[PATCH 00/11] IB/hfi1: Additional fixes for 4.6

[PATCH 00/11] IB/hfi1: Additional fixes for 4.6

[Dennis Dalessandro]

These are some more fixes we've identified that we would like to make it into
4.6 if possible. Two of these are related to the pinned page cache series which
is undergoing discussion if that series does not go these two from Mitko below
would not apply.

Otherwise this will apply on your hfi1 branch on GitHub.

Can also be seen in my GitHub at:
https://github.com/ddalessa/kernel/tree/for-4.6.

---

Dean Luick (7):
      IB/hfi1: Fix sysfs file offset usage
      IB/hfi1: Fix i2c resource reservation checks
      IB/hfi1: Fix QOS num_vl bit width
      IB/hfi1: Remove invalid QOS check
      IB/hfi1: Fix QOS rule mappings
      IB/hfi1: Correctly obtain the full service class
      IB/hfi1: Simplify init_qpmap_table()

Mike Marciniszyn (1):
      IB/rdmavt: Fix adaptive pio hang

Mitko Haralanov (2):
      IB/hfi1: Prevent NULL pointer deferences in caching code
      IB/hfi1: Fix deadlock caused by locking with wrong scope

Sebastian Sanchez (1):
      IB/hfi1: Adjust default MTU to be 10KB


 drivers/infiniband/hw/hfi1/chip.c         |   59 ++++++++++++-----------------
 drivers/infiniband/hw/hfi1/hfi.h          |    6 +--
 drivers/infiniband/hw/hfi1/mmu_rb.c       |   40 ++++++++++++--------
 drivers/infiniband/hw/hfi1/mmu_rb.h       |    3 +
 drivers/infiniband/hw/hfi1/qp.c           |    6 ++-
 drivers/infiniband/hw/hfi1/qsfp.c         |    8 ++--
 drivers/infiniband/hw/hfi1/sysfs.c        |    4 +-
 drivers/infiniband/hw/hfi1/user_exp_rcv.c |    9 ++--
 drivers/infiniband/hw/hfi1/user_sdma.c    |   24 ++++++++----
 include/rdma/rdmavt_qp.h                  |    5 +-
 10 files changed, 89 insertions(+), 75 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

[PATCH 01/11] IB/hfi1: Fix sysfs file offset usage

[Dennis Dalessandro]

From: Dean Luick <dean.luick-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

Two sysfs files do not pay attention to the file offset when
reading data. Fix that.

Reviewed-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dean Luick <dean.luick-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Jubin John <jubin.john-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/infiniband/hw/hfi1/sysfs.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/infiniband/hw/hfi1/sysfs.c b/drivers/infiniband/hw/hfi1/sysfs.c
index c7f1271..8cd6df8 100644
--- a/drivers/infiniband/hw/hfi1/sysfs.c
+++ b/drivers/infiniband/hw/hfi1/sysfs.c
@@ -84,7 +84,7 @@ static ssize_t read_cc_table_bin(struct file *filp, struct kobject *kobj,
 		rcu_read_unlock();
 		return -EINVAL;
 	}
-	memcpy(buf, &cc_state->cct, count);
+	memcpy(buf, (void *)&cc_state->cct + pos, count);
 	rcu_read_unlock();
 
 	return count;
@@ -131,7 +131,7 @@ static ssize_t read_cc_setting_bin(struct file *filp, struct kobject *kobj,
 		rcu_read_unlock();
 		return -EINVAL;
 	}
-	memcpy(buf, &cc_state->cong_setting, count);
+	memcpy(buf, (void *)&cc_state->cong_setting + pos, count);
 	rcu_read_unlock();
 
 	return count;

--
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

[PATCH 02/11] IB/hfi1: Fix i2c resource reservation checks

[Dennis Dalessandro]

From: Dean Luick <dean.luick-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

The i2c and qsfp read/write routines should check for the resource
reservation of the incoming argument target rather than the implicit
target of the hardware HFI.

Reviewed-by: Easwar Hariharan <easwar.hariharan-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dean Luick <dean.luick-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Jubin John <jubin.john-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/infiniband/hw/hfi1/qsfp.c |    8 ++++----
 1 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/infiniband/hw/hfi1/qsfp.c b/drivers/infiniband/hw/hfi1/qsfp.c
index 9ed1963..ac03d80 100644
--- a/drivers/infiniband/hw/hfi1/qsfp.c
+++ b/drivers/infiniband/hw/hfi1/qsfp.c
@@ -96,7 +96,7 @@ int i2c_write(struct hfi1_pportdata *ppd, u32 target, int i2c_addr, int offset,
 {
 	int ret;
 
-	if (!check_chip_resource(ppd->dd, qsfp_resource(ppd->dd), __func__))
+	if (!check_chip_resource(ppd->dd, i2c_target(target), __func__))
 		return -EACCES;
 
 	/* make sure the TWSI bus is in a sane state */
@@ -162,7 +162,7 @@ int i2c_read(struct hfi1_pportdata *ppd, u32 target, int i2c_addr, int offset,
 {
 	int ret;
 
-	if (!check_chip_resource(ppd->dd, qsfp_resource(ppd->dd), __func__))
+	if (!check_chip_resource(ppd->dd, i2c_target(target), __func__))
 		return -EACCES;
 
 	/* make sure the TWSI bus is in a sane state */
@@ -192,7 +192,7 @@ int qsfp_write(struct hfi1_pportdata *ppd, u32 target, int addr, void *bp,
 	int ret;
 	u8 page;
 
-	if (!check_chip_resource(ppd->dd, qsfp_resource(ppd->dd), __func__))
+	if (!check_chip_resource(ppd->dd, i2c_target(target), __func__))
 		return -EACCES;
 
 	/* make sure the TWSI bus is in a sane state */
@@ -276,7 +276,7 @@ int qsfp_read(struct hfi1_pportdata *ppd, u32 target, int addr, void *bp,
 	int ret;
 	u8 page;
 
-	if (!check_chip_resource(ppd->dd, qsfp_resource(ppd->dd), __func__))
+	if (!check_chip_resource(ppd->dd, i2c_target(target), __func__))
 		return -EACCES;
 
 	/* make sure the TWSI bus is in a sane state */

--
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

[PATCH 03/11] IB/rdmavt: Fix adaptive pio hang

[Dennis Dalessandro]

From: Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

The RVT_S_WAIT_PIO_DRAIN flag was missing from
the set of flags indicating a qp is waiting
on a resource.

This caused the sleep/wakeup for adaptive pio
drain to lose a wakeup "hanging" a QP.

Reviewed-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 include/rdma/rdmavt_qp.h |    5 +++--
 1 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/include/rdma/rdmavt_qp.h b/include/rdma/rdmavt_qp.h
index 497e590..0e1ff2a 100644
--- a/include/rdma/rdmavt_qp.h
+++ b/include/rdma/rdmavt_qp.h
@@ -117,8 +117,9 @@
 /*
  * Wait flags that would prevent any packet type from being sent.
  */
-#define RVT_S_ANY_WAIT_IO (RVT_S_WAIT_PIO | RVT_S_WAIT_TX | \
-	RVT_S_WAIT_DMA_DESC | RVT_S_WAIT_KMEM)
+#define RVT_S_ANY_WAIT_IO \
+	(RVT_S_WAIT_PIO | RVT_S_WAIT_PIO_DRAIN | RVT_S_WAIT_TX | \
+	 RVT_S_WAIT_DMA_DESC | RVT_S_WAIT_KMEM)
 
 /*
  * Wait flags that would prevent send work requests from making progress.

--
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

[PATCH 04/11] IB/hfi1: Prevent NULL pointer deferences in caching code

[Dennis Dalessandro]

From: Mitko Haralanov <mitko.haralanov-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

There is a potential kernel crash when the MMU notifier calls the
invalidation routines in the hfi1 pinned page caching code for sdma.

The invalidation routine could call the remove callback
for the node, which in turn ends up dereferencing the
current task_struct to get a pointer to the mm_struct.
However, the mm_struct pointer could be NULL resulting in
the following backtrace:

    BUG: unable to handle kernel NULL pointer dereference at 00000000000000a8
    IP: [<ffffffffa041f75a>] sdma_rb_remove+0xaa/0x100 [hfi1]
    15
    task: ffff88085e66e080 ti: ffff88085c244000 task.ti: ffff88085c244000
    RIP: 0010:[<ffffffffa041f75a>]  [<ffffffffa041f75a>] sdma_rb_remove+0xaa/0x100 [hfi1]
    RSP: 0000:ffff88085c245878  EFLAGS: 00010002
    RAX: 0000000000000000 RBX: ffff88105b9bbd40 RCX: ffffea003931a830
    RDX: 0000000000000004 RSI: ffff88105754a9c0 RDI: ffff88105754a9c0
    RBP: ffff88085c245890 R08: ffff88105b9bbd70 R09: 00000000fffffffb
    R10: ffff88105b9bbd58 R11: 0000000000000013 R12: ffff88105754a9c0
    R13: 0000000000000001 R14: 0000000000000001 R15: ffff88105b9bbd40
    FS:  0000000000000000(0000) GS:ffff88107ef40000(0000) knlGS:0000000000000000
    CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
    CR2: 00000000000000a8 CR3: 0000000001a0b000 CR4: 00000000001407e0
    Stack:
     ffff88105b9bbd40 ffff88080ec481a8 ffff88080ec481b8 ffff88085c2458c0
     ffffffffa03fa00e ffff88080ec48190 ffff88080ed9cd00 0000000001024000
     0000000000000000 ffff88085c245920 ffffffffa03fa0e7 0000000000000282
    Call Trace:
     [<ffffffffa03fa00e>] __mmu_rb_remove.isra.5+0x5e/0x70 [hfi1]
     [<ffffffffa03fa0e7>] mmu_notifier_mem_invalidate+0xc7/0xf0 [hfi1]
     [<ffffffffa03fa143>] mmu_notifier_page+0x13/0x20 [hfi1]
     [<ffffffff81156dd0>] __mmu_notifier_invalidate_page+0x50/0x70
     [<ffffffff81140bbb>] try_to_unmap_one+0x20b/0x470
     [<ffffffff81141ee7>] try_to_unmap_anon+0xa7/0x120
     [<ffffffff81141fad>] try_to_unmap+0x4d/0x60
     [<ffffffff8111fd7b>] shrink_page_list+0x2eb/0x9d0
     [<ffffffff81120ab3>] shrink_inactive_list+0x243/0x490
     [<ffffffff81121491>] shrink_lruvec+0x4c1/0x640
     [<ffffffff81121641>] shrink_zone+0x31/0x100
     [<ffffffff81121b0f>] kswapd_shrink_zone.constprop.62+0xef/0x1c0
     [<ffffffff811229e3>] kswapd+0x403/0x7e0
     [<ffffffff811225e0>] ? shrink_all_memory+0xf0/0xf0
     [<ffffffff81068ac0>] kthread+0xc0/0xd0
     [<ffffffff81068a00>] ? insert_kthread_work+0x40/0x40
     [<ffffffff814ff8ec>] ret_from_fork+0x7c/0xb0
     [<ffffffff81068a00>] ? insert_kthread_work+0x40/0x40

To correct this, the mm_struct passed to us by the MMU notifier is
used (which is what should have been done to begin with). This avoids
the broken derefences and ensures that the correct mm_struct is used.

