Re: [PATCH rdma-rc 1/3] RDMA/core: Postpone uobject cleanup on failure till FD close

[Leon Romanovsky <leon@kernel.org>]

On Tue, Oct 27, 2020 at 01:55:08PM -0300, Jason Gunthorpe wrote:
> On Mon, Oct 12, 2020 at 07:55:58AM +0300, Leon Romanovsky wrote:
> > @@ -543,17 +537,9 @@ static int __must_check destroy_hw_idr_uobject(struct ib_uobject *uobj,
> >  			     struct uverbs_obj_idr_type, type);
> >  	int ret = idr_type->destroy_object(uobj, why, attrs);
> >
> > -	/*
> > -	 * We can only fail gracefully if the user requested to destroy the
> > -	 * object or when a retry may be called upon an error.
> > -	 * In the rest of the cases, just remove whatever you can.
> > -	 */
> > -	if (ib_is_destroy_retryable(ret, why, uobj))
> > +	if (ret)
> >  		return ret;
> >
> > -	if (why == RDMA_REMOVE_ABORT)
> > -		return 0;
>
> This shouldn't be deleted..

Yes, this is what we noticed too.

I'll add your version for a couple of days to our regression.

Thanks