MiNT-Net 1.03 beta 2 released

[tlang@dialup.becom.net (Torsten Lang)]

Hi there,
Bernd Ebert found a serious bug in the AMD LANCE drivers in the MiNT-Net 
1.03 beta 1 package. I will upload a new release (net-1.03beta2.tar.gz)
to ftp.uni-giessen.de within the next hours.
A few words concerning the bug: The driver version released in MiNT-Net
1.03 beta 1 (first released on the MiNT'98 distribution) has got a new
feature: It uses more than one buffer for transmit (configurable, must
be a power of 2). So the driver will upload packets to the transmit
ring until it is full and then will enqueue them (the old release just
uploaded one packet and waited until the LANCE chip transmitted it).
In every TX interrupt the driver will look if there any enqueued packets
and - if yes - will upload as many as possible to the transmit ring.
Unfortunately the interrupts weren't locked at the right places - so
it happened from time to time that the lance_output() and the lance_int()
routines wrote to the same tx buffer.
In this version the bug seems to be fixed. 
Some words concerning the speed: Under NFS where big blocks are transmitted
at once (8K) the speed on a TT increased from around 140KB/s to 250KB/s
compared to the old driver which only used one TX buffer.

BTW, the NE2000 driver (commercial pci version) is also available in a
new version which now supports the Panther II (ISA adaptor for (MEGA)
ST(E)), the Milan and Hades ISA bus, the low level pci bios (Hades)
and the high level pci bios (Milan/Hades). On pci boards it uses 32bit i/o
and should be a little bit faster than the old version.

Regards,
   Torsten

P. S. I'm sorry that I didn't have the time to fix the MiNT-Net package
      concerning the land, teardrop etc. attacks. Please let me know
      against which kind of attacks MiNT-Net is sensitive.