Query: No IDE DMA for IBM 365X with PIIX chipset?

Query: No IDE DMA for IBM 365X with PIIX chipset?

[Grant Coady]

Hi there,

Have an old Thinkpad 365X laptop that 'hdparm -i' tells me is 
running mdma2 but it refuses to set dma mode.  2.6.16.18 also 
refuses to set dma.

With 2.432-hf32.6 running 'hdparm -c 1 /dev/hda' speeds up the 
IDE interface from ~5MB/s to ~7.5MB/s for a reading an unmounted 
500MB partition to /dev/null.

Is there something I forgot to do to get DMA operation, or this 
one of those early PIIX that Alan is doing new drivers for?  

I've yet to try development kernels since udev/pcmciautils not 
setup on the thing (yet).

Comment stripped .config and dmesg for 2.4, 2.6 at:
  <http://bugsplatter.mine.nu/test/boxen/hal/>

Thanks,
Grant.

appended:
  dmesg excerpt
  IDE config
  uname -r
  hdparm -v /dev/hda
  lspci -vvvx
  /proc/interrupts
  /proc/dma

dmesg:
...
ide: Assuming 33MHz system bus speed for PIO modes; override with idebus=xx
PIIXa: IDE controller at PCI slot 00:01.0
PIIXa: chipset revision 2
PIIXa: not 100% native mode: will probe irqs later
PIIXa: neither IDE port enabled (BIOS)
hda: TOSHIBA MK6014MAP, ATA DISK drive
ide0 at 0x1f0-0x1f7,0x3f6 on irq 14
hda: attached ide-disk driver.
hda: host protected area => 1
hda: 11733120 sectors (6007 MB), CHS=776/240/63
Partition check:
 hda: hda1 hda2 hda3 hda4 < hda5 hda6 hda7 hda8 >

2.4 IDE settings:
-------------------- IDE, ATA and ATAPI Block devices ------------------
 <*> Enhanced IDE/MFM/RLL disk/cdrom/tape/floppy support
 --- Please see Documentation/ide.txt for help/info on IDE drives
 [ ]   Use old disk-only driver on primary interface
 [ ]   Support for SATA (deprecated; conflicts with libata SATA driver)
 <*>   Include IDE/ATA-2 DISK support
 [ ]     Use multi-mode by default
 [ ]     Auto-Geometry Resizing support
 <M>   PCMCIA IDE support			<<== for CompactFlash (/dev/hdc)
 < >   Cardbus IDE support (Delkin/ASKA/Workbit)
 < >   Include IDE/ATAPI CDROM support
 < >   Include IDE/ATAPI TAPE support
 < >   Include IDE/ATAPI FLOPPY support
 [ ]   IDE Taskfile Access
 --- IDE chipset support/bugfixes
 [ ]   CMD640 chipset bugfix/support
 [ ]   ISA-PNP EIDE support
 [*]   PCI IDE chipset support
 [ ]     Generic PCI IDE Chipset Support	<<== on/off: no difference
 [ ]     Sharing PCI IDE interrupts support
 [*]     Generic PCI bus-master DMA support
 [ ]     Boot off-board chipsets first support
 [ ]       Force enable legacy 2.0.X HOSTS to use DMA
 [*]       Use PCI DMA by default when available
 [ ]     Enable DMA only for disks
 ... (in between off)
 <*>     Intel PIIXn chipsets support
 ... (rest are off)

root@stinkpad:~# uname -r
2.4.32-hf32.6
root@stinkpad:~# hdparm -v /dev/hda

/dev/hda:
 multcount    = 16 (on)
 IO_support   =  1 (32-bit)
 unmaskirq    =  0 (off)
 using_dma    =  0 (off)
 keepsettings =  0 (off)
 readonly     =  0 (off)
 readahead    =  8 (on)
 geometry     = 776/240/63, sectors = 11733120, start = 0
root@stinkpad:~# hdparm -X mdma2 -d 1 /dev/hda

/dev/hda:
 setting using_dma to 1 (on)
 HDIO_SET_DMA failed: Operation not permitted
 setting xfermode to 34 (multiword DMA mode2)
 using_dma    =  0 (off)
root@stinkpad:~# lspci -vvvx
00:00.0 Host bridge: Intel Corporation 430MX - 82437MX Mob. System Ctrlr (MTSC) & 82438MX Data Path (MTDP) (rev 02)
        Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B-
        Status: Cap- 66Mhz- UDF- FastB2B- ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort+ >SERR- <PERR-
        Latency: 32
00: 86 80 35 12 06 00 00 22 02 00 00 06 00 20 00 00
10: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
20: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
30: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

