Implement flock(2) in VFS2

LockFD is the generic implementation that can be embedded in
FileDescriptionImpl implementations. Unique lock ID is
maintained in vfs.FileDescription and is created on demand.

Updates #1480

PiperOrigin-RevId: 315604825

4 files changed, 77 insertions, 30 deletions

diff --git a/pkg/sentry/syscalls/linux/sys_file.go b/pkg/sentry/syscalls/linux/sys_file.go
index 35a98212a..8347617bd 100644
--- a/pkg/sentry/syscalls/linux/sys_file.go
+++ b/pkg/sentry/syscalls/linux/sys_file.go
@@ -998,9 +998,6 @@ func Fcntl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall
 			return 0, nil, err
 		}
 
-		// The lock uid is that of the Task's FDTable.
-		lockUniqueID := lock.UniqueID(t.FDTable().ID())
-
 		// These locks don't block; execute the non-blocking operation using the inode's lock
 		// context directly.
 		switch flock.Type {
@@ -1010,12 +1007,12 @@ func Fcntl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall
 			}
 			if cmd == linux.F_SETLK {
 				// Non-blocking lock, provide a nil lock.Blocker.
-				if !file.Dirent.Inode.LockCtx.Posix.LockRegion(lockUniqueID, lock.ReadLock, rng, nil) {
+				if !file.Dirent.Inode.LockCtx.Posix.LockRegion(t.FDTable(), lock.ReadLock, rng, nil) {
 					return 0, nil, syserror.EAGAIN
 				}
 			} else {
 				// Blocking lock, pass in the task to satisfy the lock.Blocker interface.
-				if !file.Dirent.Inode.LockCtx.Posix.LockRegion(lockUniqueID, lock.ReadLock, rng, t) {
+				if !file.Dirent.Inode.LockCtx.Posix.LockRegion(t.FDTable(), lock.ReadLock, rng, t) {
 					return 0, nil, syserror.EINTR
 				}
 			}
@@ -1026,18 +1023,18 @@ func Fcntl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall
 			}
 			if cmd == linux.F_SETLK {
 				// Non-blocking lock, provide a nil lock.Blocker.
-				if !file.Dirent.Inode.LockCtx.Posix.LockRegion(lockUniqueID, lock.WriteLock, rng, nil) {
+				if !file.Dirent.Inode.LockCtx.Posix.LockRegion(t.FDTable(), lock.WriteLock, rng, nil) {
 					return 0, nil, syserror.EAGAIN
 				}
 			} else {
 				// Blocking lock, pass in the task to satisfy the lock.Blocker interface.
-				if !file.Dirent.Inode.LockCtx.Posix.LockRegion(lockUniqueID, lock.WriteLock, rng, t) {
+				if !file.Dirent.Inode.LockCtx.Posix.LockRegion(t.FDTable(), lock.WriteLock, rng, t) {
 					return 0, nil, syserror.EINTR
 				}
 			}
 			return 0, nil, nil
 		case linux.F_UNLCK:
-			file.Dirent.Inode.LockCtx.Posix.UnlockRegion(lockUniqueID, rng)
+			file.Dirent.Inode.LockCtx.Posix.UnlockRegion(t.FDTable(), rng)
 			return 0, nil, nil
 		default:
 			return 0, nil, syserror.EINVAL
@@ -2157,22 +2154,6 @@ func Flock(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall
 	nonblocking := operation&linux.LOCK_NB != 0
 	operation &^= linux.LOCK_NB
 
-	// flock(2):
-	// Locks created by flock() are associated with an open file table entry. This means that
-	// duplicate file descriptors (created by, for example, fork(2) or dup(2)) refer to the
-	// same lock, and this lock may be modified or released using any of these descriptors. Furthermore,
-	// the lock is released either by an explicit LOCK_UN operation on any of these duplicate
-	// descriptors, or when all such descriptors have been closed.
-	//
-	// If a process uses open(2) (or similar) to obtain more than one descriptor for the same file,
-	// these descriptors are treated independently by flock(). An attempt to lock the file using
-	// one of these file descriptors may be denied by a lock that the calling process has already placed via
-	// another descriptor.
-	//
-	// We use the File UniqueID as the lock UniqueID because it needs to reference the same lock across dup(2)
-	// and fork(2).
-	lockUniqueID := lock.UniqueID(file.UniqueID)
-
 	// A BSD style lock spans the entire file.
 	rng := lock.LockRange{
 		Start: 0,
@@ -2183,29 +2164,29 @@ func Flock(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall
 	case linux.LOCK_EX:
 		if nonblocking {
 			// Since we're nonblocking we pass a nil lock.Blocker implementation.
-			if !file.Dirent.Inode.LockCtx.BSD.LockRegion(lockUniqueID, lock.WriteLock, rng, nil) {
+			if !file.Dirent.Inode.LockCtx.BSD.LockRegion(file, lock.WriteLock, rng, nil) {
 				return 0, nil, syserror.EWOULDBLOCK
 			}
 		} else {
 			// Because we're blocking we will pass the task to satisfy the lock.Blocker interface.
-			if !file.Dirent.Inode.LockCtx.BSD.LockRegion(lockUniqueID, lock.WriteLock, rng, t) {
+			if !file.Dirent.Inode.LockCtx.BSD.LockRegion(file, lock.WriteLock, rng, t) {
 				return 0, nil, syserror.EINTR
 			}
 		}
 	case linux.LOCK_SH:
 		if nonblocking {
 			// Since we're nonblocking we pass a nil lock.Blocker implementation.
-			if !file.Dirent.Inode.LockCtx.BSD.LockRegion(lockUniqueID, lock.ReadLock, rng, nil) {
+			if !file.Dirent.Inode.LockCtx.BSD.LockRegion(file, lock.ReadLock, rng, nil) {
 				return 0, nil, syserror.EWOULDBLOCK
 			}
 		} else {
 			// Because we're blocking we will pass the task to satisfy the lock.Blocker interface.
-			if !file.Dirent.Inode.LockCtx.BSD.LockRegion(lockUniqueID, lock.ReadLock, rng, t) {
+			if !file.Dirent.Inode.LockCtx.BSD.LockRegion(file, lock.ReadLock, rng, t) {
 				return 0, nil, syserror.EINTR
 			}
 		}
 	case linux.LOCK_UN:
-		file.Dirent.Inode.LockCtx.BSD.UnlockRegion(lockUniqueID, rng)
+		file.Dirent.Inode.LockCtx.BSD.UnlockRegion(file, rng)
 	default:
 		// flock(2): EINVAL operation is invalid.
 		return 0, nil, syserror.EINVAL
diff --git a/pkg/sentry/syscalls/linux/vfs2/BUILD b/pkg/sentry/syscalls/linux/vfs2/BUILD
index c0d005247..9f93f4354 100644
--- a/pkg/sentry/syscalls/linux/vfs2/BUILD
+++ b/pkg/sentry/syscalls/linux/vfs2/BUILD
@@ -14,6 +14,7 @@ go_library(
         "getdents.go",
         "inotify.go",
         "ioctl.go",
+        "lock.go",
         "memfd.go",
         "mmap.go",
         "mount.go",
@@ -42,6 +43,7 @@ go_library(
         "//pkg/fspath",
         "//pkg/gohacks",
         "//pkg/sentry/arch",
+        "//pkg/sentry/fs/lock",
         "//pkg/sentry/fsbridge",
         "//pkg/sentry/fsimpl/eventfd",
         "//pkg/sentry/fsimpl/pipefs",
diff --git a/pkg/sentry/syscalls/linux/vfs2/lock.go b/pkg/sentry/syscalls/linux/vfs2/lock.go
new file mode 100644
index 000000000..bf19028c4
--- /dev/null
+++ b/pkg/sentry/syscalls/linux/vfs2/lock.go
@@ -0,0 +1,64 @@
+// Copyright 2020 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 vfs2
+
+import (
+	"gvisor.dev/gvisor/pkg/abi/linux"
+	"gvisor.dev/gvisor/pkg/sentry/arch"
+	"gvisor.dev/gvisor/pkg/sentry/fs/lock"
+	"gvisor.dev/gvisor/pkg/sentry/kernel"
+	"gvisor.dev/gvisor/pkg/syserror"
+)
+
+// Flock implements linux syscall flock(2).
+func Flock(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
+	fd := args[0].Int()
+	operation := args[1].Int()
+
+	file := t.GetFileVFS2(fd)
+	if file == nil {
+		// flock(2): EBADF fd is not an open file descriptor.
+		return 0, nil, syserror.EBADF
+	}
+	defer file.DecRef()
+
+	nonblocking := operation&linux.LOCK_NB != 0
+	operation &^= linux.LOCK_NB
+
+	var blocker lock.Blocker
+	if !nonblocking {
+		blocker = t
+	}
+
+	switch operation {
+	case linux.LOCK_EX:
+		if err := file.LockBSD(t, lock.WriteLock, blocker); err != nil {
+			return 0, nil, err
+		}
+	case linux.LOCK_SH:
+		if err := file.LockBSD(t, lock.ReadLock, blocker); err != nil {
+			return 0, nil, err
+		}
+	case linux.LOCK_UN:
+		if err := file.UnlockBSD(t); err != nil {
+			return 0, nil, err
+		}
+	default:
+		// flock(2): EINVAL operation is invalid.
+		return 0, nil, syserror.EINVAL
+	}
+
+	return 0, nil, nil
+}
diff --git a/pkg/sentry/syscalls/linux/vfs2/vfs2.go b/pkg/sentry/syscalls/linux/vfs2/vfs2.go
index 7b6e7571a..954c82f97 100644
--- a/pkg/sentry/syscalls/linux/vfs2/vfs2.go
+++ b/pkg/sentry/syscalls/linux/vfs2/vfs2.go
@@ -62,7 +62,7 @@ func Override() {
 	s.Table[55] = syscalls.Supported("getsockopt", GetSockOpt)
 	s.Table[59] = syscalls.Supported("execve", Execve)
 	s.Table[72] = syscalls.Supported("fcntl", Fcntl)
-	delete(s.Table, 73) // flock
+	s.Table[73] = syscalls.Supported("fcntl", Flock)
 	s.Table[74] = syscalls.Supported("fsync", Fsync)
 	s.Table[75] = syscalls.Supported("fdatasync", Fdatasync)
 	s.Table[76] = syscalls.Supported("truncate", Truncate)