- This is a 'front-port' of the already existing patch on v1.21 and
v1.20, but applied on top of what Gitea has done to rework the LTA
mechanism. Forgejo will stick with the reworked mechanism by the Forgejo
Security team for the time being. The removal of legacy code (AES-GCM) has been
left out.
- The current architecture is inherently insecure, because you can
construct the 'secret' cookie value with values that are available in
the database. Thus provides zero protection when a database is
dumped/leaked.
- This patch implements a new architecture that's inspired from: [Paragonie Initiative](https://paragonie.com/blog/2015/04/secure-authentication-php-with-long-term-persistence#secure-remember-me-cookies).
- Integration testing is added to ensure the new mechanism works.
- Removes a setting, because it's not used anymore.
(cherry picked from commit e3d6622a63)
(cherry picked from commit fef1a6dac5)
(cherry picked from commit b0c5165145)
(cherry picked from commit 7ad51b9f8d)
(cherry picked from commit 64f053f383)
(cherry picked from commit e11dcc60f2)
use backticks to avoid backslash
(cherry picked from commit 34212791ee)
(cherry picked from commit bde9473c69)
(cherry picked from commit d4deb43084)
(cherry picked from commit 08e91649b0)
(cherry picked from commit 2b988e5415)
[TESTS] auth LinkAccount test coverage (squash)
(cherry picked from commit a2b2e3066b)
(cherry picked from commit 841d1b5073)
(cherry picked from commit 35da630ad8)
(cherry picked from commit caf2dc4fa7)
(cherry picked from commit 6eb81e67ba)
(cherry picked from commit d59757239f)
(cherry picked from commit 38a121b688)
(cherry picked from commit 20613874ee)
(cherry picked from commit 6d2705e108)
(cherry picked from commit f177b72814)
(cherry picked from commit 75e1fc4c83)
(cherry picked from commit ba64fa9867)
(cherry picked from commit 0b8ab0893e)
(cherry picked from commit 1419d11435)
(cherry picked from commit 38766847e0)
(cherry picked from commit 6f23426a6a)
(cherry picked from commit 9e0ff9ca54)
(cherry picked from commit 353f3601c3)
(cherry picked from commit 6e4ae401d8)
(cherry picked from commit 1a7afe4153)
(cherry picked from commit f9f3e0cc02)
(cherry picked from commit 22fd0337f3)
(cherry picked from commit ee57e138d1)
(cherry picked from commit 21f9b7e73d)
The function `GetByBean` has an obvious defect that when the fields are
empty values, it will be ignored. Then users will get a wrong result
which is possibly used to make a security problem.
To avoid the possibility, this PR removed function `GetByBean` and all
references.
And some new generic functions have been introduced to be used.
The recommand usage like below.
```go
// if query an object according id
obj, err := db.GetByID[Object](ctx, id)
// query with other conditions
obj, err := db.Get[Object](ctx, builder.Eq{"a": a, "b":b})
```
## Bug in Gitea
I ran into this bug when I accidentally used the wrong redirect URL for
the oauth2 provider when using mssql. But the oauth2 provider still got
added.
Most of the time, we use `Delete(&some{id: some.id})` or
`In(condition).Delete(&some{})`, which specify the conditions. But the
function uses `Delete(source)` when `source.Cfg` is a `TEXT` field and
not empty. This will cause xorm `Delete` function not working in mssql.
61ff91f960/models/auth/source.go (L234-L240)
## Reason
Because the `TEXT` field can not be compared in mssql, xorm doesn't
support it according to [this
PR](https://gitea.com/xorm/xorm/pulls/2062)
[related
code](b23798dc98/internal/statements/statement.go (L552-L558))
in xorm
```go
if statement.dialect.URI().DBType == schemas.MSSQL && (col.SQLType.Name == schemas.Text ||
col.SQLType.IsBlob() || col.SQLType.Name == schemas.TimeStampz) {
if utils.IsValueZero(fieldValue) {
continue
}
return nil, fmt.Errorf("column %s is a TEXT type with data %#v which cannot be as compare condition", col.Name, fieldValue.Interface())
}
}
```
When using the `Delete` function in xorm, the non-empty fields will
auto-set as conditions(perhaps some special fields are not?). If `TEXT`
field is not empty, xorm will return an error. I only found this usage
after searching, but maybe there is something I missing.
---------
Co-authored-by: delvh <dev.lh@web.de>
The steps to reproduce it.
