Files
versitygw/backend/scoutfs/scoutfs_compat.go
T
Ben McClelland 29256f0ab5 feat: add optional data-integrity ETag mode for posix/scoutfs
Introduce an opt-in mode that derives ETags from checksums instead of relying on
MD5 assumptions, aligning with AWS-compatible scenarios where ETag should be
treated as an opaque identifier rather than a content-MD5 signal.

This can significantly reduce CPU load on the server to disable MD5 checksums
when we are already validating data with other checksums.

This change adds --data-integrity-etag and VGW_DATA_INTEGRITY_ETAG support for
both posix and scoutfs backends. In this mode, PUT object ETags, multipart part
ETags, and completed multipart object ETags are checksum-derived.
2026-07-31 14:58:21 -07:00

595 lines
16 KiB
Go

// Copyright 2023 Versity Software
// This file is 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.
//go:build linux && amd64
package scoutfs
import (
"context"
"encoding/json"
"errors"
"fmt"
"io/fs"
"os"
"path/filepath"
"strconv"
"strings"
"syscall"
"github.com/aws/aws-sdk-go-v2/service/s3"
"github.com/aws/aws-sdk-go-v2/service/s3/types"
"github.com/pkg/xattr"
"github.com/versity/scoutfs-go"
"github.com/versity/versitygw/auth"
"github.com/versity/versitygw/backend"
"github.com/versity/versitygw/backend/meta"
"github.com/versity/versitygw/backend/posix"
"github.com/versity/versitygw/debuglogger"
"github.com/versity/versitygw/s3err"
"github.com/versity/versitygw/s3response"
)
type ScoutFS struct {
*posix.Posix
rootfd *os.File
rootdir string
// glaciermode enables the following behavior:
// GET object: if file offline, return invalid object state
// HEAD object: if file offline, set obj storage class to GLACIER
// if file offline and staging, x-amz-restore: ongoing-request="true"
// if file offline and not staging, x-amz-restore: ongoing-request="false"
// if file online, x-amz-restore: ongoing-request="false", expiry-date="Fri, 2 Dec 2050 00:00:00 GMT"
// note: this expiry-date is not used but provided for client glacier compatibility
// ListObjects: if file offline, set obj storage class to GLACIER
// RestoreObject: add batch stage request to file
glaciermode bool
// disableNoArchive is used to disable setting scoutam noarchive flag
// on multipart parts. This is enabled by default to prevent archive
// copies of temporary multipart parts.
disableNoArchive bool
// enable posix level bucket name validations, not needed if the
// frontend handlers are already validating bucket names
validateBucketName bool
// projectIDEnabled enables setting projectid of new buckets and objects
// to the account project id when non-0
projectIDEnabled bool
}
func New(rootdir string, opts ScoutfsOpts) (*ScoutFS, error) {
metastore := meta.XattrMeta{}
p, err := posix.New(rootdir, metastore, posix.PosixOpts{
ChownUID: opts.ChownUID,
ChownGID: opts.ChownGID,
BucketLinks: opts.BucketLinks,
NewDirPerm: opts.NewDirPerm,
VersioningDir: opts.VersioningDir,
ValidateBucketNames: opts.ValidateBucketNames,
Concurrency: opts.Concurrency,
CopyObjectThreshold: opts.CopyObjectThreshold,
DefaultEtag: opts.DefaultEtag,
DataIntegrityEtag: opts.DataIntegrityEtag,
})
if err != nil {
return nil, err
}
f, err := os.Open(rootdir)
if err != nil {
return nil, fmt.Errorf("open %v: %w", rootdir, err)
}
setProjectID := opts.SetProjectID
if opts.SetProjectID {
setProjectID = fGetFormatVersion(f).AtLeast(versionScoutFsV2)
if !setProjectID {
fmt.Println("WARNING:")
fmt.Println("Disabling ProjectIDs for unsupported FS format version")
fmt.Println("See documentation for format version upgrades")
}
}
return &ScoutFS{
Posix: p,
rootfd: f,
rootdir: rootdir,
glaciermode: opts.GlacierMode,
disableNoArchive: opts.DisableNoArchive,
projectIDEnabled: setProjectID,
}, nil
}
const (
stageComplete = "ongoing-request=\"false\", expiry-date=\"Fri, 2 Dec 2050 00:00:00 GMT\""
stageInProgress = "ongoing-request=\"true\""
)
const (
// ScoutFS special xattr types
systemPrefix = "scoutfs.hide."
