Summary:
Ref T11469. This isn't directly related, but has been on my radar for a while: building SSH keyfiles (particular for installs with a lot of keys, like ours) can be fairly slow.
At least one cluster instance is making multiple clone requests per second. While that should probably be rate limited separately, caching this should mitigate the impact of these requests.
This is pretty straightforward to cache since it's exactly the same every time, and only changes when users modify SSH keys (which is rare).
Test Plan:
- Ran `bin/auth-ssh`, saw authfile generate.
- Ran it again, saw it read from cache.
- Changed an SSH key.
- Ran it again, saw it regenerate.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T11469
Differential Revision: https://secure.phabricator.com/D16744
Summary:
Fixes T11203. Subversion handling of `SVN_SSH` commands requires some additional finesse for nonstandard remote SSH ports.
We get `domain.com:port` in the command, so parse it out if it's present.
Test Plan: @enckse confirmed this locally in T11203.
Reviewers: chad
Reviewed By: chad
Subscribers: enckse
Maniphest Tasks: T11203
Differential Revision: https://secure.phabricator.com/D16172
Summary:
In D16167 I required users to be logged in to be "activated", but this logic doesn't account for that properly when the user is an Almanac device (a cluster host connecting to another cluster host).
Don't do this check for device connections, they can always establish sessions.
Test Plan: Will push.
Reviewers: chad
Reviewed By: chad
Differential Revision: https://secure.phabricator.com/D16170
Summary:
Ref T10547. This has been around for a while but I was never able to reproduce it. I caught a repro case in the cluster recently and I think this is the right fix.
We tell Subversion to run `ssh-connect` instead of `ssh` so we can provide options and credentials, by using `SVN_SSH` in the environment. Subversion will sometimes kill the SSH tunnel subprocess aggressively with SIGTERM -- as of writing, you can search for `SIGTERM` in `make_tunnel()` here:
http://svn.apache.org/repos/asf/subversion/trunk/subversion/libsvn_ra_svn/client.c
By default, when a PHP process gets SIGTERM it just exits immediately, without running destructors or shutdown functions. Since destructors/shutdown functions don't run, `TempFile` doesn't get a chance to remove the file.
I don't have a clear picture of //when// Subversion sends SIGTERM to the child process. I can't really get this to trigger locally via `svn`, although I was able to get it to trigger explicitly. So I'm only about 95% sure this fixes it, but it seems likely.
Test Plan:
Locally, I couldn't get this to reproduce "normally" even knowing the cause (maybe Subversion doesn't do the SIGTERM stuff on OSX?) but I was able to get it to reproduce reliabily by adding `posix_kill(getmypid(), SIGTERM);` to the body of the script.
With that added, running the script with `PHABRICATOR_CREDENTIAL=PHID-CDTL-...` in the environment reliably left straggler temporary files.
Adding `declare()` and a signal handler fixed this: the script now runs the `TempFile` destructor and longer leaves the stragglers around.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T10547
Differential Revision: https://secure.phabricator.com/D16102
Summary:
Ref T11098. This primarily fixes Conduit calls to `*.edit` methods failing when trying to access user preferences.
(The actual access is a little weird, since it seems like we're building some UI stuff inside a policy query, but that's an issue for another time.)
To fix this, consolidate the "we're about to run some kind of request with this user" code and run it consistently for web, conduit, and SSH sessions.
Additionally, make sure we swap things to the user's translation.
Test Plan:
- Ran `maniphest.edit` via `arc call-conduit`, no more settings exception.
- Set translation to ALL CAPS, got all caps output from `ssh` and Conduit.
Reviewers: avivey, chad
Reviewed By: chad
Maniphest Tasks: T11098
Differential Revision: https://secure.phabricator.com/D16066
Summary:
Ref T10917. Currently, when you delete an SSH key, we really truly delete it forever.
This isn't very consistent with other applications, but we built this stuff a long time ago before we were as rigorous about retaining data and making it auditable.
In partiular, destroying data isn't good for auditing after security issues, since it means we can't show you logs of any changes an attacker might have made to your keys.
To prepare to improve this, stop destoying data. This will allow later changes to become transaction-oriented and show normal transaction logs.
The tricky part here is that we have a `UNIQUE KEY` on the public key part of the key.
Instead, I changed this to `UNIQUE (key, isActive)`, where `isActive` is a nullable boolean column. This works because MySQL does not enforce "unique" if part of the key is `NULL`.
So you can't have two rows with `("A", 1)`, but you can have as many rows as you want with `("A", null)`. This lets us keep the "each key may only be active for one user/object" rule without requiring us to delete any data.
Test Plan:
- Ran schema changes.
- Viewed public keys.
- Tried to add a duplicate key, got rejected (already associated with another object).
- Deleted SSH key.
- Verified that the key was no longer actually deleted from the database, just marked inactive (in future changes, I'll update the UI to be more clear about this).
- Uploaded a new copy of the same public key, worked fine (no duplicate key rejection).
- Tried to upload yet another copy, got rejected.
- Generated a new keypair.
- Tried to upload a duplicate to an Almanac device, got rejected.
- Generated a new pair for a device.
- Trusted a device key.
- Untrusted a device key.
- "Deleted" a device key.
- Tried to trust a deleted device key, got "inactive" message.
- Ran `bin/ssh-auth`, got good output with unique keys.
- Ran `cat ~/.ssh/id_rsa.pub | ./bin/ssh-auth-key`, got good output with one key.
- Used `auth.querypublickeys` Conduit method to query keys, got good active keys.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T10917
Differential Revision: https://secure.phabricator.com/D15943
Summary:
Ref T4292. When you run `git fetch` and connect to, say, `repo001.west.company.com`, we'll look at the current version of the repository in other nodes in the cluster.
If `repo002.east.company.com` has a newer version of the repository, we'll fetch that version first, then respond to your request.
To do this, we need to run `git fetch repo002.east.company.com ...` and have that connect to the other host and be able to fetch data.
This change allows us to run `PHABRICATOR_AS_DEVICE=1 git fetch ...` to use device credentials to do this fetch. (Device credentials are already supported and used, they just always connect as a user right now, but these fetches should be doable without having a user. We will have a valid user when you run `git fetch` yourself, but we won't have one if the daemons notice that a repository is out of date and want to update it, so the update code should not depend on having a user.)
Test Plan:
```
$ PHABRICATOR_AS_DEVICE=1 ./bin/ssh-connect local.phacility.com
Warning: Permanently added 'local.phacility.com' (RSA) to the list of known hosts.
PTY allocation request failed on channel 0
phabricator-ssh-exec: Welcome to Phabricator.
You are logged in as device/daemon.phacility.net.
You haven't specified a command to run. This means you're requesting an interactive shell, but Phabricator does not provide an interactive shell over SSH.
Usually, you should run a command like `git clone` or `hg push` rather than connecting directly with SSH.
Supported commands are: conduit, git-lfs-authenticate, git-receive-pack, git-upload-pack, hg, svnserve.
Connection to local.phacility.com closed.
```
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T4292
Differential Revision: https://secure.phabricator.com/D15755
Summary: Fixes T10317. If we failed to load `$device`, it will be null, so we won't be able to call `getName()` on it.
Test Plan: `SSH_CLIENT='127.0.0.1' ./bin/ssh-exec --phabricator-ssh-device xyz` no longer fatals
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T10317
Differential Revision: https://secure.phabricator.com/D15235
Summary: Use `PhutilClassMaQuery` instead of `PhutilSymbolLoader`, mostly for consistency. Depends on D13588.
Test Plan: Poked around a bunch of pages.
