Re: [PATCH 4/4] hwmon: vcnl3020: add hwmon driver for intrusion sensor

[Ivan Mikhaylov]

On Wed, 2021-05-05 at 07:02 -0700, Guenter Roeck wrote:
> On Tue, May 04, 2021 at 10:46:53PM +0300, Ivan Mikhaylov wrote:
> > On Fri, 2021-04-30 at 09:38 -0700, Guenter Roeck wrote:
> > > On Fri, Apr 30, 2021 at 06:24:19PM +0300, Ivan Mikhaylov wrote:
> > > > Intrusion status detection via Interrupt Status Register.
> > > > 
> > > > Signed-off-by: Ivan Mikhaylov <i.mikhaylov@xxxxxxxxx>
> > > 
> > > I think this should, if at all, be handled using the
> > > iio->hwmon bridge (or, in other words, require a solution
> > > which is not chip specific).
> > 
> > Thanks a lot for suggestion, it's actually looks what's needed here instead
> > of
> > this driver. Anyways, there is no IIO_PROXIMITY support inside supported
> > types
> > in iio_hwmon.c. Should I add additional case inside this driver for
> > IIO_PROXIMITY type?
> > 
> > > I am also not sure if "proximity" is really appropriate to use
> > > for intrusion detection in the sense of hardware monitoring.
> > > This would require a proximity sensor within a chassis, which
> > > would be both overkill and unlikely to happen in the real world.
> > > "Intrusion", in hardware monitoring context, means "someone
> > > opened the chassis", not "someone got [too] close".
> > > 
> > 
> > I'm not sure either but it exists :) And it's exactly for this purpose:
> > "someone opened the chassis", "how near/far is cover?".
> > 
> 
> The cost for VCNL3020, for a full reel with 3,300 chips, is $1.17 per chip
> at Mouser. A mechanical switch costs a couple of cents. A single proximity
> sensor won't cover all parts of a chassis; one would likely need several
> chips to be sure that are no blind spots (if that is even possible - I don't
> think it is in any of my PC chassis due to mechanical limitations). This
> is on top of programming, which would be sensitive to generating false
> alarms (or missing alarms, for that matter). That sounds quite impractical
> and expensive to me. I'd really like to see the actual use case where a
> proximity sensor (or set of proximity sensors) is used for intrusion
> detection in the sense of hardware monitoring - not just the technical
> possibility of doing so, but an actual use case (as in "this vendor,
> in this chassis, is doing it").
> 
> Thanks,
> Guenter


Guenter, VCNL3020 is indeed used as an intrusion detection sensor at least in
one real design. That is YADRO VESNIN Rev. C where the proximity sensor is
installed in a very tight space on an nvme switch board where installation of a
mechanical switch was not possible without substantial redesign of the existing
other components that would cost a lot more than the price of VCNL3020.

VESNIN is a very tight-packed design of 4 x POWER8 CPUs, up to 8TB of RAM, and 26 nvme disks, all that in just 2U.
* https://imgur.com/a/wU9wEd4