Re: [quicwg/base-drafts] Required state for retaining unacked RETIRE_CONNECTION_ID frames is unbound (#3509)

Mike Bishop <> Mon, 23 March 2020 20:12 UTC

Return-Path: <>
Received: from localhost (localhost []) by (Postfix) with ESMTP id 1D6C43A0D3D for <>; Mon, 23 Mar 2020 13:12:26 -0700 (PDT)
X-Virus-Scanned: amavisd-new at
X-Spam-Flag: NO
X-Spam-Score: -3.099
X-Spam-Status: No, score=-3.099 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_HIGH=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_IMAGE_ONLY_32=0.001, HTML_MESSAGE=0.001, MAILING_LIST_MULTI=-1, SPF_HELO_NONE=0.001, SPF_PASS=-0.001] autolearn=ham autolearn_force=no
Authentication-Results: (amavisd-new); dkim=pass (1024-bit key)
Received: from ([]) by localhost ( []) (amavisd-new, port 10024) with ESMTP id ziZHxsXTIBzr for <>; Mon, 23 Mar 2020 13:12:23 -0700 (PDT)
Received: from ( []) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by (Postfix) with ESMTPS id BE2EC3A0899 for <>; Mon, 23 Mar 2020 13:12:23 -0700 (PDT)
Date: Mon, 23 Mar 2020 13:12:22 -0700
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed;; s=pf2014; t=1584994342; bh=7o3++Etp4ivNOJD6hFUNmvaumCnyFvFxzFGsTgOemIo=; h=Date:From:Reply-To:To:Cc:In-Reply-To:References:Subject:List-ID: List-Archive:List-Post:List-Unsubscribe:From; b=l7FT6qyDggQbjOHxOGRIMMraZpGBlsw3bc7fc9BahkJNr044IiSZYW5xEhh9Wj/Fc aRBfjwrGk2WsQzwV9pfQ6PBm1zTs7lQhluGA7GURlgBxDKU0DDKooJweRkfFgYiajQ XoHMRHopxtzrTNToAGv1Xwsn95Z4Ka/4Uae6yEVY=
From: Mike Bishop <>
Reply-To: quicwg/base-drafts <>
To: quicwg/base-drafts <>
Cc: Subscribed <>
Message-ID: <quicwg/base-drafts/issues/3509/>
In-Reply-To: <quicwg/base-drafts/issues/>
References: <quicwg/base-drafts/issues/>
Subject: Re: [quicwg/base-drafts] Required state for retaining unacked RETIRE_CONNECTION_ID frames is unbound (#3509)
Mime-Version: 1.0
Content-Type: multipart/alternative; boundary="--==_mimepart_5e79182633d8_34663fb804ecd9601309a2"; charset="UTF-8"
Content-Transfer-Encoding: 7bit
Precedence: list
X-GitHub-Sender: MikeBishop
X-GitHub-Recipient: quic-issues
X-GitHub-Reason: subscribed
X-Auto-Response-Suppress: All
Archived-At: <>
X-Mailman-Version: 2.1.29
List-Id: Notification list for GitHub issues related to the QUIC WG <>
List-Unsubscribe: <>, <>
List-Archive: <>
List-Post: <>
List-Help: <>
List-Subscribe: <>, <>
X-List-Received-Date: Mon, 23 Mar 2020 20:12:50 -0000

So the issue here is fundamentally that you can induce the peer to send retransmittable frames which aren't subject to some kind of flow control, and this is the only instance we have of that.

STOP_SENDING and failing to acknowledge the RESET_STREAM appears to have a similar profile, except that the number of RESET_STREAMs you could have outstanding is bounded by the number of streams the peer has opened, which they control.

You could execute the same attack at a larger scale by choosing particular regions of stream data to never acknowledge, forcing the peer to keep them in memory for retransmission, but the peer could stop sending until that region gets through (applying backpressure) or simply reset the stream.  Since a RESET_STREAM is smaller than the data region and only one is needed per stream, this collapses into the previous case.

What's different here is that the endpoint being attacked doesn't choose whether to retire these CIDs and has no mechanism to apply backpressure.

It seems like there are a couple paths to resolving this.
- As you suggest, we could define an explicit frame that grants the additional CID space, and implementations would not send / could withhold this frame until the RCID is ACK'd.  This creates a backpressure mechanism and has the side benefit of allowing an implementation to vary the number of CIDs it wants to track over time, but the disadvantage of requiring an extra round trip before CIDs can be replaced in the normal case.
- We could mimic the "collapsing" behavior above by defining a frame that retires all CIDs up to a given sequence number.  Then the implementation under memory pressure chooses to retire a few extra CIDs and collapses the whole set waiting to be retransmitted to a single frame.

You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub: