In the Linux kernel, the following vulnerability has been resolved:
scsi: qla2xxx: Unlink NVMe unsol ctx before freeing on LS reject error
qla_nvme_xmt_ls_rsp() obtains uctx, which was linked into
fcport->unsol_ctx_head by qla2xxx_process_purls_iocb() and is still linked
when the NVMe transport calls back to transmit the LS response. On the
error (out:) path the function frees uctx with kfree() but never removes
it from the list. This leaves a freed node in fcport->unsol_ctx_head: the
next list_add_tail() for that fcport writes through the freed node, and a
subsequent list_del() can corrupt the list or panic.
Unlink uctx with list_del() before kfree() on the error path, matching the
other free sites in qla_nvme_release_lsrsp_cmd_kref() and
qla2xxx_process_purls_pkt(). qla2x00_rel_sp() in the failure path only
returns the SRB to its pool and does not invoke sp->put_fn, so the out:
path is the sole free and uctx is always still linked there.
CVSS Vector: CVSS:3.1/AV:A/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
CVSS Score: 8.8
AV:A - Unsolicited FC-NVMe LS frames are posted as PT_LS4_UNSOL; qla24xx_msix_rsp_q() (or qla2xxx_msix_rsp_q → qla_do_work) runs qla24xx_process_response_queue() → qla2xxx_process_purls_iocb(), then DPC qla2xxx_process_purls_pkt() → nvme_fc_rcv_ls_req() and nvme_fc_xmt_ls_rsp() call qla_nvme_xmt_ls_rsp(). FC/FCoE is a non-routable SAN fabric, so Adjacent is the highest defensible vector.
AC:L - qla_nvme_xmt_ls_rsp() kfree()s uctx on out: when qla2x00_get_sp(GFP_ATOMIC) fails, qla2x00_start_sp() fails after five -EAGAIN retries, or fcport->deleted is set. A peer floods CREATE_ASSOCIATION LS onto base_qpair or sends LS then LOGO so deleted is set before nvme_fc_handle_ls_rqst_work() calls xmt_ls_rsp; both sides are attacker-driven.
PR:N - qla2xxx_process_purls_iocb() looks up the sender with qla2x00_find_fcport_by_nportid(incl_deleted=0) and handles PT_LS4_UNSOL in the MSI-X/DPC path with no Linux credential or capability check. FC-SP/EDIF is opt-in; being a logged-in NVMe N_Port is fabric membership, not a privilege on the victim host.
UI:N - qla24xx_queue_purex_item() sets PROCESS_PUREX_IOCB; qla2xxx_process_purls_pkt() and the nvme_fc rport lsrcv_work then call qla_nvme_xmt_ls_rsp() with no victim mount, open, or administrator action.
S:U - kfree() of uctx while it remains on fcport->unsol_ctx_head corrupts that per-fcport list and the host kernel heap inside the qla2xxx driver. Impact stays in the same kernel security authority; it is not a VM escape or IOMMU/DMA bypass.
C:H - After qla_nvme_xmt_ls_rsp() kfree()s uctx, the next qla2xxx_process_purls_iocb() list_add_tail() and a later list_del() in qla_nvme_release_lsrsp_cmd_kref() or qla2xxx_process_purls_pkt() operate on the freed qla_nvme_unsol_ctx. That use-after-free of a kmalloc object enables disclosure of reused kernel heap.
I:H - list_add_tail(&uctx->elem, &fcport->unsol_ctx_head) writes next/prev through the already-freed node, and list_del() of a neighbor stores those pointers into adjacent heap objects. Combined with attacker-driven uctx allocations of known size, that is a kernel write primitive usable for control-flow hijack.
A:H - list_add_tail()/list_del() on the dangling uctx->elem oopses on LIST_POISON or panics on a corrupted next/prev, and the remote peer can repeat PT_LS4_UNSOL to hit the out: path again, taking down the host and its FC-NVMe storage path.
| Attack Vector |
Adjacent Network |
Scope |
Unchanged |
| Attack Complexity |
Low |
Confidentiality Impact |
High |
| Privileges Required |
None |
Integrity Impact |
High |
| User Interaction |
None |
Availability Impact |
High |
AV:A - Unsolicited FC-NVMe LS frames are posted as PT_LS4_UNSOL; qla24xx_msix_rsp_q() (or qla2xxx_msix_rsp_q → qla_do_work) runs qla24xx_process_response_queue() → qla2xxx_process_purls_iocb(), then DPC qla2xxx_process_purls_pkt() → nvme_fc_rcv_ls_req() and nvme_fc_xmt_ls_rsp() call qla_nvme_xmt_ls_rsp(). FC/FCoE is a non-routable SAN fabric, so Adjacent is the highest defensible vector.
AC:L - qla_nvme_xmt_ls_rsp() kfree()s uctx on out: when qla2x00_get_sp(GFP_ATOMIC) fails, qla2x00_start_sp() fails after five -EAGAIN retries, or fcport->deleted is set. A peer floods CREATE_ASSOCIATION LS onto base_qpair or sends LS then LOGO so deleted is set before nvme_fc_handle_ls_rqst_work() calls xmt_ls_rsp; both sides are attacker-driven.
PR:N - qla2xxx_process_purls_iocb() looks up the sender with qla2x00_find_fcport_by_nportid(incl_deleted=0) and handles PT_LS4_UNSOL in the MSI-X/DPC path with no Linux credential or capability check. FC-SP/EDIF is opt-in; being a logged-in NVMe N_Port is fabric membership, not a privilege on the victim host.
UI:N - qla24xx_queue_purex_item() sets PROCESS_PUREX_IOCB; qla2xxx_process_purls_pkt() and the nvme_fc rport lsrcv_work then call qla_nvme_xmt_ls_rsp() with no victim mount, open, or administrator action.
S:U - kfree() of uctx while it remains on fcport->unsol_ctx_head corrupts that per-fcport list and the host kernel heap inside the qla2xxx driver. Impact stays in the same kernel security authority; it is not a VM escape or IOMMU/DMA bypass.
C:H - After qla_nvme_xmt_ls_rsp() kfree()s uctx, the next qla2xxx_process_purls_iocb() list_add_tail() and a later list_del() in qla_nvme_release_lsrsp_cmd_kref() or qla2xxx_process_purls_pkt() operate on the freed qla_nvme_unsol_ctx. That use-after-free of a kmalloc object enables disclosure of reused kernel heap.
I:H - list_add_tail(&uctx->elem, &fcport->unsol_ctx_head) writes next/prev through the already-freed node, and list_del() of a neighbor stores those pointers into adjacent heap objects. Combined with attacker-driven uctx allocations of known size, that is a kernel write primitive usable for control-flow hijack.
A:H - list_add_tail()/list_del() on the dangling uctx->elem oopses on LIST_POISON or panics on a corrupted next/prev, and the remote peer can repeat PT_LS4_UNSOL to hit the out: path again, taking down the host and its FC-NVMe storage path.
CVSS 3.1