2019-09-26 10:21:21

by Rayagonda Kokatanur

[permalink] [raw]
Subject: [PATCH v2 1/1] i2c: iproc: Add i2c repeated start capability

From: Lori Hikichi <[email protected]>

Enable handling of i2c repeated start. The current code
handles a multi msg i2c transfer as separate i2c bus
transactions. This change will now handle this case
using the i2c repeated start protocol. The number of msgs
in a transfer is limited to two, and must be a write
followed by a read.

Signed-off-by: Lori Hikichi <[email protected]>
Signed-off-by: Rayagonda Kokatanur <[email protected]>
Signed-off-by: Icarus Chau <[email protected]>
Signed-off-by: Ray Jui <[email protected]>
Signed-off-by: Shivaraj Shetty <[email protected]>
---
changes from v1:
- Address code review comment from Wolfram Sang

drivers/i2c/busses/i2c-bcm-iproc.c | 63 ++++++++++++++++++++++++++++++--------
1 file changed, 50 insertions(+), 13 deletions(-)

diff --git a/drivers/i2c/busses/i2c-bcm-iproc.c b/drivers/i2c/busses/i2c-bcm-iproc.c
index d7fd76b..e478db7 100644
--- a/drivers/i2c/busses/i2c-bcm-iproc.c
+++ b/drivers/i2c/busses/i2c-bcm-iproc.c
@@ -81,6 +81,7 @@
#define M_CMD_PROTOCOL_MASK 0xf
#define M_CMD_PROTOCOL_BLK_WR 0x7
#define M_CMD_PROTOCOL_BLK_RD 0x8
+#define M_CMD_PROTOCOL_PROCESS 0xa
#define M_CMD_PEC_SHIFT 8
#define M_CMD_RD_CNT_SHIFT 0
#define M_CMD_RD_CNT_MASK 0xff
@@ -675,13 +676,20 @@ static int bcm_iproc_i2c_xfer_wait(struct bcm_iproc_i2c_dev *iproc_i2c,
return 0;
}

-static int bcm_iproc_i2c_xfer_single_msg(struct bcm_iproc_i2c_dev *iproc_i2c,
- struct i2c_msg *msg)
+/*
+ * If 'process_call' is true, then this is a multi-msg transfer that requires
+ * a repeated start between the messages.
+ * More specifically, it must be a write (reg) followed by a read (data).
+ * The i2c quirks are set to enforce this rule.
+ */
+static int bcm_iproc_i2c_xfer_internal(struct bcm_iproc_i2c_dev *iproc_i2c,
+ struct i2c_msg *msgs, bool process_call)
{
int i;
u8 addr;
u32 val, tmp, val_intr_en;
unsigned int tx_bytes;
+ struct i2c_msg *msg = &msgs[0];

/* check if bus is busy */
if (!!(iproc_i2c_rd_reg(iproc_i2c,
@@ -707,14 +715,29 @@ static int bcm_iproc_i2c_xfer_single_msg(struct bcm_iproc_i2c_dev *iproc_i2c,
val = msg->buf[i];

/* mark the last byte */
- if (i == msg->len - 1)
- val |= BIT(M_TX_WR_STATUS_SHIFT);
+ if (!process_call && (i == msg->len - 1))
+ val |= 1 << M_TX_WR_STATUS_SHIFT;

iproc_i2c_wr_reg(iproc_i2c, M_TX_OFFSET, val);
}
iproc_i2c->tx_bytes = tx_bytes;
}

+ /* Process the read message if this is process call */
+ if (process_call) {
+ msg++;
+ iproc_i2c->msg = msg; /* point to second msg */
+
+ /*
+ * The last byte to be sent out should be a slave
+ * address with read operation
+ */
+ addr = i2c_8bit_addr_from_msg(msg);
+ /* mark it the last byte out */
+ val = addr | (1 << M_TX_WR_STATUS_SHIFT);
+ iproc_i2c_wr_reg(iproc_i2c, M_TX_OFFSET, val);
+ }
+
/* mark as incomplete before starting the transaction */
if (iproc_i2c->irq)
reinit_completion(&iproc_i2c->done);
@@ -733,7 +756,7 @@ static int bcm_iproc_i2c_xfer_single_msg(struct bcm_iproc_i2c_dev *iproc_i2c,
* underrun interrupt, which will be triggerred when the TX FIFO is
* empty. When that happens we can then pump more data into the FIFO
*/
- if (!(msg->flags & I2C_M_RD) &&
+ if (!process_call && !(msg->flags & I2C_M_RD) &&
msg->len > iproc_i2c->tx_bytes)
val_intr_en |= BIT(IE_M_TX_UNDERRUN_SHIFT);

@@ -743,6 +766,8 @@ static int bcm_iproc_i2c_xfer_single_msg(struct bcm_iproc_i2c_dev *iproc_i2c,
*/
val = BIT(M_CMD_START_BUSY_SHIFT);
if (msg->flags & I2C_M_RD) {
+ u32 protocol;
+
iproc_i2c->rx_bytes = 0;
if (msg->len > M_RX_FIFO_MAX_THLD_VALUE)
iproc_i2c->thld_bytes = M_RX_FIFO_THLD_VALUE;
@@ -758,7 +783,10 @@ static int bcm_iproc_i2c_xfer_single_msg(struct bcm_iproc_i2c_dev *iproc_i2c,
/* enable the RX threshold interrupt */
val_intr_en |= BIT(IE_M_RX_THLD_SHIFT);

