This cycle, a lot of workqueue changes including some that are significant
and invasive.
- During v6.6 cycle, unbound workqueues were updated so that they are more
topology aware and flexible, which among other things improved workqueue
behavior on modern multi-L3 CPUs. In the process, 636b927eba
("workqueue: Make unbound workqueues to use per-cpu pool_workqueues")
switched unbound workqueues to use per-CPU frontend pool_workqueues as a
part of increasing front-back mapping flexibility.
An unwelcome side effect of this change was that this made max concurrency
enforcement per-CPU blowing up the maximum number of allowed concurrent
executions. I incorrectly assumed that this wouldn't cause practical
problems as most unbound workqueue users are self-regulate max
concurrency; however, there definitely are which don't (e.g. on IO paths)
and the drastic increase in the allowed max concurrency led to noticeable
perf regressions in some use cases.
This is now addressed by separating out max concurrency enforcement to a
separate struct - wq_node_nr_active - which makes @max_active consistently
mean system-wide max concurrency regardless of the number of CPUs or
(finally) NUMA nodes. This is a rather invasive and, in places, a bit
clunky; however, the clunkiness rises from the the inherent requirement to
handle the disagreement between the execution locality domain and max
concurrency enforcement domain on some modern machines. See 5797b1c189
("workqueue: Implement system-wide nr_active enforcement for unbound
workqueues") for more details.
- BH workqueue support is added. They are similar to per-CPU workqueues but
execute work items in the softirq context. This is expected to replace
tasklet. However, currently, it's missing the ability to disable and
enable work items which is needed to convert many tasklet users. To avoid
crowding this merge window too much, this will be included in the next
merge window. A separate pull request will be sent for the couple
conversion patches that are currently pending.
- Waiman plugged a long-standing hole in workqueue CPU isolation where
ordered workqueues didn't follow wq_unbound_cpumask updates. Ordered
workqueues now follow the same rules as other unbound workqueues.
- More CPU isolation improvements: Juri fixed another deficit in workqueue
isolation where unbound rescuers don't respect wq_unbound_cpumask.
Leonardo fixed delayed_work timers firing on isolated CPUs.
- Other misc changes.
-----BEGIN PGP SIGNATURE-----
iIQEABYKACwWIQTfIjM1kS57o3GsC/uxYfJx3gVYGQUCZe7JCQ4cdGpAa2VybmVs
Lm9yZwAKCRCxYfJx3gVYGcnqAP9UP8zEM1la19cilhboDumxmRWyRpV/egFOqsMP
Y5PuoAEAtsBJtQWtm5w46+y+fk3nK2ugXlQio2gH0qQcxX6SdgQ=
=/ovv
-----END PGP SIGNATURE-----
Merge tag 'wq-for-6.9' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq
Pull workqueue updates from Tejun Heo:
"This cycle, a lot of workqueue changes including some that are
significant and invasive.
- During v6.6 cycle, unbound workqueues were updated so that they are
more topology aware and flexible, which among other things improved
workqueue behavior on modern multi-L3 CPUs. In the process, commit
636b927eba ("workqueue: Make unbound workqueues to use per-cpu
pool_workqueues") switched unbound workqueues to use per-CPU
frontend pool_workqueues as a part of increasing front-back mapping
flexibility.
An unwelcome side effect of this change was that this made max
concurrency enforcement per-CPU blowing up the maximum number of
allowed concurrent executions. I incorrectly assumed that this
wouldn't cause practical problems as most unbound workqueue users
are self-regulate max concurrency; however, there definitely are
which don't (e.g. on IO paths) and the drastic increase in the
allowed max concurrency led to noticeable perf regressions in some
use cases.
This is now addressed by separating out max concurrency enforcement
to a separate struct - wq_node_nr_active - which makes @max_active
consistently mean system-wide max concurrency regardless of the
number of CPUs or (finally) NUMA nodes. This is a rather invasive
and, in places, a bit clunky; however, the clunkiness rises from
the the inherent requirement to handle the disagreement between the
execution locality domain and max concurrency enforcement domain on
some modern machines.
See commit 5797b1c189 ("workqueue: Implement system-wide
nr_active enforcement for unbound workqueues") for more details.
- BH workqueue support is added.
They are similar to per-CPU workqueues but execute work items in
the softirq context. This is expected to replace tasklet. However,
currently, it's missing the ability to disable and enable work
items which is needed to convert many tasklet users. To avoid
crowding this merge window too much, this will be included in the
next merge window. A separate pull request will be sent for the
couple conversion patches that are currently pending.
- Waiman plugged a long-standing hole in workqueue CPU isolation
where ordered workqueues didn't follow wq_unbound_cpumask updates.
Ordered workqueues now follow the same rules as other unbound
workqueues.
- More CPU isolation improvements: Juri fixed another deficit in
workqueue isolation where unbound rescuers don't respect
wq_unbound_cpumask. Leonardo fixed delayed_work timers firing on
isolated CPUs.
- Other misc changes"
* tag 'wq-for-6.9' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq: (54 commits)
workqueue: Drain BH work items on hot-unplugged CPUs
workqueue: Introduce from_work() helper for cleaner callback declarations
workqueue: Control intensive warning threshold through cmdline
workqueue: Make @flags handling consistent across set_work_data() and friends
workqueue: Remove clear_work_data()
workqueue: Factor out work_grab_pending() from __cancel_work_sync()
workqueue: Clean up enum work_bits and related constants
workqueue: Introduce work_cancel_flags
workqueue: Use variable name irq_flags for saving local irq flags
workqueue: Reorganize flush and cancel[_sync] functions
workqueue: Rename __cancel_work_timer() to __cancel_timer_sync()
workqueue: Use rcu_read_lock_any_held() instead of rcu_read_lock_held()
workqueue: Cosmetic changes
workqueue, irq_work: Build fix for !CONFIG_IRQ_WORK
workqueue: Fix queue_work_on() with BH workqueues
async: Use a dedicated unbound workqueue with raised min_active
workqueue: Implement workqueue_set_min_active()
workqueue: Fix kernel-doc comment of unplug_oldest_pwq()
workqueue: Bind unbound workqueue rescuer to wq_unbound_cpumask
kernel/workqueue: Let rescuers follow unbound wq cpumask changes
...
Currently, all macros are reexported with #[macro_export] only, which
means that to access `new_work!` from the workqueue, you need to import
it from the path `kernel::new_work` instead of importing it from the
workqueue module like all other items in the workqueue. By adding
reexports of the macros, it becomes possible to import the macros from
the correct modules.
It's still possible to import the macros from the root, but I don't
think we can do anything about that.
There is no functional change. This is merely a code cleanliness
improvement.
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Reviewed-by: Trevor Gross <tmgross@umich.edu>
Reviewed-by: Martin Rodriguez Reboredo <yakoyoku@gmail.com>
Tested-by: Boqun Feng <boqun.feng@gmail.com>
Link: https://lore.kernel.org/r/20240129145837.1419880-1-aliceryhl@google.com
[ Removed new `use kernel::prelude::*`s, reworded title. ]
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
Remove explicit targets for doclinks in cases where rustdoc can
determine the correct target by itself. The goal is to reduce unneeded
verbosity in the source code.
Signed-off-by: Valentin Obst <kernel@valentinobst.de>
Reviewed-by: Trevor Gross <tmgross@umich.edu>
Reviewed-by: Martin Rodriguez Reboredo <yakoyoku@gmail.com>
Reviewed-by: Alice Ryhl <aliceryhl@google.com>
Link: https://lore.kernel.org/r/20240131-doc-fixes-v3-v3-11-0c8af94ed7de@valentinobst.de
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
Fix places where comments include code fragments that are not enclosed
in backticks.
Signed-off-by: Valentin Obst <kernel@valentinobst.de>
Reviewed-by: Trevor Gross <tmgross@umich.edu>
Reviewed-by: Martin Rodriguez Reboredo <yakoyoku@gmail.com>
Reviewed-by: Alice Ryhl <aliceryhl@google.com>
Link: https://lore.kernel.org/r/20240131-doc-fixes-v3-v3-8-0c8af94ed7de@valentinobst.de
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
Fixes multiple trivial typos in documentation and comments of the
kernel crate.
allocator:
- Fix a trivial list item alignment issue in the last SAFETY comment of
`krealloc_aligned`.
init:
- Replace 'type' with 'trait' in the doc comments of the `PinInit` and
`Init` traits.
- Add colons before starting lists.
- Add spaces between the type and equal sign to respect the code
formatting rules in example code.
- End a sentence with a full stop instead of a colon.
ioctl:
- Replace 'an' with 'a' where appropriate.
str:
- Replace 'Return' with 'Returns' in the doc comment of `bytes_written`
as the text describes what the function does.
sync/lock:
- Fix a trivial list item alignment issue in the Safety section of the
`Backend` trait's description.
sync/lock/spinlock:
- The code in this module operates on spinlocks, not mutexes. Thus,
replace 'mutex' with 'spinlock' in the SAFETY comment of `unlock`.
workqueue:
- Replace "wont" with "won't" in the doc comment of `__enqueue`.
Signed-off-by: Valentin Obst <kernel@valentinobst.de>
Reviewed-by: Trevor Gross <tmgross@umich.edu>
Reviewed-by: Martin Rodriguez Reboredo <yakoyoku@gmail.com>
Reviewed-by: Alice Ryhl <aliceryhl@google.com>
Link: https://lore.kernel.org/r/20240131-doc-fixes-v3-v3-1-0c8af94ed7de@valentinobst.de
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
Commit e563d0a7cd ("workqueue: Break up enum definitions and give
names to the types") gives a name to the `enum` where `WORK_CPU_UNBOUND`
was defined, so `bindgen` changes its output from e.g.:
pub type _bindgen_ty_10 = core::ffi::c_uint;
pub const WORK_CPU_UNBOUND: _bindgen_ty_10 = 64;
to e.g.:
pub type wq_misc_consts = core::ffi::c_uint;
pub const wq_misc_consts_WORK_CPU_UNBOUND: wq_misc_consts = 64;
Thus update Rust's side to match the change (which requires a slight
reformat of the code), fixing the build error.
Closes: https://lore.kernel.org/rust-for-linux/CANiq72=9PZ89bCAVX0ZV4cqrYSLoZWyn-d_K4KpBMHjwUMdC3A@mail.gmail.com/
Fixes: e563d0a7cd ("workqueue: Break up enum definitions and give names to the types")
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Some of our links use relative paths in order to point to files in the
source tree, e.g.:
//! C header: [`include/linux/printk.h`](../../../../include/linux/printk.h)
/// [`struct mutex`]: ../../../../include/linux/mutex.h
These are problematic because they are hard to maintain and do not support
`O=` builds.
Instead, provide support for `srctree`-relative links, e.g.:
//! C header: [`include/linux/printk.h`](srctree/include/linux/printk.h)
/// [`struct mutex`]: srctree/include/linux/mutex.h
The links are fixed after `rustdoc` generation to be based on the absolute
path to the source tree.
Essentially, this is the automatic version of Tomonori's fix [1],
suggested by Gary [2].
Suggested-by: Gary Guo <gary@garyguo.net>
Reported-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
Closes: https://lore.kernel.org/r/20231026.204058.2167744626131849993.fujita.tomonori@gmail.com [1]
Fixes: 48fadf4400 ("docs: Move rustdoc output, cross-reference it")
Link: https://lore.kernel.org/rust-for-linux/20231026154525.6d14b495@eugeo/ [2]
Reviewed-by: Martin Rodriguez Reboredo <yakoyoku@gmail.com>
Reviewed-by: Benno Lossin <benno.lossin@proton.me>
Link: https://lore.kernel.org/r/20231215235428.243211-1-ojeda@kernel.org
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
This adds two examples of how to use the workqueue. The first example
shows how to use it when you only have one `work_struct` field, and the
second example shows how to use it when you have multiple `work_struct`
fields.
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Reviewed-by: Martin Rodriguez Reboredo <yakoyoku@gmail.com>
Reviewed-by: Gary Guo <gary@garyguo.net>
Reviewed-by: "Andreas Hindborg (Samsung)" <nmi@metaspace.dk>
Reviewed-by: Benno Lossin <benno.lossin@proton.me>
Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
This adds a convenience method that lets you spawn a closure for
execution on a workqueue. This will be the most convenient way to use
workqueues, but it is fallible because it needs to allocate memory.
