
insmod usage:

insmod dmfe			;;auto mode
insmod mode=0			;;Force to 10M Half Duplex Mode
insmod mode=1			;;Force to 100M Half Duplex Mode
insmod mode=4			;;Force to 10M Full Duplex Mode
insmod mode=5			;;Force to 100M Full Duplex Mode
insmod debug=1			;;Enable debug message
insmod cr6set=xxxxxxxx		;;Set CR6 default operating mode
	cr6set=0x100		;;Enable RX single port
	cr6set=0x002c0000	;;Enable HomeNet
insmod chkmode=0		;;Disable check mode for DM9102


date		version		description
---------	-------		--------------------------------------
01-18--2000	v1.28		Release Version:
				After netbench, file copy , compare test. it
				did not find problem. The detail see 
				c:\wang\linux\netbench

01-14--2000	v1.27f		Test Version:
				process send_filter_frame() used TX queue method

01-14--2000	v1.27e		Test Version:
				Used the tx queue to do the software one
				packet to solve RedHat6.1 performance low 
				problem.
				It need more Tx polling command to transmit
				sometimes.

10-08--99	v1.27		Release Version:
				Like as V1.26K, except do not display media type
				Support SROM format V3.0/V4.0
				Test kernel 2.0.36, 2.2.05, 2.2.12, 2.3.18
				10MHF, copy to/from over night OK.

10-07--99	v1.26K		Test Version:
				Add PCI_BIOS32.H for kernel 2.0.36

10-06--99	v1.26J		Test Version:
				Fixed phy_addr=1, to solve SROM format V3.0
				and V4.0 problem.
				Let DEVICE = device for under kernel 2.3.00
				else    DEVICE = net_device 
				compiler test: v2.2.5, v2.2.12, v2.3.18

09-29--99	v1.26I		Test Version:
				Deleted some unnecessary include files
				100M full duplex mode test, win98/cal2.2:
				date 9/29~9/30
				copy to/from many big files over 12hours,OK

09-27--99	v1.26H		Test Version:
				Solve problem: IO size at clean_module(),
				release IO regision must be the same range.
				Callback Time : 1 S"
				Tx timeout : 2 S"
				100M Half Duplex Mode test, win98/caldera2.2:
				date: 9/27/99 ~ 9/29/99
				1. copy to/from many big files over 24 hours,OK
				2. copy to/from thusronds files over 24 hours,OK

09-15--99	v1.26G		Test Version:
				Solve problem. At DM9102, IO size 128,
				DM9102A/DM9132 IO size 256. previous version
				used 256 always and it did not work on some PC.
08-04--99	v1.26F		Test Version:
				Solve problem. After a dynamic reset, it
				did not fill all MC address to MAC.
08-04--99	v1.26E		Test Version:
				Display the wrong packet for SERCOMM
08-04--99	v1.26B		Test Version:
				Support DM9102A+DM9181
				Addition: Tx statistic counter : collision
07-29--99	v1.26A		Test Version:
				Support DM9132 using 1 perfect and 64 bits
				hash_table.
07-22--99	v1.26		Same to V.125E. This is a release version.
07-26--99	v1.25F		Enable CR0 bit 20/22/23 (New BL, TX/RX 
				descriptor burst mode). Test version.
07-21--99	v1.25E		DM9102 don't need the check mode. New 
				parameter "chkmode"
07-20--99	v1.25D		1. DM9102A don't need this. DM9102 need :
				Check the data error problem for DM9102,
				If found the data error packet, let Dm9102
				work at the single port when rx packet <30000
				packets. Else let Chip works at dual port.
				2.init_module parse "mode" used another method.
07-16--99	v1.25C		Support DM9102AF(DM9132), Check Chip revision
				and do CR12 conversion. DM9102AF CR12 bit1=0
				link ok, bit1=1, link failed.
07-06--99	v1.25B		Init db->last_rx=jiffies, when rx packet coming
				at Rx handler.
07-01--99	v1.25A		When Rx interrupt too much(over 5000h per s")
				, I disabled the RX interrupt and enabled
				RX buffer unavailable interrupt. When RX traffic
				lower than 5000h, I enabled the RX interrupt 
				again.
06-28--99	v1.25		1.dynamic_reset() stop dm9102 before free the 
				Rx resource.
				2.Un-show the media type in dmfe_timer()
				3.Check CR8, if in the timer callback interval,
				no packet coming but CR8 had value, RESET the
				device.
06-22--99	v1.24C		dmfe_init_dm910x() reset dm9102 phyxcer.
06-16--99	v1.24A		kmalloc() need GFP_DMA flag, to allocate
				physical memory
06-14--99	v1.24		Solve the page_fault problem when remove
				dmfe module.
06-10--99	v1.23e		dmfe_free_rxbuffer() modified
06-09--99	v1.23b		After dynamic_reset(), let rx_error_cnt=0
				In Serrcom environment, It happens
				dmfe_dynamic_reset() very soon.
06-01--99	v1.23a		Addition a option "cr6set", user could set
				CR6 default value. 
				<eg> insmod dmfe cr6set=0x100 : enable rx fifo
				single port function

05-25--99	v1.23		Used new library. Kernel version over 2.035
				used "pci_" to replace "pcibios_"
				Reset DM9102 when rdes0 not F/L happen too fast.

05-20--99	v1.22		Solve page fault on kernel 2.20 when remove
				module.
				In dmfe_stop(), After we released the allocated
				memory, we need to clear the pointer.

05-07-99	v1.21		Solve the 8259 interrupt edge problem.
				Disable all interrupt in CR7, when enter
				interrupt handler. Resore the CR7 original
				value before leave the interrupt handler.

04-21-99	v1.20		Solve the copy and compare file problem
				on 100M mode.
				dmfe_start_xmit(): if tbusy=1 & tx_packet_cnt=0
				let tbusy=0 and keep transmitting

04-13-99	v1.10		Solve the transmit double lock problem.
				Let transmit timeout time=4S"
				Solve 100M under run error,
				CR6_DEFAULT enable SF/MII port
				dmfe_start_xmit() return 1 when dev->tbusy=1

12-30-98	v1.00		For slackware gcc compiler, merge all
				independent line to one line for #define

10-28-98	v0.9a		Beta version and released to TM
				Known bug: using UNIX FTP, it will happen
				"double unlock/lock device queue"
