Re: [PATCH v2 1/4] RDMA/core: add bio_vec based RDMA read/write API

[Leon Romanovsky <leon@kernel.org>]

On Wed, Jan 21, 2026 at 09:42:17AM +0100, Christoph Hellwig wrote:
> On Tue, Jan 20, 2026 at 09:31:21AM -0500, Chuck Lever wrote:
> > From: Chuck Lever <chuck.lever@oracle.com>
> > 
> > The existing rdma_rw_ctx_init() API requires callers to construct a
> > scatterlist, which is then DMA-mapped page by page. Callers that
> > already have data in bio_vec form (such as the NVMe-oF target) must
> > first convert to scatterlist, adding overhead and complexity.
> > 
> > Introduce rdma_rw_ctx_init_bvec() and rdma_rw_ctx_destroy_bvec() to
> > accept bio_vec arrays directly. The new helpers use dma_map_phys()
> > for hardware RDMA devices and virtual addressing for software RDMA
> > devices (rxe, siw), avoiding intermediate scatterlist construction.
> > 
> > Memory registration (MR) path support is deferred to a follow-up
> > series; callers requiring MR-based transfers (iWARP devices or
> > force_mr=1) receive -EOPNOTSUPP and should use the scatterlist API.
> > 
> > Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
> > ---
> >  drivers/infiniband/core/rw.c | 210 +++++++++++++++++++++++++++++++++++
> >  include/rdma/ib_verbs.h      |  42 +++++++
> >  include/rdma/rw.h            |  10 ++
> >  3 files changed, 262 insertions(+)
> > 
> > diff --git a/drivers/infiniband/core/rw.c b/drivers/infiniband/core/rw.c
> > index 6354ddf2a274..59f32fecf3df 100644
> > --- a/drivers/infiniband/core/rw.c
> > +++ b/drivers/infiniband/core/rw.c
> > @@ -4,6 +4,7 @@
> >   */
> >  #include <linux/memremap.h>
> >  #include <linux/moduleparam.h>
> > +#include <linux/overflow.h>
> >  #include <linux/slab.h>
> >  #include <linux/pci-p2pdma.h>
> >  #include <rdma/mr_pool.h>
> > @@ -274,6 +275,111 @@ static int rdma_rw_init_single_wr(struct rdma_rw_ctx *ctx, struct ib_qp *qp,
> >  	return 1;
> >  }
> >  
> > +static int rdma_rw_init_single_wr_bvec(struct rdma_rw_ctx *ctx,
> > +		struct ib_qp *qp, const struct bio_vec *bvec,
> 
> Nit: maybe rename bvec to bvecs or bvec_table to make it clear this
> is the base array that the iter operates on?
> 
> > +		struct bvec_iter *iter,
> > +		u64 remote_addr, u32 rkey, enum dma_data_direction dir)
> 
> Nit 2: this can be condensed a little:
> 
> > +		struct bvec_iter *iter, u64 remote_addr, u32 rkey,
> > +		enum dma_data_direction dir)
> 
> > +static int rdma_rw_init_map_wrs_bvec(struct rdma_rw_ctx *ctx, struct ib_qp *qp,
> > +		const struct bio_vec *bvec, u32 nr_bvec,
> > +		struct bvec_iter *iter,
> > +		u64 remote_addr, u32 rkey, enum dma_data_direction dir)

Don't you both think these functions take too many arguments? It might be
worth introducing something like "struct rdma_rw_init_attrs" and passing
that instead.

At the end, these functions are for you to use.

Thanks

> 
> Same here.
> 
> Otherwise looks good:
> 
> Reviewed-by: Christoph Hellwig <hch@lst.de>
>