[PATCH 0/8] crypto: hisilicon - supports to configure function's QoS for ACC

[PATCH 0/8] crypto: hisilicon - supports to configure function's QoS for ACC

[Kai Ye]

Based on the Token bucket algorithm. The QM can limit the throughput
of functions in Kunpeng930. By the device debugfs node that named
"alg_qos", the ACC driver supports to configue each function's QoS
in the host. Based on the communication between pf and vf, the driver
supports reading each function's QoS in the host and VM also by the
debugfs node.

Kai Ye (8):
  crypto: hisilicon/qm - supports writing QoS int the Host
  crypto: hisilicon/qm - add the "alg_qos" file node
  crypto: hisilicon/qm - merges the work initialization process into a
    single function
  crypto: hisilicon/qm - add pf ping single vf function
  crypto: hisilicon/qm - supports to inquiry each function's QoS
  crypto: hisilicon/sec - adds the max shaper type rate
  crypto: hisilicon/hpre - adds the max shaper type rate
  crypto: hisilicon/zip - adds the max shaper type rate

 drivers/crypto/hisilicon/hpre/hpre_main.c |   7 +
 drivers/crypto/hisilicon/qm.c             | 909 +++++++++++++++++++++++-------
 drivers/crypto/hisilicon/qm.h             |  14 +
 drivers/crypto/hisilicon/sec2/sec_main.c  |   7 +
 drivers/crypto/hisilicon/zip/zip_main.c   |  11 +
 5 files changed, 757 insertions(+), 191 deletions(-)

-- 
2.7.4

[PATCH 1/8] crypto: hisilicon/qm - supports writing QoS int the host

[Kai Ye]

Based on the Token bucket algorithm. The HAC driver supports to configure
each function's QoS in the host. The driver supports writing QoS by the
debugfs node that named "alg_qos". The qos value is 1~1000.

Signed-off-by: Kai Ye <yekai13@huawei.com>
---
 drivers/crypto/hisilicon/qm.c | 298 +++++++++++++++++++++++++++++++++++++++++-
 drivers/crypto/hisilicon/qm.h |  14 ++
 2 files changed, 310 insertions(+), 2 deletions(-)

diff --git a/drivers/crypto/hisilicon/qm.c b/drivers/crypto/hisilicon/qm.c
index f79d43c..f2e5d0b 100644
--- a/drivers/crypto/hisilicon/qm.c
+++ b/drivers/crypto/hisilicon/qm.c
@@ -240,6 +240,32 @@
 #define QM_DRIVER_REMOVING		0
 #define QM_RST_SCHED			1
 #define QM_RESETTING			2
+#define QM_QOS_PARAM_NUM		2
+#define QM_QOS_VAL_NUM			1
+#define QM_QOS_BDF_PARAM_NUM		4
+#define QM_QOS_MAX_VAL			1000
+#define QM_QOS_RATE			100
+#define QM_QOS_EXPAND_RATE		1000
+#define QM_SHAPER_CIR_B_MASK		GENMASK(7, 0)
+#define QM_SHAPER_CIR_U_MASK		GENMASK(10, 8)
+#define QM_SHAPER_CIR_S_MASK		GENMASK(14, 11)
+#define QM_SHAPER_FACTOR_CIR_U_SHIFT	8
+#define QM_SHAPER_FACTOR_CIR_S_SHIFT	11
+#define QM_SHAPER_FACTOR_CBS_B_SHIFT	15
+#define QM_SHAPER_FACTOR_CBS_S_SHIFT	19
+#define QM_SHAPER_CBS_B			1
+#define QM_SHAPER_CBS_S			16
+#define QM_SHAPER_VFT_OFFSET		6
+#define QM_QOS_MIN_ERROR_RATE		5
+#define QM_QOS_TYPICAL_NUM		8
+#define QM_SHAPER_MIN_CBS_S		8
+#define QM_QOS_TICK			0x300U
+#define QM_QOS_DIVISOR_CLK		0x1f40U
+#define QM_QOS_MAX_CIR_B		200
+#define QM_QOS_MIN_CIR_B		100
+#define QM_QOS_MAX_CIR_U		6
+#define QM_QOS_MAX_CIR_S		11
+#define QM_QOS_VAL_MAX_LEN		32
 
 #define QM_MK_CQC_DW3_V1(hop_num, pg_sz, buf_sz, cqe_sz) \
 	(((hop_num) << QM_CQ_HOP_NUM_SHIFT)	| \
@@ -280,6 +306,7 @@
 enum vft_type {
 	SQC_VFT = 0,
 	CQC_VFT,
+	SHAPER_VFT,
 };
 
 enum acc_err_result {
@@ -288,6 +315,11 @@ enum acc_err_result {
 	ACC_ERR_RECOVERED,
 };
 
+enum qm_alg_type {
+	ALG_TYPE_0,
+	ALG_TYPE_1,
+};
+
 enum qm_mb_cmd {
 	QM_PF_FLR_PREPARE = 0x01,
 	QM_PF_SRST_PREPARE,
@@ -460,6 +492,11 @@ static const char * const qp_s[] = {
 	"none", "init", "start", "stop", "close",
 };
 
+static const u32 typical_qos_val[QM_QOS_TYPICAL_NUM] = {100, 250, 500, 1000,
+						10000, 25000, 50000, 100000};
+static const u32 typical_qos_cbs_s[QM_QOS_TYPICAL_NUM] = {9, 10, 11, 12, 16,
+							 17, 18, 19};
+
 static bool qm_avail_state(struct hisi_qm *qm, enum qm_state new)
 {
 	enum qm_state curr = atomic_read(&qm->status.flags);
@@ -899,8 +936,69 @@ static void qm_init_prefetch(struct hisi_qm *qm)
 	writel(page_type, qm->io_base + QM_PAGE_SIZE);
 }
 
+/*
+ * the formula:
+ * IR = X Mbps if ir = 1 means IR = 100 Mbps, if ir = 10000 means = 10Gbps
+ *
+ *		        IR_b * (2 ^ IR_u) * 8
+ * IR(Mbps) * 10 ^ -3 = -------------------------
+ *		        Tick * (2 ^ IR_s)
+ */
+static u32 acc_shaper_para_calc(u64 cir_b, u64 cir_u, u64 cir_s)
+{
+	return ((cir_b * QM_QOS_DIVISOR_CLK) * (1 << cir_u)) /
+					(QM_QOS_TICK * (1 << cir_s));
+}
+
+static u32 acc_shaper_calc_cbs_s(u32 ir)
+{
+	int i;
+
+	if (ir < typical_qos_val[0])
+		return QM_SHAPER_MIN_CBS_S;
+
+	for (i = 1; i < QM_QOS_TYPICAL_NUM; i++) {
+		if (ir >= typical_qos_val[i - 1] && ir < typical_qos_val[i])
+			return typical_qos_cbs_s[i - 1];
+	}
+
+	return typical_qos_cbs_s[QM_QOS_TYPICAL_NUM - 1];
+}
+
+static int qm_get_shaper_para(u32 ir, struct qm_shaper_factor *factor)
+{
+	u32 cir_b, cir_u, cir_s, ir_calc;
+	u32 error_rate;
+
+	factor->cbs_s = acc_shaper_calc_cbs_s(ir);
+
+	for (cir_b = QM_QOS_MIN_CIR_B; cir_b <= QM_QOS_MAX_CIR_B; cir_b++) {
+		for (cir_u = 0; cir_u <= QM_QOS_MAX_CIR_U; cir_u++) {
+			for (cir_s = 0; cir_s <= QM_QOS_MAX_CIR_S; cir_s++) {
+				/** the formula is changed to:
+				 *	   IR_b * (2 ^ IR_u) * DIVISOR_CLK
+				 * IR(Mbps) = -------------------------
+				 *	       768 * (2 ^ IR_s)
+				 */
+				ir_calc = acc_shaper_para_calc(cir_b, cir_u,
+							       cir_s);
+				error_rate = QM_QOS_EXPAND_RATE * (u32)abs(ir_calc - ir) / ir;
+				if (error_rate <= QM_QOS_MIN_ERROR_RATE) {
+					factor->cir_b = cir_b;
+					factor->cir_u = cir_u;
+					factor->cir_s = cir_s;
+
+					return 0;
+				}
+			}
+		}
+	}
+
+	return -EINVAL;
+}
+
 static void qm_vft_data_cfg(struct hisi_qm *qm, enum vft_type type, u32 base,
-			    u32 number)
+			    u32 number, struct qm_shaper_factor *factor)
 {
 	u64 tmp = 0;
 
@@ -929,6 +1027,15 @@ static void qm_vft_data_cfg(struct hisi_qm *qm, enum vft_type type, u32 base,
 				tmp = QM_CQC_VFT_VALID;
 			}
 			break;
+		case SHAPER_VFT:
+			if (qm->ver >= QM_HW_V3) {
+				tmp = factor->cir_b |
+				(factor->cir_u << QM_SHAPER_FACTOR_CIR_U_SHIFT) |
+				(factor->cir_s << QM_SHAPER_FACTOR_CIR_S_SHIFT) |
+				(QM_SHAPER_CBS_B << QM_SHAPER_FACTOR_CBS_B_SHIFT) |
+				(factor->cbs_s << QM_SHAPER_FACTOR_CBS_S_SHIFT);
+			}
+			break;
 		}
 	}
 
@@ -939,6 +1046,7 @@ static void qm_vft_data_cfg(struct hisi_qm *qm, enum vft_type type, u32 base,
 static int qm_set_vft_common(struct hisi_qm *qm, enum vft_type type,
 			     u32 fun_num, u32 base, u32 number)
 {
+	struct qm_shaper_factor *factor = &qm->factor[fun_num];
 	unsigned int val;
 	int ret;
 
@@ -950,9 +1058,12 @@ static int qm_set_vft_common(struct hisi_qm *qm, enum vft_type type,
 
 	writel(0x0, qm->io_base + QM_VFT_CFG_OP_WR);
 	writel(type, qm->io_base + QM_VFT_CFG_TYPE);
+	if (type == SHAPER_VFT)
+		fun_num |= base << QM_SHAPER_VFT_OFFSET;
+
 	writel(fun_num, qm->io_base + QM_VFT_CFG);
 
-	qm_vft_data_cfg(qm, type, base, number);
+	qm_vft_data_cfg(qm, type, base, number, factor);
 
 	writel(0x0, qm->io_base + QM_VFT_CFG_RDY);
 	writel(0x1, qm->io_base + QM_VFT_CFG_OP_ENABLE);
@@ -962,6 +1073,27 @@ static int qm_set_vft_common(struct hisi_qm *qm, enum vft_type type,
 					  POLL_TIMEOUT);
 }
 
+static int qm_shaper_init_vft(struct hisi_qm *qm, u32 fun_num)
+{
+	int ret, i;
+
+	qm->factor[fun_num].func_qos = QM_QOS_MAX_VAL;
+	ret = qm_get_shaper_para(QM_QOS_MAX_VAL * QM_QOS_RATE, &qm->factor[fun_num]);
+	if (ret) {
+		dev_err(&qm->pdev->dev, "failed to calculate shaper parameter!\n");
+		return ret;
+	}
+	writel(qm->type_rate, qm->io_base + QM_SHAPER_CFG);
+	for (i = ALG_TYPE_0; i <= ALG_TYPE_1; i++) {
+		/* The base number of queue reuse for different alg type */
+		ret = qm_set_vft_common(qm, SHAPER_VFT, fun_num, i, 1);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
 /* The config should be conducted after qm_dev_mem_reset() */
 static int qm_set_sqc_cqc_vft(struct hisi_qm *qm, u32 fun_num, u32 base,
 			      u32 number)
@@ -974,7 +1106,21 @@ static int qm_set_sqc_cqc_vft(struct hisi_qm *qm, u32 fun_num, u32 base,
 			return ret;
 	}
 
+	/* init default shaper qos val */
+	if (qm->ver >= QM_HW_V3) {
+		ret = qm_shaper_init_vft(qm, fun_num);
+		if (ret)
+			goto back_sqc_cqc;
+	}
+
 	return 0;
+back_sqc_cqc:
+	for (i = SQC_VFT; i <= CQC_VFT; i++) {
+		ret = qm_set_vft_common(qm, i, fun_num, 0, 0);
+		if (ret)
+			return ret;
+	}
+	return ret;
 }
 
 static int qm_get_vft_v2(struct hisi_qm *qm, u32 *base, u32 *number)
@@ -3133,6 +3279,7 @@ void hisi_qm_uninit(struct hisi_qm *qm)
 	struct device *dev = &pdev->dev;
 
 	qm_cmd_uninit(qm);
+	kfree(qm->factor);
 	down_write(&qm->qps_lock);
 
 	if (!qm_avail_state(qm, QM_CLOSE)) {
@@ -3862,6 +4009,149 @@ static int qm_clear_vft_config(struct hisi_qm *qm)
 	return 0;
 }
 
