Which Sata-controller card?

[Questions]

Garance A Drosihn wrote:

> At 9:33 AM +0100 11/22/04, Stephan Fiebrandt wrote:
>
>>
>> If you just need to expand your Mainboard with a SATA Controller,
>> all you need is a poor card with a Silicon chip on it (Sil 3112 etc).
>> I've seen cards starting at 10$. Almost all low-cost cards have this
>> chip series, meanwhile on 5.3 it should work now without any problems.
>> The WRITE_DMA issues should be solved and commited to 5.3.
>
>
> Unless something happened in the past week or so, you can still have
> WRITE_DMA problems with the SiL 3112 if you connect a "fake SATA"
> drive to it, such as some of the Western Digital drives.  These are
> really ATA drives with a little conversion-chip on them to implement
> SATA.  These drives seem to work okay on more expensive SATA
> controllers, but I had a *lot* of headaches with one connected to a
> SiL-3112 controller.  I suspect that the same people who are willing
> to live with a $10 disk controller are also going to be tempted to
> buy the less-expensive "fake-SATA" hard drives...
>
>> Have also a look at the 5.3 hardware list: 
>> http://www.freebsd.org/releases/5.3R/hardware-i386.html#DISK.
>
>
I've had in my test a Sil Image running on RELENG_5 cvs from last week 
with a Maxtor 160 SATA (sorry, don't have the model no handy atm) and it 
worked fine without any WRITE_DMA issues, cuz i was curious after a 
longer thread in freebsd-current mailinglist about WRITE_DMA problems. I 
also have to point out, that 5.3-R and RELENG_5 cvs are different :) 
maybe something got commit into it last week, to be honest, i did not 
follow up the changes (shame, yes i know..). Maybe my SATA disk is just 
not a "fake-SATA".
I agree that these "fake-SATA" and cheap Sil 3112 controller might not 
work proper together. But this looks like an hardware incompatibility 
issues than a driver malfunction.
At my main mashine, i have an onboard Intel 6300ESB Controller, and had 
no WRITE_DMA's with any SATA disks yet.

Greetings,

Stephan Fiebrandt