Page MenuHomeFreeBSD

pbio: Remove the pbio(4) driver
AcceptedPublic

Authored by jhb on Mon, Oct 12, 6:30 PM.
Tags
None
Referenced Files
F175934490: D60662.diff
Wed, Oct 14, 3:12 PM
F175924419: D60662.id.diff
Wed, Oct 14, 1:23 PM
F175868604: D60662.id189640.diff
Wed, Oct 14, 1:17 AM
F175848501: D60662.id189640.diff
Tue, Oct 13, 9:01 PM
F175847938: D60662.id189640.diff
Tue, Oct 13, 8:55 PM
Unknown Object (File)
Tue, Oct 13, 4:46 PM
Unknown Object (File)
Tue, Oct 13, 11:22 AM
Unknown Object (File)
Tue, Oct 13, 6:16 AM

Details

Reviewers
kib
imp
Group Reviewers
srcmgr
Summary

While this is listed in the MI sys/conf/files, this driver is de facto
i386-only as the supported hardware is an ISA card and the manpage is
tagged as i386-only.

Diff Detail

Repository
rG FreeBSD src repository
Lint
Lint Skipped
Unit
Tests Skipped
Build Status
Buildable 78080
Build 74963: arc lint + arc unit

Event Timeline

jhb requested review of this revision.Mon, Oct 12, 6:30 PM

This once was important for embedded work that people did with 386 machines. But it's an ISA card and there's much better alternatives these days.

This revision is now accepted and ready to land.Mon, Oct 12, 10:15 PM

This can work on amd64, ISA machines exist(ed). The code is not even too rotten, it uses make_dev_s() for instance.
There are typical bugs in the read functions.
But I do not have an opinion.

IMO this one warrants some specific attention to it (e.g. referencing it explicitly in the HEADS-UP mails rather than just included in "i386-only drivers are being removed" as it theoretically could be used on amd64. But yes, I suspect this has not been used on other than i386 at all, or at least not in the last decade+

IMO this one warrants some specific attention to it (e.g. referencing it explicitly in the HEADS-UP mails rather than just included in "i386-only drivers are being removed" as it theoretically could be used on amd64. But yes, I suspect this has not been used on other than i386 at all, or at least not in the last decade+

I suspect you're right on both fronts. We should give a heads up maybe for all these devices, but this is one that just might have people interested in it going away.

Google told me that they were sold in the 90s and in the early 2000s the commercial cards were discontinued in favor of PCI versions (which this driver never supported). Last sale was like 20 years ago.

OTOH, there's still a market for these cards, though it's a replacement market only for people with legacy systems that need a new card to keep the legacy setup going. Maybe some of those systems are FreeBSD (but they'd likely be FreeBSD 3 or 4 systems, though FreeBSD 3 needs a backport). For slow applications (like turning on lights, sequencing shows, etc) it would be plenty fast. But today, the stuff this card was used for is 'table stakes' for any SOC and one of the bazillion embedded arm or similar boards.

@imp yeah for me it's less about "people are probably using this" than a general "FreeBSD might silently drop support for something that I care about"

@imp yeah for me it's less about "people are probably using this" than a general "FreeBSD might silently drop support for something that I care about"

We already dropped support in 15.0. This is not new news, this is just removing something we stopped shipping a year ago. I mean, what should the release note be: "We removed the source to something we didn't ship in the previous release"? Is that a meaningful thing to say?