rclone/backend/s3/s3.go

// Package s3 provides an interface to Amazon S3 object storage
package s3

//go:generate go run gen_setfrom.go -o setfrom.go

import (
	"context"
	"crypto/md5"
	"crypto/tls"
	"encoding/base64"
	"encoding/hex"
	"encoding/json"
	"encoding/xml"
	"errors"
	"fmt"
	"io"
	"math"
	"net/http"
	"net/url"
	"path"
	"regexp"
	"slices"
	"sort"
	"strconv"
	"strings"
	"sync"
	"time"

	"github.com/aws/aws-sdk-go-v2/aws"
	v4signer "github.com/aws/aws-sdk-go-v2/aws/signer/v4"
	awsconfig "github.com/aws/aws-sdk-go-v2/config"
	"github.com/aws/aws-sdk-go-v2/credentials"
	"github.com/aws/aws-sdk-go-v2/feature/s3/manager"
	"github.com/aws/aws-sdk-go-v2/service/s3"
	"github.com/aws/aws-sdk-go-v2/service/s3/types"
	"github.com/aws/smithy-go"
	"github.com/aws/smithy-go/logging"
	"github.com/aws/smithy-go/middleware"
	smithyhttp "github.com/aws/smithy-go/transport/http"
	"github.com/ncw/swift/v2"

	"golang.org/x/net/http/httpguts"
	"golang.org/x/sync/errgroup"

	"github.com/rclone/rclone/fs"
	"github.com/rclone/rclone/fs/accounting"
	"github.com/rclone/rclone/fs/chunksize"
	"github.com/rclone/rclone/fs/config"
	"github.com/rclone/rclone/fs/config/configmap"
	"github.com/rclone/rclone/fs/config/configstruct"
	"github.com/rclone/rclone/fs/fserrors"
	"github.com/rclone/rclone/fs/fshttp"
	"github.com/rclone/rclone/fs/hash"
	"github.com/rclone/rclone/fs/list"
	"github.com/rclone/rclone/fs/operations"
	"github.com/rclone/rclone/lib/atexit"
	"github.com/rclone/rclone/lib/bucket"
	"github.com/rclone/rclone/lib/encoder"
	"github.com/rclone/rclone/lib/multipart"
	"github.com/rclone/rclone/lib/pacer"
	"github.com/rclone/rclone/lib/pool"
	"github.com/rclone/rclone/lib/readers"
	"github.com/rclone/rclone/lib/rest"
	"github.com/rclone/rclone/lib/version"
)

// Register with Fs
func init() {
	fs.Register(addProvidersToInfo(&fs.RegInfo{
		Name:        "s3",
		Description: "Amazon S3 Compliant Storage Providers including ",
		NewFs:       NewFs,
		CommandHelp: commandHelp,
		Config: func(ctx context.Context, name string, m configmap.Mapper, config fs.ConfigIn) (*fs.ConfigOut, error) {
			switch config.State {
			case "":
				return nil, setEndpointValueForIDriveE2(m)
			}
			return nil, fmt.Errorf("unknown state %q", config.State)
		},
		MetadataInfo: &fs.MetadataInfo{
			System: systemMetadataInfo,
			Help:   `User metadata is stored as x-amz-meta- keys. S3 metadata keys are case insensitive and are always returned in lower case.`,
		},
		Options: []fs.Option{{
			Name: fs.ConfigProvider,
			Help: "Choose your S3 provider.",
		}, {
			Name:    "env_auth",
			Help:    "Get AWS credentials from runtime (environment variables or EC2/ECS meta data if no env vars).\n\nOnly applies if access_key_id and secret_access_key is blank.",
			Default: false,
			Examples: []fs.OptionExample{{
				Value: "false",
				Help:  "Enter AWS credentials in the next step.",
			}, {
				Value: "true",
				Help:  "Get AWS credentials from the environment (env vars or IAM).",
			}},
		}, {
			Name:      "access_key_id",
			Help:      "AWS Access Key ID.\n\nLeave blank for anonymous access or runtime credentials.",
			Sensitive: true,
		}, {
			Name:      "secret_access_key",
			Help:      "AWS Secret Access Key (password).\n\nLeave blank for anonymous access or runtime credentials.",
			Sensitive: true,
		}, {
			Name: "region",
			Help: "Region to connect to.\n\nLeave blank if you are using an S3 clone and you don't have a region.",
		}, {
			Name: "endpoint",
			Help: "Endpoint for S3 API.\n\nRequired when using an S3 clone.",
		}, {
			Name: "location_constraint",
			Help: "Location constraint - must be set to match the Region.\n\nLeave blank if not sure. Used when creating buckets only.",
		}, {
			Name: "acl",
			Help: `Canned ACL used when creating buckets and storing or copying objects.

This ACL is used for creating objects and if bucket_acl isn't set, for creating buckets too.

For more info visit https://docs.aws.amazon.com/AmazonS3/latest/dev/acl-overview.html#canned-acl

Note that this ACL is applied when server-side copying objects as S3
doesn't copy the ACL from the source but rather writes a fresh one.

If the acl is an empty string then no X-Amz-Acl: header is added and
the default (private) will be used.
`,
		}, {
			Name: "bucket_acl",
			Help: `Canned ACL used when creating buckets.

For more info visit https://docs.aws.amazon.com/AmazonS3/latest/dev/acl-overview.html#canned-acl

Note that this ACL is applied when only when creating buckets.  If it
isn't set then "acl" is used instead.

If the "acl" and "bucket_acl" are empty strings then no X-Amz-Acl:
header is added and the default (private) will be used.
`,
			Advanced: true,
			Examples: []fs.OptionExample{{
				Value: "private",
				Help:  "Owner gets FULL_CONTROL.\nNo one else has access rights (default).",
			}, {
				Value: "public-read",
				Help:  "Owner gets FULL_CONTROL.\nThe AllUsers group gets READ access.",
			}, {
				Value: "public-read-write",
				Help:  "Owner gets FULL_CONTROL.\nThe AllUsers group gets READ and WRITE access.\nGranting this on a bucket is generally not recommended.",
			}, {
				Value: "authenticated-read",
				Help:  "Owner gets FULL_CONTROL.\nThe AuthenticatedUsers group gets READ access.",
			}},
		}, {
			Name:     "requester_pays",
			Help:     "Enables requester pays option when interacting with S3 bucket.",
			Default:  false,
			Advanced: true,
		}, {
			Name: "server_side_encryption",
			Help: "The server-side encryption algorithm used when storing this object in S3.",
		}, {
			Name:     "sse_customer_algorithm",
			Help:     "If using SSE-C, the server-side encryption algorithm used when storing this object in S3.",
			Advanced: true,
			Examples: []fs.OptionExample{{
				Value: "",
				Help:  "None",
			}, {
				Value: "AES256",
				Help:  "AES256",
			}},
		}, {
			Name: "sse_kms_key_id",
			Help: "If using KMS ID you must provide the ARN of Key.",
			Examples: []fs.OptionExample{{
				Value: "",
				Help:  "None",
			}, {
				Value: "arn:aws:kms:us-east-1:*",
				Help:  "arn:aws:kms:*",
			}},
			Sensitive: true,
		}, {
			Name: "sse_customer_key",
			Help: `To use SSE-C you may provide the secret encryption key used to encrypt/decrypt your data.

Alternatively you can provide --sse-customer-key-base64.`,
			Advanced: true,
			Examples: []fs.OptionExample{{
				Value: "",
				Help:  "None",
			}},
			Sensitive: true,
		}, {
			Name: "sse_customer_key_base64",
			Help: `If using SSE-C you must provide the secret encryption key encoded in base64 format to encrypt/decrypt your data.

Alternatively you can provide --sse-customer-key.`,
			Advanced: true,
			Examples: []fs.OptionExample{{
				Value: "",
				Help:  "None",
			}},
			Sensitive: true,
		}, {
			Name: "sse_customer_key_md5",
			Help: `If using SSE-C you may provide the secret encryption key MD5 checksum (optional).

If you leave it blank, this is calculated automatically from the sse_customer_key provided.
`,
			Advanced: true,
			Examples: []fs.OptionExample{{
				Value: "",
				Help:  "None",
			}},
			Sensitive: true,
		}, {
			Name: "storage_class",
			Help: "The storage class to use when storing new objects in S3.",
		}, {
			Name: "upload_cutoff",
			Help: `Cutoff for switching to chunked upload.

Any files larger than this will be uploaded in chunks of chunk_size.
The minimum is 0 and the maximum is 5 GiB.`,
			Default:  defaultUploadCutoff,
			Advanced: true,
		}, {
			Name: "chunk_size",
			Help: `Chunk size to use for uploading.

When uploading files larger than upload_cutoff or files with unknown
size (e.g. from "rclone rcat" or uploaded with "rclone mount" or google
photos or google docs) they will be uploaded as multipart uploads
using this chunk size.

Note that "--s3-upload-concurrency" chunks of this size are buffered
in memory per transfer.

If you are transferring large files over high-speed links and you have
enough memory, then increasing this will speed up the transfers.

Rclone will automatically increase the chunk size when uploading a
large file of known size to stay below the 10,000 chunks limit.

Files of unknown size are uploaded with the configured
chunk_size. Since the default chunk size is 5 MiB and there can be at
most 10,000 chunks, this means that by default the maximum size of
a file you can stream upload is 48 GiB.  If you wish to stream upload
larger files then you will need to increase chunk_size.

Increasing the chunk size decreases the accuracy of the progress
statistics displayed with "-P" flag. Rclone treats chunk as sent when
it's buffered by the AWS SDK, when in fact it may still be uploading.
A bigger chunk size means a bigger AWS SDK buffer and progress
reporting more deviating from the truth.
`,
			Default:  minChunkSize,
			Advanced: true,
		}, {
			Name: "max_upload_parts",
			Help: `Maximum number of parts in a multipart upload.

This option defines the maximum number of multipart chunks to use
when doing a multipart upload.

This can be useful if a service does not support the AWS S3
specification of 10,000 chunks.

Rclone will automatically increase the chunk size when uploading a
large file of a known size to stay below this number of chunks limit.
`,
			Default:  maxUploadParts,
			Advanced: true,
		}, {
			Name: "copy_cutoff",
			Help: `Cutoff for switching to multipart copy.

Any files larger than this that need to be server-side copied will be
copied in chunks of this size.

The minimum is 0 and the maximum is 5 GiB.`,
			Default:  fs.SizeSuffix(maxSizeForCopy),
			Advanced: true,
		}, {
			Name: "disable_checksum",
			Help: `Don't store MD5 checksum with object metadata.

Normally rclone will calculate the MD5 checksum of the input before
uploading it so it can add it to metadata on the object. This is great
for data integrity checking but can cause long delays for large files
to start uploading.`,
			Default:  false,
			Advanced: true,
		}, {
			Name: "shared_credentials_file",
			Help: `Path to the shared credentials file.

If env_auth = true then rclone can use a shared credentials file.

If this variable is empty rclone will look for the
"AWS_SHARED_CREDENTIALS_FILE" env variable. If the env value is empty
it will default to the current user's home directory.

    Linux/OSX: "$HOME/.aws/credentials"
    Windows:   "%USERPROFILE%\.aws\credentials"
`,
			Advanced: true,
		}, {
			Name: "profile",
			Help: `Profile to use in the shared credentials file.

If env_auth = true then rclone can use a shared credentials file. This
variable controls which profile is used in that file.

If empty it will default to the environment variable "AWS_PROFILE" or
"default" if that environment variable is also not set.
`,
			Advanced: true,
		}, {
			Name:      "session_token",
			Help:      "An AWS session token.",
			Advanced:  true,
			Sensitive: true,
		}, {
			Name: "upload_concurrency",
			Help: `Concurrency for multipart uploads and copies.

This is the number of chunks of the same file that are uploaded
concurrently for multipart uploads and copies.

If you are uploading small numbers of large files over high-speed links
and these uploads do not fully utilize your bandwidth, then increasing
this may help to speed up the transfers.`,
			Default:  4,
			Advanced: true,
		}, {
			Name: "force_path_style",
			Help: `If true use path style access if false use virtual hosted style.

If this is true (the default) then rclone will use path style access,
if false then rclone will use virtual path style. See [the AWS S3
docs](https://docs.aws.amazon.com/AmazonS3/latest/dev/UsingBucket.html#access-bucket-intro)
for more info.

Some providers (e.g. AWS, Aliyun OSS, Netease COS, or Tencent COS) require this set to
false - rclone will do this automatically based on the provider
setting.

Note that if your bucket isn't a valid DNS name, i.e. has '.' or '_' in,
you'll need to set this to true.
`,
			Default:  true,
			Advanced: true,
		}, {
			Name: "v2_auth",
			Help: `If true use v2 authentication.

If this is false (the default) then rclone will use v4 authentication.
If it is set then rclone will use v2 authentication.

Use this only if v4 signatures don't work, e.g. pre Jewel/v10 CEPH.`,
			Default:  false,
			Advanced: true,
		}, {
			Name: "use_dual_stack",
			Help: `If true use AWS S3 dual-stack endpoint (IPv6 support).

See [AWS Docs on Dualstack Endpoints](https://docs.aws.amazon.com/AmazonS3/latest/userguide/dual-stack-endpoints.html)`,
			Default:  false,
			Advanced: true,
		}, {
			Name: "use_accelerate_endpoint",
			Help: `If true use the AWS S3 accelerated endpoint.

See: [AWS S3 Transfer acceleration](https://docs.aws.amazon.com/AmazonS3/latest/dev/transfer-acceleration-examples.html)`,
			Default:  false,
			Advanced: true,
		}, {
			Name:     "use_arn_region",
			Help:     `If true, enables arn region support for the service.`,
			Default:  false,
			Advanced: true,
		}, {
			Name: "leave_parts_on_error",
			Help: `If true avoid calling abort upload on a failure, leaving all successfully uploaded parts on S3 for manual recovery.

It should be set to true for resuming uploads across different sessions.

WARNING: Storing parts of an incomplete multipart upload counts towards space usage on S3 and will add additional costs if not cleaned up.
`,
			Default:  false,
			Advanced: true,
		}, {
			Name: "list_chunk",
			Help: `Size of listing chunk (response list for each ListObject S3 request).

This option is also known as "MaxKeys", "max-items", or "page-size" from the AWS S3 specification.
Most services truncate the response list to 1000 objects even if requested more than that.
In AWS S3 this is a global maximum and cannot be changed, see [AWS S3](https://docs.aws.amazon.com/cli/latest/reference/s3/ls.html).
In Ceph, this can be increased with the "rgw list buckets max chunk" option.
`,
			Default:  1000,
			Advanced: true,
		}, {
			Name: "list_version",
			Help: `Version of ListObjects to use: 1,2 or 0 for auto.

When S3 originally launched it only provided the ListObjects call to
enumerate objects in a bucket.

However in May 2016 the ListObjectsV2 call was introduced. This is
much higher performance and should be used if at all possible.

If set to the default, 0, rclone will guess according to the provider
set which list objects method to call. If it guesses wrong, then it
may be set manually here.
`,
			Default:  0,
			Advanced: true,
		}, {
			Name: "list_url_encode",
			Help: `Whether to url encode listings: true/false/unset

Some providers support URL encoding listings and where this is
available this is more reliable when using control characters in file
names. If this is set to unset (the default) then rclone will choose
according to the provider setting what to apply, but you can override
rclone's choice here.
`,
			Default:  fs.Tristate{},
			Advanced: true,
		}, {
			Name: "no_check_bucket",
			Help: `If set, don't attempt to check the bucket exists or create it.

This can be useful when trying to minimise the number of transactions
rclone does if you know the bucket exists already.

It can also be needed if the user you are using does not have bucket
creation permissions. Before v1.52.0 this would have passed silently
due to a bug.
`,
			Default:  false,
			Advanced: true,
		}, {
			Name: "no_head",
			Help: `If set, don't HEAD uploaded objects to check integrity.

This can be useful when trying to minimise the number of transactions
rclone does.

Setting it means that if rclone receives a 200 OK message after
uploading an object with PUT then it will assume that it got uploaded
properly.

In particular it will assume:

- the metadata, including modtime, storage class and content type was as uploaded
- the size was as uploaded

It reads the following items from the response for a single part PUT:

- the MD5SUM
- The uploaded date

For multipart uploads these items aren't read.

If an source object of unknown length is uploaded then rclone **will** do a
HEAD request.

Setting this flag increases the chance for undetected upload failures,
in particular an incorrect size, so it isn't recommended for normal
operation. In practice the chance of an undetected upload failure is
very small even with this flag.
`,
			Default:  false,
			Advanced: true,
		}, {
			Name:     "no_head_object",
			Help:     `If set, do not do HEAD before GET when getting objects.`,
			Default:  false,
			Advanced: true,
		}, {
			Name:     config.ConfigEncoding,
			Help:     config.ConfigEncodingHelp,
			Advanced: true,
			// Any UTF-8 character is valid in a key, however it can't handle
			// invalid UTF-8 and / have a special meaning.
			//
			// The SDK can't seem to handle uploading files called '.'
			//
			// FIXME would be nice to add
			// - initial / encoding
			// - doubled / encoding
			// - trailing / encoding
			// so that AWS keys are always valid file names
			Default: encoder.EncodeInvalidUtf8 |
				encoder.EncodeSlash |
				encoder.EncodeDot,
		}, {
			Name:     "memory_pool_flush_time",
			Default:  fs.Duration(time.Minute),
			Advanced: true,
			Hide:     fs.OptionHideBoth,
			Help:     `How often internal memory buffer pools will be flushed. (no longer used)`,
		}, {
			Name:     "memory_pool_use_mmap",
			Default:  false,
			Advanced: true,
			Hide:     fs.OptionHideBoth,
			Help:     `Whether to use mmap buffers in internal memory pool. (no longer used)`,
		}, {
			Name:     "disable_http2",
			Default:  false,
			Advanced: true,
			Help: `Disable usage of http2 for S3 backends.

There is currently an unsolved issue with the s3 (specifically minio) backend
and HTTP/2.  HTTP/2 is enabled by default for the s3 backend but can be
disabled here.  When the issue is solved this flag will be removed.

See: https://github.com/rclone/rclone/issues/4673, https://github.com/rclone/rclone/issues/3631

`,
		}, {
			Name: "download_url",
			Help: `Custom endpoint for downloads.
This is usually set to a CloudFront CDN URL as AWS S3 offers
cheaper egress for data downloaded through the CloudFront network.`,
			Advanced: true,
		}, {
			Name:     "directory_markers",
			Default:  false,
			Advanced: true,
			Help: `Upload an empty object with a trailing slash when a new directory is created

Empty folders are unsupported for bucket based remotes, this option creates an empty
object ending with "/", to persist the folder.
`,
		}, {
			Name: "use_multipart_etag",
			Help: `Whether to use ETag in multipart uploads for verification

This should be true, false or left unset to use the default for the provider.
`,
			Default:  fs.Tristate{},
			Advanced: true,
		}, {
			Name: "use_unsigned_payload",
			Help: `Whether to use an unsigned payload in PutObject

Rclone has to avoid the AWS SDK seeking the body when calling
PutObject. The AWS provider can add checksums in the trailer to avoid
seeking but other providers can't.

This should be true, false or left unset to use the default for the provider.
`,
			Default:  fs.Tristate{},
			Advanced: true,
		}, {
			Name: "use_presigned_request",
			Help: `Whether to use a presigned request or PutObject for single part uploads

If this is false rclone will use PutObject from the AWS SDK to upload
an object.

Versions of rclone < 1.59 use presigned requests to upload a single
part object and setting this flag to true will re-enable that
functionality. This shouldn't be necessary except in exceptional
circumstances or for testing.
`,
			Default:  false,
			Advanced: true,
		}, {
			Name: "use_data_integrity_protections",
			Help: `If true use AWS S3 data integrity protections.

See [AWS Docs on Data Integrity Protections](https://docs.aws.amazon.com/sdkref/latest/guide/feature-dataintegrity.html)`,
			Default:  fs.Tristate{},
			Advanced: true,
		}, {
			Name:     "versions",
			Help:     "Include old versions in directory listings.",
			Default:  false,
			Advanced: true,
		}, {
			Name: "version_at",
			Help: `Show file versions as they were at the specified time.

The parameter should be a date, "2006-01-02", datetime "2006-01-02
15:04:05" or a duration for that long ago, eg "100d" or "1h".

Note that when using this no file write operations are permitted,
so you can't upload files or delete them.

See [the time option docs](/docs/#time-options) for valid formats.
`,
			Default:  fs.Time{},
			Advanced: true,
		}, {
			Name: "version_deleted",
			Help: `Show deleted file markers when using versions.

This shows deleted file markers in the listing when using versions. These will appear
as 0 size files. The only operation which can be performed on them is deletion.

Deleting a delete marker will reveal the previous version.

Deleted files will always show with a timestamp.
`,
			Default:  false,
			Advanced: true,
		}, {
			Name: "decompress",
			Help: `If set this will decompress gzip encoded objects.

