Re: [PATCH] doc: coresight: correct usage for '/dev/cpu_dma_latency'

From: Leo Yan
Date: Fri Sep 15 2017 - 06:24:01 EST

Hi Kim,

On Wed, Aug 23, 2017 at 11:05:28AM -0500, Kim Phillips wrote:
> On Wed, 23 Aug 2017 15:23:18 +0800
> Leo Yan <leo.yan@xxxxxxxxxx> wrote:
> > Cc: Kim Phillips <kim.phillips@xxxxxxx>
> > Reported-by: Kim Phillips <kim.phillips@xxxxxxx>
> Thanks; typically only the latter is needed.
> > Set latency request to /dev/cpu_dma_latency to disable all CPUs specific idle
> > -states (if latency = 0uS then disable all idle states):
> > -# echo "what_ever_latency_you_need_in_uS" > /dev/cpu_dma_latency
> > +states (if latency = 0uS then CPU Idle governor selects idle state0, so this
> > +means 'WFI' is still enabled but other deeper states have be disabled, this
> > +can avoid power off CPUs):
> > +# exec 3<> /dev/cpu_dma_latency; echo "what_ever_latency_you_need_in_uS" >&3
> not a fan of the "what_ever_latency_you_need_in_uS" (including and
> especially the quotes which can create ambiguity in the user's mind):
> just put a cut-n-pasteable example, clarifying the typically-default
> value you chose, and its unit, in the text.
> More to the point, how did you test this? Are you sure that that the
> value being echoed isn't being interpreted as a binary number?

I checked for this with manually adding log in kernel function, if we
use 'echo' command and input string with 4 chars length then the
parameter cannot be parsed correctly by copy_from_user(); if input
string with other length, the string will be parsed with hexadecimal

> kernel/power/qos.c:pm_qos_power_read() looks to be looking for a 32-bit
> binary integer, but I'm not sure if that's where it gets read.
> Certainly, this 2013 article uses a C example to write a binary integer:
> Please double-check, thanks,

Thanks for the reminding. C code with 's32' type can pass correct
value into kernel.

Have sent new version patch, please help review.

Leo Yan