On Sat, 5 Jan 2002, Vojtech Pavlik wrote:
> On Sat, Jan 05, 2002 at 05:17:28PM +0100, Edward Stempel wrote:
> > On Tue, 1 Jan 2002, Vojtech Pavlik wrote:
> >
> > > On Tue, Jan 01, 2002 at 11:03:52PM +0100, Edward Stempel wrote:
> > > > In my BIOS-setup there is an option "PCI latency" which defaults to
> > > > 32. First I changed it to 128. That worked perfect. Later I have decreased
> > > > it to 64, 48 and 40. But 40 still gave some problems (a little bit
> > > > distortion of my sound). After that I tried 44, but the disk was to slow
> > > > with this setting (30 MB/s instead of 40 MB/s). So now I have a PCI
> > > > latency of 48. The strange thing is that it seems not to work
> > > > directly. After the system boot when I do hdparm -t, I get about 5
> > > > MB/s. But then when I switch using_dma off and on again using hdparm, I
> > > > get the full 40 MB/s and no distortion of my es1371. Any ideas what might
> > > > cause this?
> > >
> > > You'd have to tell me all other settings in your setup first so I can
> > > guess the cause. What options do you use? Is DMA enabled automatically?
> > > Etc, etc ...
> >
> > I am not sure which settings you need. But I compiled a kernel with:
> > - Generic PCI IDE chipset support
> > - Sharing PCI IDE interrupt support
> > - Generic PCI bus-master DMA support
> > - Use PCI DMA by default when available
> > - VIA82CXXX chipset support
> >
> > After booting this kernel hdparm -v reports that I am using dma, but
> > hdparm -t only reports about 5 MB/s data transfer rate. When I switch the
> > using of dma off and on with hdparm -d the transfer rate becomes about
> > 40MB/s
>
> What does hdparm -i say before and after? Is there any difference (the
> asterisk)?
Sorry, I cannot reproduce this behaviour again. (I just did a reboot) I am
not sure what happened, as I am using the same kernel and bios-setup. Kind
of strange non-deterministic behaviour here. However, as far as can
remember (it has been some days ago) I did not see something different in
hdparm -i before ore after the switching of and on. Both times udma-5 were
reported by hdparm.
As I cannot reproduce this behaviour now, I suggest we forget about
it. Maybe I did something wird when this happened. I will contact you
whenever I notice the same behaviour again.
>
> > The PCI latency is set in the BIOS setup, and according to /proc/pci this
> > latency is applied to all PCI devices except my matrox G550 video card,
> > which has a latency of 64.
> >
> >
> > I hope this explains something.
>
> Not much yet.
>
> >
> >
> >
> > >
> > > >
> > > > Regards,
> > > >
> > > > Edward
> > > >
> > > >
> > > >
> > > >
> > > > ---------- Forwarded message ----------
> > > > Date: Tue, 1 Jan 2002 18:14:41 +0100
> > > > From: Vojtech Pavlik <vojtech@suse.cz>
> > > > To: Edward Stempel <eazstempel@cal009001.student.utwente.nl>
> > > > Cc: linux-kernel mailinglist <linux-kernel@vger.kernel.org>
> > > > Subject: Re: DMA conflicts with soundcard for ide driver via82cxxx
> > > >
> > > > On Fri, Dec 28, 2001 at 05:10:43PM +0100, Edward Stempel wrote:
> > > >
> > > > > Excellent! That solved my problem.
> > > >
> > > > What exactly did you have to change? It might be worth to include the
> > > > changing of the latency setting in the kernel.
> > > >
> > > > >
> > > > > Thankx
> > > > >
> > > > > Edward
> > > > >
> > > > >
> > > > >
> > > > >
> > > > > Date: Fri, 28 Dec 2001 10:15:05 +0100
> > > > > From: Vojtech Pavlik <vojtech@suse.cz>
> > > > > To: Edward Stempel <eazstempel@cal009001.student.utwente.nl>
> > > > > Cc: linux-kernel mailinglist <linux-kernel@vger.kernel.org>
> > > > > Subject: Re: DMA conflicts with soundcard for ide driver via82cxxx
> > > > >
> > > > > On Fri, Dec 28, 2001 at 03:32:34AM +0100, Edward Stempel wrote:
> > > > >
> > > > > > I have an Asus a7v266 mother board and an Ensoniq sound card in it.
> > > > > > The ide chipset is a VIA VT8233 that is capable of UDMA100. So I built a
> > > > > > kernel with the es1371 sound driver and the via82cxxx ide driver
> > > > > > configured in it. Actually I tried the kernel 2.4.17 first, and the latest
> > > > > > I tried is 4.5.1 with the latest patch (patch-2.5.2-pre3) applied to it.
> > > > > > I also tried the kernel 2.5.1 with Vojtech patch (via-3.33.diff from
> > > > > > his email dated 2001-12-23 23:20:48) applied to it, with the same
> > > > > > (negative) results.
> > > > > >
> > > > > > The good thing is that hdparm reports appr. 40 MB/sec when using DMA and
> > > > > > about 6 MB/sec when not using DMA.
> > > > > > Unfortunately using DMA for ide results in some ugly distortion of the
> > > > > > sound from my soundcard whenever some IO to the disk is done. :((
> > > > > >
> > > > > > I have assigned different interrupts to the PCI-cards (ide is
> > > > > > on-board) and I even changed the sound card's PCI slot, so it shared
> > > > > > its interrupt with another device (acpi instead of USB). It did not solve
> > > > > > the problem. Because the problem only occurs when switching on using_dma
> > > > > > on the ide driver, I think it is a DMA problem with the ide driver. It may
> > > > > > be the es1371 driver as well off course, but I suspect it is the ide
> > > > > > driver (or chipset).
> > > > > >
> > > > > > Reading the list archive from linux-kernel, I discovered there have been
> > > > > > more problems with DMA using this chipset, but I did not find anyone
> > > > > > having the same problem as I have now.
> > > > > >
> > > > > > Has someone also dealt with these problems, or can someone help me
> > > > > > solving this problem? Please help!
> > > > > >
> > > > > > Below are some outputs using kernel 5.1 with patch-2.5.2-pre3.
> > > > >
> > > > > You may try changing the PCI latency settings on either the IDE
> > > > > controller or the sound card. Other than that, I don't know how to help.
> > > > >
> > > > > --
> > > > > Vojtech Pavlik
> > > > > SuSE Labs
> > > >
> > > > --
> > > > Vojtech Pavlik
> > > > SuSE Labs
> > >
> > >
> >
>
>
-
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/
This archive was generated by hypermail 2b29 : Mon Jan 07 2002 - 21:00:29 EST