Re: ensuring a machine's buses have unique names

[Markus Armbruster <armbru@redhat.com>]

BALATON Zoltan <balaton@eik.bme.hu> writes:

> On Wed, 22 Sep 2021, Markus Armbruster wrote:
>> BALATON Zoltan <balaton@eik.bme.hu> writes:
>>> To me it looks like device code can't really set a globally unique
>>> name on creating the bus without getting some help from upper
>>> levels. So maybe naming busses should be done by qdev (or whatever is
>>> handling this) instead of passing the name as an argument to
>>> qbus_create or only use that name as a unique component within the
>>> device and append it to a unique name for the device.
>>
>> Have you read the whole thread?  qdev does come up with a name when
>
> No I haven't. This just got my attention because I'm responsible for
> adding ati-vga.ddc and sm501.i2c and some ppc440 stuff so I was
> wondering what could I do better bur otherwise I did not check the
> whole thread so just ignore what I said if it's not useful in this
> context.

I wouldn't call it "not useful".

What you could do better in future code: pass a null bus name.  This is
not obvious.  As Peter noted "we created a family of easy-to-misuse
APIs".

[...]