kernel/git/stable/linux.git/drivers/nvme, branch linux-rolling-lts

nvmet-tcp: reject unsolicited H2CData PDUs

2026-09-11T09:49:15+00:00

commit db62b35cbca052860c519cbcabe7650708528738 upstream. nvmet_tcp_handle_h2c_data_pdu() accepts an H2CData PDU after only checking that its TTAG is a valid in-range command index and that the command's data buffers are mapped. It never checks that the target has actually solicited that data by sending an R2T for the command. A remote host can abuse this. It submits a write command that takes the R2T path and, before the target transmits the R2T, sends an H2CData PDU for that command's tag. The data completes the command early, and when the command then fails synchronously (e.g. a length mismatch caught by nvmet_check_transfer_len()), it is completed a second time. Each completion calls nvmet_tcp_queue_response(), so the same command is added to queue->resp_list twice while it is still linked; the second llist_add() makes the node point to itself (lentry->next == lentry). nvmet_tcp_process_resp_list() then walks that self-referential node and adds the command to resp_send_list twice. With CONFIG_DEBUG_LIST this trips the "list_add double add" check (kernel BUG); without it the loop never terminates and the nvmet_tcp workqueue wedges (soft-lockup). It is remotely triggerable and needs no authentication on an allow_any_host subsystem. Track whether an R2T has been transmitted for a command and reject an H2CData PDU that arrives before it. The flag is cleared on command reuse (nvmet_tcp_get_cmd() zeroes cmd->flags) and stays set across the multiple H2CData PDUs of a single solicited transfer. Fixes: 872d26a391da ("nvmet-tcp: add NVMe over TCP target driver") Cc: stable@vger.kernel.org Reviewed-by: Sagi Grimberg Signed-off-by: Shivam Kumar Signed-off-by: Keith Busch Signed-off-by: Greg Kroah-Hartman

nvmet-tcp: fix out-of-bounds write when receiving an over-long PDU

2026-09-11T09:49:15+00:00

commit 14cc5a7e77731497d5bea70f3bb05df7eda982e4 upstream. nvmet_tcp_try_recv_pdu() reads a PDU header into the fixed 128-byte queue->pdu union, then computes the remaining payload length as queue->left = hdr->hlen - queue->offset + hdgst; and reads that many more bytes into &queue->pdu + queue->offset, without ever bounding the result against sizeof(queue->pdu). A struct nvme_tcp_icreq_pdu is itself 128 bytes, exactly the size of the union. Once a header digest has been negotiated (hdgst = 4), a second ICReq passes the hlen == nvmet_tcp_pdu_size() check but yields queue->left = 128 - 8 + 4 = 124, so bytes 8..132 are written into the 128-byte buffer -- 4 bytes past its end, over queue->hdr_digest and queue->data_digest. Those bytes are attacker-controlled (an ICReq carries no digest), and the duplicate ICReq is only rejected later, after the overflow. A remote unauthenticated host can thus corrupt kernel memory adjacent to the receive buffer. Reject any PDU whose declared length would read past the end of queue->pdu before the second recv. Fixes: 872d26a391da ("nvmet-tcp: add NVMe over TCP target driver") Assisted-by: Claude:claude-opus-4-8 Signed-off-by: Shivam Kumar Cc: stable@vger.kernel.org Reviewed-by: Sagi Grimberg Signed-off-by: Keith Busch Signed-off-by: Greg Kroah-Hartman

nvmet-auth: Synchronize timeout work during SQ teardown

2026-09-11T09:49:15+00:00

commit eaa948c0e19b1bb2d93262207bca0c3d19cc3406 upstream. nvmet_auth_sq_free() cancels auth_expired_work with cancel_delayed_work(). If the work has already started, cancellation does not wait for the callback. Transport teardown can consequently free or reuse the queue containing struct nvmet_sq while nvmet_auth_expired_work() still accesses that SQ. Add a teardown-specific helper that synchronously drains the delayed work before freeing authentication state, and use it from nvmet_sq_destroy(). Keep the non-synchronous helper for in-band authentication state cleanup, where the SQ owner remains alive. Fixes: 1a70200f404a ("nvmet-auth: expire authentication sessions") Cc: stable@vger.kernel.org Signed-off-by: Kazuki Hanai Reviewed-by: Sagi Grimberg Reviewed-by: Christoph Hellwig Signed-off-by: Keith Busch Signed-off-by: Greg Kroah-Hartman

