We don't need to send the number of voluntary context switches (nvcsw)
from BPF to user-space, as this information is already accessible in
user-space via procfs. Sending this data would only create unnecessary
overhead for schedulers that don't require it, and those that do can
easily retrieve it through procfs.
Therefore, drop this metric from scx_rustland_core and change
scx_rustland implementing an interactive task classifier fully in the
user-space part of the scheduler.
Also drop some options that are not provide any significant benefit
(also in preparation of a bigger refactoring to define a better API for
the user-space framework).
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
Explicitly import timespec to fix the following potential build error:
error[E0422]: cannot find struct, variant or union type `timespec` in this scope
--> src/bpf.rs:365:35
|
365 | sched_ss_repl_period: timespec {
| ^^^^^^^^ not found in this scope
|
help: consider importing this struct
|
6 + use libc::timespec;
|
This fixes issue #469.
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
sched_ext is about to be merged upstream. There are some compatibility
breaking changes and we're making the current sched_ext/for-6.11
1edab907b57d ("sched_ext/scx_qmap: Pick idle CPU for direct dispatch on
!wakeup enqueues") the baseline.
Tag everything except scx_mitosis as 1.0.0. As scx_mitosis is still in early
development and is currently temporarily disabled, only the patchlevel is
bumped.
We may end up selecting an invalid CPU (according to the task's cpumask)
when dispatching the task via dispatch_direct_cpu().
When this happens simply return an error and do not dispatch the task
and let the caller handle the error: in the context of select_cpu() we
can simply ignore the dispatch and return the target CPU; in the context
of FIFO mode dispatch we can fallback to SCX_DSQ_LOCAL if the target CPU
is not valid.
This fixes issue #353.
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Kick CPUs in the dispatch path only when needed (typically when tasks
are bounced to other CPUs).
Moreover, avoid to consume all the tasks dispatched at once.
This seems to reduce the BPF overhead (according to bpftop), going from
~10% CPU usage down to ~6% CPU usage of rustland_dispatch() on an over
commissioned system, without introducing any measureable performance
regression.
Tested-by: SoulHarsh007 <harsh.peshwani@outlook.com>
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Allow to dispatch tasks directly (bypassing the user-space scheduler)
only when the scheduler is operating in FIFO mode.
On an over-commissioned system, directly dispatching tasks can only
increase OS noise. These tasks can get a brief priority boost and an
extended time slice just because they found an idle CPU, which can lead
to erratic behavior.
This is particularly problematic when measuring performance stability,
such as evaluating the frames-per-second (fps) of a video game on an
overloaded system.
In such cases, it's better to bounce all tasks to the user-space
scheduler, that will ensure a better level of fairness and smoother
performance.
Moreover, get rid of the second chance dispatch logic introduced in
commit 4791d862 ("scx_rustland_core: second chance CPU migration"). This
seems to provide benefits only on certain architectures (Intel), but it
can introduces lags in others (AMD).
Tested-by: SoulHarsh007 <harsh.peshwani@outlook.com>
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Report dispatch_direct_cpu() events in the trace, like any other
dispatch-related event.
Tested-by: SoulHarsh007 <harsh.peshwani@outlook.com>
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
The dependency of the buddy-alloc crate [1] seems to cause some troubles
with packaging, mostly because the selftests for the crate are failing
when it's compiled in release mode.
For example:
$ cargo test --release -- --nocapture
thread 'tests::fast_alloc::test_basic_malloc' panicked at src/tests/fast_alloc.rs:25:13:
assertion `left == right` failed
left: 0
right: 42
Some of these failures with BuddyAlloc can be fixed by using a memory
arena buffer aligned to page size.
However, some test failures with FastAlloc persist that cannot be
resolved merely by aligning the pre-allocated memory arena to the page
size, as mentioned in [2].
The concern is that this may potentially lead to actual memory bugs.
Therefore, it seems safer to refactor the custom allocator code to
simply use BuddyAlloc, dropping FastAlloc completely.
