[PATCH] dmaengine: qcom: gpi: Set link_rx bit on GO TRE for rx operation

[PATCH] dmaengine: qcom: gpi: Set link_rx bit on GO TRE for rx operation

[Vijaya Krishna Nivarthi]

As per GSI spec, link_rx bit is to be set on GO TRE on tx
channel whenever there is going to be a DMA TRE on rx
channel. This is currently set for duplex operation only.

Set the bit for rx operation as well.

Signed-off-by: Vijaya Krishna Nivarthi <quic_vnivarth@quicinc.com>
---
 drivers/dma/qcom/gpi.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/dma/qcom/gpi.c b/drivers/dma/qcom/gpi.c
index 061add8..59a36cb 100644
--- a/drivers/dma/qcom/gpi.c
+++ b/drivers/dma/qcom/gpi.c
@@ -1756,6 +1756,7 @@ static int gpi_create_spi_tre(struct gchan *chan, struct gpi_desc *desc,
 		tre->dword[3] = u32_encode_bits(TRE_TYPE_GO, TRE_FLAGS_TYPE);
 		if (spi->cmd == SPI_RX) {
 			tre->dword[3] |= u32_encode_bits(1, TRE_FLAGS_IEOB);
+			tre->dword[3] |= u32_encode_bits(1, TRE_FLAGS_LINK);
 		} else if (spi->cmd == SPI_TX) {
 			tre->dword[3] |= u32_encode_bits(1, TRE_FLAGS_CHAIN);
 		} else { /* SPI_DUPLEX */
-- 
Qualcomm INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by the Linux Foundation.

Re: [PATCH] dmaengine: qcom: gpi: Set link_rx bit on GO TRE for rx operation

[Doug Anderson]

Hi,

On Wed, Nov 30, 2022 at 4:20 AM Vijaya Krishna Nivarthi
<quic_vnivarth@quicinc.com> wrote:
>
> As per GSI spec, link_rx bit is to be set on GO TRE on tx
> channel whenever there is going to be a DMA TRE on rx
> channel. This is currently set for duplex operation only.
>
> Set the bit for rx operation as well.
>
> Signed-off-by: Vijaya Krishna Nivarthi <quic_vnivarth@quicinc.com>
> ---
>  drivers/dma/qcom/gpi.c | 1 +
>  1 file changed, 1 insertion(+)

I don't feel qualified to actually give this a review since I don't
know anything about the details of GSI/GPI. It seems simple enough so
I'll just assume that Bjorn will land it. Ideally someone else at
Qualcomm would give you a Reviewed-by tag.

One drive-by comment, though, is that I would say that your patch
description lacks an answer to the question: "So what?"

In other words, what is broken today? Does everything work fine today
but some bit counter looked over your shoulder and told you that you
were a bad person for not setting that bit? Did the lunar lander catch
fire (despite the lack of Oxygen on the moon!) because it started
using the RX transfer mode to talk to its fuel valve system and the RX
transfer mode never worked? ...or maybe everything today works but the
super secret Qualcomm SDM9002 (shhhh!) chip needs this bit set? Help
people looking at your patch be able to decide if it's important for
them to pick to their kernel tree! :-)

-Doug

Re: [PATCH] dmaengine: qcom: gpi: Set link_rx bit on GO TRE for rx operation

[Vijaya Krishna Nivarthi]

On 12/2/2022 4:17 AM, Doug Anderson wrote:
> Hi,
>
> On Wed, Nov 30, 2022 at 4:20 AM Vijaya Krishna Nivarthi
> <quic_vnivarth@quicinc.com> wrote:
>> As per GSI spec, link_rx bit is to be set on GO TRE on tx
>> channel whenever there is going to be a DMA TRE on rx
>> channel. This is currently set for duplex operation only.
>>
>> Set the bit for rx operation as well.
>>
>> Signed-off-by: Vijaya Krishna Nivarthi <quic_vnivarth@quicinc.com>
>> ---
>>   drivers/dma/qcom/gpi.c | 1 +
>>   1 file changed, 1 insertion(+)
> I don't feel qualified to actually give this a review since I don't
> know anything about the details of GSI/GPI. It seems simple enough so
> I'll just assume that Bjorn will land it. Ideally someone else at
> Qualcomm would give you a Reviewed-by tag.
>
> One drive-by comment, though, is that I would say that your patch
> description lacks an answer to the question: "So what?"
>
> In other words, what is broken today? Does everything work fine today
> but some bit counter looked over your shoulder and told you that you
> were a bad person for not setting that bit? Did the lunar lander catch
> fire (despite the lack of Oxygen on the moon!) because it started
> using the RX transfer mode to talk to its fuel valve system and the RX
> transfer mode never worked? ...or maybe everything today works but the
> super secret Qualcomm SDM9002 (shhhh!) chip needs this bit set? Help
> people looking at your patch be able to decide if it's important for
> them to pick to their kernel tree! :-)

:-) Thank you very much for the input. Will keep that in mind for next.


Some info...

rx is actually broken for spi gsi. A plain rx operation would crash 
right now and patch for same would come next; There are 3 further 
patches as well for gsi.

I started with simplest patch hoping it would get in quick.

While trying to bring up EC use case with spi gsi, we made some changes 
to conform to GSI spec, made some progress but still couldn't get it to 
work.

In the meantime we switched to SE DMA as interim.

> -Doug

Re: [PATCH] dmaengine: qcom: gpi: Set link_rx bit on GO TRE for rx operation

[Vijaya Krishna Nivarthi]

Gentle reminder to review/approve please.

Thank you...


On 11/30/2022 5:50 PM, Vijaya Krishna Nivarthi wrote:
> As per GSI spec, link_rx bit is to be set on GO TRE on tx
> channel whenever there is going to be a DMA TRE on rx
> channel. This is currently set for duplex operation only.
>
> Set the bit for rx operation as well.
>
> Signed-off-by: Vijaya Krishna Nivarthi <quic_vnivarth@quicinc.com>
> ---
>   drivers/dma/qcom/gpi.c | 1 +
>   1 file changed, 1 insertion(+)
>
> diff --git a/drivers/dma/qcom/gpi.c b/drivers/dma/qcom/gpi.c
> index 061add8..59a36cb 100644
> --- a/drivers/dma/qcom/gpi.c
> +++ b/drivers/dma/qcom/gpi.c
> @@ -1756,6 +1756,7 @@ static int gpi_create_spi_tre(struct gchan *chan, struct gpi_desc *desc,
>   		tre->dword[3] = u32_encode_bits(TRE_TYPE_GO, TRE_FLAGS_TYPE);
>   		if (spi->cmd == SPI_RX) {
>   			tre->dword[3] |= u32_encode_bits(1, TRE_FLAGS_IEOB);
> +			tre->dword[3] |= u32_encode_bits(1, TRE_FLAGS_LINK);
>   		} else if (spi->cmd == SPI_TX) {
>   			tre->dword[3] |= u32_encode_bits(1, TRE_FLAGS_CHAIN);
>   		} else { /* SPI_DUPLEX */

Re: [PATCH] dmaengine: qcom: gpi: Set link_rx bit on GO TRE for rx operation

[Doug Anderson]

Hi,

On Fri, Dec 16, 2022 at 4:15 AM Vijaya Krishna Nivarthi
<quic_vnivarth@quicinc.com> wrote:
>
> Gentle reminder to review/approve please.
>
> Thank you...

I was expecting you to send a v2 where you beefed up the patch
description to add the "why". Maybe others are as well? I suppose with
that I could add a weak Reviewed-by to the patch.

This is also a bad time for patches to be applied--it's right in the
middle of the merge window. Usually the best time to send reminders is
between -rc1 and -rc5. Officially "Fixes" could be applied at any time
(even during the merge window), but unless it's urgent that usually
doesn't happen.

Speaking of which, given your updated description, I'd say that this
_is_ a fix. Perhaps you should add a "Fixes" tag? That might actually
help it get applied sooner. ;-)

-Doug