Re: IDE: PDC20275 turning on/off DMA dangerous?

[Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>]

On 12/19/05, Clemens Koller <clemens.koller@anagramm.de> wrote:
> Hello, Bartolmiej!
>
> >>I am working on an embedded ppc (mpc8540) using a pretty common Promise IDE
> >>PCI controller w/ a PDC20275 on it (it's called Ultra TX2).
> >>I have an otherwise good Maxtor 6B120P0 (160GB) connected to it.
> >>
> >>But sometimes (expecially with more than zero disk-i/o-load), when I
> >>turn on DMA by
> >>
> >>$hdparm -X69 -d1 /dev/hda
> >>I get
> >>
> >>hda: task_out_intr: status=0x58 { DriveReady SeekComplete DataRequest }
> >>ide: failed opcode was: unknown
> >>hda: CHECK for good STATUS
> >>
> >>And when I turn off DMA with
> >>$hdparm -d0 /dev/hda
> >>I get sometimes a
> >>
> >>hda: DMA disabled
> >>
> >>which is fine but sometimes I also get:
> >>
> >>hda: status error: status=0x58 { DriveReady SeekComplete DataRequest }
> >>ide: failed opcode was: unknown
> >>hda: drive not ready for command
> >>hda: CHECK for good STATUS
> >>
> >>which is not so nice.
> >>Can you tell me if this is dangerous?
> >
> >
> > Is there any particular reason why you are using hdparm with '-d' and '-X'?
>
> I want to test it with DMA before fixing it into the kernel.
>
> > Your IDE host driver (pdc202xx_new in this case) should configure
> > best xfer mode and enable DMA so you shouldn't need to use hdparm.
>
> I didn't enable automatic turning on of DMA it in the kernel by default
> in the past because I had problems with this hdd controller and
> interrupts (some PCI IRQ mapping issues). The interrupt issues are
> solved now, so I've tried to enable DMA. DMA works pretty fine,
> I was just worried about the severity of the message. It doesn't
> really tell anything useful for non-ide-hackers. (whether it's
> dangerous or not.)
>
> > Given that IDE driver code for changing xfer mode and DMA setting
> > is racy and actually quite hard to fix, it will probably be removed in
> > the future (after auditing IDE host drivers).
>
> So, I guess the answer is: DMA itself seems to work and it
> isn't really dangerous - it's working as it's supposed to.

Yes.

> But if you do it the way you do, you can run into problems due
> to races. Right? So, if I enable DMA as usual in the kernel,
> I don't usually risk my data?!

Yes, just use default settings.

Thanks,
Bartlomiej
-
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/