linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Leon Romanovsky <leon@kernel.org>
To: Doug Ledford <dledford@redhat.com>, Jason Gunthorpe <jgg@nvidia.com>
Cc: Leon Romanovsky <leonro@nvidia.com>,
	Adit Ranadive <aditr@vmware.com>,
	Ariel Elior <aelior@marvell.com>,
	Bernard Metzler <bmt@zurich.ibm.com>,
	Christian Benvenuti <benve@cisco.com>,
	Dennis Dalessandro <dennis.dalessandro@cornelisnetworks.com>,
	Gal Pressman <galpress@amazon.com>,
	linux-kernel@vger.kernel.org, linux-rdma@vger.kernel.org,
	Michal Kalderon <mkalderon@marvell.com>,
	Mike Marciniszyn <mike.marciniszyn@cornelisnetworks.com>,
	Mustafa Ismail <mustafa.ismail@intel.com>,
	Naresh Kumar PBS <nareshkumar.pbs@broadcom.com>,
	Nelson Escobar <neescoba@cisco.com>,
	Potnuri Bharat Teja <bharat@chelsio.com>,
	Selvin Xavier <selvin.xavier@broadcom.com>,
	Shiraz Saleem <shiraz.saleem@intel.com>,
	Steve Wise <larrystevenwise@gmail.com>,
	VMware PV-Drivers <pv-drivers@vmware.com>,
	Weihang Li <liweihang@huawei.com>,
	Wenpeng Liang <liangwenpeng@huawei.com>,
	Yishai Hadas <yishaih@nvidia.com>,
	Zhu Yanjun <zyjzyj2000@gmail.com>
Subject: [PATCH rdma-next v1 7/9] RDMA/rdmavt: Decouple QP and SGE lists allocations
Date: Fri, 23 Jul 2021 14:39:49 +0300	[thread overview]
Message-ID: <f66c1e20ccefba0db3c69c58ca9c897f062b4d1c.1627040189.git.leonro@nvidia.com> (raw)
In-Reply-To: <cover.1627040189.git.leonro@nvidia.com>

From: Leon Romanovsky <leonro@nvidia.com>

The rdmavt QP has fields that are both needed for the control and data
path. Such mixed declaration caused to the very specific allocation flow
with kzalloc_node and SGE list embedded into the struct rvt_qp.

This patch separates QP creation to two: regular memory allocation for
the control path and specific code for the SGE list, while the access to
the later is performed through derefenced pointer.

Such pointer and its context are expected to be in the cache, so
performance difference is expected to be negligible, if any exists.

Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
---
 drivers/infiniband/sw/rdmavt/qp.c | 13 ++++++++-----
 include/rdma/rdmavt_qp.h          |  2 +-
 2 files changed, 9 insertions(+), 6 deletions(-)

diff --git a/drivers/infiniband/sw/rdmavt/qp.c b/drivers/infiniband/sw/rdmavt/qp.c
index e9f3d356b361..14900860985c 100644
--- a/drivers/infiniband/sw/rdmavt/qp.c
+++ b/drivers/infiniband/sw/rdmavt/qp.c
@@ -1078,7 +1078,7 @@ struct ib_qp *rvt_create_qp(struct ib_pd *ibpd,
 	int err;
 	struct rvt_swqe *swq = NULL;
 	size_t sz;
-	size_t sg_list_sz;
+	size_t sg_list_sz = 0;
 	struct ib_qp *ret = ERR_PTR(-ENOMEM);
 	struct rvt_dev_info *rdi = ib_to_rvt(ibpd->device);
 	void *priv = NULL;
@@ -1126,8 +1126,6 @@ struct ib_qp *rvt_create_qp(struct ib_pd *ibpd,
 		if (!swq)
 			return ERR_PTR(-ENOMEM);
 
-		sz = sizeof(*qp);
-		sg_list_sz = 0;
 		if (init_attr->srq) {
 			struct rvt_srq *srq = ibsrq_to_rvtsrq(init_attr->srq);
 
@@ -1137,10 +1135,13 @@ struct ib_qp *rvt_create_qp(struct ib_pd *ibpd,
 		} else if (init_attr->cap.max_recv_sge > 1)
 			sg_list_sz = sizeof(*qp->r_sg_list) *
 				(init_attr->cap.max_recv_sge - 1);
-		qp = kzalloc_node(sz + sg_list_sz, GFP_KERNEL,
-				  rdi->dparms.node);
+		qp = kzalloc_node(sizeof(*qp), GFP_KERNEL, rdi->dparms.node);
 		if (!qp)
 			goto bail_swq;
+		qp->r_sg_list =
+			kzalloc_node(sg_list_sz, GFP_KERNEL, rdi->dparms.node);
+		if (!qp->r_sg_list)
+			goto bail_qp;
 		qp->allowed_ops = get_allowed_ops(init_attr->qp_type);
 