00:01.0 ISA bridge: Intel Corporation 82371FB PIIX ISA [Triton I] (rev 02)
        Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B-
        Status: Cap- 66Mhz- UDF- FastB2B+ ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort- >SERR- <PERR-
        Latency: 0
00: 86 80 2e 12 07 00 80 02 02 00 01 06 00 00 00 00
10: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
20: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
30: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

00:03.0 VGA compatible controller: Trident Microsystems TGUI 9320 (rev e3) (prog-if 00 [VGA])
        Control: I/O+ Mem+ BusMaster- SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B-
        Status: Cap- 66Mhz- UDF- FastB2B- ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort- >SERR- <PERR-
        Interrupt: pin A routed to IRQ 0
        Region 0: Memory at 08000000 (32-bit, non-prefetchable) [size=2M]
        Region 1: Memory at 08200000 (32-bit, non-prefetchable) [size=64K]
        Expansion ROM at 000c0000 [disabled] [size=64K]
00: 23 10 20 93 03 00 00 02 e3 00 00 03 00 00 00 00
10: 00 00 00 08 00 00 20 08 00 00 00 00 00 00 00 00
20: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
30: 00 00 0c 00 00 00 00 00 00 00 00 00 00 01 00 00

root@stinkpad:~# cat /proc/interrupts
           CPU0
  0:     106675          XT-PIC  timer
  1:          3          XT-PIC  keyboard
  2:          0          XT-PIC  cascade
  9:        490          XT-PIC  xirc2ps_cs
 12:          3          XT-PIC  PS/2 Mouse
 14:       3820          XT-PIC  ide0
 15:          4          XT-PIC  i82365
NMI:          0
ERR:          0

root@stinkpad:~# cat /proc/dma
 4: cascade

Re: Query: No IDE DMA for IBM 365X with PIIX chipset?

[Mark Lord]

Grant Coady wrote:
> Hi there,
> 
> Have an old Thinkpad 365X laptop that 'hdparm -i' tells me is 
> running mdma2 but it refuses to set dma mode.  2.6.16.18 also 
> refuses to set dma.
...

No luck with "hdparm -d1 /dev/hda" ??

> ...
> ide: Assuming 33MHz system bus speed for PIO modes; override with idebus=xx
> PIIXa: IDE controller at PCI slot 00:01.0
> PIIXa: chipset revision 2
> PIIXa: not 100% native mode: will probe irqs later
> PIIXa: neither IDE port enabled (BIOS)
> hda: TOSHIBA MK6014MAP, ATA DISK drive
> ide0 at 0x1f0-0x1f7,0x3f6 on irq 14
> hda: attached ide-disk driver.
> hda: host protected area => 1
> hda: 11733120 sectors (6007 MB), CHS=776/240/63
> Partition check:
>  hda: hda1 hda2 hda3 hda4 < hda5 hda6 hda7 hda8 >
...

Mmm.. there's a curious line: "PIIXa: neither IDE port enabled (BIOS)".
Can you enable DMA in the BIOS?  The IDE driver seems to be unwilling
to set up the bus-master DMA (BMDMA) portion of the chip unless
it was already initialized by the BIOS (paranoia, I suppose, or maybe a bug).

Cheers

Re: Query: No IDE DMA for IBM 365X with PIIX chipset?

[Grant Coady]

On Sun, 28 May 2006 08:06:52 -0400, Mark Lord <lkml@rtr.ca> wrote:

>Grant Coady wrote:
>> Hi there,
>> 
>> Have an old Thinkpad 365X laptop that 'hdparm -i' tells me is 
>> running mdma2 but it refuses to set dma mode.  2.6.16.18 also 
>> refuses to set dma.
>...
>
>No luck with "hdparm -d1 /dev/hda" ??

~# hdparm -d1 /dev/hda

/dev/hda:
 setting using_dma to 1 (on)
 HDIO_SET_DMA failed: Operation not permitted
 using_dma    =  0 (off)

2.6.16.18 gives the same result, am I supposed to try hdparm -d1 
early in the startup scripts, before / gets remounted rw?

