The conversation view used on phones built its own sanitizer hook that
only caught `<img src>` starting with a lowercase scheme and inline
style `url()`, and its iframe CSP always allowed http(s) images. With
remote content blocked (the default `ask` policy) uppercase schemes,
srcset, `<source>`, video posters and background attributes still
loaded, so trackers fired when a phone user opened a thread.
The view now uses the desktop viewer's `blockExternalResourcesOnNode`,
and the strict iframe CSP is shared as `emailIframeCsp()` and applied
whenever remote content is blocked, together with `no-referrer`.
A newsletter's List-Unsubscribe `mailto:` was sent from the user's own
account as written: every `to=` field became a recipient, and subject
and body were the sender's choice. The confirmation said only
"Unsubscribe from this sender?", so one click could mail
"I resign" to someone's boss and HR.
The unsubscribe now goes to the single address in the URL (a list of
addresses is refused and `to=` is ignored), the subject is kept on one
line and both subject and body are capped. Both confirmations show the
recipient, subject and body before anything is sent.
The theme sanitizer was a short denylist. It missed protocol-relative
`url(//host)`, schemes and keywords spelled with CSS escapes
(`url(\68ttps:...)`, `@\69mport`), `image-set("https://...")` and
`@font-face` sources, so a theme could load remote resources from the
app document. `validateThemeSelectors` was never called, so theme CSS
could also target `input[type=password]` or restyle real UI. The admin
routes and the zip validator only sanitized when the safety check
failed, and CSS returned by the `onThemeBeforeApply` plugin hook was
injected as-is.
- every `url()` except a `#fragment` and every `image-set()` is removed
with its arguments; a banned construct spelled with escapes drops the
whole stylesheet
- theme CSS keeps only `:root` and `.dark` rules (no nested rules),
plus `@font-face`, `@keyframes` and `@media`/`@supports` wrappers
- sanitizing is unconditional everywhere, including plugin-transformed
CSS; compiled token values may not contain escapes or `url()`
Attachment thumbnails were object URLs typed with the sender's MIME
type. Object URLs share the webmail origin, so an `image/svg+xml`
attachment opened from its thumbnail ("Open image in new tab", or
dragged to the tab strip) ran the sender's script as the webmail in
Chromium and Firefox. The drag-out URL and Files thumbnails had the
same shape.
`fetchBlobAsObjectUrl` now re-types every blob to an inert type, and
image thumbnails in mail and Files go through `imageBlobUrl`, which
hands an SVG over as a `data:` URL: it still renders in the `<img>`,
but opened on its own it gets an opaque origin.
Only the host API methods checked a plugin's permissions. Every hook
and slot the sandbox reported was wired as-is, so a plugin declaring no
permissions could rewrite recipients and body of every outgoing mail
through `onTransformOutgoingEmail`, replace displayed bodies through
`onRenderEmailBody`, or read all fetched mail through
`onEmailsFetched`. Admin-installed plugins are managed, so users never
saw a consent prompt for any of it.
Each hook and slot now maps to a permission in
`lib/plugin-sandbox/permissions.ts`, checked the same way as API calls
(declared, and managed or granted). Hooks that change what is sent need
`email:send`, display takeovers need `email:render-takeover`, and a hook
or slot missing from the table is refused. Logout and account switches
stay implicit so plugins can always drop per-account state.
The Stalwart admin check only bailed while `allowCustomJmapEndpoint` was
on. A context cookie minted while it was on keeps naming the
user-chosen server, and the envelope has no expiry, so after the admin
turned the switch off that server could still answer the admin probe
and mint a Bulwark admin session. The check now also requires the
cookie's server to be an admin-configured one.
The Stalwart admin check probed `x:Account/query`. Stalwart's default
tenant-admin role carries `sysAccount*`, so an administrator of a single
tenant passed and, in the default `auto` mode, was signed into the
Bulwark dashboard, which configures the whole instance for every
tenant. The `auto|password|off` switch added for #870 left the default
at `auto`, so the advisory never held on a default install.
The probe is now `x:Tenant/query`, which needs `sysTenantQuery`: by
default only the server-wide superuser holds it.
GetFile answered on the webmail origin with the Content-Type stored in
the WOPI token, which for a mail attachment comes straight from the
launch request and for a Files node is a user-settable property. No
CSP, nosniff or disposition was added, since `/api/*` skips the proxy.
A user could mint a token for their own blob typed `text/html` and have
a victim open the GetFile URL, running script in the webmail origin.
GetFile now always sends `application/octet-stream` with
`Content-Disposition: attachment`, `nosniff` and a sandboxing CSP, and
refuses browser loads (any `Sec-Fetch-Dest` other than `empty`). The
editor fetches it server-to-server, so it is unaffected.
A MAPI property in a TNEF attachment declares how many values follow.