nvme: add missing SRCU grace period in error path

2026-09-11T09:49:14+00:00

commit ef248d5de4469fb6bbaf8dbe0c4c47800080d648 upstream. nvme_alloc_ns() error path at out_unlink_ns removes ns from the namespace head siblings list with list_del_rcu(&ns->siblings) but does not wait for SRCU readers before freeing the namespace struct. Multipath code iterates the head->list under srcu_read_lock() in nvme_find_path() and nvme_mpath_revalidate_paths(), so a concurrent reader can still hold a reference to ns when kfree(ns) runs. The normal removal path in nvme_ns_remove() correctly calls synchronize_srcu(&ns->head->srcu) after list_del_rcu() to wait for in-progress readers. Add the same grace period in the error path. Fixes: ed754e5deeb1 ("nvme: track shared namespaces") Cc: stable@vger.kernel.org Signed-off-by: Tristan Madani Reviewed-by: Sagi Grimberg Reviewed-by: John Garry Reviewed-by: Christoph Hellwig Signed-off-by: Keith Busch Signed-off-by: Greg Kroah-Hartman

nvme-tcp: check the data direction of a C2HData PDU

2026-09-11T09:49:14+00:00

commit f83af377c148f6ad94b41c0e8313f12adf45e1c1 upstream. nvme_tcp_handle_c2h_data() finds the request by command id and checks that it has a payload, but it does not check that the command asked for data to be read. A controller that answers a write command with C2HData therefore reaches nvme_tcp_recv_data(), where _copy_to_iter() hits WARN_ON_ONCE(i->data_source) and returns 0. The receive path turns that into -EFAULT and resets the controller. No data is copied, so this is not memory corruption. What a controller gets is a kernel warning it can raise at will, which is fatal on a host booted with panic_on_warn. The send path already knows the direction - it consults rq_data_dir() when it builds a command - and nvme_tcp_handle_r2t() checks the length and the offset of the request it names. The C2HData path does not check the direction at all. Reject a C2HData PDU whose command is not a read. Rejecting it fails the command and resets the controller, as the neighbouring check in this function does; what goes away is the warning. [ 6.885580] ------------[ cut here ]------------ [ 6.886457] WARNING: lib/iov_iter.c:193 at _copy_to_iter+0x289/0x1330, CPU#0: kworker/0:1H/71 [ 6.888137] CPU: 0 UID: 0 PID: 71 Comm: kworker/0:1H Not tainted 7.2.0-rc5-NVMETCP-gf5098b6bae76 #1 PREEMPT(lazy) [ 6.891165] Workqueue: nvme_tcp_wq nvme_tcp_io_work [ 6.891875] RIP: 0010:_copy_to_iter+0x289/0x1330 [ 6.903739] Call Trace: [ 6.904085] [ 6.909254] __skb_datagram_iter+0x433/0x820 [ 6.911026] skb_copy_datagram_iter+0x37/0x120 [ 6.911622] nvme_tcp_recv_skb+0xa07/0x4320 [ 6.913378] __tcp_read_sock+0x1ab/0x810 [ 6.915788] nvme_tcp_try_recv+0x152/0x1e0 [ 6.918222] nvme_tcp_io_work+0x1e4/0x6c0 [ 6.926906] [ 6.927226] ---[ end trace 0000000000000000 ]--- [ 6.927878] nvme nvme0: queue 1 failed to copy request 0x71 data [ 6.928709] nvme nvme0: receive failed: -14 Fixes: 3f2304f8c6d6 ("nvme-tcp: add NVMe over TCP host driver") Cc: stable@vger.kernel.org Reviewed-by: Christoph Hellwig Signed-off-by: Yehyeong Lee Signed-off-by: Keith Busch Signed-off-by: Greg Kroah-Hartman

