# Which Problems Are Solved
We noticed logging where 500: Internal Server errors were returned from
the token endpoint, mostly for the `refresh_token` grant. The error was
thrown by the database as it received non-UTF8 strings for token IDs
Zitadel uses symmetric encryption for opaque tokens, including refresh
tokens. Encrypted values are base64 encoded. It appeared to be possible
to send garbage base64 to the token endpoint, which will pass decryption
and string-splitting. In those cases the resulting ID is not a valid
UTF-8 string.
Invalid non-UTF8 strings are now rejected during token decryption.
# How the Problems Are Solved
- `AESCrypto.DecryptString()` checks if the decrypted bytes only contain
valid UTF-8 characters before converting them into a string.
- `AESCrypto.Decrypt()` is unmodified and still allows decryption on
non-UTF8 byte strings.
- `FromRefreshToken` now uses `DecryptString` instead of `Decrypt`
# Additional Changes
- Unit tests added for `FromRefreshToken` and
`AESCrypto.DecryptString()`.
- Fuzz tests added for `FromRefreshToken` and
`AESCrypto.DecryptString()`. This was to pinpoint the problem
- Testdata with values that resulted in invalid strings are committed.
In the pipeline this results in the Fuzz tests to execute as regular
unit-test cases. As we don't use the `-fuzz` flag in the pipeline no
further fuzzing is performed.
# Additional Context
- Closes#7765
- https://go.dev/doc/tutorial/fuzz
# Which Problems Are Solved
During triggering of the fields table WriteTooOld errors can occure when
using cockroachdb.
# How the Problems Are Solved
The statements exclusively lock the projection before they start to
insert data by using `FOR UPDATE`.
# Which Problems Are Solved
The text appears to contradict the statement in the page on oauth.net.
# How the Problems Are Solved
The text has been updated to reflect the statement in the page on
oauth.net.
# Additional Changes
None
# Additional Context
The page [OAUTH2.0 for mobile and native
apps](https://oauth.net/2/native-apps/) linked just above the text that
was changed states:
> It describes things like not allowing the third-party application to
open an embedded web view which is more susceptible to phishing attacks,
as well as platform-specific recommendations on how to do so.
Co-authored-by: Max Peintner <max@caos.ch>
# Which Problems Are Solved
We noticed logging where 500: Internal Server errors were returned from
the token endpoint, mostly for the `refresh_token` grant. The error was
thrown by the database as it received non-UTF8 strings for token IDs
Zitadel uses symmetric encryption for opaque tokens, including refresh
tokens. Encrypted values are base64 encoded. It appeared to be possible
to send garbage base64 to the token endpoint, which will pass decryption
and string-splitting. In those cases the resulting ID is not a valid
UTF-8 string.
Invalid non-UTF8 strings are now rejected during token decryption.
# How the Problems Are Solved
- `AESCrypto.DecryptString()` checks if the decrypted bytes only contain
valid UTF-8 characters before converting them into a string.
- `AESCrypto.Decrypt()` is unmodified and still allows decryption on
non-UTF8 byte strings.
- `FromRefreshToken` now uses `DecryptString` instead of `Decrypt`
# Additional Changes
- Unit tests added for `FromRefreshToken` and
`AESCrypto.DecryptString()`.
- Fuzz tests added for `FromRefreshToken` and
`AESCrypto.DecryptString()`. This was to pinpoint the problem
- Testdata with values that resulted in invalid strings are committed.
In the pipeline this results in the Fuzz tests to execute as regular
unit-test cases. As we don't use the `-fuzz` flag in the pipeline no
further fuzzing is performed.
# Additional Context
- Closes#7765
- https://go.dev/doc/tutorial/fuzz
# Which Problems Are Solved
ZITADEL currently selects the instance context based on a HTTP header
(see https://github.com/zitadel/zitadel/issues/8279#issue-2399959845 and
checks it against the list of instance domains. Let's call it instance
or API domain.
For any context based URL (e.g. OAuth, OIDC, SAML endpoints, links in
emails, ...) the requested domain (instance domain) will be used. Let's
call it the public domain.