flagskey = systemPrefix + "sam_flags"
)
const (
// ScoutAM Flags
// Staging - file requested stage
Staging uint64 = 1 << iota
// StageFail - all copies failed to stage
StageFail
// NoArchive - no archive copies of file should be made
NoArchive
// ExtCacheRequested means file policy requests Ext Cache
ExtCacheRequested
// ExtCacheDone means this file ext cache copy has been
// created already (and possibly pruned, so may not exist)
ExtCacheDone
)
func (s *ScoutFS) Shutdown() {
s.Posix.Shutdown()
s.rootfd.Close()
}
func (*ScoutFS) String() string {
return "ScoutFS Gateway"
}
func (s *ScoutFS) CreateBucket(ctx context.Context, input *s3.CreateBucketInput, acl []byte) error {
err := s.Posix.CreateBucket(ctx, input, acl)
if err != nil {
return err
}
if s.projectIDEnabled {
acct, ok := ctx.Value("account").(auth.Account)
if !ok {
acct = auth.Account{}
}
if !isValidProjectID(acct.ProjectID) {
// early return to avoid the open if we dont have a valid
// project id
return nil
}
f, err := os.Open(*input.Bucket)
if err != nil {
debuglogger.InternalError(fmt.Errorf("create bucket %q set project id - open: %v",
*input.Bucket, err))
return nil
}
err = s.setProjectID(f, acct.ProjectID)
f.Close()
if err != nil {
debuglogger.InternalError(fmt.Errorf("create bucket %q set project id: %v",
*input.Bucket, err))
}
}
return nil
}
func (s *ScoutFS) HeadObject(ctx context.Context, input *s3.HeadObjectInput) (*s3.HeadObjectOutput, error) {
res, err := s.Posix.HeadObject(ctx, input)
if err != nil {
return nil, err
}
if s.glaciermode {
objPath := filepath.Join(*input.Bucket, *input.Key)
stclass := types.StorageClassStandard
requestOngoing := stageComplete
// Check if there are any offline exents associated with this file.
// If so, we will set storage class to glacier.
st, err := scoutfs.StatMore(objPath)
if errors.Is(err, fs.ErrNotExist) {
return nil, s3err.GetAPIError(s3err.ErrNoSuchKey)
}
if err != nil {
return nil, fmt.Errorf("stat more: %w", err)
}
if st.Offline_blocks != 0 {
stclass = types.StorageClassGlacier
requestOngoing = ""
ok, err := isStaging(objPath)
if errors.Is(err, fs.ErrNotExist) {
return nil, s3err.GetAPIError(s3err.ErrNoSuchKey)
}
if err != nil {
return nil, fmt.Errorf("check stage status: %w", err)
}
if ok {
requestOngoing = stageInProgress
}
}
res.Restore = &requestOngoing
res.StorageClass = stclass
}
return res, nil
}
func (s *ScoutFS) PutObject(ctx context.Context, po s3response.PutObjectInput) (s3response.PutObjectOutput, error) {
acct, ok := ctx.Value("account").(auth.Account)
if !ok {
acct = auth.Account{}
}
return s.Posix.PutObjectWithPostFunc(ctx, po, func(f *os.File) error {
err := s.setProjectID(f, acct.ProjectID)
if err != nil {
debuglogger.InternalError(fmt.Errorf("put object %v/%v set project id: %v",
filepath.Join(*po.Bucket, *po.Key), acct.ProjectID, err))
}
return nil
})
}
func (s *ScoutFS) UploadPart(ctx context.Context, input *s3.UploadPartInput) (*s3.UploadPartOutput, error) {
acct, ok := ctx.Value("account").(auth.Account)
if !ok {
acct = auth.Account{}
}
return s.Posix.UploadPartWithPostFunc(ctx, input,
func(f *os.File) error {
if !s.disableNoArchive {
err := setNoArchive(f)
if err != nil {
return fmt.Errorf("set noarchive: %w", err)
}
}
err := s.setProjectID(f, acct.ProjectID)
if err != nil {
return fmt.Errorf("set project id %v: %w", acct.ProjectID, err)
}
return nil
})
}
const scoutfsBlockSize = 4096
// CompleteMultipartUpload scoutfs complete upload uses scoutfs move blocks
// ioctl to not have to read and copy the part data to the final object. This
// saves a read and write cycle for all mutlipart uploads.
func (s *ScoutFS) CompleteMultipartUpload(ctx context.Context, input *s3.CompleteMultipartUploadInput) (s3response.CompleteMultipartUploadResult, string, error) {
acct, ok := ctx.Value("account").(auth.Account)
if !ok {
acct = auth.Account{}
}
totalParts := len(input.MultipartUpload.Parts)
callCount := 0
return s.Posix.CompleteMultipartUploadWithCopy(ctx, input,
func(from *os.File, to *os.File) (bool, error) {
callCount++
isLast := callCount == totalParts
// May fail if the files are not 4K aligned; check for alignment.
