1 // Copyright © 2021 Intel Corporation 2 // 3 // SPDX-License-Identifier: Apache-2.0 4 // 5 6 #![allow(clippy::undocumented_unsafe_blocks)] 7 8 use once_cell::sync::Lazy; 9 use serde_json::Value; 10 use ssh2::Session; 11 use std::env; 12 use std::ffi::OsStr; 13 use std::io; 14 use std::io::{Read, Write}; 15 use std::net::TcpListener; 16 use std::net::TcpStream; 17 use std::os::unix::fs::PermissionsExt; 18 use std::os::unix::io::{AsRawFd, FromRawFd}; 19 use std::path::Path; 20 use std::process::{Child, Command, ExitStatus, Output, Stdio}; 21 use std::str::FromStr; 22 use std::sync::Mutex; 23 use std::thread; 24 use std::time::Duration; 25 use std::{fmt, fs}; 26 use vmm_sys_util::tempdir::TempDir; 27 use wait_timeout::ChildExt; 28 29 #[derive(Debug)] 30 pub enum WaitTimeoutError { 31 Timedout, 32 ExitStatus, 33 General(std::io::Error), 34 } 35 36 #[derive(Debug)] 37 pub enum Error { 38 Parsing(std::num::ParseIntError), 39 SshCommand(SshCommandError), 40 WaitForBoot(WaitForBootError), 41 EthrLogFile(std::io::Error), 42 EthrLogParse, 43 FioOutputParse, 44 Iperf3Parse, 45 Spawn(std::io::Error), 46 WaitTimeout(WaitTimeoutError), 47 } 48 49 impl From<SshCommandError> for Error { 50 fn from(e: SshCommandError) -> Self { 51 Self::SshCommand(e) 52 } 53 } 54 55 pub struct GuestNetworkConfig { 56 pub guest_ip: String, 57 pub l2_guest_ip1: String, 58 pub l2_guest_ip2: String, 59 pub l2_guest_ip3: String, 60 pub host_ip: String, 61 pub guest_mac: String, 62 pub l2_guest_mac1: String, 63 pub l2_guest_mac2: String, 64 pub l2_guest_mac3: String, 65 pub tcp_listener_port: u16, 66 } 67 68 pub const DEFAULT_TCP_LISTENER_MESSAGE: &str = "booted"; 69 pub const DEFAULT_TCP_LISTENER_PORT: u16 = 8000; 70 pub const DEFAULT_TCP_LISTENER_TIMEOUT: i32 = 120; 71 72 #[derive(Debug)] 73 pub enum WaitForBootError { 74 EpollWait(std::io::Error), 75 Listen(std::io::Error), 76 EpollWaitTimeout, 77 WrongGuestAddr, 78 Accept(std::io::Error), 79 } 80 81 impl GuestNetworkConfig { 82 pub fn wait_vm_boot(&self, custom_timeout: Option<i32>) -> Result<(), WaitForBootError> { 83 let start = std::time::Instant::now(); 84 // The 'port' is unique per 'GUEST' and listening to wild-card ip avoids retrying on 'TcpListener::bind()' 85 let listen_addr = format!("0.0.0.0:{}", self.tcp_listener_port); 86 let expected_guest_addr = self.guest_ip.as_str(); 87 let mut s = String::new(); 88 let timeout = match custom_timeout { 89 Some(t) => t, 90 None => DEFAULT_TCP_LISTENER_TIMEOUT, 91 }; 92 93 match (|| -> Result<(), WaitForBootError> { 94 let listener = 95 TcpListener::bind(listen_addr.as_str()).map_err(WaitForBootError::Listen)?; 96 listener 97 .set_nonblocking(true) 98 .expect("Cannot set non-blocking for tcp listener"); 99 100 // Reply on epoll w/ timeout to wait for guest connections faithfully 101 let epoll_fd = epoll::create(true).expect("Cannot create epoll fd"); 102 // Use 'File' to enforce closing on 'epoll_fd' 103 let _epoll_file = unsafe { fs::File::from_raw_fd(epoll_fd) }; 104 epoll::ctl( 105 epoll_fd, 106 epoll::ControlOptions::EPOLL_CTL_ADD, 107 listener.as_raw_fd(), 108 epoll::Event::new(epoll::Events::EPOLLIN, 0), 109 ) 110 .expect("Cannot add 'tcp_listener' event to epoll"); 111 let mut events = [epoll::Event::new(epoll::Events::empty(), 0); 1]; 112 loop { 113 let num_events = match epoll::wait(epoll_fd, timeout * 1000_i32, &mut events[..]) { 114 Ok(num_events) => Ok(num_events), 115 Err(e) => match e.raw_os_error() { 116 Some(libc::EAGAIN) | Some(libc::EINTR) => continue, 117 _ => Err(e), 118 }, 119 } 120 .map_err(WaitForBootError::EpollWait)?; 121 if num_events == 0 { 122 return Err(WaitForBootError::EpollWaitTimeout); 123 } 124 break; 125 } 126 127 match listener.accept() { 128 Ok((_, addr)) => { 129 // Make sure the connection is from the expected 'guest_addr' 130 if addr.ip() != std::net::IpAddr::from_str(expected_guest_addr).unwrap() { 131 s = format!( 132 "Expecting the guest ip '{}' while being connected with ip '{}'", 133 expected_guest_addr, 134 addr.ip() 135 ); 136 return Err(WaitForBootError::WrongGuestAddr); 137 } 138 139 Ok(()) 140 } 141 Err(e) => { 142 s = "TcpListener::accept() failed".to_string(); 143 Err(WaitForBootError::Accept(e)) 144 } 145 } 146 })() { 147 Err(e) => { 148 let duration = start.elapsed(); 149 eprintln!( 150 "\n\n==== Start 'wait_vm_boot' (FAILED) ====\n\n\ 151 duration =\"{duration:?}, timeout = {timeout}s\"\n\ 152 listen_addr=\"{listen_addr}\"\n\ 153 expected_guest_addr=\"{expected_guest_addr}\"\n\ 154 message=\"{s}\"\n\ 155 error=\"{e:?}\"\n\ 156 \n==== End 'wait_vm_boot' outout ====\n\n" 157 ); 158 159 Err(e) 160 } 161 Ok(_) => Ok(()), 162 } 163 } 164 } 165 166 pub enum DiskType { 167 OperatingSystem, 168 CloudInit, 169 } 170 171 pub trait DiskConfig { 172 fn prepare_files(&mut self, tmp_dir: &TempDir, network: &GuestNetworkConfig); 173 fn prepare_cloudinit(&self, tmp_dir: &TempDir, network: &GuestNetworkConfig) -> String; 174 fn disk(&self, disk_type: DiskType) -> Option<String>; 175 } 176 177 #[derive(Clone)] 178 pub struct UbuntuDiskConfig { 179 osdisk_path: String, 180 cloudinit_path: String, 181 image_name: String, 182 } 183 184 impl UbuntuDiskConfig { 185 pub fn new(image_name: String) -> Self { 186 UbuntuDiskConfig { 187 image_name, 188 osdisk_path: String::new(), 189 cloudinit_path: String::new(), 190 } 191 } 192 } 193 194 pub struct WindowsDiskConfig { 195 image_name: String, 196 osdisk_path: String, 197 loopback_device: String, 198 windows_snapshot_cow: String, 199 windows_snapshot: String, 200 } 201 202 impl WindowsDiskConfig { 203 pub fn new(image_name: String) -> Self { 204 WindowsDiskConfig { 205 image_name, 206 osdisk_path: String::new(), 207 loopback_device: String::new(), 208 windows_snapshot_cow: String::new(), 209 windows_snapshot: String::new(), 210 } 211 } 212 } 213 214 impl Drop for WindowsDiskConfig { 215 fn drop(&mut self) { 216 // dmsetup remove windows-snapshot-1 217 std::process::Command::new("dmsetup") 218 .arg("remove") 219 .arg(self.windows_snapshot.as_str()) 220 .output() 221 .expect("Expect removing Windows snapshot with 'dmsetup' to succeed"); 222 223 // dmsetup remove windows-snapshot-cow-1 224 std::process::Command::new("dmsetup") 225 .arg("remove") 226 .arg(self.windows_snapshot_cow.as_str()) 227 .output() 228 .expect("Expect removing Windows snapshot CoW with 'dmsetup' to succeed"); 229 230 // losetup -d <loopback_device> 231 std::process::Command::new("losetup") 232 .args(["-d", self.loopback_device.as_str()]) 233 .output() 234 .expect("Expect removing loopback device to succeed"); 235 } 236 } 237 238 impl DiskConfig for UbuntuDiskConfig { 239 fn prepare_cloudinit(&self, tmp_dir: &TempDir, network: &GuestNetworkConfig) -> String { 240 let cloudinit_file_path = 241 String::from(tmp_dir.as_path().join("cloudinit").to_str().unwrap()); 242 243 let cloud_init_directory = tmp_dir.as_path().join("cloud-init").join("ubuntu"); 244 245 fs::create_dir_all(&cloud_init_directory) 246 .expect("Expect creating cloud-init directory to succeed"); 247 248 let source_file_dir = std::env::current_dir() 249 .unwrap() 250 .join("test_data") 251 .join("cloud-init") 252 .join("ubuntu") 253 .join("ci"); 254 255 ["meta-data"].iter().for_each(|x| { 256 rate_limited_copy(source_file_dir.join(x), cloud_init_directory.join(x)) 257 .expect("Expect copying cloud-init meta-data to succeed"); 258 }); 259 260 let mut user_data_string = String::new(); 261 fs::File::open(source_file_dir.join("user-data")) 262 .unwrap() 263 .read_to_string(&mut user_data_string) 264 .expect("Expected reading user-data file in to succeed"); 265 user_data_string = user_data_string.replace( 266 "@DEFAULT_TCP_LISTENER_MESSAGE", 267 DEFAULT_TCP_LISTENER_MESSAGE, 268 ); 269 user_data_string = user_data_string.replace("@HOST_IP", &network.host_ip); 270 user_data_string = 271 user_data_string.replace("@TCP_LISTENER_PORT", &network.tcp_listener_port.to_string()); 272 273 fs::File::create(cloud_init_directory.join("user-data")) 274 .unwrap() 275 .write_all(user_data_string.as_bytes()) 276 .expect("Expected writing out user-data to succeed"); 277 278 let mut network_config_string = String::new(); 279 280 fs::File::open(source_file_dir.join("network-config")) 281 .unwrap() 282 .read_to_string(&mut network_config_string) 283 .expect("Expected reading network-config file in to succeed"); 284 285 network_config_string = network_config_string.replace("192.168.2.1", &network.host_ip); 286 network_config_string = network_config_string.replace("192.168.2.2", &network.guest_ip); 287 network_config_string = network_config_string.replace("192.168.2.3", &network.l2_guest_ip1); 288 network_config_string = network_config_string.replace("192.168.2.4", &network.l2_guest_ip2); 289 network_config_string = network_config_string.replace("192.168.2.5", &network.l2_guest_ip3); 290 network_config_string = 291 network_config_string.replace("12:34:56:78:90:ab", &network.guest_mac); 292 network_config_string = 293 network_config_string.replace("de:ad:be:ef:12:34", &network.l2_guest_mac1); 294 network_config_string = 295 network_config_string.replace("de:ad:be:ef:34:56", &network.l2_guest_mac2); 296 network_config_string = 297 network_config_string.replace("de:ad:be:ef:56:78", &network.l2_guest_mac3); 298 299 fs::File::create(cloud_init_directory.join("network-config")) 300 .unwrap() 301 .write_all(network_config_string.as_bytes()) 302 .expect("Expected writing out network-config to succeed"); 303 304 std::process::Command::new("mkdosfs") 305 .args(["-n", "CIDATA"]) 306 .args(["-C", cloudinit_file_path.as_str()]) 307 .arg("8192") 308 .output() 309 .expect("Expect creating disk image to succeed"); 310 311 ["user-data", "meta-data", "network-config"] 312 .iter() 313 .for_each(|x| { 314 std::process::Command::new("mcopy") 315 .arg("-o") 316 .args(["-i", cloudinit_file_path.as_str()]) 317 .args(["-s", cloud_init_directory.join(x).to_str().unwrap(), "::"]) 318 .output() 319 .expect("Expect copying files to disk image to succeed"); 320 }); 321 322 cloudinit_file_path 323 } 324 325 fn prepare_files(&mut self, tmp_dir: &TempDir, network: &GuestNetworkConfig) { 326 let mut workload_path = dirs::home_dir().unwrap(); 327 workload_path.push("workloads"); 328 329 let mut osdisk_base_path = workload_path; 330 osdisk_base_path.push(&self.image_name); 331 332 let osdisk_path = String::from(tmp_dir.as_path().join("osdisk.img").to_str().unwrap()); 333 let cloudinit_path = self.prepare_cloudinit(tmp_dir, network); 334 335 rate_limited_copy(osdisk_base_path, &osdisk_path) 336 .expect("copying of OS source disk image failed"); 337 338 self.cloudinit_path = cloudinit_path; 339 self.osdisk_path = osdisk_path; 340 } 341 342 fn disk(&self, disk_type: DiskType) -> Option<String> { 343 match disk_type { 344 DiskType::OperatingSystem => Some(self.osdisk_path.clone()), 345 DiskType::CloudInit => Some(self.cloudinit_path.clone()), 346 } 347 } 348 } 349 350 impl DiskConfig for WindowsDiskConfig { 351 fn prepare_cloudinit(&self, _tmp_dir: &TempDir, _network: &GuestNetworkConfig) -> String { 352 String::new() 353 } 354 355 fn prepare_files(&mut self, tmp_dir: &TempDir, _network: &GuestNetworkConfig) { 356 let mut workload_path = dirs::home_dir().unwrap(); 357 workload_path.push("workloads"); 358 359 let mut osdisk_path = workload_path; 360 osdisk_path.push(&self.image_name); 361 362 let osdisk_blk_size = fs::metadata(osdisk_path) 363 .expect("Expect retrieving Windows image metadata") 364 .len() 365 >> 9; 366 367 let snapshot_cow_path = 368 String::from(tmp_dir.as_path().join("snapshot_cow").to_str().unwrap()); 369 370 // Create and truncate CoW file for device mapper 371 let cow_file_size: u64 = 1 << 30; 372 let cow_file_blk_size = cow_file_size >> 9; 373 let cow_file = std::fs::File::create(snapshot_cow_path.as_str()) 374 .expect("Expect creating CoW image to succeed"); 375 cow_file 376 .set_len(cow_file_size) 377 .expect("Expect truncating CoW image to succeed"); 378 379 // losetup --find --show /tmp/snapshot_cow 380 let loopback_device = std::process::Command::new("losetup") 381 .arg("--find") 382 .arg("--show") 383 .arg(snapshot_cow_path.as_str()) 384 .output() 385 .expect("Expect creating loopback device from snapshot CoW image to succeed"); 386 387 self.loopback_device = String::from_utf8_lossy(&loopback_device.stdout) 388 .trim() 389 .to_string(); 390 391 let random_extension = tmp_dir.as_path().file_name().unwrap(); 392 let windows_snapshot_cow = format!( 393 "windows-snapshot-cow-{}", 394 random_extension.to_str().unwrap() 395 ); 396 397 // dmsetup create windows-snapshot-cow-1 --table '0 2097152 linear /dev/loop1 0' 398 std::process::Command::new("dmsetup") 399 .arg("create") 400 .arg(windows_snapshot_cow.as_str()) 401 .args([ 402 "--table", 403 format!("0 {} linear {} 0", cow_file_blk_size, self.loopback_device).as_str(), 404 ]) 405 .output() 406 .expect("Expect creating Windows snapshot CoW with 'dmsetup' to succeed"); 407 408 let windows_snapshot = format!("windows-snapshot-{}", random_extension.to_str().unwrap()); 409 410 // dmsetup mknodes 411 std::process::Command::new("dmsetup") 412 .arg("mknodes") 413 .output() 414 .expect("Expect device mapper nodes to be ready"); 415 416 // dmsetup create windows-snapshot-1 --table '0 41943040 snapshot /dev/mapper/windows-base /dev/mapper/windows-snapshot-cow-1 P 8' 417 std::process::Command::new("dmsetup") 418 .arg("create") 419 .arg(windows_snapshot.as_str()) 420 .args([ 421 "--table", 422 format!( 423 "0 {} snapshot /dev/mapper/windows-base /dev/mapper/{} P 8", 424 osdisk_blk_size, 425 windows_snapshot_cow.as_str() 426 ) 427 .as_str(), 428 ]) 429 .output() 430 .expect("Expect creating Windows snapshot with 'dmsetup' to succeed"); 431 432 // dmsetup mknodes 433 std::process::Command::new("dmsetup") 434 .arg("mknodes") 435 .output() 436 .expect("Expect device mapper nodes to be ready"); 437 438 self.osdisk_path = format!("/dev/mapper/{windows_snapshot}"); 439 self.windows_snapshot_cow = windows_snapshot_cow; 440 self.windows_snapshot = windows_snapshot; 441 } 442 443 fn disk(&self, disk_type: DiskType) -> Option<String> { 444 match disk_type { 445 DiskType::OperatingSystem => Some(self.osdisk_path.clone()), 446 DiskType::CloudInit => None, 447 } 448 } 449 } 450 451 pub fn rate_limited_copy<P: AsRef<Path>, Q: AsRef<Path>>(from: P, to: Q) -> io::Result<u64> { 452 for i in 0..10 { 453 let free_bytes = unsafe { 454 let mut stats = std::mem::MaybeUninit::zeroed(); 455 let fs_name = std::ffi::CString::new("/tmp").unwrap(); 456 libc::statvfs(fs_name.as_ptr(), stats.as_mut_ptr()); 457 458 let free_blocks = stats.assume_init().f_bfree; 459 let block_size = stats.assume_init().f_bsize; 460 461 free_blocks * block_size 462 }; 463 464 // Make sure there is at least 6 GiB of space 465 if free_bytes < 6 << 30 { 466 eprintln!("Not enough space on disk ({free_bytes}). Attempt {i} of 10. Sleeping."); 467 thread::sleep(std::time::Duration::new(60, 0)); 468 continue; 469 } 470 471 match fs::copy(&from, &to) { 472 Err(e) => { 473 if let Some(errno) = e.raw_os_error() { 474 if errno == libc::ENOSPC { 475 eprintln!("Copy returned ENOSPC. Attempt {i} of 10. Sleeping."); 476 thread::sleep(std::time::Duration::new(60, 0)); 477 continue; 478 } 479 } 480 return Err(e); 481 } 482 Ok(i) => return Ok(i), 483 } 484 } 485 Err(io::Error::last_os_error()) 486 } 487 488 pub fn handle_child_output( 489 r: Result<(), std::boxed::Box<dyn std::any::Any + std::marker::Send>>, 490 output: &std::process::Output, 491 ) { 492 use std::os::unix::process::ExitStatusExt; 493 if r.is_ok() && output.status.success() { 494 return; 495 } 496 497 match output.status.code() { 498 None => { 499 // Don't treat child.kill() as a problem 500 if output.status.signal() == Some(9) && r.is_ok() { 501 return; 502 } 503 504 eprintln!( 505 "==== child killed by signal: {} ====", 506 output.status.signal().unwrap() 507 ); 508 } 509 Some(code) => { 510 eprintln!("\n\n==== child exit code: {code} ===="); 511 } 512 } 513 514 eprintln!( 515 "\n\n==== Start child stdout ====\n\n{}\n\n==== End child stdout ====", 516 String::from_utf8_lossy(&output.stdout) 517 ); 518 eprintln!( 519 "\n\n==== Start child stderr ====\n\n{}\n\n==== End child stderr ====", 520 String::from_utf8_lossy(&output.stderr) 521 ); 522 523 panic!("Test failed") 524 } 525 526 #[derive(Debug)] 527 pub struct PasswordAuth { 528 pub username: String, 529 pub password: String, 530 } 531 532 pub const DEFAULT_SSH_RETRIES: u8 = 6; 533 pub const DEFAULT_SSH_TIMEOUT: u8 = 10; 534 535 #[derive(Debug)] 536 pub enum SshCommandError { 537 Connection(std::io::Error), 538 Handshake(ssh2::Error), 539 Authentication(ssh2::Error), 540 ChannelSession(ssh2::Error), 541 Command(ssh2::Error), 542 ExitStatus(ssh2::Error), 543 NonZeroExitStatus(i32), 544 FileRead(std::io::Error), 545 FileMetadata(std::io::Error), 546 ScpSend(ssh2::Error), 547 WriteAll(std::io::Error), 548 SendEof(ssh2::Error), 549 WaitEof(ssh2::Error), 550 } 551 552 fn scp_to_guest_with_auth( 553 path: &Path, 554 remote_path: &Path, 555 auth: &PasswordAuth, 556 ip: &str, 557 retries: u8, 558 timeout: u8, 559 ) -> Result<(), SshCommandError> { 560 let mut counter = 0; 561 loop { 562 match (|| -> Result<(), SshCommandError> { 563 let tcp = 564 TcpStream::connect(format!("{ip}:22")).map_err(SshCommandError::Connection)?; 565 let mut sess = Session::new().unwrap(); 566 sess.set_tcp_stream(tcp); 567 sess.handshake().map_err(SshCommandError::Handshake)?; 568 569 sess.userauth_password(&auth.username, &auth.password) 570 .map_err(SshCommandError::Authentication)?; 571 assert!(sess.authenticated()); 572 573 let content = fs::read(path).map_err(SshCommandError::FileRead)?; 574 let mode = fs::metadata(path) 575 .map_err(SshCommandError::FileMetadata)? 576 .permissions() 577 .mode() 578 & 0o777; 579 580 let mut channel = sess 581 .scp_send(remote_path, mode as i32, content.len() as u64, None) 582 .map_err(SshCommandError::ScpSend)?; 583 channel 584 .write_all(&content) 585 .map_err(SshCommandError::WriteAll)?; 586 channel.send_eof().map_err(SshCommandError::SendEof)?; 587 channel.wait_eof().map_err(SshCommandError::WaitEof)?; 588 589 // Intentionally ignore these results here as their failure 590 // does not precipitate a repeat 591 let _ = channel.close(); 592 let _ = channel.wait_close(); 593 594 Ok(()) 595 })() { 596 Ok(_) => break, 597 Err(e) => { 598 counter += 1; 599 if counter >= retries { 600 eprintln!( 601 "\n\n==== Start scp command output (FAILED) ====\n\n\ 602 path =\"{path:?}\"\n\ 603 remote_path =\"{remote_path:?}\"\n\ 604 auth=\"{auth:#?}\"\n\ 605 ip=\"{ip}\"\n\ 606 error=\"{e:?}\"\n\ 607 \n==== End scp command outout ====\n\n" 608 ); 609 610 return Err(e); 611 } 612 } 613 }; 614 thread::sleep(std::time::Duration::new((timeout * counter).into(), 0)); 615 } 616 Ok(()) 617 } 618 619 pub fn scp_to_guest( 620 path: &Path, 621 remote_path: &Path, 622 ip: &str, 623 retries: u8, 624 timeout: u8, 625 ) -> Result<(), SshCommandError> { 626 scp_to_guest_with_auth( 627 path, 628 remote_path, 629 &PasswordAuth { 630 username: String::from("cloud"), 631 password: String::from("cloud123"), 632 }, 633 ip, 634 retries, 635 timeout, 636 ) 637 } 638 639 pub fn ssh_command_ip_with_auth( 640 command: &str, 641 auth: &PasswordAuth, 642 ip: &str, 643 retries: u8, 644 timeout: u8, 645 ) -> Result<String, SshCommandError> { 646 let mut s = String::new(); 647 648 let mut counter = 0; 649 loop { 650 match (|| -> Result<(), SshCommandError> { 651 let tcp = 652 TcpStream::connect(format!("{ip}:22")).map_err(SshCommandError::Connection)?; 653 let mut sess = Session::new().unwrap(); 654 sess.set_tcp_stream(tcp); 655 sess.handshake().map_err(SshCommandError::Handshake)?; 656 657 sess.userauth_password(&auth.username, &auth.password) 658 .map_err(SshCommandError::Authentication)?; 659 assert!(sess.authenticated()); 660 661 let mut channel = sess 662 .channel_session() 663 .map_err(SshCommandError::ChannelSession)?; 664 channel.exec(command).map_err(SshCommandError::Command)?; 665 666 // Intentionally ignore these results here as their failure 667 // does not precipitate a repeat 668 let _ = channel.read_to_string(&mut s); 669 let _ = channel.close(); 670 let _ = channel.wait_close(); 671 672 let status = channel.exit_status().map_err(SshCommandError::ExitStatus)?; 673 674 if status != 0 { 675 Err(SshCommandError::NonZeroExitStatus(status)) 676 } else { 677 Ok(()) 678 } 679 })() { 680 Ok(_) => break, 681 Err(e) => { 682 counter += 1; 683 if counter >= retries { 684 eprintln!( 685 "\n\n==== Start ssh command output (FAILED) ====\n\n\ 686 command=\"{command}\"\n\ 687 auth=\"{auth:#?}\"\n\ 688 ip=\"{ip}\"\n\ 689 output=\"{s}\"\n\ 690 error=\"{e:?}\"\n\ 691 \n==== End ssh command outout ====\n\n" 692 ); 693 694 return Err(e); 695 } 696 } 697 }; 698 thread::sleep(std::time::Duration::new((timeout * counter).into(), 0)); 699 } 700 Ok(s) 701 } 702 703 pub fn ssh_command_ip( 704 command: &str, 705 ip: &str, 706 retries: u8, 707 timeout: u8, 708 ) -> Result<String, SshCommandError> { 709 ssh_command_ip_with_auth( 710 command, 711 &PasswordAuth { 712 username: String::from("cloud"), 713 password: String::from("cloud123"), 714 }, 715 ip, 716 retries, 717 timeout, 718 ) 719 } 720 721 pub fn exec_host_command_status(command: &str) -> ExitStatus { 722 std::process::Command::new("bash") 723 .args(["-c", command]) 724 .status() 725 .unwrap_or_else(|_| panic!("Expected '{command}' to run")) 726 } 727 728 pub fn exec_host_command_output(command: &str) -> Output { 729 std::process::Command::new("bash") 730 .args(["-c", command]) 731 .output() 732 .unwrap_or_else(|_| panic!("Expected '{command}' to run")) 733 } 734 735 pub const PIPE_SIZE: i32 = 32 << 20; 736 737 static NEXT_VM_ID: Lazy<Mutex<u8>> = Lazy::new(|| Mutex::new(1)); 738 739 pub struct Guest { 740 pub tmp_dir: TempDir, 741 pub disk_config: Box<dyn DiskConfig>, 742 pub network: GuestNetworkConfig, 743 } 744 745 // Safe to implement as we know we have no interior mutability 746 impl std::panic::RefUnwindSafe for Guest {} 747 748 impl Guest { 749 pub fn new_from_ip_range(mut disk_config: Box<dyn DiskConfig>, class: &str, id: u8) -> Self { 750 let tmp_dir = TempDir::new_with_prefix("/tmp/ch").unwrap(); 751 752 let network = GuestNetworkConfig { 753 guest_ip: format!("{class}.{id}.2"), 754 l2_guest_ip1: format!("{class}.{id}.3"), 755 l2_guest_ip2: format!("{class}.{id}.4"), 756 l2_guest_ip3: format!("{class}.{id}.5"), 757 host_ip: format!("{class}.{id}.1"), 758 guest_mac: format!("12:34:56:78:90:{id:02x}"), 759 l2_guest_mac1: format!("de:ad:be:ef:12:{id:02x}"), 760 l2_guest_mac2: format!("de:ad:be:ef:34:{id:02x}"), 761 l2_guest_mac3: format!("de:ad:be:ef:56:{id:02x}"), 762 tcp_listener_port: DEFAULT_TCP_LISTENER_PORT + id as u16, 763 }; 764 765 disk_config.prepare_files(&tmp_dir, &network); 766 767 Guest { 768 tmp_dir, 769 disk_config, 770 network, 771 } 772 } 773 774 pub fn new(disk_config: Box<dyn DiskConfig>) -> Self { 775 let mut guard = NEXT_VM_ID.lock().unwrap(); 776 let id = *guard; 777 *guard = id + 1; 778 779 Self::new_from_ip_range(disk_config, "192.168", id) 780 } 781 782 pub fn default_net_string(&self) -> String { 783 format!( 784 "tap=,mac={},ip={},mask=255.255.255.0", 785 self.network.guest_mac, self.network.host_ip 786 ) 787 } 788 789 pub fn default_net_string_w_iommu(&self) -> String { 790 format!( 791 "tap=,mac={},ip={},mask=255.255.255.0,iommu=on", 792 self.network.guest_mac, self.network.host_ip 793 ) 794 } 795 796 pub fn default_net_string_w_mtu(&self, mtu: u16) -> String { 797 format!( 798 "tap=,mac={},ip={},mask=255.255.255.0,mtu={}", 799 self.network.guest_mac, self.network.host_ip, mtu 800 ) 801 } 802 803 pub fn ssh_command(&self, command: &str) -> Result<String, SshCommandError> { 804 ssh_command_ip( 805 command, 806 &self.network.guest_ip, 807 DEFAULT_SSH_RETRIES, 808 DEFAULT_SSH_TIMEOUT, 809 ) 810 } 811 812 #[cfg(target_arch = "x86_64")] 813 pub fn ssh_command_l1(&self, command: &str) -> Result<String, SshCommandError> { 814 ssh_command_ip( 815 command, 816 &self.network.guest_ip, 817 DEFAULT_SSH_RETRIES, 818 DEFAULT_SSH_TIMEOUT, 819 ) 820 } 821 822 #[cfg(target_arch = "x86_64")] 823 pub fn ssh_command_l2_1(&self, command: &str) -> Result<String, SshCommandError> { 824 ssh_command_ip( 825 command, 826 &self.network.l2_guest_ip1, 827 DEFAULT_SSH_RETRIES, 828 DEFAULT_SSH_TIMEOUT, 829 ) 830 } 831 832 #[cfg(target_arch = "x86_64")] 833 pub fn ssh_command_l2_2(&self, command: &str) -> Result<String, SshCommandError> { 834 ssh_command_ip( 835 command, 836 &self.network.l2_guest_ip2, 837 DEFAULT_SSH_RETRIES, 838 DEFAULT_SSH_TIMEOUT, 839 ) 840 } 841 842 #[cfg(target_arch = "x86_64")] 843 pub fn ssh_command_l2_3(&self, command: &str) -> Result<String, SshCommandError> { 844 ssh_command_ip( 845 command, 846 &self.network.l2_guest_ip3, 847 DEFAULT_SSH_RETRIES, 848 DEFAULT_SSH_TIMEOUT, 849 ) 850 } 851 852 pub fn api_create_body(&self, cpu_count: u8, kernel_path: &str, kernel_cmd: &str) -> String { 853 format! {"{{\"cpus\":{{\"boot_vcpus\":{},\"max_vcpus\":{}}},\"payload\":{{\"kernel\":\"{}\",\"cmdline\": \"{}\"}},\"net\":[{{\"ip\":\"{}\", \"mask\":\"255.255.255.0\", \"mac\":\"{}\"}}], \"disks\":[{{\"path\":\"{}\"}}, {{\"path\":\"{}\"}}]}}", 854 cpu_count, 855 cpu_count, 856 kernel_path, 857 kernel_cmd, 858 self.network.host_ip, 859 self.network.guest_mac, 860 self.disk_config.disk(DiskType::OperatingSystem).unwrap().as_str(), 861 self.disk_config.disk(DiskType::CloudInit).unwrap().as_str(), 862 } 863 } 864 865 pub fn get_cpu_count(&self) -> Result<u32, Error> { 866 self.ssh_command("grep -c processor /proc/cpuinfo")? 867 .trim() 868 .parse() 869 .map_err(Error::Parsing) 870 } 871 872 pub fn get_total_memory(&self) -> Result<u32, Error> { 873 self.ssh_command("grep MemTotal /proc/meminfo | grep -o \"[0-9]*\"")? 874 .trim() 875 .parse() 876 .map_err(Error::Parsing) 877 } 878 879 #[cfg(target_arch = "x86_64")] 880 pub fn get_total_memory_l2(&self) -> Result<u32, Error> { 881 self.ssh_command_l2_1("grep MemTotal /proc/meminfo | grep -o \"[0-9]*\"")? 882 .trim() 883 .parse() 884 .map_err(Error::Parsing) 885 } 886 887 pub fn get_numa_node_memory(&self, node_id: usize) -> Result<u32, Error> { 888 self.ssh_command( 889 format!( 890 "grep MemTotal /sys/devices/system/node/node{node_id}/meminfo \ 891 | cut -d \":\" -f 2 | grep -o \"[0-9]*\"" 892 ) 893 .as_str(), 894 )? 895 .trim() 896 .parse() 897 .map_err(Error::Parsing) 898 } 899 900 pub fn wait_vm_boot(&self, custom_timeout: Option<i32>) -> Result<(), Error> { 901 self.network 902 .wait_vm_boot(custom_timeout) 903 .map_err(Error::WaitForBoot) 904 } 905 906 pub fn check_numa_node_cpus(&self, node_id: usize, cpus: Vec<usize>) -> Result<(), Error> { 907 for cpu in cpus.iter() { 908 let cmd = format!("[ -d \"/sys/devices/system/node/node{node_id}/cpu{cpu}\" ]"); 909 self.ssh_command(cmd.as_str())?; 910 } 911 912 Ok(()) 913 } 914 915 pub fn check_numa_node_distances( 916 &self, 917 node_id: usize, 918 distances: &str, 919 ) -> Result<bool, Error> { 920 let cmd = format!("cat /sys/devices/system/node/node{node_id}/distance"); 921 if self.ssh_command(cmd.as_str())?.trim() == distances { 922 Ok(true) 923 } else { 924 Ok(false) 925 } 926 } 927 928 pub fn check_numa_common( 929 &self, 930 mem_ref: Option<&[u32]>, 931 node_ref: Option<&[Vec<usize>]>, 932 distance_ref: Option<&[&str]>, 933 ) { 934 if let Some(mem_ref) = mem_ref { 935 // Check each NUMA node has been assigned the right amount of 936 // memory. 937 for (i, &m) in mem_ref.iter().enumerate() { 938 assert!(self.get_numa_node_memory(i).unwrap_or_default() > m); 939 } 940 } 941 942 if let Some(node_ref) = node_ref { 943 // Check each NUMA node has been assigned the right CPUs set. 944 for (i, n) in node_ref.iter().enumerate() { 945 self.check_numa_node_cpus(i, n.clone()).unwrap(); 946 } 947 } 948 949 if let Some(distance_ref) = distance_ref { 950 // Check each NUMA node has been assigned the right distances. 951 for (i, &d) in distance_ref.iter().enumerate() { 952 assert!(self.check_numa_node_distances(i, d).unwrap()); 953 } 954 } 955 } 956 957 #[cfg(target_arch = "x86_64")] 958 pub fn check_sgx_support(&self) -> Result<(), Error> { 959 self.ssh_command( 960 "cpuid -l 0x7 -s 0 | tr -s [:space:] | grep -q 'SGX: \ 961 Software Guard Extensions supported = true'", 962 )?; 963 self.ssh_command( 964 "cpuid -l 0x7 -s 0 | tr -s [:space:] | grep -q 'SGX_LC: \ 965 SGX launch config supported = true'", 966 )?; 967 self.ssh_command( 968 "cpuid -l 0x12 -s 0 | tr -s [:space:] | grep -q 'SGX1 \ 969 supported = true'", 970 )?; 971 972 Ok(()) 973 } 974 975 pub fn get_pci_bridge_class(&self) -> Result<String, Error> { 976 Ok(self 977 .ssh_command("cat /sys/bus/pci/devices/0000:00:00.0/class")? 978 .trim() 979 .to_string()) 980 } 981 982 pub fn get_pci_device_ids(&self) -> Result<String, Error> { 983 Ok(self 984 .ssh_command("cat /sys/bus/pci/devices/*/device")? 985 .trim() 986 .to_string()) 987 } 988 989 pub fn get_pci_vendor_ids(&self) -> Result<String, Error> { 990 Ok(self 991 .ssh_command("cat /sys/bus/pci/devices/*/vendor")? 992 .trim() 993 .to_string()) 994 } 995 996 pub fn does_device_vendor_pair_match( 997 &self, 998 device_id: &str, 999 vendor_id: &str, 1000 ) -> Result<bool, Error> { 1001 // We are checking if console device's device id and vendor id pair matches 1002 let devices = self.get_pci_device_ids()?; 1003 let devices: Vec<&str> = devices.split('\n').collect(); 1004 let vendors = self.get_pci_vendor_ids()?; 1005 let vendors: Vec<&str> = vendors.split('\n').collect(); 1006 1007 for (index, d_id) in devices.iter().enumerate() { 1008 if *d_id == device_id { 1009 if let Some(v_id) = vendors.get(index) { 1010 if *v_id == vendor_id { 1011 return Ok(true); 1012 } 1013 } 1014 } 1015 } 1016 1017 Ok(false) 1018 } 1019 1020 pub fn check_vsock(&self, socket: &str) { 1021 // Listen from guest on vsock CID=3 PORT=16 1022 // SOCKET-LISTEN:<domain>:<protocol>:<local-address> 1023 let guest_ip = self.network.guest_ip.clone(); 1024 let listen_socat = thread::spawn(move || { 1025 ssh_command_ip("sudo socat - SOCKET-LISTEN:40:0:x00x00x10x00x00x00x03x00x00x00x00x00x00x00 > vsock_log", &guest_ip, DEFAULT_SSH_RETRIES, DEFAULT_SSH_TIMEOUT).unwrap(); 1026 }); 1027 1028 // Make sure socat is listening, which might take a few second on slow systems 1029 thread::sleep(std::time::Duration::new(10, 0)); 1030 1031 // Write something to vsock from the host 1032 assert!