Sylwester,

Thanks for the review...

On Fri, Jan 11, 2013 at 2:12 PM, Sylwester Nawrocki
[off-list ref] wrote:

quoted

-       ret = of_alias_get_id(np, "i2c");
-       if (ret<  0) {
-               dev_err(&pdev->dev, "failed to get alias id, errno %d\n",
ret);
-               return ret;
-       }
-       pdev->id = ret;
+       pdev->id = -1;

I think assignments like this are illegal. At this point the device is
already registered and modifying pdev->id can cause issues at the core
code.

Good catch.  I think the old code is just as illegal since it was also
assigning to pdev->id, but I'm happy to spin the patch.

It might be better to have the bus number stored in struct pxa_i2c,
initialized with pdev->id value for non-dt and now with -1 for dt case.

I'll just init i2c->adap.nr before calling i2c_pxa_probe_dt().  Then
i2c_pxa_probe_dt() can adjust the number.  Hopefully this is OK.

I realize it is not something your patch is intended to deal with,
but since you're touching this code maybe it's worth to fix that issue
as well ?

Sure, I've spun it.  While doing this I realized a few other issues
relating to the ID number.  Specifically:

* We were using the id (as %u) when creating 'i2c->adap.name'.  This
probably gave a very nonsensical ID.  I'm just going to remove the
device number from the adap.name.  That matches what i2c-s3c2410.c
does.  As far as I can tell the adap.name isn't actually used for
much.

* The name was being used in request_irq().  I've changed this to be
like i2c-s3c2410 and use dev_name().  On my current board that means
that the name comes from the table in of_platform_populate().  ...but
that table is supposed to be going away.  If I remove that i2c parts
of that table I get names like '12c60000.i2c', which should be fine
for passing to request_irq().


-Doug