[PATCH v2] net: stmmac: dwmac-qcom-ethqos: Enable TBS on all queues but 0

[PATCH v2] net: stmmac: dwmac-qcom-ethqos: Enable TBS on all queues but 0

[Abhishek Chauhan]

TSO and TBS cannot co-exist. TBS requires special descriptor to be
allocated at bootup. Initialising Tx queues at probe to support
TSO and TBS can help in allocating those resources at bootup.

TX queues with TBS can support etf qdisc hw offload.

This is similar to the patch raised by NXP <3b12ec8f618e>
<"net: stmmac: dwmac-imx: set TSO/TBS TX queues default settings">

Changes since v1:
- Subject is changed as per upstream guidelines
- Added a reference of a similar change done by NXP in
  body of the commit message

Signed-off-by: Abhishek Chauhan <quic_abchauha@quicinc.com>
---
 drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c b/drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c
index 31631e3f89d0..d2f9b8f6c027 100644
--- a/drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c
+++ b/drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c
@@ -728,7 +728,7 @@ static int qcom_ethqos_probe(struct platform_device *pdev)
 	struct stmmac_resources stmmac_res;
 	struct device *dev = &pdev->dev;
 	struct qcom_ethqos *ethqos;
-	int ret;
+	int ret, i;
 
 	ret = stmmac_get_platform_resources(pdev, &stmmac_res);
 	if (ret)
@@ -822,6 +822,10 @@ static int qcom_ethqos_probe(struct platform_device *pdev)
 		plat_dat->serdes_powerdown  = qcom_ethqos_serdes_powerdown;
 	}
 
+	/*Enable TSO on queue0 and enable TBS on rest of the queues*/
+	for (i = 1; i < plat_dat->tx_queues_to_use; i++)
+		plat_dat->tx_queues_cfg[i].tbs_en = 1;
+
 	return devm_stmmac_pltfr_probe(pdev, plat_dat, &stmmac_res);
 }
 
-- 
2.25.1

Re: [PATCH v2] net: stmmac: dwmac-qcom-ethqos: Enable TBS on all queues but 0

[Jeff Johnson]

On 2/6/2024 4:10 PM, Abhishek Chauhan wrote:
> TSO and TBS cannot co-exist. TBS requires special descriptor to be
> allocated at bootup. Initialising Tx queues at probe to support
> TSO and TBS can help in allocating those resources at bootup.
> 
> TX queues with TBS can support etf qdisc hw offload.
> 
> This is similar to the patch raised by NXP <3b12ec8f618e>
> <"net: stmmac: dwmac-imx: set TSO/TBS TX queues default settings">

note that there is a standard way to refer to a prior patch, in your case:
3b12ec8f618e ("net: stmmac: dwmac-imx: set TSO/TBS TX queues default
settings")

(note this format is defined in the context of the Fixes tag at
<https://www.kernel.org/doc/html/latest/process/submitting-patches.html#using-reported-by-tested-by-reviewed-by-suggested-by-and-fixes>)

/jeff

Re: [PATCH v2] net: stmmac: dwmac-qcom-ethqos: Enable TBS on all queues but 0

[Jakub Kicinski]

On Wed, 7 Feb 2024 09:26:05 -0800 Jeff Johnson wrote:
> > This is similar to the patch raised by NXP <3b12ec8f618e>
> > <"net: stmmac: dwmac-imx: set TSO/TBS TX queues default settings">  
> 
> note that there is a standard way to refer to a prior patch, in your case:
> 3b12ec8f618e ("net: stmmac: dwmac-imx: set TSO/TBS TX queues default
> settings")

Yes, please fix.

> (note this format is defined in the context of the Fixes tag at
> <https://www.kernel.org/doc/html/latest/process/submitting-patches.html#using-reported-by-tested-by-reviewed-by-suggested-by-and-fixes>)

A fixes tag would be great. But we can't point at 3b12ec8f618e, right?
Can someone explain what the user-visible problem is?
TBS cannot be used? Device reinit is require to enable it?

