1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
use crate::prelude::*;
use core::num::TryFromIntError;
use holochain_serialized_bytes::prelude::*;
use parking_lot::Mutex;
use std::sync::Arc;
use wasmer::Instance;
use wasmer::Memory;
use wasmer::Value;

/// write a slice of bytes to the guest in a safe-ish way
///
/// a naive approach would look like this:
///
/// ```ignore
/// let view: MemoryView<u8> = ctx.memory(0).view();
/// unsafe {
///       std::ptr::copy_nonoverlapping(
///         slice.as_ptr(),
///         view.as_ptr().add(guest_ptr as usize) as *mut u8,
///         slice.len(),
///     );
/// }
/// ```
///
/// the guest memory is part of the host memory, so we get the host's pointer to the start of the
/// guest's memory with view.as_ptr() then we add the guest's pointer to where it wants to see the
/// written bytes then copy the slice directly across.
///
/// the problem with this approach is that the guest_ptr typically needs to be provided by the
/// allocator in the guest wasm in order to be safe for the guest's consumption, but a malicious
/// guest could provide bogus guest_ptr values that point outside the bounds of the guest memory.
/// the naive host would then corrupt its own memory by copying bytes... wherever, basically.
///
/// a better approach is to use wasmer's WasmPtr abstraction, which checks against the memory
/// bounds of the guest based on the input type and can be dereferenced to a [Cell] slice that we
/// can write to more safely.
///
/// @see https://docs.rs/wasmer-runtime-core/0.17.0/src/wasmer_runtime_core/memory/ptr.rs.html#120
///
/// this is still not completely safe in the face of shared memory and threads, etc.
///
/// the guest needs to provide a pointer to a pre-allocated (e.g. by forgetting a Vec<u8>) region
/// of the guest's memory that it is safe for the host to write to.
///
/// it is the host's responsibility to tell the guest the length of the allocation that is needed
/// and the guest's responsibility to correctly reserve an allocation to be written into.
///
/// write_bytes() takes a slice of bytes and writes it to the position at the guest pointer
///
/// as the byte slice cannot be co-ordinated by the compiler (because the host and guest have
/// different compilers and allocators) we prefix the allocation with a WasmSize length value.
///
/// for example, if we wanted to write the slice &[1, 2, 3] then we'd take the length of the slice,
/// 3 as a WasmSize, which is u32, i.e. a 3_u32 and convert it to an array of u8 bytes as
/// [ 3_u8, 0_u8, 0_u8, 0_u8 ] and concatenate it to our original [ 1_u8, 2_u8, 3_u8 ].
/// this gives the full array of bytes to write as:
///
/// ```ignore
/// [ 3_u8, 0_u8, 0_u8, 0_u8, 1_u8, 2_u8, 3_u8 ]
/// ```
///
/// this allows us to read back the byte slice given only a GuestPtr because the read operation
/// can do the inverse in a single step by reading the length inline
///
/// it also requires the host and the guest to both adopt this convention and read/write the
/// additional 4 byte prefix in order to read/write the real payload correctly
///
/// @see read_bytes()
pub fn write_bytes(
    memory: &Memory,
    guest_ptr: GuestPtr,
    slice: &[u8],
) -> Result<(), wasmer_engine::RuntimeError> {
    #[cfg(feature = "debug_memory")]
    tracing::debug!(
        "writing bytes from host to guest at: {} {}",
        guest_ptr as u32,
        slice.len() as u32
    );

    let ptr: WasmPtr<u8, Array> = WasmPtr::new(guest_ptr as _);
    // write the length prefix immediately before the slice at the guest pointer position
    for (byte, cell) in slice.iter().zip(
        ptr.deref(memory, 0 as GuestPtr, slice.len() as Len)
            .ok_or(wasm_error!(WasmErrorInner::Memory))?
            .iter(),
    ) {
        cell.set(*byte)
    }

    Ok(())
}

/// read a slice of bytes from the guest in a safe-ish way
///
/// a naive approach would look like this:
///
/// ```ignore
/// let view: MemoryView<u8> = ctx.memory(0).view();
/// unsafe {
///     std::slice::from_raw_parts::<u8>(
///         view.as_ptr().add(guest_ptr as usize) as _,
///         len as _
///     )
/// }.to_vec()
/// ```
///
/// this is similar to the naive write_slice approach and has similar problems
/// @see write_slice()
///
/// a better approach is to use an immutable deref from a WasmPtr, which checks against memory
/// bounds for the guest, and map over the whole thing to a Vec<u8>
///
/// this does the inverse of write_bytes to read a vector of arbitrary length given only a single
/// GuestPtr value
///
/// it reads the first 4 u8 bytes at the GuestPtr position and interprets them as a single u32
/// value representing a Len which is the length of the return Vec<u8> to read at position
/// GuestPtr + 4
///
/// using the example in write_bytes(), if we had written
///
/// ```ignore
/// [ 3_u8, 0_u8, 0_u8, 0_u8, 1_u8, 2_u8, 3_u8 ]
/// ```
///
/// and this returned a GuestPtr to `5678` then we would read it back by taking the first 4 bytes
/// at `5678` which would be `[ 3_u8, 0_u8, 0_u8, 0_u8 ]` which we interpret as the length `3_u32`.
///
/// we then read the length 3 bytes from position `5682` (ptr + 4) to get our originally written
/// bytes of `[ 1_u8, 2_u8, 3_u8 ]`.
pub fn read_bytes(
    memory: &Memory,
    guest_ptr: GuestPtr,
    len: Len,
) -> Result<Vec<u8>, wasmer_engine::RuntimeError> {
    #[cfg(feature = "debug_memory")]
    tracing::debug!(
        "reading bytes from guest to host at: {} {}",
        guest_ptr as u32,
        len as u32
    );

    let ptr: WasmPtr<u8, Array> = WasmPtr::new(guest_ptr as _);
    Ok(ptr
        .deref(memory, 0, len as _)
        .ok_or(wasm_error!(WasmErrorInner::Memory))?
        .iter()
        .map(|cell| cell.get())
        .collect::<Vec<u8>>())
}

