Re: [PATCH 1/1] Add strong pullup emulation to w1-gpio master driver.

[Evgeny Boger]

11/12/2013 12:01 PM, David Fries:
> On Tue, Nov 12, 2013 at 05:07:14AM +0400, Evgeny Boger wrote:
> > +David Fries <david@xxxxxxxxx>
> >
> > Hi David,
> >
> > Would you please comment on this?
>
> On Mon, Nov 11, 2013 at 06:36:54PM +0400, Evgeny Boger wrote:
> >   Strong pullup is emulated by driving pin logic high after write
> >   command when
> >   using tri-state push-pull GPIO.
> Not knowing the hardware involved, is driving the logic high a
> stronger pullup than the normal weak pullup input high?  Meaning it
> was already being left high, just with a lessor pullup and this will
> provide a stronger one?




Sure. The push-pull GPIO on common SoC's are usually able to provide up 
to 10 mA of current.




> On Tue, Nov 12, 2013 at 03:09:36AM +0400, Evgeniy Polyakov wrote:
> > > + msleep(pdata->pullup_duration);
> > This doesn't look like a good idea - kernel will sleep for that long
> > not doing usual w1 job
> Not speaking for Evgeny Boger, but I'm thinking that's intended here.
> The original strong pullup code change 6a158c0de791a81 I wrote will
> msleep in w1_post_write when a hardware pullup isn't available, while
> the hardware ds2490 ds9490r_set_pullup sleeps for the strong pullup
> using spu_sleep variable.  The user requests a strong pullup for a
> given time and any other operations on the bus will interrupt the
> strong pullup, so locking any other operations sounds desired.
>
> > 11/12/2013 05:03 AM, Evgeniy Polyakov:
> > > Hi
> > >
> > > 12.11.2013, 03:32, "Evgeny Boger" <eugenyboger@xxxxxxxxx>:
> > > > > Why did you drop this check? It has nothing with w1-gpio driver
> > > > This check prevents master from implementing "set_pullup"  provided it does support only "write_bit" method.
> > > > The comment above states that
> > > > >   w1_io.c would need to support calling set_pullup before - * the last write_bit operation of a w1_write_8 which it currently - * doesn't.
> > > > which is kind of strange, since it describes what w1_io.c actually does support.
> > > >
> > > > w1_write_8 (w1_io.c:154, https://github.com/torvalds/linux/blob/master/drivers/w1/w1_io.c#L154):
> > > > >                  for (i = 0; i < 8; ++i) {
> > > > >                          if (i == 7)
> > > > >                                  w1_pre_write(dev);
> > > > >                          w1_touch_bit(dev, (byte >> i) & 0x1);
> > > > >                  }
> > > > It seems like w1_write_8() calls w1_pre_write(), which in turn calls set_pullup() just before the last write_bit().
> I'm not seeing any harm in removing this check and clear
> master->set_pullup.  It doesn't seem correct for this code to override
> a master that claims to provide something of a stronger pullup.  It's
> been about five years since I wrote that code, I think it was just to
> protect against a stupid master.
>
> With this patch the last w1_write_bit will go logic 1, for 64 or 10 us
> before returning, then w1_gpio_set_pullup is called to enable the
> strong pullup.  What I wouldn't know is if in that last bit if the
> logic 1 would be a go up to the strong pullup, or if it would finish
> that time slot with a weak pullup and then go to a strong pullup.  I
> would have to dig into the timing specifications much more than I have
> time to right now to say what is supposed to happen.  The 18b20
> datasheet lists, "The DQ line must be switched over to the strong
> pullup within 10 us maximum after issuing any protocol that involves
> copying the E2 memory or initiates temperature conversions."  It isn't
> clear where that 10 us starts from.  You might try to dig around and
> see if that last bit written should go to weak pullup 1 or strong
> pullup 1.  It would take more changes if it should go right to a
> strong pullup.


I wasn't able to find any support for the latter statement.
It looks like the strong pull-up should be enabled *after* the last bit 
has been sent
so no need to set strong pull-up there.

However setting strong pullup for last bit makes sense just to ensure we
fit to 10us time window.

On the other hand, I didn't experienced any problems with the proposed
implementation.





> > > > I'm not sure why this check was there in the first place.
> > > Please add author of those lines to clarify things.
> > > This doesn't look obvious to me

--
 Evgeny
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/