On Sat, Jun 21, 2003 at 12:44:35PM -0400, leam wrote:
> Still using dump, sorry.  :) Trying to use the dump script to write to a 
> cleanly made partition and I get:
> 
>  extract file ./tar_files/base.tgz
>    DUMP: readBlocks: lseek fails: Invalid argument
>    DUMP: rawread: lseek fails
>    DUMP: rawread: lseek2 fails: Invalid argument!
>    DUMP: rawread: lseek2 fails: Invalid argument!
>    DUMP: rawread: lseek2 fails: Invalid argument!
>    DUMP: rawread: lseek2 fails: Invalid argument!
>    DUMP: rawread: lseek2 fails: Invalid argument!
>    DUMP: rawread: lseek2 fails: Invalid argument!
>    DUMP: rawread: lseek2 fails: Invalid argument!
>    DUMP: rawread: lseek2 fails: Invalid argument!
>    DUMP: SIGSEGV: ABORTING!
>      DUMP: SIGSEGV: ABORTING!
>      DUMP: SIGSEGV: ABORTING!
>  DUMP: DUMP: SIGSEGV: ABORTING!
>  SIGSEGV: ABORTING!
>  Segmentation fault - core dumped
> 
> This works in 1.6T. I thought 1.6.1 had fixed the dump issue. Or am I 
> hitting something else?
> 

The problem I'm aware was manifested in silent data corruption during
dump, and it was iommu streaming cache or in other words ultra2 for sure
and other higher models. I didn't requested pullup for that primarily
because I didn't have 1.6 system handy and time to make 'full circle'.
So that problem is still in 1.6.1 (unless I missed Martin's requests).

Your problem appears differently.
Is there a possibility that you're trying 16T executable against
1.6.1 kernel? I'd try ktrace on dump to see what's going on with
those lseek arguments.

	Andrey