linux-kernel - Re: [PATCH 15/16] iio: adc: max1027: Support software triggers

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: <20210905104313.66a6a59a@jic23-huawei>
Date:   Sun, 5 Sep 2021 10:43:13 +0100
From:   Jonathan Cameron <jic23@...nel.org>
To:     Miquel Raynal <miquel.raynal@...tlin.com>
Cc:     Lars-Peter Clausen <lars@...afoo.de>,
        Thomas Petazzoni <thomas.petazzoni@...tlin.com>,
        linux-iio@...r.kernel.org, <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 15/16] iio: adc: max1027: Support software triggers

On Thu, 2 Sep 2021 17:21:00 +0200
Miquel Raynal <miquel.raynal@...tlin.com> wrote:

> Hi Jonathan,
> 
> Jonathan Cameron <jic23@...nel.org> wrote on Mon, 30 Aug 2021 11:50:46
> +0100:
> 
> > On Wed, 18 Aug 2021 13:11:38 +0200
> > Miquel Raynal <miquel.raynal@...tlin.com> wrote:
> >   
> > > Now that max1027_trigger_handler() has been freed from handling hardware
> > > triggers EOC situations, we can use it for what it has been designed in
> > > the first place: trigger software originated conversions.    
> > 
> > As mentioned earlier, this is not how I'd normally expect this sort of
> > case to be handled. I'd be expecting the cnvst trigger to still be calling
> > this function and the function to do the relevant check to ensure it
> > knows the data is already available in that case.  
> 
> I tried to follow your advice and Nuno's regarding this, I hope the new
> version will match your expectations (new version coming soon).
> However if my changes do not match, I will probably need more guidance
> to understand in deep what you suggest.
> 
> > > In other
> > > words, when userspace initiates a conversion with a sysfs trigger or a
> > > hrtimer trigger, we must do all configuration steps, ie:
> > > 1- Configuring the trigger
> > > 2- Configuring the channels to scan
> > > 3- Starting the conversion (actually done automatically by step 2 in
> > >    this case)
> > > 4- Waiting for the conversion to end
> > > 5- Retrieving the data from the ADC
> > > 6- Push the data to the IIO core and notify it
> > > 
> > > Add the missing steps to this helper and drop the trigger verification
> > > hook otherwise software triggers would simply not be accepted at all.
> > > 
> > > Signed-off-by: Miquel Raynal <miquel.raynal@...tlin.com>
> > > ---
> > >  drivers/iio/adc/max1027.c | 26 ++++++++++++++------------
> > >  1 file changed, 14 insertions(+), 12 deletions(-)
> > > 
> > > diff --git a/drivers/iio/adc/max1027.c b/drivers/iio/adc/max1027.c
> > > index 8c5995ae59f2..bb437e43adaf 100644
> > > --- a/drivers/iio/adc/max1027.c
> > > +++ b/drivers/iio/adc/max1027.c
> > > @@ -413,17 +413,6 @@ static int max1027_debugfs_reg_access(struct iio_dev *indio_dev,
> > >  	return spi_write(st->spi, val, 1);
> > >  }
> > >  
> > > -static int max1027_validate_trigger(struct iio_dev *indio_dev,
> > > -				    struct iio_trigger *trig)
> > > -{
> > > -	struct max1027_state *st = iio_priv(indio_dev);
> > > -
> > > -	if (st->trig != trig)
> > > -		return -EINVAL;
> > > -
> > > -	return 0;
> > > -}
> > > -
> > >  static int max1027_set_cnvst_trigger_state(struct iio_trigger *trig, bool state)
> > >  {
> > >  	struct iio_dev *indio_dev = iio_trigger_get_drvdata(trig);
> > > @@ -512,7 +501,21 @@ static irqreturn_t max1027_trigger_handler(int irq, void *private)
> > >  
> > >  	pr_debug("%s(irq=%d, private=0x%p)\n", __func__, irq, private);
> > >  
> > > +	ret = max1027_configure_trigger(indio_dev);    
> > 
> > I'd not expect to see this ever time.  The configuration shouldn't change
> > from one call of this function to the next.  
> 
> True, this is not needed.
> 
> > > +	if (ret)
> > > +		goto out;
> > > +
> > > +	ret = max1027_configure_chans_to_scan(indio_dev);    
> > 
> > This should also not change unless it is also responsible for the 'go' signal.
> > If that's true then it is badly named.  
> 
> It's responsible for the go signal, I renamed it "configure_and_start".
> 
> However, just for my own understanding, when would I be supposed to
> configure the channels requested by the user otherwise?

For buffered operation update_scan_mask callback.
There is a quirk where that's not called on the buffer disable path because
it's not always obvious what to 'reset to'.

J
> 
> Thanks,
> Miquèl