From: Sinan Kaya <okaya@codeaurora.org>
To: linux-rdma@vger.kernel.org, timur@codeaurora.org, sulrich@codeaurora.org
Cc: linux-arm-msm@vger.kernel.org,
linux-arm-kernel@lists.infradead.org,
Sinan Kaya <okaya@codeaurora.org>,
Faisal Latif <faisal.latif@intel.com>,
Doug Ledford <dledford@redhat.com>,
Jason Gunthorpe <jgg@ziepe.ca>,
linux-kernel@vger.kernel.org
Subject: [PATCH v4 5/6] IB/nes: Eliminate duplicate barriers on weakly-ordered archs
Date: Mon, 19 Mar 2018 22:47:47 -0400 [thread overview]
Message-ID: <1521514068-8856-6-git-send-email-okaya@codeaurora.org> (raw)
In-Reply-To: <1521514068-8856-1-git-send-email-okaya@codeaurora.org>
Code includes barrier() followed by writel(). writel() already has a
barrier on some architectures like arm64.
This ends up CPU observing two barriers back to back before executing the
register write.
Create a new wrapper function with relaxed write operator. Use the new
wrapper when a write is following a barrier().
Signed-off-by: Sinan Kaya <okaya@codeaurora.org>
---
drivers/infiniband/hw/nes/nes.h | 5 +++++
drivers/infiniband/hw/nes/nes_hw.c | 21 ++++++++++++++-------
drivers/infiniband/hw/nes/nes_mgt.c | 15 ++++++++++-----
drivers/infiniband/hw/nes/nes_nic.c | 2 +-
drivers/infiniband/hw/nes/nes_utils.c | 3 ++-
drivers/infiniband/hw/nes/nes_verbs.c | 5 +++--
6 files changed, 35 insertions(+), 16 deletions(-)
diff --git a/drivers/infiniband/hw/nes/nes.h b/drivers/infiniband/hw/nes/nes.h
index 00c27291..85e007d 100644
--- a/drivers/infiniband/hw/nes/nes.h
+++ b/drivers/infiniband/hw/nes/nes.h
@@ -387,6 +387,11 @@ static inline void nes_write_indexed(struct nes_device *nesdev, u32 reg_index, u
spin_unlock_irqrestore(&nesdev->indexed_regs_lock, flags);
}
+static inline void nes_write32_relaxed(void __iomem *addr, u32 val)
+{
+ writel_relaxed(val, addr);
+}
+
static inline void nes_write32(void __iomem *addr, u32 val)
{
writel(val, addr);
diff --git a/drivers/infiniband/hw/nes/nes_hw.c b/drivers/infiniband/hw/nes/nes_hw.c
index 18a7de1..568e17d 100644
--- a/drivers/infiniband/hw/nes/nes_hw.c
+++ b/drivers/infiniband/hw/nes/nes_hw.c
@@ -1257,7 +1257,8 @@ int nes_destroy_cqp(struct nes_device *nesdev)
barrier();
/* Ring doorbell (5 WQEs) */
- nes_write32(nesdev->regs+NES_WQE_ALLOC, 0x05800000 | nesdev->cqp.qp_id);
+ nes_write32_relaxed(nesdev->regs+NES_WQE_ALLOC,
+ 0x05800000 | nesdev->cqp.qp_id);
spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
@@ -1594,7 +1595,8 @@ static void nes_replenish_nic_rq(struct nes_vnic *nesvnic)
atomic_dec(&nesvnic->rx_skbs_needed);
barrier();
if (++rx_wqes_posted == 255) {
- nes_write32(nesdev->regs+NES_WQE_ALLOC, (rx_wqes_posted << 24) | nesnic->qp_id);
+ nes_write32_relaxed(nesdev->regs+NES_WQE_ALLOC,
+ (rx_wqes_posted << 24) | nesnic->qp_id);
rx_wqes_posted = 0;
}
} else {
@@ -1612,7 +1614,8 @@ static void nes_replenish_nic_rq(struct nes_vnic *nesvnic)
} while (atomic_read(&nesvnic->rx_skbs_needed));
barrier();
if (rx_wqes_posted)
- nes_write32(nesdev->regs+NES_WQE_ALLOC, (rx_wqes_posted << 24) | nesnic->qp_id);
+ nes_write32_relaxed(nesdev->regs+NES_WQE_ALLOC,
+ (rx_wqes_posted << 24) | nesnic->qp_id);
nesnic->replenishing_rq = 0;
}
@@ -1795,7 +1798,8 @@ int nes_init_nic_qp(struct nes_device *nesdev, struct net_device *netdev)
barrier();
/* Ring doorbell (2 WQEs) */
- nes_write32(nesdev->regs+NES_WQE_ALLOC, 0x02800000 | nesdev->cqp.qp_id);
+ nes_write32_relaxed(nesdev->regs+NES_WQE_ALLOC,
+ 0x02800000 | nesdev->cqp.qp_id);
spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
nes_debug(NES_DBG_INIT, "Waiting for create NIC QP%u to complete.\n",
@@ -1844,7 +1848,8 @@ int nes_init_nic_qp(struct nes_device *nesdev, struct net_device *netdev)
do {
counter = min(wqe_count, ((u32)255));
wqe_count -= counter;
- nes_write32(nesdev->regs+NES_WQE_ALLOC, (counter << 24) | nesvnic->nic.qp_id);
+ nes_write32_relaxed(nesdev->regs+NES_WQE_ALLOC,
+ (counter << 24) | nesvnic->nic.qp_id);
} while (wqe_count);
timer_setup(&nesvnic->rq_wqes_timer, nes_rq_wqes_timeout, 0);
nes_debug(NES_DBG_INIT, "NAPI support Enabled\n");
@@ -1988,7 +1993,8 @@ void nes_destroy_nic_qp(struct nes_vnic *nesvnic)
barrier();
/* Ring doorbell (2 WQEs) */
- nes_write32(nesdev->regs+NES_WQE_ALLOC, 0x02800000 | nesdev->cqp.qp_id);
+ nes_write32_relaxed(nesdev->regs+NES_WQE_ALLOC,
+ 0x02800000 | nesdev->cqp.qp_id);
spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
nes_debug(NES_DBG_SHUTDOWN, "Waiting for CQP, cqp_head=%u, cqp.sq_head=%u,"
@@ -3064,7 +3070,8 @@ static void nes_cqp_ce_handler(struct nes_device *nesdev, struct nes_hw_cq *cq)
cqp_request, le32_to_cpu(cqp_wqe->wqe_words[NES_CQP_WQE_OPCODE_IDX])&0x3f, head);
/* Ring doorbell (1 WQEs) */
barrier();
- nes_write32(nesdev->regs+NES_WQE_ALLOC, 0x01800000 | nesdev->cqp.qp_id);
+ nes_write32_relaxed(nesdev->regs+NES_WQE_ALLOC,
+ 0x01800000 | nesdev->cqp.qp_id);
}
spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
diff --git a/drivers/infiniband/hw/nes/nes_mgt.c b/drivers/infiniband/hw/nes/nes_mgt.c
index 21e0ebd..5c5073c 100644
--- a/drivers/infiniband/hw/nes/nes_mgt.c
+++ b/drivers/infiniband/hw/nes/nes_mgt.c
@@ -96,7 +96,8 @@ static void nes_replenish_mgt_rq(struct nes_vnic_mgt *mgtvnic)
atomic_dec(&mgtvnic->rx_skbs_needed);
barrier();
if (++rx_wqes_posted == 255) {
- nes_write32(nesdev->regs + NES_WQE_ALLOC, (rx_wqes_posted << 24) | nesmgt->qp_id);
+ nes_write32_relaxed(nesdev->regs + NES_WQE_ALLOC,
+ (rx_wqes_posted << 24) | nesmgt->qp_id);
rx_wqes_posted = 0;
}
} else {
@@ -115,7 +116,8 @@ static void nes_replenish_mgt_rq(struct nes_vnic_mgt *mgtvnic)
} while (atomic_read(&mgtvnic->rx_skbs_needed));
barrier();
if (rx_wqes_posted)
- nes_write32(nesdev->regs + NES_WQE_ALLOC, (rx_wqes_posted << 24) | nesmgt->qp_id);
+ nes_write32_relaxed(nesdev->regs + NES_WQE_ALLOC,
+ (rx_wqes_posted << 24) | nesmgt->qp_id);