First, create a new oauth2 source.
Then, a user login with this oauth2 source.
Disable the oauth2 source.
Visit users -> settings -> security, 500 will be displayed.
This is because this page only load active Oauth2 sources but not all
Oauth2 sources.
Closes #27455
> The mechanism responsible for long-term authentication (the 'remember
me' cookie) uses a weak construction technique. It will hash the user's
hashed password and the rands value; it will then call the secure cookie
code, which will encrypt the user's name with the computed hash. If one
were able to dump the database, they could extract those two values to
rebuild that cookie and impersonate a user. That vulnerability exists
from the date the dump was obtained until a user changed their password.
>
> To fix this security issue, the cookie could be created and verified
using a different technique such as the one explained at
https://paragonie.com/blog/2015/04/secure-authentication-php-with-long-term-persistence#secure-remember-me-cookies.
The PR removes the now obsolete setting `COOKIE_USERNAME`.
This PR removed `unittest.MainTest` the second parameter
`TestOptions.GiteaRoot`. Now it detects the root directory by current
working directory.
---------
Co-authored-by: wxiaoguang <wxiaoguang@gmail.com>
This PR is an extended implementation of #25189 and builds upon the
proposal by @hickford in #25653, utilizing some ideas proposed
internally by @wxiaoguang.
Mainly, this PR consists of a mechanism to pre-register OAuth2
applications on startup, which can be enabled or disabled by modifying
the `[oauth2].DEFAULT_APPLICATIONS` parameter in app.ini. The OAuth2
applications registered this way are being marked as "locked" and
neither be deleted nor edited over UI to prevent confusing/unexpected
behavior. Instead, they're being removed if no longer enabled in config.
The implemented mechanism can also be used to pre-register other OAuth2
applications in the future, if wanted.
Co-authored-by: hickford <mirth.hickford@gmail.com>
Co-authored-by: wxiaoguang <wxiaoguang@gmail.com>
---------
Co-authored-by: M Hickford <mirth.hickford@gmail.com>
Co-authored-by: wxiaoguang <wxiaoguang@gmail.com>
## Changes
- Adds the following high level access scopes, each with `read` and
`write` levels:
- `activitypub`
- `admin` (hidden if user is not a site admin)
- `misc`
- `notification`
- `organization`
- `package`
- `issue`
- `repository`
- `user`
- Adds new middleware function `tokenRequiresScopes()` in addition to
`reqToken()`
- `tokenRequiresScopes()` is used for each high-level api section
- _if_ a scoped token is present, checks that the required scope is
included based on the section and HTTP method
- `reqToken()` is used for individual routes
- checks that required authentication is present (but does not check
scope levels as this will already have been handled by
`tokenRequiresScopes()`
- Adds migration to convert old scoped access tokens to the new set of
scopes
- Updates the user interface for scope selection
### User interface example
<img width="903" alt="Screen Shot 2023-05-31 at 1 56 55 PM"
src="https://github.com/go-gitea/gitea/assets/23248839/654766ec-2143-4f59-9037-3b51600e32f3">
<img width="917" alt="Screen Shot 2023-05-31 at 1 56 43 PM"
src="https://github.com/go-gitea/gitea/assets/23248839/1ad64081-012c-4a73-b393-66b30352654c">
## tokenRequiresScopes Design Decision
- `tokenRequiresScopes()` was added to more reliably cover api routes.
For an incoming request, this function uses the given scope category
(say `AccessTokenScopeCategoryOrganization`) and the HTTP method (say
`DELETE`) and verifies that any scoped tokens in use include
`delete:organization`.
- `reqToken()` is used to enforce auth for individual routes that
require it. If a scoped token is not present for a request,
`tokenRequiresScopes()` will not return an error
## TODO
- [x] Alphabetize scope categories
- [x] Change 'public repos only' to a radio button (private vs public).
Also expand this to organizations
- [X] Disable token creation if no scopes selected. Alternatively, show
warning
- [x] `reqToken()` is missing from many `POST/DELETE` routes in the api.
`tokenRequiresScopes()` only checks that a given token has the correct
scope, `reqToken()` must be used to check that a token (or some other
auth) is present.
- _This should be addressed in this PR_
- [x] The migration should be reviewed very carefully in order to
minimize access changes to existing user tokens.
- _This should be addressed in this PR_
- [x] Link to api to swagger documentation, clarify what
read/write/delete levels correspond to
- [x] Review cases where more than one scope is needed as this directly
deviates from the api definition.