Reviewed-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Reviewed-by: Dean Luick <dean.luick-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Mitko Haralanov <mitko.haralanov-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/infiniband/hw/hfi1/mmu_rb.c       |   24 ++++++++++++++----------
 drivers/infiniband/hw/hfi1/mmu_rb.h       |    3 ++-
 drivers/infiniband/hw/hfi1/user_exp_rcv.c |    9 +++++----
 drivers/infiniband/hw/hfi1/user_sdma.c    |   24 ++++++++++++++++--------
 4 files changed, 37 insertions(+), 23 deletions(-)

diff --git a/drivers/infiniband/hw/hfi1/mmu_rb.c b/drivers/infiniband/hw/hfi1/mmu_rb.c
index c7ad016..eac4d04 100644
--- a/drivers/infiniband/hw/hfi1/mmu_rb.c
+++ b/drivers/infiniband/hw/hfi1/mmu_rb.c
@@ -71,6 +71,7 @@ static inline void mmu_notifier_range_start(struct mmu_notifier *,
 					    struct mm_struct *,
 					    unsigned long, unsigned long);
 static void mmu_notifier_mem_invalidate(struct mmu_notifier *,
+					struct mm_struct *,
 					unsigned long, unsigned long);
 static struct mmu_rb_node *__mmu_rb_search(struct mmu_rb_handler *,
 					   unsigned long, unsigned long);
@@ -137,7 +138,7 @@ void hfi1_mmu_rb_unregister(struct rb_root *root)
 			rbnode = rb_entry(node, struct mmu_rb_node, node);
 			rb_erase(node, root);
 			if (handler->ops->remove)
-				handler->ops->remove(root, rbnode, false);
+				handler->ops->remove(root, rbnode, NULL);
 		}
 	}
 
@@ -201,14 +202,14 @@ static struct mmu_rb_node *__mmu_rb_search(struct mmu_rb_handler *handler,
 }
 
 static void __mmu_rb_remove(struct mmu_rb_handler *handler,
-			    struct mmu_rb_node *node, bool arg)
+			    struct mmu_rb_node *node, struct mm_struct *mm)
 {
 	/* Validity of handler and node pointers has been checked by caller. */
 	hfi1_cdbg(MMU, "Removing node addr 0x%llx, len %u", node->addr,
 		  node->len);
 	__mmu_int_rb_remove(node, handler->root);
 	if (handler->ops->remove)
-		handler->ops->remove(handler->root, node, arg);
+		handler->ops->remove(handler->root, node, mm);
 }
 
 struct mmu_rb_node *hfi1_mmu_rb_search(struct rb_root *root, unsigned long addr,
@@ -237,7 +238,7 @@ void hfi1_mmu_rb_remove(struct rb_root *root, struct mmu_rb_node *node)
 		return;
 
 	spin_lock_irqsave(&handler->lock, flags);
-	__mmu_rb_remove(handler, node, false);
+	__mmu_rb_remove(handler, node, NULL);
 	spin_unlock_irqrestore(&handler->lock, flags);
 }
 
@@ -260,7 +261,7 @@ unlock:
 static inline void mmu_notifier_page(struct mmu_notifier *mn,
 				     struct mm_struct *mm, unsigned long addr)
 {
-	mmu_notifier_mem_invalidate(mn, addr, addr + PAGE_SIZE);
+	mmu_notifier_mem_invalidate(mn, mm, addr, addr + PAGE_SIZE);
 }
 
 static inline void mmu_notifier_range_start(struct mmu_notifier *mn,
@@ -268,25 +269,28 @@ static inline void mmu_notifier_range_start(struct mmu_notifier *mn,
 					    unsigned long start,
 					    unsigned long end)
 {
-	mmu_notifier_mem_invalidate(mn, start, end);
+	mmu_notifier_mem_invalidate(mn, mm, start, end);
 }
 
 static void mmu_notifier_mem_invalidate(struct mmu_notifier *mn,
+					struct mm_struct *mm,
 					unsigned long start, unsigned long end)
 {
 	struct mmu_rb_handler *handler =
 		container_of(mn, struct mmu_rb_handler, mn);
 	struct rb_root *root = handler->root;
-	struct mmu_rb_node *node;
+	struct mmu_rb_node *node, *ptr = NULL;
 	unsigned long flags;
 
 	spin_lock_irqsave(&handler->lock, flags);
-	for (node = __mmu_int_rb_iter_first(root, start, end - 1); node;
-	     node = __mmu_int_rb_iter_next(node, start, end - 1)) {
+	for (node = __mmu_int_rb_iter_first(root, start, end - 1);
+	     node; node = ptr) {
+		/* Guard against node removal. */
+		ptr = __mmu_int_rb_iter_next(node, start, end - 1);
 		hfi1_cdbg(MMU, "Invalidating node addr 0x%llx, len %u",
 			  node->addr, node->len);
 		if (handler->ops->invalidate(root, node))
-			__mmu_rb_remove(handler, node, true);
+			__mmu_rb_remove(handler, node, mm);
 	}
 	spin_unlock_irqrestore(&handler->lock, flags);
 }
diff --git a/drivers/infiniband/hw/hfi1/mmu_rb.h b/drivers/infiniband/hw/hfi1/mmu_rb.h
index f8523fd..19a306e 100644
--- a/drivers/infiniband/hw/hfi1/mmu_rb.h
+++ b/drivers/infiniband/hw/hfi1/mmu_rb.h
@@ -59,7 +59,8 @@ struct mmu_rb_node {
 struct mmu_rb_ops {
 	bool (*filter)(struct mmu_rb_node *, unsigned long, unsigned long);
 	int (*insert)(struct rb_root *, struct mmu_rb_node *);
-	void (*remove)(struct rb_root *, struct mmu_rb_node *, bool);
+	void (*remove)(struct rb_root *, struct mmu_rb_node *,
+		       struct mm_struct *);
 	int (*invalidate)(struct rb_root *, struct mmu_rb_node *);
 };
 
diff --git a/drivers/infiniband/hw/hfi1/user_exp_rcv.c b/drivers/infiniband/hw/hfi1/user_exp_rcv.c
index 0861e09..5b72849 100644
--- a/drivers/infiniband/hw/hfi1/user_exp_rcv.c
+++ b/drivers/infiniband/hw/hfi1/user_exp_rcv.c
@@ -87,7 +87,8 @@ static u32 find_phys_blocks(struct page **, unsigned, struct tid_pageset *);
 static int set_rcvarray_entry(struct file *, unsigned long, u32,
 			      struct tid_group *, struct page **, unsigned);
 static int mmu_rb_insert(struct rb_root *, struct mmu_rb_node *);
-static void mmu_rb_remove(struct rb_root *, struct mmu_rb_node *, bool);
+static void mmu_rb_remove(struct rb_root *, struct mmu_rb_node *,
+			  struct mm_struct *);
 static int mmu_rb_invalidate(struct rb_root *, struct mmu_rb_node *);
 static int program_rcvarray(struct file *, unsigned long, struct tid_group *,
 			    struct tid_pageset *, unsigned, u16, struct page **,
@@ -899,7 +900,7 @@ static int unprogram_rcvarray(struct file *fp, u32 tidinfo,
 	if (!node || node->rcventry != (uctxt->expected_base + rcventry))
 		return -EBADF;
 	if (HFI1_CAP_IS_USET(TID_UNMAP))
-		mmu_rb_remove(&fd->tid_rb_root, &node->mmu, false);
+		mmu_rb_remove(&fd->tid_rb_root, &node->mmu, NULL);
 	else
 		hfi1_mmu_rb_remove(&fd->tid_rb_root, &node->mmu);
 
