To: Leif Neland , cc: tomppa_ERASE@fidata.fi, fabio_ERASE@cesar.unicamp.br, gcrutchr_ERASE@nightflight.com, j_ERASE@uriah.heep.sax.de, jc_ERASE@irbs.com, julian_ERASE@freebsd.org, kuku_ERASE@gilberto.physik.rwth-aachen.de, nikm_ERASE@ixa.net, tomppa_ERASE@fidata.fi, wilko_ERASE@yedi.iaf.nl, Greg Lehey , Julian H. Stacey , scsi_ERASE@FreeBSD.ORG bcc: freebsd-questions_ERASE@FreeBSD.ORG Subject: Re: bytes gets substituted on write From: Julian H. Stacey From: Greg Lehey > Date: Wed, 26 Nov 1997 13:10:32 +1030 > Message-id: <19971126131032.61824_ERASE@lemis.com> Greg Lehey wrote: > On Wed, Nov 26, 1997 at 01:06:04AM +0100, Leif Neland wrote: > > Here I go again: > > > > When copying a large file, suddenly 8 bytes in a row gets substituted > > with 0xff! > > When I repeat the copy, it's the same 8 bytes. > > Where are they in the block? > > > It also happens, when I either use ftp on this machine to get a file > > or use ftp on another to put the file. > > System: 486dx33Mhz, 8MB ram, Old Adaptec, 4 scsi's, the one with fbsd > > is a maxtor "scsi-0". > > Would the controller be a 154[02]A? People have seen this before. I > can't recall, but I seem to think that 8 bytes is about the reconnect > length for the controller. Are they located on 8-byte boundaries? > > Try tweaking your controller parameters. I don't know if this would > help, but it could help us work out what's wrong here. > > Greg Hi Leif, cc Greg et all, If it's a 1542A let me know ! I have a directory of notes on the 1542a & FreeBSD (see below). Brief summary of symptoms: Intermittent burst of 8 * 0xFF bytes on occasional multiples of (I recall) 1or4K boundaries on drives >= sd1 , mainly when multiple devices active (I think). Others apart from me have seen similar. I have an A, a B, and a C, but no complaints about the B or C. If your problem is with a B or C, I have nothing useful to contribute, in that case I suggest you just discuss with . If you have an A with problems, please mail me, and also CC the following: tomppa_ERASE@fidata.fi, fabio_ERASE@cesar.unicamp.br, gcrutchr_ERASE@nightflight.com, j_ERASE@uriah.heep.sax.de, jc_ERASE@irbs.com, julian_ERASE@freebsd.org, kuku_ERASE@gilberto.physik.rwth-aachen.de, nikm_ERASE@ixa.net, tomppa_ERASE@fidata.fi, wilko_ERASE@yedi.iaf.nl, Greg Lehey , Julian H. Stacey who all have, or have had an interest in FreeBSD & 1542A scsi cards. I'm no scsi expert, just someone who's put a lot of time into observing the 1542A 8 0xFF bytes bug, (& even ripped off cache chips & replaced, & thrown out a perfectly good 500M drive, all seeking to track this down (It'd be cheaper to Axe my 1542A into final submission ;-). If it's a 1542A card, please read my archive of mail, try my testblock/ & 8f/ & let us know your results, etc. General data testing tool: http://........../~jhs/src/bsd/jhs/bin/public/testblock/ Specific 1542A corrupt data testing search tool (use with find): http://.........../~jhs/src/bsd/jhs/bin/public/8f/ Archive of mail (will upload in next days (as of 1997.11.26)): http://........./~jhs/1542a/ Julian -- Julian H. Stacey jhs@