In cases of proxied setups, all exposed domains (public domains) require
the domain to be managed as instance domain.
This can either be done using the "ExternalDomain" in the runtime config
or via system API, which requires a validation through CustomerPortal on
zitadel.cloud.
# How the Problems Are Solved
- Two new headers / header list are added:
- `InstanceHostHeaders`: an ordered list (first sent wins), which will
be used to match the instance.
(For backward compatibility: the `HTTP1HostHeader`, `HTTP2HostHeader`
and `forwarded`, `x-forwarded-for`, `x-forwarded-host` are checked
afterwards as well)
- `PublicHostHeaders`: an ordered list (first sent wins), which will be
used as public host / domain. This will be checked against a list of
trusted domains on the instance.
- The middleware intercepts all requests to the API and passes a
`DomainCtx` object with the hosts and protocol into the context
(previously only a computed `origin` was passed)
- HTTP / GRPC server do not longer try to match the headers to instances
themself, but use the passed `http.DomainContext` in their interceptors.
- The `RequestedHost` and `RequestedDomain` from authz.Instance are
removed in favor of the `http.DomainContext`
- When authenticating to or signing out from Console UI, the current
`http.DomainContext(ctx).Origin` (already checked by instance
interceptor for validity) is used to compute and dynamically add a
`redirect_uri` and `post_logout_redirect_uri`.
- Gateway passes all configured host headers (previously only did
`x-zitadel-*`)
- Admin API allows to manage trusted domain
# Additional Changes
None
# Additional Context
- part of #8279
- open topics:
- "single-instance" mode
- Console UI
# Which Problems Are Solved
The current v3alpha actions APIs don't exactly adhere to the [new
resources API
design](https://zitadel.com/docs/apis/v3#standard-resources).
# How the Problems Are Solved
- **Breaking**: The current v3alpha actions APIs are removed. This is
breaking.
- **Resource Namespace**: New v3alpha actions APIs for targets and
executions are added under the namespace /resources.
- **Feature Flag**: New v3alpha actions APIs still have to be activated
using the actions feature flag
- **Reduced Executions Overhead**: Executions are managed similar to
settings according to the new API design: an empty list of targets
basically makes an execution a Noop. So a single method, SetExecution is
enough to cover all use cases. Noop executions are not returned in
future search requests.
- **Compatibility**: The executions created with previous v3alpha APIs
are still available to be managed with the new executions API.
# Additional Changes
- Removed integration tests which test executions but rely on readable
targets. They are added again with #8169
# Additional Context
Closes#8168
# Which Problems Are Solved
ZITADEL administrators can enable a setting called "Ignoring unknown
usernames" which helps mitigate attacks that try to guess/enumerate
usernames. If enabled, ZITADEL will show the password prompt even if the
user doesn't exist and report "Username or Password invalid".
Due to a implementation change to prevent deadlocks calling the
database, the flag would not be correctly respected in all cases and an
attacker would gain information if an account exist within ZITADEL,
since the error message shows "object not found" instead of the generic
error message.
# How the Problems Are Solved
- Proper check of the error using an error function / type and
`errors.Is`
# Additional Changes
None.
# Additional Context
- raised in a support request
Co-authored-by: Silvan <silvan.reusser@gmail.com>
(cherry picked from commit a1d24353db)
# Which Problems Are Solved
ZITADEL uses HTML for emails and renders certain information such as
usernames dynamically. That information can be entered by users or
administrators. Due to a missing output sanitization, these emails could
include malicious code.
This may potentially lead to a threat where an attacker, without
privileges, could send out altered notifications that are part of the
registration processes. An attacker could create a malicious link, where
the injected code would be rendered as part of the email.
During investigation of this issue a related issue was found and
mitigated, where on the user's detail page the username was not
sanitized and would also render HTML, giving an attacker the same
vulnerability.
While it was possible to inject HTML including javascript, the execution
of such scripts would be prevented by most email clients and the Content
Security Policy in Console UI.
# How the Problems Are Solved
- All arguments used for email are sanitized (`html.EscapeString`)
- The email text no longer `html.UnescapeString` (HTML in custom text is
still possible)
- Console no longer uses `[innerHtml]` to render the username
# Additional Changes
None
# Additional Context
- raised via email
---------
Co-authored-by: peintnermax <max@caos.ch>
(cherry picked from commit 189505c80f)
# Which Problems Are Solved
ZITADEL administrators can enable a setting called "Ignoring unknown
usernames" which helps mitigate attacks that try to guess/enumerate
usernames. If enabled, ZITADEL will show the password prompt even if the
user doesn't exist and report "Username or Password invalid".
Due to a implementation change to prevent deadlocks calling the
database, the flag would not be correctly respected in all cases and an
attacker would gain information if an account exist within ZITADEL,
since the error message shows "object not found" instead of the generic
error message.
# How the Problems Are Solved
- Proper check of the error using an error function / type and
`errors.Is`
# Additional Changes
None.
# Additional Context
- raised in a support request
Co-authored-by: Silvan <silvan.reusser@gmail.com>
# Which Problems Are Solved
ZITADEL uses HTML for emails and renders certain information such as
usernames dynamically. That information can be entered by users or
administrators. Due to a missing output sanitization, these emails could
include malicious code.
This may potentially lead to a threat where an attacker, without
privileges, could send out altered notifications that are part of the
registration processes. An attacker could create a malicious link, where
the injected code would be rendered as part of the email.
During investigation of this issue a related issue was found and
mitigated, where on the user's detail page the username was not
sanitized and would also render HTML, giving an attacker the same
vulnerability.
While it was possible to inject HTML including javascript, the execution
of such scripts would be prevented by most email clients and the Content
Security Policy in Console UI.
# How the Problems Are Solved
- All arguments used for email are sanitized (`html.EscapeString`)
- The email text no longer `html.UnescapeString` (HTML in custom text is
still possible)
- Console no longer uses `[innerHtml]` to render the username
# Additional Changes
None
# Additional Context
- raised via email
---------
Co-authored-by: peintnermax <max@caos.ch>
This fixes a problem where the org settings were hidden.
The console reads the context from either a query param or the local
storage. When one context was found, it executed a single request with
orgId filter. This let to a single org and then to a hidden org setting,
as we hide org settings for instances with a single result.
(cherry picked from commit 51210c8e34)
This fixes a problem where the org settings were hidden.
The console reads the context from either a query param or the local
storage. When one context was found, it executed a single request with
orgId filter. This let to a single org and then to a hidden org setting,
as we hide org settings for instances with a single result.
# Which Problems Are Solved
The v2beta services are stable but not GA.
# How the Problems Are Solved
The v2beta services are copied to v2. The corresponding v1 and v2beta
services are deprecated.
# Additional Context
Closes#7236
---------
Co-authored-by: Elio Bischof <elio@zitadel.com>
(cherry picked from commit 7d2d85f57c)
# Which Problems Are Solved
The v2beta services are stable but not GA.
# How the Problems Are Solved
The v2beta services are copied to v2. The corresponding v1 and v2beta
services are deprecated.
# Additional Context
Closes#7236
---------
Co-authored-by: Elio Bischof <elio@zitadel.com>
# Which Problems Are Solved
I fixed more spelling and grammar misstakes in the German language
files.
# Additional Context
- Follow-up for PR #8240
Co-authored-by: Fabi <fabienne@zitadel.com>
# Which Problems Are Solved
The initial load of the login UI with dark mode preference
(prefers-color-scheme: dark) first rendered the logo configured for
light mode. Also switching from dark to light or vice versa would result
in the same behavior.
This was due to a mixed logic of server (based on cookie) and client
(prefers-color-scheme and cookie) deciding which mode to render.
# How the Problems Are Solved
- Since the main logic of which mode to use (`prefers-color-scheme`) can
only be achieve client side, both logos will be served in the HTML and
either will be rendered based on CSS.
# Additional Changes
None
# Additional Context
- closes#2085
# Which Problems Are Solved
The default terms of service and privacy policy links are applied to all
new ZITADEL instances, also for self hosters. However, the links
contents don't apply to self-hosters.
