mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Sasha Levin <sashal@kernel.org>
To: linux-kernel@vger.kernel.org, stable@vger.kernel.org
Cc: Mike Marciniszyn <mike.marciniszyn@intel.com>,
	Kaike Wan <kaike.wan@intel.com>,
	Dennis Dalessandro <dennis.dalessandro@intel.com>,
	Doug Ledford <dledford@redhat.com>,
	Sasha Levin <sashal@kernel.org>,
	linux-rdma@vger.kernel.org
Subject: [PATCH AUTOSEL 5.1 79/95] IB/hfi1: Use aborts to trigger RC throttling
Date: Wed, 26 Jun 2019 20:30:04 -0400	[thread overview]
Message-ID: <20190627003021.19867-79-sashal@kernel.org> (raw)
In-Reply-To: <20190627003021.19867-1-sashal@kernel.org>

From: Mike Marciniszyn <mike.marciniszyn@intel.com>

[ Upstream commit 4bb02e9572af1383038d83ad196d7166c515f2ee ]

SDMA and pio flushes will cause a lot of packets to be transmitted
after a link has gone down, using a lot of CPU to retransmit
packets.

Fix for RC QPs by recognizing the flush status and:
- Forcing a timer start
- Putting the QP into a "send one" mode

Fixes: 7724105686e7 ("IB/hfi1: add driver files")
Reviewed-by: Kaike Wan <kaike.wan@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/infiniband/hw/hfi1/rc.c    | 30 ++++++++++++++++++++++++++++++
 drivers/infiniband/hw/hfi1/verbs.c | 10 ++++++----
 drivers/infiniband/hw/hfi1/verbs.h |  1 +
 3 files changed, 37 insertions(+), 4 deletions(-)

diff --git a/drivers/infiniband/hw/hfi1/rc.c b/drivers/infiniband/hw/hfi1/rc.c
index 82f101878e33..24cbac277bf0 100644
--- a/drivers/infiniband/hw/hfi1/rc.c
+++ b/drivers/infiniband/hw/hfi1/rc.c
@@ -1703,6 +1703,36 @@ static void reset_sending_psn(struct rvt_qp *qp, u32 psn)
 	}
 }
 
+/**
+ * hfi1_rc_verbs_aborted - handle abort status
+ * @qp: the QP
+ * @opah: the opa header
+ *
+ * This code modifies both ACK bit in BTH[2]
+ * and the s_flags to go into send one mode.
+ *
+ * This serves to throttle the send engine to only
+ * send a single packet in the likely case the
+ * a link has gone down.
+ */
+void hfi1_rc_verbs_aborted(struct rvt_qp *qp, struct hfi1_opa_header *opah)
+{
+	struct ib_other_headers *ohdr = hfi1_get_rc_ohdr(opah);
+	u8 opcode = ib_bth_get_opcode(ohdr);
+	u32 psn;
+
+	/* ignore responses */
+	if ((opcode >= OP(RDMA_READ_RESPONSE_FIRST) &&
+	     opcode <= OP(ATOMIC_ACKNOWLEDGE)) ||
+	    opcode == TID_OP(READ_RESP) ||
+	    opcode == TID_OP(WRITE_RESP))
+		return;
+
+	psn = ib_bth_get_psn(ohdr) | IB_BTH_REQ_ACK;
+	ohdr->bth[2] = cpu_to_be32(psn);
+	qp->s_flags |= RVT_S_SEND_ONE;
+}
+
 /*
  * This should be called with the QP s_lock held and interrupts disabled.
  */
diff --git a/drivers/infiniband/hw/hfi1/verbs.c b/drivers/infiniband/hw/hfi1/verbs.c
index 55a56b3d7f83..8d64972c6226 100644
--- a/drivers/infiniband/hw/hfi1/verbs.c
+++ b/drivers/infiniband/hw/hfi1/verbs.c
@@ -638,6 +638,8 @@ static void verbs_sdma_complete(
 		struct hfi1_opa_header *hdr;
 
 		hdr = &tx->phdr.hdr;
+		if (unlikely(status == SDMA_TXREQ_S_ABORTED))
+			hfi1_rc_verbs_aborted(qp, hdr);
 		hfi1_rc_send_complete(qp, hdr);
 	}
 	spin_unlock(&qp->s_lock);
@@ -1095,15 +1097,15 @@ int hfi1_verbs_send_pio(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
 			       &ps->s_txreq->phdr.hdr, ib_is_sc5(sc5));
 
