Impact

There is a bug in Wasmtime's implementation of it's pooling instance allocator where when a linear memory is reused for another instance the initial heap snapshot of the prior instance can be visible, erroneously to the next instance. The pooling instance allocator in Wasmtime works by preallocating virtual memory for a fixed number of instances to reside in and then new instantiations pick a slot to use. Most conventional modules additionally have an initial copy-on-write "heap image" which is mapped in Wasmtime into the linear memory slot. When a heap slot is deallocated Wasmtime resets all of its contents back to the initial state but it does not unmap the image in case the next instance is an instantiation of the same module.

The bug in Wasmtime occurs when a slot in the pooling allocator previously was used for a module with a heap image, meaning that its current state of memory contains the initial heap contents of that module. If the next instantiation within that slot does not itself contain a heap image then Wasmtime would leave the old heap image in place erroneously and continue with instantiation. This means that instantiations of modules without a heap image can see the initial heap image of the prior instantiation within that slot.

Heap images in Wasmtime are created by precomputing WebAssembly data segments into one large mapping to be placed into linear memory at a particular offset. Most modules produced by toolchains today will have a heap image and an initialization snapshot. Creating a module without a heap image would require a hand-crafted .wat file or a specially crafted source program. This consequence means that this bug is highly unlikely to be accidentally triggered and would otherwise require an intentional trigger with a hand-crafted module.

One important part of this vulnerability is Wasmtime is highly likely to segfault when the slot is reused again with a module that itself has an initialization image. For example if module A has a heap initialization image and module B does not have a heap initialization image, then the following sequence of events could happen if they all are instantiated into the same instance slot:

Module A is instantiated, and then deallocated. This leaves A's heap image in place, reset to its initial contents. Module B is instantiated and erroneously can see the initial heap contents of A. Module B is then deallocated and the entire heap is unmapped and reset back to zero. Module A is instantiated again, but the state tracking the slot did not account for module B so it thinks the module image is still mapped and proceeds with instantiation. Any action on A's part to access linear memory will then trap and if the host accesses A's memory it will segfault because the data that's supposed to be mapped is all unmapped.

Adding this all together this means that in practice modules must be deliberately crafted to not have an initial heap image to view the contents of a prior image. If this module is instantiated though then when the slot is reused the next, likely image-using, module will believe its memory is mapped when it isn't, causing the host to segfault on unmapped memory it believed was mapped.

Patches

This bug has been patched and users should upgrade to Wasmtime 2.0.2.

Workarounds

Trigging this bug requires the pooling allocator to be configured and for copy-on-write heap images to also be enabled. Pooling allocation is not enabled by default but copy-on-write heap images are. Mitigations for this bug include:

Disabling the pooling allocator - note that pooling allocation is not enabled by default in Wasmtime Disabling the memory-init-cow feature or with Config::memoryinitcow

References

Config::allocationstrategy - configuration required to enable the pooling allocator. Config::memoryinitcow - configuration required to enable or disable copy-on-write (this is enabled by default). Mailing list announcement Patch for release-2.0.0 branch Patch for main

For more information

If you have any questions or comments about this advisory:

Reach out to us on the Bytecode Alliance Zulip chat Open an issue in the bytecodealliance/wasmtime repository

Impact

Wasmtime versions from 10.0.0 to 12.0.1 contain a miscompilation of the WebAssembly i64x2.shrs instruction on x8664 platforms when the shift amount is a constant value that is larger than 32. Only x8664 is affected so all other targets are not affected by this. The miscompilation results in the instruction producing an incorrect result, namely the low 32-bits of the second lane of the vector are derived from the low 32-bits of the second lane of the input vector instead of the high 32-bits. The primary impact of this issue is that any WebAssembly program using the i64x2.shrs with a constant shift amount larger than 32 may produce an incorrect result.

This issue is not an escape from the WebAssembly sandbox. Execution of WebAssembly guest programs will still behave correctly with respect to memory sandboxing and isolation from the host. Wasmtime considers non-spec-compliant behavior as a security issue nonetheless.

This issue was discovered through fuzzing of Wasmtime's code generator Cranelift.

