removal of token-ring infrastructure coming soon

[Brooks Davis]

On Wed, Mar 28, 2018 at 09:16:23AM -0700, Rodney W. Grimes wrote:
> > On Tue, Mar 27, 2018 at 04:56:32PM -0700, Rodney W. Grimes wrote:
> > > > I have posted a revision which removes support for token-ring networking
> > > > from the tree.  There have been no such devices for some time.
> > > > 
> > > > https://reviews.freebsd.org/D14875
> > > > 
> > > 
> > > Arcnet coming soon?
> > > and probably FDDI?
> > 
> > That's my plan.  I started with token ring as it doesn't even have
> > drivers (Arcnet and FDDI have one driver each, but ISA and 32-bit PCI
> > respectively AFACT).
> 
> I fully support that arcnet and FDDI are dead.
> 
> 
> Please be a bit carefull with what your calling "32 bit PCI" as
> those are also often embedded devices in chipsets that may not
> have cards avaliable, but are infact built in to systems.

I absolutely agree.  In this case it's indicative of the fact that this
"high-performance" interface didn't get new silicon since the fastest bus
it ever ran on was was 32-bit PCI.  It's only one of several indicators.

> Also I have become aware that axing the bt946 support may not
> be a real good idea, as that is/was a common hypervisor emulation
> device and just recently found myself using it in to access
> a disk that was created with a scsi controller that had odd
> translation and the ata layer would not do the right thing
> for me.  (32 sector, 64 head, which is not really odd for
> scsi, thats common translation for many scsi cards).
> 
> I ended up using vmware with a bt946 emulation on scsi with
> the drive image attached to it, and booted FreeBSD from an
> ahci attached disk.  If the bt946 drive was missing from 
> 11.1 I would of been stuck.
> 
> As far as I am aware both vmware and virtualbox have support
> for bt946 emulation, which is also a superset of aha154x.

One of the problems we have in the project is that we don't have a
way to track how popular devices are and what sort of applications
they ended up in.  It would be nice if we could find a way to collect
information including:

Busses: ISA, PCI, PCI-X, PCIe
Form factors: ISA, PCMCIA, CardBus, PCI, PCI-X, PCIe, embedded[0], emulated
Last ship or product announcement date:
Ship volume:
FreeBSD users with meaningful deployments:
[0] Deployments at scale, not just parts available.

I don't know how we'd got about collecting this information, but it
would be good to find a way to croudsource it.

The cost of maintaining each individual device isn't usually all that
high, but it's really frustrating to spend the majority of time on an
infrastructure change fixing issues in devices that shouldn't have
survived the last release branch (or the one before that).  Devices or
technologies that failed in the market are the worst because they tend
to have drivers with cargo-cult copy-and-paste from the wrong example
that never gets fixed.

-- Brooks
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 455 bytes
Desc: not available
URL: <http://lists.freebsd.org/pipermail/freebsd-net/attachments/20180328/20d3b4fc/attachment.sig>