14 files changed, 250 insertions, 136 deletions

diff --git a/drivers/infiniband/hw/hfi1/affinity.c b/drivers/infiniband/hw/hfi1/affinity.c
index 79575ee873f2..0566393e5aba 100644
--- a/drivers/infiniband/hw/hfi1/affinity.c
+++ b/drivers/infiniband/hw/hfi1/affinity.c
@@ -47,7 +47,6 @@
 #include <linux/topology.h>
 #include <linux/cpumask.h>
 #include <linux/module.h>
-#include <linux/cpumask.h>
 
 #include "hfi.h"
 #include "affinity.h"
@@ -682,7 +681,7 @@ int hfi1_set_sdma_affinity(struct hfi1_devdata *dd, const char *buf,
 			   size_t count)
 {
 	struct hfi1_affinity_node *entry;
-	struct cpumask mask;
+	cpumask_var_t mask;
 	int ret, i;
 
 	spin_lock(&node_affinity.lock);
@@ -692,19 +691,24 @@ int hfi1_set_sdma_affinity(struct hfi1_devdata *dd, const char *buf,
 	if (!entry)
 		return -EINVAL;
 
-	ret = cpulist_parse(buf, &mask);
+	ret = zalloc_cpumask_var(&mask, GFP_KERNEL);
+	if (!ret)
+		return -ENOMEM;
+
+	ret = cpulist_parse(buf, mask);
 	if (ret)
-		return ret;
+		goto out;
 
-	if (!cpumask_subset(&mask, cpu_online_mask) || cpumask_empty(&mask)) {
+	if (!cpumask_subset(mask, cpu_online_mask) || cpumask_empty(mask)) {
 		dd_dev_warn(dd, "Invalid CPU mask\n");
-		return -EINVAL;
+		ret = -EINVAL;
+		goto out;
 	}
 
 	mutex_lock(&sdma_affinity_mutex);
 	/* reset the SDMA interrupt affinity details */
 	init_cpu_mask_set(&entry->def_intr);
-	cpumask_copy(&entry->def_intr.mask, &mask);
+	cpumask_copy(&entry->def_intr.mask, mask);
 	/*
 	 * Reassign the affinity for each SDMA interrupt.
 	 */
@@ -720,8 +724,9 @@ int hfi1_set_sdma_affinity(struct hfi1_devdata *dd, const char *buf,
 		if (ret)
 			break;
 	}
-
 	mutex_unlock(&sdma_affinity_mutex);
+out:
+	free_cpumask_var(mask);
 	return ret ? ret : strnlen(buf, PAGE_SIZE);
 }
 
diff --git a/drivers/infiniband/hw/hfi1/chip.c b/drivers/infiniband/hw/hfi1/chip.c
index b32638d58ae8..cc38004cea42 100644
--- a/drivers/infiniband/hw/hfi1/chip.c
+++ b/drivers/infiniband/hw/hfi1/chip.c
@@ -9490,6 +9490,78 @@ static void init_lcb(struct hfi1_devdata *dd)
 	write_csr(dd, DC_LCB_CFG_TX_FIFOS_RESET, 0x00);
 }
 
+/*
+ * Perform a test read on the QSFP.  Return 0 on success, -ERRNO
+ * on error.
+ */
+static int test_qsfp_read(struct hfi1_pportdata *ppd)
+{
+	int ret;
+	u8 status;
+
+	/* report success if not a QSFP */
+	if (ppd->port_type != PORT_TYPE_QSFP)
+		return 0;
+
+	/* read byte 2, the status byte */
+	ret = one_qsfp_read(ppd, ppd->dd->hfi1_id, 2, &status, 1);
+	if (ret < 0)
+		return ret;
+	if (ret != 1)
+		return -EIO;
+
+	return 0; /* success */
+}
+
+/*
+ * Values for QSFP retry.
+ *
+ * Give up after 10s (20 x 500ms).  The overall timeout was empirically
+ * arrived at from experience on a large cluster.
+ */
+#define MAX_QSFP_RETRIES 20
+#define QSFP_RETRY_WAIT 500 /* msec */
+
+/*
+ * Try a QSFP read.  If it fails, schedule a retry for later.
+ * Called on first link activation after driver load.
+ */
+static void try_start_link(struct hfi1_pportdata *ppd)
+{
+	if (test_qsfp_read(ppd)) {
+		/* read failed */
+		if (ppd->qsfp_retry_count >= MAX_QSFP_RETRIES) {
+			dd_dev_err(ppd->dd, "QSFP not responding, giving up\n");
+			return;
+		}
+		dd_dev_info(ppd->dd,
+			    "QSFP not responding, waiting and retrying %d\n",
+			    (int)ppd->qsfp_retry_count);
+		ppd->qsfp_retry_count++;
+		queue_delayed_work(ppd->hfi1_wq, &ppd->start_link_work,
+				   msecs_to_jiffies(QSFP_RETRY_WAIT));
+		return;
+	}
+	ppd->qsfp_retry_count = 0;
+
+	/*
+	 * Tune the SerDes to a ballpark setting for optimal signal and bit
+	 * error rate.  Needs to be done before starting the link.
+	 */
+	tune_serdes(ppd);
+	start_link(ppd);
+}
+
+/*
+ * Workqueue function to start the link after a delay.
+ */
+void handle_start_link(struct work_struct *work)
+{
+	struct hfi1_pportdata *ppd = container_of(work, struct hfi1_pportdata,
+						  start_link_work.work);
+	try_start_link(ppd);
+}
+
 int bringup_serdes(struct hfi1_pportdata *ppd)
 {
 	struct hfi1_devdata *dd = ppd->dd;
@@ -9525,14 +9597,8 @@ int bringup_serdes(struct hfi1_pportdata *ppd)
 		set_qsfp_int_n(ppd, 1);
 	}
 
-	/*
-	 * Tune the SerDes to a ballpark setting for
-	 * optimal signal and bit error rate
-	 * Needs to be done before starting the link
-	 */
-	tune_serdes(ppd);
-
-	return start_link(ppd);
+	try_start_link(ppd);
+	return 0;
 }
 
 void hfi1_quiet_serdes(struct hfi1_pportdata *ppd)
@@ -9549,6 +9615,10 @@ void hfi1_quiet_serdes(struct hfi1_pportdata *ppd)
 	ppd->driver_link_ready = 0;
 	ppd->link_enabled = 0;
 
+	ppd->qsfp_retry_count = MAX_QSFP_RETRIES; /* prevent more retries */
+	flush_delayed_work(&ppd->start_link_work);
+	cancel_delayed_work_sync(&ppd->start_link_work);
+
 	ppd->offline_disabled_reason =
 			HFI1_ODR_MASK(OPA_LINKDOWN_REASON_SMA_DISABLED);
 	set_link_down_reason(ppd, OPA_LINKDOWN_REASON_SMA_DISABLED, 0,
@@ -12865,7 +12935,7 @@ fail:
  */
 static int set_up_context_variables(struct hfi1_devdata *dd)
 {
-	int num_kernel_contexts;
+	unsigned long num_kernel_contexts;
 	int total_contexts;
 	int ret;
 	unsigned ngroups;
@@ -12894,9 +12964,9 @@ static int set_up_context_variables(struct hfi1_devdata *dd)
 	 */
 	if (num_kernel_contexts > (dd->chip_send_contexts - num_vls - 1)) {
 		dd_dev_err(dd,
-			   "Reducing # kernel rcv contexts to: %d, from %d\n",
+			   "Reducing # kernel rcv contexts to: %d, from %lu\n",
 			   (int)(dd->chip_send_contexts - num_vls - 1),
-			   (int)num_kernel_contexts);
+			   num_kernel_contexts);
 		num_kernel_contexts = dd->chip_send_contexts - num_vls - 1;
 	}
 	/*
diff --git a/drivers/infiniband/hw/hfi1/chip.h b/drivers/infiniband/hw/hfi1/chip.h
index ed11107c50fe..e29573769efc 100644
--- a/drivers/infiniband/hw/hfi1/chip.h
+++ b/drivers/infiniband/hw/hfi1/chip.h
@@ -706,6 +706,7 @@ void handle_link_up(struct work_struct *work);
 void handle_link_down(struct work_struct *work);
 void handle_link_downgrade(struct work_struct *work);
 void handle_link_bounce(struct work_struct *work);
+void handle_start_link(struct work_struct *work);
 void handle_sma_message(struct work_struct *work);
 void reset_qsfp(struct hfi1_pportdata *ppd);
 void qsfp_event(struct work_struct *work);
diff --git a/drivers/infiniband/hw/hfi1/debugfs.c b/drivers/infiniband/hw/hfi1/debugfs.c
index dbab9d9cc288..5e9be16f6cd3 100644
--- a/drivers/infiniband/hw/hfi1/debugfs.c
+++ b/drivers/infiniband/hw/hfi1/debugfs.c
@@ -59,6 +59,40 @@
 
 static struct dentry *hfi1_dbg_root;
 
+/* wrappers to enforce srcu in seq file */
+static ssize_t hfi1_seq_read(
+	struct file *file,
+	char __user *buf,
+	size_t size,
+	loff_t *ppos)
+{
+	struct dentry *d = file->f_path.dentry;
+	int srcu_idx;
+	ssize_t r;
+
+	r = debugfs_use_file_start(d, &srcu_idx);
+	if (likely(!r))
+		r = seq_read(file, buf, size, ppos);
+	debugfs_use_file_finish(srcu_idx);
+	return r;
+}
+
+static loff_t hfi1_seq_lseek(
+	struct file *file,
+	loff_t offset,
+	int whence)
+{
+	struct dentry *d = file->f_path.dentry;
+	int srcu_idx;
+	loff_t r;
+
+	r = debugfs_use_file_start(d, &srcu_idx);
+	if (likely(!r))
+		r = seq_lseek(file, offset, whence);
+	debugfs_use_file_finish(srcu_idx);
+	return r;
+}
+
 #define private2dd(file) (file_inode(file)->i_private)
 #define private2ppd(file) (file_inode(file)->i_private)
 
@@ -87,8 +121,8 @@ static int _##name##_open(struct inode *inode, struct file *s) \
 static const struct file_operations _##name##_file_ops = { \
 	.owner   = THIS_MODULE, \
 	.open    = _##name##_open, \
-	.read    = seq_read, \
-	.llseek  = seq_lseek, \
+	.read    = hfi1_seq_read, \
+	.llseek  = hfi1_seq_lseek, \
 	.release = seq_release \
 }
 
@@ -105,11 +139,9 @@ do { \
 	DEBUGFS_FILE_CREATE(#name, parent, data, &_##name##_file_ops, S_IRUGO)
 
 static void *_opcode_stats_seq_start(struct seq_file *s, loff_t *pos)
-__acquires(RCU)
 {
 	struct hfi1_opcode_stats_perctx *opstats;
 
