Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 16 May 1995 07:26:33 -0400 (EDT)
From:      Peter Dufault <dufault@hda.com>
To:        y0001415@ws.rz.tu-bs.de (Martin Butkus)
Cc:        freebsd-bugs@FreeBSD.org
Subject:   Re: big files written to scsi harddisk got corrupted
Message-ID:  <199505161126.HAA24513@hda.com>
In-Reply-To: <199505161010.DAA05529@freefall.cdrom.com> from "Martin Butkus" at May 16, 95 12:10:37 pm

next in thread | previous in thread | raw e-mail | index | archive | help
Martin Butkus writes:
> 
> Hello.
> 
> I think I have discovered a bug in the february SNAP of FreeBSD. After a 
> vanilla install, I did
> 
> yes > test.dat
> 
> in an xterm and killed yes when test.dat had a size of 100MB or so. When I 
> looked at the file I found that sometimes characters got replaced by 0xff's.
> If such a corruption happens, it always happens to 8 consecutive characters.
> When I tried to determine at which position in the file such a corruption
> occurs, I found that the positions (in bytes) are always divisible by 2048.
> I then ktrace'd the I/O of yes and there were no errors in the ktrace log file.
> I enabled SCSI debugging too and the error level returned from the drive was
> always 0x0 (no error). I have checked cables and termination but this had
> no effect on the bug. The problem could be found on my Quantum drive as well
> as on the Syquest. I presume this is a bug in FreeBSD as the drive is 
> new and the controller is known to have worked flawlessly in another machine.
> In case you need it here is my configuration:
> 
> intel 486/50
> 16 MB RAM
> Quantum lightning 730S rev 241E hard disk
> Syquest 270MB removable disk (currently not installed)
> Adaptec 1542 A SCSI controller (I use its floppy disk controller, too)
  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

This is suspect in the newer kernels, perhaps due to better clustering
and bigger I/O transfers.

Marc Ramirez was kind enough to send me a 1542A, but I won't have a
system to plug it into (with full length slots) until later this week.

Peter

-- 
Peter Dufault               Real Time Machine Control and Simulation
HD Associates, Inc.         Voice: 508 433 6936
dufault@hda.com             Fax:   508 433 5267



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199505161126.HAA24513>