moby--moby/pkg/idtools/idtools_unix.go

232 lines
6.7 KiB
Go
Raw Normal View History

// +build !windows
package idtools // import "github.com/docker/docker/pkg/idtools"
import (
"bytes"
"fmt"
"io"
"os"
"path/filepath"
"strings"
"sync"
"syscall"
"github.com/docker/docker/pkg/system"
"github.com/opencontainers/runc/libcontainer/user"
)
var (
entOnce sync.Once
getentCmd string
)
func mkdirAs(path string, mode os.FileMode, owner Identity, mkAll, chownExisting bool) error {
// make an array containing the original path asked for, plus (for mkAll == true)
// all path components leading up to the complete path that don't exist before we MkdirAll
// so that we can chown all of them properly at the end. If chownExisting is false, we won't
// chown the full directory path if it exists
var paths []string
stat, err := system.Stat(path)
if err == nil {
if !stat.IsDir() {
Simplify/fix MkdirAll usage This subtle bug keeps lurking in because error checking for `Mkdir()` and `MkdirAll()` is slightly different wrt to `EEXIST`/`IsExist`: - for `Mkdir()`, `IsExist` error should (usually) be ignored (unless you want to make sure directory was not there before) as it means "the destination directory was already there" - for `MkdirAll()`, `IsExist` error should NEVER be ignored. Mostly, this commit just removes ignoring the IsExist error, as it should not be ignored. Also, there are a couple of cases then IsExist is handled as "directory already exist" which is wrong. As a result, some code that never worked as intended is now removed. NOTE that `idtools.MkdirAndChown()` behaves like `os.MkdirAll()` rather than `os.Mkdir()` -- so its description is amended accordingly, and its usage is handled as such (i.e. IsExist error is not ignored). For more details, a quote from my runc commit 6f82d4b (July 2015): TL;DR: check for IsExist(err) after a failed MkdirAll() is both redundant and wrong -- so two reasons to remove it. Quoting MkdirAll documentation: > MkdirAll creates a directory named path, along with any necessary > parents, and returns nil, or else returns an error. If path > is already a directory, MkdirAll does nothing and returns nil. This means two things: 1. If a directory to be created already exists, no error is returned. 2. If the error returned is IsExist (EEXIST), it means there exists a non-directory with the same name as MkdirAll need to use for directory. Example: we want to MkdirAll("a/b"), but file "a" (or "a/b") already exists, so MkdirAll fails. The above is a theory, based on quoted documentation and my UNIX knowledge. 3. In practice, though, current MkdirAll implementation [1] returns ENOTDIR in most of cases described in #2, with the exception when there is a race between MkdirAll and someone else creating the last component of MkdirAll argument as a file. In this very case MkdirAll() will indeed return EEXIST. Because of #1, IsExist check after MkdirAll is not needed. Because of #2 and #3, ignoring IsExist error is just plain wrong, as directory we require is not created. It's cleaner to report the error now. Note this error is all over the tree, I guess due to copy-paste, or trying to follow the same usage pattern as for Mkdir(), or some not quite correct examples on the Internet. [1] https://github.com/golang/go/blob/f9ed2f75/src/os/path.go Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
2017-09-25 19:39:36 +00:00
return &os.PathError{Op: "mkdir", Path: path, Err: syscall.ENOTDIR}
}
if !chownExisting {
return nil
}
// short-circuit--we were called with an existing directory and chown was requested
return lazyChown(path, owner.UID, owner.GID, stat)
}
if os.IsNotExist(err) {
paths = []string{path}
}
if mkAll {
// walk back to "/" looking for directories which do not exist
// and add them to the paths array for chown after creation
dirPath := path
for {
dirPath = filepath.Dir(dirPath)
if dirPath == "/" {
break
}
if _, err := os.Stat(dirPath); err != nil && os.IsNotExist(err) {
paths = append(paths, dirPath)
}
}
Simplify/fix MkdirAll usage This subtle bug keeps lurking in because error checking for `Mkdir()` and `MkdirAll()` is slightly different wrt to `EEXIST`/`IsExist`: - for `Mkdir()`, `IsExist` error should (usually) be ignored (unless you want to make sure directory was not there before) as it means "the destination directory was already there" - for `MkdirAll()`, `IsExist` error should NEVER be ignored. Mostly, this commit just removes ignoring the IsExist error, as it should not be ignored. Also, there are a couple of cases then IsExist is handled as "directory already exist" which is wrong. As a result, some code that never worked as intended is now removed. NOTE that `idtools.MkdirAndChown()` behaves like `os.MkdirAll()` rather than `os.Mkdir()` -- so its description is amended accordingly, and its usage is handled as such (i.e. IsExist error is not ignored). For more details, a quote from my runc commit 6f82d4b (July 2015): TL;DR: check for IsExist(err) after a failed MkdirAll() is both redundant and wrong -- so two reasons to remove it. Quoting MkdirAll documentation: > MkdirAll creates a directory named path, along with any necessary > parents, and returns nil, or else returns an error. If path > is already a directory, MkdirAll does nothing and returns nil. This means two things: 1. If a directory to be created already exists, no error is returned. 2. If the error returned is IsExist (EEXIST), it means there exists a non-directory with the same name as MkdirAll need to use for directory. Example: we want to MkdirAll("a/b"), but file "a" (or "a/b") already exists, so MkdirAll fails. The above is a theory, based on quoted documentation and my UNIX knowledge. 3. In practice, though, current MkdirAll implementation [1] returns ENOTDIR in most of cases described in #2, with the exception when there is a race between MkdirAll and someone else creating the last component of MkdirAll argument as a file. In this very case MkdirAll() will indeed return EEXIST. Because of #1, IsExist check after MkdirAll is not needed. Because of #2 and #3, ignoring IsExist error is just plain wrong, as directory we require is not created. It's cleaner to report the error now. Note this error is all over the tree, I guess due to copy-paste, or trying to follow the same usage pattern as for Mkdir(), or some not quite correct examples on the Internet. [1] https://github.com/golang/go/blob/f9ed2f75/src/os/path.go Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
2017-09-25 19:39:36 +00:00
if err := system.MkdirAll(path, mode, ""); err != nil {
return err
}
} else {
if err := os.Mkdir(path, mode); err != nil && !os.IsExist(err) {
return err
}
}
// even if it existed, we will chown the requested path + any subpaths that
// didn't exist when we called MkdirAll
for _, pathComponent := range paths {
if err := lazyChown(pathComponent, owner.UID, owner.GID, nil); err != nil {
return err
}
}
return nil
}
// CanAccess takes a valid (existing) directory and a uid, gid pair and determines
// if that uid, gid pair has access (execute bit) to the directory
func CanAccess(path string, pair Identity) bool {
statInfo, err := system.Stat(path)
if err != nil {
return false
}
fileMode := os.FileMode(statInfo.Mode())
permBits := fileMode.Perm()
return accessible(statInfo.UID() == uint32(pair.UID),
statInfo.GID() == uint32(pair.GID), permBits)
}
func accessible(isOwner, isGroup bool, perms os.FileMode) bool {
if isOwner && (perms&0100 == 0100) {
return true
}
if isGroup && (perms&0010 == 0010) {
return true
}
if perms&0001 == 0001 {
return true
}
return false
}
// LookupUser uses traditional local system files lookup (from libcontainer/user) on a username,
// followed by a call to `getent` for supporting host configured non-files passwd and group dbs
func LookupUser(username string) (user.User, error) {
// first try a local system files lookup using existing capabilities
usr, err := user.LookupUser(username)
if err == nil {
return usr, nil
}
// local files lookup failed; attempt to call `getent` to query configured passwd dbs
usr, err = getentUser(fmt.Sprintf("%s %s", "passwd", username))
if err != nil {
return user.User{}, err
}
return usr, nil
}
// LookupUID uses traditional local system files lookup (from libcontainer/user) on a uid,
// followed by a call to `getent` for supporting host configured non-files passwd and group dbs
func LookupUID(uid int) (user.User, error) {
// first try a local system files lookup using existing capabilities
usr, err := user.LookupUid(uid)
if err == nil {
return usr, nil
}
// local files lookup failed; attempt to call `getent` to query configured passwd dbs
return getentUser(fmt.Sprintf("%s %d", "passwd", uid))
}
func getentUser(args string) (user.User, error) {
reader, err := callGetent(args)
if err != nil {
return user.User{}, err
}
users, err := user.ParsePasswd(reader)
if err != nil {
return user.User{}, err
}
if len(users) == 0 {
return user.User{}, fmt.Errorf("getent failed to find passwd entry for %q", strings.Split(args, " ")[1])
}
return users[0], nil
}
// LookupGroup uses traditional local system files lookup (from libcontainer/user) on a group name,
// followed by a call to `getent` for supporting host configured non-files passwd and group dbs
func LookupGroup(groupname string) (user.Group, error) {
// first try a local system files lookup using existing capabilities
group, err := user.LookupGroup(groupname)
if err == nil {
return group, nil
}
// local files lookup failed; attempt to call `getent` to query configured group dbs
return getentGroup(fmt.Sprintf("%s %s", "group", groupname))
}
// LookupGID uses traditional local system files lookup (from libcontainer/user) on a group ID,
// followed by a call to `getent` for supporting host configured non-files passwd and group dbs
func LookupGID(gid int) (user.Group, error) {
// first try a local system files lookup using existing capabilities
group, err := user.LookupGid(gid)
if err == nil {
return group, nil
}
// local files lookup failed; attempt to call `getent` to query configured group dbs
return getentGroup(fmt.Sprintf("%s %d", "group", gid))
}
func getentGroup(args string) (user.Group, error) {
reader, err := callGetent(args)
if err != nil {
return user.Group{}, err
}
groups, err := user.ParseGroup(reader)
if err != nil {
return user.Group{}, err
}
if len(groups) == 0 {
return user.Group{}, fmt.Errorf("getent failed to find groups entry for %q", strings.Split(args, " ")[1])
}
return groups[0], nil
}
func callGetent(args string) (io.Reader, error) {
entOnce.Do(func() { getentCmd, _ = resolveBinary("getent") })
// if no `getent` command on host, can't do anything else
if getentCmd == "" {
return nil, fmt.Errorf("")
}
out, err := execCmd(getentCmd, args)
if err != nil {
exitCode, errC := system.GetExitCode(err)
if errC != nil {
return nil, err
}
switch exitCode {
case 1:
return nil, fmt.Errorf("getent reported invalid parameters/database unknown")
case 2:
terms := strings.Split(args, " ")
return nil, fmt.Errorf("getent unable to find entry %q in %s database", terms[1], terms[0])
case 3:
return nil, fmt.Errorf("getent database doesn't support enumeration")
default:
return nil, err
}
}
return bytes.NewReader(out), nil
}
// lazyChown performs a chown only if the uid/gid don't match what's requested
// Normally a Chown is a no-op if uid/gid match, but in some cases this can still cause an error, e.g. if the
// dir is on an NFS share, so don't call chown unless we absolutely must.
func lazyChown(p string, uid, gid int, stat *system.StatT) error {
if stat == nil {
var err error
stat, err = system.Stat(p)
if err != nil {
return err
}
}
if stat.UID() == uint32(uid) && stat.GID() == uint32(gid) {
return nil
}
return os.Chown(p, uid, gid)
}