To achieve this, the entire BuddyAlloc code has been directly included
in scx_rustland_core, referencing the original project and its MIT
licensing information (with the entire code still distributed under the
GPLv2 license).
Then the code has been slightly modified to remove FastAlloc and the
external dependency on the buddy-alloc crate has been dropped.
From a performance perspective this change doesn't seem to introduce any
measurable regression.
[1] https://github.com/jjyr/buddy-alloc
[2] https://github.com/jjyr/buddy-alloc/issues/16
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Fix a potential race condition that might lead to a task being
dispatched without kicking the target CPU, which could result in a
potential stall.
With this applied, scx_rustland has been running without any stall for
about 18 hours on a system where the issue was previously quite easy to
reproduce.
Moreover, clarify a couple of comments in the dispatch path.
This fixes issue #353.
Tested-by: SoulHarsh007 <harsh.peshwani@outlook.com>
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
In preparation of upstreaming, let's set the min version requirement at the
released v6.9 kernels. Drop __COMPAT_SCX_KICK_IDLE. The open helper macros
now check the existence of SCX_KICK_IDLE and abort if not.
In preparation of upstreaming, let's set the min version requirement at the
released v6.9 kernels. Drop __COMPAT_scx_bpf_switch_call(). The open helper
macros now check the existence of SCX_OPS_SWITCH_PARTIAL and abort if not.
The same change was previously applied with commit 8820af8d
("scx_rustland: enable user-space scheduler to preempt other tasks").
However, it was reverted in commit 732ba490 ("scx_rustland: avoid using
SCX_ENQ_PREEMPT") with the introduction of the global dynamic time
slice (inversely proportional to the number of running tasks), because
it already provided sufficient context switch opportunities, negating
any advantage of the preemption.
With the introduction of the virtual time slice in commit 6f4cd853
("scx_rustland: introduce virtual time slice"), re-introducing the
ability for the user-space scheduler to preempt other tasks now appears
beneficial again.
As confirmed by experimental results, this change helps prevent
potential audio cracking issues and enhances overall system
responsiveness, resulting in a 4-5% increase in fps performing the
usual benchmark of gaming while recompiling the kernel.
Tested-by: SoulHarsh007 <harsh.peshwani@outlook.com>
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Tasks dispatched directly using SCX_DSQ_LOCAL may receive excessively
high priority compared to those dispatched by the user-space scheduler.
To avoid this priority disparity, dispatch tasks to the per-CPU DSQs
using direct dispatch and reserve SCX_DSQ_LOCAL for per-CPU kthreads
only.
Tested-by: Tested-by: SoulHarsh007 <harsh.peshwani@outlook.com>
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Make restart handling with user_exit_info simpler and consistently use the
load and report macros consistently across the rust schedulers. This makes
all schedulers automatically handle auto restarts from CPU hotplug events.
Note that this is necessary even for scx_lavd which has CPU hotplug
operations as CPU hotplug operations which took place between skel open and
scheduler init can still trigger restart.
As reported in #319, we may get a build failure in presence of musl,
that requires additional parameters in sched_param.
Fix by adding a proper conditional to support both gnu libc and musl
libc.
This fixes#319.
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
The setting of ops->dispatch_max_batch leads to a too large allocated
size for percpu allocator, and it will be unhappy if we want a size
larger than PCPU_MIN_UNIT_SIZE. Reduce MAX_ENQUEUED_TASKS for fix.
Implement a second-chance migration in select_cpu(): after a task has
been dispatched directly do not try to migrate it immediately on a
different CPU, but force it to stay on prev_cpu for another round.
This seems quite effective on certain architectures (such as on a system
with 11th Gen Intel(R) Core(TM) i7-1195G7 @ 2.90GHz), and it can provide
noticeable benefits with gaming or WebGL applications (such as
https://webglsamples.org/aquarium/aquarium.html) under regular workload
conditions (around +5% fps).
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
The shared DSQ is typically used to prioritize tasks and dispatch them
on the first CPU available, so consume from the shared DSQ before the
local CPU DSQ.
