I have a machine running with an ICH6R Intel chipset.
I've been watching the LKML for info on my ICH6R
chipset's AHCI mode support. It has been a month since
the last SATA status report so I figured I'd write to
ask: Does anyone know when the preliminary AHCI driver
will be integrated into libata mainline? If so I
assume it will be a patch against the 2.6.x kernel?
Also, are there plans to release a version of iswraid
for ICH6R against 2.6.x?
Kevin Tarr
Jeff Garzik wrote:
> Serial ATA (SATA) for Linux
> status report
> May 10, 2004
> Intel ICH6 ("AHCI")
> -------------------
> Summary: Per-device queues, full SATA control
including > hotplug
> and PM.
> libata driver status: "looks like ICH5" support
available in ata_piix.
> Preliminary driver with full AHCI support now
exists, and is being
> integrated into libata mainline.
__________________________________
Do you Yahoo!?
Friends. Fun. Try the all-new Yahoo! Messenger.
http://messenger.yahoo.com/
Kevin Tarr wrote:
> I have a machine running with an ICH6R Intel chipset.
> I've been watching the LKML for info on my ICH6R
> chipset's AHCI mode support. It has been a month since
> the last SATA status report so I figured I'd write to
> ask: Does anyone know when the preliminary AHCI driver
> will be integrated into libata mainline? If so I
> assume it will be a patch against the 2.6.x kernel?
My deadline for integrating AHCI into Red Hat's kernel is measured in
days, if that gives you an indication ;-)
> Also, are there plans to release a version of iswraid
> for ICH6R against 2.6.x?
Have you tried Carl-Daniel's raiddetect? 2.6 does not include
ataraid-based drivers, preferred a Device Mapper (DM) approach instead.
Jeff
On Thu, 10 Jun 2004, Jeff Garzik wrote:
>Have you tried Carl-Daniel's raiddetect? 2.6 does not include
>ataraid-based drivers, preferred a Device Mapper (DM) approach instead.
Have you looked at it lately? It's a nice start but far from finished.
It will not build an array or tell you how to build one yourself. (Yet)
And it's completely userland (part of udev) so it's a complicated
initrd-required path.
If you can read code (lkml... I'll assume everyone can), the header files
provide the on-disk metadata, so you can figure out the appropriate dm table
and/or mdadm config to get the job done. For example:
mdadm --build /dev/md/d0 --chunk=16 --level=0 --raid-devices=4 /dev/sd[abcd]
or a dm-table:
0 1250327228 striped 4 32 /dev/sda 0 /dev/sdb 0 /dev/sdc 0 /dev/sdd 0
works for me (4x160G SATA drives on a SI3114 in raid0 mode.) The same
md setup can be done via the kernel cmdline to boot into the array.
However, there aren't any code pieces in the kernel for reading any of
the various ataraid metadata formats and setting things up. (again, /yet/)
--Ricky
PS: Only the "dm" version is 100% safe. /dev/md/d0 exposes the entire
disk, metadata sectors included.
On Thu, 10 Jun 2004, Ricky Beam wrote:
>... For example:
> mdadm --build /dev/md/d0 --chunk=16 --level=0 --raid-devices=4 /dev/sd[abcd]
>or a dm-table:
> 0 1250327228 striped 4 32 /dev/sda 0 /dev/sdb 0 /dev/sdc 0 /dev/sdd 0
>works for me (4x160G SATA drives on a SI3114 in raid0 mode.) The same
>md setup can be done via the kernel cmdline to boot into the array.
And that cmdline is... (wait for it, *pause*, wait for it)
md=d0,0,2,0,/dev/sda,/dev/sdb,/dev/sdc,/dev/sdd
--Ricky
Jeff Garzik wrote:
> My deadline for integrating AHCI into Red Hat's kernel is measured in
> days, if that gives you an indication ;-)
Since we're on the topic of new libata drivers, how is the Marvell
driver coming along? I'm getting several server units with a 4-port
version on-board in the not-so-distant future, it would be nice if they
could use all their drive bays ;-)
On Fri, 2004-06-11 at 03:30, Andre Tomt wrote:
> Since we're on the topic of new libata drivers, how is the Marvell
> driver coming along? I'm getting several server units with a 4-port
> version on-board in the not-so-distant future, it would be nice if they
> could use all their drive bays ;-)
Though not as useful as a libata driver (and not GPL, though the license
is entirely unrestrictive), there is an open source driver for the
Marvell chipsets:
http://www.highpoint-tech.com/BIOS%20%2B%20Driver/rr1820a/Linux/rr182x-openbuild-v1.02.tgz
It wont build on 2.6 due to cli/sti (v. easy to fix though - its just
the irq locking), and it only supports 8 channel chips (only a few
#defines for PCI ids and number of ports). Intend to fix it up and test
it next week if the libata driver not out, as I have a few of these. The
highpoint card is the first PCI-X SATA card I have actually managed to
get hold of, but unlike other highpoint cards is not their chipset:
03:02.0 SCSI storage controller: Marvell MV88SX5081 8-port SATA I PCI-X
Controller (rev 03)
Jeff Garzik wrote:
............................
>
> My deadline for integrating AHCI into Red Hat's kernel is measured in
> days, if that gives you an indication ;-)
>
............
> Jeff
>
>
How about PATA support? I've got a bunch of promise 20375 cards I'd love
to use.
You know this is like 1st grade. How about me, teacher??
sean
Hello,
Justin Cormack wrote:
> On Fri, 2004-06-11 at 03:30, Andre Tomt wrote:
>
>>Since we're on the topic of new libata drivers, how is the Marvell
>>driver coming along? I'm getting several server units with a 4-port
>>version on-board in the not-so-distant future, it would be nice if they
>>could use all their drive bays ;-)
>
We have those as well. (Supermicro P4SCT+ which have a rev 03 88sx8041
part.)
>
> Though not as useful as a libata driver (and not GPL, though the license
> is entirely unrestrictive), there is an open source driver for the
> Marvell chipsets:
>
> http://www.highpoint-tech.com/BIOS%20%2B%20Driver/rr1820a/Linux/rr182x-openbuild-v1.02.tgz
>
> It wont build on 2.6 due to cli/sti (v. easy to fix though - its just
> the irq locking), and it only supports 8 channel chips (only a few
> #defines for PCI ids and number of ports). Intend to fix it up and test
> it next week if the libata driver not out, as I have a few of these. The
> highpoint card is the first PCI-X SATA card I have actually managed to
> get hold of, but unlike other highpoint cards is not their chipset:
>
> 03:02.0 SCSI storage controller: Marvell MV88SX5081 8-port SATA I PCI-X
> Controller (rev 03)
>
I have a totally different set of source code files for the
MV88SX50[4|8][0|1] chips, also open-source from Marvell (although I have
to check the exact license).
Looking at the Highpoint source code, they seem to have taken some of
the Marvell source code (mv*.*) and adapted it for their RAID system.
I am interested in doing some maintenance work on the Marvell driver,
which indeed mostly needs attention to proper locking. Also, it did
not yet support the rev 03 hardware when I downloaded it.
I found the driver source on ftp://ftp.supermicro.com, but it seems to
have been removed lately. Supermicro forwards me to Adaptec who writes
the BIOS raid system (ugh).
Regards,
Leon.
>
> Justin Cormack wrote:
> > On Fri, 2004-06-11 at 03:30, Andre Tomt wrote:
> >
> >>Since we're on the topic of new libata drivers, how is the Marvell
> >>driver coming along? I'm getting several server units with a 4-port
> >>version on-board in the not-so-distant future, it would be nice if they
> >>could use all their drive bays ;-)
> >
> We have those as well. (Supermicro P4SCT+ which have a rev 03 88sx8041
> part.)
I have the same chipset.
> >
> > Though not as useful as a libata driver (and not GPL, though the license
> > is entirely unrestrictive), there is an open source driver for the
> > Marvell chipsets:
> >
> > http://www.highpoint-tech.com/BIOS%20%2B%20Driver/rr1820a/Linux/rr182x-openbuild-v1.02.tgz
> >
> > It wont build on 2.6 due to cli/sti (v. easy to fix though - its just
> > the irq locking), and it only supports 8 channel chips (only a few
> > #defines for PCI ids and number of ports). Intend to fix it up and test
> > it next week if the libata driver not out, as I have a few of these. The
> > highpoint card is the first PCI-X SATA card I have actually managed to
> > get hold of, but unlike other highpoint cards is not their chipset:
> >
> > 03:02.0 SCSI storage controller: Marvell MV88SX5081 8-port SATA I PCI-X
> > Controller (rev 03)
> >
> I have a totally different set of source code files for the
> MV88SX50[4|8][0|1] chips, also open-source from Marvell (although I have
> to check the exact license).
ah. Can you check the license in detail? See if it is distributable and
modifiable?
> Looking at the Highpoint source code, they seem to have taken some of
> the Marvell source code (mv*.*) and adapted it for their RAID system.
I was intending to strip out the raid stuff anyway, as the highpoint stuff is
binary and of no interest to me.
> I am interested in doing some maintenance work on the Marvell driver,
> which indeed mostly needs attention to proper locking. Also, it did
> not yet support the rev 03 hardware when I downloaded it.
The highpoint one does. It would be useful to diff the shared bits.
> I found the driver source on ftp://ftp.supermicro.com, but it seems to
> have been removed lately. Supermicro forwards me to Adaptec who writes
> the BIOS raid system (ugh).
Ah I never saw this. In their faq supermicro say there is only binary linux
support.
I have some time next week to look at this. If nothing else it would give
a reference for checking a libata driver. Can you check the license on your
files and see if you can send them.
Justin