Skip to contentSkip to conversation

Documentation/rv: Explain epoll and aborted sleeps

RFC PATCH In review Chao Liu 3 participants 3 messages Jul 29 – Aug 4

Subsystems (3)
lore
Share & tools
lore

Chao Liu

author
#0 lore raw

RFC PATCH v2 1/1 Documentation/rv: Explain epoll and aborted sleeps

epoll_wait() is a valid sleeping reason for real-time tasks because it uses PI-aware locking, but the rtapp sleep monitor documentation only discusses clock_nanosleep() and futexes. Document it.

ABORT_SLEEP represents a task restoring TASK_RUNNING before entering the scheduler. Since the task does not actually block, it becomes runnable again without a wakeup sequence unsafe for real-time. Document this behavior.

This RFC is based on Nam Cao's pending "rv: rtapp monitor update" v2 series:

https://lore.kernel.org/r/[email protected]

---
Changes in v2:
- Explain why epoll_wait() is RT-safe in terms of PI-aware locking.
- Explain that ABORT_SLEEP does not require an RT-unsafe wakeup sequence.
- Update the commit message accordingly.

Link to v1:
https://lore.kernel.org/linux-trace-kernel/[email protected]/

 Documentation/trace/rv/monitor_rtapp.rst | 6 ++++++
 1 file changed, 6 insertions(+)
Documentation/trace/rv/monitor_rtapp.rst +6 -0
index 238b59395ff5..b95994ade14a 100644
@@ -67,6 +67,8 @@ thread to sleep for one of the following reasons:
     variables as safe for real-time. As an alternative, the librtpi library
     exists to provide a conditional variable implementation that is correct for
     real-time applications in Linux.
+  - Real-time thread waiting for events using `epoll_wait`, which is a
+    real-time-safe syscall for sleeping as it uses PI-aware locking.
 
 Beside the reason for sleeping, the eventual waker should also be
 real-time-safe. Namely, one of:
@@ -114,6 +116,10 @@ The monitor's specification is::
   ALLOWLIST = BLOCK_ON_RT_MUTEX
            or FUTEX_LOCK_PI
 
+`ABORT_SLEEP` represents a task restoring its state to `TASK_RUNNING` before
+entering the scheduler. In this case, the task does not actually block, so the
+task is back to runnable without any wakeup sequence unsafe for real-time.
+
 Beside the scenarios described above, this specification also defines an allow list
 to handle some special cases:
 

base-commit: 248951ddc14de84de3910f9b13f51491a8cd91df
prerequisite-patch-id: d8b6c952a954662852e6b0684e0a3386eaab7f41
prerequisite-patch-id: 940f7637aeabe4ba15a09c048c6f48138406f332
prerequisite-patch-id: b99812692691f76d825778e01a2e771e058febd6
prerequisite-patch-id: 512c300dce2cb74ac9cc007f9234aa507ef5c008

Signature · 1 line
2.54.0
maintainer Reviewed-by on v2 1/1
#1 lore raw
Quoted 9 lines · Chao Liu

epoll_wait() is a valid sleeping reason for real-time tasks because it uses PI-aware locking, but the rtapp sleep monitor documentation only discusses clock_nanosleep() and futexes. Document it.

ABORT_SLEEP represents a task restoring TASK_RUNNING before entering the scheduler. Since the task does not actually block, it becomes runnable again without a wakeup sequence unsafe for real-time. Document this behavior.

Looks good, thanks! Please next time you send a patch, even if it's an RFC, put the extra comments after the --- . This way I can pull it directly without changing the commit message because git am ignores everything after the --- . Also, no big deal, but for simple patches like this one, and even after getting comments, there's no need to mark as RFC.

I'd appreciate an ack from Nam, then I might be able to squeeze this before the merge window together with the last fixes.

Thanks,
Gabriele

Quoted 52 lines

This RFC is based on Nam Cao's pending "rv: rtapp monitor update" v2 series:

https://lore.kernel.org/r/[email protected]

---
Changes in v2:
- Explain why epoll_wait() is RT-safe in terms of PI-aware locking.
- Explain that ABORT_SLEEP does not require an RT-unsafe wakeup sequence.
- Update the commit message accordingly.

Link to v1:
https://lore.kernel.org/linux-trace-kernel/[email protected]/

 Documentation/trace/rv/monitor_rtapp.rst | 6 ++++++

 1 file changed, 6 insertions(+)
Documentation/trace/rv/monitor_rtapp.rst +0 -0
index 238b59395ff5..b95994ade14a 100644
@@ -67,6 +67,8 @@ thread to sleep for one of the following reasons:

     variables as safe for real-time. As an alternative, the librtpi library exists to provide a conditional variable implementation that is correct for
     real-time applications in Linux.
+  - Real-time thread waiting for events using epoll_wait, which is a +    real-time-safe syscall for sleeping as it uses PI-aware locking.
 Beside the reason for sleeping, the eventual waker should also be real-time-safe. Namely, one of:

Documentation/trace/rv/monitor_rtapp.rst +0 -0
@@ -114,6 +116,10 @@ The monitor's specification is::

   ALLOWLIST = BLOCK_ON_RT_MUTEX
            or FUTEX_LOCK_PI

+ABORT_SLEEP represents a task restoring its state to TASK_RUNNING before +entering the scheduler. In this case, the task does not actually block, so the

Documentation/trace/rv/monitor_rtapp.rst +2 -0
+task is back to runnable without any wakeup sequence unsafe for real-time.
+

 Beside the scenarios described above, this specification also defines an allow list
 to handle some special cases:

base-commit: 248951ddc14de84de3910f9b13f51491a8cd91df
prerequisite-patch-id: d8b6c952a954662852e6b0684e0a3386eaab7f41
prerequisite-patch-id: 940f7637aeabe4ba15a09c048c6f48138406f332
prerequisite-patch-id: b99812692691f76d825778e01a2e771e058febd6
prerequisite-patch-id: 512c300dce2cb74ac9cc007f9234aa507ef5c008

Reviewed-by
#2 lore raw
Quoted 2 lines · Gabriele Monaco

I'd appreciate an ack from Nam, then I might be able to squeeze this before the merge window together with the last fixes.