* Remove usage of `BTreeMap` for compiler flags
No need for a datastructure here really, a simple list with static
strings works alright.
* Fix winch compile and a warning
* Fix test compile
* c-api: Remove reexport of wasmtime crate
This is a follow-up to #6765 to remove this reexport since it was
originally added to use both the C API and the `wasmtime` crate in the
same downstream crate, but this should be possible through Cargo with:
[dependencies]
wasmtime = "13"
wasmtime-c-api = { version = "13", package = "wasmtime-c-api" }
and that way `wasmtime::*` is available as well as `wasmtime_c_api::*`
* Add `pub use wasmtime;`
* Enable threads, multi-memory, and relaxed-simd by default
This commit enables by default three proposals that were advanced to
stage 4 in the proposal process at last week's in-person CG meeting.
These proposals are:
* `threads` - atomic instructions and shared memory
* `multi-memory` - the ability to have more than one linear memory
* `relaxed-simd` - new simd instructions that may differ across
platforms
These proposals are now all enabled by default in Wasmtime. Note that
they can still be selectively disabled via `Config` options.
* Fix an x64-specific test
This turns out to be quite simple: the fundamental operation during
egraph-based optimization is to *merge* eclasses, which is an assertion
that their value is equal. Since the values of either side of the merge
are equal, a fact about one side is a fact about the other, and
vice-versa.
Since we only support *one* fact at most per value, we can't take the
union of all facts; instead, if one side is missing a fact, or if both
sides have exactly the same fact, we keep that one; otherwise we go to a
special "conflict" fact that can't support any check. This is edging
closer to a fact-lattice, but I opted not to go there with a full
meet-function merge yet, for simplicity. (It's a little complex because
of the "minimum fact" we can know about a value based on its type -- if
we're going to do something better, I think we should account for that
too.)
In any case, that complexity mostly isn't needed, because the two cases
that happen in reality are (i) opt rules rewrite to a new node, which
will have no facts at all, so facts just propagate; or (ii) GVN merges
two values in the input program into one, but if both are the same
value, in practice the Wasm PCC frontend (for example) should be
producing the same facts on both values, so the merge is trivial.
* PCC: add facts to global values, parse and print them. No verification yet.
Co-authored-by: Nick Fitzgerald <fitzgen@gmail.com>
* PCC: propagate facts on GV loads and check them.
Co-authored-by: Nick Fitzgerald <fitzgen@gmail.com>
* PCC: support propagating facts on iteratively-elaborated GVs as well.
Co-authored-by: Nick Fitzgerald <fitzgen@gmail.com>
* PCC: fix up Wasmtime uses of GVs after refactors to memflags handling.
Co-authored-by: Nick Fitzgerald <fitzgen@gmail.com>
* PCC: working end-to-end for static memories!
Co-authored-by: Nick Fitzgerald <fitzgen@gmail.com>
* PCC: add toplevel Wasmtime option `-C enable-pcc=y`.
* Fix filetests build.
* Review feedback, and blessed test updates due to GV legalization changes.
---------
Co-authored-by: Nick Fitzgerald <fitzgen@gmail.com>
This is needed for soundness when verifying accesses to memtype fields:
it's not enough to know that we're accessing an offset in `0` up to
`field_offset` inclusive, we need to know the access is actually to
`field_offset`.
The simplest change that validates this turned out to be the most
general one: making ranges two-sided rather than one-sided. The
transform is *mostly* mechanical, but a few new tests verify that ranges
are updated on both sides, and some fail-tests verify that "fuzzily
imprecise" pointers to struct fields fail to validate.
Some notable changes here are:
* The `wasm-tools` crates have enabled the `relaxed-simd`, `threads`,
and `multi-memory` proposals by default. For now I've left these
disabled-by-default in Wasmtime to get enabled in a future PR.
* The `wast` crate has support for parsing `thread` and `wait`
constructs from the `threads` proposal for WebAssembly. They're left
unimplemented for now and return errors. This will get filled in in a
future update.
This removes the need for the awkward "max-range fact is subsumed by
anything" rule noted by @fitzgen in [this
comment](https://github.com/bytecodealliance/wasmtime/pull/7231#discussion_r1358573147).
It also makes checking a little more efficient and logically clear, as
only the facts that the frontend/producer added are verified, rather
than all default facts as well.
* riscv64: Refactor `rotl` rules
Move from `inst.isle` to `lower.isle` since it's the only caller,
reorganize the rules to be a bit cleaner, add immediate shifting
specializations.
* riscv64: Refactor `rotr` lowerings
Same as the prior `rotl` lowerings, move the rules to `lower.isle` and
additionally add constant rules.
* Fix shift-by-128
* Remove empty comments
* cranelift: Add egraph rules for `bswap`
WebAssembly doesn't currently have a byte-swapping instruction so this
commit pattern matches what LLVM currently generates to simplify to a
single `bswap` instruction in CLIF which can be lowered within each
backend to respective native instructions.
* Use an `optimize` instead of `compile` test
This commit optimizes some `select_spectre_guard` patterns which are
frequently generated with the wasm backend when dynamic bounds checks
are enabled by pattern-matching when one of the values being selected is
zero. This enables shaving off a few instructions which can be constant
folded away.
* Support set_fuel in store APIs
Fixes: https://github.com/bytecodealliance/wasmtime/issues/5109
Signed-off-by: Tyler Rockwood <rockwood@redpanda.com>
* rename set_fuel to reset_fuel
To make it more clear that consumed fuel is being reset.
Signed-off-by: Tyler Rockwood <rockwood@redpanda.com>
* update out of date documentation for fuel in C API
Signed-off-by: Tyler Rockwood <rockwood@redpanda.com>
---------
Signed-off-by: Tyler Rockwood <rockwood@redpanda.com>
This commit removes a clause in `register_module` which was preventing
`perf` profiling the first function in a module because it was located
at address 0. This check is relatively dated at this point and I think I
originally added it in error (and/or it was copied from somewhere else),
so this commit removes it.
Some things I noticed from #7239 which are very much not critical but I
figure might be nice-to-haves:
* Move the logging configuration to the `wasmtime-cli-flags` crate with
the other logging configuration to keep it in one place.
* Remove `pretty_env_logger` since `tracing-subscriber` probably
supplants it.
* Don't explicitly inherit env vars in tests since that happens
automatically with `Command`.
* riscv64: Constants are always sign-extended
Skip generating extra sign-extension instructions in this case because
the materialization of a constant will implicitly sign-extend into the
entire register.
* riscv64: Rename `lower_int_compare` helper
Try to reserve `lower_*` as taking a thing and producing a `*Reg`.
Rename this helper to `is_nonzero_cmp` to represent how it's testing for
nonzero and producing a comparison.
* riscv64: Rename some float comparison helpers
* `FCmp` -> `FloatCompare`
* `emit_fcmp` -> `fcmp_to_float_compare`
* `lower_fcmp` -> `lower_float_compare`
Make some room for upcoming integer comparison functions.
* riscv64: Remove `ordered` helper
This is only used by one lowering so inline its definition directly.
* riscv64: Remove the `Icmp` pseudo-instruction
This commit is the first in a few steps to reimplement #6112 and #7113.
The `Icmp` pseudo-instruction is removed here and necessary
functionality is all pushed into ISLE lowerings. This enables deleting
the `lower_br_icmp` helper in `emit.rs` as it's no longer necessary,
meaning that all conditional branches should ideally be generated in
lowering rather than pseudo-instructions to benefit from various
optimizations.
Currently the lowering is the bare-bones minimum to get things working.
This involved creating a helper to lower an `IntegerCompare` into an
`XReg` via negation/swapping args/etc. In generated code this removes
branches and makes `icmp` a straight-line instruction for non-128-bit
arguments.
* riscv64: Remove an unused helper
* riscv64: Optimize comparisons with 0
Use the `x0` register which is always zero where possible and avoid
unnecessary `xor`s against this register.
* riscv64: Specialize `a < $imm`
* riscv64: Optimize Equal/NotEqual against constants
* riscv64: Optimize LessThan with constant argument
* Optimize `a <= $imm`
* riscv64: Optimize `a >= $imm`
* riscv64: Add comment for new helper
* Use i64 in icmp optimizations
Matches the sign-extension that happens at the hardware layer.
* Correct some sign extensions
* riscv64: Don't assume immediates are extended
* riscv64: Fix encoding for `c.addi4spn`
* riscv64: Remove `icmp` lowerings which modify constants
These aren't correct and will need updating
* Add regression test
* riscv64: Fix handling unsigned comparisons with constants
---------
Co-authored-by: Afonso Bordado <afonsobordado@az8.co>
In #7220, we began logging the arguments of WebAssembly `wait` and
`notify` as these instructions are executed. This change prints the
addresses using hexadecimal, which is simply more convenient when
dealing with addressing.
* winch: Add known a subset of known libcalls and improve call emission
This change is a follow up to:
- https://github.com/bytecodealliance/wasmtime/pull/7155
- https://github.com/bytecodealliance/wasmtime/pull/7035
One of the objectives of this change is to make it easy to emit
function calls at the MacroAssembler layer, for cases in which it's
challenging to know ahead-of-time if a particular functionality can be
achieved natively (e.g. rounding and SSE4.2). The original implementation
of function call emission, made this objective difficult to achieve and
it was also difficult to reason about.
I decided to simplify the overall approach to function calls as part of
this PR; in essence, the `call` module now exposes a single function
`FnCall::emit` which is reponsible of gathtering the dependencies and
orchestrating the emission of the call. This new approach deliberately
avoids holding any state regarding the function call for simplicity.
This change also standardizes the usage of `Callee` as the main
entrypoint for function call emission, as of this change 4 `Callee`
types exist (`Local`, `Builtin`, `Import`, `FuncRef`), each callee kind
is mappable to a `CalleeKind` which is the materialized version of
a callee which Cranelift understands.
This change also moves the creation of the `BuiltinFunctions` to the
`ISA` level given that they can be safely used accross multiple function
compilations.
Finally, this change also introduces support for some of the
"well-known" libcalls and hooks those libcalls at the
`MacroAssembler::float_round` callsite.
--
prtest:full
* Review comments
* Remove unnecessary `into_iter`
* Fix remaining lifetime parameter names
* Update to arbitrary 1.3.1
And use workspace dependencies for arbitrary.
* Prune cargo vet's supply-chain files
This is the mechanical changes made by running `cargo vet prune` which was
suggested to me when I ran `cargo vet`.
* PCC: add checking of struct field types on loads and stores.
Co-authored-by: Nick Fitzgerald <fitzgen@gmail.com>
* PCC: check that blockparams uphold facts appropriately.
Co-authored-by: Nick Fitzgerald <fitzgen@gmail.com>
* PCC: infer facts based on types' value ranges; add a mock vmctx example.
Co-authored-by: Nick Fitzgerald <fitzgen@gmail.com>
* Fix formatting.
* Add two fixes required by new default facts:
- Allow any fact at all to subsume a trivially-true range fact that
states a bitslice is within the maximum range for its width.
- Clamp the maximum value of a bitslice at the maximum value for the
slice's width.
---------
Co-authored-by: Nick Fitzgerald <fitzgen@gmail.com>
* wit: delete wasi logging package
missed this in a code review
* command extended: use include keyword, delete import of logging
* sync http wit definitions
* delete redundant cli test
cli_panic is almost exactly the same as cli_exit_panic, so delete the first one
* split stream pollable lifetimes test into two preview2 tests
* PCC: define memory types and add some examples in filetests (no parsing yet).
* PCC: add a notion of memory types.
Co-authored-by: Nick Fitzgerald <fitzgen@gmail.com>
* Transition `points_to` to a `memory` (static memory) memory-type.
* Review feedback.
---------
Co-authored-by: Nick Fitzgerald <fitzgen@gmail.com>
When troubleshooting deadlocks in WebAssembly modules, it is important
to understand which `wait` instructions are still pending a `notify`.
It would be nice to have some kind of `--warn-deadlock-after=1s` flag
available that would poll the parking lot for `wait`s hanging past the
time limit, but I realized the real value would be to tie the `wait`
instruction (through CLIF) to the original source code, if debug
information were available. This did not seem to be entirely feasible,
since CLIF loses the original Wasm source context (is this true?) and I
was not confident that we would be able to use `addr2line` to map from
Wasm instructions to source (e.g., see @cfallin's
[issue](https://github.com/gimli-rs/addr2line/issues/265)).
Instead, this change simply logs each valid `wait` and `notify`
execution, leaving it to the user to figure out which one is hanging
(should not be too difficult) and how to map this back to their source
code (more difficult).
* Introduce StackMemory and StackMemoryCreator
This allows custom implementations of stack memory to be plugged into
the async functionality for wasmtime. Currently, stacks are always
mmapped, and this custom allocator allows embedders to use any memory
they would like.
The new APIs are also exposed in the C api.
This has no effect on windows, as our hands are tied with fibers there.
Signed-off-by: Tyler Rockwood <rockwood@redpanda.com>
* Add test for custom host memory
Signed-off-by: Tyler Rockwood <rockwood@redpanda.com>
* fix allocator test
Signed-off-by: Tyler Rockwood <rockwood@redpanda.com>
* Address review comments
Signed-off-by: Tyler Rockwood <rockwood@redpanda.com>
* Fix lint warnings
prtest:full
Signed-off-by: Tyler Rockwood <rockwood@redpanda.com>
* fix windows lint warning
prtest:full
Signed-off-by: Tyler Rockwood <rockwood@redpanda.com>
---------
Signed-off-by: Tyler Rockwood <rockwood@redpanda.com>
* Update MSRV to 1.71.0
With the 1.73.0 release today update our MSRV to a one-higher version.
This will cause 1.73.0 to be tested in CI.
* Fix a warning on windows
* Don't store tables as `&'static mut`
Use `SendSyncPtr` instead (a small wrapper around `NonNull`) which
appeases MIRI.
* Move around some internal methods and internalize what we can.
* Rename all `*_resource` methods to drop the suffix.
* Lean more into "wrong table usage is a trap" and remove the `delete`
logic which re-inserted a parent link, if needed, on downcast errors.