-	rcu_read_lock();
 	if (*pos >= ARRAY_SIZE(opstats->stats))
 		return NULL;
 	return pos;
@@ -126,9 +158,7 @@ static void *_opcode_stats_seq_next(struct seq_file *s, void *v, loff_t *pos)
 }
 
 static void _opcode_stats_seq_stop(struct seq_file *s, void *v)
-__releases(RCU)
 {
-	rcu_read_unlock();
 }
 
 static int _opcode_stats_seq_show(struct seq_file *s, void *v)
@@ -223,28 +253,32 @@ DEBUGFS_SEQ_FILE_OPEN(ctx_stats)
 DEBUGFS_FILE_OPS(ctx_stats);
 
 static void *_qp_stats_seq_start(struct seq_file *s, loff_t *pos)
-__acquires(RCU)
+	__acquires(RCU)
 {
 	struct qp_iter *iter;
 	loff_t n = *pos;
 
-	rcu_read_lock();
 	iter = qp_iter_init(s->private);
+
+	/* stop calls rcu_read_unlock */
+	rcu_read_lock();
+
 	if (!iter)
 		return NULL;
 
-	while (n--) {
+	do {
 		if (qp_iter_next(iter)) {
 			kfree(iter);
 			return NULL;
 		}
-	}
+	} while (n--);
 
 	return iter;
 }
 
 static void *_qp_stats_seq_next(struct seq_file *s, void *iter_ptr,
 				loff_t *pos)
+	__must_hold(RCU)
 {
 	struct qp_iter *iter = iter_ptr;
 
@@ -259,7 +293,7 @@ static void *_qp_stats_seq_next(struct seq_file *s, void *iter_ptr,
 }
 
 static void _qp_stats_seq_stop(struct seq_file *s, void *iter_ptr)
-__releases(RCU)
+	__releases(RCU)
 {
 	rcu_read_unlock();
 }
@@ -281,12 +315,10 @@ DEBUGFS_SEQ_FILE_OPEN(qp_stats)
 DEBUGFS_FILE_OPS(qp_stats);
 
 static void *_sdes_seq_start(struct seq_file *s, loff_t *pos)
-__acquires(RCU)
 {
 	struct hfi1_ibdev *ibd;
 	struct hfi1_devdata *dd;
 
-	rcu_read_lock();
 	ibd = (struct hfi1_ibdev *)s->private;
 	dd = dd_from_dev(ibd);
 	if (!dd->per_sdma || *pos >= dd->num_sdma)
@@ -306,9 +338,7 @@ static void *_sdes_seq_next(struct seq_file *s, void *v, loff_t *pos)
 }
 
 static void _sdes_seq_stop(struct seq_file *s, void *v)
-__releases(RCU)
 {
-	rcu_read_unlock();
 }
 
 static int _sdes_seq_show(struct seq_file *s, void *v)
@@ -335,11 +365,9 @@ static ssize_t dev_counters_read(struct file *file, char __user *buf,
 	struct hfi1_devdata *dd;
 	ssize_t rval;
 
-	rcu_read_lock();
 	dd = private2dd(file);
 	avail = hfi1_read_cntrs(dd, NULL, &counters);
 	rval =  simple_read_from_buffer(buf, count, ppos, counters, avail);
-	rcu_read_unlock();
 	return rval;
 }
 
@@ -352,11 +380,9 @@ static ssize_t dev_names_read(struct file *file, char __user *buf,
 	struct hfi1_devdata *dd;
 	ssize_t rval;
 
-	rcu_read_lock();
 	dd = private2dd(file);
 	avail = hfi1_read_cntrs(dd, &names, NULL);
 	rval =  simple_read_from_buffer(buf, count, ppos, names, avail);
-	rcu_read_unlock();
 	return rval;
 }
 
@@ -379,11 +405,9 @@ static ssize_t portnames_read(struct file *file, char __user *buf,
 	struct hfi1_devdata *dd;
 	ssize_t rval;
 
-	rcu_read_lock();
 	dd = private2dd(file);
 	avail = hfi1_read_portcntrs(dd->pport, &names, NULL);
 	rval = simple_read_from_buffer(buf, count, ppos, names, avail);
-	rcu_read_unlock();
 	return rval;
 }
 
@@ -396,11 +420,9 @@ static ssize_t portcntrs_debugfs_read(struct file *file, char __user *buf,
 	struct hfi1_pportdata *ppd;
 	ssize_t rval;
 
-	rcu_read_lock();
 	ppd = private2ppd(file);
 	avail = hfi1_read_portcntrs(ppd, NULL, &counters);
 	rval = simple_read_from_buffer(buf, count, ppos, counters, avail);
-	rcu_read_unlock();
 	return rval;
 }
 
@@ -430,16 +452,13 @@ static ssize_t asic_flags_read(struct file *file, char __user *buf,
 	int used;
 	int i;
 
-	rcu_read_lock();
 	ppd = private2ppd(file);
 	dd = ppd->dd;
 	size = PAGE_SIZE;
 	used = 0;
 	tmp = kmalloc(size, GFP_KERNEL);
-	if (!tmp) {
-		rcu_read_unlock();
+	if (!tmp)
 		return -ENOMEM;
-	}
 
 	scratch0 = read_csr(dd, ASIC_CFG_SCRATCH);
 	used += scnprintf(tmp + used, size - used,
@@ -466,7 +485,6 @@ static ssize_t asic_flags_read(struct file *file, char __user *buf,
 	used += scnprintf(tmp + used, size - used, "Write bits to clear\n");
 
 	ret = simple_read_from_buffer(buf, count, ppos, tmp, used);
-	rcu_read_unlock();
 	kfree(tmp);
 	return ret;
 }
@@ -482,15 +500,12 @@ static ssize_t asic_flags_write(struct file *file, const char __user *buf,
 	u64 scratch0;
 	u64 clear;
 