Patches

Wasmtime versions 10.0.2, 11.0.2, and 12.0.2 are all patched to no longer have this miscompilation.

Workarounds

This issue only affects x8664 hosts and the only workaround is to either scan for this pattern in wasm modules which is nontrivial or to disable the SIMD proposal for WebAssembly. Users prior to 10.0.0 are unaffected by this vulnerability.

References

PR which introduced this bug to Wasmtime Mailing list announcement

Impact

Wasmtime's implementation of managing per-instance state, such as tables and memories, contains LLVM-level undefined behavior. This undefined behavior was found to cause runtime-level issues when compiled with LLVM 16 which causes some writes, which are critical for correctness, to be optimized away. Vulnerable versions of Wasmtime compiled with Rust 1.70, which is currently in beta, or later are known to have incorrectly compiled functions. Versions of Wasmtime compiled with the current Rust stable release, 1.69, and prior are not known at this time to have any issues, but can theoretically exhibit potential issues.

The underlying problem is that Wasmtime's runtime state for an instance involves a Rust-defined structure called Instance which has a trailing VMContext structure after it. This VMContext structure has a runtime-defined layout that is unique per-module. This representation cannot be expressed with safe code in Rust so unsafe code is required to maintain this state. The code doing this, however, has methods which take &self as an argument but modify data in the VMContext part of the allocation. This means that pointers derived from &self are mutated. This is typically not allowed, except in the presence of UnsafeCell, in Rust. When compiled to LLVM these functions have noalias readonly parameters which means it's UB to write through the pointers.

Wasmtime's internal representation and management of VMContext has been updated to use &mut self methods where appropriate. Additionally verification tools for unsafe code in Rust, such as cargo miri, are planned to be executed on the main branch soon to fix any Rust-level issues that may be exploited in future compiler versions.

Precomplied binaries available for Wasmtime from GitHub releases have been compiled with at most LLVM 15 so are not known to be vulnerable. As mentioned above, however, it's still recommended to update.

Patches

Wasmtime version 6.0.2, 7.0.1, and 8.0.1 have been issued which contain the patch necessary to work correctly on LLVM 16 and have no known UB on LLVM 15 and earlier.

Workarounds

If Wasmtime is compiled with Rust 1.69 and prior, which use LLVM 15, then there are no known issues. There is a theoretical possibility for UB to exploited, however, so it's recommended that users upgrade to a patched version of Wasmtime. Users using beta Rust (1.70 at this time) or nightly Rust (1.71 at this time) must update to a patched version to work correctly.

References

GitHub Advisory Mailing list announcement

For more information

If you have any questions or comments about this advisory:

Reach out to us on the Bytecode Alliance Zulip chat Open an issue in the bytecodealliance/wasmtime repository

Impact

Wasmtime's code generation backend, Cranelift, has a bug on x8664 platforms for the WebAssembly i8x16.select instruction which will produce the wrong results when the same operand is provided to the instruction and some of the selected indices are greater than 16. There is an off-by-one error in the calculation of the mask to the pshufb instruction which causes incorrect results to be returned if lanes are selected from the second vector.

The impact of this miscompilation is that the WebAssembly instruction can produce incorrect results for the i8x16.select instruction. This should have no effect on embedders and does not represent a sandbox escape, for example. Guest programs, however, may behave unexpectedly due to the incorrect result of this instruction. In extreme cases if a guest program is handling untrusted input then the guest program may deviate from its intended execution, for example calling an imported host function with different arguments than intended. This still does not impact embedders, however, because there is no form of privilege escalation with the guest.

At this time it's expected that this codegen pattern doesn't show up in the wild that often. LLVM-generated modules, for example, do not appear to conventionally or idiomatically generate code which would hit this bug. It is possible, however, to still write code which triggers this, so it's recommended for embedders to analyze existing modules to see if any are affected.

Patches

This codegen bug has been fixed in Wasmtime 6.0.1, 5.0.1, and 4.0.1. Users are recommended to upgrade to these updated versions.

Workarounds

If upgrading is not an option for you at this time, you can avoid this miscompilation by disabling the Wasm simd proposal

rust config.wasmsimd(false);

Additionally the bug is only present on x8664 hosts. Other platforms such as AArch64 and s390x are not affected.

References

The WebAssembly simd proposal Mailing list announcement GitHub advisory Commit to fix this issue on Wasmtime's main branch

For more information

If you have any questions or comments about this advisory:

Reach out to us on the Bytecode Alliance Zulip chat Open an issue in the bytecodealliance/wasmtime repository

Impact

Wasmtime's code generator, Cranelift, has a bug on x8664 targets where address-mode computation mistakenly would calculate a 35-bit effective address instead of WebAssembly's defined 33-bit effective address. This bug means that, with default codegen settings, a wasm-controlled load/store operation could read/write addresses up to 35 bits away from the base of linear memory. Wasmtime's default sandbox settings provide up to 6G of protection from the base of linear memory to guarantee that any memory access in that range will be semantically correct. Due to this bug, however, addresses up to 0xffffffff 8 + 0x7ffffffc = 36507222004 = ~34G bytes away from the base of linear memory are possible from guest code. This means that the virtual memory 6G away from the base of linear memory up to ~34G away can be read/written by a malicious module.

This out of bounds read/write is not semantically correct and poses a threat as an arbitrary read/write within ~34G of linear memory away from the base of a wasm module's linear memory. A guest module can, without the knowledge of the embedder, read/write memory in this region. The memory may belong to other WebAssembly instances when using the pooling allocator, for example. The memory may also belong to the embedder, depending on address layout.

Embedders do not have a necessarily reliable means of detecting when this happens. Wasm loads/stores are allowed to cause machine segfaults meaning that an invalid read/write would be translated to a nominal WebAssembly trap. This means that a malicious module in the worst case silently reads/writes memory outside its bounds and in the "best" case looks like a normal "something trapped here" during its execution. This makes it difficult to retroactively determine whether this bug has been exploited on hosts. Affected embedders are recommended to analyze preexisting wasm modules to see if they're affected by the incorrect codegen rules and possibly correlate that with an anomalous number of traps during historical execution to locate possibly suspicious modules.

The specific bug in Cranelift's x8664 backend is that a WebAssembly address which is left-shifted by a constant amount from 1 to 3 will get folded into x8664's addressing modes which perform shifts. For example (i32.load (i32.shl (local.get 0) (i32.const 3))) loads from the WebAssembly address $local0 << 3. When translated to Cranelift the $local0 << 3 computation, a 32-bit value, is zero-extended to a 64-bit value and then added to the base address of linear memory. Cranelift would generate an instruction of the form movl (%base, %local0, 8), %dst which calculates %base + %local0 << 3. The bug here, however, is that the address computation happens with 64-bit values, where the $local0 << 3 computation was supposed to be truncated to a 32-bit value. This means that %local0, which can use up to 32-bits for an address, gets 3 extra bits of address space to be accessible via this movl instruction.

The fix in Cranelift is to remove the erroneous lowering rules in the backend which handle these zero-extended expressions. The above example is then translated to movl %local0, %temp; shl $3, %temp; movl (%base, %temp), %dst which correctly truncates the intermediate computation of %local0 << 3 to 32-bits inside the %temp register which is then added to the %base value.

Patches

Wasmtime version 4.0.1, 5.0.1, and 6.0.1 have been released and have all been patched to no longer contain the erroneous lowering rules.

Workarounds

While updating Wasmtime is recommended, there are a number of possible workarounds that embedders can employ to mitigate this issue if updating is not possible. Note that none of these workarounds are on-by-default and require explicit configuration:

The Config::staticmemorymaximumsize(0) option can be used to force all accesses to linear memory to be explicitly bounds-checked. This will perform a bounds check separately from the address-mode computation which correctly calculates the effective address of a load/store. Note that this can have a large impact on the execution performance of WebAssembly modules. The Config::staticmemoryguardsize(1 << 36) option can be used to greatly increase the guard pages placed after linear memory. This will guarantee that memory accesses up-to-34G away are guaranteed to be semantically correct by reserving unmapped memory for the instance. Note that this reserves a very large amount of virtual memory per-instances and can greatly reduce the maximum number of concurrent instances being run. If using a non-x8664 host is possible, then that will also work around this bug. This bug does not affect Wasmtime's or Cranelift's AArch64 backend, for example.

References

Config::staticmemorymaximumsize Config::staticmemoryguardsize Mailing list announcement GitHub advisory Commit to fix this issue on Wasmtime's main branch

For more information

If you have any questions or comments about this advisory:

Reach out to us on the Bytecode Alliance Zulip chat Open an issue in the bytecodealliance/wasmtime repository

Impact

There is a bug in Wasmtime's C API implementation where the definition of the wasmtimetrapcode does not match its declared signature in the wasmtime/trap.h header file. This discrepancy causes the function implementation to perform a 4-byte write into a 1-byte buffer provided by the caller. This can lead to three zero bytes being written beyond the 1-byte location provided by the caller.

Patches

This bug has been patched and users should upgrade to Wasmtime 2.0.2.

Workarounds

This can be worked around by providing a 4-byte buffer casted to a 1-byte buffer when calling wasmtimetrapcode. Users of the wasmtime crate are not affected by this issue, only users of the C API function wasmtimetrapcode are affected.

References

Definition of wasmtimetrapcode Mailing list announcement Patch to fix for main branch

For more information

If you have any questions or comments about this advisory:

Reach out to us on the Bytecode Alliance Zulip chat Open an issue in the bytecodealliance/wasmtime repository

Impact

There is a bug in Wasmtime's implementation of its pooling instance allocator when the allocator is configured to give WebAssembly instances a maximum of zero pages of memory. In this configuration the virtual memory mapping for WebAssembly memories did not meet the compiler-required configuration requirements for safely executing WebAssembly modules. Wasmtime's default settings require virtual memory page faults to indicate that wasm reads/writes are out-of-bounds, but the pooling allocator's configuration would not create an appropriate virtual memory mapping for this meaning out of bounds reads/writes can successfully read/write memory unrelated to the wasm sandbox within range of the base address of the memory mapping created by the pooling allocator.

This bug can only be triggered by setting InstanceLimits::memorypages to zero. This is expected to be a very rare configuration since this means that wasm modules cannot allocate any pages of linear memory. All wasm modules produced by all current toolchains are highly likely to use linear memory, so it's expected to be unlikely that this configuration is set to zero by any production embedding of Wasmtime, hence the low severity of this bug despite the critical consequences.

Patches

This bug has been patched and users should upgrade to Wasmtime 2.0.2.

Workarounds

One way to mitigate this issue is to disable usage of the pooling allocator. Note that the pooling allocator is not enabled by default.

This bug can also only be worked around by increasing the memorypages allotment when configuring the pooling allocator to a value greater than zero. If an embedding wishes to still prevent memory from actually being used then the Store::limiter method can be used to dynamically disallow growth of memory beyond 0 bytes large. Note that the default memorypages value is greater than zero.

This bug is not applicable with the default settings of the wasmtime crate.

References

Config::allocationstrategy - configuration required to enable the pooling allocator. InstanceLimits::memorypages - configuration field that, when zero, exhibits this bug. Store::limiter - means of limiting memory without using memorypages Mailing list announcement Patch for the release-2.0.0 branch

For more information

If you have any questions or comments about this advisory:

Reach out to us on the Bytecode Alliance Zulip chat Open an issue in the bytecodealliance/wasmtime repository

Impact

There was a bug in Wasmtime's code generator, Cranelift, for AArch64 targets where constant divisors could result in incorrect division results at runtime. The translation rules for constants did not take into account whether sign- or zero-extension should happen, which resulted in an incorrect value being placed into a register when a division was encountered. For example, a constant 32-bit unsigned divisor of 0xfffffffe would be incorrectly sign-extended to 64-bits to 0xfffffffffffffffe. Any kind of division of operands smaller than 64 bits is implemented with a 64-bit division instruction which would then result in an incorrect result because the divisor was larger than expected.