# How the Problems Are Solved
The links are removed from the DefaultInstance section in the
*defaults.yaml* file.
By default, the links are not shown anymore in the hosted login pages.
They can still be configured using the privacy policy.
# Additional Context
- Found because of a support request
# Which Problems Are Solved
- Adds delete phone endpoint to v2 api
# How the Problems Are Solved
- Adds new endpoint with DELETE method to /v2beta/users/:userId/phone
which removes currently set phone number
# Additional Changes
- Added integration test for new endpoint.
# Additional Context
- Solves
https://discord.com/channels/927474939156643850/1255557862286032996
This reverts commit e126ccc9aa.
# Which Problems Are Solved
#8295 introduced the possibility to handle idps on a single callback,
but broke current setups.
# How the Problems Are Solved
- Revert the change until a proper solution is found. Revert is needed
as docs were also changed.
# Additional Changes
None.
# Additional Context
- relates to #8295
# Which Problems Are Solved
The default vue logout URL doesn't work with the @zitadel/vue defaults
# How the Problems Are Solved
A trailing slash is added to the default logout URL.
# Additional Context
Found while testing https://github.com/zitadel/zitadel-vue/pull/53
This reduces the initial payload of the `listMyProjectOrgs` to get the
active org by setting the limit of the initial orgs to 100.
Partial of #8272
---------
Co-authored-by: Livio Spring <livio.a@gmail.com>
# Which Problems Are Solved
#8291 added backwards compatibilty for users who were created through
the user V2 API and want to sign in to the login UI.
There were however to issues, where users might be prompted to set a
password even if they already had one set or they would not be able to
submit the email verification code.
# How the Problems Are Solved
- Replaced `SearchUserAuthMethods `with `ListUserAuthMethodTypes` to
check for set up auth methods.
- Fixed page / javascript to disable submit button.
# Additional Changes
- Changed `ListActiveUserAuthMethodTypes ` to `ListUserAuthMethodTypes`
and a `activeOnly` boolean parameter
# Additional Context
- relates to #8291
- noticed internally on QA
# Which Problems Are Solved
Bigger systems need to process many events during the initialisation
phase of the `eventstore.fields`-table. During setup these calls can
time out.
# How the Problems Are Solved
Changed the default behaviour of these projections to not time out and
increased the bulk limit.
(cherry picked from commit 6c1df83071)
# Which Problems Are Solved
- `pgxpool -> pgx` dependency throws "MaxSize must be >= 1" on init if
`postgres.MaxOpenConns` isn't set in the ZItadel config
# How the Problems Are Solved
Only override the `MaxConns` with the Zitadel configured `MaxOpenConns`
if greater than 0 (default value). The default `MaxConns` [is derived by
`pgxpool`](ea9610f672/pgxpool/pool.go (L309-L324))
itself in a sensible way, but somewhat undocumented: checks for explicit
config in connection url or config and falls back on max(num_cpus, 4).
# Additional Changes
Applied same check in cockroach config
# Additional Context
This is likely a regression from the changes in
https://github.com/zitadel/zitadel/pull/8325
(cherry picked from commit e009ed9fe4)
# Which Problems Are Solved
The success and failure URLs post IDP intent are limited to 200
characters. This is very low given the standard for URL lengths is much
higher
https://www.baeldung.com/cs/max-url-length#maximum-length-of-a-url-in-different-browsers
Name of the Browser | URL Length
-- | --
Google Chrome | Maximum of 2048 characters for a URL
Mozilla Firefox | The address bar no longer shows the URL after 65536
characters
Internet Explorer | Maximum length of the URL is 2083 characters
Safari | Supports up to 80000 characters for a URL
Opera | Can display up to 190000 characters in its address bar
Apache | Supports a maximum of 4000 characters
The user post SSO login w/ Zitadel, sees deep links to out app e.g.
`www.mycompany.com/idp/success?deeplink=app/mypage/id/test#thing=abcdef`
and these are usually greater than 200 characters.
