Root/Documentation/scsi/in2000.txt

1
2UPDATE NEWS: version 1.33 - 26 Aug 98
3
4   Interrupt management in this driver has become, over
5   time, increasingly odd and difficult to explain - this
6   has been mostly due to my own mental inadequacies. In
7   recent kernels, it has failed to function at all when
8   compiled for SMP. I've fixed that problem, and after
9   taking a fresh look at interrupts in general, greatly
10   reduced the number of places where they're fiddled
11   with. Done some heavy testing and it looks very good.
12   The driver now makes use of the __initfunc() and
13   __initdata macros to save about 4k of kernel memory.
14   Once again, the same code works for both 2.0.xx and
15   2.1.xx kernels.
16
17UPDATE NEWS: version 1.32 - 28 Mar 98
18
19   Removed the check for legal IN2000 hardware versions:
20   It appears that the driver works fine with serial
21   EPROMs (the 8-pin chip that defines hardware rev) as
22   old as 2.1, so we'll assume that all cards are OK.
23
24UPDATE NEWS: version 1.31 - 6 Jul 97
25
26   Fixed a bug that caused incorrect SCSI status bytes to be
27   returned from commands sent to LUNs greater than 0. This
28   means that CDROM changers work now! Fixed a bug in the
29   handling of command-line arguments when loaded as a module.
30   Also put all the header data in in2000.h where it belongs.
31   There are no longer any differences between this driver in
32   the 2.1.xx source tree and the 2.0.xx tree, as of 2.0.31
33   and 2.1.45 (or is it .46?) - this makes things much easier
34   for me...
35
36UPDATE NEWS: version 1.30 - 14 Oct 96
37
38   Fixed a bug in the code that sets the transfer direction
39   bit (DESTID_DPD in the WD_DESTINATION_ID register). There
40   are quite a few SCSI commands that do a write-to-device;
41   now we deal with all of them correctly. Thanks to Joerg
42   Dorchain for catching this one.
43
44UPDATE NEWS: version 1.29 - 24 Sep 96
45
46   The memory-mapped hardware on the card is now accessed via
47   the 'readb()' and 'readl()' macros - required by the new
48   memory management scheme in the 2.1.x kernel series.
49   As suggested by Andries Brouwer, 'bios_param()' no longer
50   forces an artificial 1023 track limit on drives. Also
51   removed some kludge-code left over from struggles with
52   older (buggy) compilers.
53
54UPDATE NEWS: version 1.28 - 07 May 96
55
56   Tightened up the "interrupts enabled/disabled" discipline
57   in 'in2000_queuecommand()' and maybe 1 or 2 other places.
58   I _think_ it may have been a little too lax, causing an
59   occasional crash during full moon. A fully functional
60   /proc interface is now in place - if you want to play
61   with it, start by doing 'cat /proc/scsi/in2000/0'. You
62   can also use it to change a few run-time parameters on
63   the fly, but it's mostly for debugging. The curious
64   should take a good look at 'in2000_proc_info()' in the
65   in2000.c file to get an understanding of what it's all
66   about; I figure that people who are really into it will
67   want to add features suited to their own needs...
68   Also, sync is now DISABLED by default.
69
70UPDATE NEWS: version 1.27 - 10 Apr 96
71
72   Fixed a well-hidden bug in the adaptive-disconnect code
73   that would show up every now and then during extreme
74   heavy loads involving 2 or more simultaneously active
75   devices. Thanks to Joe Mack for keeping my nose to the
76   grindstone on this one.
77
78UPDATE NEWS: version 1.26 - 07 Mar 96
79
80   1.25 had a nasty bug that bit people with swap partitions
81   and tape drives. Also, in my attempt to guess my way
82   through Intel assembly language, I made an error in the
83   inline code for IO writes. Made a few other changes and
84   repairs - this version (fingers crossed) should work well.
85
86UPDATE NEWS: version 1.25 - 05 Mar 96
87
88   Kernel 1.3.70 interrupt mods added; old kernels still OK.
89   Big help from Bill Earnest and David Willmore on speed
90   testing and optimizing: I think there's a real improvement
91   in this area.
92   New! User-friendly command-line interface for LILO and
93   module loading - the old method is gone, so you'll need
94   to read the comments for 'setup_strings' near the top
95   of in2000.c. For people with CDROM's or other devices
96   that have a tough time with sync negotiation, you can
97   now selectively disable sync on individual devices -
98   search for the 'nosync' keyword in the command-line
99   comments. Some of you disable the BIOS on the card, which
100   caused the auto-detect function to fail; there is now a
101   command-line option to force detection of a ROM-less card.
102
103UPDATE NEWS: version 1.24a - 24 Feb 96
104
105   There was a bug in the synchronous transfer code. Only
106   a few people downloaded before I caught it - could have
107   been worse.
108
109UPDATE NEWS: version 1.24 - 23 Feb 96
110
111   Lots of good changes. Advice from Bill Earnest resulted
112   in much better detection of cards, more efficient usage
113   of the fifo, and (hopefully) faster data transfers. The
114   jury is still out on speed - I hope it's improved some.
115   One nifty new feature is a cool way of doing disconnect/
116   reselect. The driver defaults to what I'm calling
117   'adaptive disconnect' - meaning that each command is
118   evaluated individually as to whether or not it should be
119   run with the option to disconnect/reselect (if the device
120   chooses), or as a "SCSI-bus-hog". When several devices
121   are operating simultaneously, disconnects are usually an
122   advantage. In a single device system, or if only 1 device
123   is being accessed, transfers usually go faster if disconnects
124   are not allowed.
125
126
127
128The default arguments (you get these when you don't give an 'in2000'
129command-line argument, or you give a blank argument) will cause
130the driver to do adaptive disconnect, synchronous transfers, and a
131minimum of debug messages. If you want to fool with the options,
132search for 'setup_strings' near the top of the in2000.c file and
133check the 'hostdata->args' section in in2000.h - but be warned! Not
134everything is working yet (some things will never work, probably).
135I believe that disabling disconnects (DIS_NEVER) will allow you
136to choose a LEVEL2 value higher than 'L2_BASIC', but I haven't
137spent a lot of time testing this. You might try 'ENABLE_CLUSTERING'
138to see what happens: my tests showed little difference either way.
139There's also a define called 'DEFAULT_SX_PER'; this sets the data
140transfer speed for the asynchronous mode. I've put it at 500 ns
141despite the fact that the card could handle settings of 376 or
142252, because higher speeds may be a problem with poor quality
143cables or improper termination; 500 ns is a compromise. You can
144choose your own default through the command-line with the
145'period' keyword.
146
147
148------------------------------------------------
149*********** DIP switch settings **************
150------------------------------------------------
151
152   sw1-1 sw1-2 BIOS address (hex)
153   -----------------------------------------
154    off off C8000 - CBFF0
155    on off D8000 - DBFF0
156    off on D0000 - D3FF0
157    on on BIOS disabled
158
159   sw1-3 sw1-4 IO port address (hex)
160   ------------------------------------
161    off off 220 - 22F
162    on off 200 - 20F
163    off on 110 - 11F
164    on on 100 - 10F
165
166   sw1-5 sw1-6 sw1-7 Interrupt
167   ------------------------------
168    off off off 15
169    off on off 14
170    off off on 11
171    off on on 10
172    on - - disabled
173
174   sw1-8 function depends on BIOS version. In earlier versions this
175   controlled synchronous data transfer support for MSDOS:
176      off = disabled
177      on = enabled
178   In later ROMs (starting with 01.3 in April 1994) sw1-8 controls
179   the "greater than 2 disk drive" feature that first appeared in
180   MSDOS 5.0 (ignored by Linux):
181      off = 2 drives maximum
182      on = 7 drives maximum
183
184   sw1-9 Floppy controller
185   --------------------------
186    off disabled
187    on enabled
188
189------------------------------------------------
190
191   I should mention that Drew Eckhardt's 'Generic NCR5380' sources
192   were my main inspiration, with lots of reference to the IN2000
193   driver currently distributed in the kernel source. I also owe
194   much to a driver written by Hamish Macdonald for Linux-m68k(!).
195   And to Eric Wright for being an ALPHA guinea pig. And to Bill
196   Earnest for 2 tons of great input and information. And to David
197   Willmore for extensive 'bonnie' testing. And to Joe Mack for
198   continual testing and feedback.
199
200
201            John Shifflett jshiffle@netcom.com
202
203

Archive Download this file



interactive