nvme-fc: fix double free of fabrics options when nvme_add_ctrl() fails

2026-09-11T09:49:14+00:00

commit 56e6279266f6962bb2d38a54397e3c605165b0c5 upstream. nvmf_create_ctrl() owns the fabrics options and frees them whenever ->create_ctrl() returns an error, so a transport must not free them on its own error paths. nvme-fc tracks this by testing ctrl->ctrl.opts in nvme_fc_ctrl_free(), which requires nvme_fc_init_ctrl() to clear that pointer on every error exit. The coupling is implicit, and commit 1a9e218195a5 ("nvme: split device add from initialization") broke it by adding a second error exit. When nvme_add_ctrl() fails, nvme_fc_init_ctrl() jumps to out_put_ctrl:, past the "ctrl->ctrl.opts = NULL" that only sits on the fail_ctrl: path, so nvme_fc_ctrl_free() frees the options and nvmf_create_ctrl() frees them a second time: BUG: KASAN: slab-use-after-free in nvmf_free_options+0x30/0x190 nvmf_free_options+0x30/0x190 drivers/nvme/host/fabrics.c:1284 nvmf_create_ctrl drivers/nvme/host/fabrics.c:1374 [inline] Freed by task 5534: nvme_fc_ctrl_free drivers/nvme/host/fc.c:2374 [inline] nvme_fc_init_ctrl+0xe17/0x1450 drivers/nvme/host/fc.c:3605 nvme_add_ctrl() fails when dev_set_name() cannot allocate, so this is reachable under memory pressure or fault injection. Without KASAN the options are freed twice. Rather than clear the pointer on the second exit as well, derive ownership the way nvme-tcp, nvme-rdma and nvme-loop do, from list membership: their free_ctrl leaves the options alone unless the controller made it onto the transport list. The list cannot simply be populated on the success path as it is there. nvme-fc runs the initial connect synchronously via flush_delayed_work(), and the controller has to be reachable on rport->ctrl_list for the whole of it: nvme_fc_unregister_remoteport() needs to find it to signal connectivity loss, nvme_fc_match_disconn_ls() matches an incoming Disconnect Association LS against ctrl->association_id, which is only assigned during that window, nvme_fc_resume_controller() needs it on remoteport re-registration, and nvme_fc_existing_controller() uses it to reject a duplicate connect racing the one in flight. Keep the insertion where it is and add a fail_unlist: label, falling into fail_ctrl:, for the error paths that run after it. The earlier error paths never reach the insertion and keep using fail_ctrl: directly, so the list is only touched where the controller is actually on it. nvme_fc_ctrl_free() cannot use the plain "goto free_ctrl" the other transports use, because it still has to put_device(), release the rport reference and free the ida entry for resources taken before the insertion. Sample list_empty() under rport->lock instead. ctrl->ctrl.opts also stays valid for the whole teardown now. That is not the bug being fixed, but it removes some fragility around the old idiom: nvme_free_ctrl() calls nvme_auth_free() before ->free_ctrl(), and ctrl_max_dhchaps() dereferences ctrl->opts without a NULL check when ctrl->dhchap_ctxs is set, which nvme-fc permits since NVMF_ALLOWED_OPTS allows the dhchap options. The nvme sysfs attributes that dereference ctrl->opts, such as hostnqn and address, evaluate their is_visible() test once at device_add() time and stay readable until cdev_device_del(). Fixes: 1a9e218195a5 ("nvme: split device add from initialization") Cc: stable@vger.kernel.org Reported-by: syzbot+f58e57380a6083c4041d@syzkaller.appspotmail.com Closes: https://syzkaller.appspot.com/bug?extid=f58e57380a6083c4041d Signed-off-by: Niklas Cassel Tested-by: Rihyeon Kim Reviewed-by: Hannes Reinecke Signed-off-by: Keith Busch Signed-off-by: Greg Kroah-Hartman