The impact of this bug is that programs executing within the WebAssembly sandbox would not behave according to the WebAssembly specification. This means that it is hypothetically possible for execution within the sandbox to go awry and WebAssembly programs could produce unexpected results. This should not impact hosts executing WebAssembly, but does affect the correctness of guest programs.

This bug was found with differential fuzzing of Wasmtime against other engines on the AArch64 platform. Fuzzing on AArch64 is not regularly performed at this time and the Wasmtime team is investigating how best to continuously fuzz AArch64 in the same manner as x8664.

Patches

This bug has been patched and users should upgrade to Wasmtime version 0.38.2.

Workarounds

If upgrading is not an option at this time, direct users of Cranelift that control the exact Cranelift instructions being compiled can avoid the vulnerability by explicitly extending constant divisors to 64 bits using either the sextend.i64 or the uextend.i64 operation.

Note, though, that this issue only affects the AArch64 targets. Other platforms are not affected.

For more information

If you have any questions or comments about this advisory:

Reach out to us on the Bytecode Alliance Zulip chat Open an issue in the bytecodealliance/wasmtime repository

Impact

Wasmtime's implementation of the SIMD proposal for WebAssembly on x8664 contained two distinct bugs in the instruction lowerings implemented in Cranelift. The aarch64 implementation of the simd proposal is not affected. The bugs were presented in the i8x16.swizzle and select WebAssembly instructions. The select instruction is only affected when the inputs are of v128 type. The correspondingly affected Cranelift instructions were swizzle and select.

The swizzle instruction lowering in Cranelift erroneously overwrote the mask input register which could corrupt a constant value, for example. This means that future uses of the same constant may see a different value than the constant itself.

The select instruction lowering in Cranelift wasn't correctly implemented for vector types that are 128-bits wide. When the condition was 0 the wrong instruction was used to move the correct input to the output of the instruction meaning that only the low 32 bits were moved and the upper 96 bits of the result were left as whatever the register previously contained (instead of the input being moved from). The select instruction worked correctly if the condition was nonzero, however.

This bug in Wasmtime's implementation of these instructions on x8664 represents an incorrect implementation of the specified semantics of these instructions according to the WebAssembly specification. The impact of this is benign for hosts running WebAssembly but represents possible vulnerabilities within the execution of a guest program. For example a WebAssembly program could take unintended branches or materialize incorrect values internally which runs the risk of exposing the program itself to other related vulnerabilities which can occur from miscompilations.

Patches

We have released Wasmtime 0.38.1 and cranelift-codegen (and other associated cranelift crates) 0.85.1 which contain the corrected implementations of these two instructions in Cranelift.

Workarounds

If upgrading is not an option for you at this time, you can avoid the vulnerability by disabling the Wasm simd proposal

rust config.wasmsimd(false);

Additionally the bug is only present on x8664 hosts. Other aarch64 hosts are not affected. Note that s390x hosts don't yet implement the simd proposal and are not affected.

References

The WebAssembly simd proposal Original test case showing the erroneous behavior Fix for the swizzle instruction Fix for the select instruction

For more information

If you have any questions or comments about this advisory:

Reach out to us on the Bytecode Alliance Zulip chat Open an issue in the bytecodealliance/wasmtime repository

There is a bug in Wasmtime's code generator, Cranelift, where functions using reference types may be incorrectly missing metadata required for runtime garbage collection (GC). This means that if a GC happens at runtime then the collector will mistakenly think some Wasm stack frames do not have live references to garbage collected values and therefore reclaim and deallocate them. The function can then subsequently continue to use the values, leading later to use-after-free bugs. This bug was introduced in Cranelift's migration to the regalloc2 register allocator in the Wasmtime 0.37.0 release on 2022-05-20. This bug has been patched and users should upgrade to Wasmtime version 0.38.2.

Mitigations for this issue can be achieved by doing one of:

Disabling the reference types proposal by passing false to wasmtime::Config::wasmreferencetypes. Downgrading to Wasmtime 0.36.0 or prior.

