[PATCH 00/17] coresight: next v4.12-rc4

[Kim Phillips]

On Mon, 12 Jun 2017 10:50:43 -0600
Mathieu Poirier <mathieu.poirier at linaro.org> wrote:

Hi Mathieu,

> On 9 June 2017 at 16:53, Kim Phillips <kim.phillips at arm.com> wrote:
> 
> > On Mon,  5 Jun 2017 14:15:01 -0600
> > Mathieu Poirier <mathieu.poirier at linaro.org> wrote:
> >
> > Hi Mathieu,
> >
> > > Please consider for inclusion in the 4.13 cycle.  Of particular interest
> > is
> > > patch 15/17, which adds a new driver under the CoreSight umbrella.
> >
> > After applying the patchseries, the devices now have self-named
> > subdirectories under them:
> >
> > $ find /sys/devices/ -name \*sink\*
> > /sys/devices/platform/20140000.etf/20140000.etf/enable_sink
> > /sys/devices/platform/20070000.etr/20070000.etr/enable_sink
> > /sys/devices/platform/20030000.tpiu/20030000.tpiu/enable_sink
> > /sys/devices/platform/20010000.etf/20010000.etf/enable_sink
> > $
> >
> 
> This is the expected behaviour.  CoreSight devices are found on the AMBA
> bus and like other AMBA devices they will show up under
> /sys/devices/platform/.

right, but there are subdevices of themselves, e.g.,
20010000.etf/20010000.etf:

kim at juno platform$ find /sys/devices/ -name \*etf\*
/sys/devices/platform/20140000.etf
/sys/devices/platform/20140000.etf/20140000.etf    <- THIS
/sys/devices/platform/20010000.etf
/sys/devices/platform/20010000.etf/20010000.etf    <- and this
/sys/devices/virtual/misc/20140000.etf
/sys/devices/virtual/misc/20010000.etf

'tree' output:

├── 20010000.etf
│   ├── 20010000.etf
│   │   ├── enable_sink
│   │   ├── mgmt
│   │   │   ├── ctl
│   │   │   ├── devid
│   │   │   ├── ffcr
│   │   │   ├── ffsr
│   │   │   ├── mode
│   │   │   ├── pscr
│   │   │   ├── rrp
│   │   │   ├── rsz
│   │   │   ├── rwp
│   │   │   ├── sts
│   │   │   └── trg
│   │   ├── power
│   │   │   ├── autosuspend_delay_ms
│   │   │   ├── control
│   │   │   ├── runtime_active_time
│   │   │   ├── runtime_status
│   │   │   └── runtime_suspended_time
│   │   ├── subsystem -> ../../../../bus/coresight
│   │   ├── trigger_cntr
│   │   └── uevent
│   ├── driver -> ../../../bus/amba/drivers/coresight-tmc
│   ├── driver_override
│   ├── id
│   ├── of_node -> ../../../firmware/devicetree/base/etf at 20010000
│   ├── power
│   │   ├── autosuspend_delay_ms
│   │   ├── control
│   │   ├── runtime_active_time
│   │   ├── runtime_status
│   │   └── runtime_suspended_time
│   ├── resource
│   ├── subsystem -> ../../../bus/amba
│   └── uevent

so things like power/ are duplicated, driver pointers missing from the
sub-node, etc.  I tried reading the documentation to find the reasoning
behind it, and didn't see any, but I may have missed it.  Normally each
platform device gets its own single node, and doesn't create a node
under itself with the same name.

If the subsystem pointer is of use, then I would have
thought 'bus/coresight' would have registered itself as a 'bus/amba'
node? 

> > Which doesn't match the coresight.txt usage documentation...
> 
> When operating CoreSight from the sysFS interface there is nothing
> preventing people from using the entries under /sys/devices/platform should
> they wish to.  It is simply easier to use /sys/bus/coresight/devices/, and
> that is well documented in coresight.txt.
> 
> > Also, do the current juno platforms not have CPU debug modules?  I'd
> > like to test the new driver.
> 
> By all means - Leo pointed out the patch adding the CPU debug entries in
> the DT  file for Juno.

Thanks, I applied it, was able to
toggle /sys/kernel/debug/coresight_cpu_debug/enable, but wasn't sure
how to trigger the cpu debug output...

Thanks,

Kim