1 // Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. 2 // 3 // Portions Copyright 2017 The Chromium OS Authors. All rights reserved. 4 // Use of this source code is governed by a BSD-style license that can be 5 // found in the LICENSE-BSD-3-Clause file. 6 // 7 // Copyright © 2020 Intel Corporation 8 // 9 // SPDX-License-Identifier: Apache-2.0 AND BSD-3-Clause 10 11 use std::collections::{BTreeMap, HashMap, VecDeque}; 12 use std::num::Wrapping; 13 use std::ops::Deref; 14 use std::os::unix::io::AsRawFd; 15 use std::path::PathBuf; 16 use std::sync::atomic::{AtomicBool, AtomicU64, Ordering}; 17 use std::sync::{Arc, Barrier}; 18 use std::{io, result}; 19 20 use anyhow::anyhow; 21 use block::async_io::{AsyncIo, AsyncIoError, DiskFile}; 22 use block::fcntl::{get_lock_state, LockError, LockType}; 23 use block::{build_serial, fcntl, Request, RequestType, VirtioBlockConfig}; 24 use rate_limiter::group::{RateLimiterGroup, RateLimiterGroupHandle}; 25 use rate_limiter::TokenType; 26 use seccompiler::SeccompAction; 27 use serde::{Deserialize, Serialize}; 28 use thiserror::Error; 29 use virtio_bindings::virtio_blk::*; 30 use virtio_bindings::virtio_config::*; 31 use virtio_bindings::virtio_ring::{VIRTIO_RING_F_EVENT_IDX, VIRTIO_RING_F_INDIRECT_DESC}; 32 use virtio_queue::{Queue, QueueOwnedT, QueueT}; 33 use vm_memory::{ByteValued, Bytes, GuestAddressSpace, GuestMemoryAtomic, GuestMemoryError}; 34 use vm_migration::{Migratable, MigratableError, Pausable, Snapshot, Snapshottable, Transportable}; 35 use vm_virtio::AccessPlatform; 36 use vmm_sys_util::eventfd::EventFd; 37 38 use super::{ 39 ActivateError, ActivateResult, EpollHelper, EpollHelperError, EpollHelperHandler, 40 Error as DeviceError, VirtioCommon, VirtioDevice, VirtioDeviceType, VirtioInterruptType, 41 EPOLL_HELPER_EVENT_LAST, 42 }; 43 use crate::seccomp_filters::Thread; 44 use crate::thread_helper::spawn_virtio_thread; 45 use crate::{GuestMemoryMmap, VirtioInterrupt}; 46 47 const SECTOR_SHIFT: u8 = 9; 48 pub const SECTOR_SIZE: u64 = 0x01 << SECTOR_SHIFT; 49 50 // New descriptors are pending on the virtio queue. 51 const QUEUE_AVAIL_EVENT: u16 = EPOLL_HELPER_EVENT_LAST + 1; 52 // New completed tasks are pending on the completion ring. 53 const COMPLETION_EVENT: u16 = EPOLL_HELPER_EVENT_LAST + 2; 54 // New 'wake up' event from the rate limiter 55 const RATE_LIMITER_EVENT: u16 = EPOLL_HELPER_EVENT_LAST + 3; 56 57 // latency scale, for reduce precision loss in calculate. 58 const LATENCY_SCALE: u64 = 10000; 59 60 pub const MINIMUM_BLOCK_QUEUE_SIZE: u16 = 2; 61 62 #[derive(Error, Debug)] 63 pub enum Error { 64 #[error("Failed to parse the request: {0}")] 65 RequestParsing(#[source] block::Error), 66 #[error("Failed to execute the request: {0}")] 67 RequestExecuting(#[source] block::ExecuteError), 68 #[error("Failed to complete the request: {0}")] 69 RequestCompleting(#[source] block::Error), 70 #[error("Missing the expected entry in the list of requests")] 71 MissingEntryRequestList, 72 #[error("The asynchronous request returned with failure")] 73 AsyncRequestFailure, 74 #[error("Failed synchronizing the file: {0}")] 75 Fsync(#[source] AsyncIoError), 76 #[error("Failed adding used index: {0}")] 77 QueueAddUsed(#[source] virtio_queue::Error), 78 #[error("Failed creating an iterator over the queue: {0}")] 79 QueueIterator(#[source] virtio_queue::Error), 80 #[error("Failed to update request status: {0}")] 81 RequestStatus(#[source] GuestMemoryError), 82 #[error("Failed to enable notification: {0}")] 83 QueueEnableNotification(#[source] virtio_queue::Error), 84 #[error("Failed to get {lock_type:?} lock for disk image {path}: {error}")] 85 LockDiskImage { 86 /// The underlying error. 87 #[source] 88 error: LockError, 89 /// The requested lock type. 90 lock_type: LockType, 91 /// The path of the disk image. 