[PATCH] Fix OpenFirmware system_time

[andreasf]

OpenFirmware-based system_time implementation returned the number of milliseconds since boot multiplied by 1000. On my PowerMac G3 300 MHz this results in values around 1600000 and due to the limited resolution often to identical values on subsequent boots.

Thus system_time becomes unusable as a PRNG seed for TCP client ports since the server would be in the Timed Wait state, resulting in a connection timeout.

[andreasf]

proposed patch to fix line endings

[andreasf]

proposed patch: of_milliseconds would never fail

[andreasf]

optional patch: Besides milliseconds there appears to be a get-msecs

[andreasf]

proposed patch: Use get-time on rtc device

[andreasf]

The attached patch series first fixes line endings (feel free to run a tool yourself, there may be more files affected - helps keep patches readable), then fixes the return value of of_milliseconds, optionally makes system_time fall back to get-msecs and as main change uses get-time for system_time, as done later in the kernel real time clock implementation.

[andreasf]

Patch 4 withdrawn due to errors later on about missing close method (and with of_close commented out the kernel hanging immediately after load). The preceding patches are not affected and could be reviewed and applied.

[andreasf]

Sorry for the noise - breakage is no longer reproducible, must've been a side effect of my playings with the frame buffer (e.g., handle vs. package issues).

[andreasf]

Applied a new run of dos2unix.pl in hrev37265, obsoleting patch 0001.

[andreasf]

Applied patch 0002 in hrev37279 and patch 0004 in hrev38465.