Some optimizations in regards to downloading Favicon's.
I also encounterd some issues with accessing some sites where the
connection got dropped or closed early. This seems a reqwest/hyper
thingy, https://github.com/hyperium/hyper/issues/2136. This is now also
fixed.
General:
- Decreased struct size
- Decreased memory allocations
- Optimized tokenizer a bit more to only emit tags when all attributes are there and are valid.
reqwest/hyper connection issue:
The following changes helped solve the connection issues to some sites.
The endresult is that some icons are now able to be downloaded always instead of sometimes.
- Enabled some extra reqwest features, `deflate` and `native-tls-alpn`
(Which do not bring in any extra crates since other crates already enabled them, but they were not active for Vaultwarden it self)
- Configured reqwest to have a max amount of idle pool connections per host
- Configured reqwest to timeout the idle connections in 10 seconds
When a user does not have an account yet and SMTP was disabled it would
set the UserOrg status still to Accepted, though that would make it
possible to verify the user by the Org Admin's.
This would fail, since the user didn't actually crated his account, and
therefor no PublicKey existed.
This PR fixes this behaviour by checking if the password is empty and if
so, puts the user to an `Invited` state instead of `Accepted`.
Fixes #3737
- Do not update `externalId` on group updates
Groups are only updated via the web-vault currently, and those do not
send the `externalId` value, and thus we need to prevent updating it.
- Refactored some other ExternalId functions
- Prevent empty `externalId` on `Collections`
- Return `externalId` for users
Fixes #3685
In a previous commit push notifications for mobile were added, but this
introduced a header guard which caused issues with anonymous endpoints.
This PR fixes this by using a uuid with only 0's.
Fixes #3607
Merging a PR from @Kurnihil into the already rebased branch.
Made some small changes to make it work with newer changes.
Some finetuning is probably still needed.
Co-authored-by: Daniele Andrei <daniele.andrei@geo-satis.com>
Co-authored-by: Kurnihil
This is a WIP for adding organization token login support.
It has basic token login and verification support, but that's about it.
This branch is a refresh of the previous version, and will contain code
from a PR based upon my previous branch.
When you are not loggedin, and have no cookie etc.. we always returned a 401.
This was mainly to allow the login page on all the sub pages, and after
login being redirected to the requested page, for these pages a 401 is a
valid response, since, you do not have access.
But for the main `/admin` page, it should just respond with a `200` and
show the login page.
This PR fixes this flow and response. It should prevent people using
Fail2ban, or other tools being triggered by only accessing the login page.
Resolves #3540
There now is a `v0.5` branch which will be the final release version
when the time is there. Switched to this instead of the `master` branch
which contains other fixes and enhancements as well (for `v0.6`).
This should solve all the endpoint issue we were having.
During setting the latest commit hash for Rocket and updating all the
other crates, there were some messages regarding the usage of `String`
for the Rocket endpoint function calls. I acted upon this message and
changed all `String` types to `&str` and modified the code where needed.
This ended up in less alloc calls, and probably also a bit less memory usage.
- Updated all the crates and commit hashes
- Modified all `String` to `&str` where applicable
- Fixed issue with domains starting with `admin`
- Fixed issue with DUO not being enabled globally anymore (regression)
- Renamed `Ciphers` to `Entries` in overview
- Improved `ADMIN_TOKEN` description
- Updated jquery-slim and datatables
Resolves #3382
Resolves #3415
Resolves discussion on #3288
This PR implements a (not yet fully released) new feature of Rocket which allows WebSockets/Upgrade connections.
No more need for multiple ports to be opened for Vaultwarden.
No explicit need for a reverse proxy to get WebSockets to work (Although I still suggest to use a reverse proxy).
- Using a git revision for Rocket, since `rocket_ws` is not yet released.
- Updated other crates as well.
- Added a connection guard to clear the WS connection from the Users list.
Fixes #685
Fixes #2917
Fixes #1424
For some reason I encountered a strange bug which resulted in sending
out multiple websocket notifications for the exact same user.
Added a `distinct()` for the query to filter out multiple uuid's.
the client will ignore the value of theses fields in case of `PBKDF2`
(whether they are unset or left from trying out `Argon2id` as KDF).
with `Argon2id` those fields should never be `null` but always in a
valid state. if they are `null` (how would that even happen?) the
client still assumes default values for `Argon2id` (i.e. m=64 and p=4)
and if they are set to something else login will fail anyway.
There was used a wrong macro to produce an error message when mailing
the user his password was reset failed. It was using `error!()` which
does not return an `Err` and aborts the rest of the code.
