3 [![Actions build status][actions-badge]][actions-url]
5 [actions-badge]: https://github.com/rust-lang/miri/workflows/CI/badge.svg?branch=master
6 [actions-url]: https://github.com/rust-lang/miri/actions
8 An experimental interpreter for [Rust][rust]'s
9 [mid-level intermediate representation][mir] (MIR). It can run binaries and
10 test suites of cargo projects and detect certain classes of
11 [undefined behavior](https://doc.rust-lang.org/reference/behavior-considered-undefined.html),
14 * Out-of-bounds memory accesses and use-after-free
15 * Invalid use of uninitialized data
16 * Violation of intrinsic preconditions (an [`unreachable_unchecked`] being
17 reached, calling [`copy_nonoverlapping`] with overlapping ranges, ...)
18 * Not sufficiently aligned memory accesses and references
19 * Violation of *some* basic type invariants (a `bool` that is not 0 or 1, for example,
20 or an invalid enum discriminant)
21 * **Experimental**: Violations of the [Stacked Borrows] rules governing aliasing
23 * **Experimental**: Data races (but no weak memory effects)
25 On top of that, Miri will also tell you about memory leaks: when there is memory
26 still allocated at the end of the execution, and that memory is not reachable
27 from a global `static`, Miri will raise an error.
29 You can use Miri to emulate programs on other targets, e.g. to ensure that
30 byte-level data manipulation works correctly both on little-endian and
31 big-endian systems. See
32 [cross-interpretation](#cross-interpretation-running-for-different-targets)
35 Miri has already discovered some [real-world bugs](#bugs-found-by-miri). If you
36 found a bug with Miri, we'd appreciate if you tell us and we'll add it to the
39 However, be aware that Miri will **not catch all cases of undefined behavior**
40 in your program, and cannot run all programs:
42 * There are still plenty of open questions around the basic invariants for some
43 types and when these invariants even have to hold. Miri tries to avoid false
44 positives here, so if your program runs fine in Miri right now that is by no
45 means a guarantee that it is UB-free when these questions get answered.
47 In particular, Miri does currently not check that integers/floats are
48 initialized or that references point to valid data.
49 * If the program relies on unspecified details of how data is laid out, it will
50 still run fine in Miri -- but might break (including causing UB) on different
51 compiler versions or different platforms.
52 * Program execution is non-deterministic when it depends, for example, on where
53 exactly in memory allocations end up, or on the exact interleaving of
54 concurrent threads. Miri tests one of many possible executions of your
55 program. You can alleviate this to some extent by running Miri with different
56 values for `-Zmiri-seed`, but that will still by far not explore all possible
58 * Miri runs the program as a platform-independent interpreter, so the program
59 has no access to most platform-specific APIs or FFI. A few APIs have been
60 implemented (such as printing to stdout) but most have not: for example, Miri
61 currently does not support SIMD or networking.
62 * Threading support is not finished yet. E.g., weak memory effects are not
63 emulated and spin loops (without syscalls) just loop forever. There is no
64 threading support on Windows.
66 [rust]: https://www.rust-lang.org/
67 [mir]: https://github.com/rust-lang/rfcs/blob/master/text/1211-mir.md
68 [`unreachable_unchecked`]: https://doc.rust-lang.org/stable/std/hint/fn.unreachable_unchecked.html
69 [`copy_nonoverlapping`]: https://doc.rust-lang.org/stable/std/ptr/fn.copy_nonoverlapping.html
70 [Stacked Borrows]: https://github.com/rust-lang/unsafe-code-guidelines/blob/master/wip/stacked-borrows.md
75 Install Miri on Rust nightly via `rustup`:
78 rustup +nightly component add miri
81 If `rustup` says the `miri` component is unavailable, that's because not all
82 nightly releases come with all tools. Check out
83 [this website](https://rust-lang.github.io/rustup-components-history) to
84 determine a nightly version that comes with Miri and install that using
85 `rustup toolchain install nightly-YYYY-MM-DD`.
87 Now you can run your project in Miri:
89 1. Run `cargo clean` to eliminate any cached dependencies. Miri needs your
90 dependencies to be compiled the right way, that would not happen if they have
91 previously already been compiled.
92 2. To run all tests in your project through Miri, use `cargo miri test`.
93 3. If you have a binary project, you can run it through Miri using `cargo miri run`.
95 The first time you run Miri, it will perform some extra setup and install some
96 dependencies. It will ask you for confirmation before installing anything.
98 `cargo miri run/test` supports the exact same flags as `cargo run/test`. For
99 example, `cargo miri test filter` only runs the tests containing `filter` in
102 You can pass arguments to Miri via `MIRIFLAGS`. For example,
103 `MIRIFLAGS="-Zmiri-disable-stacked-borrows" cargo miri run` runs the program
104 without checking the aliasing of references.
106 When compiling code via `cargo miri`, the `cfg(miri)` config flag is set for code
107 that will be interpret under Miri. You can use this to ignore test cases that fail
108 under Miri because they do things Miri does not support:
112 #[cfg_attr(miri, ignore)]
113 fn does_not_work_on_miri() {
114 tokio::run(futures::future::ok::<_, ()>(()));
118 There is no way to list all the infinite things Miri cannot do, but the
119 interpreter will explicitly tell you when it finds something unsupported:
122 error: unsupported operation: can't call foreign function: bind
124 = help: this is likely not a bug in the program; it indicates that the program \
125 performed an operation that the interpreter does not support
128 ### Cross-interpretation: running for different targets
130 Miri can not only run a binary or test suite for your host target, it can also
131 perform cross-interpretation for arbitrary foreign targets: `cargo miri run
132 --target x86_64-unknown-linux-gnu` will run your program as if it was a Linux
133 program, no matter your host OS. This is particularly useful if you are using
134 Windows, as the Linux target is much better supported than Windows targets.
136 You can also use this to test platforms with different properties than your host
137 platform. For example `cargo miri test --target mips64-unknown-linux-gnuabi64`
138 will run your test suite on a big-endian target, which is useful for testing
139 endian-sensitive code.
141 ### Running Miri on CI
143 To run Miri on CI, make sure that you handle the case where the latest nightly
144 does not ship the Miri component because it currently does not build. For
145 example, you can use the following snippet to always test with the latest
146 nightly that *does* come with Miri:
149 MIRI_NIGHTLY=nightly-$(curl -s https://rust-lang.github.io/rustup-components-history/x86_64-unknown-linux-gnu/miri)
150 echo "Installing latest nightly with Miri: $MIRI_NIGHTLY"
151 rustup set profile minimal
152 rustup override set "$MIRI_NIGHTLY"
153 rustup component add miri
160 When using the above instructions, you may encounter a number of confusing compiler
163 ### "note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace"
165 You may see this when trying to get Miri to display a backtrace. By default, Miri
166 doesn't expose any environment to the program, so running
167 `RUST_BACKTRACE=1 cargo miri test` will not do what you expect.
169 To get a backtrace, you need to disable isolation
170 [using `-Zmiri-disable-isolation`][miri-flags]:
173 RUST_BACKTRACE=1 MIRIFLAGS="-Zmiri-disable-isolation" cargo miri test
176 #### "found possibly newer version of crate `std` which `<dependency>` depends on"
178 Your build directory may contain artifacts from an earlier build that have/have
179 not been built for Miri. Run `cargo clean` before switching from non-Miri to
180 Miri builds and vice-versa.
182 #### "found crate `std` compiled by an incompatible version of rustc"
184 You may be running `cargo miri` with a different compiler version than the one
185 used to build the custom libstd that Miri uses, and Miri failed to detect that.
186 Try deleting `~/.cache/miri`.
188 #### "no mir for `std::rt::lang_start_internal`"
190 This means the sysroot you are using was not compiled with Miri in mind. This
191 should never happen when you use `cargo miri` because that takes care of setting
192 up the sysroot. If you are using `miri` (the Miri driver) directly, see the
193 [contributors' guide](CONTRIBUTING.md) for how to use `./miri` to best do that.
196 ## Miri `-Z` flags and environment variables
197 [miri-flags]: #miri--z-flags-and-environment-variables
199 Miri adds its own set of `-Z` flags, which are usually set via the `MIRIFLAGS`
200 environment variable:
202 * `-Zmiri-compare-exchange-weak-failure-rate=<rate>` changes the failure rate of
203 `compare_exchange_weak` operations. The default is `0.8` (so 4 out of 5 weak ops will fail).
204 You can change it to any value between `0.0` and `1.0`, where `1.0` means it
205 will always fail and `0.0` means it will never fail.
206 * `-Zmiri-disable-abi-check` disables checking [function ABI]. Using this flag
208 * `-Zmiri-disable-alignment-check` disables checking pointer alignment, so you
209 can focus on other failures, but it means Miri can miss bugs in your program.
210 Using this flag is **unsound**.
211 * `-Zmiri-disable-data-race-detector` disables checking for data races. Using
212 this flag is **unsound**.
213 * `-Zmiri-disable-stacked-borrows` disables checking the experimental
214 [Stacked Borrows] aliasing rules. This can make Miri run faster, but it also
215 means no aliasing violations will be detected. Using this flag is **unsound**
216 (but the affected soundness rules are experimental).
217 * `-Zmiri-disable-validation` disables enforcing validity invariants, which are
218 enforced by default. This is mostly useful to focus on other failures (such
219 as out-of-bounds accesses) first. Setting this flag means Miri can miss bugs
220 in your program. However, this can also help to make Miri run faster. Using
221 this flag is **unsound**.
222 * `-Zmiri-disable-isolation` disables host isolation. As a consequence,
223 the program has access to host resources such as environment variables, file
224 systems, and randomness.
225 * `-Zmiri-isolation-error=<action>` configures Miri's response to operations
226 requiring host access while isolation is enabled. `abort`, `hide`, `warn`,
227 and `warn-nobacktrace` are the supported actions. The default is to `abort`,
228 which halts the machine. Some (but not all) operations also support continuing
229 execution with a "permission denied" error being returned to the program.
230 `warn` prints a full backtrace when that happen; `warn-nobacktrace` is less
231 verbose. `hide` hides the warning entirely.
232 * `-Zmiri-env-exclude=<var>` keeps the `var` environment variable isolated from
233 the host so that it cannot be accessed by the program. Can be used multiple
234 times to exclude several variables. On Windows, the `TERM` environment
235 variable is excluded by default.
236 * `-Zmiri-ignore-leaks` disables the memory leak checker, and also allows some
237 remaining threads to exist when the main thread exits.
238 * `-Zmiri-measureme=<name>` enables `measureme` profiling for the interpreted program.
239 This can be used to find which parts of your program are executing slowly under Miri.
240 The profile is written out to a file with the prefix `<name>`, and can be processed
241 using the tools in the repository https://github.com/rust-lang/measureme.
242 * `-Zmiri-panic-on-unsupported` will makes some forms of unsupported functionality,
243 such as FFI and unsupported syscalls, panic within the context of the emulated
244 application instead of raising an error within the context of Miri (and halting
245 execution). Note that code might not expect these operations to ever panic, so
246 this flag can lead to strange (mis)behavior.
247 * `-Zmiri-seed=<hex>` configures the seed of the RNG that Miri uses to resolve
248 non-determinism. This RNG is used to pick base addresses for allocations.
249 When isolation is enabled (the default), this is also used to emulate system
250 entropy. The default seed is 0. **NOTE**: This entropy is not good enough
251 for cryptographic use! Do not generate secret keys in Miri or perform other
252 kinds of cryptographic operations that rely on proper random numbers.
253 * `-Zmiri-symbolic-alignment-check` makes the alignment check more strict. By
254 default, alignment is checked by casting the pointer to an integer, and making
255 sure that is a multiple of the alignment. This can lead to cases where a
256 program passes the alignment check by pure chance, because things "happened to
257 be" sufficiently aligned -- there is no UB in this execution but there would
258 be UB in others. To avoid such cases, the symbolic alignment check only takes
259 into account the requested alignment of the relevant allocation, and the
260 offset into that allocation. This avoids missing such bugs, but it also
261 incurs some false positives when the code does manual integer arithmetic to
262 ensure alignment. (The standard library `align_to` method works fine in both
263 modes; under symbolic alignment it only fills the middle slice when the
264 allocation guarantees sufficient alignment.)
