[PATCH AUTOSEL 4.20 058/304] net/mlx5: EQ, Use the right place to store/read IRQ affinity hint

[PATCH AUTOSEL 4.20 058/304] net/mlx5: EQ, Use the right place to store/read IRQ affinity hint

[Sasha Levin]

From: Saeed Mahameed <saeedm@mellanox.com>

[ Upstream commit 1e86ace4c140fd5a693e266c9b23409358f25381 ]

Currently the cpu affinity hint mask for completion EQs is stored and
read from the wrong place, since reading and storing is done from the
same index, there is no actual issue with that, but internal irq_info
for completion EQs stars at MLX5_EQ_VEC_COMP_BASE offset in irq_info
array, this patch changes the code to use the correct offset to store
and read the IRQ affinity hint.

Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/net/ethernet/mellanox/mlx5/core/en_main.c |  2 +-
 drivers/net/ethernet/mellanox/mlx5/core/main.c    | 14 ++++++++------
 include/linux/mlx5/driver.h                       |  2 +-
 3 files changed, 10 insertions(+), 8 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
index b70cb6fd164c..9577d0657839 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
@@ -1771,7 +1771,7 @@ static void mlx5e_close_cq(struct mlx5e_cq *cq)
 
 static int mlx5e_get_cpu(struct mlx5e_priv *priv, int ix)
 {
-	return cpumask_first(priv->mdev->priv.irq_info[ix].mask);
+	return cpumask_first(priv->mdev->priv.irq_info[ix + MLX5_EQ_VEC_COMP_BASE].mask);
 }
 
 static int mlx5e_open_tx_cqs(struct mlx5e_channel *c,
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/main.c b/drivers/net/ethernet/mellanox/mlx5/core/main.c
index 28132c7dc05f..d5cea0a36e6a 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/main.c
@@ -640,18 +640,19 @@ u64 mlx5_read_internal_timer(struct mlx5_core_dev *dev)
 static int mlx5_irq_set_affinity_hint(struct mlx5_core_dev *mdev, int i)
 {
 	struct mlx5_priv *priv  = &mdev->priv;
-	int irq = pci_irq_vector(mdev->pdev, MLX5_EQ_VEC_COMP_BASE + i);
+	int vecidx = MLX5_EQ_VEC_COMP_BASE + i;
+	int irq = pci_irq_vector(mdev->pdev, vecidx);
 
-	if (!zalloc_cpumask_var(&priv->irq_info[i].mask, GFP_KERNEL)) {
+	if (!zalloc_cpumask_var(&priv->irq_info[vecidx].mask, GFP_KERNEL)) {
 		mlx5_core_warn(mdev, "zalloc_cpumask_var failed");
 		return -ENOMEM;
 	}
 
 	cpumask_set_cpu(cpumask_local_spread(i, priv->numa_node),
-			priv->irq_info[i].mask);
+			priv->irq_info[vecidx].mask);
 
 	if (IS_ENABLED(CONFIG_SMP) &&
-	    irq_set_affinity_hint(irq, priv->irq_info[i].mask))
+	    irq_set_affinity_hint(irq, priv->irq_info[vecidx].mask))
 		mlx5_core_warn(mdev, "irq_set_affinity_hint failed, irq 0x%.4x", irq);
 
 	return 0;
@@ -659,11 +660,12 @@ static int mlx5_irq_set_affinity_hint(struct mlx5_core_dev *mdev, int i)
 
 static void mlx5_irq_clear_affinity_hint(struct mlx5_core_dev *mdev, int i)
 {
+	int vecidx = MLX5_EQ_VEC_COMP_BASE + i;
 	struct mlx5_priv *priv  = &mdev->priv;
-	int irq = pci_irq_vector(mdev->pdev, MLX5_EQ_VEC_COMP_BASE + i);
+	int irq = pci_irq_vector(mdev->pdev, vecidx);
 
 	irq_set_affinity_hint(irq, NULL);
-	free_cpumask_var(priv->irq_info[i].mask);
+	free_cpumask_var(priv->irq_info[vecidx].mask);
 }
 
 static int mlx5_irq_set_affinity_hints(struct mlx5_core_dev *mdev)
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index aa5963b5d38e..7d4ed995b4ce 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -1309,7 +1309,7 @@ enum {
 static inline const struct cpumask *
 mlx5_get_vector_affinity_hint(struct mlx5_core_dev *dev, int vector)
 {
-	return dev->priv.irq_info[vector].mask;
+	return dev->priv.irq_info[vector + MLX5_EQ_VEC_COMP_BASE].mask;
 }
 
 #endif /* MLX5_DRIVER_H */
-- 
2.19.1

[PATCH AUTOSEL 4.20 070/304] RDMA/core: Sync unregistration with netlink commands

[Sasha Levin]

From: Parav Pandit <parav@mellanox.com>

[ Upstream commit 01b671170d7f82b959dad6d5dbb44d7a915e647d ]

When the rdma device is getting removed, get resource info can race with
device removal, as below:

      CPU-0                                  CPU-1
    --------                               --------
    rdma_nl_rcv_msg()
       nldev_res_get_cq_dumpit()
          mutex_lock(device_lock);
          get device reference
          mutex_unlock(device_lock);        [..]
                                            ib_unregister_device()
                                            /* Valid reference to
                                             * device->dev exists.
                                             */
                                             ib_dealloc_device()

          [..]
          provider->fill_res_entry();

Even though device object is not freed, fill_res_entry() can get called on
device which doesn't have a driver anymore. Kernel core device reference
count is not sufficient, as this only keeps the structure valid, and
doesn't guarantee the driver is still loaded.

Similar race can occur with device renaming and device removal, where
device_rename() tries to rename a unregistered device. While this is fine
for devices of a class which are not net namespace aware, but it is
incorrect for net namespace aware class coming in subsequent series.  If a
class is net namespace aware, then the below [1] call trace is observed in
above situation.

Therefore, to avoid the race, keep a reference count and let device
unregistration wait until all netlink users drop the reference.

[1] Call trace:
kernfs: ns required in 'infiniband' for 'mlx5_0'
WARNING: CPU: 18 PID: 44270 at fs/kernfs/dir.c:842 kernfs_find_ns+0x104/0x120
libahci i2c_core mlxfw libata dca [last unloaded: devlink]
RIP: 0010:kernfs_find_ns+0x104/0x120
Call Trace:
kernfs_find_and_get_ns+0x2e/0x50
sysfs_rename_link_ns+0x40/0xb0
device_rename+0xb2/0xf0
ib_device_rename+0xb3/0x100 [ib_core]
nldev_set_doit+0x165/0x190 [ib_core]
rdma_nl_rcv_msg+0x249/0x250 [ib_core]
? netlink_deliver_tap+0x8f/0x3e0
rdma_nl_rcv+0xd6/0x120 [ib_core]
netlink_unicast+0x17c/0x230
netlink_sendmsg+0x2f0/0x3e0
sock_sendmsg+0x30/0x40
__sys_sendto+0xdc/0x160

Fixes: da5c85078215 ("RDMA/nldev: add driver-specific resource tracking")
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/infiniband/core/core_priv.h |  1 +
 drivers/infiniband/core/device.c    | 26 ++++++++++++++++++++++----
 drivers/infiniband/core/nldev.c     | 20 ++++++++++----------
 include/rdma/ib_verbs.h             |  8 +++++++-
 4 files changed, 40 insertions(+), 15 deletions(-)

diff --git a/drivers/infiniband/core/core_priv.h b/drivers/infiniband/core/core_priv.h
index bb9007a0cca7..d97d39a7537c 100644
--- a/drivers/infiniband/core/core_priv.h
+++ b/drivers/infiniband/core/core_priv.h
@@ -296,6 +296,7 @@ static inline int ib_mad_enforce_security(struct ib_mad_agent_private *map,
 #endif
 
 struct ib_device *ib_device_get_by_index(u32 ifindex);
+void ib_device_put(struct ib_device *device);
 /* RDMA device netlink */
 void nldev_init(void);
 void nldev_exit(void);
diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c
index 87eb4f2cdd7d..0027b0d79b09 100644
--- a/drivers/infiniband/core/device.c
+++ b/drivers/infiniband/core/device.c
@@ -145,7 +145,8 @@ static struct ib_device *__ib_device_get_by_index(u32 index)
 }
 
 /*
- * Caller is responsible to return refrerence count by calling put_device()
+ * Caller must perform ib_device_put() to return the device reference count
+ * when ib_device_get_by_index() returns valid device pointer.
  */
 struct ib_device *ib_device_get_by_index(u32 index)
 {
@@ -153,13 +154,21 @@ struct ib_device *ib_device_get_by_index(u32 index)
 
 	down_read(&lists_rwsem);
 	device = __ib_device_get_by_index(index);
-	if (device)
-		get_device(&device->dev);
-
+	if (device) {
+		/* Do not return a device if unregistration has started. */
+		if (!refcount_inc_not_zero(&device->refcount))
+			device = NULL;
+	}
 	up_read(&lists_rwsem);
 	return device;
 }
 
