i2c: mv64xxx: Implement I2C_M_RECV_LEN and I2C_FUNC_SMBUS_READ_BLOCK_DATA
PATCH 1/1 i2c: mv64xxx: Implement I2C_M_RECV_LEN and I2C_FUNC_SMBUS_READ_BLOCK_DATA
The i2c_mv64xxx driver doesn't implement the I2C_M_REC_LEN function essential to allow blocks with variable length to be read from an i2c
slave.
This is needed to implement the SMBus Read Block Data function.
This patch implements the function by changing the bytes_left and msg len on the fly if the flag is specified.
It has been successfully tested on Allwinner A33 with a special i2c chip that returns variable length blocks on reading.
--- drivers/i2c/busses/i2c-mv64xxx.c | 67 +++++++++++++++++++++++++------- 1 file changed, 53 insertions(+), 14 deletions(-)
drivers/i2c/busses/i2c-mv64xxx.c +53 -14
@@ -128,6 +128,7 @@ struct mv64xxx_i2c_data {u32 addr1;u32 addr2;u32 bytes_left; +u32 effective_length;u32 byte_posn;u32 send_stop;u32 block;
@@ -333,7 +334,18 @@ static void mv64xxx_i2c_send_start(struct mv64xxx_i2c_data *drv_data) { drv_data->msg = drv_data->msgs; drv_data->byte_posn =0 ; - drv_data->bytes_left = drv_data->msg->len; + +/* If we should retrieve the length from the buffer, make sure */ +/* to read enough bytes to avoid sending the */ +/* STOP bit after the read if the first byte */ +if (drv_data->msg->flags &I2C_M_RECV_LEN ) { + drv_data->effective_length =-1 ; + drv_data->bytes_left =3 ; + }else { + drv_data->effective_length = drv_data->msg->len; + drv_data->bytes_left = drv_data->msg->len; + } + drv_data->aborting =0 ; drv_data->rc =0 ;
@@ -342,6 +354,42 @@ static void mv64xxx_i2c_send_start(struct mv64xxx_i2c_data *drv_data) drv_data->reg_base + drv_data->reg_offsets.control); } +static void +mv64xxx_i2c_do_send_stop (struct mv64xxx_i2c_data *drv_data) +{ + drv_data->cntl_bits &= ~MV64XXX_I2C_REG_CONTROL_INTEN ; +writel (drv_data->cntl_bits |MV64XXX_I2C_REG_CONTROL_STOP , + drv_data->reg_base + drv_data->reg_offsets.control); + drv_data->block =0 ; +if (drv_data->errata_delay) +udelay (5 ); + +wake_up (&drv_data->waitq); +} + +static void +mv64xxx_i2c_do_read_data (struct mv64xxx_i2c_data *drv_data) +{ +u8 data; + + data =readl (drv_data->reg_base + drv_data->reg_offsets.data); + drv_data->msg->buf[drv_data->byte_posn++] = data; + +if (drv_data->effective_length ==-1 ) { +/* length=0 should not be allowed, but is indeed possible. +* To avoid locking the chip, we keep reading at least 2 bytes +*/ +if (data <1 ) + data =1 ; + drv_data->effective_length = data+1 ; + drv_data->bytes_left = data+1 ; + drv_data->msg->len = data+1 ; + } + +writel (drv_data->cntl_bits, + drv_data->reg_base + drv_data->reg_offsets.control); +} +static void mv64xxx_i2c_do_action (struct mv64xxx_i2c_data *drv_data) {
@@ -392,23 +440,13 @@ mv64xxx_i2c_do_action(struct mv64xxx_i2c_data *drv_data)break ;case MV64XXX_I2C_ACTION_RCV_DATA : - drv_data->msg->buf[drv_data->byte_posn++] = -readl (drv_data->reg_base + drv_data->reg_offsets.data); -writel (drv_data->cntl_bits, - drv_data->reg_base + drv_data->reg_offsets.control); +mv64xxx_i2c_do_read_data (drv_data);break ;case MV64XXX_I2C_ACTION_RCV_DATA_STOP : drv_data->msg->buf[drv_data->byte_posn++] =readl (drv_data->reg_base + drv_data->reg_offsets.data); - drv_data->cntl_bits &= ~MV64XXX_I2C_REG_CONTROL_INTEN ; -writel (drv_data->cntl_bits |MV64XXX_I2C_REG_CONTROL_STOP , - drv_data->reg_base + drv_data->reg_offsets.control); - drv_data->block =0 ; -if (drv_data->errata_delay) -udelay (5 ); - -wake_up (&drv_data->waitq); +mv64xxx_i2c_do_send_stop (drv_data);break ;case MV64XXX_I2C_ACTION_INVALID :
@@ -706,7 +744,8 @@ mv64xxx_i2c_can_offload(struct mv64xxx_i2c_data *drv_data)static u32 mv64xxx_i2c_functionality (struct i2c_adapter *adap) { -return I2C_FUNC_I2C |I2C_FUNC_10BIT_ADDR |I2C_FUNC_SMBUS_EMUL ; +return I2C_FUNC_I2C |I2C_FUNC_10BIT_ADDR | +I2C_FUNC_SMBUS_READ_BLOCK_DATA |I2C_FUNC_SMBUS_EMUL ; } static int
Signature · 1 line
2.17.1
Quoted 5 lines · Federico Fuga
The
i2c_mv64xxxdriver doesn't implement theI2C_M_REC_LENfunction essential to allow blocks with variable length to be read from an i2c
slave.
This is needed to implement the SMBus Read Block Data function.
Quoted 8 lines
This patch implements the function by changing the
bytes_leftand msg len on the fly if the flag is specified.It has been successfully tested on Allwinner A33 with a special i2c chip that returns variable length blocks on reading.
Signed-off-by Federico Fuga <[email protected]>---
Gregory, any comment? I can't say much about the implementation. In general, this is a nice feature to have...
Quoted 117 lines
drivers/i2c/busses/i2c-mv64xxx.c | 67 +++++++++++++++++++++++++------- 1 file changed, 53 insertions(+), 14 deletions(-)drivers/i2c/busses/i2c-mv64xxx.c +53
-14@@ -128,6 +128,7 @@ struct mv64xxx_i2c_data {u32 addr1;u32 addr2;u32 bytes_left; +u32 effective_length;u32 byte_posn;u32 send_stop;u32 block;@@ -333,7 +334,18 @@ static void mv64xxx_i2c_send_start(struct mv64xxx_i2c_data *drv_data) { drv_data->msg = drv_data->msgs; drv_data->byte_posn =0 ; - drv_data->bytes_left = drv_data->msg->len; + +/* If we should retrieve the length from the buffer, make sure */ +/* to read enough bytes to avoid sending the */ +/* STOP bit after the read if the first byte */ +if (drv_data->msg->flags &I2C_M_RECV_LEN ) { + drv_data->effective_length =-1 ; + drv_data->bytes_left =3 ; + }else { + drv_data->effective_length = drv_data->msg->len; + drv_data->bytes_left = drv_data->msg->len; + } + drv_data->aborting =0 ; drv_data->rc =0 ;@@ -342,6 +354,42 @@ static void mv64xxx_i2c_send_start(struct mv64xxx_i2c_data *drv_data) drv_data->reg_base + drv_data->reg_offsets.control); } +static void +mv64xxx_i2c_do_send_stop (struct mv64xxx_i2c_data *drv_data) +{ + drv_data->cntl_bits &= ~MV64XXX_I2C_REG_CONTROL_INTEN ; +writel (drv_data->cntl_bits |MV64XXX_I2C_REG_CONTROL_STOP , + drv_data->reg_base + drv_data->reg_offsets.control); + drv_data->block =0 ; +if (drv_data->errata_delay) +udelay (5 ); + +wake_up (&drv_data->waitq); +} + +static void +mv64xxx_i2c_do_read_data (struct mv64xxx_i2c_data *drv_data) +{ +u8 data; + + data =readl (drv_data->reg_base + drv_data->reg_offsets.data); + drv_data->msg->buf[drv_data->byte_posn++] = data; + +if (drv_data->effective_length ==-1 ) { +/* length=0 should not be allowed, but is indeed possible. +* To avoid locking the chip, we keep reading at least 2 bytes +*/ +if (data <1 ) + data =1 ; + drv_data->effective_length = data+1 ; + drv_data->bytes_left = data+1 ; + drv_data->msg->len = data+1 ; + } + +writel (drv_data->cntl_bits, + drv_data->reg_base + drv_data->reg_offsets.control); +} +static void mv64xxx_i2c_do_action (struct mv64xxx_i2c_data *drv_data) {@@ -392,23 +440,13 @@ mv64xxx_i2c_do_action(struct mv64xxx_i2c_data *drv_data)break ;case MV64XXX_I2C_ACTION_RCV_DATA : - drv_data->msg->buf[drv_data->byte_posn++] = -readl (drv_data->reg_base + drv_data->reg_offsets.data); -writel (drv_data->cntl_bits, - drv_data->reg_base + drv_data->reg_offsets.control); +mv64xxx_i2c_do_read_data (drv_data);break ;case MV64XXX_I2C_ACTION_RCV_DATA_STOP : drv_data->msg->buf[drv_data->byte_posn++] =readl (drv_data->reg_base + drv_data->reg_offsets.data); - drv_data->cntl_bits &= ~MV64XXX_I2C_REG_CONTROL_INTEN ; -writel (drv_data->cntl_bits |MV64XXX_I2C_REG_CONTROL_STOP , - drv_data->reg_base + drv_data->reg_offsets.control); - drv_data->block =0 ; -if (drv_data->errata_delay) -udelay (5 ); - -wake_up (&drv_data->waitq); +mv64xxx_i2c_do_send_stop (drv_data);break ;case MV64XXX_I2C_ACTION_INVALID :@@ -706,7 +744,8 @@ mv64xxx_i2c_can_offload(struct mv64xxx_i2c_data *drv_data)static u32 mv64xxx_i2c_functionality (struct i2c_adapter *adap) { -return I2C_FUNC_I2C |I2C_FUNC_10BIT_ADDR |I2C_FUNC_SMBUS_EMUL ; +return I2C_FUNC_I2C |I2C_FUNC_10BIT_ADDR | +I2C_FUNC_SMBUS_READ_BLOCK_DATA |I2C_FUNC_SMBUS_EMUL ; } static intSignature · 1 line
2.17.1
Hello Wolfram,
Quoted 9 lines · Federico Fuga
Quoted 8 lines
The
i2c_mv64xxxdriver doesn't implement theI2C_M_REC_LENfunction essential to allow blocks with variable length to be read from an i2c
slave.
This is needed to implement the SMBus Read Block Data function.This patch implements the function by changing the
bytes_leftand msg len on the fly if the flag is specified.
Quoted 8 lines
Quoted 5 lines
It has been successfully tested on Allwinner A33 with a special i2c chip that returns variable length blocks on reading.
Signed-off-by Federico Fuga <[email protected]>---
Gregory, any comment? I can't say much about the implementation. In general, this is a nice feature to have...
I thought I was already sent a reviewed by, as I reviewed this patch before Frederico actually sent it.
