Commit b7c23d30 authored by David Hildenbrand's avatar David Hildenbrand Committed by Sasha Levin

KVM: s390: reinjection of irqs can fail in the tpi handler

[ Upstream commit 15462e37 ]

The reinjection of an I/O interrupt can fail if the list is at the limit
and between the dequeue and the reinjection, another I/O interrupt is
injected (e.g. if user space floods kvm with I/O interrupts).

This patch avoids this memory leak and returns -EFAULT in this special
case. This error is not recoverable, so let's fail hard. This can later
be avoided by not dequeuing the interrupt but working directly on the
locked list.
Signed-off-by: default avatarDavid Hildenbrand <dahi@linux.vnet.ibm.com>
Cc: stable@vger.kernel.org # 3.16+
Signed-off-by: default avatarChristian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: default avatarSasha Levin <sasha.levin@oracle.com>
parent 19881aff
......@@ -930,10 +930,10 @@ int kvm_s390_inject_vm(struct kvm *kvm,
return rc;
}
void kvm_s390_reinject_io_int(struct kvm *kvm,
int kvm_s390_reinject_io_int(struct kvm *kvm,
struct kvm_s390_interrupt_info *inti)
{
__inject_vm(kvm, inti);
return __inject_vm(kvm, inti);
}
int kvm_s390_inject_vcpu(struct kvm_vcpu *vcpu,
......
......@@ -148,7 +148,7 @@ int __must_check kvm_s390_inject_vcpu(struct kvm_vcpu *vcpu,
int __must_check kvm_s390_inject_program_int(struct kvm_vcpu *vcpu, u16 code);
struct kvm_s390_interrupt_info *kvm_s390_get_io_int(struct kvm *kvm,
u64 cr6, u64 schid);
void kvm_s390_reinject_io_int(struct kvm *kvm,
int kvm_s390_reinject_io_int(struct kvm *kvm,
struct kvm_s390_interrupt_info *inti);
int kvm_s390_mask_adapter(struct kvm *kvm, unsigned int id, bool masked);
......
......@@ -278,7 +278,10 @@ static int handle_tpi(struct kvm_vcpu *vcpu)
* instruction is suppressed from the guest's view: reinject the
* interrupt.
*/
kvm_s390_reinject_io_int(vcpu->kvm, inti);
if (kvm_s390_reinject_io_int(vcpu->kvm, inti)) {
kfree(inti);
rc = -EFAULT;
}
/* don't set the cc, a pgm irq was injected or we drop to user space */
return rc ? -EFAULT : 0;
}
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment