diff --git a/runsc/cmd/gofer.go b/runsc/cmd/gofer.go
index 08cdc76c4b..7df103afa4 100644
--- a/runsc/cmd/gofer.go
+++ b/runsc/cmd/gofer.go
@@ -32,9 +32,11 @@ import (
 	"gvisor.dev/gvisor/pkg/log"
 	"gvisor.dev/gvisor/pkg/sentry/devices/tpuproxy/vfio"
 	"gvisor.dev/gvisor/pkg/unet"
+	"gvisor.dev/gvisor/pkg/urpc"
 	"gvisor.dev/gvisor/runsc/boot"
 	"gvisor.dev/gvisor/runsc/cmd/util"
 	"gvisor.dev/gvisor/runsc/config"
+	"gvisor.dev/gvisor/runsc/container"
 	"gvisor.dev/gvisor/runsc/flag"
 	"gvisor.dev/gvisor/runsc/fsgofer"
 	"gvisor.dev/gvisor/runsc/fsgofer/filter"
@@ -91,6 +93,7 @@ type Gofer struct {
 
 	specFD        int
 	mountsFD      int
+	rpcFD         int
 	profileFDs    profile.FDArgs
 	syncFDs       goferSyncFDs
 	stopProfiling func()
@@ -123,6 +126,7 @@ func (g *Gofer) SetFlags(f *flag.FlagSet) {
 	f.IntVar(&g.devIoFD, "dev-io-fd", -1, "optional FD to connect /dev gofer server")
 	f.IntVar(&g.specFD, "spec-fd", -1, "required fd with the container spec")
 	f.IntVar(&g.mountsFD, "mounts-fd", -1, "mountsFD is the file descriptor to write list of mounts after they have been resolved (direct paths, no symlinks).")
+	f.IntVar(&g.rpcFD, "rpc-fd", -1, "mountsFD is the RPC file descriptor.")
 
 	// Add synchronization FD flags.
 	g.syncFDs.setFlags(f)
@@ -153,8 +157,16 @@ func (g *Gofer) Execute(_ context.Context, f *flag.FlagSet, args ...any) subcomm
 	g.syncFDs.syncNVProxy()
 	g.syncFDs.syncUsernsForRootless()
 
+	rpcClntSock, err := unet.NewSocket(g.rpcFD)
+	if err != nil {
+		util.Fatalf("creating rpc socket: %v", err)
+	}
+
+	rpcClnt := urpc.NewClient(rpcClntSock)
+	defer rpcClnt.Close()
+
 	if g.setUpRoot {
-		if err := g.setupRootFS(spec, conf); err != nil {
+		if err := g.setupRootFS(spec, conf, rpcClnt); err != nil {
 			util.Fatalf("Error setting up root FS: %v", err)
 		}
 		if !conf.TestOnlyAllowRunAsCurrentUserWithoutChroot {
@@ -162,6 +174,7 @@ func (g *Gofer) Execute(_ context.Context, f *flag.FlagSet, args ...any) subcomm
 			defer cleanupUnmounter()
 		}
 	}
+	rpcClnt.Close()
 	if g.applyCaps {
 		overrides := g.syncFDs.flags()
 		overrides["apply-caps"] = "false"
@@ -369,7 +382,7 @@ func (g *Gofer) writeMounts(mounts []specs.Mount) error {
 // It is protected by selinux rules.
 const procFDBindMount = "/proc/fs"
 
-func (g *Gofer) setupRootFS(spec *specs.Spec, conf *config.Config) error {
+func (g *Gofer) setupRootFS(spec *specs.Spec, conf *config.Config, rpcClnt *urpc.Client) error {
 	// Convert all shared mounts into slaves to be sure that nothing will be
 	// propagated outside of our namespace.
 	procPath := "/proc"
@@ -437,7 +450,7 @@ func (g *Gofer) setupRootFS(spec *specs.Spec, conf *config.Config) error {
 	}
 