-	rcu_read_lock();
 	ppd = private2ppd(file);
 	dd = ppd->dd;
 
 	buff = kmalloc(count + 1, GFP_KERNEL);
-	if (!buff) {
-		ret = -ENOMEM;
-		goto do_return;
-	}
+	if (!buff)
+		return -ENOMEM;
 
 	ret = copy_from_user(buff, buf, count);
 	if (ret > 0) {
@@ -523,8 +538,6 @@ static ssize_t asic_flags_write(struct file *file, const char __user *buf,
 
  do_free:
 	kfree(buff);
- do_return:
-	rcu_read_unlock();
 	return ret;
 }
 
@@ -538,18 +551,14 @@ static ssize_t qsfp_debugfs_dump(struct file *file, char __user *buf,
 	char *tmp;
 	int ret;
 
-	rcu_read_lock();
 	ppd = private2ppd(file);
 	tmp = kmalloc(PAGE_SIZE, GFP_KERNEL);
-	if (!tmp) {
-		rcu_read_unlock();
+	if (!tmp)
 		return -ENOMEM;
-	}
 
 	ret = qsfp_dump(ppd, tmp, PAGE_SIZE);
 	if (ret > 0)
 		ret = simple_read_from_buffer(buf, count, ppos, tmp, ret);
-	rcu_read_unlock();
 	kfree(tmp);
 	return ret;
 }
@@ -565,7 +574,6 @@ static ssize_t __i2c_debugfs_write(struct file *file, const char __user *buf,
 	int offset;
 	int total_written;
 
-	rcu_read_lock();
 	ppd = private2ppd(file);
 
 	/* byte offset format: [offsetSize][i2cAddr][offsetHigh][offsetLow] */
@@ -573,16 +581,12 @@ static ssize_t __i2c_debugfs_write(struct file *file, const char __user *buf,
 	offset = *ppos & 0xffff;
 
 	/* explicitly reject invalid address 0 to catch cp and cat */
-	if (i2c_addr == 0) {
-		ret = -EINVAL;
-		goto _return;
-	}
+	if (i2c_addr == 0)
+		return -EINVAL;
 
 	buff = kmalloc(count, GFP_KERNEL);
-	if (!buff) {
-		ret = -ENOMEM;
-		goto _return;
-	}
+	if (!buff)
+		return -ENOMEM;
 
 	ret = copy_from_user(buff, buf, count);
 	if (ret > 0) {
@@ -602,8 +606,6 @@ static ssize_t __i2c_debugfs_write(struct file *file, const char __user *buf,
 
  _free:
 	kfree(buff);
- _return:
-	rcu_read_unlock();
 	return ret;
 }
 
@@ -632,7 +634,6 @@ static ssize_t __i2c_debugfs_read(struct file *file, char __user *buf,
 	int offset;
 	int total_read;
 
-	rcu_read_lock();
 	ppd = private2ppd(file);
 
 	/* byte offset format: [offsetSize][i2cAddr][offsetHigh][offsetLow] */
@@ -640,16 +641,12 @@ static ssize_t __i2c_debugfs_read(struct file *file, char __user *buf,
 	offset = *ppos & 0xffff;
 
 	/* explicitly reject invalid address 0 to catch cp and cat */
-	if (i2c_addr == 0) {
-		ret = -EINVAL;
-		goto _return;
-	}
+	if (i2c_addr == 0)
+		return -EINVAL;
 
 	buff = kmalloc(count, GFP_KERNEL);
-	if (!buff) {
-		ret = -ENOMEM;
-		goto _return;
-	}
+	if (!buff)
+		return -ENOMEM;
 
 	total_read = i2c_read(ppd, target, i2c_addr, offset, buff, count);
 	if (total_read < 0) {
@@ -669,8 +666,6 @@ static ssize_t __i2c_debugfs_read(struct file *file, char __user *buf,
 
  _free:
 	kfree(buff);
- _return:
-	rcu_read_unlock();
 	return ret;
 }
 
@@ -697,26 +692,20 @@ static ssize_t __qsfp_debugfs_write(struct file *file, const char __user *buf,
 	int ret;
 	int total_written;
 
-	rcu_read_lock();
-	if (*ppos + count > QSFP_PAGESIZE * 4) { /* base page + page00-page03 */
-		ret = -EINVAL;
-		goto _return;
-	}
+	if (*ppos + count > QSFP_PAGESIZE * 4) /* base page + page00-page03 */
+		return -EINVAL;
 
 	ppd = private2ppd(file);
 
 	buff = kmalloc(count, GFP_KERNEL);
-	if (!buff) {
-		ret = -ENOMEM;
-		goto _return;
-	}
+	if (!buff)
+		return -ENOMEM;
 
 	ret = copy_from_user(buff, buf, count);
 	if (ret > 0) {
 		ret = -EFAULT;
 		goto _free;
 	}
-
 	total_written = qsfp_write(ppd, target, *ppos, buff, count);
 	if (total_written < 0) {
 		ret = total_written;
@@ -729,8 +718,6 @@ static ssize_t __qsfp_debugfs_write(struct file *file, const char __user *buf,
 
  _free:
 	kfree(buff);
- _return:
-	rcu_read_unlock();
 	return ret;
 }
 
@@ -757,7 +744,6 @@ static ssize_t __qsfp_debugfs_read(struct file *file, char __user *buf,
 	int ret;
 	int total_read;
 
-	rcu_read_lock();
 	if (*ppos + count > QSFP_PAGESIZE * 4) { /* base page + page00-page03 */
 		ret = -EINVAL;
 		goto _return;
@@ -790,7 +776,6 @@ static ssize_t __qsfp_debugfs_read(struct file *file, char __user *buf,
  _free:
 	kfree(buff);
  _return:
-	rcu_read_unlock();
 	return ret;
 }
 
@@ -1006,7 +991,6 @@ void hfi1_dbg_ibdev_exit(struct hfi1_ibdev *ibd)
 	debugfs_remove_recursive(ibd->hfi1_ibdev_dbg);
 out:
 	ibd->hfi1_ibdev_dbg = NULL;
-	synchronize_rcu();
 }
 
 /*
@@ -1031,9 +1015,7 @@ static const char * const hfi1_statnames[] = {
 };
 
 static void *_driver_stats_names_seq_start(struct seq_file *s, loff_t *pos)
-__acquires(RCU)
 {
-	rcu_read_lock();
 	if (*pos >= ARRAY_SIZE(hfi1_statnames))
 		return NULL;
 	return pos;
@@ -1051,9 +1033,7 @@ static void *_driver_stats_names_seq_next(
 }
 
 static void _driver_stats_names_seq_stop(struct seq_file *s, void *v)
-__releases(RCU)
 {
-	rcu_read_unlock();
 }
 
 static int _driver_stats_names_seq_show(struct seq_file *s, void *v)
@@ -1069,9 +1049,7 @@ DEBUGFS_SEQ_FILE_OPEN(driver_stats_names)
 DEBUGFS_FILE_OPS(driver_stats_names);
 
 static void *_driver_stats_seq_start(struct seq_file *s, loff_t *pos)
-__acquires(RCU)
 {
-	rcu_read_lock();
 	if (*pos >= ARRAY_SIZE(hfi1_statnames))
 		return NULL;
 	return pos;
@@ -1086,9 +1064,7 @@ static void *_driver_stats_seq_next(struct seq_file *s, void *v, loff_t *pos)
 }
 
 static void _driver_stats_seq_stop(struct seq_file *s, void *v)
-__releases(RCU)
 {
-	rcu_read_unlock();
 }
 
 static u64 hfi1_sps_ints(void)
diff --git a/drivers/infiniband/hw/hfi1/driver.c b/drivers/infiniband/hw/hfi1/driver.c
index 8246dc7d0573..303f10555729 100644
--- a/drivers/infiniband/hw/hfi1/driver.c
+++ b/drivers/infiniband/hw/hfi1/driver.c
@@ -888,14 +888,15 @@ void set_all_slowpath(struct hfi1_devdata *dd)
 }
 
 static inline int set_armed_to_active(struct hfi1_ctxtdata *rcd,
-				      struct hfi1_packet packet,
+				      struct hfi1_packet *packet,
 				      struct hfi1_devdata *dd)
 {
 	struct work_struct *lsaw = &rcd->ppd->linkstate_active_work;
-	struct hfi1_message_header *hdr = hfi1_get_msgheader(packet.rcd->dd,
-							     packet.rhf_addr);
+	struct hfi1_message_header *hdr = hfi1_get_msgheader(packet->rcd->dd,
+							     packet->rhf_addr);
+	u8 etype = rhf_rcv_type(packet->rhf);
 
-	if (hdr2sc(hdr, packet.rhf) != 0xf) {
+	if (etype == RHF_RCV_TYPE_IB && hdr2sc(hdr, packet->rhf) != 0xf) {
 		int hwstate = read_logical_state(dd);
 
 		if (hwstate != LSTATE_ACTIVE) {
@@ -979,7 +980,7 @@ int handle_receive_interrupt(struct hfi1_ctxtdata *rcd, int thread)
 			/* Auto activate link on non-SC15 packet receive */
 			if (unlikely(rcd->ppd->host_link_state ==
 				     HLS_UP_ARMED) &&
-			    set_armed_to_active(rcd, packet, dd))
+			    set_armed_to_active(rcd, &packet, dd))
 				goto bail;
 			last = process_rcv_packet(&packet, thread);
 		}
diff --git a/drivers/infiniband/hw/hfi1/file_ops.c b/drivers/infiniband/hw/hfi1/file_ops.c
index 1ecbec192358..7e03ccd2554d 100644
--- a/drivers/infiniband/hw/hfi1/file_ops.c
+++ b/drivers/infiniband/hw/hfi1/file_ops.c
@@ -183,6 +183,7 @@ static int hfi1_file_open(struct inode *inode, struct file *fp)
 	if (fd) {
 		fd->rec_cpu_num = -1; /* no cpu affinity by default */
 		fd->mm = current->mm;
+		atomic_inc(&fd->mm->mm_count);
 	}
 
