nano_timer will never expired @ARC @ARM when started and tested from ISR

[zephyrbot]

Reported by Sharron LIU:

Configuration

arm, arc
Arduino 101, Arduino Due
nano
zephyr-kernel-1.1.0-160324-ca662c1

Bug detailed description

Nanokernel object timer is never expired, when invoked from ISR:
nano_timer_start()
nano_timer_test()
Refer to line #67 #68 in "src/main.c" from attached test cases.
NOTE: The issue is observed @arc and @arm, not observed @x86.
NOTE: If comment out above two lines, the test case passed.

Testcase to Reproduce

Attached "test_bat.tgz"

Steps to Reproduce

1. cd zephyr
   source zephyr-env.sh
2. download and untar attached test cases "test_bat.tgz"
   zephyr$ tar -zxvf test_bat.tgz
3. cd test_bat
   @arc: make pristine;make BOARD=arduino_101_sss ARCH=arc flash
   @arm: make pristine;make BOARD=arduino_due ARCH=arm;bossac --port=ttyACM0 -e -w -v -b outdir/zephyr.bin
4. verify from UART output test hang at "nano_timer: isr"

Expected Result

UART output:
tc_start()
...
ROJECT EXECUTION SUCCESSFUL

Actual Result

test hang at "nano_timer: isr".
then 20s later board reset (watchdog?).

User Impact

nano_timer not function.

(Imported from Jira ZEP-147)

[zephyrbot]

by Sharron LIU:

• test_bat.tgz (size: 2509)

[zephyrbot]

by Benjamin Walsh:

Is it possible that your ISR is running at the same or higher priority than the ISR of the timer driver on those platforms ? If so, the timer interrupt will never fire and the system clock will never get updated in the kernel.

On the quark_se ARC, there are only two levels of interrupts as well. I'm pretty sure they run the timer as a FIRQ (priority 0), so in that case, if you install your ISR as a FIRQ, the timer ISR will never fire either.

[zephyrbot]

by Sharron LIU:

You're right. I update the test code slightly, to allow timer IRQ raising, now nano_timer is tested from ISR and expired. Thanks.
Now the whole case passed @arc and @arm.

[zephyrbot]

by Mark Linkmeyer:

Correcting the priority field