* Store the `ValRaw` type in little-endian format
This commit changes the internal representation of the `ValRaw` type to
an unconditionally little-endian format instead of its current
native-endian format. The documentation and various accessors here have
been updated as well as the associated trampolines that read `ValRaw`
to always work with little-endian values, converting to the host
endianness as necessary.
The motivation for this change originally comes from the implementation
of the component model that I'm working on. One aspect of the component
model's canonical ABI is how variants are passed to functions as
immediate arguments. For example for a component model function:
```
foo: function(x: expected<i32, f64>)
```
This translates to a core wasm function:
```wasm
(module
(func (export "foo") (param i32 i64)
;; ...
)
)
```
The first `i32` parameter to the core wasm function is the discriminant
of whether the result is an "ok" or an "err". The second `i64`, however,
is the "join" operation on the `i32` and `f64` payloads. Essentially
these two types are unioned into one type to get passed into the function.
Currently in the implementation of the component model my plan is to
construct a `*mut [ValRaw]` to pass through to WebAssembly, always
invoking component exports through host trampolines. This means that the
implementation for `Result<T, E>` needs to do the correct "join"
operation here when encoding a particular case into the corresponding
`ValRaw`.
I personally found this particularly tricky to do structurally. The
solution that I settled on with fitzgen was that if `ValRaw` was always
stored in a little endian format then we could employ a trick where when
encoding a variant we first set all the `ValRaw` slots to zero, then the
associated case we have is encoding. Afterwards the `ValRaw` values are
already encoded into the correct format as if they'd been "join"ed.
For example if we were to encode `Ok(1i32)` then this would produce
`ValRaw { i32: 1 }`, which memory-wise is equivalent to `ValRaw { i64: 1 }`
if the other bytes in the `ValRaw` are guaranteed to be zero. Similarly
storing `ValRaw { f64 }` is equivalent to the storage required for
`ValRaw { i64 }` here in the join operation.
Note, though, that this equivalence relies on everything being
little-endian. Otherwise the in-memory representations of `ValRaw { i32: 1 }`
and `ValRaw { i64: 1 }` are different.
That motivation is what leads to this change. It's expected that this is
a low-to-zero cost change in the sense that little-endian platforms will
see no change and big-endian platforms are already required to
efficiently byte-swap loads/stores as WebAssembly requires that.
Additionally the `ValRaw` type is an esoteric niche use case primarily
used for accelerating the C API right now, so it's expected that not
many users will have to update for this change.
* Track down some more endianness conversions
Relevant to Wasmtime, this fixes undefined references to `utimensat` and
`futimens` on macOS 10.12 and earlier. See bytecodealliance/rustix#157
for details.
It also contains a fix for s390x which isn't currently needed by Wasmtime
itself, but which is needed to make rustix's own testsuite pass on s390x,
which helps people packaging rustix for use in Wasmtime. See
bytecodealliance/rustix#277 for details.
This makes the generator more similar to `wasm-smith` where it is keeping track
of what is on the stack and making choices about what instructions are valid to
generate given the current stack state. This should in theory allow the
generator to emit GC calls while there are live refs on the stack.
Fixes#3917
This commit adds a method, `Module::initialize_copy_on_write_image`,
which will explicitly create the copy-on-write image during the method
call instead of lazily deferring it until the first instantiation of a
module. Given the relative expense of creation of a memfd on Linux this
can otherwise run the risk of unnaturally perturbing the
time-of-instantiation of a module. Additionally whenever lazy
initialization is provided in an API it's typically good practice to
also have an optionally-called forced initialization.
The documentation for the `wasm-spec-interpreter` was not up-to-date,
causing some confusion on non-Ubuntu machines. This change adds the
correct dependencies to install and includes the `libgmp` path for
Fedora by default (i.e., `/lib64`).
* Update wasm-tools crates
This commit updates the wasm-tools family of crates as used in Wasmtime.
Notably this brings in the update which removes module linking support
as well as a number of internal refactorings around names and such
within wasmparser itself. This updates all of the wasm translation
support which binds to wasmparser as appropriate.
Other crates all had API-compatible changes for at least what Wasmtime
used so no further changes were necessary beyond updating version
requirements.
* Update a test expectation
* Refactor away the `Instantiator` type in Wasmtime
This internal type in Wasmtime was primarily used for the module linking
proposal to handle instantiation of many instances and refactor out the
sync and async parts to minimize duplication. With the removal of the
module linking proposal, however, this type isn't really necessary any
longer. In working to implement the component model proposal I was
looking already to refactor this and I figured it'd be good to land that
ahead of time on `main` separate of other refactorings.
