When serving FTP with --auth-proxy, the obscured password was cached in a
driver-global map keyed only by the username. Two sessions that logged in
with the same username but different credentials shared one map entry, so a
later login overwrote it and every subsequent operation on the earlier,
still-authenticated session was re-authorized with the later session's
credential and executed against the later session's backend.
Bind the credential to the FTP session by storing the obscured password in
the per-session goftp Session.Data map instead, so each session always
resolves the backend it authenticated for.
With --auth-proxy set and --auth-key unset, serve s3 registered every client
supplied access key ID with an empty secret and verified the SigV4 signature
against that, so anyone could sign a request for an arbitrary access key ID with
an empty secret and be let in. The proxy program was only ever given the access
key ID (as both user and pass) so it had nothing with which to authenticate the
client either.
An S3 client never sends its secret, only a signature made with it, so the
server has to know the secret to check the request. The auth proxy protocol as
been changed to handle this. For serve s3 the proxy program is given just the
access key ID as the user (no pass or public_key) and must return the matching
secret as _secret_access_key in its output. rclone verifies the request's
signature against that secret, refusing the request if the proxy rejects the
access key ID, doesn't return a secret or returns an empty one, or the signature
doesn't match. The secret is only used for this server's own verification and is
never registered with gofakes3, so other serve s3 instances in the same process
don't honour it.
The proxy's answers are cached. If a signature fails against a cached secret the
proxy is consulted again so a rotated secret takes effect immediately - but only
for a signature mismatch, and at most once every 10 seconds per access key ID
and client IP, so a stream of bad signatures can't make the proxy program run
for every request. A rotation never shuts down the cached backend under requests
still using it. A cached answer is checked with the proxy again once it is 5
minutes old even if in constant use, so revoking an access key ID takes effect
within 5 minutes.
This means --auth-key is no longer needed with --auth-proxy: it is ignored and a
warning is given at startup if both are set. The proxy is the source of truth
for both the credentials and the backend they map to. Presigned URLs (credential
in the query string) are now recognised by the proxy middleware too. The auth
proxy docs are added to serve s3.
Note that the serve s3 auth proxy protocol has changed. The proxy program is now
given the access key ID as "user" (it was previously given an MD5 hash of it,
with the access key ID as "pass") and must return the matching secret as
"_secret_access_key".
This needs gofakes3 v0.0.9 for signature.V4SignVerifyWithSecret.
gofakes3 kept the keys given with --auth-key in a store global to the process,
so when more than one serve s3 was running in one rclone (eg started via the rc)
each accepted the others' credentials and a client with the key for one server
could read and write the backend of another.
This updates gofakes3 to v0.0.9 which keeps auth keys per instance and adds a
test that two servers only accept their own keys.
From v1.70.0, an SFTP server started through the rc serve/start API with
a per-server proxyOpt.AuthProxy decided whether to enable proxy
authentication by checking the process-global proxy.Opt.AuthProxy
instead of the supplied proxyOpt.AuthProxy. In the normal rc case the
global is empty, so the auth proxy was silently ignored: the server
either failed to start with "no authorization found" or authenticated
against the local authorized_keys file instead of routing each login
through the proxy the operator configured.
The serve Provider refactor (f425f8d46) fixed the constructor by building the
provider from the supplied proxyOpt, but the authorized-keys handling in
configure() still consulted the global option. Make it depend on whether
proxy mode is actually active, and add a regression test for the
per-server option.
From v1.70.0 until the serve Provider refactor (f425f8d46), an S3 server started
through the rc serve/start API with a per-server proxyOpt.AuthProxy
decided whether to enable proxy authentication by checking the
process-global proxy.Opt.AuthProxy instead of the supplied
proxyOpt.AuthProxy. In the normal rc case the global is empty, so the
auth proxy was silently ignored and the server served the fixed
filesystem supplied to serve/start rather than routing each access key
to the backend chosen by the proxy, bypassing the operator's intended
per-key authorization.
The Provider refactor fixed this incidentally by building the provider
from the proxyOpt passed to the constructor. This adds a regression test
so the per-server option cannot silently stop working again, and only
logs "allowing anonymous access" when neither an auth key nor an auth
proxy is configured so the log reflects the effective mode.
From v1.70.0 until the serve Provider refactor (f425f8d46), an FTP server started
through the rc serve/start API with a per-server proxyOpt.AuthProxy
decided whether to enable proxy authentication by checking the
process-global proxy.Opt.AuthProxy instead of the supplied
proxyOpt.AuthProxy. In the normal rc case the global is empty, so the
auth proxy was silently ignored and the server fell back to its
fixed-backend mode, whose default account accepts user "anonymous" with
any password - a complete authentication bypass.
The Provider refactor fixed this incidentally by building the provider
from the proxyOpt passed to the constructor. This adds a regression test
so the per-server option cannot silently stop working again.
The multipart reorder-buffer admission trusted the client-declared part length.
A negative length was accepted, and `buffered + size` could overflow int64 for
a huge declared length, wrapping the running total negative and admitting
further parts past --multipart-streaming-buffer-limit.
Reject a negative length and use the overflow-safe comparison `size <=
bufferLimit - buffered` so an untrusted Content-Length can neither poison nor
overflow the budget.
Streamed multipart UploadPart called Reserve(contentLength) before reading any
body bytes, so the pool immediately allocated one 1 MiB page per MiB of the
client-declared Content-Length (or X-Amz-Decoded-Content-Length). An client
could declare a huge part size, send no body, and force an arbitrarily large
allocation without paying the bandwidth cost of the declared body.
Drop the Reserve so the pool-backed buffer grows a page at a time as the body
is actually read: memory now tracks the bytes received, not the unverified
header.
A client which started a multipart upload and vanished without either
completing or aborting it used to hold on to its resources forever.
Incomplete multipart uploads which have had no activity for
--multipart-expiry (default 24h) are now aborted and cleaned up
exactly as if the client had called AbortMultipartUpload, with a
NOTICE logged.
An upload with a part still being received is never expired, and each
completed part restarts the clock. Late operations on an expired
upload fail with NoSuchUpload, as they do on real S3 when a lifecycle
rule has aborted the upload.
Set --multipart-expiry 0 to keep incomplete uploads forever.
Multipart uploads used to be streamed directly to the remote with their
own PutStream machinery, bypassing the VFS, a design left over from
before the VFS could abandon a streaming write.
They are now written through the VFS exactly like plain object PUTs in
every cache mode. The parts are written, in part-number order, to a
temporary object which is renamed into place server-side on
completion.
With the default --vfs-cache-mode off the parts stream through the VFS
to the remote as they arrive. With --vfs-cache-mode writes or above
they are buffered in the VFS cache and uploaded by its write-back.
User visible changes:
- Multipart uploads now show in rclone's transfer stats and obey
--bwlimit (previously they bypassed both).
- Remotes without streaming upload support now spool the upload to a
temporary file on local disk instead of buffering it in memory.
- Multipart uploads are never buffered in memory because of missing
remote capabilities - only --disable-multipart-streaming does that.
- Remotes that upload atomically now also write to a temporary object
renamed into place, so an in-progress multipart upload is no longer
briefly visible under its final key.
- On the few remotes with no server-side move or copy the parts are
written straight to the final object in all cache modes.
- With --vfs-cache-mode writes, plain PUTs and multipart uploads to the
same key go through the same cache entry, so an earlier PUT still in
the write-back window can no longer be written back over a newer
multipart upload.
- Failed write-backs are retried by the VFS without the client having
to restart the upload, and completed objects are served from the
cache for read-after-write.
The temporary objects that uploads are written to before being renamed
into place are now named .rclone_temp_put_* and .rclone_temp_multipart_*,
and the whole .rclone_temp_ prefix is reserved: any object whose name
starts with it is hidden from S3 listings. This gives a single pattern
for cleaning up leftovers from killed servers:
rclone delete --min-age 24h --include ".rclone_temp_*" remote:path
The .rclone_multipart_upload_* objects rclone v1.75 used are still
hidden from listings so leftovers from an older server stay invisible
to S3 clients.
The mtime metadata fallback was nested inside the X-Amz-Meta-Mtime
branch, so it only ran when X-Amz-Meta-Mtime was present but invalid -
and then set the modtime from the invalid value's failed parse rather
than parsing mtime. An object PUT with only mtime metadata kept the
upload time as its modtime.
