[U-Boot] [PATCH] ppc: transform init_sequence into a function.

[Joakim Tjernlund]

Scott Wood <scottwood at freescale.com> wrote on 2010/12/07 01:21:44:
>
> On Tue, 7 Dec 2010 01:07:30 +0100
> Joakim Tjernlund <joakim.tjernlund at transmode.se> wrote:
>
> > Scott Wood <scottwood at freescale.com> wrote on 2010/12/06 23:49:04:
> > >
> > > On Tue, 7 Dec 2010 09:36:40 +1100
> > > Graeme Russ <graeme.russ at gmail.com> wrote:
> > >
> > > > On Tue, Dec 7, 2010 at 8:33 AM, Scott Wood <scottwood at freescale.com> wrote:
> > > > > I think it's easier with the function pointers -- if you want to debug
> > > > > a hang in that phase of the boot, just have the loop print the address
> > > > > of each function before it calls it.
> > > >
> > > > I agree, but you can't print the address before you have console output.
> > >
> > > It's usually not too hard to hack something together, even if it's too
> > > early for normal console output -- but I'd expect most problems to be
> > > either before the initialization list, or after the console is working.
> > >
> > > > I notice that console_init_f() can be up to 13th in the list of initialisation
> > > > functions - How often is that the case? There seems to be a lot of SDRAM
> > > > initialisation prior to getting console output which, to me, seems a little
> > > > strange - surely console output can be achieved earlier (even if it is using
> > > > a hard-coded baud rate)
> > >
> > > I don't see "a lot of SDRAM initialization" -- there's
> > > adjust_sdram_tbs_8xx, but that's really just setting up a couple clock
> > > registers (and it's only for 8xx).  It's not the real SDRAM init.
> >
> > Scott, listen to yourself. You are proposing that one should turn the
> > code inside out and scan the map files just do this simple thing.
>
> It looks like you're the one turning this code inside out. :-)

Some days it feels so :)

>
> No need to scan anything yourself; let the tools (gdb/addr2line/etc) do
> it.

gdb yes, addr2line/etc no. Needing to resort to external tools
when gdb and a sensible code layout would do the trick? No thanks.

>
> > What is so valuable with func ptrs that you think it is worth it?
>
> As I said, I think it's at least as easy to debug the way it is.  And
> you admitted you found the new way uglier...

You haven't tried my way yet. Suppose we littered the code with
function ptrs, would still feel it is easier using addr2line/etc each and
every function call?

>
> Plus, maybe someday we'll get real section-list initfuncs.

One day that hasn't come fore years yet and may never come and probably
fixable with weak functions.