This commit removes the `Instantiator` type in the `instance` module.
The type was already private to Wasmtime so this shouldn't have any
impact on consumers. This allows simplifying various code paths to avoid
another abstraction. The meat of instantiation is moved to
`Instance::new_raw` which should be reusable for the component model as
well.
One bug is actually fixed in this commit as well where
`Linker::instantiate` and `InstancePre::instantiate` failed to check
that async support was disabled on a store. This means that they could
have led to a panic if used with an async store and a start function
called an async import (or an async resource limiter yielded). A few
tests were updated with this.
* Review comments
* Upgrade all crates to the Rust 2021 edition
I've personally started using the new format strings for things like
`panic!("some message {foo}")` or similar and have been upgrading crates
on a case-by-case basis, but I think it probably makes more sense to go
ahead and blanket upgrade everything so 2021 features are always
available.
* Fix compile of the C API
* Fix a warning
* Fix another warning
* Bump to 0.36.0
* Add a two-week delay to Wasmtime's release process
This commit is a proposal to update Wasmtime's release process with a
two-week delay from branching a release until it's actually officially
released. We've had two issues lately that came up which led to this proposal:
* In #3915 it was realized that changes just before the 0.35.0 release
weren't enough for an embedding use case, but the PR didn't meet the
expectations for a full patch release.
* At Fastly we were about to start rolling out a new version of Wasmtime
when over the weekend the fuzz bug #3951 was found. This led to the
desire internally to have a "must have been fuzzed for this long"
period of time for Wasmtime changes which we felt were better
reflected in the release process itself rather than something about
Fastly's own integration with Wasmtime.
This commit updates the automation for releases to unconditionally
create a `release-X.Y.Z` branch on the 5th of every month. The actual
release from this branch is then performed on the 20th of every month,
roughly two weeks later. This should provide a period of time to ensure
that all changes in a release are fuzzed for at least two weeks and
avoid any further surprises. This should also help with any last-minute
changes made just before a release if they need tweaking since
backporting to a not-yet-released branch is much easier.
Overall there are some new properties about Wasmtime with this proposal
as well:
* The `main` branch will always have a section in `RELEASES.md` which is
listed as "Unreleased" for us to fill out.
* The `main` branch will always be a version ahead of the latest
release. For example it will be bump pre-emptively as part of the
release process on the 5th where if `release-2.0.0` was created then
the `main` branch will have 3.0.0 Wasmtime.
* Dates for major versions are automatically updated in the
`RELEASES.md` notes.
The associated documentation for our release process is updated and the
various scripts should all be updated now as well with this commit.
* Add notes on a security patch
* Clarify security fixes shouldn't be previewed early on CI
This splits the existing `lookup_by_declaration` function into a
lookup-per-type-of-item. This refactor ends up cleaning up a fair bit of
code in the `wasmtime` crate by removing a number of `unreachable!()`
blocks which are now no longer necessary.
* Remove duplicate `TypeTables` type
This was once needed historically but it is no longer needed.
* Make the internals of `TypeTables` private
Instead of reaching internally for the `wasm_signatures` map an `Index`
implementation now exists to indirect accesses through the type of the
index being accessed. For the component model this table of types will
grow a number of other tables and this'll assist in consuming sites not
having to worry so much about which map they're reaching into.
* Update to rustix 0.33.5, to fix a link error on Android
This updates to rustix 0.33.5, which includes bytecodealliance/rustix#258,
which fixes bytecodealliance/rustix#256, a link error on Android.
Fixes#3965.
* Bump the rustix versions in the Cargo.toml files too.
Updates the OpenVINO backend for wasi-nn to pre-configure all inputs as `NHWC`. This is not a long-term fix but no worse than the status quo, which configures all input tensors to the same layout. This change updates the CI script to use the latest version of OpenVINO. Closes#3948.
Co-authored-by: Andrew Brown <andrew.brown@intel.com>
My previous PR at #3958 accidentally removed the only way to get type
information from a `wasmtime_module_t`, so this commit re-adds methods
back in to continue to be able to get import/export information from a
compiled module.
* Remove the module linking implementation in Wasmtime
This commit removes the experimental implementation of the module
linking WebAssembly proposal from Wasmtime. The module linking is no
longer intended for core WebAssembly but is instead incorporated into
the component model now at this point. This means that very large parts
of Wasmtime's implementation of module linking are no longer applicable
and would change greatly with an implementation of the component model.
