Re: [PATCH v3 1/4] ACPI: Add acpi_pr_<level>() interfaces

[Toshi Kani <toshi.kani@hp.com>]

On Thu, 2012-07-26 at 15:50 -0600, Bjorn Helgaas wrote:
> On Thu, Jul 26, 2012 at 3:43 PM, Joe Perches <joe@perches.com> wrote:
> > On Thu, 2012-07-26 at 15:37 -0600, Bjorn Helgaas wrote:
> >>     PNP0C01:00: new device for \_SB_.PCI0.ISA_.MBIO
> >>
> >> I fiddled with this a while ago; it would look something like this:
> > []
> >> +static noinline_for_stack
> >> +char *acpi_name_string(char *buf, char *end, acpi_handle handle,
> >> +                    struct printf_spec spec, const char *fmt)
> >> +{
> >> +     acpi_status status;
> >> +     struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER, NULL };
> >> +     u32 type = ACPI_SINGLE_NAME;
> >> +     char *p = buf;
> >> +
> >> +     if (fmt[0] == 'A')
> >> +             type = ACPI_FULL_PATHNAME;
> >
> > maybe if (fmt[1] == 'f')
> >
> >> @@ -982,6 +1007,9 @@ char *pointer(const char *fmt, char *buf, char *end, void *ptr,
> >>       }
> >>
> >>       switch (*fmt) {
> >> +     case 'A':
> >> +     case 'a':
> >> +             return acpi_name_string(buf, end, ptr, spec, fmt);
> >
> > There are only so many letters, it might be better to
> > just use 'a' and another 'f' after that if necessary
> > for "full".
> >
> > And of course that should be #ifdef'd too
> 
> Yes.  I'm hesitant about this approach in general, because I don't
> think printing the ACPI path is something we should be doing often.
> It's not like a struct resource or a MAC address, where there are
> dozens or hundreds of users.  I really think we should only print ACPI
> paths in one or two places, so adding a %p extension would waste a
> letter and encourage the wrong behavior.

That's a good point.  I agree.  So, let's continue to use
acpi_pr_<level>() for printing ACPI device path.  The use of this
interface is limited anyway. 

Thanks,
-Toshi