 		RCU_INIT_POINTER(qp->next, NULL);
@@ -1328,6 +1329,7 @@ struct ib_qp *rvt_create_qp(struct ib_pd *ibpd,
 
 bail_qp:
 	kfree(qp->s_ack_queue);
+	kfree(qp->r_sg_list);
 	kfree(qp);
 
 bail_swq:
@@ -1762,6 +1764,7 @@ int rvt_destroy_qp(struct ib_qp *ibqp, struct ib_udata *udata)
 	kvfree(qp->r_rq.kwq);
 	rdi->driver_f.qp_priv_free(rdi, qp);
 	kfree(qp->s_ack_queue);
+	kfree(qp->r_sg_list);
 	rdma_destroy_ah_attr(&qp->remote_ah_attr);
 	rdma_destroy_ah_attr(&qp->alt_ah_attr);
 	free_ud_wq_attr(qp);
diff --git a/include/rdma/rdmavt_qp.h b/include/rdma/rdmavt_qp.h
index 8275954f5ce6..2e58d5e6ac0e 100644
--- a/include/rdma/rdmavt_qp.h
+++ b/include/rdma/rdmavt_qp.h
@@ -444,7 +444,7 @@ struct rvt_qp {
 	/*
 	 * This sge list MUST be last. Do not add anything below here.
 	 */
-	struct rvt_sge r_sg_list[] /* verified SGEs */
+	struct rvt_sge *r_sg_list /* verified SGEs */
 		____cacheline_aligned_in_smp;
 };
 
-- 
2.31.1


  parent reply	other threads:[~2021-07-23 11:40 UTC|newest]

Thread overview: 12+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-07-23 11:39 [PATCH rdma-next v1 0/9] QP allocation changes Leon Romanovsky
2021-07-23 11:39 ` [PATCH rdma-next v1 1/9] RDMA/hns: Don't skip IB creation flow for regular RC QP Leon Romanovsky
2021-07-23 11:39 ` [PATCH rdma-next v1 2/9] RDMA/hns: Don't overwrite supplied QP attributes Leon Romanovsky
2021-07-23 11:39 ` [PATCH rdma-next v1 3/9] RDMA/efa: Remove double QP type assignment Leon Romanovsky
2021-07-23 11:39 ` [PATCH rdma-next v1 4/9] RDMA/mlx5: Cancel pkey work before destroying device resources Leon Romanovsky
2021-07-23 11:39 ` [PATCH rdma-next v1 5/9] RDMA/mlx5: Delete device resource mutex that didn't protect anything Leon Romanovsky
2021-07-23 11:39 ` [PATCH rdma-next v1 6/9] RDMA/mlx5: Rework custom driver QP type creation Leon Romanovsky
2021-07-23 11:39 ` Leon Romanovsky [this message]
2021-07-23 11:39 ` [PATCH rdma-next v1 8/9] RDMA: Globally allocate and release QP memory Leon Romanovsky
2021-07-29 20:20   ` Nikolova, Tatyana E
2021-07-23 11:39 ` [PATCH rdma-next v1 9/9] RDMA/mlx5: Drop in-driver verbs object creations Leon Romanovsky
2021-08-03 17:22 ` [PATCH rdma-next v1 0/9] QP allocation changes Jason Gunthorpe

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=f66c1e20ccefba0db3c69c58ca9c897f062b4d1c.1627040189.git.leonro@nvidia.com \
    --to=leon@kernel.org \
    --cc=aditr@vmware.com \
    --cc=aelior@marvell.com \
    --cc=benve@cisco.com \
    --cc=bharat@chelsio.com \
    --cc=bmt@zurich.ibm.com \
    --cc=dennis.dalessandro@cornelisnetworks.com \
    --cc=dledford@redhat.com \
    --cc=galpress@amazon.com \
    --cc=jgg@nvidia.com \
    --cc=larrystevenwise@gmail.com \
    --cc=leonro@nvidia.com \
    --cc=liangwenpeng@huawei.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-rdma@vger.kernel.org \
    --cc=liweihang@huawei.com \
    --cc=mike.marciniszyn@cornelisnetworks.com \
    --cc=mkalderon@marvell.com \
    --cc=mustafa.ismail@intel.com \
    --cc=nareshkumar.pbs@broadcom.com \
    --cc=neescoba@cisco.com \
    --cc=pv-drivers@vmware.com \
    --cc=selvin.xavier@broadcom.com \
    --cc=shiraz.saleem@intel.com \
    --cc=yishaih@nvidia.com \
    --cc=zyjzyj2000@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).