Received-SPF: pass (google.com: domain of linux-kernel+bounces-63413-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:40f1:3f00::1 as permitted sender) client-ip=2604:1380:40f1:3f00::1;
Received-SPF: Pass (protection.outlook.com: domain of nvidia.com designates
 216.228.117.160 as permitted sender) receiver=protection.outlook.com;
 client-ip=216.228.117.160; helo=mail.nvidia.com; pr=C
From: Shravan Kumar Ramani <shravankr@nvidia.com>
To: Hans de Goede <hdegoede@redhat.com>, Ilpo Jarvinen
	<ilpo.jarvinen@linux.intel.com>, Vadim Pasternak <vadimp@nvidia.com>, "David
 Thompson" <davthompson@nvidia.com>
CC: Shravan Kumar Ramani <shravankr@nvidia.com>,
	<platform-driver-x86@vger.kernel.org>, <linux-kernel@vger.kernel.org>
Subject: [PATCH v3 3/4] platform/mellanox: mlxbf-pmc: Add support for 64-bit counters and cycle count
Date: Tue, 13 Feb 2024 06:15:27 -0500
Message-ID: <ce1140840b09df3fe2b85e3f87089f2e0d0877d0.1707808180.git.shravankr@nvidia.com>
In-Reply-To: <cover.1707808180.git.shravankr@nvidia.com>
References: <cover.1707808180.git.shravankr@nvidia.com>
Precedence: bulk
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Content-Type: text/plain
X-MS-Exchange-CrossTenant-OriginalArrivalTime: 13 Feb 2024 11:15:57.6391
 (UTC)
X-MS-Exchange-CrossTenant-Network-Message-Id: 7bf02be4-92ce-49d8-60b0-08dc2c852705
X-MS-Exchange-CrossTenant-Id: 43083d15-7273-40c1-b7db-39efd9ccc17a
X-MS-Exchange-CrossTenant-OriginalAttributedTenantConnectingIp: TenantId=43083d15-7273-40c1-b7db-39efd9ccc17a;Ip=[216.228.117.160];Helo=[mail.nvidia.com]
X-MS-Exchange-CrossTenant-AuthSource:
	SJ5PEPF000001CA.namprd05.prod.outlook.com
X-MS-Exchange-CrossTenant-AuthAs: Anonymous
X-MS-Exchange-CrossTenant-FromEntityHeader: HybridOnPrem
X-MS-Exchange-Transport-CrossTenantHeadersStamped: SA1PR12MB9247

Add support for programming any counter to monitor the cycle count.
Since counting of cycles using 32-bit ocunters would result in frequent
wraparounds, add the ability to combine 2 adjacent 32-bit counters to
form 1 64-bit counter.
Both these features are supported by BlueField-3 PMC hardware, hence
the required bit-fields are exposed by the driver via sysfs to allow
the user to configure as needed.

Signed-off-by: Shravan Kumar Ramani <shravankr@nvidia.com>
Reviewed-by: David Thompson <davthompson@nvidia.com>
Reviewed-by: Vadim Pasternak <vadimp@nvidia.com>
---
 drivers/platform/mellanox/mlxbf-pmc.c | 134 ++++++++++++++++++++++++++
 1 file changed, 134 insertions(+)

diff --git a/drivers/platform/mellanox/mlxbf-pmc.c b/drivers/platform/mellanox/mlxbf-pmc.c
index 250405bb59a7..e2f11c0c63e9 100644
--- a/drivers/platform/mellanox/mlxbf-pmc.c
+++ b/drivers/platform/mellanox/mlxbf-pmc.c
@@ -88,6 +88,8 @@
 #define MLXBF_PMC_CRSPACE_PERFMON_CTL(n) (n * MLXBF_PMC_CRSPACE_PERFMON_REG0_SZ)
 #define MLXBF_PMC_CRSPACE_PERFMON_EN BIT(30)
 #define MLXBF_PMC_CRSPACE_PERFMON_CLR BIT(28)
+#define MLXBF_PMC_CRSPACE_PERFMON_UOC GENMASK(15, 0)
+#define MLXBF_PMC_CRSPACE_PERFMON_COUNT_CLOCK(n) (MLXBF_PMC_CRSPACE_PERFMON_CTL(n) + 0x4)
 #define MLXBF_PMC_CRSPACE_PERFMON_VAL0(n) (MLXBF_PMC_CRSPACE_PERFMON_CTL(n) + 0xc)
 
 /**
@@ -114,6 +116,8 @@ struct mlxbf_pmc_attribute {
  * @attr_event: Attributes for "event" sysfs files
  * @attr_event_list: Attributes for "event_list" sysfs files
  * @attr_enable: Attributes for "enable" sysfs files
+ * @attr_use_odd_counter: Attributes for "use_odd_counter" sysfs files
+ * @attr_count_clock: Attributes for "count_clock" sysfs files
  * @block_attr: All attributes needed for the block
  * @block_attr_grp: Attribute group for the block
  */
@@ -126,6 +130,8 @@ struct mlxbf_pmc_block_info {
 	struct mlxbf_pmc_attribute *attr_event;
 	struct mlxbf_pmc_attribute attr_event_list;
 	struct mlxbf_pmc_attribute attr_enable;
+	struct mlxbf_pmc_attribute attr_use_odd_counter;
+	struct mlxbf_pmc_attribute attr_count_clock;
 	struct attribute *block_attr[MLXBF_PMC_MAX_ATTRS];
 	struct attribute_group block_attr_grp;
 };
@@ -1751,6 +1757,103 @@ static ssize_t mlxbf_pmc_enable_store(struct device *dev,
 	return count;
 }
 
+/* Show function for "use_odd_counter" sysfs files - only for crspace */
+static ssize_t mlxbf_pmc_use_odd_counter_show(struct device *dev,
+					      struct device_attribute *attr, char *buf)
+{
+	struct mlxbf_pmc_attribute *attr_use_odd_counter = container_of(
+		attr, struct mlxbf_pmc_attribute, dev_attr);
+	unsigned int blk_num;
+	u32 value, reg;
+
+	blk_num = attr_use_odd_counter->nr;
+
+	if (mlxbf_pmc_readl(pmc->block[blk_num].mmio_base +
+			MLXBF_PMC_CRSPACE_PERFMON_CTL(pmc->block[blk_num].counters),
+			&reg))
+		return -EINVAL;
+
+	value = FIELD_GET(MLXBF_PMC_CRSPACE_PERFMON_UOC, reg);
+
+	return sysfs_emit(buf, "%u\n", value);
+}
+
+/* Store function for "use_odd_counter" sysfs files - only for crspace */
+static ssize_t mlxbf_pmc_use_odd_counter_store(struct device *dev,
+					       struct device_attribute *attr,
+					       const char *buf, size_t count)
+{
+	struct mlxbf_pmc_attribute *attr_use_odd_counter = container_of(
+		attr, struct mlxbf_pmc_attribute, dev_attr);
+	unsigned int blk_num;
+	u32 uoc, reg;
+	int err;
+
+	blk_num = attr_use_odd_counter->nr;
+
+	err = kstrtouint(buf, 0, &uoc);
+	if (err < 0)
+		return err;
+
+	err = mlxbf_pmc_readl(pmc->block[blk_num].mmio_base +
+		MLXBF_PMC_CRSPACE_PERFMON_CTL(pmc->block[blk_num].counters),
+		&reg);
+	if (err)
+		return -EINVAL;
+
+	reg &= ~MLXBF_PMC_CRSPACE_PERFMON_UOC;
+	reg |= FIELD_PREP(MLXBF_PMC_CRSPACE_PERFMON_UOC, uoc);
+
+	mlxbf_pmc_write(pmc->block[blk_num].mmio_base +
+		MLXBF_PMC_CRSPACE_PERFMON_CTL(pmc->block[blk_num].counters),
+		MLXBF_PMC_WRITE_REG_32, reg);
+
+	return count;
+}
+
+/* Show function for "count_clock" sysfs files - only for crspace */
+static ssize_t mlxbf_pmc_count_clock_show(struct device *dev,
+					  struct device_attribute *attr, char *buf)
+{
+	struct mlxbf_pmc_attribute *attr_count_clock = container_of(
+		attr, struct mlxbf_pmc_attribute, dev_attr);
+	unsigned int blk_num;
+	u32 reg;
+
+	blk_num = attr_count_clock->nr;
+
+	if (mlxbf_pmc_readl(pmc->block[blk_num].mmio_base +
+			MLXBF_PMC_CRSPACE_PERFMON_COUNT_CLOCK(pmc->block[blk_num].counters),
+			&reg))
+		return -EINVAL;
+
+	return sysfs_emit(buf, "%u\n", reg);
+}
+
+/* Store function for "count_clock" sysfs files - only for crspace */
+static ssize_t mlxbf_pmc_count_clock_store(struct device *dev,
+					   struct device_attribute *attr,
+					   const char *buf, size_t count)
+{
+	struct mlxbf_pmc_attribute *attr_count_clock = container_of(
+		attr, struct mlxbf_pmc_attribute, dev_attr);
+	unsigned int blk_num;
+	u32 reg;
+	int err;
+
+	blk_num = attr_count_clock->nr;
+
+	err = kstrtouint(buf, 0, &reg);
+	if (err < 0)
+		return err;
+
+	mlxbf_pmc_write(pmc->block[blk_num].mmio_base +
+		MLXBF_PMC_CRSPACE_PERFMON_COUNT_CLOCK(pmc->block[blk_num].counters),
+		MLXBF_PMC_WRITE_REG_32, reg);
+
+	return count;
+}
+
 /* Populate attributes for blocks with counters to monitor performance */
 static int mlxbf_pmc_init_perftype_counter(struct device *dev, unsigned int blk_num)
 {
@@ -1784,6 +1887,37 @@ static int mlxbf_pmc_init_perftype_counter(struct device *dev, unsigned int blk_
 		attr = NULL;
 	}
 
+	if (pmc->block[blk_num].type == MLXBF_PMC_TYPE_CRSPACE) {
+		/*
+		 * Couple adjacent odd and even 32-bit counters to form 64-bit counters
+		 * using "use_odd_counter" sysfs which has one bit per even counter.
+		 */
+		attr = &pmc->block[blk_num].attr_use_odd_counter;
+		attr->dev_attr.attr.mode = 0644;
+		attr->dev_attr.show = mlxbf_pmc_use_odd_counter_show;
+		attr->dev_attr.store = mlxbf_pmc_use_odd_counter_store;
+		attr->nr = blk_num;
+		attr->dev_attr.attr.name = devm_kasprintf(dev, GFP_KERNEL,
+							  "use_odd_counter");
+		if (!attr->dev_attr.attr.name)
+			return -ENOMEM;
+		pmc->block[blk_num].block_attr[++i] = &attr->dev_attr.attr;
+		attr = NULL;
+
+		/* Program crspace counters to count clock cycles using "count_clock" sysfs */
+		attr = &pmc->block[blk_num].attr_count_clock;
+		attr->dev_attr.attr.mode = 0644;
+		attr->dev_attr.show = mlxbf_pmc_count_clock_show;
+		attr->dev_attr.store = mlxbf_pmc_count_clock_store;
+		attr->nr = blk_num;
+		attr->dev_attr.attr.name = devm_kasprintf(dev, GFP_KERNEL,
+							  "count_clock");
+		if (!attr->dev_attr.attr.name)
+			return -ENOMEM;
+		pmc->block[blk_num].block_attr[++i] = &attr->dev_attr.attr;
+		attr = NULL;
+	}
+
 	pmc->block[blk_num].attr_counter = devm_kcalloc(
 		dev, pmc->block[blk_num].counters,
 		sizeof(struct mlxbf_pmc_attribute), GFP_KERNEL);
-- 
2.30.1