+static int qm_func_shaper_enable(struct hisi_qm *qm, u32 fun_index, u32 qos)
+{
+	struct device *dev = &qm->pdev->dev;
+	u32 ir = qos * QM_QOS_RATE;
+	int ret, total_vfs, i;
+
+	total_vfs = pci_sriov_get_totalvfs(qm->pdev);
+	if (fun_index > total_vfs)
+		return -EINVAL;
+
+	qm->factor[fun_index].func_qos = qos;
+
+	ret = qm_get_shaper_para(ir, &qm->factor[fun_index]);
+	if (ret) {
+		dev_err(dev, "failed to calculate shaper parameter!\n");
+		return -EINVAL;
+	}
+
+	for (i = ALG_TYPE_0; i <= ALG_TYPE_1; i++) {
+		/* The base number of queue reuse for different alg type */
+		ret = qm_set_vft_common(qm, SHAPER_VFT, fun_index, i, 1);
+		if (ret) {
+			dev_err(dev, "type: %d, failed to set shaper vft!\n", i);
+			return -EINVAL;
+		}
+	}
+
+	return 0;
+}
+
+static ssize_t qm_qos_value_init(const char *buf, unsigned long *val)
+{
+	int buflen = strlen(buf);
+	int ret, i;
+
+	for (i = 0; i < buflen; i++) {
+		if (!isdigit(buf[i]))
+			return -EINVAL;
+	}
+
+	ret = sscanf(buf, "%ld", val);
+	if (ret != QM_QOS_VAL_NUM)
+		return -EINVAL;
+
+	return 0;
+}
+
+static ssize_t qm_algqos_write(struct file *filp, const char __user *buf,
+			       size_t count, loff_t *pos)
+{
+	struct hisi_qm *qm = filp->private_data;
+	char tbuf[QM_DBG_READ_LEN];
+	int tmp1, bus, device, function;
+	char tbuf_bdf[QM_DBG_READ_LEN] = {0};
+	char val_buf[QM_QOS_VAL_MAX_LEN] = {0};
+	unsigned int fun_index;
+	unsigned long val = 0;
+	int len, ret;
+
+	if (qm->fun_type == QM_HW_VF)
+		return -EINVAL;
+
+	/* Mailbox and reset cannot be operated at the same time */
+	if (test_and_set_bit(QM_RESETTING, &qm->misc_ctl)) {
+		pci_err(qm->pdev, "dev resetting, write alg qos failed!\n");
+		return -EAGAIN;
+	}
+
+	if (*pos != 0) {
+		ret = 0;
+		goto err_get_status;
+	}
+
+	if (count >= QM_DBG_READ_LEN) {
+		ret = -ENOSPC;
+		goto err_get_status;
+	}
+
+	len = simple_write_to_buffer(tbuf, QM_DBG_READ_LEN - 1, pos, buf, count);
+	if (len < 0) {
+		ret = len;
+		goto err_get_status;
+	}
+
+	tbuf[len] = '\0';
+	ret = sscanf(tbuf, "%s %s", tbuf_bdf, val_buf);
+	if (ret != QM_QOS_PARAM_NUM) {
+		ret = -EINVAL;
+		goto err_get_status;
+	}
+
+	ret = qm_qos_value_init(val_buf, &val);
+	if (val == 0 || val > QM_QOS_MAX_VAL || ret) {
+		pci_err(qm->pdev, "input qos value is error, please set 1~1000!\n");
+		ret = -EINVAL;
+		goto err_get_status;
+	}
+
+	ret = sscanf(tbuf_bdf, "%d:%x:%d.%d", &tmp1, &bus, &device, &function);
+	if (ret != QM_QOS_BDF_PARAM_NUM) {
+		pci_err(qm->pdev, "input pci bdf value is error!\n");
+		ret = -EINVAL;
+		goto err_get_status;
+	}
+
+	fun_index = device * 8 + function;
+
+	ret = qm_func_shaper_enable(qm, fun_index, val);
+	if (ret) {
+		pci_err(qm->pdev, "failed to enable function shaper!\n");
+		ret = -EINVAL;
+		goto err_get_status;
+	}
+
+	ret =  count;
+
+err_get_status:
+	clear_bit(QM_RESETTING, &qm->misc_ctl);
+	return ret;
+}
+
+static const struct file_operations qm_algqos_fops = {
+	.owner = THIS_MODULE,
+	.open = simple_open,
+	.write = qm_algqos_write,
+};
+
+/**
+ * hisi_qm_set_algqos_init() - Initialize function qos debugfs files.
+ * @qm: The qm for which we want to add debugfs files.
+ *
+ * Create function qos debugfs files.
+ */
+static void hisi_qm_set_algqos_init(struct hisi_qm *qm)
+{
+	if (qm->fun_type == QM_HW_PF)
+		debugfs_create_file("alg_qos", 0644, qm->debug.debug_root,
+				    qm, &qm_algqos_fops);
+	else
+		debugfs_create_file("alg_qos", 0444, qm->debug.debug_root,
+				    qm, &qm_algqos_fops);
+}
+
 /**
  * hisi_qm_sriov_enable() - enable virtual functions
  * @pdev: the PCIe device
@@ -3916,6 +4206,7 @@ EXPORT_SYMBOL_GPL(hisi_qm_sriov_enable);
 int hisi_qm_sriov_disable(struct pci_dev *pdev, bool is_frozen)
 {
 	struct hisi_qm *qm = pci_get_drvdata(pdev);
+	int total_vfs = pci_sriov_get_totalvfs(qm->pdev);
 
 	if (pci_vfs_assigned(pdev)) {
 		pci_err(pdev, "Failed to disable VFs as VFs are assigned!\n");
@@ -3929,6 +4220,9 @@ int hisi_qm_sriov_disable(struct pci_dev *pdev, bool is_frozen)
 	}
 
 	pci_disable_sriov(pdev);
+	/* clear vf function shaper configure array */
+	memset(qm->factor + 1, 0, sizeof(struct qm_shaper_factor) * total_vfs);
+
 	return qm_clear_vft_config(qm);
 }
 EXPORT_SYMBOL_GPL(hisi_qm_sriov_disable);
diff --git a/drivers/crypto/hisilicon/qm.h b/drivers/crypto/hisilicon/qm.h
index 8a36bad..035eaf8 100644
--- a/drivers/crypto/hisilicon/qm.h
+++ b/drivers/crypto/hisilicon/qm.h
@@ -76,6 +76,9 @@
 #define QM_Q_DEPTH			1024
 #define QM_MIN_QNUM                     2
 #define HISI_ACC_SGL_SGE_NR_MAX		255
+#define QM_SHAPER_CFG			0x100164
+#define QM_SHAPER_ENABLE		BIT(30)
+#define QM_SHAPER_TYPE1_OFFSET		10
 
 /* page number for queue file region */
 #define QM_DOORBELL_PAGE_NR		1
@@ -148,6 +151,14 @@ struct qm_debug {
 	struct debugfs_file files[DEBUG_FILE_NUM];
 };
 
+struct qm_shaper_factor {
+	u32 func_qos;
+	u64 cir_b;
+	u64 cir_u;
+	u64 cir_s;
+	u64 cbs_s;
+};
+
 struct qm_dma {
 	void *va;
 	dma_addr_t dma;
@@ -262,6 +273,9 @@ struct hisi_qm {
 	resource_size_t db_phys_base;
 	struct uacce_device *uacce;
 	int mode;
+	struct qm_shaper_factor *factor;
+	u32 mb_qos;
+	u32 type_rate;
 };
 
 struct hisi_qp_status {
-- 
2.7.4

[PATCH 2/8] crypto: hisilicon/qm - add the "alg_qos" file node

[Kai Ye]

1. Just move the code as needed.
2. Add the "alg_qos" file node in the qm debug sysfs.

Signed-off-by: Kai Ye <yekai13@huawei.com>
---
 drivers/crypto/hisilicon/qm.c | 284 +++++++++++++++++++++++-------------------
 1 file changed, 155 insertions(+), 129 deletions(-)

diff --git a/drivers/crypto/hisilicon/qm.c b/drivers/crypto/hisilicon/qm.c
index f2e5d0b..171ddbb 100644
--- a/drivers/crypto/hisilicon/qm.c
+++ b/drivers/crypto/hisilicon/qm.c
@@ -3145,62 +3145,6 @@ static int hisi_qp_memory_init(struct hisi_qm *qm, size_t dma_size, int id)
 	return 0;
 }
 
-static int hisi_qm_memory_init(struct hisi_qm *qm)
-{
-	struct device *dev = &qm->pdev->dev;
-	size_t qp_dma_size, off = 0;
-	int i, ret = 0;
-
-#define QM_INIT_BUF(qm, type, num) do { \
-	(qm)->type = ((qm)->qdma.va + (off)); \
-	(qm)->type##_dma = (qm)->qdma.dma + (off); \
-	off += QMC_ALIGN(sizeof(struct qm_##type) * (num)); \
-} while (0)
-
-	idr_init(&qm->qp_idr);
-	qm->qdma.size = QMC_ALIGN(sizeof(struct qm_eqe) * QM_EQ_DEPTH) +
-			QMC_ALIGN(sizeof(struct qm_aeqe) * QM_Q_DEPTH) +
-			QMC_ALIGN(sizeof(struct qm_sqc) * qm->qp_num) +
-			QMC_ALIGN(sizeof(struct qm_cqc) * qm->qp_num);
-	qm->qdma.va = dma_alloc_coherent(dev, qm->qdma.size, &qm->qdma.dma,
-					 GFP_ATOMIC);
-	dev_dbg(dev, "allocate qm dma buf size=%zx)\n", qm->qdma.size);
-	if (!qm->qdma.va)
-		return -ENOMEM;
-
-	QM_INIT_BUF(qm, eqe, QM_EQ_DEPTH);
-	QM_INIT_BUF(qm, aeqe, QM_Q_DEPTH);
-	QM_INIT_BUF(qm, sqc, qm->qp_num);
-	QM_INIT_BUF(qm, cqc, qm->qp_num);
-
-	qm->qp_array = kcalloc(qm->qp_num, sizeof(struct hisi_qp), GFP_KERNEL);
-	if (!qm->qp_array) {
-		ret = -ENOMEM;
-		goto err_alloc_qp_array;
-	}
-
-	/* one more page for device or qp statuses */
-	qp_dma_size = qm->sqe_size * QM_Q_DEPTH +
-		      sizeof(struct qm_cqe) * QM_Q_DEPTH;
-	qp_dma_size = PAGE_ALIGN(qp_dma_size);
-	for (i = 0; i < qm->qp_num; i++) {
-		ret = hisi_qp_memory_init(qm, qp_dma_size, i);
-		if (ret)
-			goto err_init_qp_mem;
-
-		dev_dbg(dev, "allocate qp dma buf size=%zx)\n", qp_dma_size);
-	}
-
-	return ret;
-
-err_init_qp_mem:
-	hisi_qp_memory_uninit(qm, i);
-err_alloc_qp_array:
-	dma_free_coherent(dev, qm->qdma.size, qm->qdma.va, qm->qdma.dma);
-
-	return ret;
-}
-
 static void hisi_qm_pre_init(struct hisi_qm *qm)
 {
 	struct pci_dev *pdev = qm->pdev;
@@ -3681,79 +3625,6 @@ static int qm_debugfs_atomic64_get(void *data, u64 *val)
 DEFINE_DEBUGFS_ATTRIBUTE(qm_atomic64_ops, qm_debugfs_atomic64_get,
 			 qm_debugfs_atomic64_set, "%llu\n");
 
-/**
- * hisi_qm_debug_init() - Initialize qm related debugfs files.
- * @qm: The qm for which we want to add debugfs files.
- *
- * Create qm related debugfs files.
- */
-void hisi_qm_debug_init(struct hisi_qm *qm)
-{
-	struct qm_dfx *dfx = &qm->debug.dfx;
-	struct dentry *qm_d;
-	void *data;
-	int i;
-
-	qm_d = debugfs_create_dir("qm", qm->debug.debug_root);
-	qm->debug.qm_d = qm_d;
-
-	/* only show this in PF */
-	if (qm->fun_type == QM_HW_PF) {
-		qm_create_debugfs_file(qm, qm->debug.debug_root, CURRENT_QM);
-		for (i = CURRENT_Q; i < DEBUG_FILE_NUM; i++)
-			qm_create_debugfs_file(qm, qm_d, i);
-	}
-
-	debugfs_create_file("regs", 0444, qm->debug.qm_d, qm, &qm_regs_fops);
-
-	debugfs_create_file("cmd", 0444, qm->debug.qm_d, qm, &qm_cmd_fops);
-
-	debugfs_create_file("status", 0444, qm->debug.qm_d, qm,
-			&qm_status_fops);
-	for (i = 0; i < ARRAY_SIZE(qm_dfx_files); i++) {
-		data = (atomic64_t *)((uintptr_t)dfx + qm_dfx_files[i].offset);
-		debugfs_create_file(qm_dfx_files[i].name,
-			0644,
-			qm_d,
-			data,
-			&qm_atomic64_ops);
-	}
-}
-EXPORT_SYMBOL_GPL(hisi_qm_debug_init);
-
-/**
- * hisi_qm_debug_regs_clear() - clear qm debug related registers.
- * @qm: The qm for which we want to clear its debug registers.
- */
-void hisi_qm_debug_regs_clear(struct hisi_qm *qm)
-{
-	struct qm_dfx_registers *regs;
-	int i;
-
-	/* clear current_qm */
-	writel(0x0, qm->io_base + QM_DFX_MB_CNT_VF);
-	writel(0x0, qm->io_base + QM_DFX_DB_CNT_VF);
-
-	/* clear current_q */
-	writel(0x0, qm->io_base + QM_DFX_SQE_CNT_VF_SQN);
-	writel(0x0, qm->io_base + QM_DFX_CQE_CNT_VF_CQN);
-
-	/*
-	 * these registers are reading and clearing, so clear them after
-	 * reading them.
-	 */
-	writel(0x1, qm->io_base + QM_DFX_CNT_CLR_CE);
-
-	regs = qm_dfx_regs;
-	for (i = 0; i < CNT_CYC_REGS_NUM; i++) {
-		readl(qm->io_base + regs->reg_offset);
-		regs++;
-	}
-
-	writel(0x0, qm->io_base + QM_DFX_CNT_CLR_CE);
-}
-EXPORT_SYMBOL_GPL(hisi_qm_debug_regs_clear);
-
 static void qm_hw_error_init(struct hisi_qm *qm)
 {
 	struct hisi_qm_err_info *err_info = &qm->err_info;
@@ -4153,6 +4024,83 @@ static void hisi_qm_set_algqos_init(struct hisi_qm *qm)
 }
 
