[PATCH v7 0/2] i2c: i2c-qcom-geni: Add Block event interrupt support

[PATCH v7 0/2] i2c: i2c-qcom-geni: Add Block event interrupt support

[Jyothi Kumar Seerapu]

The I2C driver gets an interrupt upon transfer completion.
When handling multiple messages in a single transfer, this
results in N interrupts for N messages, leading to significant
software interrupt latency.

To mitigate this latency, utilize Block Event Interrupt (BEI)
mechanism. Enabling BEI instructs the hardware to prevent interrupt
generation and BEI is disabled when an interrupt is necessary.

Large I2C transfer can be divided into chunks of messages internally.
Interrupts are not expected for the messages for which BEI bit set,
only the last message triggers an interrupt, indicating the completion of
N messages. This BEI mechanism enhances overall transfer efficiency.

BEI optimizations are currently implemented for I2C write transfers only,
as there is no use case for multiple I2C read messages in a single transfer
at this time.


v6 -> v7:
   - The design has been modified to configure BEI for interrupt
     generation either:
     After the last I2C message, if sufficient TREs are available, or
     After a specific I2C message, when no further TREs are available.
   - dma_buf and dma_addr for multi descriptor support is changed from
     static allocation to dynmic allocation.
   - In i2c_gpi_cb_result function, for multi descriptor case, instead of invoking
     complete for everry 8 messages completions, changed the logic to Invoke 'complete'
     for every I2C callback (for submitted I2C messages).
   - For I2C multi descriptor case, updated 'gi2c_gpi_xfer->dma_buf' and
     'gi2c_gpi_xfer->dma_addr' for unmappping in geni_i2c_gpi_multi_desc_unmap.
   - In the GPI driver, passed the flags argumnetr to the gpi_create_i2c_tre function and
     so avoided using external variables for DMA_PREP_INTERRUPT status.
   - Updated documentation removed for "struct geni_i2c_dev" as per the review comments.

v5 -> v6:
   - Instead of using bei_flag, moved the logic to use with DMA
     supported flags like DMA_PREP_INTERRUPT.
   - Additional parameter comments removed from geni_i2c_gpi_multi_desc_unmap
     function documentation.

v4 -> v5:
   -  BEI flag naming changed from flags to bei_flag.
   -  QCOM_GPI_BLOCK_EVENT_IRQ macro is removed from qcom-gpi-dma.h
      file, and Block event support is checked with bei_flag.
   -  Documentation added for "struct geni_i2c_dev".

v3 -> v4:
  - API's added for Block event interrupt with multi descriptor support is
    moved from qcom-gpi-dma.h file to I2C geni qcom driver file.
  - gpi_multi_xfer_timeout_handler function is moved from GPI driver to
    I2C driver.
  - geni_i2c_gpi_multi_desc_xfer structure is added as a member of
    struct geni_i2c_dev.
  - Removed the changes of making I2C driver is dependent on GPI driver.

v2 -> v3:
  - Updated commit description
  - In I2C GENI driver, for i2c_gpi_cb_result moved the logic of
    "!is_tx_multi_xfer" to else part.
  - MIN_NUM_OF_MSGS_MULTI_DESC changed from 4 to 2
  - Changes of I2C GENI driver to depend on the GPI driver moved
    to patch3.
  - Renamed gpi_multi_desc_process to gpi_multi_xfer_timeout_handler
  - Added description for newly added changes in "qcom-gpi-dma.h" file.

v1 -> v2:
  - DT changes are reverted for adding dma channel size as a new arg of
    dma-cells property.
  - DT binding change reveted for dma channel size as a new arg of
    dma-cells property.
  - In GPI driver, reverted the changes to parse the channel TRE size
    from device tree.
  - Made the changes in QCOM I2C geni driver to support the BEI
    functionality with the existing TRE size of 64.
  - Made changes in QCOM I2C geni driver as per the review comments.
  - Fixed Kernel test robot reported compiltion issues.


Jyothi Kumar Seerapu (2):
  dmaengine: qcom: gpi: Add GPI Block event interrupt support
  i2c: i2c-qcom-geni: Add Block event interrupt support

 drivers/dma/qcom/gpi.c             |  11 +-
 drivers/i2c/busses/i2c-qcom-geni.c | 248 ++++++++++++++++++++++++++---
 2 files changed, 233 insertions(+), 26 deletions(-)

-- 
2.34.1

[PATCH v7 1/2] 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
N interrupts for N messages, leading to significant software
interrupt latency.

To mitigate this latency, utilize Block Event Interrupt (BEI) mechanism.
Enabling BEI instructs the GSI hardware to prevent interrupt generation
and BEI is disabled when an interrupt is necessary.

Large I2C transfer can be divided into chunks of messages internally.
Interrupts are not expected for the messages for which BEI bit set,
only the last message triggers an interrupt, indicating the completion of
N messages. This BEI mechanism enhances overall transfer efficiency.

This BEI mechanism enhances overall transfer efficiency.

Signed-off-by: Jyothi Kumar Seerapu <quic_jseerapu@quicinc.com>
---

v6 -> v7:
   - The design has been modified to configure BEI for interrupt
     generation either:
     After the last I2C message, if sufficient TREs are available, or
     After a specific I2C message, when no further TREs are available.
   - In the GPI driver, passed the flags argumnetr to the gpi_create_i2c_tre function
     and so avoided using external variables for DMA_PREP_INTERRUPT status.

v5 ->v6:
  - For updating the block event interrupt bit, instead of relying on
    bei_flag, decision check is moved with DMA_PREP_INTERRUPT flag.

v4 -> v5:
  - BEI flag naming changed from flags to bei_flag.
  - QCOM_GPI_BLOCK_EVENT_IRQ macro is removed from qcom-gpi-dma.h
    file, and Block event interrupt support is checked with bei_flag.

v3 -> v4:
  - API's added for Block event interrupt with multi descriptor support for
    I2C is moved from qcom-gpi-dma.h file to I2C geni qcom driver file.
  - gpi_multi_xfer_timeout_handler function is moved from GPI driver to
    I2C driver.

v2-> v3:
   - Renamed gpi_multi_desc_process to gpi_multi_xfer_timeout_handler
   - MIN_NUM_OF_MSGS_MULTI_DESC changed from 4 to 2
   - Added documentation for newly added changes in "qcom-gpi-dma.h" file
   - Updated commit description.

v1 -> v2:
   - Changed dma_addr type from array of pointers to array.
   - To support BEI functionality with the TRE size of 64 defined in GPI driver,
     updated QCOM_GPI_MAX_NUM_MSGS to 16 and NUM_MSGS_PER_IRQ to 4.

 drivers/dma/qcom/gpi.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/drivers/dma/qcom/gpi.c b/drivers/dma/qcom/gpi.c
index 8e87738086b2..66bfea1f156d 100644
--- a/drivers/dma/qcom/gpi.c
+++ b/drivers/dma/qcom/gpi.c
@@ -1619,7 +1619,8 @@ gpi_peripheral_config(struct dma_chan *chan, struct dma_slave_config *config)
 }
 
 static int gpi_create_i2c_tre(struct gchan *chan, struct gpi_desc *desc,
-			      struct scatterlist *sgl, enum dma_transfer_direction direction)
+			      struct scatterlist *sgl, enum dma_transfer_direction direction,
+			      unsigned long flags)
 {
 	struct gpi_i2c_config *i2c = chan->config;
 	struct device *dev = chan->gpii->gpi_dev->dev;
@@ -1684,6 +1685,9 @@ 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 (!(flags & DMA_PREP_INTERRUPT))
+			tre->dword[3] |= u32_encode_bits(1, TRE_FLAGS_BEI);
 	}
 
 	for (i = 0; i < tre_idx; i++)
@@ -1827,6 +1831,9 @@ gpi_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
 		return NULL;
 	}
 
+	if (!(flags & DMA_PREP_INTERRUPT) && (nr - nr_tre < 2))
+		return NULL;
+
 	gpi_desc = kzalloc(sizeof(*gpi_desc), GFP_NOWAIT);
 	if (!gpi_desc)
 		return NULL;
@@ -1835,7 +1842,7 @@ gpi_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
 	if (gchan->protocol == QCOM_GPI_SPI) {
 		i = gpi_create_spi_tre(gchan, gpi_desc, sgl, direction);
 	} else if (gchan->protocol == QCOM_GPI_I2C) {
-		i = gpi_create_i2c_tre(gchan, gpi_desc, sgl, direction);
+		i = gpi_create_i2c_tre(gchan, gpi_desc, sgl, direction, flags);
 	} else {
 		dev_err(dev, "invalid peripheral: %d\n", gchan->protocol);
 		kfree(gpi_desc);
-- 
2.34.1

[PATCH v7 2/2] i2c: i2c-qcom-geni: Add Block event interrupt support

[Jyothi Kumar Seerapu]

The I2C driver gets an interrupt upon transfer completion.
When handling multiple messages in a single transfer, this
results in N interrupts for N messages, leading to significant
software interrupt latency.

To mitigate this latency, utilize Block Event Interrupt (BEI)
mechanism. Enabling BEI instructs the hardware to prevent interrupt
generation and BEI is disabled when an interrupt is necessary.

Large I2C transfer can be divided into chunks of messages internally.
Interrupts are not expected for the messages for which BEI bit set,
only the last message triggers an interrupt, indicating the completion of
N messages. This BEI mechanism enhances overall transfer efficiency.

BEI optimizations are currently implemented for I2C write transfers only,
as there is no use case for multiple I2C read messages in a single transfer
at this time.

Signed-off-by: Jyothi Kumar Seerapu <quic_jseerapu@quicinc.com>
---

v6 -> v7:
   - The design has been modified to configure BEI for interrupt
     generation either:
     After the last I2C message, if sufficient TREs are available, or
     After a specific I2C message, when no further TREs are available.
   - dma_buf and dma_addr for multi descriptor support is changed from
     static allocation to dynmic allocation.
   - In i2c_gpi_cb_result function, for multi descriptor case, instead of invoking
     complete for everry 8 messages completions, changed the logic to Invoke 'complete'
     for every I2C callback (for submitted I2C messages).
   - For I2C multi descriptor case, updated 'gi2c_gpi_xfer->dma_buf' and
     'gi2c_gpi_xfer->dma_addr' for unmappping in geni_i2c_gpi_multi_desc_unmap.
   - Updated documentation removed for "struct geni_i2c_dev" as per the review comments.

v5 -> v6:
  - Instead of using bei_flag, moved the logic to use with DMA
    supported flags like DMA_PREP_INTERRUPT.
  - Additional parameter comments removed from geni_i2c_gpi_multi_desc_unmap
    function documentation.

v4 -> v5:
  - Block event interrupt flag naming changed from flags to bei_flag.
  - Documentation added for "struct geni_i2c_dev".

v3 -> v4:
  - API's added for Block event interrupt with multi descriptor support for
    I2C is moved from qcom-gpi-dma.h file to I2C geni qcom driver file.
  - gpi_multi_xfer_timeout_handler function is moved from GPI driver to
    I2C driver.
  - geni_i2c_gpi_multi_desc_xfer structure is added as a member of
    struct geni_i2c_dev.

v2 -> v3:
   - In i2c_gpi_cb_result function, moved the logic of
    "!is_tx_multi_xfer" to else.
   - MIN_NUM_OF_MSGS_MULTI_DESC changed from 4 to 2
   - Updated commit description

v1 -> v2:
   - Moved gi2c_gpi_xfer->msg_idx_cnt to separate local variable.
   - Updated goto labels for error scenarios in geni_i2c_gpi function
   - memset tx_multi_xfer to 0.
   - Removed passing current msg index to geni_i2c_gpi
   - Fixed kernel test robot reported compilation issues.

 drivers/i2c/busses/i2c-qcom-geni.c | 248 ++++++++++++++++++++++++++---
 1 file changed, 224 insertions(+), 24 deletions(-)

diff --git a/drivers/i2c/busses/i2c-qcom-geni.c b/drivers/i2c/busses/i2c-qcom-geni.c
index ff2289b52c84..ab8ad12c140e 100644
--- a/drivers/i2c/busses/i2c-qcom-geni.c
+++ b/drivers/i2c/busses/i2c-qcom-geni.c
@@ -77,6 +77,25 @@ enum geni_i2c_err_code {
 #define XFER_TIMEOUT		HZ
 #define RST_TIMEOUT		HZ
 
+#define QCOM_I2C_MIN_NUM_OF_MSGS_MULTI_DESC	2
+
+/**
+ * struct geni_i2c_gpi_multi_desc_xfer - Structure for multi transfer support
+ *
+ * @msg_idx_cnt: Current message index being processed in the transfer
+ * @unmap_msg_cnt: Number of messages that have been unmapped
+ * @irq_cnt: Number of transfer completion interrupts received
+ * @dma_buf: Array of virtual addresses for DMA-safe buffers
+ * @dma_addr: Array of DMA addresses corresponding to the buffers
+ */
+struct geni_i2c_gpi_multi_desc_xfer {
+	u32 msg_idx_cnt;
+	u32 unmap_msg_cnt;
+	u32 irq_cnt;
+	void **dma_buf;
+	dma_addr_t *dma_addr;
+};
+
 struct geni_i2c_dev {
 	struct geni_se se;
 	u32 tx_wm;
@@ -99,6 +118,9 @@ struct geni_i2c_dev {
 	struct dma_chan *rx_c;
 	bool gpi_mode;
 	bool abort_done;
+	bool is_tx_multi_desc_xfer;
+	u32 num_msgs;
+	struct geni_i2c_gpi_multi_desc_xfer i2c_multi_desc_config;
 };
 
 struct geni_i2c_desc {
@@ -499,6 +521,7 @@ static int geni_i2c_tx_one_msg(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
 static void i2c_gpi_cb_result(void *cb, const struct dmaengine_result *result)
 {
 	struct geni_i2c_dev *gi2c = cb;
+	struct geni_i2c_gpi_multi_desc_xfer *tx_multi_xfer;
 
 	if (result->result != DMA_TRANS_NOERROR) {
 		dev_err(gi2c->se.dev, "DMA txn failed:%d\n", result->result);
@@ -507,6 +530,11 @@ static void i2c_gpi_cb_result(void *cb, const struct dmaengine_result *result)
 		dev_dbg(gi2c->se.dev, "DMA xfer has pending: %d\n", result->residue);
 	}
 
+	if (gi2c->is_tx_multi_desc_xfer) {
+		tx_multi_xfer = &gi2c->i2c_multi_desc_config;
+		tx_multi_xfer->irq_cnt++;
+	}
+
 	complete(&gi2c->done);
 }
 
@@ -525,7 +553,72 @@ static void geni_i2c_gpi_unmap(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
 	}
 }
 