 pio_bail:
+	spin_lock_irqsave(&qp->s_lock, flags);
 	if (qp->s_wqe) {
-		spin_lock_irqsave(&qp->s_lock, flags);
 		rvt_send_complete(qp, qp->s_wqe, wc_status);
-		spin_unlock_irqrestore(&qp->s_lock, flags);
 	} else if (qp->ibqp.qp_type == IB_QPT_RC) {
-		spin_lock_irqsave(&qp->s_lock, flags);
+		if (unlikely(wc_status == IB_WC_GENERAL_ERR))
+			hfi1_rc_verbs_aborted(qp, &ps->s_txreq->phdr.hdr);
 		hfi1_rc_send_complete(qp, &ps->s_txreq->phdr.hdr);
-		spin_unlock_irqrestore(&qp->s_lock, flags);
 	}
+	spin_unlock_irqrestore(&qp->s_lock, flags);
 
 	ret = 0;
 
diff --git a/drivers/infiniband/hw/hfi1/verbs.h b/drivers/infiniband/hw/hfi1/verbs.h
index 62ace0b2d17a..1714c0f6475d 100644
--- a/drivers/infiniband/hw/hfi1/verbs.h
+++ b/drivers/infiniband/hw/hfi1/verbs.h
@@ -415,6 +415,7 @@ void hfi1_rc_hdrerr(
 
 u8 ah_to_sc(struct ib_device *ibdev, struct rdma_ah_attr *ah_attr);
 
+void hfi1_rc_verbs_aborted(struct rvt_qp *qp, struct hfi1_opa_header *opah);
 void hfi1_rc_send_complete(struct rvt_qp *qp, struct hfi1_opa_header *opah);
 
 void hfi1_ud_rcv(struct hfi1_packet *packet);
-- 
2.20.1


  parent reply	other threads:[~2019-06-27  0:34 UTC|newest]

