Live › vulnerability › CVE-2026-58494
Go to dashboardWasmtime's wasmtime-wasi lets a WASI guest create a hard link to a read-only file in a preopen that allows writes, then write through that link to overwrite the host file. An attacker who controls a WASI guest can modify host files that were exposed as FilePerms::READ.
Summary written from the sources below. Check the originals before acting.
First seen by InfraStack 2026-10-23 19:11 UTC, 5h before NVD. All times on this page are UTC.
| Product | Ecosystem | Affected | Fixed in |
|---|---|---|---|
| wasmtime-wasi | cargo | < 24.0.11 | 24.0.11 |
| wasmtime-wasi | cargo | >= 25.0.0, < 36.0.12 | 36.0.12 |
| wasmtime-wasi | cargo | >= 37.0.0, < 45.0.3 | 45.0.3 |
| wasmtime-wasi | cargo | >= 46.0.0, < 46.0.1 | 46.0.1 |
24.0.11, 36.0.12, 45.0.3, or 46.0.1 depending on your release lineEach row is the first time that source carried this signal. The analyst’s rows are its own findings.
Wasmtime's WASI filesystem implementation lets a guest with a read-only source file capability overwrite that file by hard-linking it into another preopen that allows file writes. An attacker who can run a WASI guest can bypass WasiCtxBuilder per-preopen FilePerms and modify host files that were exposed as FilePerms::READ.
This attack is possible through the wasip1, wasip2, and wasip3 interfaces to the WASI filesystem.
This attack is also possible through an indirection of renaming a hardlink into another preopen.
WasiCtxBuilder::preopened_dir documents file_perms as the maximum permissions usable for files in that preopen. However, hard-link creation and renaming only checked directory mutation for permission, and not whether the file permissions matched.
This makes an attack possible where:
target exists in a preopen with only FilePerms::READalias, with destination to the target, is created in a preopen with FilePerms::READ | FilePerms::WRITEalias for writing, in which they can overwrite targetThe same attack is possible by way of renaming:
target exists in a preopen with only FilePerms::READalias, with destination to the target, is created in that same preopenalias file is renamed into a preopen with FilePerms::READ | FilePerms::WRITEalias for writing, in which they can overwrite targetSymbolic links are not a vector for this attack because the wasi-filesystem interfaces already enforces that the destination for a symlink is under the same preopen. We do not believe there are any other vectors to this attack because there are no other wasi filesystem interfaces that take multiple borrows of a descriptor resource for e.g. the source and destination in different descriptors.
The definition of Dir::link_at and Dir::rename_at in wasmtime-wasi's src/filesystem.rs now have an additional permissions check that ensures the source and destination values for the directory permissions perms and the file permissions file_perms are identical.
if self.perms != new_dir.perms || self.file_perms != new_dir.file_perms {
return Err(ErrorCode::NotPermitted);
}
Only wasmtime-wasi embeddings that use a combination of DirPerms::MUTATE with FilePerms::READ, as well as using a second preopen with higher permissions (FilePerms::WRITE as well), are affected by this bug, e.g. those that use in the WasiCtxBuilder:
builder.preopened_dir("readwrite", "readwrite", DirPerms::READ | DirPerms::MUTATE, FilePerms::READ | FilePerms::WRITE);
builder.preopened_dir("readonly", "readonly", DirPerms::READ | DirPerms::MUTATE, FilePerms::READ);
In particular, the Wasmtime project's wasmtime-cli's use of wasmtime-wasi is not affected, because it always sets FilePerms::all() for all preopens.
This page as .md · .json · in your agent: get_signal("CVE-2026-58494")
GitHub Advisory Database, CC-BY 4.0 · NVD data is provided by NIST and is public domain. Posts quoted belong to their authors.