* Refactor the test-programs test suite
This commit is a large refactoring that reorganizes `test-programs` and
how we tests wasms in Wasmtime. Often writing tests requires complicated
interactions with the guest which can't be done via hand-written `*.wat`
syntax and requires a compiler to get engaged. For this purpose Wasmtime
currently has the `crates/test-programs/*` test suite which builds files
from source and then runs the tests. This has been somewhat cumbersome
in the past though and it's not been easy to extend this over time, so
this commit attempts to address this.
The scheme implemented in this PR looks like:
* All wasm test programs live in `crates/test-programs/src/bin/*.rs`.
All of them, no exceptions.
* Wasm tests have shared support located at
`crates/test-programs/src/lib.rs` and its submodules, such as bindings
generation for WASI.
* Wasm tests are built by a new `crates/test-programs/artifacts` crate.
This crate compiles modules and additionally creates components for
all test programs. The crate itself only records the path to these
outputs and a small amount of testing support, but otherwise doesn't
interact with `wasmtime`-the-crate itself.
* All tests in `crates/test-programs/tests/*.rs` have moved. For example
wasi-http tests now live at `crates/wasi-http/tests/*.rs`. Legacy
tests of wasi-common now live at `crates/wasi-common/tests/*.rs`.
Modern tests for preview2 live at `crates/wasi/tests/*.rs`.
* Wasm tests are bucketed based on their filename prefix. For example
`preview1_*` is tested in wasi-common and wasmtime-wasi. The
`preview2_*` prefix is only tested with wasmtime-wasi, however.
* A new `cli_*` prefix is used to execute tests as part of
`tests/all/main.rs`. This is a new submodule in
`tests/all/cli_tests.rs` which executes these components on the
command line. Many old "command" tests were migrated here.
* Helper macros are generated to assert that a test suite is run in its
entirety. This way if a `preview1_*` test is added it's asserted to
get added to both wasi-common and wasmtime-wasi in the various modes
they run tests.
Overall this moved a number of tests around and refactored some edges of
the tests, but this should not lose any tests (except one that wasn't
actually testing anything). Additionally the hope is that it's much
easier to add tests in the future. The process is to add a new file in
`crates/test-programs/src/bin/*.rs` named appropriately. For example a
preview2 executable is `preview2_*` and a CLI tests is `cli_*`. When
building the test suite an error is generated in the appropriate module
then of "please write a test here", and then a test is written in the
same manner as the other tests in the module.
* Remove no-longer-needed fetches
prtest:full
* I'm worried wasi is running low on semicolons
* Add the WASI target in all CI actions
* Add unknown-unknown target on all CI builders too
* Fix building test artifacts under miri
Need to avoid wrappers for these cross-compiled targets
* Break circular dependency for packaging
Don't use the workspace dep for `wasmtime-wasi` since it injects a
version, instead use a `path = '..'` dependency to fool Cargo into
dropping the dependency during the package phase.
* Fix some merge conflicts with tests
* Fix rebase for new tests
* Remove stray comment
* Fix some flaky tests
* Fix network tests in synchronous mode
This commit is an attempt to fix some networking tests in synchronous
mode in our test suite. Currently networking tests don't actually run in
synchronous mode on CI which is why no failures have been surfaced yet,
but the refactoring in #7182 is going to start doing this.
Currently the `udp_sample_application.rs` test blocks infinitely in
synchronous mode for me locally, most of the time. This appears to be an
interaction between how Tokio handles readiness and how we're
entering the event loop. We're effectively entering the Tokio event loop
with a future that's always ready which ends up starving Tokio of
otherwise performing its background work such as updating flags for
readiness of reading/writing.
The fix here is to add a yield at the start of an `in_tokio` block which
is used in synchronous mode. This is a kludge fix but the intention is
to enable Tokio to have a chance to update readiness flags and process
events from epoll/kqueue/etc.
An additional fix to this issue is WebAssembly/wasi-sockets#64 where the
test is waiting on `READABLE` or `WRITABLE`, but in this specific case
it should only wait on `READABLE`. If it waited on just this then that
would also fix this issue. Nevertheless having a `yield_now` is expected
to have little-to-no overhead and otherwise fix this edge case of an
always-ready future.
* Fix passing empty arguments on the CLI
* Add another blocking accept
* Update crates/test-programs/src/bin/api_proxy.rs
Co-authored-by: Trevor Elliott <awesomelyawesome@gmail.com>
---------
Co-authored-by: Trevor Elliott <awesomelyawesome@gmail.com>
Alex Crichton
Chris Fallin
Jeffrey Charles
Afonso Bordado
Pat Hickey
Tyler Rockwood
Andrew Brown
Saúl Cabrera
Nick Fitzgerald
Kelly Thomas Kline
Muhammad Hamza
Mathew Shen
Ryan Levick