Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Introduce user-defined wrapdb mirror #13953

Open
wants to merge 5 commits into
base: master
Choose a base branch
from

Conversation

klokik
Copy link
Contributor

@klokik klokik commented Nov 24, 2024

This consolidates all queries to the wrapdb to go through the open_wrapdburl() function. The function handles a domain-specific URL scheme, wrapdb. When encountered, it substitutes the scheme with https and an authority(net loc) with either the upstream wrapdb address or a user-defined one stored in the subprojects/wrap-mirrors.json file.

The user should use meson wrap set-db-address <mirror-address> command to create the file. It looks a bit more secure to me than using an environment variable.

Justification:

  • wrapdb servers may not always be available
  • internal organization policies prohibit using external sources during the build process
  • wraps for non-public proprietary subprojects

This consolidates all queries to the wrapdb to go through the open_wrapdburl()
function. The function handles a domain-specific url scheme
wrapdb. When encountered, it substitutes the scheme with https and
an authority(netloc) with either upstream wrapdb address or
a user-defined one from the MESON_WRAPDB_MIRROR environment variable.
@jpakkane
Copy link
Member

user-defined one from the MESON_WRAPDB_MIRROR environment variable.

This has the potential to be a really nasty way of hijacking user queries for malware injection. Not to mention all the other ways in which environment variables are terrible for storing persistent configuration.

@klokik
Copy link
Contributor Author

klokik commented Nov 25, 2024

@jpakkane you mean adding more malware, once attacker already gained control over the users shell environment? Or getting the first handful of malware by user getting source code from the untrustworthy mirror?

I'd be glad to hear for alternative more secure options. Providing the mirror address as a meson wrap command argument on each invocation would be acceptable for me, but less convenient in day-to-day use.
Other than that meson does not seem to use any persistent configuration. This does not belong to the machine files either.
How about a meson wrap set-mirror (address) that will be stored in the subprojects dir, along the local version of db?

Getting code from untrusted sources IMO is a user problem. I honestly don't expect anyone to use random third party mirror found on the search engine.

@klokik
Copy link
Contributor Author

klokik commented Nov 25, 2024

Ok, I can see this being abused on a build machine, as a stepping ladder of a more complex attack.What's worse - it will affect users that are not aware of this variable.

@klokik klokik force-pushed the wrapdb-mirror branch 3 times, most recently from 8e6b1d6 to babf222 Compare November 25, 2024 22:11
Environment variables are prone to shell injection attacks, especially
on build servers that many people may have access to. Changing address in
a file requires an explicit command that is supposedly under the control
of a trusted developer.
@klokik
Copy link
Contributor Author

klokik commented Nov 25, 2024

@jpakkane, I changed it to use an explicit user-provided address, unaffected by environment variables, limited to a single workspace.