 /**
+ * hisi_qm_debug_init() - Initialize qm related debugfs files.
+ * @qm: The qm for which we want to add debugfs files.
+ *
+ * Create qm related debugfs files.
+ */
+void hisi_qm_debug_init(struct hisi_qm *qm)
+{
+	struct qm_dfx *dfx = &qm->debug.dfx;
+	struct dentry *qm_d;
+	void *data;
+	int i;
+
+	qm_d = debugfs_create_dir("qm", qm->debug.debug_root);
+	qm->debug.qm_d = qm_d;
+
+	/* only show this in PF */
+	if (qm->fun_type == QM_HW_PF) {
+		qm_create_debugfs_file(qm, qm->debug.debug_root, CURRENT_QM);
+		for (i = CURRENT_Q; i < DEBUG_FILE_NUM; i++)
+			qm_create_debugfs_file(qm, qm->debug.qm_d, i);
+	}
+
+	debugfs_create_file("regs", 0444, qm->debug.qm_d, qm, &qm_regs_fops);
+
+	debugfs_create_file("cmd", 0600, qm->debug.qm_d, qm, &qm_cmd_fops);
+
+	debugfs_create_file("status", 0444, qm->debug.qm_d, qm,
+			&qm_status_fops);
+	for (i = 0; i < ARRAY_SIZE(qm_dfx_files); i++) {
+		data = (atomic64_t *)((uintptr_t)dfx + qm_dfx_files[i].offset);
+		debugfs_create_file(qm_dfx_files[i].name,
+			0644,
+			qm_d,
+			data,
+			&qm_atomic64_ops);
+	}
+
+	if (qm->ver >= QM_HW_V3)
+		hisi_qm_set_algqos_init(qm);
+}
+EXPORT_SYMBOL_GPL(hisi_qm_debug_init);
+
+/**
+ * hisi_qm_debug_regs_clear() - clear qm debug related registers.
+ * @qm: The qm for which we want to clear its debug registers.
+ */
+void hisi_qm_debug_regs_clear(struct hisi_qm *qm)
+{
+	struct qm_dfx_registers *regs;
+	int i;
+
+	/* clear current_qm */
+	writel(0x0, qm->io_base + QM_DFX_MB_CNT_VF);
+	writel(0x0, qm->io_base + QM_DFX_DB_CNT_VF);
+
+	/* clear current_q */
+	writel(0x0, qm->io_base + QM_DFX_SQE_CNT_VF_SQN);
+	writel(0x0, qm->io_base + QM_DFX_CQE_CNT_VF_CQN);
+
+	/*
+	 * these registers are reading and clearing, so clear them after
+	 * reading them.
+	 */
+	writel(0x1, qm->io_base + QM_DFX_CNT_CLR_CE);
+
+	regs = qm_dfx_regs;
+	for (i = 0; i < CNT_CYC_REGS_NUM; i++) {
+		readl(qm->io_base + regs->reg_offset);
+		regs++;
+	}
+
+	/* clear clear_enable */
+	writel(0x0, qm->io_base + QM_DFX_CNT_CLR_CE);
+}
+EXPORT_SYMBOL_GPL(hisi_qm_debug_regs_clear);
+
+/**
  * hisi_qm_sriov_enable() - enable virtual functions
  * @pdev: the PCIe device
  * @max_vfs: the number of virtual functions to enable
@@ -5389,6 +5337,84 @@ static int hisi_qm_pci_init(struct hisi_qm *qm)
 	return ret;
 }
 
+static int hisi_qp_alloc_memory(struct hisi_qm *qm)
+{
+	struct device *dev = &qm->pdev->dev;
+	size_t qp_dma_size;
+	int i, ret;
+
+	qm->qp_array = kcalloc(qm->qp_num, sizeof(struct hisi_qp), GFP_KERNEL);
+	if (!qm->qp_array)
+		return -ENOMEM;
+
+	/* one more page for device or qp statuses */
+	qp_dma_size = qm->sqe_size * QM_Q_DEPTH +
+		      sizeof(struct qm_cqe) * QM_Q_DEPTH;
+	qp_dma_size = PAGE_ALIGN(qp_dma_size) + PAGE_SIZE;
+	for (i = 0; i < qm->qp_num; i++) {
+		ret = hisi_qp_memory_init(qm, qp_dma_size, i);
+		if (ret)
+			goto err_init_qp_mem;
+
+		dev_dbg(dev, "allocate qp dma buf size=%zx)\n", qp_dma_size);
+	}
+
+	return 0;
+err_init_qp_mem:
+	hisi_qp_memory_uninit(qm, i);
+
+	return ret;
+}
+
+static int hisi_qm_memory_init(struct hisi_qm *qm)
+{
+	struct device *dev = &qm->pdev->dev;
+	int ret, total_vfs;
+	size_t off = 0;
+
+	total_vfs = pci_sriov_get_totalvfs(qm->pdev);
+	qm->factor = kcalloc(total_vfs + 1, sizeof(struct qm_shaper_factor), GFP_KERNEL);
+	if (!qm->factor)
+		return -ENOMEM;
+
+#define QM_INIT_BUF(qm, type, num) do { \
+	(qm)->type = ((qm)->qdma.va + (off)); \
+	(qm)->type##_dma = (qm)->qdma.dma + (off); \
+	off += QMC_ALIGN(sizeof(struct qm_##type) * (num)); \
+} while (0)
+
+	idr_init(&qm->qp_idr);
+	qm->qdma.size = QMC_ALIGN(sizeof(struct qm_eqe) * QM_EQ_DEPTH) +
+			QMC_ALIGN(sizeof(struct qm_aeqe) * QM_Q_DEPTH) +
+			QMC_ALIGN(sizeof(struct qm_sqc) * qm->qp_num) +
+			QMC_ALIGN(sizeof(struct qm_cqc) * qm->qp_num);
+	qm->qdma.va = dma_alloc_coherent(dev, qm->qdma.size, &qm->qdma.dma,
+					 GFP_ATOMIC);
+	dev_dbg(dev, "allocate qm dma buf size=%zx)\n", qm->qdma.size);
+	if (!qm->qdma.va) {
+		ret =  -ENOMEM;
+		goto err_alloc_qdma;
+	}
+
+	QM_INIT_BUF(qm, eqe, QM_EQ_DEPTH);
+	QM_INIT_BUF(qm, aeqe, QM_Q_DEPTH);
+	QM_INIT_BUF(qm, sqc, qm->qp_num);
+	QM_INIT_BUF(qm, cqc, qm->qp_num);
+
+	ret = hisi_qp_alloc_memory(qm);
+	if (ret)
+		goto err_alloc_qp_array;
+
+	return 0;
+
+err_alloc_qp_array:
+	dma_free_coherent(dev, qm->qdma.size, qm->qdma.va, qm->qdma.dma);
+err_alloc_qdma:
+	kfree(qm->factor);
+
+	return ret;
+}
+
 /**
  * hisi_qm_init() - Initialize configures about qm.
  * @qm: The qm needing init.
-- 
2.7.4

[PATCH 3/8] crypto: hisilicon/qm - merges the work initialization process into a single function

[Kai Ye]

Merges the work initialization process into a single function from
qm initialization.

Signed-off-by: Kai Ye <yekai13@huawei.com>
---
 drivers/crypto/hisilicon/qm.c | 18 +++++++++++-------
 1 file changed, 11 insertions(+), 7 deletions(-)

diff --git a/drivers/crypto/hisilicon/qm.c b/drivers/crypto/hisilicon/qm.c
index 171ddbb..9b8e09e 100644
--- a/drivers/crypto/hisilicon/qm.c
+++ b/drivers/crypto/hisilicon/qm.c
@@ -5337,6 +5337,16 @@ static int hisi_qm_pci_init(struct hisi_qm *qm)
 	return ret;
 }
 
+static void hisi_qm_init_work(struct hisi_qm *qm)
+{
+	INIT_WORK(&qm->work, qm_work_process);
+	if (qm->fun_type == QM_HW_PF)
+		INIT_WORK(&qm->rst_work, hisi_qm_controller_reset);
+
+	if (qm->ver > QM_HW_V2)
+		INIT_WORK(&qm->cmd_process, qm_cmd_process);
+}
+
 static int hisi_qp_alloc_memory(struct hisi_qm *qm)
 {
 	struct device *dev = &qm->pdev->dev;
@@ -5452,13 +5462,7 @@ int hisi_qm_init(struct hisi_qm *qm)
 	if (ret)
 		goto err_alloc_uacce;
 
-	INIT_WORK(&qm->work, qm_work_process);
-	if (qm->fun_type == QM_HW_PF)
-		INIT_WORK(&qm->rst_work, hisi_qm_controller_reset);
-
-	if (qm->ver >= QM_HW_V3)
-		INIT_WORK(&qm->cmd_process, qm_cmd_process);
-
+	hisi_qm_init_work(qm);
 	qm_cmd_init(qm);
 	atomic_set(&qm->status.flags, QM_INIT);
 
-- 
2.7.4

[PATCH 4/8] crypto: hisilicon/qm - add pf ping single vf function

[Kai Ye]

According to the function communication, add pf ping single
vf function to be used in the vf read QoS.

