{ "schema_version": "1.6.1", "id": "RPI-CVE-2026-97409", "modified": "2026-10-09T11:39:44Z", "upstream": [ "CVE-2026-97409" ], "published": "2026-09-24T16:08:25Z", "summary": "nvme-fc: Do not cancel requests in io target before it is initialized", "details": "In the Linux kernel, the following vulnerability has been resolved:\n\nnvme-fc: Do not cancel requests in io target before it is initialized\n\nA new nvme-fc controller in CONNECTING state sees admin request timeout\nschedules ctrl->ioerr_work to abort inflight requests. This ends up\ncalling __nvme_fc_abort_outstanding_ios() which aborts requests in both\nadmin and io tagsets. In case fc_ctrl->tag_set was not initialized we\nsee the warning below. This is because ctrl.queue_count is initialized\nearly in nvme_fc_alloc_ctrl().\n\nnvme nvme0: NVME-FC{0}: starting error recovery Connectivity Loss\nINFO: trying to register non-static key.\nThe code is fine but needs lockdep annotation, or maybe\nlpfc 0000:ab:00.0: queue 0 connect admin queue failed (-6).\nyou didn't initialize this object before use?\nturning off the locking correctness validator.\nWorkqueue: nvme-reset-wq nvme_fc_ctrl_ioerr_work [nvme_fc]\nCall Trace:\n \n dump_stack_lvl+0x57/0x80\n register_lock_class+0x567/0x580\n __lock_acquire+0x330/0xb90\n lock_acquire.part.0+0xad/0x210\n blk_mq_tagset_busy_iter+0xf9/0xc00\n __nvme_fc_abort_outstanding_ios+0x23f/0x320 [nvme_fc]\n nvme_fc_ctrl_ioerr_work+0x172/0x210 [nvme_fc]\n process_one_work+0x82c/0x1450\n worker_thread+0x5ee/0xfd0\n kthread+0x3a0/0x750\n ret_from_fork+0x439/0x670\n ret_from_fork_asm+0x1a/0x30\n \n\nUpdate the check in __nvme_fc_abort_outstanding_ios() confirm that io\ntagset was created before iterating over busy requests. Also make sure\nto cancel ctrl->ioerr_work before removing io tagset.", "severity": [ { "type": "CVSS_V3", "score": "CVSS:3.1/AV:A/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": null, "compiled_filter": false } } } } ], "references": [ { "type": "ADVISORY", "url": "https://www.cve.org/CVERecord?id=CVE-2026-97409" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/5488e2706c58444a24e15773237dcc0bedcd771f" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/da930308b4bc24c335574e85ca84ff70a988cf05" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/dd39064d2948b14dbe4f364fe4b3c9cfc632cc21" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/ee38469f88492df99e1d97f03aa40ecfd218934f" } ], "database_specific": { "source": "https://git.kernel.org/pub/scm/linux/security/vulns.git", "upstream_modified": "2026-10-03T11:04:09Z", "severity": { "source": "kernel CNA", "cvss_v3_score": 8.8, "cvss_v3_severity": "High", "scenario": "AV:A - A rogue NVMe-FC target stalls or errors the admin-queue FCP from nvmf_connect_admin_queue() in nvme_fc_create_association(); nvme_fc_fcpio_done() queues ctrl->ioerr_work, and nvme_fc_ctrl_ioerr_work() -> nvme_fc_error_recovery() (CONNECTING) reaches the patched __nvme_fc_abort_outstanding_ios(). FC/FCoE is a non-routable SAN fabric, so Adjacent.\nAC:L - nvme_fc_alloc_ctrl() sets queue_count > 1 before nvme_alloc_io_tag_set(), so CONNECTING has a zeroed embedded tag_set. The attacker creates that window by not completing the admin Connect FCP until nvme_fc_timeout()/nvme_fc_fcpio_done() runs; __nvme_fc_abort_outstanding_ios() then deterministically calls blk_mq_tagset_busy_iter(&ctrl->tag_set).\nPR:N - The attacker is an unauthenticated NVMe-FC N_Port; the host transport does not authenticate the remote port (fabric zoning only). nvmf-autoconnect writes /dev/nvme-fabrics as root after FC_EVENT=nvmediscovery, so the fabric peer needs no host account or capability.\nUI:N - nvme_fc_register_remoteport() calls nvme_fc_signal_discovery_scan(), which emits KOBJ_CHANGE with FC_EVENT=nvmediscovery; nvmf-autoconnect and nvmefc-boot-connections then write /dev/nvme-fabrics with no interactive administrator. The admin-FCP stall is the attacker's own action.\nS:U - srcu_read_lock() on the uninitialized tags_srcu and the tag_set teardown UAF corrupt host kernel per-CPU data and blk-mq tag memory in the same kernel; no VM, IOMMU, or sandbox boundary is crossed.\nC:H - blk_mq_tagset_busy_iter() takes srcu_read_lock(&ctrl->tag_set.tags_srcu) on a kzalloc-zeroed tag_set, so __srcu_read_lock() this_cpu_inc()s a NULL srcu_ctrp. After nvme_alloc_io_tag_set(), nvme_fc_create_io_queues()'s error path also calls nvme_remove_io_tag_set() while nvme_reset_wq still iterates the live tag_set \u2014 a heap UAF.\nI:H - The NULL-srcu this_cpu_inc corrupts per-CPU words at offset 0 (x86 current_task is PER_CPU_CACHE_HOT). Concurrent blk_mq_free_tag_set() versus blk_mq_tagset_busy_iter() is a UAF of tag_set->tags and the SRCU sda; slab reuse of those objects yields a write/control-flow primitive.\nA:H - The report is lockdep in blk_mq_tagset_busy_iter() from __nvme_fc_abort_outstanding_ios() on the zeroed tag_set. Without lockdep, this_cpu_inc of a NULL per-CPU pointer and the nvme_remove_io_tag_set() UAF oops or panic the host and take down the NVMe-FC storage path." } } }