linux-rdma.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH 0/7] IB/hfi1: Another round of fixes
@ 2016-10-25 20:12 Dennis Dalessandro
       [not found] ` <20161025194241.15765.8903.stgit-9QXIwq+3FY+1XWohqUldA0EOCMrvLtNR@public.gmane.org>
  0 siblings, 1 reply; 8+ messages in thread
From: Dennis Dalessandro @ 2016-10-25 20:12 UTC (permalink / raw)
  To: dledford-H+wXaHxf7aLQT0dZR+AlfA
  Cc: Mike Marciniszyn, Dean Luick, Jakub Pawlak, Ira Weiny,
	Krzysztof Blaszkowski, Tymoteusz Kielan,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA, Dan Carpenter,
	Sebastian Sanchez, Jianxin Xiong, Dasaratharaman Chandramouli

Doug,

Here is another set of mostly small fixes. I think it could go in the RC, but if
not I can respin for next.

Can also be found in GitHub repo at:
https://github.com/ddalessa/kernel/tree/for-4.9

---

Dasaratharaman Chandramouli (1):
      IB/hfi1: Fix ECN processing in prescan_rxq

Dennis Dalessandro (1):
      IB/hfi1: Remove incorrect IS_ERR check

Jakub Pawlak (1):
      IB/hfi1: Fix status error code for unsupported packets

Jianxin Xiong (1):
      IB/hfi1: Prevent hardware counter names from being cut off

Krzysztof Blaszkowski (1):
      IB/hfi1: Relocate rcvhdrcnt module parameter check.

Sebastian Sanchez (2):
      IB/hfi1: Get rid of divide in pio buffer allocator
      IB/hfi1: Optimize pio_buf and send_context structs


 drivers/infiniband/hw/hfi1/chip.c      |    2 +
 drivers/infiniband/hw/hfi1/chip.h      |    3 ++
 drivers/infiniband/hw/hfi1/driver.c    |   37 +++++++++++++++++++--------
 drivers/infiniband/hw/hfi1/init.c      |   44 +++++++++++++++++++-------------
 drivers/infiniband/hw/hfi1/pio.c       |   16 +++++++-----
 drivers/infiniband/hw/hfi1/pio.h       |   30 ++++++++++++----------
 drivers/infiniband/hw/hfi1/pio_copy.c  |   22 ++++++++--------
 drivers/infiniband/hw/hfi1/user_sdma.c |    2 +
 8 files changed, 93 insertions(+), 63 deletions(-)

--
-Denny
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

^ permalink raw reply	[flat|nested] 8+ messages in thread

* [PATCH 1/7] IB/hfi1: Relocate rcvhdrcnt module parameter check.
       [not found] ` <20161025194241.15765.8903.stgit-9QXIwq+3FY+1XWohqUldA0EOCMrvLtNR@public.gmane.org>
@ 2016-10-25 20:12   ` Dennis Dalessandro
  2016-10-25 20:12   ` [PATCH 2/7] IB/hfi1: Fix status error code for unsupported packets Dennis Dalessandro
                     ` (5 subsequent siblings)
  6 siblings, 0 replies; 8+ messages in thread
From: Dennis Dalessandro @ 2016-10-25 20:12 UTC (permalink / raw)
  To: dledford-H+wXaHxf7aLQT0dZR+AlfA
  Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, Tymoteusz Kielan, Dean Luick,
	Krzysztof Blaszkowski

From: Krzysztof Blaszkowski <krzysztof.blaszkowski-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

Validate the rcvhdrcnt module parameter in a single function at module
load time. This allows proper error reporting.

Reviewed-by: Dean Luick <dean.luick-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Krzysztof Blaszkowski <krzysztof.blaszkowski-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Tymoteusz Kielan <tymoteusz.kielan-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/infiniband/hw/hfi1/init.c |   44 ++++++++++++++++++++++---------------
 1 files changed, 26 insertions(+), 18 deletions(-)

diff --git a/drivers/infiniband/hw/hfi1/init.c b/drivers/infiniband/hw/hfi1/init.c
index c317e69..3920bbf 100644
--- a/drivers/infiniband/hw/hfi1/init.c
+++ b/drivers/infiniband/hw/hfi1/init.c
@@ -262,13 +262,6 @@ struct hfi1_ctxtdata *hfi1_create_ctxtdata(struct hfi1_pportdata *ppd, u32 ctxt,
 		}
 		rcd->eager_base = base * dd->rcv_entries.group_size;
 