Co-developed-by: Gary Guo <gary@garyguo.net>
Signed-off-by: Gary Guo <gary@garyguo.net>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Reviewed-by: Martin Rodriguez Reboredo <yakoyoku@gmail.com>
Reviewed-by: "Andreas Hindborg (Samsung)" <nmi@metaspace.dk>
Reviewed-by: Benno Lossin <benno.lossin@proton.me>
Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
This implements the `WorkItemPointer` trait for the pointer types that
you are likely to use the workqueue with. The `Arc` type is for
reference counted objects, and the `Pin<Box<T>>` type is for objects
where the caller has exclusive ownership of the object.
Co-developed-by: Gary Guo <gary@garyguo.net>
Signed-off-by: Gary Guo <gary@garyguo.net>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Reviewed-by: Martin Rodriguez Reboredo <yakoyoku@gmail.com>
Reviewed-by: "Andreas Hindborg (Samsung)" <nmi@metaspace.dk>
Reviewed-by: Benno Lossin <benno.lossin@proton.me>
Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
The main challenge with defining `work_struct` fields is making sure
that the function pointer stored in the `work_struct` is appropriate for
the work item type it is embedded in. It needs to know the offset of the
`work_struct` field being used (even if there are several!) so that it
can do a `container_of`, and it needs to know the type of the work item
so that it can call into the right user-provided code. All of this needs
to happen in a way that provides a safe API to the user, so that users
of the workqueue cannot mix up the function pointers.
There are three important pieces that are relevant when doing this:
* The pointer type.
* The work item struct. This is what the pointer points at.
* The `work_struct` field. This is a field of the work item struct.
This patch introduces a separate trait for each piece. The pointer type
is given a `WorkItemPointer` trait, which pointer types need to
implement to be usable with the workqueue. This trait will be
implemented for `Arc` and `Box` in a later patch in this patchset.
Implementing this trait is unsafe because this is where the
`container_of` operation happens, but user-code will not need to
implement it themselves.
The work item struct should then implement the `WorkItem` trait. This
trait is where user-code specifies what they want to happen when a work
item is executed. It also specifies what the correct pointer type is.
Finally, to make the work item struct know the offset of its
`work_struct` field, we use a trait called `HasWork<T, ID>`. If a type
implements this trait, then the type declares that, at the given offset,
there is a field of type `Work<T, ID>`. The trait is marked unsafe
because the OFFSET constant must be correct, but we provide an
`impl_has_work!` macro that can safely implement `HasWork<T>` on a type.
The macro expands to something that only compiles if the specified field
really has the type `Work<T>`. It is used like this:
```
struct MyWorkItem {
work_field: Work<MyWorkItem, 1>,
}
impl_has_work! {
impl HasWork<MyWorkItem, 1> for MyWorkItem { self.work_field }
}
```
Note that since the `Work` type is annotated with an id, you can have
several `work_struct` fields by using a different id for each one.
Co-developed-by: Gary Guo <gary@garyguo.net>
Signed-off-by: Gary Guo <gary@garyguo.net>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Reviewed-by: Benno Lossin <benno.lossin@proton.me>
Reviewed-by: Martin Rodriguez Reboredo <yakoyoku@gmail.com>
Reviewed-by: Andreas Hindborg <a.hindborg@samsung.com>
Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
We provide these methods because it lets us access these queues from
Rust without using unsafe code.
These methods return `&'static Queue`. References annotated with the
'static lifetime are used when the referent will stay alive forever.
That is ok for these queues because they are global variables and cannot
be destroyed.
Signed-off-by: Wedson Almeida Filho <walmeida@microsoft.com>
Co-developed-by: Alice Ryhl <aliceryhl@google.com>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Reviewed-by: Martin Rodriguez Reboredo <yakoyoku@gmail.com>
Reviewed-by: Gary Guo <gary@garyguo.net>
Reviewed-by: "Andreas Hindborg (Samsung)" <nmi@metaspace.dk>
Reviewed-by: Benno Lossin <benno.lossin@proton.me>
Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Define basic low-level bindings to a kernel workqueue. The API defined
here can only be used unsafely. Later commits will provide safe
wrappers.
Co-developed-by: Gary Guo <gary@garyguo.net>
Signed-off-by: Gary Guo <gary@garyguo.net>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Reviewed-by: Martin Rodriguez Reboredo <yakoyoku@gmail.com>
Reviewed-by: "Andreas Hindborg (Samsung)" <nmi@metaspace.dk>
Reviewed-by: Benno Lossin <benno.lossin@proton.me>
Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Linus Torvalds