Re: [PATCH] KVM: x86: Reduce refcount if single_open() fails in kvm_mmu_rmaps_stat_open()

[Hou Wenlong]

On Fri, Oct 14, 2022 at 10:09:55PM +0800, Peter Xu wrote:
> On Fri, Oct 14, 2022 at 03:55:11PM +0800, Hou Wenlong wrote:
> > Refcount is increased before calling single_open() in
> > kvm_mmu_rmaps_stat_open(), If single_open() fails, refcount should be
> > restored, otherwise the vm couldn't be destroyed.
> > 
> > Fixes: 3bcd0662d66fd ("KVM: X86: Introduce mmu_rmaps_stat per-vm debugfs file")
> > Signed-off-by: Hou Wenlong <houwenlong.hwl@xxxxxxxxxxxx>
> > ---
> >  arch/x86/kvm/debugfs.c | 7 ++++++-
> >  1 file changed, 6 insertions(+), 1 deletion(-)
> > 
> > diff --git a/arch/x86/kvm/debugfs.c b/arch/x86/kvm/debugfs.c
> > index cfed36aba2f7..412ed1b341fa 100644
> > --- a/arch/x86/kvm/debugfs.c
> > +++ b/arch/x86/kvm/debugfs.c
> > @@ -162,7 +162,12 @@ static int kvm_mmu_rmaps_stat_open(struct inode *inode, struct file *file)
> >  	if (!kvm_get_kvm_safe(kvm))
> >  		return -ENOENT;
> >  
> > -	return single_open(file, kvm_mmu_rmaps_stat_show, kvm);
> > +	if (single_open(file, kvm_mmu_rmaps_stat_show, kvm)) {
> > +		kvm_put_kvm(kvm);
> > +		return -ENOMEM;
> > +	}
> 
> Thanks for fixing this.  Would it be nicer to wire the single_open retval
> (even though it'll only fail with -ENOMEM with current code base)?
> 
I agree, it would be nicer to return the single_open() retval directly.
I just follow how failure of simple_attr_open() is handled in kvm_debugfs_open(),
but I think it should be changed too. 

> -- 
> Peter Xu