[PATCH] i2c: qcom-geni: Fix GPI DMA buffer sync-back

[PATCH] i2c: qcom-geni: Fix GPI DMA buffer sync-back

[Robin Reckmann]

Fix i2c transfers using GPI DMA mode for all message types that do not set
the I2C_M_DMA_SAFE flag (e.g. SMBus "read byte").

In this case a bounce buffer is returned by i2c_get_dma_safe_msg_buf(),
and it has to synced back to the message after the transfer is done.

Add missing assignment of dma buffer in geni_i2c_gpi().

Set xferred in i2c_put_dma_safe_msg_buf() to true in case of no error to
ensure the sync-back of this dma buffer to the message.

Signed-off-by: Robin Reckmann <robin.reckmann@gmail.com>
---
 drivers/i2c/busses/i2c-qcom-geni.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/drivers/i2c/busses/i2c-qcom-geni.c b/drivers/i2c/busses/i2c-qcom-geni.c
index 6ac402ea58fb..d3541e94794e 100644
--- a/drivers/i2c/busses/i2c-qcom-geni.c
+++ b/drivers/i2c/busses/i2c-qcom-geni.c
@@ -484,12 +484,12 @@ static void geni_i2c_gpi_unmap(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
 {
 	if (tx_buf) {
 		dma_unmap_single(gi2c->se.dev->parent, tx_addr, msg->len, DMA_TO_DEVICE);
-		i2c_put_dma_safe_msg_buf(tx_buf, msg, false);
+		i2c_put_dma_safe_msg_buf(tx_buf, msg, !gi2c->err);
 	}
 
 	if (rx_buf) {
 		dma_unmap_single(gi2c->se.dev->parent, rx_addr, msg->len, DMA_FROM_DEVICE);
-		i2c_put_dma_safe_msg_buf(rx_buf, msg, false);
+		i2c_put_dma_safe_msg_buf(rx_buf, msg, !gi2c->err);
 	}
 }
 
@@ -553,6 +553,7 @@ static int geni_i2c_gpi(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
 	desc->callback_param = gi2c;
 
 	dmaengine_submit(desc);
+	*buf = dma_buf;
 	*dma_addr_p = addr;
 
 	return 0;
-- 
2.25.1

Re: [PATCH] i2c: qcom-geni: Fix GPI DMA buffer sync-back

[Caleb Connolly]

On 07/08/2022 15:04, Robin Reckmann wrote:
> Fix i2c transfers using GPI DMA mode for all message types that do not set
> the I2C_M_DMA_SAFE flag (e.g. SMBus "read byte").
>
> In this case a bounce buffer is returned by i2c_get_dma_safe_msg_buf(),
> and it has to synced back to the message after the transfer is done.
>
> Add missing assignment of dma buffer in geni_i2c_gpi().
>
> Set xferred in i2c_put_dma_safe_msg_buf() to true in case of no error to
> ensure the sync-back of this dma buffer to the message.
Thanks for sending this, it fixes GPI DMA on the PocoPhone F1 and Pixel 3!
>
> Signed-off-by: Robin Reckmann <robin.reckmann@gmail.com>Reviewed-by: Caleb Connolly <caleb@connolly.tech>
> ---
>   drivers/i2c/busses/i2c-qcom-geni.c | 5 +++--
>   1 file changed, 3 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/i2c/busses/i2c-qcom-geni.c b/drivers/i2c/busses/i2c-qcom-geni.c
> index 6ac402ea58fb..d3541e94794e 100644
> --- a/drivers/i2c/busses/i2c-qcom-geni.c
> +++ b/drivers/i2c/busses/i2c-qcom-geni.c
> @@ -484,12 +484,12 @@ static void geni_i2c_gpi_unmap(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
>   {
>   	if (tx_buf) {
>   		dma_unmap_single(gi2c->se.dev->parent, tx_addr, msg->len, DMA_TO_DEVICE);
> -		i2c_put_dma_safe_msg_buf(tx_buf, msg, false);
> +		i2c_put_dma_safe_msg_buf(tx_buf, msg, !gi2c->err);
>   	}
>
>   	if (rx_buf) {
>   		dma_unmap_single(gi2c->se.dev->parent, rx_addr, msg->len, DMA_FROM_DEVICE);
> -		i2c_put_dma_safe_msg_buf(rx_buf, msg, false);
> +		i2c_put_dma_safe_msg_buf(rx_buf, msg, !gi2c->err);
>   	}
>   }
>
> @@ -553,6 +553,7 @@ static int geni_i2c_gpi(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
>   	desc->callback_param = gi2c;
>
>   	dmaengine_submit(desc);
> +	*buf = dma_buf;
>   	*dma_addr_p = addr;
>
>   	return 0;
> --
> 2.25.1
>