Signed-off-by: Kai Ye <yekai13@huawei.com>
---
 drivers/crypto/hisilicon/qm.c | 36 ++++++++++++++++++++++++++++++++++++
 1 file changed, 36 insertions(+)

diff --git a/drivers/crypto/hisilicon/qm.c b/drivers/crypto/hisilicon/qm.c
index 9b8e09e..72648dc 100644
--- a/drivers/crypto/hisilicon/qm.c
+++ b/drivers/crypto/hisilicon/qm.c
@@ -2142,6 +2142,42 @@ static void qm_trigger_pf_interrupt(struct hisi_qm *qm)
 	writel(val, qm->io_base + QM_IFC_INT_SET_V);
 }
 
+static int qm_ping_single_vf(struct hisi_qm *qm, u64 cmd, u32 fun_num)
+{
+	struct device *dev = &qm->pdev->dev;
+	struct qm_mailbox mailbox;
+	int cnt = 0;
+	u64 val;
+	int ret;
+
+	qm_mb_pre_init(&mailbox, QM_MB_CMD_SRC, cmd, fun_num, 0);
+	mutex_lock(&qm->mailbox_lock);
+	ret = qm_mb_nolock(qm, &mailbox);
+	if (ret) {
+		dev_err(dev, "failed to send command to vf(%u)!\n", fun_num);
+		goto err_unlock;
+	}
+
+	qm_trigger_vf_interrupt(qm, fun_num);
+	while (true) {
+		msleep(QM_WAIT_DST_ACK);
+		val = readq(qm->io_base + QM_IFC_READY_STATUS);
+		/* if VF respond, PF notifies VF successfully. */
+		if (!(val & BIT(fun_num)))
+			goto err_unlock;
+
+		if (++cnt > QM_MAX_PF_WAIT_COUNT) {
+			dev_err(dev, "failed to get response from VF(%u)!\n", fun_num);
+			ret = -ETIMEDOUT;
+			break;
+		}
+	}
+
+err_unlock:
+	mutex_unlock(&qm->mailbox_lock);
+	return ret;
+}
+
 static int qm_ping_all_vfs(struct hisi_qm *qm, u64 cmd)
 {
 	struct device *dev = &qm->pdev->dev;
-- 
2.7.4

[PATCH 5/8] crypto: hisilicon/qm - supports to inquiry each function's QoS

[Kai Ye]

1. The ACC driver supports to inquiry each function's QoS in the Host
and VM. The driver supports reading QoS by the device debug SysFS
attribute file "alg_qos", like "cat alg_qos".
2. Modify the communication process between pf and vf as needed.

Signed-off-by: Kai Ye <yekai13@huawei.com>
---
 drivers/crypto/hisilicon/qm.c | 181 ++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 174 insertions(+), 7 deletions(-)

diff --git a/drivers/crypto/hisilicon/qm.c b/drivers/crypto/hisilicon/qm.c
index 72648dc..636432c 100644
--- a/drivers/crypto/hisilicon/qm.c
+++ b/drivers/crypto/hisilicon/qm.c
@@ -51,6 +51,7 @@
 #define QM_MB_CMD_DATA_ADDR_L		0x304
 #define QM_MB_CMD_DATA_ADDR_H		0x308
 #define QM_MB_PING_ALL_VFS		0xffff
+#define QM_MB_CMD_DATA_SHIFT		32
 #define QM_MB_CMD_DATA_MASK		GENMASK(31, 0)
 
 /* sqc shift */
@@ -185,6 +186,7 @@
 
 /* interfunction communication */
 #define QM_IFC_READY_STATUS		0x100128
+#define QM_IFC_C_STS_M			0x10012C
 #define QM_IFC_INT_SET_P		0x100130
 #define QM_IFC_INT_CFG			0x100134
 #define QM_IFC_INT_SOURCE_P		0x100138
@@ -256,6 +258,7 @@
 #define QM_SHAPER_CBS_B			1
 #define QM_SHAPER_CBS_S			16
 #define QM_SHAPER_VFT_OFFSET		6
+#define WAIT_FOR_QOS_VF			100
 #define QM_QOS_MIN_ERROR_RATE		5
 #define QM_QOS_TYPICAL_NUM		8
 #define QM_SHAPER_MIN_CBS_S		8
@@ -328,6 +331,8 @@ enum qm_mb_cmd {
 	QM_VF_PREPARE_FAIL,
 	QM_VF_START_DONE,
 	QM_VF_START_FAIL,
+	QM_PF_SET_QOS,
+	QM_VF_GET_QOS,
 };
 
 struct qm_cqe {
@@ -2124,7 +2129,7 @@ static void qm_trigger_vf_interrupt(struct hisi_qm *qm, u32 fun_num)
 	u32 val;
 
 	val = readl(qm->io_base + QM_IFC_INT_CFG);
-	val |= ~QM_IFC_SEND_ALL_VFS;
+	val &= ~QM_IFC_SEND_ALL_VFS;
 	val |= fun_num;
 	writel(val, qm->io_base + QM_IFC_INT_CFG);
 
@@ -3946,6 +3951,139 @@ static int qm_func_shaper_enable(struct hisi_qm *qm, u32 fun_index, u32 qos)
 	return 0;
 }
 
