On Fri, Sep 2, 2022 at 3:42 PM Manikanta Pubbisetty
<quic_mpubbise@xxxxxxxxxxx> wrote:
>
> On 9/2/2022 6:50 PM, Robert Marko wrote:
> > On Fri, Sep 2, 2022 at 3:18 PM Manikanta Pubbisetty
> > <quic_mpubbise@xxxxxxxxxxx> wrote:
> >>
> >> On 9/2/2022 6:18 PM, Kalle Valo wrote:
> >>> Manikanta Pubbisetty <quic_mpubbise@xxxxxxxxxxx> writes:
> >>>
> >>>> Enable threaded NAPI on WCN6750. Unlike traditional NAPI poll which
> >>>> runs in softirq context and on the core which scheduled the NAPI,
> >>>> threaded NAPI makes use of kernel threads which are under direct
> >>>> control of the scheduler and helps in balancing the NAPI processing
> >>>> load across multiple CPUs thereby improving throughput.
> >>>>
> >>>> In the case of WCN6750, enabling threaded NAPI has improved
> >>>> 160 MHz RX throughput by nearly 400 Mbps. This should give similar
> >>>> gains for other ath11k devices as well, therefore enable threaded
> >>>> NAPI on all other devices.
> >>>>
> >>>> Tested-on: WCN6750 hw1.0 AHB WLAN.MSL.1.0.1-00887-QCAMSLSWPLZ-1
> >>>>
> >>>> Signed-off-by: Manikanta Pubbisetty <quic_mpubbise@xxxxxxxxxxx>
> >>>> ---
> >>>> drivers/net/wireless/ath/ath11k/ahb.c | 1 +
> >>>> drivers/net/wireless/ath/ath11k/pcic.c | 1 +
> >>>> 2 files changed, 2 insertions(+)
> >>>>
> >>>> diff --git a/drivers/net/wireless/ath/ath11k/ahb.c b/drivers/net/wireless/ath/ath11k/ahb.c
> >>>> index d7d33d5cdfc5..e44e2f29a88f 100644
> >>>> --- a/drivers/net/wireless/ath/ath11k/ahb.c
> >>>> +++ b/drivers/net/wireless/ath/ath11k/ahb.c
> >>>> @@ -314,6 +314,7 @@ static void ath11k_ahb_ext_irq_enable(struct ath11k_base *ab)
> >>>> struct ath11k_ext_irq_grp *irq_grp = &ab->ext_irq_grp[i];
> >>>>
> >>>> if (!irq_grp->napi_enabled) {
> >>>> + dev_set_threaded(&irq_grp->napi_ndev, true);
> >>>> napi_enable(&irq_grp->napi);
> >>>> irq_grp->napi_enabled = true;
> >>>> }
> >>>> diff --git a/drivers/net/wireless/ath/ath11k/pcic.c b/drivers/net/wireless/ath/ath11k/pcic.c
> >>>> index cf12b98c480d..c703db19de51 100644
> >>>> --- a/drivers/net/wireless/ath/ath11k/pcic.c
> >>>> +++ b/drivers/net/wireless/ath/ath11k/pcic.c
> >>>> @@ -440,6 +440,7 @@ void ath11k_pcic_ext_irq_enable(struct ath11k_base *ab)
> >>>> struct ath11k_ext_irq_grp *irq_grp = &ab->ext_irq_grp[i];
> >>>>
> >>>> if (!irq_grp->napi_enabled) {
> >>>> + dev_set_threaded(&irq_grp->napi_ndev, true);
> >>>> napi_enable(&irq_grp->napi);
> >>>> irq_grp->napi_enabled = true;
> >>>> }
> >>>
> >>> The commit log claims that this enabled _only_ on WCN6750 but aren't we
> >>> enabling it on all ath11k hardware, or am I missing something? I admit I
> >>> didn't check this very carefully.
> >>>
> >>> (reads the commit log one more time)
> >>>
> >>> Ah, in the last sentence you mention that it's enabled on all hardware.
> >>> That's quite easy to miss and the commit log is quite misleading, please
> >>> emphasise already in the title and the first sentence that this is for
> >>> all hardware.
> >>
> >> My Bad, yes you right. The patch was made initially only for WCN6750 and
> >> was enabled later for all devices.
> >>
> >>>
> >>> Also more testing would be nice. Enabling something like this with
> >>> testing only on one hardware family (WCN7850) can be risky. I always get
> >>> warm fuzzy feelings if a patch is tested with all three hardware
> >>> families we currently support:
> >>>
> >>> * IPQ8074 etc
> >>> * QCA6390 etc
> >>> * WCN7850
> >>>
> >>
> >> WCN7850 should be an ath12k device If I'm correct.
> >>
> >> Regardless of the chip family, even I feel that the tput changes like
> >> these should be tested on all the chipsets. Availability of the hardware
> >> and time are something which are stopping me in testing the changes on
> >> all supported targets.
> >>
> >> As I said, I had made the changes only to WCN6750 initially (using a
> >> hw_param). Can we take that approach for now and enable this for other
> >> targets only if required & upon thorough testing?
> >
> > I can tell you that on IPQ8074 threaded NAPI really improves perfromance.
> >
>
> Great. Do you have any test results on IPQ8074?
I dont have full test results, but on Poco F2 Pro as the client @80MHz AX
I got ~720Mbps compared to ~550Mbps before.
I can tell you that in OpenWrt, we have had it enabled for 6+ months
at this point
and its been really good.
Regards,
Robert
>
> Thanks,
> Manikanta
--
Robert Marko
Staff Embedded Linux Engineer
Sartura Ltd.
Lendavska ulica 16a
10000 Zagreb, Croatia
Email: robert.marko@xxxxxxxxxx
Web: www.sartura.hr