Reviewers: #blessed_reviewers, epriestley
Reviewed By: #blessed_reviewers, epriestley
Subscribers: epriestley, Korvin
Differential Revision: https://secure.phabricator.com/D13589
Summary:
Ref T8387. Adds new mailing list users.
This doesn't migrate anything yet. I also need to update the "Email Addresses" panel to let administrators change the list address.
Test Plan:
- Created and edited a mailing list user.
- Viewed profile.
- Viewed People list.
- Searched for lists / nonlists.
- Grepped for all uses of `getIsDisabled()` / `getIsSystemAgent()` and added relevant corresponding behaviors.
- Hit the web/api/ssh session blocks.
Reviewers: btrahan
Reviewed By: btrahan
Subscribers: eadler, tycho.tatitscheff, epriestley
Maniphest Tasks: T8387
Differential Revision: https://secure.phabricator.com/D13123
Summary:
Ref T6941. In the cluster (and in other reasonable setups) we've separated SSH load balancers from HTTP load balancers.
In particular, ELBs will not let you load balance port 22, so this is likely a reasonable/common issue in larger clusters in AWS.
Allow users to specify an alternate host for SSH traffic.
Test Plan: Set host to someting different, saw it reflected in UI.
Reviewers: btrahan
Reviewed By: btrahan
Subscribers: epriestley
Maniphest Tasks: T6941
Differential Revision: https://secure.phabricator.com/D11800
Summary: Fixes T7034. Like HTTP, proxy requests to the correct host if a repository has an Almanac service host.
Test Plan: Ran VCS requests through the proxy.
Reviewers: btrahan
Reviewed By: btrahan
Subscribers: epriestley
Maniphest Tasks: T7034
Differential Revision: https://secure.phabricator.com/D11543
Summary: Ref T7034. This is a second special case, like commit hooks, where we need some help from Phabricator to make instance identity knowable.
Test Plan: Connected to an instance and ran SSH commands.
Reviewers: btrahan
Reviewed By: btrahan
Subscribers: epriestley
Maniphest Tasks: T7034
Differential Revision: https://secure.phabricator.com/D11539
Summary:
Ref T5833. This fixes a few weird things with this table:
- A bunch of columns were nullable for no reason.
- We stored an MD5 hash of the key (unusual) but never used it and callers were responsible for manually populating it.
- We didn't perform known-key-text lookups by using an index.
Test Plan:
- Ran migrations.
- Faked duplicate keys, saw them clean up correctly.
- Added new keys.
- Generated new keys.
- Used `bin/auth-ssh` and `bin/auth-ssh-key`.
Reviewers: btrahan
Reviewed By: btrahan
Subscribers: epriestley
Maniphest Tasks: T5833
Differential Revision: https://secure.phabricator.com/D10805
Summary:
Ref T5833. Currently, SSH keys are associated only with users, and are a bit un-modern. I want to let Almanac Devices have SSH keys so devices in a cluster can identify to one another.
For example, with hosted installs, initialization will go something like this:
- A request comes in for `company.phacility.com`.
- A SiteSource (from D10787) makes a Conduit call to Almanac on the master install to check if `company` is a valid install and pull config if it is.
- This call can be signed with an SSH key which identifies a trusted Almanac Device.
In the cluster case, a web host can make an authenticated call to a repository host with similar key signing.
To move toward this, put a proper Query class on top of SSH key access (this diff). In following diffs, I'll:
- Rename `userPHID` to `objectPHID`.
- Move this to the `auth` database.
- Provide UI for device/key association.
An alternative approach would be to build some kind of special token layer in Conduit, but I think that would be a lot harder to manage in the hosting case. This gives us a more direct attack on trusting requests from machines and recognizing machines as first (well, sort of second-class) actors without needing things like fake user accounts.
Test Plan:
- Added and removed SSH keys.
- Added and removed SSH keys from a bot account.
- Tried to edit an unonwned SSH key (denied).
- Ran `bin/ssh-auth`, got sensible output.
