GHSA-hc7m-r6v8-hg9q CVE-2025-64345 RUSTSEC-2025-0118

Wasmtime provides unsound API access to a WebAssembly shared linear memory

Low crates.io

Severity: 1.8 / 10; Low · CVSS 3.1
Exploit chance (EPSS): 0.12%; Chance of exploitation in the next 30 days. Higher than 1.61% of scored vulnerabilities.
Known exploited: No; Not in CISA's catalog
Weakness: CWE-362 Concurrent Execution using Shared Resource with Improper Synchronization ('Race Condition')
Fix: 38.0.4; Upgrade wasmtime to this version or later

Description

Impact

Wasmtime's Rust embedder API contains an unsound interaction where a WebAssembly shared linear memory could be viewed as a type which provides safe access to the host (Rust) to the contents of the linear memory. This is not sound for shared linear memories, which could be modified in parallel, and this could lead to a data race in the host.

Wasmtime has a wasmtime::Memory type which represents linear memories in a WebAssembly module. Wasmtime also has wasmtime::SharedMemory, however, which represents shared linear memories introduced in the WebAssembly threads proposal. The API of SharedMemory does not provide accessors which return &[u8] in Rust, for example, as that's not a sound type signature when other threads could be modifying memory. The wasmtime::Memory type, however, does provide this API as it's intended to be used with non-shared memories where static knowledge is available that no concurrent or parallel reads or writes are happening. This means that it's not sound to represent a shared linear memory with wasmtime::Memory and it must instead be represented with wasmtime::SharedMemory.

There were two different, erroneous, methods of creating a wasmtime::Memory which represents a shared memory however:

The wasmtime::Memory::new constructor takes a MemoryType which could be shared. This function did not properly reject shared memory types and require usage of SharedMemory::new instead.
Capturing a core dump with WebAssembly would expose all linear memories a wasmtime::Memory. This means that a core dump would perform an unsynchronized read of shared linear memory, possibly leading to data races.

This is a bug in Wasmtime's safe Rust API. It should not be possible to cause unsoundness with Wasmtime's embedding API if unsafe is not used. Embeddings which do not use the wasm threads proposal nor created shared memories nor actually share shared memories across threads are unaffected. Only if shared memories are created across threads might an embedding be affected.

Patches

Patch releases have been issued for all supported versions of Wasmtime, notably: 24.0.5, 36.0.3, 37.0.3, and 38.0.4. These releases reject creation of shared memories via Memory::new and shared memories are now excluded from core dumps.

Workarounds

Embeddings affected by this issue should use SharedMemory::new instead of Memory::new to create shared memories. Affected embeddings should also disable core dumps if they are unable to upgrade. Note that core dumps are disabled by default but the wasm threads proposal (and shared memory) is enabled by default. It's recommended to upgrade to a patched version of Wasmtime, however.

Affected versions

crates.io

Package	Vulnerable	Fixed in
wasmtime	`>= 38.0.0, < 38.0.4`	`38.0.4`
wasmtime	`>= 37.0.0, < 37.0.3`	`37.0.3`
wasmtime	`>= 26.0.0, < 36.0.3`	`36.0.3`
wasmtime	`< 24.0.5`	`24.0.5`
wasmtime	`>= 0.0.0-0, < 24.0.5` `>= 25.0.0, < 36.0.3` `>= 37.0.0, < 37.0.3` `>= 38.0.0, < 38.0.4`	`24.0.5` `36.0.3` `37.0.3` `38.0.4`

How to fix

Move to the fixed version or later with your package manager, update the lockfile, and run your tests. If the package only arrives through another dependency, upgrade that dependency first. If none of its releases pulls a fixed version yet, force the version with your package manager's override.

wasmtime crates.io → `38.0.4`

cargo

cargo update -p wasmtime --precise 38.0.4

Severity breakdown

CVSS 3.1 1.8

CVSS:3.1/AV:L/AC:H/PR:H/UI:R/S:U/C:N/I:L/A:N

Attack vector: Local
Attack complexity: High
Privileges required: High
User interaction: Required
Scope: Unchanged
Confidentiality: None
Integrity: Low
Availability: None

Weakness

CWE-362: Concurrent Execution using Shared Resource with Improper Synchronization ('Race Condition')

The product contains a concurrent code sequence that requires temporary, exclusive access to a shared resource, but a timing window exists in which the shared resource can be modified by another code sequence operating concurrently.