So you can even add a
Thanks,
Gregory
Quoted 118 lines
Quoted 118 lines
drivers/i2c/busses/i2c-mv64xxx.c | 67 +++++++++++++++++++++++++------- 1 file changed, 53 insertions(+), 14 deletions(-)drivers/i2c/busses/i2c-mv64xxx.c +53
-14@@ -128,6 +128,7 @@ struct mv64xxx_i2c_data {u32 addr1;u32 addr2;u32 bytes_left; +u32 effective_length;u32 byte_posn;u32 send_stop;u32 block;@@ -333,7 +334,18 @@ static void mv64xxx_i2c_send_start(struct mv64xxx_i2c_data *drv_data) { drv_data->msg = drv_data->msgs; drv_data->byte_posn =0 ; - drv_data->bytes_left = drv_data->msg->len; + +/* If we should retrieve the length from the buffer, make sure */ +/* to read enough bytes to avoid sending the */ +/* STOP bit after the read if the first byte */ +if (drv_data->msg->flags &I2C_M_RECV_LEN ) { + drv_data->effective_length =-1 ; + drv_data->bytes_left =3 ; + }else { + drv_data->effective_length = drv_data->msg->len; + drv_data->bytes_left = drv_data->msg->len; + } + drv_data->aborting =0 ; drv_data->rc =0 ;@@ -342,6 +354,42 @@ static void mv64xxx_i2c_send_start(struct mv64xxx_i2c_data *drv_data) drv_data->reg_base + drv_data->reg_offsets.control); } +static void +mv64xxx_i2c_do_send_stop (struct mv64xxx_i2c_data *drv_data) +{ + drv_data->cntl_bits &= ~MV64XXX_I2C_REG_CONTROL_INTEN ; +writel (drv_data->cntl_bits |MV64XXX_I2C_REG_CONTROL_STOP , + drv_data->reg_base + drv_data->reg_offsets.control); + drv_data->block =0 ; +if (drv_data->errata_delay) +udelay (5 ); + +wake_up (&drv_data->waitq); +} + +static void +mv64xxx_i2c_do_read_data (struct mv64xxx_i2c_data *drv_data) +{ +u8 data; + + data =readl (drv_data->reg_base + drv_data->reg_offsets.data); + drv_data->msg->buf[drv_data->byte_posn++] = data; + +if (drv_data->effective_length ==-1 ) { +/* length=0 should not be allowed, but is indeed possible. +* To avoid locking the chip, we keep reading at least 2 bytes +*/ +if (data <1 ) + data =1 ; + drv_data->effective_length = data+1 ; + drv_data->bytes_left = data+1 ; + drv_data->msg->len = data+1 ; + } + +writel (drv_data->cntl_bits, + drv_data->reg_base + drv_data->reg_offsets.control); +} +static void mv64xxx_i2c_do_action (struct mv64xxx_i2c_data *drv_data) {@@ -392,23 +440,13 @@ mv64xxx_i2c_do_action(struct mv64xxx_i2c_data *drv_data)break ;case MV64XXX_I2C_ACTION_RCV_DATA : - drv_data->msg->buf[drv_data->byte_posn++] = -readl (drv_data->reg_base + drv_data->reg_offsets.data); -writel (drv_data->cntl_bits, - drv_data->reg_base + drv_data->reg_offsets.control); +mv64xxx_i2c_do_read_data (drv_data);break ;case MV64XXX_I2C_ACTION_RCV_DATA_STOP : drv_data->msg->buf[drv_data->byte_posn++] =readl (drv_data->reg_base + drv_data->reg_offsets.data); - drv_data->cntl_bits &= ~MV64XXX_I2C_REG_CONTROL_INTEN ; -writel (drv_data->cntl_bits |MV64XXX_I2C_REG_CONTROL_STOP , - drv_data->reg_base + drv_data->reg_offsets.control); - drv_data->block =0 ; -if (drv_data->errata_delay) -udelay (5 ); - -wake_up (&drv_data->waitq); +mv64xxx_i2c_do_send_stop (drv_data);break ;case MV64XXX_I2C_ACTION_INVALID :@@ -706,7 +744,8 @@ mv64xxx_i2c_can_offload(struct mv64xxx_i2c_data *drv_data)static u32 mv64xxx_i2c_functionality (struct i2c_adapter *adap) { -return I2C_FUNC_I2C |I2C_FUNC_10BIT_ADDR |I2C_FUNC_SMBUS_EMUL ; +return I2C_FUNC_I2C |I2C_FUNC_10BIT_ADDR | +I2C_FUNC_SMBUS_READ_BLOCK_DATA |I2C_FUNC_SMBUS_EMUL ; } static intSignature · 1 line
2.17.1
Signature · 3 lines
Gregory Clement, Bootlin Embedded Linux and Kernel engineering http://bootlin.com
Hello all,
Quoted 22 lines
Il giorno 13 mar 2020, alle ore 21:48, Gregory CLEMENT <[email protected]> ha scritto:
Hello Wolfram,
Quoted 9 lines · Federico Fuga
Quoted 8 lines
The
i2c_mv64xxxdriver doesn't implement theI2C_M_REC_LENfunction essential to allow blocks with variable length to be read from an i2c slave.
This is needed to implement the SMBus Read Block Data function.This patch implements the function by changing the
bytes_leftand msg len on the fly if the flag is specified.Quoted 8 lines
Quoted 5 lines
It has been successfully tested on Allwinner A33 with a special i2c chip that returns variable length blocks on reading.
Signed-off-by Federico Fuga <[email protected]>---
Gregory, any comment? I can't say much about the implementation. In general, this is a nice feature to have...
Quoted 8 lines
I thought I was already sent a reviewed by, as I reviewed this patch before Frederico actually sent it.
So you can even add a
Acked-by Gregory CLEMENT <[email protected]>Thanks,
For what it worth, I can confirm that after many weeks of use the patch works as expected without any issue.
Federico
drivers/i2c/busses/i2c-mv64xxx.c static void mv64xxx_i2c_send_start(struct mv64xxx_i2c_data *drv_data) · line 340
Quoted 4 lines
drivers/i2c/busses/i2c-mv64xxx.c +3
-0+/* If we should retrieve the length from the buffer, make sure */ +/* to read enough bytes to avoid sending the */ +/* STOP bit after the read if the first byte */
Please stick to Kernel commenting style.
drivers/i2c/busses/i2c-mv64xxx.c static void mv64xxx_i2c_send_start(struct mv64xxx_i2c_data *drv_data) · line 381
Quoted 3 lines
drivers/i2c/busses/i2c-mv64xxx.c +3
-0+/* length=0 should not be allowed, but is indeed possible. +* To avoid locking the chip, we keep reading at least 2 bytes +*/
Ditto.
drivers/i2c/busses/i2c-mv64xxx.c static void mv64xxx_i2c_send_start(struct mv64xxx_i2c_data *drv_data) · line 386
Quoted 3 lines
drivers/i2c/busses/i2c-mv64xxx.c +3
-0+ drv_data->effective_length = data+1 ; + drv_data->bytes_left = data+1 ; + drv_data->msg->len = data+1 ;
Also, checkpatch complains about this if you run it with "--strict'
CHECKPATCH
CHECK: spaces preferred around that '+' (ctx:VxV) #81: FILE: drivers/i2c/busses/i2c-mv64xxx.c:384:
+ drv_data->effective_length = data+1;
^
CHECK: spaces preferred around that '+' (ctx:VxV)
#82: FILE: drivers/i2c/busses/i2c-mv64xxx.c:385:
+ drv_data->bytes_left = data+1;
^
CHECK: spaces preferred around that '+' (ctx:VxV)
#83: FILE: drivers/i2c/busses/i2c-mv64xxx.c:386:
+ drv_data->msg->len = data+1;
^
I think the complaints make sense.
drivers/i2c/busses/i2c-mv64xxx.c mv64xxx_i2c_can_offload(struct mv64xxx_i2c_data *drv_data) · line 748
Quoted 3 lines
drivers/i2c/busses/i2c-mv64xxx.c +2
-1-return I2C_FUNC_I2C |I2C_FUNC_10BIT_ADDR |I2C_FUNC_SMBUS_EMUL ; +return I2C_FUNC_I2C |I2C_FUNC_10BIT_ADDR | +I2C_FUNC_SMBUS_READ_BLOCK_DATA |I2C_FUNC_SMBUS_EMUL ;
Likely I2C_FUNC_SMBUS_BLOCK_PROC_CALL will work as well, but I assume you can't test it?
Otherwise, looks good.