A video's preview and a picture's thumbnail are made inside the request that posts the message, and
nothing bounded how long either could take.
- The video preview filter also selects any frame from 5 seconds on. Rails' filter takes the second
frame it selects, which a video with a single keyframe and no scene change only gives at its end, so
ffmpeg decoded all of it.
- TimeLimitedVideoPreviewer gives ffmpeg 10 seconds of wall-clock time, kills it past that, and reports
a failed preview, so the message is posted without one.
- Pictures and videos above 250 megapixels, or whose size couldn't be read, get no preview: decoding
costs in proportion to the pixels, however small the file.
- The view shows a preview only if it was made when the message was posted. Its URL used to make it on
view, so a preview that failed or was skipped would be attempted again on every view. The cached
presentation's version goes up, so cached messages pick this up.
- A video's poster is made, when the message is posted, at the size the view shows it. The full-size WebP
made until now wasn't shown anywhere, and encoding it costs in proportion to the frame's pixels.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Bj8KnxpTf9sj2Ysa8aLAVa
Only a failure while Net::HTTP checks a reused idle socket means the
push service closed it, and a new connection can take the push. A
failure while connecting again, such as a certificate for another
name, or before the request is written on a new connection, is raised
as it is, so the subscription is invalidated as before. After the
request is written, a dropped connection raises ConnectionLost on new
connections too.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0142qgjggdJ2KDdGk7RF9Xm9
Since push delivery was pinned to the IP resolved and guarded for it,
every push opens a new TCP and TLS connection: Net::HTTP::Persistent
looks the host up itself and can't be pinned. The handshake is one or
two extra round trips for every push.
WebPush::Connections keeps the pinned connections open for 30 seconds
and hands one out again only to a delivery whose own, fresh resolution
returned the same address for the same host. Net::HTTP only ever
reconnects to that address, so no request goes to an address the guard
didn't just approve. A connection the push service closed while idle is
replaced before the push is written; once it's written, a dropped
connection raises ConnectionLost instead of sending the push twice or
invalidating the subscription. A delivery without a resolved IP is no
longer sent at all, and net-http-persistent goes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0142qgjggdJ2KDdGk7RF9Xm9
Each push notification carries the subscriber's unread room count as its
badge. The pool built it per subscription, loading the user and counting
their unread memberships: two queries for every subscriber, all in the job
before the deliveries reach the threads. With 1,000 subscribed members the
job spent ~180 ms and 2,000 queries there; with 5,000, a second.
The pool now counts the unread rooms of a whole batch with one grouped query
and hands each subscription its badge; nothing else in the notification needs
the user. The queries still run before the work is posted to the threads,
which run outside the Rails executor. Push::Subscription#notification still
counts by itself when no badge is given, as for the test notification.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0142qgjggdJ2KDdGk7RF9Xm9
Main strips a javascript: href from a preview node before it's parsed, so
tell legacy Trix attachments apart by their filename instead. escapeHTML
moved to the string helpers, and the unfurling system test now drives
the Lexxy editor.
A domain name ends in a word, which is what keeps it from reading as an
address. "0x7f.0.0.1" carries a dot and a letter, so the previous shape
check let it through while a browser fetched 127.0.0.1.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0186eyzivcTn6wqjEE4Wnxdt
A browser rewrites the many spellings of an address into one before it
fetches, so "http://2130706433/rooms/1" arrives at 127.0.0.1 while a
comparison here still reads the digits. A preview names a page on the
public internet, so require its host to look like a domain name and leave
the rewriting race alone.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0186eyzivcTn6wqjEE4Wnxdt
Ruby leaves a percent-escape in URI#host, so "https://%77ww.example.com"
read as a different host than the one Campfire answers on while a browser
unescaped it straight back to us. A host that carries an escape, or a
trailing dot, is now measured the way the browser will read it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0186eyzivcTn6wqjEE4Wnxdt
A preview belongs to the page it previews, so both URLs point somewhere
else. An absolute URL on our own host passed the scheme and host checks,
and every reader's browser fetched it with their session attached, which
turns a message into a GET request made on the reader's behalf.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0186eyzivcTn6wqjEE4Wnxdt
Ruby parses "https:/rooms/1" as an HTTPS URL with no host, and a browser
resolves it against whatever origin Campfire is served from, so the scheme
check alone still let a message body aim the preview at a path here.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0186eyzivcTn6wqjEE4Wnxdt
A link preview's link and image come from attributes on the message body,
which the composer fills in from the unfurl the server performed. A body
written by hand can put anything in those attributes, and the preview
partial rendered them as they were.