- _This should be addressed in this PR_
- For example:
```go
m.Group("/users/{username}/orgs", func() {
m.Get("", reqToken(), org.ListUserOrgs)
m.Get("/{org}/permissions", reqToken(), org.GetUserOrgsPermissions)
}, tokenRequiresScopes(auth_model.AccessTokenScopeCategoryUser,
auth_model.AccessTokenScopeCategoryOrganization),
context_service.UserAssignmentAPI())
```
## Future improvements
- [ ] Add required scopes to swagger documentation
- [ ] Redesign `reqToken()` to be opt-out rather than opt-in
- [ ] Subdivide scopes like `repository`
- [ ] Once a token is created, if it has no scopes, we should display
text instead of an empty bullet point
- [ ] If the 'public repos only' option is selected, should read
categories be selected by default
Closes #24501
Closes #24799
Co-authored-by: Jonathan Tran <jon@allspice.io>
Co-authored-by: Kyle D <kdumontnu@gmail.com>
Co-authored-by: silverwind <me@silverwind.io>
minio/sha256-simd provides additional acceleration for SHA256 using
AVX512, SHA Extensions for x86 and ARM64 for ARM.
It provides a drop-in replacement for crypto/sha256 and if the
extensions are not available it falls back to standard crypto/sha256.
---------
Signed-off-by: Andrew Thornton <art27@cantab.net>
Co-authored-by: John Olheiser <john.olheiser@gmail.com>
The API to create tokens is missing the ability to set the required
scopes for tokens, and to show them on the API and on the UI.
This PR adds this functionality.
Signed-off-by: Andrew Thornton <art27@cantab.net>
This PR adds the support for scopes of access tokens, mimicking the
design of GitHub OAuth scopes.
The changes of the core logic are in `models/auth` that `AccessToken`
struct will have a `Scope` field. The normalized (no duplication of
scope), comma-separated scope string will be stored in `access_token`
table in the database.
In `services/auth`, the scope will be stored in context, which will be
used by `reqToken` middleware in API calls. Only OAuth2 tokens will have
granular token scopes, while others like BasicAuth will default to scope
`all`.
A large amount of work happens in `routers/api/v1/api.go` and the
corresponding `tests/integration` tests, that is adding necessary scopes
to each of the API calls as they fit.
- [x] Add `Scope` field to `AccessToken`
- [x] Add access control to all API endpoints
- [x] Update frontend & backend for when creating tokens
- [x] Add a database migration for `scope` column (enable 'all' access
to past tokens)
I'm aiming to complete it before Gitea 1.19 release.
Fixes #4300
- Move the file `compare.go` and `slice.go` to `slice.go`.
- Fix `ExistsInSlice`, it's buggy
- It uses `sort.Search`, so it assumes that the input slice is sorted.
- It passes `func(i int) bool { return slice[i] == target })` to
`sort.Search`, that's incorrect, check the doc of `sort.Search`.
- Conbine `IsInt64InSlice(int64, []int64)` and `ExistsInSlice(string,
[]string)` to `SliceContains[T]([]T, T)`.
- Conbine `IsSliceInt64Eq([]int64, []int64)` and `IsEqualSlice([]string,
[]string)` to `SliceSortedEqual[T]([]T, T)`.
- Add `SliceEqual[T]([]T, T)` as a distinction from
`SliceSortedEqual[T]([]T, T)`.
- Redesign `RemoveIDFromList([]int64, int64) ([]int64, bool)` to
`SliceRemoveAll[T]([]T, T) []T`.
- Add `SliceContainsFunc[T]([]T, func(T) bool)` and
`SliceRemoveAllFunc[T]([]T, func(T) bool)` for general use.
- Add comments to explain why not `golang.org/x/exp/slices`.
- Add unit tests.
`hex.EncodeToString` has better performance than `fmt.Sprintf("%x",
[]byte)`, we should use it as much as possible.
I'm not an extreme fan of performance, so I think there are some
exceptions:
- `fmt.Sprintf("%x", func(...)[N]byte())`
- We can't slice the function return value directly, and it's not worth
adding lines.
```diff
func A()[20]byte { ... }
- a := fmt.Sprintf("%x", A())
- a := hex.EncodeToString(A()[:]) // invalid
+ tmp := A()
+ a := hex.EncodeToString(tmp[:])
```
- `fmt.Sprintf("%X", []byte)`
- `strings.ToUpper(hex.EncodeToString(bytes))` has even worse
performance.
