Re: obsd: nothing but trouble on ...

[Darik Horn <darik_(_at_)_gulf_(_dot_)_waterways_(_dot_)_geeky_(_dot_)_net>]

> "aha_scsi_cmd cannot map" which comes from the Adaptec AHA-1542 driver.

I'll put a meetoo in on that error.

I thought that my SCSI problems were being caused by a cdr burner.  
However, with only a SCSI disk on the bus, I can can provoke "aha_scsi_cmd
cannot map" by compiling things from the ports tree.  Any disk-intensive
job that runs for more than ten minutes will crash the computer.

The error appears on the console twice.  The machine responds to pings
afterwards, but everything else is dead.

[]

On Thu, 5 Oct 2000, Laurence Moore wrote:

> This sounds very familiar.
> 
> I have experienced a similar problem with an old 486MB using a Hypertec
> AMD-586 not that this hardware is a problem as I can run Solaris 7 and
> various flavours of Linux on it without any problem. However OpenBSD 2.7
> is another story. I haven't filed a bug report on it yet as I have other
> matters to attend to so it isn't a priority but I do get an error message
> "aha_scsi_cmd cannot map" which comes from the Adaptec AHA-1542 driver.
> 
> The symptom I observe is identical to yours and I can reproduce when
> running two concurrent compiles on this computer. I can also make it occur
> when Samba, Squid & Postfix are running but doing absolutly nothing (i.e.
> they are idle processes) and comiling a kernel. It took me ages to workout
> why I was having this problem when compiling a kernel. I was able to
> successfuly compile without a freeze when the above services were all
> shutdown.
> 
> Now, down to what the cause of my problem is. Fortunatly I was running
> performance meter on my Sun monitoring the OpenBSD system. During a
> compilation of the kernel I saw "swap" activity and at this precise
> instant almost all the lines dropped to zero. Of course the network was
> still running as I was continuing to receive updates from rstatd but the
> console appeared frozen. I could CTRL-C out of the make process and land
> back to the prompt however doing an ls or running any other command that
> wants to access the disc it then freezes. Remember that error message I
> get! Basically disc I/O operations have come to a grinding halt so nothing
> more functions other than processes that are running from physical memory.
> I did leave the computer overnight once after it had frozen to find that
> in the morning it had paniced - to be expected if a process was to be
> swapped back in and the disk wasn't accessable.
> 
> In summary of my observation, there is very little swap activiy in my
> system but maybe after 1, sometimes 2 or 3 swap occurences the system will
> freeze.
> 
> If you have a Solaris system on your network you could use the performance
> meter to monitor your OpenBSD system. It would be interesting to see if
> you observe the same as I have when the freeze occurs.
> 
> 
> Do you have any SCSI controllers in your Compaqs or are you using the IDE
> interface.
> 
> Cheers,
> 
> Larry.
> 
> On Thu, 5 Oct 2000, gaw zay wrote:
> 
> > 
> > Compaq Prolinea 4/25s
> > 
> > I have a stack of these laying around and I have tried obsd installs on 4
> > of them. On all of them there is weirdness. obsd freezes up after a short
> > time on all of them, sometimes you can still ping the machine but can't
> > use the console or log in remotly or even use a simple network service
> > such as finger. This happens [the freezes] when I do certain things like
> > this:
> > 
> > start setiathome
> > 
> > pkd_add 'name of large package'
> > 
> > The machine is still pingable 12 hours later but control never comes back.
> > 
> > The only thing in common between all these machines is the hard drive,
> > which might be the trouble, I suppose. It's a 406M HD and I don't have any
> > other HDs big enough to install on. :-( (I could try a minimal install on
> > a 120M drive I suppose)
> > 
> > The only other mention of this line of computers on this list that I can
> > find was someone mentioning they have one of these and obsd only sees
> > 16M of 24M of RAM (same thing happens to me on one of the boxes)
> > 
> > If anyone can tip me off on how to get more debug info short of compiling
> > a new kernel since I doubt I could compile anything on there so I could
> > submit more info on this it'd be appreciated. (In my mind there should be
> > a kernel with verbose debugging facilities available for every port)
> > 
> > thanks,
> > 
> > brian 
> > 
> > 
> > 
> > 
> 
>