@@ -965,7 +966,7 @@ static void unlock_exp_tids(struct hfi1_ctxtdata *uctxt,
 					continue;
 				if (HFI1_CAP_IS_USET(TID_UNMAP))
 					mmu_rb_remove(&fd->tid_rb_root,
-						      &node->mmu, false);
+						      &node->mmu, NULL);
 				else
 					hfi1_mmu_rb_remove(&fd->tid_rb_root,
 							   &node->mmu);
@@ -1032,7 +1033,7 @@ static int mmu_rb_insert(struct rb_root *root, struct mmu_rb_node *node)
 }
 
 static void mmu_rb_remove(struct rb_root *root, struct mmu_rb_node *node,
-			  bool notifier)
+			  struct mm_struct *mm)
 {
 	struct hfi1_filedata *fdata =
 		container_of(root, struct hfi1_filedata, tid_rb_root);
diff --git a/drivers/infiniband/hw/hfi1/user_sdma.c b/drivers/infiniband/hw/hfi1/user_sdma.c
index 46e254d..b4c91a0 100644
--- a/drivers/infiniband/hw/hfi1/user_sdma.c
+++ b/drivers/infiniband/hw/hfi1/user_sdma.c
@@ -300,7 +300,8 @@ static int defer_packet_queue(
 static void activate_packet_queue(struct iowait *, int);
 static bool sdma_rb_filter(struct mmu_rb_node *, unsigned long, unsigned long);
 static int sdma_rb_insert(struct rb_root *, struct mmu_rb_node *);
-static void sdma_rb_remove(struct rb_root *, struct mmu_rb_node *, bool);
+static void sdma_rb_remove(struct rb_root *, struct mmu_rb_node *,
+			   struct mm_struct *);
 static int sdma_rb_invalidate(struct rb_root *, struct mmu_rb_node *);
 
 static struct mmu_rb_ops sdma_rb_ops = {
@@ -1066,8 +1067,10 @@ static int pin_vector_pages(struct user_sdma_request *req,
 	rb_node = hfi1_mmu_rb_search(&pq->sdma_rb_root,
 				     (unsigned long)iovec->iov.iov_base,
 				     iovec->iov.iov_len);
-	if (rb_node)
+	if (rb_node && !IS_ERR(rb_node))
 		node = container_of(rb_node, struct sdma_mmu_node, rb);
+	else
+		rb_node = NULL;
 
 	if (!node) {
 		node = kzalloc(sizeof(*node), GFP_KERNEL);
@@ -1505,7 +1508,7 @@ static void user_sdma_free_request(struct user_sdma_request *req, bool unpin)
 				&req->pq->sdma_rb_root,
 				(unsigned long)req->iovs[i].iov.iov_base,
 				req->iovs[i].iov.iov_len);
-			if (!mnode)
+			if (!mnode || IS_ERR(mnode))
 				continue;
 
 			node = container_of(mnode, struct sdma_mmu_node, rb);
@@ -1550,7 +1553,7 @@ static int sdma_rb_insert(struct rb_root *root, struct mmu_rb_node *mnode)
 }
 
 static void sdma_rb_remove(struct rb_root *root, struct mmu_rb_node *mnode,
-			   bool notifier)
+			   struct mm_struct *mm)
 {
 	struct sdma_mmu_node *node =
 		container_of(mnode, struct sdma_mmu_node, rb);
@@ -1560,14 +1563,19 @@ static void sdma_rb_remove(struct rb_root *root, struct mmu_rb_node *mnode,
 	node->pq->n_locked -= node->npages;
 	spin_unlock(&node->pq->evict_lock);
 
-	unpin_vector_pages(notifier ? NULL : current->mm, node->pages,
-			   node->npages);
+	/*
+	 * If mm is set, we are being called by the MMU notifier and we
+	 * should not pass a mm_struct to unpin_vector_page(). This is to
+	 * prevent a deadlock when hfi1_release_user_pages() attempts to
+	 * take the mmap_sem, which the MMU notifier has already taken.
+	 */
+	unpin_vector_pages(mm ? NULL : current->mm, node->pages, node->npages);
 	/*
 	 * If called by the MMU notifier, we have to adjust the pinned
 	 * page count ourselves.
 	 */
-	if (notifier)
-		current->mm->pinned_vm -= node->npages;
+	if (mm)
+		mm->pinned_vm -= node->npages;
 	kfree(node);
 }
 

--
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

[PATCH 05/11] IB/hfi1: Fix deadlock caused by locking with wrong scope

[Dennis Dalessandro]

From: Mitko Haralanov <mitko.haralanov-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

The locking around the interval RB tree is designed to prevent
access to the tree while it's being modified. The locking in its
current form is too overzealous, which is causing a deadlock in
certain cases with the following backtrace:

