fix(files): ending every session ends every API token, and the share links hold their folder

- A password reset or change, "log out everywhere" and a revoked untrusted password revoke all of the
  account's API tokens too (audited): one minted from a stolen session no longer outlives the answer.
- A bearer token or link grant with a non-ASCII character is refused (401/403), not a 500.
- Making a token's grants or a folder link holds the folder, so a racing delete cannot fail the
  foreign key; a link's view count is one atomic UPDATE.
- A folder link's card counts exactly the files its page lists (one `links.folder_files`).
- A file with no extension is served without a trailing dot.
- 0106's downgrade names a Hub folder so it never collides with one of the user's folders.
- The speed baseline reads share tokens through the app (the database keeps them hashed/encrypted).
- Tests: the token scope on the real session stack, the folder poster, the Hub refused for a link,
  0106's downgrade data step, 0107 without a key; docs/comments follow (three folder kinds, folder
  links, key change, the cleanup job's token warnings).
This commit is contained in:
2026-10-15 04:58:16 +02:00
parent 86712b6b0e
commit 827df7494c
23 changed files with 399 additions and 96 deletions
+9 -7
View File
@@ -1,7 +1,7 @@
"""The folders of the Files library — where each of a user's files sits.
A folder is a row, never a directory: on disk a file stays where the download layout put it (one
shared asset can be held by several users), so a rename or a move is a column update. Two kinds:
shared asset can be held by several users), so a rename or a move is a column update. Three kinds:
- a SOURCE folder is made automatically, one per origin a user has files from (YouTube, Plex clips,
uploads, one per other site). Its `source_key` names the origin; the client renders its name
@@ -13,11 +13,11 @@ shared asset can be held by several users), so a rename or a move is a column up
Every file of a user is in exactly one folder. The folder a new file lands in is decided here
(`source_key_for` + `source_folder`), at every place a job is created, so the rule lives once. A file
moves only into one of the user's own folders or back to its own source folder (`move`): a YouTube
file in the TikTok folder would be a lie the folder's name tells.
moves only into one of the user's own folders, their Hub folder, or back to its own source folder
(`move`): a YouTube file in the TikTok folder would be a lie the folder's name tells.
Retention is the folder's: a source folder keeps a file `download_retention_days` (the value when the
folder was made; the user can change it), Uploads and own folders keep it until it is deleted. A
folder was made; the user can change it), Uploads, own folders and the Hub keep it until it is deleted. A
done file carries its own `expires_at`, stamped by `done_expiry` when it finishes — unless the user
kept it — and its `ready_at`, the moment it finished (`stamp_done`).
"""
@@ -202,8 +202,9 @@ def own_folder(db: Session, user_id: int, folder_id: int, *, hold: bool = True)
def upload_target(db: Session, user_id: int, folder_id: int, *, hold: bool = True) -> FileFolder:
"""The folder an upload asked to land in, if it may: one of the user's own folders, or their
Uploads folder. A file put in a site's folder would claim an origin it does not have."""
"""The folder an upload asked to land in, if it may: one of the user's own folders, their Hub
folder, or their Uploads folder. A file put in a site's folder would claim an origin it does not
have."""
folder = own_folder(db, user_id, folder_id, hold=hold)
if folder.kind == FolderKind.SOURCE and folder.source_key != SourceKey.UPLOAD:
raise FolderError("file_folder_not_its_source")
@@ -225,7 +226,8 @@ def set_kept(db: Session, job: DownloadJob, kept: bool) -> None:
def move(db: Session, job: DownloadJob, folder: FileFolder) -> None:
"""Put a file in `folder` — one of its user's own folders, or the source folder it came from.
"""Put a file in `folder` — one of its user's own folders, their Hub folder, or the source folder
it came from.
The caller has checked that `folder` is the job's user's."""
if folder.kind == FolderKind.SOURCE and folder.source_key != source_key_for(db, job):
raise FolderError("file_folder_not_its_source")
+39 -3
View File
@@ -1,7 +1,8 @@
"""Public "share by link" helpers for the download center.
A DownloadLink is a capability URL: the unguessable token grants anyone read access to one
download's file on the login-free `/watch/{token}` page. So the token is never stored as written:
download's file, or to the files in one of its owner's own folders (`folder_files`), on the
login-free `/watch/{token}` page. So the token is never stored as written:
the link is found by its SHA-256, and the owner's copy is Fernet-encrypted — a database copy alone
opens nothing. Without that copy (no TOKEN_ENCRYPTION_KEY, or a key since changed) the link still
opens; only its address cannot be shown again, and the owner makes a new link to share.
