diff mbox series

[net-next,1/3] ibmvnic: Assign IRQ affinity hints to device queues

Message ID 20221110213218.28662-2-nnac123@linux.ibm.com (mailing list archive)
State Accepted
Commit 44fbc1b6e0e291df1d31f5080777ce66a72ef37a
Delegated to: Netdev Maintainers
Headers show
Series ibmvnic: Introduce affinity hint support | expand

Checks

Context Check Description
netdev/tree_selection success Clearly marked for net-next
netdev/fixes_present success Fixes tag not required for -next series
netdev/subject_prefix success Link
netdev/cover_letter success Series has a cover letter
netdev/patch_count success Link
netdev/header_inline success No static functions without inline keyword in header files
netdev/build_32bit success Errors and warnings before: 0 this patch: 0
netdev/cc_maintainers warning 9 maintainers not CCed: npiggin@gmail.com pabeni@redhat.com davem@davemloft.net danymadden@us.ibm.com christophe.leroy@csgroup.eu edumazet@google.com kuba@kernel.org linuxppc-dev@lists.ozlabs.org mpe@ellerman.id.au
netdev/build_clang success Errors and warnings before: 0 this patch: 0
netdev/module_param success Was 0 now: 0
netdev/verify_signedoff success Signed-off-by tag matches author and committer
netdev/check_selftest success No net selftest shell script
netdev/verify_fixes success No Fixes tag
netdev/build_allmodconfig_warn success Errors and warnings before: 0 this patch: 0
netdev/checkpatch warning WARNING: line length of 81 exceeds 80 columns
netdev/kdoc success Errors and warnings before: 0 this patch: 0
netdev/source_inline success Was 0 now: 0

Commit Message

Nick Child Nov. 10, 2022, 9:32 p.m. UTC
Assign affinity hints to ibmvnic device queue interrupts.
Affinity hints are assigned and removed during sub-crq init and
teardown, respectively. This update should improve latency if
utilized as interrupt lines and processing are more equally
distributed among CPU's. This implementation is based on the
virtio_net driver.

Signed-off-by: Thomas Falcon <tlfalcon@linux.ibm.com>
Signed-off-by: Dany Madden <drt@linux.ibm.com>
Signed-off-by: Nick Child <nnac123@linux.ibm.com>
Reviewed-by: Rick Lindsley <ricklind@linux.ibm.com>
Reviewed-by: Haren Myneni <haren@linux.ibm.com>
---
 drivers/net/ethernet/ibm/ibmvnic.c | 140 +++++++++++++++++++++++++++++
 drivers/net/ethernet/ibm/ibmvnic.h |   1 +
 2 files changed, 141 insertions(+)
diff mbox series

Patch

diff --git a/drivers/net/ethernet/ibm/ibmvnic.c b/drivers/net/ethernet/ibm/ibmvnic.c
index 9282381a438f..0c969bdaf94d 100644
--- a/drivers/net/ethernet/ibm/ibmvnic.c
+++ b/drivers/net/ethernet/ibm/ibmvnic.c
@@ -68,6 +68,7 @@ 
 #include <linux/workqueue.h>
 #include <linux/if_vlan.h>
 #include <linux/utsname.h>
+#include <linux/cpu.h>
 
 #include "ibmvnic.h"
 
@@ -171,6 +172,132 @@  static int send_version_xchg(struct ibmvnic_adapter *adapter)
 	return ibmvnic_send_crq(adapter, &crq);
 }
 
+static void ibmvnic_clean_queue_affinity(struct ibmvnic_adapter *adapter,
+					 struct ibmvnic_sub_crq_queue *queue)
+{
+	if (!(queue && queue->irq))
+		return;
+
+	cpumask_clear(queue->affinity_mask);
+
+	if (irq_set_affinity_and_hint(queue->irq, NULL))
+		netdev_warn(adapter->netdev,
+			    "%s: Clear affinity failed, queue addr = %p, IRQ = %d\n",
+			    __func__, queue, queue->irq);
+}
+
+static void ibmvnic_clean_affinity(struct ibmvnic_adapter *adapter)
+{
+	struct ibmvnic_sub_crq_queue **rxqs;
+	struct ibmvnic_sub_crq_queue **txqs;
+	int num_rxqs, num_txqs;
+	int rc, i;
+
+	rc = 0;
+	rxqs = adapter->rx_scrq;
+	txqs = adapter->tx_scrq;
+	num_txqs = adapter->num_active_tx_scrqs;
+	num_rxqs = adapter->num_active_rx_scrqs;
+
+	netdev_dbg(adapter->netdev, "%s: Cleaning irq affinity hints", __func__);
+	if (txqs) {
+		for (i = 0; i < num_txqs; i++)
+			ibmvnic_clean_queue_affinity(adapter, txqs[i]);
+	}
+	if (rxqs) {
+		for (i = 0; i < num_rxqs; i++)
+			ibmvnic_clean_queue_affinity(adapter, rxqs[i]);
+	}
+}
+
+static int ibmvnic_set_queue_affinity(struct ibmvnic_sub_crq_queue *queue,
+				      unsigned int *cpu, int *stragglers,
+				      int stride)
+{
+	cpumask_var_t mask;
+	int i;
+	int rc = 0;
+
+	if (!(queue && queue->irq))
+		return rc;
+
+	/* cpumask_var_t is either a pointer or array, allocation works here */
+	if (!zalloc_cpumask_var(&mask, GFP_KERNEL))
+		return -ENOMEM;
+
+	/* while we have extra cpu give one extra to this irq */
+	if (*stragglers) {
+		stride++;
+		(*stragglers)--;
+	}
+	/* atomic write is safer than writing bit by bit directly */
+	for (i = 0; i < stride; i++) {
+		cpumask_set_cpu(*cpu, mask);
+		*cpu = cpumask_next_wrap(*cpu, cpu_online_mask,
+					 nr_cpu_ids, false);
+	}
+	/* set queue affinity mask */
+	cpumask_copy(queue->affinity_mask, mask);
+	rc = irq_set_affinity_and_hint(queue->irq, queue->affinity_mask);
+	free_cpumask_var(mask);
+
+	return rc;
+}
+
+/* assumes cpu read lock is held */
+static void ibmvnic_set_affinity(struct ibmvnic_adapter *adapter)
+{
+	struct ibmvnic_sub_crq_queue **rxqs = adapter->rx_scrq;
+	struct ibmvnic_sub_crq_queue **txqs = adapter->tx_scrq;
+	struct ibmvnic_sub_crq_queue *queue;
+	int num_rxqs = adapter->num_active_rx_scrqs;
+	int num_txqs = adapter->num_active_tx_scrqs;
+	int total_queues, stride, stragglers, i;
+	unsigned int num_cpu, cpu;
+	int rc = 0;
+
+	netdev_dbg(adapter->netdev, "%s: Setting irq affinity hints", __func__);
+	if (!(adapter->rx_scrq && adapter->tx_scrq)) {
+		netdev_warn(adapter->netdev,
+			    "%s: Set affinity failed, queues not allocated\n",
+			    __func__);
+		return;
+	}
+
+	total_queues = num_rxqs + num_txqs;
+	num_cpu = num_online_cpus();
+	/* number of cpu's assigned per irq */
+	stride = max_t(int, num_cpu / total_queues, 1);
+	/* number of leftover cpu's */
+	stragglers = num_cpu >= total_queues ? num_cpu % total_queues : 0;
+	/* next available cpu to assign irq to */
+	cpu = cpumask_next(-1, cpu_online_mask);
+
+	for (i = 0; i < num_txqs; i++) {
+		queue = txqs[i];
+		rc = ibmvnic_set_queue_affinity(queue, &cpu, &stragglers,
+						stride);
+		if (rc)
+			goto out;
+	}
+
+	for (i = 0; i < num_rxqs; i++) {
+		queue = rxqs[i];
+		rc = ibmvnic_set_queue_affinity(queue, &cpu, &stragglers,
+						stride);
+		if (rc)
+			goto out;
+	}
+
+out:
+	if (rc) {
+		netdev_warn(adapter->netdev,
+			    "%s: Set affinity failed, queue addr = %p, IRQ = %d, rc = %d.\n",
+			    __func__, queue, queue->irq, rc);
+		ibmvnic_clean_affinity(adapter);
+	}
+}
+
 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
 			  unsigned long length, unsigned long *number,
 			  unsigned long *irq)
@@ -3626,6 +3753,8 @@  static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
 	if (!adapter->tx_scrq || !adapter->rx_scrq)
 		return -EINVAL;
 
+	ibmvnic_clean_affinity(adapter);
+
 	for (i = 0; i < adapter->req_tx_queues; i++) {
 		netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
 		rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
@@ -3675,6 +3804,7 @@  static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
 			 DMA_BIDIRECTIONAL);
 	free_pages((unsigned long)scrq->msgs, 2);
+	free_cpumask_var(scrq->affinity_mask);
 	kfree(scrq);
 }
 
@@ -3695,6 +3825,8 @@  static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
 		dev_warn(dev, "Couldn't allocate crq queue messages page\n");
 		goto zero_page_failed;
 	}
+	if (!zalloc_cpumask_var(&scrq->affinity_mask, GFP_KERNEL))
+		goto cpumask_alloc_failed;
 
 	scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
 					 DMA_BIDIRECTIONAL);
@@ -3747,6 +3879,8 @@  static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
 			 DMA_BIDIRECTIONAL);
 map_failed:
+	free_cpumask_var(scrq->affinity_mask);
+cpumask_alloc_failed:
 	free_pages((unsigned long)scrq->msgs, 2);
 zero_page_failed:
 	kfree(scrq);
@@ -3758,6 +3892,7 @@  static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
 {
 	int i;
 
+	ibmvnic_clean_affinity(adapter);
 	if (adapter->tx_scrq) {
 		for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
 			if (!adapter->tx_scrq[i])
@@ -4035,6 +4170,11 @@  static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
 			goto req_rx_irq_failed;
 		}
 	}
+
+	cpus_read_lock();
+	ibmvnic_set_affinity(adapter);
+	cpus_read_unlock();
+
 	return rc;
 
 req_rx_irq_failed:
diff --git a/drivers/net/ethernet/ibm/ibmvnic.h b/drivers/net/ethernet/ibm/ibmvnic.h
index e5c6ff3d0c47..6720fec1ae67 100644
--- a/drivers/net/ethernet/ibm/ibmvnic.h
+++ b/drivers/net/ethernet/ibm/ibmvnic.h
@@ -825,6 +825,7 @@  struct ibmvnic_sub_crq_queue {
 	atomic_t used;
 	char name[32];
 	u64 handle;
+	cpumask_var_t affinity_mask;
 } ____cacheline_aligned;
 
 struct ibmvnic_long_term_buff {