Now the two keys are checked independently, as TouchObject already
does.
streamPart did not check whether the upload had been torn down, but
AbortMultipartUpload sets the reorder buffer map to nil, so an abort
arriving while a part body was still being received panicked with an
assignment to a nil map once the part was buffered.
Now a part whose upload has been aborted or completed under it is
rejected with NoSuchUpload.
A PUT which failed part way through removed the object at the
destination key. As well as differing from real S3 (where a failed PUT
never affects the stored object), this raced with the client's
automatic retry of the same PUT: the retry stored the object and
returned 200 OK, then the failed first attempt's cleanup deleted it,
silently losing an acknowledged upload. The interrupted upload could
also be committed as a truncated object, since closing the write handle
gave the streaming upload a clean end of stream.
Now a failed or interrupted PUT never disturbs the object at the key:
- The object at the key is never removed on error.
- On backends where a partial upload is visible at its final name
(PartialUploads), and when the VFS cache mode is writes or above, the
upload is written to a temporary object which is renamed into place
on success and removed on failure, as streamed multipart uploads
already do. Backends which upload atomically are still streamed
straight to the destination.
- An interrupted or short body fails the upload via
WriteFileHandle.CloseWithError instead of committing truncated data,
and a body which ends cleanly short of its declared size is rejected
with IncompleteBody.
The auth proxy was only given the user and their password or public
key, so a proxy program had no way to restrict logins to particular
networks, or to record where an authentication attempt came from.
The JSON sent to the program now has a client_ip key holding the bare
IP the client connected from, with the port stripped so IPv6 arrives
as 2001:db8::1 rather than [2001:db8::1]:52344. An IPv4-mapped IPv6
address is reported as plain IPv4 so that a client arriving over a
dual-stack listener still matches IPv4 networks. The key is omitted
when the client has no IP address.
The IP is also mixed into the backend cache key. That is needed as the
program is only run on a cache miss, so a client from a
non-allowlisted address presenting valid credentials within the 5
minute cache lifetime would get a cache hit and be let in without the
program being consulted at all.
The macOS NFS client sends SETATTR after SYMLINK, which arrives as
Chmod/Chown on the link path. These opened the target with vfs.Open,
which follows symlinks - a freshly created symlink usually dangles, so
the open failed with ENOENT, surfaced to the client as NFS3ERR_IO even
though the link was created.
Add path-based VFS.Chmod and VFS.Chown mirroring VFS.Chtimes. They do
not follow symlinks (lstat semantics, matching VFS.Stat) and return
ENOSYS when the node exists, since the VFS stores neither permissions
nor ownership; serve nfs calls them and masks ENOSYS as before.
Fixes#9627
A request path beginning with "../" escaped the path the server was
started on, letting a client list, read, create, overwrite and delete
objects outside it.
The check added for CVE-2026-59733 rejected non-canonical paths by
comparing them with path.Clean, but path.Clean cannot resolve leading
".." elements in a relative path so it leaves them in place and the
comparison comes out equal. Only interior traversal such as "a/../../x"
was rejected. Whether a path then escaped depended on the backend:
those which join the root with the remote before encoding it - webdav,
ftp, sftp, http and memory - resolved the ".." away, while local and s3
encode the dot elements first and were unaffected.
A bare "." was accepted for the same reason, which on bucket backends
addresses the served directory's own key.
Validate with io/fs.ValidPath instead, which rejects ".", ".." and empty
elements wherever they appear. The empty path stays valid as the root of
the API, and "." is excluded explicitly because ValidPath accepts it as
the root of an FS.
Three ways a single client could deny service to everyone else connected to
the same serve sftp process:
A session "subsystem" request payload is a length-prefixed string, but it was
sliced at a fixed offset without checking its length, so a client sending a
truncated payload panicked the out-of-band request goroutine and killed the
process. Decode the payload instead, the way the neighbouring "exec" request
already does.
Rejecting a request then left the goroutine handling that channel waiting
forever to learn what kind of channel it was, because nothing was ever sent
on the channel it waits on. A client could open channels in a loop making
unsupported requests and grow the server's goroutines and memory without
bound. Signal the waiter when the requests run out so the channel is torn
down.
