[Arm-netbook] I Have An Possible Chance To Meet An ARM Senior Manager

[luke.leighton]

On Mon, Oct 8, 2012 at 1:43 PM, Alexander Stephen Thomas Ross
<arm-netbook at aross.me> wrote:
> Re: BOIS, uboot.
> Ok I see that I am wrong.

 it would be great if you were right!  but the sheer diversity (viz:
chaos) in the ARM world makes it impossible to "impose" any kind of
order (BIOS).

 i think someone mentioned that ARM, internally, had an initiative
where the AMBA Bus would have "identification information" on it,
similar to PCI (and USB).  in this way, the theory went, you'd be able
to write dynamic drivers similar to how USB and PCI detect the devices
and load drivers conformant to standards for that particular device,
even though it was on-chip.

 they then put it out to the various SoC Fabless Semiconductor
houses.... and they FUCKED IT UP!!  none of them could implement the
addressing correctly or agree or implement the data format correctly;
many of the companies simply zero'd out the memory area reserved for
identification information.

 and you know what?  they're absolutely right not to have bothered.

 unfortunately, the logic goes like this: "it's our chip. we're
implementing only one of them.  unlike PCI and USB, nobody is going to
be able to add extra AMBA Bus peripherals onto *our* chip's silicon,
because it's silicon!  if there are going to be any extra devices,
we'll add them, and *we* will know how to program them.  ergo, why
should we care about this AMBA 'id' standard?  that's a software
problem.  software can sort it out".

 ... except of course that means that if every ARM SoC vendor does the
exact same thing, you have the dog's dinner situation where linus
torvalds, on calling a meeting of all the Architecture Teams, looks at
the assembled masses, and says "over half of you are for ARM CPUs.
why are you all here!  get yourselves organised!  go away and next
time send only one representative!" which should tell you everything
you need to know about how linus completely fails to understand how
ARM CPUs are made and why the situation is such a mess at the
device-driver level.

> Thank you all for correcting me. Got any more
> good reading for me?

 oof.

 ok, correct me if i'm wrong on this, but i get the feeling that what
you're actually asking is, "i can see this is is a completely
dis-satisfactory situation.  what i would like to know is this: i'd
like to read more about what the *solution* is, so that i can feel
better"

 ... would that be about right, at all? :)

 because if so, my answer would be: look, don't dwell on it.  we
_know_ that there's no BIOS-like solution, that u-boot makes things
worse not better, that Russell King and the other ARM Linux Kernel
developers are getting severely overloaded to the point where many of
them are beginning to give up BUT what we need is to find a solution
that ties things together *despite* all of the disparity.

 the only thing that we have quotes control quotes over is at a higher
level.  defining hardware standards - groupings of interfaces into
common "buses" - that products can be designed around.

 other than that: if you really did mean "is there any more reading
out there", i've posted occasionally on slashdot and on here, but i
don't mind answering questions again, it's always "new" then.

 so, ask away, ok?

/peace

l.