Hi Bjorn, > This changes a number of fundamental parts of the design: > - use a shared ring buffer for all read data > - use per file position counter to map each open file into the buffer > - no read locking at all! reading is just a matter of copying data > from the ring buffer > - using an ever-increasing file position counter mapping into the > buffer > - try hard to let every client start reading at a message boundary > to support packetized management protocols better > > Signed-off-by: Bjørn Mork <bjorn@xxxxxxx> > --- > This is more of a request for comments thing at the moment. It changes > a lot in one go. But there really isn't much that can be done in > steps here. Once you put the ring buffer there then the read locking > goes away. > > FWIW, this Works For Me, both using the AT command based real Device > Management interfaces on the Ericsson F3507g modem, and using the > QMI command based CDC ECM lookalike interface on the Huawei E392 > modem why are you bothering with allowing multiple readers. I would just make the access to this device exclusive. I know that this way you can have some easy debugging and testing, but at the end of the day there will only one process controlling this device. Same argumentation as for the echo support applies here. If we just leave this and the echo stuff out of it, you are almost there with these patches. And then we could focus to actually write some QMI tools and support to utilize this functionality. Regards Marcel -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