265 * `-Zmiri-track-alloc-id=<id>` shows a backtrace when the given allocation is
266 being allocated or freed. This helps in debugging memory leaks and
268 * `-Zmiri-track-call-id=<id>` shows a backtrace when the given call id is
269 assigned to a stack frame. This helps in debugging UB related to Stacked
270 Borrows "protectors".
271 * `-Zmiri-track-pointer-tag=<tag>` shows a backtrace when the given pointer tag
272 is popped from a borrow stack (which is where the tag becomes invalid and any
273 future use of it will error). This helps you in finding out why UB is
274 happening and where in your code would be a good place to look for it.
275 * `-Zmiri-track-raw-pointers` makes Stacked Borrows track a pointer tag even for
276 raw pointers. This can make valid code fail to pass the checks, but also can
277 help identify latent aliasing issues in code that Miri accepts by default. You
278 can recognize false positives by `<untagged>` occurring in the message -- this
279 indicates a pointer that was cast from an integer, so Miri was unable to track
280 this pointer. Note that it is not currently guaranteed that code that works
281 with `-Zmiri-track-raw-pointers` also works without
282 `-Zmiri-track-raw-pointers`, but for the vast majority of code, this will be the case.
284 [function ABI]: https://doc.rust-lang.org/reference/items/functions.html#extern-function-qualifier
286 Some native rustc `-Z` flags are also very relevant for Miri:
288 * `-Zmir-opt-level` controls how many MIR optimizations are performed. Miri
289 overrides the default to be `0`; be advised that using any higher level can
290 make Miri miss bugs in your program because they got optimized away.
291 * `-Zalways-encode-mir` makes rustc dump MIR even for completely monomorphic
292 functions. This is needed so that Miri can execute such functions, so Miri
293 sets this flag per default.
294 * `-Zmir-emit-retag` controls whether `Retag` statements are emitted. Miri
295 enables this per default because it is needed for [Stacked Borrows].
297 Moreover, Miri recognizes some environment variables:
299 * `MIRI_LOG`, `MIRI_BACKTRACE` control logging and backtrace printing during
300 Miri executions, also [see "Testing the Miri driver" in `CONTRIBUTING.md`][testing-miri].
301 * `MIRIFLAGS` (recognized by `cargo miri` and the test suite) defines extra
302 flags to be passed to Miri.
303 * `MIRI_SYSROOT` (recognized by `cargo miri` and the test suite)
304 indicates the sysroot to use. To do the same thing with `miri`
305 directly, use the `--sysroot` flag.
306 * `MIRI_TEST_TARGET` (recognized by the test suite) indicates which target
307 architecture to test against. `miri` and `cargo miri` accept the `--target`
308 flag for the same purpose.
310 The following environment variables are *internal* and must not be used by
311 anyone but Miri itself. They are used to communicate between different Miri
312 binaries, and as such worth documenting:
314 * `MIRI_BE_RUSTC` can be set to `host` or `target`. It tells the Miri driver to
315 actually not interpret the code but compile it like rustc would. With `target`, Miri sets
316 some compiler flags to prepare the code for interpretation; with `host`, this is not done.
317 This environment variable is useful to be sure that the compiled `rlib`s are compatible
319 * `MIRI_CALLED_FROM_XARGO` is set during the Miri-induced `xargo` sysroot build,
320 which will re-invoke `cargo-miri` as the `rustc` to use for this build.
321 * `MIRI_CALLED_FROM_RUSTDOC` when set to any value tells `cargo-miri` that it is
322 running as a child process of `rustdoc`, which invokes it twice for each doc-test
323 and requires special treatment, most notably a check-only build before interpretation.
324 This is set by `cargo-miri` itself when running as a `rustdoc`-wrapper.
325 * `MIRI_CWD` when set to any value tells the Miri driver to change to the given
326 directory after loading all the source files, but before commencing
327 interpretation. This is useful if the interpreted program wants a different
328 working directory at run-time than at build-time.
329 * `MIRI_VERBOSE` when set to any value tells the various `cargo-miri` phases to
330 perform verbose logging.
332 [testing-miri]: CONTRIBUTING.md#testing-the-miri-driver
334 ## Miri `extern` functions
336 Miri provides some `extern` functions that programs can import to access
337 Miri-specific functionality:
342 /// Miri-provided extern function to mark the block `ptr` points to as a "root"
343 /// for some static memory. This memory and everything reachable by it is not
344 /// considered leaking even if it still exists when the program terminates.