# How the Problems Are Solved
Replace this example text with a concise list of changes that this PR
introduces.
- Update validation check for SuccessURL length on v2 and v3 APIs to a
max of 2048 characters
- Update validation check for FailureURL length on v2 and v3 APIs to a
max of 2048 characters
I didn't find any docs t update reflecting this change in length
# Additional Context
Fixes the error
```<PreparedRequest [POST]>, status: 400, reason: Bad Request, response: {"code":3,"message":"invalid StartIdentityProviderIntentRequest.Urls: embedded message failed validation | caused by: invalid RedirectURLs.SuccessUrl: value length must be between 1 and 200 runes, inclusive"}```
(cherry picked from commit d90db6693c)
# Which Problems Are Solved
The connection pool of go uses a high amount of database connections.
# How the Problems Are Solved
The standard lib connection pool was replaced by `pgxpool.Pool`
# Additional Changes
The `db.BeginTx`-spans are removed because they cause to much noise in
the traces.
# Additional Context
- part of https://github.com/zitadel/zitadel/issues/7639
(cherry picked from commit 99c645cc60)
# Which Problems Are Solved
While #8285 also checked for `+proto` and `+json` grpc content types, it
accidentally matched all grpc-web requests to grpc.
# How the Problems Are Solved
- fixed the regex by checking for an exact match (added start `^` and
end `$` anchors)
# Additional Changes
None
# Additional Context
- relates to #8285
(cherry picked from commit d7c0ec282a)
# Which Problems Are Solved
ZITADEL returned a 404 Unimplemented error if the client sent
'application/grpc+proto' or 'application/grpc+json' which are both valid
content types.
# How the Problems Are Solved
changed the header matcher to regexp
# Additional Context
Problem occured in
https://github.com/zitadel/typescript/tree/grpc-transport
(cherry picked from commit aa273ad000)
# Which Problems Are Solved
During performance testing of the `eventstore.fields` table we found
some long running queries which searched for the aggregate id.
# How the Problems Are Solved
A new index was added to the `eventstore.fields`-table called
`f_aggregate_object_type_idx`.
# Additional Changes
None
# Additional Context
- Table was added in https://github.com/zitadel/zitadel/pull/8191
- Part of https://github.com/zitadel/zitadel/issues/7639
(cherry picked from commit 23bebc7e30)
# Which Problems Are Solved
- `pgxpool -> pgx` dependency throws "MaxSize must be >= 1" on init if
`postgres.MaxOpenConns` isn't set in the ZItadel config
# How the Problems Are Solved
Only override the `MaxConns` with the Zitadel configured `MaxOpenConns`
if greater than 0 (default value). The default `MaxConns` [is derived by
`pgxpool`](ea9610f672/pgxpool/pool.go (L309-L324))
itself in a sensible way, but somewhat undocumented: checks for explicit
config in connection url or config and falls back on max(num_cpus, 4).
# Additional Changes
Applied same check in cockroach config
# Additional Context
This is likely a regression from the changes in
https://github.com/zitadel/zitadel/pull/8325
# Which Problems Are Solved
The success and failure URLs post IDP intent are limited to 200
characters. This is very low given the standard for URL lengths is much
higher
https://www.baeldung.com/cs/max-url-length#maximum-length-of-a-url-in-different-browsers
Name of the Browser | URL Length
-- | --
Google Chrome | Maximum of 2048 characters for a URL
Mozilla Firefox | The address bar no longer shows the URL after 65536
characters
Internet Explorer | Maximum length of the URL is 2083 characters
Safari | Supports up to 80000 characters for a URL
Opera | Can display up to 190000 characters in its address bar
Apache | Supports a maximum of 4000 characters
The user post SSO login w/ Zitadel, sees deep links to out app e.g.
`www.mycompany.com/idp/success?deeplink=app/mypage/id/test#thing=abcdef`
and these are usually greater than 200 characters.
# How the Problems Are Solved
Replace this example text with a concise list of changes that this PR
introduces.