How to prevent it

Architecture and Design: In languages that support it, use synchronization primitives. Only wrap these around critical code to minimize the impact on performance.
Architecture and Design: Use thread-safe capabilities such as the data access abstraction in Spring.
Architecture and Design: Minimize the usage of shared resources in order to remove as much complexity as possible from the control flow and to reduce the likelihood of unexpected conditions occurring. Additionally, this will minimize the amount of synchronization necessary and may even help to reduce the likelihood of a denial of service where an attacker may be able to repeatedly trigger a critical section (CWE-400).
Implementation: When using multithreading and operating on shared variables, only use thread-safe functions.
Implementation: Use atomic operations on shared variables. Be wary of innocent-looking constructs such as "x++". This may appear atomic at the code layer, but it is actually non-atomic at the instruction layer, since it involves a read, followed by a computation, followed by a write.
Implementation: Use a mutex if available, but be sure to avoid related weaknesses such as CWE-412.
Implementation: Avoid double-checked locking (CWE-609) and other implementation errors that arise when trying to avoid the overhead of synchronization.
Implementation: Disable interrupts or signals over critical parts of the code, but also make sure that the code does not go into a large or infinite loop.
Implementation: Use the volatile type modifier for critical variables to avoid unexpected compiler optimization or reordering. This does not necessarily solve the synchronization problem, but it can help.
Architecture and Design, Operation: Run your code using the lowest privileges that are required to accomplish the necessary tasks [REF-76]. If possible, create isolated accounts with limited privileges that are only used for a single task. That way, a successful attack will not immediately give the attacker access to the rest of the software or its environment. For example, database applications rarely need to run as the database administrator, especially in day-to-day operations.

References

Advisories

nvd.nist.gov/vuln/detail/CVE-2025-64345

Packages

Timeline

Nov 11, 2025 Published
Oct 18, 2026 Last updated by the source
Oct 31, 2026 Latest EPSS score

Advisory	Severity	Why it's related
RUSTSEC-2026-0424 Unchecked memory usage in transferring borrowed resources between components	Not rated yet	Also affects wasmtime
RUSTSEC-2026-0423 Cross-engine type confusion with host-created struct-types-with-supertypes	Not rated yet	Also affects wasmtime
RUSTSEC-2026-0422 Typed functions with `i31` or `nullref` arguments may panic due to insufficient type-check	Low	Also affects wasmtime
RUSTSEC-2026-0379 Wasmtime: Cranelift backend on aarch64 can jump to incorrect target on large Wasm `br_table` instruction	Critical	Also affects wasmtime
GHSA-9pvw-hwfg-92h5 Wasmtime: Panic lowering a host-created component model stream into a guest that expects a different payload type	Medium	Also affects wasmtime
GHSA-897f-g5gm-fmxm Wasmtime: Bit-packing in Cranelift compiler causes silent overflow and miscompilation	Critical	Also affects wasmtime
RUSTSEC-2026-0364 Exception tags can cause host data structures to be unaligned	Not rated yet	Also affects wasmtime
RUSTSEC-2026-0363 Component-model handle tables are capped only per instance and charged to no limit	Medium	Also affects wasmtime

Is this in your code?

MergeHarbor checks every repository's dependencies against this database, ranks what matters by exploitability and reachability, routes it to the owning team and opens the fix pull request.

Check your repos with MergeHarbor