Re: kvm_set_user_memory_region() doesn't check ioctl return value

[Paolo Bonzini]

On 07/01/20 15:52, Peter Maydell wrote:
> Coverity has just pointed out (CID 1412229) that the function
> kvm_set_user_memory_region() in accel/kvm/kvm-all.c calls
> kvm_vm_ioctl(s, KVM_SET_USER_MEMORY_REGION, ...) twice, but
> it only checks the ioctl return value for failure in the
> second case. Shouldn't it be checking both calls?

The first ioctl is basically setting up things so that the second can
succeed.  My understanding of the code (and the comment above the first
call especially) is that if the first ioctl fails, it will cause the
second to fail as well.

However, it would indeed by cleaner to check both.

Paolo

> (This code seems to have been in place for years so I guess
> Coverity has only just decided that kvm_vm_ioctl() passes its
> 'needs return value checking' heuristic...)