    Kernel panic - not syncing: Watchdog detected hard LOCKUP on cpu 0
    CPU: 0 PID: 5836 Comm: IMB-MPI1 Tainted: G           O 3.12.18-wfr+ #1
     0000000000000000 ffff88087f206c50 ffffffff814f1caa ffffffff817b53f0
     ffff88087f206cc8 ffffffff814ecd56 0000000000000010 ffff88087f206cd8
     ffff88087f206c78 0000000000000000 0000000000000000 0000000000001662
    Call Trace:
     <NMI>  [<ffffffff814f1caa>] dump_stack+0x45/0x56
     [<ffffffff814ecd56>] panic+0xc2/0x1cb
     [<ffffffff810d4370>] ? restart_watchdog_hrtimer+0x50/0x50
     [<ffffffff810d4432>] watchdog_overflow_callback+0xc2/0xd0
     [<ffffffff81109b4e>] __perf_event_overflow+0x8e/0x2b0
     [<ffffffff8110a714>] perf_event_overflow+0x14/0x20
     [<ffffffff8101c906>] intel_pmu_handle_irq+0x1b6/0x390
     [<ffffffff814f927b>] perf_event_nmi_handler+0x2b/0x50
     [<ffffffff814f8ad8>] nmi_handle.isra.3+0x88/0x180
     [<ffffffff814f8d39>] do_nmi+0x169/0x310
     [<ffffffff814f8177>] end_repeat_nmi+0x1e/0x2e
     [<ffffffff81272600>] ? unmap_single+0x30/0x30
     [<ffffffff814f780d>] ? _raw_spin_lock_irqsave+0x2d/0x40
     [<ffffffff814f780d>] ? _raw_spin_lock_irqsave+0x2d/0x40
     [<ffffffff814f780d>] ? _raw_spin_lock_irqsave+0x2d/0x40
     <<EOE>>  <IRQ>  [<ffffffffa056c4a8>] hfi1_mmu_rb_search+0x38/0x70 [hfi1]
     [<ffffffffa05919cb>] user_sdma_free_request+0xcb/0x120 [hfi1]
     [<ffffffffa0593393>] user_sdma_txreq_cb+0x263/0x350 [hfi1]
     [<ffffffffa057fad7>] ? sdma_txclean+0x27/0x1c0 [hfi1]
     [<ffffffffa0593130>] ? user_sdma_send_pkts+0x1710/0x1710 [hfi1]
     [<ffffffffa057fdd6>] sdma_make_progress+0x166/0x480 [hfi1]
     [<ffffffff810762c9>] ? ttwu_do_wakeup+0x19/0xd0
     [<ffffffffa0581c7e>] sdma_engine_interrupt+0x8e/0x100 [hfi1]
     [<ffffffffa0546bdd>] sdma_interrupt+0x5d/0xa0 [hfi1]
     [<ffffffff81097e57>] handle_irq_event_percpu+0x47/0x1d0
     [<ffffffff81098017>] handle_irq_event+0x37/0x60
     [<ffffffff8109aa5f>] handle_edge_irq+0x6f/0x120
     [<ffffffff810044af>] handle_irq+0xbf/0x150
     [<ffffffff8104c9b7>] ? irq_enter+0x17/0x80
     [<ffffffff8150168d>] do_IRQ+0x4d/0xc0
     [<ffffffff814f7c6a>] common_interrupt+0x6a/0x6a
     <EOI>  [<ffffffff81073524>] ? finish_task_switch+0x54/0xe0
     [<ffffffff814f56c6>] __schedule+0x3b6/0x7e0
     [<ffffffff810763a6>] __cond_resched+0x26/0x30
     [<ffffffff814f5eda>] _cond_resched+0x3a/0x50
     [<ffffffff814f4f82>] down_write+0x12/0x30
     [<ffffffffa0591619>] hfi1_release_user_pages+0x69/0x90 [hfi1]
     [<ffffffffa059173a>] sdma_rb_remove+0x9a/0xc0 [hfi1]
     [<ffffffffa056c00d>] __mmu_rb_remove.isra.5+0x5d/0x70 [hfi1]
     [<ffffffffa056c536>] hfi1_mmu_rb_remove+0x56/0x70 [hfi1]
     [<ffffffffa059427b>] hfi1_user_sdma_process_request+0x74b/0x1160 [hfi1]
     [<ffffffffa055c763>] hfi1_aio_write+0xc3/0x100 [hfi1]
     [<ffffffff8116a14c>] do_sync_readv_writev+0x4c/0x80
     [<ffffffff8116b58b>] do_readv_writev+0xbb/0x230
     [<ffffffff811a9da1>] ? fsnotify+0x241/0x320
     [<ffffffff81073524>] ? finish_task_switch+0x54/0xe0
     [<ffffffff8116b795>] vfs_writev+0x35/0x60
     [<ffffffff8116b8c9>] SyS_writev+0x49/0xc0
     [<ffffffff810cd876>] ? __audit_syscall_exit+0x1f6/0x2a0
     [<ffffffff814ff992>] system_call_fastpath+0x16/0x1b

As evident from the backtrace above, the process was being put to sleep
while holding the lock.

Limiting the scope of the lock only to the RB tree operation fixes the
above error allowing for proper locking and the process being put to
sleep when needed.

Reviewed-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Reviewed-by: Dean Luick <dean.luick-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Mitko Haralanov <mitko.haralanov-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/infiniband/hw/hfi1/mmu_rb.c |   16 +++++++++++-----
 1 files changed, 11 insertions(+), 5 deletions(-)

diff --git a/drivers/infiniband/hw/hfi1/mmu_rb.c b/drivers/infiniband/hw/hfi1/mmu_rb.c
index eac4d04..b3f0682 100644
--- a/drivers/infiniband/hw/hfi1/mmu_rb.c
+++ b/drivers/infiniband/hw/hfi1/mmu_rb.c
@@ -177,7 +177,7 @@ unlock:
 	return ret;
 }
 
-/* Caller must host handler lock */
+/* Caller must hold handler lock */
 static struct mmu_rb_node *__mmu_rb_search(struct mmu_rb_handler *handler,
 					   unsigned long addr,
 					   unsigned long len)
@@ -201,13 +201,19 @@ static struct mmu_rb_node *__mmu_rb_search(struct mmu_rb_handler *handler,
 	return node;
 }
 
