I just read on an Intel site
(http://64.143.3.64/downloads/drivers/845/perform/linux/udma.htmthat) "ICH4
requires kernel 2.5.12 or later to enable any DMA mode". Can you guys
support or refute this? No wonder I'm having problems with my DMA device on
the ASUS P4PE (using Intel 845PE and ICH4 chipsets)! Are there any patches,
by chance, against a 2.4.20-8 that will give our system DMA support? Or
maybe a patch for 2.4.21? kernel.org shows that the latest (albeit beta)
kernel is 2.6.0-test2 . . . I hestiate to use that, because we would like
something more stable to ship with our product.
Thanks,
Kathy Frazier
Senior Software Engineer
Max Daetwyler Corporation-Dayton Division
2133 Lyons Road
Miamisburg, OH 45342
Tel #: 937.439-1582 ext 6158
Fax #: 937.439-1592
Email: [email protected]
http://www.daetwyler.com
On Mon, 28 Jul 2003, Kathy Frazier wrote:
> I just read on an Intel site
> (http://64.143.3.64/downloads/drivers/845/perform/linux/udma.htmthat) "ICH4
> requires kernel 2.5.12 or later to enable any DMA mode". Can you guys
> support or refute this? No wonder I'm having problems with my DMA device on
> the ASUS P4PE (using Intel 845PE and ICH4 chipsets)! Are there any patches,
> by chance, against a 2.4.20-8 that will give our system DMA support? Or
> maybe a patch for 2.4.21? kernel.org shows that the latest (albeit beta)
> kernel is 2.6.0-test2 . . . I hestiate to use that, because we would like
> something more stable to ship with our product.
I'm running 2.4.21 on this particular machine, and know it worked under
2.4.20.
Bus 0, device 31, function 1:
IDE interface: Intel Corp. 82801DB ICH4 IDE (rev 1).
IRQ 10.
I/O at 0xf000 [0xf00f].
Non-prefetchable 32 bit memory at 0x20000000 [0x200003ff].
mike:~# hdparm /dev/hda
/dev/hda:
multcount = 16 (on)
IO_support = 3 (32-bit w/sync)
unmaskirq = 0 (off)
using_dma = 1 (on)
mike:~# hdparm -Tt /dev/hda
/dev/hda:
Timing buffer-cache reads: 128 MB in 0.37 seconds =345.95 MB/sec
Timing buffered disk reads: 64 MB in 1.88 seconds = 34.04 MB/sec
ASUS P4B533 here, and other similar machines work as well.
Mike
>>On Mon, 28 Jul 2003, Kathy Frazier wrote:
>> I just read on an Intel site
>> (http://64.143.3.64/downloads/drivers/845/perform/linux/udma.htmthat)
"ICH4
>> requires kernel 2.5.12 or later to enable any DMA mode". Can you guys
>> support or refute this? No wonder I'm having problems with my DMA device
on
On Mon, 28 Jul 2003, Mike Dresser wrote:
>I'm running 2.4.21 on this particular machine, and know it worked under
>2.4.20.
> Bus 0, device 31, function 1:
> IDE interface: Intel Corp. 82801DB ICH4 IDE (rev 1).
> IRQ 10.
> I/O at 0xf000 [0xf00f].
> Non-prefetchable 32 bit memory at 0x20000000 [0x200003ff].
>mike:~# hdparm /dev/hda
>/dev/hda:
> multcount = 16 (on)
> IO_support = 3 (32-bit w/sync)
> unmaskirq = 0 (off)
> using_dma = 1 (on)
I checked the results of the hdparm on my system and it shows that hda has
DMA on . . . Hmmm. I have seen a lot of rumblings concerning DMA and IDE
devices in the archives of this mailing list. Is it possible that the ide
drivers now set up the hardware correctly for the DMA? But somehow the O/S
does not support ICH4 for other DMA devices?
Kathy
On Mon, 28 Jul 2003, Kathy Frazier wrote:
> I checked the results of the hdparm on my system and it shows that hda has
> DMA on . . . Hmmm. I have seen a lot of rumblings concerning DMA and IDE
> devices in the archives of this mailing list. Is it possible that the ide
> drivers now set up the hardware correctly for the DMA? But somehow the O/S
> does not support ICH4 for other DMA devices?
>
> Kathy
>
Ok, what is the DMA device? Hard drive? Can you give details?
Mike
>Ok, what is the DMA device? Hard drive? Can you give details?
Mike,
It's a proprietary board that we use to allow the PC to send blocks of data
to some industrial equipment. We developed the hardware and Linux driver
in-house. This same board works (under Linux) on a MoBo using the Intel
815E chipset (Pentium III) with an IHC2 I/O Controller Hub. This is the
system I did _all_ my stress testing in. The plan was to ship our product
with these ASUS P4PE MoBos (using Intel 845PE and ICH4 controller) and were
un-pleasantly surprise when it didn't work.
Kathy
On Mon, 28 Jul 2003, Kathy Frazier wrote:
> >Ok, what is the DMA device? Hard drive? Can you give details?
>
> Mike,
>
> It's a proprietary board that we use to allow the PC to send blocks of data
> to some industrial equipment. We developed the hardware and Linux driver
> in-house. This same board works (under Linux) on a MoBo using the Intel
> 815E chipset (Pentium III) with an IHC2 I/O Controller Hub. This is the
> system I did _all_ my stress testing in. The plan was to ship our product
> with these ASUS P4PE MoBos (using Intel 845PE and ICH4 controller) and were
> un-pleasantly surprise when it didn't work.
>
> Kathy
>
Ok, you're definately well beyond my abilities to help then.
Hopefully someone else on the list can.