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Provide a knob in scx_rustland_core to automatically turn the scheduler
into a simple FIFO when the system is underutilized.
This choice is based on the assumption that, in the case of system
underutilization (less tasks running than the amount of available CPUs),
the best scheduling policy is FIFO.
With this option enabled the scheduler starts in FIFO mode. If most of
the CPUs are busy (nr_running >= num_cpus - 1), the scheduler
immediately exits from FIFO mode and starts to apply the logic
implemented by the user-space component. Then the scheduler can switch
back to FIFO if there are no tasks waiting to be scheduled (evaluated
using a moving average).
This option can be enabled/disabled by the user-space scheduler using
the fifo_sched parameter in BpfScheduler: if set, the BPF component will
periodically check for system utilization and switch back and forth to
FIFO mode based on that.
This allows to improve performance of workloads that are using a small
amount of the available CPUs in the system, while still maintaining the
same good level of performance for interactive tasks when the system is
over commissioned.
In certain video games, such as Baldur's Gate 3 or Counter-Strike 2,
running in "normal" system conditions, we can experience a boost in fps
of approximately 4-8% with this change applied.
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Simplify the CPU idle selection logic relying on the built-in logic.
If something can be improved in this logic it should be done in the
backend, changing the default idle selection logic, rustland doesn't
need to do anything special here for now.
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
This merge included additional commits that were supposed to be included
in a separate pull request and have nothing to do with the fifo-mode
changes.
Therefore, revert the whole pull request and create a separate one with
the correct list of commits required to implement this feature.
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
The shared DSQ is typically used to prioritize tasks and dispatch them
on the first CPU available, so consume from the shared DSQ before the
local CPU DSQ.
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Provide a knob in scx_rustland_core to automatically turn the scheduler
into a simple FIFO when the system is underutilized.
This choice is based on the assumption that, in the case of system
underutilization (less tasks running than the amount of available CPUs),
the best scheduling policy is FIFO.
With this option enabled the scheduler starts in FIFO mode. If most of
the CPUs are busy (nr_running >= num_cpus - 1), the scheduler
immediately exits from FIFO mode and starts to apply the logic
implemented by the user-space component. Then the scheduler can switch
back to FIFO if there are no tasks waiting to be scheduled (evaluated
using a moving average).
This option can be enabled/disabled by the user-space scheduler using
the fifo_sched parameter in BpfScheduler: if set, the BPF component will
periodically check for system utilization and switch back and forth to
FIFO mode based on that.
This allows to improve performance of workloads that are using a small
amount of the available CPUs in the system, while still maintaining the
same good level of performance for interactive tasks when the system is
over commissioned.
In certain video games, such as Baldur's Gate 3 or Counter-Strike 2,
running in "normal" system conditions, we can experience a boost in fps
of approximately 4-8% with this change applied.
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Simplify the CPU idle selection logic relying on the built-in logic.
If something can be improved in this logic it should be done in the
backend, changing the default idle selection logic, rustland doesn't
need to do anything special here for now.
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
scx_rustland has a function called get_cpu_owner() in BPF which
currently has no callers. There's nothing wrong with the function, but
it causes a warning due to an unused function. Let's just annotate it
with __maybe_unused to tell the compiler that it's not a problem.
Signed-off-by: David Vernet <void@manifault.com>
C SCX_OPS_ATTACH() and rust scx_ops_attach() macros were not calling
.attach() and were only attaching the struct_ops. This meant that all
non-struct_ops BPF programs contained in the skels were never attached which
breaks e.g. scx_layered.
Let's fix it by adding .attach() invocation the the attach macros.
Introduce a low-power mode to force the scheduler to operate in a very
non-work conserving way, causing a significant saving in terms of power
consumption, while still providing a good level of responsiveness in the
system.
This option can be enabled in scx_rustland via the --low_power / -l
option.
The idea is to not immediately re-kick a CPU when it enters an idle
state, but do that only if there are no other tasks running in the
system.
