[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20190225185015.GD21863@ziepe.ca>
Date: Mon, 25 Feb 2019 11:50:15 -0700
From: Jason Gunthorpe <jgg@...pe.ca>
To: Gal Pressman <galpress@...zon.com>
Cc: Shiraz Saleem <shiraz.saleem@...el.com>, dledford@...hat.com,
davem@...emloft.net, linux-rdma@...r.kernel.org,
netdev@...r.kernel.org, mustafa.ismail@...el.com,
jeffrey.t.kirsher@...el.com, Yossi Leybovich <sleybo@...zon.com>
Subject: Re: [RFC v1 12/19] RDMA/irdma: Implement device supported verb APIs
On Sun, Feb 24, 2019 at 04:35:02PM +0200, Gal Pressman wrote:
> > +/**> + * irdma_disassociate_ucontext - Disassociate user context> + * @context: ib user context> + */> +static void irdma_disassociate_ucontext(struct ib_ucontext *context)
> > +{
> > +}
>
> What's the motivation for a nop callback (over not implementing the
> function)?
This is my fault, I didn't finish yet and conver disassociate_ucontext
into a flags once they were all made empty.
> > + ret = irdma_alloc_rsrc(iwdev->rf,
> > + iwdev->rf->allocated_mrs, iwdev->rf->max_mr,
> > + &stag_index, &next_stag_index);
> > + if (!ret) {
> > + stag = stag_index << IRDMA_CQPSQ_STAG_IDX_S;
> > + stag |= driver_key;
> > + stag += (u32)consumer_key;
> > + irdma_add_devusecount(iwdev);
> > + }
>
> This is confusing IMHO, better to test for 'if (ret)' and keep the main flow
> unindented.
Yes please follow the standard 'success oriented flow'
Jason
Powered by blists - more mailing lists