92 path: PathBuf, 93 }, 94 } 95 96 pub type Result<T> = result::Result<T, Error>; 97 98 // latency will be records as microseconds, average latency 99 // will be save as scaled value. 100 #[derive(Clone)] 101 pub struct BlockCounters { 102 read_bytes: Arc<AtomicU64>, 103 read_ops: Arc<AtomicU64>, 104 read_latency_min: Arc<AtomicU64>, 105 read_latency_max: Arc<AtomicU64>, 106 read_latency_avg: Arc<AtomicU64>, 107 write_bytes: Arc<AtomicU64>, 108 write_ops: Arc<AtomicU64>, 109 write_latency_min: Arc<AtomicU64>, 110 write_latency_max: Arc<AtomicU64>, 111 write_latency_avg: Arc<AtomicU64>, 112 } 113 114 impl Default for BlockCounters { 115 fn default() -> Self { 116 BlockCounters { 117 read_bytes: Arc::new(AtomicU64::new(0)), 118 read_ops: Arc::new(AtomicU64::new(0)), 119 read_latency_min: Arc::new(AtomicU64::new(u64::MAX)), 120 read_latency_max: Arc::new(AtomicU64::new(u64::MAX)), 121 read_latency_avg: Arc::new(AtomicU64::new(u64::MAX)), 122 write_bytes: Arc::new(AtomicU64::new(0)), 123 write_ops: Arc::new(AtomicU64::new(0)), 124 write_latency_min: Arc::new(AtomicU64::new(u64::MAX)), 125 write_latency_max: Arc::new(AtomicU64::new(u64::MAX)), 126 write_latency_avg: Arc::new(AtomicU64::new(u64::MAX)), 127 } 128 } 129 } 130 131 struct BlockEpollHandler { 132 queue_index: u16, 133 queue: Queue, 134 mem: GuestMemoryAtomic<GuestMemoryMmap>, 135 disk_image: Box<dyn AsyncIo>, 136 disk_nsectors: u64, 137 interrupt_cb: Arc<dyn VirtioInterrupt>, 138 serial: Vec<u8>, 139 kill_evt: EventFd, 140 pause_evt: EventFd, 141 writeback: Arc<AtomicBool>, 142 counters: BlockCounters, 143 queue_evt: EventFd, 144 inflight_requests: VecDeque<(u16, Request)>, 145 rate_limiter: Option<RateLimiterGroupHandle>, 146 access_platform: Option<Arc<dyn AccessPlatform>>, 147 read_only: bool, 148 host_cpus: Option<Vec<usize>>, 149 } 150 151 impl BlockEpollHandler { 152 fn process_queue_submit(&mut self) -> Result<()> { 153 let queue = &mut self.queue; 154 155 while let Some(mut desc_chain) = queue.pop_descriptor_chain(self.mem.memory()) { 156 let mut request = Request::parse(&mut desc_chain, self.access_platform.as_ref()) 157 .map_err(Error::RequestParsing)?; 158 159 // For virtio spec compliance 160 // "A device MUST set the status byte to VIRTIO_BLK_S_IOERR for a write request 161 // if the VIRTIO_BLK_F_RO feature if offered, and MUST NOT write any data." 162 if self.read_only 163 && (request.request_type == RequestType::Out 164 || request.request_type == RequestType::Flush) 165 { 166 desc_chain 167 .memory() 168 .write_obj(VIRTIO_BLK_S_IOERR, request.status_addr) 169 .map_err(Error::RequestStatus)?; 170 171 // If no asynchronous operation has been submitted, we can 172 // simply return the used descriptor. 173 queue 174 .add_used(desc_chain.memory(), desc_chain.head_index(), 0) 175 .map_err(Error::QueueAddUsed)?; 176 queue 177 .enable_notification(self.mem.memory().deref()) 178 .map_err(Error::QueueEnableNotification)?; 179 continue; 180 } 181 182 if let Some(rate_limiter) = &mut self.rate_limiter { 183 // If limiter.consume() fails it means there is no more TokenType::Ops 184 // budget and rate limiting is in effect. 185 if !rate_limiter.consume(1, TokenType::Ops) { 186 // Stop processing the queue and return this descriptor chain to the 187 // avail ring, for later processing. 188 queue.go_to_previous_position(); 189 break; 190 } 191 // Exercise the rate limiter only if this request is of data transfer type. 192 if request.request_type == RequestType::In 193 || request.request_type == RequestType::Out 194 { 195 let mut bytes = Wrapping(0); 196 for (_, data_len) in &request.data_descriptors { 197 bytes += Wrapping(*data_len as u64); 198 } 199 200 // If limiter.consume() fails it means there is no more TokenType::Bytes 201 // budget and rate limiting is in effect. 202 if !rate_limiter.consume(bytes.0, TokenType::Bytes) { 203 // Revert the OPS consume(). 