Date: Fri, 24 May 1996 10:43:35 -0500 (CDT) From: fredriks@mcs.com To: FreeBSD-gnats-submit@freebsd.org Subject: kern/1245: scsi tape driver write-protect and eject handling is broken Message-ID: <199605241543.KAA18414@fredriks.pr.mcs.net> Resent-Message-ID: <199605241550.IAA21594@freefall.freebsd.org>
next in thread | raw e-mail | index | archive | help
>Number: 1245
>Category: kern
>Synopsis: scsi tape driver write-protet and eject handling is broken
>Confidential: no
>Severity: serious
>Priority: medium
>Responsible: freebsd-bugs
>State: open
>Class: sw-bug
>Submitter-Id: current-users
>Arrival-Date: Fri May 24 08:50:01 PDT 1996
>Last-Modified:
>Originator: & Fredriksen
>Organization:
Flaaklypa Hackers
>Release: FreeBSD 2.2-CURRENT i386
>Environment:
FreeBSD 2.2-CURRENT as of 5/23/96 (sup & compile)
>Description:
Tape device driver does not eject, nor report write-protect properly.
ecject:
The problem with eject is that there are devices out there,
that will not eject if the device has been reserved. The
current st.c reserves the device upon open and releases it
upon close. However the release happens after the unload,
so the eject will not occur. Also if you issue a
mt -f /dev/rstx offline
command that will not eject either, since no release happens.
The fix included below, fixes both of these problems.
write-protect detection:
The scsi tape driver registers its own error handling
routine that gets called when the device returns an
error. This routine checks for the Filemark, End of Medium
and Illegal Length Indicator bits to be set, and if they
are and there are bytes left to be written to tape you
get an EIO back. This routine needs to check for sense-key
of 0x7 (Data Protect) first. The Exabyte 8500 sets the
EOM bit (which is probably a bug on their part). Checking
for sense-key of 0x7 should by-pass these kinds of bugs
and accurately detect write-protected medium on all drives.
density codes:
Added density code for the Exabyte 8200, 8200C, 8500 and 8500C.
These where patches that Eric J. Haug (ejh@eas.slu.edu)
posted on current for the Exabyte 8505.
>How-To-Repeat:
>Fix:
Index: scsiconf.h
===================================================================
RCS file: /home/ncvs/src/sys/scsi/scsiconf.h,v
retrieving revision 1.41
diff -r1.41 scsiconf.h
234c234
< /* 7*/ char density;
---
> /* 7*/ u_int8_t density;
549a550,553
> #define EXB_8200 0x14
> #define EXB_8500 0x15
> #define EXB_8200C 0x8c
> #define EXB_8500C 0x90
Index: st.c
===================================================================
RCS file: /home/ncvs/src/sys/scsi/st.c,v
retrieving revision 1.68
diff -r1.68 st.c
71c71,74
< #define SCSI_2_MAX_DENSITY_CODE 0x17 /* maximum density code specified
---
> #ifdef EXB_8500C
> # define SCSI_2_MAX_DENSITY_CODE EXB_8500C
> #else
> # define SCSI_2_MAX_DENSITY_CODE 0x17 /* maximum density code specified
72a76
> #endif
498a503,509
>
> /* has to happen before the unload, otherwise media will
> * not be ejected when the EJECT flag is given.
> */
>
> scsi_prevent(sc_link, PR_ALLOW, SCSI_SILENT);
>
513d523
< scsi_prevent(sc_link, PR_ALLOW, SCSI_SILENT);
1113a1124,1126
> /* Some devices will not eject unless*/
> /* the device gets released first */
> scsi_prevent(sc_link, PR_ALLOW, SCSI_SILENT);
1147a1161
> goto try_new_value;
1149d1162
< goto try_new_value;
1817c1830,1834
< if ((sense->error_code & SSD_ERRCODE) != 0x70) {
---
> key = sense->ext.extended.flags & SSD_KEY;
>
> if ((sense->error_code & SSD_ERRCODE) != 0x70 || key == 0x7) {
> /* we want the generic code to hadle */
> /* write protect too */
1819a1837,1838
>
>
1893d1911
< key = sense->ext.extended.flags & SSD_KEY;
>Audit-Trail:
>Unformatted:
>Repeat-By:
With an Exabyte 8500 tape drive:
mt -f /dev/rst2 offline
should eject the tape, but doesn't
tar -tvf /dev/erst2
should eject the tape upon close, but doesn't
tar -cvf /dev/rstx and have a write protected cartridge in the
device. You will get EIO instead of EACCESS.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199605241543.KAA18414>