{ "schema_version": "1.6.1", "id": "RPI-CVE-2024-56640", "modified": "2026-10-09T11:39:44Z", "upstream": [ "CVE-2024-56640" ], "published": "2024-12-27T15:03:32Z", "summary": "net/smc: fix LGR and link use-after-free issue", "details": "In the Linux kernel, the following vulnerability has been resolved:\n\nnet/smc: fix LGR and link use-after-free issue\n\nWe encountered a LGR/link use-after-free issue, which manifested as\nthe LGR/link refcnt reaching 0 early and entering the clear process,\nmaking resource access unsafe.\n\n refcount_t: addition on 0; use-after-free.\n WARNING: CPU: 14 PID: 107447 at lib/refcount.c:25 refcount_warn_saturate+0x9c/0x140\n Workqueue: events smc_lgr_terminate_work [smc]\n Call trace:\n refcount_warn_saturate+0x9c/0x140\n __smc_lgr_terminate.part.45+0x2a8/0x370 [smc]\n smc_lgr_terminate_work+0x28/0x30 [smc]\n process_one_work+0x1b8/0x420\n worker_thread+0x158/0x510\n kthread+0x114/0x118\n\nor\n\n refcount_t: underflow; use-after-free.\n WARNING: CPU: 6 PID: 93140 at lib/refcount.c:28 refcount_warn_saturate+0xf0/0x140\n Workqueue: smc_hs_wq smc_listen_work [smc]\n Call trace:\n refcount_warn_saturate+0xf0/0x140\n smcr_link_put+0x1cc/0x1d8 [smc]\n smc_conn_free+0x110/0x1b0 [smc]\n smc_conn_abort+0x50/0x60 [smc]\n smc_listen_find_device+0x75c/0x790 [smc]\n smc_listen_work+0x368/0x8a0 [smc]\n process_one_work+0x1b8/0x420\n worker_thread+0x158/0x510\n kthread+0x114/0x118\n\nIt is caused by repeated release of LGR/link refcnt. One suspect is that\nsmc_conn_free() is called repeatedly because some smc_conn_free() from\nserver listening path are not protected by sock lock.\n\ne.g.\n\nCalls under socklock | smc_listen_work\n-------------------------------------------------------\nlock_sock(sk) | smc_conn_abort\nsmc_conn_free | \\- smc_conn_free\n\\- smcr_link_put | \\- smcr_link_put (duplicated)\nrelease_sock(sk)\n\nSo here add sock lock protection in smc_listen_work() path, making it\nexclusive with other connection operations.", "severity": [ { "type": "CVSS_V3", "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H" } ], "affected": [ { "package": { "ecosystem": "Raspberry Pi OS:12", "name": "linux" }, "versions": [ "1:6.1.47-1+rpt4" ], "ecosystem_specific": { "binaries": [ { "binary_name": "linux-image-6.1.0-rpi3-rpi-2712", "binary_version": "1:6.1.47-1+rpt4" }, { "binary_name": "linux-image-6.1.0-rpi3-rpi-v6", "binary_version": "1:6.1.47-1+rpt4" }, { "binary_name": "linux-image-6.1.0-rpi3-rpi-v7", "binary_version": "1:6.1.47-1+rpt4" }, { "binary_name": "linux-image-6.1.0-rpi3-rpi-v7l", "binary_version": "1:6.1.47-1+rpt4" }, { "binary_name": "linux-image-6.1.0-rpi3-rpi-v8", "binary_version": "1:6.1.47-1+rpt4" } ] }, "database_specific": { "per_version": { "1:6.1.47-1+rpt4": { "linux_commit": "655fc658a15ae7a6f37103754adb39ba52a9a14e", "upstream_version": "6.1.47", "assessment": "commit ancestry (strak rule)", "upstream_fixed": "6.1.120", "compiled_filter": false } } } } ], "references": [ { "type": "ADVISORY", "url": "https://www.cve.org/CVERecord?id=CVE-2024-56640" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/f502a88fdd415647a1f2dc45fac71b9c522a052b" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/0cf598548a6c36d90681d53c6b77d52363f2f295" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/673d606683ac70bc074ca6676b938bff18635226" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/6f0ae06a234a78ae137064f2c89135ac078a00eb" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/2c7f14ed9c19ec0f149479d1c2842ec1f9bf76d7" } ], "database_specific": { "source": "https://git.kernel.org/pub/scm/linux/security/vulns.git", "upstream_modified": "2026-08-05T11:52:36Z", "severity": { "source": "kernel CNA", "cvss_v3_score": 9.8, "cvss_v3_severity": "Critical", "scenario": "AV:N - The bug is in `smc_listen_work()`, the SMC server-side CLC handshake worker that is queued from `smc_tcp_listen_work()` for every inbound TCP connection carrying the SMC option, and it processes CLC Proposal/Confirm/Decline messages sent by the remote peer over routable TCP/IP. A remote host that can reach an AF_SMC listener drives the entire vulnerable path.\nAC:L - The attacker controls both sides of the race: `smc_clc_wait_msg()` (net/smc/smc_clc.c:784-794) accepts a CLC DECLINE at any point, and with the attacker-chosen `SMC_FIRST_CONTACT_MASK` bit set it calls `smc_lgr_terminate_sched()`, scheduling `__smc_lgr_terminate()` (which takes lock_sock and frees conn state) concurrently with the unlocked `smc_listen_work()` \u2192 `smc_conn_abort()` \u2192 `smc_conn_free()` path, whose `conn->freed` guard is an unsynchronized bitfield. The attacker can open unlimited parallel connections into the same link group on the separate `smc_hs_wq` and retry indefinitely at no cost.\nPR:N - The CLC handshake in `smc_listen_work()` runs entirely before any application-level authentication and requires no credentials \u2014 the attacker only needs to complete a TCP connection to the SMC listener. No capability, namespace, or local account is involved.\nUI:N - The listen worker is queued automatically from the TCP accept path by `smc_clcsock_data_ready()`/`smc_tcp_listen_work()`; no action by the local user or administrator is needed beyond running the SMC server that is already listening.\nS:U - The corruption is confined to kernel heap objects (`struct smc_link_group`, `struct smc_link`) within the same kernel security authority, with no crossing of a VM, IOMMU, or sandbox boundary.\nC:H - The duplicate `smcr_link_put()`/`smc_lgr_put()` drives the refcounts to zero early, so the `kzalloc`'d `struct smc_link_group` is `kfree()`d while `__smc_lgr_terminate()`, `smc_listen_work()`, and the LLC/CDC paths still dereference it \u2014 a classic use-after-free read of a generic kmalloc slab object that an attacker can reallocate and shape, leaking link-group state including `peer_systemid` and RDMA `rtokens` (remote keys/addresses).\nI:H - Post-free writes into the reallocated object include `lgr->sync_err`/`terminating` bitfields, `refcount_inc()`, and list/rb-tree manipulation, while the freed structure embeds `struct work_struct terminate_work`, `struct delayed_work free_work`, and per-link `link_down_wrk` \u2014 function pointers that `schedule_work()`/`smc_lgr_schedule_free_work()` will subsequently invoke, giving a heap-spray path to control-flow hijack.\nA:H - The commit documents real production `refcount_t: underflow`/`addition on 0` warnings in `smcr_link_put()` and `__smc_lgr_terminate()`; refcount saturation plus use of freed link-group memory yields oopses and kernel panics, and the condition is remotely and repeatably triggerable." } } }