Permissions denied when enabling DMA

[Wayne Gemmell]

Hi, I'm running the bf24 kernel and trying to enable DMA on my hard drive.
My first question is, is DMA already activated?

running hdparm /dev/hdc yields
using_dma = 0 (off)
while running hdparm -i /dev/hdc yields
DMA modes: mdma0 mdma1 mdma2 udma0 udma1 udma2
and hdparm -I /dev/hdc outputs
DMA: mdma0 mdma1 mdma2 udma0 udma1 udma2 udma3 udma4 *udma5
Cycle time: min=120ns recommended=120ns
Could somebody please help me make sence of these readings?

Now when I run hdparm -d1 /dev/hdc I get

Wuler:/var/www/Digi-CRAUDphp/Venue# hdparm -d1 /dev/hdc
/dev/hdc:
setting using_dma to 1 (on)
HDIO_SET_DMA failed: Operation not permitted
using_dma = 0 (off)

Has anybody got any ideas?

Regards
Wayne Gemmell


--
To UNSUBSCRIBE, email to [email]debian-user-request@lists.debian.org[/email]
with a subject of "unsubscribe". Trouble? Contact [email]listmaster@lists.debian.org[/email]

[Alvin Oga]

hi wayn

you need to make sure the the ide chipset
swupported by the kernel you're using

- look at the motherboard, figure out what chipset
and than check your kernel config options

- try installing that module for the chipset

c ya
alvin


On Fri, 1 Aug 2003, Wayne Gemmell wrote:

> Hi, I'm running the bf24 kernel and trying to enable DMA on my hard drive.
> My first question is, is DMA already activated?
>
> running hdparm /dev/hdc yields
> using_dma = 0 (off)
> while running hdparm -i /dev/hdc yields
> DMA modes: mdma0 mdma1 mdma2 udma0 udma1 udma2
> and hdparm -I /dev/hdc outputs
> DMA: mdma0 mdma1 mdma2 udma0 udma1 udma2 udma3 udma4 *udma5
> Cycle time: min=120ns recommended=120ns
> Could somebody please help me make sence of these readings?
>
> Now when I run hdparm -d1 /dev/hdc I get
>
> Wuler:/var/www/Digi-CRAUDphp/Venue# hdparm -d1 /dev/hdc
> /dev/hdc:
> setting using_dma to 1 (on)
> HDIO_SET_DMA failed: Operation not permitted
> using_dma = 0 (off)
>

--
To UNSUBSCRIBE, email to [email]debian-user-request@lists.debian.org[/email]
with a subject of "unsubscribe". Trouble? Contact [email]listmaster@lists.debian.org[/email]

[Andrew McGuinness]

Alvin Oga wrote:

> hi wayn
>
> you need to make sure the the ide chipset
> swupported by the kernel you're using
>
> - look at the motherboard, figure out what chipset
> and than check your kernel config options
>
> - try installing that module for the chipset
>

"Look at the motherboard" is a little awkward. Try:

# lspci -v | more

and look for IDE information.

In my case, the output includes:

00:11.1 IDE interface: VIA Technologies, Inc. Bus Master IDE (rev 06)
(prog-if 8a [Master SecP PriP])
Subsystem: Micro-star International Co Ltd: Unknown device 7380
Flags: bus master, medium devsel, latency 32
I/O ports at ac00 [size=16]
Capabilities: [c0] Power Management version 2



Beyond that, Alvin's advice is good :-)

--
Andrew


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - [url]http://enigmail.mozdev.org[/url]

iD8DBQE/K5JBqEms0aLHW4URAgyyAJ9LwCQz8hA+1tp/bPn36eRF0UE/qQCcCUx0
rZ6Ct20U4KYBpUs+yh8Hh9Y=
=qpNI
-----END PGP SIGNATURE-----

[Wayne Gemmell]

On Saturday 02 August 2003 12:28, Andrew McGuinness wrote:

> Alvin Oga wrote:

> > hi wayn
> >
> > you need to make sure the the ide chipset
> > swupported by the kernel you're using
> >

My output for "lspci -v | more" is

00:11.1 IDE interface: VIA Technologies, Inc. Bus Master IDE (rev 06)
(prog-if 8a [Master SecP PriP])
Subsystem: VIA Technologies, Inc. Bus Master IDE
Flags: bus master, medium devsel, latency 32, IRQ 16
I/O ports at fc00 [size=16]
Capabilities: [c0] Power Management version 2

> > - look at the motherboard, figure out what chipset
> > and than check your kernel config options
> >
> > - try installing that module for the chipset

>

now in modconf the only ide modules are for ide-tape.
Am I looking in the right place? Below is what I think is the relevant part of
the kernel config file.