This resulted in the users password still being resetted, but not being
notified. This PR fixes this by using `err!()`. Also, do not set the
user object as mutable until it really is needed.
Second, when a user was using the new Argon2id KDF with custom values
like memory and parallelism, that would have rendered the password
incorrect. The endpoint which should return all the data did not
returned all the new Argon2id values.
Fixes #3388
Co-authored-by: Stefan Melmuk <509385+stefan0xC@users.noreply.github.com>
- Updated workflows to use new checkout version
This probably fixes the curl download for hadolint also.
- Updated crates including Rocket to the latest rc3 :party:
- Applied 2 nightly clippy lints to prevent future clippy issues.
Since we now use the `ClientIp` Guard on a lot more places, it also
increases the size of binary, and the macro generated code because of
this extra Guard. By merging the `ClientIp` Guard with the several
`Header` guards we have it reduces the amount of code generated
(including LLVM IR), but also a small speedup in build time.
I also spotted some small `json!()` optimizations which also reduced the
amount of code generated.
- Changed MSRV to v1.65.
Discussed this with @dani-garcia, and we will support **N-2**.
This is/will be the same as for the `time` crate we use.
Also updated the wiki regarding this https://github.com/dani-garcia/vaultwarden/wiki/Building-binary
- Removed backtrace crate in favor of `std::backtrace` stable since v1.65
- Updated Rust to v1.67.1
- Updated all the crates
- Updated the GHA action versions
- Adjusted the GHA MSRV build to extract the MSRV from `Cargo.toml`
There was a small bug left in regards to the web-vault v2023.2.0 fixes.
This PR fixes the left items. I think all should be addressed now.
When editing a User, you were not able to see or edit groups, or see
wich collections a user bellonged to.
Fixes #3311
Rocket automatically implements a HEAD route when there's a matching GET
route, but relying on this behavior also means a spurious error gets
logged due to <https://github.com/SergioBenitez/Rocket/issues/1098>.
Add explicit HEAD routes for `/` and `/alive` to prevent uptime monitoring
services from generating error messages like `No matching routes for HEAD /`.
With these new routes, `HEAD /` only checks that the server can respond over
the network, while `HEAD /alive` also checks that the database connection is
alive, similar to `GET /alive`.
Added support for Argon2 hashing support for the `ADMIN_TOKEN` instead
of only supporting a plain text string.
The hash must be a PHC string which can be generated via the `argon2`
CLI **or** via the also built-in hash command in Vaultwarden.
You can simply run `vaultwarden hash` to generate a hash based upon a
password the user provides them self.
Added a warning during startup and within the admin settings panel is
the `ADMIN_TOKEN` is not an Argon2 hash.
Within the admin environment a user can ignore that warning and it will
not be shown for at least 30 days. After that the warning will appear
again unless the `ADMIN_TOKEN` has be converted to an Argon2 hash.
I have also tested this on my RaspberryPi 2b and there the `Bitwarden`
preset takes almost 4.5 seconds to generate/verify the Argon2 hash.
Using the `OWASP` preset it is below 1 second, which I think should be
fine for low-graded hardware. If it is needed people could use lower
memory settings, but in those cases I even doubt Vaultwarden it self
would run. They can always use the `argon2` CLI and generate a faster hash.
- Updated datatables
- Added NTP Time check
- Added Collections, Groups and Events count for orgs
- Renamed `Items` to `Ciphers`
- Some small style updates
- Supports the new Collection/Group/User editing UI's
- Support `/partial` endpoint for cipher updating to allow folder and favorite update for read-only ciphers.
- Prevent `Favorite`, `Folder`, `read-only` and `hide-passwords` from being added to the organizational sync.
- Added and corrected some `Object` key's to the output json.
Fixes #3279
In the org vault view, the Bitwarden web vault currently tries to fetch the
groups for an org regardless of whether it claims to have group support.
If this errors out, no vault items are displayed.
I messed up with identation sorry it's my first PR
Fix Collection Read Only access for groups
Fix Collection Read Only access for groups
With indentation modification
With existing groups configured within an org, deleting that org would
fail because of Foreign Key issues.
This PR fixes this by making sure the groups get deleted before the org does.
Fixes #3247
During the client API login we need to have a `device_identifier`, `device_name` and `device_type`.
When these were not provided Vaultwarden would panic.
This PR add checks for these fields and makes sure it returns a better error message instead of causing a panic.
When an icon will not be downloaded due to matching a configured
blacklist, ensure that the log message indicates the type of blacklist
that was matched.
the client does not send the key on every update of an emergency access
contact so the field would be emptied on a change of the wait days or access level.