Separately, the request handlers - and reads, writes and closes on the file
handles they return - run on pkg/sftp packet worker goroutines which have no
panic recovery. A panic raised by a backend while serving one request took the
process down with it. Recover panics at that boundary, log them with a stack
trace, and return them to the requesting client as an error instead.
Addresses GHSA-6jcg-q3wp-x2f4.
The builtin authentication compared the configured username and password
with ==, whose run time depends on how much of the value matches, giving a
timing side-channel that could in principle help guess the password.
serve sftp, serve s3 and the auth proxy already use subtle.ConstantTimeCompare
so bring serve ftp in line with them. An empty configured password
still accepts any password.
Addresses GHSA-mfvx-7rcj-9m5g finding 4.
The mount served fine on Linux and macOS but *BSD kernel NFS clients
refused it with "Authentication error". The MOUNT reply only ever
offered AUTH_NULL, and the OpenBSD/FreeBSD/NetBSD clients won't mount a
server unless AUTH_UNIX is among the offered flavors.
Add AUTH_UNIX to the advertised list. The server still doesn't inspect
the credential (there's no per-user access control here) so the AUTH_UNIX
cred the client then sends is read as an opaque blob and ignored, exactly
as the AUTH_NULL one was. No behaviour change for existing Linux/macOS
clients, and OpenBSD now mounts and reads files over the share.
Verified on a real OpenBSD 7.9 arm64 VM: registered rclone's server in
portmap and ran "mount_nfs -T localhost:/ /mnt"; the mount now succeeds
(MOUNT and GetAttr RPCs go through) and files read back correctly through
the mount, where before it stopped at the auth stage.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Previously, GET requests for a directory URL always returned an HTML
directory listing. There was no way to suppress this, unlike
`serve webdav` which has supported --disable-dir-list since #4191.
This adds the same flag to `serve http`. When set, GET requests for
directory URLs return 404 instead of a listing, while file downloads
continue to work normally.
Based on the approach suggested in #6306.
Fixes#4000
This refactors all 8 serve protocols to use `rc.ParseOptions` for VFS
and protocol options decoding. It also implements parameter validation in the
main runner using `rc.CheckParamsUsed`.
- Update dlna, ftp, http, nfs, restic, s3, sftp, and webdav to call
rc.ParseOptions, enabling nested option block support.
- Remove unused configstruct imports.
- Update startRc in cmd/serve/rc.go to copy input parameters and
call rc.CheckParamsUsed to reject unknown parameters.
- Add TestRcStartFlatNestedAndUnknownRejection to cmd/serve/rc_test.go.
- Update inline documentation in serve/start command help to include
nested blocks information (vfsOpt, proxyOpt, opt) and a new WebDAV example.
The unauthenticated DLNA control endpoint decoded arbitrary SOAP bodies
into an in-memory XML field. A LAN client could send a large request
and exhaust the server's memory.
Limit SOAP request bodies to 1 MiB and return 413 when the limit is
exceeded.
Streamed multipart uploads had two problems when the client uploaded parts
concurrently:
Parts arriving ahead of the next part needed by the backend stream were buffered
in memory without limit, acknowledging each part as soon as it was received. A
client uploading faster than the backend could drain would therefore balloon the
server's memory to the size of the upload. Buffering is now bounded a new
--multipart-streaming-buffer-limit flag (default 256Mi, 0 for unlimited): a part
that would take the buffer over the limit is not read until the stream drains,
applying backpressure to the client instead of using unbounded memory.
A part uploaded again with the same number - typically a client retrying after
its request timed out - left a stale copy in the reorder buffer which made
CompleteMultipartUpload fail with InvalidPart, aborting the whole upload.
Re-uploaded parts are now handled properly: a copy still in the buffer is
replaced, an identical copy of an already streamed part is accepted as a no-op,
and only replacing an already streamed part with different content (which the
in-order stream cannot honour) is rejected.
Go 1.27 changes compress/flate output, which also changes archive/zip
byte output. The HTTP zip download tests currently compare raw zip bytes
against golden files, so they fail even though the generated zip archives
contain the expected files.
Compare zip entries and decompressed contents instead of the exact
compressed byte stream.
Signed-off-by: Mikel Olasagasti Uranga <mikel@olasagasti.info>
Docker may re-send Create requests for volumes that already exist,
especially after a plugin restart. Previously this returned
ErrVolumeExists which Docker surfaced as "volume name must be unique".
