Hi, On Tue, Oct 30, 2012 at 12:50:52PM +0000, Paul Walmsley wrote: > On Tue, 30 Oct 2012, Felipe Balbi wrote: > > > the wmb() patch should help, but you said it doesn't, right ? > > Correct. Not sure why you think that would help; the problem is almost because of how the driver is behaving... The only way for omap_i2c_wait_for_bb() to timeout is if the bus is really still busy; meaning that either transfer hasn't fully completed (some data still in FIFO), or we didn't send STOP or Repeated START condition on the bus. > certainly PM-idle related, given that the problem goes away by reverting > Jean's commit 3db11feffc1ad2ab9dea27789e6b5b3032827adc "ARM: OMAP: convert > I2C driver to PM QoS for MPU latency constraints". > > The problem is consistent and easy to reproduce here. I don't doubt that :-) What I think, however, is that PM QoS just made the problem a bit easier to trigger for whatever reason. Maybe because pm qos defers constraint updates, or something similar ? > > without seeing debugging information on the failing case, it'll be very > > difficult to sort this one out provided noone else seems to be able to > > reproduce it. > > Aaro sees it also, so it's definitely not a local artifact. Again I don't doubt it. > It's not very surprising that the problem doesn't appear with the extra > logging enabled, since the extra logging presumably keeps the system out > of idle. > > Anyway, you are welcome to the userspace, bootloader, and X-loader here if > you would like to try to reproduce it. They've been sent to Jean and > Kevin and Shubhrajyoti already. Simply request it and I will send links. I can have a look, sure, but I don't have the same platform as you have. I have a beagle XM revB (according to the stickers on the board) which I could use to try to reproduce, if I manage to, then hooray :-) In that case, please give me also a link to your uImage so I can make sure to use the exact same setup. cheers -- balbi
Attachment:
signature.asc
Description: Digital signature