>> ide: Assuming 33MHz system bus speed for PIO modes; override with idebus=xx
>> PIIXa: IDE controller at PCI slot 00:01.0
>> PIIXa: chipset revision 2
>> PIIXa: not 100% native mode: will probe irqs later
>> PIIXa: neither IDE port enabled (BIOS)
>> hda: TOSHIBA MK6014MAP, ATA DISK drive
>> ide0 at 0x1f0-0x1f7,0x3f6 on irq 14
>> hda: attached ide-disk driver.
>> hda: host protected area => 1
>> hda: 11733120 sectors (6007 MB), CHS=776/240/63
>> Partition check:
>>  hda: hda1 hda2 hda3 hda4 < hda5 hda6 hda7 hda8 >
>...
>
>Mmm.. there's a curious line: "PIIXa: neither IDE port enabled (BIOS)".
>Can you enable DMA in the BIOS?

The IBM pretty BIOS with a bird shaped mouse pointer flapping wings 
as it moves?  ;)  No, the BIOS hasn't much config options.

Installed the ms-dos PS2 utility, no joy there either:
PS2 DISK sets HDD idle standby delay in minutes
PS2 DMA selects DMA channel for audio and printer

Nothing looks like HDD DMA enable in there.  Same HDD in another 
box via IDE adapter does udma2 (well, it wanted to go faster but 
I slowed it down with hdparm -X to suit the HDD spec.).

>  The IDE driver seems to be unwilling
>to set up the bus-master DMA (BMDMA) portion of the chip unless
>it was already initialized by the BIOS (paranoia, I suppose, or maybe a bug).

Perhaps a buglet, why I cc'd Alan.  I'll check if 2.6.17-rc5 boots 
on the thing...

Thanks,
Grant.

Re: Query: No IDE DMA for IBM 365X with PIIX chipset?

[Alan Cox]

On Sul, 2006-05-28 at 15:29 +1000, Grant Coady wrote:
> PIIXa: chipset revision 2
> PIIXa: not 100% native mode: will probe irqs later
> PIIXa: neither IDE port enabled (BIOS)

It thinks the chip has not been activated, and then falls back to the
legacy driver. Could be incorrect enable checks or other problems.

> 00:01.0 ISA bridge: Intel Corporation 82371FB PIIX ISA [Triton I] (rev 02)
>         Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B-
>         Status: Cap- 66Mhz- UDF- FastB2B+ ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort- >SERR- <PERR-
>         Latency: 0
> 00: 86 80 2e 12 07 00 80 02 02 00 01 06 00 00 00 00
> 10: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
> 20: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
> 30: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

82371FB, whee thats prehistoric 8)

I don't actually have any support for the 371FB PIIX in either driver as
I've not been able to find a source for the data sheet to the chip. It
may work if added to the drivers/scsi/pata_oldpiix identifiers in the
2.6.17rc5-mm kernel. Would be useful to know as I don't know anyone else
with that chip any more 8)

Alan

Re: Query: No IDE DMA for IBM 365X with PIIX chipset?

[Ondrej Zary]

Alan Cox wrote:
> On Sul, 2006-05-28 at 15:29 +1000, Grant Coady wrote:
>> PIIXa: chipset revision 2
>> PIIXa: not 100% native mode: will probe irqs later
>> PIIXa: neither IDE port enabled (BIOS)
> 
> It thinks the chip has not been activated, and then falls back to the
> legacy driver. Could be incorrect enable checks or other problems.
> 
>> 00:01.0 ISA bridge: Intel Corporation 82371FB PIIX ISA [Triton I] (rev 02)
>>         Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B-
>>         Status: Cap- 66Mhz- UDF- FastB2B+ ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort- >SERR- <PERR-
>>         Latency: 0
>> 00: 86 80 2e 12 07 00 80 02 02 00 01 06 00 00 00 00
>> 10: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>> 20: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>> 30: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
> 
> 82371FB, whee thats prehistoric 8)
> 
> I don't actually have any support for the 371FB PIIX in either driver as
> I've not been able to find a source for the data sheet to the chip. It
> may work if added to the drivers/scsi/pata_oldpiix identifiers in the
> 2.6.17rc5-mm kernel. Would be useful to know as I don't know anyone else
> with that chip any more 8)
> 

Isn't that this one? 
http://www.intel.com/design/chipsets/datashts/290550.htm
Used in i430FX chipset. I have some boards with that and they always 
worked fine. Haven't tried any recent kernels, though.

-- 
Ondrej Zary

Re: Query: No IDE DMA for IBM 365X with PIIX chipset?

[Mark Lord]

Alan Cox wrote:
>>
> 82371FB, whee thats prehistoric 8)
> 
> I don't actually have any support for the 371FB PIIX in either driver as
> I've not been able to find a source for the data sheet to the chip. It
> may work if added to the drivers/scsi/pata_oldpiix identifiers in the
> 2.6.17rc5-mm kernel. Would be useful to know as I don't know anyone else
> with that chip any more 8)

That's the original Intel "triton" chipset.
I have a spare printed Intel document for the chipset (Intel #290519-001)
which I can mail you (Alan).  Email me privately with a postal address.

Cheers

Re: Query: No IDE DMA for IBM 365X with PIIX chipset?

[Alan Cox]

On Iau, 2006-06-01 at 13:22 -0400, Mark Lord wrote:
> That's the original Intel "triton" chipset.
> I have a spare printed Intel document for the chipset (Intel #290519-001)
> which I can mail you (Alan).  Email me privately with a postal address.

>From the other docs it appears 0x122E is the ISA bridge and this laptop
has 0x122E (PIIX bridge) and an 82437MX system controller, but no PIIX
IDE. That actually suggests its more like the "MPIIX" which has a PIO
only IDE controller existing (logically anyway) on the ISA side of the
system.

That would explain the observed behaviour and fit with the pattern of
PCI identifiers. Now to hunt 82437 docs.

(In the mean time try adding the 0x1235 id to the 2.6.17-mm kernel in
drivers/scsi/pata_mpiix and see if that works with the new libata layer
not drivers/ide).

Alan

Re: Query: No IDE DMA for IBM 365X with PIIX chipset?

[Grant Coady]

On Thu, 01 Jun 2006 17:55:28 +0200, Ondrej Zary <linux@rainbow-software.org> wrote:

>Alan Cox wrote:
>> On Sul, 2006-05-28 at 15:29 +1000, Grant Coady wrote:
>>> PIIXa: chipset revision 2
>>> PIIXa: not 100% native mode: will probe irqs later
>>> PIIXa: neither IDE port enabled (BIOS)
>> 
>> It thinks the chip has not been activated, and then falls back to the
>> legacy driver. Could be incorrect enable checks or other problems.
>> 
>>> 00:01.0 ISA bridge: Intel Corporation 82371FB PIIX ISA [Triton I] (rev 02)
>>>         Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B-
>>>         Status: Cap- 66Mhz- UDF- FastB2B+ ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort- >SERR- <PERR-
>>>         Latency: 0
>>> 00: 86 80 2e 12 07 00 80 02 02 00 01 06 00 00 00 00
>>> 10: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>>> 20: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>>> 30: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>> 
>> 82371FB, whee thats prehistoric 8)

Only a little ten-year-old...
>> 
>> I don't actually have any support for the 371FB PIIX in either driver as
>> I've not been able to find a source for the data sheet to the chip. It
>> may work if added to the drivers/scsi/pata_oldpiix identifiers in the
>> 2.6.17rc5-mm kernel. Would be useful to know as I don't know anyone else
>> with that chip any more 8)

Boots 2.6.16.18, 2.6.17rc5 but panics with 2.6.17rc5-mm2 :(  

>Isn't that this one? 
>http://www.intel.com/design/chipsets/datashts/290550.htm
>Used in i430FX chipset. I have some boards with that and they always 
>worked fine. Haven't tried any recent kernels, though.

Yep:

2.2.2. DID-DEVICE IDENTIFICATION REGISTER (Function 0)
Address Offset: 02-03h
Default Value: 122Eh (PIIX)		<<==

~# lspci -n
00:00.0 Class 0600: 8086:1235 (rev 02)
00:01.0 Class 0601: 8086:122e (rev 02)	<<==
00:03.0 Class 0300: 1023:9320 (rev e3)

Grant.

Re: Query: No IDE DMA for IBM 365X with PIIX chipset?

[Ondrej Zary]

Alan Cox wrote:
> On Iau, 2006-06-01 at 13:22 -0400, Mark Lord wrote:
>> That's the original Intel "triton" chipset.
>> I have a spare printed Intel document for the chipset (Intel #290519-001)
>> which I can mail you (Alan).  Email me privately with a postal address.
> 
>>From the other docs it appears 0x122E is the ISA bridge and this laptop
> has 0x122E (PIIX bridge) and an 82437MX system controller, but no PIIX
> IDE. That actually suggests its more like the "MPIIX" which has a PIO
> only IDE controller existing (logically anyway) on the ISA side of the
> system.
> 
> That would explain the observed behaviour and fit with the pattern of
> PCI identifiers. Now to hunt 82437 docs.
> 
> (In the mean time try adding the 0x1235 id to the 2.6.17-mm kernel in
> drivers/scsi/pata_mpiix and see if that works with the new libata layer
> not drivers/ide).

Maybe http://stuff.mit.edu/afs/sipb/contrib/doc/specs/unfiled/i82437MX.pdf

-- 
Ondrej Zary