-static int geni_i2c_gpi(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
+/**
+ * geni_i2c_gpi_multi_desc_unmap() - Unmaps DMA buffers post multi message TX transfers
+ * @gi2c: I2C dev handle
+ * @msgs: Array of I2C  messages
+ * @peripheral: Pointer to gpi_i2c_config
+ */
+static void geni_i2c_gpi_multi_desc_unmap(struct geni_i2c_dev *gi2c, struct i2c_msg msgs[],
+					  struct gpi_i2c_config *peripheral)
+{
+	u32 msg_xfer_cnt, wr_idx = 0;
+	struct geni_i2c_gpi_multi_desc_xfer *tx_multi_xfer = &gi2c->i2c_multi_desc_config;
+
+	msg_xfer_cnt = gi2c->err ? tx_multi_xfer->msg_idx_cnt : tx_multi_xfer->irq_cnt;
+
+	/* Unmap the processed DMA buffers based on the received interrupt count */
+	for (; tx_multi_xfer->unmap_msg_cnt < msg_xfer_cnt; tx_multi_xfer->unmap_msg_cnt++) {
+		wr_idx = tx_multi_xfer->unmap_msg_cnt;
+		geni_i2c_gpi_unmap(gi2c, &msgs[wr_idx],
+				   tx_multi_xfer->dma_buf[wr_idx],
+				   tx_multi_xfer->dma_addr[wr_idx],
+				   NULL, (dma_addr_t)NULL);
+
+		if (tx_multi_xfer->unmap_msg_cnt == gi2c->num_msgs - 1) {
+			kfree(tx_multi_xfer->dma_buf);
+			kfree(tx_multi_xfer->dma_addr);
+			break;
+		}
+	}
+}
+
+/**
+ * geni_i2c_gpi_multi_xfer_timeout_handler() - Handles multi message transfer timeout
+ * @dev: Pointer to the corresponding dev node
+ * @multi_xfer: Pointer to the geni_i2c_gpi_multi_desc_xfer
+ * @transfer_timeout_msecs: Timeout value in milliseconds
+ * @transfer_comp: Completion object of the transfer
+ *
+ * This function waits for the completion of each processed transfer messages
+ * based on the interrupts generated upon transfer completion.
+ *
+ * Return: On success returns 0, -ETIMEDOUT on timeout.
+ */
+static int geni_i2c_gpi_multi_xfer_timeout_handler(struct device *dev,
+						   struct geni_i2c_gpi_multi_desc_xfer *multi_xfer,
+						   u32 transfer_timeout_msecs,
+						   struct completion *transfer_comp)
+{
+	int i;
+	u32 time_left;
+
+	for (i = 0; i < multi_xfer->msg_idx_cnt - 1; i++) {
+		reinit_completion(transfer_comp);
+
+		if (multi_xfer->msg_idx_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;
+			}
+		}
+	}
+	return 0;
+}
+
+static int geni_i2c_gpi(struct geni_i2c_dev *gi2c, struct i2c_msg msgs[],
 			struct dma_slave_config *config, dma_addr_t *dma_addr_p,
 			void **buf, unsigned int op, struct dma_chan *dma_chan)
 {
@@ -537,26 +630,45 @@ static int geni_i2c_gpi(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
 	enum dma_transfer_direction dma_dirn;
 	struct dma_async_tx_descriptor *desc;
 	int ret;
+	struct geni_i2c_gpi_multi_desc_xfer *gi2c_gpi_xfer;
+	dma_cookie_t cookie;
+	u32 msg_idx;
 
 	peripheral = config->peripheral_config;
+	gi2c_gpi_xfer = &gi2c->i2c_multi_desc_config;
+	msg_idx = gi2c_gpi_xfer->msg_idx_cnt;
 
-	dma_buf = i2c_get_dma_safe_msg_buf(msg, 1);
-	if (!dma_buf)
-		return -ENOMEM;
+	dma_buf = i2c_get_dma_safe_msg_buf(&msgs[msg_idx], 1);
+	if (!dma_buf) {
+		ret = -ENOMEM;
+		goto out;
+	}
 
 	if (op == I2C_WRITE)
 		map_dirn = DMA_TO_DEVICE;
 	else
 		map_dirn = DMA_FROM_DEVICE;
 
-	addr = dma_map_single(gi2c->se.dev->parent, dma_buf, msg->len, map_dirn);
+	addr = dma_map_single(gi2c->se.dev->parent, dma_buf,
+			      msgs[msg_idx].len, map_dirn);
 	if (dma_mapping_error(gi2c->se.dev->parent, addr)) {
-		i2c_put_dma_safe_msg_buf(dma_buf, msg, false);
-		return -ENOMEM;
+		i2c_put_dma_safe_msg_buf(dma_buf, &msgs[msg_idx], false);
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	if (gi2c->is_tx_multi_desc_xfer) {
+		flags = DMA_CTRL_ACK;
+
+		/* BEI bit to be cleared for last TRE */
+		if (msg_idx == gi2c->num_msgs - 1)
+			flags |= DMA_PREP_INTERRUPT;
+	} else {
+		flags = DMA_PREP_INTERRUPT | DMA_CTRL_ACK;
 	}
 
