Re: [PATCH] hw/i2c: flatten pca954x mux device

[Patrick Venture <venture@google.com>]

[-- Attachment #1: Type: text/plain, Size: 3673 bytes --]

On Wed, Feb 2, 2022 at 9:31 AM Philippe Mathieu-Daudé <f4bug@amsat.org>
wrote:

> On 2/2/22 17:40, Patrick Venture wrote:
>
> >     Philippe,
> >
> >     I0202 08:29:45.380384  6641 stream.go:31] qemu: child buses at
> >     "pca9546": "channel[*]", "channel[*]", "channel[*]", "channel[*]"
> >
> >     Ok, so that's interesting.  In one system (using qom-list) it's
> >     correct, but then when using it to do path assignment
> >     (qdev-monitor), it fails...
> >
> >     I'm not as fond of the name i2c-bus.%d, since they're referred to as
> >     channels in the datasheet.  If I do the manual name creation, can I
> >     keep the name channel or should I pivot over?
> >
> >     Thanks
> >
> >
> >             -- >8 --
> >             diff --git a/hw/i2c/i2c_mux_pca954x.c
> b/hw/i2c/i2c_mux_pca954x.c
> >             index f9ce633b3a..a9517b612a 100644
> >             --- a/hw/i2c/i2c_mux_pca954x.c
> >             +++ b/hw/i2c/i2c_mux_pca954x.c
> >             @@ -189,9 +189,11 @@ static void pca954x_init(Object *obj)
> >
> >                    /* SMBus modules. Cannot fail. */
> >                    for (i = 0; i < c->nchans; i++) {
> >             +        g_autofree gchar *bus_name =
> >             g_strdup_printf("i2c.%d", i);
> >             +
> >                        /* start all channels as disabled. */
> >                        s->enabled[i] = false;
> >             -        s->bus[i] = i2c_init_bus(DEVICE(s), "channel[*]");
> >             +        s->bus[i] = i2c_init_bus(DEVICE(s), bus_name);
> >                    }
> >                }
> >
> >             ---
> >
> >             (look at HMP 'info qtree' output).
> >
> >              >       }
> >              >   }
> >
> >             With the change:
> >             Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org
> >             <mailto:f4bug@amsat.org>>
> >             Tested-by: Philippe Mathieu-Daudé <f4bug@amsat.org
> >             <mailto:f4bug@amsat.org>>
> >
> >
> > Just saw your reply, and found a bunch of other non-spam in my spam
> > folder.  I sent the message to the anti-spam team, hopefully that'll
> > resolve this for myself and presumably others.
>
> Thanks. I suppose the problem is the amsat.org domain.
>

Yours aren't the only ones I've missed, but who knows.


>
> > I definitely see the same result with the qdev-monitor, but was really
> > surprised that the qom-list worked.  I'll explicitly set the name, and
> > i2c.%d is fine.  The detail that they're channels is not really
> > important to the end user presumably.
>
> I agree it is better to follow datasheets, thus I am fine if you
> change and use channel. How would it look like? "channel.0"?
> FYI qdev busses are described in docs/qdev-device-use.txt.
>

Thanks.  I went with i2c.%d in v2, since I figured it wasn't super
important.


>
> We should be able to plug a device using some command line
> such "-device i2c_test_dev,bus=channel.0,addr=0x55".
> I wonder how to select the base PCA9548 ...
>

So I have been working on that, and I have been running into a different
issue, but related.

/smbus[1]/i2c-bus/pca9546/i2c.0 works to add a device via command line.

However, if there are two pca9546s on that main bus.  So if i do:

/smbus[1]/i2c-bus/child[0]/i2c.0 it'll respond that there is no child[0],
but rather includes "pca9546, pca9546"


>
> Maybe we need to pass the PCA ID to pca954x_init(), so we can
> name "channel.2.0" for the 1st channel on the 2nd PCA?
>

It sounds like you're thinking about the same problem overall.


>
> Regards,
>
> Phil.
>

[-- Attachment #2: Type: text/html, Size: 5548 bytes --]