lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20231127202704.1263376-10-tom.zanussi@linux.intel.com>
Date:   Mon, 27 Nov 2023 14:26:59 -0600
From:   Tom Zanussi <tom.zanussi@...ux.intel.com>
To:     herbert@...dor.apana.org.au, davem@...emloft.net,
        fenghua.yu@...el.com, vkoul@...nel.org
Cc:     dave.jiang@...el.com, tony.luck@...el.com,
        wajdi.k.feghali@...el.com, james.guilford@...el.com,
        kanchana.p.sridhar@...el.com, vinodh.gopal@...el.com,
        giovanni.cabiddu@...el.com, pavel@....cz,
        linux-kernel@...r.kernel.org, linux-crypto@...r.kernel.org,
        dmaengine@...r.kernel.org
Subject: [PATCH v10 09/14] crypto: iaa - Add per-cpu workqueue table with rebalancing

The iaa compression/decompression algorithms in later patches need a
way to retrieve an appropriate IAA workqueue depending on how close
the associated IAA device is to the current cpu.

For this purpose, add a per-cpu array of workqueues such that an
appropriate workqueue can be retrieved by simply accessing the per-cpu
array.

Whenever a new workqueue is bound to or unbound from the iaa_crypto
driver, the available workqueues are 'rebalanced' such that work
submitted from a particular CPU is given to the most appropriate
workqueue available.  There currently isn't any way for the user to
tweak the way this is done internally - if necessary, knobs can be
added later for that purpose.  Current best practice is to configure
and bind at least one workqueue for each IAA device, but as long as
there is at least one workqueue configured and bound to any IAA device
in the system, the iaa_crypto driver will work, albeit most likely not
as efficiently.

[ Based on work originally by George Powley, Jing Lin and Kyung Min
Park ]

Signed-off-by: Tom Zanussi <tom.zanussi@...ux.intel.com>
---
 drivers/crypto/intel/iaa/iaa_crypto.h      |   7 +
 drivers/crypto/intel/iaa/iaa_crypto_main.c | 222 +++++++++++++++++++++
 2 files changed, 229 insertions(+)

diff --git a/drivers/crypto/intel/iaa/iaa_crypto.h b/drivers/crypto/intel/iaa/iaa_crypto.h
index 5d1fff7f4b8e..c25546fa87f7 100644
--- a/drivers/crypto/intel/iaa/iaa_crypto.h
+++ b/drivers/crypto/intel/iaa/iaa_crypto.h
@@ -27,4 +27,11 @@ struct iaa_device {
 	struct list_head		wqs;
 };
 
+struct wq_table_entry {
+	struct idxd_wq **wqs;
+	int	max_wqs;
+	int	n_wqs;
+	int	cur_wq;
+};
+
 #endif
diff --git a/drivers/crypto/intel/iaa/iaa_crypto_main.c b/drivers/crypto/intel/iaa/iaa_crypto_main.c
index 9a662ae49106..925b8fea0d06 100644
--- a/drivers/crypto/intel/iaa/iaa_crypto_main.c
+++ b/drivers/crypto/intel/iaa/iaa_crypto_main.c
@@ -22,6 +22,46 @@
 
 /* number of iaa instances probed */
 static unsigned int nr_iaa;
+static unsigned int nr_cpus;
+static unsigned int nr_nodes;
+static unsigned int nr_cpus_per_node;
+
+/* Number of physical cpus sharing each iaa instance */
+static unsigned int cpus_per_iaa;
+
+/* Per-cpu lookup table for balanced wqs */
+static struct wq_table_entry __percpu *wq_table;
+
+static void wq_table_add(int cpu, struct idxd_wq *wq)
+{
+	struct wq_table_entry *entry = per_cpu_ptr(wq_table, cpu);
+
+	if (WARN_ON(entry->n_wqs == entry->max_wqs))
+		return;
+
+	entry->wqs[entry->n_wqs++] = wq;
+
+	pr_debug("%s: added iaa wq %d.%d to idx %d of cpu %d\n", __func__,
+		 entry->wqs[entry->n_wqs - 1]->idxd->id,
+		 entry->wqs[entry->n_wqs - 1]->id, entry->n_wqs - 1, cpu);
+}
+
+static void wq_table_free_entry(int cpu)
+{
+	struct wq_table_entry *entry = per_cpu_ptr(wq_table, cpu);
+
+	kfree(entry->wqs);
+	memset(entry, 0, sizeof(*entry));
+}
+
+static void wq_table_clear_entry(int cpu)
+{
+	struct wq_table_entry *entry = per_cpu_ptr(wq_table, cpu);
+
+	entry->n_wqs = 0;
+	entry->cur_wq = 0;
+	memset(entry->wqs, 0, entry->max_wqs * sizeof(struct idxd_wq *));
+}
 
 static LIST_HEAD(iaa_devices);
 static DEFINE_MUTEX(iaa_devices_lock);
@@ -141,6 +181,53 @@ static void del_iaa_wq(struct iaa_device *iaa_device, struct idxd_wq *wq)
 	}
 }
 
+static void clear_wq_table(void)
+{
+	int cpu;
+
+	for (cpu = 0; cpu < nr_cpus; cpu++)
+		wq_table_clear_entry(cpu);
+
+	pr_debug("cleared wq table\n");
+}
+
+static void free_wq_table(void)
+{
+	int cpu;
+
+	for (cpu = 0; cpu < nr_cpus; cpu++)
+		wq_table_free_entry(cpu);
+
+	free_percpu(wq_table);
+
+	pr_debug("freed wq table\n");
+}
+
+static int alloc_wq_table(int max_wqs)
+{
+	struct wq_table_entry *entry;
+	int cpu;
+
+	wq_table = alloc_percpu(struct wq_table_entry);
+	if (!wq_table)
+		return -ENOMEM;
+
+	for (cpu = 0; cpu < nr_cpus; cpu++) {
+		entry = per_cpu_ptr(wq_table, cpu);
+		entry->wqs = kcalloc(max_wqs, sizeof(struct wq *), GFP_KERNEL);
+		if (!entry->wqs) {
+			free_wq_table();
+			return -ENOMEM;
+		}
+
+		entry->max_wqs = max_wqs;
+	}
+
+	pr_debug("initialized wq table\n");
+
+	return 0;
+}
+
 static int save_iaa_wq(struct idxd_wq *wq)
 {
 	struct iaa_device *iaa_device, *found = NULL;
@@ -193,6 +280,8 @@ static int save_iaa_wq(struct idxd_wq *wq)
 
 	if (WARN_ON(nr_iaa == 0))
 		return -EINVAL;
+
+	cpus_per_iaa = (nr_nodes * nr_cpus_per_node) / nr_iaa;
 out:
 	return 0;
 }
@@ -207,6 +296,116 @@ static void remove_iaa_wq(struct idxd_wq *wq)
 			break;
 		}
 	}
+
+	if (nr_iaa)
+		cpus_per_iaa = (nr_nodes * nr_cpus_per_node) / nr_iaa;
+	else
+		cpus_per_iaa = 0;
+}
+
+static int wq_table_add_wqs(int iaa, int cpu)
+{
+	struct iaa_device *iaa_device, *found_device = NULL;
+	int ret = 0, cur_iaa = 0, n_wqs_added = 0;
+	struct idxd_device *idxd;
+	struct iaa_wq *iaa_wq;
+	struct pci_dev *pdev;
+	struct device *dev;
+
+	list_for_each_entry(iaa_device, &iaa_devices, list) {
+		idxd = iaa_device->idxd;
+		pdev = idxd->pdev;
+		dev = &pdev->dev;
+
+		if (cur_iaa != iaa) {
+			cur_iaa++;
+			continue;
+		}
+
+		found_device = iaa_device;
+		dev_dbg(dev, "getting wq from iaa_device %d, cur_iaa %d\n",
+			found_device->idxd->id, cur_iaa);
+		break;
+	}
+
+	if (!found_device) {
+		found_device = list_first_entry_or_null(&iaa_devices,
+							struct iaa_device, list);
+		if (!found_device) {
+			pr_debug("couldn't find any iaa devices with wqs!\n");
+			ret = -EINVAL;
+			goto out;
+		}
+		cur_iaa = 0;
+
+		idxd = found_device->idxd;
+		pdev = idxd->pdev;
+		dev = &pdev->dev;
+		dev_dbg(dev, "getting wq from only iaa_device %d, cur_iaa %d\n",
+			found_device->idxd->id, cur_iaa);
+	}
+
+	list_for_each_entry(iaa_wq, &found_device->wqs, list) {
+		wq_table_add(cpu, iaa_wq->wq);
+		pr_debug("rebalance: added wq for cpu=%d: iaa wq %d.%d\n",
+			 cpu, iaa_wq->wq->idxd->id, iaa_wq->wq->id);
+		n_wqs_added++;
+	};
+
+	if (!n_wqs_added) {
+		pr_debug("couldn't find any iaa wqs!\n");
+		ret = -EINVAL;
+		goto out;
+	}
+out:
+	return ret;
+}
+
+/*
+ * Rebalance the wq table so that given a cpu, it's easy to find the
+ * closest IAA instance.  The idea is to try to choose the most
+ * appropriate IAA instance for a caller and spread available
+ * workqueues around to clients.
+ */
+static void rebalance_wq_table(void)
+{
+	const struct cpumask *node_cpus;
+	int node, cpu, iaa = -1;
+
+	if (nr_iaa == 0)
+		return;
+
+	pr_debug("rebalance: nr_nodes=%d, nr_cpus %d, nr_iaa %d, cpus_per_iaa %d\n",
+		 nr_nodes, nr_cpus, nr_iaa, cpus_per_iaa);
+
+	clear_wq_table();
+
+	if (nr_iaa == 1) {
+		for (cpu = 0; cpu < nr_cpus; cpu++) {
+			if (WARN_ON(wq_table_add_wqs(0, cpu))) {
+				pr_debug("could not add any wqs for iaa 0 to cpu %d!\n", cpu);
+				return;
+			}
+		}
+
+		return;
+	}
+
+	for_each_online_node(node) {
+		node_cpus = cpumask_of_node(node);
+
+		for (cpu = 0; cpu < nr_cpus_per_node; cpu++) {
+			int node_cpu = cpumask_nth(cpu, node_cpus);
+
+			if ((cpu % cpus_per_iaa) == 0)
+				iaa++;
+
+			if (WARN_ON(wq_table_add_wqs(iaa, node_cpu))) {
+				pr_debug("could not add any wqs for iaa %d to cpu %d!\n", iaa, cpu);
+				return;
+			}
+		}
+	}
 }
 
