Commit Graph

63 Commits

Author SHA1 Message Date
Tejun Heo
6f81409df4 Bump versions
- scx_utils bumped from 0.6.0 to 0.7.0.

- Repo and rust schedulers get a PATCH level bump.
2024-04-02 10:58:50 -10:00
Tejun Heo
b7b402836d scx_utils: Sync libbpf SHA1s with the rest 2024-04-02 10:35:21 -10:00
Tejun Heo
59bbd800c1 compat: Implement scx_utils::compat and fix up scx_layered
Implement scx_utils::compat to match C's scx/compat.h and update
scx_layered. Other rust scheds are still broken.
2024-04-02 07:08:56 -10:00
David Vernet
1bd990fb87
topology: Fall back to cache 0
As described in https://github.com/sched-ext/scx/issues/195, apparently
some chips don't export information about their cache topology. There's
not much we can do if we don't have that information, so let's just
assume a unified cache per node if that happens.

Andrea suggested this patch -- I'm applying exactly what he proposed,
with a slightly modified comment.

Suggested-by: Andrea Righi <andrea.righi@canonical.com>
Signed-off-by: David Vernet <void@manifault.com>
2024-03-21 11:10:55 -05:00
David Vernet
3ad0fff855
Merge pull request #188 from sched-ext/topology-fix-single-cpu
topology: support single CPU systems
2024-03-14 13:16:59 -05:00
David Vernet
4520514fe8
rusty: Account for disabled but offline CPUs
As described in https://bugzilla.kernel.org/show_bug.cgi?id=218109,
https://github.com/sched-ext/scx/issues/147 and
https://github.com/sched-ext/sched_ext/issues/69, AMD chips can
sometimes report fully disabled CPUs as offline, which causes us to
count them when looking at /sys/devices/system/cpu/possible.

Additionally, systems can have holes in their active CPU maps. For
example, a system with CPUs 0, 1, 2, 3 possible, may have only 0 and 2
active. To address this, we need to do a few things:

1. Update topology.rs to be clear that it's returning the number of
   _possible_ CPUs in the system. Also update Topology to only record
   online CPUs when creating its span and iterating over sysfs when
   creating domains. It was previously trying to record when a CPU was
   online, but this was actually broken as the topology directory isn't
   present in sysfs when the CPU is offline.

2. Schedulers should not be relying on nr_possible_cpus for anything
   other than interacting with per-CPU data (e.g. for stats extraction),
   or e.g. verifying maximum sizes of statically sized arrays in BPF. It
   should _not_ be used for e.g. performing load calculations, etc. With
   that said, we'll also need to update schedulers to not rely on the
   nr_possible_cpus figure being exported by the topology crate. We do
   that for rusty in this patch, but don't fix any of the others other
   than updating how they call topology.rs.

3. Account for the fact that LLC IDs may be non-contiguous. For example,
   if there is a single core in an LLC, then if we assign LLC IDs to
   domains, then the domain IDs won't be contiguous. This doesn't fit
   our current model which is used by e.g. infeasible_weights.rs. We'll
   update some of the code in rusty to accomodate this, but we'll need
   to do more.

4. Update schedulers to properly reset themselves in the event of a
   hotplug event. We'll take care of that in a follow-on change.

Signed-off-by: David Vernet <void@manifault.com>
2024-03-14 11:15:28 -05:00
David Vernet
bc0336d727
cpumask: Add bitwise ops for cpumask
We implement functions or(), and(), and xor() for cpumasks, but we
should also implement the bitwise ops for those operations in case
people prefer that syntax.

Signed-off-by: David Vernet <void@manifault.com>
2024-03-14 11:02:01 -05:00
David Vernet
84a202e2a0
topology: Skip offline CPUs
Offline CPUs don't have a /sys/devices/system/cpu/cpuN/topology
directory, so let's just skip them if they're not online. Schedulers are
expected to detect hotplug, and handle gracefully restarting.

Signed-off-by: David Vernet <void@manifault.com>
2024-03-14 11:02:01 -05:00
David Vernet
583696f940
topology: Include last CPU in online
We're iterating from min..max cpu in cpus_online(), but that's not
inclusive of the max CPU. Let's also include that so we don't think that
last CPU is offline.

