{"id":"CVE-2026-98366","published":"2026-10-06T09:18:30.823","lastModified":"2026-10-07T07:17:10.960","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nRDMA/rxe: validate access flags before swapping the MR's PD\n\nrxe_rereg_user_mr() reassigns mr->ibmr.pd first and only then\nvalidates the IB_MR_REREG_ACCESS argument:\n\n\tif (flags & IB_MR_REREG_PD) {\n\t\trxe_put(old_pd);\n\t\trxe_get(pd);\n\t\tmr->ibmr.pd = ibpd;\n\t}\n\n\tif (flags & IB_MR_REREG_ACCESS) {\n\t\tif (access & ~RXE_ACCESS_SUPPORTED_MR)\n\t\t\treturn ERR_PTR(-EOPNOTSUPP);\n\t\tmr->access = access;\n\t}\n\nBoth flags pass the entry check because RXE_MR_REREG_SUPPORTED is\nIB_MR_REREG_PD | IB_MR_REREG_ACCESS, so a caller can reach the access\ncheck with mr->ibmr.pd already reassigned.\n\nmr->ibmr.pd is owned by the core, which adjusts pd->usecnt only on the\nsuccess path: ib_uverbs_rereg_mr() jumps to put_new_uobj on a driver error\nwithout undoing the reassignment, so mr->pd == new_pd while the usecnts\nstill charge the MR to orig_pd. ib_dereg_mr_user() then decrements\nnew_pd, whose count can reach zero while a memory window still references\nit; uverbs_free_pd() frees the PD on that count alone and rxe_mw_cleanup()\nwrites to freed memory:\n\n  BUG: KASAN: slab-use-after-free in __rxe_put+0x31/0xa0\n  Write of size 4 at addr ffff8881301dd690 by task rxe_poc/591\n   __rxe_put+0x31/0xa0\n   rxe_mw_cleanup+0x42/0x200\n   __rxe_cleanup+0x115/0x370\n   rxe_dealloc_mw+0x4c/0x80\n  Allocated by task 591:\n   ib_uverbs_alloc_pd+0x258/0x540\n  Freed by task 591:\n   ib_dealloc_pd_user+0x174/0x210\n   uverbs_free_pd+0x8d/0xc0\n   ib_uverbs_dealloc_pd+0x18e/0x1d0\n\nValidate the access flags before mutating any state so the callback either\napplies every requested change or none.","cvssScore":7.8,"cvssSeverity":"HIGH","cvssVector":"CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H","cwes":[],"vendors":[],"products":[],"references":[{"url":"https://git.kernel.org/stable/c/08f12745cb72981aa2cabe214af0c7a42a856f0a","tags":[]},{"url":"https://git.kernel.org/stable/c/4dd7a53f1c5b44693c26dac4b9b5bd5bb9d604c8","tags":[]},{"url":"https://git.kernel.org/stable/c/7230cc456d4bb2221c20c5f1d22b38b8576aa16f","tags":[]},{"url":"https://git.kernel.org/stable/c/ae36a5b609ae79f4de966328b78d2584be9719a4","tags":[]},{"url":"https://git.kernel.org/stable/c/fe602c91a52e161ace4afd5bdb8f33270c554c6f","tags":[]}],"exploitRefs":[],"hasPoc":false,"ai":{"summary":"The Linux kernel's RDMA/rxe driver has a race condition where memory region access flags are validated after the protection domain (PD) is already reassigned, potentially leading to use-after-free vulnerabilities when a PD is freed while still referenced by active memory windows.","exploitability":"Exploitation requires careful crafting of memory registration operations with specific flag combinations, likely needing kernel-level access or a local user context with RDMA capabilities, but the vulnerability is relatively straightforward to trigger once the right conditions are met.","blast_radius":"If exploited, this vulnerability could lead to kernel crashes, memory corruption, and potentially privilege escalation in systems running RDMA-enabled kernels with active memory window operations.","remediation":"Apply the kernel patch that fixes the order of PD reassignment and access flag validation in rxe_rereg_user_mr() function, specifically addressing the race condition between PD reference counting and memory window cleanup.","detection":"No reliable host or network indicator is derivable from the published description.","tags":["kernel","rdma","memory-corruption"],"model":"garnet-coder","analyzedAt":"2026-10-14T09:02:31.029Z"}}