..\..\libcpu\arm\cortex-m3\context_rvds.S(207):
	warning: A1581W: Added 2 bytes of padding at address 0xd6

vector_undef and vector_dabt

When RT-Thread is running stand alone, it forgot to enable the
distributor of GIC.

This piece of code will enable the interrupt early before switching to
the first thread. Although it is harmless, but not prefect.

Unit of "byte" is more intuitive than "long".

As the svc stack is the stack of threads, there is no need to allocate a
separate stack for the startup. Reuse the IRQ stack should be OK.

Tested on rm48 board.

for 2013-02-20 aozima commmit "port for gcc", but the commit is
not tested,and the kernel is breakdown in context_gcc.S, the file
is a copy from cortex-m3,but not port for cortex-m0, so i complete
this port for aozima, test it in stm32f0discovery board ,
and it works fine.

We could not handle any traps except IRQ/FIQ.

format the s3c24x0 serial.c and mini2440 rtconfig.py

Signed-off-by: Nvisitor83 <root@wolflouis.(none)>

You need to turn on RT_VFP_LAZY_STACKING in rtconfig.h. By default, RTT
will turn on VFP for all threads and stack all the VFP registers. When
doing lazy stacking, VFP will only be turned on for the thread who
issued VFP instructions.

Currently, if a thread turned on VFP, it cannot be turned off. RTT will
never know what time the thread doesn't need VFP any more. The thread
might could turn off the VFP for it self in proper time.

VFP instructions only have IA(Increment After)/DB(Decrement Before)
mode. To keep consistency, just change STM/LDM to DB/IA accordingly.