346 /// `ptr` has to point to the beginning of an allocated block.
347 fn miri_static_root(ptr: *const u8);
349 /// Miri-provided extern function to obtain a backtrace of the current call stack.
350 /// This returns a boxed slice of pointers - each pointer is an opaque value
351 /// that is only useful when passed to `miri_resolve_frame`
352 /// The `flags` argument must be `0`.
353 fn miri_get_backtrace(flags: u64) -> Box<[*mut ()]>;
355 /// Miri-provided extern function to resolve a frame pointer obtained
356 /// from `miri_get_backtrace`. The `flags` argument must be `0`,
357 /// and `MiriFrame` should be declared as follows:
361 /// struct MiriFrame {
362 /// // The name of the function being executed, encoded in UTF-8
364 /// // The filename of the function being executed, encoded in UTF-8
365 /// filename: Box<[u8]>,
366 /// // The line number currently being executed in `filename`, starting from '1'.
368 /// // The column number currently being executed in `filename`, starting from '1'.
370 /// // The function pointer to the function currently being executed.
371 /// // This can be compared against function pointers obtained by
372 /// // casting a function (e.g. `my_fn as *mut ()`)
377 /// The fields must be declared in exactly the same order as they appear in `MiriFrame` above.
378 /// This function can be called on any thread (not just the one which obtained `frame`).
379 fn miri_resolve_frame(frame: *mut (), flags: u64) -> MiriFrame;
381 /// Miri-provided extern function to begin unwinding with the given payload.
383 /// This is internal and unstable and should not be used; we give it here
384 /// just to be complete.
385 fn miri_start_panic(payload: *mut u8) -> !;
389 ## Contributing and getting help
391 If you want to contribute to Miri, great! Please check out our
392 [contribution guide](CONTRIBUTING.md).
394 For help with running Miri, you can open an issue here on
395 GitHub or use the [Miri stream on the Rust Zulip][zulip].
397 [zulip]: https://rust-lang.zulipchat.com/#narrow/stream/269128-miri
401 This project began as part of an undergraduate research course in 2015 by
402 @solson at the [University of Saskatchewan][usask]. There are [slides] and a
403 [report] available from that project. In 2016, @oli-obk joined to prepare miri
404 for eventually being used as const evaluator in the Rust compiler itself
405 (basically, for `const` and `static` stuff), replacing the old evaluator that
406 worked directly on the AST. In 2017, @RalfJung did an internship with Mozilla
407 and began developing miri towards a tool for detecting undefined behavior, and
408 also using miri as a way to explore the consequences of various possible
409 definitions for undefined behavior in Rust. @oli-obk's move of the miri engine
410 into the compiler finally came to completion in early 2018. Meanwhile, later
411 that year, @RalfJung did a second internship, developing miri further with
412 support for checking basic type invariants and verifying that references are
413 used according to their aliasing restrictions.
415 [usask]: https://www.usask.ca/
416 [slides]: https://solson.me/miri-slides.pdf
417 [report]: https://solson.me/miri-report.pdf
419 ## Bugs found by Miri
421 Miri has already found a number of bugs in the Rust standard library and beyond, which we collect here.
425 * [`Debug for vec_deque::Iter` accessing uninitialized memory](https://github.com/rust-lang/rust/issues/53566)
426 * [`Vec::into_iter` doing an unaligned ZST read](https://github.com/rust-lang/rust/pull/53804)
427 * [`From<&[T]> for Rc` creating a not sufficiently aligned reference](https://github.com/rust-lang/rust/issues/54908)
428 * [`BTreeMap` creating a shared reference pointing to a too small allocation](https://github.com/rust-lang/rust/issues/54957)
429 * [`Vec::append` creating a dangling reference](https://github.com/rust-lang/rust/pull/61082)
430 * [Futures turning a shared reference into a mutable one](https://github.com/rust-lang/rust/pull/56319)
431 * [`str` turning a shared reference into a mutable one](https://github.com/rust-lang/rust/pull/58200)
432 * [`rand` performing unaligned reads](https://github.com/rust-random/rand/issues/779)
433 * [The Unix allocator calling `posix_memalign` in an invalid way](https://github.com/rust-lang/rust/issues/62251)
434 * [`getrandom` calling the `getrandom` syscall in an invalid way](https://github.com/rust-random/getrandom/pull/73)
435 * [`Vec`](https://github.com/rust-lang/rust/issues/69770) and [`BTreeMap`](https://github.com/rust-lang/rust/issues/69769) leaking memory under some (panicky) conditions
436 * [`beef` leaking memory](https://github.com/maciejhirsz/beef/issues/12)
437 * [`EbrCell` using uninitialized memory incorrectly](https://github.com/Firstyear/concread/commit/b15be53b6ec076acb295a5c0483cdb4bf9be838f#diff-6282b2fc8e98bd089a1f0c86f648157cR229)
438 * [TiKV performing an unaligned pointer access](https://github.com/tikv/tikv/issues/7613)
439 * [`servo_arc` creating a dangling shared reference](https://github.com/servo/servo/issues/26357)
440 * [TiKV constructing out-of-bounds pointers (and overlapping mutable references)](https://github.com/tikv/tikv/pull/7751)
441 * [`encoding_rs` doing out-of-bounds pointer arithmetic](https://github.com/hsivonen/encoding_rs/pull/53)
442 * [TiKV using `Vec::from_raw_parts` incorrectly](https://github.com/tikv/agatedb/pull/24)
443 * Incorrect doctests for [`AtomicPtr`](https://github.com/rust-lang/rust/pull/84052) and [`Box::from_raw_in`](https://github.com/rust-lang/rust/pull/84053)
445 Violations of [Stacked Borrows] found that are likely bugs (but Stacked Borrows is currently just an experiment):
447 * [`VecDeque::drain` creating overlapping mutable references](https://github.com/rust-lang/rust/pull/56161)
448 * Various `BTreeMap` problems
449 * [`BTreeMap` iterators creating mutable references that overlap with shared references](https://github.com/rust-lang/rust/pull/58431)
450 * [`BTreeMap::iter_mut` creating overlapping mutable references](https://github.com/rust-lang/rust/issues/73915)
451 * [`BTreeMap` node insertion using raw pointers outside their valid memory area](https://github.com/rust-lang/rust/issues/78477)
452 * [`LinkedList` cursor insertion creating overlapping mutable references](https://github.com/rust-lang/rust/pull/60072)
453 * [`Vec::push` invalidating existing references into the vector](https://github.com/rust-lang/rust/issues/60847)
454 * [`align_to_mut` violating uniqueness of mutable references](https://github.com/rust-lang/rust/issues/68549)
455 * [`sized-chunks` creating aliasing mutable references](https://github.com/bodil/sized-chunks/issues/8)
456 * [`String::push_str` invalidating existing references into the string](https://github.com/rust-lang/rust/issues/70301)
457 * [`ryu` using raw pointers outside their valid memory area](https://github.com/dtolnay/ryu/issues/24)
458 * [ink! creating overlapping mutable references](https://github.com/rust-lang/miri/issues/1364)
459 * [TiKV creating overlapping mutable reference and raw pointer](https://github.com/tikv/tikv/pull/7709)
460 * [Windows `Env` iterator using a raw pointer outside its valid memory area](https://github.com/rust-lang/rust/pull/70479)
461 * [`VecDeque::iter_mut` creating overlapping mutable references](https://github.com/rust-lang/rust/issues/74029)
462 * [Various standard library aliasing issues involving raw pointers](https://github.com/rust-lang/rust/pull/78602)
463 * [`<[T]>::copy_within` using a loan after invalidating it](https://github.com/rust-lang/rust/pull/85610)
467 Licensed under either of
469 * Apache License, Version 2.0 ([LICENSE-APACHE](LICENSE-APACHE) or
470 http://www.apache.org/licenses/LICENSE-2.0)
471 * MIT license ([LICENSE-MIT](LICENSE-MIT) or
472 http://opensource.org/licenses/MIT)
478 Unless you explicitly state otherwise, any contribution intentionally submitted
479 for inclusion in the work by you shall be dual licensed as above, without any
480 additional terms or conditions.