Date: Tue, 18 Jun 2002 11:04:35 -0400 From: Brian T.Schellenberger <bts@babbleon.org> To: msch@snafu.de, "freebsd-stable@FreeBSD.ORG" <freebsd-stable@FreeBSD.ORG> Subject: Re: ATA Atapi 4.6 Release Message-ID: <20020618150435.02FBDBB2C@i8k.babbleon.org> In-Reply-To: <E17KJst-0003up-00@smart.eusc.inter.net> References: <E17KJst-0003up-00@smart.eusc.inter.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tuesday 18 June 2002 10:23 am, Matthias Schuendehuette wrote:
| Hello all,
|
| I haven't read the whole thread so far, but...
|
| I would encourage some posters to cool down a bit. There are only a few
| types of ATA-Disks that support Tagged Queuing and they all work with
| Write Cache nearly as fast as with WC+TQ.
|
| I made some comparison (admittedly with a SCSI-Disk :-) between TQ and
| WC and I found out, that, when copying the whole ports-tree f.i., TQ
| and WC have nearly the same performance gain and TQ+WC adds another 10%
| performance... so, no big deal IMHO.
|
| Of course, safety is another point, and I prefer TQ over WC exactly
| because of this! But - to be honest - I prefer SCSI over ATA because of
| this ;-)
|
| So, in the end I think it's a at least possible decision to release 4.6
| with ATA 'as is'. It's switched off by default and shouldn't frustrate
| any new users. It should be stated clearly and 'loud', that ATA Tagged
| Queuing in not working on many chipsets and that the situation is
| expected to improve during 4.6-STABLE.
|
| And BTW: I don't think that it's helpful to focus on that ATA TQ-Issue
| so hard and to imply (at least between the lines of some postings) that
| Soren's driver is all crap. It's obviously the best driver we *have*
| and obviously the most knowledgeable ATA-Programmer who's working for
| 'our' FreeBSD and we should appreciate that.
The objection to releasing it as it was done lay in the fact that some
hardware worked before and it doesn't now. And this certainly is an
undesirable outcome, I believe everybody would agree.
Even though I was a prime beneficiary of the new ATA code (I have a
hot-swappable ATA device) and in fact was a beta-tester for the MFC code, I
myself have my doubts about whether it should have shipped when it introduced
a known regression like that. It might have been better to have shipped both
old and new code and had a kernel config option to use the new code or
something that otherwise could have been selected by the benficieries without
messing up other people.
It is a prime claim of FreeBSD that it is stable and careful in contrast to
that 'other' predominant free Unix-like O/S that runs on PC hardward, which
is more of the "life in the fast-lane" O/S.
I think it would be good if FreeBSD continued to pursue its strengths.
On the other hand, it can't just stay still.
It is a cunundrum; hopefully all the people involved in the FreeBSD
decision-making will listen to all the feedback and use it to help make the
best decisions moving forward.
I will say that a lot was done *right* with ATA code. It was MFC'ed rather
early in the 4.6 time line; it was pretty widely beta-tested before it was
even MFC'ed; and Soren (sorry, i don't know how to make the latter right)
seemed to be quite responsive.
It is quite unfortunate that the workaround wan't worked about before 4.6
shipped so that it would have been "pre-integrated" or something (I myself
might suffer the underlying bug but since I was already setting the DMA modes
I would never have seen the problem. Life's funny sometimes, eh?), but it
didn't. I never installed 4.5 because KDE was messed up; these things happen.
Hopefully all that can be done will be done to improve things next time
'round.
--
Brian T. Schellenberger . . . . . . . bts@wnt.sas.com (work)
Brian, the man from Babble-On . . . . bts@babbleon.org (personal)
http://www.babbleon.org
http://www.eff.org http://www.programming-freedom.org
If you smell the smoke you don't need to be told what you've got to do;
Yet there's a certain breed, so very in-between, they'd rather take a
vote. -- DEVO -- Here To Go
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-stable" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20020618150435.02FBDBB2C>