All of lore.kernel.org
 help / color / mirror / Atom feed
From: Hariprasad Shenai <hariprasad@chelsio.com>
To: netdev@vger.kernel.org, linux-rdma@vger.kernel.org
Cc: davem@davemloft.net, roland@purestorage.com, kumaras@chelsio.com,
	dm@chelsio.com, swise@opengridcomputing.com, leedom@chelsio.com,
	santosh@chelsio.com, hariprasad@chelsio.com,
	nirranjan@chelsio.com
Subject: [PATCH net-next 20/31] iw_cxgb4: adjust tcp snd/rcv window based on link speed.
Date: Wed, 26 Feb 2014 20:36:59 +0530	[thread overview]
Message-ID: <1393427230-14532-21-git-send-email-hariprasad@chelsio.com> (raw)
In-Reply-To: <1393427230-14532-1-git-send-email-hariprasad@chelsio.com>

From: Steve Wise <swise@opengridcomputing.com>

40G devices need a bigger windows, so default 40G devices to snd 512K
rcv 1024K.

Fixed a bug that shows up with recv window sizes that exceed the size of
the RCV_BUFSIZ field in opt0 (>= 1024K :).  If the recv window exceeds
this, then we specify the max possible in opt0, add add the rest in via
a RX_DATA_ACK credits.

Added module option named adjust_win, defaulted to 1, that allows
disabling the 40G window bump.  This allows a user to specify the exact
default window sizes via module options snd_win and rcv_win.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
---
 drivers/infiniband/hw/cxgb4/cm.c            |   63 +++++++++++++++++++++++++--
 drivers/infiniband/hw/cxgb4/iw_cxgb4.h      |    2 +
 drivers/net/ethernet/chelsio/cxgb4/t4_msg.h |    1 +
 3 files changed, 62 insertions(+), 4 deletions(-)

diff --git a/drivers/infiniband/hw/cxgb4/cm.c b/drivers/infiniband/hw/cxgb4/cm.c
index 452ae3a..81fbc6e 100644
--- a/drivers/infiniband/hw/cxgb4/cm.c
+++ b/drivers/infiniband/hw/cxgb4/cm.c
@@ -134,6 +134,11 @@ static int snd_win = 128 * 1024;
 module_param(snd_win, int, 0644);
 MODULE_PARM_DESC(snd_win, "TCP send window in bytes (default=128KB)");
 
+static int adjust_win = 1;
+module_param(adjust_win, int, 0644);
+MODULE_PARM_DESC(adjust_win,
+		 "Adjust TCP window based on link speed (default=1)");
+
 static struct workqueue_struct *workq;
 
 static struct sk_buff_head rxq;
@@ -465,7 +470,7 @@ static void send_flowc(struct c4iw_ep *ep, struct sk_buff *skb)
 	flowc->mnemval[5].mnemonic = FW_FLOWC_MNEM_RCVNXT;
 	flowc->mnemval[5].val = cpu_to_be32(ep->rcv_seq);
 	flowc->mnemval[6].mnemonic = FW_FLOWC_MNEM_SNDBUF;
-	flowc->mnemval[6].val = cpu_to_be32(snd_win);
+	flowc->mnemval[6].val = cpu_to_be32(ep->snd_win);
 	flowc->mnemval[7].mnemonic = FW_FLOWC_MNEM_MSS;
 	flowc->mnemval[7].val = cpu_to_be32(ep->emss);
 	/* Pad WR to 16 byte boundary */
@@ -547,6 +552,7 @@ static int send_connect(struct c4iw_ep *ep)
 	struct sockaddr_in *ra = (struct sockaddr_in *)&ep->com.remote_addr;
 	struct sockaddr_in6 *la6 = (struct sockaddr_in6 *)&ep->com.local_addr;
 	struct sockaddr_in6 *ra6 = (struct sockaddr_in6 *)&ep->com.remote_addr;
+	int win;
 