It looks tempting to store the address inside subprojects/wrap.json along with the revisions it provides, but this will require breaking file layout and annoyance for the end-users.

```

You will be limited to the wraps available on the mirror, only one source is used at a time.
The address is stored in `subprojects/wrapdb-mirrors.json`, remove the file to use upstream address.
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I am not sure I like the UI for this feature. I think that the use case you have in mind for it is very different from what I have in mind. Why isn't a command line argument to meson wrap enough? If this is a functionality intended mostly for corporate environment where access to the "upstream" wrap database is somehow forbidden, doesn't a network level redirect provide a transparent solution for the same problem?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Command-line argument will have to be set for each of the update and install sub-commands, and transparent fallback can not happen unless the address is stored somewhere.

May I propose an optional positional argument to the update-db command that will create the "persistent" file?

Actually, you caught my mental confusion on what what it is for. I'm calling it a mirror, but it is not what I want it for - I want a different wrap-db-source.

This is indeed for a corporate environment, but a simple redirect is not acceptable. All sources should be stored on the company-controlled machines, wraps and source updates have to be audited to an extent. Wraps for the subprojects developed by other teams, that have no interest in using meson yet are to be served from this source as well.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is indeed for a corporate environment, but a simple redirect is not acceptable. All sources should be stored on the company-controlled machines, wraps and source updates have to be audited to an extent. Wraps for the subprojects developed by other teams, that have no interest in using meson yet are to be served from this source as well.

I don't understand why a redirect is not acceptable. If that is your policy, most likely you have measures in place that forbid access to https://wrapdb.mesonbuild.com. You just need to change those measures to redirect access to that resource to https://wrapdb.localnet or whatever you like. Then you just serve your own wrap-db from that URL. Your wrap-db may contain your hand-picked selection of wraps. Why would this not work?

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

May I propose an optional positional argument to the update-db command that will create the "persistent" file?

AFAIK running meson wrap update-db is not required to use other meson wrap commands. Therefore, making the configuration of the wrap database URL dependent on running it seems at least counterintuitive.

One thing I haven't understood yet is whether you want a setting that is per system, per user, per project, or per specific wrap. Depending on what is the use case the configuration would need to be stored in a different location.

I am also not convinced that there is the need of a meson command to set the WrapDB URL instead that a simple configuration file format.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You just need to change those measures to redirect access to that resource to https://wrapdb.localnet or whatever you like. Then you just serve your own wrap-db from that URL. Your wrap-db may contain your hand-picked selection of wraps. Why would this not work?

  1. It will break TLS
  2. Let's say that policy is only enforced on the CI machines
  3. I have no control over this redirect
  4. meson is overly attached to this specific address hardcoded in a few places

Copy link
Contributor Author

@klokik klokik Nov 27, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

One thing I haven't understood yet is whether you want a setting that is per system, per user, per project, or per specific wrap. Depending on what is the use case the configuration would need to be stored in a different location.

Ideally setting wrapdb source per project with all of its subprojects. + A single instance of wrap server on the network.

AFAIK running meson wrap update-db is not required to use other meson wrap commands. Therefore, making the configuration of the wrap database URL dependent on running it seems at least counterintuitive.

Certainly, it will not require update-db with the changes proposed in this MR. Only commands that query wrapdb need the update command whether with the source set or not. I don't think that one is supposed to have wrapdb.json file checked-in, and without it status/info/search/list won't work. I see what you mean now: other commands do query release data directly from the server bypassing local wrapdb.json file. I don't like it.

Copy link
Contributor Author

@klokik klokik Nov 28, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

May I propose an optional positional argument to the update-db command that will create the "persistent" file?

I did it, but I don't like it. Will revert to a separate wrap subcommand.

If for whatever reason you want to use self-hosted or proxied Wrap database *(since 1.X.X)* you may set mirror address to be used for the workspace:

```console
$ meson wrap set-db-mirror user:[email protected]:8080
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I really don't like that this does not allow to set a protocol. For now the wrap database speaks only https, but I can imagine a future where a different protocol is implemented.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Like ftp, sftp, git? This will require parsing urls differently. An ssh/git user might want to pass a key somehow. Doable, but not sure if I want to open this can of worms yet.

Having a file: or plain http: protocol will be convenient for local testing, but not really useful otherwise.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't understand what you mean with "opining this can of worms". You are making the location of the wrap database configurable. What I am asking is to do not assume that the protocol is https and make the protocol part of the URL part of the configuration string instead of hard-coding it to be https. Where are the worms in this request?

Copy link
Contributor Author

@klokik klokik Nov 28, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Wish granted. There is enforcement of https in the code with the assumption that the only other alternative is http. I didn't get it right yet. WIP.

This changes moves responsibility of setting wrapdb source to the
`wrap update-db` command. Improve wording.
Extend source address to accept other protocols. urllib can work with
http(s), file, ftp protocols, so this sould work, but type annotations
are not correct for the latter two. And due to enforcement of the https for
secure connections protocol selection options are still limited.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants