* Update to latest wasm-tools crates
This commit pushes through the full update of the wasm-tools crates
through Wasmtime. There are two major features which changed, both
related to components, which required updates in Wasmtime:
* Resource types are now implemented in wasm-tools and they're not yet
implemented in Wasmtime so I've stubbed out the integration point with
panics as reminders to come back and implement them.
* There are new validation rules about how aggregate types must be
named. This doesn't affect runtime internals at all but was done on
behalf of code generators. This did however affect a number of tests
which have to ensure that types are exported.
* Fix more tests
* Add vet entries
* Switch wasmtime-wasi-http to using Wasmtime's version
This should use the same versioning scheme as all the other `wasmtime-*`
crates.
* Fixup more directives
This commit updates the "integrating with Wasmtime" link in
crates/wiggle/README.md which currently results in a 404 Not Found.
I hope the link in this commit is correct, but if not hopefully someone
will point out the correct page this link should point to if this is not
the case.
Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
* delete adapter src/main.o: this was accidentally left out of #165
* move adapter, byte-array, and verify to a new workspace
* rename byte-array crate to a name available on crates.io
* add a readme for verify, also give it a slightly better name
* CI: wit dep check in its own step, verify before publish, trim down publication
* reactor-tests: delete deps symlinks
* reactor-tests: manage wit with wit-deps
* test: dont set default toolchain to nightly
* wit-deps lock adapter
* wit-deps lock reactor-tests
wit-deps doesnt manage these for some reason
* x64: Add non-SSE4.1 lowerings for `extractlane`
This affects some specializations of extract-to-memory which involved
adding guards, and for the general cases involved adding specialized
lowerings that don't use the `pextr*` instructions.
Additionally this also removes the optimization where an
extend-to-larger-width operation was removed after an `extractlane`
since now the lowerings for `extractlane` may conditionally
auto-zero-extend and may not. To avoid any subtle future bugs here the
optimization is removed, and if needed in the future it's hoped by that
point it's either easier to add back in or it's more clear why the
maintenance burden should be required.
* Add more comments
* Update coredump generation in the cli to use wasm_encoder
* Add deny.toml exception for wasm-encoder 0.25.0
* add missing newline
* update custom section in fuzzing crate
Instead of passing environment variables from the spec file to the
sandboxed environment, we set them for the execution environment of the
process (therefore they were not available for WASM test application).
It was an alias, defined in `build.rs`, for
cfg(any(feature = "cranelift", feature = "winch"))
But it wasn't actually saving us much typing because we also have to have
#[cfg_attr(nightlydoc, doc(cfg(any(feature = "cranelift", feature = "winch"))))]
for all the public APIs that are gated on a compiler being present so that the
API docs show which methods require which features.
So I'm removing it and using the full `cfg(any(..))` form instead.
This allows us to remove a `build.rs` that is otherwise unused and helps
`rust-analyzer` give better code completions and jump-to-definitions and things.
* Fold guest profiling flags into `--profile` CLI option
This commit removes the `--profile-guest` and `--profile-guest-interval`
CLI flags and folds them into the preexisting `--profile` flag.
Specifying guest profiling can now be done with `--profile guest` which
has a default path/interval to write to. The path/interval can be
configured with `--profile guest,path.json` and `--profile
guest,path.json,10ms`.
* Update src/commands/run.rs
Co-authored-by: Nick Fitzgerald <fitzgen@gmail.com>
---------
Co-authored-by: Nick Fitzgerald <fitzgen@gmail.com>
* Reorganize profiling-related code
This commit is a bit of reorganization around profiling-related code in
Wasmtime with the aim of eventually simplifying it a bit more. The
changes here are:
* All exposed agents are removed and instead only constructor functions
returning trait objects are now exposed.
* All `*_disabled.rs` files and modules are removed in favor of a
function that returns a result (less trait impls).
* All `*_linux.rs` files where renamed to just `*.rs`. (less files in
play)
* The `pid` and `tid` arguments were removed as they were only used by
the jitdump profiler and now that manages it internally.
* Registering an entire ELF image is now part of the trait rather than
buried within the trampoline code in Wasmtime.
* Remove DWARF support from jitdump
In general Wasmtime's support for DWARF is not great so this is rarely
used and at least in my experience this hasn't been necessary to get
good information from perf. This commit removes the processing here
which while probably useful is probably not necessary and otherwise
makes the jidump agent a bit of an odd-one-out relative among the other
agents.
