Date: Thu, 15 Dec 2005 11:22:34 +0100 From: Kern Sibbald <kern@sibbald.com> To: freebsd-scsi@freebsd.org Subject: Tape drive handling on FreeBSD Message-ID: <200512151122.34949.kern@sibbald.com>
next in thread | raw e-mail | index | archive | help
Hello,
We correponded a couple of years ago when Bacula was experiencing data loss
at the end of tapes -- it turned out to be a pthreads bug that suppressed
the early end of tape warning.
I have a new problem, and if you cannot help, please
direct me in the right direction.
Basically Bacula (network backup program) needs to open() the tape drive so
that it can read/write it. On both Linux and Solaris, it is possible to open
the drive with O_NONBLOCK and get a descriptor that can be used for ioctl()
calls and providing there is a tape in the drive, it can be used for read()
and write().
On FreeBSD, if there is no tape in the drive, the OS always immediately
returns errno=ENXIO "Device not configured". This means that on FreeBSD, if
there is no tape in the drive, that drive is totally unusable by Bacula. Now
it seems to me that there are two solutions:
1. Modify Bacula and system dependent code that opens the control device to
see if a tape drive is really there or not, and then rewrite the tape driver
code to deal with the fact that if you cannot open a device, it may really be
there, and you should continue trying to open it between asking the user to
mount it. This is clearly possible.
2. Consider implementing something in FreeBSD as exists on at least Linux and
Solaris -- i.e. a means to open the drive and get a valid file descriptor.
If you read/write/rewind/... a drive opened and there is no tape, it should
either return EIO or better ENOMEDIUM.
Comments?
--
Best regards,
Kern
(">
/\
V_V
-------------------------------------------------------
--
Best regards,
Kern
(">
/\
V_V
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200512151122.34949.kern>