English
资源说明：kernel module for EPLIP (Enhanced PLIP)
version 0.5.6 Default operation mode now is ECP FIFO DMA if the dma channel was detected.
No more support for AUTODIN II CRC frame checksumming.
Added new module parameter "hwaddr" which is used to set rightmost
four octets of the ETH MAC address(it's usefull for using eplip with DHCP|BOOTP).
EPLIP is no longer Point to Point and runs with ARP( This was necessary in order to
connect EPLIP to WEPLIP - because NDIS does not allow MEDIA_802_3 and PPP flags on the same device ).
So you should no longer bring the eplipX with ifconfig eplipX xxx.xxx.xxx.xxx Pointopoint xxx.xxx.xxx.xxx up
but rather like a normal ETH net interface.
version 0.4
Added Makefile/Rules.make to build EPLIP driver.
Don't use build.sh script anymore, just type "make" to build
and "make install" to copy eplip.o into "/lib/modules/$VERSION/kernel/drivers/net" directory
version 0.4pre1:
All written below for eplip-0.3 applies to 0.4pre1. Here I'll list only
changes made from 0.3 to 0.4:
version 0.3:
version 0.1:
This is the first version of Enhanced Parallel Line IP.( 0.1 ).
It is based on the original PLIP driver, but is intended to run
on ECP ports using a so called "FAST ECP" cable.( see WIRING in the dist tree).
On my test machines this scratch gets more than 5 times faster transfers than PLIP.
( 60-70 KB/s raw transfers with PLIP - 300-400 KB/s EPLIP - Actually I get 30K ftping
from remote to /dev/null with PLIP and 350-370K with EPLIP )
I.What you need:
0. The source tarball for eplip -eplip-0.3.tar.gz
1. "FAST ECP" cable: as described in the file WIRING.( without 1k resistors on data lines)
2. linux-2.2 or 2.4 kernel sources.
3. To build the driver issue ./build.sh from within the source directory( You don't have to , and
it is __not__ advisable to extract the tarball in the kernel source tree - if gcc complains about includes
edit the build.sh file and make sure you include path is set properly by this script )
If you C Compiler complains about unrecognized "mpreferred-stack-boundary" option edit the "build.sh" and
comment the line:
ALIGN_CFLAGS="-mpreferred-stack-boundary="$STK_ALIGN
4. insmod ./eplip.o io=0x378 ( if you ECP is @ iobase 0x378 - or replace the __io__ param with the appropriate one )
You may need to supply the DMA channel( etc. insmod ./eplip.o io=0x378 dma=3 ) if the this is not
detected automatically( currently DMA autodetection fails for MIOs integrated in VIA686A South bridge -
this was tested and found to not work on an ABIT KT7 ).
5. if this succeeds and an ECP capable port was detected @ this address you may proceeded with 6
6. ifconfig eplip0 xxx.xxx.xxx.xxx. pointopoint xxx.xxx.xxx.xxx up
for example : ifconfig eplip0 192.168.1.1 pointpoint 192.168.1.2 up
If this succeeds activate eplip0 on the second machine. Make sure you routing is set up properly.
and try pinging from 192.168.1.1 the 192.168.2
7. If this does not work __read__ the PLIP mini howto in your documentation or if you do not have it
look at http://www.linuxdoc.org for it. EPLIP is much like PLIP, but first read this file to the __END__
( there are many limitations in this version ).
II.Limitations:
1. Works only on x86 linux
2. Does not work for PCI based ECP( or MIOs that have ECP ), or at least will not work if
the PnP subsystem has assigned second IOEXTENT Range IO address != IOBASE+0x400.
This won't need any fixing when eplip becomes parport-ish.
3. Does not work for ECPs which have FIFOs with more than 8bit PWORD.
This also will be fixed. If I find a 16/32 bit ECP to test with or if someone fix it :)
As of version 0.5.6 this should work if eplip was compiled with EPLIP_16bit_FIFO_SUPPORT
( uncomment MACRODEFS+=EPLIP_16bit_FIFO_SUPPORT in the Makefile ), but it's not tested.
4. Could not coexist with a parport driver. Make sure you've built your kernel with parport and parport_pc
as modules, and you do not have grabbed the port you are going to use with eplip.
This will __NEVER__ be fixed. The reason is IEEE1284 handshaking uses Transfer Negotiation That is
__good__ for host<-->peripheral device connections, but if eplip follows the IEEE1284 phase transition
scheme you'll probably get you port's output latches blown, since the IDLE phase is defined as
setting direction to __WRITE__ and output is not tristated.Since the both sides are __not__ peripheral
devices, both have to be in REVERSE IDLE state, instead of being in FORWARD IDLE. This violates the IEEE1284 protocol
and since a normal implementation of driver supporting IEEE1284-94 ECP will not allow this, EPLIP
uses it's own implementation of ECP driving software.
NOTE: It is not impossible to implement a higher level transport protocol based on IEEE1284-94
aware parport driver, but this will require asymmetric connection ( Like with the Windows Direct Cable
connection - one computer is HOST and the other is GUEST - actually it is not a guest but just
behaves like a peripheral device ).
The EPLIP approach is symmetric: Both sides behave as a "peripheral" when sending and as a "host"
when receiving.
5. Many bugs probably.
Angel Valkov( whitefang.dir.bg)
本源码包内暂不包含可直接显示的源代码文件，请下载源码包。
