{ "dataType": "CVE_RECORD", "dataVersion": "5.2", "cveMetadata": { "cveId": "CVE-2025-21710", "assignerOrgId": "416baaa9-dc9f-4396-8d5f-8c081fb06d67", "state": "PUBLISHED", "assignerShortName": "Linux", "dateReserved": "2024-12-29T08:45:45.752Z", "datePublished": "2025-02-27T02:07:23.112Z", "dateUpdated": "2026-08-05T11:53:49.495Z" }, "containers": { "cna": { "providerMetadata": { "orgId": "416baaa9-dc9f-4396-8d5f-8c081fb06d67", "shortName": "Linux", "dateUpdated": "2026-08-05T11:53:49.495Z" }, "descriptions": [ { "lang": "en", "value": "In the Linux kernel, the following vulnerability has been resolved:\n\ntcp: correct handling of extreme memory squeeze\n\nTesting with iperf3 using the \"pasta\" protocol splicer has revealed\na problem in the way tcp handles window advertising in extreme memory\nsqueeze situations.\n\nUnder memory pressure, a socket endpoint may temporarily advertise\na zero-sized window, but this is not stored as part of the socket data.\nThe reasoning behind this is that it is considered a temporary setting\nwhich shouldn't influence any further calculations.\n\nHowever, if we happen to stall at an unfortunate value of the current\nwindow size, the algorithm selecting a new value will consistently fail\nto advertise a non-zero window once we have freed up enough memory.\nThis means that this side's notion of the current window size is\ndifferent from the one last advertised to the peer, causing the latter\nto not send any data to resolve the sitution.\n\nThe problem occurs on the iperf3 server side, and the socket in question\nis a completely regular socket with the default settings for the\nfedora40 kernel. We do not use SO_PEEK or SO_RCVBUF on the socket.\n\nThe following excerpt of a logging session, with own comments added,\nshows more in detail what is happening:\n\n// tcp_v4_rcv(->)\n// tcp_rcv_established(->)\n[5201<->39222]: ==== Activating log @ net/ipv4/tcp_input.c/tcp_data_queue()/5257 ====\n[5201<->39222]: tcp_data_queue(->)\n[5201<->39222]: DROPPING skb [265600160..265665640], reason: SKB_DROP_REASON_PROTO_MEM\n [rcv_nxt 265600160, rcv_wnd 262144, snt_ack 265469200, win_now 131184]\n [copied_seq 259909392->260034360 (124968), unread 5565800, qlen 85, ofoq 0]\n [OFO queue: gap: 65480, len: 0]\n[5201<->39222]: tcp_data_queue(<-)\n[5201<->39222]: __tcp_transmit_skb(->)\n [tp->rcv_wup: 265469200, tp->rcv_wnd: 262144, tp->rcv_nxt 265600160]\n[5201<->39222]: tcp_select_window(->)\n[5201<->39222]: (inet_csk(sk)->icsk_ack.pending & ICSK_ACK_NOMEM) ? --> TRUE\n [tp->rcv_wup: 265469200, tp->rcv_wnd: 262144, tp->rcv_nxt 265600160]\n returning 0\n[5201<->39222]: tcp_select_window(<-)\n[5201<->39222]: ADVERTISING WIN 0, ACK_SEQ: 265600160\n[5201<->39222]: [__tcp_transmit_skb(<-)\n[5201<->39222]: tcp_rcv_established(<-)\n[5201<->39222]: tcp_v4_rcv(<-)\n\n// Receive queue is at 85 buffers and we are out of memory.\n// We drop the incoming buffer, although it is in sequence, and decide\n// to send an advertisement with a window of zero.\n// We don't update tp->rcv_wnd and tp->rcv_wup accordingly, which means\n// we unconditionally shrink the window.\n\n[5201<->39222]: tcp_recvmsg_locked(->)\n[5201<->39222]: __tcp_cleanup_rbuf(->) tp->rcv_wup: 265469200, tp->rcv_wnd: 262144, tp->rcv_nxt 265600160\n[5201<->39222]: [new_win = 0, win_now = 131184, 2 * win_now = 262368]\n[5201<->39222]: [new_win >= (2 * win_now) ? --> time_to_ack = 0]\n[5201<->39222]: NOT calling tcp_send_ack()\n [tp->rcv_wup: 265469200, tp->rcv_wnd: 262144, tp->rcv_nxt 265600160]\n[5201<->39222]: __tcp_cleanup_rbuf(<-)\n [rcv_nxt 265600160, rcv_wnd 262144, snt_ack 265469200, win_now 131184]\n [copied_seq 260040464->260040464 (0), unread 5559696, qlen 85, ofoq 0]\n returning 6104 bytes\n[5201<->39222]: tcp_recvmsg_locked(<-)\n\n// After each read, the algorithm for calculating the new receive\n// window in __tcp_cleanup_rbuf() finds it is too small to advertise\n// or to update tp->rcv_wnd.\n// Meanwhile, the peer thinks the window is zero, and will not send\n// any more data to trigger an update from the interrupt mode side.\n\n[5201<->39222]: tcp_recvmsg_locked(->)\n[5201<->39222]: __tcp_cleanup_rbuf(->) tp->rcv_wup: 265469200, tp->rcv_wnd: 262144, tp->rcv_nxt 265600160\n[5201<->39222]: [new_win = 262144, win_now = 131184, 2 * win_n\n---truncated---" } ], "metrics": [ { "cvssV3_1": { "version": "3.1", "vectorString": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:L/A:H", "baseScore": 8.2, "baseSeverity": "HIGH" }, "scenarios": [ { "lang": "en", "value": "AV:N - The bug is in the core TCP receive/output path (`tcp_select_window` via `tcp_data_queue`) and is triggered purely by the traffic pattern of a remote peer driving the receive socket into memory squeeze. Any routable TCP service — HTTP, SSH, ksmbd, nfsd, database — is reachable over the network.