Marvell PATA-SATA bridge meets 2.4.x

Marvell PATA-SATA bridge meets 2.4.x

[Rumi Szabolcs]

Hello!

A while ago I reported a problem with the 2.4.22 kernel and the
tiny Marvell PATA to SATA bridge chip that is used on many of
the now-not-so-recent motherboards which don't have native
SATA ports in their southbridges.

As it can be seen below, a native SATA150 drive is connected
to a SATA port implemented using that Marvell chip hooked up
to the ICH4's parallel ATA133 port and this way the drive is
only recognized (and used) as UDMA33:

ICH4: IDE controller at PCI slot 00:1f.1
ICH4: chipset revision 2
ICH4: not 100% native mode: will probe irqs later
    ide0: BM-DMA at 0xf000-0xf007, BIOS settings: hda:pio, hdb:pio
    ide1: BM-DMA at 0xf008-0xf00f, BIOS settings: hdc:DMA, hdd:pio
hdc: ST3160023AS, ATA DISK drive
blk: queue c04a1ff4, I/O limit 4095Mb (mask 0xffffffff)
ide1 at 0x170-0x177,0x376 on irq 15
hdc: attached ide-disk driver.
hdc: host protected area => 1
hdc: 312581808 sectors (160042 MB) w/8192KiB Cache, CHS=19457/255/63, UDMA(33)

As far as I can remember someone (Jeff Garzik?) suspected the
SATA cable not being recognized as a 80-conductor thus >=UDMA66
capable cable. Then it was told that there is a fix underway that
will be included in the 2.4.23 kernel. The above snippet shows
that the 2.4.25 kernel still has this problem. Any comments?

Regards,
Szabolcs Rumi

Re: Marvell PATA-SATA bridge meets 2.4.x

[Mike Fedyk]

Rumi Szabolcs wrote:
> Hello!
> 
> A while ago I reported a problem with the 2.4.22 kernel and the
> tiny Marvell PATA to SATA bridge chip that is used on many of
> the now-not-so-recent motherboards which don't have native
> SATA ports in their southbridges.
> 
> As it can be seen below, a native SATA150 drive is connected
> to a SATA port implemented using that Marvell chip hooked up
> to the ICH4's parallel ATA133 port and this way the drive is
> only recognized (and used) as UDMA33:
> 
> ICH4: IDE controller at PCI slot 00:1f.1
> ICH4: chipset revision 2
> ICH4: not 100% native mode: will probe irqs later
>     ide0: BM-DMA at 0xf000-0xf007, BIOS settings: hda:pio, hdb:pio
>     ide1: BM-DMA at 0xf008-0xf00f, BIOS settings: hdc:DMA, hdd:pio
> hdc: ST3160023AS, ATA DISK drive
> blk: queue c04a1ff4, I/O limit 4095Mb (mask 0xffffffff)
> ide1 at 0x170-0x177,0x376 on irq 15
> hdc: attached ide-disk driver.
> hdc: host protected area => 1
> hdc: 312581808 sectors (160042 MB) w/8192KiB Cache, CHS=19457/255/63, UDMA(33)
> 
> As far as I can remember someone (Jeff Garzik?) suspected the
> SATA cable not being recognized as a 80-conductor thus >=UDMA66
> capable cable. Then it was told that there is a fix underway that
> will be included in the 2.4.23 kernel. The above snippet shows
> that the 2.4.25 kernel still has this problem. Any comments?

You want to use a 2.6 kernel and talk to Bart, and Jeff about this...

Re: Marvell PATA-SATA bridge meets 2.4.x

[Rumi Szabolcs]

On Sat, 06 Mar 2004 19:55:00 -0800
Mike Fedyk <mfedyk@matchmail.com> wrote:

> Rumi Szabolcs wrote:
> > As it can be seen below, a native SATA150 drive is connected
> > to a SATA port implemented using that Marvell chip hooked up
> > to the ICH4's parallel ATA133 port and this way the drive is
> > only recognized (and used) as UDMA33:
> > 
> > hdc: 312581808 sectors (160042 MB) w/8192KiB Cache, CHS=19457/255/63, UDMA(33)
> > 
> > As far as I can remember someone (Jeff Garzik?) suspected the
> > SATA cable not being recognized as a 80-conductor thus >=UDMA66
> > capable cable. Then it was told that there is a fix underway that
> > will be included in the 2.4.23 kernel. The above snippet shows
> > that the 2.4.25 kernel still has this problem. Any comments?
> 
> You want to use a 2.6 kernel and talk to Bart, and Jeff about this...

Well, I don't really want a 2.6 kernel on that machine (yet) because
in my opinion it is not stable enough for a production system.

Would it be hard to fix that in 2.4?

Regards,
Szabolcs Rumi

Re: Marvell PATA-SATA bridge meets 2.4.x

[Mike Fedyk]

Rumi Szabolcs wrote:
> On Sat, 06 Mar 2004 19:55:00 -0800
> Mike Fedyk <mfedyk@matchmail.com> wrote:
> 
> 
>>Rumi Szabolcs wrote:
>>
>>>As it can be seen below, a native SATA150 drive is connected
>>>to a SATA port implemented using that Marvell chip hooked up
>>>to the ICH4's parallel ATA133 port and this way the drive is
>>>only recognized (and used) as UDMA33:
>>>
>>>hdc: 312581808 sectors (160042 MB) w/8192KiB Cache, CHS=19457/255/63, UDMA(33)
>>>
>>>As far as I can remember someone (Jeff Garzik?) suspected the
>>>SATA cable not being recognized as a 80-conductor thus >=UDMA66
>>>capable cable. Then it was told that there is a fix underway that
>>>will be included in the 2.4.23 kernel. The above snippet shows
>>>that the 2.4.25 kernel still has this problem. Any comments?
>>
>>You want to use a 2.6 kernel and talk to Bart, and Jeff about this...
> 
> 
> Well, I don't really want a 2.6 kernel on that machine (yet) because
> in my opinion it is not stable enough for a production system.

What problems are you having?

> 
> Would it be hard to fix that in 2.4?

Dunno.

Re: Marvell PATA-SATA bridge meets 2.4.x

[Rumi Szabolcs]

On Mon, 08 Mar 2004 14:07:43 -0800
Mike Fedyk <mfedyk@matchmail.com> wrote:

> >>You want to use a 2.6 kernel and talk to Bart, and Jeff about this...
> > 
> > Well, I don't really want a 2.6 kernel on that machine (yet) because
> > in my opinion it is not stable enough for a production system.
> 
> What problems are you having?

I'm just tracking lkml and seeing all the serious bug reports in 2.6
each day. Maybe 2.6.3 is really rock stable compared to what 2.4.3 was
like but compared to VMS 6.2 it looks about as stable as a barrel of
nitroglycerine, so I thought I better wait for at least 2.6.10
before even trying to put it on anything in production.

By the way, I wrote "in my opinion" to avoid starting a flame war on
that "is 2.6 stable already or not?" thing, now it seems like I've
failed... ;-)

So, until the reliability of 2.6 becomes acceptable for everyone I think
many people would appreciate fixes/backports for 2.4, especially for
those problems which don't take too much of an effort to fix.

Regards,
Szabolcs Rumi

Re: Marvell PATA-SATA bridge meets 2.4.x

[Mike Fedyk]

Rumi Szabolcs wrote:
> On Mon, 08 Mar 2004 14:07:43 -0800
> Mike Fedyk <mfedyk@matchmail.com> wrote:
> 
> 
>>>>You want to use a 2.6 kernel and talk to Bart, and Jeff about this...
>>>
>>>Well, I don't really want a 2.6 kernel on that machine (yet) because
>>>in my opinion it is not stable enough for a production system.
>>
>>What problems are you having?
> 
> 
> I'm just tracking lkml and seeing all the serious bug reports in 2.6
> each day. Maybe 2.6.3 is really rock stable compared to what 2.4.3 was
> like but compared to VMS 6.2 it looks about as stable as a barrel of
> nitroglycerine, so I thought I better wait for at least 2.6.10
> before even trying to put it on anything in production.
> 
> By the way, I wrote "in my opinion" to avoid starting a flame war on
> that "is 2.6 stable already or not?" thing, now it seems like I've
> failed... ;-)
> 
> So, until the reliability of 2.6 becomes acceptable for everyone I think
> many people would appreciate fixes/backports for 2.4, especially for
> those problems which don't take too much of an effort to fix.

Ok, that's up to you.  I'd still suggest you run it on one not so 
important machine so you get used to it.

Mike

Re: Marvell PATA-SATA bridge meets 2.4.x

[Andre Hedrick]

It was me who pointed this fact out.

ATA-7 fails to address SATA w/ cable detect and now it reverts back to
ATA-5 mess :-(

I will post a patch for 2.4 for siimage.[ch] later showing how to fudge
the driver.

Cheers,


Andre Hedrick
LAD Storage Consulting Group

On Sat, 6 Mar 2004, Mike Fedyk wrote:

> Rumi Szabolcs wrote:
> > Hello!
> > 
> > A while ago I reported a problem with the 2.4.22 kernel and the
> > tiny Marvell PATA to SATA bridge chip that is used on many of
> > the now-not-so-recent motherboards which don't have native
> > SATA ports in their southbridges.
> > 
> > As it can be seen below, a native SATA150 drive is connected
> > to a SATA port implemented using that Marvell chip hooked up
> > to the ICH4's parallel ATA133 port and this way the drive is
> > only recognized (and used) as UDMA33:
> > 
> > ICH4: IDE controller at PCI slot 00:1f.1
> > ICH4: chipset revision 2
> > ICH4: not 100% native mode: will probe irqs later
> >     ide0: BM-DMA at 0xf000-0xf007, BIOS settings: hda:pio, hdb:pio
> >     ide1: BM-DMA at 0xf008-0xf00f, BIOS settings: hdc:DMA, hdd:pio
> > hdc: ST3160023AS, ATA DISK drive
> > blk: queue c04a1ff4, I/O limit 4095Mb (mask 0xffffffff)
> > ide1 at 0x170-0x177,0x376 on irq 15
> > hdc: attached ide-disk driver.
> > hdc: host protected area => 1
> > hdc: 312581808 sectors (160042 MB) w/8192KiB Cache, CHS=19457/255/63, UDMA(33)
> > 
> > As far as I can remember someone (Jeff Garzik?) suspected the
> > SATA cable not being recognized as a 80-conductor thus >=UDMA66
> > capable cable. Then it was told that there is a fix underway that
> > will be included in the 2.4.23 kernel. The above snippet shows
> > that the 2.4.25 kernel still has this problem. Any comments?
> 
> You want to use a 2.6 kernel and talk to Bart, and Jeff about this...
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>