+static u32 qm_get_shaper_vft_qos(struct hisi_qm *qm, u32 fun_index)
+{
+	u64 cir_u = 0, cir_b = 0, cir_s = 0;
+	u64 shaper_vft, ir_calc, ir;
+	unsigned int val;
+	u32 error_rate;
+	int ret;
+
+	ret = readl_relaxed_poll_timeout(qm->io_base + QM_VFT_CFG_RDY, val,
+					 val & BIT(0), POLL_PERIOD,
+					 POLL_TIMEOUT);
+	if (ret)
+		return 0;
+
+	writel(0x1, qm->io_base + QM_VFT_CFG_OP_WR);
+	writel(SHAPER_VFT, qm->io_base + QM_VFT_CFG_TYPE);
+	writel(fun_index, qm->io_base + QM_VFT_CFG);
+
+	writel(0x0, qm->io_base + QM_VFT_CFG_RDY);
+	writel(0x1, qm->io_base + QM_VFT_CFG_OP_ENABLE);
+
+	ret = readl_relaxed_poll_timeout(qm->io_base + QM_VFT_CFG_RDY, val,
+					 val & BIT(0), POLL_PERIOD,
+					 POLL_TIMEOUT);
+	if (ret)
+		return 0;
+
+	shaper_vft = readl(qm->io_base + QM_VFT_CFG_DATA_L) |
+		  ((u64)readl(qm->io_base + QM_VFT_CFG_DATA_H) << 32);
+
+	cir_b = shaper_vft & QM_SHAPER_CIR_B_MASK;
+	cir_u = shaper_vft & QM_SHAPER_CIR_U_MASK;
+	cir_u = cir_u >> QM_SHAPER_FACTOR_CIR_U_SHIFT;
+
+	cir_s = shaper_vft & QM_SHAPER_CIR_S_MASK;
+	cir_s = cir_s >> QM_SHAPER_FACTOR_CIR_S_SHIFT;
+
+	ir_calc = acc_shaper_para_calc(cir_b, cir_u, cir_s);
+
+	ir = qm->factor[fun_index].func_qos * QM_QOS_RATE;
+
+	error_rate = QM_QOS_EXPAND_RATE * (u32)abs(ir_calc - ir) / ir;
+	if (error_rate > QM_QOS_MIN_ERROR_RATE) {
+		pci_err(qm->pdev, "error_rate: %u, get function qos is error!\n", error_rate);
+		return 0;
+	}
+
+	return ir;
+}
+
+static void qm_vf_get_qos(struct hisi_qm *qm, u32 fun_num)
+{
+	struct device *dev = &qm->pdev->dev;
+	u64 mb_cmd;
+	u32 qos;
+	int ret;
+
+	qos = qm_get_shaper_vft_qos(qm, fun_num);
+	if (!qos) {
+		dev_err(dev, "function(%u) failed to get qos by PF!\n", fun_num);
+		return;
+	}
+
+	mb_cmd = QM_PF_SET_QOS | (u64)qos << QM_MB_CMD_DATA_SHIFT;
+	ret = qm_ping_single_vf(qm, mb_cmd, fun_num);
+	if (ret)
+		dev_err(dev, "failed to send cmd to VF(%u)!\n", fun_num);
+}
+
+static int qm_vf_read_qos(struct hisi_qm *qm)
+{
+	int cnt = 0;
+	int ret;
+
+	/* reset mailbox qos val */
+	qm->mb_qos = 0;
+
+	/* vf ping pf to get function qos */
+	if (qm->ops->ping_pf) {
+		ret = qm->ops->ping_pf(qm, QM_VF_GET_QOS);
+		if (ret) {
+			pci_err(qm->pdev, "failed to send cmd to PF to get qos!\n");
+			return ret;
+		}
+	}
+
+	while (true) {
+		msleep(QM_WAIT_DST_ACK);
+		if (qm->mb_qos)
+			break;
+
+		if (++cnt > QM_MAX_VF_WAIT_COUNT) {
+			pci_err(qm->pdev, "PF ping VF timeout!\n");
+			return  -ETIMEDOUT;
+		}
+	}
+
+	return ret;
+}
+
+static ssize_t qm_algqos_read(struct file *filp, char __user *buf,
+			       size_t count, loff_t *pos)
+{
+	struct hisi_qm *qm = filp->private_data;
+	char tbuf[QM_DBG_READ_LEN];
+	u32 qos_val, ir;
+	int ret;
+
+	/* Mailbox and reset cannot be operated at the same time */
+	if (test_and_set_bit(QM_RESETTING, &qm->misc_ctl)) {
+		pci_err(qm->pdev, "dev resetting, read alg qos failed!\n");
+		return  -EAGAIN;
+	}
+
+	if (qm->fun_type == QM_HW_PF) {
+		ir = qm_get_shaper_vft_qos(qm, 0);
+	} else {
+		ret = qm_vf_read_qos(qm);
+		if (ret)
+			goto err_get_status;
+		ir = qm->mb_qos;
+	}
+
+	qos_val = ir / QM_QOS_RATE;
+	ret = scnprintf(tbuf, QM_DBG_READ_LEN, "%u\n", qos_val);
+
+	ret =  simple_read_from_buffer(buf, count, pos, tbuf, ret);
+
+err_get_status:
+	clear_bit(QM_RESETTING, &qm->misc_ctl);
+	return ret;
+}
+
 static ssize_t qm_qos_value_init(const char *buf, unsigned long *val)
 {
 	int buflen = strlen(buf);
@@ -4040,6 +4178,7 @@ static ssize_t qm_algqos_write(struct file *filp, const char __user *buf,
 static const struct file_operations qm_algqos_fops = {
 	.owner = THIS_MODULE,
 	.open = simple_open,
+	.read = qm_algqos_read,
 	.write = qm_algqos_write,
 };
 
@@ -5149,10 +5288,8 @@ static void qm_pf_reset_vf_process(struct hisi_qm *qm,
 	qm_reset_bit_clear(qm);
 }
 
-static void qm_cmd_process(struct work_struct *cmd_process)
+static void qm_handle_cmd_msg(struct hisi_qm *qm, u32 fun_num)
 {
-	struct hisi_qm *qm = container_of(cmd_process,
-					struct hisi_qm, cmd_process);
 	struct device *dev = &qm->pdev->dev;
 	u64 msg;
 	u32 cmd;
@@ -5162,8 +5299,8 @@ static void qm_cmd_process(struct work_struct *cmd_process)
 	 * Get the msg from source by sending mailbox. Whether message is got
 	 * successfully, destination needs to ack source by clearing the interrupt.
 	 */
-	ret = qm_get_mb_cmd(qm, &msg, 0);
-	qm_clear_cmd_interrupt(qm, 0);
+	ret = qm_get_mb_cmd(qm, &msg, fun_num);
+	qm_clear_cmd_interrupt(qm, BIT(fun_num));
 	if (ret) {
 		dev_err(dev, "failed to get msg from source!\n");
 		return;
@@ -5177,12 +5314,42 @@ static void qm_cmd_process(struct work_struct *cmd_process)
 	case QM_PF_SRST_PREPARE:
 		qm_pf_reset_vf_process(qm, QM_SOFT_RESET);
 		break;
+	case QM_VF_GET_QOS:
+		qm_vf_get_qos(qm, fun_num);
+		break;
+	case QM_PF_SET_QOS:
+		qm->mb_qos = msg >> QM_MB_CMD_DATA_SHIFT;
+		break;
 	default:
-		dev_err(dev, "unsupported cmd %u sent by PF!\n", cmd);
+		dev_err(dev, "unsupported cmd %u sent by function(%u)!\n", cmd, fun_num);
 		break;
 	}
 }
 
+static void qm_cmd_process(struct work_struct *cmd_process)
+{
+	struct hisi_qm *qm = container_of(cmd_process,
+					struct hisi_qm, cmd_process);
+	u32 vfs_num = qm->vfs_num;
+	u64 val;
+	u32 i;
+
+	if (qm->fun_type == QM_HW_PF) {
+		val = readq(qm->io_base + QM_IFC_INT_SOURCE_P);
+		if (!val)
+			return;
+
+		for (i = 1; i <= vfs_num; i++) {
+			if (val & BIT(i))
+				qm_handle_cmd_msg(qm, i);
+		}
+
+		return;
+	}
+
+	qm_handle_cmd_msg(qm, 0);
+}
+
 /**
  * hisi_qm_alg_register() - Register alg to crypto and add qm to qm_list.
  * @qm: The qm needs add.
-- 
2.7.4

[PATCH 6/8] crypto: hisilicon/sec - adds the max shaper type rate

[Kai Ye]

The SEC driver support configure each function's QoS in the Host
for Kunpeng930. The SEC driver needs to configure the maximum shaper
type rate.