When a non sqlite database is used, loading the admin interface fails
because the backup button is not generated.
This PR is solves it by checking if the elements are valid.
Also made some other changes and fixed some eslint errors.
Showing `_post` errors is better now.
Update jquery to latest version.
Fixes #3166
We also need to validate the note sizes on key-rotation.
If we do not validate them before we store them, that could lead to a
partial or total loss of the password vault. Validating these
restrictions before actually processing them to store/replace the
existing ciphers should prevent this.
There was also a small bug when using web-sockets. The client which is
triggering the password/key-rotation change should not be forced to
logout via a web-socket request. That is something the client will
handle it self. Refactored the logout notification to either send the
device uuid or not on specific actions.
Fixes #3152
- Change default Password Hash KDF Storage from 100_000 to 600_000 iterations
- Update Password Hash when the default iteration value is different
- Validate password_iterations
- Validate client-side KDF to prevent it from being set lower than 100_000
The Organization member overview supports showing an icon if the user
has MFA enabled or not. This PR adds this feature.
This is very useful if you want to enable force mfa for example.
As mentioned in #3111, using a very very large vault causes some issues.
Mainly because of a SQLite limit, but, it could also cause issue on
MariaDB/MySQL or PostgreSQL. It also uses a lot of memory, and memory
allocations.
This PR solves this by removing the need of all the cipher_uuid's just
to gather the correct attachments.
It will use the user_uuid and org_uuid's to get all attachments linked
to both, weither the user has access to them or not. This isn't an
issue, since the matching is done per cipher and the attachment data is
only returned if there is a matching cipher to where the user has access to.
I also modified some code to be able to use `::with_capacity(n)` where
possible. This prevents re-allocations if the `Vec` increases size,
which will happen a lot if there are a lot of ciphers.
According to my tests measuring the time it takes to sync, it seems to
have lowered the duration a bit more.
Fixes #3111
Recent versions of the Bitwarden clients (see bitwarden/clients#3574)
won't parse non-JSON responses. The most noticeable consequence is that
`/api/accounts/revision-date` responses won't be parsed, leading to
`/api/sync` always being called, even when it's not necessary.
When importing to Vaultwarden (or Bitwarden) notes larger then 10_000
encrypted characters are invalid. This because it for one isn't
compatible with Bitwarden. And some clients tend to break on very large
notes.
We already added a check for this limit when adding a single cipher, but
this caused issues during import, and could cause a partial imported
vault. Bitwarden does some validations before actually running it
through the import process and generates a special error message which
helps the user indicate which items are invalid during the import.
This PR adds that validation check and returns the same kind of error.
Fixes #3048
As kinda discussed here #3090, the messages regarding loading the
configuration files is a bit strange or unclear. There have been some
other reports regarding this in the past, but wasn't that big a of a
deal.
But to make the whole process it bit more nice, this PR adjusts the way
it reports issues and some small changes to the messages to make it all
a bit more clear.
- Do not report a missing `.env` file, but only send a message when using one.
- Exit instead of Panic, a panic causes a stacktrace, which isn't needed
here. I'm using a exit code 255 here so it is different to the other
exit's we use.
- Exit on more issues, since if we continue, it could cause
configuration issues if the user thinks all is fine.
- Use the actual env file used in the messages instead of `.env`.
- Added a **INFO** message when loading the `config.json`.
This makes it consistent with the info message for loading the env file.
Resolves #3090
Previously the websocket notifications were using `app_id` as the
`ContextId`. This was incorrect and should have been the device_uuid
from the client device executing the request. The clients will ignore
the websocket request if the uuid matches. This also fixes some issues
with the Desktop client which is able to modify attachments within the
same screen and causes an issue when saving the attachment afterwards.
Also changed the way to handle removed attachments, since that causes an
error saving the vault cipher afterwards, complaining about a missing
attachment. Bitwarden ignores this, and continues with the remaining
attachments (if any). This also fixes #2591 .
Further some more websocket notifications have been added to some other
functions which enhance the user experience.
- Logout users when deauthed, changed password, rotated keys
- Trigger OrgSyncKeys on user confirm and removal
- Added some extra to the send feature
Also renamed UpdateTypes to match Bitwarden naming.
If the `YUBICO_SERVER` is defined to an empty string, the whole yubikey
implementation doesn't work anymore.
This PR adds a check for this variable that it at least starts with `https://`.
Resolves #3003
- Removed `unsafe-inline` for javascript from CSP.
