KVM: X86: Fix scan ioapic use-before-initialization
authorWanpeng Li <wanpengli@tencent.com>
Tue, 20 Nov 2018 08:34:18 +0000 (16:34 +0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 5 Dec 2018 18:41:23 +0000 (19:41 +0100)
commit e97f852fd4561e77721bb9a4e0ea9d98305b1e93 upstream.

Reported by syzkaller:

 BUG: unable to handle kernel NULL pointer dereference at 00000000000001c8
 PGD 80000003ec4da067 P4D 80000003ec4da067 PUD 3f7bfa067 PMD 0
 Oops: 0000 [#1] PREEMPT SMP PTI
 CPU: 7 PID: 5059 Comm: debug Tainted: G           OE     4.19.0-rc5 #16
 RIP: 0010:__lock_acquire+0x1a6/0x1990
 Call Trace:
  lock_acquire+0xdb/0x210
  _raw_spin_lock+0x38/0x70
  kvm_ioapic_scan_entry+0x3e/0x110 [kvm]
  vcpu_enter_guest+0x167e/0x1910 [kvm]
  kvm_arch_vcpu_ioctl_run+0x35c/0x610 [kvm]
  kvm_vcpu_ioctl+0x3e9/0x6d0 [kvm]
  do_vfs_ioctl+0xa5/0x690
  ksys_ioctl+0x6d/0x80
  __x64_sys_ioctl+0x1a/0x20
  do_syscall_64+0x83/0x6e0
  entry_SYSCALL_64_after_hwframe+0x49/0xbe

The reason is that the testcase writes hyperv synic HV_X64_MSR_SINT6 msr
and triggers scan ioapic logic to load synic vectors into EOI exit bitmap.
However, irqchip is not initialized by this simple testcase, ioapic/apic
objects should not be accessed.
This can be triggered by the following program:

    #define _GNU_SOURCE

    #include <endian.h>
    #include <stdint.h>
    #include <stdio.h>
    #include <stdlib.h>
    #include <string.h>
    #include <sys/syscall.h>
    #include <sys/types.h>
    #include <unistd.h>

    uint64_t r[3] = {0xffffffffffffffff, 0xffffffffffffffff, 0xffffffffffffffff};

    int main(void)
    {
     syscall(__NR_mmap, 0x20000000, 0x1000000, 3, 0x32, -1, 0);
     long res = 0;
     memcpy((void*)0x20000040, "/dev/kvm", 9);
     res = syscall(__NR_openat, 0xffffffffffffff9c, 0x20000040, 0, 0);
     if (res != -1)
     r[0] = res;
     res = syscall(__NR_ioctl, r[0], 0xae01, 0);
     if (res != -1)
     r[1] = res;
     res = syscall(__NR_ioctl, r[1], 0xae41, 0);
     if (res != -1)
     r[2] = res;
     memcpy(
     (void*)0x20000080,
     "\x01\x00\x00\x00\x00\x5b\x61\xbb\x96\x00\x00\x40\x00\x00\x00\x00\x01\x00"
     "\x08\x00\x00\x00\x00\x00\x0b\x77\xd1\x78\x4d\xd8\x3a\xed\xb1\x5c\x2e\x43"
     "\xaa\x43\x39\xd6\xff\xf5\xf0\xa8\x98\xf2\x3e\x37\x29\x89\xde\x88\xc6\x33"
     "\xfc\x2a\xdb\xb7\xe1\x4c\xac\x28\x61\x7b\x9c\xa9\xbc\x0d\xa0\x63\xfe\xfe"
     "\xe8\x75\xde\xdd\x19\x38\xdc\x34\xf5\xec\x05\xfd\xeb\x5d\xed\x2e\xaf\x22"
     "\xfa\xab\xb7\xe4\x42\x67\xd0\xaf\x06\x1c\x6a\x35\x67\x10\x55\xcb",
     106);
     syscall(__NR_ioctl, r[2], 0x4008ae89, 0x20000080);
     syscall(__NR_ioctl, r[2], 0xae80, 0);
     return 0;
    }

This patch fixes it by bailing out scan ioapic if ioapic is not initialized in
kernel.

Reported-by: Wei Wu <ww9210@gmail.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Radim Krčmář <rkrcmar@redhat.com>
Cc: Wei Wu <ww9210@gmail.com>
Signed-off-by: Wanpeng Li <wanpengli@tencent.com>
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
arch/x86/kvm/x86.c

index 7c4d02d..f243296 100644 (file)
@@ -6885,7 +6885,8 @@ static void vcpu_scan_ioapic(struct kvm_vcpu *vcpu)
        else {
                if (kvm_x86_ops->sync_pir_to_irr && vcpu->arch.apicv_active)
                        kvm_x86_ops->sync_pir_to_irr(vcpu);
-               kvm_ioapic_scan_entry(vcpu, vcpu->arch.ioapic_handled_vectors);
+               if (ioapic_in_kernel(vcpu->kvm))
+                       kvm_ioapic_scan_entry(vcpu, vcpu->arch.ioapic_handled_vectors);
        }
        bitmap_or((ulong *)eoi_exit_bitmap, vcpu->arch.ioapic_handled_vectors,
                  vcpu_to_synic(vcpu)->vec_bitmap, 256);