Re: [hatari-devel] Linux-68k and REQUEST SENSE

[Uwe Seimet <Uwe.Seimet@xxxxxxxxx>]

Hi Eero,

I was referring to the Hatari logfile, i.e. a current version of what's
checked in in git. The one where you find "Strange REQUEST SENSE", even
though nothing is strange with this request.
The file is doc/m68k-linux.txt. The log file also says "SCSI BUS reset".

>=> Linux IDE driver works fine with single EXT2 partition image, but it seem
>like there could be also a problem of Linux SCSI driver expecting also a
>partition table?

I have already pointed this out earlier: If a bus reset is logged, like in
the log I am referring to, this is a low-level issue with the SCSI signals.
This is a completely different layer than the filesystem layer.
What I also mentioned earlier: Linux requires a partition table, just like
any other OS. Your image is probably just the image of a single partition,
but you need a full drive image with an Atari MBR partition table.

What's interesting about your log is that Linux uses PERSISTENT RESERVE IN,
which is not supported by Hatari, but I am quite sure its availaibility is
not mandatory.
This command is newer than SCSI-2. This confirms my assessment that having
SCSI-1 the default setting for Hatari is not that good.
When trying to run Linux, you should use SCSI-2. Just like IMO for the
Atari.

So all in all these are the things that I think are somehow wrong:

1. The image is not a valid drive image with a partition table.
2. There seems to be a low-level signal issue with the SCSI protocol.
3. The REQUEST SENSE command sent is not strange, the logging is just wrong.
4. Linux is aware of newer SCSI standards than SCSI-2 and tries to use
commands not covered by SCSI-2. Therefore, Hatari should be configured
with the latest SCSI level it supports, in order to accommodate Linux as
good as possible.

Hope that helps.

Best regards

Uwe