Age | Commit message (Collapse) | Author |
|
* max-au/erts/dirty_scheduler_shutdown/PR-2172/OTP-15690:
erts: release dirty runqueue lock before entering endless loop when BEAM is shutting down
|
|
shutting down
This patch fixes a problem happening when BEAM is shutting down. It is possible for a dirty scheduler to take the lock, and keep it, when the system is shutting down. It may also happen that a normal scheduler decides to schedule some dirty job (example is major garbage collection that results in migrating the process into dirty CPU queue), and hangs trying to take the lock that will never be released.
To fix the problem, either release the lock before entering endless wait loop, or reverse the order in which schedulers are stopped. Either fix works, and, of course, it works even better to apply both.
|
|
* sverker/heart-nice-exit/OTP-15599:
erts: Avoid heart killing a nicely exiting emulator
|
|
Symptom:
Heart kills exiting emulator before is has flushed all ports
and with HEART_KILL_SIGNAL=SIGABRT it may also produce
unnecessary core dumps from doing init:reboot() for example.
Problem:
Heart port is closed together with all the others in handle_reap_ports()
which is detected by heart OS process.
Solution 1:
Leave the heart port alone in handle_reap_ports() and let it be closed
by OS when emulator exists. It doesn't need to be flushed anyway.
Solution 2:
When heart OS process gets EOF on connection let it wait max 5 seconds
for emulator process to self terminate before trying to kill it.
|
|
* lukas/erts/scheduler-pollset-fixes/OTP-15538:
erts: Fix getting of poll events on linux >= 4.15.0
erts: Use reduction based polling for starved poll-set
erts: Fix pollset test cases
|
|
When the schedulers never go to sleep (and thus never polls)
it may be that the fds in schedulers poll-sets are never polled.
Before this commit, this was solved by starting a timer when an
overload was detected. This had issues as overloads were not always
detected in time. So this commit reverts to the pre OTP-21 behaviour
so keep a global counter makes that the poll is called when it should.
|
|
* rickard/dirty_scheduler_collapse/OTP-15509:
Fix bug causing dirty scheduler sleeper list inconsistency
|
|
rickard/dirty_scheduler_collapse/maint-21/OTP-15509
* rickard/dirty_scheduler_collapse/OTP-15509:
Fix bug causing dirty scheduler sleeper list inconsistency
|
|
|
|
At start of the VM a poll-set that the schedulers
will check is created where fds that have triggered
many (at the moment, many means 10) times without
being deselected inbetween. In this scheduler specific
poll-set fds do not use ONESHOT, which means that the
number of syscalls goes down dramatically for such fds.
This pollset is introduced in order to handle fds that
are used by the erlang distribution and that never
change their state from {active, true}.
This pollset only handles ready_input events,
ready_output is still handled by the poll threads.
During overload, polling the scheduler poll-set is done
on a 10ms timer.
|
|
The poll thread does a lot of waking up and then going
back to sleep. A large part of the waking up is managing
thread progress and a large part of that was using thread
specific data to get the thread progress data pointer.
With this refactor the tpd is passed to each of the functions
which greatly decreases the number of ethr_get_tsd calls
which in turn halves the CPU usage of the poller thread in
certain scenarios.
|
|
process_info(self(), ...)
It is possible that a process has to yield before completing process_info BIF when it runs out of reductions. If this BIF is called by the process itself, it does not send a signal but executes in the context of a process. If it has to yield, it turns F_LOCAL_SIGS_ONLY flag on, which means new signals won't be fetched from the outer message queue.
When the same process needs to execute dirty system code (e.g. dirty GC) it has to be run on a dirty scheduler. However signals enqueued into outer queue cause it to be rescheduled on a normal scheduler. F_LOCAL_SIGS_ONLY prevent outer queue signals delivery, creating an endless rescheduling loop.
This commit disengages F_LOCAL_SIG_ONLY if process needs to execute dirty code in order to complete signal delivery and allow process to be moved to dirty run queue.
|
|
* maint-21:
Updated OTP version
Update release notes
Update version numbers
Fix missing 'in' trace events during 'running' trace
|
|
'in' trace events could be lost when a process had to be
rescheduled on another scheduler type (normal <-> dirty).
|
|
A lot of erts internal messages used behind APIs to create
non-blocking calls, e.g. port_command, would cause the seq_trace
token to be cleared from the caller when it should not.
This commit fixes that and adds asserts that makes sure
that all messages sent have to correct token set.
Fixes: ERL-602
|
|
* john/erts/fix-dirty-reschedule-bug/OTP-15154:
Move to a dirty scheduler even when we have pending system tasks
|
|
into maint-20
* john/erts/fix-process-schedule-after-free/OTP-15067/ERL-573:
Don't enqueue system tasks if target process is in fail_state
Fix erroneous schedule of freed/exiting processes
Fix deadlock in run queue evacuation
Fix memory leak of processes that died in the run queue
|
|
When a system task was enqueued on a process between being
scheduled and the check altered in this commit, we'd run dirty
code on a normal scheduler as the RUNNING_SYS flag wasn't set
and we wouldn't migrate back.
This change migrates us to a dirty scheduler instead, which will
immediately bounce us back to a normal scheduler where
RUNNING_SYS will be set appropriately.
