[mythtv-users] setpci, interrupts (Was: SOLVED System Load/Performance Question)

[Todd Ignasiak]

On 2/25/06, Yan-Fa Li <yanfali at gmail.com> wrote:
>
> On 2/25/06, Osma Ahvenlampi <oa at iki.fi> wrote:
> > > > I'm trying to reset the latency_timer on my on-board nforce2 sound
> to
> > > > 32 instead of 0, but it doesn't seem to work. I issue the command
> > > > "setpci -v -s 00:06.0 latency_timer=20 as root, but lspci still
> shows
> > > > the latency as being 0. Any ideas?
> >
> > I wonder if someone might be able to advice on a somewhat related
> > problem. I also have an nforce2 IGP system, but I'm suffering from
> > graphics related instability. The X server sometimes crashes, sometimes
> > locks up intermittently. It's 100% repeatable when trying to do anything
> > OpenGL related (mythtv opengl vsync, glxgears, etc), and otherwise very
> > unpredictable. The same box has, however, already run MythTV otherwise
> > fine for over a year, and OpenGL has worked before.
> >
> > I've gone through a lot of stuff - different NVidia driver versions (now
> > using 7676, but tried many of the 8xxx releases as well), system
> > temperatures (fairly high in the Silverstone LC11 case, but not totally
> > overboard), turning off AGP modes, underclocking the graphics, etc. The
> > thing I'm beginning to suspect might be the reason is a shared interrupt
> > between the nvidia IGP and one of my DVB capture cards. However, I
> > haven't figured out a way to change that. I do have free IRQ lines in
> > the system, but no free PCI slots (the small case leaves me no
> > flexibility in this respect).
> >
> > I'm using an ASUS A7N8X-VM/400, with two Technisat AirStar 2 PCI cards.
> > These are the only components in the system, with all other functions
> > integrated on the M/B. The BIOS does not give me a lot of options. I
> > haven't tried ripping out the DVB cards to verify my suspicion, since
> > this box is the only TV receiver we have...
> >
> > So, how could I move some functions out of the IRQ lines 10 and 11, to
> > any of the free ones, like 3, 6, 7, or 13? Some hardware info below..
> >
>
> You don't mention what distribution or kernel you are running, there
> was a problem IO-APIC lockup problem with older 2.4 kernels on NForce2
> boards, but I doubt you are running anything that old.
>
> Try recompiling your kernel and enabling CONFIG_X86_UP_IOAPIC in your
> kernel config.  This will make it switch to using different interrupt
> controller code rather than the legacy XT PIC that your board is
> currently using, here's the output from my SIS755 mythbackend, the
> high errors are being caused by a known phantom IRQ problem on this
> board which is harmless:
>
> [root at mythbackend ~]# cat /proc/interrupts
>            CPU0
>   0:  355696015    IO-APIC-edge  timer
>   1:        284    IO-APIC-edge  i8042
>   7:          2    IO-APIC-edge  parport0
>   8:          0    IO-APIC-edge  rtc
>   9:          1   IO-APIC-level  acpi
> 12:        110    IO-APIC-edge  i8042
> 15:   12754335    IO-APIC-edge  ide1
> 169:    9252860   IO-APIC-level  ide2, ide3, ide4, ide5, libata
> 177:   10711697   IO-APIC-level  eth0
> 185:   10314019   IO-APIC-level  ivtv0
> 193:          0   IO-APIC-level  ehci_hcd:usb1
> 201:          0   IO-APIC-level  ohci_hcd:usb2
> 209:          0   IO-APIC-level  ohci_hcd:usb3
> 217:          0   IO-APIC-level  ohci_hcd:usb4
> NMI:      75895
> LOC:  355714402
> ERR:      85708
> MIS:          0
>
> Switching to APIC should allow your peripherals to use "level"
> detected PCI interrupts instead of "edge", which are much more
> efficient.




I'm using the APIC mode on my x86-64 system.  But, I still get several
devices bunched up on one IRQ, with others not being used at all.     Is
there any way to tell the Linux kernel to distribute the interrupts
differently?  Or, is that hardwired in the PCI configuration?   Also, the
old IRQ's for standard devices (IRQ 3/4 for serial, IRQ 7 for Parallel, etc)
never seem to get re-used.

Is there any way to optimize this in Linux?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mythtv.org/pipermail/mythtv-users/attachments/20060225/f784795f/attachment.htm