CVE-2026-64527

*** Pendiente de traducción *** In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> drm/hyperv: validate VMBus packet size in receive callback<br /> <br /> hyperv_receive_sub() reads msg-&gt;vid_hdr.type and dispatches into one<br /> of four message-type branches without knowing how many bytes the host<br /> wrote into hv-&gt;recv_buf. The completion path then runs<br /> memcpy(hv-&gt;init_buf, msg, VMBUS_MAX_PACKET_SIZE), so the consumer that<br /> wakes on wait_for_completion_timeout() can read up to 16 KiB of<br /> residue from a prior message as if it were the response payload.<br /> <br /> Pass bytes_recvd into hyperv_receive_sub() and reject any packet that<br /> does not cover the pipe + synthvid header. A single switch on<br /> msg-&gt;vid_hdr.type then computes the type-specific payload size: the<br /> three completion-driving types (SYNTHVID_VERSION_RESPONSE,<br /> SYNTHVID_RESOLUTION_RESPONSE, SYNTHVID_VRAM_LOCATION_ACK) fall through<br /> to a shared exit that requires that size before memcpy/complete, while<br /> SYNTHVID_FEATURE_CHANGE validates its own payload and returns before<br /> reading is_dirt_needed. Unknown types are dropped.<br /> <br /> SYNTHVID_RESOLUTION_RESPONSE is variable length: the host fills<br /> resolution_count entries, not the full SYNTHVID_MAX_RESOLUTION_COUNT<br /> array. Validate the fixed prefix first so resolution_count can be<br /> read, bound it against the array, then require only the count-sized<br /> array, so the shorter responses the host actually sends are accepted.<br /> <br /> Only run the sub-handler when vmbus_recvpacket() returned success. The<br /> memcpy length is bytes_recvd, which is bounded by VMBUS_MAX_PACKET_SIZE<br /> only on a successful receive; on -ENOBUFS vmbus_recvpacket() instead<br /> reports the required length, which can exceed hv-&gt;recv_buf, so copying<br /> bytes_recvd would read and write past the 16 KiB buffers. Gating on the<br /> success return keeps the copy bounded. The nonzero-return path is itself<br /> a malformed-message case and is now logged rather than silently skipped;<br /> channel recovery is not attempted.<br /> <br /> Rejected packets are reported via drm_err_ratelimited() rather than<br /> silently dropped, matching the CoCo-hardened pattern in<br /> hv_kvp_onchannelcallback().