(exec_host_command_status(&format!( 1033 "echo -e \"CONNECT 16\\nHelloWorld!\" | socat - UNIX-CONNECT:{socket}" 1034 )) 1035 .success()); 1036 1037 // Wait for the thread to terminate. 1038 listen_socat.join().unwrap(); 1039 1040 assert_eq!( 1041 self.ssh_command("cat vsock_log").unwrap().trim(), 1042 "HelloWorld!" 1043 ); 1044 } 1045 1046 #[cfg(target_arch = "x86_64")] 1047 pub fn check_nvidia_gpu(&self) { 1048 assert!(self.ssh_command("nvidia-smi").unwrap().contains("Tesla T4")); 1049 } 1050 1051 pub fn reboot_linux(&self, current_reboot_count: u32, custom_timeout: Option<i32>) { 1052 let list_boots_cmd = "sudo last | grep -c reboot"; 1053 let boot_count = self 1054 .ssh_command(list_boots_cmd) 1055 .unwrap() 1056 .trim() 1057 .parse::<u32>() 1058 .unwrap_or_default(); 1059 1060 assert_eq!(boot_count, current_reboot_count + 1); 1061 self.ssh_command("sudo reboot").unwrap(); 1062 1063 self.wait_vm_boot(custom_timeout).unwrap(); 1064 let boot_count = self 1065 .ssh_command(list_boots_cmd) 1066 .unwrap() 1067 .trim() 1068 .parse::<u32>() 1069 .unwrap_or_default(); 1070 assert_eq!(boot_count, current_reboot_count + 2); 1071 } 1072 1073 pub fn enable_memory_hotplug(&self) { 1074 self.ssh_command("echo online | sudo tee /sys/devices/system/memory/auto_online_blocks") 1075 .unwrap(); 1076 } 1077 1078 pub fn check_devices_common( 1079 &self, 1080 socket: Option<&String>, 1081 console_text: Option<&String>, 1082 pmem_path: Option<&String>, 1083 ) { 1084 // Check block devices are readable 1085 self.ssh_command("sudo dd if=/dev/vda of=/dev/null bs=1M iflag=direct count=1024") 1086 .unwrap(); 1087 self.ssh_command("sudo dd if=/dev/vdb of=/dev/null bs=1M iflag=direct count=8") 1088 .unwrap(); 1089 // Check if the rng device is readable 1090 self.ssh_command("sudo head -c 1000 /dev/hwrng > /dev/null") 1091 .unwrap(); 1092 // Check vsock 1093 if let Some(socket) = socket { 1094 self.check_vsock(socket.as_str()); 1095 } 1096 // Check if the console is usable 1097 if let Some(console_text) = console_text { 1098 let console_cmd = format!("echo {console_text} | sudo tee /dev/hvc0"); 1099 self.ssh_command(&console_cmd).unwrap(); 1100 } 1101 // The net device is 'automatically' exercised through the above 'ssh' commands 1102 1103 // Check if the pmem device is usable 1104 if let Some(pmem_path) = pmem_path { 1105 assert_eq!( 1106 self.ssh_command(&format!("ls {pmem_path}")).unwrap().trim(), 1107 pmem_path 1108 ); 1109 assert_eq!( 1110 self.ssh_command(&format!("sudo mount {pmem_path} /mnt")) 1111 .unwrap(), 1112 "" 1113 ); 1114 assert_eq!(self.ssh_command("ls /mnt").unwrap(), "lost+found\n"); 1115 self.ssh_command("echo test123 | sudo tee /mnt/test") 1116 .unwrap(); 1117 assert_eq!(self.ssh_command("sudo umount /mnt").unwrap(), ""); 1118 assert_eq!(self.ssh_command("ls /mnt").unwrap(), ""); 1119 1120 assert_eq!( 1121 self.ssh_command(&format!("sudo mount {pmem_path} /mnt")) 1122 .unwrap(), 1123 "" 1124 ); 1125 assert_eq!( 1126 self.ssh_command("sudo cat /mnt/test || true") 1127 .unwrap() 1128 .trim(), 1129 "test123" 1130 ); 1131 self.ssh_command("sudo rm /mnt/test").unwrap(); 1132 assert_eq!(self.ssh_command("sudo umount /mnt").unwrap(), ""); 1133 } 1134 } 1135 } 1136 1137 pub enum VerbosityLevel { 1138 Warn, 1139 Info, 1140 Debug, 1141 } 1142 1143 impl Default for VerbosityLevel { 1144 fn default() -> Self { 1145 Self::Warn 1146 } 1147 } 1148 1149 impl ToString for VerbosityLevel { 1150 fn to_string(&self) -> String { 1151 use VerbosityLevel::*; 1152 match self { 1153 Warn => "".to_string(), 1154 Info => "-v".to_string(), 1155 Debug => "-v -v".to_string(), 1156 } 1157 } 1158 } 1159 1160 pub struct GuestCommand<'a> { 1161 command: Command, 1162 guest: &'a Guest, 1163 capture_output: bool, 1164 print_cmd: bool, 1165 verbosity: VerbosityLevel, 1166 } 1167 1168 impl<'a> GuestCommand<'a> { 1169 pub fn new(guest: &'a Guest) -> Self { 1170 Self::new_with_binary_path(guest, &clh_command("cloud-hypervisor")) 1171 } 1172 1173 pub fn new_with_binary_path(guest: &'a Guest, binary_path: &str) -> Self { 1174 Self { 1175 command: Command::new(binary_path), 1176 guest, 1177 capture_output: false, 1178 print_cmd: true, 1179 verbosity: VerbosityLevel::Info, 1180 } 1181 } 1182 1183 pub fn verbosity(&mut self, verbosity: VerbosityLevel) -> &mut Self { 1184 self.verbosity = verbosity; 1185 self 1186 } 1187 1188 pub fn capture_output(&mut self) -> &mut Self { 1189 self.capture_output = true; 1190 self 1191 } 1192 1193 pub fn set_print_cmd(&mut self, print_cmd: bool) -> &mut Self { 1194 self.print_cmd = print_cmd; 1195 self 1196 } 1197 1198 pub fn spawn(&mut self) -> io::Result<Child> { 1199 use VerbosityLevel::*; 1200 match &self.verbosity { 1201 Warn => {} 1202 Info => { 1203 self.command.arg("-v"); 1204 } 1205 Debug => { 1206 self.command.args(["-v", "-v"]); 1207 } 1208 }; 1209 1210 if self.print_cmd { 1211 println!( 1212 "\n\n==== Start cloud-hypervisor command-line ====\n\n\ 1213 {:?}\n\ 1214 \n==== End cloud-hypervisor command-line ====\n\n", 1215 self.command 1216 ); 1217 } 1218 1219 if self.capture_output { 1220 let child = self 1221 .command 1222 .stderr(Stdio::piped()) 1223 .stdout(Stdio::piped()) 1224 .spawn() 1225 .unwrap(); 1226 1227 let fd = child.stdout.as_ref().unwrap().as_raw_fd(); 1228 let pipesize = unsafe { libc::fcntl(fd, libc::F_SETPIPE_SZ, PIPE_SIZE) }; 1229 if pipesize == -1 { 1230 return Err(io::Error::last_os_error()); 1231 } 1232 let fd = child.stderr.as_ref().unwrap().as_raw_fd(); 1233 let pipesize1 = unsafe { libc::fcntl(fd, libc::F_SETPIPE_SZ, PIPE_SIZE) }; 1234 if pipesize1 == -1 { 1235 return Err(io::Error::last_os_error()); 1236 } 1237 1238 if pipesize >= PIPE_SIZE && pipesize1 >= PIPE_SIZE { 1239 Ok(child) 1240 } else { 1241 Err(std::io::Error::new( 1242 std::io::ErrorKind::Other, 1243 format!