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
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
// Copyright 2018 The ChromiumOS Authors
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.

use std::collections::BTreeMap;
use std::sync::Arc;

#[cfg(target_arch = "x86_64")]
use acpi_tables::sdt::SDT;
use anyhow::anyhow;
use anyhow::Result;
use base::Event;
use base::Protection;
use base::RawDescriptor;
use hypervisor::MemCacheType;
use sync::Mutex;
use vm_control::VmMemorySource;
use vm_memory::GuestAddress;
use vm_memory::GuestMemory;

use super::*;
use crate::pci::MsixConfig;
use crate::pci::MsixStatus;
use crate::pci::PciAddress;
use crate::pci::PciBarConfiguration;
use crate::pci::PciBarIndex;
use crate::pci::PciCapability;
use crate::virtio::queue::QueueConfig;

#[derive(Clone, Copy, Debug, PartialEq, Eq)]
pub enum VirtioTransportType {
    Pci,
    Mmio,
}

#[derive(Clone)]
pub struct SharedMemoryRegion {
    /// The id of the shared memory region. A device may have multiple regions, but each
    /// must have a unique id. The meaning of a particular region is device-specific.
    pub id: u8,
    pub length: u64,
}

/// Trait for mapping memory into the device's shared memory region.
pub trait SharedMemoryMapper: Send {
    /// Maps the given |source| into the shared memory region at |offset|.
    fn add_mapping(
        &mut self,
        source: VmMemorySource,
        offset: u64,
        prot: Protection,
        cache: MemCacheType,
    ) -> Result<()>;

    /// Removes the mapping beginning at |offset|.
    fn remove_mapping(&mut self, offset: u64) -> Result<()>;

    fn as_raw_descriptor(&self) -> Option<RawDescriptor> {
        None
    }
}

/// Trait for virtio devices to be driven by a virtio transport.
///
/// The lifecycle of a virtio device is to be moved to a virtio transport, which will then query the
/// device. Once the guest driver has configured the device, `VirtioDevice::activate` will be called
/// and all the events, memory, and queues for device operation will be moved into the device.
/// Optionally, a virtio device can implement device reset in which it returns said resources and
/// resets its internal.
pub trait VirtioDevice: Send {
    /// Returns a label suitable for debug output.
    fn debug_label(&self) -> String {
        format!("virtio-{}", self.device_type())
    }

    /// A vector of device-specific file descriptors that must be kept open
    /// after jailing. Must be called before the process is jailed.
    fn keep_rds(&self) -> Vec<RawDescriptor>;

    /// The virtio device type.
    fn device_type(&self) -> DeviceType;

    /// The maximum size of each queue that this device supports.
    fn queue_max_sizes(&self) -> &[u16];

    /// The number of interrupts used by this device.
    fn num_interrupts(&self) -> usize {
        self.queue_max_sizes().len()
    }

    /// The set of feature bits that this device supports in addition to the base features.
    fn features(&self) -> u64 {
        0
    }

    /// Acknowledges that this set of features should be enabled.
    fn ack_features(&mut self, value: u64) {
        let _ = value;
    }

    /// Reads this device configuration space at `offset`.
    fn read_config(&self, offset: u64, data: &mut [u8]) {
        let _ = offset;
        let _ = data;
    }

    /// Writes to this device configuration space at `offset`.
    fn write_config(&mut self, offset: u64, data: &[u8]) {
        let _ = offset;
        let _ = data;
    }

    /// Activates this device for real usage.
    fn activate(
        &mut self,
        mem: GuestMemory,
        interrupt: Interrupt,
        queues: BTreeMap<usize, Queue>,
    ) -> Result<()>;

    /// Optionally deactivates this device. If the reset method is
    /// not able to reset the virtio device, or the virtio device model doesn't
    /// implement the reset method, an `Err` value is returned to indicate
    /// the reset is not successful. Otherwise `Ok(())` should be returned.
    fn reset(&mut self) -> Result<()> {
        Err(anyhow!("reset not implemented for {}", self.debug_label()))
    }

    /// Returns any additional BAR configuration required by the device.
    fn get_device_bars(&mut self, _address: PciAddress) -> Vec<PciBarConfiguration> {
        Vec::new()
    }

    /// Returns any additional capabiltiies required by the device.
    fn get_device_caps(&self) -> Vec<Box<dyn PciCapability>> {
        Vec::new()
    }

    /// Invoked when the device is sandboxed.
    fn on_device_sandboxed(&mut self) {}

    fn control_notify(&self, _behavior: MsixStatus) {}

    #[cfg(target_arch = "x86_64")]
    fn generate_acpi(
        &mut self,
        _pci_address: &Option<PciAddress>,
        sdts: Vec<SDT>,
    ) -> Option<Vec<SDT>> {
        Some(sdts)
    }

    /// Reads from a BAR region mapped in to the device.
    /// * `addr` - The guest address inside the BAR.
    /// * `data` - Filled with the data from `addr`.
    fn read_bar(&mut self, _bar_index: PciBarIndex, _offset: u64, _data: &mut [u8]) {}

    /// Writes to a BAR region mapped in to the device.
    /// * `addr` - The guest address inside the BAR.
    /// * `data` - The data to write.
    fn write_bar(&mut self, _bar_index: PciBarIndex, _offset: u64, _data: &[u8]) {}

    /// Returns the PCI address where the device will be allocated.
    /// Returns `None` if any address is good for the device.
    fn pci_address(&self) -> Option<PciAddress> {
        None
    }

    /// Returns the Virtio transport type: PCI (default for crosvm) or MMIO.
    fn transport_type(&self) -> VirtioTransportType {
        VirtioTransportType::Pci
    }

    /// Returns the device's shared memory region if present.
    fn get_shared_memory_region(&self) -> Option<SharedMemoryRegion> {
        None
    }

    /// If true, VFIO passthrough devices can access descriptors mapped into
    /// this region by mapping the corresponding addresses from this device's
    /// PCI bar into their IO address space with virtio-iommu.
    ///
    /// NOTE: Not all vm_control::VmMemorySource types are supported.
    /// NOTE: Not yet compatible with PrepareSharedMemoryRegion (aka fixed mapping).
    fn expose_shmem_descriptors_with_viommu(&self) -> bool {
        false
    }

    /// Provides the trait object used to map files into the device's shared
    /// memory region.
    ///
    /// If `get_shared_memory_region` returns `Some`, then this will be called
    /// before `activate`.
    fn set_shared_memory_mapper(&mut self, _mapper: Box<dyn SharedMemoryMapper>) {}

    /// Provides the base address of the shared memory region, if one is present. Will
    /// be called before `activate`.
    ///
    /// NOTE: Mappings in shared memory regions should be accessed via offset, rather
    /// than via raw guest physical address. This function is only provided so
    /// devices can remain backwards compatible with older drivers.
    fn set_shared_memory_region_base(&mut self, _addr: GuestAddress) {}

    /// Pause all processing.
    ///
    /// Gives up the queues so that a higher layer can potentially snapshot them. The
    /// implementations should also drop the `Interrupt` and queues `Event`s that were given along
    /// with the queues originally.
    ///
    /// Unlike `Suspendable::sleep`, this is not idempotent. Attempting to sleep while already
    /// asleep is an error.
    fn virtio_sleep(&mut self) -> anyhow::Result<Option<BTreeMap<usize, Queue>>> {
        anyhow::bail!("virtio_sleep not implemented for {}", self.debug_label());
    }

    /// Resume all processing.
    ///
    /// If the device's queues are active, then the queues and associated data will is included.
    ///
    /// Unlike `Suspendable::wake`, this is not idempotent. Attempting to wake while already awake
    /// is an error.
    fn virtio_wake(
        &mut self,
        _queues_state: Option<(GuestMemory, Interrupt, BTreeMap<usize, Queue>)>,
    ) -> anyhow::Result<()> {
        anyhow::bail!("virtio_wake not implemented for {}", self.debug_label());
    }

    /// Snapshot current state. Device must be asleep.
    fn virtio_snapshot(&mut self) -> anyhow::Result<serde_json::Value> {
        anyhow::bail!("virtio_snapshot not implemented for {}", self.debug_label());
    }

    /// Restore device state from a snapshot.
    /// TODO(b/280607404): Vhost user will need fds passed to the device process.
    fn virtio_restore(&mut self, _data: serde_json::Value) -> anyhow::Result<()> {
        anyhow::bail!("virtio_restore not implemented for {}", self.debug_label());
    }

    /// Returns true if the device uses the vhost user protocol.
    fn is_vhost_user(&self) -> bool {
        false
    }

    /// Vhost user device specific restore to be called instead of `virtio_restore`. This will
    /// rewire irqfds, queue_evts, start up the worker if needed, and send a RESTORE request to
    /// the device process.
    fn vhost_user_restore(
        &mut self,
        _data: serde_json::Value,
        _queue_configs: &[QueueConfig],
        _queue_evts: Option<Vec<Event>>,
        _interrupt: Option<Interrupt>,
        _mem: GuestMemory,
        _msix_config: &Arc<Mutex<MsixConfig>>,
        _device_activated: bool,
    ) -> anyhow::Result<()> {
        anyhow::bail!(
            "vhost_user_restore not implemented for {}",
            self.debug_label()
        );
    }

    // Returns a tuple consisting of the non-arch specific part of the OpenFirmware path,
    // represented as bytes, and the boot index of a device. The non-arch specific part of path for
    // a virtio-blk device, for example, would consist of everything after the first '/' below:
    // pci@i0cf8/scsi@6[,3]/disk@0,0
    //    ^           ^  ^       ^ ^
    //    |           |  |       fixed
    //    |           | (PCI function related to disk (optional))
    // (x86 specf  (PCI slot holding disk)
    //  root at sys
    //  bus port)
    fn bootorder_fw_cfg(&self, _pci_address: u8) -> Option<(Vec<u8>, usize)> {
        None
    }
}

// General tests that should pass on all suspendables.
// Do implement device-specific tests to validate the functionality of the device.
// Those tests are not a replacement for regular tests. Only an extension specific to the trait's
// basic functionality.
/// `name` is the name of the test grouping. Can be anything unique within the same crate.
/// `dev` is a block that returns a created virtio device.
/// ``num_queues` is the number of queues to be created.
/// `modfun` is the function name of the function that would modify the device. The function call
/// should modify the device so that a snapshot taken after the function call would be different
/// from a snapshot taken before the function call.
#[macro_export]
macro_rules! suspendable_virtio_tests {
    ($name:ident, $dev: expr, $num_queues:literal, $modfun:expr) => {
        mod $name {
            use $crate::virtio::QueueConfig;

            use super::*;

            fn memory() -> GuestMemory {
                GuestMemory::new(&[(GuestAddress(0u64), 4 * 1024 * 1024)])
                    .expect("Creating guest memory failed.")
            }

            fn interrupt() -> Interrupt {
                Interrupt::new_for_test()
            }

            fn create_queues(
                num_queues: usize,
                queue_size: u16,
                mem: &GuestMemory,
            ) -> BTreeMap<usize, Queue> {
                let mut queues = BTreeMap::new();
                for i in 0..num_queues {
                    // activate with queues of an arbitrary size.
                    let mut queue = QueueConfig::new(queue_size, 0);
                    queue.set_ready(true);
                    let queue = queue
                        .activate(mem, Event::new().unwrap())
                        .expect("QueueConfig::activate");
                    queues.insert(i, queue);
                }
                queues
            }

            #[test]
            fn test_unactivated_sleep_snapshot_wake() {
                let (_ctx, mut device) = $dev();
                let sleep_result = device.virtio_sleep().expect("failed to sleep");
                assert!(sleep_result.is_none());
                device.virtio_snapshot().expect("failed to snapshot");
                device.virtio_wake(None).expect("failed to wake");
            }

            #[test]
            fn test_sleep_snapshot_wake() {
                let (_ctx, mut device) = $dev();
                let mem = memory();
                let interrupt = interrupt();
                let queues = create_queues(
                    $num_queues,
                    device
                        .queue_max_sizes()
                        .first()
                        .cloned()
                        .expect("missing queue size"),
                    &mem,
                );
                device
                    .activate(mem.clone(), interrupt.clone(), queues)
                    .expect("failed to activate");
                let sleep_result = device
                    .virtio_sleep()
                    .expect("failed to sleep")
                    .expect("missing queues while sleeping");
                device.virtio_snapshot().expect("failed to snapshot");
                device
                    .virtio_wake(Some((mem.clone(), interrupt.clone(), sleep_result)))
                    .expect("failed to wake");
            }

            #[test]
            fn test_suspend_mod_restore() {
                let (mut context, mut device) = $dev();
                let mem = memory();
                let interrupt = interrupt();
                let queues = create_queues(
                    $num_queues,
                    device
                        .queue_max_sizes()
                        .first()
                        .cloned()
                        .expect("missing queue size"),
                    &mem,
                );
                device
                    .activate(mem.clone(), interrupt.clone(), queues)
                    .expect("failed to activate");
                let sleep_result = device
                    .virtio_sleep()
                    .expect("failed to sleep")
                    .expect("missing queues while sleeping");
                // Modify device before snapshotting.
                $modfun(&mut context, &mut device);
                let snap = device
                    .virtio_snapshot()
                    .expect("failed to take initial snapshot");
                device
                    .virtio_wake(Some((mem.clone(), interrupt.clone(), sleep_result)))
                    .expect("failed to wake");
                let (_, device) = &mut $dev();
                device
                    .virtio_restore(snap.clone())
                    .expect("failed to restore");
                let snap2 = device
                    .virtio_snapshot()
                    .expect("failed to take snapshot after mod");
                assert_eq!(snap, snap2);
            }
        }
    };
}