On Fri, Mar 27, 2015 at 06:14:28AM -0700, Guenter Roeck wrote:
> On 03/27/2015 06:01 AM, Wolfram Sang wrote:
> >On Fri, Mar 27, 2015 at 05:51:11AM -0700, Guenter Roeck wrote:
> >>On 03/27/2015 01:09 AM, Wolfram Sang wrote:
> >>>
> >>>>just to give you an update: I do have some code, but it is a bit messy,
> >>>>and it doesn't work well for ds2482 (the chip behind it still hangs up
> >>>>if I access it in parallel through i2c-dev). On top of that, it causes
> >>>>pretty significant slow-downs when accessing other devices on the same
> >>>>bus at the same time. Not surprising, I guess, since it expands the scope
> >>>>of the bus lock significantly.
> >>>
> >>>Just to get a better idea: Did you try taking the adapter_lock before
> >>>the two SMBus command which needed to be concatenated (and use
> >>>smbus_xfer directly)?
> >>>
> >>I did. I didn't use smbus_xfer directly, though, but introduced lockless
> >>versions of the various smbus commands, and kept using those.
> >
> >And then the chip still hangs? Or was that the performance penalty here?
> >
> Parallel access to a second eeprom chip on the same bus was much slower
> than before.

Interesting. I wonder what is the reason, I would have expected just a
small delay. Would you mind sending the patches for the non-locked smbus
routines? Would be nice to have that around in case I or someone else
find some time to try as well.

> Also, the new code did not solve the problem for ds2482 (completely unrelated
> to the at24 driver of course). Even with proper locking, the chip ended up
> hanging after some parallel accesses through i2c-dev. Granted, ds2482 is
> a difficult beast, but it is still annoying how access through i2c-dev
> can mess it up.

I assume you basically replaced the access_lock with the adapter_lock
with this one?

> 
> The latter is what bothered me more: What is the point of all this if we
> still can not ensure correct operation ?

Yeah, this is not good at all.

How do you use i2c-dev BTW? i2c_rdwr_msgs? What about iterating over all
msgs in that and check for busy addresses?