nvme-fabrics: fix DHCHAP secret leak on parse failure

2026-09-11T09:49:14+00:00

commit afdee49a1b88ed9bb44e2b30e855297c169bcc53 upstream. nvmf_parse_options() duplicates dhchap_secret and dhchap_ctrl_secret with match_strdup() before validating the DHHC-1: representation. If validation fails, the parser returns -EINVAL before the temporary string in p is assigned to opts->dhchap_secret or opts->dhchap_ctrl_secret. nvmf_create_ctrl() subsequently frees opts, but nvmf_free_options() cannot release the unassigned temporary string. Each rejected option therefore leaks one allocation. This is easy to miss because valid secrets transfer ownership to opts and are freed normally, while the malformed-secret path still returns the expected -EINVAL to userspace. With CONFIG_NVME_HOST_AUTH enabled, the leak is reachable before the required-option checks and transport lookup. No NVMe-oF target or working transport connection is required; for example, repeatedly writing dhchap_secret=BAD or dhchap_ctrl_secret=BAD to /dev/nvme-fabrics deterministically takes the leaking parse path. Free the temporary string before leaving both validation error paths. Use kfree_sensitive() because the copied option may contain secret material even when its representation is rejected, matching the sensitive cleanup used for stored DHCHAP secrets. Fixes: f50fff73d620 ("nvme: implement In-Band authentication") Cc: stable@vger.kernel.org Reviewed-by: Christoph Hellwig Signed-off-by: Xu Rao Signed-off-by: Keith Busch Signed-off-by: Greg Kroah-Hartman

nvme-tcp: reject a read that transferred too few bytes

2026-09-07T15:22:46+00:00

commit 7fa3f73f6c8ddc5f0425b50fb2a626a782ef7d12 upstream. nvme_tcp_recv_data() completes a request once the current C2HData PDU has been consumed. Nothing compares the total bytes received against the length the command asked for: struct nvme_tcp_request has no receive-side counter, queue->data_remaining is per queue, and blk_mq_end_request() completes for blk_rq_bytes(rq) unconditionally with no residual concept anywhere above. A controller can therefore answer a 4096-byte read with 512 bytes and have it reported as a complete read; user space then gets 4096 bytes of which 3584 are whatever was already in the page. I reproduced that with a test target. Count the bytes received and refuse to complete a successful read whose count does not match, at the two NVME_TCP_F_DATA_SUCCESS paths and in nvme_tcp_process_nvme_cqe(). The success test shifts req->status right by one, because the driver keeps the wire value there and shifts it on completion, so the check must see what the completion path will see. Only REQ_OP_READ is checked, because there the length comes from the sectors the request covers; a passthrough command is built by its submitter, which picks both command and buffer, so the kernel has nothing to compare against. Fixes: 3f2304f8c6d6 ("nvme-tcp: add NVMe over TCP host driver") Cc: stable@vger.kernel.org Signed-off-by: Yehyeong Lee Signed-off-by: Keith Busch Signed-off-by: Greg Kroah-Hartman

nvme-tcp: fix host memory disclosure on R2T for a read command

2026-09-07T15:22:46+00:00

commit 6efbc52237facda35d2d874fe1765bb4839275d8 upstream. nvme_tcp_handle_r2t() does not check the direction of the request the R2T refers to. A malicious controller can send an R2T for a READ and the host will answer it: nvme_tcp_setup_h2c_data_pdu() builds the H2CData header and nvme_tcp_try_send_data() sends the request's data buffer. That buffer is the READ destination, so its contents go to the controller. The command then completes normally and nothing is logged. Against a test controller that answers every READ with an R2T, a 4096 byte buffered read returned all 4096 bytes, split over two R2Ts. The pages contained stale kernel data, including an array of struct page pointers. Reject an R2T for a request that is not a write. Fixes: 3f2304f8c6d6 ("nvme-tcp: add NVMe over TCP host driver") Cc: stable@vger.kernel.org Signed-off-by: Yehyeong Lee Signed-off-by: Keith Busch Signed-off-by: Greg Kroah-Hartman