The admin interface now uses files instead of inline javascript.
- Modified javascript to work not being inline.
- Run eslint over javascript and fixed some items.
- Added a `to_json` Handlebars helper.
Used at the diagnostics page.
- Changed `AdminTemplateData` struct to be smaller.
The `config` was always added, but only used at one page.
Same goes for `can_backup` and `version`.
- Also inlined CSS.
We can't remove the `unsafe-inline` from css, because that seems to
break the web-vault currently. That might need some further checks.
But for now the 404 page and all the admin pages are clear of inline scripts and styles.
- Put groups support behind a feature flag, and disabled by default.
The reason is that it has some known issues, but we want to keep
optimizing this feature. Putting it behind a feature flag could help
some users, and the developers into optimizing this feature without to
much trouble.
Further:
- Updates Rust to v1.66.0
- Updated GHA workflows
- Updated Alpine to 3.17
- Updated jquery to v3.6.2
- Moved jdenticon.js to load at the bottom, fixes an issue on chromium
- Added autocomplete attribute to admin login password field
- Added some extra CSP options (Tested this on Safari, Firefox, Chrome, Bitwarden Desktop)
- Moved uppercase convertion from runtime to compile-time using `paste`
for building the environment variables, lowers heap allocations.
When audit logging was introduced there entered a small bug preventing
the recover-2fa from working.
This PR fixes that by add a new headers check to extract the device-type
when possible and use that for the logging.
Fixes #2985
The yubico crate uses blocking reqwest, and we called the `verify` from
a async thread. To prevent issues we need to wrap it within a
`spawn_blocking`.
It looks like Bitwarden, in-the-end, didn't changed the export feature
on v2022.11.0, and now have put in on v2023.1.0.
This patch now changes that to the same version.
Before those new clients are being released, we should see if they
changed that again, and adjust where needed.
This changes the masking function to hide a bit more information from
the generated support string. It will still keep showing the `://` for
example, and `,`, but other characters will be hidden.
Also did some small changes on some key's which all showed up as
`Internal` on the Settings page.
Fixes #2929
Using the `inner_join` seems to cause issues, even though i have tested
it. Strangely it does cause issues. Reverting it back to `left_join`
seems to solve the issue for me.
Fixes #2975
This PR should fix the managers and group link.
Although i think there might be a cleaner sollution, there are a lot of
other items to fix here which we should do in time.
But for now, with theh group support already merged, this fix should at
least help solving issue #2932.
Fixes #2932
Currently when you login into the admin, and then directly hit the save
button, it will come with a re-post/re-submit warning.
This has to do with the `window.location.reload()` function, which
triggers the admin login POST again.
By changing the way to reload the page, we prevent this repost.
As discussed in #2937, this will limit the amount of encrypted
characters to 10.000 characters, same as Bitwarden.
This will not break current ciphers which exceed this limit, but it will prevent those
ciphers from being updated.
Fixes #2937
- Updated Rust deps and one small change regarding chrono
- Updated bootstrap 5 css
- Updated datatables
- Replaced identicon.js with jdenticon.
identicon.js is unmaintained ( https://github.com/stewartlord/identicon.js/issues/52 )
The icon's are very different, but nice. It also doesn't need custom
code to find and update the icons our selfs.
- Several cleanups and code optimizations for Emergency Access
- Fixed a race-condition regarding jobs for Emergency Access
- Some other small changes like `allow(clippy::)` removals
Fixes #2925
This PR adds query logging support as an optional feature.
It is only allowed during development/debug builds, and will abort when
used during a `--release` build.
For this feature to be fully activated you also need to se an
environment variable `QUERY_LOGGER=1` to activate the debug log-level
for this crate, else there will be no output.
The reason for this PR is that sometimes it is useful to be able to see
the generated queries, like when debugging an issue, or trying to
optimize a query. Currently i always added this code when needed, but
having this a part of the code could benifit other developers too who
maybe need this.
instead of creating the parent folders to a sqlite database
vaultwarden should just exit if it does not.
this should fix issues like #2835 when a wrongly configured
`DATABASE_URL` falls back to using sqlite
Updated the logo's so the `V` is better visible.
Also the cog it self is better now, the previous version wasn't fully round.
These versions also are used with the PR to update the web-vault and use these logo's.
Also updated the images in the static folder.
This PR adds event/audit logging support for organizations.
By default this feature is disabled, since it does log a lot and adds
extra database transactions.
All events are touched except a few, since we do not support those
features (yet), like SSO for example.
This feature is tested with multiple clients and all database types.