 	fp->private_data = fd;
@@ -222,7 +223,7 @@ static long hfi1_file_ioctl(struct file *fp, unsigned int cmd,
 		ret = assign_ctxt(fp, &uinfo);
 		if (ret < 0)
 			return ret;
-		setup_ctxt(fp);
+		ret = setup_ctxt(fp);
 		if (ret)
 			return ret;
 		ret = user_init(fp);
@@ -779,6 +780,7 @@ static int hfi1_file_close(struct inode *inode, struct file *fp)
 	mutex_unlock(&hfi1_mutex);
 	hfi1_free_ctxtdata(dd, uctxt);
 done:
+	mmdrop(fdata->mm);
 	kobject_put(&dd->kobj);
 	kfree(fdata);
 	return 0;
diff --git a/drivers/infiniband/hw/hfi1/hfi.h b/drivers/infiniband/hw/hfi1/hfi.h
index 1000e0fd96d9..325ec211370f 100644
--- a/drivers/infiniband/hw/hfi1/hfi.h
+++ b/drivers/infiniband/hw/hfi1/hfi.h
@@ -605,6 +605,7 @@ struct hfi1_pportdata {
 	struct work_struct freeze_work;
 	struct work_struct link_downgrade_work;
 	struct work_struct link_bounce_work;
+	struct delayed_work start_link_work;
 	/* host link state variables */
 	struct mutex hls_lock;
 	u32 host_link_state;
@@ -659,6 +660,7 @@ struct hfi1_pportdata {
 	u8 linkinit_reason;
 	u8 local_tx_rate;	/* rate given to 8051 firmware */
 	u8 last_pstate;		/* info only */
+	u8 qsfp_retry_count;
 
 	/* placeholders for IB MAD packet settings */
 	u8 overrun_threshold;
@@ -1272,9 +1274,26 @@ static inline int hdr2sc(struct hfi1_message_header *hdr, u64 rhf)
 	       ((!!(rhf_dc_info(rhf))) << 4);
 }
 
+#define HFI1_JKEY_WIDTH       16
+#define HFI1_JKEY_MASK        (BIT(16) - 1)
+#define HFI1_ADMIN_JKEY_RANGE 32
+
+/*
+ * J_KEYs are split and allocated in the following groups:
+ *   0 - 31    - users with administrator privileges
+ *  32 - 63    - kernel protocols using KDETH packets
+ *  64 - 65535 - all other users using KDETH packets
+ */
 static inline u16 generate_jkey(kuid_t uid)
 {
-	return from_kuid(current_user_ns(), uid) & 0xffff;
+	u16 jkey = from_kuid(current_user_ns(), uid) & HFI1_JKEY_MASK;
+
+	if (capable(CAP_SYS_ADMIN))
+		jkey &= HFI1_ADMIN_JKEY_RANGE - 1;
+	else if (jkey < 64)
+		jkey |= BIT(HFI1_JKEY_WIDTH - 1);
+
+	return jkey;
 }
 
 /*
@@ -1656,7 +1675,6 @@ struct cc_state *get_cc_state_protected(struct hfi1_pportdata *ppd)
 struct hfi1_devdata *hfi1_init_dd(struct pci_dev *,
 				  const struct pci_device_id *);
 void hfi1_free_devdata(struct hfi1_devdata *);
-void cc_state_reclaim(struct rcu_head *rcu);
 struct hfi1_devdata *hfi1_alloc_devdata(struct pci_dev *pdev, size_t extra);
 
 /* LED beaconing functions */
@@ -1788,7 +1806,7 @@ extern unsigned int hfi1_max_mtu;
 extern unsigned int hfi1_cu;
 extern unsigned int user_credit_return_threshold;
 extern int num_user_contexts;
-extern unsigned n_krcvqs;
+extern unsigned long n_krcvqs;
 extern uint krcvqs[];
 extern int krcvqsset;
 extern uint kdeth_qp;
diff --git a/drivers/infiniband/hw/hfi1/init.c b/drivers/infiniband/hw/hfi1/init.c
index a358d23ecd54..384b43d2fd49 100644
--- a/drivers/infiniband/hw/hfi1/init.c
+++ b/drivers/infiniband/hw/hfi1/init.c
@@ -94,7 +94,7 @@ module_param_array(krcvqs, uint, &krcvqsset, S_IRUGO);
 MODULE_PARM_DESC(krcvqs, "Array of the number of non-control kernel receive queues by VL");
 
 /* computed based on above array */
-unsigned n_krcvqs;
+unsigned long n_krcvqs;
 
 static unsigned hfi1_rcvarr_split = 25;
 module_param_named(rcvarr_split, hfi1_rcvarr_split, uint, S_IRUGO);
@@ -500,6 +500,7 @@ void hfi1_init_pportdata(struct pci_dev *pdev, struct hfi1_pportdata *ppd,
 	INIT_WORK(&ppd->link_downgrade_work, handle_link_downgrade);
 	INIT_WORK(&ppd->sma_message_work, handle_sma_message);
 	INIT_WORK(&ppd->link_bounce_work, handle_link_bounce);
+	INIT_DELAYED_WORK(&ppd->start_link_work, handle_start_link);
 	INIT_WORK(&ppd->linkstate_active_work, receive_interrupt_work);
 	INIT_WORK(&ppd->qsfp_info.qsfp_work, qsfp_event);
 
@@ -1333,7 +1334,7 @@ static void cleanup_device_data(struct hfi1_devdata *dd)
 		spin_unlock(&ppd->cc_state_lock);
 
 		if (cc_state)
-			call_rcu(&cc_state->rcu, cc_state_reclaim);
+			kfree_rcu(cc_state, rcu);
 	}
 
 	free_credit_return(dd);
diff --git a/drivers/infiniband/hw/hfi1/mad.c b/drivers/infiniband/hw/hfi1/mad.c
index 1263abe01999..7ffc14f21523 100644
--- a/drivers/infiniband/hw/hfi1/mad.c
+++ b/drivers/infiniband/hw/hfi1/mad.c
@@ -1819,6 +1819,11 @@ static int __subn_get_opa_cable_info(struct opa_smp *smp, u32 am, u8 *data,
 	u32 len = OPA_AM_CI_LEN(am) + 1;
 	int ret;
 