Mike
On Llu, 2003-07-28 at 22:02, Kathy Frazier wrote:
> 815E chipset (Pentium III) with an IHC2 I/O Controller Hub. This is the
> system I did _all_ my stress testing in. The plan was to ship our product
> with these ASUS P4PE MoBos (using Intel 845PE and ICH4 controller) and were
> un-pleasantly surprise when it didn't work.
Later 2.4 supports ICH4 UDMA
Alan,
Thanks for your reply! Later? Could you be more specific? What version?
Management is chomping at the bit here. We have invested a lot of effort to
move to Linux and get away from Solaris!
Thanks for your help!
Kathy
-----Original Message-----
From: [email protected]
[mailto:[email protected]]On Behalf Of Alan Cox
Sent: Monday, July 28, 2003 3:19 PM
To: Kathy Frazier
Cc: Mike Dresser; Linux Kernel Mailing List
Subject: RE: DMA not supported with Intel ICH4 I/O controller?
On Llu, 2003-07-28 at 22:02, Kathy Frazier wrote:
> 815E chipset (Pentium III) with an IHC2 I/O Controller Hub. This is the
> system I did _all_ my stress testing in. The plan was to ship our product
> with these ASUS P4PE MoBos (using Intel 845PE and ICH4 controller) and
were
> un-pleasantly surprise when it didn't work.
Later 2.4 supports ICH4 UDMA
Kathy> It's a proprietary board that we use to allow the PC to
Kathy> send blocks of data to some industrial equipment. We
Kathy> developed the hardware and Linux driver in-house. This
Kathy> same board works (under Linux) on a MoBo using the Intel
Kathy> 815E chipset (Pentium III) with an IHC2 I/O Controller Hub.
Kathy> This is the system I did _all_ my stress testing in. The
Kathy> plan was to ship our product with these ASUS P4PE MoBos
Kathy> (using Intel 845PE and ICH4 controller) and were
Kathy> un-pleasantly surprise when it didn't work.
It sounds like your board is acting as a PCI bus master. This is
completely different from DMA for the IDE controller. External PCI
bus masters should be supported by any version of Linux that works on
the motherboard at all.
However there are of course many differences between an 815E and an
845PE motherboard, and between the ICH2 and ICH4. You may have
borderline PCI compliance or signal integrity issues that only cause
problems on the P4PE motherboard. The BIOS on the P4PE may be setting
your device up differently from the 815E motherboard. Your device
might be confusing the BIOS on the P4PE so that the IRQ routing
information (eg in ACPI tables) is screwed up. And so on.
However, I have not heard of any generic problems with external PCI
bus masters and the ICH4.
Best,
Roland
"Kathy Frazier" <[email protected]> writes:
> Thanks for your reply! Later? Could you be more specific? What
> version?
as far as I can tell, 2.4.21 doesn't support it, but 2.4.22-pre8
mentions the chipset in drivers/ide/pci/piix.h. try downloading
2.4.21 and patch it up to 2.4.22-pre8 and see if it works. :-)
> Management is chomping at the bit here.
you have my sympathy, for what it's worth.
[ ... ]
--
Terje
On Llu, 2003-07-28 at 23:02, Kathy Frazier wrote:
> Alan,
>
> Thanks for your reply! Later? Could you be more specific? What version?
> Management is chomping at the bit here. We have invested a lot of effort to
> move to Linux and get away from Solaris!
2.4.20 or later should do it. Red Hat 9 current errata, probably the current errata
kernels of other vendors too.
hdparm -d /dev/hda will tell you
Roland,
Thanks for your response!
>It sounds like your board is acting as a PCI bus master. This is
>completely different from DMA for the IDE controller. External PCI
That's correct, we are a PCI bus master device.
>The BIOS on the P4PE may be setting
>your device up differently from the 815E motherboard. Your device
>might be confusing the BIOS on the P4PE so that the IRQ routing
>information (eg in ACPI tables) is screwed up. And so on.
With respect to the IRQ routing: If the routing was messed up, wouldn't
this mean that no IRQs would get through on that device? We are getting
some . . .
>However, I have not heard of any generic problems with external PCI
>bus masters and the ICH4.
. . . thanks for the feedback.
Regards,
Kathy
>> kernel.org shows that the latest (albeit beta) kernel
>> is 2.6.0-test2 . . . I hestiate to use that, because
>> we would like something more stable to ship with our product.
>The 2.6.0-test2 kernel is very good for a test release.
>Most problems are with drivers for old hardware not
>being updated yet.
Good info. Thanks.
>Also, in all seriousness, try a different PCI slot.
>You may be having electrical problems.
We've tried that . . .
Regards,
Kathy
Terje,
> as far as I can tell, 2.4.21 doesn't support it, but 2.4.22-pre8
> mentions the chipset in drivers/ide/pci/piix.h. try downloading
> 2.4.21 and patch it up to 2.4.22-pre8 and see if it works. :-)
Thanks. But, from what I'm hearing, this problem with the ICH4 is only
concerned with UDMA of IDE devices, so it doesn't sound like an O/S upgrade
will help my DMA bus master device (it's NOT an IDE device).
>> Management is chomping at the bit here.
> you have my sympathy, for what it's worth.
lol. Ahhh, some humor! Thanks!
Regards,
Kathy
On Tue, 29 Jul 2003, Alan Cox wrote:
> 2.4.20 or later should do it. Red Hat 9 current errata, probably the current errata
> kernels of other vendors too.
>
> hdparm -d /dev/hda will tell you
>
I had thought it was a hard drive as well, since that's what everyone
elses problem with DMA turns out to be. Turns out it's a piece of
hardware they're working on that has problems.
Unless I misunderstood again :)
Mike