_23_16 - base] = (msg->address >> 16) &
> > +                                                 SWAUX_ADDR_19_16_MASK;
> > +       addr_len[PAGE0_SWAUX_ADDR_23_16 - base] |= (msg->request << 4) &
> > +                                                  SWAUX_CMD_MASK;
>
> optional nit: Probably you could get rid of the mask for the request.
> After all, you're storing it to a thing that's a byte (so bits above
> bit 7 will implicitly be masked) and you're left shifting by 4 (so
> bits 0-3 will implicitly be masked) so this just makes it uglier. ;-)
>
Fixed in v6.
PTAL.
> optional nit: In theory you could also get rid of the
> SWAUX_ADDR_19_16_MASK and if you really wanted to you could error
> check that the address wasn't bigger than 20-bits since giving an
> error for an invalid address would actually be better than silently
> masking it anyway...
>
Fixed in v6.
PTAL.
>
> > +       if (len && (request == DP_AUX_NATIVE_READ ||
> > +                   request == DP_AUX_I2C_READ)) {
> > +               /* Read from the internal FIFO buffer */
> > +               for (i = 0; i < len; i++) {
> > +                       ret = regmap_read(map, PAGE0_SWAUX_RDATA,
> > +                                         (unsigned int *)(buf + i));
>
> The cast to "unsigned int *" looks wrong to me. You can't just cast
> like this for a number of reasons. Go back to reading into a local
> variable and copy the byte into your buffer.
>
Previously I was not 100% sure about this change either.
Now I'm sure it is bad after some experiments.
In v6, I reverted to how this was handled in v3.
PTAL.

>
> Other than the regmap_read() this looks fine to me. If you send a v6
> with that fixed I'll plan to wait a day or two and then apply it with
> Sam's tags.
>
> -Doug