Keep the link and the image only when they parse as absolute http or https
URLs, so nothing in a message body can aim either one at another scheme or
at a path on this Campfire, and render the title and the description as
text.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0186eyzivcTn6wqjEE4Wnxdt
The bot HTTP API carries the bot key as a URL path segment
(/rooms/:room_id/:bot_key/...). config.filter_parameters redacts query
and form parameters but never path segments, so the key was written
verbatim to the request log (the "Started POST ..." line) and to any
log line echoing the pagination Link header.
Add a log formatter that redacts the bot-key path segment wherever it
appears in a formatted line, and wire it into the production logger.
* Guard push-subscription endpoints against SSRF
Web push delivery POSTed to the endpoint URL a user supplied when
registering a subscription, with no scheme, host, or private-network
check -- unlike the OpenGraph unfurl path, which already routes through
the shared SSRF address policy (surfguard). Any authenticated user could
register a subscription whose endpoint pointed at an internal address and
have the server fetch it on every chat message: blind SSRF for internal
recon and reachability probing, plus a thread-pool DoS on the delivery
pool.
Validate the endpoint when the subscription is saved: it must be HTTPS,
its host must belong to a known browser push service (allowlist), and it
must resolve to a public IP. On every delivery, re-resolve the host and
pin the connection to that public IP so a later DNS rebind can't redirect
the request to an internal address. If no public IP resolves at delivery
time -- a rebind, or a subscription that predates this validation --
delivery is skipped rather than falling back to re-resolving the raw
host.
Adds model, controller, and delivery-pinning tests, plus a DNS stub
helper for deterministic resolution in tests.
* Route push endpoint guarding through RestrictedHTTP::PrivateNetworkGuard
The endpoint SSRF check already delegated its private-network classification
to surfguard, but called Surfguard.resolve_public_ips directly rather than
through RestrictedHTTP::PrivateNetworkGuard -- the hostname-in, address-out
shim #241 established as the app's single guarded-outbound entry point and
that Opengraph::Fetch resolves through. Route push resolution through the same
guard so once-campfire keeps one place that resolves and classifies outbound
addresses. Behavior is unchanged: the guard raises Violation when a permitted
host resolves only to blocked addresses (previously an empty list -> nil) and
propagates Surfguard::Unresolvable for a host that resolves to nothing; both
map to a nil endpoint IP, which fails validation and skips delivery. The
allowlist, HTTPS/443 constraints, and per-delivery IP pinning are unchanged.
* Address push-SSRF review: defer DNS off enqueue path, disable proxy on pinned path, revalidate on re-registration
- Resolve the guarded endpoint IP lazily inside WebPush::Notification#deliver
(on the bounded delivery worker) instead of eagerly when the notification is
built on the serial enqueue path, so a slow resolver can't stall the push job
before any delivery starts. resolved_endpoint_ip only reads the already-loaded
endpoint attribute, so it is safe off the AR connection.
- Pin the delivery socket with an explicit nil proxy address so http_proxy/
https_proxy can't route the request through a proxy that re-resolves the host
and defeats the ipaddr pin.
- Revalidate an existing subscription on re-registration so a row predating
endpoint validation gets the same 422 as a fresh create instead of being kept
alive by touch.
- Regression tests: resolution deferred to delivery, pin survives proxy env,
legacy invalid row rejected with 422.
* Bound the web-push delivery queue (max_queue, not the ignored queue_size)
Concurrent::ThreadPoolExecutor takes :max_queue; :queue_size was silently
ignored, leaving the delivery backlog unbounded (max_queue: 0). A flood of
valid push subscriptions could accumulate queued deliveries without limit --
more acute now that each delivery task also resolves DNS. Using max_queue: 10000
activates the intended cap; overflow raises RejectedExecutionError under the
default :abort policy, which deliver_later already rescues (push is best-effort,
retried on the next message).
* Trim push-SSRF guard comments to match house style
Apply the review suggestions on the push-subscription SSRF guard: replace
the verbose rationale comments with terse one-liners (or drop them where
the code speaks for itself). No behavior change -- the Surfguard-shim
routing, per-delivery public-IP pin, and bounded delivery queue are
untouched.
* Take the SSRF address policy from surfguard instead of keeping our own copy
Four other apps carried this same classification and the five had drifted into
four different ideas of what "internal" means. It now comes from the surfguard
gem, which is their union. resolve and Violation keep their shapes, so the
opengraph callers are unchanged.
Two verdicts change.
SIIT (::ffff:0:0:0/96) is now recognised. It is the third way an IPv4 address
rides inside an IPv6 one and the only one Ruby has no predicate for --
ipv4_mapped?, ipv4_compat?, private?, loopback? and link_local? are all false
for ::ffff:0:a9fe:a9fe, so it fell through to the IPv6 branch unrecognised and
reached the metadata endpoint. Note the extra group: ::ffff:0:0:0/96 is not the
IPv4-mapped ::ffff:0:0/96 the guard already refused, and the two do not overlap.
The RFC 8215 local-use NAT64 block is now refused whole rather than decoded.
Reading its low 32 bits as an embedded IPv4 is only correct for a /96 Pref64;
the block can host any length from /32 to /96 and the position is not
recoverable from the address alone (RFC 6052 2.2), so the decode reads the
wrong octets. It is never globally routed, so refusing it costs nothing. The
well-known /96 is still decoded and re-checked, so DNS64 for public sites on
IPv6-only hosts keeps working.
Resolution moves from Resolv.getaddress to Resolv.getaddresses, so the guard
sees every address a host answers with rather than only the first.
* Distinguish a DNS lookup failure from a private-IP block in the guard
Advance the surfguard pin so resolve_public_ips raises Unresolvable when a
host resolves to nothing and returns an empty list only when it resolves to a
blocked address. The shim lets Unresolvable propagate as a lookup failure --
matching the old Resolv.getaddress behavior -- and reserves Violation for a
resolved-but-blocked address, so a transient DNS miss is no longer reported as
an SSRF attempt.
The guard uses IPAddr but relied on something else loading it first;
require it explicitly. And the rebinding tests lost their port assertion
when the matchers were loosened for newer Net::HTTP keyword args, so
check the port alongside the IP again.
The RFC1918, loopback, and link-local ranges in DISALLOWED_IPV4
duplicated the private?/loopback?/link_local? checks that run right
before the list scan, so they could never be the deciding factor. Keep
only the ranges the predicates don't catch.
Matches the fizzy guard: the local-use NAT64 prefix (64:ff9b:1::/48,
RFC 8215) embeds an IPv4 target in its low 32 bits just like the
well-known prefix, so run it through the same embedded-IPv4 recheck.
Local-use NAT64 to a public address now resolves (keeping unfurls
working for self-hosters on such networks) while local-use NAT64 to an
internal address stays blocked.
The local-use NAT64 prefix (64:ff9b:1::/48, RFC 8215) embeds an IPv4
target like the well-known prefix does, but at a deployment-chosen
position we can't extract, so block the whole range outright.
Also block the IPv6 benchmarking range (2001:2::/48, RFC 5180) to match
the IPv4 benchmarking block on 198.18.0.0/15.
The guard blocked the usual private, loopback, and link-local ranges (and
the IPv4-mapped/-compatible IPv6 forms), but let through NAT64, 6to4, and
Teredo addresses, which can point at an internal IPv4, and CGNAT.
Now it pulls the IPv4 out of a NAT64 address and checks that (so NAT64 to a
public site still works), blocks 6to4 and Teredo outright, and adds the
missing IPv4 and IPv6 ranges.
Adds ipv4_mapped? and ipv4_compat? checks to PrivateNetworkGuard.private_ip?
to block SSRF bypass attempts using IPv6 address formats like:
- ::ffff:169.254.169.254 (IPv4-mapped)
- ::169.254.169.254 (IPv4-compatible)
These formats could previously bypass the link_local? check since Ruby
treats them as IPv6 addresses, not IPv4.
Ref: HackerOne #3481701
* Bump Ruby to 3.4.5
* Update dependencies
* Adjust for Rails 8 and Ruby 3.5 API changes
* Mark params strings as mutable in prepapration for frozen strings in Ruby 3.5
* Update test for HTML5 sanitizer
With Rails 7.1 the HTML5 sanitizer became the default, this breakts this test because the old sanitizer used to delete unpermitted nodes, while the new one returns their content
The final string is safe, but different then it used to be in Rails 7.0
* Remove direct Turbo tesh helpers require & parallelize tests
* Fix Zeitwerk issues with rails extensions
* Update Resque setup for Redis 5+
* Remove unused views
* Remove GID v1 handler
GPT on behalf of DHH
Marcello Costagliola
Stanko K.R.
Rosa Gutierrez
Jeremy Daer
Donal McBreen
Kevin McConnell