The main purpose of this is to remove Wasmtime's reliance on the support
for module-linking in `wasmparser` and tooling crates. With this
reliance removed we can move over to the `component-model` branch of
`wasmparser` and use the updated support for the component model.
Additionally given the trajectory of the component model proposal the
embedding API of Wasmtime will not look like what it looks like today
for WebAssembly. For example the core wasm `Instance` will not change
and instead a `Component` is likely to be added instead.
Some more rationale for this is in #3941, but the basic idea is that I
feel that it's not going to be viable to develop support for the
component model on a non-`main` branch of Wasmtime. Additionaly I don't
think it's viable, for the same reasons as `wasm-tools`, to support the
old module linking proposal and the new component model at the same
time.
This commit takes a moment to not only delete the existing module
linking implementation but some abstractions are also simplified. For
example module serialization is a bit simpler that there's only one
module. Additionally instantiation is much simpler since the only
initializer we have to deal with are imports and nothing else.
Closes#3941
* Fix doc link
* Update comments
* Instead of simply panicking, return an error when we attempt to resume on a dying fiber.
This situation should never occur in the existing code base, but can be
triggered if support for running outside async code in a call hook.
* Shift `async_cx()` to return an `Option`, reflecting if the fiber is dying.
This should never happen in the existing code base, but is a nice
forward-looking guard. The current implementations simply lift the
trap that would eventually be produced by such an operation into
a `Trap` (or similar) at the invocation of `async_cx()`.
* Add support for using `async` call hooks.
This retains the ability to do non-async hooks. Hooks end up being
implemented as an async trait with a handler call, to get around some
issues passing around async closures. This change requires some of
the prior changes to handle picking up blocked tasks during fiber
shutdown, to avoid some panics during timeouts and other such events.
* More fully specify a doc link, to avoid a doc-building error.
* Revert the use of catchable traps on cancellation of a fiber; turn them into expect()/unwrap().
The justification for this revert is that (a) these events shouldn't
happen, and (b) they wouldn't be catchable by wasm anyways.
* Replace a duplicated check in `async` hook evaluation with a single check.
This also moves the checks inside of their respective Async variants,
meaning that if you're using an async-enabled version of wasmtime but
using the synchronous versions of the callbacks, you won't pay any
penalty for validating the async context.
* Use `match &mut ...` insead of `ref mut`.
* Add some documentation on why/when `async_cx` can return None.
* Add two simple test cases for async call hooks.
* Fix async_cx() to check both the box and the value for current_poll_cx.
In the prior version, we only checked that the box had not been cleared,
but had not ensured that there was an actual context for us to use. This
updates the check to validate both, returning None if the inner context
is missing. This allows us to skip a validation check inside `block_on`,
since all callers will have run through the `async_cx` check prior to
arrival.
* Tweak the timeout test to address PR suggestions.
* Add a test about dropping async hooks while suspended
Should help exercise that the check for `None` is properly handled in a
few more locations.
Co-authored-by: Alex Crichton <alex@alexcrichton.com>
This commit exposes some various details and config options for having
finer-grain control over mlock-ing the memory of modules. This amounts
to three different changes being present in this commit:
* A new `Module::image_range` API is added to expose the range in host
memory of where the compiled image resides. This enables embedders to
make mlock-ing decisions independently of Wasmtime. Otherwise though
there's not too much useful that can be done with this range
information at this time.
* A new `Config::force_memory_init_memfd` option has been added. This
option is used to force the usage of `memfd_create` on Linux even when
the original module comes from a file on disk. With mlock-ing the main
purpose for Wasmtime is likely to be avoiding major page faults that
go back to disk, so this is another major source of avoiding page
faults by ensuring that the initialization contents of memory are
always in RAM.
* The `memory_images` field of a `Module` has gone back to being lazily
created on the first instantiation, effectively reverting #3914. This
enables embedders to defer the creation of the image to as late as
possible to allow modules to be created from precompiled images
without actually loading all the contents of the data segments from
disk immediately.
These changes are all somewhat low-level controls which aren't intended
to be generally used by embedders. If fine-grained control is desired
though it's hoped that these knobs provide what's necessary to be
achieved.
In today's installment of the Wondrous Adventures of What Are the Actual
Limits on the Pooling Allocator Required to Run the Spec Tests a fuzz
bug was found where the instance size wasn't big enough to run
`names.wast`. Today's episode is similar to prior episodes where a limit
is bumped until the test passes.
