vpp-manager: rebind interfaces to kernel if link is missing

If VPP is killed abruptly (e.g. due to OOM), it does not shut down
gracefully and does not restore interface bindings. As a result,
interfaces may remain bound to a DPDK driver and no longer appear
as Linux network devices.

When the expected interface is not found, attempt to rebind PCI
devices back to the kernel driver and retry the lookup once.

Because we don't die gracefully, pingCalicoVPP() won't be called
and agent will never restart. So we call it in the beginning to
kill previous stale agents if any.

Author: hedibouattour

config/config.go

@@ -22,6 +22,7 @@ import (
 	"net"
 	"os"
 	"os/exec"
+	"path/filepath"
 	"regexp"
 	"sort"
 	"strconv"
@@ -726,13 +727,67 @@ type LinuxInterfaceState struct {
 	TapSwIfIndex uint32
 }
 
+func bindPCIDevicesToKernel() error {
+	drivers := []string{"igb_uio", "uio_pci_generic", "vfio-pci"}
+	removed := false
+
+	for _, driver := range drivers {
+		pattern := filepath.Join("/sys/bus/pci/drivers", driver, "*")
+		matches, err := filepath.Glob(pattern)
+		if err != nil {
+			return err
+		}
+
+		for _, f := range matches {
+			configPath := filepath.Join(f, "config")
+
+			// Skip if config does not exist
+			if _, err := os.Stat(configPath); err != nil {
+				continue
+			}
+
+			// Check if config file is in use via fuser
+			cmd := exec.Command("fuser", "-s", configPath)
+			if err := cmd.Run(); err == nil {
+				// fuser found a user, skip
+				continue
+			}
+
+			// Write "1" to remove file
+			removePath := filepath.Join(f, "remove")
+			if err := os.WriteFile(removePath, []byte("1"), 0200); err != nil {
+				return err
+			}
+
+			removed = true
+		}
+	}
+
+	// If any device was removed, rescan PCI bus
+	if removed {
+		if err := os.WriteFile("/sys/bus/pci/rescan", []byte("1"), 0200); err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
 func LoadInterfaceConfigFromLinux(interfaceName string) (*LinuxInterfaceState, error) {
 	conf := LinuxInterfaceState{
 		TapSwIfIndex: ^uint32(0), // in case we forget to set it
 	}
 	link, err := netlink.LinkByName(interfaceName)
 	if err != nil {
-		return nil, errors.Wrapf(err, "cannot find interface named %s", interfaceName)
+		// attempt binding PCI devices to kernel
+		bindErr := bindPCIDevicesToKernel()
+		if bindErr != nil {
+			return nil, errors.Wrapf(err, "cannot find interface named %s, cannot bind pci devices to kernel: %v", interfaceName, bindErr)
+		}
+		link, err = netlink.LinkByName(interfaceName)
+		if err != nil {
+			return nil, errors.Wrapf(err, "cannot find interface named %s after binding devices to kernel", interfaceName)
+		}
 	}
 	conf.IsUp = (link.Attrs().Flags & net.FlagUp) != 0
 	if conf.IsUp {

vpp-manager/vpp_runner.go

@@ -944,6 +944,16 @@ func (v *VppRunner) runVpp() (err error) {
 		}
 		vppProcess = vppCmd.Process
 	}
+	/**
+	* Ensure any stale Calico VPP agents are terminated by calling pingCalicoVPP().
+	* This handles cases where a previous VPP instance was killed abruptly and
+	* didn't restore its configuration.
+	* Must be done before writing the info file to make sure the new agent
+	* doesn't respond to SIGUSR1 and avoid killing it */
+	err = v.pingCalicoVpp()
+	if err != nil {
+		log.Errorf("Error pinging calico-vpp: %v", err)
+	}
 
 	defer func() {
 		if r := recover(); r != nil {