Adding the ControllerManagementFaultConfiguration to the mix

Author: gurasinghMS

vm/devices/storage/disk_nvme/nvme_driver/src/tests.rs

@@ -11,8 +11,9 @@ use inspect::InspectMut;
 use mesh::CellUpdater;
 use nvme::NvmeControllerCaps;
 use nvme_resources::fault::AdminQueueFaultConfig;
-use nvme_resources::fault::FaultConfiguration;
+use nvme_resources::fault::ControllerManagementFaultConfig;
 use nvme_resources::fault::FaultBehaviour;
+use nvme_resources::fault::FaultConfiguration;
 use nvme_spec::AdminOpcode;
 use nvme_spec::Cap;
 use nvme_spec::Command;
@@ -50,6 +51,7 @@ async fn test_nvme_command_fault(driver: DefaultDriver) {
                 AdminOpcode::CREATE_IO_COMPLETION_QUEUE.0,
                 FaultBehaviour::Update(output_cmd),
             ),
+            controller_management_fault: ControllerManagementFaultConfig::new(),
         },
     )
     .await;

vm/devices/storage/nvme_resources/src/fault.rs

@@ -30,7 +30,7 @@ pub struct ControllerManagementFaultConfig {
     pub controller_management_fault_enable: FaultBehaviour<bool>,
 }
 
-#[derive(MeshPayload)]
+#[derive(MeshPayload, Clone)]
 /// A buildable fault configuration for the controller management interface (cc.en(), csts.rdy(), ... )
 pub struct ControllerManagementFaultConfig {
     /// Fault to apply to cc.en() bit during enablement

vm/devices/storage/nvme_test/src/pci.rs

@@ -30,6 +30,7 @@ use guestmem::GuestMemory;
 use guid::Guid;
 use inspect::Inspect;
 use inspect::InspectMut;
+use nvme_resources::fault::FaultBehaviour;
 use nvme_resources::fault::FaultConfiguration;
 use parking_lot::Mutex;
 use pci_core::capabilities::msix::MsixEmulator;
@@ -157,7 +158,7 @@ impl NvmeFaultController {
             max_cqs: caps.max_io_queues,
             qe_sizes: Arc::clone(&qe_sizes),
             subsystem_id: caps.subsystem_id,
-            fault_configuration,
+            fault_configuration: fault_configuration.clone(),
         });
 
         Self {
@@ -166,6 +167,7 @@ impl NvmeFaultController {
             registers: RegState::new(),
             workers: admin,
             qe_sizes,
+            fault_configuration,
         }
     }
 
@@ -343,8 +345,23 @@ impl NvmeFaultController {
 
         if cc.en() != self.registers.cc.en() {
             if cc.en() {
-                // If any, perform the fault at the beginning of the enable sequence.
-                if 
+                // If any fault was configured for cc.en() process it here
+                match self
+                    .fault_configuration
+                    .controller_management_fault
+                    .controller_management_fault_enable
+                {
+                    FaultBehaviour::Delay(duration) => {
+                        std::thread::sleep(duration);
+                    }
+                    FaultBehaviour::Drop => {
+                        tracelimit::warn_ratelimited!(
+                            "Dropping enable command due to fault injection"
+                        );
+                        return;
+                    }
+                    _ => {} // Update is not yet configured for this fault. Treat that as a default action for now
+                }
 
                 // Some drivers will write zeros to IOSQES and IOCQES, assuming that the defaults will work.
                 if cc.iocqes() == 0 {
@@ -389,8 +406,6 @@ impl NvmeFaultController {
             }
         }
 
-
-
         self.registers.cc = cc;
         *self.qe_sizes.lock() = IoQueueEntrySizes {
             sqe_bits: cc.iosqes(),
@@ -437,6 +452,7 @@ impl ChangeDeviceState for NvmeFaultController {
             registers,
             qe_sizes,
             workers,
+            fault_configuration: _,
         } = self;
         workers.reset().await;
         cfg_space.reset();