(unknown charset) Re: Updated SATA support patches for Intel ICH & Promise cards

From: (unknown charset) Doug Ambrisko <ambrisko_at_ambrisko.com>
Date: Tue, 4 May 2004 13:48:41 -0700 (PDT)
Søren Schmidt writes:
| Doug Ambrisko wrote:
| > These patches have been updated to -current as of the past weekend.
| > They increase the robustness of SATA disk errors (media errors and
| > drive going away such as drives that spin up and down).  Also has
| > code to support hot swap on both controllers.
| > 
| > This is only HW support and does not address other ata/FreeBSD issues.
| > That isn't quite right ... some ata changes have been made to avoid
| > panics in probes etc. to support the HW changes.
| > 
| > It would be nice to get this into FreeBSD in some form then I can
| > help with other issues.
| 
| I do have Intel HW here as well now, (big thanks go to FreeBSD Systems 
| and Yahoo for making that possible), and I am looking into the matter.
| I have most of the problems solved regarding support of the chipsets, 
| but very differently from your work.
| I havn't looked into hotswapping SATA much yet, as there isn't much HW 
| around that actually supports that yet, but that will most likely chage 
| soon.

Hot swaping is pretty trivial and a fall out of properly supporting
SATA drives since you have to read the registers and do reset etc.
of the controller and channel.  Seems like all SATA drives support
this especially for things like Intel when they just expose the SATA
registers so you can read the state of the PHY.  I guess a month
or so of figuring SATA issues don't matter.  Promise has their own ways.

You might notice that hot-swap is run into when a drive has a problem 
and spins down and up since the PHY has to re-sync to it and reports
the drive went away then later comes back.

Note my work hasn't been hidden from view but I guess you mostly
ignored it.  You could have told me how you wanted it and I could
have moved it in that direction.  I tried to follow your existing
methods as much as I could figure out.
 
| > I have -stable code to read and write the Intel RAID meta-data
| > but not Adaptec.  My latest -stable patches supports Intel/
| > Promise SATA hot-swap and the Intel RAID meta-data with ata-raid.
| 
| As I said I have code for reading the Adaptec case (which btw is used on 
| all the Intel boards I've seen so far) and that will get in when the 
| Intel stuff gets committed as well.

Good thing I didn't bother with the Adaptec format. 

| I'm sorry if this sounds a bit harsh, but your work conflicts pretty 
| much head on with the way I want this to move along. However code to 
| read/write the RAID config (AMI way I'd guess) would be welcomed as I 
| only have the descriptions for that done, but no code yet...

Oh well .. I guess I will drop trying to make this stuff work and
stop wasting time.  It would have been nice to move forward past
the HW support then fix some of the other issues in FreeBSD with
ata drives.  I guess not.  Have fun.

Now my employer will have to figure out how we migrate to 5.X when 
the time comes to get around various issues with 5.X that I know about
and other hidden issues that I and our QA stuff could look for.  This 
will set those plans back.

Later,

Doug A.
Received on Tue May 04 2004 - 11:48:47 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:37:53 UTC