Fixes #229
currently, if the admin guard fails the user will get a 404 page.
and when the session times out after 20 minutes post methods will
give the reason "undefined" as a response while generating the support
string will fail without any user feedback.
this commit changes the error handling on admin pages
* by removing the reliance on Rockets forwarding and making the login
page an explicit route that can be redirected to from all admin pages
* by removing the obsolete and mostly unused Referer struct we can
redirect the user back to the requested admin page directley
* by providing an error message for json requests the
`get_diagnostics_config` and all post methods can return a more
comprehensible message and the user can be alerted
* the `admin_url()` function can be simplified because rfc2616 has been
obsoleted by rfc7231 in 2014 (and also by the recently released
rfc9110) which allows relative urls in the Location header.
c.f. https://www.rfc-editor.org/rfc/rfc7231#section-7.1.2 and
https://www.rfc-editor.org/rfc/rfc9110#section-10.2.2
When a icon blacklist regex was configured to not check for a domain, it
still did a DNS lookup first. This could cause a DNS leakage for these
regex blocked domains.
This PR resolves this issue by first checking the regex, and afterwards
the other checks.
Fixes #2909
In the upcomming web-vault and other clients they changed the register
endpoint from `/api/accounts/register` to `/identity/register`.
This PR adds the new endpoint to already be compatible with the new
clients.
Fixes #2889
The DuckDuckGo email service is not supported for self-hosted servers.
This option is already hidden via the latest web-vault.
This PR also removes some server side headers.
Fixes #2828
It was possible to disable the admin security via the admin interface.
This is kinda insecure as mentioned in #2761.
This PR set this value as read-only and admin's need to set the correct ENV variable.
Currently saved settings which do override this are still valid though.
If an admin want's this removed, they either need to reset the config,
or change the value in the `config.json` file.
Fixes #2761
All uses of `get_random()` were in the form of:
`&get_random(vec![0u8; SIZE])`
with `SIZE` being a constant.
Building a `Vec` is unnecessary for two reasons. First, it uses a
very short-lived dynamic memory allocation. Second, a `Vec` is a
resizable object, which is useless in those context when random
data have a fixed size and will only be read.
`get_random_bytes()` takes a constant as a generic parameter and
returns an array with the requested number of random bytes.
Stack safety analysis: the random bytes will be allocated on the
caller stack for a very short time (until the encoding function has
been called on the data). In some cases, the random bytes take
less room than the `Vec` did (a `Vec` is 24 bytes on a 64 bit
computer). The maximum used size is 180 bytes, which makes it
for 0.008% of the default stack size for a Rust thread (2MiB),
so this is a non-issue.
Also, most of the uses of those random bytes are to encode them
using an `Encoding`. The function `crypto::encode_random_bytes()`
generates random bytes and encode them with the provided
`Encoding`, leading to code deduplication.
`generate_id()` has also been converted to use a constant generic
parameter as well since the length of the requested String is always
a constant.
Added a new endpoint which the currently beta client for at least
Android v2022.10.1 seems to be calling, and crashes with the response we
currently provide
Fixes #2890
Fixes #2891
Fixes #2892
Since v2022.9.x the org export uses a different endpoint.
But, since v2022.11.x this endpoint will return a different format.
See: https://github.com/bitwarden/clients/pull/3641 and https://github.com/bitwarden/server/pull/2316
To support both version in the case of users having an older client
either web-vault or cli this PR checks the version and responds using
the correct format. If no version can be determined it will use the new
format as a default.
if `SIGNUPS_VERIFY` is enabled new users that have been invited have
their onboarding flow interrupted because they have to first verify
their mail address before they can join an organization.
we can skip the extra verication of the email address when signing up
because a valid invitation token already means that the email address is
working and we don't allow invited users to signup with a different
address.
unfortunately, this is not possible with emergency access invitations
at the moment as they are handled differently.
Set SMTP_EMBED_IMAGES option to false if you don't want to attach images
to the mail.
NOTE: If you have customized the template files `email_header.hbs` and
`email_footer.hbs` you can replace `{url}/vw_static/` to `{img_url}`
to support both URL schemes
If you add a new user that has already been Invited to another
organization they will be Accepted automatically. This should not be
possible because they cannot be Confirmed until they have completed
their registration. It is also not necessary because their invitation
will be accepted automatically once they register.
- The Master Password Hint input has changed it's location to the
password update form. This PR updates the the code to process this.
- Also changed the `ProfileData` struct to exclude `Culture` and
`MasterPasswordHint`, since both are not used at all, and when not
defined they will also not be allocated.
Fixes #2833