- Update validation check for SuccessURL length on v2 and v3 APIs to a
max of 2048 characters
- Update validation check for FailureURL length on v2 and v3 APIs to a
max of 2048 characters
I didn't find any docs t update reflecting this change in length
# Additional Context
Fixes the error
```<PreparedRequest [POST]>, status: 400, reason: Bad Request, response: {"code":3,"message":"invalid StartIdentityProviderIntentRequest.Urls: embedded message failed validation | caused by: invalid RedirectURLs.SuccessUrl: value length must be between 1 and 200 runes, inclusive"}```
# Which Problems Are Solved
The mirror command read the configurations in the wrong order
# How the Problems Are Solved
The Pre execution run of `mirror` reads the default config first and
then applies the custom configs
# Which Problems Are Solved
The connection pool of go uses a high amount of database connections.
# How the Problems Are Solved
The standard lib connection pool was replaced by `pgxpool.Pool`
# Additional Changes
The `db.BeginTx`-spans are removed because they cause to much noise in
the traces.
# Additional Context
- part of https://github.com/zitadel/zitadel/issues/7639
# Which Problems Are Solved
Bigger systems need to process many events during the initialisation
phase of the `eventstore.fields`-table. During setup these calls can
time out.
# How the Problems Are Solved
Changed the default behaviour of these projections to not time out and
increased the bulk limit.
# Which Problems Are Solved
User created through the User V2 API without any authentication method
and possibly unverified email address was not able to login through the
current hosted login UI.
An unverified email address would result in a mail verification and not
an initialization mail like it would with the management API. Also the
login UI would then require the user to enter the init code, which the
user never received.
# How the Problems Are Solved
- When verifying the email through the login UI, it will check for
existing auth methods (password, IdP, passkeys). In case there are none,
the user will be prompted to set a password.
- When a user was created through the V2 API with a verified email and
no auth method, the user will be prompted to set a password in the login
UI.
- Since setting a password requires a corresponding code, the code will
be generated and sent when login in.
# Additional Changes
- Changed `RequestSetPassword` to get the codeGenerator from the
eventstore instead of getting it from query.
# Additional Context
- closes https://github.com/zitadel/zitadel/issues/6600
- closes https://github.com/zitadel/zitadel/issues/8235
---------
Co-authored-by: Stefan Benz <46600784+stebenz@users.noreply.github.com>
# Which Problems Are Solved
Both the login UI and the IdP intent flow have their own IdP callback
endpoints.
This makes configuration hard to impossible (e.g. Github only allows one
endpoint) for customers.
# How the Problems Are Solved
- The login UI prefixes the `state` parameter when creating an auth /
SAML request.
- All requests now use the `/idp/callback` or the corresponding
variation (e.g. SAML)
- On callback, the state, resp. its prefix is checked. In case of the
login UI prefix, the request will be forwarded to the existing login UI
handler without the prefix state.
Existing setups will therefore not be affected and also requests started
before this release can be handled without any impact.
- Console only lists the "new" endpoint(s). Any
`/login/externalidp/callback` is removed.
# Additional Changes
- Cleaned up some images from the IdP documentation.
- fix the error handling in `handleExternalNotFoundOptionCheck`
# Additional Context
- closes#8236
# Which Problems Are Solved
ListOrgs has no option to select for organizations specific to Ids.
# How the Problems Are Solved
Add OrgIDQuery to ListOrgs.
# Additional Changes
Clean up double mapping for the OrgQueries.
# Additional Context
- noted internally while checking performance issues (in Console)
# Which Problems Are Solved
In the console a wrong translation is shown for the password re auth
time, it is shown that it is the time after which users have to renew
their password, but it is the time, after which they have to
authenticate again with their password
# How the Problems Are Solved
Translations are changed in all languages
Co-authored-by: Livio Spring <livio.a@gmail.com>
# Which Problems Are Solved
Logs the type of sonyflake strategy used for generating unique machine
IDs
# How the Problems Are Solved
- Created function to log machine id strategy on the start up logs
# Additional Changes
- Added public function for retrieving current strategy set by
configuration
# Additional Context
- Closes#7750