| Age | Commit message (Collapse) | Author |
|---|
|
Dropping a timer on the millisecond that it was scheduled for, when it was on
the pending list, could result in a panic previously, as we did not record the
pending-list state in cached_when.
Hopefully fixes: ZcashFoundation/zebra#1452
|
|
Previously, the runtime shutdown logic would first-hand control over all cores
to a single thread, which would sequentially shut down all tasks on the core
and then wait for them to complete.
This could deadlock when one task is waiting for a later core's task to
complete. For example, in the newly added test, we have a `block_in_place` task
that is waiting for another task to be dropped. If the latter task adds its
core to the shutdown list later than the former, we end up waiting forever for
the `block_in_place` task to complete.
Additionally, there also was a bug wherein we'd attempt to park on the parker
after shutting it down which was fixed as part of the refactors above.
This change restructures the code to bring all tasks to a halt (and do any
parking needed) before we collapse to a single thread to avoid this deadlock.
There was also an issue in which canceled tasks would not unpark the
originating thread, due to what appears to be some sort of optimization gone
wrong. This has been fixed to be much more conservative in selecting when not
to unpark the source thread (this may be too conservative; please take a look
at the changes to `release()`).
Fixes: #2789
|
|
|
|
More-or-less a half-rewrite of the current time driver, supporting the
use of intrusive futures for timer registration.
Fixes: #3028, #3069
|
|
tokio:
merge rt-core and rt-util as rt
rename rt-threaded to rt-multi-thread
tokio-util:
rename rt-core to rt
Closes #2942
|
|
Co-authored-by: Alice Ryhl <alice@ryhl.io>
Co-authored-by: Carl Lerche <me@carllerche.com>
|
|
|
|
|
|
* remove unnecessary wheel::Poll
the timer wheel uses the `wheel::Poll` struct as input when
advancing the timer to the next time step. the `Poll` struct
contains an instant representing the time step to advance to
and also contains an optional and mutable reference to an
`Expiration` struct. from what I can tell, the latter field
is only used in the context of polling the wheel and does not
need to be exposed outside of that method. without the
expiration field the `Poll` struct is nothing more than a
wrapper around the instant being polled. this change removes
the `Poll` struct and updates integration points accordingly.
* remove Stack trait in favor of concrete Stack implementation
* remove timer Registration struct
|
|
|
|
Fixes: #2696
Co-authored-by: Mikail Bagishov <bagishov.mikail@yandex.ru>
|
|
JoinHandle of threads created by the pool are now tracked and properly joined at
shutdown. If the thread does not return within the timeout, then it's not joined and
left to the OS for cleanup.
Also, break a cycle between wakers held by the timer and the runtime.
Fixes #2641, #2535
|
|
Closes #1953
|
|
* Update doc comments
* Remove trailing whitespace
|
|
Applying a suggestion from #2617 to make the sentence more clear.
|
|
This patch updates the coop logic so that the budget is only decremented
if a future makes progress (that is, if it returns `Ready`). This is
realized by restoring the budget to its former value after
`poll_proceed` _unless_ the caller indicates that it made progress.
The thinking here is that we always want tasks to make progress when we
poll them. With the way things were, if a task polled 128 resources that
could make no progress, and just returned `Pending`, then a 129th
resource that _could_ make progress would not be polled. Worse yet, this
could manifest as a deadlock, if the first 128 resources were all
_waiting_ for the 129th resource, since it would _never_ be polled.
The downside of this change is that `Pending` resources now do not take
up any part of the budget, even though they _do_ take up time on the
executor. If a task is particularly aggressive (or unoptimized), and
polls a large number of resources that cannot make progress whenever it
is polled, then coop will allow it to run potentially much longer before
yielding than it could before. The impact of this should be relatively
contained though, because tasks that behaved in this way in the past
probably ignored `Pending` _anyway_, so whether a resource returned
`Pending` due to coop or due to lack of progress may not make a
difference to it.
|
|
If a Delay has been polled, then the task that polled it may be waiting
for a notification. If the delay gets reset to a time in the past, then
it immediately becomes elapsed, so it should notify the relevant task.
|
|
Fixes a test from a PR that was written before the recent loom upgrade.
A change in the details how loom executes models resulted in the test to
start failing. The fix is to reduce the number of iterations performed
by the test.
|
|
This commit improves the memory ordering in the implementation of
Inner's increment function. The former code did a sequentially
consistent load of self.num, then entered a loop with a sequentially
consistent compare and swap on the same, bailing out with and Err only
if the loaded value was MAX_TIMEOUTS. The use of SeqCst means that all
threads must observe all relevant memory operations in the same order,
implying synchronization between all CPUs.
This commit adjusts the implementation in two key ways. First, the
initial load of self.num is now down with Relaxed ordering. If two
threads entered this code simultaneously, formerly, tokio required
that one proceed before the other, negating their parallelism. Now,
either thread may proceed without coordination. Second, the SeqCst
compare_and_swap is changed to a Release, Relaxed
compare_exchange_weak. The first memory ordering referrs to success:
if the value is swapped the load of that value for comparison will be
Relaxed and the store will be Release. The second memory ordering
referrs to failure: if the value is not swapped the load is
Relaxed. The _weak variant may spuriously fail but will generate
better code.
These changes mean that it is possible for more loops to be taken per
call than strictly necessary but with greater parallelism available on
this operation, improved energy consumption as CPUs don't have to
coordinate as much.
|
|
The resume function was breaking the guarantee that Instants should
never be less than any previously measured Instants when created.
Altered the pause and resume function such that they will not break this
guarantee. After resume, the time should continue from where it left
off.
Created test to prove that the advanced function still works as
expected.
Added additional tests for the pause/advance/resume functions.
|
|
A single call to `poll` on a top-level task may potentially do a lot of
work before it returns `Poll::Pending`. If a task runs for a long period
of time without yielding back to the executor, it can starve other tasks
waiting on that executor to execute them, or drive underlying resources.
See for example rust-lang/futures-rs#2047, rust-lang/futures-rs#1957,
and rust-lang/futures-rs#869. Since Rust does not have a runtime, it is
difficult to forcibly preempt a long-running task.
Consider a future like this one:
```rust
use tokio::stream::StreamExt;
async fn drop_all<I: Stream>(input: I) {
while let Some(_) = input.next().await {}
}
```
It may look harmless, but consider what happens under heavy load if the
input stream is _always_ ready. If we spawn `drop_all`, the task will
never yield, and will starve other tasks and resources on the same
executor.
This patch adds a `coop` module that provides an opt-in mechanism for
futures to cooperate with the executor to avoid starvation. This
alleviates the problem above:
```
use tokio::stream::StreamExt;
async fn drop_all<I: Stream>(input: I) {
while let Some(_) = input.next().await {
tokio::coop::proceed().await;
}
}
```
The call to [`proceed`] will coordinate with the executor to make sure
that every so often control is yielded back to the executor so it can
run other tasks.
The implementation uses a thread-local counter that simply counts how
many "cooperation points" we have passed since the task was first
polled. Once the "budget" has been spent, any subsequent points will
return `Poll::Pending`, eventually making the top-level task yield. When
it finally does yield, the executor resets the budget before
running the next task.
The budget per task poll is currently hard-coded to 128. Eventually, we
may want to make it dynamic as more cooperation points are added. The
number 128 was chosen more or less arbitrarily to balance the cost of
yielding unnecessarily against the time an executor may be "held up".
At the moment, all the tokio leaf futures ("resources") call into coop,
but external futures have no way of doing so. We probably want to
continue limiting coop points to leaf futures in the future, but may
want to also enable third-party leaf futures to cooperate to benefit the
ecosystem as a whole. This is reflected in the methods marked as `pub`
in `mod coop` (even though the module is only `pub(crate)`). We will
likely also eventually want to expose `coop::limit`, which enables
sub-executors and manual `impl Future` blocks to avoid one sub-task
spending all of their poll budget.
Benchmarks (see tokio-rs/tokio#2160) suggest that the overhead of `coop`
is marginal.
|
|
This requires fixing a few warnings.
|
|
|
|
|
|
Tweak context to remove more fns and usage of `Option`. Remove
`ThreadContext` struct as it is reduced to just `Handle`. Avoid passing
around individual driver handles and instead limit to the
`runtime::Handle` struct.
|
|
This patch improves the behavior of frozen time (a testing utility made
available with the `test-util` feature flag). Instead of of requiring
`time::advance` to be called in order to advance the value returned by
`Instant::now`, calls to `time::Driver::park_timeout` will use the
provided duration to advance the time.
This is the desired behavior as the timeout is used to indicate when the
next scheduled delay needs to be fired.
|
|
Previously, thread-locals used by the various drivers were situated
with the driver code. This resulted in state being spread out and many
thread-locals being required to run a runtime.
This PR coalesces the thread-locals into a single struct.
|
|
also add an async version of `fs::canonicalize`
|
|
## Motivation
#1800 removed the lazy binding of `Delay`s to timers. With the removal of the
logic required for that, `HandlePriv` is no longer needed. This PR removes the
use of `HandlePriv`.
A `TODO` was also removed that would panic if when registering a new `Delay`
the current timer handle was full. That has been fixed to now immediately
transition that `Delay` to an error state that can be handled in a similar way
to other error states.
Signed-off-by: Kevin Leimkuhler <kleimkuhler@icloud.com>
|
|
|
|
* runtime: cleanup and add config options
This patch finishes the cleanup as part of the transition to Tokio 0.2.
A number of changes were made to take advantage of having all Tokio
types in a single crate. Also, fixes using Tokio types from
`spawn_blocking`.
* Many threads, one resource driver
Previously, in the threaded scheduler, a resource driver (mio::Poll /
timer combo) was created per thread. This was more or less fine, except
it required balancing across the available drivers. When using a
resource driver from **outside** of the thread pool, balancing is
tricky. The change was original done to avoid having a dedicated driver
thread.
Now, instead of creating many resource drivers, a single resource driver
is used. Each scheduler thread will attempt to "lock" the resource
driver before parking on it. If the resource driver is already locked,
the thread uses a condition variable to park. Contention should remain
low as, under load, the scheduler avoids using the drivers.
* Add configuration options to enable I/O / time
New configuration options are added to `runtime::Builder` to allow
enabling I/O and time drivers on a runtime instance basis. This is
useful when wanting to create lightweight runtime instances to execute
compute only tasks.
* Bug fixes
The condition variable parker is updated to the same algorithm used in
`std`. This is motivated by some potential deadlock cases discovered by
`loom`.
The basic scheduler is fixed to fairly schedule tasks. `push_front` was
accidentally used instead of `push_back`.
I/O, time, and spawning now work from within `spawn_blocking` closures.
* Misc cleanup
The threaded scheduler is no longer generic over `P :Park`. Instead, it
is hard coded to a specific parker. Tests, including loom tests, are
updated to use `Runtime` directly. This provides greater coverage.
The `blocking` module is moved back into `runtime` as all usage is
within `runtime` itself.
|
|
## Motivation
Similar to #1666, it is no longer necessary to lazily register delays with the
executions default timer. All delays are expected to be created from within a
runtime, and should panic if not done so.
## Solution
`tokio::time` now assumes there to be a `CURRENT_TIMER` set when creating a
delay; this can be assumed if called within a tokio runtime. If there is no
current timer, the application will panic with a "no current timer" message.
## Follow-up
Similar to #1666, `HandlePriv` can probably be removed, but this mainly prepares
for 0.2 API changes. Because it is not in the public API, this can be done in a
following change.
Signed-off-by: Kevin Leimkuhler <kleimkuhler@icloud.com>
|
|
In an effort to reach API stability, the `tokio` crate is shedding its
_public_ dependencies on crates that are either a) do not provide a
stable (1.0+) release with longevity guarantees or b) match the `tokio`
release cadence. Of course, implementing `std` traits fits the
requirements.
The on exception, for now, is the `Stream` trait found in `futures_core`.
It is expected that this trait will not change much and be moved into `std.
Since Tokio is not yet going reaching 1.0, I feel that it is acceptable to maintain
a dependency on this trait given how foundational it is.
Since the `Stream` implementation is optional, types that are logically
streams provide `async fn next_*` functions to obtain the next value.
Avoiding the `next()` name prevents fn conflicts with `StreamExt::next()`.
Additionally, some misc cleanup is also done:
- `tokio::io::io` -> `tokio::io::util`.
- `delay` -> `delay_until`.
- `Timeout::new` -> `timeout(...)`.
- `signal::ctrl_c()` returns a future instead of a stream.
- `{tcp,unix}::Incoming` is removed (due to lack of `Stream` trait).
- `time::Throttle` is removed (due to lack of `Stream` trait).
- Fix: `mpsc::UnboundedSender::send(&self)` (no more conflict with `Sink` fns).
|
|
This patch started as an effort to make `time::Timer` private. However, in an
effort to get the build compiling again, more and more changes were made. This
probably should have been broken up, but here we are. I will attempt to
summarize the changes here.
* Feature flags are reorganized to make clearer. `net-driver` becomes
`io-driver`. `rt-current-thread` becomes `rt-core`.
* The `Runtime` can be created without any executor. This replaces `enter`. It
also allows creating I/O / time drivers that are standalone.
* `tokio::timer` is renamed to `tokio::time`. This brings it in line with `std`.
* `tokio::timer::Timer` is renamed to `Driver` and made private.
* The `clock` module is removed. Instead, an `Instant` type is provided. This
type defaults to calling `std::time::Instant`. A `test-util` feature flag can
be used to enable hooking into time.
* The `blocking` module is moved to the top level and is cleaned up.
* The `task` module is moved to the top level.
* The thread-pool's in-place blocking implementation is cleaned up.
* `runtime::Spawner` is renamed to `runtime::Handle` and can be used to "enter"
a runtime context.
|