Now if a volume with the same name already exists, Create returns
success (no-op), matching the Docker volume plugin protocol's
expectation of idempotent operations.
When restoring multiple volumes from saved state, each volume's
filesystem setup (including connecting to the remote) was done
sequentially. Now volumes are restored concurrently, so a single
slow or unreachable remote doesn't delay the restoration of other
volumes.
When restoring volumes from saved state during plugin startup, a slow
or unreachable remote could block indefinitely in fs.NewFs. Add a
30-second per-volume timeout context so that individual volume failures
are logged and skipped rather than blocking the entire plugin.
Previously, restoreState in NewDriver would restore volumes AND perform
FUSE mounts synchronously before the Docker plugin socket was created.
This caused two problems:
1. The monChan was created after restoreState, but mount() sends on
monChan, causing a deadlock (send on nil channel blocks forever).
2. Even with the channel fix, slow or hanging mounts during state
restore would block the server socket from being created, causing
Docker to time out after ~13 seconds with "no such file or
directory" for the plugin socket.
Fix by:
- Moving monChan creation and monitor goroutine start before
restoreState
- Splitting state restore into two phases: restoreState (metadata +
filesystem setup only) and RestoreMounts (actual FUSE mounts)
- Calling RestoreMounts asynchronously after the server starts
listening
- Performing mount restoration concurrently across volumes
Fixes#9231
The volume plugin parses the remote option as a trusted connection
string, which can run local commands via backend options. Spell out
that access to the unix or TCP socket is equivalent to command
execution as the serving user, document the unix socket permissions,
and warn that the TCP socket is unauthenticated.
Before this change a streamed multipart upload wrote its parts
straight to the object's final path on the underlying remote. That
meant an in-progress upload overwrote any object already stored under
that name, and aborting or failing the upload destroyed it. The
opposite of the S3 guarantee that an object only changes on a
successful CompleteMultipartUpload.
Remotes that upload atomically already (PartialUploads is false, e.g.
object stores) are safe to stream straight to the destination, so they
still do. Remotes where a partial upload is visible (PartialUploads is
true, e.g. local) now stream the parts to a temporary object instead
and move it, server-side, into its final place only when the upload
completes. A failed or aborted upload then just removes the temporary
object and leaves any pre-existing object untouched.
The temporary-object path needs the remote to support a server-side
move or copy in addition to PutStream uploads fall back to being
buffered in memory as before. The temporary objects are named with a
leading ".rclone_multipart_upload_" and hidden from listings.
Before this change when an aborted upload had overwritten a
pre-existing object of the same name, this left a ghost of that object
in every listing.
This invalidate the VFS cache on all the multipart upload abort paths,
so listings reflect what is actually on the underlying Fs.
A user could reach another user's private repository by sending a path
such as /<me>/../<victim>/config. The authorization check compares the
first path segment against the authenticated user, while the backend
object key was built from the raw, un-cleaned URL path.
Reject any non-canonical request path so the authorization segment and
the backend object key can no longer disagree.
Fixes GHSA-fqj9-69pf-6pjg
S3 object keys are opaque names that may legally contain `..` segments. `serve
s3` built backend paths with `path.Join(bucket, key)`, which normalised the key
so a request such as `GET /bucket/../root-secret.txt` resolved to a file outside
the selected bucket elsewhere under the serve root. Listing prefixes and
multipart uploads were affected also.
This did not allow reading of files outside the root, but did allow reading of
files in the root which normally aren't visible; only directories are visible as
buckets normally.
Because `serve s3` maps keys to file paths it cannot represent every opaque S3
key, so rather than normalising keys (which would alias distinct keys onto one
file as well as allow traversal) it now rejects any key that is not already in
canonical path form - containing `..`, `.`, `//` or a leading or trailing slash
- with a 400 Bad Request, as MinIO does. Directory listing prefixes are
validated the same way but allow the empty bucket-root prefix and an optional
trailing slash.
Fixes: GHSA-8v25-v8p6-qf7v
The --disable-zip flag was registered manually and was missing from
OptionsInfo, so it could not be set over the rc interface. Move it
into OptionsInfo like serve webdav does, which keeps the command line
flag and also makes it settable via rc.
Per RFC 4918 section 10.6, when the Overwrite header is omitted from a
COPY or MOVE request the resource MUST treat the request as if
Overwrite: T had been sent.
The upstream golang.org/x/net/webdav library mishandles this for MOVE
by checking == "T" instead of != "F", so an absent header is treated
as Overwrite: F and the request fails with 412 Precondition Failed.
Normalise the header to T in the rclone WebDAV server before
delegating to the upstream handler when the client did not send one.
This restores RFC-compliant default behaviour and can be removed once
the upstream fix in golang/go#66059 lands and the golang.org/x/net
dependency is bumped.
Fixes#9496
After an upload (notably multipart) to a slow backing remote, the file
lives in the VFS and is returned by ListBucket, but node.DirEntry() stays
nil until the --vfs-write-back writeback completes. HeadObject and
GetObject returned gofakes3.KeyNotFound while it was nil, so a HEAD/GET in
that window 404'd even though the object existed.
getFileHashByte already falls back to hashing the VFS cache when the
backing object isn't available yet. Drop the early nil return, pass the
node (not the fs.Object) to getFileHashByte, and take the Content-Type
from fs.MimeTypeFromName when the backing object isn't there yet.
TestS3Minio brings up a minio container via the fstest/testserver
framework, which exec's bash init.d scripts that shell out to docker.
This is not available on all platforms - Windows has no POSIX shell to
run the scripts, and macOS CI runners have no docker daemon - which
caused the build to fail there.
Add testy.SkipUnlessDocker to detect whether the framework can run and
skip the test when it cannot.
Previously serve s3 buffered every part of a multipart upload in memory
(in the gofakes3 S3 library) and concatenated them when the upload
completed, so memory use grew with the size of the upload.
serve s3 now streams the parts, in part-number order, into a single
PutStream upload to the underlying remote, which performs its own upload
with bounded memory. The whole file is never held in memory - memory use
is bounded by the parts in flight. This works for any remote that
supports PutStream (nearly all, including crypt) and for any part size,
so clients that don't produce uniform-sized parts (for example
PostgreSQL backup tools such as pgBarman and pgBackRest) work too.
Parts must arrive in ascending, contiguous part-number order; parts
uploaded out of order are buffered until their turn, and there is no
per-part retry (a failure aborts the whole upload). These trade-offs are
documented.
Passing --disable-multipart-streaming, or using a remote without
PutStream, reverts to buffering the parts in memory (the previous
behaviour); a one-off NOTICE is logged the first time this happens.
Fixes#7453
Run still uses a fresh local directory as the backing Fs that the
server wraps. RunWithBackend takes an extra remote name (e.g.
"TestS3Minio:") and uses a random subdirectory of that remote instead,
starting the matching fstest/testserver/init.d script on the way in
and tearing it down on the way out.
AuthProxy is only run for the local backend.
Before this change, if the user changed their password or public-key
and the auth proxy script returned updated config parameters for the
backend (eg a rotated api_key) rclone would continue to re-use the old
backend with the old config parameters out of the fscache.
This was because both the VFS cache and the fs/cache key were derived
from the user name only, so a change in the user's password or
public-key did not invalidate the cached backend.
Fix this by deriving the cache key from the user plus a hash of the
password/public-key, so a credential change forces a fresh backend.
The hash uses a per-process random HMAC key so the fragment that
appears in logs cannot be brute-forced offline.
When a SETSTAT request set the access and modification times, the
handler passed the modification time for both, discarding the
requested access time. Pass the requested access time through instead.
The VFS currently ignores the atime, but it might use it one day.
The statvfs@openssh.com extension was advertised but returned an
unsupported status, so clients couldn't query the amount of free and
used space. Implement it using the VFS Statfs method, which reports the
backend's usage where the backend supports About.
Nick Craig-Wood
Hakan İSMAİLandNick Craig-Wood
am-at-enrollvbandGitHub
SillyZirandNick Craig-Wood
Leon BrocardandNick Craig-Wood
Hakan İSMAİLandNick Craig-Wood
Acts1631andGitHub
Mikel Olasagasti UrangaandNick Craig-Wood
Sanjay SanthanamandGitHub
blackflytechandNick Craig-Wood
maxandNick Craig-Wood