It is possible to upload objects to S3 with "Content-Encoding: gzip"
set. Normally rclone will download these files as compressed objects.

If this flag is set then rclone will decompress these files with
"Content-Encoding: gzip" as they are received. This means that rclone
can't check the size and hash but the file contents will be decompressed.
`,
			Advanced: true,
			Default:  false,
		}, {
			Name: "might_gzip",
			Help: strings.ReplaceAll(`Set this if the backend might gzip objects.

Normally providers will not alter objects when they are downloaded. If
an object was not uploaded with |Content-Encoding: gzip| then it won't
be set on download.

However some providers may gzip objects even if they weren't uploaded
with |Content-Encoding: gzip| (eg Cloudflare).

A symptom of this would be receiving errors like

    ERROR corrupted on transfer: sizes differ NNN vs MMM

If you set this flag and rclone downloads an object with
Content-Encoding: gzip set and chunked transfer encoding, then rclone
will decompress the object on the fly.

If this is set to unset (the default) then rclone will choose
according to the provider setting what to apply, but you can override
rclone's choice here.
`, "|", "`"),
			Default:  fs.Tristate{},
			Advanced: true,
		}, {
			Name: "use_accept_encoding_gzip",
			Help: strings.ReplaceAll(`Whether to send |Accept-Encoding: gzip| header.

By default, rclone will append |Accept-Encoding: gzip| to the request to download
compressed objects whenever possible.

However some providers such as Google Cloud Storage may alter the HTTP headers, breaking
the signature of the request.

A symptom of this would be receiving errors like

	SignatureDoesNotMatch: The request signature we calculated does not match the signature you provided.

In this case, you might want to try disabling this option.
`, "|", "`"),
			Default:  fs.Tristate{},
			Advanced: true,
		}, {
			Name:     "no_system_metadata",
			Help:     `Suppress setting and reading of system metadata`,
			Advanced: true,
			Default:  false,
		}, {
			Name:     "sts_endpoint",
			Help:     "Endpoint for STS (deprecated).\n\nLeave blank if using AWS to use the default endpoint for the region.",
			Advanced: true,
			Hide:     fs.OptionHideBoth,
		}, {
			Name: "use_already_exists",
			Help: strings.ReplaceAll(`Set if rclone should report BucketAlreadyExists errors on bucket creation.

At some point during the evolution of the s3 protocol, AWS started
returning an |AlreadyOwnedByYou| error when attempting to create a
bucket that the user already owned, rather than a
|BucketAlreadyExists| error.

Unfortunately exactly what has been implemented by s3 clones is a
little inconsistent, some return |AlreadyOwnedByYou|, some return
|BucketAlreadyExists| and some return no error at all.

This is important to rclone because it ensures the bucket exists by
creating it on quite a lot of operations (unless
|--s3-no-check-bucket| is used).

If rclone knows the provider can return |AlreadyOwnedByYou| or returns
no error then it can report |BucketAlreadyExists| errors when the user
attempts to create a bucket not owned by them. Otherwise rclone
ignores the |BucketAlreadyExists| error which can lead to confusion.

This should be automatically set correctly for all providers rclone
knows about - please make a bug report if not.
`, "|", "`"),
			Default:  fs.Tristate{},
			Advanced: true,
		}, {
			Name: "use_multipart_uploads",
			Help: `Set if rclone should use multipart uploads.

You can change this if you want to disable the use of multipart uploads.
This shouldn't be necessary in normal operation.

This should be automatically set correctly for all providers rclone
knows about - please make a bug report if not.
`,
			Default:  fs.Tristate{},
			Advanced: true,
		}, {
			Name: "use_x_id",
			Help: `Set if rclone should add x-id URL parameters.

You can change this if you want to disable the AWS SDK from
adding x-id URL parameters.

This shouldn't be necessary in normal operation.

This should be automatically set correctly for all providers rclone
knows about - please make a bug report if not.
`,
			Default:  fs.Tristate{},
			Advanced: true,
		}, {
			Name: "sign_accept_encoding",
			Help: `Set if rclone should include Accept-Encoding as part of the signature.

You can change this if you want to stop rclone including
Accept-Encoding as part of the signature.

This shouldn't be necessary in normal operation.

This should be automatically set correctly for all providers rclone
knows about - please make a bug report if not.
`,
			Default:  fs.Tristate{},
			Advanced: true,
		}, {
			Name: "directory_bucket",
			Help: strings.ReplaceAll(`Set to use AWS Directory Buckets

If you are using an AWS Directory Bucket then set this flag.

This will ensure no |Content-Md5| headers are sent and ensure |ETag|
headers are not interpreted as MD5 sums. |X-Amz-Meta-Md5chksum| will
be set on all objects whether single or multipart uploaded.

This also sets |no_check_bucket = true|.

Note that Directory Buckets do not support:

- Versioning
- |Content-Encoding: gzip|

Rclone limitations with Directory Buckets:

- rclone does not support creating Directory Buckets with |rclone mkdir|
- ... or removing them with |rclone rmdir| yet
- Directory Buckets do not appear when doing |rclone lsf| at the top level.
- Rclone can't remove auto created directories yet. In theory this should
  work with |directory_markers = true| but it doesn't.
- Directories don't seem to appear in recursive (ListR) listings.
`, "|", "`"),
			Default:  false,
			Advanced: true,
		}, {
			Name: "sdk_log_mode",
			Help: strings.ReplaceAll(`Set to debug the SDK

This can be set to a comma separated list of the following functions:

- |Signing|
- |Retries|
- |Request|
- |RequestWithBody|
- |Response|
- |ResponseWithBody|
- |DeprecatedUsage|
- |RequestEventMessage|
- |ResponseEventMessage|

Use |Off| to disable and |All| to set all log levels. You will need to
use |-vv| to see the debug level logs.
`, "|", "`"),
			Default:  sdkLogMode(0),
			Advanced: true,
		}, {
			Name: "ibm_api_key",
			Help: "IBM API Key to be used to obtain IAM token",
		}, {
			Name: "ibm_resource_instance_id",
			Help: "IBM service instance id",
		},
		}}))
}

// Constants
const (
	metaMtime   = "mtime"     // the meta key to store mtime in - e.g. X-Amz-Meta-Mtime
	metaMD5Hash = "md5chksum" // the meta key to store md5hash in
	// The maximum size of object we can COPY - this should be 5 GiB but is < 5 GB for b2 compatibility
	// See https://forum.rclone.org/t/copying-files-within-a-b2-bucket/16680/76
	maxSizeForCopy      = 4768 * 1024 * 1024
	maxUploadParts      = 10000 // maximum allowed number of parts in a multi-part upload
	minChunkSize        = fs.SizeSuffix(1024 * 1024 * 5)
	defaultUploadCutoff = fs.SizeSuffix(200 * 1024 * 1024)
	maxUploadCutoff     = fs.SizeSuffix(5 * 1024 * 1024 * 1024)
	minSleep            = 10 * time.Millisecond           // In case of error, start at 10ms sleep.
	maxExpireDuration   = fs.Duration(7 * 24 * time.Hour) // max expiry is 1 week
)

type (
	sdkLogMode        = fs.Bits[sdkLogModeChoices]
	sdkLogModeChoices struct{}
)

func (sdkLogModeChoices) Choices() []fs.BitsChoicesInfo {
	return []fs.BitsChoicesInfo{
		{Bit: uint64(0), Name: "Off"},
		{Bit: uint64(aws.LogSigning), Name: "Signing"},
		{Bit: uint64(aws.LogRetries), Name: "Retries"},
		{Bit: uint64(aws.LogRequest), Name: "Request"},
		{Bit: uint64(aws.LogRequestWithBody), Name: "RequestWithBody"},
		{Bit: uint64(aws.LogResponse), Name: "Response"},
		{Bit: uint64(aws.LogResponseWithBody), Name: "ResponseWithBody"},
		{Bit: uint64(aws.LogDeprecatedUsage), Name: "DeprecatedUsage"},
		{Bit: uint64(aws.LogRequestEventMessage), Name: "RequestEventMessage"},
		{Bit: uint64(aws.LogResponseEventMessage), Name: "ResponseEventMessage"},
		{Bit: math.MaxUint64, Name: "All"},
	}
}

// globals
var (
	errNotWithVersionAt = errors.New("can't modify or delete files in --s3-version-at mode")
)

// system metadata keys which this backend owns
var systemMetadataInfo = map[string]fs.MetadataHelp{
	"cache-control": {
		Help:    "Cache-Control header",
		Type:    "string",
		Example: "no-cache",
	},
	"content-disposition": {
		Help:    "Content-Disposition header",
		Type:    "string",
		Example: "inline",
	},
	"content-encoding": {
		Help:    "Content-Encoding header",
		Type:    "string",
		Example: "gzip",
	},
	"content-language": {
		Help:    "Content-Language header",
		Type:    "string",
		Example: "en-US",
	},
	"content-type": {
		Help:    "Content-Type header",
		Type:    "string",
		Example: "text/plain",
	},
	// "tagging": {
	// 	Help:    "x-amz-tagging header",
	// 	Type:    "string",
	// 	Example: "tag1=value1&tag2=value2",
	// },
	"tier": {
		Help:     "Tier of the object",
		Type:     "string",
		Example:  "GLACIER",
		ReadOnly: true,
	},
	"mtime": {
		Help:    "Time of last modification, read from rclone metadata",
		Type:    "RFC 3339",
		Example: "2006-01-02T15:04:05.999999999Z07:00",
	},
	"btime": {
		Help:     "Time of file birth (creation) read from Last-Modified header",
		Type:     "RFC 3339",
		Example:  "2006-01-02T15:04:05.999999999Z07:00",
		ReadOnly: true,
	},
}

// Options defines the configuration for this backend
type Options struct {
	Provider                    string               `config:"provider"`
	EnvAuth                     bool                 `config:"env_auth"`
	AccessKeyID                 string               `config:"access_key_id"`
	SecretAccessKey             string               `config:"secret_access_key"`
	Region                      string               `config:"region"`
	Endpoint                    string               `config:"endpoint"`
	STSEndpoint                 string               `config:"sts_endpoint"`
	UseDualStack                bool                 `config:"use_dual_stack"`
	LocationConstraint          string               `config:"location_constraint"`
	ACL                         string               `config:"acl"`
	BucketACL                   string               `config:"bucket_acl"`
	RequesterPays               bool                 `config:"requester_pays"`
	ServerSideEncryption        string               `config:"server_side_encryption"`
	SSEKMSKeyID                 string               `config:"sse_kms_key_id"`
	SSECustomerAlgorithm        string               `config:"sse_customer_algorithm"`
	SSECustomerKey              string               `config:"sse_customer_key"`
	SSECustomerKeyBase64        string               `config:"sse_customer_key_base64"`
	SSECustomerKeyMD5           string               `config:"sse_customer_key_md5"`
	StorageClass                string               `config:"storage_class"`
	UploadCutoff                fs.SizeSuffix        `config:"upload_cutoff"`
	CopyCutoff                  fs.SizeSuffix        `config:"copy_cutoff"`
	ChunkSize                   fs.SizeSuffix        `config:"chunk_size"`
	MaxUploadParts              int                  `config:"max_upload_parts"`
	DisableChecksum             bool                 `config:"disable_checksum"`
	SharedCredentialsFile       string               `config:"shared_credentials_file"`
	Profile                     string               `config:"profile"`
	SessionToken                string               `config:"session_token"`
	UploadConcurrency           int                  `config:"upload_concurrency"`
	ForcePathStyle              bool                 `config:"force_path_style"`
	V2Auth                      bool                 `config:"v2_auth"`
	UseAccelerateEndpoint       bool                 `config:"use_accelerate_endpoint"`
	UseARNRegion                bool                 `config:"use_arn_region"`
	LeavePartsOnError           bool                 `config:"leave_parts_on_error"`
	ListChunk                   int32                `config:"list_chunk"`
	ListVersion                 int                  `config:"list_version"`
	ListURLEncode               fs.Tristate          `config:"list_url_encode"`
	NoCheckBucket               bool                 `config:"no_check_bucket"`
	NoHead                      bool                 `config:"no_head"`
	NoHeadObject                bool                 `config:"no_head_object"`
	Enc                         encoder.MultiEncoder `config:"encoding"`
	DisableHTTP2                bool                 `config:"disable_http2"`
	DownloadURL                 string               `config:"download_url"`
	DirectoryMarkers            bool                 `config:"directory_markers"`
	UseMultipartEtag            fs.Tristate          `config:"use_multipart_etag"`
	UsePresignedRequest         bool                 `config:"use_presigned_request"`
	UseDataIntegrityProtections fs.Tristate          `config:"use_data_integrity_protections"`
	Versions                    bool                 `config:"versions"`
	VersionAt                   fs.Time              `config:"version_at"`
	VersionDeleted              bool                 `config:"version_deleted"`
	Decompress                  bool                 `config:"decompress"`
	MightGzip                   fs.Tristate          `config:"might_gzip"`
	UseAcceptEncodingGzip       fs.Tristate          `config:"use_accept_encoding_gzip"`
	NoSystemMetadata            bool                 `config:"no_system_metadata"`
	UseAlreadyExists            fs.Tristate          `config:"use_already_exists"`
	UseMultipartUploads         fs.Tristate          `config:"use_multipart_uploads"`
	UseUnsignedPayload          fs.Tristate          `config:"use_unsigned_payload"`
	SDKLogMode                  sdkLogMode           `config:"sdk_log_mode"`
	DirectoryBucket             bool                 `config:"directory_bucket"`
	IBMAPIKey                   string               `config:"ibm_api_key"`
	IBMInstanceID               string               `config:"ibm_resource_instance_id"`
	UseXID                      fs.Tristate          `config:"use_x_id"`
	SignAcceptEncoding          fs.Tristate          `config:"sign_accept_encoding"`
}

// Fs represents a remote s3 server
type Fs struct {
	name           string          // the name of the remote
	root           string          // root of the bucket - ignore all objects above this
	opt            Options         // parsed options
	ci             *fs.ConfigInfo  // global config
	ctx            context.Context // global context for reading config
	features       *fs.Features    // optional features
	c              *s3.Client      // the connection to the s3 server
	rootBucket     string          // bucket part of root (if any)
	rootDirectory  string          // directory part of root (if any)
	cache          *bucket.Cache   // cache for bucket creation status
	pacer          *fs.Pacer       // To pace the API calls
	srv            *http.Client    // a plain http client
	srvRest        *rest.Client    // the rest connection to the server
	etagIsNotMD5   bool            // if set ETags are not MD5s
	versioningMu   sync.Mutex
	versioning     fs.Tristate // if set bucket is using versions
	warnCompressed sync.Once   // warn once about compressed files
}

// Object describes a s3 object
type Object struct {
	// Will definitely have everything but meta which may be nil
	//
	// List will read everything but meta & mimeType - to fill
	// that in you need to call readMetaData
	fs           *Fs               // what this object is part of
	remote       string            // The remote path
	md5          string            // md5sum of the object
	bytes        int64             // size of the object
	lastModified time.Time         // Last modified
	meta         map[string]string // The object metadata if known - may be nil - with lower case keys
	mimeType     string            // MimeType of object - may be ""
	versionID    *string           // If present this points to an object version

	// Metadata as pointers to strings as they often won't be present
	storageClass       *string // e.g. GLACIER
	cacheControl       *string // Cache-Control: header
	contentDisposition *string // Content-Disposition: header
	contentEncoding    *string // Content-Encoding: header
	contentLanguage    *string // Content-Language: header
}

// safely dereference the pointer, returning a zero T if nil
func deref[T any](p *T) T {
	if p == nil {
		var zero T
		return zero
	}
	return *p
}

// gets an http status code from err or returns -1
func getHTTPStatusCode(err error) int {
	var httpErr interface{ HTTPStatusCode() int }
	if errors.As(err, &httpErr) {
		return httpErr.HTTPStatusCode()
	}
	return -1
}

// ------------------------------------------------------------

// Name of the remote (as passed into NewFs)
func (f *Fs) Name() string {
	return f.name
}

// Root of the remote (as passed into NewFs)
func (f *Fs) Root() string {
	return f.root
}

// String converts this Fs to a string
func (f *Fs) String() string {
	if f.rootBucket == "" {
		return "S3 root"
	}
	if f.rootDirectory == "" {
		return fmt.Sprintf("S3 bucket %s", f.rootBucket)
	}
	return fmt.Sprintf("S3 bucket %s path %s", f.rootBucket, f.rootDirectory)
}

// Features returns the optional features of this Fs
func (f *Fs) Features() *fs.Features {
	return f.features
}

// retryErrorCodes is a slice of error codes that we will retry
// See: https://docs.aws.amazon.com/AmazonS3/latest/API/ErrorResponses.html
var retryErrorCodes = []int{
	429, // Too Many Requests
	500, // Internal Server Error - "We encountered an internal error. Please try again."
	503, // Service Unavailable/Slow Down - "Reduce your request rate"
}

// S3 is pretty resilient, and the built in retry handling is probably sufficient
// as it should notice closed connections and timeouts which are the most likely
// sort of failure modes
func (f *Fs) shouldRetry(ctx context.Context, err error) (bool, error) {
	if fserrors.ContextError(ctx, &err) {
		return false, err
	}
	// https://github.com/aws/aws-sdk-go-v2/blob/main/CHANGELOG.md#error-handling
	// If this is an awserr object, try and extract more useful information to determine if we should retry
	var awsError smithy.APIError
	if errors.As(err, &awsError) {
		// Simple case, check the original embedded error in case it's generically retryable
		if fserrors.ShouldRetry(awsError) {
			return true, err
		}
		// If it is a timeout then we want to retry that
		if awsError.ErrorCode() == "RequestTimeout" {
			return true, err
		}
	}
	// Check http status code if available
	if httpStatusCode := getHTTPStatusCode(err); httpStatusCode > 0 {
		// 301 if wrong region for bucket - can only update if running from a bucket
		if f.rootBucket != "" {
			if httpStatusCode == http.StatusMovedPermanently {
				urfbErr := f.updateRegionForBucket(ctx, f.rootBucket)
				if urfbErr != nil {
					fs.Errorf(f, "Failed to update region for bucket: %v", urfbErr)
					return false, err
				}
				return true, err
			}
		}
		if slices.Contains(retryErrorCodes, httpStatusCode) {
			return true, err
		}
	}
	// Ok, not an awserr, check for generic failure conditions
	return fserrors.ShouldRetry(err), err
}

// parsePath parses a remote 'url'
func parsePath(path string) (root string) {
	root = strings.Trim(path, "/")
	return
}

// split returns bucket and bucketPath from the rootRelativePath
// relative to f.root
func (f *Fs) split(rootRelativePath string) (bucketName, bucketPath string) {
	bucketName, bucketPath = bucket.Split(bucket.Join(f.root, rootRelativePath))
	if f.opt.DirectoryMarkers && strings.HasSuffix(bucketPath, "//") {
		bucketPath = bucketPath[:len(bucketPath)-1]
	}
	return f.opt.Enc.FromStandardName(bucketName), f.opt.Enc.FromStandardPath(bucketPath)
}

// split returns bucket and bucketPath from the object
func (o *Object) split() (bucket, bucketPath string) {
	bucket, bucketPath = o.fs.split(o.remote)
	// If there is an object version, then the path may have a
	// version suffix, if so remove it.
	//
	// If we are unlucky enough to have a file name with a valid
	// version path where this wasn't required (eg using
	// --s3-version-at) then this will go wrong.
	if o.versionID != nil {
		_, bucketPath = version.Remove(bucketPath)
	}
	return bucket, bucketPath
}

// getClient makes an http client according to the options
func getClient(ctx context.Context, opt *Options) *http.Client {
	// TODO: Do we need cookies too?
	t := fshttp.NewTransportCustom(ctx, func(t *http.Transport) {
		if opt.DisableHTTP2 {
			t.TLSNextProto = map[string]func(string, *tls.Conn) http.RoundTripper{}
		}
	})
	return &http.Client{
		Transport: t,
	}
}

// Fixup the request if needed.
//
// Google Cloud Storage alters the Accept-Encoding header, which
// breaks the v2 request signature. This is set with opt.SignAcceptEncoding.
//
// It also doesn't like the x-id URL parameter SDKv2 puts in so we
// remove that too. This is set with opt.UseXID.Value.
//
// See https://github.com/aws/aws-sdk-go-v2/issues/1816.
// Adapted from: https://github.com/aws/aws-sdk-go-v2/issues/1816#issuecomment-1927281540
func fixupRequest(o *s3.Options, opt *Options) {
	type ignoredHeadersKey struct{}
	headers := []string{"Accept-Encoding"}

	fixup := middleware.FinalizeMiddlewareFunc(
		"FixupRequest",
		func(ctx context.Context, in middleware.FinalizeInput, next middleware.FinalizeHandler) (out middleware.FinalizeOutput, metadata middleware.Metadata, err error) {
			req, ok := in.Request.(*smithyhttp.Request)
			if !ok {
				return out, metadata, fmt.Errorf("fixupRequest: unexpected request middleware type %T", in.Request)
			}

			if !opt.SignAcceptEncoding.Value {
				// Delete headers from being signed - will restore later
				ignored := make(map[string]string, len(headers))
				for _, h := range headers {
					ignored[h] = req.Header.Get(h)
					req.Header.Del(h)
				}

				// Store ignored on context
				ctx = middleware.WithStackValue(ctx, ignoredHeadersKey{}, ignored)
			}

			if !opt.UseXID.Value {
				// Remove x-id
				if query := req.URL.Query(); query.Has("x-id") {
					query.Del("x-id")
					req.URL.RawQuery = query.Encode()
				}
			}

			return next.HandleFinalize(ctx, in)
		},
	)

	// Restore headers if necessary
	restore := middleware.FinalizeMiddlewareFunc(
		"FixupRequestRestoreHeaders",
		func(ctx context.Context, in middleware.FinalizeInput, next middleware.FinalizeHandler) (out middleware.FinalizeOutput, metadata middleware.Metadata, err error) {
			req, ok := in.Request.(*smithyhttp.Request)
			if !ok {
				return out, metadata, fmt.Errorf("fixupRequest: unexpected request middleware type %T", in.Request)
			}

			if !opt.SignAcceptEncoding.Value {
				// Restore ignored from ctx
				ignored, _ := middleware.GetStackValue(ctx, ignoredHeadersKey{}).(map[string]string)
				for k, v := range ignored {
					req.Header.Set(k, v)
				}
			}

			return next.HandleFinalize(ctx, in)
		},
	)

	o.APIOptions = append(o.APIOptions, func(stack *middleware.Stack) error {
		if err := stack.Finalize.Insert(fixup, "Signing", middleware.Before); err != nil {
			return err
		}
		if err := stack.Finalize.Insert(restore, "Signing", middleware.After); err != nil {
			return err
		}
		return nil
	})
}

// A logger for the S3 SDK
type s3logger struct{}

// Logf is expected to support the standard fmt package "verbs".
func (s3logger) Logf(classification logging.Classification, format string, v ...any) {
	switch classification {
	default:
	case logging.Debug:
		fs.Debugf("S3 SDK", format, v...)
	case logging.Warn:
		fs.Infof("S3 SDK", format, v...)
	}
}

// s3Connection makes a connection to s3
func s3Connection(ctx context.Context, opt *Options, client *http.Client) (s3Client *s3.Client, provider *Provider, err error) {
	ci := fs.GetConfig(ctx)
	var awsConfig aws.Config
	// Make the default static auth
	v := aws.Credentials{
		AccessKeyID:     opt.AccessKeyID,
		SecretAccessKey: opt.SecretAccessKey,
		SessionToken:    opt.SessionToken,
	}
	awsConfig.Credentials = &credentials.StaticCredentialsProvider{Value: v}

	// Try to fill in the config from the environment if env_auth=true
	if opt.EnvAuth && opt.AccessKeyID == "" && opt.SecretAccessKey == "" {

		configOpts := []func(*awsconfig.LoadOptions) error{}
		// Set the name of the profile if supplied
		if opt.Profile != "" {
			configOpts = append(configOpts, awsconfig.WithSharedConfigProfile(opt.Profile))
		}
		// Set the shared config file if supplied
		if opt.SharedCredentialsFile != "" {
			configOpts = append(configOpts, awsconfig.WithSharedConfigFiles([]string{opt.SharedCredentialsFile}))
		}
		awsConfig, err = awsconfig.LoadDefaultConfig(ctx, configOpts...)
		if err != nil {
			return nil, nil, fmt.Errorf("couldn't load configuration with env_auth=true: %w", err)
		}

	} else {
		switch {
		case opt.Provider == "IBMCOS" && opt.V2Auth:
			awsConfig.Credentials = &NoOpCredentialsProvider{}
			fs.Debugf(nil, "Using IBM IAM")
		case opt.AccessKeyID == "" && opt.SecretAccessKey == "":
			// if no access key/secret and iam is explicitly disabled then fall back to anon interaction
			awsConfig.Credentials = aws.AnonymousCredentials{}
			fs.Debugf(nil, "Using anonymous credentials - did you mean to set env_auth=true?")
		case opt.AccessKeyID == "":
			return nil, nil, errors.New("access_key_id not found")
		case opt.SecretAccessKey == "":
			return nil, nil, errors.New("secret_access_key not found")
		default:
			// static credentials are already set
		}
	}

	if opt.Region == "" {
		opt.Region = "us-east-1"
	}

	provider = loadProvider(opt.Provider)
	if provider == nil {
		fs.Logf("s3", "s3 provider %q not known - please set correctly", opt.Provider)
		provider = loadProvider("Other")
	}

	setQuirks(opt, provider)
	awsConfig.RetryMaxAttempts = ci.LowLevelRetries
	awsConfig.HTTPClient = client

	options := []func(*s3.Options){}
	options = append(options, func(s3Opt *s3.Options) {
		s3Opt.UsePathStyle = opt.ForcePathStyle
		s3Opt.UseAccelerate = opt.UseAccelerateEndpoint
		s3Opt.UseARNRegion = opt.UseARNRegion
		// FIXME maybe this should be a tristate so can default to DualStackEndpointStateUnset?
		if opt.UseDualStack {
			s3Opt.EndpointOptions.UseDualStackEndpoint = aws.DualStackEndpointStateEnabled
		} else {
			s3Opt.EndpointOptions.UseDualStackEndpoint = aws.DualStackEndpointStateDisabled
		}
		if !opt.UseDataIntegrityProtections.Value {
			s3Opt.RequestChecksumCalculation = aws.RequestChecksumCalculationWhenRequired
			s3Opt.ResponseChecksumValidation = aws.ResponseChecksumValidationWhenRequired
		}
		// FIXME not ported from SDK v1 - not sure what this does
		// s3Opt.UsEast1RegionalEndpoint = endpoints.RegionalS3UsEast1Endpoint
	})

	if opt.Region != "" {
		awsConfig.Region = opt.Region
	}
	if opt.STSEndpoint != "" {
		// FIXME not sure if anyone is using this
		// Haven't figured out how to do it with the v2 SDK
		return nil, nil, errors.New("--s3-sts-endpoint is no longer supported with the v2 SDK - please make an issue")
	}
	if opt.Endpoint != "" {
		if !strings.HasPrefix(opt.Endpoint, "http") {
			opt.Endpoint = "https://" + opt.Endpoint
		}
		options = append(options, func(s3Opt *s3.Options) {
			s3Opt.BaseEndpoint = &opt.Endpoint
		})
	}

	if opt.V2Auth || opt.Region == "other-v2-signature" {
		fs.Debugf(nil, "Using v2 auth")
		if opt.Provider == "IBMCOS" && opt.IBMAPIKey != "" && opt.IBMInstanceID != "" {
			options = append(options, func(s3Opt *s3.Options) {
				s3Opt.HTTPSignerV4 = &IbmIamSigner{APIKey: opt.IBMAPIKey, InstanceID: opt.IBMInstanceID}
			})
		} else {
			options = append(options, func(s3Opt *s3.Options) {
				s3Opt.HTTPSignerV4 = &v2Signer{opt: opt}
			})
		}
	}

	// Fixup the request if needed
	if !opt.UseXID.Value || !opt.SignAcceptEncoding.Value {
		options = append(options, func(o *s3.Options) {
			fixupRequest(o, opt)
		})
	}

	// Enable SDK logging if requested
	if opt.SDKLogMode != 0 {
		awsConfig.ClientLogMode = aws.ClientLogMode(opt.SDKLogMode)
		awsConfig.Logger = s3logger{}
	}

	c := s3.NewFromConfig(awsConfig, options...)
	return c, provider, nil
}

func checkUploadChunkSize(cs fs.SizeSuffix) error {
	if cs < minChunkSize {
		return fmt.Errorf("%s is less than %s", cs, minChunkSize)
	}
	return nil
}

func (f *Fs) setUploadChunkSize(cs fs.SizeSuffix) (old fs.SizeSuffix, err error) {
	err = checkUploadChunkSize(cs)
	if err == nil {
		old, f.opt.ChunkSize = f.opt.ChunkSize, cs
	}
	return
}

func checkCopyCutoff(cs fs.SizeSuffix) error {
	minCopySize := fs.SizeSuffixBase
	if cs < minCopySize {
		return fmt.Errorf("value is too small (%v is less than %v)", cs, minCopySize)
	}
	return nil
}

func checkUploadCutoff(cs fs.SizeSuffix) error {
	if cs > maxUploadCutoff {
		return fmt.Errorf("%s is greater than %s", cs, maxUploadCutoff)
	}
	return nil
}

func (f *Fs) setUploadCutoff(cs fs.SizeSuffix) (old fs.SizeSuffix, err error) {
	if f.opt.Provider != "Rclone" {
		err = checkUploadCutoff(cs)
	}
	if err == nil {
		old, f.opt.UploadCutoff = f.opt.UploadCutoff, cs
	}
	return
}

func (f *Fs) setCopyCutoff(cs fs.SizeSuffix) (old fs.SizeSuffix, err error) {
	if f.opt.CopyCutoff == math.MaxInt64 {
		return f.opt.CopyCutoff, fmt.Errorf("--s3-copy-cutoff not supported: %w", fs.ErrorNotImplemented)
	}
	err = checkUploadChunkSize(cs)
	if err == nil {
		old, f.opt.CopyCutoff = f.opt.CopyCutoff, cs
	}
	return
}

// setEndpointValueForIDriveE2 gets user region endpoint against the Access Key details by calling the API
func setEndpointValueForIDriveE2(m configmap.Mapper) (err error) {
	value, ok := m.Get(fs.ConfigProvider)
	if !ok || value != "IDrive" {
		return
	}
	value, ok = m.Get("access_key_id")
	if !ok || value == "" {
		return
	}
	client := &http.Client{Timeout: time.Second * 3}
	// API to get user region endpoint against the Access Key details: https://www.idrive.com/e2/guides/get_region_endpoint
	resp, err := client.Post("https://api.idrivee2.com/api/service/get_region_end_point",
		"application/json",
		strings.NewReader(`{"access_key": `+strconv.Quote(value)+`}`))
	if err != nil {
		return
	}
	defer fs.CheckClose(resp.Body, &err)
	decoder := json.NewDecoder(resp.Body)
	var data = &struct {
		RespCode   int    `json:"resp_code"`
		RespMsg    string `json:"resp_msg"`
		DomainName string `json:"domain_name"`
	}{}
	if err = decoder.Decode(data); err == nil && data.RespCode == 0 {
		m.Set("endpoint", data.DomainName)
	}
	return
}

// Set the provider quirks
//
// There should be no testing against opt.Provider anywhere in the
// code except in here to localise the setting of the quirks.
//
// Run the integration tests to check you have the quirks correct.
//
//	go test -v -remote NewS3Provider:
func setQuirks(opt *Options, provider *Provider) {
	// Set tristate to the ultimate default value or the override
	// in provider.Quirks. Pass in the ultimate default as value.
	set := func(tristate *fs.Tristate, value bool, override *bool) {
		if override != nil {
			value = *override
		}
		if !tristate.Valid {
			tristate.Valid = true
			tristate.Value = value
		}
	}

	// Set Path Style vs Virtual Host style
	var virtualHostStyle = true // Default use bucket.provider.com instead of putting the bucket in the URL
	if provider.Quirks.ForcePathStyle != nil {
		// "don't force path style" so inverted
		virtualHostStyle = !*provider.Quirks.ForcePathStyle
	}
	if virtualHostStyle || opt.UseAccelerateEndpoint {
		opt.ForcePathStyle = false
	}

	// Set the correct list version if not manually set
	if opt.ListVersion == 0 {
		if provider.Quirks.ListVersion != nil {
			opt.ListVersion = *provider.Quirks.ListVersion
		} else {
			opt.ListVersion = 2
		}
	}

	// Set the copy cutoff if not manually set
	// Check equality with strings as config values get round tripped via strings
	if opt.CopyCutoff.String() == fs.SizeSuffix(maxSizeForCopy).String() {
		if provider.Quirks.CopyCutoff != nil {
			opt.CopyCutoff = fs.SizeSuffix(*provider.Quirks.CopyCutoff)
		}
	}

	// Clip the max upload parts to the quirk
	if provider.Quirks.MaxUploadParts != nil {
		opt.MaxUploadParts = min(opt.MaxUploadParts, *provider.Quirks.MaxUploadParts)
	}

	// Clip the chunk size to the quirk
	if provider.Quirks.MinChunkSize != nil {
		opt.ChunkSize = max(opt.ChunkSize, fs.SizeSuffix(*provider.Quirks.MinChunkSize))
	}

	// Set new style Tristate quirks
	set(&opt.ListURLEncode, true, provider.Quirks.ListURLEncode)
	set(&opt.UseMultipartEtag, true, provider.Quirks.UseMultipartEtag)
	set(&opt.UseAcceptEncodingGzip, true, provider.Quirks.UseAcceptEncodingGzip)
	set(&opt.UseDataIntegrityProtections, false, provider.Quirks.UseDataIntegrityProtections)
	set(&opt.MightGzip, true, provider.Quirks.MightGzip)
	set(&opt.UseAlreadyExists, true, provider.Quirks.UseAlreadyExists)
	set(&opt.UseMultipartUploads, true, provider.Quirks.UseMultipartUploads)
	if !opt.UseMultipartUploads.Value {
		opt.UploadCutoff = math.MaxInt64
	}
	set(&opt.UseUnsignedPayload, true, provider.Quirks.UseUnsignedPayload)
	set(&opt.UseXID, true, provider.Quirks.UseXID)
	set(&opt.SignAcceptEncoding, true, provider.Quirks.SignAcceptEncoding)
}

// setRoot changes the root of the Fs
func (f *Fs) setRoot(root string) {
	f.root = parsePath(root)
	f.rootBucket, f.rootDirectory = bucket.Split(f.root)
}

// NewFs constructs an Fs from the path, bucket:path
func NewFs(ctx context.Context, name, root string, m configmap.Mapper) (fs.Fs, error) {
	// Parse config into Options struct
	opt := new(Options)
	err := configstruct.Set(m, opt)
	if err != nil {
		return nil, err
	}
	err = checkUploadChunkSize(opt.ChunkSize)
	if err != nil {
		return nil, fmt.Errorf("s3: chunk size: %w", err)
	}
	err = checkUploadCutoff(opt.UploadCutoff)
	if err != nil {
		return nil, fmt.Errorf("s3: upload cutoff: %w", err)
	}
	err = checkCopyCutoff(opt.CopyCutoff)
	if err != nil {
		return nil, fmt.Errorf("s3: --s3-copy-cutoff: %w", err)
	}
	if opt.Versions && opt.VersionAt.IsSet() {
		return nil, errors.New("s3: can't use --s3-versions and --s3-version-at at the same time")
	}
	if opt.BucketACL == "" {
		opt.BucketACL = opt.ACL
	}
	if opt.SSECustomerKeyBase64 != "" && opt.SSECustomerKey != "" {
		return nil, errors.New("s3: can't use sse_customer_key and sse_customer_key_base64 at the same time")
	} else if opt.SSECustomerKeyBase64 != "" {
		// Decode the base64-encoded key and store it in the SSECustomerKey field
		decoded, err := base64.StdEncoding.DecodeString(opt.SSECustomerKeyBase64)
		if err != nil {
			return nil, fmt.Errorf("s3: Could not decode sse_customer_key_base64: %w", err)
		}
		opt.SSECustomerKey = string(decoded)
	} else {
		// Encode the raw key as base64
		opt.SSECustomerKeyBase64 = base64.StdEncoding.EncodeToString([]byte(opt.SSECustomerKey))
	}
	if opt.SSECustomerKey != "" && opt.SSECustomerKeyMD5 == "" {
		// calculate CustomerKeyMD5 if not supplied
		md5sumBinary := md5.Sum([]byte(opt.SSECustomerKey))
		opt.SSECustomerKeyMD5 = base64.StdEncoding.EncodeToString(md5sumBinary[:])
	}
	srv := getClient(ctx, opt)
	c, provider, err := s3Connection(ctx, opt, srv)
	if err != nil {
		return nil, err
	}

	ci := fs.GetConfig(ctx)
	pc := fs.NewPacer(ctx, pacer.NewS3(pacer.MinSleep(minSleep)))
	// Set pacer retries to 2 (1 try and 1 retry) because we are
	// relying on SDK retry mechanism, but we allow 2 attempts to
	// retry directory listings after XMLSyntaxError
	pc.SetRetries(2)

	f := &Fs{
		name:    name,
		opt:     *opt,
		ci:      ci,
		ctx:     ctx,
		c:       c,
		pacer:   pc,
		cache:   bucket.NewCache(),
		srv:     srv,
		srvRest: rest.NewClient(fshttp.NewClient(ctx)),
	}
	if opt.ServerSideEncryption == "aws:kms" || opt.SSECustomerAlgorithm != "" {
		// From: https://docs.aws.amazon.com/AmazonS3/latest/API/RESTCommonResponseHeaders.html
		//
		// Objects encrypted by SSE-S3 or plaintext have ETags that are an MD5
		// digest of their data.
		//
		// Objects encrypted by SSE-C or SSE-KMS have ETags that are not an
		// MD5 digest of their object data.
		f.etagIsNotMD5 = true
	}
	if opt.DirectoryBucket {
		// Objects uploaded to directory buckets appear to have random ETags
		//
		// This doesn't appear to be documented
		f.etagIsNotMD5 = true
		// The normal API doesn't work for creating directory buckets, so don't try
		f.opt.NoCheckBucket = true
	}
	f.setRoot(root)
	f.features = (&fs.Features{
		ReadMimeType:      true,
		WriteMimeType:     true,
		ReadMetadata:      true,
		WriteMetadata:     true,
		UserMetadata:      true,
		BucketBased:       true,
		BucketBasedRootOK: true,
		SetTier:           provider.StorageClass.Len() > 0,
		GetTier:           provider.StorageClass.Len() > 0,
		SlowModTime:       true,
	}).Fill(ctx, f)
	if opt.Provider == "AWS" {
		f.features.DoubleSlash = true
	}
	if opt.Provider == "Rabata" {
		f.features.Copy = nil
	}
	if opt.DirectoryMarkers {
		f.features.CanHaveEmptyDirectories = true
	}
	// f.listMultipartUploads()
	if !opt.UseMultipartUploads.Value {
		fs.Debugf(f, "Disabling multipart uploads")
		f.features.OpenChunkWriter = nil
	}

	if f.rootBucket != "" && f.rootDirectory != "" && !opt.NoHeadObject && !strings.HasSuffix(root, "/") {
		// Check to see if the (bucket,directory) is actually an existing file
		oldRoot := f.root
		newRoot, leaf := path.Split(oldRoot)
		f.setRoot(newRoot)
		_, err := f.NewObject(ctx, leaf)
		if errors.Is(err, fs.ErrorObjectNotFound) {
			// File doesn't exist or is a directory so return old f
			f.setRoot(oldRoot)
			return f, nil
		}
		if err != nil {
			return nil, err
		}
		// return an error with an fs which points to the parent
		return f, fs.ErrorIsFile
	}
	return f, nil
}

// getMetaDataListing gets the metadata from the object unconditionally from the listing
//
// This is needed to find versioned objects from their paths.
//
// It may return info == nil and err == nil if a HEAD would be more appropriate
func (f *Fs) getMetaDataListing(ctx context.Context, wantRemote string) (info *types.Object, versionID *string, err error) {
	bucket, bucketPath := f.split(wantRemote)

	// Strip the version string off if using versions
	if f.opt.Versions {
		var timestamp time.Time
		timestamp, bucketPath = version.Remove(bucketPath)
		// If the path had no version string return no info, to force caller to look it up
		if timestamp.IsZero() {
			return nil, nil, nil
		}
	}

	err = f.list(ctx, listOpt{
		bucket:       bucket,
		directory:    bucketPath,
		prefix:       f.rootDirectory,
		recurse:      true,
		withVersions: f.opt.Versions,
		findFile:     true,
		versionAt:    f.opt.VersionAt,
		hidden:       f.opt.VersionDeleted,
	}, func(gotRemote string, object *types.Object, objectVersionID *string, isDirectory bool) error {
		if isDirectory {
			return nil
		}
		if wantRemote != gotRemote {
			return nil
		}
		info = object
		versionID = objectVersionID
		return errEndList // read only 1 item
	})
	if err != nil {
		if err == fs.ErrorDirNotFound {
			return nil, nil, fs.ErrorObjectNotFound
		}
		return nil, nil, err
	}
	if info == nil {
		return nil, nil, fs.ErrorObjectNotFound
	}
	return info, versionID, nil
}

// stringClone clones the string s into new memory. This is useful to
// stop us keeping references to small strings carved out of large XML
// responses.
func stringClone(s string) *string {
	var sNew = strings.Clone(s)
	return &sNew
}

// stringClonePointer clones the string pointed to by sp into new
// memory. This is useful to stop us keeping references to small
// strings carved out of large XML responses.
func stringClonePointer(sp *string) *string {
	if sp == nil {
		return nil
	}
	var s = strings.Clone(*sp)
	return &s
}

// Return an Object from a path
//
// If it can't be found it returns the error ErrorObjectNotFound.
func (f *Fs) newObjectWithInfo(ctx context.Context, remote string, info *types.Object, versionID *string) (obj fs.Object, err error) {
	o := &Object{
		fs:     f,
		remote: remote,
	}
	if info == nil && ((f.opt.Versions && version.Match(remote)) || f.opt.VersionAt.IsSet()) {
		// If versions, have to read the listing to find the correct version ID
		info, versionID, err = f.getMetaDataListing(ctx, remote)
		if err != nil {
			return nil, err
		}
	}
	if info != nil {
		// Set info but not meta
		if info.LastModified == nil {
			fs.Logf(o, "Failed to read last modified")
			o.lastModified = time.Now()
		} else {
			o.lastModified = *info.LastModified
		}
		o.setMD5FromEtag(deref(info.ETag))
		o.bytes = deref(info.Size)
		o.storageClass = stringClone(string(info.StorageClass))
		o.versionID = stringClonePointer(versionID)
		// If is delete marker, show that metadata has been read as there is none to read
		if info.Size == isDeleteMarker {
			o.meta = map[string]string{}
		}
	} else if !o.fs.opt.NoHeadObject {
		err := o.readMetaData(ctx) // reads info and meta, returning an error
		if err != nil {
			return nil, err
		}
	}
	return o, nil
}

// NewObject finds the Object at remote.  If it can't be found
// it returns the error fs.ErrorObjectNotFound.
func (f *Fs) NewObject(ctx context.Context, remote string) (fs.Object, error) {
	return f.newObjectWithInfo(ctx, remote, nil, nil)
}

// Gets the bucket location
func (f *Fs) getBucketLocation(ctx context.Context, bucket string) (string, error) {
	region, err := manager.GetBucketRegion(ctx, f.c, bucket)
	if err != nil {
		return "", err
	}
	return region, nil
}

// Updates the region for the bucket by reading the region from the
// bucket then updating the session.
func (f *Fs) updateRegionForBucket(ctx context.Context, bucket string) error {
	region, err := f.getBucketLocation(ctx, bucket)
	if err != nil {
		return fmt.Errorf("reading bucket location failed: %w", err)
	}
	if f.opt.Endpoint != "" {
		return fmt.Errorf("can't set region to %q as endpoint is set", region)
	}
	if f.opt.Region == region {
		return fmt.Errorf("region is already %q - not updating", region)
	}

	// Make a new session with the new region
	oldRegion := f.opt.Region
	f.opt.Region = region
	c, _, err := s3Connection(f.ctx, &f.opt, f.srv)
	if err != nil {
		return fmt.Errorf("creating new session failed: %w", err)
	}
	f.c = c

	fs.Logf(f, "Switched region to %q from %q", region, oldRegion)
	return nil
}

// Common interface for bucket listers
type bucketLister interface {
	List(ctx context.Context) (resp *s3.ListObjectsV2Output, versionIDs []*string, err error)
	URLEncodeListings(bool)
}

// V1 bucket lister
type v1List struct {
	f   *Fs
	req s3.ListObjectsInput
}

// Create a new V1 bucket lister
func (f *Fs) newV1List(req *s3.ListObjectsV2Input) bucketLister {
	l := &v1List{
		f: f,
	}
	// Convert v2 req into v1 req
	//structs.SetFrom(&l.req, req)
	setFrom_s3ListObjectsInput_s3ListObjectsV2Input(&l.req, req)
	return l
}

// List a bucket with V1 listing
func (ls *v1List) List(ctx context.Context) (resp *s3.ListObjectsV2Output, versionIDs []*string, err error) {
	respv1, err := ls.f.c.ListObjects(ctx, &ls.req)
	if err != nil {
		return nil, nil, err
	}

	// Set up the request for next time
	ls.req.Marker = respv1.NextMarker
	if deref(respv1.IsTruncated) && ls.req.Marker == nil {
		if len(respv1.Contents) == 0 {
			return nil, nil, errors.New("s3 protocol error: received listing v1 with IsTruncated set, no NextMarker and no Contents")
		}
		// Use the last Key received if no NextMarker and isTruncated
		ls.req.Marker = respv1.Contents[len(respv1.Contents)-1].Key

	}

	// If we are URL encoding then must decode the marker
	if ls.req.Marker != nil && ls.req.EncodingType == types.EncodingTypeUrl {
		*ls.req.Marker, err = url.QueryUnescape(*ls.req.Marker)
		if err != nil {
			return nil, nil, fmt.Errorf("failed to URL decode Marker %q: %w", *ls.req.Marker, err)
		}
	}

	// convert v1 resp into v2 resp
	resp = new(s3.ListObjectsV2Output)
	//structs.SetFrom(resp, respv1)
	setFrom_s3ListObjectsV2Output_s3ListObjectsOutput(resp, respv1)

	return resp, nil, nil
}

// URL Encode the listings
func (ls *v1List) URLEncodeListings(encode bool) {
	if encode {
		ls.req.EncodingType = types.EncodingTypeUrl
	} else {
		ls.req.EncodingType = types.EncodingType("")
	}
}

// V2 bucket lister
type v2List struct {
	f   *Fs
	req s3.ListObjectsV2Input
}

// Create a new V2 bucket lister
func (f *Fs) newV2List(req *s3.ListObjectsV2Input) bucketLister {
	return &v2List{
		f:   f,
		req: *req,
	}
}

// Do a V2 listing
func (ls *v2List) List(ctx context.Context) (resp *s3.ListObjectsV2Output, versionIDs []*string, err error) {
	resp, err = ls.f.c.ListObjectsV2(ctx, &ls.req)
	if err != nil {
		return nil, nil, err
	}
	if deref(resp.IsTruncated) && (resp.NextContinuationToken == nil || *resp.NextContinuationToken == "") {
		return nil, nil, errors.New("s3 protocol error: received listing v2 with IsTruncated set and no NextContinuationToken. Should you be using `--s3-list-version 1`?")
	}
	ls.req.ContinuationToken = resp.NextContinuationToken
	return resp, nil, nil
}

// URL Encode the listings
func (ls *v2List) URLEncodeListings(encode bool) {
	if encode {
		ls.req.EncodingType = types.EncodingTypeUrl
	} else {
		ls.req.EncodingType = types.EncodingType("")
	}
}

// Versions bucket lister
type versionsList struct {
	f              *Fs
	req            s3.ListObjectVersionsInput
	versionAt      time.Time // set if we want only versions before this
	usingVersionAt bool      // set if we need to use versionAt
	hidden         bool      // set to see hidden versions
	lastKeySent    string    // last Key sent to the receiving function
}

// Create a new Versions bucket lister
func (f *Fs) newVersionsList(req *s3.ListObjectsV2Input, hidden bool, versionAt time.Time) bucketLister {
	l := &versionsList{
		f:              f,
		versionAt:      versionAt,
		usingVersionAt: !versionAt.IsZero(),
		hidden:         hidden,
	}
	// Convert v2 req into withVersions req
	//structs.SetFrom(&l.req, req)
	setFrom_s3ListObjectVersionsInput_s3ListObjectsV2Input(&l.req, req)
	return l
}

// Any types.Object or types.ObjectVersion with this as their Size are delete markers
var isDeleteMarker = new(int64)

// Compare two types.ObjectVersions, sorted alphabetically by key with
// the newest first if the Keys match or the one with IsLatest set if
// everything matches.
func versionLess(a, b *types.ObjectVersion) bool {
	if a == nil || a.Key == nil || a.LastModified == nil {
		return true
	}
	if b == nil || b.Key == nil || b.LastModified == nil {
		return false
	}
	if *a.Key < *b.Key {
		return true
	}
	if *a.Key > *b.Key {
		return false
	}
	dt := a.LastModified.Sub(*b.LastModified)
	if dt > 0 {
		return true
	}
	if dt < 0 {
		return false
	}
	if deref(a.IsLatest) {
		return true
	}
	return false
}

// Merge the DeleteMarkers into the Versions.
//
// These are delivered by S3 sorted by key then by LastUpdated
// newest first but annoyingly the SDK splits them up into two
// so we need to merge them back again
//
// We do this by converting the s3.DeleteEntry into
// types.ObjectVersion with Size = isDeleteMarker to tell them apart
//
// We then merge them back into the Versions in the correct order
func mergeDeleteMarkers(oldVersions []types.ObjectVersion, deleteMarkers []types.DeleteMarkerEntry) (newVersions []types.ObjectVersion) {
	newVersions = make([]types.ObjectVersion, 0, len(oldVersions)+len(deleteMarkers))
	for _, deleteMarker := range deleteMarkers {
		var obj types.ObjectVersion
		//structs.SetFrom(obj, deleteMarker)
		setFrom_typesObjectVersion_typesDeleteMarkerEntry(&obj, &deleteMarker)
		obj.Size = isDeleteMarker
		for len(oldVersions) > 0 && versionLess(&oldVersions[0], &obj) {
			newVersions = append(newVersions, oldVersions[0])
			oldVersions = oldVersions[1:]
		}
		newVersions = append(newVersions, obj)
	}
	// Merge any remaining versions
	newVersions = append(newVersions, oldVersions...)
	return newVersions
}

// List a bucket with versions
func (ls *versionsList) List(ctx context.Context) (resp *s3.ListObjectsV2Output, versionIDs []*string, err error) {
	respVersions, err := ls.f.c.ListObjectVersions(ctx, &ls.req)
	if err != nil {
		return nil, nil, err
	}

	// Set up the request for next time
	ls.req.KeyMarker = respVersions.NextKeyMarker
	ls.req.VersionIdMarker = respVersions.NextVersionIdMarker
	if deref(respVersions.IsTruncated) && ls.req.KeyMarker == nil {
		return nil, nil, errors.New("s3 protocol error: received versions listing with IsTruncated set with no NextKeyMarker")
	}

	// If we are URL encoding then must decode the marker
	if ls.req.KeyMarker != nil && ls.req.EncodingType == types.EncodingTypeUrl {
		*ls.req.KeyMarker, err = url.QueryUnescape(*ls.req.KeyMarker)
		if err != nil {
			return nil, nil, fmt.Errorf("failed to URL decode KeyMarker %q: %w", *ls.req.KeyMarker, err)
		}
	}

	// convert Versions resp into v2 resp
	resp = new(s3.ListObjectsV2Output)
	//structs.SetFrom(resp, respVersions)
	setFrom_s3ListObjectsV2Output_s3ListObjectVersionsOutput(resp, respVersions)

	// Merge in delete Markers as types.ObjectVersion if we need them
	if ls.hidden || ls.usingVersionAt {
		respVersions.Versions = mergeDeleteMarkers(respVersions.Versions, respVersions.DeleteMarkers)
	}

	// Convert the Versions and the DeleteMarkers into an array of types.Object
	//
	// These are returned in the order that they are stored with the most recent first.
	// With the annoyance that the Versions and DeleteMarkers are split into two
	objs := make([]types.Object, 0, len(respVersions.Versions))
	for _, objVersion := range respVersions.Versions {
		if ls.usingVersionAt {
			if objVersion.LastModified.After(ls.versionAt) {
				// Ignore versions that were created after the specified time
				continue
			}
			if *objVersion.Key == ls.lastKeySent {
				// Ignore versions before the already returned version
				continue
			}
		}
		ls.lastKeySent = *objVersion.Key
		// Don't send delete markers if we don't want hidden things
		if !ls.hidden && objVersion.Size == isDeleteMarker {
			continue
		}
		var obj types.Object
		//structs.SetFrom(obj, objVersion)
		setFrom_typesObject_typesObjectVersion(&obj, &objVersion)
		// Adjust the file names
		if !ls.usingVersionAt && (!deref(objVersion.IsLatest) || objVersion.Size == isDeleteMarker) {
			if obj.Key != nil && objVersion.LastModified != nil {
				*obj.Key = version.Add(*obj.Key, *objVersion.LastModified)
			}
		}
		objs = append(objs, obj)
		versionIDs = append(versionIDs, objVersion.VersionId)
	}

	resp.Contents = objs
	return resp, versionIDs, nil
}

// URL Encode the listings
func (ls *versionsList) URLEncodeListings(encode bool) {
	if encode {
		ls.req.EncodingType = types.EncodingTypeUrl
	} else {
		ls.req.EncodingType = types.EncodingType("")
	}
}

// listFn is called from list to handle an object.
type listFn func(remote string, object *types.Object, versionID *string, isDirectory bool) error

// errEndList is a sentinel used to end the list iteration now.
// listFn should return it to end the iteration with no errors.
var errEndList = errors.New("end list")

// list options
type listOpt struct {
	bucket        string  // bucket to list
	directory     string  // directory with bucket
	prefix        string  // prefix to remove from listing
	addBucket     bool    // if set, the bucket is added to the start of the remote
	recurse       bool    // if set, recurse to read sub directories
	withVersions  bool    // if set, versions are produced
	hidden        bool    // if set, return delete markers as objects with size == isDeleteMarker
	findFile      bool    // if set, it will look for files called (bucket, directory)
	versionAt     fs.Time // if set only show versions <= this time
	noSkipMarkers bool    // if set return dir marker objects
	restoreStatus bool    // if set return restore status in listing too
}

// list lists the objects into the function supplied with the opt
// supplied.
func (f *Fs) list(ctx context.Context, opt listOpt, fn listFn) error {
	if opt.prefix != "" {
		opt.prefix += "/"
	}
	if !opt.findFile {
		if opt.directory != "" && (opt.prefix == "" && !bucket.IsAllSlashes(opt.directory) || opt.prefix != "" && !strings.HasSuffix(opt.directory, "/")) {
			opt.directory += "/"
		}
	}
	delimiter := ""
	if !opt.recurse {
		delimiter = "/"
	}
	// URL encode the listings so we can use control characters in object names
	// See: https://github.com/aws/aws-sdk-go/issues/1914
	//
	// However this doesn't work perfectly under Ceph (and hence DigitalOcean/Dreamhost) because
	// it doesn't encode CommonPrefixes.
	// See: https://tracker.ceph.com/issues/41870
	//
	// This does not work under IBM COS also: See https://github.com/rclone/rclone/issues/3345
	// though maybe it does on some versions.
	//
	// This does work with minio but was only added relatively recently
	// https://github.com/minio/minio/pull/7265
	//
	// So we enable only on providers we know supports it properly, all others can retry when a
	// XML Syntax error is detected.
	urlEncodeListings := f.opt.ListURLEncode.Value
	req := s3.ListObjectsV2Input{
		Bucket:    &opt.bucket,
		Delimiter: &delimiter,
		Prefix:    &opt.directory,
		MaxKeys:   &f.opt.ListChunk,
	}
	if opt.restoreStatus {
		req.OptionalObjectAttributes = []types.OptionalObjectAttributes{types.OptionalObjectAttributesRestoreStatus}
	}
	if f.opt.RequesterPays {
		req.RequestPayer = types.RequestPayerRequester
	}
	var listBucket bucketLister
	switch {
	case opt.withVersions || opt.versionAt.IsSet():
		listBucket = f.newVersionsList(&req, opt.hidden, time.Time(opt.versionAt))
	case f.opt.ListVersion == 1:
		listBucket = f.newV1List(&req)
	default:
		listBucket = f.newV2List(&req)
	}
	foundItems := 0
	for {
		var resp *s3.ListObjectsV2Output
		var err error
		var versionIDs []*string
		err = f.pacer.Call(func() (bool, error) {

			listBucket.URLEncodeListings(urlEncodeListings)
			resp, versionIDs, err = listBucket.List(ctx)
			if err != nil && !urlEncodeListings {
				var xmlErr *xml.SyntaxError
				if errors.As(err, &xmlErr) {
					// Retry the listing with URL encoding as there were characters that XML can't encode
					urlEncodeListings = true
					fs.Debugf(f, "Retrying listing because of characters which can't be XML encoded")
					return true, err
				}
			}
			return f.shouldRetry(ctx, err)
		})
		if err != nil {
			if getHTTPStatusCode(err) == http.StatusNotFound {
				err = fs.ErrorDirNotFound
			}
			if f.rootBucket == "" {
				// if listing from the root ignore wrong region requests returning
				// empty directory
				// 301 if wrong region for bucket
				if getHTTPStatusCode(err) == http.StatusMovedPermanently {
					fs.Errorf(f, "Can't change region for bucket %q with no bucket specified", opt.bucket)
					return nil
				}
			}
			return err
		}
		if !opt.recurse {
			foundItems += len(resp.CommonPrefixes)
			for _, commonPrefix := range resp.CommonPrefixes {
				if commonPrefix.Prefix == nil {
					fs.Logf(f, "Nil common prefix received")
					continue
				}
				remote := *commonPrefix.Prefix
				if urlEncodeListings {
					remote, err = url.QueryUnescape(remote)
					if err != nil {
						fs.Logf(f, "failed to URL decode %q in listing common prefix: %v", *commonPrefix.Prefix, err)
						continue
					}
				}
				remote = f.opt.Enc.ToStandardPath(remote)
				if !strings.HasPrefix(remote, opt.prefix) {
					fs.Logf(f, "Odd directory name received %q", remote)
					continue
				}
				remote = remote[len(opt.prefix):]
				// Trim one slash off the remote name
				remote, _ = strings.CutSuffix(remote, "/")
				if remote == "" || bucket.IsAllSlashes(remote) {
					remote += "/"
				}
				if opt.addBucket {
					remote = bucket.Join(opt.bucket, remote)
				}
				err = fn(remote, &types.Object{Key: &remote}, nil, true)
				if err != nil {
					if err == errEndList {
						return nil
					}
					return err
				}
			}
		}
		foundItems += len(resp.Contents)
		for i, object := range resp.Contents {
			remote := *stringClone(deref(object.Key))
			if urlEncodeListings {
				remote, err = url.QueryUnescape(remote)
				if err != nil {
					fs.Logf(f, "failed to URL decode %q in listing: %v", deref(object.Key), err)
					continue
				}
			}
			remote = f.opt.Enc.ToStandardPath(remote)
			if !strings.HasPrefix(remote, opt.prefix) {
				fs.Logf(f, "Odd name received %q", remote)
				continue
			}
			isDirectory := (remote == "" || strings.HasSuffix(remote, "/")) && object.Size != nil && *object.Size == 0
			// is this a directory marker?
			if isDirectory {
				if opt.noSkipMarkers {
					// process directory markers as files
					isDirectory = false
				} else if remote == f.opt.Enc.ToStandardPath(opt.directory) {
					// Don't insert the root directory
					continue
				}
			}
			remote = remote[len(opt.prefix):]
			if isDirectory {
				// process directory markers as directories
				remote, _ = strings.CutSuffix(remote, "/")
			}
			if opt.addBucket {
				remote = bucket.Join(opt.bucket, remote)
			}
			if versionIDs != nil {
				err = fn(remote, &object, versionIDs[i], isDirectory)
			} else {
				err = fn(remote, &object, nil, isDirectory)
			}
			if err != nil {
				if err == errEndList {
					return nil
				}
				return err
			}
		}
		if !deref(resp.IsTruncated) {
			break
		}
	}
	if f.opt.DirectoryMarkers && foundItems == 0 && opt.directory != "" {
		// Determine whether the directory exists or not by whether it has a marker
		req := s3.HeadObjectInput{
			Bucket: &opt.bucket,
			Key:    &opt.directory,
		}
		_, err := f.headObject(ctx, &req)
		if err != nil {
			if err == fs.ErrorObjectNotFound {
				return fs.ErrorDirNotFound
			}
			return err
		}
	}
	return nil
}

// Convert a list item into a DirEntry
func (f *Fs) itemToDirEntry(ctx context.Context, remote string, object *types.Object, versionID *string, isDirectory bool) (fs.DirEntry, error) {
	if isDirectory {
		size := int64(0)
		if object.Size != nil {
			size = *object.Size
		}
		d := fs.NewDir(remote, time.Time{}).SetSize(size)
		return d, nil
	}
	o, err := f.newObjectWithInfo(ctx, remote, object, versionID)
	if err != nil {
		return nil, err
	}
	return o, nil
}

// listDir lists files and directories to out
func (f *Fs) listDir(ctx context.Context, bucket, directory, prefix string, addBucket bool, callback func(fs.DirEntry) error) (err error) {
	// List the objects and directories
	err = f.list(ctx, listOpt{
		bucket:       bucket,
		directory:    directory,
		prefix:       prefix,
		addBucket:    addBucket,
		withVersions: f.opt.Versions,
		versionAt:    f.opt.VersionAt,
		hidden:       f.opt.VersionDeleted,
	}, func(remote string, object *types.Object, versionID *string, isDirectory bool) error {
		entry, err := f.itemToDirEntry(ctx, remote, object, versionID, isDirectory)
		if err != nil {
			return err
		}
		if entry != nil {
			return callback(entry)
		}
		return nil
	})
	if err != nil {
		return err
	}
	// bucket must be present if listing succeeded
	f.cache.MarkOK(bucket)
	return nil
}

