home *** CD-ROM | disk | FTP | other *** search
- On Nov 23, 8:57pm, Patrick Vervoorn wrote:
- > I have tried the brand-new vmunix-040.713 kernel, but it, much to my
- > regret, doesn't do much on my Amiga 4000/040, though comparably a lot more
- > than previous kernels... :)
-
- Ah - someone finally willing to try it out on the A4000.
-
- > The first problem is (probably) the AGA chipset; if I boot with it enabled,
- > the console screen that appears is a bunch of garbage. Selecting the ECS
- > setting in the 3.0 bootmenu fixes this. Typing
-
- That would make sense - the current kernel wouldn't know anything about
- the AGA stuff.
-
- > 1> RunShit loadbsd.713 vmunix-040.713
- >
- > also fixes this problem. RunShit is a rather useful "getting euro-demos to
- > work"-utility. Too bad I have to use it with NetBSD :)
-
- Is the source to RunShit available? If it's easy to select the ECS mode
- from software, loadbsd could be modified to do this. This would be
- a workaround until the kernel can initialize things properly.
-
- > The second problem happens during booting; since I know of no way to
- > redirect the console output to a file (and I don't even have a NetBSD
- > partition yet (my HD died sunday :()), I have written it on a piece of
- > paper and will type it from this piece of paper; I might have missed
- > something or used the wrong character case (upper/lower), so there might be
- > some errors. All goes well upto and including the first line I quote, after
- > that the sh*t hits the fan...
- >
- > ---------------------------------------------------------------------------
- >
- > using 102 buffers blabla....
- > unexpected trap (vector offset 30) from 10009
-
- This is rather strange, as that trap is defined as "reserved, undefined"
- according to my 68040 manual. I'll have to look at what's at that pc
- location.
-
- > pid 0: kernel format exception
-
- Looks like a stack frame got clobbered, or maybe the 040 doesn't accept
- the stack frame from the previous exception.
-
- > trap: bad kernel access at 10
- > trap type 8, code = 485, v = c0
- > pid = 0, pc = 00059F68, ps = 2018, sfc = 0001, dfc = 0001
- > registers:
- > 0 1 2 3 4 5 6 7
- > preg FFFFFFF7 FFFFFFFF 00000001 00000000 0000008A 0000000C 0000000D 00000000
- > areg 00000000 01C0F850 000A5E50 00000000 0009A068 00000000 FFFFFEAE 0EFFFFFC
-
- This looks like a fault in the kernel trying to access location 0x10
- (or 0xc0 - the "at 10" and "v= c0" should be printing out the same value).
- Again, I'll have to check the pc location and see what was going on.
-
- > kernel stack (FFFFFD8C)
- >
- > <kernel stack dump *NOT* hand-copied :)>
-
- I don't blame you :-)
-
- > panic: MMU fault
- > hit any key to boot/dump
- >
- > <this repeats indefinately when hitting a key>
-
- It's probably not a good idea to hit a key at this point until the
- kernel gets far enough to be able to write a dump file. Since there's
- nothing to analyze a dump file yet anyway, it's probably best to just
- reboot at this point.
-
- > All this on an Amiga 4000/040, 8MB FAST, 2MB CHIP, the plain SeaCrate
- > 120meg HD and nothing else in it. It is of course possible it crashes
- > because it finds no supported host-adapter; after my HD crash I removed my
- > GVP HC8+ (it slowed down the booting by about half a minute every time I
- > reset due to no devices on the SCSI bus)... If this is so, it might prove
- > useful to give some more meaningfull messages during bootup, indicating
- > this problem...
-
- I'm not sure, but you might be able to disable the autoboot on the GVP
- HC8+ card which would allow you to leave the card in the machine without
- a hard disk connected. I ran mine that way for a while. If I wanted to
- use the controller, I could just run binddrivers (I don't normally run it
- on bootup since I have nothing that needs it at the moment). Then again,
- I don't think this would work on the next kernel - I modified the autoconfig
- of the GVP devices to try to distinguish the Series II controller and
- the I/O extender card. I think the Series II controller with the autoboot
- disabled would look exactly like the I/O extender. [I'm going to disable
- the autoboot on my GVP this weekend since I want to try out a new ADOS
- driver for it, so I'll find out if it will work.]