Thanks
Wayne







CONFIG_BLK_DEV_IDE=y
CONFIG_BLK_DEV_IDEDISK=y
CONFIG_BLK_DEV_IDECD=y
CONFIG_BLK_DEV_IDETAPE=m
CONFIG_BLK_DEV_IDEFLOPPY=y
CONFIG_BLK_DEV_IDESCSI=m
CONFIG_BLK_DEV_CMD640=y
CONFIG_BLK_DEV_RZ1000=y
CONFIG_BLK_DEV_IDEPCI=y
CONFIG_IDEPCI_SHARE_IRQ=y
CONFIG_BLK_DEV_IDEDMA_PCI=y
CONFIG_BLK_DEV_ADMA=y
CONFIG_BLK_DEV_IDEDMA=y
CONFIG_BLK_DEV_AEC62XX=y
CONFIG_BLK_DEV_ALI15X3=y
CONFIG_BLK_DEV_AMD74XX=y
CONFIG_BLK_DEV_CMD64X=y
CONFIG_BLK_DEV_CS5530=y
CONFIG_BLK_DEV_HPT34X=y
CONFIG_BLK_DEV_HPT366=y
CONFIG_BLK_DEV_PIIX=y
CONFIG_BLK_DEV_PDC202XX=y
CONFIG_PDC202XX_BURST=y
CONFIG_BLK_DEV_SVWKS=y
CONFIG_BLK_DEV_SIS5513=y
CONFIG_BLK_DEV_SLC90E66=y
CONFIG_BLK_DEV_VIA82CXXX=y
CONFIG_IDE_CHIPSETS=y
CONFIG_BLK_DEV_4DRIVES=y
CONFIG_BLK_DEV_ALI14XX=y
CONFIG_BLK_DEV_DTC2278=y
CONFIG_BLK_DEV_HT6560B=y
CONFIG_BLK_DEV_PDC4030=y
CONFIG_BLK_DEV_QD65XX=y
CONFIG_BLK_DEV_UMC8672=y
CONFIG_BLK_DEV_IDE_MODES=y
CONFIG_BLK_DEV_ATARAID=y
CONFIG_BLK_DEV_ATARAID_PDC=y
CONFIG_BLK_DEV_ATARAID_HPT=y


--
To UNSUBSCRIBE, email to [email]debian-user-request@lists.debian.org[/email]
with a subject of "unsubscribe". Trouble? Contact [email]listmaster@lists.debian.org[/email]

[Johann Koenig]

On Sat, 02 Aug 2003 13:50:08 +0200
Wayne Gemmell <wayneg@ananzi.co.za> wrote:

> > > you need to make sure the the ide chipset
> > > swupported by the kernel you're using
> > >

> My output for "lspci -v | more" is
>
> 00:11.1 IDE interface: VIA Technologies, Inc. Bus Master IDE
> (rev 06)
> (prog-if 8a [Master SecP PriP])
> Subsystem: VIA Technologies, Inc. Bus Master IDE
> Flags: bus master, medium devsel, latency 32, IRQ 16
> I/O ports at fc00 [size=16]
> Capabilities: [c0] Power Management version 2

Thats odd, it doesn't seem to include any useful model/version information. Mine gives:

00:11.1 IDE interface: VIA Technologies, Inc. VT82C586/B/686A/B PIPC Bus Master

The module for VIA 82Cxxx is built into the bf24 kernel, so that works for me. However, it is the only VIA module I am aware of. Is your board very new?

Someone else mentioned DMA was not available in bf24. This is not true. It works for supported chipsets.
--
-johann koenig
now playing: Amazing Transparent Man - The Ocean Is A Fuck Of A Long Way
To Swim
Today is Prickle-Prickle, the 68th day of Confusion in the YOLD 3169
My public pgp key: [url]http://mental-graffiti.com/pgp/johannkoenig.pgp[/url]


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)

iD8DBQE/K9caFk4duwOzQpURAgJrAJ4hRzp+jQNt+HF7V2D5h0447aF7Eg CeN3vj
2uUJMonSFFBFAYx/FXV/0gQ=
=jMHK
-----END PGP SIGNATURE-----

[Marc Wilson]

On Fri, Aug 01, 2003 at 11:00:56PM +0200, Wayne Gemmell wrote:

> running hdparm /dev/hdc yields
> using_dma = 0 (off)

The kernel didn't enable DMA on the interface the drive is connected to.