// listBuckets lists the buckets to out
func (f *Fs) listBuckets(ctx context.Context) (entries fs.DirEntries, err error) {
	req := s3.ListBucketsInput{}
	var resp *s3.ListBucketsOutput
	err = f.pacer.Call(func() (bool, error) {
		resp, err = f.c.ListBuckets(ctx, &req)
		return f.shouldRetry(ctx, err)
	})
	if err != nil {
		return nil, err
	}
	for _, bucket := range resp.Buckets {
		bucketName := f.opt.Enc.ToStandardName(deref(bucket.Name))
		f.cache.MarkOK(bucketName)
		d := fs.NewDir(bucketName, deref(bucket.CreationDate))
		entries = append(entries, d)
	}
	return entries, nil
}

// List the objects and directories in dir into entries.  The
// entries can be returned in any order but should be for a
// complete directory.
//
// dir should be "" to list the root, and should not have
// trailing slashes.
//
// This should return ErrDirNotFound if the directory isn't
// found.
func (f *Fs) List(ctx context.Context, dir string) (entries fs.DirEntries, err error) {
	return list.WithListP(ctx, dir, f)
}

// ListP lists the objects and directories of the Fs starting
// from dir non recursively into out.
//
// dir should be "" to start from the root, and should not
// have trailing slashes.
//
// This should return ErrDirNotFound if the directory isn't
// found.
//
// It should call callback for each tranche of entries read.
// These need not be returned in any particular order.  If
// callback returns an error then the listing will stop
// immediately.
func (f *Fs) ListP(ctx context.Context, dir string, callback fs.ListRCallback) error {
	list := list.NewHelper(callback)
	bucket, directory := f.split(dir)
	if bucket == "" {
		if directory != "" {
			return fs.ErrorListBucketRequired
		}
		entries, err := f.listBuckets(ctx)
		if err != nil {
			return err
		}
		for _, entry := range entries {
			err = list.Add(entry)
			if err != nil {
				return err
			}
		}
	} else {
		err := f.listDir(ctx, bucket, directory, f.rootDirectory, f.rootBucket == "", list.Add)
		if err != nil {
			return err
		}
	}
	return list.Flush()
}

// ListR lists the objects and directories of the Fs starting
// from dir recursively into out.
//
// dir should be "" to start from the root, and should not
// have trailing slashes.
//
// This should return ErrDirNotFound if the directory isn't
// found.
//
// It should call callback for each tranche of entries read.
// These need not be returned in any particular order.  If
// callback returns an error then the listing will stop
// immediately.
//
// Don't implement this unless you have a more efficient way
// of listing recursively than doing a directory traversal.
func (f *Fs) ListR(ctx context.Context, dir string, callback fs.ListRCallback) (err error) {
	bucket, directory := f.split(dir)
	list := list.NewHelper(callback)
	listR := func(bucket, directory, prefix string, addBucket bool) error {
		return f.list(ctx, listOpt{
			bucket:       bucket,
			directory:    directory,
			prefix:       prefix,
			addBucket:    addBucket,
			recurse:      true,
			withVersions: f.opt.Versions,
			versionAt:    f.opt.VersionAt,
			hidden:       f.opt.VersionDeleted,
		}, func(remote string, object *types.Object, versionID *string, isDirectory bool) error {
			entry, err := f.itemToDirEntry(ctx, remote, object, versionID, isDirectory)
			if err != nil {
				return err
			}
			return list.Add(entry)
		})
	}
	if bucket == "" {
		entries, err := f.listBuckets(ctx)
		if err != nil {
			return err
		}
		for _, entry := range entries {
			err = list.Add(entry)
			if err != nil {
				return err
			}
			bucket := entry.Remote()
			err = listR(bucket, "", f.rootDirectory, true)
			if err != nil {
				return err
			}
			// bucket must be present if listing succeeded
			f.cache.MarkOK(bucket)
		}
	} else {
		err = listR(bucket, directory, f.rootDirectory, f.rootBucket == "")
		if err != nil {
			return err
		}
		// bucket must be present if listing succeeded
		f.cache.MarkOK(bucket)
	}
	return list.Flush()
}

// Put the Object into the bucket
func (f *Fs) Put(ctx context.Context, in io.Reader, src fs.ObjectInfo, options ...fs.OpenOption) (fs.Object, error) {
	// Temporary Object under construction
	fs := &Object{
		fs:     f,
		remote: src.Remote(),
	}
	return fs, fs.Update(ctx, in, src, options...)
}

// PutStream uploads to the remote path with the modTime given of indeterminate size
func (f *Fs) PutStream(ctx context.Context, in io.Reader, src fs.ObjectInfo, options ...fs.OpenOption) (fs.Object, error) {
	return f.Put(ctx, in, src, options...)
}

// Check if the bucket exists
//
// NB this can return incorrect results if called immediately after bucket deletion
func (f *Fs) bucketExists(ctx context.Context, bucket string) (bool, error) {
	req := s3.HeadBucketInput{
		Bucket: &bucket,
	}
	err := f.pacer.Call(func() (bool, error) {
		_, err := f.c.HeadBucket(ctx, &req)
		return f.shouldRetry(ctx, err)
	})
	if err == nil {
		return true, nil
	}
	if getHTTPStatusCode(err) == http.StatusNotFound {
		return false, nil
	}
	return false, err
}

// Create directory marker file and parents
func (f *Fs) createDirectoryMarker(ctx context.Context, bucket, dir string) error {
	if !f.opt.DirectoryMarkers || bucket == "" {
		return nil
	}

	// Object to be uploaded
	o := &Object{
		fs: f,
		meta: map[string]string{
			metaMtime: swift.TimeToFloatString(time.Now()),
		},
	}

	for {
		_, bucketPath := f.split(dir)
		// Don't create the directory marker if it is the bucket or at the very root
		if bucketPath == "" {
			break
		}
		o.remote = dir + "/"

		// Check to see if object already exists
		_, err := o.headObject(ctx)
		if err == nil {
			return nil
		}

		// Upload it if not
		fs.Debugf(o, "Creating directory marker")
		content := io.Reader(strings.NewReader(""))
		err = o.Update(ctx, content, o)
		if err != nil {
			return fmt.Errorf("creating directory marker failed: %w", err)
		}

		// Now check parent directory exists
		dir = path.Dir(dir)
		if dir == "/" || dir == "." {
			break
		}
	}

	return nil
}

// Mkdir creates the bucket if it doesn't exist
func (f *Fs) Mkdir(ctx context.Context, dir string) error {
	bucket, _ := f.split(dir)
	e := f.makeBucket(ctx, bucket)
	if e != nil {
		return e
	}
	return f.createDirectoryMarker(ctx, bucket, dir)
}

// mkdirParent creates the parent bucket/directory if it doesn't exist
func (f *Fs) mkdirParent(ctx context.Context, remote string) error {
	remote, _ = strings.CutSuffix(remote, "/")
	dir := path.Dir(remote)
	if dir == "/" || dir == "." {
		dir = ""
	}
	return f.Mkdir(ctx, dir)
}

// makeBucket creates the bucket if it doesn't exist
func (f *Fs) makeBucket(ctx context.Context, bucket string) error {
	if f.opt.NoCheckBucket {
		return nil
	}
	return f.cache.Create(bucket, func() error {
		req := s3.CreateBucketInput{
			Bucket: &bucket,
			ACL:    types.BucketCannedACL(f.opt.BucketACL),
		}
		if f.opt.LocationConstraint != "" {
			req.CreateBucketConfiguration = &types.CreateBucketConfiguration{
				LocationConstraint: types.BucketLocationConstraint(f.opt.LocationConstraint),
			}
		}
		err := f.pacer.Call(func() (bool, error) {
			_, err := f.c.CreateBucket(ctx, &req)
			return f.shouldRetry(ctx, err)
		})
		if err == nil {
			fs.Infof(f, "Bucket %q created with ACL %q", bucket, f.opt.BucketACL)
		}
		var awsErr smithy.APIError
		if errors.As(err, &awsErr) {
			switch awsErr.ErrorCode() {
			case "BucketAlreadyOwnedByYou":
				err = nil
			case "BucketAlreadyExists", "BucketNameUnavailable":
				if f.opt.UseAlreadyExists.Value {
					// We can trust BucketAlreadyExists to mean not owned by us, so make it non retriable
					err = fserrors.NoRetryError(err)
				} else {
					// We can't trust BucketAlreadyExists to mean not owned by us, so ignore it
					err = nil
				}
			}
		}
		return err
	}, func() (bool, error) {
		return f.bucketExists(ctx, bucket)
	})
}

// Rmdir deletes the bucket if the fs is at the root
//
// Returns an error if it isn't empty
func (f *Fs) Rmdir(ctx context.Context, dir string) error {
	bucket, directory := f.split(dir)
	// Remove directory marker file
	if f.opt.DirectoryMarkers && bucket != "" && dir != "" {
		o := &Object{
			fs:     f,
			remote: dir + "/",
		}
		fs.Debugf(o, "Removing directory marker")
		err := o.Remove(ctx)
		if err != nil {
			return fmt.Errorf("removing directory marker failed: %w", err)
		}
	}
	if bucket == "" || directory != "" {
		return nil
	}
	return f.cache.Remove(bucket, func() error {
		req := s3.DeleteBucketInput{
			Bucket: &bucket,
		}
		err := f.pacer.Call(func() (bool, error) {
			_, err := f.c.DeleteBucket(ctx, &req)
			return f.shouldRetry(ctx, err)
		})
		if err == nil {
			fs.Infof(f, "Bucket %q deleted", bucket)
		}
		return err
	})
}

// Precision of the remote
func (f *Fs) Precision() time.Duration {
	return time.Nanosecond
}

// pathEscape escapes s as for a URL path.  It uses rest.URLPathEscape
// but also escapes '+' for S3 and Digital Ocean spaces compatibility
func pathEscape(s string) string {
	return strings.ReplaceAll(rest.URLPathEscape(s), "+", "%2B")
}

// copy does a server-side copy
//
// It adds the boiler plate to the req passed in and calls the s3
// method
func (f *Fs) copy(ctx context.Context, req *s3.CopyObjectInput, dstBucket, dstPath, srcBucket, srcPath string, src *Object) error {
	req.Bucket = &dstBucket
	req.ACL = types.ObjectCannedACL(f.opt.ACL)
	req.Key = &dstPath
	source := pathEscape(bucket.Join(srcBucket, srcPath))
	if src.versionID != nil {
		source += fmt.Sprintf("?versionId=%s", *src.versionID)
	}
	req.CopySource = &source
	if f.opt.RequesterPays {
		req.RequestPayer = types.RequestPayerRequester
	}
	if f.opt.ServerSideEncryption != "" {
		req.ServerSideEncryption = types.ServerSideEncryption(f.opt.ServerSideEncryption)
	}
	if f.opt.SSECustomerAlgorithm != "" {
		req.SSECustomerAlgorithm = &f.opt.SSECustomerAlgorithm
		req.CopySourceSSECustomerAlgorithm = &f.opt.SSECustomerAlgorithm
	}
	if f.opt.SSECustomerKeyBase64 != "" {
		req.SSECustomerKey = &f.opt.SSECustomerKeyBase64
		req.CopySourceSSECustomerKey = &f.opt.SSECustomerKeyBase64
	}
	if f.opt.SSECustomerKeyMD5 != "" {
		req.SSECustomerKeyMD5 = &f.opt.SSECustomerKeyMD5
		req.CopySourceSSECustomerKeyMD5 = &f.opt.SSECustomerKeyMD5
	}
	if f.opt.SSEKMSKeyID != "" {
		req.SSEKMSKeyId = &f.opt.SSEKMSKeyID
	}
	if req.StorageClass == types.StorageClass("") && f.opt.StorageClass != "" {
		req.StorageClass = types.StorageClass(f.opt.StorageClass)
	}

	if src.bytes >= int64(f.opt.CopyCutoff) {
		return f.copyMultipart(ctx, req, dstBucket, dstPath, srcBucket, srcPath, src)
	}
	return f.pacer.Call(func() (bool, error) {
		_, err := f.c.CopyObject(ctx, req)
		return f.shouldRetry(ctx, err)
	})
}

func calculateRange(partSize, partIndex, numParts, totalSize int64) string {
	start := partIndex * partSize
	var ends string
	if partIndex == numParts-1 {
		if totalSize >= 1 {
			ends = strconv.FormatInt(totalSize-1, 10)
		}
	} else {
		ends = strconv.FormatInt(start+partSize-1, 10)
	}
	return fmt.Sprintf("bytes=%v-%v", start, ends)
}

func (f *Fs) copyMultipart(ctx context.Context, copyReq *s3.CopyObjectInput, dstBucket, dstPath, srcBucket, srcPath string, src *Object) (err error) {
	info, err := src.headObject(ctx)
	if err != nil {
		return err
	}

	req := &s3.CreateMultipartUploadInput{}

	// Fill in the request from the head info
	//structs.SetFrom(req, info)
	setFrom_s3CreateMultipartUploadInput_s3HeadObjectOutput(req, info)

	// If copy metadata was set then set the Metadata to that read
	// from the head request
	if copyReq.MetadataDirective == types.MetadataDirectiveCopy {
		copyReq.Metadata = info.Metadata
	}

	// Overwrite any from the copyReq
	//structs.SetFrom(req, copyReq)
	setFrom_s3CreateMultipartUploadInput_s3CopyObjectInput(req, copyReq)

	req.Bucket = &dstBucket
	req.Key = &dstPath

	var cout *s3.CreateMultipartUploadOutput
	if err := f.pacer.Call(func() (bool, error) {
		var err error
		cout, err = f.c.CreateMultipartUpload(ctx, req)
		return f.shouldRetry(ctx, err)
	}); err != nil {
		return err
	}
	uid := cout.UploadId

	defer atexit.OnError(&err, func() {
		// Try to abort the upload, but ignore the error.
		fs.Debugf(src, "Cancelling multipart copy")
		_ = f.pacer.Call(func() (bool, error) {
			_, err := f.c.AbortMultipartUpload(context.Background(), &s3.AbortMultipartUploadInput{
				Bucket:       &dstBucket,
				Key:          &dstPath,
				UploadId:     uid,
				RequestPayer: req.RequestPayer,
			})
			return f.shouldRetry(ctx, err)
		})
	})()

	srcSize := src.bytes
	partSize := int64(f.opt.CopyCutoff)
	numParts := (srcSize-1)/partSize + 1

	fs.Debugf(src, "Starting  multipart copy with %d parts", numParts)

	var (
		parts   = make([]types.CompletedPart, numParts)
		g, gCtx = errgroup.WithContext(ctx)
	)
	g.SetLimit(f.opt.UploadConcurrency)
	for partNum := int32(1); int64(partNum) <= numParts; partNum++ {
		// Fail fast, in case an errgroup managed function returns an error
		// gCtx is cancelled. There is no point in uploading all the other parts.
		if gCtx.Err() != nil {
			break
		}
		partNum := partNum // for closure
		g.Go(func() error {
			var uout *s3.UploadPartCopyOutput
			uploadPartReq := &s3.UploadPartCopyInput{}
			//structs.SetFrom(uploadPartReq, copyReq)
			setFrom_s3UploadPartCopyInput_s3CopyObjectInput(uploadPartReq, copyReq)
			uploadPartReq.Bucket = &dstBucket
			uploadPartReq.Key = &dstPath
			uploadPartReq.PartNumber = &partNum
			uploadPartReq.UploadId = uid
			uploadPartReq.CopySourceRange = aws.String(calculateRange(partSize, int64(partNum-1), numParts, srcSize))
			err := f.pacer.Call(func() (bool, error) {
				uout, err = f.c.UploadPartCopy(gCtx, uploadPartReq)
				return f.shouldRetry(gCtx, err)
			})
			if err != nil {
				return err
			}
			parts[partNum-1] = types.CompletedPart{
				PartNumber: &partNum,
				ETag:       uout.CopyPartResult.ETag,
			}
			return nil
		})
	}

	err = g.Wait()
	if err != nil {
		return err
	}

	return f.pacer.Call(func() (bool, error) {
		_, err := f.c.CompleteMultipartUpload(ctx, &s3.CompleteMultipartUploadInput{
			Bucket: &dstBucket,
			Key:    &dstPath,
			MultipartUpload: &types.CompletedMultipartUpload{
				Parts: parts,
			},
			RequestPayer:         req.RequestPayer,
			SSECustomerAlgorithm: req.SSECustomerAlgorithm,
			SSECustomerKey:       req.SSECustomerKey,
			SSECustomerKeyMD5:    req.SSECustomerKeyMD5,
			UploadId:             uid,
		})
		return f.shouldRetry(ctx, err)
	})
}

// Copy src to this remote using server-side copy operations.
//
// This is stored with the remote path given.
//
// It returns the destination Object and a possible error.
//
// Will only be called if src.Fs().Name() == f.Name()
//
// If it isn't possible then return fs.ErrorCantCopy
func (f *Fs) Copy(ctx context.Context, src fs.Object, remote string) (fs.Object, error) {
	if f.opt.VersionAt.IsSet() {
		return nil, errNotWithVersionAt
	}
	dstBucket, dstPath := f.split(remote)
	err := f.mkdirParent(ctx, remote)
	if err != nil {
		return nil, err
	}
	srcObj, ok := src.(*Object)
	if !ok {
		fs.Debugf(src, "Can't copy - not same remote type")
		return nil, fs.ErrorCantCopy
	}

	srcBucket, srcPath := srcObj.split()
	req := s3.CopyObjectInput{
		MetadataDirective: types.MetadataDirectiveCopy,
	}

	// Update the metadata if it is in use
	if ci := fs.GetConfig(ctx); ci.Metadata {
		ui, err := srcObj.prepareUpload(ctx, src, fs.MetadataAsOpenOptions(ctx), true)
		if err != nil {
			return nil, fmt.Errorf("failed to prepare upload: %w", err)
		}
		setFrom_s3CopyObjectInput_s3PutObjectInput(&req, ui.req)
		req.MetadataDirective = types.MetadataDirectiveReplace
	}

	err = f.copy(ctx, &req, dstBucket, dstPath, srcBucket, srcPath, srcObj)
	if err != nil {
		return nil, err
	}
	return f.NewObject(ctx, remote)
}

// Hashes returns the supported hash sets.
func (f *Fs) Hashes() hash.Set {
	return hash.Set(hash.MD5)
}

// PublicLink generates a public link to the remote path (usually readable by anyone)
func (f *Fs) PublicLink(ctx context.Context, remote string, expire fs.Duration, unlink bool) (link string, err error) {
	if strings.HasSuffix(remote, "/") {
		return "", fs.ErrorCantShareDirectories
	}
	obj, err := f.NewObject(ctx, remote)
	if err != nil {
		return "", err
	}
	o := obj.(*Object)
	if expire > maxExpireDuration {
		fs.Logf(f, "Public Link: Reducing expiry to %v as %v is greater than the max time allowed", maxExpireDuration, expire)
		expire = maxExpireDuration
	}
	bucket, bucketPath := f.split(remote)
	httpReq, err := s3.NewPresignClient(f.c).PresignGetObject(ctx, &s3.GetObjectInput{
		Bucket:    &bucket,
		Key:       &bucketPath,
		VersionId: o.versionID,
	}, s3.WithPresignExpires(time.Duration(expire)))
	if err != nil {
		return "", err
	}
	return httpReq.URL, nil
}

var commandHelp = []fs.CommandHelp{{
	Name:  "restore",
	Short: "Restore objects from GLACIER or INTELLIGENT-TIERING archive tier.",
	Long: `This command can be used to restore one or more objects from GLACIER to normal
storage or from INTELLIGENT-TIERING Archive Access / Deep Archive Access tier
to the Frequent Access tier.

Usage examples:

` + "```console" + `
rclone backend restore s3:bucket/path/to/ --include /object -o priority=PRIORITY -o lifetime=DAYS
rclone backend restore s3:bucket/path/to/directory -o priority=PRIORITY -o lifetime=DAYS
rclone backend restore s3:bucket -o priority=PRIORITY -o lifetime=DAYS
rclone backend restore s3:bucket/path/to/directory -o priority=PRIORITY
` + "```" + `

This flag also obeys the filters. Test first with --interactive/-i or --dry-run
flags.

` + "```console" + `
rclone --interactive backend restore --include "*.txt" s3:bucket/path -o priority=Standard -o lifetime=1
` + "```" + `

All the objects shown will be marked for restore, then:

` + "```console" + `
rclone backend restore --include "*.txt" s3:bucket/path -o priority=Standard -o lifetime=1
` + "```" + `

It returns a list of status dictionaries with Remote and Status
keys. The Status will be OK if it was successful or an error message
if not.

` + "```json" + `
[
    {
        "Status": "OK",
        "Remote": "test.txt"
    },
    {
        "Status": "OK",
        "Remote": "test/file4.txt"
    }
]
` + "```",
	Opts: map[string]string{
		"priority": "Priority of restore: Standard|Expedited|Bulk",
		"lifetime": `Lifetime of the active copy in days, ignored for INTELLIGENT-TIERING
storage.`,
		"description": "The optional description for the job.",
	},
}, {
	Name:  "restore-status",
	Short: "Show the status for objects being restored from GLACIER or INTELLIGENT-TIERING.",
	Long: `This command can be used to show the status for objects being restored from
GLACIER to normal storage or from INTELLIGENT-TIERING Archive Access / Deep
Archive Access tier to the Frequent Access tier.

Usage examples:

` + "```console" + `
rclone backend restore-status s3:bucket/path/to/object
rclone backend restore-status s3:bucket/path/to/directory
rclone backend restore-status -o all s3:bucket/path/to/directory
` + "```" + `

This command does not obey the filters.

It returns a list of status dictionaries:

` + "```json" + `
[
    {
        "Remote": "file.txt",
        "VersionID": null,
        "RestoreStatus": {
            "IsRestoreInProgress": true,
            "RestoreExpiryDate": "2023-09-06T12:29:19+01:00"
        },
        "StorageClass": "GLACIER"
    },
    {
        "Remote": "test.pdf",
        "VersionID": null,
        "RestoreStatus": {
            "IsRestoreInProgress": false,
            "RestoreExpiryDate": "2023-09-06T12:29:19+01:00"
        },
        "StorageClass": "DEEP_ARCHIVE"
    },
    {
        "Remote": "test.gz",
        "VersionID": null,
        "RestoreStatus": {
            "IsRestoreInProgress": true,
            "RestoreExpiryDate": "null"
        },
        "StorageClass": "INTELLIGENT_TIERING"
    }
]
` + "```",
	Opts: map[string]string{
		"all": "If set then show all objects, not just ones with restore status.",
	},
}, {
	Name:  "list-multipart-uploads",
	Short: "List the unfinished multipart uploads.",
	Long: `This command lists the unfinished multipart uploads in JSON format.

Usage examples:

` + "```console" + `
rclone backend list-multipart s3:bucket/path/to/object
` + "```" + `

It returns a dictionary of buckets with values as lists of unfinished
multipart uploads.

You can call it with no bucket in which case it lists all bucket, with
a bucket or with a bucket and path.

` + "```json" + `
{
    "rclone": [
        {
            "Initiated": "2020-06-26T14:20:36Z",
            "Initiator": {
                "DisplayName": "XXX",
                "ID": "arn:aws:iam::XXX:user/XXX"
            },
            "Key": "KEY",
            "Owner": {
                "DisplayName": null,
                "ID": "XXX"
            },
            "StorageClass": "STANDARD",
            "UploadId": "XXX"
        }
    ],
    "rclone-1000files": [],
    "rclone-dst": []
}
` + "```",
}, {
	Name:  "cleanup",
	Short: "Remove unfinished multipart uploads.",
	Long: `This command removes unfinished multipart uploads of age greater than
max-age which defaults to 24 hours.

Note that you can use --interactive/-i or --dry-run with this command to see
what it would do.

Usage examples:

` + "```console" + `
rclone backend cleanup s3:bucket/path/to/object
rclone backend cleanup -o max-age=7w s3:bucket/path/to/object
` + "```" + `

Durations are parsed as per the rest of rclone, 2h, 7d, 7w etc.`,
	Opts: map[string]string{
		"max-age": "Max age of upload to delete.",
	},
}, {
	Name:  "cleanup-hidden",
	Short: "Remove old versions of files.",
	Long: `This command removes any old hidden versions of files
on a versions enabled bucket.

Note that you can use --interactive/-i or --dry-run with this command to see
what it would do.

Usage example:

` + "```console" + `
rclone backend cleanup-hidden s3:bucket/path/to/dir
` + "```",
}, {
	Name:  "versioning",
	Short: "Set/get versioning support for a bucket.",
	Long: `This command sets versioning support if a parameter is
passed and then returns the current versioning status for the bucket
supplied.

Usage examples:

` + "```console" + `
rclone backend versioning s3:bucket # read status only
rclone backend versioning s3:bucket Enabled
rclone backend versioning s3:bucket Suspended
` + "```" + `

It may return "Enabled", "Suspended" or "Unversioned". Note that once
versioning has been enabled the status can't be set back to "Unversioned".`,
}, {
	Name:  "set",
	Short: "Set command for updating the config parameters.",
	Long: `This set command can be used to update the config parameters
for a running s3 backend.

Usage examples:

` + "```console" + `
rclone backend set s3: [-o opt_name=opt_value] [-o opt_name2=opt_value2]
rclone rc backend/command command=set fs=s3: [-o opt_name=opt_value] [-o opt_name2=opt_value2]
rclone rc backend/command command=set fs=s3: -o session_token=X -o access_key_id=X -o secret_access_key=X
` + "```" + `

The option keys are named as they are in the config file.

This rebuilds the connection to the s3 backend when it is called with
the new parameters. Only new parameters need be passed as the values
will default to those currently in use.

It doesn't return anything.`,
}}

// Command the backend to run a named command
//
// The command run is name
// args may be used to read arguments from
// opts may be used to read optional arguments from
//
// The result should be capable of being JSON encoded
// If it is a string or a []string it will be shown to the user
// otherwise it will be JSON encoded and shown to the user like that
func (f *Fs) Command(ctx context.Context, name string, arg []string, opt map[string]string) (out any, err error) {
	switch name {
	case "restore":
		req := s3.RestoreObjectInput{
			//Bucket:         &f.rootBucket,
			//Key:            &encodedDirectory,
			RestoreRequest: &types.RestoreRequest{},
		}
		if lifetime := opt["lifetime"]; lifetime != "" {
			ilifetime, err := strconv.ParseInt(lifetime, 10, 32)
			if err != nil {
				return nil, fmt.Errorf("bad lifetime: %w", err)
			}
			ilifetime32 := int32(ilifetime)
			req.RestoreRequest.Days = &ilifetime32
		}
		if priority := opt["priority"]; priority != "" {
			req.RestoreRequest.GlacierJobParameters = &types.GlacierJobParameters{
				Tier: types.Tier(priority),
			}
		}
		if description := opt["description"]; description != "" {
			req.RestoreRequest.Description = &description
		}
		type status struct {
			Status string
			Remote string
		}
		var (
			outMu sync.Mutex
			out   = []status{}
		)
		err = operations.ListFn(ctx, f, func(obj fs.Object) {
			// Remember this is run --checkers times concurrently
			o, ok := obj.(*Object)
			st := status{Status: "OK", Remote: obj.Remote()}
			defer func() {
				outMu.Lock()
				out = append(out, st)
				outMu.Unlock()
			}()
			if operations.SkipDestructive(ctx, obj, "restore") {
				return
			}
			if !ok {
				st.Status = "Not an S3 object"
				return
			}
			if o.storageClass == nil || (*o.storageClass != "GLACIER" && *o.storageClass != "DEEP_ARCHIVE" && *o.storageClass != "INTELLIGENT_TIERING") {
				st.Status = "Not GLACIER or DEEP_ARCHIVE or INTELLIGENT_TIERING storage class"
				return
			}
			bucket, bucketPath := o.split()
			reqCopy := req
			if *o.storageClass == "INTELLIGENT_TIERING" {
				reqCopy.RestoreRequest.Days = nil
			}
			reqCopy.Bucket = &bucket
			reqCopy.Key = &bucketPath
			reqCopy.VersionId = o.versionID
			err = f.pacer.Call(func() (bool, error) {
				_, err = f.c.RestoreObject(ctx, &reqCopy)
				return f.shouldRetry(ctx, err)
			})
			if err != nil {
				st.Status = err.Error()
			}
		})
		if err != nil {
			return out, err
		}
		return out, nil
	case "restore-status":
		_, all := opt["all"]
		return f.restoreStatus(ctx, all)
	case "list-multipart-uploads":
		return f.listMultipartUploadsAll(ctx)
	case "cleanup":
		maxAge := 24 * time.Hour
		if opt["max-age"] != "" {
			maxAge, err = fs.ParseDuration(opt["max-age"])
			if err != nil {
				return nil, fmt.Errorf("bad max-age: %w", err)
			}
		}
		return nil, f.cleanUp(ctx, maxAge)
	case "cleanup-hidden":
		return nil, f.CleanUpHidden(ctx)
	case "versioning":
		return f.setGetVersioning(ctx, arg...)
	case "set":
		newOpt := f.opt
		err := configstruct.Set(configmap.Simple(opt), &newOpt)
		if err != nil {
			return nil, fmt.Errorf("reading config: %w", err)
		}
		c, _, err := s3Connection(f.ctx, &newOpt, f.srv)
		if err != nil {
			return nil, fmt.Errorf("updating session: %w", err)
		}
		f.c = c
		f.opt = newOpt
		keys := []string{}
		for k := range opt {
			keys = append(keys, k)
		}
		fs.Logf(f, "Updated config values: %s", strings.Join(keys, ", "))
		return nil, nil
	default:
		return nil, fs.ErrorCommandNotFound
	}
}

// Returned from "restore-status"
type restoreStatusOut struct {
	Remote        string
	VersionID     *string
	RestoreStatus *types.RestoreStatus
	StorageClass  types.ObjectStorageClass
}

// Recursively enumerate the current fs to find objects with a restore status
func (f *Fs) restoreStatus(ctx context.Context, all bool) (out []restoreStatusOut, err error) {
	fs.Debugf(f, "all = %v", all)
	bucket, directory := f.split("")
	out = []restoreStatusOut{}
	err = f.list(ctx, listOpt{
		bucket:        bucket,
		directory:     directory,
		prefix:        f.rootDirectory,
		addBucket:     f.rootBucket == "",
		recurse:       true,
		withVersions:  f.opt.Versions,
		versionAt:     f.opt.VersionAt,
		hidden:        f.opt.VersionDeleted,
		restoreStatus: true,
	}, func(remote string, object *types.Object, versionID *string, isDirectory bool) error {
		entry, err := f.itemToDirEntry(ctx, remote, object, versionID, isDirectory)
		if err != nil {
			return err
		}
		if entry != nil {
			if o, ok := entry.(*Object); ok && (all || object.RestoreStatus != nil) {
				out = append(out, restoreStatusOut{
					Remote:        o.remote,
					VersionID:     o.versionID,
					RestoreStatus: object.RestoreStatus,
					StorageClass:  object.StorageClass,
				})
			}
		}
		return nil
	})
	if err != nil {
		return nil, err
	}
	// bucket must be present if listing succeeded
	f.cache.MarkOK(bucket)
	return out, nil
}

// listMultipartUploads lists all outstanding multipart uploads for (bucket, key)
//
// Note that rather lazily we treat key as a prefix so it matches
// directories and objects. This could surprise the user if they ask
// for "dir" and it returns "dirKey"
func (f *Fs) listMultipartUploads(ctx context.Context, bucket, key string) (uploads []types.MultipartUpload, err error) {
	var (
		keyMarker      *string
		uploadIDMarker *string
	)
	uploads = []types.MultipartUpload{}
	for {
		req := s3.ListMultipartUploadsInput{
			Bucket:         &bucket,
			MaxUploads:     &f.opt.ListChunk,
			KeyMarker:      keyMarker,
			UploadIdMarker: uploadIDMarker,
			Prefix:         &key,
		}
		var resp *s3.ListMultipartUploadsOutput
		err = f.pacer.Call(func() (bool, error) {
			resp, err = f.c.ListMultipartUploads(ctx, &req)
			return f.shouldRetry(ctx, err)
		})
		if err != nil {
			return nil, fmt.Errorf("list multipart uploads bucket %q key %q: %w", bucket, key, err)
		}
		uploads = append(uploads, resp.Uploads...)
		if !deref(resp.IsTruncated) {
			break
		}
		keyMarker = resp.NextKeyMarker
		uploadIDMarker = resp.NextUploadIdMarker
	}
	return uploads, nil
}

func (f *Fs) listMultipartUploadsAll(ctx context.Context) (uploadsMap map[string][]types.MultipartUpload, err error) {
	uploadsMap = make(map[string][]types.MultipartUpload)
	bucket, directory := f.split("")
	if bucket != "" {
		uploads, err := f.listMultipartUploads(ctx, bucket, directory)
		if err != nil {
			return uploadsMap, err
		}
		uploadsMap[bucket] = uploads
		return uploadsMap, nil
	}
	entries, err := f.listBuckets(ctx)
	if err != nil {
		return uploadsMap, err
	}
	for _, entry := range entries {
		bucket := entry.Remote()
		uploads, listErr := f.listMultipartUploads(ctx, bucket, "")
		if listErr != nil {
			err = listErr
			fs.Errorf(f, "%v", err)
		}
		uploadsMap[bucket] = uploads
	}
	return uploadsMap, err
}

// cleanUpBucket removes all pending multipart uploads for a given bucket over the age of maxAge
func (f *Fs) cleanUpBucket(ctx context.Context, bucket string, maxAge time.Duration, uploads []types.MultipartUpload) (err error) {
	fs.Infof(f, "cleaning bucket %q of pending multipart uploads older than %v", bucket, maxAge)
	for _, upload := range uploads {
		if upload.Initiated != nil && upload.Key != nil && upload.UploadId != nil {
			age := time.Since(*upload.Initiated)
			what := fmt.Sprintf("pending multipart upload for bucket %q key %q dated %v (%v ago)", bucket, *upload.Key, upload.Initiated, age)
			if age > maxAge {
				fs.Infof(f, "removing %s", what)
				if operations.SkipDestructive(ctx, what, "remove pending upload") {
					continue
				}
				req := s3.AbortMultipartUploadInput{
					Bucket:   &bucket,
					UploadId: upload.UploadId,
					Key:      upload.Key,
				}
				_, abortErr := f.c.AbortMultipartUpload(ctx, &req)
				if abortErr != nil {
					err = fmt.Errorf("failed to remove %s: %w", what, abortErr)
					fs.Errorf(f, "%v", err)
				}
			} else {
				fs.Debugf(f, "ignoring %s", what)
			}
		}
	}
	return err
}

// CleanUp removes all pending multipart uploads
func (f *Fs) cleanUp(ctx context.Context, maxAge time.Duration) (err error) {
	uploadsMap, err := f.listMultipartUploadsAll(ctx)
	if err != nil {
		return err
	}
	for bucket, uploads := range uploadsMap {
		cleanErr := f.cleanUpBucket(ctx, bucket, maxAge, uploads)
		if err != nil {
			fs.Errorf(f, "Failed to cleanup bucket %q: %v", bucket, cleanErr)
			err = cleanErr
		}
	}
	return err
}

// Read whether the bucket is versioned or not
func (f *Fs) isVersioned(ctx context.Context) bool {
	f.versioningMu.Lock()
	defer f.versioningMu.Unlock()
	if !f.versioning.Valid {
		_, _ = f.setGetVersioning(ctx)
		fs.Debugf(f, "bucket is versioned: %v", f.versioning.Value)
	}
	return f.versioning.Value
}

// Set or get bucket versioning.
//
// Pass no arguments to get, or pass "Enabled" or "Suspended"
//
// Updates f.versioning
func (f *Fs) setGetVersioning(ctx context.Context, arg ...string) (status types.BucketVersioningStatus, err error) {
	if len(arg) > 1 {
		return "", errors.New("too many arguments")
	}
	if f.rootBucket == "" {
		return "", errors.New("need a bucket")
	}
	if len(arg) == 1 {
		var versioning = types.VersioningConfiguration{
			Status: types.BucketVersioningStatus(arg[0]),
		}
		// Disabled is indicated by the parameter missing
		if versioning.Status == types.BucketVersioningStatus("Disabled") {
			versioning.Status = types.BucketVersioningStatus("")
		}
		req := s3.PutBucketVersioningInput{
			Bucket:                  &f.rootBucket,
			VersioningConfiguration: &versioning,
		}
		err := f.pacer.Call(func() (bool, error) {
			_, err = f.c.PutBucketVersioning(ctx, &req)
			return f.shouldRetry(ctx, err)
		})
		if err != nil {
			return "", err
		}
	}
	req := s3.GetBucketVersioningInput{
		Bucket: &f.rootBucket,
	}
	var resp *s3.GetBucketVersioningOutput
	err = f.pacer.Call(func() (bool, error) {
		resp, err = f.c.GetBucketVersioning(ctx, &req)
		return f.shouldRetry(ctx, err)
	})
	f.versioning.Valid = true
	f.versioning.Value = false
	if err != nil {
		fs.Errorf(f, "Failed to read versioning status, assuming unversioned: %v", err)
		return "", err
	}
	if len(resp.Status) == 0 {
		return "Unversioned", err
	}
	f.versioning.Value = true
	return resp.Status, err
}

// CleanUp removes all pending multipart uploads older than 24 hours
func (f *Fs) CleanUp(ctx context.Context) (err error) {
	return f.cleanUp(ctx, 24*time.Hour)
}

// purge deletes all the files and directories
//
// if oldOnly is true then it deletes only non current files.
//
// Implemented here so we can make sure we delete old versions.
func (f *Fs) purge(ctx context.Context, dir string, oldOnly bool) error {
	if f.opt.VersionAt.IsSet() {
		return errNotWithVersionAt
	}
	bucket, directory := f.split(dir)
	if bucket == "" {
		return errors.New("can't purge from root")
	}
	versioned := f.isVersioned(ctx)
	if !versioned && oldOnly {
		fs.Infof(f, "bucket is not versioned so not removing old versions")
		return nil
	}
	var errReturn error
	var checkErrMutex sync.Mutex
	var checkErr = func(err error) {
		if err == nil {
			return
		}
		checkErrMutex.Lock()
		defer checkErrMutex.Unlock()
		if errReturn == nil {
			errReturn = err
		}
	}

	// Delete Config.Transfers in parallel
	delChan := make(fs.ObjectsChan, f.ci.Transfers)
	delErr := make(chan error, 1)
	go func() {
		delErr <- operations.DeleteFiles(ctx, delChan)
	}()
	checkErr(f.list(ctx, listOpt{
		bucket:        bucket,
		directory:     directory,
		prefix:        f.rootDirectory,
		addBucket:     f.rootBucket == "",
		recurse:       true,
		withVersions:  versioned,
		hidden:        true,
		noSkipMarkers: true,
	}, func(remote string, object *types.Object, versionID *string, isDirectory bool) error {
		if isDirectory {
			return nil
		}
		// If the root is a dirmarker it will have lost its trailing /
		if remote == "" {
			remote = "/"
		}
		oi, err := f.newObjectWithInfo(ctx, remote, object, versionID)
		if err != nil {
			fs.Errorf(object, "Can't create object %+v", err)
			return nil
		}
		tr := accounting.Stats(ctx).NewCheckingTransfer(oi, "checking")
		// Work out whether the file is the current version or not
		isCurrentVersion := !versioned || !version.Match(remote)
		fs.Debugf(nil, "%q version %v", remote, version.Match(remote))
		if oldOnly && isCurrentVersion {
			// Check current version of the file
			if object.Size == isDeleteMarker {
				fs.Debugf(remote, "Deleting current version (id %q) as it is a delete marker", deref(versionID))
				delChan <- oi
			} else {
				fs.Debugf(remote, "Not deleting current version %q", deref(versionID))
			}
		} else {
			if object.Size == isDeleteMarker {
				fs.Debugf(remote, "Deleting delete marker (id %q)", deref(versionID))
			} else {
				fs.Debugf(remote, "Deleting (id %q)", deref(versionID))
			}
			delChan <- oi
		}
		tr.Done(ctx, nil)
		return nil
	}))
	close(delChan)
	checkErr(<-delErr)

	if !oldOnly {
		checkErr(f.Rmdir(ctx, dir))
	}
	return errReturn
}

// Purge deletes all the files and directories including the old versions.
func (f *Fs) Purge(ctx context.Context, dir string) error {
	return f.purge(ctx, dir, false)
}

// CleanUpHidden deletes all the hidden files.
func (f *Fs) CleanUpHidden(ctx context.Context) error {
	return f.purge(ctx, "", true)
}

// ------------------------------------------------------------

// Fs returns the parent Fs
func (o *Object) Fs() fs.Info {
	return o.fs
}

// Return a string version
func (o *Object) String() string {
	if o == nil {
		return "<nil>"
	}
	return o.remote
}

// Remote returns the remote path
func (o *Object) Remote() string {
	return o.remote
}

var matchMd5 = regexp.MustCompile(`^[0-9a-f]{32}$`)

// Set the MD5 from the etag
func (o *Object) setMD5FromEtag(etag string) {
	if o.fs.etagIsNotMD5 {
		o.md5 = ""
		return
	}
	if etag == "" {
		o.md5 = ""
		return
	}
	hash := strings.Trim(strings.ToLower(etag), `"`)
	// Check the etag is a valid md5sum
	if !matchMd5.MatchString(hash) {
		o.md5 = ""
		return
	}
	o.md5 = hash
}

// Hash returns the Md5sum of an object returning a lowercase hex string
func (o *Object) Hash(ctx context.Context, t hash.Type) (string, error) {
	if t != hash.MD5 {
		return "", hash.ErrUnsupported
	}
	// If decompressing, erase the hash
	if o.bytes < 0 {
		return "", nil
	}
	// If we haven't got an MD5, then check the metadata
	if o.md5 == "" {
		err := o.readMetaData(ctx)
		if err != nil {
			return "", err
		}
	}
	return o.md5, nil
}

