Uh oh!
There was an error while loading. Please reload this page.
[Flight Reply] Early bailout if backing entry for Blob deserialization is not a Blob - #36055
Merged
Conversation
react-sizebot
commented
Mar 17, 2026
Comparing: c80a075...a46c5af Critical size changesIncludes critical production bundles, as well as any change greater than 2%:
Significant size changesIncludes any change greater than 0.2%: (No significant changes) |
eps1lon
marked this pull request as ready for review
March 17, 2026 10:45
unstubbable
approved these changes
Mar 17, 2026
ericadalton124-stack
commented
Mar 17, 2026
via email
Thanks for reporting this I'll take a look at it …On Tue, Mar 17, 2026, 10:51 AM eps1lon ***@***.***> wrote:
Merged #36055 <#36055> into main.
—
Reply to this email directly, view it on GitHub
<#36055 (comment)>, or
unsubscribe
<https://github.com/notifications/unsubscribe-auth/B7BPWU64U3DGUFBWUDP5NQ34REUZXAVCNFSM6AAAAACWUOH5H2VHI2DSMVQWIX3LMV45UABCJFZXG5LFIV3GK3TUJZXXI2LGNFRWC5DJN5XDWMRTGYZTSMJSHA4TAMA>
.
You are receiving this because you are subscribed to this thread.Message
ID: ***@***.***>
|
SyMind pushed a commit
to SyMind/react
that referenced
this pull request
Aug 11, 2026
* [Flight] Restore standard React version placeholder * [Flight Reply] Early bailout if backing entry for Blob deserialization is not a Blob (react#36055) Co-authored-by: Hendrik Liebau <mail@hendrik-liebau.de> (cherry picked from commit 12ba7d8) * [Flight] Avoid consuming cyclic models multiple times Co-authored-by: "Sebastian \"Sebbie\" Silbermann" <sebastian.silbermann@vercel.com> (cherry picked from commit 672b242) * [FlightReply] Type hardening and performance improvements Co-authored-by: Hendrik Liebau <mail@hendrik-liebau.de> (cherry picked from commit 795203e) * [FlightReply] Don't drop FormData entries in `decodeReplyFromBusboy` (react#36468) Fixes a regression from react#36425 where referenced `FormData` entries can be dropped by `decodeReplyFromBusboy` when files are interleaved with text fields in the payload. `decodeReplyFromBusboy` queues text fields that arrive while a file is being streamed and flushes them after the last file's `'end'`, working around busboy emitting `'end'` deferred relative to subsequent `'field'` events. With multiple files interleaved with text, this loses the relative order of the affected text entries. The reorder was a long-standing but invisible issue — entries came back in the wrong order but were all present — until react#36425 tightened how referenced FormData entries are collected from the backing store to rely on them being contiguous. With that assumption violated, referenced FormDatas can now come back with some entries dropped. The pattern is most easily surfaced through `useActionState` actions that return the submitted `FormData` as part of their state. This replaces the tail-flush with a linked list of pending files. Text fields that arrive while a file is in flight are queued on the tail file's `queuedFields`; fields that arrive when the list is empty resolve immediately. `flush()` walks from the head, resolving each completed file followed by its queued fields, and stops at the first file that hasn't ended yet. The backing FormData now matches the payload's order, restoring the contiguity assumption (and fixing the long-standing reorder as a side effect). The same change is applied to all five copies in `react-server-dom-{webpack,turbopack,parcel,esm,unbundled}`. Two new tests cover the multi-file interleave. fixesvercel/next.js#93822 (cherry picked from commit b91823e) * [FlightReply] Performance improvements when decoding (react#37087) This fixes security vulnerabilities in Server Functions. (cherry picked from commit 1dd4ecb) * [Flight Reply] Align Rspack decoders with upstream changes Mirror the selected Reply changes into the Rspack-owned browser, edge, and Node adapters after the dependency-closed upstream backports. Forward caller-provided array size limits, preserve multipart field/file order, and settle failed async iterators without recursive error re-entry. Add public behavior coverage across every Rspack decoder, action-selection path, cyclic collection type, iterator settlement, and Busboy ordering direction. This is a tactical source-parity change for the existing proposal branch; it does not make that branch current with React main. --------- Co-authored-by: Sebastian "Sebbie" Silbermann <sebastian.silbermann@vercel.com> Co-authored-by: Hendrik Liebau <mail@hendrik-liebau.de>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for freeto join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Additional security hardening to bail out early for malformed Server Action payloads.
The
$B(Blob) case inparseModelStringreturned whateverFormData.get()returned without validating its type. SinceFormData.get()returns either a string or a File/Blob, an attacker could store a large string in a FormData slot and reference it via$B, bypassing thebumpArrayCountsize guard that applies to regular string values. However, this does not lead to a real attack vector because it doesn't produce amplification. For that, it would need to be combined with regular references in nested arrays, which are covered by the array counting mitigation.As a defense-in-depth mitigation, this adds an
instanceof Blobcheck that rejects non-Blob backing entries. Bumping the array count for Blob size is not necessary because real Blobs are opaque handles — they don't expand during common operations likeJSON.stringify(which produces{}),.flat(), or.toString(). The data only materializes through explicit reads like.text()or.arrayBuffer(), and since all references to the same$BID resolve to the same object, there is no memory duplication.