Tips on tracing bwmon?

Tips on tracing bwmon?

[Dr. David Alan Gilbert]

Hi all,
  I was looking for some tips on if I can use bwmon to see what interconnect
bandwidth is going through the dwc3 usb controllers on my x1 based Asus Zenbook A14.

  I've got usb-ncm gadget wired between a pair of them and I'd
like to see what stats are available;
(See details of my gadget stuff setup in
 https://lore.kernel.org/linux-arm-msm/atJn2jfI_cZy3UMo@gallifrey/T/#u )
And that's struggling to get much over 4Gbit/s over the network interface.

One reason I'm curious is because the dwc3-qcom glue sets an average
and a peak icc limit, and I wanted to know what the difference was and
what the type of timescales were for how long it allowed a peak.

Are there any docs I should look at?

I did:
   trace-cmd stream -e icc_bwmon:qcom_bwmon_update > log
on both machines and during the run I'm seeing events on exactly 3 of the
monitors;
24091000.pmu, 240b5400.pmu, 240b6400.pmu

(In the numbers below I've grepped for that monitor and sed'ed out
upto the 'measu_kbps=' part so I can then sort on the meas_kbps)

My reading of the interconnects_summary is these PMUs are
somewhere under the llcc_mc@interconnect-1, and in some
cases under both chm_apps@  and qns_llcc

So how do I interpret these - can I infer the bandwidth the usb controller
is using or just have a feel for the level above?
(Here the iperf -s server is running on 'dragon' and the client is
running on 'soup' - I find the partial symmetry between 240b[56]400.pmu
on the two machines promising it's measuring something useful!)

This is running a 7.30-rc6 with a bunch of stuff from Jens Glathe
on top.

Thanks in advance,

Dave (numbers below)

24091000.pmu
  on the llcc_mc@interconnect-1 (which also has the a400000.usb)
  
   dragon/server:
        lowest:    16000 up_kbps=800001 down_kbps=800000
       heighest: 2816000 up_kbps=3072001 down_kbps=2188000
        h-l:     2800000         2272000           1388000

   soup/client:
         lowest: 160000 up_kbps=800001 down_kbps=800000
        highest: 6256000 up_kbps=6835201 down_kbps=6220800               
        h-l      6096000         6035200           5420800


240b5400.pmu
  on the chm_apps@26400000.interconnect (which also has the a400000.usb)
        also under qns_llcc@26400000.interconnect
  
    dragon/server:
         lowest: 1328000 up_kbps=4800001 down_kbps=4800000
       heighest: 5616000 up_kbps=7464001 down_kbps=4800000 
            h-l  4288000         2664000                 0

    soup/client:
         lowest: 1440000 up_kbps=4800001 down_kbps=4800000
        highest: 4480000 up_kbps=4800001 down_kbps=4800000
            h-l  3040000               0                 0
                                                          
240b6400.pmu
  same as for 240b5400
  
    dragon/server:
         lowest:  352000 up_kbps=4800001 down_kbps=4800000
        highest: 3568000 up_kbps=4800001 down_kbps=4800000
            h-l  3216000               0                 0

    soup/client:
         lowest: 3216000 up_kbps=4800001 down_kbps=4800000
       heighest: 8736000 up_kbps=9600001 down_kbps=7464000
           h-l   5520000         4800000           2664000

-- 
 -----Open up your eyes, open up your mind, open up your code -------   
/ Dr. David Alan Gilbert    |       Running GNU/Linux       | Happy  \ 
\        dave @ treblig.org |                               | In Hex /
 \ _________________________|_____ http://www.treblig.org   |_______/

Re: Tips on tracing bwmon?

[Konrad Dybcio]

On 10/20/26 4:27 PM, Dr. David Alan Gilbert wrote:
> Hi all,
>   I was looking for some tips on if I can use bwmon to see what interconnect
> bandwidth is going through the dwc3 usb controllers on my x1 based Asus Zenbook A14.
> 
>   I've got usb-ncm gadget wired between a pair of them and I'd
> like to see what stats are available;
> (See details of my gadget stuff setup in
>  https://lore.kernel.org/linux-arm-msm/atJn2jfI_cZy3UMo@gallifrey/T/#u )
> And that's struggling to get much over 4Gbit/s over the network interface.
> 
> One reason I'm curious is because the dwc3-qcom glue sets an average
> and a peak icc limit, and I wanted to know what the difference was and
> what the type of timescales were for how long it allowed a peak.
> 
> Are there any docs I should look at?
> 
> I did:
>    trace-cmd stream -e icc_bwmon:qcom_bwmon_update > log
> on both machines and during the run I'm seeing events on exactly 3 of the
> monitors;
> 24091000.pmu, 240b5400.pmu, 240b6400.pmu
> 
> (In the numbers below I've grepped for that monitor and sed'ed out
> upto the 'measu_kbps=' part so I can then sort on the meas_kbps)
> 
> My reading of the interconnects_summary is these PMUs are
> somewhere under the llcc_mc@interconnect-1, and in some
> cases under both chm_apps@  and qns_llcc

These BWMONs only monitor the CPU-DRAM paths

Konrad

Re: Tips on tracing bwmon?

[Dr. David Alan Gilbert]

* Konrad Dybcio (konrad.dybcio@oss.qualcomm.com) wrote:
> On 10/20/26 4:27 PM, Dr. David Alan Gilbert wrote:
> > Hi all,
> >   I was looking for some tips on if I can use bwmon to see what interconnect
> > bandwidth is going through the dwc3 usb controllers on my x1 based Asus Zenbook A14.
> > 
> >   I've got usb-ncm gadget wired between a pair of them and I'd
> > like to see what stats are available;
> > (See details of my gadget stuff setup in
> >  https://lore.kernel.org/linux-arm-msm/atJn2jfI_cZy3UMo@gallifrey/T/#u )
> > And that's struggling to get much over 4Gbit/s over the network interface.
> > 
> > One reason I'm curious is because the dwc3-qcom glue sets an average
> > and a peak icc limit, and I wanted to know what the difference was and
> > what the type of timescales were for how long it allowed a peak.
> > 
> > Are there any docs I should look at?
> > 
> > I did:
> >    trace-cmd stream -e icc_bwmon:qcom_bwmon_update > log
> > on both machines and during the run I'm seeing events on exactly 3 of the
> > monitors;
> > 24091000.pmu, 240b5400.pmu, 240b6400.pmu
> > 
> > (In the numbers below I've grepped for that monitor and sed'ed out
> > upto the 'measu_kbps=' part so I can then sort on the meas_kbps)
> > 
> > My reading of the interconnects_summary is these PMUs are
> > somewhere under the llcc_mc@interconnect-1, and in some
> > cases under both chm_apps@  and qns_llcc
> 
> These BWMONs only monitor the CPU-DRAM paths

Thanks for the reply and explanation.

Are there other BWMONs that should show the USB path bandwidth,
and if so how would I watch them?

Thanks again,

> Konrad

Dave
-- 
 -----Open up your eyes, open up your mind, open up your code -------   
/ Dr. David Alan Gilbert    |       Running GNU/Linux       | Happy  \ 
\        dave @ treblig.org |                               | In Hex /
 \ _________________________|_____ http://www.treblig.org   |_______/

Re: Tips on tracing bwmon?

[Konrad Dybcio]