Thread overview: 95+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-06-27  0:28 [PATCH AUTOSEL 5.1 01/95] ARM: dts: dra76x: Disable rtc target module Sasha Levin
2019-06-27  0:28 ` [PATCH AUTOSEL 5.1 02/95] ARM: dts: dra76x: Disable usb4_tm " Sasha Levin
2019-06-27  0:28 ` [PATCH AUTOSEL 5.1 03/95] ARM: dts: dra71x: Disable rtc " Sasha Levin
2019-06-27  0:28 ` [PATCH AUTOSEL 5.1 04/95] ARM: dts: dra71x: Disable usb4_tm " Sasha Levin
2019-06-27  0:28 ` [PATCH AUTOSEL 5.1 05/95] soc: brcmstb: Fix error path for unsupported CPUs Sasha Levin
2019-06-27  0:28 ` [PATCH AUTOSEL 5.1 06/95] soc: bcm: brcmstb: biuctrl: Register writes require a barrier Sasha Levin
2019-06-27  0:28 ` [PATCH AUTOSEL 5.1 07/95] Input: elantech - enable middle button support on 2 ThinkPads Sasha Levin
2019-06-27  0:28 ` [PATCH AUTOSEL 5.1 08/95] bpf: fix out-of-bounds read in __bpf_skc_lookup Sasha Levin
2019-06-27  0:28 ` [PATCH AUTOSEL 5.1 09/95] samples, bpf: fix to change the buffer size for read() Sasha Levin
2019-06-27  0:28 ` [PATCH AUTOSEL 5.1 10/95] samples, bpf: suppress compiler warning Sasha Levin
2019-06-27  0:28 ` [PATCH AUTOSEL 5.1 11/95] bpf, riscv: clear target register high 32-bits for and/or/xor on ALU32 Sasha Levin
2019-06-27  0:28 ` [PATCH AUTOSEL 5.1 12/95] bpf: sockmap, restore sk_write_space when psock gets dropped Sasha Levin
2019-06-27  0:28 ` [PATCH AUTOSEL 5.1 13/95] mac80211: fix rate reporting inside cfg80211_calculate_bitrate_he() Sasha Levin
2019-06-27  0:28 ` [PATCH AUTOSEL 5.1 14/95] bpf: sockmap, fix use after free from sleep in psock backlog workqueue Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 15/95] soundwire: stream: fix out of boundary access on port properties Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 16/95] staging:iio:ad7150: fix threshold mode config bit Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 17/95] iio: imu: st_lsm6dsx: fix PM support for st_lsm6dsx i2c controller Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 18/95] iio: temperature: mlx90632 Relax the compatibility check Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 19/95] mac80211: mesh: fix RCU warning Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 20/95] mac80211: free peer keys before vif down in mesh Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 21/95] ARM: dts: Drop bogus CLKSEL for timer12 on dra7 Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 22/95] mwifiex: Fix possible buffer overflows at parsing bss descriptor Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 23/95] mwifiex: Abort at too short BSS descriptor element Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 24/95] bpf, riscv: clear high 32 bits for ALU32 add/sub/neg/lsh/rsh/arsh Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 25/95] iwlwifi: fix load in rfkill flow for unified firmware Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 26/95] iwlwifi: clear persistence bit according to device family Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 27/95] iwlwifi: fix AX201 killer sku loading firmware issue Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 28/95] iwlwifi: Fix double-free problems in iwl_req_fw_callback() Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 29/95] mwifiex: Fix heap overflow in mwifiex_uap_parse_tail_ies() Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 30/95] bpf: udp: ipv6: Avoid running reuseport's bpf_prog from __udp6_lib_err Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 31/95] bpf: udp: Avoid calling reuseport's bpf_prog from udp_gro Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 32/95] netfilter: ipv6: nf_defrag: fix leakage of unqueued fragments Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 33/95] tools: bpftool: Fix JSON output when lookup fails Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 34/95] soundwire: stream: fix bad unlock balance Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 35/95] soundwire: intel: set dai min and max channels correctly Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 36/95] netfilter: ipv6: nf_defrag: accept duplicate fragments again Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 37/95] dt-bindings: can: mcp251x: add mcp25625 support Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 38/95] can: mcp251x: add support for mcp25625 Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 39/95] can: m_can: implement errata "Needless activation of MRAF irq" Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 40/95] can: af_can: Fix error path of can_init() Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 41/95] can: flexcan: Remove unneeded registration message Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 42/95] net: phy: rename Asix Electronics PHY driver Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 43/95] ibmvnic: Do not close unopened driver during reset Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 44/95] ibmvnic: Refresh device multicast list after reset Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 45/95] ibmvnic: Fix unchecked return codes of memory allocations Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 46/95] ARM: dts: am335x phytec boards: Fix cd-gpios active level Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 47/95] s390/boot: disable address-of-packed-member warning Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 48/95] bpf: lpm_trie: check left child of last leftmost node for NULL Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 49/95] RISC-V: defconfig: enable clocks, serial console Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 50/95] drm/vmwgfx: Honor the sg list segment size limitation Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 51/95] drm/vmwgfx: fix a warning due to missing dma_parms Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 52/95] riscv: Fix udelay in RV32 Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 53/95] Input: imx_keypad - make sure keyboard can always wake up system Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 54/95] xdp: check device pointer before clearing Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 55/95] KVM: arm/arm64: vgic: Fix kvm_device leak in vgic_its_destroy Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 56/95] mlxsw: spectrum: Disallow prio-tagged packets when PVID is removed Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 57/95] KVM: nVMX: use correct clean fields when copying from eVMCS Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 58/95] bpf: fix div64 overflow tests to properly detect errors Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 59/95] ARM: davinci: da850-evm: call regulator_has_full_constraints() Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 60/95] ARM: davinci: da8xx: specify dma_coherent_mask for lcdc Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 61/95] gpu: ipu-v3: image-convert: Fix input bytesperline width/height align Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 62/95] gpu: ipu-v3: image-convert: Fix input bytesperline for packed formats Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 63/95] gpu: ipu-v3: image-convert: Fix image downsize coefficients Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 64/95] mac80211: only warn once on chanctx_conf being NULL Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 65/95] mac80211: do not start any work during reconfigure flow Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 66/95] cfg80211: util: fix bit count off by one Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 67/95] cfg80211: report measurement start TSF correctly Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 68/95] bpf, devmap: Fix premature entry free on destroying map Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 69/95] bpf, devmap: Add missing bulk queue free Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 70/95] bpf, devmap: Add missing RCU read lock on flush Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 71/95] bpf, x64: fix stack layout of JITed bpf code Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 72/95] qmi_wwan: add support for QMAP padding in the RX path Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 73/95] qmi_wwan: avoid RCU stalls on device disconnect when in QMAP mode Sasha Levin
2019-06-27  0:29 ` [PATCH AUTOSEL 5.1 74/95] qmi_wwan: extend permitted QMAP mux_id value range Sasha Levin
2019-06-27  0:30 ` [PATCH AUTOSEL 5.1 75/95] bpf: fix nested bpf tracepoints with per-cpu data Sasha Levin
2019-06-27  0:30 ` [PATCH AUTOSEL 5.1 76/95] mmc: core: complete HS400 before checking status Sasha Levin
2019-06-27  0:30 ` [PATCH AUTOSEL 5.1 77/95] block: fix page leak when merging to same page Sasha Levin
2019-06-27  0:30 ` [PATCH AUTOSEL 5.1 78/95] IB/hfi1: Create inline to get extended headers Sasha Levin
2019-06-27  0:30 ` Sasha Levin [this message]
2019-06-27  0:30 ` [PATCH AUTOSEL 5.1 80/95] IB/hfi1: Wakeup QPs orphaned on wait list after flush Sasha Levin
2019-06-27  0:30 ` [PATCH AUTOSEL 5.1 81/95] IB/hfi1: Handle wakeup of orphaned QPs for pio Sasha Levin
2019-06-27  0:30 ` [PATCH AUTOSEL 5.1 82/95] IB/hfi1: Handle port down properly in pio Sasha Levin
2019-06-27  0:30 ` [PATCH AUTOSEL 5.1 83/95] md: fix for divide error in status_resync Sasha Levin
2019-06-27  0:30 ` [PATCH AUTOSEL 5.1 84/95] bnx2x: Check if transceiver implements DDM before access Sasha Levin
2019-06-27  0:30 ` [PATCH AUTOSEL 5.1 85/95] drm: return -EFAULT if copy_to_user() fails Sasha Levin
2019-06-27  0:30 ` [PATCH AUTOSEL 5.1 86/95] ip6_tunnel: allow not to count pkts on tstats by passing dev as NULL Sasha Levin
2019-06-27  0:30 ` [PATCH AUTOSEL 5.1 87/95] net: lio_core: fix potential sign-extension overflow on large shift Sasha Levin
2019-06-27  0:30 ` [PATCH AUTOSEL 5.1 88/95] scsi: qedi: Check targetname while finding boot target information Sasha Levin
2019-06-27  0:30 ` [PATCH AUTOSEL 5.1 89/95] powerpc: enable a 30-bit ZONE_DMA for 32-bit pmac Sasha Levin
2019-06-27  0:30 ` [PATCH AUTOSEL 5.1 90/95] quota: fix a problem about transfer quota Sasha Levin
2019-06-27  0:30 ` [PATCH AUTOSEL 5.1 91/95] fanotify: update connector fsid cache on add mark Sasha Levin
2019-06-27  0:30 ` [PATCH AUTOSEL 5.1 92/95] net: dsa: mv88e6xxx: fix shift of FID bits in mv88e6185_g1_vtu_loadpurge() Sasha Levin
2019-06-27  0:30 ` [PATCH AUTOSEL 5.1 93/95] KVM: arm/arm64: Fix emulated ptimer irq injection Sasha Levin
2019-06-27  0:30 ` [PATCH AUTOSEL 5.1 94/95] NFS4: Only set creation opendata if O_CREAT Sasha Levin
2019-06-27  0:30 ` [PATCH AUTOSEL 5.1 95/95] net :sunrpc :clnt :Fix xps refcount imbalance on the error path Sasha Levin

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=20190627003021.19867-79-sashal@kernel.org \
    --to=sashal@kernel.org \
    --cc=dennis.dalessandro@intel.com \
    --cc=dledford@redhat.com \
    --cc=kaike.wan@intel.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-rdma@vger.kernel.org \
    --cc=mike.marciniszyn@intel.com \
    --cc=stable@vger.kernel.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

all inboxes | Powered by JetHome®