[BUGS] Worth persuing or try NetBSD?

[Andrew Reilly]

Hi Andy,

On Mon, 28 Dec 2009 13:38:30 +1000
Andy Farkas <chuzzwassa at gmail.com> wrote:

> Hello arm-chair FreeBSD kernel hackers!
> 
> My main gateway box has been running FreeBSD 4.12 for quite a while until
> I recently decided to upgrade it, via the source upgrade route:

There are a couple of large-ish issues with trying to get to
-STABLE or -CURRENT from 4-vintage, I think.  You seem to have
hit one involving specific driver support, but you might also run
into problems (or at least inconvenience or bit-rot) relating to
the move to UFS2.  Also, I think that rc-ng will make most of
your existing set-up and config (/etc) excitingly out-of-date.
However I dare say you'll be able to manage those if you can get
past the driver issue.

Does your gateway box have any possibility of access to a CD-ROM
drive?  If it does, I would strongly suggest that the fastest and
least troublesome way to get to where you want to go will be to
backup your config and user files, and do a reformat and install
from the sysinstall ISO of the version of choice.  At the very
least that will give you a taste to see if more recent versions
deal more nicely with your AMR controller than 6.1.

>  1/ cvsup RELENG_5_5_0_RELEASE, make (GENERIC) world, install world, reboot.
>  2/ cvsup RELENG_6_0_0_RELEASE, make (GENERIC) world, install world, reboot.
>  3/ cvsup RELENG_6_1_0_RELEASE, make (GENERIC) world, install world, reboot.
> 
> 6.1-RELEASE is where the problem started. Processes started hanging during
> disk I/O. 6.0-R works flawlessly and is able to do buildworlds without fail.
> 
> The disk controller is a:
> 
> amr0 at pci0:6:0:	class=0x018000 card=0x00000000 chip=0x9010101e rev=0x03 hdr=0x00
>     vendor   = 'American Megatrends Inc.'
>     device   = 'MegaRAID 428 Ultra Fast Wide SCSI RAID Controller'
>     class    = mass storage
> 
> 
> After some research, I discovered there was a "mega update" merged into the
> amr(4) driver between 6.0-R and 6.1-R. So this is what I am concentrating on.

Any particular reason to keep trying 6.1, rather than jumping
forward to 6-last (or 7- or 8-), to see if the problem has been
solved in the mean time?

> Is it worth persuing this to debug the driver for such an old piece of
> hardware in an old Pentium-Pro 200MHz? Or would it be worth trying NetBSD
> instead?

I'm running 8-STABLE on a box nearly that old, albeit without the
fancy disk interface, so it should be reasonable.  In a sense,
some of the improved efficiencies in the more modern kernels
should make old hardware work better (for some values of
better).  Certainly there are plenty of people running both
-CURRENT and current versions of NetBSD on low-powered gateway or
NAS boxes.  I used to run a diskless VIA-eden NetBSD system as a
gateway, up to about version 4, and that worked fine, but
lazyness and a broken access point has resulted in me replacing
it with an off-the shelf LinkSys system, so I haven't done any
NetBSD hacking for quite a while.  Version 5 sounds pretty slick,
though.

> Perhaps I should post this to the -fs or -hackers mlist.... is there a
> -scsi list?  .../me checks...

You might want to try -fs, or perhaps try to contact the amr
driver developer(s) directly, or perhaps submit a send-pr (or
check the gnats database for existing reports?)

Good luck!

Cheers,

-- 
Andrew