Re: [PATCH v3 1/1] ACPI: scan: Use unique number for instance_no

[Rafael J. Wysocki]

On Mon, Mar 22, 2021 at 4:02 PM Andy Shevchenko
<andy.shevchenko@xxxxxxxxx> wrote:
>
> On Mon, Mar 22, 2021 at 4:57 PM Rafael J. Wysocki <rafael@xxxxxxxxxx> wrote:
> >
> > On Fri, Mar 19, 2021 at 8:21 PM Andy Shevchenko
> > <andriy.shevchenko@xxxxxxxxxxxxxxx> wrote:
> > >
> > > The decrementation of acpi_device_bus_id->instance_no
> > > in acpi_device_del() is incorrect, because it may cause
> > > a duplicate instance number to be allocated next time
> > > a device with the same acpi_device_bus_id is added.
> > >
> > > Replace above mentioned approach by using IDA framework.
>
> ...
>
> > > +       result = ida_simple_get(&acpi_device_bus_id->instance_ida, 0, 255, GFP_KERNEL);
> >
> > This is ida_alloc_range(ida, start, (end) - 1, gfp), so I think it
> > should be 256 above, instead of 255.
>
> Ah, good catch!
>
>
> > While at it, though, there can be more than 256 CPU devices easily on
> > contemporary systems, so I would use a greater number here.  Maybe
> > 4096 and define a symbol for it?
>
> I was thinking about it, but there is a problem with the device name,
> since it will break a lot of code,

What problem is there?

> And taking into account that currently we don't change the behaviour
> it is good enough per se as a fix.
>
> That said, we may extend by an additional patch with a logic like this:
>
> res = ida_get(4096)
> if (res < 0)
>   return res;
> if (res >= 256)
>  use %04x
> else
>  use %02x
>
> Would it make sense to you?

I'm not sure why not to always use %02x ?  It doesn't truncate numbers
longer than 2 digits AFAICS.