Re: sio: lots of silo overflows on Asus K8V with Moxa Smartio C104H/PCI

On Sat, 1 May 2004, Burkard Meyendriesch wrote:

> On Sat, 1 May 2004 14:23:13 +1000 (EST) Bruce Evans wrote:
> > ...
> > locking.  Try the following patch if you still have shared interrupts:
> > ...
> > I don't expect this to fix your problem.
>
> Thank you for the patch! I just installed it in a new kernel. Result:
>
> /var/log/messages after boot -v:
> ..
> May  1 09:07:31 Reineke kernel: puc0: <Moxa Technologies, Smartio C104H/PCI> port 0xa000-0xa00f,0xa400-0xa43f,0xa800-0xa87f irq 19 at device 14.0 on pci0
> May  1 09:07:31 Reineke kernel: puc0: Reserved 0x40 bytes for rid 0x18 type 4 at 0xa400
> May  1 09:07:31 Reineke kernel: sio4: <Moxa Technologies, Smartio C104H/PCI> on puc0
> May  1 09:07:31 Reineke kernel: sio4: type 16550A
> May  1 09:07:31 Reineke kernel: sio4: unable to activate interrupt in fast mode - using normal mode
> ..

Keep trying to make it use a fast interrupt.  Most likely you need
PUC_FASTINTR.

> My Hardware is an Asus K8V Athlon64 board; in my BIOS I enabled "ACPI 2.0"
> and "ACPI APIC".
> >
> >     19200 bps = 1920 cps = at least 1920/16 = 120 interrupts/second
> >
> > Here 16 is the input fifo size for the usual case of a 16550A UART.
> > Use systat -v to see the transient interrupt rate.
> >
> At 115200 bps my puc interrupt rate is at about 300 ints/s, but I see
> peeks up to 1500 ints/s. iostat 10 gives me an average char input rate
> of 2300 chars/s. The silo overflow rate is about 1 per 5 seconds.

It could still be an interrupt storm.  I didn't noticed that you had
hz = 1024.  This gives about 102 interrupts/second from correct, but
the storm handling isn't quite right and can permit more interrupts
than intended.

Another reson to try using a fast interrupt is that interrupt storm
protection doesn't apply to them, so they should make the problem
obvious by hanging the system :-).

> At 9600 bps the puc interrupt rate is about 105 ints/s with 730 char/s
> input character rate. At this low speed I don't see any silo overflows.

105 interrupts/second looks very like storm handling.  The normal
interrupt rate for 9600 bps input sustained is 960/14 = 68.6, and 9600
bps sustained should give 960 cps.  Output increases the interrupt
rate a little but should affect input speed.

> Btw, systat -v shows a constant high interrupt rate of 1024 0: clk;
> is this normal?

1024 is normal for amd64 only. "0: clk" is normal misformatting.
RELENG_4 has the correct formatting "clk irq0".

Bruce