 static int iaa_crypto_probe(struct idxd_dev *idxd_dev)
@@ -215,6 +414,7 @@ static int iaa_crypto_probe(struct idxd_dev *idxd_dev)
 	struct idxd_device *idxd = wq->idxd;
 	struct idxd_driver_data *data = idxd->data;
 	struct device *dev = &idxd_dev->conf_dev;
+	bool first_wq = false;
 	int ret = 0;
 
 	if (idxd->state != IDXD_DEV_ENABLED)
@@ -245,10 +445,19 @@ static int iaa_crypto_probe(struct idxd_dev *idxd_dev)
 
 	mutex_lock(&iaa_devices_lock);
 
+	if (list_empty(&iaa_devices)) {
+		ret = alloc_wq_table(wq->idxd->max_wqs);
+		if (ret)
+			goto err_alloc;
+		first_wq = true;
+	}
+
 	ret = save_iaa_wq(wq);
 	if (ret)
 		goto err_save;
 
+	rebalance_wq_table();
+
 	mutex_unlock(&iaa_devices_lock);
 out:
 	mutex_unlock(&wq->wq_lock);
@@ -256,6 +465,10 @@ static int iaa_crypto_probe(struct idxd_dev *idxd_dev)
 	return ret;
 
 err_save:
+	if (first_wq)
+		free_wq_table();
+err_alloc:
+	mutex_unlock(&iaa_devices_lock);
 	idxd_drv_disable_wq(wq);
 err:
 	wq->type = IDXD_WQT_NONE;
@@ -273,7 +486,12 @@ static void iaa_crypto_remove(struct idxd_dev *idxd_dev)
 	mutex_lock(&iaa_devices_lock);
 
 	remove_iaa_wq(wq);
+
 	idxd_drv_disable_wq(wq);
+	rebalance_wq_table();
+
+	if (nr_iaa == 0)
+		free_wq_table();
 
 	mutex_unlock(&iaa_devices_lock);
 	mutex_unlock(&wq->wq_lock);
@@ -295,6 +513,10 @@ static int __init iaa_crypto_init_module(void)
 {
 	int ret = 0;
 
+	nr_cpus = num_online_cpus();
+	nr_nodes = num_online_nodes();
+	nr_cpus_per_node = nr_cpus / nr_nodes;
+
 	ret = idxd_driver_register(&iaa_crypto_driver);
 	if (ret) {
 		pr_debug("IAA wq sub-driver registration failed\n");
-- 
2.34.1

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