204 rate_limiter.manual_replenish(1, TokenType::Ops); 205 // Stop processing the queue and return this descriptor chain to the 206 // avail ring, for later processing. 207 queue.go_to_previous_position(); 208 break; 209 } 210 }; 211 } 212 213 request.set_writeback(self.writeback.load(Ordering::Acquire)); 214 215 if request 216 .execute_async( 217 desc_chain.memory(), 218 self.disk_nsectors, 219 self.disk_image.as_mut(), 220 &self.serial, 221 desc_chain.head_index() as u64, 222 ) 223 .map_err(Error::RequestExecuting)? 224 { 225 self.inflight_requests 226 .push_back((desc_chain.head_index(), request)); 227 } else { 228 desc_chain 229 .memory() 230 .write_obj(VIRTIO_BLK_S_OK as u8, request.status_addr) 231 .map_err(Error::RequestStatus)?; 232 233 // If no asynchronous operation has been submitted, we can 234 // simply return the used descriptor. 235 queue 236 .add_used(desc_chain.memory(), desc_chain.head_index(), 0) 237 .map_err(Error::QueueAddUsed)?; 238 queue 239 .enable_notification(self.mem.memory().deref()) 240 .map_err(Error::QueueEnableNotification)?; 241 } 242 } 243 244 Ok(()) 245 } 246 247 fn try_signal_used_queue(&mut self) -> result::Result<(), EpollHelperError> { 248 if self 249 .queue 250 .needs_notification(self.mem.memory().deref()) 251 .map_err(|e| { 252 EpollHelperError::HandleEvent(anyhow!( 253 "Failed to check needs_notification: {:?}", 254 e 255 )) 256 })? 257 { 258 self.signal_used_queue().map_err(|e| { 259 EpollHelperError::HandleEvent(anyhow!("Failed to signal used queue: {:?}", e)) 260 })?; 261 } 262 263 Ok(()) 264 } 265 266 fn process_queue_submit_and_signal(&mut self) -> result::Result<(), EpollHelperError> { 267 self.process_queue_submit().map_err(|e| { 268 EpollHelperError::HandleEvent(anyhow!("Failed to process queue (submit): {:?}", e)) 269 })?; 270 271 self.try_signal_used_queue() 272 } 273 274 #[inline] 275 fn find_inflight_request(&mut self, completed_head: u16) -> Result<Request> { 276 // This loop neatly handles the fast path where the completions are 277 // in order (it turns into just a pop_front()) and the 1% of the time 278 // (analysis during boot) where slight out of ordering has been 279 // observed e.g. 280 // Submissions: 1 2 3 4 5 6 7 281 // Completions: 2 1 3 5 4 7 6 282 // In this case find the corresponding item and swap it with the front 283 // This is a O(1) operation and is prepared for the future as it it likely 284 // the next completion would be for the one that was skipped which will 285 // now be the new front. 286 for (i, (head, _)) in self.inflight_requests.iter().enumerate() { 287 if head == &completed_head { 288 return Ok(self.inflight_requests.swap_remove_front(i).unwrap().1); 289 } 290 } 291 292 Err(Error::MissingEntryRequestList) 293 } 294 295 fn process_queue_complete(&mut self) -> Result<()> { 296 let mem = self.mem.memory(); 297 let mut read_bytes = Wrapping(0); 298 let mut write_bytes = Wrapping(0); 299 let mut read_ops = Wrapping(0); 300 let mut write_ops = Wrapping(0); 301 302 while let Some((user_data, result)) = self.disk_image.next_completed_request() { 303 let desc_index = user_data as u16; 304 305 let mut request = self.find_inflight_request(desc_index)?; 306 307 request.complete_async().map_err(Error::RequestCompleting)?; 308 309 let latency = request.start.elapsed().as_micros() as u64; 310 let read_ops_last = self.counters.read_ops.load(Ordering::Relaxed); 311 let write_ops_last = self.counters.write_ops.load(Ordering::Relaxed); 312 let read_max = self.counters.read_latency_max.load(Ordering::Relaxed); 313 let write_max = self.counters.write_latency_max.load(Ordering::Relaxed); 314 let mut read_avg = self.counters.read_latency_avg.load(Ordering::Relaxed); 315 let mut write_avg = self.counters.write_latency_avg.load(Ordering::Relaxed); 316 let (status, len) = if result >= 0 { 317 match request.request_type { 318 RequestType::In => { 319 for (_, data_len) in &request.data_descriptors { 320 read_bytes += Wrapping(*data_len as u64); 321 } 322 read_ops += Wrapping(1); 323 if latency < self.counters.read_latency_min.load(Ordering::Relaxed) { 324 self.counters 325 .read_latency_min 326 .store(latency, Ordering::Relaxed); 327 } 328 if latency > read_max || read_max == u64::MAX { 329 self.counters 330 .read_latency_max 331 .store(latency, Ordering::Relaxed); 332 } 333 334 // Special case the first real latency report 335 read_avg = if read_avg == u64::MAX { 336 latency * LATENCY_SCALE 337 } else { 338 // Cumulative average is guaranteed to be 339 // positive if being calculated properly 340 (read_avg as i64 341 + ((latency * LATENCY_SCALE) as i64 - read_avg as i64) 342 / (read_ops_last + read_ops.0) as i64) 343 .try_into() 344 .unwrap() 345 }; 346 } 347 RequestType::Out => { 348 if !request.writeback { 349 self.disk_image.fsync(None).map_err(Error::Fsync)?; 350 } 351 for (_, data_len) in &request.data_descriptors { 352 write_bytes += Wrapping(*data_len as u64); 353 } 354 write_ops += Wrapping(1); 355 if latency < self.counters.write_latency_min.load(Ordering::Relaxed) { 356 self.counters 357 .write_latency_min 358 .store(latency, Ordering::Relaxed); 359 } 360 if latency > write_max || write_max == u64::MAX { 361 self.counters 362 .write_latency_max 363 .store(latency, Ordering::Relaxed); 364 } 365 366 // Special case the first real latency report 367 write_avg = if write_avg == u64::MAX { 368 latency * LATENCY_SCALE 369 } else { 370 // Cumulative average is guaranteed to be 371 // positive if being calculated properly 372 (write_avg as i64 373 + ((latency * LATENCY_SCALE) as i64 - write_avg as i64) 374 / (write_ops_last + write_ops.0) as i64) 375 .try_into() 376 .unwrap() 377 } 378 } 379 _ => {} 380 } 381 382 self.counters 383 .read_latency_avg 384 .store(read_avg, Ordering::Relaxed); 385 386 self.counters 387 .write_latency_avg 388 .store(write_avg, Ordering::Relaxed); 389 390 (VIRTIO_BLK_S_OK as u8, result as u32) 391 } else { 392 error!( 393 "Request failed: {:x?} {:?}", 394 request, 395 io::Error::from_raw_os_error(-result) 396 ); 397 return Err(Error::AsyncRequestFailure); 398 }; 399 400 mem.write_obj(status, request.status_addr) 401 .map_err(Error::RequestStatus)?; 402 403 let queue = &mut self.queue; 404 405 queue 406 .add_used(mem.deref(), desc_index, len) 407 .map_err(Error::QueueAddUsed)?; 408 queue 409 .enable_notification(mem.deref()) 410 .map_err(Error::QueueEnableNotification)?; 411 } 412 413 self.counters 414 .write_bytes 415 .fetch_add(write_bytes.0, Ordering::AcqRel); 416 self.counters 417 .write_ops 418 .fetch_add(write_ops.0, Ordering::AcqRel); 419 420 self.counters 421 .read_bytes 422 .fetch_add(read_bytes.0, Ordering::AcqRel); 423 self.counters 424 .read_ops 425 .fetch_add(read_ops.0, Ordering::AcqRel); 426 427 Ok(()) 428 } 429 430 fn signal_used_queue(&self) -> result::Result<(), DeviceError> { 431 self.interrupt_cb 432 .trigger(VirtioInterruptType::Queue(self.queue_index)) 433 .map_err(|e| { 434 error!("Failed to signal used queue: {:?}", e); 435 DeviceError::FailedSignalingUsedQueue(e) 436 }) 437 } 438 439 fn set_queue_thread_affinity(&self) { 440 // Prepare the CPU set the current queue thread is expected to run onto. 441 let cpuset = self.host_cpus.as_ref().map(|host_cpus| { 442 // SAFETY: all zeros is a valid pattern 443 let mut cpuset: libc::cpu_set_t = unsafe { std::mem::zeroed() }; 444 // SAFETY: FFI call, trivially safe 445 unsafe { libc::CPU_ZERO(&mut cpuset) }; 446 for host_cpu in host_cpus { 447 // SAFETY: FFI call, trivially safe 448 unsafe { libc::CPU_SET(*host_cpu, &mut cpuset) }; 449 } 450 cpuset 451 }); 452 453 // Schedule the thread to run on the expected CPU set 454 if let Some(cpuset) = cpuset.as_ref() { 455 // SAFETY: FFI call with correct arguments 456 let ret = unsafe { 457 libc::sched_setaffinity( 458 0, 459 std::mem::size_of::<libc::cpu_set_t>(), 460 cpuset as *const libc::cpu_set_t, 461 ) 462 }; 463 464 if ret != 0 { 465 error!