Security: session hardening
Refresh token rotation with theft detection (family_id)
- Added family_id column to auth.sessions (migration 20260507000000_session_family.sql) grouping all tokens issued from the same login into a family
- On refresh, the new session inherits the parent's family_id
- If a revoked token is replayed (indicates the token was stolen after rotation), the entire family is immediately invalidated and a warning is logged — forcing re-authentication on all devices
SameSite=Strict on refresh cookie
- Access cookie stays SameSite=Lax (needed for top-level navigation)
- Refresh cookie upgraded to SameSite=Strict — it is only ever used for explicit POST to /api/auth/refresh, never via cross-site navigation
Refresh token TTL: 30 days → 7 days
- With rotation, active sessions auto-renew and effectively never expire
- Inactive sessions expire after 7 days instead of 30, reducing the theft window
After successful POST /api/s/{token}/verify, the server issues a
short-lived signed JWT cookie (oxi_share_unlock_<token>; HttpOnly;
SameSite=Lax; 1h TTL). Subsequent /api/s/{token} and
/api/s/{token}/download requests honour the cookie to bypass the
password gate, so password-protected file downloads work end-to-end.
- New module src/infrastructure/services/share_unlock_cookie.rs:
issue_jwt, verify_jwt, extract_from_cookie_header, build_set_cookie
+ 10 unit tests.
- New ShareService methods issue_unlock_jwt and
get_shared_link_with_unlock; trait method get_shared_link_by_token
delegates to a private fetch_share_resolved with an
allow_password_protected flag.
- access_shared_item, verify_shared_item_password, and
download_shared_file honour the unlock cookie.
Reuses OXICLOUD_JWT_SECRET, no new env var. Auth-token JWTs and
unlock-cookie JWTs cannot be confused: auth requires
username/email/role/jti claims, unlock has only sub/exp/iat.
Cross-share replay rejected via claims.sub == requested_token check.
a drop outside of the browser will:
- upload the file if only 1 file selected
- upload a .zip of the directory or multiple selection (browsers do not permit multiple upload yet)
note: I had to create a new handler because a post request to /api/batch/download is possible via JS but it will create a memory blob in the browser, during the drag action.
This may exhaust the browser's memory if heavy files
This will initiate zip creation from the server even if drop is canceled
The best approach is to add a handler supporting GET calls, this call will be triggered by the browser on drop action outside of it's window
a drop outside of the browser will:
- upload the file if only 1 file selected
- upload a .zip of the directory or multiple selection (browsers do not permit multiple upload yet)
note: I had to create a new handler because a post request to /api/batch/download is possible via JS but it will create a memory blob in the browser, during the drag action.
This may exhaust the browser's memory if heavy files
This will initiate zip creation from the server even if drop is canceled
The best approach is to add a handler supporting GET calls, this call will be triggered by the browser on drop action outside of it's window
The PROPFIND handlers (added in #153) handle the /caldav/{username}/ and
/carddav/{username}/ prefixes returned by discovery, but do so via an
expensive database call to distinguish usernames from UUIDs. The remaining
handlers (PUT, GET, DELETE, REPORT, PROPPATCH) don't handle the prefix
at all, causing 500 errors when standard clients (DAVx5, Apple Calendar,
Thunderbird) follow the discovery URLs.
This commit:
1. Adds a strip_username_prefix() helper that uses a UUID parse check
to distinguish collection IDs from usernames (no DB call needed).
2. Applies it to PUT, GET, DELETE, REPORT, and PROPPATCH handlers in
both caldav_handler.rs and carddav_handler.rs.
3. Refactors PROPFIND in both files to use the same UUID heuristic for
consistency, avoiding unnecessary database queries.
Includes 7 unit tests for the helper function.
Share links now point to /s/{token} (was /api/s/{token}) and render a
proper HTML page instead of raw JSON.
Changes:
- static/share.html: standalone public share page
- static/css/views/share-public.css: share page styles
- static/js/views/public/publicShare.js: client-side logic that fetches
share metadata via /api/s/{token}, handles password-protected shares,
and renders file download / folder info
- build.rs: include share.html in the HTML embed pipeline
- web/mod.rs: serve /s/{token} route (unauthenticated)
- share_dto.rs: generate URLs as /s/{token} instead of /api/s/{token}
- share_handler.rs: new download_shared_file() handler that validates the
share token and streams file content without requiring authentication
- routes.rs: mount GET /api/s/{token}/download (public, uses AppState)
Three changes to fix the immediate-logout issue reported by multiple
Docker users:
1. Add explicit `credentials: 'same-origin'` to the login fetch call.
This was the only fetch in the entire codebase missing it. While
modern browsers default to 'same-origin', some privacy configs or
older engines may default to 'omit', silently dropping Set-Cookie
headers from the login response.
2. Post-login cookie verification: after a successful login, the
frontend now checks that the CSRF cookie (non-HttpOnly, readable
by JS) was actually stored before redirecting. If the browser
rejected the cookies, a clear error message is shown explaining
the OXICLOUD_COOKIE_SECURE / HTTP mismatch.
3. Server-side diagnostic: the login handler now warns in logs when
Secure cookies are set on a request that didn't arrive via HTTPS
(no X-Forwarded-Proto: https header), pointing admins to the
OXICLOUD_COOKIE_SECURE=false fix.
Root cause: users who set OXICLOUD_BASE_URL=https://... (or have
OXICLOUD_COOKIE_SECURE=true) but access via plain HTTP get cookies
with the Secure flag, which browsers silently reject over HTTP.
Thumbnails are now keyed by blob_hash on disk so identical files share
a single set of thumbnails (icon/preview/large). For 4000 duplicate
files with the same content, this reduces thumbnail storage from 12,000
files to just 3.
Changes:
- get_thumbnail_path() keys by blob_hash instead of file_id
- get_thumbnail(), get_cached_thumbnail(), generate_all_sizes_background()
accept blob_hash parameter for disk dedup
- generate_all_sizes_background() fast path: if blob-hash thumbnails
already exist on disk, skip image processing entirely and just
populate moka cache for the new file_id
- delete_thumbnails() only invalidates moka cache (shared disk
thumbnails must not be deleted when one file is removed)
- delete_blob_thumbnails() added for GC; garbage_collect() now cleans
up orphaned thumbnail files alongside blob files
- External thumbnails (video frames) stored as ext-{file_id}.jpg
since they are client-generated and not dedup-able
- ThumbnailPort trait updated with blob_hash parameters
- All handler call sites updated (file_handler, preview_handler)
- Tests updated for new signatures
- Add utoipa v5 dependency with ToSchema derives on all REST API DTOs
- Annotate free-function handlers with #[utoipa::path] (trash, share, favorites, recent)
- Create ApiDoc struct with OpenApi derive registering 37 schemas across 7 tags
- Add generate-openapi binary outputting resources/gen/openapi.json
- Serve OpenAPI spec at GET /api/openapi.json (public, no auth)
- Add justfile with common dev commands (build, test, lint, check, openapi, db)
When file or folder rename/move operations failed in WebDAV handlers,
all errors were incorrectly converted to HTTP 500 Internal Server Error
using AppError::internal_error(). This masked specific error types:
- AlreadyExists errors should return 409 CONFLICT
- NotFound errors should return 404 NOT FOUND
- AccessDenied errors should return 403 FORBIDDEN
Changed error handling to use AppError::from() which preserves the
original DomainError type and maps to appropriate HTTP status codes.
Fixes handling of incorrect status code for file rename failure.
Adds storage quota checking to the WebDAV PUT handler, which was missing
while present in other upload handlers (regular upload and chunked upload).
The quota check happens after the file is spooled to a temp file (so we
know the exact size) but before it's moved to permanent storage. If the
quota is exceeded, the temp file is cleaned up and a 507 Insufficient
Storage error is returned.
Fixes#104
Fix two issues causing login loop after successful admin setup:
1. CSP blocking inline styles: The frontend JavaScript dynamically sets
inline styles (e.g., element.style.display = 'none') for UI state
management. The CSP header only allowed 'self' for style-src, blocking
these dynamic styles. Added 'unsafe-inline' to style-src directive.
2. Session refresh 401 errors: The cookie Secure flag defaulted to true
when OXICLOUD_BASE_URL was not set, causing cookies to not be sent
over HTTP in Docker deployments. Changed the default to false when
the base URL is not explicitly set to HTTPS, with clear logging to
guide users to set OXICLOUD_COOKIE_SECURE=true for production.
Fixes#203
The constant 10 * 1024 * 1024 * 1024 (10 GB) overflows on 32-bit systems
where usize is 32-bit (max ~4GB). This caused compilation failures on
ARMv7 architecture.
Fix by using architecture-appropriate limits:
- 64-bit: 10 GB (unchanged)
- 32-bit: 1 GB (safe maximum for 32-bit usize)
Fixes#206
The Nextcloud Android client compares ETags before and after upload to
verify its write landed. OxiCloud was returning the stable file UUID as
the ETag, which never changed on content updates, causing false
SYNC_CONFLICT errors on every upload.
Five fixes applied:
1. Thread blob_hash (SHA-256) through File entity, FileDto, all read/write
queries, and all WebDAV/PROPFIND responses as the ETag — changes on
every content update, no DB migration needed.
2. Honor X-OC-Mtime header: parse the client-supplied mtime and use it
for updated_at via COALESCE(to_timestamp($n), NOW()).
3. Disable phantom checksum capability (preferredUploadType/supportedTypes)
that the server never actually implemented, stopping retry loops.
4. Add nc:creation_time and nc:upload_time to PROPFIND responses.
5. Return oc-etag header in chunked upload MOVE (assemble) responses.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Fix all clippy lints (collapsible if, clone on Copy, needless borrow,
redundant bindings, unused params) and apply rustfmt across the codebase.
Update test mocks to match Uuid-based trait signatures.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace all 3 ImageFormat::WebP encode sites with JpegEncoder q=80.
