{"id":"CVE-2026-90095","published":"2026-09-17T17:17:01.123","lastModified":"2026-09-17T17:17:01.123","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nfuse: Fix the condition to enable over-io-uring\n\nThe existing condition in fuse_uring_cmd() is there only to avoid\ndisabling io-uring for connections that already run with it, missing\nwas a condition to refuse any IORING_OP_URING_CMD if the\nconnection/channel didn't get enabled because of missing FUSE_INIT\nreply flag FUSE_OVER_IO_URING. Without the reply flag the barrier in\nfuse_uring_ready() doesn't work and IO could already be going on and\ncause deadlock states (at a minimum one between fch->bg_lock and\nqueue->lock).\n\nThe change itself is trivial, but brings behavior change,\nFUSE_OVER_IO_URING has to be set in the FUSE_INIT_REPLY by fuse servers\nto accept any IORING_OP_URING_CMD. Libfuse does that and the only\nnon-libfuse implementation I found (fractal-fuse) also does it.\nQemu patches for fuse-io-uring are not merged yet, as far as I know.\n\nMoved up is the smp_load_acquire(&fch->initialized) check, as a\nfuse-server implementation might try to setup io-uring before FUSE_INIT\nis processed and might have gotten -EOPNOTSUPP instead of -EAGAIN.\n\nAlso fixed is a stale comment that explains the handling of the\nFUSE_OVER_IO_URING flag in early RFC versions.\n\nIf there should be a report from any library or application we\nprobably need to revert this commit.","cvssScore":null,"cvssSeverity":null,"cvssVector":null,"cwes":[],"vendors":[],"products":[],"references":[{"url":"https://git.kernel.org/stable/c/1f59015e958174e89be58cc8db16d70a60d17255","tags":[]},{"url":"https://git.kernel.org/stable/c/8f9a725d89711ad027f6b7183586ef91528f106d","tags":[]}],"exploitRefs":[],"hasPoc":false,"ai":null}