\nAC:L - The attacker controls both preconditions: skb truesize amplification lets a flood of small segments exhaust `sk_rmem_alloc`/`tcp_mem` while consuming little sequence space, which reliably leaves `tcp_receive_window()` above half the maximum at squeeze time — exactly the state that makes the `new_window >= 2 * rcv_window_now` test in `__tcp_cleanup_rbuf()` permanently unsatisfiable — and the attacker then simply declines to send zero-window probes, which is the only recovery path.\nPR:N - No authentication or privilege is needed; the condition is reached in `tcp_data_queue()` on established connections before any application-layer authentication, so any unauthenticated client of any TCP listener can drive it.\nUI:N - Triggering requires only that the target socket receive attacker-supplied data; no action by a local user or administrator is involved.\nS:U - The inconsistent window state and the resulting stall are confined to the kernel TCP stack and the sockets it owns, with no crossing of a security authority boundary.\nC:N - No uninitialized, out-of-bounds, or freed memory is read or transmitted; the defect is limited to bookkeeping of `tp->rcv_wnd`/`tp->rcv_wup` and produces no disclosure of kernel or other sockets' data.\nI:L - The attacker causes the connection's protocol state to be left incorrect — the stored receive window diverges from what was advertised, and the fix explicitly notes the pre-fix path transmits regressing `th->ack_seq` values that are \"wrong from a protocol viewpoint\" — pinning the peer's `snd_wnd` at a bogus zero; the modification is bounded and its consequences are not attacker-directed.\nA:H - The connection enters a permanent deadlock — the receive queue empties but no window update is ever sent and the peer never transmits again — indefinitely pinning a socket and, for blocking servers, a worker thread; repeated across many connections this exhausts a server's connection and thread pools, and TCP keepalive is disabled by default so nothing breaks the stall." } ] } ], "affected": [ { "product": "Linux", "vendor": "Linux", "defaultStatus": "unaffected", "repo": "https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git", "programFiles": [ "net/ipv4/tcp_output.c" ], "versions": [ { "version": "e2142825c120d4317abf7160a0fc34b3de532586", "lessThan": "b01e7ceb35dcb7ffad413da657b78c3340a09039", "status": "affected", "versionType": "git" }, { "version": "e2142825c120d4317abf7160a0fc34b3de532586", "lessThan": "1dd823a46e25ffde1492c391934f69a9e5eb574f", "status": "affected", "versionType": "git" }, { "version": "e2142825c120d4317abf7160a0fc34b3de532586", "lessThan": "b4055e2fe96f4ef101d8af0feb056d78d77514ff", "status": "affected", "versionType": "git" }, { "version": "e2142825c120d4317abf7160a0fc34b3de532586", "lessThan": "8c670bdfa58e48abad1d5b6ca1ee843ca91f7303", "status": "affected", "versionType": "git" } ] }, { "product": "Linux", "vendor": "Linux", "defaultStatus": "affected", "repo": "https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git", "programFiles": [ "net/ipv4/tcp_output.c" ], "versions": [ { "version": "6.6", "status": "affected" }, { "version": "0", "lessThan": "6.6", "status": "unaffected", "versionType": "semver" }, { "version": "6.6.76", "lessThanOrEqual": "6.6.*", "status": "unaffected", "versionType": "semver" }, { "version": "6.12.13", "lessThanOrEqual": "6.12.*", "status": "unaffected", "versionType": "semver" }, { "version": "6.13.2", "lessThanOrEqual": "6.13.*", "status": "unaffected", "versionType": "semver" }, { "version": "6.14", "lessThanOrEqual": "*", "status": "unaffected", "versionType": "original_commit_for_fix" } ] } ], "cpeApplicability": [ { "nodes": [ { "operator": "OR", "negate": false, "cpeMatch": [ { "vulnerable": true, "criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*", "versionStartIncluding": "6.6", "versionEndExcluding": "6.6.76" }, { "vulnerable": true, "criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*", "versionStartIncluding": "6.6", "versionEndExcluding": "6.12.13" }, { "vulnerable": true, "criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*", "versionStartIncluding": "6.6", "versionEndExcluding": "6.13.2" }, { "vulnerable": true, "criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*", "versionStartIncluding": "6.6", "versionEndExcluding": "6.14" } ] } ] } ], "references": [ { "url": "https://git.kernel.org/stable/c/b01e7ceb35dcb7ffad413da657b78c3340a09039" }, { "url": "https://git.kernel.org/stable/c/1dd823a46e25ffde1492c391934f69a9e5eb574f" }, { "url": "https://git.kernel.org/stable/c/b4055e2fe96f4ef101d8af0feb056d78d77514ff" }, { "url": "https://git.kernel.org/stable/c/8c670bdfa58e48abad1d5b6ca1ee843ca91f7303" } ], "title": "tcp: correct handling of extreme memory squeeze", "x_generator": { "engine": "bippy-1.2.0" } } } }