* Remove now no-longer-needed `dbg_image` argument
* Only grab the jitdump lock once-per-module
Refactor slightly to account for this.
* Fill in the `tid` argument to jitump
This has been the same as `self.pid` for quite some time but with
`rustix` it's pretty easy to get access to the current thread id.
* Merge module/trampoline registration for profilers
Add a second argument to registration of an entire module for custom
names to get functions named correctly, and otherwise profilers now only
need to look at individual functions.
* Fixup vtune support
* Delete no-longer-needed accessors
Closes#6328
* Remove unused import
* Fix a portability issue with u32-vs-i32
Bake in the `*mut VMContext` to `&mut Instance` translation into the
macro-generated trampolines to avoid the need to use
`Instance::from_vmctx` with an extra level of indentation everywhere.
Additionally some libcalls are now entirely safe code as their one
unsafe operation was the `VMContext` to `Instance` translation.
* Make Wasmtime compatible with Stacked Borrows in MIRI
The fact that Wasmtime executes correctly under Tree Borrows but not
Stacked Borrows is a bit suspect and given what I've since learned about
the aliasing models I wanted to give it a stab to get things working
with Stacked Borrows. It turns out that this wasn't all that difficult,
but required two underlying changes:
* First the implementation of `Instance::vmctx` is now specially crafted
in an intentional way to preserve the provenance of the returned
pointer. This way all `&Instance` pointers will return a `VMContext`
pointer with the same provenance and acquiring the pointer won't
accidentally invalidate all prior pointers.
* Second the conversion from `VMContext` to `Instance` has been updated
to work with provenance and such. Previously the conversion looked
like `&mut VMContext -> &mut Instance`, but I think this didn't play
well with MIRI because `&mut VMContext` has no provenance over any
data since it's zero-sized. Instead now the conversion is from `*mut
VMContext` to `&mut Instance` where we know that `*mut VMContext` has
provenance over the entire instance allocation. This shuffled a fair
bit around to handle the new closure-based API to prevent escaping
pointers, but otherwise no major change other than the structure and
the types in play.
This commit additionally picks up a dependency on the `sptr` crate which
is a crate for prototyping strict-provenance APIs in Rust. This is I
believe intended to be upstreamed into Rust one day (it's in the
standard library as a Nightly-only API right now) but in the meantime
this is a stable alternative.
* Clean up manual `unsafe impl Send` impls
This commit adds a new wrapper type `SendSyncPtr<T>` which automatically
impls the `Send` and `Sync` traits based on the `T` type contained.
Otherwise it works similarly to `NonNull<T>`. This helps clean up a
number of manual annotations of `unsafe impl {Send,Sync} for ...`
throughout the runtime.
* Remove pointer-to-integer casts with tables
In an effort to enable MIRI's "strict provenance" mode this commit
removes the integer-to-pointer casts in the runtime `Table`
implementation for Wasmtime. Most of the bits were already there to
track all this, so this commit plumbed around the various pointer types
and with the help of the `sptr` crate preserves the provenance of all
related pointers.
* Remove integer-to-pointer casts in CoW management
The `MemoryImageSlot` type stored a `base: usize` field mostly because I
was too lazy to have a `Send`/`Sync` type as a pointer, so this commit
updates it to use `SendSyncPtr<u8>` and then plumbs the pointer-ness
throughout the implementation. This removes all integer-to-pointer casts
and has pointers stores as actual pointers when they're at rest.
* Remove pointer-to-integer casts in "raw" representations
This commit changes the "raw" representation of `Func` and `ExternRef`
to a `*mut c_void` instead of the previous `usize`. This is done to
satisfy MIRI's requirements with strict provenance, properly marking the
intermediate value as a pointer rather than round-tripping through
integers.
* Minor remaining cleanups
* Switch to Stacked Borrows for MIRI on CI
Additionally enable the strict-provenance features to force warnings
emitted today to become errors.
* Fix a typo
* Replace a negative offset with `sub`
* Comment the sentinel value
* Use NonNull::dangling
In #6338 that PR is bouncing on CI due to Windows running out of disk
space. Locally a `./ci/run-tests.sh` compile produces a ~13G build
directory. Testing on CI it looks like Windows builders have ~13G of
free space on them for GitHub Actions. I think something in that PR has
tipped us just over the edge of requiring too much space, although I'm
not sure exactly what.