- Ran `bin/ssh-auth-key`, got sensible output.
Reviewers: btrahan
Reviewed By: btrahan
Subscribers: epriestley
Maniphest Tasks: T5833
Differential Revision: https://secure.phabricator.com/D10790
Summary: Fixes T6006. These didn't get caught by D10392.
Test Plan: Forced migration to re-run; ran SSH commands against Phabricator.
Auditors: btrahan
Summary: Ran `arc lint --apply-patches --everything` over rP, mainly to change double quotes to single quotes where appropriate. These changes also validate that the `ArcanistXHPASTLinter::LINT_DOUBLE_QUOTE` rule is working as expected.
Test Plan: Eyeballed it.
Reviewers: #blessed_reviewers, epriestley
Reviewed By: #blessed_reviewers, epriestley
Subscribers: epriestley, Korvin, hach-que
Differential Revision: https://secure.phabricator.com/D9431
Summary: We currently print a fairly vague, technical message which is ambiguous about indicating success or an error if you aren't familiar with SSH.
Test Plan:
$ ssh -T dweller@localhost
phabricator-ssh-exec: Welcome to Phabricator.
You are logged in as epriestley.
You haven't specified a command to run. This means you're requesting an interactive shell, but Phabricator does not provide an interactive shell over SSH.
Usually, you should run a command like `git clone` or `hg push` rather than connecting directly with SSH.
Supported commands are: conduit, git-receive-pack, git-upload-pack, hg, svnserve.
Reviewers: btrahan, dctrwatson
Reviewed By: dctrwatson
CC: aran
Differential Revision: https://secure.phabricator.com/D7854
Summary:
The documentation is explicit that Git does not pass this flag:
The $GIT_SSH command will be given exactly two arguments: the username@host (or just host) from the URL and the shell command to execute on that remote system.
This isn't true; it does. Accommodate it.
I'll see if I can fix this in the upstream, too.
Test Plan: Ran various `ssh-connect` commands with -p, etc.
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Differential Revision: https://secure.phabricator.com/D7680
Summary: If no ssh keys are in phabricator, bin/ssh-auth errors with undefined `$lines`. This fixes that case and explicitly tells the user no rows were found.
Test Plan: Ran bin/ssh-auth before and after change with no ssh keys in the system. Error goes away after change.
Reviewers: #blessed_reviewers, epriestley
Reviewed By: epriestley
CC: epriestley, aran, Korvin
Differential Revision: https://secure.phabricator.com/D7675
Summary: Fixes T2230. This isn't a total walk in the park to configure, but should work for early adopters now.
Test Plan: Read documentation, browsed UI.
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Maniphest Tasks: T2230
Differential Revision: https://secure.phabricator.com/D7634
Summary: Fixes T4122. Ref T2230. Instead of storing credentials on each repository, store them in Passphrase. This allows easy creation/management of many repositories which share credentials.
Test Plan:
- Upgraded repositories.
- Created and edited repositories.
- Pulled HTTP and SSH repositories.
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Maniphest Tasks: T2230, T4122
Differential Revision: https://secure.phabricator.com/D7629
Summary:
This cleans up some garbage:
- We were specifying environmental variables with `X=y git ...`, but now have `setEnv()` on both `ExecFuture` and `PhutilExecPassthru`. Use `setEnv()`.
- We were specifying the working directory with `(cd %s && git ...)`, but now have `setCWD()` on both `ExecFuture` and `PhutilExecPassthru`. Use `setCWD()`.
- We were specifying the Git credentials with `ssh-agent -c (ssh-add ... && git ...)`. We can do this more cleanly with `GIT_SSH`. Use `GIT_SSH`.
- Since we have to write a script for `GIT_SSH` anyway, use the same script for Subversion and Mercurial.
