Can't enable DMA when using IDE devices on SATA controllers

[Marcelo Magno T. Sales]

Hi,

I have several IBM desktops and some notebooks in which the IDE interface is a 
SATA controller, most of them Intel. Here's lspci output:
-------------
[marcelo.sales@sf002698 ~]$ /sbin/lspci | grep IDE
00:1f.2 IDE interface: Intel Corporation 82801FB/FW (ICH6/ICH6W) SATA 
Controller (rev 03)
-------------
On these machines, I can't enable DMA on IDE devices. hdparm shows:
-------------
[marcelo.sales@sf002698 ~]$ /sbin/hdparm /dev/hdc
/dev/hdc:
 IO_support   =  0 (default 16-bit)
 unmaskirq    =  0 (off)
 using_dma    =  0 (off)
 keepsettings =  0 (off)
 readonly     =  0 (off)
 readahead    = 256 (on)
 HDIO_GETGEO failed: Inappropriate ioctl for device
-------------
When I try to enable DMA, I get:
-------------
[marcelo.sales@sf002698 ~]$ su -c '/sbin/hdparm -d1 /dev/hdc'
/dev/hdc:
 setting using_dma to 1 (on)
 HDIO_SET_DMA failed: Operation not permitted
 using_dma    =  0 (off)
-------------
I thought this was caused by lack of kernel support for the SATA controller, 
but I've compiled the FC kernel and the vanilla kernel several times, with 
different settings, and have got no luck on this. libata is built and loaded, 
as shown by lsmod:
[marcelo.sales@sf002698 ~]$ /sbin/lsmod | grep libata
libata                 58065  1 ata_piix
scsi_mod              129001  3 sg,libata,sd_mod

What needs to be done to enable DMA support? I'm using kernel 
2.6.16-1.2122_FC5smp.

Thanks,

Marcelo

[wwp]

Hello Marcelo,


On Wed, 31 May 2006 13:06:54 -0300 "Marcelo Magno T. Sales"
<marcelo.sales(a)sefaz.pe.gov.br&gt; wrote:

>  Hi,
>  
>  I have several IBM desktops and some notebooks in which the IDE interface
>  is a SATA controller, most of them Intel. Here's lspci output:
>  -------------
>  [marcelo.sales@sf002698 ~]$ /sbin/lspci | grep IDE
>  00:1f.2 IDE interface: Intel Corporation 82801FB/FW (ICH6/ICH6W) SATA 
>  Controller (rev 03)
>  -------------
>  On these machines, I can't enable DMA on IDE devices. hdparm shows:
>  -------------
>  [marcelo.sales@sf002698 ~]$ /sbin/hdparm /dev/hdc
>  /dev/hdc:
>   IO_support   =  0 (default 16-bit)
>   unmaskirq    =  0 (off)
>   using_dma    =  0 (off)
>   keepsettings =  0 (off)
>   readonly     =  0 (off)
>   readahead    = 256 (on)
>   HDIO_GETGEO failed: Inappropriate ioctl for device
>  -------------
>  When I try to enable DMA, I get:
>  -------------
>  [marcelo.sales@sf002698 ~]$ su -c '/sbin/hdparm -d1 /dev/hdc'
>  /dev/hdc:
>   setting using_dma to 1 (on)
>   HDIO_SET_DMA failed: Operation not permitted
>   using_dma    =  0 (off)
>  -------------
>  I thought this was caused by lack of kernel support for the SATA
>  controller, but I've compiled the FC kernel and the vanilla kernel several
>  times, with different settings, and have got no luck on this. libata is
>  built and loaded, as shown by lsmod:
>  [marcelo.sales@sf002698 ~]$ /sbin/lsmod | grep libata
>  libata                 58065  1 ata_piix
>  scsi_mod              129001  3 sg,libata,sd_mod
>  
>  What needs to be done to enable DMA support? I'm using kernel 
>  2.6.16-1.2122_FC5smp. 
Does this mean that none of the replies you got to your previous email
provided a solution?


Regards,

-- 
wwp

[Marcelo Magno T. Sales]

Hi, wwp

>  Does this mean that none of the replies you got to your previous
> email
>  provided a solution? 
This one did not work at all:
>  To get good performance I have to add the following to the boot
> kernel
>  parameters:
>  combined_mode=libata atapi_enabled=1 
And your suggestion ("Here I added hdc=noprobe to the kernel boot options 
(grub)") seems to solve the problem with the IDE hard drives (which now 
appear as if they were SCSI), but the DVD drives are still too slow... I'll 
research this a little more.

Thanks,

Marcelo

[Mikkel]

Marcelo Magno T. Sales wrote:
>  Hi,
>  
>  I have several IBM desktops and some notebooks in which the IDE interface is a 
>  SATA controller, most of them Intel. Here's lspci output:
>  -------------
>  [marcelo.sales@sf002698 ~]$ /sbin/lspci | grep IDE
>  00:1f.2 IDE interface: Intel Corporation 82801FB/FW (ICH6/ICH6W) SATA 
>  Controller (rev 03)
>  -------------
>  On these machines, I can't enable DMA on IDE devices. hdparm shows:
>  -------------
>  [marcelo.sales@sf002698 ~]$ /sbin/hdparm /dev/hdc
>  /dev/hdc:
>   IO_support   =  0 (default 16-bit)
>   unmaskirq    =  0 (off)
>   using_dma    =  0 (off)
>   keepsettings =  0 (off)
>   readonly     =  0 (off)
>   readahead    = 256 (on)
>   HDIO_GETGEO failed: Inappropriate ioctl for device
>  -------------
>  When I try to enable DMA, I get:
>  -------------
>  [marcelo.sales@sf002698 ~]$ su -c '/sbin/hdparm -d1 /dev/hdc'
>  /dev/hdc:
>   setting using_dma to 1 (on)
>   HDIO_SET_DMA failed: Operation not permitted
>   using_dma    =  0 (off)
>  -------------
>  I thought this was caused by lack of kernel support for the SATA controller, 
>  but I've compiled the FC kernel and the vanilla kernel several times, with 
>  different settings, and have got no luck on this. libata is built and loaded, 
>  as shown by lsmod:
>  [marcelo.sales@sf002698 ~]$ /sbin/lsmod | grep libata
>  libata                 58065  1 ata_piix
>  scsi_mod              129001  3 sg,libata,sd_mod
>  
>  What needs to be done to enable DMA support? I'm using kernel 
>  2.6.16-1.2122_FC5smp.
>  
>  Thanks,
>  
>  Marcelo
>   
Hdparm can NOT control DMA setting for SATA drives. What I have done
is to add "hdc=noprobe" to the Grub boot line for the kernel. hdc is
then treated as a SCSI device, and it uses the BIOS DMA settings.
Hdparm will still say that DMA is not turned on for the drive, but
DMA will be used. (It also says that DMA is off for /dev/sda.)

This was covered in another thread here not that log ago. There is
still work to be done so that hdparm will correctly report things
for SATA devices...

Mikkel
-- 

  Do not meddle in the affairs of dragons,
for thou art crunchy and taste good with Ketchup!

[Marcelo Magno T. Sales]

Hi, Mikkel

Em Quarta 31 Maio 2006 15:59, Mikkel L. Ellertson escreveu:
>  Hdparm can NOT control DMA setting for SATA drives. What I have done
>  is to add "hdc=noprobe" to the Grub boot line for the kernel. hdc is
>  then treated as a SCSI device, and it uses the BIOS DMA settings.
>  Hdparm will still say that DMA is not turned on for the drive, but
>  DMA will be used. (It also says that DMA is off for /dev/sda.)
>  
Thanks. This seems to have solved the problem for the hard drives, but I have 
yet to find out if there's something more that must be done for the DVD 
drives, which are still too slow.

[]'s
Marcelo

[Mikkel]

Marcelo Magno T. Sales wrote:
>  Hi, Mikkel
>  
>  Em Quarta 31 Maio 2006 15:59, Mikkel L. Ellertson escreveu:
> > Hdparm can NOT control DMA setting for SATA drives. What I have done
> > is to add "hdc=noprobe" to the Grub boot line for the kernel. hdc is
> > then treated as a SCSI device, and it uses the BIOS DMA settings.
> > Hdparm will still say that DMA is not turned on for the drive, but
> > DMA will be used. (It also says that DMA is off for /dev/sda.)
> >
>  Thanks. This seems to have solved the problem for the hard drives, but I have 
>  yet to find out if there's something more that must be done for the DVD 
>  drives, which are still too slow.
>  
>  []'s
>  Marcelo
>   
What devices do the DVD drives show up as? You may have to add more
noprobe options. I am using it on a laptop, so I only have the hard
drive and the DVD-RW drive.

Mikkel
-- 

  Do not meddle in the affairs of dragons,
for thou art crunchy and taste good with Ketchup!