The parser looped that many times even when a value could not be read
and nothing was consumed, so a 31-byte winmail.dat claiming 0xFFFFFFFF
values froze the tab for about 13 seconds on every open, longer with
more such properties. The count is now clamped to what the block can
hold and the loop stops as soon as a value consumes no bytes.
The Authentication-Results parser matched `dmarc=` and `dkim=` anywhere
in all headers joined together. A sender who used `dmarc=pass@...` as
the envelope address got a green DMARC pass for a message that failed,
because the server echoes that address in the SPF clause before its
real `dmarc=fail`. A sender could also add their own
Authentication-Results header to an unsigned message and get a DKIM and
DMARC pass. Calendar invites from such a message were shown as trusted.
The parser now splits each header on `;` outside quoted strings and
comments and reads the method only at the start of a part. DKIM, DMARC
and iprev come from the topmost header, which the receiving server
wrote; lower headers can only escalate SPF to a failure.
The origin gate in the proxy only covered `/api/auth/*`. The JMAP
passthrough, `/api/settings`, `/api/wopi/launch`,
`/api/plugin-approval-status`, `/api/fetch-ical`, `/api/translate`,
`/api/caldav/discover`, `/api/calendar-agenda` and `/api/webdav` all act
with the session cookie and had no origin check. A page on a same-site
sibling origin receives SameSite=Lax cookies, so it could run arbitrary
JMAP as the signed-in user, including sharing mailboxes to a third
account.
The proxy now gates every `/api/` route except the WOPI file endpoints
the office editor calls server-to-server, and each of these handlers
checks again. The proxy also decides on the decoded path, so an escaped
spelling such as `/api/%61uth/session` no longer skips the gate.
An impersonation handoff sealed `<target>%<master>` and the master
password into the session cookie, and `PUT /api/auth/session` hands that
cookie's contents to the browser on every restore. Anyone who received
a handoff could read a credential that signs in as every mailbox and
into Stalwart administration, with no expiry and no audit trail.
The impersonate route now uses the master credential server-side, once,
to create an app password on the target mailbox that expires after
8 hours. Only that app password reaches the cookies. Signing out
revokes it through a sealed grant cookie, and PUT refuses any session
that still holds the configured master password, so cookies minted
before this change are dropped instead of returned.
The URL guard behind fetch-ical, OAuth discovery, branding and the
server-side JMAP fetches blocked only the single address 127.0.0.1 plus
the RFC 1918, link-local and ULA subnets. Any logged-in user could make
the server fetch 127.0.0.2, 100.64.0.0/10, 198.18.0.0/15, multicast,
reserved space or a NAT64/6to4/Teredo address that wraps an internal
IPv4 address.
The telemetry guard already rejected all of these (GHSA-m7j8), so its
range check moves to `lib/security/ip-ranges.ts` and both guards use it.
`guardedLookup` inherits the wider list, so the connect-time rebinding
check covers the new ranges too.
The composer uploaded any file and learned about `maxSizeUpload` (per
blob) or `maxSizeAttachmentsPerEmail` (per message) only from a failed
upload or a failed send. Files over the per-file limit, and files that
would push the message past the per-message limit, are now skipped
with a message naming the limit. The limits come from the account the
message is composed in.
From 0.16.21 Stalwart fails the whole CalendarEvent/set with
`forbidden` when the account may not send scheduling messages, so an
event with attendees was not saved at all and the calendar showed only
"Failed to save event".
Such a refusal now surfaces as SchedulingDeniedError, and the calendar
names the server's reason and offers to save the event without sending
the invitations.
Bulwark asks for a 90-day push subscription, but Stalwart clamps
`expires` to 7 days. The background renewal ran once per page load,
so a tab or installed app left open for more than a week let the
subscription lapse and notifications stopped without a hint.
The renewal now runs at most once a day instead of once per load, and
also when the tab comes back to the foreground.
The share dialogs and the availability view listed principals with an
unbounded Principal/query whose ids fed one Principal/get. Once the
directory held more principals than `maxObjectsInGet` (500 on
Stalwart), the server refused the get with `requestTooLarge` and the
list came back empty. The listing now pages in steps of
`maxObjectsInGet`.
"Mark all as read" paged through the unread messages by position, but
each page it marked dropped out of the unread query, so every other
page was skipped: with more unread mail than one page (500), part of it
stayed unread. The next page now starts after the messages that were
left unread (those in excluded folders) only.
Neither mark-as-read loop checked the Email/set result. A refused
update made "mark folder as read" fetch the same unread ids forever;
it now stops with an error. Marking as spam and "not spam" also
ignored `notUpdated` (e.g. a Junk folder the server no longer knows)
and showed success; they now report the failure.