+void ib_device_put(struct ib_device *device)
+{
+	if (refcount_dec_and_test(&device->refcount))
+		complete(&device->unreg_completion);
+}
+
 static struct ib_device *__ib_device_get_by_name(const char *name)
 {
 	struct ib_device *device;
@@ -293,6 +302,8 @@ struct ib_device *ib_alloc_device(size_t size)
 	rwlock_init(&device->client_data_lock);
 	INIT_LIST_HEAD(&device->client_data_list);
 	INIT_LIST_HEAD(&device->port_list);
+	refcount_set(&device->refcount, 1);
+	init_completion(&device->unreg_completion);
 
 	return device;
 }
@@ -641,6 +652,13 @@ void ib_unregister_device(struct ib_device *device)
 	struct ib_client_data *context, *tmp;
 	unsigned long flags;
 
+	/*
+	 * Wait for all netlink command callers to finish working on the
+	 * device.
+	 */
+	ib_device_put(device);
+	wait_for_completion(&device->unreg_completion);
+
 	mutex_lock(&device_mutex);
 
 	down_write(&lists_rwsem);
diff --git a/drivers/infiniband/core/nldev.c b/drivers/infiniband/core/nldev.c
index ff6468e7fe79..77a0f1e1576f 100644
--- a/drivers/infiniband/core/nldev.c
+++ b/drivers/infiniband/core/nldev.c
@@ -632,13 +632,13 @@ static int nldev_get_doit(struct sk_buff *skb, struct nlmsghdr *nlh,
 
 	nlmsg_end(msg, nlh);
 
-	put_device(&device->dev);
+	ib_device_put(device);
 	return rdma_nl_unicast(msg, NETLINK_CB(skb).portid);
 
 err_free:
 	nlmsg_free(msg);
 err:
-	put_device(&device->dev);
+	ib_device_put(device);
 	return err;
 }
 
@@ -668,7 +668,7 @@ static int nldev_set_doit(struct sk_buff *skb, struct nlmsghdr *nlh,
 		err = ib_device_rename(device, name);
 	}
 
-	put_device(&device->dev);
+	ib_device_put(device);
 	return err;
 }
 
@@ -752,14 +752,14 @@ static int nldev_port_get_doit(struct sk_buff *skb, struct nlmsghdr *nlh,
 		goto err_free;
 
 	nlmsg_end(msg, nlh);
-	put_device(&device->dev);
+	ib_device_put(device);
 
 	return rdma_nl_unicast(msg, NETLINK_CB(skb).portid);
 
 err_free:
 	nlmsg_free(msg);
 err:
-	put_device(&device->dev);
+	ib_device_put(device);
 	return err;
 }
 
@@ -816,7 +816,7 @@ static int nldev_port_get_dumpit(struct sk_buff *skb,
 	}
 
 out:
-	put_device(&device->dev);
+	ib_device_put(device);
 	cb->args[0] = idx;
 	return skb->len;
 }
@@ -855,13 +855,13 @@ static int nldev_res_get_doit(struct sk_buff *skb, struct nlmsghdr *nlh,
 		goto err_free;
 
 	nlmsg_end(msg, nlh);
-	put_device(&device->dev);
+	ib_device_put(device);
 	return rdma_nl_unicast(msg, NETLINK_CB(skb).portid);
 
 err_free:
 	nlmsg_free(msg);
 err:
-	put_device(&device->dev);
+	ib_device_put(device);
 	return ret;
 }
 
@@ -1054,7 +1054,7 @@ next:		idx++;
 	if (!filled)
 		goto err;
 
-	put_device(&device->dev);
+	ib_device_put(device);
 	return skb->len;
 
 res_err:
@@ -1065,7 +1065,7 @@ err:
 	nlmsg_cancel(skb, nlh);
 
 err_index:
-	put_device(&device->dev);
+	ib_device_put(device);
 	return ret;
 }
 
diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
index 9c0c2132a2d6..64626b32107b 100644
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@@ -56,7 +56,7 @@
 #include <linux/string.h>
 #include <linux/slab.h>
 #include <linux/netdevice.h>
-
+#include <linux/refcount.h>
 #include <linux/if_link.h>
 #include <linux/atomic.h>
 #include <linux/mmu_notifier.h>
@@ -2605,6 +2605,12 @@ struct ib_device {
 
 	const struct uverbs_object_tree_def *const *driver_specs;
 	enum rdma_driver_id		driver_id;
+	/*
+	 * Provides synchronization between device unregistration and netlink
+	 * commands on a device. To be used only by core.
+	 */
+	refcount_t refcount;
+	struct completion unreg_completion;
 };
 
 struct ib_client {
-- 
2.19.1

[PATCH AUTOSEL 4.20 134/304] IB/hfi1: Unreserve a reserved request when it is completed

