From bbdf18351f9de1e588b8a98372c9b9159baf3241 Mon Sep 17 00:00:00 2001
From: iximeow <iximeow@oxide.computer>
Date: Thu, 12 Mar 2026 19:01:47 +0000
Subject: [PATCH] phd: collect core when killing non-booting guest

this may or may not prove useful in practice; if we're lucky something
got funky in device emulation and we can see a stuck thread. on the
other hand, if we're unlucky the guest is stuck in a loop and all we see
is one vCPU was running while everything else was idle.
---
 phd-tests/framework/src/test_vm/mod.rs    | 10 +++++-
 phd-tests/framework/src/test_vm/server.rs | 39 ++++++++++++++++++++++-
 2 files changed, 47 insertions(+), 2 deletions(-)

diff --git a/phd-tests/framework/src/test_vm/mod.rs b/phd-tests/framework/src/test_vm/mod.rs
index 98686861b..32a997c9e 100644
--- a/phd-tests/framework/src/test_vm/mod.rs
+++ b/phd-tests/framework/src/test_vm/mod.rs
@@ -887,7 +887,15 @@ impl TestVm {
         .instrument(info_span!("wait_to_boot"));
 
         match timeout(timeout_duration, boot).await {
-            Err(_) => anyhow::bail!("timed out while waiting to boot"),
+            Err(_) => {
+                error!(
+                    "Guest did not boot after {}ms! Collecting core..",
+                    timeout_duration.as_millis()
+                );
+                let proc = self.server.as_ref().unwrap();
+                proc.core();
+                anyhow::bail!("timed out while waiting to boot")
+            }
             Ok(inner) => {
                 inner.context("executing guest login sequence")?;
             }
diff --git a/phd-tests/framework/src/test_vm/server.rs b/phd-tests/framework/src/test_vm/server.rs
index be88a665b..737f6bad7 100644
--- a/phd-tests/framework/src/test_vm/server.rs
+++ b/phd-tests/framework/src/test_vm/server.rs
@@ -8,11 +8,12 @@ use std::{
     fmt::Debug,
     net::{SocketAddr, SocketAddrV4},
     os::unix::process::CommandExt,
+    time::SystemTime,
 };
 
 use anyhow::Result;
 use camino::{Utf8Path, Utf8PathBuf};
-use tracing::{debug, info};
+use tracing::{debug, info, warn};
 
 use crate::log_config::LogConfig;
 
@@ -44,6 +45,7 @@ pub struct ServerProcessParameters<'a> {
 pub struct PropolisServer {
     server: Option<std::process::Child>,
     address: SocketAddrV4,
+    output_dir: Utf8PathBuf,
 }
 
 impl PropolisServer {
@@ -117,6 +119,9 @@ impl PropolisServer {
         let server = PropolisServer {
             server: Some(server_cmd.spawn()?),
             address: server_addr,
+            // Stash the same output directory in case the framework has to
+            // write any files on behalf of the test run.
+            output_dir: output_dir.to_owned(),
         };
 
         info!(
@@ -130,6 +135,38 @@ impl PropolisServer {
         self.address
     }
 
+    /// Collect a core of this server process, placing it in the same output
+    /// directory as other artifacts of this test.
+    pub(super) fn core(&self) {
+        let Some(server_proc) = self.server.as_ref() else {
+            warn!("Tried to produce a core without a propolis-server?");
+            return;
+        };
+
+        let core_name = format!(
+            "core-{}",
+            SystemTime::now()
+                .duration_since(SystemTime::UNIX_EPOCH)
+                .expect("Time is gone, the song is over")
+                .as_millis()
+        );
+        let core_path = self.output_dir.join(core_name);
+
+        std::process::Command::new("pfexec")
+            .args([
+                "gcore".as_ref(),
+                "-o".as_ref(),
+                core_path.as_os_str(),
+                server_proc.id().to_string().as_ref(),
+            ])
+            .spawn()
+            .expect("can try to gcore a process")
+            .wait()
+            .expect("can gcore a propolis-server we spawned");
+
+        warn!("core written to {}", core_path);
+    }
+
     /// Kills this server process if it hasn't been killed already.
     pub(super) fn kill(&mut self) {
         let Some(mut server) = self.server.take() else {