{"id":"CVE-2026-97994","published":"2026-09-25T11:17:27.637","lastModified":"2026-09-25T11:17:27.637","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nvhost/vdpa: reject VRING_NUM larger than device max\n\nvhost_vring_set_num() accepts any non-zero power-of-two queue size that\nfits in 16 bits. vhost-vdpa then passes that value to set_vq_num()\nwithout comparing it with get_vq_num_max().\n\nA process with access to /dev/vhost-vdpa-* can therefore configure a\nqueue larger than the device advertises. With vdpa_sim, the worker can\nwalk descriptors beyond the mapped descriptor ring. KASAN reports a\n16-byte out-of-bounds read, corresponding to one vring_desc, in the\nvringh IOTLB path:\n\n  BUG: KASAN: out-of-bounds in _copy_from_iter\n  Read of size 16\n  copy_from_iotlb\n  copydesc_iotlb\n  vringh_getdesc_iotlb\n  vdpasim_net_work\n\nCache get_vq_num_max() immediately after reset. Some backends derive\nit from writable queue-size state, so querying it after SET_NUM may\nreturn the current size instead of the device capability. Invalidate\nthe cached value before reset so a failed reset leaves SET_NUM\ndisabled.\n\nFor VHOST_SET_VRING_NUM, copy the complete vring state once and use\nthe same index and size for validation, vq->num, and set_vq_num().\nThis ensures that validation and use operate on the same copied values.","cvssScore":null,"cvssSeverity":null,"cvssVector":null,"cwes":[],"vendors":[],"products":[],"references":[{"url":"https://git.kernel.org/stable/c/1d09201d81b7d3e276860cc7b7dbf6c6cbe8e9b4","tags":[]},{"url":"https://git.kernel.org/stable/c/59522639a7d71cff4e20d594d0b9ea30dd0c77e0","tags":[]},{"url":"https://git.kernel.org/stable/c/68232102f20fc961327fb9e0f605a7eaadf030a9","tags":[]},{"url":"https://git.kernel.org/stable/c/ccb1dc7c527f8c925925cf92afc76ae590dac311","tags":[]}],"exploitRefs":[],"hasPoc":false,"ai":null}