Change all license headers to comply with REUSE specification.
Fix #16132
Co-authored-by: flynnnnnnnnnn <flynnnnnnnnnn@github>
Co-authored-by: John Olheiser <john.olheiser@gmail.com>
Fix #19513
This PR introduce a new db method `InTransaction(context.Context)`,
and also builtin check on `db.TxContext` and `db.WithTx`.
There is also a new method `db.AutoTx` has been introduced but could be used by other PRs.
`WithTx` will always open a new transaction, if a transaction exist in context, return an error.
`AutoTx` will try to open a new transaction if no transaction exist in context.
That means it will always enter a transaction if there is no error.
Co-authored-by: delvh <dev.lh@web.de>
Co-authored-by: 6543 <6543@obermui.de>
The OAuth spec [defines two types of
client](https://datatracker.ietf.org/doc/html/rfc6749#section-2.1),
confidential and public. Previously Gitea assumed all clients to be
confidential.
> OAuth defines two client types, based on their ability to authenticate
securely with the authorization server (i.e., ability to
> maintain the confidentiality of their client credentials):
>
> confidential
> Clients capable of maintaining the confidentiality of their
credentials (e.g., client implemented on a secure server with
> restricted access to the client credentials), or capable of secure
client authentication using other means.
>
> **public
> Clients incapable of maintaining the confidentiality of their
credentials (e.g., clients executing on the device used by the resource
owner, such as an installed native application or a web browser-based
application), and incapable of secure client authentication via any
other means.**
>
> The client type designation is based on the authorization server's
definition of secure authentication and its acceptable exposure levels
of client credentials. The authorization server SHOULD NOT make
assumptions about the client type.
https://datatracker.ietf.org/doc/html/rfc8252#section-8.4
> Authorization servers MUST record the client type in the client
registration details in order to identify and process requests
accordingly.
Require PKCE for public clients:
https://datatracker.ietf.org/doc/html/rfc8252#section-8.1
> Authorization servers SHOULD reject authorization requests from native
apps that don't use PKCE by returning an error message
Fixes #21299
Co-authored-by: wxiaoguang <wxiaoguang@gmail.com>
Co-authored-by: Lunny Xiao <xiaolunwen@gmail.com>
A lot of our code is repeatedly testing if individual errors are
specific types of Not Exist errors. This is repetitative and unnecesary.
`Unwrap() error` provides a common way of labelling an error as a
NotExist error and we can/should use this.
This PR has chosen to use the common `io/fs` errors e.g.
`fs.ErrNotExist` for our errors. This is in some ways not completely
correct as these are not filesystem errors but it seems like a
reasonable thing to do and would allow us to simplify a lot of our code
to `errors.Is(err, fs.ErrNotExist)` instead of
`package.IsErr...NotExist(err)`
I am open to suggestions to use a different base error - perhaps
`models/db.ErrNotExist` if that would be felt to be better.
Signed-off-by: Andrew Thornton <art27@cantab.net>
Co-authored-by: delvh <dev.lh@web.de>
Adds the settings pages to create OAuth2 apps also to the org settings
and allows to create apps for orgs.
Refactoring: the oauth2 related templates are shared for
instance-wide/org/user, and the backend code uses `OAuth2CommonHandlers`
to share code for instance-wide/org/user.
Co-authored-by: wxiaoguang <wxiaoguang@gmail.com>
WebAuthn have updated their specification to set the maximum size of the
CredentialID to 1023 bytes. This is somewhat larger than our current
size and therefore we need to migrate.
The PR changes the struct to add CredentialIDBytes and migrates the CredentialID string
to the bytes field before another migration drops the old CredentialID field. Another migration
renames this field back.
Fix #20457
Signed-off-by: Andrew Thornton <art27@cantab.net>
v208.go is seriously broken as it misses an ID() check. We need to no-op and remigrate all of the u2f keys.
See #18756
Signed-off-by: Andrew Thornton <art27@cantab.net>
- Switch to use `CryptoRandomBytes` instead of `CryptoRandomString`, OAuth's secrets are copied pasted and don't need to avoid dubious characters etc.
- `CryptoRandomBytes` gives  `CryptoRandomString` gives  possible states.
- Add a prefix, such that code scanners can easily grep these in source code.
- 32 Bytes + prefix