( 466 "Failed scheduling the virtqueue thread {} on the expected CPU set: {}", 467 self.queue_index, 468 io::Error::last_os_error() 469 ) 470 } 471 } 472 } 473 474 fn run( 475 &mut self, 476 paused: Arc<AtomicBool>, 477 paused_sync: Arc<Barrier>, 478 ) -> result::Result<(), EpollHelperError> { 479 let mut helper = EpollHelper::new(&self.kill_evt, &self.pause_evt)?; 480 helper.add_event(self.queue_evt.as_raw_fd(), QUEUE_AVAIL_EVENT)?; 481 helper.add_event(self.disk_image.notifier().as_raw_fd(), COMPLETION_EVENT)?; 482 if let Some(rate_limiter) = &self.rate_limiter { 483 helper.add_event(rate_limiter.as_raw_fd(), RATE_LIMITER_EVENT)?; 484 } 485 self.set_queue_thread_affinity(); 486 helper.run(paused, paused_sync, self)?; 487 488 Ok(()) 489 } 490 } 491 492 impl EpollHelperHandler for BlockEpollHandler { 493 fn handle_event( 494 &mut self, 495 _helper: &mut EpollHelper, 496 event: &epoll::Event, 497 ) -> result::Result<(), EpollHelperError> { 498 let ev_type = event.data as u16; 499 match ev_type { 500 QUEUE_AVAIL_EVENT => { 501 self.queue_evt.read().map_err(|e| { 502 EpollHelperError::HandleEvent(anyhow!("Failed to get queue event: {:?}", e)) 503 })?; 504 505 let rate_limit_reached = self.rate_limiter.as_ref().is_some_and(|r| r.is_blocked()); 506 507 // Process the queue only when the rate limit is not reached 508 if !rate_limit_reached { 509 self.process_queue_submit_and_signal()? 510 } 511 } 512 COMPLETION_EVENT => { 513 self.disk_image.notifier().read().map_err(|e| { 514 EpollHelperError::HandleEvent(anyhow!("Failed to get queue event: {:?}", e)) 515 })?; 516 517 self.process_queue_complete().map_err(|e| { 518 EpollHelperError::HandleEvent(anyhow!( 519 "Failed to process queue (complete): {:?}", 520 e 521 )) 522 })?; 523 524 let rate_limit_reached = self.rate_limiter.as_ref().is_some_and(|r| r.is_blocked()); 525 526 // Process the queue only when the rate limit is not reached 527 if !rate_limit_reached { 528 self.process_queue_submit().map_err(|e| { 529 EpollHelperError::HandleEvent(anyhow!( 530 "Failed to process queue (submit): {:?}", 531 e 532 )) 533 })?; 534 } 535 self.try_signal_used_queue()?; 536 } 537 RATE_LIMITER_EVENT => { 538 if let Some(rate_limiter) = &mut self.rate_limiter { 539 // Upon rate limiter event, call the rate limiter handler 540 // and restart processing the queue. 541 rate_limiter.event_handler().map_err(|e| { 542 EpollHelperError::HandleEvent(anyhow!( 543 "Failed to process rate limiter event: {:?}", 544 e 545 )) 546 })?; 547 548 self.process_queue_submit_and_signal()? 549 } else { 550 return Err(EpollHelperError::HandleEvent(anyhow!( 551 "Unexpected 'RATE_LIMITER_EVENT' when rate_limiter is not enabled." 552 ))); 553 } 554 } 555 _ => { 556 return Err(EpollHelperError::HandleEvent(anyhow!( 557 "Unexpected event: {}", 558 ev_type 559 ))); 560 } 561 } 562 Ok(()) 563 } 564 } 565 566 /// Virtio device for exposing block level read/write operations on a host file. 567 pub struct Block { 568 common: VirtioCommon, 569 id: String, 570 disk_image: Box<dyn DiskFile>, 571 disk_path: PathBuf, 572 disk_nsectors: u64, 573 config: VirtioBlockConfig, 574 writeback: Arc<AtomicBool>, 575 counters: BlockCounters, 576 seccomp_action: SeccompAction, 577 rate_limiter: Option<Arc<RateLimiterGroup>>, 578 exit_evt: EventFd, 579 read_only: bool, 580 serial: Vec<u8>, 581 queue_affinity: BTreeMap<u16, Vec<usize>>, 582 } 583 584 #[derive(Serialize, Deserialize)] 585 pub struct BlockState { 586 pub disk_path: String, 587 pub disk_nsectors: u64, 588 pub avail_features: u64, 589 pub acked_features: u64, 590 pub config: VirtioBlockConfig, 591 } 592 593 impl Block { 594 /// Create a new virtio block device that operates on the given file. 595 #[allow(clippy::too_many_arguments)] 596 pub fn new( 597 id: String, 598 mut disk_image: Box<dyn DiskFile>, 599 disk_path: PathBuf, 600 read_only: bool, 601 iommu: bool, 602 num_queues: usize, 603 queue_size: u16, 604 serial: Option<String>, 605 seccomp_action: SeccompAction, 606 rate_limiter: Option<Arc<RateLimiterGroup>>, 607 exit_evt: EventFd, 608 state: Option<BlockState>, 609 queue_affinity: BTreeMap<u16, Vec<usize>>, 610 ) -> io::Result<Self> { 611 let (disk_nsectors, avail_features, acked_features, config, paused) = 612 if let Some(state) = state { 613 info!("Restoring virtio-block {}", id); 614 ( 615 state.disk_nsectors, 616 state.avail_features, 617 state.acked_features, 618 state.config, 619 true, 620 ) 621 } else { 622 let disk_size = disk_image 623 .size() 624 .map_err(|e| io::Error::other(format!("Failed getting disk size: {e}")))?; 625 if disk_size % SECTOR_SIZE != 0 { 626 warn!( 627 "Disk size {} is not a multiple of sector size {}; \ 628 the remainder will not be visible to the guest.", 629 disk_size, SECTOR_SIZE 630 ); 631 } 632 633 let mut avail_features = (1u64 << VIRTIO_F_VERSION_1) 634 | (1u64 << VIRTIO_BLK_F_FLUSH) 635 | (1u64 << VIRTIO_BLK_F_CONFIG_WCE) 636 | (1u64 << VIRTIO_BLK_F_BLK_SIZE) 637 | (1u64 << VIRTIO_BLK_F_TOPOLOGY) 638 | (1u64 << VIRTIO_BLK_F_SEG_MAX) 639 | (1u64 << VIRTIO_RING_F_EVENT_IDX) 640 | (1u64 << VIRTIO_RING_F_INDIRECT_DESC); 641 if iommu { 642 avail_features |= 1u64 << VIRTIO_F_IOMMU_PLATFORM; 643 } 644 645 if read_only { 646 avail_features |= 1u64 << VIRTIO_BLK_F_RO; 647 } 648 649 let topology = disk_image.topology(); 650 info!("Disk topology: {:?}", topology); 651 652 let logical_block_size = if topology.logical_block_size > 512 { 653 topology.logical_block_size 654 } else { 655 512 656 }; 657 658 // Calculate the exponent that maps physical block to logical block 659 let mut physical_block_exp = 0; 660 let mut size = logical_block_size; 661 while size < topology.physical_block_size { 662 physical_block_exp += 1; 663 size <<= 1; 664 } 665 666 let disk_nsectors = disk_size / SECTOR_SIZE; 667 let mut config = VirtioBlockConfig { 668 capacity: disk_nsectors, 669 writeback: 1, 670 blk_size: topology.logical_block_size as u32, 671 physical_block_exp, 672 min_io_size: (topology.minimum_io_size / logical_block_size) as u16, 673 opt_io_size: (topology.optimal_io_size / logical_block_size) as u32, 674 seg_max: (queue_size - MINIMUM_BLOCK_QUEUE_SIZE) as u32, 675 ..Default::default() 676 }; 677 678 if num_queues > 1 { 679 avail_features |= 1u64 << VIRTIO_BLK_F_MQ; 680 config.num_queues = num_queues as u16; 681 } 682 683 (disk_nsectors, avail_features, 0, config, false) 684 }; 685 686 let serial = serial 687 .map(Vec::from) 688 .unwrap_or_else(|| build_serial(&disk_path)); 689 690 Ok(Block { 691 common: VirtioCommon { 692 device_type: VirtioDeviceType::Block as u32, 693 avail_features, 694 acked_features, 695 paused_sync: Some(Arc::new(Barrier::new(num_queues + 1))), 696 queue_sizes: vec![queue_size; num_queues], 697 min_queues: 1, 698 paused: Arc::new(AtomicBool::new(paused)), 699 ..Default::default() 700 }, 701 id, 702 disk_image, 703 disk_path, 704 disk_nsectors, 705 config, 706 writeback: Arc::new(AtomicBool::new(true)), 707 counters: BlockCounters::default(), 708 seccomp_action, 709 rate_limiter, 710 exit_evt, 711 read_only, 712 serial, 713 queue_affinity, 714 }) 715 } 716 717 /// Tries to set an advisory lock for the corresponding disk image. 718 pub fn try_lock_image(&mut self) -> Result<()> { 719 let lock_type = match self.read_only { 720 true => LockType::Read, 721 false => LockType::Write, 722 }; 723 log::debug!( 724 "Attempting to acquire {lock_type:?} lock for disk image id={},path={}", 725 self.id, 726 self.disk_path.display() 727 ); 728 let fd = self.disk_image.fd(); 729 fcntl::try_acquire_lock(fd, lock_type).map_err(|error| { 730 let current_lock = get_lock_state(fd); 731 // Don't propagate the error to the outside, as it is not useful at all. Instead, 732 // we try to log additional help to the user. 733 if let Ok(current_lock) = current_lock { 734 log::error!