-		/* Validate and initialize Rcv Hdr Q variables */
-		if (rcvhdrcnt % HDRQ_INCREMENT) {
-			dd_dev_err(dd,
-				   "ctxt%u: header queue count %d must be divisible by %lu\n",
-				   rcd->ctxt, rcvhdrcnt, HDRQ_INCREMENT);
-			goto bail;
-		}
 		rcd->rcvhdrq_cnt = rcvhdrcnt;
 		rcd->rcvhdrqentsize = hfi1_hdrq_entsize;
 		/*
@@ -1399,6 +1392,29 @@ static void postinit_cleanup(struct hfi1_devdata *dd)
 	hfi1_free_devdata(dd);
 }
 
+static int init_validate_rcvhdrcnt(struct device *dev, uint thecnt)
+{
+	if (thecnt <= HFI1_MIN_HDRQ_EGRBUF_CNT) {
+		hfi1_early_err(dev, "Receive header queue count too small\n");
+		return -EINVAL;
+	}
+
+	if (thecnt > HFI1_MAX_HDRQ_EGRBUF_CNT) {
+		hfi1_early_err(dev,
+			       "Receive header queue count cannot be greater than %u\n",
+			       HFI1_MAX_HDRQ_EGRBUF_CNT);
+		return -EINVAL;
+	}
+
+	if (thecnt % HDRQ_INCREMENT) {
+		hfi1_early_err(dev, "Receive header queue count %d must be divisible by %lu\n",
+			       thecnt, HDRQ_INCREMENT);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
 static int init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 {
 	int ret = 0, j, pidx, initfail;
@@ -1409,18 +1425,10 @@ static int init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	HFI1_CAP_LOCK();
 
 	/* Validate some global module parameters */
-	if (rcvhdrcnt <= HFI1_MIN_HDRQ_EGRBUF_CNT) {
-		hfi1_early_err(&pdev->dev, "Header queue  count too small\n");
-		ret = -EINVAL;
-		goto bail;
-	}
-	if (rcvhdrcnt > HFI1_MAX_HDRQ_EGRBUF_CNT) {
-		hfi1_early_err(&pdev->dev,
-			       "Receive header queue count cannot be greater than %u\n",
-			       HFI1_MAX_HDRQ_EGRBUF_CNT);
-		ret = -EINVAL;
+	ret = init_validate_rcvhdrcnt(&pdev->dev, rcvhdrcnt);
+	if (ret)
 		goto bail;
-	}
+
 	/* use the encoding function as a sanitization check */
 	if (!encode_rcv_header_entry_size(hfi1_hdrq_entsize)) {
 		hfi1_early_err(&pdev->dev, "Invalid HdrQ Entry size %u\n",

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

^ permalink raw reply related	[flat|nested] 8+ messages in thread

* [PATCH 2/7] IB/hfi1: Fix status error code for unsupported packets
       [not found] ` <20161025194241.15765.8903.stgit-9QXIwq+3FY+1XWohqUldA0EOCMrvLtNR@public.gmane.org>
  2016-10-25 20:12   ` [PATCH 1/7] IB/hfi1: Relocate rcvhdrcnt module parameter check Dennis Dalessandro
@ 2016-10-25 20:12   ` Dennis Dalessandro
  2016-10-25 20:12   ` [PATCH 3/7] IB/hfi1: Fix ECN processing in prescan_rxq Dennis Dalessandro
                     ` (4 subsequent siblings)
  6 siblings, 0 replies; 8+ messages in thread
From: Dennis Dalessandro @ 2016-10-25 20:12 UTC (permalink / raw)
  To: dledford-H+wXaHxf7aLQT0dZR+AlfA
  Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, Jakub Pawlak

From: Jakub Pawlak <jakub.pawlak-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

Set the status code BAD_L2 when unsupported type of packet
is received and dropped.

Reviewed-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Jakub Pawlak <jakub.pawlak-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/infiniband/hw/hfi1/chip.h   |    3 +++
 drivers/infiniband/hw/hfi1/driver.c |   17 +++++++++++++++--
 2 files changed, 18 insertions(+), 2 deletions(-)

diff --git a/drivers/infiniband/hw/hfi1/chip.h b/drivers/infiniband/hw/hfi1/chip.h
index 9234525..043fd21 100644
--- a/drivers/infiniband/hw/hfi1/chip.h
+++ b/drivers/infiniband/hw/hfi1/chip.h
@@ -320,6 +320,9 @@
 /* DC_DC8051_CFG_MODE.GENERAL bits */
 #define DISABLE_SELF_GUID_CHECK 0x2
 
+/* Bad L2 frame error code */
+#define BAD_L2_ERR      0x6
+
 /*
  * Eager buffer minimum and maximum sizes supported by the hardware.
  * All power-of-two sizes in between are supported as well.
diff --git a/drivers/infiniband/hw/hfi1/driver.c b/drivers/infiniband/hw/hfi1/driver.c
index 6563e4d..dadd35e 100644
--- a/drivers/infiniband/hw/hfi1/driver.c
+++ b/drivers/infiniband/hw/hfi1/driver.c
@@ -1360,12 +1360,25 @@ int process_receive_ib(struct hfi1_packet *packet)
 
 int process_receive_bypass(struct hfi1_packet *packet)
 {
+	struct hfi1_devdata *dd = packet->rcd->dd;
+
 	if (unlikely(rhf_err_flags(packet->rhf)))
 		handle_eflags(packet);
 
-	dd_dev_err(packet->rcd->dd,
+	dd_dev_err(dd,
 		   "Bypass packets are not supported in normal operation. Dropping\n");
-	incr_cntr64(&packet->rcd->dd->sw_rcv_bypass_packet_errors);
+	incr_cntr64(&dd->sw_rcv_bypass_packet_errors);
+	if (!(dd->err_info_rcvport.status_and_code & OPA_EI_STATUS_SMASK)) {
+		u64 *flits = packet->ebuf;
+
+		if (flits && !(packet->rhf & RHF_LEN_ERR)) {
+			dd->err_info_rcvport.packet_flit1 = flits[0];
+			dd->err_info_rcvport.packet_flit2 =
+				packet->tlen > sizeof(flits[0]) ? flits[1] : 0;
+		}
+		dd->err_info_rcvport.status_and_code |=
+			(OPA_EI_STATUS_SMASK | BAD_L2_ERR);
+	}
 	return RHF_RCV_CONTINUE;
 }
 

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

^ permalink raw reply related	[flat|nested] 8+ messages in thread

* [PATCH 3/7] IB/hfi1: Fix ECN processing in prescan_rxq
       [not found] ` <20161025194241.15765.8903.stgit-9QXIwq+3FY+1XWohqUldA0EOCMrvLtNR@public.gmane.org>
  2016-10-25 20:12   ` [PATCH 1/7] IB/hfi1: Relocate rcvhdrcnt module parameter check Dennis Dalessandro
  2016-10-25 20:12   ` [PATCH 2/7] IB/hfi1: Fix status error code for unsupported packets Dennis Dalessandro
@ 2016-10-25 20:12   ` Dennis Dalessandro
  2016-10-25 20:12   ` [PATCH 4/7] IB/hfi1: Get rid of divide in pio buffer allocator Dennis Dalessandro
                     ` (3 subsequent siblings)
  6 siblings, 0 replies; 8+ messages in thread
From: Dennis Dalessandro @ 2016-10-25 20:12 UTC (permalink / raw)
  To: dledford-H+wXaHxf7aLQT0dZR+AlfA
  Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, Ira Weiny,
	Dasaratharaman Chandramouli

From: Dasaratharaman Chandramouli <dasaratharaman.chandramouli-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

When processing ECN via the prescan_rxq path, some fields in the packet
structure are passed uninitialized. This can potentially
cause NULL pointer exceptions during ECN handling.

Reviewed-by: Ira Weiny <ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Reviewed-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/infiniband/hw/hfi1/driver.c |   20 +++++++++++---------
 1 files changed, 11 insertions(+), 9 deletions(-)

diff --git a/drivers/infiniband/hw/hfi1/driver.c b/drivers/infiniband/hw/hfi1/driver.c
index dadd35e..c5efff2 100644
--- a/drivers/infiniband/hw/hfi1/driver.c
+++ b/drivers/infiniband/hw/hfi1/driver.c
@@ -599,7 +599,6 @@ static void __prescan_rxq(struct hfi1_packet *packet)
 					 dd->rhf_offset;
 		struct rvt_qp *qp;
 		struct ib_header *hdr;
-		struct ib_other_headers *ohdr;
 		struct rvt_dev_info *rdi = &dd->verbs_dev.rdi;
 		u64 rhf = rhf_to_cpu(rhf_addr);
 		u32 etype = rhf_rcv_type(rhf), qpn, bth1;
@@ -615,18 +614,21 @@ static void __prescan_rxq(struct hfi1_packet *packet)
 		if (etype != RHF_RCV_TYPE_IB)
 			goto next;
 
-		hdr = hfi1_get_msgheader(dd, rhf_addr);
+		packet->hdr = hfi1_get_msgheader(dd, rhf_addr);
+		hdr = packet->hdr;
 
 		lnh = be16_to_cpu(hdr->lrh[0]) & 3;
 
-		if (lnh == HFI1_LRH_BTH)
-			ohdr = &hdr->u.oth;
-		else if (lnh == HFI1_LRH_GRH)
-			ohdr = &hdr->u.l.oth;
-		else
+		if (lnh == HFI1_LRH_BTH) {
+			packet->ohdr = &hdr->u.oth;
+		} else if (lnh == HFI1_LRH_GRH) {
+			packet->ohdr = &hdr->u.l.oth;
+			packet->rcv_flags |= HFI1_HAS_GRH;
+		} else {
 			goto next; /* just in case */
+		}
 
-		bth1 = be32_to_cpu(ohdr->bth[1]);
+		bth1 = be32_to_cpu(packet->ohdr->bth[1]);
 		is_ecn = !!(bth1 & (HFI1_FECN_SMASK | HFI1_BECN_SMASK));
 
 		if (!is_ecn)
@@ -646,7 +648,7 @@ static void __prescan_rxq(struct hfi1_packet *packet)
 
 		/* turn off BECN, FECN */
 		bth1 &= ~(HFI1_FECN_SMASK | HFI1_BECN_SMASK);
-		ohdr->bth[1] = cpu_to_be32(bth1);
+		packet->ohdr->bth[1] = cpu_to_be32(bth1);
 next:
 		update_ps_mdata(&mdata, rcd);
 	}

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

^ permalink raw reply related	[flat|nested] 8+ messages in thread

* [PATCH 4/7] IB/hfi1: Get rid of divide in pio buffer allocator
       [not found] ` <20161025194241.15765.8903.stgit-9QXIwq+3FY+1XWohqUldA0EOCMrvLtNR@public.gmane.org>
                     ` (2 preceding siblings ...)
  2016-10-25 20:12   ` [PATCH 3/7] IB/hfi1: Fix ECN processing in prescan_rxq Dennis Dalessandro
@ 2016-10-25 20:12   ` Dennis Dalessandro
  2016-10-25 20:12   ` [PATCH 5/7] IB/hfi1: Optimize pio_buf and send_context structs Dennis Dalessandro
                     ` (2 subsequent siblings)
  6 siblings, 0 replies; 8+ messages in thread
From: Dennis Dalessandro @ 2016-10-25 20:12 UTC (permalink / raw)
  To: dledford-H+wXaHxf7aLQT0dZR+AlfA
  Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, Mike Marciniszyn,
	Sebastian Sanchez

From: Sebastian Sanchez <sebastian.sanchez-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

The div instruction shows costly in profiles.

Reviewed-by: Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Sebastian Sanchez <sebastian.sanchez-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/infiniband/hw/hfi1/pio.c |   11 +++++++----
 drivers/infiniband/hw/hfi1/pio.h |    1 +
 2 files changed, 8 insertions(+), 4 deletions(-)

diff --git a/drivers/infiniband/hw/hfi1/pio.c b/drivers/infiniband/hw/hfi1/pio.c
index e60fe3e..00a0e9e 100644
--- a/drivers/infiniband/hw/hfi1/pio.c
+++ b/drivers/infiniband/hw/hfi1/pio.c
@@ -1242,6 +1242,7 @@ int sc_enable(struct send_context *sc)
 	sc->free = 0;
 	sc->alloc_free = 0;
 	sc->fill = 0;
+	sc->fill_wrap = 0;
 	sc->sr_head = 0;
 	sc->sr_tail = 0;
 	sc->flags = 0;
@@ -1385,7 +1386,7 @@ struct pio_buf *sc_buffer_alloc(struct send_context *sc, u32 dw_len,
 	unsigned long flags;
 	unsigned long avail;
 	unsigned long blocks = dwords_to_blocks(dw_len);
-	unsigned long start_fill;
+	u32 fill_wrap;
 	int trycount = 0;
 	u32 head, next;
 