This is caught fairly reliably by
process_SUITE:system_task_failed_enqueue on machines with a lot of
cores.
|
|
into john/erts/merge-OTP-15067
|
|
The fail state wasn't re-checked in the state change loop; only
the FREE state was checked. In addition to that, we would leave
the task in the queue when bailing out which could lead to a
double-free.
This commit backports active_sys_enqueue from master to make it
easier to merge onwards.
|
|
When scheduled out, the process was never checked for the FREE state
before rescheduling, which meant that a system task could sneak in
and cause a double-free later on.
|
|
* sverker/system-profile-bug/OTP-15085:
erts: Fix bug in system_profile
|
|
If scheduler_data is not set correctly on normal schedulers
the code in erts_schedule_time_break and possibly others
will trigger asserts.
|
|
|
|
|
|
* sverker/system-profile-bug/OTP-15085:
erts: Fix bug in system_profile
|
|
seen to cause redundant {profile,_,active,_,_} messages
when process is terminating.
|
|
* rickard/delete_process_schedule/OTP-15081:
Do not hold runq lock while deleting a process
|
|
|
|
|
|
|
|
|
|
If no message/signal is sent (to same destination)
then monitor signal is flushed when process is scheduled out.
|
|
Improve memory instrumentation
OTP-15024
OTP-14961
|
|
This commit replaces the old memory instrumentation with a new
implementation that scans carriers instead of wrapping
erts_alloc/erts_free. The old implementation could not extract
information without halting the emulator, had considerable runtime
overhead, and the memory maps it produced were noisy and lacked
critical information.
Since the new implementation walks through existing data structures
there's no longer a need to start the emulator with special flags to
get information about carrier utilization/fragmentation. Memory
fragmentation is also easier to diagnose as it's presented on a
per-carrier basis which eliminates the need to account for "holes"
between mmap segments.
To help track allocations, each allocation can now be tagged with
what it is and who allocated it at the cost of one extra word per
allocation. This is controlled on a per-allocator basis with the
+M<S>atags option, and is enabled by default for binary_alloc and
driver_alloc (which is also used by NIFs).
|
|
proc->run_queue detected as uninitialized by valgrind
but seems harmless in practice as it's not used for proxy processes.
"Bug" introduced in OTP-17 by ca0425c6ff85262bc15367f5fd9cbc51cde52b20
and made worse (but still harmless) in master for OTP-21
at fbb10ebc4a37555c7ea7f99e14286d862993976a.
|
|
This may be of interest in crash dumps and allows the upcoming
allocation tagging feature to track allocations on a per-NIF basis.
Note that this is only updated when user code calls a NIF; it's not
altered when the emulator calls NIFs during code upgrades or
tracing.
|
|
* rickard/process_info/OTP-14966:
New process_info() implementation using signals
|
|
|
|
* rickard/signals/OTP-14589:
Fix seq trace
Fix bad assert
|
|
* john/erts/bwt-wt-dirty-schedulers/OTP-14959:
Add +sbwt/+swt analogues for dirty schedulers
|
|
|
|
Sharing these settings for all schedulers can degrade performance,
so it makes sense to be able to configure them separately.
This also changes the default busy-wait time to "short" for both
kinds of dirty schedulers.
|
|
|
|
Communication between Erlang processes has conceptually always been
performed through asynchronous signaling. The runtime system
implementation has however previously preformed most operation
synchronously. In a system with only one true thread of execution, this
is not problematic (often the opposite). In a system with multiple threads
of execution (as current runtime system implementation with SMP support)
it becomes problematic. This since it often involves locking of structures
when updating them which in turn cause resource contention. Utilizing
true asynchronous communication often avoids these resource contention
issues.
The case that triggered this change was contention on the link lock due
to frequent updates of the monitor trees during communication with a
frequently used server. The signal order delivery guarantees of the
language makes it hard to change the implementation of only some signals
to use true asynchronous signaling. Therefore the implementations
of (almost) all signals have been changed.
Currently the following signals have been implemented as true
asynchronous signals:
- Message signals
- Exit signals
- Monitor signals
- Demonitor signals
- Monitor triggered signals (DOWN, CHANGE, etc)
- Link signals
- Unlink signals
- Group leader signals
All of the above already defined as asynchronous signals in the
language. The implementation of messages signals was quite
asynchronous to begin with, but had quite strict delivery constraints
due to the ordering guarantees of signals between a pair of processes.
The previously used message queue partitioned into two halves has been
replaced by a more general signal queue partitioned into three parts
that service all kinds of signals. More details regarding the signal
queue can be found in comments in the erl_proc_sig_queue.h file.
The monitor and link implementations have also been completely replaced
in order to fit the new asynchronous signaling implementation as good
as possible. More details regarding the new monitor and link
implementations can be found in the erl_monitor_link.h file.
|
|
* rickard/dirty-work-gone/OTP-14978:
Reschedule on ordinary scheduler if dirty work is gone
|
|
|
|
* rickard/remove-approx-started/OTP-14975:
Remove process start time for crash dumps
|
|
Bug introduced in commit fbb10ebc4a37555c7ea7f99e14286d862993976a
|
|
|