("Can't get {lock_type:?} lock for {} as there is already a {current_lock:?} lock", self.disk_path.display()); 735 } else { 736 log::error!("Can't get {lock_type:?} lock for {}, but also can't determine the current lock state", self.disk_path.display()); 737 } 738 Error::LockDiskImage { 739 path: self.disk_path.clone(), 740 error, 741 lock_type, 742 } 743 })?; 744 log::info!( 745 "Acquired {lock_type:?} lock for disk image id={},path={}", 746 self.id, 747 self.disk_path.display() 748 ); 749 Ok(()) 750 } 751 752 /// Releases the advisory lock held for the corresponding disk image. 753 pub fn unlock_image(&mut self) -> Result<()> { 754 // It is very unlikely that this fails; 755 // Should we remove the Result to simplify the error propagation on 756 // higher levels? 757 fcntl::clear_lock(self.disk_image.fd()).map_err(|error| Error::LockDiskImage { 758 path: self.disk_path.clone(), 759 error, 760 lock_type: LockType::Unlock, 761 }) 762 } 763 764 fn state(&self) -> BlockState { 765 BlockState { 766 disk_path: self.disk_path.to_str().unwrap().to_owned(), 767 disk_nsectors: self.disk_nsectors, 768 avail_features: self.common.avail_features, 769 acked_features: self.common.acked_features, 770 config: self.config, 771 } 772 } 773 774 fn update_writeback(&mut self) { 775 // Use writeback from config if VIRTIO_BLK_F_CONFIG_WCE 776 let writeback = if self.common.feature_acked(VIRTIO_BLK_F_CONFIG_WCE.into()) { 777 self.config.writeback == 1 778 } else { 779 // Else check if VIRTIO_BLK_F_FLUSH negotiated 780 self.common.feature_acked(VIRTIO_BLK_F_FLUSH.into()) 781 }; 782 783 info!( 784 "Changing cache mode to {}", 785 if writeback { 786 "writeback" 787 } else { 788 "writethrough" 789 } 790 ); 791 self.writeback.store(writeback, Ordering::Release); 792 } 793 794 #[cfg(fuzzing)] 795 pub fn wait_for_epoll_threads(&mut self) { 796 self.common.wait_for_epoll_threads(); 797 } 798 } 799 800 impl Drop for Block { 801 fn drop(&mut self) { 802 if let Some(kill_evt) = self.common.kill_evt.take() { 803 // Ignore the result because there is nothing we can do about it. 804 let _ = kill_evt.write(1); 805 } 806 self.common.wait_for_epoll_threads(); 807 } 808 } 809 810 impl VirtioDevice for Block { 811 fn device_type(&self) -> u32 { 812 self.common.device_type 813 } 814 815 fn queue_max_sizes(&self) -> &[u16] { 816 &self.common.queue_sizes 817 } 818 819 fn features(&self) -> u64 { 820 self.common.avail_features 821 } 822 823 fn ack_features(&mut self, value: u64) { 824 self.common.ack_features(value) 825 } 826 827 fn read_config(&self, offset: u64, data: &mut [u8]) { 828 self.read_config_from_slice(self.config.as_slice(), offset, data); 829 } 830 831 fn write_config(&mut self, offset: u64, data: &[u8]) { 832 // The "writeback" field is the only mutable field 833 let writeback_offset = 834 (&self.config.writeback as *const _ as u64) - (&self.config as *const _ as u64); 835 if offset != writeback_offset || data.len() != std::mem::size_of_val(&self.config.writeback) 836 { 837 error!( 838 "Attempt to write to read-only field: offset {:x} length {}", 839 offset, 840 data.len() 841 ); 842 return; 843 } 844 845 self.config.writeback = data[0]; 846 self.update_writeback(); 847 } 848 849 fn activate( 850 &mut self, 851 mem: GuestMemoryAtomic<GuestMemoryMmap>, 852 interrupt_cb: Arc<dyn VirtioInterrupt>, 853 mut queues: Vec<(usize, Queue, EventFd)>, 854 ) -> ActivateResult { 855 self.common.activate(&queues, &interrupt_cb)?; 856 857 self.update_writeback(); 858 859 let mut epoll_threads = Vec::new(); 860 let event_idx = self.common.feature_acked(VIRTIO_RING_F_EVENT_IDX.into()); 861 862 for i in 0..queues.len() { 863 let (_, mut queue, queue_evt) = queues.remove(0); 864 queue.set_event_idx(event_idx); 865 866 let queue_size = queue.size(); 867 let (kill_evt, pause_evt) = self.common.dup_eventfds(); 868 let queue_idx = i as u16; 869 870 let mut handler = BlockEpollHandler { 871 queue_index: queue_idx, 872 queue, 873 mem: mem.clone(), 874 disk_image: self 875 .disk_image 876 .new_async_io(queue_size as u32) 877 .map_err(|e| { 878 error!