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
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.
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