-
- If the fast memory on the A4000 starts at 0x07000000 or higher, the current
- kernel will think it's running on an A3000 and attempt to configure the
- hardware as such. Specifically, it will try to access the real time clock
- [does anybody know if the RTC on the A4000 is the same or different than
- the RTC on the A3000?], and it will also attempt to access the built-in
- SCSI on the A3000. The next kernel will have a patchable location
- (_a3000_flag) that can be patched to 0 to disable the A3000 configuration
- setup. I suspect that the A4000 may need a special configuation setup.
-
- > I hope this means something to all you kernel hackers; I'm not one of them,
- > though I've tried to compile the kernel under AmigaDOS. To no avail, which
- > brings me to my third problem;
-
- Since I've been doing the 040 support, it does mean something to me and
- I will do what I can to help get it running on the A4000. Since I don't
- have any access to an A4000, it's not too easy to do.
-
- > - Until my new HD arrives (hopefully at or before the end of this week), I
- > can't even try to write the rootfs to my HD. Does the current rootfs still
- > work with the '040 patches (or a '040 kernel)? Or is it hopelessly
- > out-of-date?
-
- I've booted my '040 kernel using the original rootfs and it works well
- enough to boot up to the root prompt and allow access to the root file
- system. Unfortunately, the 713 and later kernels have changed some
- data structures so that the old newfs will not run. It should be easy
- enough to transfer the new newfs to the root partition.
-
- > - Could the author of the rather handy (or so the message claimed) "raw
- > partition write"- util (I forgot its AND the author's name: sorry about
- > that) please upload it to ftp.eunet.ch? I don't have a spare HD and will
- > add NetBSD to my new HD and would like to, when possible, keep the other
- > partitions intact.. ;-)
-
- If you are talking about the dcp program by Chris Hooper, it should be
- available at ftp.mtu.edu in the directory /pub/cdh. It's much easier to
- use dcp than filetodev [I never bothered trying to use filetodev.]
-
- > - Is there a way to check where you write the rootfs? I know the method to
- > calculate the blocknumber where you should write it is in the FAQ (and I
- > understand it fully), but I suppose you could also use the calculated
- > numbers with the 'devtofile' command to read the blocks you intend to
- > overwrite into a file. Can this file be examined to check if you are indeed
- > 'poking' at the right spot? What should I look for in this file?
-
- If you have the RDB information correct, then dcp will access the correct
- location.
-
- > - A question about the current GVP SCSI driver. Does it take advantage of
- > possible 16-bit RAM which I could put on the HC8+? This would be the
- > fastest possible temporary buffer (of minimally 2MB) for the HC8+ to DMA
- > the data to and the fastest possible buffer to CPU-copy it out of. (Besides
- > having 32-bit RAM within the ZORRO-II range, something which is, AFAIK,
- > impossible on the A4000/040).
-
- No, the current GVP SCSI driver does not take advantage of the Zorro II
- memory. You need to patch _scsi_no_dma if your fast memory is outside
- the Zorro II address space. I've had some thoughts on doing some things
- dynamically for both the GVP and A2091 drivers so that it will use DMA
- if the buffer is withing the Zorro II address space and use PIO if the
- buffer isn't. I'd also like to try setting up a chip memory buffer that
- could be used to do DMA if the buffer is outside the Zorro II address
- space (there isn't any convenient way at the moment to have it use 16
- bit memory, which would be better - but DMA using chip memory should
- be better than PIO).
-
- > And, to ease my decision, when will there be a NetBSD driver for either the
- > A4091 or the Z3 Fastlane? :) (or any other Fast-SCSI-2 adapter that might
- > emerge.)
-
- I've got the 53C710 [got that number correct this time!] drive reorganized
- to allow supporting different cards, so the A4091 support should be easy
- to add once all the required hardware details are known.
-
- Michael
-
- --
- Michael L. Hitch INTERNET: osymh@montana.edu
- Computer Consultant BITNET: OSYMH@MTSUNIX1.BITNET
- Office of Systems and Computing Services
- Montana State University Bozeman, MT USA
-
-