 	// Replace the current spec, with the clean spec with symlinks resolved.
-	if err := g.setupMounts(conf, spec.Mounts, root, procPath); err != nil {
+	if err := g.setupMounts(conf, spec.Mounts, root, procPath, rpcClnt); err != nil {
 		util.Fatalf("error setting up FS: %v", err)
 	}
 
@@ -487,7 +500,7 @@ func (g *Gofer) setupRootFS(spec *specs.Spec, conf *config.Config) error {
 // setupMounts bind mounts all mounts specified in the spec in their correct
 // location inside root. It will resolve relative paths and symlinks. It also
 // creates directories as needed.
-func (g *Gofer) setupMounts(conf *config.Config, mounts []specs.Mount, root, procPath string) error {
+func (g *Gofer) setupMounts(conf *config.Config, mounts []specs.Mount, root, procPath string, rpcClnt *urpc.Client) error {
 	mountIdx := 1 // First index is for rootfs.
 	for _, m := range mounts {
 		if !specutils.IsGoferMount(m) {
@@ -511,7 +524,19 @@ func (g *Gofer) setupMounts(conf *config.Config, mounts []specs.Mount, root, pro
 		}
 
 		log.Infof("Mounting src: %q, dst: %q, flags: %#x", m.Source, dst, flags)
-		if err := specutils.SafeSetupAndMount(m.Source, dst, m.Type, flags, procPath); err != nil {
+		src := m.Source
+		var fd *os.File
+		if unix.Access(src, unix.R_OK); err != nil {
+			var res container.OpenMountResult
+			if err := rpcClnt.Call("goferRPC.OpenMount", &m, &res); err != nil {
+				return fmt.Errorf("opening %s: %w", m.Source, err)
+			}
+			fd = res.Files[0]
+			src = fmt.Sprintf("%s/self/fd/%d", procPath, fd)
+		}
+		err = specutils.SafeSetupAndMount(src, dst, m.Type, flags, procPath)
+		fd.Close()
+		if err != nil {
 			return fmt.Errorf("mounting %+v: %v", m, err)
 		}
 
diff --git a/runsc/container/BUILD b/runsc/container/BUILD
index 135a918cbb..ca7b6473c5 100644
--- a/runsc/container/BUILD
+++ b/runsc/container/BUILD
@@ -9,6 +9,7 @@ go_library(
     name = "container",
     srcs = [
         "container.go",
+        "gofer_rpc.go",
         "hook.go",
         "state_file.go",
         "status.go",
@@ -28,6 +29,8 @@ go_library(
         "//pkg/sighandling",
         "//pkg/state/statefile",
         "//pkg/sync",
+        "//pkg/unet",
+        "//pkg/urpc",
         "//runsc/boot",
         "//runsc/cgroup",
         "//runsc/config",
diff --git a/runsc/container/container.go b/runsc/container/container.go
index a87c48c865..123fa452c6 100644
--- a/runsc/container/container.go
+++ b/runsc/container/container.go
@@ -41,6 +41,8 @@ import (
 	"gvisor.dev/gvisor/pkg/sentry/pgalloc"
 	"gvisor.dev/gvisor/pkg/sighandling"
 	"gvisor.dev/gvisor/pkg/state/statefile"
+	"gvisor.dev/gvisor/pkg/unet"
+	"gvisor.dev/gvisor/pkg/urpc"
 	"gvisor.dev/gvisor/runsc/boot"
 	"gvisor.dev/gvisor/runsc/cgroup"
 	"gvisor.dev/gvisor/runsc/config"
@@ -1272,6 +1274,25 @@ func (c *Container) createGoferProcess(spec *specs.Spec, conf *config.Config, bu
 	}
 	donations.DonateAndClose("mounts-fd", mountsGofer)
 
+	rpcServ, rpcClnt, err := unet.SocketPair(false)
+	if err != nil {
+		return nil, nil, nil, fmt.Errorf("failed to create an rpc socket pair: %w", err)
+	}
+	rpcClntFD, _ := rpcClnt.Release()
+	donations.DonateAndClose("rpc-fd", os.NewFile(uintptr(rpcClntFD), "gofer-rpc"))
+	rpcPidCh := make(chan int, 1)
+	defer close(rpcPidCh)
+	go func() {
+		pid := <-rpcPidCh
+		if pid == 0 {
+			rpcServ.Close()
+			return
+		}
+		s := urpc.NewServer()
+		s.Register(&goferRPC{goferPID: pid})
+		s.StartHandling(rpcServ)
+	}()
+
 	// Count the number of mounts that needs an IO file.
 	ioFileCount := 0
 	for _, cfg := range c.GoferMountConfs {
@@ -1370,6 +1391,7 @@ func (c *Container) createGoferProcess(spec *specs.Spec, conf *config.Config, bu
 	log.Infof("Gofer started, PID: %d", cmd.Process.Pid)
 	c.GoferPid = cmd.Process.Pid
 	c.goferIsChild = true
+	rpcPidCh <- cmd.Process.Pid
 
 	// Set up and synchronize rootless mode userns mappings.
 	if rootlessEUID {
diff --git a/runsc/container/gofer_rpc.go b/runsc/container/gofer_rpc.go
new file mode 100644
index 0000000000..9898ee56b4
--- /dev/null
+++ b/runsc/container/gofer_rpc.go
@@ -0,0 +1,103 @@
+// Copyright 2024 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package container
+
+import (
+	"fmt"
+	"os"
+	"runtime"
+	"sync"
+
+	specs "github.com/opencontainers/runtime-spec/specs-go"
+	"golang.org/x/sys/unix"
+	"gvisor.dev/gvisor/pkg/urpc"
+)
+
+type openMountRequest struct {
+	mount  *specs.Mount
+	result *OpenMountResult
+	done   chan error
+}
+
+type goferRPC struct {
+	mu                sync.Mutex
+	openMountRequests chan *openMountRequest
+	goferPID          int
+}
+
+// OpenMountResult is a result of the rpcp.OpenMount call.
+type OpenMountResult struct {
+	urpc.FilePayload
+}
+
+func (rpc *goferRPC) handleRequest(req *openMountRequest) {
+	defer close(req.done)
+	fd, err := os.OpenFile(req.mount.Source, unix.O_PATH|unix.O_CLOEXEC, 0)
+	if err != nil {
+		req.done <- err
+		return
+	}
+	req.result.Files = []*os.File{fd}
+}
+
+func (rpc *goferRPC) openMountLoop() error {
+	if err := unix.Unshare(unix.CLONE_FS); err != nil {
+		return fmt.Errorf("open mount thread: %w", err)
+	}
+	nsFd, err := os.Open(fmt.Sprintf("/proc/%d/ns/mnt", rpc.goferPID))
+	if err != nil {
+		return fmt.Errorf("open mount thread: open container mntns: %w", err)
+	}
+	defer nsFd.Close()
+	if err := unix.Setns(int(nsFd.Fd()), unix.CLONE_NEWNS); err != nil {
+		return fmt.Errorf("open mount thread: join container mntns: %w", err)
+	}
+	for req := range rpc.openMountRequests {
+		rpc.handleRequest(req)
+	}
+	return nil
+}
+
+// OpenMount is a helper rpc call that a gofer process uses when it can't
+// open/create a mount.
+func (rpc *goferRPC) OpenMount(m *specs.Mount, res *OpenMountResult) error {
+	rpc.mu.Lock()
+	defer rpc.mu.Unlock()
+
+	if rpc.openMountRequests == nil {
+		rpc.openMountRequests = make(chan *openMountRequest)
+		go func() {
+			// This goroutine holds the current threads forever. It
+			// never exits, because child proccesses can set
+			// PDEATHSIG. It can't serve other go-routines, because
+			// it does unshare CLONE_FS.
+			runtime.LockOSThread()
+			if err := rpc.openMountLoop(); err != nil {
+				for req := range rpc.openMountRequests {
+					req.done <- err
+				}
+			}
+			panic("unreachable")
+		}()
+	}
+	req := openMountRequest{
+		mount:  m,
+		result: res,
+		done:   make(chan error),
+	}
+	rpc.openMountRequests <- &req
+	err := <-req.done
+	return err
+}