The serial port of the HP Brio BA 200 generates receive frame errors at a rate severe enough to make the ppp0 connection fail at speeds over 58K (15% to 30% receive frame error rate as reported by ifconfig). The error rate is less severe at lower speeds, but is still evident at a 1% to 2% rate at 14.4K. See Red Hat Service Request Number 145382 for workarounds that have already been tried. The problem is evident on two separate new Brio BA200 PC's. One has a 433 Celeron and the other a 466 Celeron. Both use the i810 chip set. The symptoms are evident under Red Hat 6.1, 6.2, and Suse 6.4. The Red Hat 6.2 Retail version has been upgraded to the 2.2.14-12.i686.rpm kernel upgrade. Almost identical symptoms are reported by another user in a Linuxcare post for an Intel SE440BX-2 motherboard with a different chip set (see http://lc.experts- exchange.com/Computers/Networking/Linux/Q_10361018.html)..... The external modem works error free on an old Gateway 2000 Pentium 60 on RH Linux 6.1 (retail version with no upgrades) used to IP masq a local network. The serial port also works fine at high speed under Windows with no equipment changes.
I finally got to the bottom of the problem with help from Red hat support -- see Red Hat service request number 145382 for a complete history. The problem appears to be an OEM hard drive made for HP by Seagate. When I use the Seagate OEM drive on an Intel CA810EAL motherboard I have the problem (although less of a problem than when using the HP motherboard; apparently the Intel board has better PCI/IDE interrupt timing specs, but still not good enough to eliminate the problem completely). When I use a WD136AA retail drive with the Intel motherboard the problem goes away (I'd test the WD drive on the HP motherboard but the HP BIOS won't handle the WD drive). This appears to be a kernel problem involving a marginal OEM hard drive and perhaps a marginal OEM motherboard in the HP Brio BA200, not a ppp problem. Based on the symptoms, I suspect a race condition involving PCI/IDE interrupt timing. In addition to the high speed serial errors the Brio generates "unexpected interrupts" and "drive reset" errors whether or not I use the serial port. In essence the Brio BA200 is bad news for the current Linux kernel and needs to be on the list of non-compliant hardware. Apparently the Brio works in Windows because Windows uses the DMA channel, avoiding PCI/IDE interrupt contention. Since the Linux kernel in its current release does not use DMA(?) all hard drive and CD traffic has to be handled through IDE interrupts, and apparently certain OEM hard drives and/or motherboards generate race conditions in that environment.
I'll mark this as "deferred" and assign it to the kernel component then.
I'd like to request that you at least show the HP Brio BA 200 as non-Linux compliant hardware on your web site. It is not compliant at the kernel level, and probably has PCI/IDE race conditions that put users at risk of data corruption, not to mention ppp problems. I wouldn't ask, except that their web site shows that this model works fine with Linux and they won't even acknowledge my messages to the contrary. I suffered through hours of troubleshooting to find their problem after being lied to by their web site marketing and ignored by their product support (not to mention that their duplicity occupied Red Hat support for some of those hours). At least Red Hat might be able to keep someone else from suffering the same fate.
I'm going to follow this one up with the web team.