* Support disabling backtraces at compile time
This commit adds support to Wasmtime to disable, at compile time, the
gathering of backtraces on traps. The `wasmtime` crate now sports a
`wasm-backtrace` feature which, when disabled, will mean that backtraces
are never collected at compile time nor are unwinding tables inserted
into compiled objects.
The motivation for this commit stems from the fact that generating a
backtrace is quite a slow operation. Currently backtrace generation is
done with libunwind and `_Unwind_Backtrace` typically found in glibc or
other system libraries. When thousands of modules are loaded into the
same process though this means that the initial backtrace can take
nearly half a second and all subsequent backtraces can take upwards of
hundreds of milliseconds. Relative to all other operations in Wasmtime
this is extremely expensive at this time. In the future we'd like to
implement a more performant backtrace scheme but such an implementation
would require coordination with Cranelift and is a big chunk of work
that may take some time, so in the meantime if embedders don't need a
backtrace they can still use this option to disable backtraces at
compile time and avoid the performance pitfalls of collecting
backtraces.
In general I tried to originally make this a runtime configuration
option but ended up opting for a compile-time option because `Trap::new`
otherwise has no arguments and always captures a backtrace. By making
this a compile-time option it was possible to configure, statically, the
behavior of `Trap::new`. Additionally I also tried to minimize the
amount of `#[cfg]` necessary by largely only having it at the producer
and consumer sites.
Also a noteworthy restriction of this implementation is that if
backtrace support is disabled at compile time then reference types
support will be unconditionally disabled at runtime. With backtrace
support disabled there's no way to trace the stack of wasm frames which
means that GC can't happen given our current implementation.
* Always enable backtraces for the C API
A recently discovered fuzz bug reported a difference in execution result
between Wasmtime and v8. The module in question had an exported function
which had multiple results, including floats. About a year ago I
reported a bug on v8 about functions with multiple results leading to
incorrect results, and it was fixed many months back but I don't think
that we ever actually pulled in that update. After updating v8 I saw the
test failure go away, so this update is being done to fix the fuzz bug
test failure I saw.
* Delete historical interruptable support in Wasmtime
This commit removes the `Config::interruptable` configuration along with
the `InterruptHandle` type from the `wasmtime` crate. The original
support for adding interruption to WebAssembly was added pretty early on
in the history of Wasmtime when there was no other method to prevent an
infinite loop from the host. Nowadays, however, there are alternative
methods for interruption such as fuel or epoch-based interruption.
One of the major downsides of `Config::interruptable` is that even when
it's not enabled it forces an atomic swap to happen when entering
WebAssembly code. This technically could be a non-atomic swap if the
configuration option isn't enabled but that produces even more branch-y
code on entry into WebAssembly which is already something we try to
optimize. Calling into WebAssembly is on the order of a dozens of
nanoseconds at this time and an atomic swap, even uncontended, can add
up to 5ns on some platforms.
The main goal of this PR is to remove this atomic swap on entry into
WebAssembly. This is done by removing the `Config::interruptable` field
entirely, moving all existing consumers to epochs instead which are
suitable for the same purposes. This means that the stack overflow check
is no longer entangled with the interruption check and perhaps one day
we could continue to optimize that further as well.
Some consequences of this change are:
* Epochs are now the only method of remote-thread interruption.
* There are no more Wasmtime traps that produces the `Interrupted` trap
code, although we may wish to move future traps to this so I left it
in place.
* The C API support for interrupt handles was also removed and bindings
for epoch methods were added.
* Function-entry checks for interruption are a tiny bit less efficient
since one check is performed for the stack limit and a second is
performed for the epoch as opposed to the `Config::interruptable`
style of bundling the stack limit and the interrupt check in one. It's
expected though that this is likely to not really be measurable.
* The old `VMInterrupts` structure is renamed to `VMRuntimeLimits`.
When using the pooling instance allocator for running spec tests we have
to make sure that the configured limits of the allocator aren't so low
as to cause the spec tests to fail due to resource exhaustion issues or
similar. This commit adds in minimum thresholds for instance size as
well as instance count. While here this goes ahead and refactors
everything here to look similar.
This commit removes the currently existing laziness-via-`OnceCell` when
a `Module` is created for creating a `ModuleMemoryImages` data
structure. Processing of data is now already shifted to compile time for
the wasm module which means that creating a `ModuleMemoryImages` is
either cheap because the module is backed by a file on disk, it's a
single `write` into the kernel to a memfd, or it's cheap as it's not
supported. This should help make module instantiation time more
deterministic, even for the first instantiation of a module.