Moving mail into a shared account used `Email/copy` with
`onSuccessDestroyOriginal`. Stalwart before 0.16.15 destroys the wrong
id with it, and from 0.16.15 it destroys the source even when the copy
failed, so a move refused for quota or a missing message deleted the
original with no copy anywhere.
The copy now runs without that flag, and the source is destroyed with
a separate Email/set only after the copy was created. A failed destroy
is reported, leaving a duplicate rather than a loss.
From 0.16.6 Stalwart refuses FileNode names with `/<>:"\|?*` or a
reserved name such as `CON` ("Name contains a forbidden character."),
and publishes the rules as `forbiddenNameChars` / `forbiddenNodeNames`.
Bulwark did not read them: the error showed up only after an upload had
started, and it aborted the rest of a multi-file or folder upload.
- new folder, new text file and rename now say which characters or
names the server does not allow, before sending anything
- uploads store an accepted variant instead: forbidden characters
become `_` and a reserved name gets `_` appended
MIME types over 30 characters (most Office formats) were always stored
as `application/octet-stream`, because Stalwart before 0.16.6 refuses
them. From 0.16.6 the limit is 255, so the real type is kept there.
Stalwart before 0.16.6 implements an older JMAP File Storage draft
whose rights are only `mayRead`, `mayWrite` and `mayShare`. Bulwark
always sent the newer `mayAddChildren` / `mayRename` / `mayDelete` /
`mayModifyContent`, so sharing a file or folder failed there with
`Invalid permission "mayAddChildren"`, and the rights such a server
reported did not match what the UI checks.
When the account's filenode capability lacks `forbiddenNameChars`
(added by the same draft update), shares are now written with
`mayWrite` set from the finer rights, and `mayWrite` read from the
server is spread over them.
Copying a folder sent one FileNode with no content, which Stalwart
0.16.0 rejected ("size") and 0.16.6+ turned into an empty folder. Copy
now recreates the subtree level by level, and copies into another
account move the file contents with `Blob/copy`.
Every write (new folder, upload, rename, move, delete, copy, share)
went to the primary account, including for nodes of a shared drive,
which the listing names "<accountId>:<nodeId>". Inside a shared drive
the server read those ids as local ones and failed or touched the
wrong node. Writes now go to the account that owns the node, with the
bare id, and a move between accounts is refused instead of being sent.
Creates pass `onExists: "rename"`, so an upload or new folder that
clashes with an existing name becomes "name (2)" instead of failing.
Stalwart before 0.16.6 ignores that argument and refuses the clash, so
the client retries with a numbered name there.
CalendarEvent/query reads `after`/`before` as local date-times in the
request's `timeZone`, and Stalwart silently drops a `Z` or offset. The
proactive alert fetch and the contact activity panel sent UTC strings,
so their windows were shifted by the user's UTC offset (two hours in
Berlin in summer). UTC and offset bounds are now converted to the wall
clock of the request time zone before the query is sent.
With server-side expansion, a range with more than
`maxRecurrenceExpansions` occurrences (a daily series over ten years)
failed the whole query with `invalidArguments`, and the calendar showed
nothing. The query now falls back to the unexpanded series, which the
calendar expands on the client as it does for older servers.
The query also stopped at a fixed 1000 ids. Without an explicit limit it
now pages until the range is complete.
Stalwart rejects a whole ContactCard over one property it does not
know, and several reached it:
- the form sent `calendarUri`, `schedulingUri` and `freeBusyUri`,
so any contact with a calendar, scheduling or free/busy link failed
to save ("Invalid property.")
- vCard import sent `source`, and Pro mode left `localAccountId` on
the card
- vCard import wrote addresses as flat `street`/`locality` fields
instead of RFC 9553 `components`
The card now passes through `lib/jmap/contact-wire.ts` on its way to
the server: client-only fields are dropped, the links are written as
`calendars` / `schedulingAddresses` (and SOURCE as a `directories`
entry), flat addresses become components, and reads map the links back
for the form and detail view.
Clearing a field (nickname, notes, a link) never reached the server:
the form set it to `undefined`, which JSON drops, and an omitted
property keeps its value. Updates now send `null` for cleared keys.
Deleting an address book that still held contacts always failed with
`addressBookHasContents`. The delete now passes
`onDestroyRemoveContents`, which the confirm dialog already promised.
Stalwart advertises `maxDelayedSend` as a fixed 30 days, but its MTA
rejects any hold beyond `futureRelease`, 7 days by default. Scheduling
a message 8 to 30 days ahead passed the picker and then failed with
"501 5.5.4 Requested hold time exceeds maximum of 604800 seconds".
On Stalwart the advertised 30 days now counts as 7. When a submission
is still rejected for its hold time, the limit from the error message
replaces it for the session, and the composer shows the translated
"later than the server allows" message instead of the raw SMTP reply.
Linus Rath
Aaron Guise
Linus Rath