KVM: x86: fix out-of-bounds accesses of rtc_eoi map
authorRadim Krčmář <rkrcmar@redhat.com>
Wed, 23 Nov 2016 20:15:27 +0000 (21:15 +0100)
committerRadim Krčmář <rkrcmar@redhat.com>
Thu, 24 Nov 2016 17:37:19 +0000 (18:37 +0100)
KVM was using arrays of size KVM_MAX_VCPUS with vcpu_id, but ID can be
bigger that the maximal number of VCPUs, resulting in out-of-bounds
access.

Found by syzkaller:

  BUG: KASAN: slab-out-of-bounds in __apic_accept_irq+0xb33/0xb50 at addr [...]
  Write of size 1 by task a.out/27101
  CPU: 1 PID: 27101 Comm: a.out Not tainted 4.9.0-rc5+ #49
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
   [...]
  Call Trace:
   [...] __apic_accept_irq+0xb33/0xb50 arch/x86/kvm/lapic.c:905
   [...] kvm_apic_set_irq+0x10e/0x180 arch/x86/kvm/lapic.c:495
   [...] kvm_irq_delivery_to_apic+0x732/0xc10 arch/x86/kvm/irq_comm.c:86
   [...] ioapic_service+0x41d/0x760 arch/x86/kvm/ioapic.c:360
   [...] ioapic_set_irq+0x275/0x6c0 arch/x86/kvm/ioapic.c:222
   [...] kvm_ioapic_inject_all arch/x86/kvm/ioapic.c:235
   [...] kvm_set_ioapic+0x223/0x310 arch/x86/kvm/ioapic.c:670
   [...] kvm_vm_ioctl_set_irqchip arch/x86/kvm/x86.c:3668
   [...] kvm_arch_vm_ioctl+0x1a08/0x23c0 arch/x86/kvm/x86.c:3999
   [...] kvm_vm_ioctl+0x1fa/0x1a70 arch/x86/kvm/../../../virt/kvm/kvm_main.c:3099

Reported-by: Dmitry Vyukov <dvyukov@google.com>
Cc: stable@vger.kernel.org
Fixes: af1bae5497b9 ("KVM: x86: bump KVM_MAX_VCPU_ID to 1023")
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
arch/x86/kvm/ioapic.c
arch/x86/kvm/ioapic.h

index 1a22de70f7f7d4e6267d262dbe153d0670895315..6e219e5c07d27c5dc41786953b1114b1e475e346 100644 (file)
@@ -94,7 +94,7 @@ static unsigned long ioapic_read_indirect(struct kvm_ioapic *ioapic,
 static void rtc_irq_eoi_tracking_reset(struct kvm_ioapic *ioapic)
 {
        ioapic->rtc_status.pending_eoi = 0;
-       bitmap_zero(ioapic->rtc_status.dest_map.map, KVM_MAX_VCPUS);
+       bitmap_zero(ioapic->rtc_status.dest_map.map, KVM_MAX_VCPU_ID);
 }
 
 static void kvm_rtc_eoi_tracking_restore_all(struct kvm_ioapic *ioapic);
index 7d2692a4965756143825d65d37688837d82aaca2..1cc6e54436dbaa71e4a68943456b9beaceec00f6 100644 (file)
@@ -42,13 +42,13 @@ struct kvm_vcpu;
 
 struct dest_map {
        /* vcpu bitmap where IRQ has been sent */
-       DECLARE_BITMAP(map, KVM_MAX_VCPUS);
+       DECLARE_BITMAP(map, KVM_MAX_VCPU_ID);
 
        /*
         * Vector sent to a given vcpu, only valid when
         * the vcpu's bit in map is set
         */
-       u8 vectors[KVM_MAX_VCPUS];
+       u8 vectors[KVM_MAX_VCPU_ID];
 };