- val |= (M_CMD_PROTOCOL_BLK_RD << M_CMD_PROTOCOL_SHIFT) |
+ protocol = process_call ?
+ M_CMD_PROTOCOL_PROCESS : M_CMD_PROTOCOL_BLK_RD;
+
+ val |= (protocol << M_CMD_PROTOCOL_SHIFT) |
(msg->len << M_CMD_RD_CNT_SHIFT);
} else {
val |= (M_CMD_PROTOCOL_BLK_WR << M_CMD_PROTOCOL_SHIFT);
@@ -774,17 +802,24 @@ static int bcm_iproc_i2c_xfer(struct i2c_adapter *adapter,
struct i2c_msg msgs[], int num)
{
struct bcm_iproc_i2c_dev *iproc_i2c = i2c_get_adapdata(adapter);
- int ret, i;
+ bool process_call = false;
+ int ret;

- /* go through all messages */
- for (i = 0; i < num; i++) {
- ret = bcm_iproc_i2c_xfer_single_msg(iproc_i2c, &msgs[i]);
- if (ret) {
- dev_dbg(iproc_i2c->device, "xfer failed\n");
- return ret;
+ if (num == 2) {
+ /* Repeated start, use process call */
+ process_call = true;
+ if (msgs[1].flags & I2C_M_NOSTART) {
+ dev_err(iproc_i2c->device, "Invalid repeated start\n");
+ return -EOPNOTSUPP;
}
}

+ ret = bcm_iproc_i2c_xfer_internal(iproc_i2c, msgs, process_call);
+ if (ret) {
+ dev_dbg(iproc_i2c->device, "xfer failed\n");
+ return ret;
+ }
+
return num;
}

@@ -806,6 +841,8 @@ static uint32_t bcm_iproc_i2c_functionality(struct i2c_adapter *adap)
};

static struct i2c_adapter_quirks bcm_iproc_i2c_quirks = {
+ .flags = I2C_AQ_COMB_WRITE_THEN_READ,
+ .max_comb_1st_msg_len = M_TX_RX_FIFO_SIZE,
.max_read_len = M_RX_MAX_READ_LEN,
};

--
1.9.1


2019-09-28 18:24:31

by Wolfram Sang

[permalink] [raw]
Subject: Re: [PATCH v2 1/1] i2c: iproc: Add i2c repeated start capability

On Thu, Sep 26, 2019 at 10:10:08AM +0530, Rayagonda Kokatanur wrote:
> From: Lori Hikichi <[email protected]>
>
> Enable handling of i2c repeated start. The current code
> handles a multi msg i2c transfer as separate i2c bus
> transactions. This change will now handle this case
> using the i2c repeated start protocol. The number of msgs
> in a transfer is limited to two, and must be a write
> followed by a read.
>
> Signed-off-by: Lori Hikichi <[email protected]>
> Signed-off-by: Rayagonda Kokatanur <[email protected]>
> Signed-off-by: Icarus Chau <[email protected]>
> Signed-off-by: Ray Jui <[email protected]>
> Signed-off-by: Shivaraj Shetty <[email protected]>
> ---
> changes from v1:
> - Address code review comment from Wolfram Sang

No, sorry, this is not a proper changelog. I review so many patches, I
can't recall what I suggested to do for every patch. Please describe
what changes you actually made. It is also better when digging through
mail archives.


Attachments:
(No filename) (1.02 kB)
signature.asc (849.00 B)
Download all attachments

2019-09-30 06:56:52

by Rayagonda Kokatanur

[permalink] [raw]
Subject: Re: [PATCH v2 1/1] i2c: iproc: Add i2c repeated start capability

On Sat, Sep 28, 2019 at 11:53 PM Wolfram Sang <[email protected]> wrote:
>
> On Thu, Sep 26, 2019 at 10:10:08AM +0530, Rayagonda Kokatanur wrote:
> > From: Lori Hikichi <[email protected]>
> >
> > Enable handling of i2c repeated start. The current code
> > handles a multi msg i2c transfer as separate i2c bus
> > transactions. This change will now handle this case
> > using the i2c repeated start protocol. The number of msgs
> > in a transfer is limited to two, and must be a write
> > followed by a read.
> >
> > Signed-off-by: Lori Hikichi <[email protected]>
> > Signed-off-by: Rayagonda Kokatanur <[email protected]>
> > Signed-off-by: Icarus Chau <[email protected]>
> > Signed-off-by: Ray Jui <[email protected]>
> > Signed-off-by: Shivaraj Shetty <[email protected]>
> > ---
> > changes from v1:
> > - Address code review comment from Wolfram Sang
>
> No, sorry, this is not a proper changelog. I review so many patches, I
> can't recall what I suggested to do for every patch. Please describe
> what changes you actually made. It is also better when digging through
> mail archives.
>
Sorry for inconvenience, I updated the changelog and resent the patch.
I have kept the patch version as v2 only. Hope that is fine.

Best regards,
Rayagonda