Signed-off-by: David Vernet <void@manifault.com>
2024-03-14 11:01:52 -05:00
Andrea Righi
a1b05c5ab3 topology: support single CPU systems
We are failing to parse /sys/devices/system/cpu/online in systems with
just one CPU, for example:

 $ vng -r --cpus 1 -- scx_rusty
 Error: Failed to parse online cpus 0

Correctly handle strings containing only a single CPU during parsing.

Fixes: c5a3b83b ("topology: Add new topology crate")
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
2024-03-14 07:46:20 +01:00
David Vernet
91cb5ce8ab
Merge pull request #178 from sched-ext/multi_numa_rusty
rusty: Implement NUMA-aware load balancing
2024-03-12 15:50:27 -05:00
Jordan Rome
ffc7b7dc4a Fetch and build bpftool by default
This pairs with the new default behavior to fetch and build libbpf
and is mostly being used so we can use the latest bpftool and libbpf.
2024-03-11 10:00:01 -07:00
David Vernet
1c3168d2a4
topology: Don't assume unique core IDs
The current topology.rs crate assumes that all cores have unique core
IDs in a system. This need not be the case, such as in certain Intel
Xeon processors which reuse core IDs in different NUMA nodes. Let's
update the crate to assume unique core IDs only per socket.

Signed-off-by: David Vernet <void@manifault.com>
2024-03-08 15:13:46 -06:00
David Vernet
12e0586fe9
cpumask: Update cpumask fmt function
The cpumask print formatter doesn't look great in its current form, which uses
the BitVec formatter under the hood:

[INFO] NUMA[00 32:<[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]>]
[INFO]         DOM[00] 32:<[1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0]>
[INFO]         DOM[01] 32:<[0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1]>

Let's just iterate over the mask and manually format the string using the
binary formatter over the slice of u64's, which renders like this:

[INFO] NUMA[00] 0b11111111111111111111111111111111]
[INFO]         DOM[00] 0b00000000111111110000000011111111
[INFO]         DOM[01] 0b11111111000000001111111100000000

Signed-off-by: David Vernet <void@manifault.com>
2024-03-08 15:11:17 -06:00
Jordan Rome
499924ead8 Add libbpf as a submodule
This is to potentinally reduce issues with folks
using different versions of libbpf at runtime.

This also:
- makes static linking of libbpf the default
- adds steps in `meson setup` to fetch libbpf and make it
2024-03-01 12:39:35 -08:00
Andrea Righi
06d8170f9f scx_utils: introduce Builder()
Introduce a Builder() class in scx_utils that can be used by other scx
crates (such as scx_rustland_core) to prevent code duplication.

Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
2024-02-28 17:49:44 +01:00
Andrea Righi
416d6a940f rust: introduce scx_rustland_core crate
Introduce a separate crate (scx_rustland_core) that can be used to
implement sched-ext schedulers in Rust that run in user-space.

This commit only provides the basic layout for the new crate and the
abstraction to the custom allocator.

In general, any scheduler that has a user-space component needs to use
the custom allocator to prevent potential deadlock conditions, caused by
page faults (a kthread needs to run to resolve the page fault, but the
scheduler is blocked waiting for the user-space page fault to be
resolved => deadlock).

However, we don't want to necessarily enforce this constraint to all the
existing Rust schedulers, some of them may do all user-space allocations
in safe paths, hence the separate scx_rustland_core crate.

Merging this code in scx_utils would force all the Rust schedulers to
use the custom allocator.

In a future commit the scx_rustland backend will be moved to
scx_rustland_core, making it a totally generic BPF scheduler framework
that can be used to implement user-space schedulers in Rust.

Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
2024-02-28 17:49:44 +01:00
David Vernet
b1dc37889f
infeasible: Add a new infeasible crate for load balancing
We want to avoid every scheduler implementation from having to implement
the solution to the infeasible weights problem, but we also want to
enable sufficient flexibility where not every program has to have the
same partition of scheduling domains, etc. To enable this, a new
infeasible crate is added which encapsulates all of the logic for being
given duty cycle and weight, and performing the necessary math to adjust
for infeasibility.

Signed-off-by: David Vernet <void@manifault.com>
2024-02-26 10:51:52 -06:00
David Vernet
f85cde7038
topology: Add maps to cores and cpus from root Topology object
For convenience, let's provide callers with a way to easily look up
cores and CPUs from the root topology object.