In this way, latency-critical tasks can be still dispatched immediately
on the other active CPUs, while CPU-bound tasks will be forced to spend
more time waiting to be scheduled, basically enforcing a special CPU
throttling mechanism that affects only the tasks that are not latency
critical.
The consequence is a reduction in the overall system throughput, but
also a significant reduction of power consumption, that can be useful
for mobile / battery-powered devices.
Test case (using `scx_rustland -l`):
- play a video game (Terraria) while recompiling the kernel
- measure game performance (fps) and core power consumption (W)
- compare the result of normal mode vs low-power mode
Result:
Game performance | Power consumption |
------------+-----------------+-------------------+
normal mode | 60 fps | 6W |
low-power mode | 60 fps | 3W |
As we can see from the result the reduction of power consumption is
quite significant (50%), while the responsiveness of the game (fps)
remains the same, that means battery life can be potentially doubled
without significantly affecting system responsiveness.
The overall throughput of the system is, of course, affected in a
negative way (kernel build is approximately 50% slower during this
test), but the goal here is to save power while still maintaining a good
level of responsiveness in the system.
For this reason the low-power mode should be considered only in
emergency conditions, for example when the system is close to completely
run out of power or simply to extend the battery life of a mobile device
without compromising its responsiveness.
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
The comment that describes rustland_update_idle() is still incorrectly
reporting an old implemention detail. Update its description for better
clarity.
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Change the BPF CPU selection logic as following:
- if the previously used CPU is idle, keep using it
- if the task is not coming from a wait state, try to stick as much as
possible to the same CPU (for better cache usage)
- if the task is waking up from a wait state rely on the sched_ext
built-int idle selection logic
This logic can be completely disabled when the full user-space mode is
enabled. In this case tasks will always be assigned to the previously
used CPU and the user-space scheduler should take care of distributing
them among the available CPUs.
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Drop the global effective time-slice and use the more fine-grained
per-task time-slice to implement the dynamic time-slice capability.
This allows to reduce the scheduler's overhead (dropping the global time
slice volatile variable shared between user-space and BPF) and it
provides a more fine-grained control on the per-task time slice.
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Replace the BPF_MAP_TYPE_QUEUE with a BPF_MAP_TYPE_USER_RINGBUF to store
the tasks dispatched from the user-space scheduler to the BPF component.
This eliminates the need of the bpf() syscalls, significantly reducing
the overhead of the user-space->kernel communication and delivering a
notable performance boost in the overall system throughput.
Based on experimental results, this change allows to reduces the scheduling
overhead by approximately 30-35% when the system is overcommitted.
This improvement has the potential to make user-space schedulers based
on scx_rustland_core viable options for real production systems.
Link: https://github.com/libbpf/libbpf-rs/pull/776
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
In order to prevent deadlock conditions user-space schedulers need to
perform memory allocations from a pool of pre-allocated memory that will
never be unmapped/reclaimed by the kernel (unevictable memory).
However, there is a special kernel sysctl setting
(vm.compact_unevictable_allowed) that allows kcompactd to reclaim
unevictable memory. This behavior should be prevented by setting
vm.compact_unevictable_allowed = 0, that is what scx_rustland_core does
transparently when a scheduler is started, restoring the previous value
when the scheduler is stopped.
Unfortunately, this is not always doable, especially when running a
scheduler inside a containerized environment or under certain
security/privilege restrictions (e.g., AppArmor confinement).
Therefore, just report a WARNING if we are unable to change this
parameter, instead of considering it a hard-failure, to allow running
scx_rustland_core schedulers also inside such limited environments.
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Given that rustland_core now supports task preemption and it has been
tested successfully, it's worhtwhile to cut a new version of the crate.
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Do not encode dispatch flags in the cpu field, but simply use a separate
"flags" field.
This makes the code much simpler and it increases the size of
dispatched_task_ctx from 24 to 32, that is probably better in terms of
cacheline allocation / performance.
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