> while running hdparm -i /dev/hdc yields
> DMA modes: mdma0 mdma1 mdma2 udma0 udma1 udma2

Not relevant... you care about the NEXT line, the one that reports
available UDMA modes.

> and hdparm -I /dev/hdc outputs
> DMA: mdma0 mdma1 mdma2 udma0 udma1 udma2 udma3 udma4 *udma5

The BIOS put the drive into UDMA5 (ATA100), which the drive supports.

> Now when I run hdparm -d1 /dev/hdc I get
> /dev/hdc:
> setting using_dma to 1 (on)
> HDIO_SET_DMA failed: Operation not permitted
> using_dma = 0 (off)

The kernel you're using doesn't have support for your IDE chipset. This
is why the kernel didn't enable DMA on its own.

Using hdparm to force DMA is a bad, bad, BAD idea. The kernel usually has
very good reasons for not enabling it.

--
Marc Wilson | You are only young once, but you can stay immature
[email]msw@cox.net[/email] | indefinitely.


--
To UNSUBSCRIBE, email to [email]debian-user-request@lists.debian.org[/email]
with a subject of "unsubscribe". Trouble? Contact [email]listmaster@lists.debian.org[/email]

[Wayne Gemmell]

Hi

>
> The kernel you're using doesn't have support for your IDE chipset. This
> is why the kernel didn't enable DMA on its own.

How would you recomend that I get support? apt-get *new-kernel*, or apt-get
source? Are there maybe modules that I should be looking for?

>
> Using hdparm to force DMA is a bad, bad, BAD idea. The kernel usually has
> very good reasons for not enabling it.

Do you mean that I shouldn't use hdparm at all? I have a laptop that is set by
default with DMA and 32bit mode set off, is it also BAD to enable these on
the laptop because I am able to. How do I make this a default on boot?

Regards
Wayne


--
To UNSUBSCRIBE, email to [email]debian-user-request@lists.debian.org[/email]
with a subject of "unsubscribe". Trouble? Contact [email]listmaster@lists.debian.org[/email]

[Anthony Campbell]

On 03 Aug 2003, Wayne Gemmell wrote:

> Hi

> >
> > The kernel you're using doesn't have support for your IDE chipset. This
> > is why the kernel didn't enable DMA on its own.

> How would you recomend that I get support? apt-get *new-kernel*, or apt-get
> source? Are there maybe modules that I should be looking for?

> >
> > Using hdparm to force DMA is a bad, bad, BAD idea. The kernel usually has
> > very good reasons for not enabling it.

> Do you mean that I shouldn't use hdparm at all? I have a laptop that is set by
> default with DMA and 32bit mode set off, is it also BAD to enable these on
> the laptop because I am able to. How do I make this a default on boot?
>
> Regards
> Wayne
>
>

I don't understand the objection to hdparm. Configuration help for the
kernel (2.4.21) says:

GENERIC PCI BUS-MASTER DMA SUPPORT

If your PCI system uses IDE drive(s) (as opposed to SCSI, say) and
is capable of bus-master DMA operation (most Pentium PCI systems),
you will want to say Y here to reduce CPU overhead. You can then use
the "hdparm" utility to enable DMA for drives for which it was not
enabled automatically. By default, DMA is not enabled automatically
for these drives, but you can change that by saying Y to the
following question "Use DMA by default when available". You can get
the latest version of the hdparm utility from
<ftp://ibiblio.org/pub/Linux/system/hardware/>.

Read the comments at the beginning of drivers/ide/ide-dma.c and the
file Documentation/ide.txt for more information.


Although this seems a little self-contradictory, it does appear to
approve of the use of hdparm if needed to get DMA working.

AC

--
[email]ac@acampbell.org.uk[/email] || [url]http://www.acampbell.org.uk[/url]
using Linux GNU/Debian || for book reviews, electronic
Windows-free zone || books and skeptical articles


--
To UNSUBSCRIBE, email to [email]debian-user-request@lists.debian.org[/email]
with a subject of "unsubscribe". Trouble? Contact [email]listmaster@lists.debian.org[/email]

[Eduard Bloch]

#include <hallo.h>
* Wayne Gemmell [Sun, Aug 03 2003, 12:05:18AM]:

> Hi

> >
> > The kernel you're using doesn't have support for your IDE chipset. This
> > is why the kernel didn't enable DMA on its own.