 	/* set the length as message for rx txn */
-	peripheral->rx_len = msg->len;
+	peripheral->rx_len = msgs[msg_idx].len;
 	peripheral->op = op;
 
 	ret = dmaengine_slave_config(dma_chan, config);
@@ -567,14 +679,21 @@ static int geni_i2c_gpi(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
 
 	peripheral->set_config = 0;
 	peripheral->multi_msg = true;
-	flags = DMA_PREP_INTERRUPT | DMA_CTRL_ACK;
 
 	if (op == I2C_WRITE)
 		dma_dirn = DMA_MEM_TO_DEV;
 	else
 		dma_dirn = DMA_DEV_TO_MEM;
 
-	desc = dmaengine_prep_slave_single(dma_chan, addr, msg->len, dma_dirn, flags);
+	desc = dmaengine_prep_slave_single(dma_chan, addr, msgs[msg_idx].len,
+					   dma_dirn, flags);
+	if (!desc && !(flags & DMA_PREP_INTERRUPT)) {
+		/* Retry with interrupt if not enough TREs */
+		flags |= DMA_PREP_INTERRUPT;
+		desc = dmaengine_prep_slave_single(dma_chan, addr, msgs[msg_idx].len,
+						   dma_dirn, flags);
+	}
+
 	if (!desc) {
 		dev_err(gi2c->se.dev, "prep_slave_sg failed\n");
 		ret = -EIO;
@@ -584,15 +703,48 @@ static int geni_i2c_gpi(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
 	desc->callback_result = i2c_gpi_cb_result;
 	desc->callback_param = gi2c;
 
-	dmaengine_submit(desc);
-	*buf = dma_buf;
-	*dma_addr_p = addr;
+	if (!((msgs[msg_idx].flags & I2C_M_RD) && op == I2C_WRITE))
+		gi2c_gpi_xfer->msg_idx_cnt++;
 
+	cookie = dmaengine_submit(desc);
+	if (dma_submit_error(cookie)) {
+		dev_err(gi2c->se.dev,
+			"%s: dmaengine_submit failed (%d)\n", __func__, cookie);
+		ret = -EINVAL;
+		goto err_config;
+	}
+
+	if (gi2c->is_tx_multi_desc_xfer) {
+		gi2c_gpi_xfer->dma_buf[msg_idx] = dma_buf;
+		gi2c_gpi_xfer->dma_addr[msg_idx] = addr;
+
+		dma_async_issue_pending(gi2c->tx_c);
+
+		if ((msg_idx == (gi2c->num_msgs - 1)) || flags & DMA_PREP_INTERRUPT) {
+			ret = geni_i2c_gpi_multi_xfer_timeout_handler(gi2c->se.dev, gi2c_gpi_xfer,
+								      XFER_TIMEOUT, &gi2c->done);
+			if (ret) {
+				dev_err(gi2c->se.dev,
+					"I2C multi write msg transfer timeout: %d\n",
+					ret);
+				gi2c->err = ret;
+				return ret;
+			}
+		}
+	} else {
+		/* Non multi descriptor message transfer */
+		*buf = dma_buf;
+		*dma_addr_p = addr;
+	}
 	return 0;
 
 err_config:
-	dma_unmap_single(gi2c->se.dev->parent, addr, msg->len, map_dirn);
-	i2c_put_dma_safe_msg_buf(dma_buf, msg, false);
+	dma_unmap_single(gi2c->se.dev->parent, addr,
+			 msgs[msg_idx].len, map_dirn);
+	i2c_put_dma_safe_msg_buf(dma_buf, &msgs[msg_idx], false);
+
+out:
+	gi2c->err = ret;
 	return ret;
 }
 
@@ -604,6 +756,7 @@ static int geni_i2c_gpi_xfer(struct geni_i2c_dev *gi2c, struct i2c_msg msgs[], i
 	unsigned long time_left;
 	dma_addr_t tx_addr, rx_addr;
 	void *tx_buf = NULL, *rx_buf = NULL;
+	struct geni_i2c_gpi_multi_desc_xfer *tx_multi_xfer;
 	const struct geni_i2c_clk_fld *itr = gi2c->clk_fld;
 
 	config.peripheral_config = &peripheral;
@@ -617,6 +770,41 @@ static int geni_i2c_gpi_xfer(struct geni_i2c_dev *gi2c, struct i2c_msg msgs[], i
 	peripheral.set_config = 1;
 	peripheral.multi_msg = false;
 
