🚀 go-pugleaf

	Having recently upgraded my kernel version from 1.1.52 to 1.1.81 and 
then applying patches when available to 1.1.90, I'm experiencing some weird 
behaviour with serial port io.

System configuration
~~~~~~~~~~~~~~~~~~~~
Machine: Hewlett Packard 386/25N, 8Mb Ram, Math Co-processor
Disks:   Internal Quantum PD170 IDE, External CDC 94161-9 SCSI on AHA1510
         host adaptor.
Serial Ports: 2 x 16450 Uarts
Kernel: 1.1.90 

Symptoms
~~~~~~~~
	The symptoms show up as a reduction in serial io speed when there
is any disk activity either to the internal IDE or external SCSI drive,
this manifests as a significant retrying of TCP/IP packets over my dial-up
slip link to my provider and also to a local machine using PPP under KA9Q.
Both serial ports are currently running at 38400 and even reducing them
down to as low as 4800 doesn't seem to cure the problem.  Incidently even
at 38400 there are no reports of serial overruns in the syslog and have
successfully used this configuration under both 1.0.9 and 1.1.52 with no
loss in throughput. 

	Having quickly looked at the kernel source it appears the
interrupt processing has changed to a queued method for most devices
(certainly serial & tty) but apparently not for disk related activity.
Could it be that this is the cause of the problem and has anybody else 
noticed these symptoms?

Forever living in hope :-)

	Steve
--
(=====)-----------------------------------------------------(=====)
 | / |       Steve Wyles             steve@g6vju.demon.co.uk | \ |
 | / | Bracknell, Berkshire, UK         +44 (0)1344 423102   | \ |
[_____]-----------------------------------------------------[_____]