Skip to content
Draft
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
63 changes: 47 additions & 16 deletions internal/provisioning/hostagent/networkmanager/network_manager.go
Original file line number Diff line number Diff line change
Expand Up @@ -154,16 +154,26 @@ func (nm *NetworkManager) loadNetworkRequest() error {
}

func (nm *NetworkManager) run() {
nm.Lock()
defer nm.Unlock()
nm.RLock()
reqs := make([]NetworkRequest, 0, len(nm.reqs))
for _, nr := range nm.reqs {
reqs = append(reqs, nr)
}
nm.RUnlock()

for uid, nr := range nm.reqs {
for _, nr := range reqs {
if err := nm.processNetworkRequest(nr); err != nil {
klog.Errorf("failed to process network request, nr: %+v, err: %v", nm.reqs[uid], err)
klog.Errorf("failed to process network request, nr: %+v, err: %v", nr, err)
}
}
}

func (nm *NetworkManager) removeRequest(uid string) {
nm.Lock()
defer nm.Unlock()
delete(nm.reqs, uid)
}

func (nm *NetworkManager) processNetworkRequest(nr NetworkRequest) error {
nn := types.NamespacedName{Namespace: nr.DPUNamespace, Name: nr.DpuName}
dpu := &provisioningv1.DPU{}
Expand All @@ -180,7 +190,7 @@ func (nm *NetworkManager) processNetworkRequest(nr NetworkRequest) error {
if err != nil && !os.IsNotExist(err) {
return fmt.Errorf("failed to remove network request file: %w", err)
}
delete(nm.reqs, nr.UID)
nm.removeRequest(nr.UID)
klog.Infof("removed VF and network request for DPU %s/%s(UID: %s)", nr.DPUNamespace, nr.DpuName, nr.UID)
return nil
}
Expand Down Expand Up @@ -270,16 +280,34 @@ func (nm *NetworkManager) processNetworkRequest(nr NetworkRequest) error {
return nil
}

func (nm *NetworkManager) AddNetworkRequest(dpu *provisioningv1.DPU) error {
nm.Lock()
defer nm.Unlock()
// lookupDevice checks preconditions and returns the PCI device for the DPU.
// Returns found=false when a network request for this DPU already exists.
func (nm *NetworkManager) lookupDevice(dpu *provisioningv1.DPU) (dev hostutil.Device, found bool, err error) {
nm.RLock()
defer nm.RUnlock()
if !nm.initialized {
return fmt.Errorf("network manager is not initialized")
} else if dpu == nil {
return hostutil.Device{}, false, fmt.Errorf("network manager is not initialized")
}
if _, ok := nm.reqs[string(dpu.UID)]; ok {
return hostutil.Device{}, false, nil
}
dev, ok := nm.devicesBySN[dpu.Spec.SerialNumber]
if !ok {
return hostutil.Device{}, false, fmt.Errorf("PCI address of device %s not found", dpu.Spec.SerialNumber)
}
return dev, true, nil
}

func (nm *NetworkManager) AddNetworkRequest(dpu *provisioningv1.DPU) error {
if dpu == nil {
return fmt.Errorf("DPU is nil")
}

if _, ok := nm.reqs[string(dpu.UID)]; ok {
dev, found, err := nm.lookupDevice(dpu)
if err != nil {
return err
}
if !found {
return nil
}

Expand All @@ -289,10 +317,6 @@ func (nm *NetworkManager) AddNetworkRequest(dpu *provisioningv1.DPU) error {
nr.SetDPUObjectMeta(*dpu)

// use the PCI address collected locally, so that it's not affected by PCI address changes
dev, ok := nm.devicesBySN[nr.SerialNumber]
if !ok {
return fmt.Errorf("PCI address of device %s not found", nr.SerialNumber)
}
nr.PCIAddress = dev.Address

numOfVFs, err := nm.getNumOfVFs(dpu)
Expand All @@ -317,10 +341,17 @@ func (nm *NetworkManager) AddNetworkRequest(dpu *provisioningv1.DPU) error {
if err := writeNetworkRequestFile(nr); err != nil {
return fmt.Errorf("failed to write network request file: %w", err)
}
nm.reqs[nr.UID] = *nr

nm.addRequest(nr)
return nil
}

func (nm *NetworkManager) addRequest(nr *NetworkRequest) {
nm.Lock()
defer nm.Unlock()
nm.reqs[nr.UID] = *nr
}

func (nm *NetworkManager) getNumOfVFs(dpu *provisioningv1.DPU) (int, error) {
flavor := &provisioningv1.DPUFlavor{}
if err := nm.Get(context.TODO(), types.NamespacedName{Namespace: dpu.Namespace, Name: dpu.Spec.DPUFlavor}, flavor); err != nil {
Expand Down
Loading