("failed to create new AsyncIo: {}", e); 879 ActivateError::BadActivate 880 })?, 881 disk_nsectors: self.disk_nsectors, 882 interrupt_cb: interrupt_cb.clone(), 883 serial: self.serial.clone(), 884 kill_evt, 885 pause_evt, 886 writeback: self.writeback.clone(), 887 counters: self.counters.clone(), 888 queue_evt, 889 // Analysis during boot shows around ~40 maximum requests 890 // This gives head room for systems with slower I/O without 891 // compromising the cost of the reallocation or memory overhead 892 inflight_requests: VecDeque::with_capacity(64), 893 rate_limiter: self 894 .rate_limiter 895 .as_ref() 896 .map(|r| r.new_handle()) 897 .transpose() 898 .unwrap(), 899 access_platform: self.common.access_platform.clone(), 900 read_only: self.read_only, 901 host_cpus: self.queue_affinity.get(&queue_idx).cloned(), 902 }; 903 904 let paused = self.common.paused.clone(); 905 let paused_sync = self.common.paused_sync.clone(); 906 907 spawn_virtio_thread( 908 &format!("{}_q{}", self.id.clone(), i), 909 &self.seccomp_action, 910 Thread::VirtioBlock, 911 &mut epoll_threads, 912 &self.exit_evt, 913 move || handler.run(paused, paused_sync.unwrap()), 914 )?; 915 } 916 917 self.common.epoll_threads = Some(epoll_threads); 918 event!("virtio-device", "activated", "id", &self.id); 919 920 Ok(()) 921 } 922 923 fn reset(&mut self) -> Option<Arc<dyn VirtioInterrupt>> { 924 let result = self.common.reset(); 925 event!("virtio-device", "reset", "id", &self.id); 926 result 927 } 928 929 fn counters(&self) -> Option<HashMap<&'static str, Wrapping<u64>>> { 930 let mut counters = HashMap::new(); 931 932 counters.insert( 933 "read_bytes", 934 Wrapping(self.counters.read_bytes.load(Ordering::Acquire)), 935 ); 936 counters.insert( 937 "write_bytes", 938 Wrapping(self.counters.write_bytes.load(Ordering::Acquire)), 939 ); 940 counters.insert( 941 "read_ops", 942 Wrapping(self.counters.read_ops.load(Ordering::Acquire)), 943 ); 944 counters.insert( 945 "write_ops", 946 Wrapping(self.counters.write_ops.load(Ordering::Acquire)), 947 ); 948 counters.insert( 949 "write_latency_min", 950 Wrapping(self.counters.write_latency_min.load(Ordering::Acquire)), 951 ); 952 counters.insert( 953 "write_latency_max", 954 Wrapping(self.counters.write_latency_max.load(Ordering::Acquire)), 955 ); 956 counters.insert( 957 "write_latency_avg", 958 Wrapping(self.counters.write_latency_avg.load(Ordering::Acquire) / LATENCY_SCALE), 959 ); 960 counters.insert( 961 "read_latency_min", 962 Wrapping(self.counters.read_latency_min.load(Ordering::Acquire)), 963 ); 964 counters.insert( 965 "read_latency_max", 966 Wrapping(self.counters.read_latency_max.load(Ordering::Acquire)), 967 ); 968 counters.insert( 969 "read_latency_avg", 970 Wrapping(self.counters.read_latency_avg.load(Ordering::Acquire) / LATENCY_SCALE), 971 ); 972 973 Some(counters) 974 } 975 976 fn set_access_platform(&mut self, access_platform: Arc<dyn AccessPlatform>) { 977 self.common.set_access_platform(access_platform) 978 } 979 } 980 981 impl Pausable for Block { 982 fn pause(&mut self) -> result::Result<(), MigratableError> { 983 self.common.pause() 984 } 985 986 fn resume(&mut self) -> result::Result<(), MigratableError> { 987 self.common.resume() 988 } 989 } 990 991 impl Snapshottable for Block { 992 fn id(&self) -> String { 993 self.id.clone() 994 } 995 996 fn snapshot(&mut self) -> std::result::Result<Snapshot, MigratableError> { 997 Snapshot::new_from_state(&self.state()) 998 } 999 } 1000 impl Transportable for Block {} 1001 impl Migratable for Block {} 1002