Issue #7462
closedauth token used in registry requests are not validated properly when behind an SSL proxy
Description
When the content app is running over HTTP deployed behind a reverse proxy running on HTTPS, the content app does not properly detect the url that the request is coming in as and fails to validate the provided auth token.
As an example, my CONTENT_ORIGIN is set to "https://hostname.example.com", a fetch to the api, would generate an auth token based on this URL:
https://hostname.example.com/pulp/container/default_organization-foo-foobar/manifests/latest
however, when the request is redirected to that URL, and proxied back to the content app, the content app tries to get the URL that is coming in and gets a url that looks like:
It seems to think that the request is coming in over 'http' since the content app is running over http. In addition, it seems to be ignoring the port #, so if CONTENT_ORIGIN was configured to a non-standard port, comparison would also fail (although this is not a concern for us at this time).
I think it makes sense to just create and validate the token based off the request path, and ignore hostname, protocol, and port
Some relevant links showing why/where this is a problem: https://docs.aiohttp.org/en/stable/web_advanced.html#deploying-behind-a-proxy https://github.com/aio-libs/aiohttp/blob/72b5344af59ccdefc1e9a9e489c222a2b2592c7d/aiohttp/web_request.py#L384-L386 https://github.com/pulp/pulp_container/blob/789979a8706c527e4c9e8d04d084b262ea78208f/pulp_container/app/models.py#L442
Updated by dkliban@redhat.com about 4 years ago
I was able to reproduce on my machine. I had to ensure that the 'CONTENT_ORIGIN' setting was set to https://localhost
Updated by ekohl about 4 years ago
A common way with reverse proxy setups is https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/X-Forwarded-Proto. Looks like Django has docs on this and https://docs.djangoproject.com/en/2.2/ref/settings/#secure-proxy-ssl-header suggests to use this setting:
SECURE_PROXY_SSL_HEADER = ('HTTP_X_FORWARDED_PROTO', 'https')
Updated by mdellweg about 4 years ago
jsherril@redhat.com wrote:
[...] I think it makes sense to just create and validate the token based off the request path, and ignore hostname, protocol, and port [...]
I agree. This would be compatible with whatever strange url rewrite rules a user might apply on his reverse proxy (well almost, but we need to assume something for the hash).
Updated by pulpbot about 4 years ago
- Status changed from NEW to POST
Added by dkliban@redhat.com about 4 years ago
Added by dkliban@redhat.com about 4 years ago
Revision b6d54914 | View on GitHub
Use the path and query string of the URL to generate a 'validate_token'.
Updated by dkliban@redhat.com about 4 years ago
- Assignee set to dkliban@redhat.com
Updated by dkliban@redhat.com about 4 years ago
- Status changed from POST to MODIFIED
Applied in changeset b6d5491405f33df4da46d07143473b3005f680a7.
Updated by pulpbot about 4 years ago
Added by dkliban@redhat.com about 4 years ago
Revision e6b9e932 | View on GitHub
Use the path and query string of the URL to generate a 'validate_token'.
Added by dkliban@redhat.com about 4 years ago
Revision e6b9e932 | View on GitHub
Use the path and query string of the URL to generate a 'validate_token'.
Updated by pulpbot about 4 years ago
- Status changed from MODIFIED to CLOSED - CURRENTRELEASE
Use the path and query string of the URL to generate a 'validate_token'.
fixes: #7462 https://pulp.plan.io/issues/7462