@@ -1428,8 +1429,11 @@ retry:
 	head = sc->sr_head;
 
 	/* "allocate" the buffer */
-	start_fill = sc->fill;
 	sc->fill += blocks;
+	fill_wrap = sc->fill_wrap;
+	sc->fill_wrap += blocks;
+	if (sc->fill_wrap >= sc->credits)
+		sc->fill_wrap = sc->fill_wrap - sc->credits;
 
 	/*
 	 * Fill the parts that the releaser looks at before moving the head.
@@ -1458,8 +1462,7 @@ retry:
 	spin_unlock_irqrestore(&sc->alloc_lock, flags);
 
 	/* finish filling in the buffer outside the lock */
-	pbuf->start = sc->base_addr + ((start_fill % sc->credits)
-							* PIO_BLOCK_SIZE);
+	pbuf->start = sc->base_addr + fill_wrap * PIO_BLOCK_SIZE;
 	pbuf->size = sc->credits * PIO_BLOCK_SIZE;
 	pbuf->end = sc->base_addr + pbuf->size;
 	pbuf->block_count = blocks;
diff --git a/drivers/infiniband/hw/hfi1/pio.h b/drivers/infiniband/hw/hfi1/pio.h
index bd19507..498b548 100644
--- a/drivers/infiniband/hw/hfi1/pio.h
+++ b/drivers/infiniband/hw/hfi1/pio.h
@@ -119,6 +119,7 @@ struct send_context {
 	unsigned long fill;		/* official alloc count */
 	unsigned long alloc_free;	/* copy of free (less cache thrash) */
 	u32 __percpu *buffers_allocated;/* count of buffers allocated */
+	u32 fill_wrap;			/* tracks fill within ring */
 	/* releaser fields */
 	spinlock_t release_lock ____cacheline_aligned_in_smp;
 	u32 sr_tail;			/* shadow ring tail */

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

^ permalink raw reply related	[flat|nested] 8+ messages in thread

* [PATCH 5/7] IB/hfi1: Optimize pio_buf and send_context structs
       [not found] ` <20161025194241.15765.8903.stgit-9QXIwq+3FY+1XWohqUldA0EOCMrvLtNR@public.gmane.org>
                     ` (3 preceding siblings ...)
  2016-10-25 20:12   ` [PATCH 4/7] IB/hfi1: Get rid of divide in pio buffer allocator Dennis Dalessandro
@ 2016-10-25 20:12   ` Dennis Dalessandro
  2016-10-25 20:12   ` [PATCH 6/7] IB/hfi1: Prevent hardware counter names from being cut off Dennis Dalessandro
  2016-10-25 20:12   ` [PATCH 7/7] IB/hfi1: Remove incorrect IS_ERR check Dennis Dalessandro
  6 siblings, 0 replies; 8+ messages in thread
From: Dennis Dalessandro @ 2016-10-25 20:12 UTC (permalink / raw)
  To: dledford-H+wXaHxf7aLQT0dZR+AlfA
  Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, Mike Marciniszyn,
	Sebastian Sanchez

From: Sebastian Sanchez <sebastian.sanchez-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

Both pio_buf and send_context structs have oversized
fields and have cachelines that can be optimized.

Reduce oversized fields for both structs.
Make sure pio_buf struct fits within a cacheline.
Move read-only fields to their own cacheline in
send_context struct.

All of this will avoid cacheline trading as the ring
progresses and pio buffers/send contexts are used.

Reviewed-by: Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Sebastian Sanchez <sebastian.sanchez-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/infiniband/hw/hfi1/pio.c      |    5 ++---
 drivers/infiniband/hw/hfi1/pio.h      |   29 +++++++++++++++--------------
 drivers/infiniband/hw/hfi1/pio_copy.c |   22 +++++++++++-----------
 3 files changed, 28 insertions(+), 28 deletions(-)

diff --git a/drivers/infiniband/hw/hfi1/pio.c b/drivers/infiniband/hw/hfi1/pio.c
index 00a0e9e..92701f1 100644
--- a/drivers/infiniband/hw/hfi1/pio.c
+++ b/drivers/infiniband/hw/hfi1/pio.c
@@ -758,6 +758,7 @@ struct send_context *sc_alloc(struct hfi1_devdata *dd, int type,
 	sc->hw_context = hw_context;
 	cr_group_addresses(sc, &dma);
 	sc->credits = sci->credits;
+	sc->size = sc->credits * PIO_BLOCK_SIZE;
 