Signed-off-by: David Vernet <void@manifault.com>
2024-02-23 13:09:02 -06:00
David Vernet
43624a87ce
rusty: Use new topology crate
Now that we have this new Topology crate, let's update Rusty to use it
instead of using the old one.

Signed-off-by: David Vernet <void@manifault.com>
2024-02-23 10:39:55 -06:00
David Vernet
c5a3b83bbd
topology: Add new topology crate
The topology.rs crate is insufficiently generic, and reflects
implementation details of scx_rusty more than it provides generic use
cases for modeling a host's topology. This adds a new topology2.rs crate
that will replace topology.rs. We have this as an intermediate commit so
that we don't bundle updating scx_rusty with adding this crate.

Signed-off-by: David Vernet <void@manifault.com>
2024-02-23 10:39:00 -06:00
David Vernet
a2f531e429
topology: Refactor topology.rs to use cpumask crate
Now that we have cpumask.rs, we can remove some logic from topology.rs
and have it create and use Cpumasks.

Signed-off-by: David Vernet <void@manifault.com>
2024-02-22 17:25:44 -06:00
David Vernet
bd15eb8e41
cpumask: Add new Cpumask crate
Let's add a Cpumask trait that schedulers can use to avoid all having to
deal directly with BitVec and the like.

Signed-off-by: David Vernet <void@manifault.com>
2024-02-22 17:24:26 -06:00
David Vernet
49065de8df
scx: Demote panic! to warn! in topology crate
We currently panic! if we're building a Topology that detects more than
two siblings on a physical core. This can and will likely happen on
multi-socket machines. Given that we're planning to add support for
detecting NUMA nodes soon, let's just demote the panic! to a warn!.

Signed-off-by: David Vernet <void@manifault.com>
2024-02-20 21:01:17 -06:00
Jordan Rome
7c32acece0 Add libbpf logging to the rust schedulers
This is to get better logs when failing to load, attach, etc.
2024-02-20 15:17:10 -08:00
David Vernet
ef8aa9ea31
add documentation
Signed-off-by: David Vernet <void@manifault.com>
2024-02-20 14:57:09 -06:00
David Vernet
8aba090d4f
rust: Add topology module to utils crate
scx_rusty has logic in the scheduler to inspect the host to
automatically build scheduling domains across every L3 cache. This would
be generically useful for many different types of schedulers, so let's
add it to the scx_utils crate so it can be used by others.

Signed-off-by: David Vernet <void@manifault.com>
2024-02-20 14:57:09 -06:00
Andrea Righi
f5a21198ad scx_utils: use c_char to prevent build failures
Use c_char to convert C strings, that is more portable across different
architectures.

This prevents a build failure on arm64 and ppc64el.

Fixes: d57a23f ("rust/scx_utils: Add user_exit_info support")
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
2024-02-11 21:42:52 +01:00
Tejun Heo
d57a23f481 rust/scx_utils: Add user_exit_info support
So that rust scheds can use the same unified exit handling as C scheds.

Signed-off-by: Tejun Heo <tj@kernel.org>
2024-01-31 10:54:15 -10:00
Tejun Heo
988b7d13c1 Bump versions
scx_exit_info change doesn't require code to be updated but breaks binary
compatbility. Bump versions and cut a new release.
2024-01-25 09:01:23 -10:00
Tejun Heo
942b0269b8 Bump versions
After updates to reflect the updated init and direct dispatch API, the
schedulers aren't compatible with older kernels. Bump versions and publish
releases.
2024-01-08 18:49:54 -10:00
Jordan Rome
6caf6c5c99 Add new archs for bpf_builder
This is to fix fedora build failures for these archs:
s390x and ppc64le

Error:
```
---- bpf_builder::tests::test_bpf_builder_new stdout ----
thread 'bpf_builder::tests::test_bpf_builder_new' panicked at src/bpf_builder.rs:592:9:
Failed to create BpfBuilder (Err(CPU arch "s390x" not found in ARCH_MAP))
```

https://koji.fedoraproject.org/koji/taskinfo?taskID=111114326
2024-01-03 10:50:33 -08:00
Tejun Heo
98773131df Bump versions to publish scx_utils fedora compat change 2023-12-29 06:58:45 +09:00
Tejun Heo
c47a4b6716 scx_utils: Explain what's going on with bindgen version and suppress deprecation warning
This is a followup to https://github.com/sched-ext/scx/pull/50. See the
comment in BpfBuilder::bindgen_bpf_intf() for details.
2023-12-29 06:56:07 +09:00
Jordan Rome
c8a721b033 Downgrade bindgen to 0.68
This is so we can package scx_utils into fedora without having
to upgrade rust-bindgen
(https://bodhi.fedoraproject.org/updates/FEDORA-2023-18e7f124e1).

To make this happen we need to stop using the `CargoCallbacks::new`
constructor which was added in 0.69. Old way seems legit according
to the docs:
https://rust-lang.github.io/rust-bindgen/non-system-libraries.html
2023-12-27 12:19:28 -08:00
Jordan Rome
e9a9d32ab6 Restructure scheds folder names
- combine c and kernel-examples as it's confusing to have both
- rename 'rust-user' and 'c-user' to just 'rust' and 'c', which is simpler
- update and fix sync-to-kernel.sh
2023-12-17 13:14:31 -08:00
Daniel Müller
fed1dae9da rust: Update libbpf-rs & libbpf-cargo to 0.22
This is a follow on to #32, which got reverted. I wrongly assumed that
scx_rusty resides in the sched_ext tree and consumes published version
of scx_utils.
With this change we update the other in-tree dependencies. I built
scx_layered & scx_rusty. I bumped scx-utils to 0.4, because the
libbpf-cargo seems to be part of the public API surface and libbpf-cargo
0.21 and 0.22 are not compatible with each other.

Signed-off-by: Daniel Müller <deso@posteo.net>
2023-12-14 14:33:58 -08:00
Tejun Heo
995520e415 Revert "scx_utils: Update libbpf-cargo to 0.22"
This reverts commit 6d8ba3e3d7.

Piotr Gorski reports build breakage after the commit:

  https://paste.cachyos.org/p/aa14709.txt
2023-12-14 10:32:25 -10:00
Daniel Müller
6d8ba3e3d7 scx_utils: Update libbpf-cargo to 0.22
It's the latest and greatest. Given the repository setup, this is likely
a breaking change from a semver perspective.
2023-12-14 11:18:27 -08:00
Tejun Heo
8a07bcc31b Bump versions and add LICENSE symlinks for scx_layered and scx_rusty 2023-12-12 11:21:08 -10:00
Davide Cavalca
21e468a491 rust: clarify license and include text 2023-12-12 13:02:13 -08:00
Tejun Heo
3acdef8f2b scx_utils: Bump version to 0.3.2
- Build fix for ubuntu clang.
2023-12-07 13:08:21 -10:00
Andrea Righi
7ceccf6516 scx_utils::BpfBuilder: properly detect clang version in Ubuntu
Apply the same logic of commit 00cd15a ("build: properly detect clang
version in Ubuntu") in scx_utils as well.

This allows to build scx_utils properly in Ubuntu.

Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
2023-12-07 23:26:21 +01:00
Tejun Heo
fa532a87f5 scx_utils: Bump version & publish
So that it doesn't cause the missing lifetime parameter warnings everywhere.
2023-12-05 14:58:51 -10:00
Tejun Heo
5422584baa scx_utils::BpfBuilder: Add 'static to BPF_H_TAR definition
Newer rustc is unhappy about lack of explicit lifetime parameter.
2023-12-05 14:15:06 -10:00
Tejun Heo
fcab460386 scx_utils: Bump version and publish
include directory structure has changed (a breaking change) and the doc had
a misleading error. Let's cut a new release.
2023-12-03 12:51:16 -10:00
Tejun Heo
d0ed7913b4 scheds: Rearrange include files to match kernel/tools/sched_ext/include
Build scripts are updated accordingly.
2023-12-03 12:47:23 -10:00
Tejun Heo
44b811831a doc: README.md and OVERVIEW.md added and other minor updates 2023-12-03 11:48:55 -10:00
Tejun Heo
12a90335b3 scx_utils: Bump version to 0.2.1 2023-12-02 23:56:51 -10:00
Tejun Heo
b38f7574ac scx_utils: Documentation and other minor updates 2023-12-02 23:56:36 -10:00