Support pipes and sockets in VFS2 gofer fs.

Named pipes and sockets can be represented in two ways in gofer fs:
1. As a file on the remote filesystem. In this case, all file operations are
   passed through 9p.
2. As a synthetic file that is internal to the sandbox. In this case, the
   dentry stores an endpoint or VFSPipe for sockets and pipes respectively,
   which replaces interactions with the remote fs through the gofer.
In gofer.filesystem.MknodAt, we attempt to call mknod(2) through 9p,
and if it fails, fall back to the synthetic version.

Updates #1200.

PiperOrigin-RevId: 308828161

1 files changed, 18 insertions, 0 deletions

diff --git a/pkg/sentry/vfs/options.go b/pkg/sentry/vfs/options.go
index 022bac127..53d364c5c 100644
--- a/pkg/sentry/vfs/options.go
+++ b/pkg/sentry/vfs/options.go
@@ -151,6 +151,24 @@ type SetStatOptions struct {
 	Stat linux.Statx
 }
 
+// BoundEndpointOptions contains options to VirtualFilesystem.BoundEndpointAt()
+// and FilesystemImpl.BoundEndpointAt().
+type BoundEndpointOptions struct {
+	// Addr is the path of the file whose socket endpoint is being retrieved.
+	// It is generally irrelevant: most endpoints are stored at a dentry that
+	// was created through a bind syscall, so the path can be stored on creation.
+	// However, if the endpoint was created in FilesystemImpl.BoundEndpointAt(),
+	// then we may not know what the original bind address was.
+	//
+	// For example, if connect(2) is called with address "foo" which corresponds
+	// a remote named socket in goferfs, we need to generate an endpoint wrapping
+	// that file. In this case, we can use Addr to set the endpoint address to
+	// "foo". Note that Addr is only a best-effort attempt--we still do not know
+	// the exact address that was used on the remote fs to bind the socket (it
+	// may have been "foo", "./foo", etc.).
+	Addr string
+}
+
 // GetxattrOptions contains options to VirtualFilesystem.GetxattrAt(),
 // FilesystemImpl.GetxattrAt(), FileDescription.Getxattr(), and
 // FileDescriptionImpl.Getxattr().