Currently, the use of the downcast method means that you have to use one
of the hard-coded types. But Enarx needs to define its own `WasiFile`
implementations. This works fine, except the resulting files cannot be
used in poll because they aren't part of the hard-coded list.
Replace this with an accessor method for the pollable type in
`WasiFile`. Because we provide a default implementation of the method
and manually implement it on all the hard-coded types, this is backwards
compatible.
Signed-off-by: Nathaniel McCallum <nathaniel@profian.com>
1. This makes it easier for implementors to deal with internal APIs.
2. This matches the signatures of the WASI Snapshot traits.
Although it is likely true that these methods would have to become
immutable in order to implement threading efficiently, threading will
impact a large number of existing traits. So this change is practical
for now with an already-unavoidable change required for threading.
Signed-off-by: Nathaniel McCallum <nathaniel@profian.com>
Previously (as in an hour ago) #3905 landed a new ability for fuzzing to
arbitrarily insert padding between functions. Running some fuzzers
locally though this instantly hit a lot of problems on AArch64 because
the arbitrary padding isn't aligned to 4 bytes like all other functions
are. To fix this issue appending functions now correctly aligns the
output as appropriate for the platform. The alignment argument for
appending was switched to `None` where `None` means "use the platform
default" and otherwise and explicit alignment can be specified for
inserting other data (like arbitrary padding or Windows unwind tables).
* fuzz: Fuzz padding between compiled functions
This commit hooks up the custom
`wasmtime_linkopt_padding_between_functions` configuration option to the
cranelift compiler into the fuzz configuration, enabling us to ensure
that randomly inserting a moderate amount of padding between functions
shouldn't tamper with any results.
* fuzz: Fuzz the `Config::generate_address_map` option
This commit adds fuzz configuration where `generate_address_map` is
either enabled or disabled, unlike how it's always enabled for fuzzing
today.
* Remove unnecessary handling of relocations
This commit removes a number of bits and pieces all related to handling
relocations in JIT code generated by Wasmtime. None of this is necessary
nowadays that the "old backend" has been removed (quite some time ago)
and relocations are no longer expected to be in the JIT code at all.
Additionally with the minimum x86_64 features required to run wasm code
it should be expected that no libcalls are required either for
Wasmtime-based JIT code.
A recent fuzz failure showed that at least `call.wast` requires a memory
larger than 10 pages, so increase the minimum number of pages that can
be used for executing spec tests.
* Implement runtime checks for compilation settings
This commit fills out a few FIXME annotations by implementing run-time
checks that when a `Module` is created it has compatible codegen
settings for the current host (as `Module` is proof of "this code can
run"). This is done by implementing new `Engine`-level methods which
validate compiler settings. These settings are validated on
`Module::new` as well as when loading serialized modules.
Settings are split into two categories, one for "shared" top-level
settings and one for ISA-specific settings. Both categories now have
allow-lists hardcoded into `Engine` which indicate the acceptable values
for each setting (if applicable). ISA-specific settings are checked with
the Rust standard library's `std::is_x86_feature_detected!` macro. Other
macros for other platforms are not stable at this time but can be added
here if necessary.
Closes#3897
* Fix fall-through logic to actually be correct
* Use a `OnceCell`, not an `AtomicBool`
* Fix some broken tests
In some use cases it is desirable to provide a custom snapshot WASI
context. Facilitate this by depending on a combination of traits
required rather than concrete type in the signature.
Signed-off-by: Roman Volosatovs <rvolosatovs@riseup.net>
If either end stack overflows we can't validate the other side since the
other side, depending on codegen settings, may have been successful, hit
a different trap, or also stack overflowed.
This commit fixes calling the call hooks configured in a store for host
functions defined with `Func::new_unchecked` or similar. I believe that
this was just an accidental oversight and there's no fundamental reason
to not support this.
Another instance similar to #3879 where when doing differential tests
the pooling allocator configuration needs to be updated to allow for a
possible table.
* Move spec interpreter fuzzing behind a Cargo feature
Building the spec interpreter requires a local installation of Ocaml and
now libgmp which isn't always available, so this enables the ability to
disable building the spec interpreter by using `cargo +nightly fuzz
build --no-default-features`. The spec interpreter is still built by
default but if fuzzers are being built locally and the spec interpreter
isn't needed then this should enable it to be relatively easily
opted-out of.
* Tweak manifest directives
This commit removes some `.unwrap()` annotations around casts between
integers to either be infallible or handle errors. This fixes a panic in
a fuzz test case that popped out for memory64-using modules. The actual
issue here is pretty benign, we were just too eager about assuming
things fit into 32-bit.