nesmgt->replenishing_rq = 0;
}
@@ -995,7 +997,8 @@ int nes_init_mgt_qp(struct nes_device *nesdev, struct net_device *netdev, struct
barrier();
/* Ring doorbell (2 WQEs) */
- nes_write32(nesdev->regs + NES_WQE_ALLOC, 0x02800000 | nesdev->cqp.qp_id);
+ nes_write32_relaxed(nesdev->regs + NES_WQE_ALLOC,
+ 0x02800000 | nesdev->cqp.qp_id);
spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
nes_debug(NES_DBG_INIT, "Waiting for create MGT QP%u to complete.\n",
@@ -1050,7 +1053,8 @@ int nes_init_mgt_qp(struct nes_device *nesdev, struct net_device *netdev, struct
do {
counter = min(wqe_count, ((u32)255));
wqe_count -= counter;
- nes_write32(nesdev->regs + NES_WQE_ALLOC, (counter << 24) | mgtvnic->mgt.qp_id);
+ nes_write32_relaxed(nesdev->regs + NES_WQE_ALLOC,
+ (counter << 24) | mgtvnic->mgt.qp_id);
} while (wqe_count);
nes_write32(nesdev->regs + NES_CQE_ALLOC, NES_CQE_ALLOC_NOTIFY_NEXT |
@@ -1124,7 +1128,8 @@ void nes_destroy_mgt(struct nes_vnic *nesvnic)
barrier();
/* Ring doorbell (2 WQEs) */
- nes_write32(nesdev->regs + NES_WQE_ALLOC, 0x02800000 | nesdev->cqp.qp_id);
+ nes_write32_relaxed(nesdev->regs + NES_WQE_ALLOC,
+ 0x02800000 | nesdev->cqp.qp_id);
spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
nes_debug(NES_DBG_SHUTDOWN, "Waiting for CQP, cqp_head=%u, cqp.sq_head=%u,"
diff --git a/drivers/infiniband/hw/nes/nes_nic.c b/drivers/infiniband/hw/nes/nes_nic.c
index 0a75164..0653da0 100644
--- a/drivers/infiniband/hw/nes/nes_nic.c
+++ b/drivers/infiniband/hw/nes/nes_nic.c
@@ -683,7 +683,7 @@ static int nes_netdev_start_xmit(struct sk_buff *skb, struct net_device *netdev)
barrier();
if (wqe_count)
- nes_write32(nesdev->regs+NES_WQE_ALLOC,
+ nes_write32_relaxed(nesdev->regs+NES_WQE_ALLOC,
(wqe_count << 24) | (1 << 23) | nesvnic->nic.qp_id);
netif_trans_update(netdev);
diff --git a/drivers/infiniband/hw/nes/nes_utils.c b/drivers/infiniband/hw/nes/nes_utils.c
index 21b4a83..79a3d98 100644
--- a/drivers/infiniband/hw/nes/nes_utils.c
+++ b/drivers/infiniband/hw/nes/nes_utils.c
@@ -661,7 +661,8 @@ void nes_post_cqp_request(struct nes_device *nesdev,
barrier();
/* Ring doorbell (1 WQEs) */
- nes_write32(nesdev->regs+NES_WQE_ALLOC, 0x01800000 | nesdev->cqp.qp_id);
+ nes_write32_relaxed(nesdev->regs+NES_WQE_ALLOC,
+ 0x01800000 | nesdev->cqp.qp_id);
barrier();
} else {
diff --git a/drivers/infiniband/hw/nes/nes_verbs.c b/drivers/infiniband/hw/nes/nes_verbs.c
index 162475a..87f8635 100644
--- a/drivers/infiniband/hw/nes/nes_verbs.c
+++ b/drivers/infiniband/hw/nes/nes_verbs.c
@@ -3310,7 +3310,7 @@ static int nes_post_send(struct ib_qp *ibqp, struct ib_send_wr *ib_wr,
while (wqe_count) {
counter = min(wqe_count, ((u32)255));
wqe_count -= counter;
- nes_write32(nesdev->regs + NES_WQE_ALLOC,
+ nes_write32_relaxed(nesdev->regs + NES_WQE_ALLOC,
(counter << 24) | 0x00800000 | nesqp->hwqp.qp_id);
}
@@ -3404,7 +3404,8 @@ static int nes_post_recv(struct ib_qp *ibqp, struct ib_recv_wr *ib_wr,
while (wqe_count) {
counter = min(wqe_count, ((u32)255));
wqe_count -= counter;
- nes_write32(nesdev->regs+NES_WQE_ALLOC, (counter<<24) | nesqp->hwqp.qp_id);
+ nes_write32_relaxed(nesdev->regs+NES_WQE_ALLOC,
+ (counter<<24) | nesqp->hwqp.qp_id);
}
spin_unlock_irqrestore(&nesqp->lock, flags);
--
2.7.4
next prev parent reply other threads:[~2018-03-20 2:47 UTC|newest]
Thread overview: 47+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-03-20 2:47 [PATCH v4 0/6] ib: Eliminate duplicate barriers on weakly-ordered archs Sinan Kaya
2018-03-20 2:47 ` [PATCH v4 1/6] RDMA/bnxt_re: " Sinan Kaya
2018-03-20 14:48 ` Jason Gunthorpe
2018-03-20 15:00 ` Sinan Kaya
2018-03-20 15:08 ` Sinan Kaya
2018-03-20 15:23 ` Jason Gunthorpe
2018-03-20 15:20 ` Jason Gunthorpe
2018-03-20 15:30 ` Sinan Kaya
2018-03-20 16:02 ` Jason Gunthorpe
2018-03-20 2:47 ` [PATCH v4 2/6] IB/mlx4: " Sinan Kaya
2018-03-20 14:48 ` Jason Gunthorpe
2018-03-20 2:47 ` [PATCH v4 3/6] RDMA/i40iw: " Sinan Kaya
2018-03-20 14:56 ` Jason Gunthorpe
2018-03-21 13:38 ` Shiraz Saleem
2018-03-21 20:02 ` Jason Gunthorpe
2018-03-21 21:01 ` Sinan Kaya
2018-03-20 2:47 ` [PATCH v4 4/6] infiniband: cxgb4: " Sinan Kaya
2018-03-20 14:51 ` Jason Gunthorpe
2018-03-20 15:10 ` Steve Wise
2018-03-20 15:38 ` Steve Wise
2018-03-22 6:44 ` kbuild test robot
2018-03-22 12:24 ` okaya
2018-03-22 12:48 ` okaya
2018-03-22 14:33 ` Sinan Kaya
2018-03-22 14:40 ` Steve Wise
2018-03-22 14:52 ` Sinan Kaya
2018-03-22 16:28 ` Steve Wise
2018-03-22 19:44 ` Casey Leedom
2018-03-22 20:16 ` Jason Gunthorpe
2018-03-22 20:45 ` Casey Leedom
2018-03-22 21:25 ` Jason Gunthorpe
2018-03-22 21:27 ` Sinan Kaya
2018-03-22 22:02 ` Casey Leedom
[not found] ` <437ab002-b8db-24aa-583e-0e61d61aaa97@codeaurora.org>
2018-03-22 18:46 ` Jason Gunthorpe
2018-03-22 18:48 ` Jason Gunthorpe
2018-03-22 18:58 ` Sinan Kaya
2018-03-23 4:14 ` kbuild test robot
2018-03-20 2:47 ` Sinan Kaya [this message]
2018-03-20 14:54 ` [PATCH v4 5/6] IB/nes: " Jason Gunthorpe
2018-03-20 15:23 ` Sinan Kaya
2018-03-20 16:01 ` Jason Gunthorpe
2018-03-20 16:08 ` Sinan Kaya
2018-03-20 16:29 ` Jason Gunthorpe
2018-03-20 2:47 ` [PATCH v4 6/6] RDMA/qedr: eliminate duplicate barriers on weakly-ordered archs #2 Sinan Kaya
2018-03-20 7:38 ` Kalderon, Michal
2018-03-20 14:55 ` Jason Gunthorpe
2018-03-21 20:08 ` [PATCH v4 0/6] ib: Eliminate duplicate barriers on weakly-ordered archs 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=1521514068-8856-6-git-send-email-okaya@codeaurora.org \
--to=okaya@codeaurora.org \
--cc=dledford@redhat.com \
--cc=faisal.latif@intel.com \
--cc=jgg@ziepe.ca \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-arm-msm@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-rdma@vger.kernel.org \
--cc=sulrich@codeaurora.org \
--cc=timur@codeaurora.org \
/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).