Re: [PATCHv8 2/5] arm64: dts: qcom: msm8998: Add Coresight support

[Suzuki K Poulose]

On 18/07/2019 10:14, Sai Prakash Ranjan wrote:
> Hi Suzuki,
>
> On 7/18/2019 1:58 PM, Suzuki K Poulose wrote:
> > Hi Sai,
> >
> >
> >   ÂÂÂÂÂÂÂ etr@6048000 {
> > > +ÂÂÂÂÂÂÂÂÂÂÂ compatible = "arm,coresight-tmc", "arm,primecell";
> > > +ÂÂÂÂÂÂÂÂÂÂÂ reg = <0x06048000 0x1000>;
> > > +
> > > +ÂÂÂÂÂÂÂÂÂÂÂ clocks = <&rpmcc RPM_SMD_QDSS_CLK>, <&rpmcc
> > > RPM_SMD_QDSS_A_CLK>;
> > > +ÂÂÂÂÂÂÂÂÂÂÂ clock-names = "apb_pclk", "atclk";
> > > +ÂÂÂÂÂÂÂÂÂÂÂ arm,scatter-gather;
> >
> > Please could you confirm that you have tested the scatter-gather mode
> > with ETR ? Either via perf/sysfs. Please could you share your results ?
> > Unless verified
> > this is going to be fatal for the system.
> >
> > Similarly for other platforms.
>
> Yes I have tested with scatter-gather mode with ETR on all platforms
> which I have posted via sysfs(not perf) before on previous versions of
> this patch series and no issues were found. And I suppose this was
> discussed in v2 of this patch series [1].

Using the sysfs doesn't guarantee that the ETR actually uses SG mode, unless
the buffer size selected is > 1M, which is why I am more interested in the
perf usage. Alternatively you may configure a larger buffer size (say, 8MB) via:

echo 0x800000 > /sys/bus/coresight/.../tmc_etr0/buffer_size


> As said in one of the series initially [1], QCOM msm downstream kernels
> have been using scatter gather mode and we haven't seen any fatal issues.
>
> [1] https://patchwork.kernel.org/patch/10769535/

I haven't seen any test results there either.

Cheers
Suzuki