// Size returns the size of an object in bytes
func (o *Object) Size() int64 {
	return o.bytes
}

func (o *Object) headObject(ctx context.Context) (resp *s3.HeadObjectOutput, err error) {
	bucket, bucketPath := o.split()
	req := s3.HeadObjectInput{
		Bucket:    &bucket,
		Key:       &bucketPath,
		VersionId: o.versionID,
	}
	return o.fs.headObject(ctx, &req)
}

func (f *Fs) headObject(ctx context.Context, req *s3.HeadObjectInput) (resp *s3.HeadObjectOutput, err error) {
	if f.opt.RequesterPays {
		req.RequestPayer = types.RequestPayerRequester
	}
	if f.opt.SSECustomerAlgorithm != "" {
		req.SSECustomerAlgorithm = &f.opt.SSECustomerAlgorithm
	}
	if f.opt.SSECustomerKeyBase64 != "" {
		req.SSECustomerKey = &f.opt.SSECustomerKeyBase64
	}
	if f.opt.SSECustomerKeyMD5 != "" {
		req.SSECustomerKeyMD5 = &f.opt.SSECustomerKeyMD5
	}
	err = f.pacer.Call(func() (bool, error) {
		var err error
		resp, err = f.c.HeadObject(ctx, req)
		return f.shouldRetry(ctx, err)
	})
	if err != nil {
		if statusCode := getHTTPStatusCode(err); statusCode == http.StatusNotFound || statusCode == http.StatusMethodNotAllowed {
			return nil, fs.ErrorObjectNotFound
		}
		return nil, err
	}
	if req.Bucket != nil {
		f.cache.MarkOK(*req.Bucket)
	}
	return resp, nil
}

// readMetaData gets the metadata if it hasn't already been fetched
//
// it also sets the info
func (o *Object) readMetaData(ctx context.Context) (err error) {
	if o.meta != nil {
		return nil
	}
	resp, err := o.headObject(ctx)
	if err != nil {
		return err
	}
	o.setMetaData(resp)
	// resp.ETag, resp.ContentLength, resp.LastModified, resp.Metadata, resp.ContentType, resp.StorageClass)
	return nil
}

// Convert S3 metadata into a map[string]string while lowercasing the
// keys
func s3MetadataToMap(s3Meta map[string]string) map[string]string {
	meta := make(map[string]string, len(s3Meta))
	for k, v := range s3Meta {
		meta[strings.ToLower(k)] = *stringClone(v)
	}
	return meta
}

// Convert our metadata back into S3 metadata
func mapToS3Metadata(meta map[string]string) map[string]string {
	return meta
}

func (o *Object) setMetaData(resp *s3.HeadObjectOutput) {
	// Ignore missing Content-Length assuming it is 0
	// Some versions of ceph do this due their apache proxies
	if resp.ContentLength != nil {
		o.bytes = *resp.ContentLength
	}
	o.setMD5FromEtag(deref(resp.ETag))
	o.meta = s3MetadataToMap(resp.Metadata)
	// Read MD5 from metadata if present
	if md5sumBase64, ok := o.meta[metaMD5Hash]; ok {
		md5sumBytes, err := base64.StdEncoding.DecodeString(md5sumBase64)
		if err != nil {
			fs.Debugf(o, "Failed to read md5sum from metadata %q: %v", md5sumBase64, err)
		} else if len(md5sumBytes) != 16 {
			fs.Debugf(o, "Failed to read md5sum from metadata %q: wrong length", md5sumBase64)
		} else {
			o.md5 = hex.EncodeToString(md5sumBytes)
		}
	}
	if resp.LastModified == nil {
		o.lastModified = time.Now()
		fs.Logf(o, "Failed to read last modified")
	} else {
		// Try to keep the maximum precision in lastModified. If we read
		// it from listings then it may have millisecond precision, but
		// if we read it from a HEAD/GET request then it will have
		// second precision.
		equalToWithinOneSecond := o.lastModified.Truncate(time.Second).Equal(resp.LastModified.Truncate(time.Second))
		newHasNs := resp.LastModified.Nanosecond() != 0
		if !equalToWithinOneSecond || newHasNs {
			o.lastModified = *resp.LastModified
		}
	}
	o.mimeType = strings.Clone(deref(resp.ContentType))

	// Set system metadata
	o.storageClass = stringClone(string(resp.StorageClass))
	o.cacheControl = stringClonePointer(resp.CacheControl)
	o.contentDisposition = stringClonePointer(resp.ContentDisposition)
	o.contentEncoding = stringClonePointer(removeAWSChunked(resp.ContentEncoding))
	o.contentLanguage = stringClonePointer(resp.ContentLanguage)

	// If decompressing then size and md5sum are unknown
	if o.fs.opt.Decompress && deref(o.contentEncoding) == "gzip" {
		o.bytes = -1
		o.md5 = ""
	}
}

// ModTime returns the modification time of the object
//
// It attempts to read the objects mtime and if that isn't present the
// LastModified returned in the http headers
func (o *Object) ModTime(ctx context.Context) time.Time {
	if o.fs.ci.UseServerModTime {
		return o.lastModified
	}
	err := o.readMetaData(ctx)
	if err != nil {
		fs.Logf(o, "Failed to read metadata: %v", err)
		return time.Now()
	}
	// read mtime out of metadata if available
	d, ok := o.meta[metaMtime]
	if !ok {
		// fs.Debugf(o, "No metadata")
		return o.lastModified
	}
	modTime, err := swift.FloatStringToTime(d)
	if err != nil {
		fs.Logf(o, "Failed to read mtime from object: %v", err)
		return o.lastModified
	}
	return modTime
}

// SetModTime sets the modification time of the local fs object
func (o *Object) SetModTime(ctx context.Context, modTime time.Time) error {
	if o.fs.opt.Provider == "Rabata" {
		// Rabata does not support copying objects
		return fs.ErrorCantSetModTime
	}
	err := o.readMetaData(ctx)
	if err != nil {
		return err
	}
	o.meta[metaMtime] = swift.TimeToFloatString(modTime)

	// Can't update metadata here, so return this error to force a recopy
	if o.storageClass != nil && (*o.storageClass == "GLACIER" || *o.storageClass == "DEEP_ARCHIVE") {
		return fs.ErrorCantSetModTime
	}

	// Copy the object to itself to update the metadata
	bucket, bucketPath := o.split()
	req := s3.CopyObjectInput{
		ContentType:       aws.String(fs.MimeType(ctx, o)), // Guess the content type
		Metadata:          mapToS3Metadata(o.meta),
		MetadataDirective: types.MetadataDirectiveReplace, // replace metadata with that passed in
	}
	if o.fs.opt.RequesterPays {
		req.RequestPayer = types.RequestPayerRequester
	}
	return o.fs.copy(ctx, &req, bucket, bucketPath, bucket, bucketPath, o)
}

// Storable raturns a boolean indicating if this object is storable
func (o *Object) Storable() bool {
	return true
}

// removeAWSChunked removes the "aws-chunked" content-coding from a
// Content-Encoding field value (RFC 9110). Comparison is case-insensitive.
// Returns nil if encoding is empty after removal.
func removeAWSChunked(pv *string) *string {
	if pv == nil {
		return nil
	}
	v := *pv
	if v == "" {
		return nil
	}
	if !strings.Contains(strings.ToLower(v), "aws-chunked") {
		return pv
	}
	parts := strings.Split(v, ",")
	out := make([]string, 0, len(parts))
	for _, p := range parts {
		tok := strings.TrimSpace(p)
		if tok == "" || strings.EqualFold(tok, "aws-chunked") {
			continue
		}
		out = append(out, tok)
	}
	if len(out) == 0 {
		return nil
	}
	v = strings.Join(out, ",")
	return &v
}

func (o *Object) downloadFromURL(ctx context.Context, bucketPath string, options ...fs.OpenOption) (in io.ReadCloser, err error) {
	url := o.fs.opt.DownloadURL + bucketPath
	var resp *http.Response
	opts := rest.Opts{
		Method:  "GET",
		RootURL: url,
		Options: options,
	}
	err = o.fs.pacer.Call(func() (bool, error) {
		resp, err = o.fs.srvRest.Call(ctx, &opts)
		return o.fs.shouldRetry(ctx, err)
	})
	if err != nil {
		return nil, err
	}

	contentLength := rest.ParseSizeFromHeaders(resp.Header)
	if contentLength < 0 {
		fs.Debugf(o, "Failed to parse file size from headers")
	}

	lastModified, err := http.ParseTime(resp.Header.Get("Last-Modified"))
	if err != nil {
		fs.Debugf(o, "Failed to parse last modified from string %s, %v", resp.Header.Get("Last-Modified"), err)
	}

	metaData := make(map[string]string)
	for key, value := range resp.Header {
		key = strings.ToLower(key)
		if after, ok := strings.CutPrefix(key, "x-amz-meta-"); ok {
			metaKey := after
			metaData[metaKey] = value[0]
		}
	}

	header := func(k string) *string {
		v := resp.Header.Get(k)
		if v == "" {
			return nil
		}
		return &v
	}

	var head = s3.HeadObjectOutput{
		ETag:               header("Etag"),
		ContentLength:      &contentLength,
		LastModified:       &lastModified,
		Metadata:           metaData,
		CacheControl:       header("Cache-Control"),
		ContentDisposition: header("Content-Disposition"),
		ContentEncoding:    header("Content-Encoding"),
		ContentLanguage:    header("Content-Language"),
		ContentType:        header("Content-Type"),
		StorageClass:       types.StorageClass(deref(header("X-Amz-Storage-Class"))),
	}
	o.setMetaData(&head)
	return resp.Body, err
}

// middleware to stop the SDK adding `Accept-Encoding: identity`
func removeDisableGzip() func(*middleware.Stack) error {
	return func(stack *middleware.Stack) error {
		_, err := stack.Finalize.Remove("DisableAcceptEncodingGzip")
		return err
	}
}

// middleware to set Accept-Encoding to how we want it
//
// This make sure we download compressed files as-is from all platforms
func (f *Fs) acceptEncoding() (APIOptions []func(*middleware.Stack) error) {
	APIOptions = append(APIOptions, removeDisableGzip())
	if f.opt.UseAcceptEncodingGzip.Value {
		APIOptions = append(APIOptions, smithyhttp.AddHeaderValue("Accept-Encoding", "gzip"))
	}
	return APIOptions
}

// Open an object for read
func (o *Object) Open(ctx context.Context, options ...fs.OpenOption) (in io.ReadCloser, err error) {
	bucket, bucketPath := o.split()

	if o.fs.opt.DownloadURL != "" {
		return o.downloadFromURL(ctx, bucketPath, options...)
	}

	req := s3.GetObjectInput{
		Bucket:    &bucket,
		Key:       &bucketPath,
		VersionId: o.versionID,
	}
	if o.fs.opt.RequesterPays {
		req.RequestPayer = types.RequestPayerRequester
	}
	if o.fs.opt.SSECustomerAlgorithm != "" {
		req.SSECustomerAlgorithm = &o.fs.opt.SSECustomerAlgorithm
	}
	if o.fs.opt.SSECustomerKeyBase64 != "" {
		req.SSECustomerKey = &o.fs.opt.SSECustomerKeyBase64
	}
	if o.fs.opt.SSECustomerKeyMD5 != "" {
		req.SSECustomerKeyMD5 = &o.fs.opt.SSECustomerKeyMD5
	}
	// httpReq, err := s3.NewPresignClient(o.fs.c).PresignGetObject(ctx, &req)
	// if err != nil {
	// 	return nil, err
	// }
	fs.FixRangeOption(options, o.bytes)

	var APIOptions []func(*middleware.Stack) error

	// Set the SDK to always download compressed files as-is
	APIOptions = append(APIOptions, o.fs.acceptEncoding()...)

	for _, option := range options {
		switch option.(type) {
		case *fs.RangeOption, *fs.SeekOption:
			_, value := option.Header()
			req.Range = &value
		case *fs.HTTPOption:
			key, value := option.Header()
			APIOptions = append(APIOptions, smithyhttp.AddHeaderValue(key, value))
		default:
			if option.Mandatory() {
				fs.Logf(o, "Unsupported mandatory option: %v", option)
			}
		}
	}

	var resp *s3.GetObjectOutput
	err = o.fs.pacer.Call(func() (bool, error) {
		var err error
		resp, err = o.fs.c.GetObject(ctx, &req, s3.WithAPIOptions(APIOptions...))
		return o.fs.shouldRetry(ctx, err)
	})
	var awsError smithy.APIError
	if errors.As(err, &awsError) {
		if awsError.ErrorCode() == "InvalidObjectState" {
			return nil, fmt.Errorf("Object in GLACIER, restore first: bucket=%q, key=%q", bucket, bucketPath)
		}
	}
	if err != nil {
		return nil, err
	}

	// read size from ContentLength or ContentRange
	size := resp.ContentLength
	if resp.ContentRange != nil {
		var contentRange = *resp.ContentRange
		slash := strings.IndexRune(contentRange, '/')
		if slash >= 0 {
			i, err := strconv.ParseInt(contentRange[slash+1:], 10, 64)
			if err == nil {
				size = &i
			} else {
				fs.Debugf(o, "Failed to find parse integer from in %q: %v", contentRange, err)
			}
		} else {
			fs.Debugf(o, "Failed to find length in %q", contentRange)
		}
	}
	var head s3.HeadObjectOutput
	//structs.SetFrom(&head, resp)
	setFrom_s3HeadObjectOutput_s3GetObjectOutput(&head, resp)
	head.ContentLength = size
	o.setMetaData(&head)

	// Decompress body if necessary
	if deref(removeAWSChunked(resp.ContentEncoding)) == "gzip" {
		if o.fs.opt.Decompress || (resp.ContentLength == nil && o.fs.opt.MightGzip.Value) {
			return readers.NewGzipReader(resp.Body)
		}
		o.fs.warnCompressed.Do(func() {
			fs.Logf(o, "Not decompressing 'Content-Encoding: gzip' compressed file. Use --s3-decompress to override")
		})
	}

	return resp.Body, nil
}

var warnStreamUpload sync.Once

// state of ChunkWriter
type s3ChunkWriter struct {
	chunkSize            int64
	size                 int64
	f                    *Fs
	bucket               *string
	key                  *string
	uploadID             *string
	multiPartUploadInput *s3.CreateMultipartUploadInput
	completedPartsMu     sync.Mutex
	completedParts       []types.CompletedPart
	eTag                 string
	versionID            string
	md5sMu               sync.Mutex
	md5s                 []byte
	ui                   uploadInfo
	o                    *Object
}

// OpenChunkWriter returns the chunk size and a ChunkWriter
//
// Pass in the remote and the src object
// You can also use options to hint at the desired chunk size
func (f *Fs) OpenChunkWriter(ctx context.Context, remote string, src fs.ObjectInfo, options ...fs.OpenOption) (info fs.ChunkWriterInfo, writer fs.ChunkWriter, err error) {
	// Temporary Object under construction
	o := &Object{
		fs:     f,
		remote: remote,
	}
	ui, err := o.prepareUpload(ctx, src, options, false)
	if err != nil {
		return info, nil, fmt.Errorf("failed to prepare upload: %w", err)
	}

	//structs.SetFrom(&mReq, req)
	var mReq s3.CreateMultipartUploadInput
	setFrom_s3CreateMultipartUploadInput_s3PutObjectInput(&mReq, ui.req)

	uploadParts := f.opt.MaxUploadParts
	if uploadParts < 1 {
		uploadParts = 1
	} else if uploadParts > maxUploadParts {
		uploadParts = maxUploadParts
	}
	size := src.Size()

	// calculate size of parts
	chunkSize := f.opt.ChunkSize

	// size can be -1 here meaning we don't know the size of the incoming file. We use ChunkSize
	// buffers here (default 5 MiB). With a maximum number of parts (10,000) this will be a file of
	// 48 GiB which seems like a not too unreasonable limit.
	if size == -1 {
		warnStreamUpload.Do(func() {
			fs.Logf(f, "Streaming uploads using chunk size %v will have maximum file size of %v",
				f.opt.ChunkSize, fs.SizeSuffix(int64(chunkSize)*int64(uploadParts)))
		})
	} else {
		chunkSize = chunksize.Calculator(src, size, uploadParts, chunkSize)
	}

	var mOut *s3.CreateMultipartUploadOutput
	err = f.pacer.Call(func() (bool, error) {
		mOut, err = f.c.CreateMultipartUpload(ctx, &mReq)
		if err == nil {
			if mOut == nil {
				err = fserrors.RetryErrorf("internal error: no info from multipart upload")
			} else if mOut.UploadId == nil {
				err = fserrors.RetryErrorf("internal error: no UploadId in multipart upload: %#v", *mOut)
			}
		}
		return f.shouldRetry(ctx, err)
	})
	if err != nil {
		return info, nil, fmt.Errorf("create multipart upload failed: %w", err)
	}

	chunkWriter := &s3ChunkWriter{
		chunkSize:            int64(chunkSize),
		size:                 size,
		f:                    f,
		bucket:               ui.req.Bucket,
		key:                  ui.req.Key,
		uploadID:             mOut.UploadId,
		multiPartUploadInput: &mReq,
		completedParts:       make([]types.CompletedPart, 0),
		ui:                   ui,
		o:                    o,
	}
	info = fs.ChunkWriterInfo{
		ChunkSize:         int64(chunkSize),
		Concurrency:       o.fs.opt.UploadConcurrency,
		LeavePartsOnError: o.fs.opt.LeavePartsOnError,
	}
	fs.Debugf(o, "open chunk writer: started multipart upload: %v", *mOut.UploadId)
	return info, chunkWriter, err
}

// add a part number and etag to the completed parts
func (w *s3ChunkWriter) addCompletedPart(partNum *int32, eTag *string) {
	w.completedPartsMu.Lock()
	defer w.completedPartsMu.Unlock()
	w.completedParts = append(w.completedParts, types.CompletedPart{
		PartNumber: partNum,
		ETag:       eTag,
	})
}

// addMd5 adds a binary md5 to the md5 calculated so far
func (w *s3ChunkWriter) addMd5(md5binary *[]byte, chunkNumber int64) {
	w.md5sMu.Lock()
	defer w.md5sMu.Unlock()
	start := chunkNumber * md5.Size
	end := start + md5.Size
	if extend := end - int64(len(w.md5s)); extend > 0 {
		w.md5s = append(w.md5s, make([]byte, extend)...)
	}
	copy(w.md5s[start:end], (*md5binary))
}

// WriteChunk will write chunk number with reader bytes, where chunk number >= 0
func (w *s3ChunkWriter) WriteChunk(ctx context.Context, chunkNumber int, reader io.ReadSeeker) (int64, error) {
	if chunkNumber < 0 {
		err := fmt.Errorf("invalid chunk number provided: %v", chunkNumber)
		return -1, err
	}
	// Only account after the checksum reads have been done
	if do, ok := reader.(pool.DelayAccountinger); ok {
		// To figure out this number, do a transfer and if the accounted size is 0 or a
		// multiple of what it should be, increase or decrease this number.
		//
		// For transfers over https the SDK does not sign the body whereas over http it does
		if len(w.f.opt.Endpoint) >= 5 && strings.EqualFold(w.f.opt.Endpoint[:5], "http:") {
			do.DelayAccounting(3)
		} else {
			do.DelayAccounting(2)
		}
	}

	// create checksum of buffer for integrity checking
	// currently there is no way to calculate the md5 without reading the chunk a 2nd time (1st read is in uploadMultipart)
	// possible in AWS SDK v2 with trailers?
	m := md5.New()
	currentChunkSize, err := io.Copy(m, reader)
	if err != nil {
		return -1, err
	}
	// If no data read and not the first chunk, don't write the chunk
	if currentChunkSize == 0 && chunkNumber != 0 {
		return 0, nil
	}
	md5sumBinary := m.Sum([]byte{})
	w.addMd5(&md5sumBinary, int64(chunkNumber))
	md5sum := base64.StdEncoding.EncodeToString(md5sumBinary)

	// S3 requires 1 <= PartNumber <= 10000
	s3PartNumber := aws.Int32(int32(chunkNumber + 1))
	uploadPartReq := &s3.UploadPartInput{
		Body:                 reader,
		Bucket:               w.bucket,
		Key:                  w.key,
		PartNumber:           s3PartNumber,
		UploadId:             w.uploadID,
		ContentMD5:           &md5sum,
		ContentLength:        aws.Int64(currentChunkSize),
		RequestPayer:         w.multiPartUploadInput.RequestPayer,
		SSECustomerAlgorithm: w.multiPartUploadInput.SSECustomerAlgorithm,
		SSECustomerKey:       w.multiPartUploadInput.SSECustomerKey,
		SSECustomerKeyMD5:    w.multiPartUploadInput.SSECustomerKeyMD5,
	}
	if w.f.opt.DirectoryBucket {
		// Directory buckets do not support "Content-Md5" header
		uploadPartReq.ContentMD5 = nil
	}
	var uout *s3.UploadPartOutput
	err = w.f.pacer.Call(func() (bool, error) {
		// rewind the reader on retry and after reading md5
		_, err = reader.Seek(0, io.SeekStart)
		if err != nil {
			return false, err
		}
		uout, err = w.f.c.UploadPart(ctx, uploadPartReq)
		if err != nil {
			if chunkNumber <= 8 {
				return w.f.shouldRetry(ctx, err)
			}
			if fserrors.ContextError(ctx, &err) {
				return false, err
			}
			// retry all chunks once have done the first few
			return true, err
		}
		return false, nil
	})
	if err != nil {
		return -1, fmt.Errorf("failed to upload chunk %d with %v bytes: %w", chunkNumber+1, currentChunkSize, err)
	}

	w.addCompletedPart(s3PartNumber, uout.ETag)

	fs.Debugf(w.o, "multipart upload wrote chunk %d with %v bytes and etag %v", chunkNumber+1, currentChunkSize, *uout.ETag)
	return currentChunkSize, err
}

// Abort the multipart upload
func (w *s3ChunkWriter) Abort(ctx context.Context) error {
	err := w.f.pacer.Call(func() (bool, error) {
		_, err := w.f.c.AbortMultipartUpload(context.Background(), &s3.AbortMultipartUploadInput{
			Bucket:       w.bucket,
			Key:          w.key,
			UploadId:     w.uploadID,
			RequestPayer: w.multiPartUploadInput.RequestPayer,
		})
		return w.f.shouldRetry(ctx, err)
	})
	if err != nil {
		return fmt.Errorf("failed to abort multipart upload %q: %w", *w.uploadID, err)
	}
	fs.Debugf(w.o, "multipart upload %q aborted", *w.uploadID)
	return err
}

// Close and finalise the multipart upload
func (w *s3ChunkWriter) Close(ctx context.Context) (err error) {
	// sort the completed parts by part number
	sort.Slice(w.completedParts, func(i, j int) bool {
		return *w.completedParts[i].PartNumber < *w.completedParts[j].PartNumber
	})
	var resp *s3.CompleteMultipartUploadOutput
	err = w.f.pacer.Call(func() (bool, error) {
		resp, err = w.f.c.CompleteMultipartUpload(ctx, &s3.CompleteMultipartUploadInput{
			Bucket: w.bucket,
			Key:    w.key,
			MultipartUpload: &types.CompletedMultipartUpload{
				Parts: w.completedParts,
			},
			RequestPayer:         w.multiPartUploadInput.RequestPayer,
			SSECustomerAlgorithm: w.multiPartUploadInput.SSECustomerAlgorithm,
			SSECustomerKey:       w.multiPartUploadInput.SSECustomerKey,
			SSECustomerKeyMD5:    w.multiPartUploadInput.SSECustomerKeyMD5,
			UploadId:             w.uploadID,
		})
		return w.f.shouldRetry(ctx, err)
	})
	if err != nil {
		return fmt.Errorf("failed to complete multipart upload %q: %w", *w.uploadID, err)
	}
	if resp != nil {
		if resp.ETag != nil {
			w.eTag = *resp.ETag
		}
		if resp.VersionId != nil {
			w.versionID = *resp.VersionId
		}
	}
	fs.Debugf(w.o, "multipart upload %q finished", *w.uploadID)
	return err
}

func (o *Object) uploadMultipart(ctx context.Context, src fs.ObjectInfo, in io.Reader, options ...fs.OpenOption) (wantETag, gotETag string, versionID *string, ui uploadInfo, err error) {
	chunkWriter, err := multipart.UploadMultipart(ctx, src, in, multipart.UploadMultipartOptions{
		Open:        o.fs,
		OpenOptions: options,
	})
	if err != nil {
		return wantETag, gotETag, versionID, ui, err
	}

	s3cw := chunkWriter.(*s3ChunkWriter)
	gotETag = *stringClone(s3cw.eTag)
	versionID = stringClone(s3cw.versionID)

	hashOfHashes := md5.Sum(s3cw.md5s)
	wantETag = fmt.Sprintf("%s-%d", hex.EncodeToString(hashOfHashes[:]), len(s3cw.completedParts))

	return wantETag, gotETag, versionID, s3cw.ui, nil
}

// Upload a single part using PutObject
func (o *Object) uploadSinglepartPutObject(ctx context.Context, req *s3.PutObjectInput, size int64, in io.Reader) (etag string, lastModified time.Time, versionID *string, err error) {
	req.Body = io.NopCloser(in)
	var options = []func(*s3.Options){}
	if o.fs.opt.UseUnsignedPayload.Value {
		options = append(options, s3.WithAPIOptions(
			// avoids operation error S3: PutObject, failed to compute payload hash: failed to seek body to start, request stream is not seekable
			v4signer.SwapComputePayloadSHA256ForUnsignedPayloadMiddleware,
		))
	}
	// Can't retry single part uploads as only have an io.Reader
	options = append(options, func(s3opt *s3.Options) {
		s3opt.RetryMaxAttempts = 1
	})
	var resp *s3.PutObjectOutput
	err = o.fs.pacer.CallNoRetry(func() (bool, error) {
		resp, err = o.fs.c.PutObject(ctx, req, options...)
		return o.fs.shouldRetry(ctx, err)
	})
	if err != nil {
		return etag, lastModified, nil, err
	}
	lastModified = time.Now()
	if resp != nil {
		etag = *stringClone(deref(resp.ETag))
		versionID = stringClonePointer(resp.VersionId)
	}
	return etag, lastModified, versionID, nil
}

// Upload a single part using a presigned request
func (o *Object) uploadSinglepartPresignedRequest(ctx context.Context, req *s3.PutObjectInput, size int64, in io.Reader) (etag string, lastModified time.Time, versionID *string, err error) {
	// Create the presigned request
	putReq, err := s3.NewPresignClient(o.fs.c).PresignPutObject(ctx, req, s3.WithPresignExpires(15*time.Minute))
	if err != nil {
		return etag, lastModified, nil, fmt.Errorf("s3 upload: sign request: %w", err)
	}

	// Set request to nil if empty so as not to make chunked encoding
	if size == 0 {
		in = nil
	}

	// create the vanilla http request
	httpReq, err := http.NewRequestWithContext(ctx, "PUT", putReq.URL, in)
	if err != nil {
		return etag, lastModified, nil, fmt.Errorf("s3 upload: new request: %w", err)
	}

	// set the headers we signed and the length
	httpReq.Header = putReq.SignedHeader
	httpReq.ContentLength = size

	var resp *http.Response
	err = o.fs.pacer.CallNoRetry(func() (bool, error) {
		var err error
		resp, err = o.fs.srv.Do(httpReq)
		if err != nil {
			return o.fs.shouldRetry(ctx, err)
		}
		body, err := rest.ReadBody(resp)
		if err != nil {
			return o.fs.shouldRetry(ctx, err)
		}
		if resp.StatusCode >= 200 && resp.StatusCode < 299 {
			return false, nil
		}
		err = fmt.Errorf("s3 upload: %s: %s", resp.Status, body)
		return fserrors.ShouldRetryHTTP(resp, retryErrorCodes), err
	})
	if err != nil {
		return etag, lastModified, nil, err
	}
	if resp != nil {
		if date, err := http.ParseTime(resp.Header.Get("Date")); err != nil {
			lastModified = date
		}
		etag = *stringClone(resp.Header.Get("Etag"))
		vID := *stringClone(resp.Header.Get("x-amz-version-id"))
		if vID != "" {
			versionID = &vID
		}
	}
	return etag, lastModified, versionID, nil
}

// Info needed for an upload
type uploadInfo struct {
	req       *s3.PutObjectInput
	md5sumHex string
}

// Prepare object for being uploaded
//
// If noHash is true the md5sum will not be calculated
func (o *Object) prepareUpload(ctx context.Context, src fs.ObjectInfo, options []fs.OpenOption, noHash bool) (ui uploadInfo, err error) {
	bucket, bucketPath := o.split()
	// Create parent dir/bucket if not saving directory marker
	if !strings.HasSuffix(o.remote, "/") {
		err := o.fs.mkdirParent(ctx, o.remote)
		if err != nil {
			return ui, err
		}
	}
	modTime := src.ModTime(ctx)

	ui.req = &s3.PutObjectInput{
		Bucket: &bucket,
		ACL:    types.ObjectCannedACL(o.fs.opt.ACL),
		Key:    &bucketPath,
	}

	// Fetch metadata if --metadata is in use
	meta, err := fs.GetMetadataOptions(ctx, o.fs, src, options)
	if err != nil {
		return ui, fmt.Errorf("failed to read metadata from source object: %w", err)
	}
	ui.req.Metadata = make(map[string]string, len(meta)+2)
	// merge metadata into request and user metadata
	for k, v := range meta {
		pv := aws.String(v)
		k = strings.ToLower(k)
		if o.fs.opt.NoSystemMetadata {
			ui.req.Metadata[k] = v
			continue
		}
		switch k {
		case "cache-control":
			ui.req.CacheControl = pv
		case "content-disposition":
			ui.req.ContentDisposition = pv
		case "content-encoding":
			ui.req.ContentEncoding = removeAWSChunked(pv)
		case "content-language":
			ui.req.ContentLanguage = pv
		case "content-type":
			ui.req.ContentType = pv
		case "x-amz-tagging":
			ui.req.Tagging = pv
		case "tier":
			// ignore
		case "mtime":
			// mtime in meta overrides source ModTime
			metaModTime, err := time.Parse(time.RFC3339Nano, v)
			if err != nil {
				fs.Debugf(o, "failed to parse metadata %s: %q: %v", k, v, err)
			} else {
				modTime = metaModTime
			}
		case "btime":
			// write as metadata since we can't set it
			ui.req.Metadata[k] = v
		default:
			ui.req.Metadata[k] = v
		}
	}

	// Set the mtime in the meta data
	ui.req.Metadata[metaMtime] = swift.TimeToFloatString(modTime)

	// read the md5sum if available
	// - for non multipart
	//    - so we can add a ContentMD5
	//    - so we can add the md5sum in the metadata as metaMD5Hash if using SSE/SSE-C
	// - for multipart provided checksums aren't disabled
	//    - so we can add the md5sum in the metadata as metaMD5Hash
	var md5sumBase64 string
	size := src.Size()
	multipart := size < 0 || size >= int64(o.fs.opt.UploadCutoff)
	if !noHash && (!multipart || !o.fs.opt.DisableChecksum) {
		ui.md5sumHex, err = src.Hash(ctx, hash.MD5)
		if err == nil && matchMd5.MatchString(ui.md5sumHex) {
			hashBytes, err := hex.DecodeString(ui.md5sumHex)
			if err == nil {
				md5sumBase64 = base64.StdEncoding.EncodeToString(hashBytes)
				if (multipart || o.fs.etagIsNotMD5) && !o.fs.opt.DisableChecksum {
					// Set the md5sum as metadata on the object if
					// - a multipart upload
					// - the Etag is not an MD5, eg when using SSE/SSE-C or directory buckets
					// provided checksums aren't disabled
					ui.req.Metadata[metaMD5Hash] = md5sumBase64
				}
			}
		}
	}

	// Set the content type if it isn't set already
	if ui.req.ContentType == nil {
		ui.req.ContentType = aws.String(fs.MimeType(ctx, src))
	}
	if size >= 0 {
		ui.req.ContentLength = &size
	}
	if md5sumBase64 != "" && !o.fs.opt.DirectoryBucket {
		ui.req.ContentMD5 = &md5sumBase64
	}
	if o.fs.opt.RequesterPays {
		ui.req.RequestPayer = types.RequestPayerRequester
	}
	if o.fs.opt.ServerSideEncryption != "" {
		ui.req.ServerSideEncryption = types.ServerSideEncryption(o.fs.opt.ServerSideEncryption)
	}
	if o.fs.opt.SSECustomerAlgorithm != "" {
		ui.req.SSECustomerAlgorithm = &o.fs.opt.SSECustomerAlgorithm
	}
	if o.fs.opt.SSECustomerKeyBase64 != "" {
		ui.req.SSECustomerKey = &o.fs.opt.SSECustomerKeyBase64
	}
	if o.fs.opt.SSECustomerKeyMD5 != "" {
		ui.req.SSECustomerKeyMD5 = &o.fs.opt.SSECustomerKeyMD5
	}
	if o.fs.opt.SSEKMSKeyID != "" {
		ui.req.SSEKMSKeyId = &o.fs.opt.SSEKMSKeyID
	}
	if o.fs.opt.StorageClass != "" {
		ui.req.StorageClass = types.StorageClass(o.fs.opt.StorageClass)
	}
	// Apply upload options
	for _, option := range options {
		key, value := option.Header()
		lowerKey := strings.ToLower(key)
		switch lowerKey {
		case "":
			// ignore
		case "cache-control":
			ui.req.CacheControl = aws.String(value)
		case "content-disposition":
			ui.req.ContentDisposition = aws.String(value)
		case "content-encoding":
			ui.req.ContentEncoding = removeAWSChunked(aws.String(value))
		case "content-language":
			ui.req.ContentLanguage = aws.String(value)
		case "content-type":
			ui.req.ContentType = aws.String(value)
		case "if-match":
			ui.req.IfMatch = aws.String(value)
		case "if-none-match":
			ui.req.IfNoneMatch = aws.String(value)
		case "x-amz-tagging":
			ui.req.Tagging = aws.String(value)
		default:
			const amzMetaPrefix = "x-amz-meta-"
			if strings.HasPrefix(lowerKey, amzMetaPrefix) {
				metaKey := lowerKey[len(amzMetaPrefix):]
				ui.req.Metadata[metaKey] = value
			} else {
				fs.Errorf(o, "Don't know how to set key %q on upload", key)
			}
		}
	}

	// Check metadata keys and values are valid
	for key, value := range ui.req.Metadata {
		if !httpguts.ValidHeaderFieldName(key) {
			fs.Errorf(o, "Dropping invalid metadata key %q", key)
			delete(ui.req.Metadata, key)
		} else if !httpguts.ValidHeaderFieldValue(value) {
			fs.Errorf(o, "Dropping invalid metadata value %q for key %q", value, key)
			delete(ui.req.Metadata, key)
		}
	}

	return ui, nil
}

// Update the Object from in with modTime and size
func (o *Object) Update(ctx context.Context, in io.Reader, src fs.ObjectInfo, options ...fs.OpenOption) error {
	if o.fs.opt.VersionAt.IsSet() {
		return errNotWithVersionAt
	}
	size := src.Size()
	multipart := size < 0 || size >= int64(o.fs.opt.UploadCutoff)

	var wantETag string        // Multipart upload Etag to check
	var gotETag string         // Etag we got from the upload
	var lastModified time.Time // Time we got from the upload
	var versionID *string      // versionID we got from the upload
	var err error
	var ui uploadInfo
	if multipart {
		wantETag, gotETag, versionID, ui, err = o.uploadMultipart(ctx, src, in, options...)
	} else {
		ui, err = o.prepareUpload(ctx, src, options, false)
		if err != nil {
			return fmt.Errorf("failed to prepare upload: %w", err)
		}

		if o.fs.opt.UsePresignedRequest {
			gotETag, lastModified, versionID, err = o.uploadSinglepartPresignedRequest(ctx, ui.req, size, in)
		} else {
			gotETag, lastModified, versionID, err = o.uploadSinglepartPutObject(ctx, ui.req, size, in)
		}
	}
	if err != nil {
		return err
	}
	// Only record versionID if we are using --s3-versions or --s3-version-at
	if o.fs.opt.Versions || o.fs.opt.VersionAt.IsSet() {
		o.versionID = versionID
	} else {
		o.versionID = nil
	}

	// User requested we don't HEAD the object after uploading it
	// so make up the object as best we can assuming it got
	// uploaded properly. If size < 0 then we need to do the HEAD.
	var head *s3.HeadObjectOutput
	if o.fs.opt.NoHead && size >= 0 {
		head = new(s3.HeadObjectOutput)
		//structs.SetFrom(head, &req)
		setFrom_s3HeadObjectOutput_s3PutObjectInput(head, ui.req)
		head.ETag = &ui.md5sumHex // doesn't matter quotes are missing
		head.ContentLength = &size
		// We get etag back from single and multipart upload so fill it in here
		if gotETag != "" {
			head.ETag = &gotETag
		}
		if lastModified.IsZero() {
			lastModified = time.Now()
		}
		head.LastModified = &lastModified
		head.VersionId = versionID
	} else {
		// Read the metadata from the newly created object
		o.meta = nil // wipe old metadata
		head, err = o.headObject(ctx)
		if err != nil {
			return err
		}
	}
	o.setMetaData(head)

	// Check multipart upload ETag if required
	if o.fs.opt.UseMultipartEtag.Value && !o.fs.etagIsNotMD5 && wantETag != "" && head.ETag != nil && *head.ETag != "" {
		gotETag := strings.Trim(strings.ToLower(*head.ETag), `"`)
		if wantETag != gotETag {
			return fmt.Errorf("multipart upload corrupted: Etag differ: expecting %s but got %s", wantETag, gotETag)
		}
		fs.Debugf(o, "Multipart upload Etag: %s OK", wantETag)
	}
	return err
}

// Remove an object
func (o *Object) Remove(ctx context.Context) error {
	if o.fs.opt.VersionAt.IsSet() {
		return errNotWithVersionAt
	}
	bucket, bucketPath := o.split()
	req := s3.DeleteObjectInput{
		Bucket:    &bucket,
		Key:       &bucketPath,
		VersionId: o.versionID,
	}
	if o.fs.opt.RequesterPays {
		req.RequestPayer = types.RequestPayerRequester
	}
	err := o.fs.pacer.Call(func() (bool, error) {
		_, err := o.fs.c.DeleteObject(ctx, &req)
		return o.fs.shouldRetry(ctx, err)
	})
	return err
}

// MimeType of an Object if known, "" otherwise
func (o *Object) MimeType(ctx context.Context) string {
	err := o.readMetaData(ctx)
	if err != nil {
		fs.Logf(o, "Failed to read metadata: %v", err)
		return ""
	}
	return o.mimeType
}

// SetTier performs changing storage class
func (o *Object) SetTier(tier string) (err error) {
	ctx := context.TODO()
	tier = strings.ToUpper(tier)
	bucket, bucketPath := o.split()
	req := s3.CopyObjectInput{
		MetadataDirective: types.MetadataDirectiveCopy,
		StorageClass:      types.StorageClass(tier),
	}
	err = o.fs.copy(ctx, &req, bucket, bucketPath, bucket, bucketPath, o)
	if err != nil {
		return err
	}
	o.storageClass = &tier
	return err
}

// GetTier returns storage class as string
func (o *Object) GetTier() string {
	if o.storageClass == nil || *o.storageClass == "" {
		return "STANDARD"
	}
	return *o.storageClass
}

// Metadata returns metadata for an object
//
// It should return nil if there is no Metadata
func (o *Object) Metadata(ctx context.Context) (metadata fs.Metadata, err error) {
	err = o.readMetaData(ctx)
	if err != nil {
		return nil, err
	}
	metadata = make(fs.Metadata, len(o.meta)+7)
	for k, v := range o.meta {
		switch k {
		case metaMtime:
			if modTime, err := swift.FloatStringToTime(v); err == nil {
				metadata["mtime"] = modTime.Format(time.RFC3339Nano)
			}
		case metaMD5Hash:
			// don't write hash metadata
		default:
			metadata[k] = v
		}
	}
	if o.mimeType != "" {
		metadata["content-type"] = o.mimeType
	}
	// metadata["x-amz-tagging"] = ""
	if !o.lastModified.IsZero() {
		metadata["btime"] = o.lastModified.Format(time.RFC3339Nano)
	}

	// Set system metadata
	setMetadata := func(k string, v *string) {
		if o.fs.opt.NoSystemMetadata {
			return
		}
		if v == nil || *v == "" {
			return
		}
		metadata[k] = *v
	}
	setMetadata("cache-control", o.cacheControl)
	setMetadata("content-disposition", o.contentDisposition)
	setMetadata("content-encoding", o.contentEncoding)
	setMetadata("content-language", o.contentLanguage)
	metadata["tier"] = o.GetTier()

	return metadata, nil
}

// Check the interfaces are satisfied
var (
	_ fs.Fs              = &Fs{}
	_ fs.Purger          = &Fs{}
	_ fs.Copier          = &Fs{}
	_ fs.PutStreamer     = &Fs{}
	_ fs.ListRer         = &Fs{}
	_ fs.ListPer         = &Fs{}
	_ fs.Commander       = &Fs{}
	_ fs.CleanUpper      = &Fs{}
	_ fs.OpenChunkWriter = &Fs{}
	_ fs.Object          = &Object{}
	_ fs.MimeTyper       = &Object{}
	_ fs.GetTierer       = &Object{}
	_ fs.SetTierer       = &Object{}
	_ fs.Metadataer      = &Object{}
)