Hi, On Thu, Jan 24, 2013 at 11:13:39AM +0200, Felipe Balbi wrote: > On Thu, Jan 24, 2013 at 11:05:11AM +0200, Aaro Koskinen wrote: > > On Wed, Jan 23, 2013 at 12:23:04PM +0200, Felipe Balbi wrote: > > > In case we loop on IRQ handler until stat is > > > finally zero, we would end up in a situation > > > where all I2C transfers would misteriously > > > timeout because we were not calling complete() > > > in that situation. > > > > This is wrong, you may still have bytes to transfer, > > but a new interrupt has not arrived yet. > > > > This patch also breaks bisection; if you just apply patches 1-2 of > > this series the I2C does not work at all. > > probably another n900-thing. What I had seen is that we could endup in a > situation where we wouldn't "break" out of the look, and rather would > fall in the "goto out" when STAT register is zero. In that case, even if > we had transferred all bytes, we wouldn't call complete() and, thus, our > commands would timeout. > > Care to send console logs of the error you have seen ? I kinda doubt we > can fall into the situation you described above where IRQ hasn't been > updated to STAT register. BTW, check what happens if you just revert that commit. -- balbi
Attachment:
signature.asc
Description: Digital signature