To address the issue this commit disables DWARF debug information
entirely on all builders on CI. Not only should this save a sliver of
compile time but this reduces a local build directory to ~4.7G, over a
50% reduction. That should keep us fitting in CI for more time to come
hopefully.
* x64: Add non-SSE 4.1 lowerings for `insertlane` instructions
This commit avoids the use of `pinsr*` when SSE 4.1 is enabled by using
alternative means of inserting values into vectors.
* Clarify comments
* Remove the initializer from a global's type information
This commit removes the `Global::initializer` field to instead only
store type information about the global rather than its initialization
state. Instead now the initializer is stored separately from the type of
the global, and only for defined globals. This removes the need in a few
locations to synthesize dummy initializers.
* Actually delete what I intended to delete
* Simplify global initializer loop
* Remove lingering references to `experimental_x64`
This hasn't been experimental for quite a long time now and these are
all mostly old vestiges of when the current backend was under
development, so remove them all as they're no longer necessary.
* Try to fix test
* Try again to fix tests
* add fuel parameter to control plane
Co-authored-by: Falk Zwimpfer <24669719+FalkZ@users.noreply.github.com>
Co-authored-by: Moritz Waser <mzrw.dev@pm.me>
* remove unused cranelift setting
Co-authored-by: Falk Zwimpfer <24669719+FalkZ@users.noreply.github.com>
Co-authored-by: Moritz Waser <mzrw.dev@pm.me>
* update fuel parameter documentation
Co-authored-by: Falk Zwimpfer <24669719+FalkZ@users.noreply.github.com>
Co-authored-by: Moritz Waser <mzrw.dev@pm.me>
* include control plane in printed test case
Co-authored-by: Falk Zwimpfer <24669719+FalkZ@users.noreply.github.com>
Co-authored-by: Moritz Waser <mzrw.dev@pm.me>
* print control plane as comment
Co-authored-by: Falk Zwimpfer <24669719+FalkZ@users.noreply.github.com>
Co-authored-by: Moritz Waser <mzrw.dev@pm.me>
* set fuel limit in cranelift settings
Co-authored-by: Falk Zwimpfer <24669719+FalkZ@users.noreply.github.com>
Co-authored-by: Moritz Waser <mzrw.dev@pm.me>
* separate cli arg parsing from arbitrary impl
`ControlPlane::new` is replaced with an `Arbitrary` implementation
and `ControlPlane::set_fuel`, as it was before.
The CLI argument parsing inside the fuzz target `cranelift-fuzzgen`
is separated from the `Arbitrary` implementation of `TestCase`.
To achieve this, the `TestCase` doesn't carry a build TargetIsa
anymore, but it it generated with an isa- and flags-builder.
The CLI argument can then modify the flags further before the
`TargetIsa` is built.
The `TestCase.isa_builder` is wrapper in an `Rc` such that optimized
test cases can share the same one and it doesn't need to be made
`Clone`.
Co-authored-by: Falk Zwimpfer <24669719+FalkZ@users.noreply.github.com>
Co-authored-by: Moritz Waser <mzrw.dev@pm.me>
* parse control planes with clif-util
Co-authored-by: Falk Zwimpfer <24669719+FalkZ@users.noreply.github.com>
Co-authored-by: Moritz Waser <mzrw.dev@pm.me>
* move parsing logic from cranelift-control to cranelift-reader
Co-authored-by: Falk Zwimpfer <24669719+FalkZ@users.noreply.github.com>
Co-authored-by: Moritz Waser <mzrw.dev@pm.me>
* remove parsing and settings plumbing
Co-authored-by: Falk Zwimpfer <24669719+FalkZ@users.noreply.github.com>
Co-authored-by: Moritz Waser <mzrw.dev@pm.me>
---------
Co-authored-by: Falk Zwimpfer <24669719+FalkZ@users.noreply.github.com>
Co-authored-by: Moritz Waser <mzrw.dev@pm.me>
Alex Crichton
Pat Hickey
Pat Hickey
Daniel Bevenius
Trevor Elliott
theothergraham
Eduardo de Moura Rodrigues
Rainy Sinclair
Marcin Kolny
Afonso Bordado
Nick Fitzgerald
Juan Jose Medina
Remo Senekowitsch