diff options
Diffstat (limited to 'pkg/sentry/syscalls')
47 files changed, 548 insertions, 375 deletions
diff --git a/pkg/sentry/syscalls/BUILD b/pkg/sentry/syscalls/BUILD index 877318fa9..18fddee76 100644 --- a/pkg/sentry/syscalls/BUILD +++ b/pkg/sentry/syscalls/BUILD @@ -8,7 +8,7 @@ go_library( "epoll.go", "syscalls.go", ], - importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/syscalls", + importpath = "gvisor.dev/gvisor/pkg/sentry/syscalls", visibility = ["//:sandbox"], deps = [ "//pkg/abi/linux", diff --git a/pkg/sentry/syscalls/epoll.go b/pkg/sentry/syscalls/epoll.go index ec1eab331..c710ec9e3 100644 --- a/pkg/sentry/syscalls/epoll.go +++ b/pkg/sentry/syscalls/epoll.go @@ -18,11 +18,11 @@ import ( "syscall" "time" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/epoll" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" - ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" - "gvisor.googlesource.com/gvisor/pkg/waiter" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/epoll" + "gvisor.dev/gvisor/pkg/sentry/kernel/kdefs" + ktime "gvisor.dev/gvisor/pkg/sentry/kernel/time" + "gvisor.dev/gvisor/pkg/waiter" ) // CreateEpoll implements the epoll_create(2) linux syscall. diff --git a/pkg/sentry/syscalls/linux/BUILD b/pkg/sentry/syscalls/linux/BUILD index 1c057526b..6e5be0158 100644 --- a/pkg/sentry/syscalls/linux/BUILD +++ b/pkg/sentry/syscalls/linux/BUILD @@ -49,7 +49,7 @@ go_library( "sys_write.go", "timespec.go", ], - importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/syscalls/linux", + importpath = "gvisor.dev/gvisor/pkg/sentry/syscalls/linux", visibility = ["//:sandbox"], deps = [ "//pkg/abi", @@ -86,6 +86,7 @@ go_library( "//pkg/sentry/syscalls", "//pkg/sentry/usage", "//pkg/sentry/usermem", + "//pkg/syserr", "//pkg/syserror", "//pkg/waiter", ], diff --git a/pkg/sentry/syscalls/linux/error.go b/pkg/sentry/syscalls/linux/error.go index 72146ea63..ac3905c5c 100644 --- a/pkg/sentry/syscalls/linux/error.go +++ b/pkg/sentry/syscalls/linux/error.go @@ -19,12 +19,12 @@ import ( "sync" "syscall" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/log" - "gvisor.googlesource.com/gvisor/pkg/metric" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/log" + "gvisor.dev/gvisor/pkg/metric" + "gvisor.dev/gvisor/pkg/sentry/fs" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/syserror" ) var ( diff --git a/pkg/sentry/syscalls/linux/flags.go b/pkg/sentry/syscalls/linux/flags.go index d83e12971..0c1b5ec27 100644 --- a/pkg/sentry/syscalls/linux/flags.go +++ b/pkg/sentry/syscalls/linux/flags.go @@ -15,8 +15,8 @@ package linux import ( - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/fs" ) // flagsToPermissions returns a Permissions object from Linux flags. diff --git a/pkg/sentry/syscalls/linux/linux64.go b/pkg/sentry/syscalls/linux/linux64.go index 5251c2463..94816b038 100644 --- a/pkg/sentry/syscalls/linux/linux64.go +++ b/pkg/sentry/syscalls/linux/linux64.go @@ -18,13 +18,13 @@ package linux import ( "syscall" - "gvisor.googlesource.com/gvisor/pkg/abi" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/syscalls" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/syscalls" + "gvisor.dev/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/syserror" ) // AUDIT_ARCH_X86_64 identifies the Linux syscall API on AMD64, and is taken @@ -351,8 +351,8 @@ var AMD64 = &kernel.SyscallTable{ 300: syscalls.ErrorWithEvent("fanotify_init", syscall.ENOSYS, "Needs CONFIG_FANOTIFY", nil), 301: syscalls.ErrorWithEvent("fanotify_mark", syscall.ENOSYS, "Needs CONFIG_FANOTIFY", nil), 302: syscalls.Undocumented("prlimit64", Prlimit64), - 303: syscalls.ErrorWithEvent("name_to_handle_at", syscall.EOPNOTSUPP, "Needs filesystem support", nil), - 304: syscalls.ErrorWithEvent("open_by_handle_at", syscall.EOPNOTSUPP, "Needs filesystem support", nil), + 303: syscalls.Error("name_to_handle_at", syscall.EOPNOTSUPP, "Not supported by gVisor filesystems", nil), + 304: syscalls.Error("open_by_handle_at", syscall.EOPNOTSUPP, "Not supported by gVisor filesystems", nil), 305: syscalls.CapError("clock_adjtime", linux.CAP_SYS_TIME, "", nil), 306: syscalls.Undocumented("syncfs", Syncfs), 307: syscalls.Undocumented("sendmmsg", SendMMsg), @@ -379,6 +379,7 @@ var AMD64 = &kernel.SyscallTable{ 326: syscalls.ErrorWithEvent("copy_file_range", syscall.ENOSYS, "", nil), 327: syscalls.Undocumented("preadv2", Preadv2), 328: syscalls.Undocumented("pwritev2", Pwritev2), + 332: syscalls.Supported("statx", Statx), }, Emulate: map[usermem.Addr]uintptr{ diff --git a/pkg/sentry/syscalls/linux/sigset.go b/pkg/sentry/syscalls/linux/sigset.go index 5438b664b..00b7e7cf2 100644 --- a/pkg/sentry/syscalls/linux/sigset.go +++ b/pkg/sentry/syscalls/linux/sigset.go @@ -17,10 +17,10 @@ package linux import ( "syscall" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/syserror" ) // copyInSigSet copies in a sigset_t, checks its size, and ensures that KILL and diff --git a/pkg/sentry/syscalls/linux/sys_aio.go b/pkg/sentry/syscalls/linux/sys_aio.go index 1b27b2415..7081d1a45 100644 --- a/pkg/sentry/syscalls/linux/sys_aio.go +++ b/pkg/sentry/syscalls/linux/sys_aio.go @@ -17,15 +17,15 @@ package linux import ( "encoding/binary" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/eventfd" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" - ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" - "gvisor.googlesource.com/gvisor/pkg/sentry/mm" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/fs" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/eventfd" + "gvisor.dev/gvisor/pkg/sentry/kernel/kdefs" + ktime "gvisor.dev/gvisor/pkg/sentry/kernel/time" + "gvisor.dev/gvisor/pkg/sentry/mm" + "gvisor.dev/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/syserror" ) // I/O commands. diff --git a/pkg/sentry/syscalls/linux/sys_capability.go b/pkg/sentry/syscalls/linux/sys_capability.go index 622cb8d0d..adf5ea5f2 100644 --- a/pkg/sentry/syscalls/linux/sys_capability.go +++ b/pkg/sentry/syscalls/linux/sys_capability.go @@ -15,11 +15,11 @@ package linux import ( - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/auth" + "gvisor.dev/gvisor/pkg/syserror" ) func lookupCaps(t *kernel.Task, tid kernel.ThreadID) (permitted, inheritable, effective auth.CapabilitySet, err error) { diff --git a/pkg/sentry/syscalls/linux/sys_epoll.go b/pkg/sentry/syscalls/linux/sys_epoll.go index 1467feb4e..14a61cfa5 100644 --- a/pkg/sentry/syscalls/linux/sys_epoll.go +++ b/pkg/sentry/syscalls/linux/sys_epoll.go @@ -17,14 +17,14 @@ package linux import ( "syscall" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/epoll" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" - "gvisor.googlesource.com/gvisor/pkg/sentry/syscalls" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" - "gvisor.googlesource.com/gvisor/pkg/waiter" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/epoll" + "gvisor.dev/gvisor/pkg/sentry/kernel/kdefs" + "gvisor.dev/gvisor/pkg/sentry/syscalls" + "gvisor.dev/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/waiter" ) // EpollCreate1 