Brian Murphy <br...@murphy.dk> writes: > On Mon, Feb 15, 2010 at 10:07 PM, Bansal, Prateek <pban...@ti.com> wrote: >> Hi Brian, >> >> Thanks for sharing the patch. We tested this on 2.6.32 kernel on DM355 >> device and it worked fine however we are yet to get this working with 2.6.30 >> kernel. >> >> One of the things that we are trying to investigate is why the transmit FIFO >> empty interrupts are not generated. If you have more visibility into this, I >> would appreciate any details. Considering we see this issue across number of >> devices OMAP-L137, DM6446 and DM355, it points to a software issue rather >> than hardware one but needs to be confirmed. >> >> > > I have it working with several kernels, It needed a little massaging > to work with the latest > as someone else was playing in the same place as me :). > > The issue the patch fixes is a hardware issue. The UART in the TI SOC > devices does not > have the same behaviour as a 16550 uart and the linux kernel driver > requires that > behaviour to function efficiently. > > The workaround in the current kernel driver is not good and hides the > problem in a very > bad way. > > I will try again next week to get a discusion going on the > linux-serial list. It seems like Alan > Cox is still active despite threats to the contary so I may send > directly to him to see if that > helps.
[sorry to jump in late on this thread...] Unfortunately, last time I checked, the 8250 driver was orphaned (no maintainer.) And since there are 8250 UARTs on *tons* of devices out there, folks are reluctant to do significant changes without lots of testing. I suggest in your repost of this patch to be also list the devices that you (or others) have validated it on. That always helps. You might also request others do do some testing on their platforms as well. Kevin _______________________________________________ Davinci-linux-open-source mailing list Davinci-linux-open-source@linux.davincidsp.com http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
