new file mode 100644
@@ -0,0 +1,10 @@
+What: /sys/kernel/debug/qat_<device>_<BDF>/qat/fw_counters
+Date: November 2023
+KernelVersion: 6.6
+Contact: qat-linux@intel.com
+Description: (RO) Read returns the number of requests sent to the FW and the number of responses
+ received from the FW for each Acceleration Engine
+ Reported firmware counters::
+
+ <N>: Number of requests sent from Acceleration Engine N to FW and responses
+ Acceleration Engine N received from FW
@@ -28,6 +28,7 @@ intel_qat-objs := adf_cfg.o \
qat_bl.o
intel_qat-$(CONFIG_DEBUG_FS) += adf_transport_debug.o \
+ adf_fw_counters.o \
adf_dbgfs.o
intel_qat-$(CONFIG_PCI_IOV) += adf_sriov.o adf_vf_isr.o adf_pfvf_utils.o \
@@ -292,6 +292,7 @@ struct adf_accel_dev {
unsigned long status;
atomic_t ref_count;
struct dentry *debugfs_dir;
+ struct dentry *fw_cntr_dbgfile;
struct list_head list;
struct module *owner;
struct adf_accel_pci accel_pci_dev;
@@ -223,6 +223,24 @@ static int adf_get_dc_capabilities(struct adf_accel_dev *accel_dev,
return 0;
}
+int adf_get_ae_fw_counters(struct adf_accel_dev *accel_dev, u16 ae, u64 *reqs, u64 *resps)
+{
+ struct icp_qat_fw_init_admin_resp resp = { };
+ struct icp_qat_fw_init_admin_req req = { };
+ int ret;
+
+ req.cmd_id = ICP_QAT_FW_COUNTERS_GET;
+
+ ret = adf_put_admin_msg_sync(accel_dev, ae, &req, &resp);
+ if (ret || resp.status)
+ return -EFAULT;
+
+ *reqs = resp.req_rec_count;
+ *resps = resp.resp_sent_count;
+
+ return 0;
+}
+
/**
* adf_send_admin_init() - Function sends init message to FW
* @accel_dev: Pointer to acceleration device.
@@ -94,6 +94,7 @@ void adf_exit_aer(void);
int adf_init_admin_comms(struct adf_accel_dev *accel_dev);
void adf_exit_admin_comms(struct adf_accel_dev *accel_dev);
int adf_send_admin_init(struct adf_accel_dev *accel_dev);
+int adf_get_ae_fw_counters(struct adf_accel_dev *accel_dev, u16 ae, u64 *reqs, u64 *resps);
int adf_init_admin_pm(struct adf_accel_dev *accel_dev, u32 idle_delay);
int adf_init_arb(struct adf_accel_dev *accel_dev);
void adf_exit_arb(struct adf_accel_dev *accel_dev);
@@ -6,6 +6,7 @@
#include "adf_cfg.h"
#include "adf_common_drv.h"
#include "adf_dbgfs.h"
+#include "adf_fw_counters.h"
/**
* adf_dbgfs_init() - add persistent debugfs entries
@@ -56,6 +57,9 @@ void adf_dbgfs_add(struct adf_accel_dev *accel_dev)
{
if (!accel_dev->debugfs_dir)
return;
+
+ if (!accel_dev->is_vf)
+ adf_fw_counters_dbgfs_add(accel_dev);
}
/**
@@ -66,4 +70,7 @@ void adf_dbgfs_rm(struct adf_accel_dev *accel_dev)
{
if (!accel_dev->debugfs_dir)
return;
+
+ if (!accel_dev->is_vf)
+ adf_fw_counters_dbgfs_rm(accel_dev);
}
new file mode 100644
@@ -0,0 +1,264 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/* Copyright(c) 2023 Intel Corporation */
+#include <linux/bitops.h>
+#include <linux/debugfs.h>
+#include <linux/err.h>
+#include <linux/fs.h>
+#include <linux/kernel.h>
+#include <linux/seq_file.h>
+#include <linux/types.h>
+
+#include "adf_accel_devices.h"
+#include "adf_common_drv.h"
+#include "adf_fw_counters.h"
+
+#define ADF_FW_COUNTERS_MAX_PADDING 16
+
+enum adf_fw_counters_types {
+ ADF_FW_REQUESTS,
+ ADF_FW_RESPONSES,
+ ADF_FW_COUNTERS_COUNT
+};
+
+static const char * const adf_fw_counter_names[] = {
+ [ADF_FW_REQUESTS] = "Requests",
+ [ADF_FW_RESPONSES] = "Responses",
+};
+
+static_assert(ARRAY_SIZE(adf_fw_counter_names) == ADF_FW_COUNTERS_COUNT);
+
+struct adf_ae_counters {
+ u16 ae;
+ u64 values[ADF_FW_COUNTERS_COUNT];
+};
+
+struct adf_fw_counters {
+ u16 ae_count;
+ struct adf_ae_counters ae_counters[];
+};
+
+static void adf_fw_counters_parse_ae_values(struct adf_ae_counters *ae_counters, u32 ae,
+ u64 req_count, u64 resp_count)
+{
+ ae_counters->ae = ae;
+ ae_counters->values[ADF_FW_REQUESTS] = req_count;
+ ae_counters->values[ADF_FW_RESPONSES] = resp_count;
+}
+
+static int adf_fw_counters_load_from_device(struct adf_accel_dev *accel_dev,
+ struct adf_fw_counters *fw_counters)
+{
+ struct adf_hw_device_data *hw_data = GET_HW_DATA(accel_dev);
+ unsigned long ae_mask;
+ unsigned int i;
+ unsigned long ae;
+
+ /* Ignore the admin AEs */
+ ae_mask = hw_data->ae_mask & ~hw_data->admin_ae_mask;
+
+ if (hweight_long(ae_mask) > fw_counters->ae_count)
+ return -EINVAL;
+
+ i = 0;
+ for_each_set_bit(ae, &ae_mask, GET_MAX_ACCELENGINES(accel_dev)) {
+ u64 req_count, resp_count;
+ int ret;
+
+ ret = adf_get_ae_fw_counters(accel_dev, ae, &req_count, &resp_count);
+ if (ret)
+ return ret;
+
+ adf_fw_counters_parse_ae_values(&fw_counters->ae_counters[i++], ae,
+ req_count, resp_count);
+ }
+
+ return 0;
+}
+
+static struct adf_fw_counters *adf_fw_counters_allocate(unsigned long ae_count)
+{
+ struct adf_fw_counters *fw_counters;
+
+ if (unlikely(!ae_count))
+ return ERR_PTR(-EINVAL);
+
+ fw_counters = kmalloc(struct_size(fw_counters, ae_counters, ae_count), GFP_KERNEL);
+ if (!fw_counters)
+ return ERR_PTR(-ENOMEM);
+
+ fw_counters->ae_count = ae_count;
+
+ return fw_counters;
+}
+
+/**
+ * adf_fw_counters_get() - Return FW counters for the provided device.
+ * @accel_dev: Pointer to a QAT acceleration device
+ *
+ * Allocates and returns a table of counters containing execution statistics
+ * for each non-admin AE available through the supplied acceleration device.
+ * The caller becomes the owner of such memory and is responsible for
+ * the deallocation through a call to kfree().
+ *
+ * Returns: a pointer to a dynamically allocated struct adf_fw_counters
+ * on success, or a negative value on error.
+ */
+static struct adf_fw_counters *adf_fw_counters_get(struct adf_accel_dev *accel_dev)
+{
+ struct adf_hw_device_data *hw_data = GET_HW_DATA(accel_dev);
+ struct adf_fw_counters *fw_counters;
+ unsigned long ae_count;
+ int ret;
+
+ if (!adf_dev_started(accel_dev)) {
+ dev_err(&GET_DEV(accel_dev), "QAT Device not started\n");
+ return ERR_PTR(-EFAULT);
+ }
+
+ /* Ignore the admin AEs */
+ ae_count = hweight_long(hw_data->ae_mask & ~hw_data->admin_ae_mask);
+
+ fw_counters = adf_fw_counters_allocate(ae_count);
+ if (IS_ERR(fw_counters))
+ return fw_counters;
+
+ ret = adf_fw_counters_load_from_device(accel_dev, fw_counters);
+ if (ret) {
+ kfree(fw_counters);
+ dev_err(&GET_DEV(accel_dev),
+ "Failed to create QAT fw_counters file table [%d].\n", ret);
+ return ERR_PTR(ret);
+ }
+
+ return fw_counters;
+}
+
+static void *qat_fw_counters_seq_start(struct seq_file *sfile, loff_t *pos)
+{
+ struct adf_fw_counters *fw_counters = sfile->private;
+
+ if (*pos == 0)
+ return SEQ_START_TOKEN;
+
+ if (*pos > fw_counters->ae_count)
+ return NULL;
+
+ return &fw_counters->ae_counters[*pos - 1];
+}
+
+static void *qat_fw_counters_seq_next(struct seq_file *sfile, void *v, loff_t *pos)
+{
+ struct adf_fw_counters *fw_counters = sfile->private;
+
+ (*pos)++;
+
+ if (*pos > fw_counters->ae_count)
+ return NULL;
+
+ return &fw_counters->ae_counters[*pos - 1];
+}
+
+static void qat_fw_counters_seq_stop(struct seq_file *sfile, void *v) {}
+
+static int qat_fw_counters_seq_show(struct seq_file *sfile, void *v)
+{
+ int i;
+
+ if (v == SEQ_START_TOKEN) {
+ seq_puts(sfile, "AE ");
+ for (i = 0; i < ADF_FW_COUNTERS_COUNT; ++i)
+ seq_printf(sfile, " %*s", ADF_FW_COUNTERS_MAX_PADDING,
+ adf_fw_counter_names[i]);
+ } else {
+ struct adf_ae_counters *ae_counters = (struct adf_ae_counters *)v;
+
+ seq_printf(sfile, "%2d:", ae_counters->ae);
+ for (i = 0; i < ADF_FW_COUNTERS_COUNT; ++i)
+ seq_printf(sfile, " %*llu", ADF_FW_COUNTERS_MAX_PADDING,
+ ae_counters->values[i]);
+ }
+ seq_putc(sfile, '\n');
+
+ return 0;
+}
+
+static const struct seq_operations qat_fw_counters_sops = {
+ .start = qat_fw_counters_seq_start,
+ .next = qat_fw_counters_seq_next,
+ .stop = qat_fw_counters_seq_stop,
+ .show = qat_fw_counters_seq_show,
+};
+
+static int qat_fw_counters_file_open(struct inode *inode, struct file *file)
+{
+ struct adf_accel_dev *accel_dev = inode->i_private;
+ struct seq_file *fw_counters_seq_file;
+ struct adf_fw_counters *fw_counters;
+ int ret;
+
+ fw_counters = adf_fw_counters_get(accel_dev);
+ if (IS_ERR(fw_counters))
+ return PTR_ERR(fw_counters);
+
+ ret = seq_open(file, &qat_fw_counters_sops);
+ if (unlikely(ret)) {
+ kfree(fw_counters);
+ return ret;
+ }
+
+ fw_counters_seq_file = file->private_data;
+ fw_counters_seq_file->private = fw_counters;
+ return ret;
+}
+
+static int qat_fw_counters_file_release(struct inode *inode, struct file *file)
+{
+ struct seq_file *seq = file->private_data;
+
+ kfree(seq->private);
+ seq->private = NULL;
+
+ return seq_release(inode, file); }
+
+static const struct file_operations qat_fw_counters_fops = {
+ .owner = THIS_MODULE,
+ .open = qat_fw_counters_file_open,
+ .read = seq_read,
+ .llseek = seq_lseek,
+ .release = qat_fw_counters_file_release,
+};
+
+/**
+ * adf_fw_counters_dbgfs_add() - Create a debugfs file containing FW
+ * execution counters.
+ * @accel_dev: Pointer to a QAT acceleration device
+ *
+ * Function creates a file to display a table with statistics for the given
+ * QAT acceleration device. The table stores device specific execution values
+ * for each AE, such as the number of requests sent to the FW and responses
+ * received from the FW.
+ *
+ * Return: void
+ */
+void adf_fw_counters_dbgfs_add(struct adf_accel_dev *accel_dev)
+{
+ accel_dev->fw_cntr_dbgfile = debugfs_create_file("fw_counters", 0400,
+ accel_dev->debugfs_dir,
+ accel_dev,
+ &qat_fw_counters_fops);
+}
+
+/**
+ * adf_fw_counters_dbgfs_rm() - Remove the debugfs file containing FW counters.
+ * @accel_dev: Pointer to a QAT acceleration device.
+ *
+ * Function removes the file providing the table of statistics for the given
+ * QAT acceleration device.
+ *
+ * Return: void
+ */
+void adf_fw_counters_dbgfs_rm(struct adf_accel_dev *accel_dev)
+{
+ debugfs_remove(accel_dev->fw_cntr_dbgfile);
+ accel_dev->fw_cntr_dbgfile = NULL;
+}
new file mode 100644
@@ -0,0 +1,11 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/* Copyright(c) 2023 Intel Corporation */
+#ifndef ADF_FW_COUNTERS_H
+#define ADF_FW_COUNTERS_H
+
+struct adf_accel_dev;
+
+void adf_fw_counters_dbgfs_add(struct adf_accel_dev *accel_dev);
+void adf_fw_counters_dbgfs_rm(struct adf_accel_dev *accel_dev);
+
+#endif