On 10/22/26 1:44 PM, Dr. David Alan Gilbert wrote:
> * Konrad Dybcio (konrad.dybcio@oss.qualcomm.com) wrote:
>> On 10/20/26 4:27 PM, Dr. David Alan Gilbert wrote:
>>> Hi all,
>>>   I was looking for some tips on if I can use bwmon to see what interconnect
>>> bandwidth is going through the dwc3 usb controllers on my x1 based Asus Zenbook A14.
>>>
>>>   I've got usb-ncm gadget wired between a pair of them and I'd
>>> like to see what stats are available;
>>> (See details of my gadget stuff setup in
>>>  https://lore.kernel.org/linux-arm-msm/atJn2jfI_cZy3UMo@gallifrey/T/#u )
>>> And that's struggling to get much over 4Gbit/s over the network interface.
>>>
>>> One reason I'm curious is because the dwc3-qcom glue sets an average
>>> and a peak icc limit, and I wanted to know what the difference was and
>>> what the type of timescales were for how long it allowed a peak.
>>>
>>> Are there any docs I should look at?
>>>
>>> I did:
>>>    trace-cmd stream -e icc_bwmon:qcom_bwmon_update > log
>>> on both machines and during the run I'm seeing events on exactly 3 of the
>>> monitors;
>>> 24091000.pmu, 240b5400.pmu, 240b6400.pmu
>>>
>>> (In the numbers below I've grepped for that monitor and sed'ed out
>>> upto the 'measu_kbps=' part so I can then sort on the meas_kbps)
>>>
>>> My reading of the interconnects_summary is these PMUs are
>>> somewhere under the llcc_mc@interconnect-1, and in some
>>> cases under both chm_apps@  and qns_llcc
>>
>> These BWMONs only monitor the CPU-DRAM paths
> 
> Thanks for the reply and explanation.
> 
> Are there other BWMONs that should show the USB path bandwidth,
> and if so how would I watch them?

Unfortunately there aren't.

Konrad

Re: Tips on tracing bwmon?

[Dr. David Alan Gilbert]

* Konrad Dybcio (konrad.dybcio@oss.qualcomm.com) wrote:
> On 10/22/26 1:44 PM, Dr. David Alan Gilbert wrote:
> > * Konrad Dybcio (konrad.dybcio@oss.qualcomm.com) wrote:
> >> On 10/20/26 4:27 PM, Dr. David Alan Gilbert wrote:
> >>> Hi all,
> >>>   I was looking for some tips on if I can use bwmon to see what interconnect
> >>> bandwidth is going through the dwc3 usb controllers on my x1 based Asus Zenbook A14.
> >>>
> >>>   I've got usb-ncm gadget wired between a pair of them and I'd
> >>> like to see what stats are available;
> >>> (See details of my gadget stuff setup in
> >>>  https://lore.kernel.org/linux-arm-msm/atJn2jfI_cZy3UMo@gallifrey/T/#u )
> >>> And that's struggling to get much over 4Gbit/s over the network interface.
> >>>
> >>> One reason I'm curious is because the dwc3-qcom glue sets an average
> >>> and a peak icc limit, and I wanted to know what the difference was and
> >>> what the type of timescales were for how long it allowed a peak.
> >>>
> >>> Are there any docs I should look at?
> >>>
> >>> I did:
> >>>    trace-cmd stream -e icc_bwmon:qcom_bwmon_update > log
> >>> on both machines and during the run I'm seeing events on exactly 3 of the
> >>> monitors;
> >>> 24091000.pmu, 240b5400.pmu, 240b6400.pmu
> >>>
> >>> (In the numbers below I've grepped for that monitor and sed'ed out
> >>> upto the 'measu_kbps=' part so I can then sort on the meas_kbps)
> >>>
> >>> My reading of the interconnects_summary is these PMUs are
> >>> somewhere under the llcc_mc@interconnect-1, and in some
> >>> cases under both chm_apps@  and qns_llcc
> >>
> >> These BWMONs only monitor the CPU-DRAM paths
> > 
> > Thanks for the reply and explanation.
> > 
> > Are there other BWMONs that should show the USB path bandwidth,
> > and if so how would I watch them?
> 
> Unfortunately there aren't.

Bah oh well; thanks for your answers.

Dave

> Konrad
-- 
 -----Open up your eyes, open up your mind, open up your code -------   
/ Dr. David Alan Gilbert    |       Running GNU/Linux       | Happy  \ 
\        dave @ treblig.org |                               | In Hex /
 \ _________________________|_____ http://www.treblig.org   |_______/