No, the installation kernels do not enable DMA by default because of an
unknown number of not-yet blacklisted data-corrupting drives.

> How would you recomend that I get support? apt-get *new-kernel*, or apt-get
> source? Are there maybe modules that I should be looking for?

What you prefer. If you know how to read english text and edit
lilo.conf, you could try with the backports from Adrian Bunk (see
apt-get.org).

> Do you mean that I shouldn't use hdparm at all? I have a laptop that is set by
> default with DMA and 32bit mode set off, is it also BAD to enable these on
> the laptop because I am able to. How do I make this a default on boot?

That's a kernel option or via hdparm in an init script or via
"echo using_dma:1 > /proc/ide/.../settings".

MfG,
Eduard.
--
Da die Trojaner offensichtlich keinerlei Anstalten machten, das
sogenannte "Trojanische Stinktier" in ihre Stadt zu ziehen, mußten sich
die Griechen wohl oder übel etwas anderes einfallen lassen.
-- Bernd Pfarr


--
To UNSUBSCRIBE, email to [email]debian-user-request@lists.debian.org[/email]
with a subject of "unsubscribe". Trouble? Contact [email]listmaster@lists.debian.org[/email]

[Shyamal Prasad]

"Marc" == Marc Wilson <msw@cox.net> writes:

Marc> Using hdparm to force DMA is a bad, bad, BAD idea. The
Marc> kernel usually has very good reasons for not enabling it.

This might be true for many kernel distributions, but does not apply
to Debian. The prebuilt Debian kernels will not turn on DMA for IDE
drives automatically (which is kinda cool because you can install on
systems that have DMA related bugs, but kinda sucks for newbies).

$ grep DMA /boot/config-2.4.18-k7
CONFIG_BLK_DEV_IDEDMA_PCI=y
CONFIG_BLK_DEV_ADMA=y
# CONFIG_IDEDMA_PCI_AUTO is not set
.....

A "standard" procedure to turn on DMA is to install the hwtools
package and then edit /etc/init.d/hwtools to run hdparm at boot up.

Cheers!
Shyamal


--
To UNSUBSCRIBE, email to [email]debian-user-request@lists.debian.org[/email]
with a subject of "unsubscribe". Trouble? Contact [email]listmaster@lists.debian.org[/email]

[Marc Wilson]

On Sun, Aug 03, 2003 at 12:05:18AM +0200, Wayne Gemmell wrote:

> > Using hdparm to force DMA is a bad, bad, BAD idea. The kernel usually has
> > very good reasons for not enabling it.

> Do you mean that I shouldn't use hdparm at all? I have a laptop that is set by
> default with DMA and 32bit mode set off, is it also BAD to enable these on
> the laptop because I am able to. How do I make this a default on boot?

No, I said using hdparm to force DMA is a bad idea. I didn't say anything
about the other things hdparm can do.

Further, if the kernel you're running defaults to DMA off, then obviously
you'd either recompile the kernel or use hdparm.

The idea is to know *why* DMA didn't get enabled, not blindly start
stomping on settings. As someone else pointed out in this thread, did it
not get enabled because there's no support for it, or because there is
support and it's not enabled by default, or because the particular
drive/controller combination is blacklisted, etc.

But for most people, it's usually easier to just stomp first and whine
later when the data goes away, I must admit. Don't be that person. :)

--
Marc Wilson | Resisting temptation is easier when you think you'll
[email]msw@cox.net[/email] | probably get another chance later on.


--
To UNSUBSCRIBE, email to [email]debian-user-request@lists.debian.org[/email]
with a subject of "unsubscribe". Trouble? Contact [email]listmaster@lists.debian.org[/email]

[Marc Wilson]

On Sun, Aug 03, 2003 at 11:46:11AM +0100, Anthony Campbell wrote:

> I don't understand the objection to hdparm. Configuration help for the
> kernel (2.4.21) says:

<sigh> I don't object to the use of hdparm. I object to people that use it
without bothering to find out why the kernel didn't do whatever they think
it should have done on its own.

Why didn't DMA get enabled by default? That should be the *first*
question, not "oh, gee, no DMA, let's go stomp on settings with hdparm."

--
Marc Wilson | MERYL STREEP is my obstetrician!
[email]msw@cox.net[/email] |


--
To UNSUBSCRIBE, email to [email]debian-user-request@lists.debian.org[/email]
with a subject of "unsubscribe". Trouble? Contact [email]listmaster@lists.debian.org[/email]