diff options
Diffstat (limited to 'pkg/lisafs')
| -rw-r--r-- | pkg/lisafs/BUILD | 1 | ||||
| -rw-r--r-- | pkg/lisafs/client.go | 55 | ||||
| -rw-r--r-- | pkg/lisafs/client_file.go | 475 | ||||
| -rw-r--r-- | pkg/lisafs/connection.go | 16 | ||||
| -rw-r--r-- | pkg/lisafs/fd.go | 28 | ||||
| -rw-r--r-- | pkg/lisafs/handlers.go | 650 | ||||
| -rw-r--r-- | pkg/lisafs/message.go | 993 | ||||
| -rw-r--r-- | pkg/lisafs/testsuite/BUILD | 20 | ||||
| -rw-r--r-- | pkg/lisafs/testsuite/testsuite.go | 637 |
9 files changed, 2871 insertions, 4 deletions
diff --git a/pkg/lisafs/BUILD b/pkg/lisafs/BUILD index 9914ed2f5..313c1756d 100644 --- a/pkg/lisafs/BUILD +++ b/pkg/lisafs/BUILD @@ -57,6 +57,7 @@ go_library( srcs = [ "channel.go", "client.go", + "client_file.go", "communicator.go", "connection.go", "control_fd_list.go", diff --git a/pkg/lisafs/client.go b/pkg/lisafs/client.go index c99f8c73d..ccf1b9f72 100644 --- a/pkg/lisafs/client.go +++ b/pkg/lisafs/client.go @@ -20,12 +20,19 @@ import ( "golang.org/x/sys/unix" "gvisor.dev/gvisor/pkg/cleanup" + "gvisor.dev/gvisor/pkg/context" "gvisor.dev/gvisor/pkg/flipcall" "gvisor.dev/gvisor/pkg/log" "gvisor.dev/gvisor/pkg/sync" "gvisor.dev/gvisor/pkg/unet" ) +const ( + // fdsToCloseBatchSize is the number of closed FDs batched before an Close + // RPC is made to close them all. fdsToCloseBatchSize is immutable. + fdsToCloseBatchSize = 100 +) + // Client helps manage a connection to the lisafs server and pass messages // efficiently. There is a 1:1 mapping between a Connection and a Client. type Client struct { @@ -53,6 +60,12 @@ type Client struct { // maxMessageSize is the maximum payload length (in bytes) that can be sent. // It is initialized on Mount and is immutable. maxMessageSize uint32 + + // fdsToClose tracks the FDs to close. It caches the FDs no longer being used + // by the client and closes them in one shot. It is not preserved across + // checkpoint/restore as FDIDs are not preserved. + fdsMu sync.Mutex + fdsToClose []FDID } // NewClient creates a new client for communication with the server. It mounts @@ -66,6 +79,7 @@ func NewClient(sock *unet.Socket, mountPath string) (*Client, *Inode, error) { channels: make([]*channel, 0, maxChans), availableChannels: make([]*channel, 0, maxChans), maxMessageSize: 1 << 20, // 1 MB for now. + fdsToClose: make([]FDID, 0, fdsToCloseBatchSize), } // Start a goroutine to check socket health. This goroutine is also @@ -245,6 +259,47 @@ func (c *Client) IsSupported(m MID) bool { return int(m) < len(c.supported) && c.supported[m] } +// CloseFDBatched either queues the passed FD to be closed or makes a batch +// RPC to close all the accumulated FDs-to-close. +func (c *Client) CloseFDBatched(ctx context.Context, fd FDID) { + c.fdsMu.Lock() + c.fdsToClose = append(c.fdsToClose, fd) + if len(c.fdsToClose) < fdsToCloseBatchSize { + c.fdsMu.Unlock() + return + } + + // Flush the cache. We should not hold fdsMu while making an RPC, so be sure + // to copy the fdsToClose to another buffer before unlocking fdsMu. + var toCloseArr [fdsToCloseBatchSize]FDID + toClose := toCloseArr[:len(c.fdsToClose)] + copy(toClose, c.fdsToClose) + + // Clear fdsToClose so other FDIDs can be appended. + c.fdsToClose = c.fdsToClose[:0] + c.fdsMu.Unlock() + + req := CloseReq{FDs: toClose} + ctx.UninterruptibleSleepStart(false) + err := c.SndRcvMessage(Close, uint32(req.SizeBytes()), req.MarshalBytes, NoopUnmarshal, nil) + ctx.UninterruptibleSleepFinish(false) + if err != nil { + log.Warningf("lisafs: batch closing FDs returned error: %v", err) + } +} + +// SyncFDs makes a Fsync RPC to sync multiple FDs. +func (c *Client) SyncFDs(ctx context.Context, fds []FDID) error { + if len(fds) == 0 { + return nil + } + req := FsyncReq{FDs: fds} + ctx.UninterruptibleSleepStart(false) + err := c.SndRcvMessage(FSync, uint32(req.SizeBytes()), req.MarshalBytes, NoopUnmarshal, nil) + ctx.UninterruptibleSleepFinish(false) + return err +} + // SndRcvMessage invokes reqMarshal to marshal the request onto the payload // buffer, wakes up the server to process the request, waits for the response // and invokes respUnmarshal with the response payload. respFDs is populated diff --git a/pkg/lisafs/client_file.go b/pkg/lisafs/client_file.go new file mode 100644 index 000000000..0f8788f3b --- /dev/null +++ b/pkg/lisafs/client_file.go @@ -0,0 +1,475 @@ +// Copyright 2021 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 lisafs + +import ( + "golang.org/x/sys/unix" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/context" + "gvisor.dev/gvisor/pkg/log" + "gvisor.dev/gvisor/pkg/marshal/primitive" +) + +// ClientFD is a wrapper around FDID that provides client-side utilities +// so that RPC making is easier. +type ClientFD struct { + fd FDID + client *Client +} + +// ID returns the underlying FDID. +func (f *ClientFD) ID() FDID { + return f.fd +} + +// Client returns the backing Client. +func (f *ClientFD) Client() *Client { + return f.client +} + +// NewFD initializes a new ClientFD. +func (c *Client) NewFD(fd FDID) ClientFD { + return ClientFD{ + client: c, + fd: fd, + } +} + +// Ok returns true if the underlying FD is ok. +func (f *ClientFD) Ok() bool { + return f.fd.Ok() +} + +// CloseBatched queues this FD to be closed on the server and resets f.fd. +// This maybe invoke the Close RPC if the queue is full. +func (f *ClientFD) CloseBatched(ctx context.Context) { + f.client.CloseFDBatched(ctx, f.fd) + f.fd = InvalidFDID +} + +// Close closes this FD immediately (invoking a Close RPC). Consider using +// CloseBatched if closing this FD on remote right away is not critical. +func (f *ClientFD) Close(ctx context.Context) error { + fdArr := [1]FDID{f.fd} + req := CloseReq{FDs: fdArr[:]} + + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(Close, uint32(req.SizeBytes()), req.MarshalBytes, NoopUnmarshal, nil) + ctx.UninterruptibleSleepFinish(false) + return err +} + +// OpenAt makes the OpenAt RPC. +func (f *ClientFD) OpenAt(ctx context.Context, flags uint32) (FDID, int, error) { + req := OpenAtReq{ + FD: f.fd, + Flags: flags, + } + var respFD [1]int + var resp OpenAtResp + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(OpenAt, uint32(req.SizeBytes()), req.MarshalUnsafe, resp.UnmarshalUnsafe, respFD[:]) + ctx.UninterruptibleSleepFinish(false) + return resp.NewFD, respFD[0], err +} + +// OpenCreateAt makes the OpenCreateAt RPC. +func (f *ClientFD) OpenCreateAt(ctx context.Context, name string, flags uint32, mode linux.FileMode, uid UID, gid GID) (Inode, FDID, int, error) { + var req OpenCreateAtReq + req.DirFD = f.fd + req.Name = SizedString(name) + req.Flags = primitive.Uint32(flags) + req.Mode = mode + req.UID = uid + req.GID = gid + + var respFD [1]int + var resp OpenCreateAtResp + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(OpenCreateAt, uint32(req.SizeBytes()), req.MarshalBytes, resp.UnmarshalUnsafe, respFD[:]) + ctx.UninterruptibleSleepFinish(false) + return resp.Child, resp.NewFD, respFD[0], err +} + +// StatTo makes the Fstat RPC and populates stat with the result. +func (f *ClientFD) StatTo(ctx context.Context, stat *linux.Statx) error { + req := StatReq{FD: f.fd} + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(FStat, uint32(req.SizeBytes()), req.MarshalUnsafe, stat.UnmarshalUnsafe, nil) + ctx.UninterruptibleSleepFinish(false) + return err +} + +// Sync makes the Fsync RPC. +func (f *ClientFD) Sync(ctx context.Context) error { + req := FsyncReq{FDs: []FDID{f.fd}} + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(FSync, uint32(req.SizeBytes()), req.MarshalBytes, NoopUnmarshal, nil) + ctx.UninterruptibleSleepFinish(false) + return err +} + +// Read makes the PRead RPC. +func (f *ClientFD) Read(ctx context.Context, dst []byte, offset uint64) (uint64, error) { + req := PReadReq{ + Offset: offset, + FD: f.fd, + Count: uint32(len(dst)), + } + + resp := PReadResp{ + // This will be unmarshalled into. Already set Buf so that we don't need to + // allocate a temporary buffer during unmarshalling. + // PReadResp.UnmarshalBytes expects this to be set. + Buf: dst, + } + + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(PRead, uint32(req.SizeBytes()), req.MarshalUnsafe, resp.UnmarshalBytes, nil) + ctx.UninterruptibleSleepFinish(false) + return uint64(resp.NumBytes), err +} + +// Write makes the PWrite RPC. +func (f *ClientFD) Write(ctx context.Context, src []byte, offset uint64) (uint64, error) { + req := PWriteReq{ + Offset: primitive.Uint64(offset), + FD: f.fd, + NumBytes: primitive.Uint32(len(src)), + Buf: src, + } + + var resp PWriteResp + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(PWrite, uint32(req.SizeBytes()), req.MarshalBytes, resp.UnmarshalUnsafe, nil) + ctx.UninterruptibleSleepFinish(false) + return resp.Count, err +} + +// MkdirAt makes the MkdirAt RPC. +func (f *ClientFD) MkdirAt(ctx context.Context, name string, mode linux.FileMode, uid UID, gid GID) (*Inode, error) { + var req MkdirAtReq + req.DirFD = f.fd + req.Name = SizedString(name) + req.Mode = mode + req.UID = uid + req.GID = gid + + var resp MkdirAtResp + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(MkdirAt, uint32(req.SizeBytes()), req.MarshalBytes, resp.UnmarshalUnsafe, nil) + ctx.UninterruptibleSleepFinish(false) + return &resp.ChildDir, err +} + +// SymlinkAt makes the SymlinkAt RPC. +func (f *ClientFD) SymlinkAt(ctx context.Context, name, target string, uid UID, gid GID) (*Inode, error) { + req := SymlinkAtReq{ + DirFD: f.fd, + Name: SizedString(name), + Target: SizedString(target), + UID: uid, + GID: gid, + } + + var resp SymlinkAtResp + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(SymlinkAt, uint32(req.SizeBytes()), req.MarshalBytes, resp.UnmarshalUnsafe, nil) + ctx.UninterruptibleSleepFinish(false) + return &resp.Symlink, err +} + +// LinkAt makes the LinkAt RPC. +func (f *ClientFD) LinkAt(ctx context.Context, targetFD FDID, name string) (*Inode, error) { + req := LinkAtReq{ + DirFD: f.fd, + Target: targetFD, + Name: SizedString(name), + } + + var resp LinkAtResp + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(LinkAt, uint32(req.SizeBytes()), req.MarshalBytes, resp.UnmarshalUnsafe, nil) + ctx.UninterruptibleSleepFinish(false) + return &resp.Link, err +} + +// MknodAt makes the MknodAt RPC. +func (f *ClientFD) MknodAt(ctx context.Context, name string, mode linux.FileMode, uid UID, gid GID, minor, major uint32) (*Inode, error) { + var req MknodAtReq + req.DirFD = f.fd + req.Name = SizedString(name) + req.Mode = mode + req.UID = uid + req.GID = gid + req.Minor = primitive.Uint32(minor) + req.Major = primitive.Uint32(major) + + var resp MknodAtResp + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(MknodAt, uint32(req.SizeBytes()), req.MarshalBytes, resp.UnmarshalUnsafe, nil) + ctx.UninterruptibleSleepFinish(false) + return &resp.Child, err +} + +// SetStat makes the SetStat RPC. +func (f *ClientFD) SetStat(ctx context.Context, stat *linux.Statx) (uint32, error, error) { + req := SetStatReq{ + FD: f.fd, + Mask: stat.Mask, + Mode: uint32(stat.Mode), + UID: UID(stat.UID), + GID: GID(stat.GID), + Size: stat.Size, + Atime: linux.Timespec{ + Sec: stat.Atime.Sec, + Nsec: int64(stat.Atime.Nsec), + }, + Mtime: linux.Timespec{ + Sec: stat.Mtime.Sec, + Nsec: int64(stat.Mtime.Nsec), + }, + } + + var resp SetStatResp + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(SetStat, uint32(req.SizeBytes()), req.MarshalUnsafe, resp.UnmarshalUnsafe, nil) + ctx.UninterruptibleSleepFinish(false) + return resp.FailureMask, unix.Errno(resp.FailureErrNo), err +} + +// WalkMultiple makes the Walk RPC with multiple path components. +func (f *ClientFD) WalkMultiple(ctx context.Context, names []string) (WalkStatus, []Inode, error) { + req := WalkReq{ + DirFD: f.fd, + Path: StringArray(names), + } + + var resp WalkResp + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(Walk, uint32(req.SizeBytes()), req.MarshalBytes, resp.UnmarshalBytes, nil) + ctx.UninterruptibleSleepFinish(false) + return resp.Status, resp.Inodes, err +} + +// Walk makes the Walk RPC with just one path component to walk. +func (f *ClientFD) Walk(ctx context.Context, name string) (*Inode, error) { + req := WalkReq{ + DirFD: f.fd, + Path: []string{name}, + } + + var inode [1]Inode + resp := WalkResp{Inodes: inode[:]} + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(Walk, uint32(req.SizeBytes()), req.MarshalBytes, resp.UnmarshalBytes, nil) + ctx.UninterruptibleSleepFinish(false) + if err != nil { + return nil, err + } + + switch resp.Status { + case WalkComponentDoesNotExist: + return nil, unix.ENOENT + case WalkComponentSymlink: + // f is not a directory which can be walked on. + return nil, unix.ENOTDIR + } + + if n := len(resp.Inodes); n > 1 { + for i := range resp.Inodes { + f.client.CloseFDBatched(ctx, resp.Inodes[i].ControlFD) + } + log.Warningf("requested to walk one component, but got %d results", n) + return nil, unix.EIO + } else if n == 0 { + log.Warningf("walk has success status but no results returned") + return nil, unix.ENOENT + } + return &inode[0], err +} + +// WalkStat makes the WalkStat RPC with multiple path components to walk. +func (f *ClientFD) WalkStat(ctx context.Context, names []string) ([]linux.Statx, error) { + req := WalkReq{ + DirFD: f.fd, + Path: StringArray(names), + } + + var resp WalkStatResp + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(WalkStat, uint32(req.SizeBytes()), req.MarshalBytes, resp.UnmarshalBytes, nil) + ctx.UninterruptibleSleepFinish(false) + return resp.Stats, err +} + +// StatFSTo makes the FStatFS RPC and populates statFS with the result. +func (f *ClientFD) StatFSTo(ctx context.Context, statFS *StatFS) error { + req := FStatFSReq{FD: f.fd} + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(FStatFS, uint32(req.SizeBytes()), req.MarshalUnsafe, statFS.UnmarshalUnsafe, nil) + ctx.UninterruptibleSleepFinish(false) + return err +} + +// Allocate makes the FAllocate RPC. +func (f *ClientFD) Allocate(ctx context.Context, mode, offset, length uint64) error { + req := FAllocateReq{ + FD: f.fd, + Mode: mode, + Offset: offset, + Length: length, + } + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(FAllocate, uint32(req.SizeBytes()), req.MarshalUnsafe, NoopUnmarshal, nil) + ctx.UninterruptibleSleepFinish(false) + return err +} + +// ReadLinkAt makes the ReadLinkAt RPC. +func (f *ClientFD) ReadLinkAt(ctx context.Context) (string, error) { + req := ReadLinkAtReq{FD: f.fd} + var resp ReadLinkAtResp + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(ReadLinkAt, uint32(req.SizeBytes()), req.MarshalUnsafe, resp.UnmarshalBytes, nil) + ctx.UninterruptibleSleepFinish(false) + return string(resp.Target), err +} + +// Flush makes the Flush RPC. +func (f *ClientFD) Flush(ctx context.Context) error { + if !f.client.IsSupported(Flush) { + // If Flush is not supported, it probably means that it would be a noop. + return nil + } + req := FlushReq{FD: f.fd} + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(Flush, uint32(req.SizeBytes()), req.MarshalUnsafe, NoopUnmarshal, nil) + ctx.UninterruptibleSleepFinish(false) + return err +} + +// Connect makes the Connect RPC. +func (f *ClientFD) Connect(ctx context.Context, sockType linux.SockType) (int, error) { + req := ConnectReq{FD: f.fd, SockType: uint32(sockType)} + var sockFD [1]int + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(Connect, uint32(req.SizeBytes()), req.MarshalUnsafe, NoopUnmarshal, sockFD[:]) + ctx.UninterruptibleSleepFinish(false) + if err == nil && sockFD[0] < 0 { + err = unix.EBADF + } + return sockFD[0], err +} + +// UnlinkAt makes the UnlinkAt RPC. +func (f *ClientFD) UnlinkAt(ctx context.Context, name string, flags uint32) error { + req := UnlinkAtReq{ + DirFD: f.fd, + Name: SizedString(name), + Flags: primitive.Uint32(flags), + } + + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(UnlinkAt, uint32(req.SizeBytes()), req.MarshalBytes, NoopUnmarshal, nil) + ctx.UninterruptibleSleepFinish(false) + return err +} + +// RenameTo makes the RenameAt RPC which renames f to newDirFD directory with +// name newName. +func (f *ClientFD) RenameTo(ctx context.Context, newDirFD FDID, newName string) error { + req := RenameAtReq{ + Renamed: f.fd, + NewDir: newDirFD, + NewName: SizedString(newName), + } + + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(RenameAt, uint32(req.SizeBytes()), req.MarshalBytes, NoopUnmarshal, nil) + ctx.UninterruptibleSleepFinish(false) + return err +} + +// Getdents64 makes the Getdents64 RPC. +func (f *ClientFD) Getdents64(ctx context.Context, count int32) ([]Dirent64, error) { + req := Getdents64Req{ + DirFD: f.fd, + Count: count, + } + + var resp Getdents64Resp + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(Getdents64, uint32(req.SizeBytes()), req.MarshalUnsafe, resp.UnmarshalBytes, nil) + ctx.UninterruptibleSleepFinish(false) + return resp.Dirents, err +} + +// ListXattr makes the FListXattr RPC. +func (f *ClientFD) ListXattr(ctx context.Context, size uint64) ([]string, error) { + req := FListXattrReq{ + FD: f.fd, + Size: size, + } + + var resp FListXattrResp + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(FListXattr, uint32(req.SizeBytes()), req.MarshalUnsafe, resp.UnmarshalBytes, nil) + ctx.UninterruptibleSleepFinish(false) + return resp.Xattrs, err +} + +// GetXattr makes the FGetXattr RPC. +func (f *ClientFD) GetXattr(ctx context.Context, name string, size uint64) (string, error) { + req := FGetXattrReq{ + FD: f.fd, + Name: SizedString(name), + BufSize: primitive.Uint32(size), + } + + var resp FGetXattrResp + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(FGetXattr, uint32(req.SizeBytes()), req.MarshalBytes, resp.UnmarshalBytes, nil) + ctx.UninterruptibleSleepFinish(false) + return string(resp.Value), err +} + +// SetXattr makes the FSetXattr RPC. +func (f *ClientFD) SetXattr(ctx context.Context, name string, value string, flags uint32) error { + req := FSetXattrReq{ + FD: f.fd, + Name: SizedString(name), + Value: SizedString(value), + Flags: primitive.Uint32(flags), + } + + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(FSetXattr, uint32(req.SizeBytes()), req.MarshalBytes, NoopUnmarshal, nil) + ctx.UninterruptibleSleepFinish(false) + return err +} + +// RemoveXattr makes the FRemoveXattr RPC. +func (f *ClientFD) RemoveXattr(ctx context.Context, name string) error { + req := FRemoveXattrReq{ + FD: f.fd, + Name: SizedString(name), + } + + ctx.UninterruptibleSleepStart(false) + err := f.client.SndRcvMessage(FRemoveXattr, uint32(req.SizeBytes()), req.MarshalBytes, NoopUnmarshal, nil) + ctx.UninterruptibleSleepFinish(false) + return err +} diff --git a/pkg/lisafs/connection.go b/pkg/lisafs/connection.go index 8dba4805f..f6e5ecb4f 100644 --- a/pkg/lisafs/connection.go +++ b/pkg/lisafs/connection.go @@ -289,6 +289,22 @@ func (c *Connection) RemoveFD(id FDID) { } } +// RemoveControlFDLocked is the same as RemoveFD with added preconditions. +// +// Preconditions: +// * server's rename mutex must at least be read locked. +// * id must be pointing to a control FD. +func (c *Connection) RemoveControlFDLocked(id FDID) { + c.fdsMu.Lock() + fd := c.removeFDLocked(id) + c.fdsMu.Unlock() + if fd != nil { + // Drop the ref held by c. This can take arbitrarily long. So do not hold + // c.fdsMu while calling it. + fd.(*ControlFD).DecRefLocked() + } +} + // removeFDLocked makes c stop tracking the passed FDID. Note that the caller // must drop ref on the returned fd (preferably without holding c.fdsMu). // diff --git a/pkg/lisafs/fd.go b/pkg/lisafs/fd.go index 9dd8ba384..cc6919a1b 100644 --- a/pkg/lisafs/fd.go +++ b/pkg/lisafs/fd.go @@ -231,7 +231,7 @@ func (fd *ControlFD) FilePath() string { return fd.FilePathLocked() } -// FilePathLocked is the same as FilePath with the additonal precondition. +// FilePathLocked is the same as FilePath with the additional precondition. // // Precondition: server's rename mutex must be at least read locked. func (fd *ControlFD) FilePathLocked() string { @@ -333,6 +333,25 @@ func (fd *OpenFD) Init(cfd *ControlFD, flags uint32, impl OpenFDImpl) { type ControlFDImpl interface { FD() *ControlFD Close(c *Connection) + Stat(c *Connection, comm Communicator) (uint32, error) + SetStat(c *Connection, comm Communicator, stat SetStatReq) (uint32, error) + Walk(c *Connection, comm Communicator, path StringArray) (uint32, error) + WalkStat(c *Connection, comm Communicator, path StringArray) (uint32, error) + Open(c *Connection, comm Communicator, flags uint32) (uint32, error) + OpenCreate(c *Connection, comm Communicator, mode linux.FileMode, uid UID, gid GID, name string, flags uint32) (uint32, error) + Mkdir(c *Connection, comm Communicator, mode linux.FileMode, uid UID, gid GID, name string) (uint32, error) + Mknod(c *Connection, comm Communicator, mode linux.FileMode, uid UID, gid GID, name string, minor uint32, major uint32) (uint32, error) + Symlink(c *Connection, comm Communicator, name string, target string, uid UID, gid GID) (uint32, error) + Link(c *Connection, comm Communicator, dir ControlFDImpl, name string) (uint32, error) + StatFS(c *Connection, comm Communicator) (uint32, error) + Readlink(c *Connection, comm Communicator) (uint32, error) + Connect(c *Connection, comm Communicator, sockType uint32) error + Unlink(c *Connection, name string, flags uint32) error + RenameLocked(c *Connection, newDir ControlFDImpl, newName string) (func(ControlFDImpl), func(), error) + GetXattr(c *Connection, comm Communicator, name string, size uint32) (uint32, error) + SetXattr(c *Connection, name string, value string, flags uint32) error + ListXattr(c *Connection, comm Communicator, size uint64) (uint32, error) + RemoveXattr(c *Connection, comm Communicator, name string) error } // OpenFDImpl contains implementation details for a OpenFD. Implementations of @@ -345,4 +364,11 @@ type ControlFDImpl interface { type OpenFDImpl interface { FD() *OpenFD Close(c *Connection) + Stat(c *Connection, comm Communicator) (uint32, error) + Sync(c *Connection) error + Write(c *Connection, comm Communicator, buf []byte, off uint64) (uint32, error) + Read(c *Connection, comm Communicator, off uint64, count uint32) (uint32, error) + Allocate(c *Connection, mode, off, length uint64) error + Flush(c *Connection) error + Getdent64(c *Connection, comm Communicator, count uint32, seek0 bool) (uint32, error) } diff --git a/pkg/lisafs/handlers.go b/pkg/lisafs/handlers.go index 9b8d8164a..82807734d 100644 --- a/pkg/lisafs/handlers.go +++ b/pkg/lisafs/handlers.go @@ -15,15 +15,23 @@ package lisafs import ( + "fmt" "path" "path/filepath" + "strings" "golang.org/x/sys/unix" "gvisor.dev/gvisor/pkg/flipcall" + "gvisor.dev/gvisor/pkg/fspath" "gvisor.dev/gvisor/pkg/log" "gvisor.dev/gvisor/pkg/marshal/primitive" ) +const ( + allowedOpenFlags = unix.O_ACCMODE | unix.O_TRUNC + setStatSupportedMask = unix.STATX_MODE | unix.STATX_UID | unix.STATX_GID | unix.STATX_SIZE | unix.STATX_ATIME | unix.STATX_MTIME +) + // RPCHandler defines a handler that is invoked when the associated message is // received. The handler is responsible for: // @@ -34,9 +42,35 @@ import ( type RPCHandler func(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) var handlers = [...]RPCHandler{ - Error: ErrorHandler, - Mount: MountHandler, - Channel: ChannelHandler, + Error: ErrorHandler, + Mount: MountHandler, + Channel: ChannelHandler, + FStat: FStatHandler, + SetStat: SetStatHandler, + Walk: WalkHandler, + WalkStat: WalkStatHandler, + OpenAt: OpenAtHandler, + OpenCreateAt: OpenCreateAtHandler, + Close: CloseHandler, + FSync: FSyncHandler, + PWrite: PWriteHandler, + PRead: PReadHandler, + MkdirAt: MkdirAtHandler, + MknodAt: MknodAtHandler, + SymlinkAt: SymlinkAtHandler, + LinkAt: LinkAtHandler, + FStatFS: FStatFSHandler, + FAllocate: FAllocateHandler, + ReadLinkAt: ReadLinkAtHandler, + Flush: FlushHandler, + Connect: ConnectHandler, + UnlinkAt: UnlinkAtHandler, + RenameAt: RenameAtHandler, + Getdents64: Getdents64Handler, + FGetXattr: FGetXattrHandler, + FSetXattr: FSetXattrHandler, + FListXattr: FListXattrHandler, + FRemoveXattr: FRemoveXattrHandler, } // ErrorHandler handles Error message. @@ -122,3 +156,613 @@ func ChannelHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32 resp.MarshalUnsafe(comm.PayloadBuf(respLen)) return respLen, nil } + +// FStatHandler handles the FStat RPC. +func FStatHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + var req StatReq + req.UnmarshalUnsafe(comm.PayloadBuf(payloadLen)) + + fd, err := c.lookupFD(req.FD) + if err != nil { + return 0, err + } + defer fd.DecRef(nil) + + switch t := fd.(type) { + case *ControlFD: + return t.impl.Stat(c, comm) + case *OpenFD: + return t.impl.Stat(c, comm) + default: + panic(fmt.Sprintf("unknown fd type %T", t)) + } +} + +// SetStatHandler handles the SetStat RPC. +func SetStatHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + if c.readonly { + return 0, unix.EROFS + } + + var req SetStatReq + req.UnmarshalUnsafe(comm.PayloadBuf(payloadLen)) + + fd, err := c.LookupControlFD(req.FD) + if err != nil { + return 0, err + } + defer fd.DecRef(nil) + + if req.Mask&^setStatSupportedMask != 0 { + return 0, unix.EPERM + } + + return fd.impl.SetStat(c, comm, req) +} + +// WalkHandler handles the Walk RPC. +func WalkHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + var req WalkReq + req.UnmarshalBytes(comm.PayloadBuf(payloadLen)) + + fd, err := c.LookupControlFD(req.DirFD) + if err != nil { + return 0, err + } + defer fd.DecRef(nil) + if !fd.IsDir() { + return 0, unix.ENOTDIR + } + for _, name := range req.Path { + if err := checkSafeName(name); err != nil { + return 0, err + } + } + + return fd.impl.Walk(c, comm, req.Path) +} + +// WalkStatHandler handles the WalkStat RPC. +func WalkStatHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + var req WalkReq + req.UnmarshalBytes(comm.PayloadBuf(payloadLen)) + + fd, err := c.LookupControlFD(req.DirFD) + if err != nil { + return 0, err + } + defer fd.DecRef(nil) + + // Note that this fd is allowed to not actually be a directory when the + // only path component to walk is "" (self). + if !fd.IsDir() { + if len(req.Path) > 1 || (len(req.Path) == 1 && len(req.Path[0]) > 0) { + return 0, unix.ENOTDIR + } + } + for i, name := range req.Path { + // First component is allowed to be "". + if i == 0 && len(name) == 0 { + continue + } + if err := checkSafeName(name); err != nil { + return 0, err + } + } + + return fd.impl.WalkStat(c, comm, req.Path) +} + +// OpenAtHandler handles the OpenAt RPC. +func OpenAtHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + var req OpenAtReq + req.UnmarshalUnsafe(comm.PayloadBuf(payloadLen)) + + // Only keep allowed open flags. + if allowedFlags := req.Flags & allowedOpenFlags; allowedFlags != req.Flags { + log.Debugf("discarding open flags that are not allowed: old open flags = %d, new open flags = %d", req.Flags, allowedFlags) + req.Flags = allowedFlags + } + + accessMode := req.Flags & unix.O_ACCMODE + trunc := req.Flags&unix.O_TRUNC != 0 + if c.readonly && (accessMode != unix.O_RDONLY || trunc) { + return 0, unix.EROFS + } + + fd, err := c.LookupControlFD(req.FD) + if err != nil { + return 0, err + } + defer fd.DecRef(nil) + if fd.IsDir() { + // Directory is not truncatable and must be opened with O_RDONLY. + if accessMode != unix.O_RDONLY || trunc { + return 0, unix.EISDIR + } + } + + return fd.impl.Open(c, comm, req.Flags) +} + +// OpenCreateAtHandler handles the OpenCreateAt RPC. +func OpenCreateAtHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + if c.readonly { + return 0, unix.EROFS + } + var req OpenCreateAtReq + req.UnmarshalBytes(comm.PayloadBuf(payloadLen)) + + // Only keep allowed open flags. + if allowedFlags := req.Flags & allowedOpenFlags; allowedFlags != req.Flags { + log.Debugf("discarding open flags that are not allowed: old open flags = %d, new open flags = %d", req.Flags, allowedFlags) + req.Flags = allowedFlags + } + + name := string(req.Name) + if err := checkSafeName(name); err != nil { + return 0, err + } + + fd, err := c.LookupControlFD(req.DirFD) + if err != nil { + return 0, err + } + defer fd.DecRef(nil) + if !fd.IsDir() { + return 0, unix.ENOTDIR + } + + return fd.impl.OpenCreate(c, comm, req.Mode, req.UID, req.GID, name, uint32(req.Flags)) +} + +// CloseHandler handles the Close RPC. +func CloseHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + var req CloseReq + req.UnmarshalBytes(comm.PayloadBuf(payloadLen)) + for _, fd := range req.FDs { + c.RemoveFD(fd) + } + + // There is no response message for this. + return 0, nil +} + +// FSyncHandler handles the FSync RPC. +func FSyncHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + var req FsyncReq + req.UnmarshalBytes(comm.PayloadBuf(payloadLen)) + + // Return the first error we encounter, but sync everything we can + // regardless. + var retErr error + for _, fdid := range req.FDs { + if err := c.fsyncFD(fdid); err != nil && retErr == nil { + retErr = err + } + } + + // There is no response message for this. + return 0, retErr +} + +func (c *Connection) fsyncFD(id FDID) error { + fd, err := c.LookupOpenFD(id) + if err != nil { + return err + } + return fd.impl.Sync(c) +} + +// PWriteHandler handles the PWrite RPC. +func PWriteHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + if c.readonly { + return 0, unix.EROFS + } + var req PWriteReq + // Note that it is an optimized Unmarshal operation which avoids any buffer + // allocation and copying. req.Buf just points to payload. This is safe to do + // as the handler owns payload and req's lifetime is limited to the handler. + req.UnmarshalBytes(comm.PayloadBuf(payloadLen)) + fd, err := c.LookupOpenFD(req.FD) + if err != nil { + return 0, err + } + if !fd.writable { + return 0, unix.EBADF + } + return fd.impl.Write(c, comm, req.Buf, uint64(req.Offset)) +} + +// PReadHandler handles the PRead RPC. +func PReadHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + var req PReadReq + req.UnmarshalUnsafe(comm.PayloadBuf(payloadLen)) + + fd, err := c.LookupOpenFD(req.FD) + if err != nil { + return 0, err + } + defer fd.DecRef(nil) + if !fd.readable { + return 0, unix.EBADF + } + return fd.impl.Read(c, comm, req.Offset, req.Count) +} + +// MkdirAtHandler handles the MkdirAt RPC. +func MkdirAtHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + if c.readonly { + return 0, unix.EROFS + } + var req MkdirAtReq + req.UnmarshalBytes(comm.PayloadBuf(payloadLen)) + + name := string(req.Name) + if err := checkSafeName(name); err != nil { + return 0, err + } + + fd, err := c.LookupControlFD(req.DirFD) + if err != nil { + return 0, err + } + defer fd.DecRef(nil) + if !fd.IsDir() { + return 0, unix.ENOTDIR + } + return fd.impl.Mkdir(c, comm, req.Mode, req.UID, req.GID, name) +} + +// MknodAtHandler handles the MknodAt RPC. +func MknodAtHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + if c.readonly { + return 0, unix.EROFS + } + var req MknodAtReq + req.UnmarshalBytes(comm.PayloadBuf(payloadLen)) + + name := string(req.Name) + if err := checkSafeName(name); err != nil { + return 0, err + } + + fd, err := c.LookupControlFD(req.DirFD) + if err != nil { + return 0, err + } + defer fd.DecRef(nil) + if !fd.IsDir() { + return 0, unix.ENOTDIR + } + return fd.impl.Mknod(c, comm, req.Mode, req.UID, req.GID, name, uint32(req.Minor), uint32(req.Major)) +} + +// SymlinkAtHandler handles the SymlinkAt RPC. +func SymlinkAtHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + if c.readonly { + return 0, unix.EROFS + } + var req SymlinkAtReq + req.UnmarshalBytes(comm.PayloadBuf(payloadLen)) + + name := string(req.Name) + if err := checkSafeName(name); err != nil { + return 0, err + } + + fd, err := c.LookupControlFD(req.DirFD) + if err != nil { + return 0, err + } + defer fd.DecRef(nil) + if !fd.IsDir() { + return 0, unix.ENOTDIR + } + return fd.impl.Symlink(c, comm, name, string(req.Target), req.UID, req.GID) +} + +// LinkAtHandler handles the LinkAt RPC. +func LinkAtHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + if c.readonly { + return 0, unix.EROFS + } + var req LinkAtReq + req.UnmarshalBytes(comm.PayloadBuf(payloadLen)) + + name := string(req.Name) + if err := checkSafeName(name); err != nil { + return 0, err + } + + fd, err := c.LookupControlFD(req.DirFD) + if err != nil { + return 0, err + } + defer fd.DecRef(nil) + if !fd.IsDir() { + return 0, unix.ENOTDIR + } + + targetFD, err := c.LookupControlFD(req.Target) + if err != nil { + return 0, err + } + return targetFD.impl.Link(c, comm, fd.impl, name) +} + +// FStatFSHandler handles the FStatFS RPC. +func FStatFSHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + var req FStatFSReq + req.UnmarshalUnsafe(comm.PayloadBuf(payloadLen)) + + fd, err := c.LookupControlFD(req.FD) + if err != nil { + return 0, err + } + defer fd.DecRef(nil) + return fd.impl.StatFS(c, comm) +} + +// FAllocateHandler handles the FAllocate RPC. +func FAllocateHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + if c.readonly { + return 0, unix.EROFS + } + var req FAllocateReq + req.UnmarshalUnsafe(comm.PayloadBuf(payloadLen)) + + fd, err := c.LookupOpenFD(req.FD) + if err != nil { + return 0, err + } + defer fd.DecRef(nil) + if !fd.writable { + return 0, unix.EBADF + } + return 0, fd.impl.Allocate(c, req.Mode, req.Offset, req.Length) +} + +// ReadLinkAtHandler handles the ReadLinkAt RPC. +func ReadLinkAtHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + var req ReadLinkAtReq + req.UnmarshalUnsafe(comm.PayloadBuf(payloadLen)) + + fd, err := c.LookupControlFD(req.FD) + if err != nil { + return 0, err + } + defer fd.DecRef(nil) + if !fd.IsSymlink() { + return 0, unix.EINVAL + } + return fd.impl.Readlink(c, comm) +} + +// FlushHandler handles the Flush RPC. +func FlushHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + var req FlushReq + req.UnmarshalUnsafe(comm.PayloadBuf(payloadLen)) + + fd, err := c.LookupOpenFD(req.FD) + if err != nil { + return 0, err + } + defer fd.DecRef(nil) + + return 0, fd.impl.Flush(c) +} + +// ConnectHandler handles the Connect RPC. +func ConnectHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + var req ConnectReq + req.UnmarshalUnsafe(comm.PayloadBuf(payloadLen)) + + fd, err := c.LookupControlFD(req.FD) + if err != nil { + return 0, err + } + defer fd.DecRef(nil) + if !fd.IsSocket() { + return 0, unix.ENOTSOCK + } + return 0, fd.impl.Connect(c, comm, req.SockType) +} + +// UnlinkAtHandler handles the UnlinkAt RPC. +func UnlinkAtHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + if c.readonly { + return 0, unix.EROFS + } + var req UnlinkAtReq + req.UnmarshalBytes(comm.PayloadBuf(payloadLen)) + + name := string(req.Name) + if err := checkSafeName(name); err != nil { + return 0, err + } + + fd, err := c.LookupControlFD(req.DirFD) + if err != nil { + return 0, err + } + defer fd.DecRef(nil) + if !fd.IsDir() { + return 0, unix.ENOTDIR + } + return 0, fd.impl.Unlink(c, name, uint32(req.Flags)) +} + +// RenameAtHandler handles the RenameAt RPC. +func RenameAtHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + if c.readonly { + return 0, unix.EROFS + } + var req RenameAtReq + req.UnmarshalBytes(comm.PayloadBuf(payloadLen)) + + newName := string(req.NewName) + if err := checkSafeName(newName); err != nil { + return 0, err + } + + renamed, err := c.LookupControlFD(req.Renamed) + if err != nil { + return 0, err + } + defer renamed.DecRef(nil) + + newDir, err := c.LookupControlFD(req.NewDir) + if err != nil { + return 0, err + } + defer newDir.DecRef(nil) + if !newDir.IsDir() { + return 0, unix.ENOTDIR + } + + // Hold RenameMu for writing during rename, this is important. + c.server.RenameMu.Lock() + defer c.server.RenameMu.Unlock() + + if renamed.parent == nil { + // renamed is root. + return 0, unix.EBUSY + } + + oldParentPath := renamed.parent.FilePathLocked() + oldPath := oldParentPath + "/" + renamed.name + if newName == renamed.name && oldParentPath == newDir.FilePathLocked() { + // Nothing to do. + return 0, nil + } + + updateControlFD, cleanUp, err := renamed.impl.RenameLocked(c, newDir.impl, newName) + if err != nil { + return 0, err + } + + c.server.forEachMountPoint(func(root *ControlFD) { + if !strings.HasPrefix(oldPath, root.name) { + return + } + pit := fspath.Parse(oldPath[len(root.name):]).Begin + root.renameRecursiveLocked(newDir, newName, pit, updateControlFD) + }) + + if cleanUp != nil { + cleanUp() + } + return 0, nil +} + +// Precondition: rename mutex must be locked for writing. +func (fd *ControlFD) renameRecursiveLocked(newDir *ControlFD, newName string, pit fspath.Iterator, updateControlFD func(ControlFDImpl)) { + if !pit.Ok() { + // fd should be renamed. + fd.clearParentLocked() + fd.setParentLocked(newDir) + fd.name = newName + if updateControlFD != nil { + updateControlFD(fd.impl) + } + return + } + + cur := pit.String() + next := pit.Next() + // No need to hold fd.childrenMu because RenameMu is locked for writing. + for child := fd.children.Front(); child != nil; child = child.Next() { + if child.name == cur { + child.renameRecursiveLocked(newDir, newName, next, updateControlFD) + } + } +} + +// Getdents64Handler handles the Getdents64 RPC. +func Getdents64Handler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + var req Getdents64Req + req.UnmarshalUnsafe(comm.PayloadBuf(payloadLen)) + + fd, err := c.LookupOpenFD(req.DirFD) + if err != nil { + return 0, err + } + defer fd.DecRef(nil) + if !fd.controlFD.IsDir() { + return 0, unix.ENOTDIR + } + + seek0 := false + if req.Count < 0 { + seek0 = true + req.Count = -req.Count + } + return fd.impl.Getdent64(c, comm, uint32(req.Count), seek0) +} + +// FGetXattrHandler handles the FGetXattr RPC. +func FGetXattrHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + var req FGetXattrReq + req.UnmarshalBytes(comm.PayloadBuf(payloadLen)) + + fd, err := c.LookupControlFD(req.FD) + if err != nil { + return 0, err + } + defer fd.DecRef(nil) + return fd.impl.GetXattr(c, comm, string(req.Name), uint32(req.BufSize)) +} + +// FSetXattrHandler handles the FSetXattr RPC. +func FSetXattrHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + if c.readonly { + return 0, unix.EROFS + } + var req FSetXattrReq + req.UnmarshalBytes(comm.PayloadBuf(payloadLen)) + + fd, err := c.LookupControlFD(req.FD) + if err != nil { + return 0, err + } + defer fd.DecRef(nil) + return 0, fd.impl.SetXattr(c, string(req.Name), string(req.Value), uint32(req.Flags)) +} + +// FListXattrHandler handles the FListXattr RPC. +func FListXattrHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + var req FListXattrReq + req.UnmarshalUnsafe(comm.PayloadBuf(payloadLen)) + + fd, err := c.LookupControlFD(req.FD) + if err != nil { + return 0, err + } + defer fd.DecRef(nil) + return fd.impl.ListXattr(c, comm, req.Size) +} + +// FRemoveXattrHandler handles the FRemoveXattr RPC. +func FRemoveXattrHandler(c *Connection, comm Communicator, payloadLen uint32) (uint32, error) { + if c.readonly { + return 0, unix.EROFS + } + var req FRemoveXattrReq + req.UnmarshalBytes(comm.PayloadBuf(payloadLen)) + + fd, err := c.LookupControlFD(req.FD) + if err != nil { + return 0, err + } + defer fd.DecRef(nil) + return 0, fd.impl.RemoveXattr(c, comm, string(req.Name)) +} + +// checkSafeName validates the name and returns nil or returns an error. +func checkSafeName(name string) error { + if name != "" && !strings.Contains(name, "/") && name != "." && name != ".." { + return nil + } + return unix.EINVAL +} diff --git a/pkg/lisafs/message.go b/pkg/lisafs/message.go index 55fd2c0b1..722afd0be 100644 --- a/pkg/lisafs/message.go +++ b/pkg/lisafs/message.go @@ -46,6 +46,94 @@ const ( // Channel requests to start a new communicational channel. Channel MID = 2 + + // FStat requests the stat(2) results for a specified file. + FStat MID = 3 + + // SetStat requests to change file attributes. Note that there is no one + // corresponding Linux syscall. This is a conglomeration of fchmod(2), + // fchown(2), ftruncate(2) and futimesat(2). + SetStat MID = 4 + + // Walk requests to walk the specified path starting from the specified + // directory. Server-side path traversal is terminated preemptively on + // symlinks entries because they can cause non-linear traversal. + Walk MID = 5 + + // WalkStat is the same as Walk, except the following differences: + // * If the first path component is "", then it also returns stat results + // for the directory where the walk starts. + // * Does not return Inode, just the Stat results for each path component. + WalkStat MID = 6 + + // OpenAt is analogous to openat(2). It does not perform any walk. It merely + // duplicates the control FD with the open flags passed. + OpenAt MID = 7 + + // OpenCreateAt is analogous to openat(2) with O_CREAT|O_EXCL added to flags. + // It also returns the newly created file inode. + OpenCreateAt MID = 8 + + // Close is analogous to close(2) but can work on multiple FDs. + Close MID = 9 + + // FSync is analogous to fsync(2) but can work on multiple FDs. + FSync MID = 10 + + // PWrite is analogous to pwrite(2). + PWrite MID = 11 + + // PRead is analogous to pread(2). + PRead MID = 12 + + // MkdirAt is analogous to mkdirat(2). + MkdirAt MID = 13 + + // MknodAt is analogous to mknodat(2). + MknodAt MID = 14 + + // SymlinkAt is analogous to symlinkat(2). + SymlinkAt MID = 15 + + // LinkAt is analogous to linkat(2). + LinkAt MID = 16 + + // FStatFS is analogous to fstatfs(2). + FStatFS MID = 17 + + // FAllocate is analogous to fallocate(2). + FAllocate MID = 18 + + // ReadLinkAt is analogous to readlinkat(2). + ReadLinkAt MID = 19 + + // Flush cleans up the file state. Its behavior is implementation + // dependent and might not even be supported in server implementations. + Flush MID = 20 + + // Connect is loosely analogous to connect(2). + Connect MID = 21 + + // UnlinkAt is analogous to unlinkat(2). + UnlinkAt MID = 22 + + // RenameAt is loosely analogous to renameat(2). + RenameAt MID = 23 + + // Getdents64 is analogous to getdents64(2). + Getdents64 MID = 24 + + // FGetXattr is analogous to fgetxattr(2). + FGetXattr MID = 25 + + // FSetXattr is analogous to fsetxattr(2). + FSetXattr MID = 26 + + // FListXattr is analogous to flistxattr(2). + FListXattr MID = 27 + + // FRemoveXattr is analogous to fremovexattr(2). + FRemoveXattr MID = 28 ) const ( @@ -256,3 +344,908 @@ type ChannelResp struct { type ErrorResp struct { errno uint32 } + +// StatReq requests the stat results for the specified FD. +// +// +marshal +type StatReq struct { + FD FDID +} + +// SetStatReq is used to set attributeds on FDs. +// +// +marshal +type SetStatReq struct { + FD FDID + _ uint32 + Mask uint32 + Mode uint32 // Only permissions part is settable. + UID UID + GID GID + Size uint64 + Atime linux.Timespec + Mtime linux.Timespec +} + +// SetStatResp is used to communicate SetStat results. It contains a mask +// representing the failed changes. It also contains the errno of the failed +// set attribute operation. If multiple operations failed then any of those +// errnos can be returned. +// +// +marshal +type SetStatResp struct { + FailureMask uint32 + FailureErrNo uint32 +} + +// WalkReq is used to request to walk multiple path components at once. This +// is used for both Walk and WalkStat. +type WalkReq struct { + DirFD FDID + Path StringArray +} + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (w *WalkReq) SizeBytes() int { + return w.DirFD.SizeBytes() + w.Path.SizeBytes() +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (w *WalkReq) MarshalBytes(dst []byte) { + w.DirFD.MarshalUnsafe(dst) + dst = dst[w.DirFD.SizeBytes():] + w.Path.MarshalBytes(dst) +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (w *WalkReq) UnmarshalBytes(src []byte) { + w.DirFD.UnmarshalUnsafe(src) + src = src[w.DirFD.SizeBytes():] + w.Path.UnmarshalBytes(src) +} + +// WalkStatus is used to indicate the reason for partial/unsuccessful server +// side Walk operations. Please note that partial/unsuccessful walk operations +// do not necessarily fail the RPC. The RPC is successful with a failure hint +// which can be used by the