 /* PIO Send Memory Address details */
 #define PIO_ADDR_CONTEXT_MASK 0xfful
@@ -1463,9 +1464,7 @@ retry:
 
 	/* finish filling in the buffer outside the lock */
 	pbuf->start = sc->base_addr + fill_wrap * PIO_BLOCK_SIZE;
-	pbuf->size = sc->credits * PIO_BLOCK_SIZE;
-	pbuf->end = sc->base_addr + pbuf->size;
-	pbuf->block_count = blocks;
+	pbuf->end = sc->base_addr + sc->size;
 	pbuf->qw_written = 0;
 	pbuf->carry_bytes = 0;
 	pbuf->carry.val64 = 0;
diff --git a/drivers/infiniband/hw/hfi1/pio.h b/drivers/infiniband/hw/hfi1/pio.h
index 498b548..867e5ff 100644
--- a/drivers/infiniband/hw/hfi1/pio.h
+++ b/drivers/infiniband/hw/hfi1/pio.h
@@ -83,43 +83,43 @@ struct pio_buf {
 	void *arg;		/* argument for cb */
 	void __iomem *start;	/* buffer start address */
 	void __iomem *end;	/* context end address */
-	unsigned long size;	/* context size, in bytes */
 	unsigned long sent_at;	/* buffer is sent when <= free */
-	u32 block_count;	/* size of buffer, in blocks */
-	u32 qw_written;		/* QW written so far */
-	u32 carry_bytes;	/* number of valid bytes in carry */
 	union mix carry;	/* pending unwritten bytes */
+	u16 qw_written;		/* QW written so far */
+	u8 carry_bytes;	/* number of valid bytes in carry */
 };
 
 /* cache line aligned pio buffer array */
 union pio_shadow_ring {
 	struct pio_buf pbuf;
-	u64 unused[16];		/* cache line spacer */
 } ____cacheline_aligned;
 
 /* per-NUMA send context */
 struct send_context {
 	/* read-only after init */
 	struct hfi1_devdata *dd;		/* device */
-	void __iomem *base_addr;	/* start of PIO memory */
 	union pio_shadow_ring *sr;	/* shadow ring */
+	void __iomem *base_addr;	/* start of PIO memory */
+	u32 __percpu *buffers_allocated;/* count of buffers allocated */
+	u32 size;			/* context size, in bytes */
 
-	struct work_struct halt_work;	/* halted context work queue entry */
-	unsigned long flags;		/* flags */
 	int node;			/* context home node */
-	int type;			/* context type */
-	u32 sw_index;			/* software index number */
-	u32 hw_context;			/* hardware context number */
-	u32 credits;			/* number of blocks in context */
 	u32 sr_size;			/* size of the shadow ring */
-	u32 group;			/* credit return group */
+	u16 flags;			/* flags */
+	u8  type;			/* context type */
+	u8  sw_index;			/* software index number */
+	u8  hw_context;			/* hardware context number */
+	u8  group;			/* credit return group */
+
 	/* allocator fields */
 	spinlock_t alloc_lock ____cacheline_aligned_in_smp;
 	u32 sr_head;			/* shadow ring head */
 	unsigned long fill;		/* official alloc count */
 	unsigned long alloc_free;	/* copy of free (less cache thrash) */
-	u32 __percpu *buffers_allocated;/* count of buffers allocated */
 	u32 fill_wrap;			/* tracks fill within ring */
+	u32 credits;			/* number of blocks in context */
+	/* adding a new field here would make it part of this cacheline */
+
 	/* releaser fields */
 	spinlock_t release_lock ____cacheline_aligned_in_smp;
 	u32 sr_tail;			/* shadow ring tail */
@@ -131,6 +131,7 @@ struct send_context {
 	u32 credit_intr_count;		/* count of credit intr users */
 	u64 credit_ctrl;		/* cache for credit control */
 	wait_queue_head_t halt_wait;    /* wait until kernel sees interrupt */
+	struct work_struct halt_work;	/* halted context work queue entry */
 };
 