// The last part is allowed to have a size that is not a multiple of 4K.
ffi, err := from.Stat()
if err != nil {
return true, fmt.Errorf("complete-mpu stat from: %w", err)
}
tfi, err := to.Stat()
if err != nil {
return true, fmt.Errorf("complete-mpu stat to: %w", err)
}
if !isLast && (ffi.Size()%scoutfsBlockSize != 0 || tfi.Size()%scoutfsBlockSize != 0) {
return true, os.ErrInvalid
}
err = s.setProjectID(to, acct.ProjectID)
if err != nil {
debuglogger.InternalError(fmt.Errorf("complete-mpu %q/%q set project id %v: %v",
*input.Bucket, *input.Key, acct.ProjectID, err))
}
err = scoutfs.MoveData(from, to)
if err != nil {
return true, fmt.Errorf("complete-mpu movedata: %w", err)
}
// once scoutfs.MoveData is successful, we are no longer
// idempotent since we have moved the extents from the
// source file
return false, nil
})
}
func (s *ScoutFS) isBucketValid(bucket string) bool {
if !s.validateBucketName {
return true
}
return backend.IsValidDirectoryName(bucket)
}
func (s *ScoutFS) GetObject(ctx context.Context, input *s3.GetObjectInput) (*s3.GetObjectOutput, error) {
bucket := *input.Bucket
object := *input.Key
if !s.isBucketValid(bucket) {
return nil, s3err.GetBucketErr(s3err.ErrInvalidBucketName, bucket)
}
_, err := os.Stat(bucket)
if errors.Is(err, fs.ErrNotExist) {
return nil, s3err.GetBucketErr(s3err.ErrNoSuchBucket, *input.Bucket)
}
if err != nil {
return nil, fmt.Errorf("stat bucket: %w", err)
}
objPath := filepath.Join(bucket, object)
fi, err := os.Stat(objPath)
if errors.Is(err, fs.ErrNotExist) || errors.Is(err, syscall.ENOTDIR) {
return nil, s3err.GetAPIError(s3err.ErrNoSuchKey)
}
if errors.Is(err, syscall.ENAMETOOLONG) {
return nil, s3err.GetAPIError(s3err.ErrKeyTooLong)
}
if err != nil {
return nil, fmt.Errorf("stat object: %w", err)
}
if strings.HasSuffix(object, "/") && !fi.IsDir() {
return nil, s3err.GetAPIError(s3err.ErrNoSuchKey)
}
if s.glaciermode {
// Check if there are any offline exents associated with this file.
// If so, we will return the InvalidObjectState error.
st, err := scoutfs.StatMore(objPath)
if errors.Is(err, fs.ErrNotExist) {
return nil, s3err.GetAPIError(s3err.ErrNoSuchKey)
}
if err != nil {
return nil, fmt.Errorf("stat more: %w", err)
}
if st.Offline_blocks != 0 {
return nil, s3err.GetAPIError(s3err.ErrInvalidObjectState)
}
}
return s.Posix.GetObject(ctx, input)
}
func (s *ScoutFS) ListObjects(ctx context.Context, input *s3.ListObjectsInput) (s3response.ListObjectsResult, error) {
if s.glaciermode {
return s.Posix.ListObjectsParametrized(ctx, input, s.glacierFileToObj)
} else {
return s.Posix.ListObjects(ctx, input)
}
}
func (s *ScoutFS) ListObjectsV2(ctx context.Context, input *s3.ListObjectsV2Input) (s3response.ListObjectsV2Result, error) {
if s.glaciermode {
return s.Posix.ListObjectsV2Parametrized(ctx, input, s.glacierFileToObj)
} else {
return s.Posix.ListObjectsV2(ctx, input)
}
}
// FileToObj function for ListObject calls that adds a Glacier storage class if the file is offline
func (s *ScoutFS) glacierFileToObj(bucket string, fetchOwner bool) backend.GetObjFunc {
posixFileToObj := s.Posix.FileToObj(bucket, fetchOwner)
return func(path string, d fs.DirEntry) (s3response.Object, error) {
res, err := posixFileToObj(path, d)
if err != nil || d.IsDir() {
return res, err
}
objPath := filepath.Join(bucket, path)
// Check if there are any offline exents associated with this file.
