gotd: implement the delete-refs capability Allow clients to run "got send -d" against gotd. Clients will send a zero-id as new id for a reference and, in the special but more common case of just deleting and not updating, no pack file will be sent. ok and tweaks by stsp@

update client state tracking in the gotd parent process The session process takes over the old state definitions under a new name ("session state"). The parent only needs to keep track of whether a client has been granted access, so it only uses two states: NEW, and ACCCESS_GRANTED which is set as soon as the auth process has granted repository access and before the session and repo_read/repo_write children are started. Because 'gotctl info' can no longer observe the session state remove support code for printing it. ok op@

remove support for showing client capabilities in 'gotctl info' The gotd parent process has lost access to client capabilities. Take the easy way out and remove related code. If needed, client capabilities can still be found in the debug log with 'gotd -v'. ok op, jamsek

add a gotd session process, split off from the parent process The new session process is able to manipulate files in the repository and keeps track of the read/write client session state. The parent process now restricts its view of the filesystem to the absolute path stored in argv[0], and combines this with unveil "x" on this path. As a result the parent process can only re-exec itself. small tweaks + ok op@

remove the gotsh group requirement from gotd; any user can now connect Repository access is now controlled by access rules in gotd.conf, and concurrent connections to the gotd socket by local users are limited by the listen process. We should keep refining our anti-DoS measures in the future, but at least we have something in place now. ok jamsek, op

introduce connection options to gotd.conf Allow administrators to tweak the default authentication and request timeouts if needed, and to tweak the limit of concurrent connections for specific user accounts. with several tweaks from and ok op@

enforce a per-uid connection limit in the gotd listen process For now the limit is set at compile-time. It will become configurable via gotd.conf soon. ok op@

move "unix" pledge promise from gotd parent to auth process The listen process now communicates the client UID/GID to the parent, and the auth process verifies this on behalf of the parent. This allows us to remove the "unix" pledge promise from the parent, removing parent access to syscalls such as listen() and accept() in the AF_UNIX domain. ok tracey@ op@

run gotd authentication in a separate child process ok op@

fork gotd repo_read/repo_write children on demand ok op, jamsek

switch gotd from chroot(2) to unveil(2) In the future, gotd will fork+exec new processes for each client connection. Using unveil instead of chroot avoids having to start such processes as root. The -portable version could use chroot(2) where no equivalent to unveil(2) exists. A future component which starts new processes will be isolated as a separate process, which could run as root in the -portable version. ok op@

add a gotd "listen" process which watches the unix socket ok op@

rename GOTD_SOCK_FILENO to GOTD_FILENO_MSG_PIPE for clarity (it's not a socket)

fix group membership check in gotd auth ok op@

implement per-repository read/write authorization rules in gotd ok op@

add gotctl(8); initially supported commands are 'info' and 'stop' This will be used by an upcoming regress test suite for gotd(8). ok tracey

introduce gotd(8), a Git repository server reachable via ssh(1) This is an initial barebones implementation which provides the absolute minimum of functionality required to serve got(1) and git(1) clients. Basic fetch/send functionality has been tested and seems to work here, but this server is not yet expected to be stable. More testing is welcome. See the man pages for setup instructions. The current design uses one reader and one writer process per repository, which will have to be extended to N readers and N writers in the future. At startup, each process will chroot(2) into its assigned repository. This works because gotd(8) can only be started as root, and will then fork+exec, chroot, and privdrop. At present the parent process runs with the following pledge(2) promises: "stdio rpath wpath cpath proc getpw sendfd recvfd fattr flock unix unveil" The parent is the only process able to modify the repository in a way that becomes visible to Git clients. The parent uses unveil(2) to restrict its view of the filesystem to /tmp and the repositories listed in the configuration file gotd.conf(5). Per-repository chroot(2) processes use "stdio rpath sendfd recvfd". The writer defers to the parent for modifying references in the repository to point at newly uploaded commits. The reader is fine without such help, because Git repositories can be read without having to create any lock-files. gotd(8) requires a dedicated user ID, which should own repositories on the filesystem, and a separate secondary group, which should not have filesystem-level repository access, and must be allowed access to the gotd(8) socket. To obtain Git repository access, users must be members of this secondary group, and must have their login shell set to gotsh(1). gotsh(1) connects to the gotd(8) socket and speaks Git-protocol towards the client on the other end of the SSH connection. gotsh(1) is not an interactive command shell. At present, authenticated clients are granted read/write access to all repositories and all references (except for the "refs/got/" and the "refs/remotes/" namespaces, which are already being protected from modification). While complicated access control mechanism are not a design goal, making it possible to safely offer anonymous Git repository access over ssh(1) is on the road map.