mirror of
https://github.com/AsahiLinux/u-boot
synced 2024-11-14 08:57:58 +00:00
69 lines
2.6 KiB
Text
69 lines
2.6 KiB
Text
|
IMPORTANT NOTE - read before defining CFG_USE_OSCCLK in your board
|
||
|
|
config file!!!
|
||
|
|
|
||
|
|
|
||
|
|
WARNING: Wrong settings of this parameter have the potential to
|
||
|
|
damage hardware by running the MBX's CPU at frequencies that exceed
|
||
|
|
it's rating and/or overdriving the it's SPLL!
|
||
|
|
|
||
|
|
|
||
|
|
Ramblings:
|
||
|
|
1) Motorola offered 12 different variants of the MBX, 6 823s and 6 860s.
|
||
|
|
2) Of these 12 variants, only 2 were entry level boards.
|
||
|
|
3) I believe that the 2 entry level boards were the only ones that
|
||
|
|
used OSCM clocking. I can't be completely certain of this at this
|
||
|
|
point.
|
||
|
|
4) Motorola never offered an MBX that ran faster than 50Mhz.
|
||
|
|
5) The 10, non-entry level boards, ran at 40Mhz.
|
||
|
|
6) The EXTCLK input has a minimum clock of 15Mhz for the 823/860.
|
||
|
|
7) Motorola no longer sells MBXs.
|
||
|
|
|
||
|
|
Based on this information, I can surmise that the default power-on
|
||
|
|
reset clocking was one of the following three options.
|
||
|
|
|
||
|
|
Multiplier SPLL Options
|
||
|
|
------------------------------------
|
||
|
|
513 OSCM is SPLL input
|
||
|
|
5 OSCM is SPLL input
|
||
|
|
1 EXTCLK is SPLL input
|
||
|
|
|
||
|
|
The forth option:
|
||
|
|
|
||
|
|
5 EXTCLK is SPLL input
|
||
|
|
|
||
|
|
is not possible on MBXs. This is because the minimum EXTCLK input
|
||
|
|
frequency is 15Mhz. 5 * 15Mhz = 75 Mhz. There was no variant that ran
|
||
|
|
above 50 Mhz.
|
||
|
|
|
||
|
|
The board I have borrowed definitely uses a multiplier of 1 for
|
||
|
|
EXTCLK and runs at 40Mhz. I even went so far as to put a scope on it.
|
||
|
|
|
||
|
|
One of the two default OSCM modes are most likely what was used on
|
||
|
|
the entry level boards to cheapen them by eliminating the external
|
||
|
|
crystal oscillator.
|
||
|
|
|
||
|
|
To add insult to injury, the stupid 860 PLPRCR register retains it's
|
||
|
|
multiplication factor through hard resets. You can't clear it out
|
||
|
|
because it is battery backed and once it is set wrong, it stays
|
||
|
|
wrong. The only way to reset it, so that it takes on it's default
|
||
|
|
multiplier is to disconnect all power including external, batteries,
|
||
|
|
as well discharging caps on the board. This precludes the fact that
|
||
|
|
your 860 may be quite DEAD by this time!
|
||
|
|
|
||
|
|
If you don't setup the multiplication factor for boards that use the
|
||
|
|
OSCM input, they won't run correctly, but at least they won't be
|
||
|
|
dead.
|
||
|
|
|
||
|
|
Addtionally, there is no good way to determine the clock input source
|
||
|
|
from CPU register data. The only way to deal with this is either hard
|
||
|
|
code it, determine the correct value with some rather NASTY timing
|
||
|
|
loops, or try to grok it from external data sources. Motorola
|
||
|
|
firmware opts for the NASTY timing loops, but needs to configure the
|
||
|
|
serial ports to do so.
|
||
|
|
|
||
|
|
|
||
|
|
You may have a legitimate need to define CFG_USE_OSCCLK if your
|
||
|
|
MBX8xx board is using the OSCM clocking mode.
|
||
|
|
|
||
|
|
You better know what you are doing here.
|