Update fast-path to detect JPEG SOI instead of RIFF/WEBP magic.
Change file extension .webp -> .jpg, Content-Type headers, and
browser toBlob. Remove unused ImageFormat import and stale comments.
The webp feature stays for DECODING uploaded WebP images.
Solution A — Cache-first path:
- get_thumbnail now tries moka (RAM) → disk BEFORE any DB query
- Ownership was verified at creation time; UUIDv4 prevents enumeration
- 146-photo timeline: 0 SQL queries instead of 146 per page load
Solution C — ETag / If-None-Match short-circuit:
- Deterministic ETag 'thumb-{id}-{size}' on every response
- If browser sends matching If-None-Match → 304 with zero I/O
- Cache-Control: immutable prevents even conditional requests
Additional improvements:
- Remove moka TTL (was 600s); thumbnails are immutable, weight-only eviction
- generate_all_sizes_background populates moka after disk write;
first GET after upload served from RAM with zero disk I/O
- DB path only taken on cache miss for images needing generation
Backend:
- 204 response: add Cache-Control: no-store so browser never caches
'no thumbnail yet' — next GET after PUT upload gets the WebP
Frontend:
- _videoThumbCache (Map): persists fileId → URL across re-renders,
so switching tabs reuses cached URLs instantly (no re-decode)
- Render: videos with cached URL skip the 204/error/decode cycle
- After PUT succeeds: swap blob URL → server ?v=1 URL so blob is GC'd
- Concurrency throttle: max 3 simultaneous video decodes to avoid
overwhelming network + CPU when gallery has many videos
- Decode queue: pending videos processed as slots free up
Backend:
- get_thumbnail: for non-images, serve cached thumb or 204 (not 400)
- store_external_thumbnail: validate + re-encode to WebP + persist
- get_cached_thumbnail: memory → disk lookup without generation
- PUT /api/files/{id}/thumbnail/{size} endpoint for client uploads
- ThumbnailPort trait: add get_cached_thumbnail + store_external_thumbnail
Frontend (photos.js):
- On <img> error for video tiles, use <video> + <canvas> to extract
a frame at 25% duration using browser's native codec
- Show frame immediately via blob URL
- Fire-and-forget PUT to server for permanent caching
- Subsequent visits serve cached WebP instantly (no re-extraction)
Zero server-side dependencies — CPU distributed across clients.
- rate_limit: cache OXICLOUD_TRUST_PROXY_HEADERS in OnceLock<bool> to avoid
syscall on every request (~500ns → ~1ns)
- auth middleware: insert Arc<CurrentUser> instead of bare CurrentUser;
all 5 extractors now clone Arc (~1ns) instead of 4 Strings (~60-100ns)
- search_service: pre-compute query.to_lowercase() once before loops,
eliminating N redundant heap allocations per search
upload_file_inner now returns (FileDto, blob_hash) tuple, propagating
the BLAKE3 hash computed during the hash-on-write spool directly to
upload_file_with_thumbnails. This removes the unnecessary
get_blob_hash() SQL query that was fetching a value already in memory.
Saves ~0.5-2ms + 1 PG pool connection per image upload.
The cached-user-data path in checkAuthentication() fired resolveHomeFolder()
and loadFiles() concurrently with refreshUserData() using non-blocking .then()
chains. When the session cookie was expired, the folder/file API calls received
401 errors before the session could be refreshed. Now awaits session validation
before loading files.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add /remote.php/dav discovery endpoint for Android app server detection
- Add /index.php/204 connectivity check endpoint (returns 204 No Content)
- Redirect login flow to nc:// deep link for mobile credential delivery
- Support GET/HEAD on folders (NC clients use as existence checks)
- Recursive MKCOL to create missing parent directories
- Fix single-file PROPFIND returning empty multistatus response
- Strip instance suffix from preview fileId (e.g. "00000326ocnca")
- Add recommendations stub endpoint
Backend: new GET /api/photos endpoint with cursor-based pagination that
queries image/video files sorted by EXIF captured_at (falling back to
created_at), joining file_metadata for sort dates.
Frontend: dense photo grid grouped by day with lazy-loaded thumbnails,
IntersectionObserver infinite scroll, multi-select with batch
download/delete, and a full-screen lightbox with prev/next navigation,
EXIF metadata display, and download/favorite/delete toolbar.
Includes navigation wiring, CSS (with dark theme), and i18n translations
for all 9 locales.
Extract EXIF orientation, GPS coordinates, camera info, and timestamps
from uploaded images using kamadak-exif. Store metadata in a new
file_metadata PG table. Apply EXIF orientation to thumbnail generation
so images display correctly. Add /api/files/{id}/metadata endpoint.
- Add resolve_webdav_path() helper that prepends user's home folder
prefix to raw WebDAV paths before routing to handlers
- Rewrite handle_mkcol to walk path segments, creating folders with
proper parent_id chain instead of relying on path-only lookup
- Fix double path stripping in update_file_streaming where
get_parent_folder_id received an already-stripped parent path
instead of the full file path it expects
Fixes#165
Edouard Vanbelle