[CentOS] Re: OT raid controller -- IT8212 = embedded core + 32KB SRAM

Mon Jun 20 02:26:49 UTC 2005
Bryan J. Smith <b.j.smith at ieee.org>

On Mon, 2005-06-20 at 11:40 +1000, John Newbigin wrote:
> Sorry for an OT post, but given the recent discussions here re raid, I 
> was wondering if anyone had any info on the
> ITE 8212 RAID controller (GigaRAID)
> This is a PATA HW RAID as found on the GigaByte GA-8I915P-PRO 
> motherboard. (I can see an IC on the board labeled GigaRAID.  Their 
> manual claims that is has it's own CPU).

First off, the separate IC means _nothing_.  Standard ATA channels have
a separate IC as well.  Then they merely hack their 16-bit BIOS with a
bit of extra RAID organization, which they must match in a software
driver.

_However_, apparently the IT8212 is _not_ a FRAID solution.

Upon further investigation, it does appear it _is_ an embedded
ASIC/microcontroller core (I'm trying to find out the details) with 32KB
of SRAM** (at least in v0.4.1) and 512 bytes/channel FIFO.  Now here
comes the tricky part.

[ **NOTE:  All current 3Ware 4-12 channel cards come with 2-4MB of SRAM
(the 2 channels seem to be 1MB of SRAM, but don't quote me), the 9500S
series adds 128MB of S_D_RAM.  Don't confuse SRAM and SDRAM.  SRAM is 0
wait state, but very large/costly.  DRAM has massive latency times, but
it's very simplistic/cheap. ]

The PCI bus arbitrator can interact with _both_ the embedded core _and_
the ATA FIFOs directly.  The former is the "RAID (intelligent)" mode,
the latter is the "pass-through" mode.  Normally FRAID cards don't have
the former, and the ATA channels are directly accessed on their own.

I've been looking through Alan Cox's 2.6.x patches and seeing that its
supposed to support both.  But from what I've seen, a lot of people are
having trouble with _both_ modes -- even though 2.6.11.x.

> I am not using it because I have SATA, but I am still curious about it's 
> HW v SW raid ability and general linux support.  They do provide a linux 
> driver but the install instructions refer to /dev/had (obviously a typo 
> so I don't think it is well tested or supported).
> I have tried google, but most references are to questions, now answers :(

There is both a "dumb" SCSI block driver (most true hardware RAID cards
are "dumb" drivers because all the intelligence is on-board) that can
drive the "intelligent" mode.

But the recommended option is to use the stock ATA/IDE driver which can
support both the "RAID (intelligent)" and "pass-through" modes.
Although it still seems to have issues at this point.

In the case of this chipset, I'd actually use the RAID functionality and
see if it works on a 2.6.x kernel.  When it's enabled, it's supported to
enable that embedded core and its 32KB of SRAM.  From the block diagram,
it looks like those components go completely unused when the RAID
functionality is not used.

IT8212F v0.4.1 Specification (the v0.4 links from posts don't work):  
http://www.ite.com.tw/product_info/file/pc/IT8212F_V0.4.1.pdf  

-- Bryan

P.S.  It should be noted that 32KB is not much cache, so I wouldn't use
it on anything but a desktop system.

-- 
Bryan J. Smith                                     b.j.smith at ieee.org 
--------------------------------------------------------------------- 
It is mathematically impossible for someone who makes more than you
to be anything but richer than you.  Any tax rate that penalizes them
will also penalize you similarly (to those below you, and then below
them).  Linear algebra, let alone differential calculus or even ele-
mentary concepts of limits, is mutually exclusive with US journalism.
So forget even attempting to explain how tax cuts work.  ;->