This fixes two specific issues:
- Previously, we were not able to set `-o StrictHostKeyChecking=no` on Git commands, so the first time you cloned a git repo the daemons would generally prompt you to add `github.com` or whatever to `known_hosts`. Since this was non-interactive, things would mysteriously hang, in effect. With `GIT_SSH`, we can specify the flag, reducing the number of ways things can go wrong.
- This adds `LANG=C`, which probably (?) forces the language to English for all commands. Apparently you need to install special language packs or something, so I don't know that this actually works, but at least two users with non-English languages have claimed it does (see <https://github.com/facebook/arcanist/pull/114> for a similar issue in Arcanist).
At some point in the future I might want to combine the Arcanist code for command execution with the Phabricator code for command execution (they share some stuff like LANG and HGPLAIN). However, credential management is kind of messy, so I'm adopting a "wait and see" approach for now. I expect to split this at least somewhat in the future, for Drydock/Automerge if nothing else.
Also I'm not sure if we use the passthru stuff at all anymore, I may just be able to delete that. I'll check in a future diff.
Test Plan: Browsed and pulled Git, Subversion and Mercurial repositories.
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Maniphest Tasks: T2230
Differential Revision: https://secure.phabricator.com/D7600
Summary:
Small step forward which improves existing stuff or lays groudwork for future stuff:
- Currently, to check for email verification, we have to single-query the email address on every page. Instead, denoramlize it into the user object.
- Migrate all the existing users.
- When the user verifies an email, mark them as `isEmailVerified` if the email is their primary email.
- Just make the checks look at the `isEmailVerified` field.
- Add a new check, `isUserActivated()`, to cover email-verified plus disabled. Currently, a non-verified-but-not-disabled user could theoretically use Conduit over SSH, if anyone deployed it. Tighten that up.
- Add an `isApproved` flag, which is always true for now. In a future diff, I want to add a default-on admin approval queue for new accounts, to prevent configuration mistakes. The way it will work is:
- When the queue is enabled, registering users are created with `isApproved = false`.
- Admins are sent an email, "[Phabricator] New User Approval (alincoln)", telling them that a new user is waiting for approval.
- They go to the web UI and approve the user.
- Manually-created accounts are auto-approved.
- The email will have instructions for disabling the queue.
I think this queue will be helpful for new installs and give them peace of mind, and when you go to disable it we have a better opportunity to warn you about exactly what that means.
Generally, I want to improve the default safety of registration, since if you just blindly coast through the path of least resistance right now your install ends up pretty open, and realistically few installs are on VPNs.
Test Plan:
- Ran migration, verified `isEmailVerified` populated correctly.
- Created a new user, checked DB for verified (not verified).
- Verified, checked DB (now verified).
- Used Conduit, People, Diffusion.
Reviewers: btrahan
Reviewed By: btrahan
CC: chad, aran
Differential Revision: https://secure.phabricator.com/D7572
Summary:
Ref T2230. The SVN protocol has a sensible protocol format with a good spec here:
http://svn.apache.org/repos/asf/subversion/trunk/subversion/libsvn_ra_svn/protocol
Particularly, compare this statement to the clown show that is the Mercurial wire protocol:
> It is possible to parse an item without knowing its type in advance.
WHAT A REASONABLE STATEMENT TO BE ABLE TO MAKE ABOUT A WIRE PROTOCOL
Although it makes substantially more sense than Mercurial, it's much heavier-weight than the Git or Mercurial protocols, since it isn't distributed.
It's also not possible to figure out if a request is a write request (or even which repository it is against) without proxying some of the protocol frames. Finally, several protocol commands embed repository URLs, and we need to reach into the protocol and translate them.
Test Plan: Ran various SVN commands over SSH (`svn log`, `svn up`, `svn commit`, etc).
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Maniphest Tasks: T2230
Differential Revision: https://secure.phabricator.com/D7556
Summary:
Ref T2230. This is substantially more complicated than Git, but mostly because Mercurial's protocol is a like 50 ad-hoc extensions cobbled together. Because we must decode protocol frames in order to determine if a request is read or write, 90% of this is implementing a stream parser for the protocol.
