Joerg Schilling wrote:
"Martin Bochnig" <[EMAIL PROTECTED]> wrote:
I miss that openness from others.
The clofi code (original idea from Knoppix) and the hsfs code haven't been
(publically!) updated since last summer (and clofi didn't build or work without
substantial modification on current ON builds) :
http://www.genunix.org/distributions/belenix_site/binfiles/hsfs_iosched.tar.gz
http://sarovar.org/project/showfiles.php?group_id=360
Let me comment on hsfs_iosched.tar.gz
The code is based on an outdated hsfs source state.
Yes, it is basically a matter of getting time to sync it up with the
latest source.
I am going to sync it up shortly.
hsfs did change massively since last year (added Joliet, ISO-9660:1999, ...
as well as new hardlink - Rock Ridge V-1.12 support).
This would be no real problem. The real problem is that the iosched code causes
a kernel panic in case you heavily use it. I did inform Moinak last year
(around September) about this problem and I did send even a kernel crash dump.
He told me he knows that understood the problem but I did not get a fix yet.
>From my tests, the code has two problems:
1) The kernel panic (see above)
I have been able to identify the possible root cause last week and
will be testing it
once I have synced up to the latest sources.
2) It slows down to ~ 1/2 the speed it gives a shoert time after the mount.
This seems to point to a coneptional problem.
Is this observed on a real CDROM device or using a loopback mounted
ISO image on
harddisk ? Also need to check whether this happens for sequential or
random I/O.
In any case there is a need to add some kstats for observing some of
the performance
statistics.
I would really like to see the code integrated in Solaris if it would not slow
down after some time of usage and if it did not cause kernel panics. The
problem is that "hsfs" is under development and for this reason, Moinak would
need to create updatewd versions and the problems need to be fixed.
As I did tell Moinak before, I am willng to review the code after it has been
fixed and Frank Hofmann told me he would also review the code.
We would need to know whether Moinak like to see the code integrated into
OpenSolaris.
Most certainly yes. I want this to go in. I will email you offline
once I have made some
changes.
Regards,
Moinak.
Jörg
_______________________________________________
opensolaris-discuss mailing list
[email protected]