+/* Caller must *not* hold handler lock. */
 static void __mmu_rb_remove(struct mmu_rb_handler *handler,
 			    struct mmu_rb_node *node, struct mm_struct *mm)
 {
+	unsigned long flags;
+
 	/* Validity of handler and node pointers has been checked by caller. */
 	hfi1_cdbg(MMU, "Removing node addr 0x%llx, len %u", node->addr,
 		  node->len);
+	spin_lock_irqsave(&handler->lock, flags);
 	__mmu_int_rb_remove(node, handler->root);
+	spin_unlock_irqrestore(&handler->lock, flags);
+
 	if (handler->ops->remove)
 		handler->ops->remove(handler->root, node, mm);
 }
@@ -232,14 +238,11 @@ struct mmu_rb_node *hfi1_mmu_rb_search(struct rb_root *root, unsigned long addr,
 void hfi1_mmu_rb_remove(struct rb_root *root, struct mmu_rb_node *node)
 {
 	struct mmu_rb_handler *handler = find_mmu_handler(root);
-	unsigned long flags;
 
 	if (!handler || !node)
 		return;
 
-	spin_lock_irqsave(&handler->lock, flags);
 	__mmu_rb_remove(handler, node, NULL);
-	spin_unlock_irqrestore(&handler->lock, flags);
 }
 
 static struct mmu_rb_handler *find_mmu_handler(struct rb_root *root)
@@ -289,8 +292,11 @@ static void mmu_notifier_mem_invalidate(struct mmu_notifier *mn,
 		ptr = __mmu_int_rb_iter_next(node, start, end - 1);
 		hfi1_cdbg(MMU, "Invalidating node addr 0x%llx, len %u",
 			  node->addr, node->len);
-		if (handler->ops->invalidate(root, node))
+		if (handler->ops->invalidate(root, node)) {
+			spin_unlock_irqrestore(&handler->lock, flags);
 			__mmu_rb_remove(handler, node, mm);
+			spin_lock_irqsave(&handler->lock, flags);
+		}
 	}
 	spin_unlock_irqrestore(&handler->lock, flags);
 }

--
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

[PATCH 06/11] IB/hfi1: Fix QOS num_vl bit width

[Dennis Dalessandro]

From: Dean Luick <dean.luick-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

The bit width for num_vls, n, needs to be calculated based on
the pow2 rounded up of the number of vls.  Otherwise num_vls of 3,
5, 6, and 7 will have misplaced QOS RSM map entries.

Reviewed-by: Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dean Luick <dean.luick-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 c29860c..78a5c32 100644
--- a/drivers/infiniband/hw/hfi1/chip.c
+++ b/drivers/infiniband/hw/hfi1/chip.c
@@ -13511,7 +13511,7 @@ static void init_qos(struct hfi1_devdata *dd, u32 first_ctxt)
 		goto bail;
 	qpns_per_vl = __roundup_pow_of_two(max_by_vl);
 	/* determine bits vl */
-	n = ilog2(num_vls);
+	n = ilog2(__roundup_pow_of_two(num_vls));
 	/* determine bits for qpn */
 	m = ilog2(qpns_per_vl);
 	if ((m + n) > 7)

--
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

[PATCH 07/11] IB/hfi1: Remove invalid QOS check

[Dennis Dalessandro]

From: Dean Luick <dean.luick-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

Remove an invalid compare of the number of QOS RSM map table entries
against the number of physical receive contexts.  The RSM map table
has its own size and has no relation to the number of physical receive
contexts.

Reviewed-by: Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dean Luick <dean.luick-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/infiniband/hw/hfi1/chip.c |    2 --
 1 files changed, 0 insertions(+), 2 deletions(-)

diff --git a/drivers/infiniband/hw/hfi1/chip.c b/drivers/infiniband/hw/hfi1/chip.c
index 78a5c32..24c96e7 100644
--- a/drivers/infiniband/hw/hfi1/chip.c
+++ b/drivers/infiniband/hw/hfi1/chip.c
@@ -13516,8 +13516,6 @@ static void init_qos(struct hfi1_devdata *dd, u32 first_ctxt)
 	m = ilog2(qpns_per_vl);
 	if ((m + n) > 7)
 		goto bail;
-	if (num_vls * qpns_per_vl > dd->chip_rcv_contexts)
-		goto bail;
 	rsmmap = kmalloc_array(NUM_MAP_REGS, sizeof(u64), GFP_KERNEL);
 	if (!rsmmap)
 		goto bail;

--
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

[PATCH 08/11] IB/hfi1: Fix QOS rule mappings

[Dennis Dalessandro]

From: Dean Luick <dean.luick-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

The QOS RSM rule mappings are off by one, referencing a kernel receive
context that does not exist.

Correctly start the QOS RSM map entries at FIRST_KERNEL_CONTEXT rather
than MIN_KERNEL_KCTXTS.  Remove the cruft that hid this.

Change the QP map table so all traffic not caught by QOS RSM goes to
the control context rather than the first QOS context.

Correct comments to match the actual code operation and intent.

Reviewed-by: Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dean Luick <dean.luick-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/infiniband/hw/hfi1/chip.c |   48 ++++++++++++++++---------------------
 1 files changed, 21 insertions(+), 27 deletions(-)

diff --git a/drivers/infiniband/hw/hfi1/chip.c b/drivers/infiniband/hw/hfi1/chip.c
index 24c96e7..690871e 100644
--- a/drivers/infiniband/hw/hfi1/chip.c
+++ b/drivers/infiniband/hw/hfi1/chip.c
@@ -12678,20 +12678,20 @@ static int set_up_context_variables(struct hfi1_devdata *dd)
 	unsigned ngroups;
 
 	/*
-	 * Kernel contexts: (to be fixed later):
-	 * - min or 2 or 1 context/numa
+	 * Kernel receive contexts:
+	 * - min of 2 or 1 context/numa (excluding control context)
 	 * - Context 0 - control context (VL15/multicast/error)
-	 * - Context 1 - default context
+	 * - Context 1 - first kernel context
+	 * - Context 2 - second kernel context
+	 * ...
 	 */
 	if (n_krcvqs)
 		/*
-		 * Don't count context 0 in n_krcvqs since
-		 * is isn't used for normal verbs traffic.
-		 *
-		 * krcvqs will reflect number of kernel
-		 * receive contexts above 0.
+		 * n_krcvqs is the sum of module parameter kernel receive
+		 * contexts, krcvqs[].  It does not include the control
+		 * context, so add that.
 		 */
-		num_kernel_contexts = n_krcvqs + MIN_KERNEL_KCTXTS - 1;
+		num_kernel_contexts = n_krcvqs + 1;
 	else
 		num_kernel_contexts = num_online_nodes() + 1;
 	num_kernel_contexts =
@@ -13476,22 +13476,17 @@ static void init_qpmap_table(struct hfi1_devdata *dd,
 /**
  * init_qos - init RX qos
  * @dd - device data
- * @first_context
- *
- * This routine initializes Rule 0 and the
- * RSM map table to implement qos.
  *
- * If all of the limit tests succeed,
- * qos is applied based on the array
- * interpretation of krcvqs where
- * entry 0 is VL0.
+ * This routine initializes Rule 0 and the RSM map table to implement
+ * quality of service (qos).
  *
- * The number of vl bits (n) and the number of qpn
- * bits (m) are computed to feed both the RSM map table
- * and the single rule.
+ * If all of the limit tests succeed, qos is applied based on the array
+ * interpretation of krcvqs where entry 0 is VL0.
  *
+ * The number of vl bits (n) and the number of qpn bits (m) are computed to
+ * feed both the RSM map table and the single rule.
  */
-static void init_qos(struct hfi1_devdata *dd, u32 first_ctxt)
+static void init_qos(struct hfi1_devdata *dd)
 {
 	u8 max_by_vl = 0;
 	unsigned qpns_per_vl, ctxt, i, qpn, n = 1, m;
@@ -13521,7 +13516,7 @@ static void init_qos(struct hfi1_devdata *dd, u32 first_ctxt)
 		goto bail;
 	memset(rsmmap, rxcontext, NUM_MAP_REGS * sizeof(u64));
 	/* init the local copy of the table */
-	for (i = 0, ctxt = first_ctxt; i < num_vls; i++) {
+	for (i = 0, ctxt = FIRST_KERNEL_KCTXT; i < num_vls; i++) {
 		unsigned tctxt;
 
 		for (qpn = 0, tctxt = ctxt;
@@ -13549,7 +13544,7 @@ static void init_qos(struct hfi1_devdata *dd, u32 first_ctxt)
 	/* add rule0 */
 	write_csr(dd, RCV_RSM_CFG /* + (8 * 0) */,
 		  RCV_RSM_CFG_ENABLE_OR_CHAIN_RSM0_MASK <<
-		  RCV_RSM_CFG_ENABLE_OR_CHAIN_RSM0_SHIFT |
+			RCV_RSM_CFG_ENABLE_OR_CHAIN_RSM0_SHIFT |
 		  2ull << RCV_RSM_CFG_PACKET_TYPE_SHIFT);
 	write_csr(dd, RCV_RSM_SELECT /* + (8 * 0) */,
 		  LRH_BTH_MATCH_OFFSET << RCV_RSM_SELECT_FIELD1_OFFSET_SHIFT |
@@ -13566,8 +13561,8 @@ static void init_qos(struct hfi1_devdata *dd, u32 first_ctxt)
 	/* Enable RSM */
 	add_rcvctrl(dd, RCV_CTRL_RCV_RSM_ENABLE_SMASK);
 	kfree(rsmmap);
-	/* map everything else to first context */
-	init_qpmap_table(dd, FIRST_KERNEL_KCTXT, MIN_KERNEL_KCTXTS - 1);
+	/* map everything else to the mcast/err/vl15 context */
+	init_qpmap_table(dd, HFI1_CTRL_CTXT, HFI1_CTRL_CTXT);
 	dd->qos_shift = n + 1;
 	return;
 bail:
@@ -13580,8 +13575,7 @@ static void init_rxe(struct hfi1_devdata *dd)
 	/* enable all receive errors */
 	write_csr(dd, RCV_ERR_MASK, ~0ull);
 	/* setup QPN map table - start where VL15 context leaves off */
-	init_qos(dd, dd->n_krcv_queues > MIN_KERNEL_KCTXTS ?
-		 MIN_KERNEL_KCTXTS : 0);
+	init_qos(dd);
 	/*
 	 * make sure RcvCtrl.RcvWcb <= PCIe Device Control
 	 * Register Max_Payload_Size (PCI_EXP_DEVCTL in Linux PCIe config

--
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

[PATCH 09/11] IB/hfi1: Correctly obtain the full service class

[Dennis Dalessandro]

From: Dean Luick <dean.luick-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

The function hdr2sc was using an unshifted mask to obtain
the 5th bit of the service class.  Correct the issue by using
the shifted mask.

Reviewed-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dean Luick <dean.luick-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/infiniband/hw/hfi1/hfi.h |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/drivers/infiniband/hw/hfi1/hfi.h b/drivers/infiniband/hw/hfi1/hfi.h
index 16cbdc4..ac553f1 100644
--- a/drivers/infiniband/hw/hfi1/hfi.h
+++ b/drivers/infiniband/hw/hfi1/hfi.h
@@ -1258,7 +1258,7 @@ void receive_interrupt_work(struct work_struct *work);
 static inline int hdr2sc(struct hfi1_message_header *hdr, u64 rhf)
 {
 	return ((be16_to_cpu(hdr->lrh[0]) >> 12) & 0xf) |
-	       ((!!(rhf & RHF_DC_INFO_MASK)) << 4);
+	       ((!!(rhf & RHF_DC_INFO_SMASK)) << 4);
 }
 
 static inline u16 generate_jkey(kuid_t uid)

--
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

[PATCH 10/11] IB/hfi1: Simplify init_qpmap_table()

[Dennis Dalessandro]

From: Dean Luick <dean.luick-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

Make init_qpmap_table() easier to understand by simplifying
the loop indexing and writing each register when it is "full",
removing the need for a follow-on register write.