client to infer server-side state. +type WalkStatus = primitive.Uint8 + +const ( + // WalkSuccess indicates that all path components were successfully walked. + WalkSuccess WalkStatus = iota + + // WalkComponentDoesNotExist indicates that the walk was prematurely + // terminated because an intermediate path component does not exist on + // server. The results of all previous existing path components is returned. + WalkComponentDoesNotExist + + // WalkComponentSymlink indicates that the walk was prematurely + // terminated because an intermediate path component was a symlink. It is not + // safe to resolve symlinks remotely (unaware of mount points). + WalkComponentSymlink +) + +// WalkResp is used to communicate the inodes walked by the server. +type WalkResp struct { + Status WalkStatus + Inodes []Inode +} + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (w *WalkResp) SizeBytes() int { + return w.Status.SizeBytes() + + (*primitive.Uint32)(nil).SizeBytes() + (len(w.Inodes) * (*Inode)(nil).SizeBytes()) +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (w *WalkResp) MarshalBytes(dst []byte) { + w.Status.MarshalUnsafe(dst) + dst = dst[w.Status.SizeBytes():] + + numInodes := primitive.Uint32(len(w.Inodes)) + numInodes.MarshalUnsafe(dst) + dst = dst[numInodes.SizeBytes():] + + MarshalUnsafeInodeSlice(w.Inodes, dst) +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (w *WalkResp) UnmarshalBytes(src []byte) { + w.Status.UnmarshalUnsafe(src) + src = src[w.Status.SizeBytes():] + + var numInodes primitive.Uint32 + numInodes.UnmarshalUnsafe(src) + src = src[numInodes.SizeBytes():] + + if cap(w.Inodes) < int(numInodes) { + w.Inodes = make([]Inode, numInodes) + } else { + w.Inodes = w.Inodes[:numInodes] + } + UnmarshalUnsafeInodeSlice(w.Inodes, src) +} + +// WalkStatResp is used to communicate stat results for WalkStat. +type WalkStatResp struct { + Stats []linux.Statx +} + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (w *WalkStatResp) SizeBytes() int { + return (*primitive.Uint32)(nil).SizeBytes() + (len(w.Stats) * linux.SizeOfStatx) +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (w *WalkStatResp) MarshalBytes(dst []byte) { + numStats := primitive.Uint32(len(w.Stats)) + numStats.MarshalUnsafe(dst) + dst = dst[numStats.SizeBytes():] + + linux.MarshalUnsafeStatxSlice(w.Stats, dst) +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (w *WalkStatResp) UnmarshalBytes(src []byte) { + var numStats primitive.Uint32 + numStats.UnmarshalUnsafe(src) + src = src[numStats.SizeBytes():] + + if cap(w.Stats) < int(numStats) { + w.Stats = make([]linux.Statx, numStats) + } else { + w.Stats = w.Stats[:numStats] + } + linux.UnmarshalUnsafeStatxSlice(w.Stats, src) +} + +// OpenAtReq is used to open existing FDs with the specified flags. +// +// +marshal +type OpenAtReq struct { + FD FDID + Flags uint32 +} + +// OpenAtResp is used to communicate the newly created FD. +// +// +marshal +type OpenAtResp struct { + NewFD FDID +} + +// +marshal +type createCommon struct { + DirFD FDID + Mode linux.FileMode + _ uint16 // Need to make struct packed. + UID UID + GID GID +} + +// OpenCreateAtReq is used to make OpenCreateAt requests. +type OpenCreateAtReq struct { + createCommon + Name SizedString + Flags primitive.Uint32 +} + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (o *OpenCreateAtReq) SizeBytes() int { + return o.createCommon.SizeBytes() + o.Name.SizeBytes() + o.Flags.SizeBytes() +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (o *OpenCreateAtReq) MarshalBytes(dst []byte) { + o.createCommon.MarshalUnsafe(dst) + dst = dst[o.createCommon.SizeBytes():] + o.Name.MarshalBytes(dst) + dst = dst[o.Name.SizeBytes():] + o.Flags.MarshalUnsafe(dst) +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (o *OpenCreateAtReq) UnmarshalBytes(src []byte) { + o.createCommon.UnmarshalUnsafe(src) + src = src[o.createCommon.SizeBytes():] + o.Name.UnmarshalBytes(src) + src = src[o.Name.SizeBytes():] + o.Flags.UnmarshalUnsafe(src) +} + +// OpenCreateAtResp is used to communicate successful OpenCreateAt results. +// +// +marshal +type OpenCreateAtResp struct { + Child Inode + NewFD FDID + _ uint32 // Need to make struct packed. +} + +// FdArray is a utility struct which implements a marshallable type for +// communicating an array of FDIDs. In memory, the array data is preceded by a +// uint32 denoting the array length. +type FdArray []FDID + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (f *FdArray) SizeBytes() int { + return (*primitive.Uint32)(nil).SizeBytes() + (len(*f) * (*FDID)(nil).SizeBytes()) +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (f *FdArray) MarshalBytes(dst []byte) { + arrLen := primitive.Uint32(len(*f)) + arrLen.MarshalUnsafe(dst) + dst = dst[arrLen.SizeBytes():] + MarshalUnsafeFDIDSlice(*f, dst) +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (f *FdArray) UnmarshalBytes(src []byte) { + var arrLen primitive.Uint32 + arrLen.UnmarshalUnsafe(src) + src = src[arrLen.SizeBytes():] + if cap(*f) < int(arrLen) { + *f = make(FdArray, arrLen) + } else { + *f = (*f)[:arrLen] + } + UnmarshalUnsafeFDIDSlice(*f, src) +} + +// CloseReq is used to close(2) FDs. +type CloseReq struct { + FDs FdArray +} + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (c *CloseReq) SizeBytes() int { + return c.FDs.SizeBytes() +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (c *CloseReq) MarshalBytes(dst []byte) { + c.FDs.MarshalBytes(dst) +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (c *CloseReq) UnmarshalBytes(src []byte) { + c.FDs.UnmarshalBytes(src) +} + +// FsyncReq is used to fsync(2) FDs. +type FsyncReq struct { + FDs FdArray +} + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (f *FsyncReq) SizeBytes() int { + return f.FDs.SizeBytes() +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (f *FsyncReq) MarshalBytes(dst []byte) { + f.FDs.MarshalBytes(dst) +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (f *FsyncReq) UnmarshalBytes(src []byte) { + f.FDs.UnmarshalBytes(src) +} + +// PReadReq is used to pread(2) on an FD. +// +// +marshal +type PReadReq struct { + Offset uint64 + FD FDID + Count uint32 +} + +// PReadResp is used to return the result of pread(2). +type PReadResp struct { + NumBytes primitive.Uint32 + Buf []byte +} + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (r *PReadResp) SizeBytes() int { + return r.NumBytes.SizeBytes() + int(r.NumBytes) +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (r *PReadResp) MarshalBytes(dst []byte) { + r.NumBytes.MarshalUnsafe(dst) + dst = dst[r.NumBytes.SizeBytes():] + copy(dst[:r.NumBytes], r.Buf[:r.NumBytes]) +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (r *PReadResp) UnmarshalBytes(src []byte) { + r.NumBytes.UnmarshalUnsafe(src) + src = src[r.NumBytes.SizeBytes():] + + // We expect the client to have already allocated r.Buf. r.Buf probably + // (optimally) points to usermem. Directly copy into that. + copy(r.Buf[:r.NumBytes], src[:r.NumBytes]) +} + +// PWriteReq is used to pwrite(2) on an FD. +type PWriteReq struct { + Offset primitive.Uint64 + FD FDID + NumBytes primitive.Uint32 + Buf []byte +} + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (w *PWriteReq) SizeBytes() int { + return w.Offset.SizeBytes() + w.FD.SizeBytes() + w.NumBytes.SizeBytes() + int(w.NumBytes) +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (w *PWriteReq) MarshalBytes(dst []byte) { + w.Offset.MarshalUnsafe(dst) + dst = dst[w.Offset.SizeBytes():] + w.FD.MarshalUnsafe(dst) + dst = dst[w.FD.SizeBytes():] + w.NumBytes.MarshalUnsafe(dst) + dst = dst[w.NumBytes.SizeBytes():] + copy(dst[:w.NumBytes], w.Buf[:w.NumBytes]) +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (w *PWriteReq) UnmarshalBytes(src []byte) { + w.Offset.UnmarshalUnsafe(src) + src = src[w.Offset.SizeBytes():] + w.FD.UnmarshalUnsafe(src) + src = src[w.FD.SizeBytes():] + w.NumBytes.UnmarshalUnsafe(src) + src = src[w.NumBytes.SizeBytes():] + + // This is an optimization. Assuming that the server is making this call, it + // is safe to just point to src rather than allocating and copying. + w.Buf = src[:w.NumBytes] +} + +// PWriteResp is used to return the result of pwrite(2). +// +// +marshal +type PWriteResp struct { + Count uint64 +} + +// MkdirAtReq is used to make MkdirAt requests. +type MkdirAtReq struct { + createCommon + Name SizedString +} + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (m *MkdirAtReq) SizeBytes() int { + return m.createCommon.SizeBytes() + m.Name.SizeBytes() +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (m *MkdirAtReq) MarshalBytes(dst []byte) { + m.createCommon.MarshalUnsafe(dst) + dst = dst[m.createCommon.SizeBytes():] + m.Name.MarshalBytes(dst) +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (m *MkdirAtReq) UnmarshalBytes(src []byte) { + m.createCommon.UnmarshalUnsafe(src) + src = src[m.createCommon.SizeBytes():] + m.Name.UnmarshalBytes(src) +} + +// MkdirAtResp is the response to a successful MkdirAt request. +// +// +marshal +type MkdirAtResp struct { + ChildDir Inode +} + +// MknodAtReq is used to make MknodAt requests. +type MknodAtReq struct { + createCommon + Name SizedString + Minor primitive.Uint32 + Major primitive.Uint32 +} + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (m *MknodAtReq) SizeBytes() int { + return m.createCommon.SizeBytes() + m.Name.SizeBytes() + m.Minor.SizeBytes() + m.Major.SizeBytes() +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (m *MknodAtReq) MarshalBytes(dst []byte) { + m.createCommon.MarshalUnsafe(dst) + dst = dst[m.createCommon.SizeBytes():] + m.Name.MarshalBytes(dst) + dst = dst[m.Name.SizeBytes():] + m.Minor.MarshalUnsafe(dst) + dst = dst[m.Minor.SizeBytes():] + m.Major.MarshalUnsafe(dst) +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (m *MknodAtReq) UnmarshalBytes(src []byte) { + m.createCommon.UnmarshalUnsafe(src) + src = src[m.createCommon.SizeBytes():] + m.Name.UnmarshalBytes(src) + src = src[m.Name.SizeBytes():] + m.Minor.UnmarshalUnsafe(src) + src = src[m.Minor.SizeBytes():] + m.Major.UnmarshalUnsafe(src) +} + +// MknodAtResp is the response to a successful MknodAt request. +// +// +marshal +type MknodAtResp struct { + Child Inode +} + +// SymlinkAtReq is used to make SymlinkAt request. +type SymlinkAtReq struct { + DirFD FDID + Name SizedString + Target SizedString + UID UID + GID GID +} + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (s *SymlinkAtReq) SizeBytes() int { + return s.DirFD.SizeBytes() + s.Name.SizeBytes() + s.Target.SizeBytes() + s.UID.SizeBytes() + s.GID.SizeBytes() +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (s *SymlinkAtReq) MarshalBytes(dst []byte) { + s.DirFD.MarshalUnsafe(dst) + dst = dst[s.DirFD.SizeBytes():] + s.Name.MarshalBytes(dst) + dst = dst[s.Name.SizeBytes():] + s.Target.MarshalBytes(dst) + dst = dst[s.Target.SizeBytes():] + s.UID.MarshalUnsafe(dst) + dst = dst[s.UID.SizeBytes():] + s.GID.MarshalUnsafe(dst) +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (s *SymlinkAtReq) UnmarshalBytes(src []byte) { + s.DirFD.UnmarshalUnsafe(src) + src = src[s.DirFD.SizeBytes():] + s.Name.UnmarshalBytes(src) + src = src[s.Name.SizeBytes():] + s.Target.UnmarshalBytes(src) + src = src[s.Target.SizeBytes():] + s.UID.UnmarshalUnsafe(src) + src = src[s.UID.SizeBytes():] + s.GID.UnmarshalUnsafe(src) +} + +// SymlinkAtResp is the response to a successful SymlinkAt request. +// +// +marshal +type SymlinkAtResp struct { + Symlink Inode +} + +// LinkAtReq is used to make LinkAt requests. +type LinkAtReq struct { + DirFD FDID + Target FDID + Name SizedString +} + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (l *LinkAtReq) SizeBytes() int { + return l.DirFD.SizeBytes() + l.Target.SizeBytes() + l.Name.SizeBytes() +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (l *LinkAtReq) MarshalBytes(dst []byte) { + l.DirFD.MarshalUnsafe(dst) + dst = dst[l.DirFD.SizeBytes():] + l.Target.MarshalUnsafe(dst) + dst = dst[l.Target.SizeBytes():] + l.Name.MarshalBytes(dst) +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (l *LinkAtReq) UnmarshalBytes(src []byte) { + l.DirFD.UnmarshalUnsafe(src) + src = src[l.DirFD.SizeBytes():] + l.Target.UnmarshalUnsafe(src) + src = src[l.Target.SizeBytes():] + l.Name.UnmarshalBytes(src) +} + +// LinkAtResp is used to respond to a successful LinkAt request. +// +// +marshal +type LinkAtResp struct { + Link Inode +} + +// FStatFSReq is used to request StatFS results for the specified FD. +// +// +marshal +type FStatFSReq struct { + FD FDID +} + +// StatFS is responded to a successful FStatFS request. +// +// +marshal +type StatFS struct { + Type uint64 + BlockSize int64 + Blocks uint64 + BlocksFree uint64 + BlocksAvailable uint64 + Files uint64 + FilesFree uint64 + NameLength uint64 +} + +// FAllocateReq is used to request to fallocate(2) an FD. This has no response. +// +// +marshal +type FAllocateReq struct { + FD FDID + _ uint32 + Mode uint64 + Offset uint64 + Length uint64 +} + +// ReadLinkAtReq is used to readlinkat(2) at the specified FD. +// +// +marshal +type ReadLinkAtReq struct { + FD FDID +} + +// ReadLinkAtResp is used to communicate ReadLinkAt results. +type ReadLinkAtResp struct { + Target SizedString +} + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (r *ReadLinkAtResp) SizeBytes() int { + return r.Target.SizeBytes() +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (r *ReadLinkAtResp) MarshalBytes(dst []byte) { + r.Target.MarshalBytes(dst) +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (r *ReadLinkAtResp) UnmarshalBytes(src []byte) { + r.Target.UnmarshalBytes(src) +} + +// FlushReq is used to make Flush requests. +// +// +marshal +type FlushReq struct { + FD FDID +} + +// ConnectReq is used to make a Connect request. +// +// +marshal +type ConnectReq struct { + FD FDID + // SockType is used to specify the socket type to connect to. As a special + // case, SockType = 0 means that the socket type does not matter and the + // requester will accept any socket type. + SockType uint32 +} + +// UnlinkAtReq is used to make UnlinkAt request. +type UnlinkAtReq struct { + DirFD FDID + Name SizedString + Flags primitive.Uint32 +} + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (u *UnlinkAtReq) SizeBytes() int { + return u.DirFD.SizeBytes() + u.Name.SizeBytes() + u.Flags.SizeBytes() +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (u *UnlinkAtReq) MarshalBytes(dst []byte) { + u.DirFD.MarshalUnsafe(dst) + dst = dst[u.DirFD.SizeBytes():] + u.Name.MarshalBytes(dst) + dst = dst[u.Name.SizeBytes():] + u.Flags.MarshalUnsafe(dst) +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (u *UnlinkAtReq) UnmarshalBytes(src []byte) { + u.DirFD.UnmarshalUnsafe(src) + src = src[u.DirFD.SizeBytes():] + u.Name.UnmarshalBytes(src) + src = src[u.Name.SizeBytes():] + u.Flags.UnmarshalUnsafe(src) +} + +// RenameAtReq is used to make Rename requests. Note that the request takes in +// the to-be-renamed file's FD instead of oldDir and oldName like renameat(2). +type RenameAtReq struct { + Renamed FDID + NewDir FDID + NewName SizedString +} + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (r *RenameAtReq) SizeBytes() int { + return r.Renamed.SizeBytes() + r.NewDir.SizeBytes() + r.NewName.SizeBytes() +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (r *RenameAtReq) MarshalBytes(dst []byte) { + r.Renamed.MarshalUnsafe(dst) + dst = dst[r.Renamed.SizeBytes():] + r.NewDir.MarshalUnsafe(dst) + dst = dst[r.NewDir.SizeBytes():] + r.NewName.MarshalBytes(dst) +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (r *RenameAtReq) UnmarshalBytes(src []byte) { + r.Renamed.UnmarshalUnsafe(src) + src = src[r.Renamed.SizeBytes():] + r.NewDir.UnmarshalUnsafe(src) + src = src[r.NewDir.SizeBytes():] + r.NewName.UnmarshalBytes(src) +} + +// Getdents64Req is used to make Getdents64 requests. +// +// +marshal +type Getdents64Req struct { + DirFD FDID + // Count is the number of bytes to read. A negative value of Count is used to + // indicate that the implementation must lseek(0, SEEK_SET) before calling + // getdents64(2). Implementations must use the absolute value of Count to + // determine the number of bytes to read. + Count int32 +} + +// Dirent64 is analogous to struct linux_dirent64. +type Dirent64 struct { + Ino primitive.Uint64 + DevMinor primitive.Uint32 + DevMajor primitive.Uint32 + Off primitive.Uint64 + Type primitive.Uint8 + Name SizedString +} + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (d *Dirent64) SizeBytes() int { + return d.Ino.SizeBytes() + d.DevMinor.SizeBytes() + d.DevMajor.SizeBytes() + d.Off.SizeBytes() + d.Type.SizeBytes() + d.Name.SizeBytes() +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (d *Dirent64) MarshalBytes(dst []byte) { + d.Ino.MarshalUnsafe(dst) + dst = dst[d.Ino.SizeBytes():] + d.DevMinor.MarshalUnsafe(dst) + dst = dst[d.DevMinor.SizeBytes():] + d.DevMajor.MarshalUnsafe(dst) + dst = dst[d.DevMajor.SizeBytes():] + d.Off.MarshalUnsafe(dst) + dst = dst[d.Off.SizeBytes():] + d.Type.MarshalUnsafe(dst) + dst = dst[d.Type.SizeBytes():] + d.Name.MarshalBytes(dst) +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (d *Dirent64) UnmarshalBytes(src []byte) { + d.Ino.UnmarshalUnsafe(src) + src = src[d.Ino.SizeBytes():] + d.DevMinor.UnmarshalUnsafe(src) + src = src[d.DevMinor.SizeBytes():] + d.DevMajor.UnmarshalUnsafe(src) + src = src[d.DevMajor.SizeBytes():] + d.Off.UnmarshalUnsafe(src) + src = src[d.Off.SizeBytes():] + d.Type.UnmarshalUnsafe(src) + src = src[d.Type.SizeBytes():] + d.Name.UnmarshalBytes(src) +} + +// Getdents64Resp is used to communicate getdents64 results. +type Getdents64Resp struct { + Dirents []Dirent64 +} + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (g *Getdents64Resp) SizeBytes() int { + ret := (*primitive.Uint32)(nil).SizeBytes() + for i := range g.Dirents { + ret += g.Dirents[i].SizeBytes() + } + return ret +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (g *Getdents64Resp) MarshalBytes(dst []byte) { + numDirents := primitive.Uint32(len(g.Dirents)) + numDirents.MarshalUnsafe(dst) + dst = dst[numDirents.SizeBytes():] + for i := range g.Dirents { + g.Dirents[i].MarshalBytes(dst) + dst = dst[g.Dirents[i].SizeBytes():] + } +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (g *Getdents64Resp) UnmarshalBytes(src []byte) { + var numDirents primitive.Uint32 + numDirents.UnmarshalUnsafe(src) + if cap(g.Dirents) < int(numDirents) { + g.Dirents = make([]Dirent64, numDirents) + } else { + g.Dirents = g.Dirents[:numDirents] + } + + src = src[numDirents.SizeBytes():] + for i := range g.Dirents { + g.Dirents[i].UnmarshalBytes(src) + src = src[g.Dirents[i].SizeBytes():] + } +} + +// FGetXattrReq is used to make FGetXattr requests. The response to this is +// just a SizedString containing the xattr value. +type FGetXattrReq struct { + FD FDID + BufSize primitive.Uint32 + Name SizedString +} + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (g *FGetXattrReq) SizeBytes() int { + return g.FD.SizeBytes() + g.BufSize.SizeBytes() + g.Name.SizeBytes() +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (g *FGetXattrReq) MarshalBytes(dst []byte) { + g.FD.MarshalUnsafe(dst) + dst = dst[g.FD.SizeBytes():] + g.BufSize.MarshalUnsafe(dst) + dst = dst[g.BufSize.SizeBytes():] + g.Name.MarshalBytes(dst) +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (g *FGetXattrReq) UnmarshalBytes(src []byte) { + g.FD.UnmarshalUnsafe(src) + src = src[g.FD.SizeBytes():] + g.BufSize.UnmarshalUnsafe(src) + src = src[g.BufSize.SizeBytes():] + g.Name.UnmarshalBytes(src) +} + +// FGetXattrResp is used to respond to FGetXattr request. +type FGetXattrResp struct { + Value SizedString +} + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (g *FGetXattrResp) SizeBytes() int { + return g.Value.SizeBytes() +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (g *FGetXattrResp) MarshalBytes(dst []byte) { + g.Value.MarshalBytes(dst) +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (g *FGetXattrResp) UnmarshalBytes(src []byte) { + g.Value.UnmarshalBytes(src) +} + +// FSetXattrReq is used to make FSetXattr requests. It has no response. +type FSetXattrReq struct { + FD FDID + Flags primitive.Uint32 + Name SizedString + Value SizedString +} + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (s *FSetXattrReq) SizeBytes() int { + return s.FD.SizeBytes() + s.Flags.SizeBytes() + s.Name.SizeBytes() + s.Value.SizeBytes() +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (s *FSetXattrReq) MarshalBytes(dst []byte) { + s.FD.MarshalUnsafe(dst) + dst = dst[s.FD.SizeBytes():] + s.Flags.MarshalUnsafe(dst) + dst = dst[s.Flags.SizeBytes():] + s.Name.MarshalBytes(dst) + dst = dst[s.Name.SizeBytes():] + s.Value.MarshalBytes(dst) +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (s *FSetXattrReq) UnmarshalBytes(src []byte) { + s.FD.UnmarshalUnsafe(src) + src = src[s.FD.SizeBytes():] + s.Flags.UnmarshalUnsafe(src) + src = src[s.Flags.SizeBytes():] + s.Name.UnmarshalBytes(src) + src = src[s.Name.SizeBytes():] + s.Value.UnmarshalBytes(src) +} + +// FRemoveXattrReq is used to make FRemoveXattr requests. It has no response. +type FRemoveXattrReq struct { + FD FDID + Name SizedString +} + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (r *FRemoveXattrReq) SizeBytes() int { + return r.FD.SizeBytes() + r.Name.SizeBytes() +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (r *FRemoveXattrReq) MarshalBytes(dst []byte) { + r.FD.MarshalUnsafe(dst) + dst = dst[r.FD.SizeBytes():] + r.Name.MarshalBytes(dst) +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (r *FRemoveXattrReq) UnmarshalBytes(src []byte) { + r.FD.UnmarshalUnsafe(src) + src = src[r.FD.SizeBytes():] + r.Name.UnmarshalBytes(src) +} + +// FListXattrReq is used to make FListXattr requests. +// +// +marshal +type FListXattrReq struct { + FD FDID + _ uint32 + Size uint64 +} + +// FListXattrResp is used to respond to FListXattr requests. +type FListXattrResp struct { + Xattrs StringArray +} + +// SizeBytes implements marshal.Marshallable.SizeBytes. +func (l *FListXattrResp) SizeBytes() int { + return l.Xattrs.SizeBytes() +} + +// MarshalBytes implements marshal.Marshallable.MarshalBytes. +func (l *FListXattrResp) MarshalBytes(dst []byte) { + l.Xattrs.MarshalBytes(dst) +} + +// UnmarshalBytes implements marshal.Marshallable.UnmarshalBytes. +func (l *FListXattrResp) UnmarshalBytes(src []byte) { + l.Xattrs.UnmarshalBytes(src) +} diff --git a/pkg/lisafs/testsuite/BUILD b/pkg/lisafs/testsuite/BUILD new file mode 100644 index 000000000..b4a542b3a --- /dev/null +++ b/pkg/lisafs/testsuite/BUILD @@ -0,0 +1,20 @@ +load("//tools:defs.bzl", "go_library") + +package( + default_visibility = ["//visibility:public"], + licenses = ["notice"], +) + +go_library( + name = "testsuite", + testonly = True, + srcs = ["testsuite.go"], + deps = [ + "//pkg/abi/linux", + "//pkg/context", + "//pkg/lisafs", + "//pkg/unet", + "@com_github_syndtr_gocapability//capability:go_default_library", + "@org_golang_x_sys//unix:go_default_library", + ], +) diff --git a/pkg/lisafs/testsuite/testsuite.go b/pkg/lisafs/testsuite/testsuite.go new file mode 100644 index 000000000..476ff76a5 --- /dev/null +++ b/pkg/lisafs/testsuite/testsuite.go @@ -0,0 +1,637 @@ +// Copyright 2021 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 testsuite provides a integration testing suite for lisafs. +// These tests are intended for servers serving the local filesystem. +package testsuite + +import ( + "bytes" + "fmt" + "io/ioutil" + "math/rand" + "os" + "testing" + "time" + + "github.com/syndtr/gocapability/capability" + "golang.org/x/sys/unix" + "gvisor.dev/gvisor/pkg/abi/linux" + "gvisor.dev/gvisor/pkg/context" + "gvisor.dev/gvisor/pkg/lisafs" + "gvisor.dev/gvisor/pkg/unet" +) + +// Tester is the client code using this test suite. This interface abstracts +// away all the caller specific details. +type Tester interface { + // NewServer returns a new instance of the tester server. + NewServer(t *testing.T) *lisafs.Server + + // LinkSupported returns true if the backing server supports LinkAt. + LinkSupported() bool + + // SetUserGroupIDSupported returns true if the backing server supports + // changing UID/GID for files. + SetUserGroupIDSupported() bool +} + +// RunAllLocalFSTests runs all local FS tests as subtests. +func RunAllLocalFSTests(t *testing.T, tester Tester) { + for name, testFn := range localFSTests { + t.Run(name, func(t *testing.T) { + runServerClient(t, tester, testFn) + }) + } +} + +type testFunc func(context.Context, *testing.T, Tester, lisafs.ClientFD) + +var localFSTests map[string]testFunc = map[string]testFunc{ + "Stat": testStat, + "RegularFileIO": testRegularFileIO, + "RegularFileOpen": testRegularFileOpen, + "SetStat": testSetStat, + "Allocate": testAllocate, + "StatFS": testStatFS, + "Unlink": testUnlink, + "Symlink": testSymlink, + "HardLink": testHardLink, + "Walk": testWalk, + "Rename": testRename, + "Mknod": testMknod, + "Getdents": testGetdents, +} + +func runServerClient(t *testing.T, tester Tester, testFn testFunc) { + mountPath, err := ioutil.TempDir(os.Getenv("TEST_TMPDIR"), "") + if err != nil { + t.Fatalf("creation of temporary mountpoint failed: %v", err) + } + defer os.RemoveAll(mountPath) + + // fsgofer should run with a umask of 0, because we want to preserve file + // modes exactly for testing purposes. + unix.Umask(0) + + serverSocket, clientSocket, err := unet.SocketPair(false) + if err != nil { + t.Fatalf("socketpair got err %v expected nil", err) + } + + server := tester.NewServer(t) + conn, err := server.CreateConnection(serverSocket, false /* readonly */) + if err != nil { + t.Fatalf("starting connection failed: %v", err) + return + } + server.StartConnection(conn) + + c, root, err := lisafs.NewClient(clientSocket, mountPath) + if err != nil { + t.Fatalf("client creation failed: %v", err) + } + + if !root.ControlFD.Ok() { + t.Fatalf("root control FD is not valid") + } + rootFile := c.NewFD(root.ControlFD) + + ctx := context.Background() + testFn(ctx, t, tester, rootFile) + closeFD(ctx, t, rootFile) + + c.Close() // This should trigger client and server shutdown. + server.Wait() +} + +func closeFD(ctx context.Context, t testing.TB, fdLisa lisafs.ClientFD) { + if err := fdLisa.Close(ctx); err != nil { + t.Errorf("failed to close FD: %v", err) + } +} + +func statTo(ctx context.Context, t *testing.T, fdLisa lisafs.ClientFD, stat *linux.Statx) { + if err := fdLisa.StatTo(ctx, stat); err != nil { + t.Fatalf("stat failed: %v", err) + } +} + +func openCreateFile(ctx context.Context, t *testing.T, fdLisa lisafs.ClientFD, name string) (lisafs.ClientFD, linux.Statx, lisafs.ClientFD, int) { + child, childFD, childHostFD, err := fdLisa.OpenCreateAt(ctx, name, unix.O_RDWR, 0777, lisafs.UID(unix.Getuid()), lisafs.GID(unix.Getgid())) + if err != nil { + t.Fatalf("OpenCreateAt failed: %v", err) + } + if childHostFD == -1 { + t.Error("no host FD donated") + } + client := fdLisa.Client() + return client.NewFD(child.ControlFD), child.Stat, fdLisa.Client().NewFD(childFD), childHostFD +} + +func openFile(ctx context.Context, t *testing.T, fdLisa lisafs.ClientFD, flags uint32, isReg bool) (lisafs.ClientFD, int) { + newFD, hostFD, err := fdLisa.OpenAt(ctx, flags) + if err != nil { + t.Fatalf("OpenAt failed: %v", err) + } + if hostFD == -1 && isReg { + t.Error("no host FD donated") + } + return fdLisa.Client().NewFD(newFD), hostFD +} + +func unlinkFile(ctx context.Context, t *testing.T, dir lisafs.ClientFD, name string, isDir bool) { + var flags uint32 + if isDir { + flags = unix.AT_REMOVEDIR + } + if err := dir.UnlinkAt(ctx, name, flags); err != nil { + t.Errorf("unlinking file %s failed: %v", name, err) + } +} + +func symlink(ctx context.Context, t *testing.T, dir lisafs.ClientFD, name, target string) (lisafs.ClientFD, linux.Statx) { + linkIno, err := dir.SymlinkAt(ctx, name, target, lisafs.UID(unix.Getuid()), lisafs.GID(unix.Getgid())) + if err != nil { + t.Fatalf("symlink failed: %v", err) + } + return dir.Client().NewFD(linkIno.ControlFD), linkIno.Stat +} + +func link(ctx context.Context, t *testing.T, dir lisafs.ClientFD, name string, target lisafs.ClientFD) (lisafs.ClientFD, linux.Statx) { + linkIno, err := dir.LinkAt(ctx, target.ID(), name) + if err != nil { + t.Fatalf("link failed: %v", err) + } + return dir.Client().NewFD(linkIno.ControlFD), linkIno.Stat +} + +func mkdir(ctx context.Context, t *testing.T, dir lisafs.ClientFD, name string) (lisafs.ClientFD, linux.Statx) { + childIno, err := dir.MkdirAt(ctx, name, 0777, lisafs.UID(unix.Getuid()), lisafs.GID(unix.Getgid())) + if err != nil { + t.Fatalf("mkdir failed: %v", err) + } + return dir.Client().NewFD(childIno.ControlFD), childIno.Stat +} + +func mknod(ctx context.Context, t *testing.T, dir lisafs.ClientFD, name string) (lisafs.ClientFD, linux.Statx) { + nodeIno, err := dir.MknodAt(ctx, name, unix.S_IFREG|0777, lisafs.UID(unix.Getuid()), lisafs.GID(unix.Getgid()), 0, 0) + if err != nil { + t.Fatalf("mknod failed: %v", err) + } + return dir.Client().NewFD(nodeIno.ControlFD), nodeIno.Stat +} + +func walk(ctx context.Context, t *testing.T, dir lisafs.ClientFD, names []string) []lisafs.Inode { + _, inodes, err := dir.WalkMultiple(ctx, names) + if err != nil { + t.Fatalf("walk failed while trying to walk components %+v: %v", names, err) + } + return inodes +} + +func walkStat(ctx context.Context, t *testing.T, dir lisafs.ClientFD, names []string) []linux.Statx { + stats, err := dir.WalkStat(ctx, names) + if err != nil { + t.Fatalf("walk failed while trying to walk components %+v: %v", names, err) + } + return stats +} + +func writeFD(ctx context.Context, t *testing.T, fdLisa lisafs.ClientFD, off uint64, buf []byte) error { + count, err := fdLisa.Write(ctx, buf, off) + if err != nil { + return err + } + if int(count) != len(buf) { + t.Errorf("partial write: buf size = %d, written = %d", len(buf), count) + } + return nil +} + +func readFDAndCmp(ctx context.Context, t *testing.T, fdLisa lisafs.ClientFD, off uint64, want []byte) { + buf := make([]byte, len(want)) + n, err := fdLisa.Read(ctx, buf, off) + if err != nil { + t.Errorf("read failed: %v", err) + return + } + if int(n) != len(want) { + t.Errorf("partial read: buf size = %d, read = %d", len(want), n) + return + } + if bytes.Compare(buf, want) != 0 { + t.Errorf("bytes read differ from what was expected: want = %v, got = %v", want, buf) + } +} + +func allocateAndVerify(ctx context.Context, t *testing.T, fdLisa lisafs.ClientFD, off uint64, length uint64) { + if err := fdLisa.Allocate(ctx, 0, off, length); err != nil { + t.Fatalf("fallocate failed: %v", err) + } + + var stat linux.Statx + statTo(ctx, t, fdLisa, &stat) + if want := off + length; stat.Size != want { + t.Errorf("incorrect file size after allocate: expected %d, got %d", off+length, stat.Size) + } +} + +func cmpStatx(t *testing.T, want, got linux.Statx) { + if got.Mask&unix.STATX_MODE != 0 && want.Mask&unix.STATX_MODE != 0 { + if got.Mode != want.Mode { + t.Errorf("mode differs: want %d, got %d", want.Mode, got.Mode) + } + } + if got.Mask&unix.STATX_INO != 0 && want.Mask&unix.STATX_INO != 0 { + if got.Ino != want.Ino { + t.Errorf("inode number differs: want %d, got %d", want.Ino, got.Ino) + } + } + if got.Mask&unix.STATX_NLINK != 0 && want.Mask&unix.STATX_NLINK != 0 { + if got.Nlink != want.Nlink { + t.Errorf("nlink differs: want %d, got %d", want.Nlink, got.Nlink) + } + } + if got.Mask&unix.STATX_UID != 0 && want.Mask&unix.STATX_UID != 0 { + if got.UID != want.UID { + t.Errorf("UID differs: want %d, got %d", want.UID, got.UID) + } + } + if got.Mask&unix.STATX_GID != 0 && want.Mask&unix.STATX_GID != 0 { + if got.GID != want.GID { + t.Errorf("GID differs: want %d, got %d", want.GID, got.GID) + } + } + if got.Mask&unix.STATX_SIZE != 0 && want.Mask&unix.STATX_SIZE != 0 { + if got.Size != want.Size { + t.Errorf("size differs: want %d, got %d", want.Size, got.Size) + } + } + if got.Mask&unix.STATX_BLOCKS != 0 && want.Mask&unix.STATX_BLOCKS != 0 { + if got.Blocks != want.Blocks { + t.Errorf("blocks differs: want %d, got %d", want.Blocks, got.Blocks) + } + } + if got.Mask&unix.STATX_ATIME != 0 && want.Mask&unix.STATX_ATIME != 0 { + if got.Atime != want.Atime { + t.Errorf("atime differs: want %d, got %d", want.Atime, got.Atime) + } + } + if got.Mask&unix.STATX_MTIME != 0 && want.Mask&unix.STATX_MTIME != 0 { + if got.Mtime != want.Mtime { + t.Errorf("mtime differs: want %d, got %d", want.Mtime, got.Mtime) + } + } + if got.Mask&unix.STATX_CTIME != 0 && want.Mask&unix.STATX_CTIME != 0 { + if got.Ctime != want.Ctime { + t.Errorf("ctime differs: want %d, got %d", want.Ctime, got.Ctime) + } + } +} + +func hasCapability(c capability.Cap) bool { + caps, err := capability.NewPid2(os.Getpid()) + if err != nil { + return false + } + if err := caps.Load(); err != nil { + return false + } + return caps.Get(capability.EFFECTIVE, c) +} + +func testStat(ctx context.Context, t *testing.T, tester Tester, root lisafs.ClientFD) { + var rootStat linux.Statx + if err := root.StatTo(ctx, &rootStat); err != nil { + t.Errorf("stat on the root dir