implements the epoll_create1(2) linux syscall. diff --git a/pkg/sentry/syscalls/linux/sys_eventfd.go b/pkg/sentry/syscalls/linux/sys_eventfd.go index ca4ead488..7dbe84884 100644 --- a/pkg/sentry/syscalls/linux/sys_eventfd.go +++ b/pkg/sentry/syscalls/linux/sys_eventfd.go @@ -17,10 +17,10 @@ package linux import ( "syscall" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/eventfd" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/fs" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/eventfd" ) const ( diff --git a/pkg/sentry/syscalls/linux/sys_file.go b/pkg/sentry/syscalls/linux/sys_file.go index 19f579930..2776fdec7 100644 --- a/pkg/sentry/syscalls/linux/sys_file.go +++ b/pkg/sentry/syscalls/linux/sys_file.go @@ -17,31 +17,30 @@ package linux import ( "syscall" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/context" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs/lock" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs/tmpfs" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/fasync" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/pipe" - ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" - "gvisor.googlesource.com/gvisor/pkg/sentry/limits" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/context" + "gvisor.dev/gvisor/pkg/sentry/fs" + "gvisor.dev/gvisor/pkg/sentry/fs/lock" + "gvisor.dev/gvisor/pkg/sentry/fs/tmpfs" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/auth" + "gvisor.dev/gvisor/pkg/sentry/kernel/fasync" + "gvisor.dev/gvisor/pkg/sentry/kernel/kdefs" + ktime "gvisor.dev/gvisor/pkg/sentry/kernel/time" + "gvisor.dev/gvisor/pkg/sentry/limits" + "gvisor.dev/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/syserror" ) // fileOpAt performs an operation on the second last component in the path. -func fileOpAt(t *kernel.Task, dirFD kdefs.FD, path string, fn func(root *fs.Dirent, d *fs.Dirent, name string) error) error { +func fileOpAt(t *kernel.Task, dirFD kdefs.FD, path string, fn func(root *fs.Dirent, d *fs.Dirent, name string, remainingTraversals uint) error) error { // Extract the last component. dir, name := fs.SplitLast(path) if dir == "/" { // Common case: we are accessing a file in the root. root := t.FSContext().RootDirectory() - err := fn(root, root, name) + err := fn(root, root, name, linux.MaxSymlinkTraversals) root.DecRef() return err } else if dir == "." && dirFD == linux.AT_FDCWD { @@ -49,19 +48,19 @@ func fileOpAt(t *kernel.Task, dirFD kdefs.FD, path string, fn func(root *fs.Dire // working directory; skip the look-up. wd := t.FSContext().WorkingDirectory() root := t.FSContext().RootDirectory() - err := fn(root, wd, name) + err := fn(root, wd, name, linux.MaxSymlinkTraversals) wd.DecRef() root.DecRef() return err } - return fileOpOn(t, dirFD, dir, true /* resolve */, func(root *fs.Dirent, d *fs.Dirent) error { - return fn(root, d, name) + return fileOpOn(t, dirFD, dir, true /* resolve */, func(root *fs.Dirent, d *fs.Dirent, remainingTraversals uint) error { + return fn(root, d, name, remainingTraversals) }) } // fileOpOn performs an operation on the last entry of the path. -func fileOpOn(t *kernel.Task, dirFD kdefs.FD, path string, resolve bool, fn func(root *fs.Dirent, d *fs.Dirent) error) error { +func fileOpOn(t *kernel.Task, dirFD kdefs.FD, path string, resolve bool, fn func(root *fs.Dirent, d *fs.Dirent, remainingTraversals uint) error) error { var ( d *fs.Dirent // The file. wd *fs.Dirent // The working directory (if required.) @@ -110,7 +109,7 @@ func fileOpOn(t *kernel.Task, dirFD kdefs.FD, path string, resolve bool, fn func return err } - err = fn(root, d) + err = fn(root, d, remainingTraversals) d.DecRef() return err } @@ -139,7 +138,7 @@ func openAt(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr, flags uint) (fd u } resolve := flags&linux.O_NOFOLLOW == 0 - err = fileOpOn(t, dirFD, path, resolve, func(root *fs.Dirent, d *fs.Dirent) error { + err = fileOpOn(t, dirFD, path, resolve, func(root *fs.Dirent, d *fs.Dirent, _ uint) error { // First check a few things about the filesystem before trying to get the file // reference. // @@ -211,7 +210,7 @@ func mknodAt(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr, mode linux.FileM return syserror.ENOENT } - return fileOpAt(t, dirFD, path, func(root *fs.Dirent, d *fs.Dirent, name string) error { + return fileOpAt(t, dirFD, path, func(root *fs.Dirent, d *fs.Dirent, name string, _ uint) error { if !fs.IsDir(d.Inode.StableAttr) { return syserror.ENOTDIR } @@ -304,45 +303,105 @@ func createAt(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr, flags uint, mod return 0, syserror.ENOENT } - err = fileOpAt(t, dirFD, path, func(root *fs.Dirent, d *fs.Dirent, name string) error { - if !fs.IsDir(d.Inode.StableAttr) { - return syserror.ENOTDIR - } - - fileFlags := linuxToFlags(flags) - // Linux always adds the O_LARGEFILE flag when running in 64-bit mode. - fileFlags.LargeFile = true + fileFlags := linuxToFlags(flags) + // Linux always adds the O_LARGEFILE flag when running in 64-bit mode. + fileFlags.LargeFile = true + + err = fileOpAt(t, dirFD, path, func(root *fs.Dirent, parent *fs.Dirent, name string, remainingTraversals uint) error { + // Resolve the name to see if it exists, and follow any + // symlinks along the way. We must do the symlink resolution + // manually because if the symlink target does not exist, we + // must create the target (and not the symlink itself). + var ( + found *fs.Dirent + err error + ) + for { + if !fs.IsDir(parent.Inode.StableAttr) { + return syserror.ENOTDIR + } - // Does this file exist already? - remainingTraversals := uint(linux.MaxSymlinkTraversals) - targetDirent, err := t.MountNamespace().FindInode(t, root, d, name, &remainingTraversals) - var newFile *fs.File - switch err { - case nil: - // The file existed. - defer targetDirent.DecRef() + // Start by looking up the dirent at 'name'. + found, err = t.MountNamespace().FindLink(t, root, parent, name, &remainingTraversals) + if err != nil { + break + } + defer found.DecRef() - // Check if we wanted to create. + // We found something (possibly a symlink). If the + // O_EXCL flag was passed, then we can immediately + // return EEXIST. if flags&linux.O_EXCL != 0 { return syserror.EEXIST } + // If we have a non-symlink, then we can proceed. + if !fs.IsSymlink(found.Inode.StableAttr) { + break + } + + // If O_NOFOLLOW was passed, then don't try to resolve + // anything. + if flags&linux.O_NOFOLLOW != 0 { + return syserror.ELOOP + } + + // Try to resolve the symlink directly to a Dirent. + var resolved *fs.Dirent + resolved, err = found.Inode.Getlink(t) + if err == nil { + // No more resolution necessary. + defer resolved.DecRef() + break + } else if err != fs.ErrResolveViaReadlink { + return err + } + + // Are we able to resolve further? + if remainingTraversals == 0 { + return syscall.ELOOP + } + + // Resolve the symlink to a path via Readlink. + path, err := found.Inode.Readlink(t) + if err != nil { + break + } + remainingTraversals-- + + // Get the new parent from the target path. + newParentPath, newName := fs.SplitLast(path) + newParent, err := t.MountNamespace().FindInode(t, root, parent, newParentPath, &remainingTraversals) + if err != nil { + break + } + defer newParent.DecRef() + + // Repeat the process with the parent and name of the + // symlink target. + parent = newParent + name = newName + } + + var newFile *fs.File + switch err { + case nil: // Like sys_open, check for a few things about the // filesystem before trying to get a reference to the // fs.File. The same constraints on Check apply. - if err := targetDirent.Inode.CheckPermission(t, flagsToPermissions(flags)); err != nil { + if err := found.Inode.CheckPermission(t, flagsToPermissions(flags)); err != nil { return err } // Should we truncate the file? if flags&linux.O_TRUNC != 0 { - if err := targetDirent.Inode.Truncate(t, targetDirent, 0); err != nil { + if err := found.Inode.Truncate(t, found, 0); err != nil { return err } } // Create a new fs.File. - newFile, err = targetDirent.Inode.GetFile(t, targetDirent, fileFlags) + newFile, err = found.Inode.GetFile(t, found, fileFlags) if err != nil { return syserror.ConvertIntr(err, kernel.ERESTARTSYS) } @@ -351,19 +410,19 @@ func createAt(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr, flags uint, mod // File does not exist. Proceed with creation. // Do we have write permissions on the parent? - if err := d.Inode.CheckPermission(t, fs.PermMask{Write: true, Execute: true}); err != nil { + if err := parent.Inode.CheckPermission(t, fs.PermMask{Write: true, Execute: true}); err != nil { return err } // Attempt a creation. perms := fs.FilePermsFromMode(mode &^ linux.FileMode(t.FSContext().Umask())) - newFile, err = d.Create(t, root, name, fileFlags, perms) + newFile, err = parent.Create(t, root, name, fileFlags, perms) if err != nil { // No luck, bail. return err } defer newFile.DecRef() - targetDirent = newFile.Dirent + found = newFile.Dirent default: return err } @@ -379,10 +438,10 @@ func createAt(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr, flags uint, mod fd = uintptr(newFD) // Queue the open inotify event. The creation event is - // automatically queued when the dirent is targetDirent. The - // open events are implemented at the syscall layer so we need - // to manually queue one here. - targetDirent.InotifyEvent(linux.IN_OPEN, 0) + // automatically queued when the dirent is found. The open + // events are implemented at the syscall layer so we need to + // manually queue one here. + found.InotifyEvent(linux.IN_OPEN, 0) return nil }) @@ -458,7 +517,7 @@ func accessAt(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr, resolve bool, m return syserror.EINVAL } - return fileOpOn(t, dirFD, path, resolve, func(root *fs.Dirent, d *fs.Dirent) error { + return fileOpOn(t, dirFD, path, resolve, func(root *fs.Dirent, d *fs.Dirent, _ uint) error { // access(2) and faccessat(2) check permissions using real // UID/GID, not effective UID/GID. // @@ -572,7 +631,7 @@ func Ioctl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall return 0, nil, err default: - ret, err := file.FileOperations.Ioctl(t, t.MemoryManager(), args) + ret, err := file.FileOperations.Ioctl(t, file, t.MemoryManager(), args) if err != nil { return 0, nil, err } @@ -626,7 +685,7 @@ func Chroot(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal return 0, nil, err } - return 0, nil, fileOpOn(t, linux.AT_FDCWD, path, true /* resolve */, func(root *fs.Dirent, d *fs.Dirent) error { + return 0, nil, fileOpOn(t, linux.AT_FDCWD, path, true /* resolve */, func(root *fs.Dirent, d *fs.Dirent, _ uint) error { // Is it a directory? if !fs.IsDir(d.Inode.StableAttr) { return syserror.ENOTDIR @@ -651,7 +710,7 @@ func Chdir(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall return 0, nil, err } - return 0, nil, fileOpOn(t, linux.AT_FDCWD, path, true /* resolve */, func(root *fs.Dirent, d *fs.Dirent) error { + return 0, nil, fileOpOn(t, linux.AT_FDCWD, path, true /* resolve */, func(root *fs.Dirent, d *fs.Dirent, _ uint) error { // Is it a directory? if !fs.IsDir(d.Inode.StableAttr) { return syserror.ENOTDIR @@ -945,17 +1004,18 @@ func Fcntl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall err := tmpfs.AddSeals(file.Dirent.Inode, args[2].Uint()) return 0, nil, err case linux.F_GETPIPE_SZ: - sz, ok := file.FileOperations.(pipe.Sizer) + sz, ok := file.FileOperations.(fs.FifoSizer) if !ok { return 0, nil, syserror.EINVAL } - return uintptr(sz.PipeSize()), nil, nil + size, err := sz.FifoSize(t, file) + return uintptr(size), nil, err case linux.F_SETPIPE_SZ: - sz, ok := file.FileOperations.(pipe.Sizer) + sz, ok := file.FileOperations.(fs.FifoSizer) if !ok { return 0, nil, syserror.EINVAL } - n, err := sz.SetPipeSize(int64(args[2].Int())) + n, err := sz.SetFifoSize(int64(args[2].Int())) return uintptr(n), nil, err default: // Everything else is not yet supported. @@ -1017,7 +1077,7 @@ func mkdirAt(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr, mode linux.FileM return err } - return fileOpAt(t, dirFD, path, func(root *fs.Dirent, d *fs.Dirent, name string) error { + return fileOpAt(t, dirFD, path, func(root *fs.Dirent, d *fs.Dirent, name string, _ uint) error { if !fs.IsDir(d.Inode.StableAttr) { return syserror.ENOTDIR } @@ -1074,7 +1134,7 @@ func rmdirAt(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr) error { return syserror.EBUSY } - return fileOpAt(t, dirFD, path, func(root *fs.Dirent, d *fs.Dirent, name string) error { + return fileOpAt(t, dirFD, path, func(root *fs.Dirent, d *fs.Dirent, name string, _ uint) error { if !fs.IsDir(d.Inode.StableAttr) { return syserror.ENOTDIR } @@ -1122,7 +1182,7 @@ func symlinkAt(t *kernel.Task, dirFD kdefs.FD, newAddr usermem.Addr, oldAddr use return syserror.ENOENT } - return fileOpAt(t, dirFD, newPath, func(root *fs.Dirent, d *fs.Dirent, name string) error { + return fileOpAt(t, dirFD, newPath, func(root *fs.Dirent, d *fs.Dirent, name string, _ uint) error { if !fs.IsDir(d.Inode.StableAttr) { return syserror.ENOTDIR } @@ -1212,7 +1272,7 @@ func linkAt(t *kernel.Task, oldDirFD kdefs.FD, oldAddr usermem.Addr, newDirFD kd } // Resolve the target directory. - return fileOpAt(t, newDirFD, newPath, func(root *fs.Dirent, newParent *fs.Dirent, newName string) error { + return fileOpAt(t, newDirFD, newPath, func(root *fs.Dirent, newParent *fs.Dirent, newName string, _ uint) error { if !fs.IsDir(newParent.Inode.StableAttr) { return syserror.ENOTDIR } @@ -1227,13 +1287,13 @@ func linkAt(t *kernel.Task, oldDirFD kdefs.FD, oldAddr usermem.Addr, newDirFD kd // Resolve oldDirFD and oldAddr to a dirent. The "resolve" argument // only applies to this name. - return fileOpOn(t, oldDirFD, oldPath, resolve, func(root *fs.Dirent, target *fs.Dirent) error { + return fileOpOn(t, oldDirFD, oldPath, resolve, func(root *fs.Dirent, target *fs.Dirent, _ uint) error { if err := mayLinkAt(t, target.Inode); err != nil { return err } // Next resolve newDirFD and newAddr to the parent dirent and name. - return fileOpAt(t, newDirFD, newPath, func(root *fs.Dirent, newParent *fs.Dirent, newName string) error { + return fileOpAt(t, newDirFD, newPath, func(root *fs.Dirent, newParent *fs.Dirent, newName string, _ uint) error { if !fs.IsDir(newParent.Inode.StableAttr) { return syserror.ENOTDIR } @@ -1300,7 +1360,7 @@ func readlinkAt(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr, bufAddr userm return 0, syserror.ENOENT } - err = fileOpOn(t, dirFD, path, false /* resolve */, func(root *fs.Dirent, d *fs.Dirent) error { + err = fileOpOn(t, dirFD, path, false /* resolve */, func(root *fs.Dirent, d *fs.Dirent, _ uint) error { // Check for Read permission. if err := d.Inode.CheckPermission(t, fs.PermMask{Read: true}); err != nil { return err @@ -1359,7 +1419,7 @@ func unlinkAt(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr) error { return syserror.ENOENT } - return fileOpAt(t, dirFD, path, func(root *fs.Dirent, d *fs.Dirent, name string) error { + return fileOpAt(t, dirFD, path, func(root *fs.Dirent, d *fs.Dirent, name string, _ uint) error { if !fs.IsDir(d.Inode.StableAttr) { return syserror.ENOTDIR } @@ -1414,7 +1474,7 @@ func Truncate(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysc return 0, nil, syserror.EFBIG } - return 0, nil, fileOpOn(t, linux.AT_FDCWD, path, true /* resolve */, func(root *fs.Dirent, d *fs.Dirent) error { + return 0, nil, fileOpOn(t, linux.AT_FDCWD, path, true /* resolve */, func(root *fs.Dirent, d *fs.Dirent, _ uint) error { if fs.IsDir(d.Inode.StableAttr) { return syserror.EISDIR } @@ -1576,7 +1636,7 @@ func chownAt(t *kernel.Task, fd kdefs.FD, addr usermem.Addr, resolve, allowEmpty return chown(t, file.Dirent, uid, gid) } - return fileOpOn(t, fd, path, resolve, func(root *fs.Dirent, d *fs.Dirent) error { + return fileOpOn(t, fd, path, resolve, func(root *fs.Dirent, d *fs.Dirent, _ uint) error { return chown(t, d, uid, gid) }) } @@ -1652,7 +1712,7 @@ func chmodAt(t *kernel.Task, fd kdefs.FD, addr usermem.Addr, mode linux.FileMode return err } - return fileOpOn(t, fd, path, true /* resolve */, func(root *fs.Dirent, d *fs.Dirent) error { + return fileOpOn(t, fd, path, true /* resolve */, func(root *fs.Dirent, d *fs.Dirent, _ uint) error { return chmod(t, d, mode) }) } @@ -1698,7 +1758,7 @@ func defaultSetToSystemTimeSpec() fs.TimeSpec { } func utimes(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr, ts fs.TimeSpec, resolve bool) error { - setTimestamp := func(root *fs.Dirent, d *fs.Dirent) error { + setTimestamp := func(root *fs.Dirent, d *fs.Dirent, _ uint) error { // Does the task own the file? if !d.Inode.CheckOwnership(t) { // Trying to set a specific time? Must be owner. @@ -1739,7 +1799,7 @@ func utimes(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr, ts fs.TimeSpec, r root := t.FSContext().RootDirectory() defer root.DecRef() - return setTimestamp(root, f.Dirent) + return setTimestamp(root, f.Dirent, linux.MaxSymlinkTraversals) } path, _, err := copyInPath(t, addr, false /* allowEmpty */) @@ -1867,7 +1927,7 @@ func renameAt(t *kernel.Task, oldDirFD kdefs.FD, oldAddr usermem.Addr, newDirFD return err } - return fileOpAt(t, oldDirFD, oldPath, func(root *fs.Dirent, oldParent *fs.Dirent, oldName string) error { + return fileOpAt(t, oldDirFD, oldPath, func(root *fs.Dirent, oldParent *fs.Dirent, oldName string, _ uint) error { if !fs.IsDir(oldParent.Inode.StableAttr) { return syserror.ENOTDIR } @@ -1879,7 +1939,7 @@ func renameAt(t *kernel.Task, oldDirFD kdefs.FD, oldAddr usermem.Addr, newDirFD return syserror.EBUSY } - return fileOpAt(t, newDirFD, newPath, func(root *fs.Dirent, newParent *fs.Dirent, newName string) error { + return fileOpAt(t, newDirFD, newPath, func(root *fs.Dirent, newParent *fs.Dirent, newName string, _ uint) error { if !fs.IsDir(newParent.Inode.StableAttr) { return syserror.ENOTDIR } @@ -2067,7 +2127,7 @@ func MemfdCreate(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.S name = memfdPrefix + name inode := tmpfs.NewMemfdInode(t, allowSeals) - dirent := fs.NewDirent(inode, name) + dirent := fs.NewDirent(t, inode, name) // Per Linux, mm/shmem.c:__shmem_file_setup(), memfd files are set up with // FMODE_READ | FMODE_WRITE. file, err := inode.GetFile(t, dirent, fs.FileFlags{Read: true, Write: true}) diff --git a/pkg/sentry/syscalls/linux/sys_futex.go b/pkg/sentry/syscalls/linux/sys_futex.go index 7cef4b50c..b9bd25464 100644 --- a/pkg/sentry/syscalls/linux/sys_futex.go +++ b/pkg/sentry/syscalls/linux/sys_futex.go @@ -17,12 +17,12 @@ package linux import ( "time" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" + ktime "gvisor.dev/gvisor/pkg/sentry/kernel/time" + "gvisor.dev/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/syserror" ) // futexWaitRestartBlock encapsulates the state required to restart futex(2) diff --git a/pkg/sentry/syscalls/linux/sys_getdents.go b/pkg/sentry/syscalls/linux/sys_getdents.go index 1b597d5bc..dea872672 100644 --- a/pkg/sentry/syscalls/linux/sys_getdents.go +++ b/pkg/sentry/syscalls/linux/sys_getdents.go @@ -19,13 +19,13 @@ import ( "io" "syscall" - "gvisor.googlesource.com/gvisor/pkg/binary" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/binary" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/fs" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/kdefs" + "gvisor.dev/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/syserror" ) // Getdents implements linux syscall getdents(2) for 64bit systems. diff --git a/pkg/sentry/syscalls/linux/sys_identity.go b/pkg/sentry/syscalls/linux/sys_identity.go index 27e765a2d..715ac45e6 100644 --- a/pkg/sentry/syscalls/linux/sys_identity.go +++ b/pkg/sentry/syscalls/linux/sys_identity.go @@ -15,10 +15,10 @@ package linux import ( - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/auth" + "gvisor.dev/gvisor/pkg/syserror" ) const ( diff --git a/pkg/sentry/syscalls/linux/sys_inotify.go b/pkg/sentry/syscalls/linux/sys_inotify.go index 20269a769..9cfa660fa 100644 --- a/pkg/sentry/syscalls/linux/sys_inotify.go +++ b/pkg/sentry/syscalls/linux/sys_inotify.go @@ -17,12 +17,12 @@ package linux import ( "syscall" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs/anon" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/fs" + "gvisor.dev/gvisor/pkg/sentry/fs/anon" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/kdefs" ) const allFlags = int(linux.IN_NONBLOCK | linux.IN_CLOEXEC) @@ -35,7 +35,7 @@ func InotifyInit1(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel. return 0, nil, syscall.EINVAL } - dirent := fs.NewDirent(anon.NewInode(t), "inotify") + dirent := fs.NewDirent(t, anon.NewInode(t), "inotify") fileFlags := fs.FileFlags{ Read: true, Write: true, @@ -107,7 +107,7 @@ func InotifyAddWatch(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kern return 0, nil, err } - err = fileOpOn(t, linux.AT_FDCWD, path, resolve, func(root *fs.Dirent, dirent *fs.Dirent) error { + err = fileOpOn(t, linux.AT_FDCWD, path, resolve, func(root *fs.Dirent, dirent *fs.Dirent, _ uint) error { // "IN_ONLYDIR: Only watch pathname if it is a directory." -- inotify(7) if onlyDir := mask&linux.IN_ONLYDIR != 0; onlyDir && !fs.IsDir(dirent.Inode.StableAttr) { return syscall.ENOTDIR diff --git a/pkg/sentry/syscalls/linux/sys_lseek.go b/pkg/sentry/syscalls/linux/sys_lseek.go index 8aadc6d8c..a3813b818 100644 --- a/pkg/sentry/syscalls/linux/sys_lseek.go +++ b/pkg/sentry/syscalls/linux/sys_lseek.go @@ -15,11 +15,11 @@ package linux import ( - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/fs" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/kdefs" + "gvisor.dev/gvisor/pkg/syserror" ) // Lseek implements linux syscall lseek(2). diff --git a/pkg/sentry/syscalls/linux/sys_mempolicy.go b/pkg/sentry/syscalls/linux/sys_mempolicy.go index 652b2c206..f5a519d8a 100644 --- a/pkg/sentry/syscalls/linux/sys_mempolicy.go +++ b/pkg/sentry/syscalls/linux/sys_mempolicy.go @@ -17,11 +17,11 @@ package linux import ( "fmt" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/syserror" ) // We unconditionally report a single NUMA node. This also means that our diff --git a/pkg/sentry/syscalls/linux/sys_mmap.go b/pkg/sentry/syscalls/linux/sys_mmap.go index 9926f0ac5..d831833bc 100644 --- a/pkg/sentry/syscalls/linux/sys_mmap.go +++ b/pkg/sentry/syscalls/linux/sys_mmap.go @@ -17,14 +17,14 @@ package linux import ( "bytes" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" - "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" - "gvisor.googlesource.com/gvisor/pkg/sentry/mm" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/kdefs" + "gvisor.dev/gvisor/pkg/sentry/memmap" + "gvisor.dev/gvisor/pkg/sentry/mm" + "gvisor.dev/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/syserror" ) // Brk implements linux syscall brk(2). @@ -180,6 +180,10 @@ func Madvise(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysca switch adv { case linux.MADV_DONTNEED: return 0, nil, t.MemoryManager().Decommit(addr, length) + case linux.MADV_DOFORK: + return 0, nil, t.MemoryManager().SetDontFork(addr, length, false) + case linux.MADV_DONTFORK: + return 0, nil, t.MemoryManager().SetDontFork(addr, length, true) case linux.MADV_HUGEPAGE, linux.MADV_NOHUGEPAGE: fallthrough case linux.MADV_MERGEABLE, linux.MADV_UNMERGEABLE: @@ -191,7 +195,7 @@ func Madvise(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysca case linux.MADV_NORMAL, linux.MADV_RANDOM, linux.MADV_SEQUENTIAL, linux.MADV_WILLNEED: // Do nothing, we totally ignore the suggestions above. return 0, nil, nil - case linux.MADV_REMOVE, linux.MADV_DOFORK, linux.MADV_DONTFORK: + case linux.MADV_REMOVE: // These "suggestions" have application-visible side effects, so we // have to indicate that we don't support them. return 0, nil, syserror.ENOSYS diff --git a/pkg/sentry/syscalls/linux/sys_mount.go b/pkg/sentry/syscalls/linux/sys_mount.go index cf613bad0..9080a10c3 100644 --- a/pkg/sentry/syscalls/linux/sys_mount.go +++ b/pkg/sentry/syscalls/linux/sys_mount.go @@ -15,12 +15,12 @@ package linux import ( - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/fs" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/syserror" ) // Mount implements Linux syscall mount(2). @@ -109,7 +109,7 @@ func Mount(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall return 0, nil, syserror.EINVAL } - return 0, nil, fileOpOn(t, linux.AT_FDCWD, targetPath, true /* resolve */, func(root *fs.Dirent, d *fs.Dirent) error { + return 0, nil, fileOpOn(t, linux.AT_FDCWD, targetPath, true /* resolve */, func(root *fs.Dirent, d *fs.Dirent, _ uint) error { return t.MountNamespace().Mount(t, d, rootInode) }) } @@ -140,7 +140,7 @@ func Umount2(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysca resolve := flags&linux.UMOUNT_NOFOLLOW != linux.UMOUNT_NOFOLLOW detachOnly := flags&linux.MNT_DETACH == linux.MNT_DETACH - return 0, nil, fileOpOn(t, linux.AT_FDCWD, path, resolve, func(root *fs.Dirent, d *fs.Dirent) error { + return 0, nil, fileOpOn(t, linux.AT_FDCWD, path, resolve, func(root *fs.Dirent, d *fs.Dirent, _ uint) error { return t.MountNamespace().Unmount(t, d, detachOnly) }) } diff --git a/pkg/sentry/syscalls/linux/sys_pipe.go b/pkg/sentry/syscalls/linux/sys_pipe.go index 036845c13..7c1bea43d 100644 --- a/pkg/sentry/syscalls/linux/sys_pipe.go +++ b/pkg/sentry/syscalls/linux/sys_pipe.go @@ -17,12 +17,12 @@ package linux import ( "syscall" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/pipe" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/kdefs" + "gvisor.dev/gvisor/pkg/sentry/kernel/pipe" + "gvisor.dev/gvisor/pkg/sentry/usermem" ) // pipe2 implements the actual system call with flags. diff --git a/pkg/sentry/syscalls/linux/sys_poll.go b/pkg/sentry/syscalls/linux/sys_poll.go index e32099dd4..ef6211218 100644 --- a/pkg/sentry/syscalls/linux/sys_poll.go +++ b/pkg/sentry/syscalls/linux/sys_poll.go @@ -17,16 +17,16 @@ package linux import ( "time" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" - ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" - "gvisor.googlesource.com/gvisor/pkg/sentry/limits" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" - "gvisor.googlesource.com/gvisor/pkg/waiter" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/fs" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/kdefs" + ktime "gvisor.dev/gvisor/pkg/sentry/kernel/time" + "gvisor.dev/gvisor/pkg/sentry/limits" + "gvisor.dev/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/waiter" ) // fileCap is the maximum allowable files for poll & select. diff --git a/pkg/sentry/syscalls/linux/sys_prctl.go b/pkg/sentry/syscalls/linux/sys_prctl.go index 1b7e5616b..9d70881fd 100644 --- a/pkg/sentry/syscalls/linux/sys_prctl.go +++ b/pkg/sentry/syscalls/linux/sys_prctl.go @@ -18,13 +18,13 @@ import ( "fmt" "syscall" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" - "gvisor.googlesource.com/gvisor/pkg/sentry/mm" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/fs" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/auth" + "gvisor.dev/gvisor/pkg/sentry/kernel/kdefs" + "gvisor.dev/gvisor/pkg/sentry/mm" ) // Prctl implements linux syscall prctl(2). diff --git a/pkg/sentry/syscalls/linux/sys_random.go b/pkg/sentry/syscalls/linux/sys_random.go index fc3959a7e..bc4c588bf 100644 --- a/pkg/sentry/syscalls/linux/sys_random.go +++ b/pkg/sentry/syscalls/linux/sys_random.go @@ -18,12 +18,12 @@ import ( "io" "math" - "gvisor.googlesource.com/gvisor/pkg/rand" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/rand" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/safemem" + "gvisor.dev/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/syserror" ) const ( diff --git a/pkg/sentry/syscalls/linux/sys_read.go b/pkg/sentry/syscalls/linux/sys_read.go index 48b0fd49d..a1965f490 100644 --- a/pkg/sentry/syscalls/linux/sys_read.go +++ b/pkg/sentry/syscalls/linux/sys_read.go @@ -17,16 +17,16 @@ package linux import ( "time" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" - ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" - "gvisor.googlesource.com/gvisor/pkg/sentry/socket" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" - "gvisor.googlesource.com/gvisor/pkg/waiter" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/fs" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/kdefs" + ktime "gvisor.dev/gvisor/pkg/sentry/kernel/time" + "gvisor.dev/gvisor/pkg/sentry/socket" + "gvisor.dev/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/waiter" ) const ( diff --git a/pkg/sentry/syscalls/linux/sys_rlimit.go b/pkg/sentry/syscalls/linux/sys_rlimit.go index 8b0379779..51e3f836b 100644 --- a/pkg/sentry/syscalls/linux/sys_rlimit.go +++ b/pkg/sentry/syscalls/linux/sys_rlimit.go @@ -15,12 +15,12 @@ package linux import ( - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/limits" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/limits" + "gvisor.dev/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/syserror" ) // rlimit describes an implementation of 'struct rlimit', which may vary from diff --git a/pkg/sentry/syscalls/linux/sys_rusage.go b/pkg/sentry/syscalls/linux/sys_rusage.go index 003d718da..1674c7445 100644 --- a/pkg/sentry/syscalls/linux/sys_rusage.go +++ b/pkg/sentry/syscalls/linux/sys_rusage.go @@ -15,12 +15,12 @@ package linux import ( - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" - "gvisor.googlesource.com/gvisor/pkg/sentry/usage" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" + ktime "gvisor.dev/gvisor/pkg/sentry/kernel/time" + "gvisor.dev/gvisor/pkg/sentry/usage" + "gvisor.dev/gvisor/pkg/syserror" ) func getrusage(t *kernel.Task, which int32) linux.Rusage { diff --git a/pkg/sentry/syscalls/linux/sys_sched.go b/pkg/sentry/syscalls/linux/sys_sched.go index 8aea03abe..434bbb322 100644 --- a/pkg/sentry/syscalls/linux/sys_sched.go +++ b/pkg/sentry/syscalls/linux/sys_sched.go @@ -17,9 +17,9 @@ package linux import ( "syscall" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" ) const ( diff --git a/pkg/sentry/syscalls/linux/sys_seccomp.go b/pkg/sentry/syscalls/linux/sys_seccomp.go index b4262162a..4885b5e40 100644 --- a/pkg/sentry/syscalls/linux/sys_seccomp.go +++ b/pkg/sentry/syscalls/linux/sys_seccomp.go @@ -17,11 +17,11 @@ package linux import ( "syscall" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/bpf" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/bpf" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/usermem" ) // userSockFprog is equivalent to Linux's struct sock_fprog on amd64. diff --git a/pkg/sentry/syscalls/linux/sys_sem.go b/pkg/sentry/syscalls/linux/sys_sem.go index 5bd61ab87..cde3b54e7 100644 --- a/pkg/sentry/syscalls/linux/sys_sem.go +++ b/pkg/sentry/syscalls/linux/sys_sem.go @@ -17,13 +17,13 @@ package linux import ( "math" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/fs" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/auth" + "gvisor.dev/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/syserror" ) const opsMax = 500 // SEMOPM diff --git a/pkg/sentry/syscalls/linux/sys_shm.go b/pkg/sentry/syscalls/linux/sys_shm.go index d0eceac7c..d57ffb3a1 100644 --- a/pkg/sentry/syscalls/linux/sys_shm.go +++ b/pkg/sentry/syscalls/linux/sys_shm.go @@ -15,11 +15,11 @@ package linux import ( - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/shm" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/shm" + "gvisor.dev/gvisor/pkg/syserror" ) // Shmget implements shmget(2). diff --git a/pkg/sentry/syscalls/linux/sys_signal.go b/pkg/sentry/syscalls/linux/sys_signal.go index 7fbeb4fcd..0104a94c0 100644 --- a/pkg/sentry/syscalls/linux/sys_signal.go +++ b/pkg/sentry/syscalls/linux/sys_signal.go @@ -18,10 +18,10 @@ import ( "math" "time" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/syserror" ) // "For a process to have permission to send a signal it must diff --git a/pkg/sentry/syscalls/linux/sys_socket.go b/pkg/sentry/syscalls/linux/sys_socket.go index 31295a6a9..ccdb079bb 100644 --- a/pkg/sentry/syscalls/linux/sys_socket.go +++ b/pkg/sentry/syscalls/linux/sys_socket.go @@ -18,18 +18,19 @@ import ( "syscall" "time" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/binary" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" - ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" - "gvisor.googlesource.com/gvisor/pkg/sentry/socket" - "gvisor.googlesource.com/gvisor/pkg/sentry/socket/control" - "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/binary" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/fs" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/kdefs" + ktime "gvisor.dev/gvisor/pkg/sentry/kernel/time" + "gvisor.dev/gvisor/pkg/sentry/socket" + "gvisor.dev/gvisor/pkg/sentry/socket/control" + "gvisor.dev/gvisor/pkg/sentry/socket/unix/transport" + "gvisor.dev/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/syserr" + "gvisor.dev/gvisor/pkg/syserror" ) // minListenBacklog is the minimum reasonable backlog for listening sockets. @@ -61,6 +62,8 @@ const controlLenOffset = 40 // to the Flags field. const flagsOffset = 48 +const sizeOfInt32 = 4 + // messageHeader64Len is the length of a MessageHeader64 struct. var messageHeader64Len = uint64(binary.Size(MessageHeader64{})) @@ -466,7 +469,7 @@ func GetSockOpt(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sy } // Call syscall implementation then copy both value and value len out. - v, e := s.GetSockOpt(t, int(level), int(name), int(optLen)) + v, e := getSockOpt(t, s, int(level), int(name), int(optLen)) if e != nil { return 0, nil, e.ToError() } @@ -487,6 +490,33 @@ func GetSockOpt(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sy return 0, nil, nil } +// getSockOpt tries to handle common socket options, or dispatches to a specific +// socket implementation. +func getSockOpt(t *kernel.Task, s socket.Socket, level, name, len int) (interface{}, *syserr.Error) { + if level == linux.SOL_SOCKET { + switch name { + case linux.SO_TYPE, linux.SO_DOMAIN, linux.SO_PROTOCOL: + if len < sizeOfInt32 { + return nil, syserr.ErrInvalidArgument + } + } + + switch name { + case linux.SO_TYPE: + _, skType, _ := s.Type() + return int32(skType), nil + case linux.SO_DOMAIN: + family, _, _ := s.Type() + return int32(family), nil + case linux.SO_PROTOCOL: + _, _, protocol := s.Type() + return int32(protocol), nil + } + } + + return s.GetSockOpt(t, level, name, len) +} + // SetSockOpt implements the linux syscall setsockopt(2). // // Note that unlike Linux, enabling SO_PASSCRED does not autobind the socket. diff --git a/pkg/sentry/syscalls/linux/sys_splice.go b/pkg/sentry/syscalls/linux/sys_splice.go index 37303606f..b6517313f 100644 --- a/pkg/sentry/syscalls/linux/sys_splice.go +++ b/pkg/sentry/syscalls/linux/sys_splice.go @@ -15,13 +15,13 @@ package linux import ( - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" - "gvisor.googlesource.com/gvisor/pkg/syserror" - "gvisor.googlesource.com/gvisor/pkg/waiter" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/fs" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/kdefs" + "gvisor.dev/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/waiter" ) // doSplice