Re: Amiga Kernel 2.4.20 Lockup
Hi Geert,
It looks like there was some discussion which was not cc'd back
to debian-68k@lists.debian.org. I cc'd m68k-build@nocrew.org
on the original post. Just a bit more to add based on the replies.
> On 16 Mar 2003, Kars de Jong wrote:
> > On Sun, 2003-03-16 at 08:38, Christian T. Steigies wrote:
> > > On Sun, Mar 16, 2003 at 12:43:10AM +0000, Lance Tagliapietra wrote:
> > > > I am able to get the 2.4.20 kernel for the Amiga to lock up
> > > > with the following command sequence as root user:
> >
> > <snip>
> >
> > > > A dmesg of my system is available at
> > > >
> > > > http://www.luminet.net/~lancetag/dmesg
> >
> > >From the dmesg:
> >
> > SCSI subsystem driver Revision: 1.00
> > wd33c93-0: chip=WD33c93A/9 no_sync=0xff no_dma=0 debug_flags=0x00
> > setup_args=,,,,,,,,,
> > Version 1.25 - 09/Jul/1997, Compiled Mar 5 2003 at 02:20:39
The CPU is a GVP 030 Combo Card with 16M RAM installed. There is a GVP
RAM card with another 6M enabled (8M installed, but 2M of the ZorroII
space is being used by the EGS Spectrum display card, as shown in the
dmesg. The ethernet card is an A2065. There is a multi I/O card that
I have not tested under linux at all yet.
I have another SCSI tape drive available for testing also, a Tandberg T3600
(150/250M) drive. But I'l like to see this one working.
> >
> > >
> > > So 2.4.20 is not less table than 2.2.20? Thats good news, I haven't had any
> > > problems yet:
> > > 02:28:57 up 4 days, 5:36, 1 user, load average: 2.07, 2.04, 2.06
> > >
> > > Lets ask the kernel experts if they know about your problem, I haven't had a
> > > tape drive connected to my Amiga for at least 3 years...
> > >
> > > IIRC your scsi device is a GVPII? That one used to freeze my system even when
> > > there was no SCSI device connected to that controller...
> >
> > I have made a lot of bugfixes to the WD33C93 driver last year (sorry,
> > didn't clean them up enough yet for commit) to make it work with the
> > revision 0 chip and an old 150 MB QIC tape drive. Before this it used to
> > lock up the system all the time. But I got tired of having to run
> > Motorola SYSV/68 SVR3 UNIX to be able to read a tape so I fixed it.
> >
> > I see you have a revision A chip, but some of the fixes were generic
> > (mostly about timing problems in the driver).
> >
> > If you're daring I can send you the patch so you can test it.
I did not realized that there are different versions of the chips that
cause required work arounds on the GVP Combo cards. I have another
of these cards, which I'll be putting in another A2000 system I have
(which is currently disassembled from the replacement of its real time
clock battery) which I'll be able to test with.
I am not set up to build kernels here. If Christian or someone else
is willing to compile/build them, I'll give them a nice stress test on
my system here.
>
> Aha, so you sound like the right person to fix the remaining issues with the
> wd33c93 driver in 2.5.x? ;-)
>
> Osamu Tomita just posted a patch to lkml to implement .eh_abort_handler and
> .eh_host_reset_handler in a generic way, but his .eh_bus_reset_handler is
> PC98-specific. Hence we still need an implementation of .eh_bus_reset_handler
> for A3000, A2091, GVP II, MVME147 (and SGI WD93), or a generic version.
One thing that does seem better about the 2.4.20 kernel on the Amiga is
support of memory. In the 2.2.10 and 2.2.20 kernels I have to use a
memfile to disable the last 1M of 32-bit memory on the GVP Combo card.
If I don't, I'll see problems like the login message does not get displayed.
I've been running with the system auto-detecting the memory size for this
kernel. I have done various memory tests under AmigaDos, so was fairly
convinced that it was not a memory chip problem, and swapping the SIMMs
around did not seem to move the problem.
Noting this, I'll try the tape backup with the memfile used. The memfile
will be available at http://www.luminet.net/~lancetag/memfile.
Thanks for taking the time to discuss this issue. Please feel free to
email me if addition information is needed.
[...]
--Lance
Reply to: