Look up another CVE · open CVE-2026-90172 in the interactive tracker
CVE-2026-90172
smb: smbdirect: destroy QP before mem pools on accept failure
Published 2026-09-17 (31 days ago) by the Linux kernel CNA. Subsystem fs/smb. Debian urgency: unimportant.
What to do, per Debian release
Debian's own status for CVE-2026-90172 in every release this site tracks, and what each one means in practice. Every status below is the one the Debian Security Team publishes for this CVE. source
| Release | Status | Version | What this means |
|---|---|---|---|
| Debian 12bookworm | not affected | 6.1.190-1 | Nothing to do. bookworm never shipped the vulnerable code. |
| Debian 13trixie | not affected | 6.12.113-1 | Nothing to do. trixie never shipped the vulnerable code. |
| Debian 14forky | fixed | 7.2.6-1 | Update the kernel and reboot. Fixed in 7.2.6-1. A running kernel keeps the old code until the machine restarts. |
| Debian unstablesid | fixed | 7.2.6-1 | Update the kernel and reboot. Fixed in 7.2.6-1. A running kernel keeps the old code until the machine restarts. |
The vulnerability
What the bug is
On the rdma_accept_failed error path of smbdirect_accept_connect_request(), the receive io posted just above is owned by the QP (recv_io is set to NULL after a successful post). The error path fell through to smbdirect_connection_destroy_mem_pools() before smbdirect_connection_destroy_qp(), so the mem pools and the recv_io slab cache were destroyed while that recv_io was still outstanding on the QP. The drain in smbdirect_connection_destroy_qp() (ib_drain_qp()) is what runs the recv completion that returns the recv_io to the free list, so destroying the pools first leaves the object outstanding at kmem_cache_destroy() time ("Slab cache still has objects") and later frees it into an already-destroyed mempool (mempool_free_bulk NULL-pointer dereference). Give rdma_accept_failed its own teardown that drains the QP first, then destroys the mem pools, and returns. The remaining labels (post_recv_io_failed onward) run before the recv_io was ever posted, so they keep the mem-pools-then-qp order. The outstanding recv_io at kmem_cache_destroy() time: [ 3487.344647] ============================================================================= [ 3487.349942] BUG smbdirect_recv_io_cache_ffff88811ba99000 (Not tainted): Objects remaining on __kmem_cache_shutdown() [ 3487.356078] ----------------------------------------------------------------------------- [ 3487.356078] [ 3487.356738] Object 0xffff8881511c3440 @offset=13376 [ 3487.358464] Allocated in mempool_alloc_noprof+0x18c/0x290 age=1194 cpu=6 pid=22254 [ 3487.361197] mempool_alloc_noprof+0x18c/0x290 [ 3487.361542] smbdirect_connection_create_mem_pools+0x405/0x780 [ 3487.361972] smbdirect_accept_connect_request+0x5a8/0x1b80 [ 3487.362359] smbdirect_listen_rdma_event_handler+0x1579/0x1b90 [ 3487.362779] cma_cm_event_handler+0x9c/0x230 [ 3487.363096] cma_ib_req_handler+0x2682/0x45d0 [ 3487.363414] cm_process_work+0x56/0x3d0 [ 3487.363676] cm_work_handler+0x8a0e/0xd000 [ 3487.367496] process_scheduled_works+0xa07/0x13a0 [ 3487.367859] worker_thread+0x7c9/0xc80 [ 3487.368148] kthread+0x341/0x430 [ 3487.368407] ret_from_fork+0x3a8/0x7a0 [ 3487.368704] ret_from_fork_asm+0x1a/0x30 [ 3487.370307] Slab 0xffffea0005447000 objects=19 used=1 fp=0xffff8881511c0040 flags=0x100000000000240(workingset|head|node=0|zone=2) [ 3487.372840] ------------[ cut here ]------------ [ 3487.373195] WARNING: mm/slub.c:1244 at __slab_err+0x1a/0x30, CPU#6: kworker/6:84/22254 [ 3487.373759] Modules linked in: [ 3487.373993] CPU: 6 UID: 0 PID: 22254 Comm: kworker/6:84 Tainted: G B 7.1.0-next-20260623+ #88 PREEMPT(lazy) [ 3487.374778] Tainted: [B]=BAD_PAGE [ 3487.377830] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.17.0-debian-1.17.0-1 04/01/2014 [ 3487.378515] Workqueue: ib_cm cm_work_handler [ 3487.378820] RIP: 0010:__slab_err+0x1a/0x30 [ 3487.379129] Code: 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 0f 1f 44 00 00 e8 36 00 00 00 bf 05 00 00 00 be 01 00 00 00 e8 f7 75 45 00 90 <0f> 0b 90 c3 cc cc cc cc cc 66 66 66 66 2e 0f 1f 84 00 00 00 00 00 [ 3487.383255] RSP: 0018:ffff888220fc7050 EFLAGS: 00010093 [ 3487.383643] RAX: ffffffff8168e60a RBX: ffff88810955e640 RCX: ffff88821c381d80 [ 3487.384158] RDX: 0000000000000000 RSI: 0000000000000008 RDI: ffffffff870fa080 [ 3487.384662] RBP: ffff888220fc7068 R08: ffffffff870fa087 R09: 1ffffffff0e1f410 [ 3487.385192] R10: dffffc0000000000 R11: fffffbfff0e1f411 R12: ffffea0005447210 [ 3487.385674] R13: ffffea0005447000 R14: ffff888220fc7068 R15: ffff88812a8ab300 [ 3487.388932] FS: 0000000000000000(0000) GS:ffff888427e76000(0000) knlGS:0000000000000000 [ 3487.389529] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 3487.389934] CR2: 00007ffcf2d84fd8 CR3: 0000000111d64006 CR4: 0000000000f72ef0 [ 3487.390440] PKRU: 55555554 [ 3487.390641] Call Trace: [ 3487.390826] <TASK> [ 3 ---truncated---
Where it lives
The fix touches these files. If the subsystem is one you do not use, the practical exposure is lower, though the package is still the vulnerable one.
fs/smb/smbdirect/accept.c
- What this is
- SMBDIRECT (RDMA Stream Transport with Read/Write-Offload, MS-SMBD)MAINTAINERS section matched on fs/smb/smbdirect/ · linux-cifs@vger.kernel.org · Maintained
Check it on the machine
grep -wE 'CONFIG_SMBDIRECT|CONFIG_SMBFS' /boot/config-$(uname -r)Reads the booted kernel's own configuration, on the machine and nowhere else. =y is built in, =m is a module, and an is not set line means the code is not there.
Show whether Debian builds this, per release
Debian does not build this code in 2 releases. The build rule could not be established for 2 releases, so nothing is claimed about those.
- bookworm, trixie?
- forky, sidn
- ?
- not established - no Makefile entry for the affected file
- n
- not enabled on any flavour - this release does not build the code
Where a release above reads "not established", no build rule for the affected file was found in that release's kernel tree. That is a gap in this lookup, not a finding that the code is absent.
- Build switches
- CONFIG_SMBDIRECT, CONFIG_SMBFS
- Reads as
- all on means the file is compiled; any one off means it is not.
From the configuration Debian publishes for every kernel flavour it builds. It covers only the architectures Debian ships a linux-config package for, and nothing about a kernel you built yourself. Debian's own status above stays authoritative about the source package.MAINTAINERS kbuild Makefiles Debian linux-config packages
Signals
- Published
- 2026-09-17 (31 days ago)source
- CVSS
- 7.5 high
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:Hsource recompute it - Attack vector
- network (AV:N)source
- EPSS
- 0.49%, higher than 40.6% of all CVEssource
- CISA KEV
- not listedthe KEV catalogue
- Debian urgency
- unimportantsource
- Subsystem
- fs/smb
Upstream fixes
The versions the CNA records as introducing and fixing this, and the commit for each. source
| Introduced | Fixed in | Commit |
|---|---|---|
| 7.1 | 7.2.6 | 959813a3752b |
| 7.1 | 7.3 | 383a9480f5f4 |
Why the CNA scored it this way
Quoted from the kernel CNA's own scoring file, word for word. A highlighted metric carries the most severe value CVSS v3.1 defines for it, which is what pushes the score up. source
- AV:NAttack vector, network
- A remote RDMA peer sends an IB CM REQ (RDMA_CM_EVENT_CONNECT_REQUEST) that cma_ib_req_handler delivers to smbdirect_listen_rdma_event_handler; ksmbd_rdma_init() listens for that on RoCEv2/IB port 445 and iWARP port 5445, which are IP-routable SMB Direct transports.
- AC:LAttack complexity, low
- smbdirect_accept_connect_request() takes rdma_accept_failed after smbdirect_connection_post_recv_io() has already posted the recv_io; the attacker makes rdma_accept() fail with a CONNECT_REQUEST whose CM REQ path/QP fields cause cma_modify_qp_rtr()/ib_send_cm_rep() to fail, so the error path is attacker-driven rather than a victim-side race.
- PR:NPrivileges required, none
- The bug runs in smbdirect_accept_connect_request() from the listen CONNECT_REQUEST handler, before smbdirect_accept_negotiate_recv_done() processes the SMB Direct negotiate request and before ksmbd_conn_handler_loop() runs SMB2 SESSION_SETUP, so no credentials are checked.
- UI:NUser interaction, none
- Once ksmbd_rdma_init() has called smbdirect_socket_listen(), the attacker triggers the buggy teardown solely by sending the RDMA CM CONNECT_REQUEST that fails rdma_accept(); no victim mount or other interaction is required.
- S:UScope, unchanged
- The oops is in the host kernel on smbdirect_connection_destroy_mem_pools()/mempool_free_bulk() during ksmbd SMB Direct accept teardown and does not cross a VM, IOMMU, or other security authority.
- C:NConfidentiality impact, none
- kmem_cache_destroy() of the per-connection smbdirect_recv_io_cache returns without freeing the still-posted recv_io, and the later mempool_free(recv_io, NULL) NULL-dereferences sc->recv_io.mem.pool in mempool_free_bulk(); that path does not return kernel memory to the RDMA peer.
- I:NIntegrity impact, none
- The outstanding recv_io is not actually released by the failed kmem_cache_destroy(), and mempool_free_bulk() oopses on the NULL pool pointer before any write through a dangling object, so there is no integrity or code-execution primitive.
- A:HAvailability impact, high
- After ib_drain_qp() in smbdirect_connection_destroy_qp() returns the posted recv_io onto the free list, smbdirect_socket_release() calls smbdirect_connection_destroy_mem_pools() again, which mempool_free()s it with sc->recv_io.mem.pool already NULL and oopses in mempool_free_bulk().