Port-atari archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: scsi freezing problem with NetBSD current



My understanding was that a lot of DMA on the TT030 is to ST-ram, but the SCSI DMA specifically does use TT-ram. Doing some quick digging, I found:
http://www.vanc.igs.net/~roughley/overview.htm#SCSI

Which says:

"The TT's SCSI direct-memory-access subsystem can send and receive information equally well, both from ST RAM and faster TT RAM, meaning that present and future software will be able to exploit SCSI peripherals efficiently."


There's a great archive of TT030 hardware info that may be useful as well:
http://dev-docs.atariforge.org/html/search.php?find=_4

As far as the DEBUG kernel below goes... On first boot I got a panic after the boot message telling me the filesystem is clean. "Mutex error: mutex_vector_enter: locking against myself" Then it had a few lines of debug spew and "panic: lock error"

So I rebooted (still with the same kernel) and now it's busy chugging away validating the filesystem. Not sure if it will make it past that or not. I've got to head off to sleep now so I'll report back with more info tomorrow.

I've been fixated on 1.6.1 for the last few months but now that the kernel bug is fixed I want to check out 4.0 again. David, if you can clarify what some of the "manual fixups" are that you mention are, that would be very helpful. The next roadblock I'm aware of is the inability to get the more recent installers to drop a working bootstrap on the drive (or even upgrade an install that already has one).

And I second your call for some heroic developer to work on fixing up the installer. =)

David Ross
dross%pobox.com@localhost

----- Original Message -----
From: "David Brownlee" <abs%NetBSD.org@localhost>
To: "T. Makinen" <tjamaloo%gmail.com@localhost>
Cc: <dross%pobox.com@localhost>; <port-atari%netbsd.org@localhost>
Sent: Tuesday, October 28, 2008 6:12 PM
Subject: Re: scsi freezing problem with NetBSD current


On Wed, 29 Oct 2008, T. Makinen wrote:

On Tue, Oct 28, 2008 at 2:17 PM, David Brownlee <abs%netbsd.org@localhost> 
wrote:

       As a less invasic change it should be possible to adjust
       scsi_dmaok() to return 0 in the CT63 accelerator case...

You're right, that way other systems are not affected by possible change.

       Hmm, Does the standard falcon System RAM and TT RAM similar
       to the TT030? If the standard falcon only has TT RAM then it
       should not be too hard to change the DMA behaviour when an
       accelerator with additional RAM is enabled.

Falcon can have up 14MB of ST-ram and 512MB (with CT60/63) of TT-ram. So,
standard Falcon has only ST-ram. I reckon that TT can have up to 12MB of
ST-ram and 256MB of TT-ram.

 Ah, so presumably both Falcon and TT030 can DMA to ST-ram only?
 Assuming so, and also assuming that NetBSD is working fine for
 scsi on the TT030, we are in one of:
 - Falcon with TT ram needs the same treatment as TT030 and the
   code is treating them differently
 - Falcon with TT ram needs different treatment to TT030 and the
   code is treating them the same
 - Falcon with TT ram needs different treatment to TT030 and the
   code is treating them differently but in a different way :)

 OK, that was stating the obvious, but sometimes that helps...
 Its probably worth taking a pass through the code looking
 for places where it treats the Falcon and TT030 differently
 - there may be a case where the Falcon case is shortcircuiting
 a check for TT-ram because its assuming it never has any.

Yes, here are links to dmesg files:
http://koti.welho.com/tmakinen/atari/CT63-netbsd-dmesg
http://koti.welho.com/tmakinen/atari/Falcon-netbsd-dmesg

 Great - thats nice and clear, ST-ram low, TT-ram >= 0x01000000

 CT63 68060:
     pmap_init:  0: 0060a000 - 00dd4000 (   8167424)
     pmap_init:  1: 01000000 - 20eb2000 ( 543670272)
     total memory = 524 MB

 Falcon 68030:
     pmap_init:  0: 00212000 - 00dd2000 (  12320768)
     total memory = 14336 KB

 Did you have to remove/disable 4MB of RAM for the CT63 case?
 Its probably not relevant, just curious.


NetBSD current kernel with DEBUG enabled pmap.c:
http://koti.welho.com/tmakinen/atari/netbsd-current-atari-pmap_DEBUG-BOOT-20081029.gz

 Great - David, could you test? :)

       As an aside - NetBSD is always looking for more developers on
       the less mainstream platforms :)

Well, I'm new to NetBSD, but I'm happy to offer any help I can :)

 Fixing the scsi bug has been an *excellent* way to start! :)

 There are plenty of different areas in which developers can help

 - Kernel: Fixing bugs, supporting new hardware (in the atari case
   that would be the NETUSBEE, which probably involves building
   one from the plans available, so thats quite a project!),
   performance tuning, or just updating the port to some of the
   newer MI (machine independent) features - such as switching the
   display, keyboard and mouse to wscons with virtual terminals

 - Installer: for recent versions of NetBSD/atari the installer
   doesn't work correctly and needs manual fixups to get the
   installed system working. Ideally this needs someone with
   the skill to fix it _and_ a machine on which they can see
   the issue...

 - X: I don't know if the Atari X server uses the blitter. If not,
   that could be a nice speed boost (Thoughits not going to help
   the TT030 users)

 - Userland/pkgsrc: A Falcon with a CT63 and a chunk of memory
   could probably run firefox (albeit not that fast :), but as
   firefox needs some C++ glue for each architecture, and I don't
   think anyone ever wrote this for gcc4 and NetBSD/m68k.

 - Documentation/help pages - Its always easier for someone else
   to get a system up and running if someone else has done it first
   and left a trail to follow

 Its a question of picking something that interests you and
 running with it. After all, this stuff is supposed to be
 fun (and satisfying)

--
 David/absolute       -- www.NetBSD.org: No hype required --





Home | Main Index | Thread Index | Old Index