failed: %v", err) + } + + if ftype := rootStat.Mode & unix.S_IFMT; ftype != unix.S_IFDIR { + t.Errorf("root inode is not a directory, file type = %d", ftype) + } +} + +func testRegularFileIO(ctx context.Context, t *testing.T, tester Tester, root lisafs.ClientFD) { + name := "tempFile" + controlFile, _, fd, hostFD := openCreateFile(ctx, t, root, name) + defer closeFD(ctx, t, controlFile) + defer closeFD(ctx, t, fd) + defer unix.Close(hostFD) + + // Test Read/Write RPCs. + data := make([]byte, 100) + rand.Read(data) + if err := writeFD(ctx, t, fd, 0, data); err != nil { + t.Fatalf("write failed: %v", err) + } + readFDAndCmp(ctx, t, fd, 0, data) + readFDAndCmp(ctx, t, fd, 50, data[50:]) + + // Make sure the host FD is configured properly. + hostReadData := make([]byte, len(data)) + if n, err := unix.Pread(hostFD, hostReadData, 0); err != nil { + t.Errorf("host read failed: %v", err) + } else if n != len(hostReadData) { + t.Errorf("partial read: buf size = %d, read = %d", len(hostReadData), n) + } else if bytes.Compare(hostReadData, data) != 0 { + t.Errorf("bytes read differ from what was expected: want = %v, got = %v", data, hostReadData) + } + + // Test syncing the writable FD. + if err := fd.Sync(ctx); err != nil { + t.Errorf("syncing the FD failed: %v", err) + } +} + +func testRegularFileOpen(ctx context.Context, t *testing.T, tester Tester, root lisafs.ClientFD) { + name := "tempFile" + controlFile, _, fd, hostFD := openCreateFile(ctx, t, root, name) + defer closeFD(ctx, t, controlFile) + defer closeFD(ctx, t, fd) + defer unix.Close(hostFD) + + // Open a readonly FD and try writing to it to get an EBADF. + roFile, roHostFD := openFile(ctx, t, controlFile, unix.O_RDONLY, true /* isReg */) + defer closeFD(ctx, t, roFile) + defer unix.Close(roHostFD) + if err := writeFD(ctx, t, roFile, 0, []byte{1, 2, 3}); err != unix.EBADF { + t.Errorf("writing to read only FD should generate EBADF, but got %v", err) + } +} + +func testSetStat(ctx context.Context, t *testing.T, tester Tester, root lisafs.ClientFD) { + name := "tempFile" + controlFile, _, fd, hostFD := openCreateFile(ctx, t, root, name) + defer closeFD(ctx, t, controlFile) + defer closeFD(ctx, t, fd) + defer unix.Close(hostFD) + + now := time.Now() + wantStat := linux.Statx{ + Mask: unix.STATX_MODE | unix.STATX_ATIME | unix.STATX_MTIME | unix.STATX_SIZE, + Mode: 0760, + UID: uint32(unix.Getuid()), + GID: uint32(unix.Getgid()), + Size: 50, + Atime: linux.NsecToStatxTimestamp(now.UnixNano()), + Mtime: linux.NsecToStatxTimestamp(now.UnixNano()), + } + if tester.SetUserGroupIDSupported() { + wantStat.Mask |= unix.STATX_UID | unix.STATX_GID + } + failureMask, failureErr, err := controlFile.SetStat(ctx, &wantStat) + if err != nil { + t.Fatalf("setstat failed: %v", err) + } + if failureMask != 0 { + t.Fatalf("some setstat operations failed: failureMask = %#b, failureErr = %v", failureMask, failureErr) + } + + // Verify that attributes were updated. + var gotStat linux.Statx + statTo(ctx, t, controlFile, &gotStat) + if gotStat.Mode&07777 != wantStat.Mode || + gotStat.Size != wantStat.Size || + gotStat.Atime.ToNsec() != wantStat.Atime.ToNsec() || + gotStat.Mtime.ToNsec() != wantStat.Mtime.ToNsec() || + (tester.SetUserGroupIDSupported() && (uint32(gotStat.UID) != wantStat.UID || uint32(gotStat.GID) != wantStat.GID)) { + t.Errorf("setStat did not update file correctly: setStat = %+v, stat = %+v", wantStat, gotStat) + } +} + +func testAllocate(ctx context.Context, t *testing.T, tester Tester, root lisafs.ClientFD) { + name := "tempFile" + controlFile, _, fd, hostFD := openCreateFile(ctx, t, root, name) + defer closeFD(ctx, t, controlFile) + defer closeFD(ctx, t, fd) + defer unix.Close(hostFD) + + allocateAndVerify(ctx, t, fd, 0, 40) + allocateAndVerify(ctx, t, fd, 20, 100) +} + +func testStatFS(ctx context.Context, t *testing.T, tester Tester, root lisafs.ClientFD) { + var statFS lisafs.StatFS + if err := root.StatFSTo(ctx, &statFS); err != nil { + t.Errorf("statfs failed: %v", err) + } +} + +func testUnlink(ctx context.Context, t *testing.T, tester Tester, root lisafs.ClientFD) { + name := "tempFile" + controlFile, _, fd, hostFD := openCreateFile(ctx, t, root, name) + defer closeFD(ctx, t, controlFile) + defer closeFD(ctx, t, fd) + defer unix.Close(hostFD) + + unlinkFile(ctx, t, root, name, false /* isDir */) + if inodes := walk(ctx, t, root, []string{name}); len(inodes) > 0 { + t.Errorf("deleted file should not be generating inodes on walk: %+v", inodes) + } +} + +func testSymlink(ctx context.Context, t *testing.T, tester Tester, root lisafs.ClientFD) { + target := "/tmp/some/path" + name := "symlinkFile" + link, linkStat := symlink(ctx, t, root, name, target) + defer closeFD(ctx, t, link) + + if linkStat.Mode&unix.S_IFMT != unix.S_IFLNK { + t.Errorf("stat return from symlink RPC indicates that the inode is not a symlink: mode = %d", linkStat.Mode) + } + + if gotTarget, err := link.ReadLinkAt(ctx); err != nil { + t.Fatalf("readlink failed: %v", err) + } else if gotTarget != target { + t.Errorf("readlink return incorrect target: expected %q, got %q", target, gotTarget) + } +} + +func testHardLink(ctx context.Context, t *testing.T, tester Tester, root lisafs.ClientFD) { + if !tester.LinkSupported() { + t.Skipf("server does not support LinkAt RPC") + } + if !hasCapability(capability.CAP_DAC_READ_SEARCH) { + t.Skipf("TestHardLink requires CAP_DAC_READ_SEARCH, running as %d", unix.Getuid()) + } + name := "tempFile" + controlFile, fileIno, fd, hostFD := openCreateFile(ctx, t, root, name) + defer closeFD(ctx, t, controlFile) + defer closeFD(ctx, t, fd) + defer unix.Close(hostFD) + + link, linkStat := link(ctx, t, root, name, controlFile) + defer closeFD(ctx, t, link) + + if linkStat.Ino != fileIno.Ino { + t.Errorf("hard linked files have different inode numbers: %d %d", linkStat.Ino, fileIno.Ino) + } + if linkStat.DevMinor != fileIno.DevMinor { + t.Errorf("hard linked files have different minor device numbers: %d %d", linkStat.DevMinor, fileIno.DevMinor) + } + if linkStat.DevMajor != fileIno.DevMajor { + t.Errorf("hard linked files have different major device numbers: %d %d", linkStat.DevMajor, fileIno.DevMajor) + } +} + +func testWalk(ctx context.Context, t *testing.T, tester Tester, root lisafs.ClientFD) { + // Create 10 nested directories. + n := 10 + curDir := root + + dirNames := make([]string, 0, n) + for i := 0; i < n; i++ { + name := fmt.Sprintf("tmpdir-%d", i) + childDir, _ := mkdir(ctx, t, curDir, name) + defer closeFD(ctx, t, childDir) + defer unlinkFile(ctx, t, curDir, name, true /* isDir */) + + curDir = childDir + dirNames = append(dirNames, name) + } + + // Walk all these directories. Add some junk at the end which should not be + // walked on. + dirNames = append(dirNames, []string{"a", "b", "c"}...) + inodes := walk(ctx, t, root, dirNames) + if len(inodes) != n { + t.Errorf("walk returned the incorrect number of inodes: wanted %d, got %d", n, len(inodes)) + } + + // Close all control FDs and collect stat results for all dirs including + // the root directory. + dirStats := make([]linux.Statx, 0, n+1) + var stat linux.Statx + statTo(ctx, t, root, &stat) + dirStats = append(dirStats, stat) + for _, inode := range inodes { + dirStats = append(dirStats, inode.Stat) + closeFD(ctx, t, root.Client().NewFD(inode.ControlFD)) + } + + // Test WalkStat which additonally returns Statx for root because the first + // path component is "". + dirNames = append([]string{""}, dirNames...) + gotStats := walkStat(ctx, t, root, dirNames) + if len(gotStats) != len(dirStats) { + t.Errorf("walkStat returned the incorrect number of statx: wanted %d, got %d", len(dirStats), len(gotStats)) + } else { + for i := range gotStats { + cmpStatx(t, dirStats[i], gotStats[i]) + } + } +} + +func testRename(ctx context.Context, t *testing.T, tester Tester, root lisafs.ClientFD) { + name := "tempFile" + tempFile, _, fd, hostFD := openCreateFile(ctx, t, root, name) + defer closeFD(ctx, t, tempFile) + defer closeFD(ctx, t, fd) + defer unix.Close(hostFD) + + tempDir, _ := mkdir(ctx, t, root, "tempDir") + defer closeFD(ctx, t, tempDir) + + // Move tempFile into tempDir. + if err := tempFile.RenameTo(ctx, tempDir.ID(), "movedFile"); err != nil { + t.Fatalf("rename failed: %v", err) + } + + inodes := walkStat(ctx, t, root, []string{"tempDir", "movedFile"}) + if len(inodes) != 2 { + t.Errorf("expected 2 files on walk but only found %d", len(inodes)) + } +} + +func testMknod(ctx context.Context, t *testing.T, tester Tester, root lisafs.ClientFD) { + name := "namedPipe" + pipeFile, pipeStat := mknod(ctx, t, root, name) + defer closeFD(ctx, t, pipeFile) + + var stat linux.Statx + statTo(ctx, t, pipeFile, &stat) + + if stat.Mode != pipeStat.Mode { + t.Errorf("mknod mode is incorrect: want %d, got %d", pipeStat.Mode, stat.Mode) + } + if stat.UID != pipeStat.UID { + t.Errorf("mknod UID is incorrect: want %d, got %d", pipeStat.UID, stat.UID) + } + if stat.GID != pipeStat.GID { + t.Errorf("mknod GID is incorrect: want %d, got %d", pipeStat.GID, stat.GID) + } +} + +func testGetdents(ctx context.Context, t *testing.T, tester Tester, root lisafs.ClientFD) { + tempDir, _ := mkdir(ctx, t, root, "tempDir") + defer closeFD(ctx, t, tempDir) + defer unlinkFile(ctx, t, root, "tempDir", true /* isDir */) + + // Create 10 files in tempDir. + n := 10 + fileStats := make(map[string]linux.Statx) + for i := 0; i < n; i++ { + name := fmt.Sprintf("file-%d", i) + newFile, fileStat := mknod(ctx, t, tempDir, name) + defer closeFD(ctx, t, newFile) + defer unlinkFile(ctx, t, tempDir, name, false /* isDir */) + + fileStats[name] = fileStat + } + + // Use opened directory FD for getdents. + openDirFile, _ := openFile(ctx, t, tempDir, unix.O_RDONLY, false /* isReg */) + defer closeFD(ctx, t, openDirFile) + + dirents := make([]lisafs.Dirent64, 0, n) + for i := 0; i < n+2; i++ { + gotDirents, err := openDirFile.Getdents64(ctx, 40) + if err != nil { + t.Fatalf("getdents failed: %v", err) + } + if len(gotDirents) == 0 { + break + } + for _, dirent := range gotDirents { + if dirent.Name != "." && dirent.Name != ".." { + dirents = append(dirents, dirent) + } + } + } + + if len(dirents) != n { + t.Errorf("got incorrect number of dirents: wanted %d, got %d", n, len(dirents)) + } + for _, dirent := range dirents { + stat, ok := fileStats[string(dirent.Name)] + if !ok { + t.Errorf("received a dirent that was not created: %+v", dirent) + continue + } + + if dirent.Type != unix.DT_REG { + t.Errorf("dirent type of %s is incorrect: %d", dirent.Name, dirent.Type) + } + if uint64(dirent.Ino) != stat.Ino { + t.Errorf("dirent ino of %s is incorrect: want %d, got %d", dirent.Name, stat.Ino, dirent.Ino) + } + if uint32(dirent.DevMinor) != stat.DevMinor { + t.Errorf("dirent dev minor of %s is incorrect: want %d, got %d", dirent.Name, stat.DevMinor, dirent.DevMinor) + } + if uint32(dirent.DevMajor) != stat.DevMajor { + t.Errorf("dirent dev major of %s is incorrect: want %d, got %d", dirent.Name, stat.DevMajor, dirent.DevMajor) + } + } +} |