Signed-off-by: Kai Ye <yekai13@huawei.com>
---
 drivers/crypto/hisilicon/sec2/sec_main.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/drivers/crypto/hisilicon/sec2/sec_main.c b/drivers/crypto/hisilicon/sec2/sec_main.c
index 8ab4e67..d120ce3 100644
--- a/drivers/crypto/hisilicon/sec2/sec_main.c
+++ b/drivers/crypto/hisilicon/sec2/sec_main.c
@@ -98,6 +98,7 @@
 
 #define SEC_SQE_MASK_OFFSET		64
 #define SEC_SQE_MASK_LEN		48
+#define SEC_SHAPER_TYPE_RATE		128
 
 struct sec_hw_error {
 	u32 int_msk;
@@ -874,6 +875,7 @@ static void sec_qm_uninit(struct hisi_qm *qm)
 
 static int sec_probe_init(struct sec_dev *sec)
 {
+	u32 type_rate = SEC_SHAPER_TYPE_RATE;
 	struct hisi_qm *qm = &sec->qm;
 	int ret;
 
@@ -881,6 +883,11 @@ static int sec_probe_init(struct sec_dev *sec)
 		ret = sec_pf_probe_init(sec);
 		if (ret)
 			return ret;
+		/* enable shaper type 0 */
+		if (qm->ver >= QM_HW_V3) {
+			type_rate |= QM_SHAPER_ENABLE;
+			qm->type_rate = type_rate;
+		}
 	}
 
 	return 0;
-- 
2.7.4

[PATCH 7/8] crypto: hisilicon/hpre - adds the max shaper type rate

[Kai Ye]

The HPRE driver support configure each function's QoS in the Host
for Kunpeng930. The HPRE driver needs to configure the maximum shaper
type rate.

Signed-off-by: Kai Ye <yekai13@huawei.com>
---
 drivers/crypto/hisilicon/hpre/hpre_main.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/drivers/crypto/hisilicon/hpre/hpre_main.c b/drivers/crypto/hisilicon/hpre/hpre_main.c
index 37c5296..8b0640f 100644
--- a/drivers/crypto/hisilicon/hpre/hpre_main.c
+++ b/drivers/crypto/hisilicon/hpre/hpre_main.c
@@ -92,6 +92,7 @@
 #define HPRE_QM_PM_FLR			BIT(11)
 #define HPRE_QM_SRIOV_FLR		BIT(12)
 
+#define HPRE_SHAPER_TYPE_RATE		128
 #define HPRE_VIA_MSI_DSM		1
 #define HPRE_SQE_MASK_OFFSET		8
 #define HPRE_SQE_MASK_LEN		24
@@ -947,6 +948,7 @@ static int hpre_pf_probe_init(struct hpre *hpre)
 
 static int hpre_probe_init(struct hpre *hpre)
 {
+	u32 type_rate = HPRE_SHAPER_TYPE_RATE;
 	struct hisi_qm *qm = &hpre->qm;
 	int ret;
 
@@ -954,6 +956,11 @@ static int hpre_probe_init(struct hpre *hpre)
 		ret = hpre_pf_probe_init(hpre);
 		if (ret)
 			return ret;
+		/* Enable shaper type 0 */
+		if (qm->ver >= QM_HW_V3) {
+			type_rate |= QM_SHAPER_ENABLE;
+			qm->type_rate = type_rate;
+		}
 	}
 
 	return 0;
-- 
2.7.4

[PATCH 8/8] crypto: hisilicon/zip - adds the max shaper type rate

[Kai Ye]

The ZIP driver support configure each function's QoS in the Host
for Kunpeng930. The ZIP driver needs to configure the maximum shaper
type rate.

Signed-off-by: Kai Ye <yekai13@huawei.com>
---
 drivers/crypto/hisilicon/zip/zip_main.c | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/drivers/crypto/hisilicon/zip/zip_main.c b/drivers/crypto/hisilicon/zip/zip_main.c
index 9e4c49c..f8482ce 100644
--- a/drivers/crypto/hisilicon/zip/zip_main.c
+++ b/drivers/crypto/hisilicon/zip/zip_main.c
@@ -102,6 +102,8 @@
 #define HZIP_PREFETCH_ENABLE		(~(BIT(26) | BIT(17) | BIT(0)))
 #define HZIP_SVA_PREFETCH_DISABLE	BIT(26)
 #define HZIP_SVA_DISABLE_READY		(BIT(26) | BIT(30))
+#define HZIP_SHAPER_RATE_COMPRESS	252
+#define HZIP_SHAPER_RATE_DECOMPRESS	229
 #define HZIP_DELAY_1_US		1
 #define HZIP_POLL_TIMEOUT_US	1000
 
@@ -823,6 +825,7 @@ static void hisi_zip_qm_uninit(struct hisi_qm *qm)
 
 static int hisi_zip_probe_init(struct hisi_zip *hisi_zip)
 {
+	u32 type_rate = HZIP_SHAPER_RATE_COMPRESS;
 	struct hisi_qm *qm = &hisi_zip->qm;
 	int ret;
 
@@ -830,6 +833,14 @@ static int hisi_zip_probe_init(struct hisi_zip *hisi_zip)
 		ret = hisi_zip_pf_probe_init(hisi_zip);
 		if (ret)
 			return ret;
+		/* enable shaper type 0 */
+		if (qm->ver >= QM_HW_V3) {
+			type_rate |= QM_SHAPER_ENABLE;
+
+			/* ZIP need to enable shaper type 1 */
+			type_rate |= HZIP_SHAPER_RATE_DECOMPRESS << QM_SHAPER_TYPE1_OFFSET;
+			qm->type_rate = type_rate;
+		}
 	}
 
 	return 0;
-- 
2.7.4

Re: [PATCH 0/8] crypto: hisilicon - supports to configure function's QoS for ACC

[Herbert Xu]

On Fri, Jun 11, 2021 at 05:06:42PM +0800, Kai Ye wrote:
> Based on the Token bucket algorithm. The QM can limit the throughput
> of functions in Kunpeng930. By the device debugfs node that named
> "alg_qos", the ACC driver supports to configue each function's QoS
> in the host. Based on the communication between pf and vf, the driver
> supports reading each function's QoS in the host and VM also by the
> debugfs node.
> 
> Kai Ye (8):
>   crypto: hisilicon/qm - supports writing QoS int the Host
>   crypto: hisilicon/qm - add the "alg_qos" file node
>   crypto: hisilicon/qm - merges the work initialization process into a
>     single function
>   crypto: hisilicon/qm - add pf ping single vf function
>   crypto: hisilicon/qm - supports to inquiry each function's QoS
>   crypto: hisilicon/sec - adds the max shaper type rate
>   crypto: hisilicon/hpre - adds the max shaper type rate
>   crypto: hisilicon/zip - adds the max shaper type rate
> 
>  drivers/crypto/hisilicon/hpre/hpre_main.c |   7 +
>  drivers/crypto/hisilicon/qm.c             | 909 +++++++++++++++++++++++-------
>  drivers/crypto/hisilicon/qm.h             |  14 +
>  drivers/crypto/hisilicon/sec2/sec_main.c  |   7 +
>  drivers/crypto/hisilicon/zip/zip_main.c   |  11 +
>  5 files changed, 757 insertions(+), 191 deletions(-)

All applied.  Thanks.
-- 
Email: Herbert Xu <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt