Hi
I have Intel(R) Xeon(R) Silver 4410Y platform. (I have 48 cores.)
The IRQ interrupts now send to all available cores as given below for my ethernet interface ‘eth3’
# find /proc/irq/ -name "*eth3*"
/proc/irq/361/ice-eth3-TxRx-0
/proc/irq/362/ice-eth3-TxRx-1
/proc/irq/363/ice-eth3-TxRx-2
/proc/irq/364/ice-eth3-TxRx-3
/proc/irq/365/ice-eth3-TxRx-4
/proc/irq/366/ice-eth3-TxRx-5
/proc/irq/367/ice-eth3-TxRx-6
/proc/irq/368/ice-eth3-TxRx-7
/proc/irq/369/ice-eth3-TxRx-8
/proc/irq/370/ice-eth3-TxRx-9
/proc/irq/371/ice-eth3-TxRx-10
/proc/irq/372/ice-eth3-TxRx-11
/proc/irq/373/ice-eth3-TxRx-12
/proc/irq/374/ice-eth3-TxRx-13
/proc/irq/375/ice-eth3-TxRx-14
/proc/irq/376/ice-eth3-TxRx-15
/proc/irq/377/ice-eth3-TxRx-16
/proc/irq/378/ice-eth3-TxRx-17
/proc/irq/379/ice-eth3-TxRx-18
/proc/irq/380/ice-eth3-TxRx-19
/proc/irq/381/ice-eth3-TxRx-20
/proc/irq/382/ice-eth3-TxRx-21
/proc/irq/383/ice-eth3-TxRx-22
/proc/irq/384/ice-eth3-TxRx-23
/proc/irq/385/ice-eth3-TxRx-24
/proc/irq/386/ice-eth3-TxRx-25
/proc/irq/387/ice-eth3-TxRx-26
/proc/irq/388/ice-eth3-TxRx-27
/proc/irq/389/ice-eth3-TxRx-28
/proc/irq/390/ice-eth3-TxRx-29
/proc/irq/391/ice-eth3-TxRx-30
/proc/irq/392/ice-eth3-TxRx-31
/proc/irq/393/ice-eth3-TxRx-32
/proc/irq/394/ice-eth3-TxRx-33
/proc/irq/395/ice-eth3-TxRx-34
/proc/irq/396/ice-eth3-TxRx-35
/proc/irq/397/ice-eth3-TxRx-36
/proc/irq/398/ice-eth3-TxRx-37
/proc/irq/399/ice-eth3-TxRx-38
/proc/irq/400/ice-eth3-TxRx-39
/proc/irq/401/ice-eth3-TxRx-40
/proc/irq/402/ice-eth3-TxRx-41
/proc/irq/403/ice-eth3-TxRx-42
/proc/irq/404/ice-eth3-TxRx-43
/proc/irq/405/ice-eth3-TxRx-44
/proc/irq/406/ice-eth3-TxRx-45
/proc/irq/407/ice-eth3-TxRx-46
/proc/irq/408/ice-eth3-TxRx-47
I need to divert all the IRQ interrupts only to CPU core 4, 6, 8, 10.
I used the utility ./set_irq_affinity 4, 6, 8, 10 eth3
The outcome when I run the above command in Intel(R) Xeon(R) CPU E5-2650 v3 @ 2.30GHz is,
eth3 4 10 → /proc/irq/85/smp_affinity
eth3 6 40 → /proc/irq/86/smp_affinity
eth3 8 100 → /proc/irq/87/smp_affinity
eth3 10 400 → /proc/irq/88/smp_affinity
Now the question is, why does the same command give a different result in Intel(R) Xeon(R) Silver 4410Y? How can I converge IRQ requests only to 4, 6, 8, 10 cores?
Please suggest your thoughts.