( 1244 "resizing pipe w/ 'fnctl' failed: stdout pipesize {pipesize}, stderr pipesize {pipesize1}" 1245 ), 1246 )) 1247 } 1248 } else { 1249 self.command.spawn() 1250 } 1251 } 1252 1253 pub fn args<I, S>(&mut self, args: I) -> &mut Self 1254 where 1255 I: IntoIterator<Item = S>, 1256 S: AsRef<OsStr>, 1257 { 1258 self.command.args(args); 1259 self 1260 } 1261 1262 pub fn default_disks(&mut self) -> &mut Self { 1263 if self.guest.disk_config.disk(DiskType::CloudInit).is_some() { 1264 self.args([ 1265 "--disk", 1266 format!( 1267 "path={}", 1268 self.guest 1269 .disk_config 1270 .disk(DiskType::OperatingSystem) 1271 .unwrap() 1272 ) 1273 .as_str(), 1274 "--disk", 1275 format!( 1276 "path={}", 1277 self.guest.disk_config.disk(DiskType::CloudInit).unwrap() 1278 ) 1279 .as_str(), 1280 ]) 1281 } else { 1282 self.args([ 1283 "--disk", 1284 format!( 1285 "path={}", 1286 self.guest 1287 .disk_config 1288 .disk(DiskType::OperatingSystem) 1289 .unwrap() 1290 ) 1291 .as_str(), 1292 ]) 1293 } 1294 } 1295 1296 pub fn default_net(&mut self) -> &mut Self { 1297 self.args(["--net", self.guest.default_net_string().as_str()]) 1298 } 1299 } 1300 1301 pub fn clh_command(cmd: &str) -> String { 1302 env::var("BUILD_TARGET").map_or( 1303 format!("target/x86_64-unknown-linux-gnu/release/{cmd}"), 1304 |target| format!("target/{target}/release/{cmd}"), 1305 ) 1306 } 1307 1308 pub fn parse_iperf3_output(output: &[u8], sender: bool, bandwidth: bool) -> Result<f64, Error> { 1309 std::panic::catch_unwind(|| { 1310 let s = String::from_utf8_lossy(output); 1311 let v: Value = serde_json::from_str(&s).expect("'iperf3' parse error: invalid json output"); 1312 1313 if bandwidth { 1314 if sender { 1315 v["end"]["sum_sent"]["bits_per_second"] 1316 .as_f64() 1317 .expect("'iperf3' parse error: missing entry 'end.sum_sent.bits_per_second'") 1318 } else { 1319 v["end"]["sum_received"]["bits_per_second"].as_f64().expect( 1320 "'iperf3' parse error: missing entry 'end.sum_received.bits_per_second'", 1321 ) 1322 } 1323 } else { 1324 // iperf does not distinguish sent vs received in this case. 1325 1326 let lost_packets = v["end"]["sum"]["lost_packets"] 1327 .as_f64() 1328 .expect("'iperf3' parse error: missing entry 'end.sum.lost_packets'"); 1329 let packets = v["end"]["sum"]["packets"] 1330 .as_f64() 1331 .expect("'iperf3' parse error: missing entry 'end.sum.packets'"); 1332 let seconds = v["end"]["sum"]["seconds"] 1333 .as_f64() 1334 .expect("'iperf3' parse error: missing entry 'end.sum.seconds'"); 1335 1336 (packets - lost_packets) / seconds 1337 } 1338 }) 1339 .map_err(|_| { 1340 eprintln!( 1341 "=============== iperf3 output ===============\n\n{}\n\n===========end============\n\n", 1342 String::from_utf8_lossy(output) 1343 ); 1344 Error::Iperf3Parse 1345 }) 1346 } 1347 1348 #[derive(Clone)] 1349 pub enum FioOps { 1350 Read, 1351 RandomRead, 1352 Write, 1353 RandomWrite, 1354 ReadWrite, 1355 RandRW, 1356 } 1357 1358 impl fmt::Display for FioOps { 1359 fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { 1360 match self { 1361 FioOps::Read => write!(f, "read"), 1362 FioOps::RandomRead => write!(f, "randread"), 1363 FioOps::Write => write!(f, "write"), 1364 FioOps::RandomWrite => write!(f, "randwrite"), 1365 FioOps::ReadWrite => write!(f, "rw"), 1366 FioOps::RandRW => write!(f, "randrw"), 1367 } 1368 } 1369 } 1370 1371 pub fn parse_fio_output(output: &str, fio_ops: &FioOps, num_jobs: u32) -> Result<f64, Error> { 1372 std::panic::catch_unwind(|| { 1373 let v: Value = 1374 serde_json::from_str(output).expect("'fio' parse error: invalid json output"); 1375 let jobs = v["jobs"] 1376 .as_array() 1377 .expect("'fio' parse error: missing entry 'jobs'"); 1378 assert_eq!( 1379 jobs.len(), 1380 num_jobs as usize, 1381 "'fio' parse error: Unexpected number of 'fio' jobs." 1382 ); 1383 1384 let (read, write) = match fio_ops { 1385 FioOps::Read | FioOps::RandomRead => (true, false), 1386 FioOps::Write | FioOps::RandomWrite => (false, true), 1387 FioOps::ReadWrite | FioOps::RandRW => (true, true), 1388 }; 1389 1390 let mut total_bps = 0_f64; 1391 for j in jobs { 1392 if read { 1393 let bytes = j["read"]["io_bytes"] 1394 .as_u64() 1395 .expect("'fio' parse error: missing entry 'read.io_bytes'"); 1396 let runtime = j["read"]["runtime"] 1397 .as_u64() 1398 .expect("'fio' parse error: missing entry 'read.runtime'") 1399 as f64 1400 / 1000_f64; 1401 total_bps += bytes as f64 / runtime; 1402 } 1403 if write { 1404 let bytes = j["write"]["io_bytes"] 1405 .as_u64() 1406 .expect("'fio' parse error: missing entry 'write.io_bytes'"); 1407 let runtime = j["write"]["runtime"] 1408 .as_u64() 1409 .expect("'fio' parse error: missing entry 'write.runtime'") 1410 as f64 1411 / 1000_f64; 1412 total_bps += bytes as f64 / runtime; 1413 } 1414 } 1415 1416 total_bps 1417 }) 1418 .map_err(|_| { 1419 eprintln!( 1420 "=============== Fio output ===============\n\n{output}\n\n===========end============\n\n" 1421 ); 1422 Error::FioOutputParse 1423 }) 1424 } 1425 1426 pub fn parse_fio_output_iops(output: &str, fio_ops: &FioOps, num_jobs: u32) -> Result<f64, Error> { 1427 std::panic::catch_unwind(|| { 1428 let v: Value = 1429 serde_json::from_str(output).expect("'fio' parse error: invalid json output"); 1430 let jobs = v["jobs"] 1431 .as_array() 1432 .expect("'fio' parse error: missing entry 'jobs'"); 1433 assert_eq!