/// Deserialize any DeserializeOwned type out of the guest from a guest pointer.
pub fn from_guest_ptr<O>(
    memory: &Memory,
    guest_ptr: GuestPtr,
    len: Len,
) -> Result<O, wasmer_engine::RuntimeError>
where
    O: serde::de::DeserializeOwned + std::fmt::Debug,
{
    let bytes = read_bytes(memory, guest_ptr, len)?;
    match holochain_serialized_bytes::decode(&bytes) {
        Ok(v) => Ok(v),
        Err(e) => {
            tracing::error!(input_type = std::any::type_name::<O>(), bytes = ?bytes, "{}", e);
            Err(wasm_error!(e.into()).into())
        }
    }
}

/// Host calling guest for the function named `call` with the given `payload` in a vector of bytes
/// result is either a vector of bytes from the guest found at the location of the returned guest
/// allocation pointer or a `WasmError`.
pub fn call<I, O>(
    instance: Arc<Mutex<Instance>>,
    f: &str,
    input: I,
) -> Result<O, wasmer_engine::RuntimeError>
where
    I: serde::Serialize + std::fmt::Debug,
    O: serde::de::DeserializeOwned + std::fmt::Debug,
{
    let instance = instance.lock();
    // The guest will use the same crate for decoding if it uses the wasm common crate.
    let payload: Vec<u8> =
        holochain_serialized_bytes::encode(&input).map_err(|e| wasm_error!(e.into()))?;

    // Get a pre-allocated guest pointer to write the input into.
    let guest_input_ptr: GuestPtr = match instance
        .exports
        .get_function("__allocate")
        .map_err(|e| wasm_error!(WasmErrorInner::CallError(e.to_string())))?
        .call(&[Value::I32(
            payload
                .len()
                .try_into()
                .map_err(|e: TryFromIntError| wasm_error!(e.into()))?,
        )])
        .map_err(|e| wasm_error!(WasmErrorInner::CallError(e.to_string())))?[0]
    {
        Value::I32(i) => i as GuestPtr,
        _ => unreachable!(),
    };

    // Write the input payload into the guest at the offset specified by the allocation.
    write_bytes(
        &instance
            .exports
            // potentially snake oil
            // https://github.com/wasmerio/wasmer/issues/2780#issuecomment-1054452629
            .get_with_generics_weak("memory")
            .map_err(|_| wasm_error!(WasmErrorInner::Memory))?,
        guest_input_ptr,
        &payload,
    )?;

    // Call the guest function with its own pointer to its input.
    // Collect the guest's pointer to its output.
    let (guest_return_ptr, len): (GuestPtr, Len) = match instance
        .exports
        .get_function(f)
        .map_err(|e| wasm_error!(WasmErrorInner::CallError(e.to_string())))?
        .call(&[
            Value::I32(
                guest_input_ptr
                    .try_into()
                    .map_err(|e: TryFromIntError| wasm_error!(e.into()))?,
            ),
            Value::I32(
                payload
                    .len()
                    .try_into()
                    .map_err(|e: TryFromIntError| wasm_error!(e.into()))?,
            ),
        ]) {
        Ok(v) => match v[0] {
            Value::I64(i) => split_u64(i as _),
            _ => unreachable!(),
        },
        Err(e) => match e.downcast::<WasmError>() {
            Ok(WasmError { file, line, error }) => match error {
                WasmErrorInner::HostShortCircuit(encoded) => {
                    return match holochain_serialized_bytes::decode(&encoded) {
                        Ok(v) => Ok(v),
                        Err(e) => {
                            tracing::error!(
                                input_type = std::any::type_name::<O>(),
                                ?encoded,
                                "{}",
                                e
                            );
                            Err(wasm_error!(e.into()).into())
                        }
                    }
                }
                _ => return Err(WasmError { file, line, error }.into()),
            },
            Err(e) => return Err(wasm_error!(WasmErrorInner::CallError(e.to_string())).into()),
        },
    };

    // We ? here to return early WITHOUT calling deallocate.
    // The host MUST discard any wasm instance that errors at this point to avoid memory leaks.
    // The WasmError in the result type here is for deserializing out of the guest.
    let return_value: Result<O, WasmError> = from_guest_ptr(
        &instance
            .exports
            // maybe snake oil but:
            // https://github.com/wasmerio/wasmer/issues/2780#issuecomment-1054452629
            .get_with_generics_weak("memory")
            .map_err(|_| wasm_error!(WasmErrorInner::Memory))?,
        guest_return_ptr,
        len,
    )?;

    // Tell the guest we are finished with the return pointer's data.
    instance
        .exports
        .get_function("__deallocate")
        .map_err(|e| wasm_error!(WasmErrorInner::CallError(e.to_string())))?
        .call(&[
            Value::I32(
                guest_return_ptr
                    .try_into()
                    .map_err(|e: TryFromIntError| wasm_error!(e.into()))?,
            ),
            Value::I32(
                len.try_into()
                    .map_err(|e: TryFromIntError| wasm_error!(e.into()))?,
            ),
        ])
        .map_err(|e| wasm_error!(WasmErrorInner::CallError(format!("{:?}", e))))?;

    return_value.map_err(|e| e.into())
}