From a090a6775613903e2d7dc05838fb9cbfaa67fa27 Mon Sep 17 00:00:00 2001
From: "pulse-triage[bot]"
 <249995291+pulse-triage[bot]@users.noreply.github.com>
Date: Sat, 3 Oct 2026 14:21:03 +0100
Subject: [PATCH] Keep guest-agent admission single-attempt for password
 sessions

The generic password-session client retries once after a 401; that implicit retry can bypass a fresh lock check inside a guest admission. Disable it only for QGA commands, retain the refusal, and prove healthy guest reads plus unchanged ordinary API session recovery.

Contract-Neutral: Completes the documented no-replay guest admission rule without adding a contract or surface; adds password refusal and healthy recovery controls.
Change-source: pulse-maintainer
---
 pkg/proxmox/guest_agent_guard.go      |  5 ++-
 pkg/proxmox/guest_agent_guard_test.go | 52 +++++++++++++++++++++++++++
 2 files changed, 56 insertions(+), 1 deletion(-)

diff --git a/pkg/proxmox/guest_agent_guard.go b/pkg/proxmox/guest_agent_guard.go
index d273d7ae02..1f8df97621 100644
--- a/pkg/proxmox/guest_agent_guard.go
+++ b/pkg/proxmox/guest_agent_guard.go
@@ -171,7 +171,10 @@ func (c *Client) getGuestAgent(ctx context.Context, path, node string, vmid int)
 	if err := c.verifyGuestAgentUnlocked(ctx, node, vmid); err != nil {
 		return nil, err
 	}
-	resp, err := c.request(ctx, http.MethodGet, path, nil)
+	// A password-session 401 must not replay this command after re-authentication:
+	// the operation lock could change between attempts. Ordinary API reads still
+	// own their usual session recovery; this guest admission is single-attempt.
+	resp, err := c.requestWithRetry(ctx, http.MethodGet, path, nil, true)
 	if err != nil {
 		// Explicit refusals/unsupported commands remain ordinary errors. An
 		// uncertain completion blocks every command, not just this method.
diff --git a/pkg/proxmox/guest_agent_guard_test.go b/pkg/proxmox/guest_agent_guard_test.go
index bf27f30353..62d77db187 100644
--- a/pkg/proxmox/guest_agent_guard_test.go
+++ b/pkg/proxmox/guest_agent_guard_test.go
@@ -189,3 +189,55 @@ func TestGuestAgentCooldownExpiresWithoutSleepOrReplay(t *testing.T) {
 		t.Fatal("healthy completed operation retained estate state")
 	}
 }
+
+func TestGuestAgentPasswordSessionDoesNotReplayRefusedCommand(t *testing.T) {
+	for _, refused := range []bool{false, true} {
+		t.Run(fmt.Sprintf("refused-%t", refused), func(t *testing.T) {
+			var authCalls, agentCalls, nodeCalls atomic.Int32
+			server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+				switch {
+				case strings.HasSuffix(r.URL.Path, "/access/ticket"):
+					authCalls.Add(1)
+					fmt.Fprint(w, `{"data":{"ticket":"fixture-ticket","CSRFPreventionToken":"fixture-csrf"}}`)
+				case strings.HasSuffix(r.URL.Path, "/config"):
+					fmt.Fprint(w, `{"data":{}}`)
+				case strings.Contains(r.URL.Path, "/agent/"):
+					if agentCalls.Add(1) == 1 && refused {
+						http.Error(w, "session refused", http.StatusUnauthorized)
+						return
+					}
+					backupAgentPayload(w, r)
+				case strings.HasSuffix(r.URL.Path, "/nodes"):
+					if nodeCalls.Add(1) == 1 {
+						http.Error(w, "expired ordinary API session", http.StatusUnauthorized)
+						return
+					}
+					fmt.Fprint(w, `{"data":[]}`)
+				default:
+					http.NotFound(w, r)
+				}
+			}))
+			defer server.Close()
+			client, err := NewClient(ClientConfig{Host: server.URL, User: "fixture@pam", Password: "fixture-password", Timeout: time.Second})
+			if err != nil {
+				t.Fatal(err)
+			}
+			_, err = client.GetVMFSInfo(context.Background(), "node", 105)
+			if refused && (err == nil || !strings.Contains(err.Error(), "401")) {
+				t.Fatalf("guest refusal lost: %v", err)
+			}
+			if !refused && err != nil {
+				t.Fatal(err)
+			}
+			if agentCalls.Load() != 1 || authCalls.Load() != 1 {
+				t.Fatalf("guest replay: commands=%d authentications=%d", agentCalls.Load(), authCalls.Load())
+			}
+			if _, err := client.GetNodes(context.Background()); err != nil {
+				t.Fatalf("ordinary API session recovery broken: %v", err)
+			}
+			if agentCalls.Load() != 1 || nodeCalls.Load() != 2 || authCalls.Load() != 2 {
+				t.Fatalf("ordinary recovery replayed guest or did not recover: guest=%d nodes=%d auth=%d", agentCalls.Load(), nodeCalls.Load(), authCalls.Load())
+			}
+		})
+	}
+}