processes, interrupts, and cpus

Discussion:

(too old to reply)

f***@protonmail.com

2024-08-20 18:10:01 UTC

i have a video capture card with 4 separate chips
when i start up i get bttv0 - bttv3
i'm getting a lot of errors in syslog on bttv2 and bttv3
like
bttv: 2: timeout: drop=3888442 irq=27910652/28054718, risc=338f048c, bits: HSYNC
bttv: 3: timeout: drop=3940265 irq=27929862/27929862, risc=33adc01c, bits: HSYNC

could it be related to interrupts and cpus
cat /proc/interrupts
CPU0 CPU1 CPU2 CPU3
16: 0 26950435 0 0 IO-APIC 16-fasteoi bttv1
17: 0 0 28079982 0 IO-APIC 17-fasteoi bttv2
18: 0 0 27950635 0 IO-APIC 18-fasteoi i801_smbus, bttv3
19: 66003 0 0 0 IO-APIC 19-fasteoi bttv0

i notice bttv2 and bttv3 are on the same cpu
and bttv3 is sharing an interrupt
is this related to my timeout errors
if so is it correctable

Dan Ritter

2024-08-20 22:30:01 UTC

Permalink

Post by f***@protonmail.com
i have a video capture card with 4 separate chips
when i start up i get bttv0 - bttv3
i'm getting a lot of errors in syslog on bttv2 and bttv3
like
bttv: 2: timeout: drop=3888442 irq=27910652/28054718, risc=338f048c, bits: HSYNC
bttv: 3: timeout: drop=3940265 irq=27929862/27929862, risc=33adc01c, bits: HSYNC
could it be related to interrupts and cpus
cat /proc/interrupts
CPU0 CPU1 CPU2 CPU3
16: 0 26950435 0 0 IO-APIC 16-fasteoi bttv1
17: 0 0 28079982 0 IO-APIC 17-fasteoi bttv2
18: 0 0 27950635 0 IO-APIC 18-fasteoi i801_smbus, bttv3
19: 66003 0 0 0 IO-APIC 19-fasteoi bttv0
i notice bttv2 and bttv3 are on the same cpu
and bttv3 is sharing an interrupt
is this related to my timeout errors
if so is it correctable

Depending on the hardware -- I don't remember the ins and outs
of most video capture cards these days -- you should be able to
set the interrupt per card either in hardware or at kernel
module load time.

This could certainly be an issue, especially on a relatively
low-powered machine; if you don't have the irqbalance daemon
installed, you might try that first.

-dsr-

f***@protonmail.com

2024-08-20 23:40:01 UTC

Permalink

Post by Dan Ritter

Depending on the hardware -- I don't remember the ins and outs
of most video capture cards these days -- you should be able to
set the interrupt per card either in hardware or at kernel
module load time.
This could certainly be an issue, especially on a relatively
low-powered machine; if you don't have the irqbalance daemon
installed, you might try that first.
-dsr-

i installed irqbalance and it dramatically reduced the errors
from 2-3 a minute to 1 every 3-4 minutes

Dan Ritter

2024-08-21 00:00:01 UTC

Permalink

Post by f***@protonmail.com

Post by Dan Ritter

Depending on the hardware -- I don't remember the ins and outs
of most video capture cards these days -- you should be able to
set the interrupt per card either in hardware or at kernel
module load time.
This could certainly be an issue, especially on a relatively
low-powered machine; if you don't have the irqbalance daemon
installed, you might try that first.
-dsr-

i installed irqbalance and it dramatically reduced the errors
from 2-3 a minute to 1 every 3-4 minutes

OK, now you know it's an IRQ sharing problem. Look up docs for
your card and driver module and see if you can push them to
non-competing IRQs.

If they are PCI rather than PCIe cards, you may need to
physically switch slots, or there may be a BIOS change
available.

-dsr-