+	gi2c->num_msgs = num;
+	gi2c->is_tx_multi_desc_xfer = false;
+
+	tx_multi_xfer = &gi2c->i2c_multi_desc_config;
+	memset(tx_multi_xfer, 0, sizeof(struct geni_i2c_gpi_multi_desc_xfer));
+
+	/*
+	 * If number of write messages are two and higher then
+	 * configure hardware for multi descriptor transfers with BEI.
+	 */
+	if (num >= QCOM_I2C_MIN_NUM_OF_MSGS_MULTI_DESC) {
+		gi2c->is_tx_multi_desc_xfer = true;
+		for (i = 0; i < num; i++) {
+			if (msgs[i].flags & I2C_M_RD) {
+				/*
+				 * Multi descriptor transfer with BEI
+				 * support is enabled for write transfers.
+				 * TODO: Add BEI optimization support for
+				 * read transfers later.
+				 */
+				gi2c->is_tx_multi_desc_xfer = false;
+				break;
+			}
+		}
+	}
+
+	if (gi2c->is_tx_multi_desc_xfer) {
+		tx_multi_xfer->dma_buf = kcalloc(num, sizeof(void *), GFP_KERNEL);
+		tx_multi_xfer->dma_addr = kcalloc(num, sizeof(dma_addr_t), GFP_KERNEL);
+		if (!tx_multi_xfer->dma_buf || !tx_multi_xfer->dma_addr) {
+			ret = -ENOMEM;
+			goto err;
+		}
+	}
+
 	for (i = 0; i < num; i++) {
 		gi2c->cur = &msgs[i];
 		gi2c->err = 0;
@@ -627,14 +815,16 @@ static int geni_i2c_gpi_xfer(struct geni_i2c_dev *gi2c, struct i2c_msg msgs[], i
 			peripheral.stretch = 1;
 
 		peripheral.addr = msgs[i].addr;
+		if (i > 0 && (!(msgs[i].flags & I2C_M_RD)))
+			peripheral.multi_msg = false;
 
-		ret =  geni_i2c_gpi(gi2c, &msgs[i], &config,
+		ret =  geni_i2c_gpi(gi2c, msgs, &config,
 				    &tx_addr, &tx_buf, I2C_WRITE, gi2c->tx_c);
 		if (ret)
 			goto err;
 
 		if (msgs[i].flags & I2C_M_RD) {
-			ret =  geni_i2c_gpi(gi2c, &msgs[i], &config,
+			ret =  geni_i2c_gpi(gi2c, msgs, &config,
 					    &rx_addr, &rx_buf, I2C_READ, gi2c->rx_c);
 			if (ret)
 				goto err;
@@ -642,18 +832,24 @@ static int geni_i2c_gpi_xfer(struct geni_i2c_dev *gi2c, struct i2c_msg msgs[], i
 			dma_async_issue_pending(gi2c->rx_c);
 		}
 
-		dma_async_issue_pending(gi2c->tx_c);
-
-		time_left = wait_for_completion_timeout(&gi2c->done, XFER_TIMEOUT);
-		if (!time_left)
-			gi2c->err = -ETIMEDOUT;
+		if (!gi2c->is_tx_multi_desc_xfer) {
+			dma_async_issue_pending(gi2c->tx_c);
+			time_left = wait_for_completion_timeout(&gi2c->done, XFER_TIMEOUT);
+			if (!time_left) {
+				dev_err(gi2c->se.dev, "%s:I2C timeout\n", __func__);
+				gi2c->err = -ETIMEDOUT;
+			}
+		}
 
 		if (gi2c->err) {
 			ret = gi2c->err;
 			goto err;
 		}
 
-		geni_i2c_gpi_unmap(gi2c, &msgs[i], tx_buf, tx_addr, rx_buf, rx_addr);
+		if (!gi2c->is_tx_multi_desc_xfer)
+			geni_i2c_gpi_unmap(gi2c, &msgs[i], tx_buf, tx_addr, rx_buf, rx_addr);
+		else if (tx_multi_xfer->unmap_msg_cnt != tx_multi_xfer->irq_cnt)
+			geni_i2c_gpi_multi_desc_unmap(gi2c, msgs, &peripheral);
 	}
 
 	return num;
@@ -662,7 +858,11 @@ static int geni_i2c_gpi_xfer(struct geni_i2c_dev *gi2c, struct i2c_msg msgs[], i
 	dev_err(gi2c->se.dev, "GPI transfer failed: %d\n", ret);
 	dmaengine_terminate_sync(gi2c->rx_c);
 	dmaengine_terminate_sync(gi2c->tx_c);
-	geni_i2c_gpi_unmap(gi2c, &msgs[i], tx_buf, tx_addr, rx_buf, rx_addr);
+	if (gi2c->is_tx_multi_desc_xfer)
+		geni_i2c_gpi_multi_desc_unmap(gi2c, msgs, &peripheral);
+	else
+		geni_i2c_gpi_unmap(gi2c, &msgs[i], tx_buf, tx_addr, rx_buf, rx_addr);
+
 	return ret;
 }
 
-- 
2.34.1

Re: [PATCH v7 1/2] dmaengine: qcom: gpi: Add GPI Block event interrupt support

[Dmitry Baryshkov]

On Wed, Sep 03, 2025 at 01:00:58PM +0530, Jyothi Kumar Seerapu wrote:
> GSI hardware generates an interrupt for each transfer completion.
> For multiple messages within a single transfer, this results in
> N interrupts for N messages, leading to significant software
> interrupt latency.
> 
> To mitigate this latency, utilize Block Event Interrupt (BEI) mechanism.
> Enabling BEI instructs the GSI hardware to prevent interrupt generation
> and BEI is disabled when an interrupt is necessary.
> 
> Large I2C transfer can be divided into chunks of messages internally.
> Interrupts are not expected for the messages for which BEI bit set,
> only the last message triggers an interrupt, indicating the completion of
> N messages. This BEI mechanism enhances overall transfer efficiency.
> 
> This BEI mechanism enhances overall transfer efficiency.

Duplicate phrase.

> 
> Signed-off-by: Jyothi Kumar Seerapu <quic_jseerapu@quicinc.com>
> ---
> 
> v6 -> v7:
>    - The design has been modified to configure BEI for interrupt
>      generation either:
>      After the last I2C message, if sufficient TREs are available, or
>      After a specific I2C message, when no further TREs are available.
>    - In the GPI driver, passed the flags argumnetr to the gpi_create_i2c_tre function
>      and so avoided using external variables for DMA_PREP_INTERRUPT status.
> 
> v5 ->v6:
>   - For updating the block event interrupt bit, instead of relying on
>     bei_flag, decision check is moved with DMA_PREP_INTERRUPT flag.
> 
> v4 -> v5:
>   - BEI flag naming changed from flags to bei_flag.
>   - QCOM_GPI_BLOCK_EVENT_IRQ macro is removed from qcom-gpi-dma.h
>     file, and Block event interrupt support is checked with bei_flag.
> 
> v3 -> v4:
>   - API's added for Block event interrupt with multi descriptor support for
>     I2C is moved from qcom-gpi-dma.h file to I2C geni qcom driver file.
>   - gpi_multi_xfer_timeout_handler function is moved from GPI driver to
>     I2C driver.
> 
> v2-> v3:
>    - Renamed gpi_multi_desc_process to gpi_multi_xfer_timeout_handler
>    - MIN_NUM_OF_MSGS_MULTI_DESC changed from 4 to 2
>    - Added documentation for newly added changes in "qcom-gpi-dma.h" file
>    - Updated commit description.
> 
> v1 -> v2:
>    - Changed dma_addr type from array of pointers to array.
>    - To support BEI functionality with the TRE size of 64 defined in GPI driver,
>      updated QCOM_GPI_MAX_NUM_MSGS to 16 and NUM_MSGS_PER_IRQ to 4.
> 
>  drivers/dma/qcom/gpi.c | 11 +++++++++--
>  1 file changed, 9 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/dma/qcom/gpi.c b/drivers/dma/qcom/gpi.c
> index 8e87738086b2..66bfea1f156d 100644
> --- a/drivers/dma/qcom/gpi.c
> +++ b/drivers/dma/qcom/gpi.c
> @@ -1619,7 +1619,8 @@ gpi_peripheral_config(struct dma_chan *chan, struct dma_slave_config *config)
>  }
>  
>  static int gpi_create_i2c_tre(struct gchan *chan, struct gpi_desc *desc,
> -			      struct scatterlist *sgl, enum dma_transfer_direction direction)
> +			      struct scatterlist *sgl, enum dma_transfer_direction direction,
> +			      unsigned long flags)
>  {
>  	struct gpi_i2c_config *i2c = chan->config;
>  	struct device *dev = chan->gpii->gpi_dev->dev;
> @@ -1684,6 +1685,9 @@ 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 (!(flags & DMA_PREP_INTERRUPT))
> +			tre->dword[3] |= u32_encode_bits(1, TRE_FLAGS_BEI);
>  	}
>  
>  	for (i = 0; i < tre_idx; i++)
> @@ -1827,6 +1831,9 @@ gpi_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
>  		return NULL;
>  	}
>  
> +	if (!(flags & DMA_PREP_INTERRUPT) && (nr - nr_tre < 2))
> +		return NULL;

Comment in the source file.

> +
>  	gpi_desc = kzalloc(sizeof(*gpi_desc), GFP_NOWAIT);
>  	if (!gpi_desc)
>  		return NULL;
> @@ -1835,7 +1842,7 @@ gpi_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
>  	if (gchan->protocol == QCOM_GPI_SPI) {
>  		i = gpi_create_spi_tre(gchan, gpi_desc, sgl, direction);
>  	} else if (gchan->protocol == QCOM_GPI_I2C) {
> -		i = gpi_create_i2c_tre(gchan, gpi_desc, sgl, direction);
> +		i = gpi_create_i2c_tre(gchan, gpi_desc, sgl, direction, flags);
>  	} else {
>  		dev_err(dev, "invalid peripheral: %d\n", gchan->protocol);
>  		kfree(gpi_desc);
> -- 
> 2.34.1
> 

-- 
With best wishes
Dmitry

Re: [PATCH v7 1/2] dmaengine: qcom: gpi: Add GPI Block event interrupt support

[Jyothi Kumar Seerapu]

On 9/3/2025 2:55 PM, Dmitry Baryshkov wrote:
> On Wed, Sep 03, 2025 at 01:00:58PM +0530, Jyothi Kumar Seerapu wrote:
>> GSI hardware generates an interrupt for each transfer completion.
>> For multiple messages within a single transfer, this results in
>> N interrupts for N messages, leading to significant software
>> interrupt latency.
>>
>> To mitigate this latency, utilize Block Event Interrupt (BEI) mechanism.
>> Enabling BEI instructs the GSI hardware to prevent interrupt generation
>> and BEI is disabled when an interrupt is necessary.
>>
>> Large I2C transfer can be divided into chunks of messages internally.
>> Interrupts are not expected for the messages for which BEI bit set,
>> only the last message triggers an interrupt, indicating the completion of
>> N messages. This BEI mechanism enhances overall transfer efficiency.
>>
>> This BEI mechanism enhances overall transfer efficiency.
> 
> Duplicate phrase.
Thanks, I will remove the duplicate phrase 'This BEI mechanism enhances 
overall transfer efficiency.' in the later patchset.

> 
>>
>> Signed-off-by: Jyothi Kumar Seerapu <quic_jseerapu@quicinc.com>
>> ---
>>
>> v6 -> v7:
>>     - The design has been modified to configure BEI for interrupt
>>       generation either:
>>       After the last I2C message, if sufficient TREs are available, or
>>       After a specific I2C message, when no further TREs are available.
>>     - In the GPI driver, passed the flags argumnetr to the gpi_create_i2c_tre function
>>       and so avoided using external variables for DMA_PREP_INTERRUPT status.
>>
>> v5 ->v6:
>>    - For updating the block event interrupt bit, instead of relying on
>>      bei_flag, decision check is moved with DMA_PREP_INTERRUPT flag.
>>
>> v4 -> v5:
>>    - BEI flag naming changed from flags to bei_flag.
>>    - QCOM_GPI_BLOCK_EVENT_IRQ macro is removed from qcom-gpi-dma.h
>>      file, and Block event interrupt support is checked with bei_flag.
>>
>> v3 -> v4:
>>    - API's added for Block event interrupt with multi descriptor support for
>>      I2C is moved from qcom-gpi-dma.h file to I2C geni qcom driver file.
>>    - gpi_multi_xfer_timeout_handler function is moved from GPI driver to
>>      I2C driver.
>>
>> v2-> v3:
>>     - Renamed gpi_multi_desc_process to gpi_multi_xfer_timeout_handler
>>     - MIN_NUM_OF_MSGS_MULTI_DESC changed from 4 to 2
>>     - Added documentation for newly added changes in "qcom-gpi-dma.h" file
>>     - Updated commit description.
>>
>> v1 -> v2:
>>     - Changed dma_addr type from array of pointers to array.
>>     - To support BEI functionality with the TRE size of 64 defined in GPI driver,
>>       updated QCOM_GPI_MAX_NUM_MSGS to 16 and NUM_MSGS_PER_IRQ to 4.
>>
>>   drivers/dma/qcom/gpi.c | 11 +++++++++--
>>   1 file changed, 9 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/dma/qcom/gpi.c b/drivers/dma/qcom/gpi.c
>> index 8e87738086b2..66bfea1f156d 100644
>> --- a/drivers/dma/qcom/gpi.c
>> +++ b/drivers/dma/qcom/gpi.c
>> @@ -1619,7 +1619,8 @@ gpi_peripheral_config(struct dma_chan *chan, struct dma_slave_config *config)
>>   }
>>   
>>   static int gpi_create_i2c_tre(struct gchan *chan, struct gpi_desc *desc,
>> -			      struct scatterlist *sgl, enum dma_transfer_direction direction)
>> +			      struct scatterlist *sgl, enum dma_transfer_direction direction,
>> +			      unsigned long flags)
>>   {
>>   	struct gpi_i2c_config *i2c = chan->config;
>>   	struct device *dev = chan->gpii->gpi_dev->dev;
>> @@ -1684,6 +1685,9 @@ 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 (!(flags & DMA_PREP_INTERRUPT))
>> +			tre->dword[3] |= u32_encode_bits(1, TRE_FLAGS_BEI);
>>   	}
>>   
>>   	for (i = 0; i < tre_idx; i++)
>> @@ -1827,6 +1831,9 @@ gpi_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
>>   		return NULL;
>>   	}
>>   
>> +	if (!(flags & DMA_PREP_INTERRUPT) && (nr - nr_tre < 2))
>> +		return NULL;
> 
> Comment in the source file.
> 
>> +
>>   	gpi_desc = kzalloc(sizeof(*gpi_desc), GFP_NOWAIT);
>>   	if (!gpi_desc)
>>   		return NULL;
>> @@ -1835,7 +1842,7 @@ gpi_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
>>   	if (gchan->protocol == QCOM_GPI_SPI) {
>>   		i = gpi_create_spi_tre(gchan, gpi_desc, sgl, direction);
>>   	} else if (gchan->protocol == QCOM_GPI_I2C) {
>> -		i = gpi_create_i2c_tre(gchan, gpi_desc, sgl, direction);
>> +		i = gpi_create_i2c_tre(gchan, gpi_desc, sgl, direction, flags);
>>   	} else {
>>   		dev_err(dev, "invalid peripheral: %d\n", gchan->protocol);
>>   		kfree(gpi_desc);
>> -- 
>> 2.34.1
>>
>