cvs commit: src/share/man/man4/man4.i386 ct.4

M. Warner Losh imp at
Wed Mar 3 09:23:15 PST 2004

In message: <20040304015305Q.matusita at>
            Makoto Matsushita <matusita at> writes:
: rik> I can't rename it cause ct name is used in Linux and in FreeBSD
: rik> as the interface name since 1996.
: I dunno what Linux community did, but 'ct' is a driver for SCSI in
: FreeBSD for a long time.  Please note that this is FreeBSD, not Linux.
: As you know, it often occurs that the driver name of Linux and FreeBSD
: is different (e.g., 'fxp' is called 'e100' in Linux); you do not need
: to worry that 'ct' in Linux is called 'ctau' or whatever in FreeBSD.
: BTW, nyan-san, do you know why there's no manpage for 'ct' SCSI driver?
: And Warner-san, how do you think about this issue?

I think that there's two drivers called ct, and they've both been
called that for a long time.  Both of the lineages stretch out into
out of tree drivers.  The ct driver has had a major number reserved
for it in FreeBSD:

Revision 1.42; 13 12:38:26 1998 UTC (5 years, 9 months ago) by jkh

Name cronyx devices correctly.

+99  ct		    Cronyx/Tau serial adaptor

But the ct driver (which replaces the bs driver) in pc98 land doesn't
seem to have come along until later:

Revision 1.192 Tue Feb 27 12:33:58 2001 UTC (3 years ago) by nyan
Added another wd33c93 based SCSI card driver which replaces the bs driver.
Now, default is still bs.

Yet on the other hand, the cronyx ct driver didn't arrive in the tree
until a few days ago (mostly due to the fact that the ct driver sat in
the pr database for a long time).  Both drivers existed out of the
tree for years before they were integrated into the tree, and have a
long history with related projects.

So both sides have good claims for the one true 'ct' driver.  Both
sides have a long history with the driver name and there's impacts of
various sorts to trying to change it now.  Both are nitch items that
have a small community, and it is hard to judge which of the
communities is larger and which is more impacted. Even if we could say
there are 27 ct users and 29 ct' users, the numbers likely would be so
close to even as to make it not a useful tie breaker (unlike, say, if
there was a port specific fxp driver for some serial hardware for a
architecture that had no pci bus).

There cannot be the case where one person would want to use both forms
of the ct driver.  The dev/ctau driver supports only ISA cards (and no
pc98 machines have ISA expansion slots).  The dev/ct driver supports
only CBUS cards, and there's no non-pc98 machines that have CBUS
slots.  While many architectures have ISA slots (i386, alpha), none
have both.

The only issue to me seems to be the man page and how/where it is
installed.  That's why I thought it would be OK to have the link in
place.  However, it is the first port specific driver whose name is
the same as a driver that's port specific for another port.  AND both
of these drivers have a man page.

Are there issues other than the man page?


More information about the cvs-all mailing list