sys/xtimer: _xtimer_now64(): fix irq_disable() return value type

[kaspar030]

Contribution description

Small mistake, large impact. Use of the wrong type for the return value of irq_disable() basically chopped off part of the ISR state.

On the fe310, this lead to a trap right after xtimer ISR exit, sometimes.

Testing procedure

Fix should be obvious.

Issues/PRs references

Fixes #13109.

[emmanuelsearch]

Nice catch @kaspar030 !

[fjmolinas]

Tested on hifive1b:

• master

2020-01-22 12:55:05,965 # ATE0-->ATE0
2020-01-22 12:55:05,982 # OK
2020-01-22 12:55:06,140 # AT+BLEINIT=0-->OK
2020-01-22 12:55:06,295 # AT+CWMODE=0-->OK
2020-01-22 12:55:06,296 # 
2020-01-22 12:55:06,304 # Help: Press s to start test, r to print it is ready
s
2020-01-22 12:55:23,420 # START
2020-01-22 12:55:23,426 # main(): This is RIOT! (Version: 2020.04-devel-19-g97a71b-HEAD)
2020-01-22 12:55:23,430 # Running test 5 times with 7 distinct sleep times
2020-01-22 12:55:23,441 # Unhandled trap:
2020-01-22 12:55:23,443 #   mcause: 0x00000001
2020-01-22 12:55:23,444 #   mepc:   0x200104e8
2020-01-22 12:55:23,446 #   mtval:  0x200104e8
2020-01-22 12:55:23,448 # *** RIOT kernel panic:
2020-01-22 12:55:23,450 # Unhandled trap
2020-01-22 12:55:23,450 # 
2020-01-22 12:55:23,451 # *** halted.
2020-01-22 12:55:23,451 # 

-pr

Bench Clock Reset Complete
2020-01-22 12:56:06,256 # 
2020-01-22 12:56:06,305 # ATE0-->ATE0
2020-01-22 12:56:06,322 # OK
2020-01-22 12:56:06,479 # AT+BLEINIT=0-->OK
2020-01-22 12:56:06,635 # AT+CWMODE=0-->OK
2020-01-22 12:56:06,636 # 
2020-01-22 12:56:06,644 # Help: Press s to start test, r to print it is ready
s
2020-01-22 12:56:10,716 # START
2020-01-22 12:56:10,722 # main(): This is RIOT! (Version: 2020.04-devel-71-g885bb84-pr-13182)
2020-01-22 12:56:10,727 # Running test 5 times with 7 distinct sleep times
2020-01-22 12:56:10,741 # Slept for 10102 us (expected: 10000 us) Offset: 102 us
2020-01-22 12:56:10,795 # Slept for 49988 us (expected: 50000 us) Offset: -12 us
2020-01-22 12:56:10,809 # Slept for 10223 us (expected: 10234 us) Offset: -11 us
2020-01-22 12:56:10,870 # Slept for 56763 us (expected: 56780 us) Offset: -17 us
2020-01-22 12:56:10,886 # Slept for 12116 us (expected: 12122 us) Offset: -6 us
2020-01-22 12:56:10,989 # Slept for 98755 us (expected: 98765 us) Offset: -10 us
2020-01-22 12:56:11,068 # Slept for 74982 us (expected: 75000 us) Offset: -18 us
2020-01-22 12:56:11,081 # Slept for 9980 us (expected: 10000 us) Offset: -20 us
2020-01-22 12:56:11,135 # Slept for 49988 us (expected: 50000 us) Offset: -12 us
2020-01-22 12:56:11,150 # Slept for 10224 us (expected: 10234 us) Offset: -10 us
2020-01-22 12:56:11,210 # Slept for 56762 us (expected: 56780 us) Offset: -18 us
2020-01-22 12:56:11,226 # Slept for 12116 us (expected: 12122 us) Offset: -6 us
2020-01-22 12:56:11,329 # Slept for 98755 us (expected: 98765 us) Offset: -10 us
2020-01-22 12:56:11,408 # Slept for 74981 us (expected: 75000 us) Offset: -19 us

[fjmolinas]

ACK! Thanks for investigating this and providing the fix @kaspar030, I'll run all tests on hifive1b ass good measure, but the fix is obvious.

[fjmolinas]

I'll use the results in #13086 as a comparison point.

[fjmolinas]

GO!, still waiting on all tests results but the change is valid on its own.

[fjmolinas]

Backport provided in #13183

[aabadie]

Nice catch @kaspar030, thanks!

[miri64]

As a reminder for our offline discussion: maybe having a Coccinelle check for that in the CI is a good idea.

[fjmolinas]

Re0ran the tests on hifive1b and got pretty much the same results as before:

ERROR:hifive1b:Tests failed: 6
Failures during test:
- [tests/cpp11_mutex](tests/cpp11_mutex/test.failed)
- [tests/gnrc_ipv6_fwd_w_sub](tests/gnrc_ipv6_fwd_w_sub/test.failed)
- [tests/pkg_utensor](tests/pkg_utensor/test.failed)
- [tests/ps_schedstatistics](tests/ps_schedstatistics/test.failed)
- [tests/xtimer_periodic_wakeup](tests/xtimer_periodic_wakeup/test.failed)
- [tests/xtimer_usleep](tests/xtimer_usleep/test.failed)