Mercurial's own parser is simpler, but relies on blocking reads. Since we don't even have methods for blocking reads right now and keeping the whole thing non-blocking is conceptually better, I made the parser nonblocking. It ends up being a lot of stuff. I made an effort to cover it reasonably well with unit tests, and to make sure we fail closed (i.e., reject requests) if there are any parts of the protocol I got wrong.
A lot of the complexity is sharable with the HTTP stuff, so it ends up being not-so-bad, just very hard to verify by inspection as clearly correct.
Test Plan:
- Ran `hg clone` over SSH.
- Ran `hg fetch` over SSH.
- Ran `hg push` over SSH, to a read-only repo (error) and a read-write repo (success).
Reviewers: btrahan, asherkin
Reviewed By: btrahan
CC: aran
Maniphest Tasks: T2230
Differential Revision: https://secure.phabricator.com/D7553
Summary: Looks like this is pretty straightforward; same as the reads except mark it as needing PUSH.
Test Plan: Ran `git push`, pushed over SSH to a hosted repo.
Reviewers: btrahan
Reviewed By: btrahan
CC: hach-que, aran
Maniphest Tasks: T2230
Differential Revision: https://secure.phabricator.com/D7425
Summary:
- Add web UI for configuring SSH hosting.
- Route git reads (`git-upload-pack` over SSH).
Test Plan:
>>> orbital ~ $ git clone ssh://127.0.0.1/
Cloning into '127.0.0.1'...
Exception: Unrecognized repository path "/". Expected a path like "/diffusion/X/".
fatal: Could not read from remote repository.
Please make sure you have the correct access rights
and the repository exists.
>>> orbital ~ $ git clone ssh://127.0.0.1/diffusion/X/
Cloning into 'X'...
Exception: No repository "X" exists!
fatal: Could not read from remote repository.
Please make sure you have the correct access rights
and the repository exists.
>>> orbital ~ $ git clone ssh://127.0.0.1/diffusion/MT/
Cloning into 'MT'...
Exception: This repository is not available over SSH.
fatal: Could not read from remote repository.
Please make sure you have the correct access rights
and the repository exists.
>>> orbital ~ $ git clone ssh://127.0.0.1/diffusion/P/
Cloning into 'P'...
Exception: TODO: Implement serve over SSH.
fatal: Could not read from remote repository.
Please make sure you have the correct access rights
and the repository exists.
Reviewers: btrahan
Reviewed By: btrahan
CC: hach-que, aran
Maniphest Tasks: T2230
Differential Revision: https://secure.phabricator.com/D7421
Summary:
Fixes T2229. This sets the stage for a patch similar to D7417, but for SSH. In particular, SSH 6.2 introduced an `AuthorizedKeysCommand` directive, which lets us do this in a mostly-reasonable way without needing users to patch sshd (if they have a recent enough version, at least).
The way the `AuthorizedKeysCommand` works is that it gets run and produces an `authorized_keys`-style file fragment. This isn't ideal, because we have to dump every key into the result, but should be fine for most installs. The earlier patch against `sshd` passes the public key itself, which allows the script to just look up the key. We might use this eventually, since it can scale much better, so I haven't removed it.
Generally, auth is split into two scripts now which mostly do the same thing:
- `ssh-auth` is the AuthorizedKeysCommand auth, which takes nothing and dumps the whole keyfile.
- `ssh-auth-key` is the slightly cleaner and more scalable (but patch-dependent) version, which takes the public key and dumps only matching options.
I also reworked the argument parsing to be a bit more sane.
Test Plan:
This is somewhat-intentionally a bit obtuse since I don't really want anyone using it yet, but basically:
- Copy `phabricator-ssh-hook.sh` to somewhere like `/usr/libexec/openssh/`, chown it `root` and chmod it `500`.
- This script should probably also do a username check in the future.
- Create a copy of `sshd_config` and fix the paths/etc. Point the KeyScript at your copy of the hook.
- Start a copy of sshd (6.2 or newer) with `-f <your config file>` and maybe `-d -d -d` to foreground and debug.
- Run `ssh -p 2222 localhost` or similar.
Specifically, I did this setup and then ran a bunch of commands like:
- `ssh host` (denied, no command)
- `ssh host ls` (denied, not supported)
- `echo '{}' | ssh host conduit conduit.ping` (works)
Reviewers: btrahan
Reviewed By: btrahan
CC: hach-que, aran
Maniphest Tasks: T2229, T2230
Differential Revision: https://secure.phabricator.com/D7419
Summary:
- Original command is in SSH_ORIGINAL_COMMAND, not normal argv.
- Use PhutilShellLexer to parse it.
- Fix a protocol encoding issue with ConduitSSHWorkflow. I think I'm going to make this protocol accept multiple commands anyway because SSH pipes are crazy expensive to build (even locally, they're ~300ms).
Test Plan: With other changes, successfully executed "arc list --conduit-uri=ssh://localhost:2222".
Reviewers: btrahan, vrana
Reviewed By: btrahan
CC: aran
Maniphest Tasks: T550
Differential Revision: https://secure.phabricator.com/D4232
Summary:
- Build "sshd-auth" (for authentication) and "sshd-exec" (for command execution) binaries. These are callable by "sshd-vcs", located [[https://github.com/epriestley/sshd-vcs | in my account on GitHub]]. They are based on precursors [[https://github.com/epriestley/sshd-vcs-glue | here on GitHub]] which I deployed for TenXer about a year ago, so I have some confidence they at least basically work.
- The problem this solves is that normally every user would need an account on a machine to connect to it, and/or their public keys would all need to be listed in `~/.authorized_keys`. This is a big pain in most installs. Software like Gitosis/Gitolite solve this problem by giving you an easy way to add public keys to `~/.authorized_keys`, but this is pretty gross.
- Roughly, instead of looking in `~/.authorized_keys` when a user connects, the patched sshd instead runs `echo <public key> | sshd-auth`. The `sshd-auth` script looks up the public key and authorizes the matching user, if they exist. It also forces sshd to run `sshd-exec` instead of a normal shell.
- `sshd-exec` receives the authenticated user and any command which was passed to ssh (like `git receive-pack`) and can route them appropriately.
- Overall, this permits a single account to be set up on a server which all Phabricator users can connect to without any extra work, and which can safely execute commands and apply appropriate permissions, and disable users when they are disabled in Phabricator and all that stuff.
- Build out "sshd-exec" to do more thorough checks and setup, and delegate command execution to Workflows (they now exist, and did not when I originally built this stuff).
- Convert @btrahan's conduit API script into a workflow and slightly simplify it (ConduitCall did not exist at the time it was written).
The next steps here on the Repository side are to implement Workflows for Git, SVN and HG wire protocols. These will mostly just proxy the protocols, but also need to enforce permissions. So the approach will basically be:
- Implement workflows for stuff like `git receive-pack`.
- These workflows will implement enough of the underlying protocol to determine what resource the user is trying to access, and whether they want to read or write it.
- They'll then do a permissons check, and kick the user out if they don't have permission to do whatever they are trying to do.
- If the user does have permission, we just proxy the rest of the transaction.
Next steps on the Conduit side are more simple:
- Make ConduitClient understand "ssh://" URLs.
Test Plan: Ran `sshd-exec --phabricator-ssh-user epriestley conduit differential.query`, etc. This will get a more comprehensive test once I set up sshd-vcs.
Reviewers: btrahan, vrana
Reviewed By: btrahan
CC: aran
Maniphest Tasks: T603, T550
Differential Revision: https://secure.phabricator.com/D4229