// If so, we will return the Glacier storage class
st, err := scoutfs.StatMore(objPath)
if errors.Is(err, fs.ErrNotExist) {
return s3response.Object{}, backend.ErrSkipObj
}
if err != nil {
return s3response.Object{}, fmt.Errorf("stat more: %w", err)
}
if st.Offline_blocks != 0 {
res.StorageClass = types.ObjectStorageClassGlacier
}
return res, nil
}
}
// RestoreObject will set stage request on file if offline and do nothing if
// file is online
func (s *ScoutFS) RestoreObject(_ context.Context, input *s3.RestoreObjectInput) error {
bucket := *input.Bucket
object := *input.Key
if !s.isBucketValid(bucket) {
return s3err.GetBucketErr(s3err.ErrInvalidBucketName, bucket)
}
_, err := os.Stat(bucket)
if errors.Is(err, fs.ErrNotExist) {
return s3err.GetBucketErr(s3err.ErrNoSuchBucket, *input.Bucket)
}
if err != nil {
return fmt.Errorf("stat bucket: %w", err)
}
err = setStaging(filepath.Join(bucket, object))
if errors.Is(err, fs.ErrNotExist) {
return s3err.GetAPIError(s3err.ErrNoSuchKey)
}
if err != nil {
return fmt.Errorf("stage object: %w", err)
}
return nil
}
func isStaging(objname string) (bool, error) {
b, err := xattr.Get(objname, flagskey)
if err != nil && !isNoAttr(err) {
return false, err
}
var flags uint64
if !isNoAttr(err) {
err = json.Unmarshal(b, &flags)
if err != nil {
return false, err
}
}
return flags&Staging == Staging, nil
}
func setFlag(objname string, flag uint64) error {
f, err := os.Open(objname)
if err != nil {
return err
}
defer f.Close()
return fsetFlag(f, flag)
}
func fsetFlag(f *os.File, flag uint64) error {
b, err := xattr.FGet(f, flagskey)
if err != nil && !isNoAttr(err) {
return err
}
var oldflags uint64
if !isNoAttr(err) {
err = json.Unmarshal(b, &oldflags)
if err != nil {
return err
}
}
newflags := oldflags | flag
if newflags == oldflags {
// no flags change, just return
return nil
}
b, err = json.Marshal(&newflags)
if err != nil {
return err
}
return xattr.FSet(f, flagskey, b)
}
func setStaging(objname string) error {
return setFlag(objname, Staging)
}
func setNoArchive(f *os.File) error {
return fsetFlag(f, NoArchive)
}
func isNoAttr(err error) bool {
xerr, ok := err.(*xattr.Error)
if ok && xerr.Err == xattr.ENOATTR {
return true
}
return false
}
func (s *ScoutFS) setProjectID(f *os.File, proj int) error {
if s.projectIDEnabled && isValidProjectID(proj) {
err := scoutfs.SetProjectID(f, uint64(proj))
if err != nil {
return fmt.Errorf("set project id: %w", err)
}
}
return nil
}
func isValidProjectID(proj int) bool {
return proj > 0
}
const (
sysscoutfs = "/sys/fs/scoutfs/"
formatversion = "format_version"
)
// GetFormatVersion returns ScoutFS version reported by sysfs
func fGetFormatVersion(f *os.File) scoutFsVersion {
fsid, err := scoutfs.GetIDs(f)
if err != nil {
return versionScoutFsNotScoutFS
}
path := filepath.Join(sysscoutfs, fsid.ShortID, formatversion)
buf, err := os.ReadFile(path)
if err != nil {
return versionScoutFsUnknown
}
str := strings.TrimSpace(string(buf))
vers, err := strconv.Atoi(str)
if err != nil {
return versionScoutFsUnknown
}
return scoutFsVersion(vers)
}
const (
// versionScoutFsUnknown is unknown version
versionScoutFsUnknown scoutFsVersion = iota
// versionScoutFsV1 is version 1
versionScoutFsV1
// versionScoutFsV2 is version 2
versionScoutFsV2
// versionScoutFsMin is minimum scoutfs version
versionScoutFsMin = versionScoutFsV1
// versionScoutFsMax is maximum scoutfs version
versionScoutFsMax = versionScoutFsV2
// versionScoutFsNotScoutFS means the target FS is not scoutfs
versionScoutFsNotScoutFS = versionScoutFsMax + 1
)
// scoutFsVersion version
type scoutFsVersion int
// AtLeast returns true if version is valid and at least b
func (a scoutFsVersion) AtLeast(b scoutFsVersion) bool {
return a.IsValid() && a >= b
}
func (a scoutFsVersion) IsValid() bool {
return a >= versionScoutFsMin && a <= versionScoutFsMax
}