riscv_common: make thread_yield_higher IRQ compatible

[bergzand]

Contribution description

This should resolve the issues mentioned here. The thread_yield_higher function now checks if it is executed from interrupt context and just set the sched_context_switch_request flag if it is in interrupt context.

Testing procedure

These tests should work again:

• tests/event_wait_timeout
• tests/thread_flags
• tests/thread_flags_xtimer.

Issues/PRs references

Fixes regressions mentioned here

[jia200x]

I guess this should be backported

[aabadie]

This PR fixes the failing tests:

./dist/tools/compile_and_test_for_board/compile_and_test_for_board.py . hifive1b --jobs=6 --applications="tests/event_wait_timeout tests/thread_flags tests/thread_flags_xtimer"
INFO:hifive1b:Saving toolchain
INFO:hifive1b.tests/event_wait_timeout:Board supported: True
INFO:hifive1b.tests/event_wait_timeout:Board has enough memory: True
INFO:hifive1b.tests/event_wait_timeout:Application has test: True
INFO:hifive1b.tests/event_wait_timeout:Run compilation
INFO:hifive1b.tests/event_wait_timeout:Run test
INFO:hifive1b.tests/event_wait_timeout:Run test.flash
INFO:hifive1b.tests/event_wait_timeout:Success
INFO:hifive1b.tests/thread_flags:Board supported: True
INFO:hifive1b.tests/thread_flags:Board has enough memory: True
INFO:hifive1b.tests/thread_flags:Application has test: True
INFO:hifive1b.tests/thread_flags:Run compilation
INFO:hifive1b.tests/thread_flags:Run test
INFO:hifive1b.tests/thread_flags:Run test.flash
INFO:hifive1b.tests/thread_flags:Success
INFO:hifive1b.tests/thread_flags_xtimer:Board supported: True
INFO:hifive1b.tests/thread_flags_xtimer:Board has enough memory: True
INFO:hifive1b.tests/thread_flags_xtimer:Application has test: True
INFO:hifive1b.tests/thread_flags_xtimer:Run compilation
INFO:hifive1b.tests/thread_flags_xtimer:Run test
INFO:hifive1b.tests/thread_flags_xtimer:Run test.flash
INFO:hifive1b.tests/thread_flags_xtimer:Success
INFO:hifive1b:Tests successful

I'm running the full test suite, just to be sure nothing else is broken.

[bergzand]

I'm running the full test suite, just to be sure nothing else is broken.

Thanks, appreciated!

[aabadie]

Now I have this:

Failures during test:
- [tests/gnrc_ipv6_fwd_w_sub](tests/gnrc_ipv6_fwd_w_sub/test.failed)
- [tests/periph_timer_short_relative_set](tests/periph_timer_short_relative_set/test.failed)
- [tests/periph_wdt](tests/periph_wdt/test.failed)
- [tests/pkg_utensor](tests/pkg_utensor/test.failed)
- [tests/ps_schedstatistics](tests/ps_schedstatistics/test.failed)
- [tests/sys_architecture](tests/sys_architecture/test.failed)
- [tests/xtimer_periodic_wakeup](tests/xtimer_periodic_wakeup/test.failed)
- [tests/xtimer_usleep](tests/xtimer_usleep/test.failed)

So no new failures!

[aabadie]

Code changes are good and are fixing the failing tests. I didn't find any regression.

ACK

[jia200x]

could you please provide a backport against 2021.01-branch?

[aabadie]

could you please provide a backport against 2021.01-branch?

it might be difficult because this PR was opened after #15718 was merged.

[jia200x]

it might be difficult because this PR was opened after #15718 was merged.

Sorry, I thought this was related to #15736. Since nobody replied #15942 (comment), I assumed it was in ;)

[bergzand]

could you please provide a backport against 2021.01-branch?

Done, see #15961.

it might be difficult because this PR was opened after #15718 was merged.

It's an artisanal handcrafted backport :)