Hi,

>>>>> On Mon, 18 Feb 2002 18:22:33 +1100 (EST), nigel@ind.tansu.com.au said:

nigel> Now, at first I thought it was the old ufs_open() code in
nigel> the Booter, so over the weekend I replaced it with libsa/ufs.c.
nigel> No change.
>> 
>> Does "No change" mean that a situation isn't better?

nigel> 	Correct. The magic number is still wrong, though it appears to
nigel> contain a large number instead of 0.

I don't know what is magic number, but what is this, partition number
or others?

nigel> 1) The ide_read code is bad.
nigel> Unlikely because, if I comment out the magic-number check,
nigel> ufs_open() reads and gets the diinodes, lists the files,
nigel> and seems to have a good attempt at reading the file data.
>> 
>> Maybe this. A kernel isn't booted up from a partition which is made with 
>> a recent sysinst.  It seems same in the case of SCSI disk.

nigel> 	Possible. It could be that OS X's 'newfs -O' doesn't in fact
nigel> create an old-style filesystem

Possible? Really? I couldn't make a kernel booted up from a partition
made with -current sysinst on SCSI disk. It was same on IDE disk.

>> Temporary, why don't you comment out the magic-number check?

nigel> * With libsa/ufs.c, it fails a few 4K reads into parsing the fs stuff.
nigel> * With the Booter's old ufs_read.c, it locates the directory's inode,
nigel>   locates the kernel, and gets a few K into reading it in before it
nigel>   reads past the end of the partition.

Umm. I don't know why... I can boot up a kernel from HFS, so it is no
problem for me exactly...

Takeshi Shibagaki
ie9t-sbgk@asahi-net.or.jp