 	wrlen = (ep->com.remote_addr.ss_family == AF_INET) ?
 			roundup(sizev4, 16) :
@@ -564,6 +570,15 @@ static int send_connect(struct c4iw_ep *ep)
 
 	cxgb4_best_mtu(ep->com.dev->rdev.lldi.mtus, ep->mtu, &mtu_idx);
 	wscale = compute_wscale(rcv_win);
+
+	/*
+	 * Specify the largest window that will fit in opt0. The
+	 * remainder will be specified in the rx_data_ack.
+	 */
+	win = ep->rcv_win >> 10;
+	if (win > RCV_BUFSIZ_MASK)
+		win = RCV_BUFSIZ_MASK;
+
 	opt0 = (nocong ? NO_CONG(1) : 0) |
 	       KEEP_ALIVE(1) |
 	       DELACK(1) |
@@ -574,7 +589,7 @@ static int send_connect(struct c4iw_ep *ep)
 	       SMAC_SEL(ep->smac_idx) |
 	       DSCP(ep->tos) |
 	       ULP_MODE(ULP_MODE_TCPDDP) |
-	       RCV_BUFSIZ(rcv_win>>10);
+	       RCV_BUFSIZ(win);
 	opt2 = RX_CHANNEL(0) |
 	       CCTRL_ECN(enable_ecn) |
 	       RSS_QUEUE_VALID | RSS_QUEUE(ep->rss_qid);
@@ -1134,6 +1149,14 @@ static int update_rx_credits(struct c4iw_ep *ep, u32 credits)
 		return 0;
 	}
 
+	/*
+	 * If we couldn't specify the entire rcv window at connection setup
+	 * due to the limit in the number of bits in the RCV_BUFSIZ field,
+	 * then add the overage in to the credits returned.
+	 */
+	if (ep->rcv_win > RCV_BUFSIZ_MASK * 1024)
+		credits += ep->rcv_win - RCV_BUFSIZ_MASK * 1024;
+
 	req = (struct cpl_rx_data_ack *) skb_put(skb, wrlen);
 	memset(req, 0, wrlen);
 	INIT_TP_WR(req, ep->hwtid);
@@ -1592,6 +1615,7 @@ static void send_fw_act_open_req(struct c4iw_ep *ep, unsigned int atid)
 	unsigned int mtu_idx;
 	int wscale;
 	struct sockaddr_in *sin;
+	int win;
 
 	skb = get_skb(NULL, sizeof(*req), GFP_KERNEL);
 	req = (struct fw_ofld_connection_wr *)__skb_put(skb, sizeof(*req));
@@ -1616,6 +1640,15 @@ static void send_fw_act_open_req(struct c4iw_ep *ep, unsigned int atid)
 	req->tcb.rcv_adv = htons(1);
 	cxgb4_best_mtu(ep->com.dev->rdev.lldi.mtus, ep->mtu, &mtu_idx);
 	wscale = compute_wscale(rcv_win);
+
+	/*
+	 * Specify the largest window that will fit in opt0. The
+	 * remainder will be specified in the rx_data_ack.
+	 */
+	win = ep->rcv_win >> 10;
+	if (win > RCV_BUFSIZ_MASK)
+		win = RCV_BUFSIZ_MASK;
+
 	req->tcb.opt0 = (__force __be64) (TCAM_BYPASS(1) |
 		(nocong ? NO_CONG(1) : 0) |
 		KEEP_ALIVE(1) |
@@ -1627,7 +1660,7 @@ static void send_fw_act_open_req(struct c4iw_ep *ep, unsigned int atid)
 		SMAC_SEL(ep->smac_idx) |
 		DSCP(ep->tos) |
 		ULP_MODE(ULP_MODE_TCPDDP) |
-		RCV_BUFSIZ(rcv_win >> 10));
+		RCV_BUFSIZ(win));
 	req->tcb.opt2 = (__force __be32) (PACE(1) |
 		TX_QUEUE(ep->com.dev->rdev.lldi.tx_modq[ep->tx_chan]) |
 		RX_CHANNEL(0) |
@@ -1665,6 +1698,17 @@ static int is_neg_adv(unsigned int status)
 	       status == CPL_ERR_KEEPALV_NEG_ADVICE;
 }
 
+static void set_tcp_window(struct c4iw_ep *ep, struct port_info *pi)
+{
+	ep->snd_win = snd_win;
+	ep->rcv_win = rcv_win;
+	if (adjust_win && pi->link_cfg.speed == 40000) {
+		ep->snd_win *= 4;
+		ep->rcv_win *= 4;
+	}
+	PDBG("%s snd_win %d rcv_win %d\n", __func__, ep->snd_win, ep->rcv_win);
+}
+
 #define ACT_OPEN_RETRY_COUNT 2
 
 static int import_ep(struct c4iw_ep *ep, int iptype, __u8 *peer_ip,
@@ -1713,6 +1757,7 @@ static int import_ep(struct c4iw_ep *ep, int iptype, __u8 *peer_ip,
 		ep->ctrlq_idx = cxgb4_port_idx(pdev);
 		ep->rss_qid = cdev->rdev.lldi.rxq_ids[
 			cxgb4_port_idx(pdev) * step];
+		set_tcp_window(ep, (struct port_info *)netdev_priv(pdev));
 		dev_put(pdev);
 	} else {
 		pdev = get_real_dev(n->dev);
@@ -1731,6 +1776,7 @@ static int import_ep(struct c4iw_ep *ep, int iptype, __u8 *peer_ip,
 			cdev->rdev.lldi.nchan;
 		ep->rss_qid = cdev->rdev.lldi.rxq_ids[
 			cxgb4_port_idx(n->dev) * step];
+		set_tcp_window(ep, (struct port_info *)netdev_priv(pdev));
 
 		if (clear_mpa_v1) {
 			ep->retry_with_mpa_v1 = 0;
@@ -1961,6 +2007,7 @@ static void accept_cr(struct c4iw_ep *ep, struct sk_buff *skb,
 	u64 opt0;
 	u32 opt2;
 	int wscale;
+	int win;
 
 	PDBG("%s ep %p tid %u\n", __func__, ep, ep->hwtid);
 	BUG_ON(skb_cloned(skb));
@@ -1968,6 +2015,14 @@ static void accept_cr(struct c4iw_ep *ep, struct sk_buff *skb,
 	skb_get(skb);
 	cxgb4_best_mtu(ep->com.dev->rdev.lldi.mtus, ep->mtu, &mtu_idx);
 	wscale = compute_wscale(rcv_win);
+
+	/*
+	 * Specify the largest window that will fit in opt0. The
+	 * remainder will be specified in the rx_data_ack.
+	 */
+	win = ep->rcv_win >> 10;
+	if (win > RCV_BUFSIZ_MASK)
+		win = RCV_BUFSIZ_MASK;
 	opt0 = (nocong ? NO_CONG(1) : 0) |
 	       KEEP_ALIVE(1) |
 	       DELACK(1) |
@@ -1978,7 +2033,7 @@ static void accept_cr(struct c4iw_ep *ep, struct sk_buff *skb,
 	       SMAC_SEL(ep->smac_idx) |
 	       DSCP(ep->tos >> 2) |
 	       ULP_MODE(ULP_MODE_TCPDDP) |
-	       RCV_BUFSIZ(rcv_win>>10);
+	       RCV_BUFSIZ(win);
 	opt2 = RX_CHANNEL(0) |
 	       RSS_QUEUE_VALID | RSS_QUEUE(ep->rss_qid);
 
diff --git a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h
index b75f8f5..3b6cea0 100644
--- a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h
+++ b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h
@@ -804,6 +804,8 @@ struct c4iw_ep {
 	u8 retry_with_mpa_v1;
 	u8 tried_with_mpa_v1;
 	unsigned int retry_count;
+	int snd_win;
+	int rcv_win;
 };
 
 static inline struct c4iw_ep *to_ep(struct iw_cm_id *cm_id)
diff --git a/drivers/net/ethernet/chelsio/cxgb4/t4_msg.h b/drivers/net/ethernet/chelsio/cxgb4/t4_msg.h
index f2738c7..330bc14 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/t4_msg.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/t4_msg.h
@@ -227,6 +227,7 @@ struct cpl_pass_open_req {
 #define DELACK(x)     ((x) << 5)
 #define ULP_MODE(x)   ((x) << 8)
 #define RCV_BUFSIZ(x) ((x) << 12)
+#define RCV_BUFSIZ_MASK 0x3FFU
 #define DSCP(x)       ((x) << 22)
 #define SMAC_SEL(x)   ((u64)(x) << 28)
 #define L2T_IDX(x)    ((u64)(x) << 36)
-- 
1.7.1

  parent reply	other threads:[~2014-02-26 15:06 UTC|newest]

Thread overview: 40+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2014-02-26 15:06 [PATCH net-next 00/31] Misc. fixes for cxgb4 and iw_cxgb4 Hariprasad Shenai
2014-02-26 15:06 ` [PATCH net-next 01/31] cxgb4: Fix some small bugs in t4_sge_init_soft() when our Page Size is 64KB Hariprasad Shenai
2014-02-26 15:06 ` [PATCH net-next 02/31] cxgb4: Add code to dump SGE registers when hitting idma hangs Hariprasad Shenai
2014-02-26 15:06 ` [PATCH net-next 03/31] cxgb4: Rectify emitting messages about SGE Ingress DMA channels being potentially stuck Hariprasad Shenai
2014-02-26 15:06 ` [PATCH net-next 04/31] cxgb4: Updates for T5 SGE's Egress Congestion Threshold Hariprasad Shenai
2014-02-26 15:06 ` [PATCH net-next 05/31] cxgb4: use spinlock_irqsave/spinlock_irqrestore for db lock Hariprasad Shenai
2014-02-26 15:06 ` [PATCH net-next 06/31] iw_cxgb4: cap CQ size at T4_MAX_IQ_SIZE Hariprasad Shenai
2014-02-26 15:06 ` [PATCH net-next 07/31] iw_cxgb4: Allow loopback connections Hariprasad Shenai
2014-02-26 15:06 ` [PATCH net-next 08/31] iw_cxgb4: release neigh entry in error paths Hariprasad Shenai
2014-02-26 15:06 ` [PATCH net-next 09/31] iw_cxgb4: Treat CPL_ERR_KEEPALV_NEG_ADVICE as negative advice Hariprasad Shenai
2014-02-26 15:06 ` [PATCH net-next 10/31] cxgb4/iw_cxgb4: Doorbell Drop Avoidance Bug Fixes Hariprasad Shenai
     [not found]   ` <1393427230-14532-11-git-send-email-hariprasad-ut6Up61K2wZBDgjK7y7TUQ@public.gmane.org>
2014-02-26 23:12     ` David Miller
     [not found]       ` <20140226.181216.1563503549713890339.davem-fT/PcQaiUtIeIZ0/mPfg9Q@public.gmane.org>
2014-02-27 17:11         ` Steve Wise
2014-02-27 17:11           ` Steve Wise
2014-02-27 18:05           ` David Miller
2014-02-27 18:21             ` Stephen Hemminger
     [not found]             ` <20140227.130544.1892840202381139797.davem-fT/PcQaiUtIeIZ0/mPfg9Q@public.gmane.org>
2014-03-04 23:22               ` Ben Hutchings
     [not found]                 ` <1393975366.16256.23.camel-nDn/Rdv9kqW9Jme8/bJn5UCKIB8iOfG2tUK59QYPAWc@public.gmane.org>
2014-03-05  5:34                   ` Hariprasad S
2014-03-05  5:34                     ` Hariprasad S
2014-02-26 15:06 ` [PATCH net-next 11/31] iw_cxgb4: use the BAR2/WC path for kernel QPs and T5 devices Hariprasad Shenai
2014-02-26 15:06 ` [PATCH net-next 12/31] iw_cxgb4: Fix incorrect BUG_ON conditions Hariprasad Shenai
2014-02-26 15:06 ` [PATCH net-next 13/31] iw_cxgb4: Mind the sq_sig_all/sq_sig_type QP attributes Hariprasad Shenai
2014-02-26 15:06 ` [PATCH net-next 14/31] iw_cxgb4: default peer2peer mode to 1 Hariprasad Shenai
2014-02-26 15:06 ` [PATCH net-next 15/31] iw_cxgb4: save the correct map length for fast_reg_page_lists Hariprasad Shenai
2014-02-26 15:06 ` [PATCH net-next 16/31] iw_cxgb4: don't leak skb in c4iw_uld_rx_handler() Hariprasad Shenai
2014-02-26 15:06 ` [PATCH net-next 17/31] iw_cxgb4: fix possible memory leak in RX_PKT processing Hariprasad Shenai
2014-02-26 15:06 ` [PATCH net-next 18/31] iw_cxgb4: ignore read reponse type 1 CQEs Hariprasad Shenai
2014-02-26 15:06 ` [PATCH net-next 19/31] iw_cxgb4: connect_request_upcall fixes Hariprasad Shenai
2014-02-26 15:06 ` Hariprasad Shenai [this message]
2014-02-26 15:07 ` [PATCH net-next 21/31] iw_cxgb4: update snd_seq when sending MPA messages Hariprasad Shenai
2014-02-26 15:07 ` [PATCH net-next 22/31] iw_cxgb4: lock around accept/reject downcalls Hariprasad Shenai
2014-02-26 15:07 ` [PATCH net-next 23/31] iw_cxgb4: drop RX_DATA packets if the endpoint is gone Hariprasad Shenai
2014-02-26 15:07 ` [PATCH net-next 24/31] iw_cxgb4: rx_data() needs to hold the ep mutex Hariprasad Shenai
2014-02-26 15:07 ` [PATCH net-next 25/31] iw_cxgb4: endpoint timeout fixes Hariprasad Shenai
2014-02-26 15:07 ` [PATCH net-next 26/31] iw_cxgb4: rmb() after reading valid gen bit Hariprasad Shenai
2014-02-26 15:07 ` [PATCH net-next 27/31] iw_cxgb4: wc_wmb() needed after DB writes Hariprasad Shenai
2014-02-26 15:07 ` [PATCH net-next 28/31] iw_cxgb4: SQ flush fix Hariprasad Shenai
2014-02-26 15:07 ` [PATCH net-next 29/31] iw_cxgb4: minor fixes/cleanup Hariprasad Shenai
2014-02-26 15:07 ` [PATCH net-next 30/31] iw_cxgb4: Max fastreg depth depends on DSGL support Hariprasad Shenai
2014-02-26 15:07 ` [PATCH net-next 31/31] Revert "cxgb4: Don't assume LSO only uses SGL path in t4_eth_xmit()" Hariprasad Shenai

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=1393427230-14532-21-git-send-email-hariprasad@chelsio.com \
    --to=hariprasad@chelsio.com \
    --cc=davem@davemloft.net \
    --cc=dm@chelsio.com \
    --cc=kumaras@chelsio.com \
    --cc=leedom@chelsio.com \
    --cc=linux-rdma@vger.kernel.org \
    --cc=netdev@vger.kernel.org \
    --cc=nirranjan@chelsio.com \
    --cc=roland@purestorage.com \
    --cc=santosh@chelsio.com \
    --cc=swise@opengridcomputing.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.