 /* send context flags */
diff --git a/drivers/infiniband/hw/hfi1/pio_copy.c b/drivers/infiniband/hw/hfi1/pio_copy.c
index aa77736..03024ce 100644
--- a/drivers/infiniband/hw/hfi1/pio_copy.c
+++ b/drivers/infiniband/hw/hfi1/pio_copy.c
@@ -129,8 +129,8 @@ void pio_copy(struct hfi1_devdata *dd, struct pio_buf *pbuf, u64 pbc,
 				dest += sizeof(u64);
 			}
 
-			dest -= pbuf->size;
-			dend -= pbuf->size;
+			dest -= pbuf->sc->size;
+			dend -= pbuf->sc->size;
 		}
 
 		/* write 8-byte non-SOP, non-wrap chunk data */
@@ -361,8 +361,8 @@ void seg_pio_copy_start(struct pio_buf *pbuf, u64 pbc,
 				dest += sizeof(u64);
 			}
 
-			dest -= pbuf->size;
-			dend -= pbuf->size;
+			dest -= pbuf->sc->size;
+			dend -= pbuf->sc->size;
 		}
 
 		/* write 8-byte non-SOP, non-wrap chunk data */
@@ -458,8 +458,8 @@ static void mid_copy_mix(struct pio_buf *pbuf, const void *from, size_t nbytes)
 			dest += sizeof(u64);
 		}
 
-		dest -= pbuf->size;
-		dend -= pbuf->size;
+		dest -= pbuf->sc->size;
+		dend -= pbuf->sc->size;
 	}
 
 	/* write 8-byte non-SOP, non-wrap chunk data */
@@ -492,7 +492,7 @@ static void mid_copy_mix(struct pio_buf *pbuf, const void *from, size_t nbytes)
 		 */
 		/* adjust if we have wrapped */
 		if (dest >= pbuf->end)
-			dest -= pbuf->size;
+			dest -= pbuf->sc->size;
 		/* jump to the SOP range if within the first block */
 		else if (pbuf->qw_written < PIO_BLOCK_QWS)
 			dest += SOP_DISTANCE;
@@ -584,8 +584,8 @@ static void mid_copy_straight(struct pio_buf *pbuf,
 			dest += sizeof(u64);
 		}
 
-		dest -= pbuf->size;
-		dend -= pbuf->size;
+		dest -= pbuf->sc->size;
+		dend -= pbuf->sc->size;
 	}
 
 	/* write 8-byte non-SOP, non-wrap chunk data */
@@ -666,7 +666,7 @@ void seg_pio_copy_mid(struct pio_buf *pbuf, const void *from, size_t nbytes)
 			 */
 			/* adjust if we've wrapped */
 			if (dest >= pbuf->end)
-				dest -= pbuf->size;
+				dest -= pbuf->sc->size;
 			/* jump to SOP range if within the first block */
 			else if (pbuf->qw_written < PIO_BLOCK_QWS)
 				dest += SOP_DISTANCE;
@@ -719,7 +719,7 @@ void seg_pio_copy_end(struct pio_buf *pbuf)
 	 */
 	/* adjust if we have wrapped */
 	if (dest >= pbuf->end)
-		dest -= pbuf->size;
+		dest -= pbuf->sc->size;
 	/* jump to the SOP range if within the first block */
 	else if (pbuf->qw_written < PIO_BLOCK_QWS)
 		dest += SOP_DISTANCE;

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

^ permalink raw reply related	[flat|nested] 8+ messages in thread

* [PATCH 6/7] IB/hfi1: Prevent hardware counter names from being cut off
       [not found] ` <20161025194241.15765.8903.stgit-9QXIwq+3FY+1XWohqUldA0EOCMrvLtNR@public.gmane.org>
                     ` (4 preceding siblings ...)
  2016-10-25 20:12   ` [PATCH 5/7] IB/hfi1: Optimize pio_buf and send_context structs Dennis Dalessandro
@ 2016-10-25 20:12   ` Dennis Dalessandro
  2016-10-25 20:12   ` [PATCH 7/7] IB/hfi1: Remove incorrect IS_ERR check Dennis Dalessandro
  6 siblings, 0 replies; 8+ messages in thread
From: Dennis Dalessandro @ 2016-10-25 20:12 UTC (permalink / raw)
  To: dledford-H+wXaHxf7aLQT0dZR+AlfA
  Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, Jianxin Xiong

From: Jianxin Xiong <jianxin.xiong-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

Increase the size of the buffer that is used to construct per-VL
and per-SDMA counter names.

