CVE-2022-48943

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> KVM: x86/mmu: make apf token non-zero to fix bug<br /> <br /> In current async pagefault logic, when a page is ready, KVM relies on<br /> kvm_arch_can_dequeue_async_page_present() to determine whether to deliver<br /> a READY event to the Guest. This function test token value of struct<br /> kvm_vcpu_pv_apf_data, which must be reset to zero by Guest kernel when a<br /> READY event is finished by Guest. If value is zero meaning that a READY<br /> event is done, so the KVM can deliver another.<br /> But the kvm_arch_setup_async_pf() may produce a valid token with zero<br /> value, which is confused with previous mention and may lead the loss of<br /> this READY event.<br /> <br /> This bug may cause task blocked forever in Guest:<br /> INFO: task stress:7532 blocked for more than 1254 seconds.<br /> Not tainted 5.10.0 #16<br /> "echo 0 &gt; /proc/sys/kernel/hung_task_timeout_secs" disables this message.<br /> task:stress state:D stack: 0 pid: 7532 ppid: 1409<br /> flags:0x00000080<br /> Call Trace:<br /> __schedule+0x1e7/0x650<br /> schedule+0x46/0xb0<br /> kvm_async_pf_task_wait_schedule+0xad/0xe0<br /> ? exit_to_user_mode_prepare+0x60/0x70<br /> __kvm_handle_async_pf+0x4f/0xb0<br /> ? asm_exc_page_fault+0x8/0x30<br /> exc_page_fault+0x6f/0x110<br /> ? asm_exc_page_fault+0x8/0x30<br /> asm_exc_page_fault+0x1e/0x30<br /> RIP: 0033:0x402d00<br /> RSP: 002b:00007ffd31912500 EFLAGS: 00010206<br /> RAX: 0000000000071000 RBX: ffffffffffffffff RCX: 00000000021a32b0<br /> RDX: 000000000007d011 RSI: 000000000007d000 RDI: 00000000021262b0<br /> RBP: 00000000021262b0 R08: 0000000000000003 R09: 0000000000000086<br /> R10: 00000000000000eb R11: 00007fefbdf2baa0 R12: 0000000000000000<br /> R13: 0000000000000002 R14: 000000000007d000 R15: 0000000000001000