--
Kind Regards,
Caleb

Re: [PATCH] i2c: qcom-geni: Fix GPI DMA buffer sync-back

[Wolfram Sang]

[-- Attachment #1: Type: text/plain, Size: 665 bytes --]

On Sun, Aug 07, 2022 at 11:04:54PM +0900, Robin Reckmann wrote:
> Fix i2c transfers using GPI DMA mode for all message types that do not set
> the I2C_M_DMA_SAFE flag (e.g. SMBus "read byte").
> 
> In this case a bounce buffer is returned by i2c_get_dma_safe_msg_buf(),
> and it has to synced back to the message after the transfer is done.
> 
> Add missing assignment of dma buffer in geni_i2c_gpi().
> 
> Set xferred in i2c_put_dma_safe_msg_buf() to true in case of no error to
> ensure the sync-back of this dma buffer to the message.
> 
> Signed-off-by: Robin Reckmann <robin.reckmann@gmail.com>

Thank you! What would be a Fixes tag for this?


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH] i2c: qcom-geni: Fix GPI DMA buffer sync-back

[Luca Weiss]

Hi Robin,

On Sun Aug 7, 2022 at 4:04 PM CEST, Robin Reckmann wrote:
> Fix i2c transfers using GPI DMA mode for all message types that do not set
> the I2C_M_DMA_SAFE flag (e.g. SMBus "read byte").
>
> In this case a bounce buffer is returned by i2c_get_dma_safe_msg_buf(),
> and it has to synced back to the message after the transfer is done.
>
> Add missing assignment of dma buffer in geni_i2c_gpi().
>
> Set xferred in i2c_put_dma_safe_msg_buf() to true in case of no error to
> ensure the sync-back of this dma buffer to the message.
>
> Signed-off-by: Robin Reckmann <robin.reckmann@gmail.com>

This makes I2C with GPI DMA work on sm6350/sm7725 fairphone-fp4!
Thanks for fixing this!

Tested-by: Luca Weiss <luca.weiss@fairphone.com>

Regards
Luca

> ---
>  drivers/i2c/busses/i2c-qcom-geni.c | 5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/i2c/busses/i2c-qcom-geni.c b/drivers/i2c/busses/i2c-qcom-geni.c
> index 6ac402ea58fb..d3541e94794e 100644
> --- a/drivers/i2c/busses/i2c-qcom-geni.c
> +++ b/drivers/i2c/busses/i2c-qcom-geni.c
> @@ -484,12 +484,12 @@ static void geni_i2c_gpi_unmap(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
>  {
>  	if (tx_buf) {
>  		dma_unmap_single(gi2c->se.dev->parent, tx_addr, msg->len, DMA_TO_DEVICE);
> -		i2c_put_dma_safe_msg_buf(tx_buf, msg, false);
> +		i2c_put_dma_safe_msg_buf(tx_buf, msg, !gi2c->err);
>  	}
>  
>  	if (rx_buf) {
>  		dma_unmap_single(gi2c->se.dev->parent, rx_addr, msg->len, DMA_FROM_DEVICE);
> -		i2c_put_dma_safe_msg_buf(rx_buf, msg, false);
> +		i2c_put_dma_safe_msg_buf(rx_buf, msg, !gi2c->err);
>  	}
>  }
>  
> @@ -553,6 +553,7 @@ static int geni_i2c_gpi(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
>  	desc->callback_param = gi2c;
>  
>  	dmaengine_submit(desc);
> +	*buf = dma_buf;
>  	*dma_addr_p = addr;
>  
>  	return 0;
> -- 
> 2.25.1

Re: [PATCH] i2c: qcom-geni: Fix GPI DMA buffer sync-back

[Konrad Dybcio]

On 7.08.2022 16:04, Robin Reckmann wrote:
> Fix i2c transfers using GPI DMA mode for all message types that do not set
> the I2C_M_DMA_SAFE flag (e.g. SMBus "read byte").
> 
> In this case a bounce buffer is returned by i2c_get_dma_safe_msg_buf(),
> and it has to synced back to the message after the transfer is done.
> 
> Add missing assignment of dma buffer in geni_i2c_gpi().
> 
> Set xferred in i2c_put_dma_safe_msg_buf() to true in case of no error to
> ensure the sync-back of this dma buffer to the message.
> 
> Signed-off-by: Robin Reckmann <robin.reckmann@gmail.com>
> ---
Makes SM8450 Xperia 1 IV boot with the touchscreen but enabled (previously
it would simply crash), but the touchscreen itself does not work yet (not
yet sure if something is still missing on my part wrt hw setup):

[    1.838819] gpi 900000.dma-controller: Error in Transaction
[    1.838944] geni_i2c 990000.i2c: DMA txn failed:3
[    1.839166] geni_i2c 990000.i2c: GPI transfer failed: -5


Still, this is a very nice improvement.

Reviewed-by: Konrad Dybcio <konrad.dybcio@somainline.org>

Konrad
>  drivers/i2c/busses/i2c-qcom-geni.c | 5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/i2c/busses/i2c-qcom-geni.c b/drivers/i2c/busses/i2c-qcom-geni.c
> index 6ac402ea58fb..d3541e94794e 100644
> --- a/drivers/i2c/busses/i2c-qcom-geni.c
> +++ b/drivers/i2c/busses/i2c-qcom-geni.c
> @@ -484,12 +484,12 @@ static void geni_i2c_gpi_unmap(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
>  {
>  	if (tx_buf) {
>  		dma_unmap_single(gi2c->se.dev->parent, tx_addr, msg->len, DMA_TO_DEVICE);
> -		i2c_put_dma_safe_msg_buf(tx_buf, msg, false);
> +		i2c_put_dma_safe_msg_buf(tx_buf, msg, !gi2c->err);
>  	}
>  
>  	if (rx_buf) {
>  		dma_unmap_single(gi2c->se.dev->parent, rx_addr, msg->len, DMA_FROM_DEVICE);
> -		i2c_put_dma_safe_msg_buf(rx_buf, msg, false);
> +		i2c_put_dma_safe_msg_buf(rx_buf, msg, !gi2c->err);
>  	}
>  }
>  
> @@ -553,6 +553,7 @@ static int geni_i2c_gpi(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
>  	desc->callback_param = gi2c;
>  
>  	dmaengine_submit(desc);
> +	*buf = dma_buf;
>  	*dma_addr_p = addr;
>  
>  	return 0;

Re: [PATCH] i2c: qcom-geni: Fix GPI DMA buffer sync-back

[Robin Reckmann]

On Mon, 8 Aug 2022, Wolfram Sang wrote:

> On Sun, Aug 07, 2022 at 11:04:54PM +0900, Robin Reckmann wrote:
>> Fix i2c transfers using GPI DMA mode for all message types that do not set
>> the I2C_M_DMA_SAFE flag (e.g. SMBus "read byte").
>>
>> In this case a bounce buffer is returned by i2c_get_dma_safe_msg_buf(),
>> and it has to synced back to the message after the transfer is done.
>>
>> Add missing assignment of dma buffer in geni_i2c_gpi().
>>
>> Set xferred in i2c_put_dma_safe_msg_buf() to true in case of no error to
>> ensure the sync-back of this dma buffer to the message.
>>
>> Signed-off-by: Robin Reckmann <robin.reckmann@gmail.com>
>
> Thank you! What would be a Fixes tag for this?
>
>
Thanks for having a look!

Fixes: d8703554f4de ("i2c: qcom-geni: Add support for GPI DMA")

Re: [PATCH] i2c: qcom-geni: Fix GPI DMA buffer sync-back

[Wolfram Sang]

[-- Attachment #1: Type: text/plain, Size: 650 bytes --]

On Sun, Aug 07, 2022 at 11:04:54PM +0900, Robin Reckmann wrote:
> Fix i2c transfers using GPI DMA mode for all message types that do not set
> the I2C_M_DMA_SAFE flag (e.g. SMBus "read byte").
> 
> In this case a bounce buffer is returned by i2c_get_dma_safe_msg_buf(),
> and it has to synced back to the message after the transfer is done.
> 
> Add missing assignment of dma buffer in geni_i2c_gpi().
> 
> Set xferred in i2c_put_dma_safe_msg_buf() to true in case of no error to
> ensure the sync-back of this dma buffer to the message.
> 
> Signed-off-by: Robin Reckmann <robin.reckmann@gmail.com>

Applied to for-current, thanks!


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]