Handle errors and crashes in VM differently to clean shutdown.
BUG=b:211704107
TEST=tools/dev_container tools/run_tests
TEST=tools/dev_container tools/run_tests --target=vm:aarch64
Change-Id: I383be89becefe300d9dc94b7d67d35269c628a39
Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/crosvm/+/3412772
Reviewed-by: Dmitry Torokhov <dtor@chromium.org>
Auto-Submit: Andrew Walbran <qwandor@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Commit-Queue: Andrew Walbran <qwandor@google.com>
diff --git a/src/linux.rs b/src/linux.rs
index 3c4b00b..8b2583a 100644
--- a/src/linux.rs
+++ b/src/linux.rs
@@ -2477,6 +2477,7 @@
mut mmio_bus: devices::Bus,
exit_evt: Event,
reset_evt: Event,
+ crash_evt: Event,
requires_pvclock_ctrl: bool,
from_main_tube: mpsc::Receiver<VcpuControl>,
use_hypervisor_signals: bool,
@@ -2561,6 +2562,7 @@
let final_event = match exit_reason {
ExitState::Stop => exit_evt,
ExitState::Reset => reset_evt,
+ ExitState::Crash => crash_evt,
};
if let Err(e) = final_event.write(1) {
error!(
@@ -2615,13 +2617,13 @@
Ok(m) => m,
Err(mpsc::RecvError) => {
error!("Failed to read from main tube in vcpu");
- return ExitState::Stop;
+ return ExitState::Crash;
}
}
}
Err(mpsc::TryRecvError::Disconnected) => {
error!("Failed to read from main tube in vcpu");
- return ExitState::Stop;
+ return ExitState::Crash;
}
};
@@ -2762,7 +2764,7 @@
hardware_entry_failure_reason,
}) => {
error!("vcpu hw run failure: {:#x}", hardware_entry_failure_reason);
- return ExitState::Stop;
+ return ExitState::Crash;
}
Ok(VcpuExit::SystemEventShutdown) => {
info!("system shutdown event on vcpu {}", cpu_id);
@@ -2786,7 +2788,7 @@
if let Some(ref ch) = to_gdb_tube {
if let Err(e) = ch.send(msg) {
error!("failed to notify breakpoint to GDB thread: {}", e);
- return ExitState::Stop;
+ return ExitState::Crash;
}
}
run_mode = VmRunMode::Breakpoint;
@@ -2798,7 +2800,7 @@
libc::EAGAIN => {}
_ => {
error!("vcpu hit unknown error: {}", e);
- return ExitState::Stop;
+ return ExitState::Crash;
}
},
}
@@ -2810,7 +2812,7 @@
// attempting to handle pause requests.
if let Err(e) = clear_signal(SIGRTMIN() + 0) {
error!("failed to clear pending signal: {}", e);
- return ExitState::Stop;
+ return ExitState::Crash;
}
} else {
vcpu.set_immediate_exit(false);
@@ -2915,6 +2917,7 @@
pub enum ExitState {
Reset,
Stop,
+ Crash,
}
pub fn run_config(cfg: Config) -> Result<ExitState> {
@@ -3135,6 +3138,7 @@
let exit_evt = Event::new().context("failed to create event")?;
let reset_evt = Event::new().context("failed to create event")?;
+ let crash_evt = Event::new().context("failed to create event")?;
let mut sys_allocator = Arch::create_system_allocator(&vm);
// Allocate the ramoops region first. AArch64::build_vm() assumes this.
@@ -3329,6 +3333,7 @@
usb_control_tube,
exit_evt,
reset_evt,
+ crash_evt,
sigchld_fd,
Arc::clone(&map_request),
gralloc,
@@ -3473,6 +3478,7 @@
#[cfg(feature = "usb")] usb_control_tube: Tube,
exit_evt: Event,
reset_evt: Event,
+ crash_evt: Event,
sigchld_fd: SignalFd,
map_request: Arc<Mutex<Option<ExternalMapping>>>,
mut gralloc: RutabagaGralloc,
@@ -3482,6 +3488,7 @@
enum Token {
Exit,
Reset,
+ Crash,
Suspend,
ChildSignal,
IrqFd { index: IrqEventIndex },
@@ -3496,6 +3503,7 @@
let wait_ctx = WaitContext::build_with(&[
(&exit_evt, Token::Exit),
(&reset_evt, Token::Reset),
+ (&crash_evt, Token::Crash),
(&linux.suspend_evt, Token::Suspend),
(&sigchld_fd, Token::ChildSignal),
])
@@ -3594,6 +3602,7 @@
(*linux.mmio_bus).clone(),
exit_evt.try_clone().context("failed to clone event")?,
reset_evt.try_clone().context("failed to clone event")?,
+ crash_evt.try_clone().context("failed to clone event")?,
linux.vm.check_capability(VmCap::PvClockSuspend),
from_main_channel,
use_hypervisor_signals,
@@ -3662,6 +3671,11 @@
exit_state = ExitState::Reset;
break 'wait;
}
+ Token::Crash => {
+ info!("vcpu crashed");
+ exit_state = ExitState::Crash;
+ break 'wait;
+ }
Token::Suspend => {
info!("VM requested suspend");
linux.suspend_evt.read().unwrap();
diff --git a/src/main.rs b/src/main.rs
index ef1b93b..3040dd6 100644
--- a/src/main.rs
+++ b/src/main.rs
@@ -2450,6 +2450,7 @@
Success,
VmReset,
VmStop,
+ VmCrash,
}
fn run_vm(args: std::env::Args) -> std::result::Result<CommandStatus, ()> {
@@ -2740,6 +2741,10 @@
info!("crosvm has exited normally due to reset request");
Ok(CommandStatus::VmReset)
}
+ Ok(platform::ExitState::Crash) => {
+ info!("crosvm has exited due to a VM crash");
+ Ok(CommandStatus::VmCrash)
+ }
Err(e) => {
error!("crosvm has exited with error: {:#}", e);
Err(())
@@ -3395,6 +3400,7 @@
let exit_code = match crosvm_main() {
Ok(CommandStatus::Success | CommandStatus::VmStop) => 0,
Ok(CommandStatus::VmReset) => 32,
+ Ok(CommandStatus::VmCrash) => 33,
Err(_) => 1,
};
std::process::exit(exit_code);