Add warning notes for known quirkly ata chipsets

[Mark Kirkwood]

Do folks think that amending the ata(4) man page, and adding warnings
beside known quirky or broken chipsets is a good idea? I'm thinking of
Si 3112 here in particular(!) - the aim being to discourage folks from
buying 'em based on the fact they appear in ata(4).

I'm happy to draft up something - altho I'd be pretty basic like:

<patch idea>
Silicon Image: SiI0680, SiI3112 (see NOTES), SiI3114, SiI3512.

Notes:
....
There have been reports of data corruption and erratic behavior with
some chipsets. This seems to be related to deficiencies with the chipset
design rather than the driver code. It is recommended to avoid these. A
list of known problematic chipsets is: SiI3112,...
</patch idea>

I don't know any others offhand - but if a) the basic idea seems
sensible and b) people mail me with well documented cases, I'd be happy
to add 'em!

Finally, I'm bringing this up now, as 6.1 is out soon, and it seemed a
good time :-)

regards

Mark
_______________________________________________
freebsd-...@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-hardware
To unsubscribe, send any mail to "freebsd-hardwa...@freebsd.org"

[Mark Kirkwood]

This is a multi-part message in MIME format.
--------------040102090200090609040804
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

Mark Kirkwood wrote:
> Do folks think that amending the ata(4) man page, and adding warnings
> beside known quirky or broken chipsets is a good idea? I'm thinking of
> Si 3112 here in particular(!) - the aim being to discourage folks from
> buying 'em based on the fact they appear in ata(4).
>
> I'm happy to draft up something - altho I'd be pretty basic like:

A tentative patch, adding a BUGS section instead of lumping it into
NOTES (see attached).

regards

Mark

--------------040102090200090609040804
Content-Type: text/plain;
name="ata4.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline;
filename="ata4.patch"

*** ata.4.orig Tue Mar 7 13:00:17 2006
--- ata.4 Tue Mar 7 15:21:12 2006
***************
*** 135,141 ****
.It ServerWorks:
ROSB4, CSB5, CSB6.
.It Silicon Image:
! SiI0680, SiI3112, SiI3114, SiI3512.
.It SiS:
SIS180, SIS181, SIS182, SIS5513, SIS530, SIS540, SIS550, SIS620, SIS630, SIS630S, SIS633, SIS635, SIS730, SIS733, SIS735, SIS745, SIS961, SIS962, SIS963, SIS964, SIS965.
.It VIA:
--- 135,141 ----
.It ServerWorks:
ROSB4, CSB5, CSB6.
.It Silicon Image:
! SiI0680, SiI3112 (see BUGS), SiI3114, SiI3512.
.It SiS:
SIS180, SIS181, SIS182, SIS5513, SIS530, SIS540, SIS550, SIS620, SIS630, SIS630S, SIS633, SIS635, SIS730, SIS733, SIS735, SIS745, SIS961, SIS962, SIS963, SIS964, SIS965.
.It VIA:
***************
*** 216,221 ****
--- 216,225 ----
This is useful in hotswap scenarios
where disks should always show up as the same numbered device,
and not depend on attach order.
+ .Sh BUGS
+ There have been reports of data corruption and erratic behavior with some
+ controllers which seem to be related to deficiencies with the chipset design
+ rather than the driver code. It is recommended to avoid these: SiI3112.
.Sh SEE ALSO
.Xr ataraid 4 ,
.Xr atacontrol 8 ,

--------------040102090200090609040804
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

_______________________________________________
freebsd-...@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-hardware
To unsubscribe, send any mail to "freebsd-hardwa...@freebsd.org"

--------------040102090200090609040804--

[Dieter]

> I don't know any others offhand

Word is that nforce4 plus Maxtor or Hitachi disks gives
data corruption. Nforce4 plus Seagate is said to be okay.

http://forums.nvidia.com/lofiversion/index.php?t8171.html

I have nforce4-ultra with 4 Seagate 7200.8 SATA drives and have
not observed any data corruption. I have have copied many
multi-GB files from one disk to another, and afterwards cmp(1)
says they are identical.

I picked the nforce4-ultra because the other AMD64 chipsets
did not have NCQ. I have not had any problems with the nf4
chipset. (But then I don't have any Maxtor or Hitachi disks
hooked to it.)

On the other hand, I *have* seen data problems with a Maxtor
PATA drive on NetBSD/alpha (not a nforce4 chipset, obviously).
Read the same sector three times and get three different results.