Siis(4) does not play well with others! :-(

[Alexander Motin]

Jim Bryant wrote:
> you have port multiplier working on that?
> 
> i have a generic board (ebay from hong kong) that looks just like the
> siis reference board for that chip (1x pcie), and it IDs during
> probe/attach as not having port multiplier capability.
> 
> does this capability only show if there is actually a port multiplier
> attached?  i would have thought that it would show as a capability even
> without.

siis(4) hardware always supports port multipliers and FIS-based switching.

> also, i have a couple of WD Green (EADS) 2TB drives attached to mine
> without problem on an intel dq45ek mini-itx mobo (Core2 Duo E8200).
> 
> i have an interest in the above question, as the goal for the mentioned
> box is to maintain it as the houshold server, and it's running out of
> space with six 2TB drives.
> 
> on a side note, is there a way to do the equiv of atacontrol spindown on
> ahci?  i get error messages on that command at startup now after
> switching to ahci a couple of months ago.  this server will ultimately
> have 15 drives in it, and i really want spindown capability in ahci.

camcontrol idle/standby/sleep ...

> Dieter BSD wrote:
>>> For now, I can only say that you are still running not the latest driver
>>> version. Latest have some delays reduced and report connect time in
>>> "us". May be some more of remaining could be replaced with interrupts
>>> (specifically on siis hardware), but I need to think and experiment more
>>> with it.
>>>     
>>
>> I was able to swap cables and put that disk on the JMB363 controller.
>> (directly, no port multiplier) I now just get read errors from the disk
>> and a "error = 5" message from g_vfs_done(), but no messages from ahci(4)
>> at all, and no other problems.
>>
>> So much for testing the disk before putting it into service.  I read the
>> entire disk, then wrote the entire disk with 0x00, 0xff, 0x55 and 0xaa,
>> reading it back each time into hexdump -C.  Put it into service, and
>> after 82 GB of real data (on a 2 TB drive) it fails.  grumble
>>
>> My todo list:
>>
>> 1) Build new kernel with the latest siis(4) and achi(4) bits.
>>
>> 2) Test if a read error causes achi(4) to lock out Ethernet for too long.
>>
>> 3) Test with achi(4) with and without a port multiplier.
>>
>> 3) Test with siis(4) with and without a port multiplier.
>>
>>  
>>> Generally, the situation when the system doesn't reboot with the reset
>>> button doesn't smell good to me.
>>>     
>>
>> FreeBSD does something that disables the soft power switch.  I don't
>> recall if I've had problems with reset before.  Stopping the PM LEDS
>> blinking but not doing a full reset seems pretty odd.  Something
>> important must not be getting reset.  I thought Tyan's mainboards were
>> supposed to be "tier 1".  :-(
>>
>> I'm still puzzled that one Ethernet responded to ping but another did
>> not.


-- 
Alexander Motin