gitolite/doc/3-faq-tips-etc.mkd

29 KiB

assorted faqs, tips, and notes on gitolite

In this document:


common errors and mistakes

  • adding repositories/ at the start of the repo name in the git clone. This error is typically made by the admin himself -- because he knows what $REPO_BASE is set to and thinks he has to provide that prefix on the client side also :-) In fact gitolite prepends $REPO_BASE internally, so you shouldn't also do the same thing!

  • being able to clone but getting errors on push. Most likely caused by a combination of:

    • you already have shell access to the server, not just "gitolite" access, and

    • you cloned using git clone git@server:repositories/repo.git (notice there's an extra "repositories/" in there?)

    In other words, you used a key that completely bypassed gitolite and went straight to the shell to do the clone.

    Please see doc/ssh-troubleshooting.mkd for what all this means.

git version dependency

Gitolite (on the server) now refuses to run if git is not at least 1.6.2.

other errors, warnings, notes...

cloning an empty repo

Cloning an empty repo is only possible with clients greater than 1.6.2. So at least one of your clients needs to have a recent git. Once at least one commit has been made, older clients can also use it

When you clone an empty repo, git seems to complain about fatal: The remote end hung up unexpectedly. However, you can ignore this, since it doesn't seem to hurt anything. [Update 2009-09-14; this has been fixed in git 1.6.4.3]

@all syntax for repos

There is a way to use the @all syntax for repos also, as described in conf/example.conf. However, there are a couple of minor cautions:

  • don't use NAME/ or such restrictions on the special @all repo. Due to the potential for defeating a crucial optimisation and slowing down all access, we do not support this.
  • don't try giving @all users some permission for @all repos

umask setting

Gitweb not able to read your repos? You can change the umask for newly created repos to something more relaxed -- see the ~/.gitolite.rc file

getting a tar file from a clone

You can clone the repo from github or indefero, then execute a make command to extract a tar file of the branch you want. Please use the make command, not a plain "git archive", because the Makefile adds a file called .GITOLITE-VERSION that will help you identify which version you are using.

git clone git://github.com/sitaramc/gitolite.git
        # (OR)
git clone git://sitaramc.indefero.net/sitaramc/gitolite.git
cd gitolite
make master.tar
# or maybe "make pu.tar"

features

Apart from the big ones listed in the top level README, and subjective ones like "better config file format", gitolite has evolved to have many useful features than the original goal of branch-level access control.

syntax and normal usage

simpler syntax

The basic syntax is simpler and cleaner but it goes beyond that: you can specify access in bits and pieces, even if they overlap.

Some access needs are best grouped by repo, some by username, and some by both. So just do all of them, and gitolite will combine all the access lists! Here's an example:

# define groups of people
@bosses     = phb1 phb2 phb3
@devs       = dev1 dev2 dev3
@interns    = int1 int2 int3

# define groups of projects
@open       = git gitolite linux rakudo
@closed     = c1 c2 c3
@topsecret  = ts1 ts2 ts3

# all bosses have read access to all projects
repo @open @closed @topsecret
    R   =   @bosses

# everyone has read access to "open" projects
repo @open
    R   =   @bosses @devs @interns

[...or any other combination you want...]

# later in the file:

# specify access for individual repos (like RW, RW+, etc)
repo c1
    [...]

[...etc...]

If you notice that @bosses are given read access to @open via both rules, do not worry that this causes some duplication or inefficiency. It doesn't :-)

See the "specify gitweb/daemon access" section below for one more example.

one user, many keys

I have a laptop and a desktop I need to access the server from. I have different private keys on them, but as far as gitolite is concerned both of them should be treated as "sitaram". How does this work?

The way it works is that you copy one pubkey as "sitaram@laptop.pub" and the other as "sitaram@desktop.pub". The part before the "@" is the username, so gitolite knows these two keys belong to the same person. The part after the "@" can be anything you like, of course; gitolite doesn't care.

Note that you don't say "sitaram@laptop" and so on in the config file -- as far as the config file is concerned there's just one user called "sitaram" -- so you only say "sitaram" there.

I think this is easier to maintain if you have to delete or change one of those keys.

However, now that sitaramc@gmail.com is also a valid username, we need to distinguish between sitaramc@gmail.com.pub and sitaramc@desktop.pub. We do that by requiring that the multi-key suffix you use (like "desktop" and "laptop") should not have a "." in it. If it does, it looks like an email address. The following table lists sample pubkey filenames and the corresponding derived usernames (which is what goes into the conf/gitolite.conf file):

  • old style multikeys; not mistaken for emails because there is no "." in hostname part

    sitaramc.pub                            sitaramc
    sitaramc@laptop.pub                     sitaramc
    sitaramc@desktop.pub                    sitaramc
    
  • new style, email keys; there is a "." in hostname part; so it's an email address

    sitaramc@gmail.com.pub                  sitaramc@gmail.com
    
  • multikeys with email address

    sitaramc@gmail.com@laptop.pub           sitaramc@gmail.com
    sitaramc@gmail.com@desktop.pub          sitaramc@gmail.com
    

security, access control, and auditing

two levels of access rights checking

Gitolite has two levels of access checks. The first check is what I will call the pre-git level. At this stage, the gl-auth-command has been invoked by sshd, and it knows just three things:

  • who,
  • what repository, and
  • what type of access (R or W)

Note that at this point no git program has entered the picture, and we have no way of knowing what ref (branch, tag, etc) he is trying to update, even if it is a "write" operation.

For a "read" operation to pass this check, the username (or @all) must have read permission (i.e., R, RW, or RW+) on at least one branch of the repo.

For a "write" operation, there is an additional restriction: lines specifying only R (read access) don't count. The user must have write access to some ref in the repo in order to pass this stage!

The second check is via a git update hook. This check only happens for write operations. By this time we know what "ref" he is trying to update, as well as the old and the new SHAs of that ref (by which we can also deduce whether it's a rewind or not). This is where the "per-branch" permissions come into play.

Each refex that allows W access (or + if this is a rewind) for this user, on this repo, is matched against the actual refname being updated. If any of the refexes match, the push succeeds. If none of them match, it fails.

Gitolite also allows "exclude" or "deny" rules. See later in this document for details.

better logging

If you have been too liberal with the permission to rewind, it has built-in logging as an emergency fallback if someone goes too far, or for audit purposes [*]. The logfile names and location are configurable, and can include the year/month/day etc in the filename for easy archival or further processing. The log file even tells you which pattern in the config file matched to allow that specific access to proceed.

[*] setting core.logAllRefUpdates true does provide a safety net against over-zealous rewinds, but it does not tell you "who". And strangely, management does not seem to share the view that "blame" is just a synonym for "annotate" ;-)]

The log lines look like this:

2009-09-19.10:24:37  +  b4e76569659939  4fb16f2a88d8b5  myrepo refs/heads/master       user2   refs/heads/master

The "+" at the start indicates a non-fast forward update, in this case from b4e76569659939 to 4fb16f2a88d8b5. So b4e76569659939 is the one to restore! Can it get easier?

The other parts of the log line are the name of the repo, the refname being updated, the user updating it, and the refex pattern (from the config file) that matched, in case you need to debug the config file itself.

"exclude" (or "deny") rules

Here is an illustrative explanation of "deny" rules. However, please be sure to read the "DENY/EXCLUDE RULES" section in conf/example.conf for important notes/caveats before using "deny" rules.

Take a look at the following snippet, which seems to say that "bruce" can write versioned tags (anything containing refs/tags/v[0-9]), but the other staffers can't:

    @staff = bruce whitfield martin
            [... and later ...]
    RW refs/tags/v[0-9]     = bruce
    RW refs/tags            = @staff

But that's not how the matching works. As long as any refex matches the refname being updated, it's a "yes". Since the second refex (which says "anything containing refs/tags") is a superset of the first one, it lets anyone on @staff create versioned tags, not just Bruce.

One way to fix this is to allow "excludes" -- some changes in syntax, combined with a rigorous, ordered, interpretation would do it.

Let's recap the existing semantics:

the first matching refex that has the permission you're looking for (W or +), results in success. A fallthrough results in failure

Here are the new semantics, with changes from the "main" one in bold:

the first matching refex that has the permission you're looking for (W or +) or a minus (-), results in success or failure, respectively. A fallthrough also results in failure

So the example we started with becomes, if you use "deny" rules:

    RW refs/tags/v[0-9]     = bruce
    -  refs/tags/v[0-9]     = @staff
    RW refs/tags            = @staff

And here's how it works:

  • for non-version tags, only the 3rd rule matches, so anyone on staff can push them
  • for version tags by bruce, the first rule matches so he can push them
  • for version tags by staffers other than bruce, the second rule matches before the third one, and it has a - as the permission, so the push fails

separating delete and rewind rights

Since the beginning, RW+ meant being able to rewind or delete a ref. My stand is that these two are fairly similar, and infact a rewind is almost the same as a delete+push (the only difference I can see is if you had core.logAllRefUpdates set, which is not a default setting).

However, there seem to be cases where it is useful to distinguish them -- situations where one of them should be restricted more than the other. (Arguments exist for both sides: restrict delete more than rewind, and vice versa).

So we now allow these two rights to be separated. Here's how:

  • branch deletion is permitted by using RWD or RW+D -- essentially the current branch permissions with a D suffixed
  • if a repo has a rule containing such a D, all RW+ permissions (for that repo) cease to permit deletion of the ref matched.

This provides the greatest backward compatibility, while also enabling the new semantics at the granularity of a repo, instead of the entire config.

Note 1: if you find that RW+ no longer allows deletion but you can't see a D permission in the rules, remember that gitolite allows a repo config to be specified in multiple places for convenience, included delegated or included files. Be sure to search everywhere :)

Note 2: a quick way to make this the default for all your repos is:

repo @all
    RWD     dummy-branch    =   foo

where foo can be either the administrator, or if you can ignore the warning message when you push, a non-existant user.

Note 3: you can combine this with the "create a branch" permissions described in the next section, as the example line in conf/example.conf shows.

separating create and push rights

This feature is similar in spirit to the previous one, so please read that section for a general understanding.

Briefly:

  • branch creation is permitted by using RWC or RW+C -- essentially the current branch permissions with a C suffixed
  • if a repo has a rule containing such a C, then the RW and RW+ permissions (for that repo) no longer permit creation of the ref matched; they will only allow pushing to an existing ref

Note: you can combine this with the "delete a branch" permissions described in the previous section, as the example line in conf/example.conf shows.

file/dir NAME based restrictions

In addition to branch-name based restrictions, gitolite also allows you to restrict what files or directories can be involved in changes being pushed. This basically uses git diff --name-only to obtain the list of files being changed, treating each filename as a "ref" to be matched.

Please see conf/example.conf for syntax and examples.

delegating parts of the config file

You can now split up the config file and delegate the authority to specify access control for their own pieces. See delegation for details.

convenience features

what repos do I have access to?

Sometimes there are too many repos, maybe even named similarly, or with the potential for typos, confusion about hyphens/underscores or upper/lower case, etc. You'd just like a simple way to know what repos you have access to.

Gitolite provides two commands (info and expand) to help you find this information; please check doc/report-output.mkd for details.

including config lines from other files

See the entry under "INCLUDE SOME OTHER FILE" in conf/example.conf.

support for git installed outside default PATH

The normal solution is to add to the system default PATH somehow, either by munging /etc/profile or by enabling PermitUserEnvironment in /etc/ssh/sshd_config and then setting the PATH in ~/.ssh/.environment. All these are security risks because they allow a lot more than just you and your git install :-)

And if you don't have root, you can't do this anyway.

The only solution till now has been to ask every client to set the config parameters remote.<name>.receivepack and remote.<name>.uploadpack. But telling every client to do so is a pain...

Gitolite lets you specify the directory in which git binaries are to be found, via a new variable ($GIT_PATH) in the "rc" file. If this variable is non-empty, it will be appended to the PATH environment variable before attempting to run git stuff.

Very easy, very simple, and completely transparent to the users :-)

Note: sometimes you have a system that already has an older "git" installed in one of the system PATHs, but you've installed a newer git in some non-standard location and want that picked up. Because of security reasons, gitolite will not prepend GIT_PATH to the PATH variable, so the older git comes first and it gets kinda frustrating!

Here's a simple workaround. Ignore the GIT_PATH variable, and directly set the full PATH in the rc file, like so:

$ENV{PATH} = "/home/sitaram/bin:$ENV{PATH}";

"personal" branches

"personal" branches are great for corporate environments, where unauthenticated pull/clone is a no-no. Since a dev workstation cannot do authentication, even work shared just between 2 devs has to go via the server. This causes the same branch name clutter as in a centralised VCS, plus setting up permissions for this becomes a chore for the admin.

gitolite lets you define a "personal" or "scratch" namespace prefix for each developer (e.g., refs/personal/<devname>/*). Just add a line like:

    RW+ personal/USER/      =   @userlist

This means I (user "sitaram") can do anything to any branch whose name starts with personal/sitaram/ assuming I'm in "userlist".

You can have any number of such lines with different prefixes (for example, using topic names instead of "personal") or even suffixes if you like. The important thing is that the "branch" name should contain /USER/ (including the slashes). At runtime this will match whoever is the current user. Access is still determined by the right hand side of course.

custom hooks and custom git config

You can specify hooks that you want to propagate to all repos, as well as per-repo "gitconfig" settings. Please see doc/2-admin.mkd and conf/example.conf for details.

bypassing gitolite

Sometimes you'll need to access one of the gitolite-managed repos directly on the server, without going through gitolite. Reasons may be some automatic updates or some other ad hoc purposes you can dream up.

Cloning a gitolite-controlled repo is easy enough -- just use the full path (typically ~/repositories/reponame.git) instead of just reponame, to compensate for gitolite not sitting in between and adding those things to the repo path.

But when you push, the update hook (which git will invoke anyway) will fail because it needs all sorts of access control info that it now doesn't have, because the push was invoked without going through gitolite.

In order to bypass the update hook, just set the GL_BYPASS_UPDATE_HOOK environment variable to "1" or something, export it, and push. I prefer not to set that variable permanently, preferring this mode instead:

GL_BYPASS_UPDATE_HOOK=1 git push

INconvenience features

deleting a repo

By design, there is no code in gitolite to delete a repo if the repo was specified by name in the config file. (Wildcard repos can be deleted by the user; see here for details).

If you do want to permanently delete a non-wildcard repo, here's what you do:

  • remove the repo from the gitolite-admin repo clone's conf/gitolite.conf file. "add" the change, commit, and push.

  • then remove the repo from ~/repositories on the server (or whatever you set $GL_REPO_BASE to in the ~/.gitolite.rc)

helping with gitweb

Although gitweb is a completely separate program, gitolite can do quite a lot to help you manage gitweb access as well; once the initial setup is complete, you can do it all from within the gitolite config file!

easier to specify gitweb "description" and gitweb/daemon access

Please see gwd for details on how to do this if you've never done this before. This section is only about how gitolite makes it easy to specify different combinations of access for different sets of repos.

Remember gitolite lets you specify the access control specs in bits and pieces, so you can keep all the daemon/gitweb access in one place, even if each repo has more specific branch-level access config specified elsewhere. Here's an example, using really short reponames because I'm lazy:

# maybe near the top of the file, for ease of access:

@only_web       = r1 r2 r3
@only_daemon    = r4 r5 r6
@web_and_daemon = r7 r8 r9

repo @only_web
    R   = gitweb
repo @only_daemon
    R   = daemon
repo @web_and_daemon
    R   = gitweb
    R   = daemon

# ...maybe much later in the file:

repo r1
    # normal developer access lists for r1 and its branches/tags in the
    # usual way

repo r2
# ...and so on...

Over and above whether a repo is even shown by gitweb, you may want to further restrict people, allowing them to view only those repos for which they have been given read access by gitolite.

This requires that:

  • you have to have some sort of HTTP auth on your web server (out of my scope, sorry!)
  • the HTTP auth should use the same username (like "sitaram") as used in the gitolite config (for the corresponding user)

Normally a superuser sets up passwords for users using the "htpasswd" command, but this is an administrative chore.

Robin Smidsrød had the great idea that, since each user already has pubkey access to git@server, this gives us a very neat way of using gitolite to let the users manage their own HTTP passwords. Here's how:

  • setup apache so that the htaccess file it looks for is owned by the "git" user
  • in the ~/.gitolite.rc file, look for the variable $HTPASSWD_FILE and point it to this file
  • tell your users to type in ssh git@server htpasswd to set or change their HTTP passwords

Of course some other authentication method can be used (e.g. mod_ldap) as long as the usernames match.

Gitweb allows you to specify a subroutine to decide on access. We use that feature and tie it to gitolite. Configuration example can be found in contrib/gitweb/.

advanced features

repos named with wildcards

Please see doc/wildcard-repositories.mkd for all the details.

admin defined commands

This requires the wildcards feature to be enabled, but is then an extremely powerful feature. See doc/admin-defined-commands.mkd.

access control for external commands

Gitolite now has a mechanism for allowing access control for arbitrary external commands, as long as they are invoked via ssh and present a server-side command that contains enough information to make an access control decision.

Note that this is incompatible with giving people shell access as described in doc/ssh-troubleshooting.mkd -- people who have shell access are not subject to this mechanism (it wouldn't make sense to try and control someone who has shell access anyway).

In general, external commands require changes in one or both the config files; the sample files in conf/ double as documentation, so you should look there for examples and usage.

Commands implemented so far are:

  • rsync
  • svnserve (see next section for a brief description; this has been contributed by Simon and Vladimir)

svnserve

If you are transitioning from SVN to gitolite, and have a lot of users using public-key authentication with SVN, this feature may be useful to you. Once you migrate all users' public keys into gitolite, you can set the $SVNSERVE variable in ~/.gitolite.rc to tie svnserve with gitolite's authentication system. Assuming you installed gitolite to the same user as the one you used for SVN, SVN connectivity will be retained, and users will be able to use both SVN and git using the same SSH configuration.

design choices

keeping the parser and the access control separate

There are two programs concerned with access control:

  • gl-auth-command, the program that is run via ~/.ssh/authorized_keys; this decides whether git should even be allowed to run (basic R/W/no access). (This one cannot decide on the branch-level access; it is not known at this point what branch is being accessed)
  • the update-hook on each repo, which decides the per-branch permissions

I have chosen to keep the relatively complex task of parsing the config file out of them to keep them simpler (and faster). So any changes to the config have to be first "compiled", and the access control programs use this "compiled" version of the config. (The compile step also refreshes ~/.ssh/authorized_keys).

If you choose the "easy install" method, all this is quite transparent to you anyway. If you cannot use the easy install and must install manually, I have clear instructions on how to set it up.