feat(identity): implement --copy-as USER[:GROUP] safe subset (P7 Wave E)

Force the receiver to apply the requested owner/group to every written
entry through the confined fd-relative identity path instead of switching
the process credentials (unsafe for the multithreaded receiver).  An
unprivileged receiver refuses the transfer up front in server_module_gate,
before STATUS_OK, so no data is written with the wrong ownership.

- new Config fields copy_as_set/copy_as_uid/copy_as_gid + defaults
- identity_parse_copy_as (name/@N/* resolution, primary-gid default,
  gid==uid fallback for numeric ids with no passwd entry); implies -M
- identity snapshot + highest-priority forcing in identity_resolve_targets
- identity_copy_as_refused() helper
- trailing config-frame block (presence int + two int32 ids, >=0 checked)
- PROTOCOL_VERSION 2.17.0 -> 2.18.0; version-sensitive tests updated
- unit tests for parse + wire round-trip/negative-id rejection
- integration TestCopyAs: unprivileged refusal + root chown assertion
- RSYNC_COMPAT.md --copy-as row updated (safe subset + divergence); README
  protocol version refreshed
This commit is contained in:
2026-09-12 11:58:37 +02:00
parent f64d252faf
commit 80dd64aae6
12 changed files with 474 additions and 18 deletions
+32 -2
View File
@@ -439,6 +439,20 @@ typedef struct Config {
* a reserved user.fastsync.stat xattr recording the source uid/gid/mode/mtime
* so a later privileged restore could re-apply them. Crosses the wire. */
bool fake_super;
/* --copy-as=USER[:GROUP] (P7 Wave E, protocol 2.18.0). Safe-subset
* implementation, a documented divergence from rsync's real identity switch:
* the receiver does NOT change its process credentials (FastSync's receiver
* is multithreaded, so a setuid/seteuid drop would be unsafe). Instead the
* receiver FORCES the ownership of every entry it writes to copy_as_uid /
* copy_as_gid through the existing confined, fd-relative identity path
* (fchown/fchownat), which REQUIRES receiver privilege (root); an
* unprivileged receiver REFUSES the whole transfer up front at the config
* handshake (never a silent wrong-ownership result). All three fields CROSS
* the wire as a trailing config-frame block so the receiver learns the
* requested ids; see the PROTOCOL_VERSION note below. */
bool copy_as_set;
int32_t copy_as_uid;
int32_t copy_as_gid;
// Phase 5: --trust-sender
/* Long-form-only, receiver-local policy. rsync's --trust-sender tells the
@@ -563,8 +577,24 @@ typedef struct Config {
* would desynchronize on the unknown frame, and the strict same-version
* handshake (config_receive rejects a mismatched version before parsing
* anything else) is what keeps a 2.17 client and a 2.16 server from ever
* reaching that state. */
#define PROTOCOL_VERSION "2.17.0"
* reaching that state.
*
* Privilege Wave (P7 Wave E): 2.17.0 -> 2.18.0.
*
* WHY the bump, grounded in the wire: --copy-as=USER[:GROUP] adds a serialized
* field to the binary config frame. The client sends, as a new trailing block
* AFTER the --iconv CONVERT_SPEC string (in config_send/config_receive), a
* presence int followed, when set, by the target uid and gid (both int32).
* The receiver needs those ids to force the ownership of every entry it writes
* (the safe-subset --copy-as model; see RSYNC_COMPAT.md), and it REQUIRES
* receiver privilege: an unprivileged receiver refuses the transfer at the
* config handshake (server_module_gate) instead of silently ignoring the flag.
* Any config-frame layout change must bump the protocol version: a peer that
* does not parse the new trailing bytes would desynchronize on the frame
* boundary, and the strict same-version handshake (config_receive rejects a
* mismatched version before parsing anything else) is what keeps a 2.18 client
* and a 2.17 server from ever reaching that state. */
#define PROTOCOL_VERSION "2.18.0"
#define DEFAULT_CHUNK_SIZE (10 * 1024 * 1024)
/* Upper bound on total basis-dir entries (rsync caps --link-dest at 20). */
#define MAX_BASIS_DIRS 64