@@ -21,9 +22,14 @@ from sqlalchemy.orm import Session
from app import security
from app.config import settings
from app.models import DownloadLink
from app.downloads import storage
from app.downloads.status import JobStatus
from app.models import DownloadJob, DownloadLink, FileFolder, MediaAsset
_GRANT_TTL = 6 * 3600 # a watch session's grant is good for 6h, then the viewer re-unlocks
# The most files a folder link lists: as many as the owner's own Files list shows (`routes.files.
# MAX_LIST`; a test holds the two equal — this layer may not import that one).
FOLDER_MAX = 500
def new_token() -> str:
@@ -48,6 +54,34 @@ def recover(link: DownloadLink) -> str | None:
return security.decrypt_optional(link.token_enc)
def in_folder(folder: FileFolder) -> tuple:
"""Which jobs a link to `folder` reaches: the done files in it NOW, the folder owner's. A file
moved out stops opening through the link at its next request; one moved in starts."""
return (
DownloadJob.folder_id == folder.id,
DownloadJob.user_id == folder.user_id,
DownloadJob.status == JobStatus.DONE,
)
def folder_files(db: Session, folder: FileFolder) -> list[tuple[DownloadJob, MediaAsset]]:
"""What a link to `folder` shows — its page lists exactly these, and its link card counts them:
each file `in_folder` whose asset is ready and whose bytes are on disk (so the list never offers
a dead row), newest first, at most `FOLDER_MAX`."""
rows = db.execute(
select(DownloadJob, MediaAsset)
.join(MediaAsset, MediaAsset.id == DownloadJob.asset_id)
.where(*in_folder(folder), MediaAsset.status == "ready")
.order_by(DownloadJob.ready_at.desc().nulls_last(), DownloadJob.id.desc())
.limit(FOLDER_MAX)
).all()
return [
(job, asset)
for job, asset in rows
if storage.safe_abs_path(settings.files_root, asset.rel_path) is not None
]
def is_expired(link: DownloadLink) -> bool:
return link.expires_at is not None and link.expires_at <= datetime.now(timezone.utc)
@@ -81,7 +115,9 @@ def check_grant(token: str, grant: str | None, password_version: int) -> bool:
return False
if exp < datetime.now(timezone.utc).timestamp():
return False
return hmac.compare_digest(sig, _sign(f"{token}.{exp}.{password_version}"))
# As bytes: compare_digest refuses (TypeError, a 500) a str with anything non-ASCII in it, and
# `sig` is whatever the URL carried.
return hmac.compare_digest(sig.encode(), _sign(f"{token}.{exp}.{password_version}").encode())
# --- serialization -------------------------------------------------------------------------
+1 -9
View File
@@ -11,13 +11,11 @@ import html
import re
from pathlib import Path
from sqlalchemy import func, select
from sqlalchemy.orm import Session
from app.config import settings
from app.downloads import links as linksmod
from app.downloads import storage
from app.downloads.status import JobStatus
from app.models import DownloadJob, DownloadLink, FileFolder, MediaAsset
_OG_BLOCK = re.compile(r"<!--OG:START-->.*?<!--OG:END-->", re.DOTALL)
@@ -157,13 +155,7 @@ def _folder_tags(db: Session, link: DownloadLink, url: str) -> str | None:
folder = db.get(FileFolder, link.folder_id)
if folder is None or not folder.name:
return None
count = db.execute(
select(func.count(DownloadJob.id)).where(
DownloadJob.folder_id == folder.id,
DownloadJob.user_id == folder.user_id,
DownloadJob.status == JobStatus.DONE,
)
).scalar_one()
count = len(linksmod.folder_files(db, folder)) # the page's own list, so the two agree
title = folder.name.strip()
about = f"{count} file{'' if count == 1 else 's'} · Shared via Siftlode"
return "\n ".join(
+5 -1
View File
@@ -294,9 +294,13 @@ def served_media_type(media_kind: str | None, container: str | None) -> str | No
def download_filename(display_name: str, container: str | None, path: Path) -> str:
"""The Content-Disposition filename for a served download: the cleaned display name carrying a single
correct extension — the asset container, else the file's own suffix — without doubling it when the
name already ends in that extension. Shared by the private download and the public watch endpoints."""
name already ends in that extension. A file with no extension at all is served without one —
never as `name.` (Windows drops a trailing dot, so the saved file would not be the one named).
Shared by the private download and the public watch endpoints."""
ext = container or path.suffix.lstrip(".")
base = display_filename(display_name)
if not ext:
return base
if base.lower().endswith(f".{ext.lower()}"):
base = base[: -(len(ext) + 1)]
return f"{base}.{ext}"