Re: CVS commit: src/sys/dev/i2c

[Paul Goyette <paul%whooppee.com@localhost>]

On Tue, 26 Aug 2008, Jared D. McNeill wrote:

> Paul Goyette wrote:
> > To be more precise, the existing i2c code includes a bus scan to
> > enumerate the devices on the bus, however this loop is currently
> > #ifdef'd out.  It is unclear (to me, at least) if this enumeration
> > has ever been tested, or if it's probing of bus addresses is both
> > reliable and non-intrusive;  I have to assume that it was #ifdef'd
> > out for a reason.  Allowing wildcarding of the i2c addr locator
>
> It was causing hangs at boot on some machines and erratic behaviour on 
> others, which was why it was disabled.

That explains it.

Since there's a good reason for not probing/enumerating the bus, and
without such enumeration a wildcarded addr locator won't work, would
it not make sense to reverse this commit?

Leaving things as they are might give users false hope.  IE, if you
configure "spdmem* at iic?" config will blindly accept it but it will
have no effect and spdmem devices will not be matched.  At least with
things as they were prior to this commit, config would give an error
message that pretty explicitly stated that wildcarding was not allowed.

> There is a limited subset of i2c devices that can be enumerated (eg. for 
> direct config), this is described in the smbus specification but I'm not sure 
> how many (if any) devices implement it.
>
> Cheers,
> Jared
>
> !DSPAM:48b4a5401769723414189!

----------------------------------------------------------------------
|   Paul Goyette   | PGP DSS Key fingerprint: |  E-mail addresses:   |
| Customer Service | FA29 0E3B 35AF E8AE 6651 |  paul%whooppee.com@localhost   |
| Network Engineer | 0786 F758 55DE 53BA 7731 | pgoyette%juniper.net@localhost |
----------------------------------------------------------------------