Hi, We are using our custom tty-driver for communicating between an embedded Linux processor and a DSP running modem software. The pppd is started by typing "pppd /dev/our_driver0". pppd opens this driver in non-blocking mode. At write, when our write queue is full, we return -EAGAIN. With low traffic there is no problem, but when load increases above PPP-link capacity, the ppp subsystem begins to throw packets. Still the communication link is working, but this situation poses two problems: 1. If using PPP compression, packets drops inside PPP causes performance drop because the compression algorithm has the be resynchronized. 2. The Linux networking code output queuing system sees the ppp0 interface as a very fast interface (instead of a slow one), preventing Linux queue backlog to happen, and thus makes it impossible to have prioritized queuing (e.g. PRIO/TBF or HTB supporting DiffServ) in front of the ppp interface. We want the PPP-link to be reliable, and relying on the Linux network code output queue to take care of traffic overload situations (with FIFO tail drop or QoS supported drops, e.g.). We would appreciate any help / suggestions for what could be wrong with our implementation: what are we doing wrong? Best regards, Arne Lie SINTEF ICT, Dept. of Communication Systems e-mail: arne.lie@xxxxxxxxx -- To unsubscribe from this list: send the line "unsubscribe linux-ppp" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
