2.4.21-rc1-ac2 Promise IDE DMA won't work

2.4.21-rc1-ac2 Promise IDE DMA won't work

[Tabris]

I looked thru marc's archives, and i'm not finding anything specific to 
this problem, tho I do believe it is probably a known issue.

ASUS A7v266-E
PDC20265
LITE-ON LTR-32123S

only device on the Promise IDE is the cd-rw drive. have four hard drives 
on the VIA IDE.

running kernel 2.4.21-rc1-ac2 + preempt + rmap15g + i2c-sensors

burning now works using cdrecord ATAPI (previous was 2.4.21-pre5-ac3 
where the ide-scsi was too unstable)

at present, reading the cd drive takes 80+% of CPU system time...

I would like to be able to enable DMA. i tried doing it manually with 
hdparm. enabling 32-bit IO or enabling DMA would both make it just not 
work.
[root@tabriel root]# hdparm /dev/hdg

/dev/hdg:
 HDIO_GET_MULTCOUNT failed: Invalid argument
 IO_support   =  0 (default 16-bit)
 unmaskirq    =  0 (off)
 using_dma    =  0 (off)
 keepsettings =  0 (off)
 readonly     =  1 (on)
 readahead    =  8 (on)
 HDIO_GETGEO failed: Invalid argument
[root@tabriel root]#
[root@tabriel root]# hdparm -I /dev/hdg

/dev/hdg:

ATAPI CD-ROM, with removable media
        Model Number:       LITE-ON LTR-32123S
        Serial Number:
        Firmware Revision:  XS0R
Standards:
        Used: ATAPI for CD-ROMs, SFF-8020i, r2.5
        Supported: CD-ROM ATAPI-2
Configuration:
        DRQ response: 50us.
        Packet size: 12 bytes
Capabilities:
        LBA, IORDY(cannot be disabled)
        DMA: *mdma0 mdma1 mdma2 udma0 udma1 udma2
             Cycle time: min=120ns recommended=120ns
        PIO: pio0 pio1 pio2 pio3 pio4
             Cycle time: no flow control=227ns  IORDY flow control=120ns
[root@tabriel root]#


any more info needed, just ask. any help much appreciated.


-- 
"Ignorance is the soil in which belief in miracles grows."
-- Robert G. Ingersoll

Re: 2.4.21-rc1-ac2 Promise IDE DMA won't work

[Andre Hedrick]

NO ATAPI DMA!

I will not write the driver core to attempt to support the various
combinations.  The ATAPI DMA engine space is used support 48bit.
Use the onboard controller for ATAPI.



On Mon, 28 Apr 2003, Tabris wrote:

> I looked thru marc's archives, and i'm not finding anything specific to 
> this problem, tho I do believe it is probably a known issue.
> 
> ASUS A7v266-E
> PDC20265
> LITE-ON LTR-32123S
> 
> only device on the Promise IDE is the cd-rw drive. have four hard drives 
> on the VIA IDE.
> 
> running kernel 2.4.21-rc1-ac2 + preempt + rmap15g + i2c-sensors
> 
> burning now works using cdrecord ATAPI (previous was 2.4.21-pre5-ac3 
> where the ide-scsi was too unstable)
> 
> at present, reading the cd drive takes 80+% of CPU system time...
> 
> I would like to be able to enable DMA. i tried doing it manually with 
> hdparm. enabling 32-bit IO or enabling DMA would both make it just not 
> work.
> [root@tabriel root]# hdparm /dev/hdg
> 
> /dev/hdg:
>  HDIO_GET_MULTCOUNT failed: Invalid argument
>  IO_support   =  0 (default 16-bit)
>  unmaskirq    =  0 (off)
>  using_dma    =  0 (off)
>  keepsettings =  0 (off)
>  readonly     =  1 (on)
>  readahead    =  8 (on)
>  HDIO_GETGEO failed: Invalid argument
> [root@tabriel root]#
> [root@tabriel root]# hdparm -I /dev/hdg
> 
> /dev/hdg:
> 
> ATAPI CD-ROM, with removable media
>         Model Number:       LITE-ON LTR-32123S
>         Serial Number:
>         Firmware Revision:  XS0R
> Standards:
>         Used: ATAPI for CD-ROMs, SFF-8020i, r2.5
>         Supported: CD-ROM ATAPI-2
> Configuration:
>         DRQ response: 50us.
>         Packet size: 12 bytes
> Capabilities:
>         LBA, IORDY(cannot be disabled)
>         DMA: *mdma0 mdma1 mdma2 udma0 udma1 udma2
>              Cycle time: min=120ns recommended=120ns
>         PIO: pio0 pio1 pio2 pio3 pio4
>              Cycle time: no flow control=227ns  IORDY flow control=120ns
> [root@tabriel root]#
> 
> 
> any more info needed, just ask. any help much appreciated.
> 
> 
> -- 
> "Ignorance is the soil in which belief in miracles grows."
> -- Robert G. Ingersoll
> 
> -
> 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/
> 

Andre Hedrick
LAD Storage Consulting Group

Re: 2.4.21-rc1-ac2 Promise IDE DMA won't work

[Tabris]

On Monday 28 April 2003 09:58 pm, Andre Hedrick wrote:
> NO ATAPI DMA!
>
> I will not write the driver core to attempt to support the various
> combinations.  The ATAPI DMA engine space is used support 48bit.
> Use the onboard controller for ATAPI.
> Andre Hedrick
> LAD Storage Consulting Group

Can I object that it came built onto the board? okok... i'll take that 
as a no for now...

Tho i'm still not quite sure how it makes a diff to be honest, unless 
you mean that the Promise and HPT will never be supported for DMA?

and the only other thing i should say is that altho i'm not exactly a 
n00b, the average user WILL expect it to work.

can i expect this to be fixed by 2.6? (yeah, i know... 2.4-ac-ide code 
is similar to 2.5-ide code)
--
tabris

Re: 2.4.21-rc1-ac2 Promise IDE DMA won't work

[Thomas Backlund]

Tabris wrote:

> On Monday 28 April 2003 09:58 pm, Andre Hedrick wrote:
>> NO ATAPI DMA!
>>
>> I will not write the driver core to attempt to support the various
>> combinations.  The ATAPI DMA engine space is used support 48bit.
>> Use the onboard controller for ATAPI.
>> Andre Hedrick
>> LAD Storage Consulting Group
> 
> Can I object that it came built onto the board? okok... i'll take that
> as a no for now...
> 
> Tho i'm still not quite sure how it makes a diff to be honest, unless
> you mean that the Promise and HPT will never be supported for DMA?
> 
> and the only other thing i should say is that altho i'm not exactly a
> n00b, the average user WILL expect it to work.
> 
> can i expect this to be fixed by 2.6? (yeah, i know... 2.4-ac-ide code
> is similar to 2.5-ide code)
> --
> tabris

The point is that you should put your ATAPI device (in this case the cd-rom)
on the VIA controllers, and place the hdd's on the promise controller,
since dma is supported for hdd's...

This is also their intended primary functions by the manufacturers...,
for example the HPT370 controller with bios version >1.1.x.xxx does
not have ATAPI support (atleast officially, I haven't tried it)


Thomas

Re: 2.4.21-rc1-ac2 Promise IDE DMA won't work

[Andre Hedrick]

The Promise chipset use a second DMA engine at offset 0x24 respective of
the channel.  Mixing an ATA and ATAPI on that channel is almost impossible
to make the corner cases work.  Next, if there us a 48-bit ATA plus ATAPI
on the channel popping between the two enignes does not look sane at all
because one has to swithc the location of the hwif->sgtable.

On Mon, 28 Apr 2003, Tabris wrote:

> On Monday 28 April 2003 09:58 pm, Andre Hedrick wrote:
> > NO ATAPI DMA!
> >
> > I will not write the driver core to attempt to support the various
> > combinations.  The ATAPI DMA engine space is used support 48bit.
> > Use the onboard controller for ATAPI.
> > Andre Hedrick
> > LAD Storage Consulting Group
> 
> Can I object that it came built onto the board? okok... i'll take that 
> as a no for now...
> 
> Tho i'm still not quite sure how it makes a diff to be honest, unless 
> you mean that the Promise and HPT will never be supported for DMA?
> 
> and the only other thing i should say is that altho i'm not exactly a 
> n00b, the average user WILL expect it to work.
> 
> can i expect this to be fixed by 2.6? (yeah, i know... 2.4-ac-ide code 
> is similar to 2.5-ide code)
> --
> tabris
> 

Andre Hedrick
LAD Storage Consulting Group

Re: 2.4.21-rc1-ac2 Promise IDE DMA won't work

[Alan Cox]

On Maw, 2003-04-29 at 21:15, Andre Hedrick wrote:
> The Promise chipset use a second DMA engine at offset 0x24 respective of
> the channel.  Mixing an ATA and ATAPI on that channel is almost impossible
> to make the corner cases work.  Next, if there us a 48-bit ATA plus ATAPI
> on the channel popping between the two enignes does not look sane at all
> because one has to swithc the location of the hwif->sgtable.

You can wrap that on the drivers for the disk stuff now but not cleanly for
the other cases. Is there any reason we can't use the ATAPI engine for
everything btw ?

Re: 2.4.21-rc1-ac2 Promise IDE DMA won't work

[Tabris]

On Tuesday 29 April 2003 04:15 pm, Andre Hedrick wrote:
> The Promise chipset use a second DMA engine at offset 0x24 respective
> of the channel.  Mixing an ATA and ATAPI on that channel is almost
> impossible to make the corner cases work.  Next, if there us a 48-bit
> ATA plus ATAPI on the channel popping between the two enignes does
> not look sane at all because one has to swithc the location of the
> hwif->sgtable.

Ok... moved the HDDs from the VIA secondary to the PDC primary (tried 
moving both, but it seems that ide=reverse doesn't work), and the 
CD-R/W to the VIA secondary.

good news, regular reads seem to use the DMA engine. bad news, CDDA 
ripping (using cdparanoia) does not.

is this a known issue? (i thought that CDDA ripping had been fixed)

Haven't tested burning yet.

Hope this all helps, and doesn't annoy too much ;)

--
tabris
-
The day advanced as if to light some work of mine; it was morning, 
and lo! now it is evening, and nothing memorable is accomplished.  
		-- H.D. Thoreau

Re: 2.4.21-rc1-ac2 Promise IDE DMA won't work

[Nicholas Wourms]

Tabris wrote:
> On Tuesday 29 April 2003 04:15 pm, Andre Hedrick wrote:
> 
>>The Promise chipset use a second DMA engine at offset 0x24 respective
>>of the channel.  Mixing an ATA and ATAPI on that channel is almost
>>impossible to make the corner cases work.  Next, if there us a 48-bit
>>ATA plus ATAPI on the channel popping between the two enignes does
>>not look sane at all because one has to swithc the location of the
>>hwif->sgtable.
> 
> 
> Ok... moved the HDDs from the VIA secondary to the PDC primary (tried 
> moving both, but it seems that ide=reverse doesn't work), and the 
> CD-R/W to the VIA secondary.
> 
> good news, regular reads seem to use the DMA engine. bad news, CDDA 
> ripping (using cdparanoia) does not.
> 
> is this a known issue? (i thought that CDDA ripping had been fixed)
> 

Have you tried Andrew Morton's ide-cd dma patch?  It seems to have 
worked well for me, at least.  A summary of what it does is included at 
the top of the patch.

http://www.zipworld.com.au/~akpm/linux/patches/2.4/2.4.20/ide-akpm.patch

Perhaps Alan might considier it for the -rc2 ac patch?

Cheers,
Nicholas

Re: 2.4.21-rc1-ac2 Promise IDE DMA won't work

[Tabris]

> On Monday 28 April 2003 09:58 pm, Andre Hedrick wrote:
>> NO ATAPI DMA!
>>
>> I will not write the driver core to attempt to support the various
>> combinations.  The ATAPI DMA engine space is used support 48bit.
>> Use the onboard controller for ATAPI.
>> Andre Hedrick
>> LAD Storage Consulting Group
> 
> Can I object that it came built onto the board? okok... i'll take that
> as a no for now...
> 
>> Tho i'm still not quite sure how it makes a diff to be honest, unless
>> you mean that the Promise and HPT will never be supported for DMA?
>> 
>> and the only other thing i should say is that altho i'm not exactly a
>> n00b, the average user WILL expect it to work.
>> 
>> can i expect this to be fixed by 2.6? (yeah, i know... 2.4-ac-ide 
>>code is similar to 2.5-ide code)
>> --
>> tabris

>The point is that you should put your ATAPI device (in this case the 
>cd-rom)
>on the VIA controllers, and place the hdd's on the promise controller,
>since dma is supported for hdd's...

>This is also their intended primary functions by the manufacturers...,
>for example the HPT370 controller with bios version >1.1.x.xxx does
>not have ATAPI support (atleast officially, I haven't tried it)
>
>
>Thomas

So, are you telling me that DMA for both my hard drives, and my CD-R/Ws 
will work, if i put my HDs on the PDC20265 controller, and my CD-R/W on 
the VIA?

or will i lose DMA on the hard drives in this way, losing performance?

the second solution, tho one _I_ could probably live with, is still a 
problem, especially for those of us who want to sell GNU/Linux to our 
customers as a 'better' solution.

btw, please cc: me in your replies, as I no longer am able to subscribe 
to the linux-kernel list. this reply i had to hack by hand from 
marc.theaimsgroup

--
tabris
-
Why don't you fix your little problem... and light this candle?
		-- Alan Shepherd, the first man into space, Gemini program