implements a blocking splice operation. @@ -48,12 +48,12 @@ func doSplice(t *kernel.Task, outFile, inFile *fs.File, opts fs.SpliceOpts, nonB if ch == nil { ch = make(chan struct{}, 1) } - if !inW && inFile.Readiness(EventMaskRead) == 0 && !inFile.Flags().NonBlocking { + if !inW && !inFile.Flags().NonBlocking { w, _ := waiter.NewChannelEntry(ch) inFile.EventRegister(&w, EventMaskRead) defer inFile.EventUnregister(&w) inW = true // Registered. - } else if !outW && outFile.Readiness(EventMaskWrite) == 0 && !outFile.Flags().NonBlocking { + } else if !outW && !outFile.Flags().NonBlocking { w, _ := waiter.NewChannelEntry(ch) outFile.EventRegister(&w, EventMaskWrite) defer outFile.EventUnregister(&w) @@ -65,6 +65,11 @@ func doSplice(t *kernel.Task, outFile, inFile *fs.File, opts fs.SpliceOpts, nonB break } + if (!inW || inFile.Readiness(EventMaskRead) != 0) && (!outW || outFile.Readiness(EventMaskWrite) != 0) { + // Something became ready, try again without blocking. + continue + } + // Block until there's data. if err = t.Block(ch); err != nil { break diff --git a/pkg/sentry/syscalls/linux/sys_stat.go b/pkg/sentry/syscalls/linux/sys_stat.go index 10fc201ef..9a5657254 100644 --- a/pkg/sentry/syscalls/linux/sys_stat.go +++ b/pkg/sentry/syscalls/linux/sys_stat.go @@ -15,14 +15,14 @@ package linux import ( - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/binary" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/binary" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/fs" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/kdefs" + "gvisor.dev/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/syserror" ) // Stat implements linux syscall stat(2). @@ -35,7 +35,7 @@ func Stat(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallC return 0, nil, err } - return 0, nil, fileOpOn(t, linux.AT_FDCWD, path, true /* resolve */, func(root *fs.Dirent, d *fs.Dirent) error { + return 0, nil, fileOpOn(t, linux.AT_FDCWD, path, true /* resolve */, func(root *fs.Dirent, d *fs.Dirent, _ uint) error { return stat(t, d, dirPath, statAddr) }) } @@ -67,7 +67,7 @@ func Fstatat(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysca // then we must resolve the final component. resolve := dirPath || flags&linux.AT_SYMLINK_NOFOLLOW == 0 - return 0, nil, fileOpOn(t, fd, path, resolve, func(root *fs.Dirent, d *fs.Dirent) error { + return 0, nil, fileOpOn(t, fd, path, resolve, func(root *fs.Dirent, d *fs.Dirent, _ uint) error { return stat(t, d, dirPath, statAddr) }) } @@ -86,7 +86,7 @@ func Lstat(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall // want to resolve the final component. resolve := dirPath - return 0, nil, fileOpOn(t, linux.AT_FDCWD, path, resolve, func(root *fs.Dirent, d *fs.Dirent) error { + return 0, nil, fileOpOn(t, linux.AT_FDCWD, path, resolve, func(root *fs.Dirent, d *fs.Dirent, _ uint) error { return stat(t, d, dirPath, statAddr) }) } @@ -132,24 +132,6 @@ func fstat(t *kernel.Task, f *fs.File, statAddr usermem.Addr) error { // t.CopyObjectOut has noticeable performance impact due to its many slice // allocations and use of reflection. func copyOutStat(t *kernel.Task, dst usermem.Addr, sattr fs.StableAttr, uattr fs.UnstableAttr) error { - var mode uint32 - switch sattr.Type { - case fs.RegularFile, fs.SpecialFile: - mode |= linux.ModeRegular - case fs.Symlink: - mode |= linux.ModeSymlink - case fs.Directory, fs.SpecialDirectory: - mode |= linux.ModeDirectory - case fs.Pipe: - mode |= linux.ModeNamedPipe - case fs.CharacterDevice: - mode |= linux.ModeCharacterDevice - case fs.BlockDevice: - mode |= linux.ModeBlockDevice - case fs.Socket: - mode |= linux.ModeSocket - } - b := t.CopyScratchBuffer(int(linux.SizeOfStat))[:0] // Dev (uint64) @@ -159,7 +141,7 @@ func copyOutStat(t *kernel.Task, dst usermem.Addr, sattr fs.StableAttr, uattr fs // Nlink (uint64) b = binary.AppendUint64(b, usermem.ByteOrder, uattr.Links) // Mode (uint32) - b = binary.AppendUint32(b, usermem.ByteOrder, mode|uint32(uattr.Perms.LinuxMode())) + b = binary.AppendUint32(b, usermem.ByteOrder, sattr.Type.LinuxType()|uint32(uattr.Perms.LinuxMode())) // UID (uint32) b = binary.AppendUint32(b, usermem.ByteOrder, uint32(uattr.Owner.UID.In(t.UserNamespace()).OrOverflow())) // GID (uint32) @@ -194,6 +176,98 @@ func copyOutStat(t *kernel.Task, dst usermem.Addr, sattr fs.StableAttr, uattr fs return err } +// Statx implements linux syscall statx(2). +func Statx(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { + fd := kdefs.FD(args[0].Int()) + pathAddr := args[1].Pointer() + flags := args[2].Int() + mask := args[3].Uint() + statxAddr := args[4].Pointer() + + if mask&linux.STATX__RESERVED > 0 { + return 0, nil, syserror.EINVAL + } + if flags&linux.AT_STATX_SYNC_TYPE == linux.AT_STATX_SYNC_TYPE { + return 0, nil, syserror.EINVAL + } + + path, dirPath, err := copyInPath(t, pathAddr, flags&linux.AT_EMPTY_PATH != 0) + if err != nil { + return 0, nil, err + } + + if path == "" { + file := t.FDMap().GetFile(fd) + if file == nil { + return 0, nil, syserror.EBADF + } + defer file.DecRef() + uattr, err := file.UnstableAttr(t) + if err != nil { + return 0, nil, err + } + return 0, nil, statx(t, file.Dirent.Inode.StableAttr, uattr, statxAddr) + } + + resolve := dirPath || flags&linux.AT_SYMLINK_NOFOLLOW == 0 + + return 0, nil, fileOpOn(t, fd, path, resolve, func(root *fs.Dirent, d *fs.Dirent, _ uint) error { + if dirPath && !fs.IsDir(d.Inode.StableAttr) { + return syserror.ENOTDIR + } + uattr, err := d.Inode.UnstableAttr(t) + if err != nil { + return err + } + return statx(t, d.Inode.StableAttr, uattr, statxAddr) + }) +} + +func statx(t *kernel.Task, sattr fs.StableAttr, uattr fs.UnstableAttr, statxAddr usermem.Addr) error { + // "[T]he kernel may return fields that weren't requested and may fail to + // return fields that were requested, depending on what the backing + // filesystem supports. + // [...] + // A filesystem may also fill in fields that the caller didn't ask for + // if it has values for them available and the information is available + // at no extra cost. If this happens, the corresponding bits will be + // set in stx_mask." -- statx(2) + // + // We fill in all the values we have (which currently does not include + // btime, see b/135608823), regardless of what the user asked for. The + // STATX_BASIC_STATS mask indicates that all fields are present except + // for btime. + + devMajor, devMinor := linux.DecodeDeviceID(uint32(sattr.DeviceID)) + s := linux.Statx{ + // TODO(b/135608823): Support btime, and then change this to + // STATX_ALL to indicate presence of btime. + Mask: linux.STATX_BASIC_STATS, + + // No attributes, and none supported. + Attributes: 0, + AttributesMask: 0, + + Blksize: uint32(sattr.BlockSize), + Nlink: uint32(uattr.Links), + UID: uint32(uattr.Owner.UID.In(t.UserNamespace()).OrOverflow()), + GID: uint32(uattr.Owner.GID.In(t.UserNamespace()).OrOverflow()), + Mode: uint16(sattr.Type.LinuxType()) | uint16(uattr.Perms.LinuxMode()), + Ino: sattr.InodeID, + Size: uint64(uattr.Size), + Blocks: uint64(uattr.Usage) / 512, + Atime: uattr.AccessTime.StatxTimestamp(), + Ctime: uattr.StatusChangeTime.StatxTimestamp(), + Mtime: uattr.ModificationTime.StatxTimestamp(), + RdevMajor: uint32(sattr.DeviceFileMajor), + RdevMinor: sattr.DeviceFileMinor, + DevMajor: uint32(devMajor), + DevMinor: devMinor, + } + _, err := t.CopyOut(statxAddr, &s) + return err +} + // Statfs implements linux syscall statfs(2). func Statfs(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { addr := args[0].Pointer() @@ -204,7 +278,7 @@ func Statfs(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal return 0, nil, err } - return 0, nil, fileOpOn(t, linux.AT_FDCWD, path, true /* resolve */, func(root *fs.Dirent, d *fs.Dirent) error { + return 0, nil, fileOpOn(t, linux.AT_FDCWD, path, true /* resolve */, func(root *fs.Dirent, d *fs.Dirent, _ uint) error { return statfsImpl(t, d, statfsAddr) }) } @@ -252,8 +326,6 @@ func statfsImpl(t *kernel.Task, d *fs.Dirent, addr usermem.Addr) error { FragmentSize: d.Inode.StableAttr.BlockSize, // Leave other fields 0 like simple_statfs does. } - if _, err := t.CopyOut(addr, &statfs); err != nil { - return err - } - return nil + _, err = t.CopyOut(addr, &statfs) + return err } diff --git a/pkg/sentry/syscalls/linux/sys_sync.go b/pkg/sentry/syscalls/linux/sys_sync.go index 4352482fb..37225735f 100644 --- a/pkg/sentry/syscalls/linux/sys_sync.go +++ b/pkg/sentry/syscalls/linux/sys_sync.go @@ -15,12 +15,12 @@ package linux import ( - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/fs" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/kdefs" + "gvisor.dev/gvisor/pkg/syserror" ) // Sync implements linux system call sync(2). diff --git a/pkg/sentry/syscalls/linux/sys_sysinfo.go b/pkg/sentry/syscalls/linux/sys_sysinfo.go index ecf88edc1..a65b560c8 100644 --- a/pkg/sentry/syscalls/linux/sys_sysinfo.go +++ b/pkg/sentry/syscalls/linux/sys_sysinfo.go @@ -15,10 +15,10 @@ package linux import ( - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/usage" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/usage" ) // Sysinfo implements the sysinfo syscall as described in man 2 sysinfo. diff --git a/pkg/sentry/syscalls/linux/sys_syslog.go b/pkg/sentry/syscalls/linux/sys_syslog.go index 9efc58d34..40c8bb061 100644 --- a/pkg/sentry/syscalls/linux/sys_syslog.go +++ b/pkg/sentry/syscalls/linux/sys_syslog.go @@ -15,9 +15,9 @@ package linux import ( - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/syserror" ) const ( diff --git a/pkg/sentry/syscalls/linux/sys_thread.go b/pkg/sentry/syscalls/linux/sys_thread.go index 26f7e8ead..9e037bd7b 100644 --- a/pkg/sentry/syscalls/linux/sys_thread.go +++ b/pkg/sentry/syscalls/linux/sys_thread.go @@ -17,12 +17,12 @@ package linux import ( "syscall" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/sched" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/sched" + "gvisor.dev/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/syserror" ) const ( diff --git a/pkg/sentry/syscalls/linux/sys_time.go b/pkg/sentry/syscalls/linux/sys_time.go index b4f2609c0..fe8725191 100644 --- a/pkg/sentry/syscalls/linux/sys_time.go +++ b/pkg/sentry/syscalls/linux/sys_time.go @@ -17,12 +17,12 @@ package linux import ( "time" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" + ktime "gvisor.dev/gvisor/pkg/sentry/kernel/time" + "gvisor.dev/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/syserror" ) // The most significant 29 bits hold either a pid or a file descriptor. diff --git a/pkg/sentry/syscalls/linux/sys_timer.go b/pkg/sentry/syscalls/linux/sys_timer.go index 04ea7a4e9..ca5ccb7c3 100644 --- a/pkg/sentry/syscalls/linux/sys_timer.go +++ b/pkg/sentry/syscalls/linux/sys_timer.go @@ -18,10 +18,10 @@ import ( "syscall" "time" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/usermem" ) const nsecPerSec = int64(time.Second) diff --git a/pkg/sentry/syscalls/linux/sys_timerfd.go b/pkg/sentry/syscalls/linux/sys_timerfd.go index ec0155cbb..ea6d44315 100644 --- a/pkg/sentry/syscalls/linux/sys_timerfd.go +++ b/pkg/sentry/syscalls/linux/sys_timerfd.go @@ -15,14 +15,14 @@ package linux import ( - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs/timerfd" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" - ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/fs" + "gvisor.dev/gvisor/pkg/sentry/fs/timerfd" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/kdefs" + ktime "gvisor.dev/gvisor/pkg/sentry/kernel/time" + "gvisor.dev/gvisor/pkg/syserror" ) // TimerfdCreate implements Linux syscall timerfd_create(2). diff --git a/pkg/sentry/syscalls/linux/sys_tls.go b/pkg/sentry/syscalls/linux/sys_tls.go index 1e8312e00..e3d1c6201 100644 --- a/pkg/sentry/syscalls/linux/sys_tls.go +++ b/pkg/sentry/syscalls/linux/sys_tls.go @@ -19,9 +19,9 @@ package linux import ( "syscall" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" ) // ArchPrctl implements linux syscall arch_prctl(2). diff --git a/pkg/sentry/syscalls/linux/sys_utsname.go b/pkg/sentry/syscalls/linux/sys_utsname.go index fa81fe10e..271ace08e 100644 --- a/pkg/sentry/syscalls/linux/sys_utsname.go +++ b/pkg/sentry/syscalls/linux/sys_utsname.go @@ -17,10 +17,10 @@ package linux import ( - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/syserror" ) // Uname implements linux syscall uname. diff --git a/pkg/sentry/syscalls/linux/sys_write.go b/pkg/sentry/syscalls/linux/sys_write.go index 1da72d606..3a5bf9ac4 100644 --- a/pkg/sentry/syscalls/linux/sys_write.go +++ b/pkg/sentry/syscalls/linux/sys_write.go @@ -17,16 +17,16 @@ package linux import ( "time" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" - ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" - "gvisor.googlesource.com/gvisor/pkg/sentry/socket" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" - "gvisor.googlesource.com/gvisor/pkg/waiter" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/fs" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/kernel/kdefs" + ktime "gvisor.dev/gvisor/pkg/sentry/kernel/time" + "gvisor.dev/gvisor/pkg/sentry/socket" + "gvisor.dev/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/waiter" ) const ( diff --git a/pkg/sentry/syscalls/linux/timespec.go b/pkg/sentry/syscalls/linux/timespec.go index fa6fcdc0b..9ba0eba7a 100644 --- a/pkg/sentry/syscalls/linux/timespec.go +++ b/pkg/sentry/syscalls/linux/timespec.go @@ -18,10 +18,10 @@ import ( "syscall" "time" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/sentry/usermem" + "gvisor.dev/gvisor/pkg/syserror" ) // copyTimespecIn copies a Timespec from the untrusted app range to the kernel. diff --git a/pkg/sentry/syscalls/syscalls.go b/pkg/sentry/syscalls/syscalls.go index 48c114232..a5f3d8407 100644 --- a/pkg/sentry/syscalls/syscalls.go +++ b/pkg/sentry/syscalls/syscalls.go @@ -28,10 +28,10 @@ import ( "fmt" "syscall" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/sentry/arch" + "gvisor.dev/gvisor/pkg/sentry/kernel" + "gvisor.dev/gvisor/pkg/syserror" ) // Supported returns a syscall that is fully supported. |