Reviewed-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Reviewed-by: Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dean Luick <dean.luick-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/infiniband/hw/hfi1/chip.c |    7 ++-----
 1 files changed, 2 insertions(+), 5 deletions(-)

diff --git a/drivers/infiniband/hw/hfi1/chip.c b/drivers/infiniband/hw/hfi1/chip.c
index 690871e..7523958 100644
--- a/drivers/infiniband/hw/hfi1/chip.c
+++ b/drivers/infiniband/hw/hfi1/chip.c
@@ -13454,20 +13454,17 @@ static void init_qpmap_table(struct hfi1_devdata *dd,
 	int i;
 	u64 ctxt = first_ctxt;
 
-	for (i = 0; i < 256;) {
+	for (i = 0; i < 256; i++) {
 		reg |= ctxt << (8 * (i % 8));
-		i++;
 		ctxt++;
 		if (ctxt > last_ctxt)
 			ctxt = first_ctxt;
-		if (i % 8 == 0) {
+		if (i % 8 == 7) {
 			write_csr(dd, regno, reg);
 			reg = 0;
 			regno += 8;
 		}
 	}
-	if (i % 8)
-		write_csr(dd, regno, reg);
 
 	add_rcvctrl(dd, RCV_CTRL_RCV_QP_MAP_ENABLE_SMASK
 			| RCV_CTRL_RCV_BYPASS_ENABLE_SMASK);

--
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

[PATCH 11/11] IB/hfi1: Adjust default MTU to be 10KB

[Dennis Dalessandro]

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

Increasing the default MTU size to 10KB improves performance
for PSM. Change the default MTU to 10KB but constrain
Verbs MTU to 8KB.

Reviewed-by: Dean Luick <dean.luick-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Reviewed-by: Mitko Haralanov <mitko.haralanov-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Reviewed-by: Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Sebastian Sanchez <sebastian.sanchez-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/infiniband/hw/hfi1/hfi.h |    4 ++--
 drivers/infiniband/hw/hfi1/qp.c  |    6 +++++-
 2 files changed, 7 insertions(+), 3 deletions(-)

diff --git a/drivers/infiniband/hw/hfi1/hfi.h b/drivers/infiniband/hw/hfi1/hfi.h
index ac553f1..ff04593 100644
--- a/drivers/infiniband/hw/hfi1/hfi.h
+++ b/drivers/infiniband/hw/hfi1/hfi.h
@@ -455,9 +455,9 @@ struct rvt_sge_state;
 #define HLS_UP (HLS_UP_INIT | HLS_UP_ARMED | HLS_UP_ACTIVE)
 
 /* use this MTU size if none other is given */
-#define HFI1_DEFAULT_ACTIVE_MTU 8192
+#define HFI1_DEFAULT_ACTIVE_MTU 10240
 /* use this MTU size as the default maximum */
-#define HFI1_DEFAULT_MAX_MTU 8192
+#define HFI1_DEFAULT_MAX_MTU 10240
 /* default partition key */
 #define DEFAULT_PKEY 0xffff
 
diff --git a/drivers/infiniband/hw/hfi1/qp.c b/drivers/infiniband/hw/hfi1/qp.c
index 29a5ad2..e68d08a 100644
--- a/drivers/infiniband/hw/hfi1/qp.c
+++ b/drivers/infiniband/hw/hfi1/qp.c
@@ -167,8 +167,12 @@ static inline int opa_mtu_enum_to_int(int mtu)
  */
 static inline int verbs_mtu_enum_to_int(struct ib_device *dev, enum ib_mtu mtu)
 {
-	int val = opa_mtu_enum_to_int((int)mtu);
+	int val;
 
+	/* Constraining 10KB packets to 8KB packets */
+	if (mtu == (enum ib_mtu)OPA_MTU_10240)
+		mtu = OPA_MTU_8192;
+	val = opa_mtu_enum_to_int((int)mtu);
 	if (val > 0)
 		return val;
 	return ib_mtu_enum_to_int(mtu);

--
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

Re: [PATCH 00/11] IB/hfi1: Additional fixes for 4.6

[Dennis Dalessandro]

On Tue, Mar 15, 2016 at 10:54:00AM -0700, Dennis Dalessandro wrote:
>These are some more fixes we've identified that we would like to make it into
>4.6 if possible. Two of these are related to the pinned page cache series which
>is undergoing discussion if that series does not go these two from Mitko below
>would not apply.

Doug,

You can drop this series. I'll resubmit since these target the files in the 
infiniband/hfi1 directory rather than staging/rdma/hfi1.

I have some additional patches as well. Perhaps it is more than we want in 
an RC. I'm going to reorder things and submit two patch sets. One that fixes 
critical bugs, and should most likely go in RC. The other we would certainly 
like to be in an RC but fix less critical issues.

-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

Re: [PATCH 00/11] IB/hfi1: Additional fixes for 4.6

[Sagi Grimberg]

Hey Dennis,

> I have some additional patches as well. Perhaps it is more than we want
> in an RC. I'm going to reorder things and submit two patch sets. One
> that fixes critical bugs, and should most likely go in RC.

I was just about to point out that the last two patches really don't
look like RC material. I'd suggest deferring these two for the next
merge window...
--
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

Re: [PATCH 00/11] IB/hfi1: Additional fixes for 4.6

[Dennis Dalessandro]

On Sun, Apr 10, 2016 at 08:25:28PM +0300, Sagi Grimberg wrote:
>Hey Dennis,
>
>>I have some additional patches as well. Perhaps it is more than we want
>>in an RC. I'm going to reorder things and submit two patch sets. One
>>that fixes critical bugs, and should most likely go in RC.
>
>I was just about to point out that the last two patches really don't
>look like RC material. I'd suggest deferring these two for the next
>merge window...

Yes I tend to agree. There are three patches from this series that I think 
need to go in the next RC:

[PATCH 03/11] IB/rdmavt: Fix adaptive pio hang
http://marc.info/?l=linux-rdma&m=145806446130951&w=2

[PATCH 04/11] IB/hfi1: Prevent NULL pointer deferences in caching code
http://marc.info/?l=linux-rdma&m=145806447530956&w=2

[PATCH 05/11] IB/hfi1: Fix deadlock caused by locking with wrong scope
http://marc.info/?l=linux-rdma&m=145806447730959&w=2

I have some others that I hope to get out really soon as well, they also fix 
major bugs. I'll re-send a new series soon.

-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