Reviewed-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Jianxin Xiong <jianxin.xiong-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/infiniband/hw/hfi1/chip.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/drivers/infiniband/hw/hfi1/chip.c b/drivers/infiniband/hw/hfi1/chip.c
index 3371361..8dee6c2 100644
--- a/drivers/infiniband/hw/hfi1/chip.c
+++ b/drivers/infiniband/hw/hfi1/chip.c
@@ -12098,7 +12098,7 @@ static void update_synth_timer(unsigned long opaque)
 	mod_timer(&dd->synth_stats_timer, jiffies + HZ * SYNTH_CNT_TIME);
 }
 
-#define C_MAX_NAME 13 /* 12 chars + one for /0 */
+#define C_MAX_NAME 16 /* 15 chars + one for /0 */
 static int init_cntrs(struct hfi1_devdata *dd)
 {
 	int i, rcv_ctxts, j;

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

^ permalink raw reply related	[flat|nested] 8+ messages in thread

* [PATCH 7/7] IB/hfi1: Remove incorrect IS_ERR check
       [not found] ` <20161025194241.15765.8903.stgit-9QXIwq+3FY+1XWohqUldA0EOCMrvLtNR@public.gmane.org>
                     ` (5 preceding siblings ...)
  2016-10-25 20:12   ` [PATCH 6/7] IB/hfi1: Prevent hardware counter names from being cut off Dennis Dalessandro
@ 2016-10-25 20:12   ` Dennis Dalessandro
  6 siblings, 0 replies; 8+ messages in thread
From: Dennis Dalessandro @ 2016-10-25 20:12 UTC (permalink / raw)
  To: dledford-H+wXaHxf7aLQT0dZR+AlfA
  Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, Dean Luick, Dan Carpenter

Remove IS_ERR check from caching code as the function being called does
not actually return error pointers.

Fixes: f19bd643dbde: "IB/hfi1: Prevent NULL pointer deferences in caching code"
Reported-by: Dan Carpenter <dan.carpenter-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>
Reviewed-by: Dean Luick <dean.luick-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/infiniband/hw/hfi1/user_sdma.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/drivers/infiniband/hw/hfi1/user_sdma.c b/drivers/infiniband/hw/hfi1/user_sdma.c
index a761f80..77697d6 100644
--- a/drivers/infiniband/hw/hfi1/user_sdma.c
+++ b/drivers/infiniband/hw/hfi1/user_sdma.c
@@ -1144,7 +1144,7 @@ static int pin_vector_pages(struct user_sdma_request *req,
 	rb_node = hfi1_mmu_rb_extract(pq->handler,
 				      (unsigned long)iovec->iov.iov_base,
 				      iovec->iov.iov_len);
-	if (rb_node && !IS_ERR(rb_node))
+	if (rb_node)
 		node = container_of(rb_node, struct sdma_mmu_node, rb);
 	else
 		rb_node = NULL;

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

^ permalink raw reply related	[flat|nested] 8+ messages in thread

end of thread, other threads:[~2016-10-25 20:12 UTC | newest]

Thread overview: 8+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2016-10-25 20:12 [PATCH 0/7] IB/hfi1: Another round of fixes Dennis Dalessandro
     [not found] ` <20161025194241.15765.8903.stgit-9QXIwq+3FY+1XWohqUldA0EOCMrvLtNR@public.gmane.org>
2016-10-25 20:12   ` [PATCH 1/7] IB/hfi1: Relocate rcvhdrcnt module parameter check Dennis Dalessandro
2016-10-25 20:12   ` [PATCH 2/7] IB/hfi1: Fix status error code for unsupported packets Dennis Dalessandro
2016-10-25 20:12   ` [PATCH 3/7] IB/hfi1: Fix ECN processing in prescan_rxq Dennis Dalessandro
2016-10-25 20:12   ` [PATCH 4/7] IB/hfi1: Get rid of divide in pio buffer allocator Dennis Dalessandro
2016-10-25 20:12   ` [PATCH 5/7] IB/hfi1: Optimize pio_buf and send_context structs Dennis Dalessandro
2016-10-25 20:12   ` [PATCH 6/7] IB/hfi1: Prevent hardware counter names from being cut off Dennis Dalessandro
2016-10-25 20:12   ` [PATCH 7/7] IB/hfi1: Remove incorrect IS_ERR check Dennis Dalessandro

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).