Date:	Fri, 2 Nov 2012 02:42:51 -0700
From:	Russ Dill <Russ.Dill@...com>
To:	balbi@...com
Cc:	Pantelis Antoniou <panto@...oniou-consulting.com>,
	Alan Cox <alan@...rguk.ukuu.org.uk>,
	"Cousson, Benoit" <b-cousson@...com>,
	Tony Lindgren <tony@...mide.com>, linux-kernel@...r.kernel.org,
	Koen Kooi <koen@...inion.thruhere.net>,
	Matt Porter <mporter@...com>, linux-omap@...r.kernel.org,
	Kevin Hilman <khilman@...com>, Paul Walmsley <paul@...an.com>
Subject: Re: [PATCH 0/3] capebus moving omap_devices to mach-omap2

On Fri, Nov 2, 2012 at 1:57 AM, Felipe Balbi <balbi@...com> wrote:
> Hi,
>
> On Thu, Nov 01, 2012 at 04:49:23PM -0700, Russ Dill wrote:
>> On Thu, Nov 1, 2012 at 3:05 PM, Felipe Balbi <balbi@...com> wrote:
>> > HI,
>> >
>> > On Thu, Nov 01, 2012 at 03:59:50PM +0200, Pantelis Antoniou wrote:
>> >> Hi Alan,
>> >>
>> >> On Nov 1, 2012, at 3:51 PM, Alan Cox wrote:
>> >>
>> >> >> What they want, and what every user wants, is I plug this board in, and
>> >> >> the driver make sure everything is loaded and ready. No, the end users
>> >> >> don't want to see any of the implementation details of how the bitfile
>> >> >> is transported; the driver can handle it.
>> >> >
>> >> > That doesn't necessarily make it a bus merely some kind of hotplug
>> >> > enumeration of devices. That should all work properly both for devices
>> >> > and busses with spi and i²c as the final bits needed for it got fixed
>> >> > some time ago.
>> >> >
>> >> > In an ideal world you don't want to be writing custom drivers for stuff.
>> >> > If your cape routes an i²c serial device to the existing system i²c
>> >> > busses then you want to just create an instance of any existing driver on
>> >> > the existing i²c bus not create a whole new layer of goop.
>> >> >
>> >> > It does need to do the plumbing and resource management for the plumbing
>> >> > but thats not the same as being a bus.
>> >> >
>> >> > Alan
>> >>
>> >>
>> >> Fair enough. But there's no such thing a 'hotplug enumeration
>> >> construct' in Linux yet, and a bus is the closest thing to it. It does
>> >> take advantage of the nice way device code matches drivers and devices
>> >> though.
>> >>
>> >> I'm afraid that having the I2C/SPI drivers doing the detection won't
>> >> work.  The capes can have arbitrary I2C/SPI devices (and even more
>> >> weird components).  There is no way to assure for example that the I2C
>> >> device responding to address 'foo' in cape A is the same I2C device
>> >> responding to the same address in cape B.
>> >
>> > your ->detect() method should take care of that.
>>
>> There isn't some magical serial number in I²C devices that a
>> ->detect() method can read and the implementation of I²C is somewhat
>> flexible. One devices read may be another devices write. A detect
>
> look at what other drivers do. You can read a revision register, you can
> write a command and see if the device responds as expected, it doesn't
> matter.

Since a "revision" register isn't required by the I²C spec, it isn't
implemented on a huge number of chips. Also, having a few dozen probe
routines come though and write to random address of every single I²C
device can a) take a really long time, and b) have quite a few
unintended side effects.

>> method that only performs reads could easily toggle a gpio that resets
>> the board, rewrite and eeprom, or set the printer on fire. If you
>
> how ? It's just a read.

Because the I²C spec is incredibly flexible. For a lot of devices,
reading from a register is done by writing the register address, and
then reading the contents. For devices that don't implement registers
in that way (such as many eeproms), this is just a write.

>> browse through various detect functions, yes, some of them key off an
>> ID, but a lot of them just check various registers to see if certain
>> bits are zero, or certain bits are one. A lot of I²C devices I've
>> dealt with have no good way of probing them, especially GPIO chips
>> (you'll notice none of the I²C GPIO expanders have detect functions)
>
> it doesn't mean it can't be done.

Really? Please, do tell how you would write a detect function for a
PCA9534. It has 4 registers, an input port registers, an output port
register, a polarity inversion register, and a configuration register.
And don't forget, since we are probing, every detect routine for every
I²C driver will have to run with every I²C address on every bus,
possibly with both address formats.

>> On top of all this the detect routine does not tell you if the alert
>> pin is connected to some IRQ, or in the case of a GPIO expander, what
>> those GPIOs are connected to, etc, etc.
>
> so what ? All you want to do with detect is figure out if the far end is
> who you think it is, not what it's doing.

If we already knew who was there, we wouldn't need a detect routine.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists