Ethercard driver status and buglist

________________
Resolved bugs remaining in the pl14 drivers:
HP driver:
The hp27*** driver incorrectly recognizes the new HP J2405A LANCE ethercard.
An extra probe check is in the new driver, pl15/hp.c

NE*000 driver:
Some NE2000 clone boards are not detected at boot time with pl12+ kernels.
This bug was initially a little confusing, because the same code finds them
later.  A recently suggested fix was to get rid of a SLOW_DOWN_IO, and seems
to fix the problem.  The new code is in pl15/ne.c.

ATP pocket driver:
Receiver will hang, allowing transmit packets bug no receive packets.
I have a new version that resets the device after 8 Tx-with-no-Rx cycles.
12/22/93 Updated version that adds time-since-last-rx to decide if the
receiver needs to be reset.  The driver now seems to be very reliable.
(A better solution is to detect this when it happens, but I haven't figured
out how to detect that.)

EtherExpress and 3c507 drivers:
Receive and transmit queues will occasionally become corrupted.
The i82586 raises an exception, and the current code noisily rewrites the
queues and resets the chip.  I have evidence that this is a hardware bug,
so changing the IE_DEBUG level to just omit the message (but still increment
the errors count in 'stats') is probably the right solution.
12.22.93 I've received a report that the driver works fine on slower machines,
but occasionally fails with fast machines.  Adding delays fixes the problems.

________________
Unresolved issues:

net_init.c:
Only the new LANCE driver uses this file.  I have a more complete version that
supports the other drivers and (among other features) incorporates Drew's new
initialization scheme.  Should these changes wait for post-1.00?

LANCE:
I reworked the Tx queue handling on the pl14 version to incorporate bounce
buffers, and I seem to have introduced a bug.  The effects can be masked by
increasing the Tx queue to the old value of 16 entries from the pl14 value of
4 entries with '#define TX_RING_LOG_SIZE 4'.
12.22.93 The old version likely had this bug as well, and I can't reproduce
the problem with a ring size of 16, so that might become the final fix.

3c509:
The driver will occasionally lose an interrupt (?).
The driver checks for this condition and recovers, so the message can be
safely ignored.  I've spent some time trying to track this down, but none of
my changes has significantly increased the reliability over the pl13 version.
Dropping the DEBUG level for 1.0 will silence this message (but still log an
error). 

AC3200
I lost the alpha tester just as the driver started working.  This is a fairly
generic 32 bit EISA shared memory 8390 (WD8013-like) design, so most of the
complexity is in the 8390 core.

Cabletron E2100 series
I made a statement that this driver was at the bottom of the list, but I
actually wrote it about a month ago.  It seems stable, but should be tested.