Описание
In the Linux kernel, the following vulnerability has been resolved:
vhost: reset the vring metadata cache on vring reconfiguration
vq->meta_iotlb[] caches the vhost_iotlb_map that backs each vring metadata region, and iotlb_access_ok() returns early on a cache hit, taking the hit as proof that the region has already been validated:
The cache is reset on VHOST_IOTLB_UPDATE and VHOST_IOTLB_INVALIDATE, on device IOTLB (re)initialisation and on vq reset, but not when VHOST_SET_VRING_ADDR replaces vq->desc, vq->avail and vq->used, nor when VHOST_SET_VRING_NUM changes the region sizes.
With a device IOTLB attached both ioctls are accepted while the vq is live, and neither validates the addresses at ioctl time: vq_access_ok() and vq_log_used_access_ok() return true early because the addresses are GIOVAs, deferring validation to prefetch time. Once the cache has been populated that deferred validation no longer runs -- vq_m...
In the Linux kernel, the following vulnerability has been resolved:
vhost: reset the vring metadata cache on vring reconfiguration
vq->meta_iotlb[] caches the vhost_iotlb_map that backs each vring metadata region, and iotlb_access_ok() returns early on a cache hit, taking the hit as proof that the region has already been validated:
The cache is reset on VHOST_IOTLB_UPDATE and VHOST_IOTLB_INVALIDATE, on device IOTLB (re)initialisation and on vq reset, but not when VHOST_SET_VRING_ADDR replaces vq->desc, vq->avail and vq->used, nor when VHOST_SET_VRING_NUM changes the region sizes.
With a device IOTLB attached both ioctls are accepted while the vq is live, and neither validates the addresses at ioctl time: vq_access_ok() and vq_log_used_access_ok() return true early because the addresses are GIOVAs, deferring validation to prefetch time. Once the cache has been populated that deferred validation no longer runs -- vq_meta_prefetch() hits the stale entry and returns true -- and vhost_vq_meta_fetch() keeps translating through the old mapping as
for an address the mapping no longer covers. vhost_copy_to_user() and vhost_copy_from_user() consume the result with __copy_to_user() and __copy_from_user(), which do not check it either, so a subsequent used ring update or descriptor fetch accesses memory outside the region the IOTLB actually maps.
Reset the metadata cache whenever the vring is reconfigured, so the new addresses are pushed back through iotlb_access_ok()'s slow path.
Ссылки
- https://nvd.nist.gov/vuln/detail/CVE-2026-74580
- https://git.kernel.org/stable/c/13fa6f32a56a386a82bd7451644c494beed034af
- https://git.kernel.org/stable/c/5224bd37e37d36076a550d99b2aebba33939fd95
- https://git.kernel.org/stable/c/54617e9119be2eb728ecdd8d977b99c99d4c498a
- https://git.kernel.org/stable/c/6fa3e9b1fe856259555a7e22f3f3082e7827fd9b
- https://git.kernel.org/stable/c/b70ebe0bba254e093dd5fd4c0c170941ce83eb85
- https://git.kernel.org/stable/c/cf363a7a02ce132ef1f58084fdb13e1a3b7da7e7
- https://git.kernel.org/stable/c/de845981da67a6b049080c87e605130b0c30adc5
- https://git.kernel.org/stable/c/f1e21108e3ddfcce62f6cad4ebd7b5674543c9e6
Связанные уязвимости
In the Linux kernel, the following vulnerability has been resolved: vhost: reset the vring metadata cache on vring reconfiguration vq->meta_iotlb[] caches the vhost_iotlb_map that backs each vring metadata region, and iotlb_access_ok() returns early on a cache hit, taking the hit as proof that the region has already been validated: if (vhost_vq_meta_fetch(vq, addr, len, type)) return true; The cache is reset on VHOST_IOTLB_UPDATE and VHOST_IOTLB_INVALIDATE, on device IOTLB (re)initialisation and on vq reset, but not when VHOST_SET_VRING_ADDR replaces vq->desc, vq->avail and vq->used, nor when VHOST_SET_VRING_NUM changes the region sizes. With a device IOTLB attached both ioctls are accepted while the vq is live, and neither validates the addresses at ioctl time: vq_access_ok() and vq_log_used_access_ok() return true early because the addresses are GIOVAs, deferring validation to prefetch time. Once the cache has been populated that deferred validation no longer runs -- vq_meta_pref...
In the Linux kernel, the following vulnerability has been resolved: vhost: reset the vring metadata cache on vring reconfiguration vq->meta_iotlb[] caches the vhost_iotlb_map that backs each vring metadata region, and iotlb_access_ok() returns early on a cache hit, taking the hit as proof that the region has already been validated: if (vhost_vq_meta_fetch(vq, addr, len, type)) return true; The cache is reset on VHOST_IOTLB_UPDATE and VHOST_IOTLB_INVALIDATE, on device IOTLB (re)initialisation and on vq reset, but not when VHOST_SET_VRING_ADDR replaces vq->desc, vq->avail and vq->used, nor when VHOST_SET_VRING_NUM changes the region sizes. With a device IOTLB attached both ioctls are accepted while the vq is live, and neither validates the addresses at ioctl time: vq_access_ok() and vq_log_used_access_ok() return true early because the addresses are GIOVAs, deferring validation to prefetch time. Once the cache has been populated that deferred validation no longer runs -- vq_meta_pref...
In the Linux kernel, the following vulnerability has been resolved: vhost: reset the vring metadata cache on vring reconfiguration vq->meta_iotlb[] caches the vhost_iotlb_map that backs each vring metadata region, and iotlb_access_ok() returns early on a cache hit, taking the hit as proof that the region has already been validated: if (vhost_vq_meta_fetch(vq, addr, len, type)) return true; The cache is reset on VHOST_IOTLB_UPDATE and VHOST_IOTLB_INVALIDATE, on device IOTLB (re)initialisation and on vq reset, but not when VHOST_SET_VRING_ADDR replaces vq->desc, vq->avail and vq->used, nor when VHOST_SET_VRING_NUM changes the region sizes. With a device IOTLB attached both ioctls are accepted while the vq is live, and neither validates the addresses at ioctl time: vq_access_ok() and vq_log_used_access_ok() return true early because the addresses are GIOVAs, deferring validation to prefetch time. Once the cache has been populated that deferred validation no longer runs -- vq_meta
In the Linux kernel, the following vulnerability has been resolved: v ...