[PATCH v1 4/5] dmaengine: qcom: gpi: Add GPI Block event interrupt support

[Jyothi Kumar Seerapu]

GSI hardware generates an interrupt for each transfer completion.
For multiple messages within a single transfer, this results
in receiving N interrupts for N messages, which can introduce
significant software interrupt latency. To mitigate this latency,
utilize Block Event Interrupt (BEI) only when an interrupt is necessary.
When using BEI, consider splitting a single multi-message transfer into
chunks of 64. This approach can enhance overall transfer time and
efficiency.

Signed-off-by: Jyothi Kumar Seerapu <quic_jseerapu@xxxxxxxxxxx>
---
 drivers/dma/qcom/gpi.c           | 49 ++++++++++++++++++++++++++++++++
 include/linux/dma/qcom-gpi-dma.h | 37 ++++++++++++++++++++++++
 2 files changed, 86 insertions(+)

diff --git a/drivers/dma/qcom/gpi.c b/drivers/dma/qcom/gpi.c
index 3c89b4a88ac1..b8ca119114d2 100644
--- a/drivers/dma/qcom/gpi.c
+++ b/drivers/dma/qcom/gpi.c
@@ -1694,6 +1694,9 @@ static int gpi_create_i2c_tre(struct gchan *chan, struct gpi_desc *desc,
 
 		tre->dword[3] = u32_encode_bits(TRE_TYPE_DMA, TRE_FLAGS_TYPE);
 		tre->dword[3] |= u32_encode_bits(1, TRE_FLAGS_IEOT);
+
+		if (i2c->flags & QCOM_GPI_BLOCK_EVENT_IRQ)
+			tre->dword[3] |= u32_encode_bits(1, TRE_FLAGS_BEI);
 	}
 
 	for (i = 0; i < tre_idx; i++)
@@ -2099,6 +2102,52 @@ static int gpi_find_avail_gpii(struct gpi_dev *gpi_dev, u32 seid)
 	return -EIO;
 }
 
+/**
+ * gpi_multi_desc_process() - Process received transfers from GSI HW
+ * @dev: pointer to the corresponding dev node
+ * @multi_xfer: pointer to the gpi_multi_xfer
+ * @num_xfers: total number of transfers
+ * @transfer_timeout_msecs: transfer timeout value
+ * @transfer_comp: completion object of the transfer
+ *
+ * This function is used to process the received transfers based on the
+ * completion events
+ *
+ * Return: On success returns 0, otherwise return error code
+ */
+int gpi_multi_desc_process(struct device *dev, struct gpi_multi_xfer *multi_xfer,
+			   u32 num_xfers, u32 transfer_timeout_msecs,
+			   struct completion *transfer_comp)
+{
+	int i;
+	u32 max_irq_cnt, time_left;
+
+	max_irq_cnt = num_xfers / NUM_MSGS_PER_IRQ;
+	if (num_xfers % NUM_MSGS_PER_IRQ)
+		max_irq_cnt++;
+
+	/*
+	 * Wait for the interrupts of the processed transfers in multiple
+	 * of 64 and for the last transfer. If the hardware is fast and
+	 * already processed all the transfers then no need to wait.
+	 */
+	for (i = 0; i < max_irq_cnt; i++) {
+		reinit_completion(transfer_comp);
+		if (max_irq_cnt != multi_xfer->irq_cnt) {
+			time_left = wait_for_completion_timeout(transfer_comp,
+								transfer_timeout_msecs);
+			if (!time_left) {
+				dev_err(dev, "%s: Transfer timeout\n", __func__);
+				return -ETIMEDOUT;
+			}
+		}
+		if (num_xfers > multi_xfer->msg_idx_cnt)
+			return 0;
+	}
+	return 0;
+}
+EXPORT_SYMBOL_GPL(gpi_multi_desc_process);
+
 /* gpi_of_dma_xlate: open client requested channel */
 static struct dma_chan *gpi_of_dma_xlate(struct of_phandle_args *args,
 					 struct of_dma *of_dma)
diff --git a/include/linux/dma/qcom-gpi-dma.h b/include/linux/dma/qcom-gpi-dma.h
index 6680dd1a43c6..ca0465627a21 100644
--- a/include/linux/dma/qcom-gpi-dma.h
+++ b/include/linux/dma/qcom-gpi-dma.h
@@ -15,6 +15,12 @@ enum spi_transfer_cmd {
 	SPI_DUPLEX,
 };
 
+#define QCOM_GPI_BLOCK_EVENT_IRQ	BIT(0)
+
+#define QCOM_GPI_MAX_NUM_MSGS		200
+#define NUM_MSGS_PER_IRQ		64
+#define MIN_NUM_OF_MSGS_MULTI_DESC	4
+
 /**
  * struct gpi_spi_config - spi config for peripheral
  *
@@ -51,6 +57,29 @@ enum i2c_op {
 	I2C_READ,
 };
 
+/**
+ * struct gpi_multi_xfer - Used for multi transfer support
+ *
+ * @msg_idx_cnt: message index for the transfer
+ * @buf_idx: dma buffer index
+ * @unmap_msg_cnt: unampped transfer index
+ * @freed_msg_cnt: freed transfer index
+ * @irq_cnt: received interrupt count
+ * @irq_msg_cnt: transfer message count for the received irqs
+ * @dma_buf: virtual address of the buffer
+ * @dma_addr: dma address of the buffer
+ */
+struct gpi_multi_xfer {
+	u32 msg_idx_cnt;
+	u32 buf_idx;
+	u32 unmap_msg_cnt;
+	u32 freed_msg_cnt;
+	u32 irq_cnt;
+	u32 irq_msg_cnt;
+	void *dma_buf[QCOM_GPI_MAX_NUM_MSGS];
+	dma_addr_t *dma_addr[QCOM_GPI_MAX_NUM_MSGS];
+};
+
 /**
  * struct gpi_i2c_config - i2c config for peripheral
  *
@@ -65,6 +94,8 @@ enum i2c_op {
  * @rx_len: receive length for buffer
  * @op: i2c cmd
  * @muli-msg: is part of multi i2c r-w msgs
+ * @flags: true for block event interrupt support
+ * @multi_xfer: indicates transfer has multi messages
  */
 struct gpi_i2c_config {
 	u8 set_config;
@@ -78,6 +109,12 @@ struct gpi_i2c_config {
 	u32 rx_len;
 	enum i2c_op op;
 	bool multi_msg;
+	u8 flags;
+	struct gpi_multi_xfer multi_xfer;
 };
 
+int gpi_multi_desc_process(struct device *dev, struct gpi_multi_xfer *multi_xfer,
+			   u32 num_xfers, u32 tranfer_timeout_msecs,
+			   struct completion *transfer_comp);
+
 #endif /* QCOM_GPI_DMA_H */
-- 
2.17.1