Re: Re: [PATCH v2] net: stmmac: dwmac-qcom-ethqos: Enable TBS on all queues but 0

[Andrew Halaney]

On Wed, Feb 07, 2024 at 10:19:34AM -0800, Jakub Kicinski wrote:
> On Wed, 7 Feb 2024 09:26:05 -0800 Jeff Johnson wrote:
> > > This is similar to the patch raised by NXP <3b12ec8f618e>
> > > <"net: stmmac: dwmac-imx: set TSO/TBS TX queues default settings">  
> > 
> > note that there is a standard way to refer to a prior patch, in your case:
> > 3b12ec8f618e ("net: stmmac: dwmac-imx: set TSO/TBS TX queues default
> > settings")
> 
> Yes, please fix.
> 
> > (note this format is defined in the context of the Fixes tag at
> > <https://www.kernel.org/doc/html/latest/process/submitting-patches.html#using-reported-by-tested-by-reviewed-by-suggested-by-and-fixes>)
> 
> A fixes tag would be great. But we can't point at 3b12ec8f618e, right?
> Can someone explain what the user-visible problem is?
> TBS cannot be used? Device reinit is require to enable it?
> 

I'm not sure you'd consider this a fix, but yes the syntax there should
be using that fixes style for sure.

This enables the ability to use TBS / etf on queues other than 0. So I'd
consider this a new feature for the Qualcomm variant of things here
personally.

Longer term it would be nice to be able to change which queues can do
what via ethtool as was discussed over here, but for now this at least
improves things and follows suit with the imx and intel variants:

    https://lore.kernel.org/netdev/c2497eef-1041-4cd0-8220-42622c8902f4@quicinc.com/


Thanks,
Andrew

Re: [PATCH v2] net: stmmac: dwmac-qcom-ethqos: Enable TBS on all queues but 0

[Andrew Halaney]

Hey ABC,

I may have swamped you with information last time when talking about
the commit message etiquette.

This looks a lot better, there's some discussion going on about the
canonical way to refer to another patch in the other thread (which I
agree with in the other thread, outside that this looks pretty good to me).

All that to say, I think you didn't understand my ask prior on the subject line:

    This is a new feature for netdev, so the Subject should have "net-next"
    in it: https://docs.kernel.org/process/maintainer-netdev.html#tl-dr

i.e. assuming no other changes:

    [PATCH net-next v2] net: stmmac: dwmac-qcom-ethqos: Enable TBS on all queues but 0

since at the moment I'm considering this a new feature. If it was a fix
(and had a Fixes: tag), you'd target "net" instead of "net-next".

I'll try and test this later today!

I'm not sure how accustomed to inline responses you are, so I'll be
verbose, but normally all responses are inline on list so look for em inline.
I'll repost my missed comment from v1 below inline.


On Tue, Feb 06, 2024 at 04:10:36PM -0800, Abhishek Chauhan wrote:
> TSO and TBS cannot co-exist. TBS requires special descriptor to be
> allocated at bootup. Initialising Tx queues at probe to support
> TSO and TBS can help in allocating those resources at bootup.
> 
> TX queues with TBS can support etf qdisc hw offload.
> 
> This is similar to the patch raised by NXP <3b12ec8f618e>
> <"net: stmmac: dwmac-imx: set TSO/TBS TX queues default settings">
> 
> Changes since v1:
> - Subject is changed as per upstream guidelines
> - Added a reference of a similar change done by NXP in
>   body of the commit message
> 
> Signed-off-by: Abhishek Chauhan <quic_abchauha@quicinc.com>
> ---
>  drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c | 6 +++++-
>  1 file changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c b/drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c
> index 31631e3f89d0..d2f9b8f6c027 100644
> --- a/drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c
> +++ b/drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c
> @@ -728,7 +728,7 @@ static int qcom_ethqos_probe(struct platform_device *pdev)
>  	struct stmmac_resources stmmac_res;
>  	struct device *dev = &pdev->dev;
>  	struct qcom_ethqos *ethqos;
> -	int ret;
> +	int ret, i;
>  
>  	ret = stmmac_get_platform_resources(pdev, &stmmac_res);
>  	if (ret)
> @@ -822,6 +822,10 @@ static int qcom_ethqos_probe(struct platform_device *pdev)
>  		plat_dat->serdes_powerdown  = qcom_ethqos_serdes_powerdown;
>  	}
>  
> +	/*Enable TSO on queue0 and enable TBS on rest of the queues*/

nitpicky: Please put spaces around the comment markers i.e. /* this */ over /*this*/

Thanks,
Andrew

> +	for (i = 1; i < plat_dat->tx_queues_to_use; i++)
> +		plat_dat->tx_queues_cfg[i].tbs_en = 1;
> +
>  	return devm_stmmac_pltfr_probe(pdev, plat_dat, &stmmac_res);
>  }
>  
> -- 
> 2.25.1
>