lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20250819193837.74c762c4@jic23-huawei>
Date: Tue, 19 Aug 2025 19:38:37 +0100
From: Jonathan Cameron <jic23@...nel.org>
To: Ben Collins <bcollins@...nel.org>
Cc: David Lechner <dlechner@...libre.com>, Nuno Sá
 <nuno.sa@...log.com>, Andy Shevchenko <andy@...nel.org>,
 linux-iio@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v5 5/5] iio: mcp9600: Add support for IIR filter

On Tue, 19 Aug 2025 19:28:51 +0100
Jonathan Cameron <jic23@...nel.org> wrote:

> On Mon, 18 Aug 2025 16:00:20 -0400
> Ben Collins <bcollins@...nel.org> wrote:
> 
> > On Mon, Aug 18, 2025 at 08:10:35PM -0500, Jonathan Cameron wrote:  
> > >     
> > > > > >  	case IIO_CHAN_INFO_SCALE:
> > > > > >  		*val = 62;
> > > > > >  		*val2 = 500000;
> > > > > >  		return IIO_VAL_INT_PLUS_MICRO;
> > > > > > +      
> > > > > If you want the extra space put it in previous patch.
> > > > >       
> > > > > >  	case IIO_CHAN_INFO_THERMOCOUPLE_TYPE:
> > > > > >  		*val = mcp9600_tc_types[data->thermocouple_type];
> > > > > >  		return IIO_VAL_CHAR;
> > > > > > +
> > > > > > +	case IIO_CHAN_INFO_LOW_PASS_FILTER_3DB_FREQUENCY:
> > > > > > +		if (data->filter_level == 0)      
> > > > > 
> > > > > Return the current requested value. An error is just going to confuse
> > > > > someone who tried to write this before enabling the filter and then
> > > > > checked to see if the write was successful.      
> > > > 
> > > > I could not get a concensus on this. On the one hand, if a user sets a
> > > > value here, would they not assume that the filter was enabled? What
> > > > about cases where a filter_type can be more than one valid type with
> > > > different available coefficients for each? What should it show then?    
> > > 
> > > So I was thinking of this like other things with 'enables' such as events.
> > > For those you always set the value first.  They don't really have a type
> > > field though (well they do but the ABI allows multiple at once unlike filters
> > > so we end up with a quite different looking ABI).
> > > 
> > > Agreed it gets challenging with multiple filter types. If it weren't for
> > > advertising the range I'd suggest just stashing whatever was written and
> > > then mapping it to nearest possible when the filter type is set.
> > > That's what the ad7124 does for changing between filters anyway
> > > though oddly it doesn't seem to have a control for filter type.
> > > 
> > > This is a good argument against the whole 'none' value for filter type
> > > - that's not much used so we could deprecate it for new drivers.
> > > 
> > > I'm not particularly keen on filter_enable but seems we are coming back
> > > around to that option to avoid this corner case.  Alternative being what
> > > you have here which isn't great for ease of use.    
> > 
> > I'm somewhat wondering if the filter frequency and frequency_available
> > attributes should not even show in sysfs unless the filter_type was
> > something other than "none".
> >   
> I'm not keen on that and trying to bolt is_visible into the mess of how
> we generate attributes would be hard and actual add and remove of attributes
> is horrible for races with userspace.
> 
> > > So for next version let's go for that. Make sure to include Documentation
> > > in a separate patch though so it's easy to see an poke holes in.    
> > 
> > Just to make sure I understand, you'd like to see a filter_enable
> > attribute and filter_type would not contain "none", then frequency and
> > frequency_available would always show something for whatever was in
> > filter_type?  
> 
> Yes.  I think that is best way forwards.  If we want to retrofit the one
> user of none to support the new ABI as well it should be easy to do.
> 

Ignore that. David convinced me otherwise.  Lets take this a bit slow
and discuss it fully in other branch of this thread.

J

> >   
> > > ABI design is a pain sometimes.    
> > 
> > The epitome of being able to paint yourself into a corner.
> >   
> Yup.
> 
> J
> 


Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