Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 27 Jun 1997 11:52:59 +0200 (MET DST)
From:      Andrzej Bialecki <abial@korin.warman.org.pl>
To:        Michael Smith <msmith@atrad.adelaide.edu.au>
Cc:        freebsd-hackers@FreeBSD.ORG
Subject:   Re: Disk built-in hw cache
Message-ID:  <Pine.NEB.3.95.970627113451.14546A-100000@korin.warman.org.pl>
In-Reply-To: <199706270205.LAA08899@genesis.atrad.adelaide.edu.au>

next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, 27 Jun 1997, Michael Smith wrote:

> Andrzej Bialecki stands accused of saying:
> > 
> > I just read certain discussion on Linux list concerning
> > bad/missing/removed disk cache in "repaired" (and sold as new) hard disks.
> > Linux prints during probing the size of disk cache (at least that what the
> > hd will tell it). How to achieve the same result in FreeBSD? There is a
> > #ifdef WDDEBUG flag in wd.c, around some interesting printfs. Will it give
> > me this info?
> 
> Uhh, this sounds pretty bogus.  Do you have a reference to anything
> authoratative on the subject?

No :-(. AFAIR from that discussion, IDE disks answer certain query by
returning the info stored somwhere on cyl 0. How this info relates to
reality is of course highly disputable (unless there are any specs on
this). How it gets updated if the chip gets bad/missing is similarly
vague.

The original story goes like this: those folks bought some disks, saw the
"0 kB cache" during startup, went to the seller who checked the disk with
his diagnostic proggy (it also said "0 kB cache"), and then he started
checking other units, using Linux boot flp & his program. To his surprise
there were quite a few such units from one batch. They inquired the
distributor and got very unconvincing (and rather nervous) response.
Finally, they replaced those disks with other ones, which according to the
proggy AND the Linux boot.flp (notice this persistent accordance) had
non-zero amount of cache.

 > 
> It's probable that some IDE disk vendors allow you to query the cache
> size using a vendor-specific command, but the thought of "removing"

Well, AFAIK *we* query the buffersize in wd.c. What it reports is quite
another story :-). Perhaps you think of a *real* test that would show the
physical amount of the cache. And this likely is very vendor-specific.

> size using a vendor-specific command, but the thought of "removing"
> the cache memory from a repaired disk is laughable.  The PCBA on a
> modern disk is worth no more than a few dollars; the _only_ economical
> means for "reapairing" it would be to throw it away and replace it.

Except when it is done by home-grown expert, and then sold as new or
a bargain.

Sincerely yours,

---
Andrzej Bialecki                  FreeBSD: Turning PCs Into Workstations
<abial@warman.org.pl>             http://www.freebsd.org
Research and Academic Network in Poland




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.NEB.3.95.970627113451.14546A-100000>