To add a little more heat to this frey, it seem to me that the
most interesting new function in this driver is that it's able to set
the hardware flow control.  As I understand it, the 'stop' signal on
the 16550 must be set by software (unless you have a slightly better
16550).  From reading all the messages that I have seen go by, it
seems to me that the heavy use of the 'ed2' (Novell 2000 compatible
ethernet card) could be part of the problem.

	The 386 in question uses almost entirely NFS mounted disks,
(local swap and root) and almost any serial activity implies ethernet
activity --- incoming ppp packets are almost always for other machines
on the net and incoming uucico packets are written to an NFS mounted
spool directory.

	Even if this were fixed properly, it would be desireable that
the serial drivers had provision for high-priority setting of the flow
off bit on stupid serial chips (which seems to be about 99% of serial
chips out there).

Dave.

-- 
----------------------------------------------------------------------------
|David Gilbert, PCI, Richmond Hill, Ontario.  | Two things can only be     |
|Mail:      dgilbert@jaywon.pci.on.ca         |  equal if and only if they |
|http://www.pci.on.ca/~dgilbert               |   are precisely opposite.  |
---------------------------------------------------------GLO----------------