Fix reset.c, timer.c, setup.c file.
Signed-off-by: NMichal Simek <monstr@monstr.eu>

This lets a dts author flag a particular timer in the system as the system timer.
If the chosen node contains a "system-timer=<&foo>" entry than that handle will
be used to determine the system timer. In no such entry exists then the first
found timer will be used (current behaviour).
Signed-off-by: NPeter A. G. Crosthwaite <peter.crosthwaite@petalogix.com>
Signed-off-by: NMichal Simek <monstr@monstr.eu>

Disintegrate asm/system.h for Microblaze.  Not compiled.
Signed-off-by: NDavid Howells <dhowells@redhat.com>
cc: microblaze-uclinux@itee.uq.edu.au

Change report in bootlog:
Origin:
xlnx,xps-intc-1.00.a #0 at 0xc8000000, num_irq=6, edge=0x4
xlnx,xps-timer-1.00.a #0 at 0xc8004000, irq=2

New:
interrupt-controller #0 at 0xc8000000, num_irq=6, edge=0x4
system-timer #0 at 0xc8004000, irq=2
Signed-off-by: NMichal Simek <monstr@monstr.eu>

Signed-off-by: NMichal Simek <monstr@monstr.eu>

eprintk macro was used for printing early_printk
messages. Early console registration was changed
that's why this is not needed.
Signed-off-by: NMichal Simek <monstr@monstr.eu>

It is necessary to call generic function for irq finding.
The main reason is that this generic function calls irq_create_of_mapping
which can add some shift because of NO_IRQ.
Signed-off-by: NMichal Simek <monstr@monstr.eu>

Calling of_find_compatible_node instead of calling private code
which does the same.
Signed-off-by: NMichal Simek <monstr@monstr.eu>

Top bit is used as garbage and it must be clear
explicitly.
It is causing the problem with soft lookup code
because it checks delays which are long when
top bit is setup.
Signed-off-by: NMichal Simek <monstr@monstr.eu>

Variables and init_microblaze_timecounter should be static.

Warning log:
CHECK   arch/microblaze/kernel/timer.c
arch/microblaze/kernel/timer.c:41:14: warning: symbol 'freq_div_hz' was not declared. Should it be static?
arch/microblaze/kernel/timer.c:42:14: warning: symbol 'timer_clock_freq' was not declared. Should it be static?
arch/microblaze/kernel/timer.c:205:12: warning: symbol 'init_microblaze_timecounter' was not declared. Should it be static?
Signed-off-by: NMichal Simek <monstr@monstr.eu>

This converts the microblaze clocksources to use clocksource_register_hz/khz

CC: Michal Simek <monstr@monstr.eu>
CC: Thomas Gleixner <tglx@linutronix.de>
Tested-by: NMichal Simek <monstr@monstr.eu>
Signed-off-by: NJohn Stultz <johnstul@us.ibm.com>

New microblaze systems uses two buses. One for memories
and flashes and the second for low-speed peripherals
which can run on different CLK. This is the reason
why the kernel is trying to read clock-frequency
directly from node. If there is then the kernel will
work with it. If not then cpu CLK is used.
Signed-off-by: NMichal Simek <monstr@monstr.eu>

Microblaze little-endian toolchain exports __MICROBLAZEEL__
which is used in the kernel to identify little/big endian.

The most of the changes are in loading values from DTB which
is always big endian.

Little endian platforms are based on new AXI bus which has
impact to early uartlite initialization.
Signed-off-by: NMichal Simek <monstr@monstr.eu>

Device-tree BSP generates longer compatible properties
that's why we can remove extra/old one.
Signed-off-by: NMichal Simek <monstr@monstr.eu>

Signed-off-by: NJoe Perches <joe@perches.com>
Signed-off-by: NMichal Simek <monstr@monstr.eu>

Kernel needs two timers because of clocksource and clockevent.
It is better to show warning message directly on early
console if available. If it isn't available kernel log buffer
contains it.
Signed-off-by: NMichal Simek <monstr@monstr.eu>

Lower shifting values ensure that shifted 32bit counter
value doesn't exceed 64bit cycle variable too fast.
Signed-off-by: NMichal Simek <monstr@monstr.eu>

I had to comment sched_clock generic function because of broken toolchain.
It is fine grain timing.
Signed-off-by: NMichal Simek <monstr@monstr.eu>

It is the same counter as we use as free running one.
I would like to use it for ftrace.
Signed-off-by: NMichal Simek <monstr@monstr.eu>

This code path doesn't test any returned pointers for NULL, leading to a bad
kernel page fault if there's no timer/intc found.

Slightly better is to BUG(), but even better still would be a printk beforehand.
Signed-off-by: NJohn Williams <john.williams@petalogix.com>
Signed-off-by: NMichal Simek <monstr@monstr.eu>

This patch modifies parameter of microblaze_read() from 'void' to 'struct
clocksource *cs', which fixes compile warning for incompatible parameter type.
Signed-off-by: NColy Li <coly.li@suse.de>
Cc: Michal Simek <monstr@monstr.eu>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: NMichal Simek <monstr@monstr.eu>

Reviewed-by: NThomas Gleixner <tglx@linutronix.de>
Reviewed-by: NIngo Molnar <mingo@elte.hu>
Reviewed-by: NStephen Neuendorffer <stephen.neuendorffer@xilinx.com>
Acked-by: NJohn Linn <john.linn@xilinx.com>
Acked-by: NJohn Williams <john.williams@petalogix.com>
Signed-off-by: NMichal Simek <monstr@monstr.eu>