nvme-tcp: do not accept C2HData based on blk_rq_payload_bytes() alone

2026-09-07T15:22:46+00:00

commit 3a4aa9e6ad3e35f8e24d5eaf38ee4d437075fb36 upstream. Commit 25e5cb780e62 ("nvme-tcp: fix possible crash in write_zeroes processing") established that blk_rq_payload_bytes() must not be read without first checking blk_rq_nr_phys_segments(), and recorded the result in nvme_tcp_setup_cmd_pdu() as req->data_len. The receive side was left as it was. The two differ for REQ_OP_WRITE_ZEROES, which has no physical segments but a non-zero blk_rq_bytes(), so setup leaves req->iter untouched while the receive gate lets a C2HData through and nvme_tcp_recv_data() copies into whatever the previous command on that tag left there. The driver-private area is zeroed only when the tag set is allocated. Reproduced with a test target that leaves a residual iterator on a tag and then sends a C2HData for a WRITE_ZEROES command on the same tag: BUG: KASAN: wild-memory-access in _copy_to_iter+0x642/0x1330 Write of size 512 at addr ffe728c2175dfa81 by task kworker/0:1H/103 CPU: 0 UID: 0 PID: 103 Comm: kworker/0:1H Not tainted 7.2.0-rc5-NVMETCP-gf5098b6bae76 #1 PREEMPT(lazy) Hardware name: QEMU Ubuntu 24.04 PC v2 (i440FX + PIIX, arch_caps fix, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014 Workqueue: nvme_tcp_wq nvme_tcp_io_work Call Trace: dump_stack_lvl+0x53/0x70 kasan_report+0xce/0x100 ? _copy_to_iter+0x642/0x1330 kasan_check_range+0x105/0x1b0 __asan_memcpy+0x3c/0x60 _copy_to_iter+0x642/0x1330 ? __pfx_sock_has_perm+0x10/0x10 ? worker_thread+0x45b/0xd10 ? __pfx__copy_to_iter+0x10/0x10 ? _raw_spin_lock_bh+0x83/0xe0 ? __pfx__raw_spin_lock_bh+0x10/0x10 __skb_datagram_iter+0xf3/0x820 ? __pfx_simple_copy_to_iter+0x10/0x10 ? __asan_memcpy+0x3c/0x60 ? skb_copy_bits+0x58d/0x830 skb_copy_datagram_iter+0x37/0x120 nvme_tcp_recv_skb+0xa07/0x4320 ? __pfx_nvme_tcp_recv_skb+0x10/0x10 __tcp_read_sock+0x1ab/0x810 ? __pfx_nvme_tcp_recv_skb+0x10/0x10 ? __pfx_lock_sock_nested+0x10/0x10 ? __pfx___tcp_read_sock+0x10/0x10 nvme_tcp_try_recv+0x152/0x1e0 ? __pfx_nvme_tcp_try_recv+0x10/0x10 ? __pfx_mutex_unlock+0x10/0x10 nvme_tcp_io_work+0x1e4/0x6c0 ? __schedule+0x181a/0x49f0 ? __pfx_nvme_tcp_io_work+0x10/0x10 process_one_work+0x633/0x1030 Keep the blk_rq_payload_bytes() test and add req->data_len to it. The old test is what rejects a C2HData naming a tag that is no longer in flight, because blk_update_request() zeroes rq->__data_len on completion; req->data_len and req->curr_bio are driver-private and survive completion, so they cannot stand in for it. Setup initialises the iterator only when both req->curr_bio and req->data_len are set, so the gate now tests the same two. Fixes: 25e5cb780e62 ("nvme-tcp: fix possible crash in write_zeroes processing") Cc: stable@vger.kernel.org Signed-off-by: Yehyeong Lee Signed-off-by: Keith Busch Signed-off-by: Greg Kroah-Hartman