+	if (dd->pport->port_type != PORT_TYPE_QSFP) {
+		smp->status |= IB_SMP_INVALID_FIELD;
+		return reply((struct ib_mad_hdr *)smp);
+	}
+
 #define __CI_PAGE_SIZE BIT(7) /* 128 bytes */
 #define __CI_PAGE_MASK ~(__CI_PAGE_SIZE - 1)
 #define __CI_PAGE_NUM(a) ((a) & __CI_PAGE_MASK)
@@ -2599,7 +2604,7 @@ static int pma_get_opa_datacounters(struct opa_pma_mad *pmp,
 	u8 lq, num_vls;
 	u8 res_lli, res_ler;
 	u64 port_mask;
-	unsigned long port_num;
+	u8 port_num;
 	unsigned long vl;
 	u32 vl_select_mask;
 	int vfi;
@@ -2633,9 +2638,9 @@ static int pma_get_opa_datacounters(struct opa_pma_mad *pmp,
 	 */
 	port_mask = be64_to_cpu(req->port_select_mask[3]);
 	port_num = find_first_bit((unsigned long *)&port_mask,
-				  sizeof(port_mask));
+				  sizeof(port_mask) * 8);
 
-	if ((u8)port_num != port) {
+	if (port_num != port) {
 		pmp->mad_hdr.status |= IB_SMP_INVALID_FIELD;
 		return reply((struct ib_mad_hdr *)pmp);
 	}
@@ -2837,7 +2842,7 @@ static int pma_get_opa_porterrors(struct opa_pma_mad *pmp,
 	 */
 	port_mask = be64_to_cpu(req->port_select_mask[3]);
 	port_num = find_first_bit((unsigned long *)&port_mask,
-				  sizeof(port_mask));
+				  sizeof(port_mask) * 8);
 
 	if (port_num != port) {
 		pmp->mad_hdr.status |= IB_SMP_INVALID_FIELD;
@@ -3010,7 +3015,7 @@ static int pma_get_opa_errorinfo(struct opa_pma_mad *pmp,
 	 */
 	port_mask = be64_to_cpu(req->port_select_mask[3]);
 	port_num = find_first_bit((unsigned long *)&port_mask,
-				  sizeof(port_mask));
+				  sizeof(port_mask) * 8);
 
 	if (port_num != port) {
 		pmp->mad_hdr.status |= IB_SMP_INVALID_FIELD;
@@ -3247,7 +3252,7 @@ static int pma_set_opa_errorinfo(struct opa_pma_mad *pmp,
 	 */
 	port_mask = be64_to_cpu(req->port_select_mask[3]);
 	port_num = find_first_bit((unsigned long *)&port_mask,
-				  sizeof(port_mask));
+				  sizeof(port_mask) * 8);
 
 	if (port_num != port) {
 		pmp->mad_hdr.status |= IB_SMP_INVALID_FIELD;
@@ -3398,7 +3403,7 @@ static void apply_cc_state(struct hfi1_pportdata *ppd)
 
 	spin_unlock(&ppd->cc_state_lock);
 
-	call_rcu(&old_cc_state->rcu, cc_state_reclaim);
+	kfree_rcu(old_cc_state, rcu);
 }
 
 static int __subn_set_opa_cong_setting(struct opa_smp *smp, u32 am, u8 *data,
@@ -3553,13 +3558,6 @@ static int __subn_get_opa_cc_table(struct opa_smp *smp, u32 am, u8 *data,
 	return reply((struct ib_mad_hdr *)smp);
 }
 
-void cc_state_reclaim(struct rcu_head *rcu)
-{
-	struct cc_state *cc_state = container_of(rcu, struct cc_state, rcu);
-
-	kfree(cc_state);
-}
-
 static int __subn_set_opa_cc_table(struct opa_smp *smp, u32 am, u8 *data,
 				   struct ib_device *ibdev, u8 port,
 				   u32 *resp_len)
diff --git a/drivers/infiniband/hw/hfi1/pio_copy.c b/drivers/infiniband/hw/hfi1/pio_copy.c
index 8c25e1b58849..3a1ef3056282 100644
--- a/drivers/infiniband/hw/hfi1/pio_copy.c
+++ b/drivers/infiniband/hw/hfi1/pio_copy.c
@@ -771,6 +771,9 @@ void seg_pio_copy_mid(struct pio_buf *pbuf, const void *from, size_t nbytes)
 			read_extra_bytes(pbuf, from, to_fill);
 			from += to_fill;
 			nbytes -= to_fill;
+			/* may not be enough valid bytes left to align */
+			if (extra > nbytes)
+				extra = nbytes;
 
 			/* ...now write carry */
 			dest = pbuf->start + (pbuf->qw_written * sizeof(u64));
@@ -798,6 +801,15 @@ void seg_pio_copy_mid(struct pio_buf *pbuf, const void *from, size_t nbytes)
 			read_low_bytes(pbuf, from, extra);
 			from += extra;
 			nbytes -= extra;
+			/*
+			 * If no bytes are left, return early - we are done.
+			 * NOTE: This short-circuit is *required* because
+			 * "extra" may have been reduced in size and "from"
+			 * is not aligned, as required when leaving this
+			 * if block.
+			 */
+			if (nbytes == 0)
+				return;
 		}
 
 		/* at this point, from is QW aligned */
diff --git a/drivers/infiniband/hw/hfi1/qp.c b/drivers/infiniband/hw/hfi1/qp.c
index a5aa3517e7d5..4e4d8317c281 100644
--- a/drivers/infiniband/hw/hfi1/qp.c
+++ b/drivers/infiniband/hw/hfi1/qp.c
@@ -656,10 +656,6 @@ struct qp_iter *qp_iter_init(struct hfi1_ibdev *dev)
 
 	iter->dev = dev;
 	iter->specials = dev->rdi.ibdev.phys_port_cnt * 2;
-	if (qp_iter_next(iter)) {
-		kfree(iter);
-		return NULL;
-	}
 
 	return iter;
 }
diff --git a/drivers/infiniband/hw/hfi1/qsfp.c b/drivers/infiniband/hw/hfi1/qsfp.c
index a207717ade2a..4e95ad810847 100644
--- a/drivers/infiniband/hw/hfi1/qsfp.c
+++ b/drivers/infiniband/hw/hfi1/qsfp.c
@@ -706,8 +706,8 @@ int get_cable_info(struct hfi1_devdata *dd, u32 port_num, u32 addr, u32 len,
 		   u8 *data)
 {
 	struct hfi1_pportdata *ppd;
-	u32 excess_len = 0;
-	int ret = 0;
+	u32 excess_len = len;
+	int ret = 0, offset = 0;
 
 	if (port_num > dd->num_pports || port_num < 1) {
 		dd_dev_info(dd, "%s: Invalid port number %d\n",
@@ -740,6 +740,34 @@ int get_cable_info(struct hfi1_devdata *dd, u32 port_num, u32 addr, u32 len,
 	}
 
 	memcpy(data, &ppd->qsfp_info.cache[addr], len);
+
+	if (addr <= QSFP_MONITOR_VAL_END &&
+	    (addr + len) >= QSFP_MONITOR_VAL_START) {
+		/* Overlap with the dynamic channel monitor range */
+		if (addr < QSFP_MONITOR_VAL_START) {
+			if (addr + len <= QSFP_MONITOR_VAL_END)
+				len = addr + len - QSFP_MONITOR_VAL_START;
+			else
+				len = QSFP_MONITOR_RANGE;
+			offset = QSFP_MONITOR_VAL_START - addr;
+			addr = QSFP_MONITOR_VAL_START;
+		} else if (addr == QSFP_MONITOR_VAL_START) {
+			offset = 0;
+			if (addr + len > QSFP_MONITOR_VAL_END)
+				len = QSFP_MONITOR_RANGE;
+		} else {
+			offset = 0;
+			if (addr + len > QSFP_MONITOR_VAL_END)
+				len = QSFP_MONITOR_VAL_END - addr + 1;
+		}
+		/* Refresh the values of the dynamic monitors from the cable */
+		ret = one_qsfp_read(ppd, dd->hfi1_id, addr, data + offset, len);
+		if (ret != len) {
+			ret = -EAGAIN;
+			goto set_zeroes;
+		}
+	}
+
 	return 0;
 
 set_zeroes:
diff --git a/drivers/infiniband/hw/hfi1/qsfp.h b/drivers/infiniband/hw/hfi1/qsfp.h
index 69275ebd9597..36cf52359848 100644
--- a/drivers/infiniband/hw/hfi1/qsfp.h
+++ b/drivers/infiniband/hw/hfi1/qsfp.h
@@ -74,6 +74,9 @@
 /* Defined fields that Intel requires of qualified cables */
 /* Byte 0 is Identifier, not checked */
 /* Byte 1 is reserved "status MSB" */
+#define QSFP_MONITOR_VAL_START 22
+#define QSFP_MONITOR_VAL_END 81
+#define QSFP_MONITOR_RANGE (QSFP_MONITOR_VAL_END - QSFP_MONITOR_VAL_START + 1)
 #define QSFP_TX_CTRL_BYTE_OFFS 86
 #define QSFP_PWR_CTRL_BYTE_OFFS 93
 #define QSFP_CDR_CTRL_BYTE_OFFS 98
diff --git a/drivers/infiniband/hw/hfi1/user_sdma.c b/drivers/infiniband/hw/hfi1/user_sdma.c
index 0ecf27903dc2..1694037d1eee 100644
--- a/drivers/infiniband/hw/hfi1/user_sdma.c
+++ b/drivers/infiniband/hw/hfi1/user_sdma.c
@@ -114,6 +114,8 @@ MODULE_PARM_DESC(sdma_comp_size, "Size of User SDMA completion ring. Default: 12
 #define KDETH_HCRC_LOWER_SHIFT    24
 #define KDETH_HCRC_LOWER_MASK     0xff
 
+#define AHG_KDETH_INTR_SHIFT 12
+
 #define PBC2LRH(x) ((((x) & 0xfff) << 2) - 4)
 #define LRH2PBC(x) ((((x) >> 2) + 1) & 0xfff)
 
@@ -1480,7 +1482,8 @@ static int set_txreq_header_ahg(struct user_sdma_request *req,
 		/* Clear KDETH.SH on last packet */
 		if (unlikely(tx->flags & TXREQ_FLAGS_REQ_LAST_PKT)) {
 			val |= cpu_to_le16(KDETH_GET(hdr->kdeth.ver_tid_offset,
-								INTR) >> 16);
+						     INTR) <<
+					   AHG_KDETH_INTR_SHIFT);
 			val &= cpu_to_le16(~(1U << 13));
 			AHG_HEADER_SET(req->ahg, diff, 7, 16, 14, val);
 		} else {