There is a use after free vulnerability in Wasmtime when both running Wasm that uses externrefs and enabling epoch interruption in Wasmtime. If you are not explicitly enabling epoch interruption (it is disabled by default) then you are not affected. If you are explicitly disabling the Wasm reference types proposal (it is enabled by default) then you are also not affected.

The use after free is caused by Cranelift failing to emit stack maps when there are safepoints inside cold blocks. Cold blocks occur when epoch interruption is enabled. Cold blocks are emitted at the end of compiled functions, and change the order blocks are emitted versus defined. This reordering accidentally caused Cranelift to skip emitting some stack maps because it expected to emit the stack maps in block definition order, rather than block emission order. When Wasmtime would eventually collect garbage, it would fail to find live references on the stack because of the missing stack maps, think that they were unreferenced garbage, and therefore reclaim them. Then after the collection ended, the Wasm code could use the reclaimed-too-early references, which is a use after free.

This bug was discovered while extending our fuzz targets for externrefs and GC in Wasmtime. The updated fuzz target thoroughly exercises these code paths and feature combinations now. We have also added a regression test for this bug. Released versions 0.34.2 and 0.35.2, which fix the vulnerability. We recommend all Wasmtime users upgrade to these patched versions. If upgrading is not an option for you at this time, you can avoid the vulnerability by either disabling the Wasm reference types proposal or by disabling epoch interruption if you were previously enabling it.

Impact

There exists a bug in the pooling instance allocator in Wasmtime's runtime where a failure to instantiate an instance for a module that defines an externref global will result in an invalid drop of a VMExternRef via an uninitialized pointer.

As instance slots may be reused between consecutive instantiations, the value of the uninitialized pointer may be from a previous instantiation and therefore under the control of an attacker via a module's initial values for its globals. If the attacker can somehow determine an address under their control inside the mapped memory representing the instance pool, it is possible to trick the runtime to call dropinplace on a trait object under the attacker's control and therefore cause remote code execution.

Exploiting the bug to cause remote code execution would be very difficult as attackers cannot determine the addresses of globals from code executing within the WebAssembly VM and the memory space for the instance pool cannot be statically determined. Operating system mitigations, such as address space layout randomization, would additionally increase the difficulty for attackers to determine useful executable code to target with an exploit. It is also very unlikely that attackers will be able to directly influence the conditions that trigger the bug as described below.

When the conditions to trigger the bug are met, however, it is much easier to exploit this bug to cause a denial of service by crashing the host with an invalid memory read.

The following engine configuration (via Config) is required to be impacted by this bug:

support for the reference types proposal must be enabled (this is the default for Config). a pooling allocation strategy must be configured via Config::allocationstrategy, which is not the default allocation strategy.

A module must be instantiated with all the following characteristics:

The module defines at least one table or memory. The module defines at least one externref global.

During instantiation, one of the following must occur to cause the instantiation to fail:

a call to mprotect or VirtualAlloc fails (e.g. out-of-memory conditions). a resource limiter was configured in the associated Store (via Store::limiter or Store::limiterasync) and the limiter returns false from the initial call to memorygrowing or tablegrowing. Stores do not have a resource limiter set by default.

This results in a partially-initialized instance being dropped and that attempts to drop the uninitialized VMExternRef representing the defined externref global.

We have reason to believe that the effective impact of this bug is relatively small because the usage of externref is still uncommon and without a resource limiter configured on the Store, which is not the default configuration, it is only possible to trigger the bug from an error returned by mprotect or VirtualAlloc.

Note that on Linux with the uffd feature enabled, it is only possible to trigger the bug from a resource limiter as the call to mprotect is skipped; if no resource limiter is used, then this configuration is not vulnerable.

Patches

The bug has been fixed in 0.34.1 and 0.33.1; users are encouraged to upgrade as soon as possible.

Workarounds

If it is not possible to upgrade to 0.34.1 or 0.33.1 of the wasmtime crate, it is recommend that support for the reference types proposal be disabled by passing false to Config::wasmreferencetypes.

Doing so will prevent modules that use externref from being loaded entirely.

For more information

If you have any questions or comments about this advisory:

Reach out to us on the Bytecode Alliance Zulip chat Open an issue in the bytecodealliance/wasmtime repository