Maybe we could have provision for a command of form mt blocksize 512 in the install ? During install of Release 2.1.6 ... I have a (aha0:2:0): "TANDBERG TDC 3800 =04Y" type 1 removable SCSI 2 st0(aha0:2:0): Sequential-Access density code 0x0, drive empty & it's running in start stop mode (non streaming) during an install, just like it does when I don't do an mt blocksize 512 after new media insertion, in normal use on a normal installed fbsd sys. The non streaming behaviour could be rather disconcerting for a non FreeBSD person as the screen just shows "Attempting to extract from SCSI tape drive" and it sounds more like it repeatedly failing & not streaming, till one hits Alt F2 & sees signs of successful extraction. BTW after the install I validated the tape on another TANDBERG TDC 3800 =04Y on the other system that wrote the tape, & it exhibited the same behaviour there too, tar t ran in lurch mode then ctrl C & mt blocksize 512 tar t then it streamed. Thus the tape was confirmed to be in correct 512 fixed block format (if it had streamed straight off before the mt blocksize 512 that would have indicated the tape was in vari sized blocking, I think) When one generates distrib tapes for others, one has to assume fixed 512 blocking type of drives as least common denominator, & if I create one of those, & recipient has (unknown to me) an enhanced drive like mine, it won't stream during install. Fix: Adding an optional mt blocksize 512 to the tape menu would solve this ... if there's space on the flop :-) Julian --- Julian H. Stacey jhs@freebsd.org http://www.freebsd.org/~jhs/
As jhs@freebsd.org wrote: > Maybe we could have provision for a command of form > mt blocksize 512 > in the install ? There is one. It's in the `options' screen. The blocksize of 512 is only more optimal than the default for your Tandberg (and only if you're using QIC-150 cartridges, things are different for QIC-525 ones). To the contrary, it would really kill performance (if not even making the installation totally impossible) for almost all modern drives. This is merely a problem with the st driver not having a quirk record for this particular drive, where (almost?) all Tandbergs require the ST_Q_SNS_HELP quirk flag set in order to work correctly. Can you please confirm that the following patch obviates the requirement to do an additional ``mt blocksize 512'' for the TDC3800 and QIC-150 tapes? Index: sys/scsi/scsiconf.c =================================================================== RCS file: /home/ncvs/src/sys/scsi/scsiconf.c,v retrieving revision 1.72 diff -u -u -r1.72 scsiconf.c --- scsiconf.c 1996/12/14 09:57:04 1.72 +++ scsiconf.c 1996/12/18 22:38:01 @@ -286,6 +286,10 @@ "st", SC_ONE_LU, ST_Q_NEEDS_PAGE_0, mode_tandberg3600 }, { + T_SEQUENTIAL, T_SEQUENTIAL, T_REMOV, "TANDBERG", " TDC 3800", "*", + "st", SC_ONE_LU, ST_Q_SNS_HLP|ST_Q_NEEDS_PAGE_0, mode_tandberg3600 + }, + { T_SEQUENTIAL, T_SEQUENTIAL, T_REMOV, "TANDBERG", " TDC 42*", "*", "st", SC_ONE_LU, ST_Q_SNS_HLP|ST_Q_NO_1024, mode_tandberg4200 }, -- cheers, J"org joerg_wunsch@uriah.heep.sax.de -- http://www.sax.de/~joerg/ -- NIC: JW11-RIPE Never trust an operating system you don't have sources for. ;-)
Hi, Reference: > From: J Wunsch <j@uriah.heep.sax.de> > Reply-to: joerg_wunsch@uriah.heep.sax.de (Joerg Wunsch) > Subject: Re: misc/2242: mt blocksize 512 in Release 2.1.6 install > Date: Wed, 18 Dec 1996 23:40:24 +0100 (MET) > Message-id: <199612182240.XAA24266@uriah.heep.sax.de> > > As jhs@freebsd.org wrote: > > > Maybe we could have provision for a command of form > > mt blocksize 512 > > in the install ? > > There is one. It's in the `options' screen. > > The blocksize of 512 is only more optimal than the default for your > Tandberg (and only if you're using QIC-150 cartridges, things are > different for QIC-525 ones). To the contrary, it would really kill > performance (if not even making the installation totally impossible) > for almost all modern drives. > > This is merely a problem with the st driver not having a quirk record > for this particular drive, where (almost?) all Tandbergs require the > ST_Q_SNS_HELP quirk flag set in order to work correctly. > > Can you please confirm that the following patch obviates the > requirement to do an additional ``mt blocksize 512'' for the TDC3800 > and QIC-150 tapes? I'll build a kernel to test it, but I recall building a release flop is a horror story that needs CVS ? in which case I may have trouble creating a release type boot flop , particularly as I no longer have a working cvs tree as its outgrown my dedicated 300 M drive , I'll get back to you Joerg, thanks. > > Index: sys/scsi/scsiconf.c > =================================================================== > RCS file: /home/ncvs/src/sys/scsi/scsiconf.c,v > retrieving revision 1.72 > diff -u -u -r1.72 scsiconf.c > --- scsiconf.c 1996/12/14 09:57:04 1.72 > +++ scsiconf.c 1996/12/18 22:38:01 > @@ -286,6 +286,10 @@ > "st", SC_ONE_LU, ST_Q_NEEDS_PAGE_0, mode_tandberg3600 > }, > { > + T_SEQUENTIAL, T_SEQUENTIAL, T_REMOV, "TANDBERG", " TDC 3800", " - *", > + "st", SC_ONE_LU, ST_Q_SNS_HLP|ST_Q_NEEDS_PAGE_0, mode_tandberg3 - 600 > + }, > + { > T_SEQUENTIAL, T_SEQUENTIAL, T_REMOV, "TANDBERG", " TDC 42*", "* - ", > "st", SC_ONE_LU, ST_Q_SNS_HLP|ST_Q_NO_1024, mode_tandberg4200 > }, > > > -- > cheers, J"org > > joerg_wunsch@uriah.heep.sax.de -- http://www.sax.de/~joerg/ -- NIC: JW11-RIPE > Never trust an operating system you don't have sources for. ;-) > > Julian -- Julian H. Stacey jhs@freebsd.org http://www.freebsd.org/~jhs/
State Changed From-To: open->closed timed out As part of our PR auditing campaign, this PR has been closed due to it's age and lack of activity on the PR.