( 1434 jobs.len(), 1435 num_jobs as usize, 1436 "'fio' parse error: Unexpected number of 'fio' jobs." 1437 ); 1438 1439 let (read, write) = match fio_ops { 1440 FioOps::Read | FioOps::RandomRead => (true, false), 1441 FioOps::Write | FioOps::RandomWrite => (false, true), 1442 FioOps::ReadWrite | FioOps::RandRW => (true, true), 1443 }; 1444 1445 let mut total_iops = 0_f64; 1446 for j in jobs { 1447 if read { 1448 let ios = j["read"]["total_ios"] 1449 .as_u64() 1450 .expect("'fio' parse error: missing entry 'read.total_ios'"); 1451 let runtime = j["read"]["runtime"] 1452 .as_u64() 1453 .expect("'fio' parse error: missing entry 'read.runtime'") 1454 as f64 1455 / 1000_f64; 1456 total_iops += ios as f64 / runtime; 1457 } 1458 if write { 1459 let ios = j["write"]["total_ios"] 1460 .as_u64() 1461 .expect("'fio' parse error: missing entry 'write.total_ios'"); 1462 let runtime = j["write"]["runtime"] 1463 .as_u64() 1464 .expect("'fio' parse error: missing entry 'write.runtime'") 1465 as f64 1466 / 1000_f64; 1467 total_iops += ios as f64 / runtime; 1468 } 1469 } 1470 1471 total_iops 1472 }) 1473 .map_err(|_| { 1474 eprintln!( 1475 "=============== Fio output ===============\n\n{output}\n\n===========end============\n\n" 1476 ); 1477 Error::FioOutputParse 1478 }) 1479 } 1480 1481 // Wait the child process for a given timeout 1482 fn child_wait_timeout(child: &mut Child, timeout: u64) -> Result<(), WaitTimeoutError> { 1483 match child.wait_timeout(Duration::from_secs(timeout)) { 1484 Err(e) => { 1485 return Err(WaitTimeoutError::General(e)); 1486 } 1487 Ok(s) => match s { 1488 None => { 1489 return Err(WaitTimeoutError::Timedout); 1490 } 1491 Some(s) => { 1492 if !s.success() { 1493 return Err(WaitTimeoutError::ExitStatus); 1494 } 1495 } 1496 }, 1497 } 1498 1499 Ok(()) 1500 } 1501 1502 pub fn measure_virtio_net_throughput( 1503 test_timeout: u32, 1504 queue_pairs: u32, 1505 guest: &Guest, 1506 receive: bool, 1507 bandwidth: bool, 1508 ) -> Result<f64, Error> { 1509 let default_port = 5201; 1510 1511 // 1. start the iperf3 server on the guest 1512 for n in 0..queue_pairs { 1513 guest.ssh_command(&format!("iperf3 -s -p {} -D", default_port + n))?; 1514 } 1515 1516 thread::sleep(Duration::new(1, 0)); 1517 1518 // 2. start the iperf3 client on host to measure RX through-put 1519 let mut clients = Vec::new(); 1520 for n in 0..queue_pairs { 1521 let mut cmd = Command::new("iperf3"); 1522 cmd.args([ 1523 "-J", // Output in JSON format 1524 "-c", 1525 &guest.network.guest_ip, 1526 "-p", 1527 &format!("{}", default_port + n), 1528 "-t", 1529 &format!("{test_timeout}"), 1530 "-i", 1531 "0", 1532 ]); 1533 // For measuring the guest transmit throughput (as a sender), 1534 // use reverse mode of the iperf3 client on the host 1535 if !receive { 1536 cmd.args(["-R"]); 1537 } 1538 // Use UDP stream to measure packets per second. The bitrate is set to 1539 // 1T to make sure it saturates the link. 1540 if !bandwidth { 1541 cmd.args(["-u", "-b", "1T"]); 1542 } 1543 let client = cmd 1544 .stderr(Stdio::piped()) 1545 .stdout(Stdio::piped()) 1546 .spawn() 1547 .map_err(Error::Spawn)?; 1548 1549 clients.push(client); 1550 } 1551 1552 let mut err: Option<Error> = None; 1553 let mut results = Vec::new(); 1554 let mut failed = false; 1555 for c in clients { 1556 let mut c = c; 1557 if let Err(e) = child_wait_timeout(&mut c, test_timeout as u64 + 5) { 1558 err = Some(Error::WaitTimeout(e)); 1559 failed = true; 1560 } 1561 1562 if !failed { 1563 // Safe to unwrap as we know the child has terminated succesffully 1564 let output = c.wait_with_output().unwrap(); 1565 results.push(parse_iperf3_output(&output.stdout, receive, bandwidth)?); 1566 } else { 1567 let _ = c.kill(); 1568 let output = c.wait_with_output().unwrap(); 1569 println!( 1570 "=============== Client output [Error] ===============\n\n{}\n\n===========end============\n\n", 1571 String::from_utf8_lossy(&output.stdout) 1572 ); 1573 } 1574 } 1575 1576 if let Some(e) = err { 1577 Err(e) 1578 } else { 1579 Ok(results.iter().sum()) 1580 } 1581 } 1582 1583 pub fn parse_ethr_latency_output(output: &[u8]) -> Result<Vec<f64>, Error> { 1584 std::panic::catch_unwind(|| { 1585 let s = String::from_utf8_lossy(output); 1586 let mut latency = Vec::new(); 1587 for l in s.lines() { 1588 let v: Value = serde_json::from_str(l).expect("'ethr' parse error: invalid json line"); 1589 // Skip header/summary lines 1590 if let Some(avg) = v["Avg"].as_str() { 1591 // Assume the latency unit is always "us" 1592 latency.push( 1593 avg.split("us").collect::<Vec<&str>>()[0] 1594 .parse::<f64>() 1595 .expect("'ethr' parse error: invalid 'Avg' entry"), 1596 ); 1597 } 1598 } 1599 1600 assert!( 1601 !latency.is_empty(), 1602 "'ethr' parse error: no valid latency data found" 1603 ); 1604 1605 latency 1606 }) 1607 .map_err(|_| { 1608 eprintln!( 1609 "=============== ethr output ===============\n\n{}\n\n===========end============\n\n", 1610 String::from_utf8_lossy(output) 1611 ); 1612 Error::EthrLogParse 1613 }) 1614 } 1615 1616 pub fn measure_virtio_net_latency(guest: &Guest, test_timeout: u32) -> Result<Vec<f64>, Error> { 1617 // copy the 'ethr' tool to the guest image 1618 let ethr_path = "/usr/local/bin/ethr"; 1619 let ethr_remote_path = "/tmp/ethr"; 1620 scp_to_guest( 1621 Path::new(ethr_path), 1622 Path::new(ethr_remote_path), 1623 &guest.network.guest_ip, 1624 //DEFAULT_SSH_RETRIES, 1625 1, 1626 DEFAULT_SSH_TIMEOUT, 1627 )?; 1628 1629 // Start the ethr server on the guest 1630 guest.ssh_command(&format!("{ethr_remote_path} -s &> /dev/null &"))?; 1631 1632 thread::sleep(Duration::new(10, 0)); 1633 1634 // Start the ethr client on the host 1635 let log_file = guest 1636 .tmp_dir 1637 .as_path() 1638 .join("ethr.client.log") 1639 .to_str() 1640 .unwrap() 1641 .to_string(); 1642 let mut c = Command::new(ethr_path) 1643 .args([ 1644 "-c", 1645 &guest.network.guest_ip, 1646 "-t", 1647 "l", 1648 "-o", 1649 &log_file, // file output is JSON format 1650 "-d", 1651 &format!("{test_timeout}s"), 1652 ]) 1653 .stderr(Stdio::piped()) 1654 .stdout(Stdio::piped()) 1655 .spawn() 1656 .map_err(Error::Spawn)?; 1657 1658 if let Err(e) = child_wait_timeout(&mut c, test_timeout as u64 + 5).map_err(Error::WaitTimeout) 1659 { 1660 let _ = c.kill(); 1661 return Err(e); 1662 } 1663 1664 // Parse the ethr latency test output 1665 let content = fs::read(log_file).map_err(Error::EthrLogFile)?; 1666 parse_ethr_latency_output(&content) 1667 } 1668