Commits


avoid copying reused deltas to delta cache file; copy from pack file instead ok op@


use sub-second precision when checking for objects/pack/ modification Convert from st.m_time (second-precision time_t) to st.m_tim (struct timespec). To compensate for the potential case where a filesystem provides resolution in seconds only, always read the directory if no pack files are known to exist. Otherwise, there is a race condition when gotd repo_write creates a new pack and a request arrives for repo_read shortly after. Caught by a regression test for gotd on empty repositories. Test failure pointed out by Mikhail. ok op@


remove trailing whitespace; patch by Josiah Frentsos


respect resource limit for number of open files when sizing the pack file cache ok op@


check size before calling mmap(2) It's only a preparatory step, as checking whether a size_t is less than SIZE_MAX is moot. In a follow-up commit, however, the `filesize' field of the struct got_pack will become off_t and these checks will kick in. This also makes consistent how we guard mmap(2) against empty files. ok and improvements stsp@


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.


refresh cached list of pack index paths while searching a packed object Previously, this list was only refreshed while trying to match an object ID prefix. Regular pack file access needs to refresh this list, too. In particular, future gotd(8) needs this to ensure that newly uploaded packfiles are picked up as expected.


move code for reading Git's config file into new file read_gitconfig_privsep.c The end goal here is to remove the dependency of repository.c on privsep.c during compilation.


add gotadmin init -b <branch> to specify repo head ref Similar to `git init -b`. Includes a change to `got import` behaviour such that "main" is no longer hardcoded by default; instead, we import to the branch resolved via the repository's HEAD reference unless `got import -b` is used, and only if HEAD cannot be resolved to a branch do we fallback to "main". includes fix plus ok from stsp@


free the return of got_repo_get_path_objects{,_pack}


plug mem leak in got_repo_pack_fds_open pack_fds_tmp is not always free'd. However, while here, simplify it to not use a temporary array. ok stsp@


plug a leak in match_loose_object in the loop of match_loose_object we allocate a string per directory entry that in some case it was free(3)'d before `continue' or `goto', but not always. Instead, use a more common idiom. ok stsp@


refresh our pack-index path list if the mtime of objects/pack has changed ok op@


don't leak memory allocated internally by bloom filters; ok jamsek


remove trailing whitespace; patch by Josiah Frentsos