[Sasha Levin]

From: Kaike Wan <kaike.wan@intel.com>

[ Upstream commit ca95f802ef5139722acc8d30aeaab6fe5bbe939e ]

Currently, When a reserved operation is completed, its entry in the send
queue will not be unreserved, which leads to the miscalculation of
qp->s_avail and thus the triggering of a WARN_ON call trace. This patch
fixes the problem by unreserving the reserved operation when it is
completed.

Fixes: 856cc4c237ad ("IB/hfi1: Add the capability for reserved operations")
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Kaike Wan <kaike.wan@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/infiniband/hw/hfi1/rc.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/infiniband/hw/hfi1/rc.c b/drivers/infiniband/hw/hfi1/rc.c
index 188aa4f686a0..ea3aac264df9 100644
--- a/drivers/infiniband/hw/hfi1/rc.c
+++ b/drivers/infiniband/hw/hfi1/rc.c
@@ -1157,6 +1157,7 @@ void hfi1_rc_send_complete(struct rvt_qp *qp, struct hfi1_opa_header *opah)
 		if (cmp_psn(wqe->lpsn, qp->s_sending_psn) >= 0 &&
 		    cmp_psn(qp->s_sending_psn, qp->s_sending_hpsn) <= 0)
 			break;
+		rvt_qp_wqe_unreserve(qp, wqe);
 		s_last = qp->s_last;
 		trace_hfi1_qp_send_completion(qp, wqe, s_last);
 		if (++s_last >= qp->s_size)
@@ -1209,6 +1210,7 @@ static struct rvt_swqe *do_rc_completion(struct rvt_qp *qp,
 		u32 s_last;
 
 		rvt_put_swqe(wqe);
+		rvt_qp_wqe_unreserve(qp, wqe);
 		s_last = qp->s_last;
 		trace_hfi1_qp_send_completion(qp, wqe, s_last);
 		if (++s_last >= qp->s_size)
-- 
2.19.1

[PATCH AUTOSEL 4.20 236/304] mlx5: update timecounter at least twice per counter overflow

[Sasha Levin]

From: Miroslav Lichvar <mlichvar@redhat.com>

[ Upstream commit 5d8678365c90b9ce1fd2243ff5ea562609f6cec1 ]

The timecounter needs to be updated at least once in half of the
cyclecounter interval to prevent timecounter_cyc2time() interpreting a
new timestamp as an old value and causing a backward jump.

This would be an issue if the timecounter multiplier was so small that
the update interval would not be limited by the 64-bit overflow in
multiplication.

Shorten the calculated interval to make sure the timecounter is updated
in time even when the system clock is slowed down by up to 10%, the
multiplier is increased by up to 10%, and the scheduled overflow check
is late by 15%.

Cc: Richard Cochran <richardcochran@gmail.com>
Cc: Ariel Levkovich <lariel@mellanox.com>
Cc: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Miroslav Lichvar <mlichvar@redhat.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
index 0d90b1b4a3d3..2d6168ee99e8 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
@@ -511,14 +511,14 @@ void mlx5_init_clock(struct mlx5_core_dev *mdev)
 			 ktime_to_ns(ktime_get_real()));
 
 	/* Calculate period in seconds to call the overflow watchdog - to make
-	 * sure counter is checked at least once every wrap around.
+	 * sure counter is checked at least twice every wrap around.
 	 * The period is calculated as the minimum between max HW cycles count
 	 * (The clock source mask) and max amount of cycles that can be
 	 * multiplied by clock multiplier where the result doesn't exceed
 	 * 64bits.
 	 */
 	overflow_cycles = div64_u64(~0ULL >> 1, clock->cycles.mult);
-	overflow_cycles = min(overflow_cycles, clock->cycles.mask >> 1);
+	overflow_cycles = min(overflow_cycles, div_u64(clock->cycles.mask, 3));
 
 	ns = cyclecounter_cyc2ns(&clock->cycles, overflow_cycles,
 				 frac, &frac);
-- 
2.19.1