RE: [patch V5 00/15] Improve /proc/interrupts further

[Michael Kelley]

From: Thomas Gleixner <tglx@xxxxxxxxxx> Sent: Wednesday, April 1, 2026 2:51 PM
> 
> This is a follow up to v4 which can be found here:
> 
>   https://lore.kernel.org/20260331071453.172185305@xxxxxxxxxx
> 
> The v1 cover letter contains a full analysis, explanation and numbers:
> 
>   https://lore.kernel.org/20260303150539.513068586@xxxxxxxxxx
> 
> TLDR:
> 
>   - The performance of reading of /proc/interrupts has been improved
>     piecewise over the years, but most of the low hanging fruit has been
>     left on the table.
> 

As I did with v2 of the patch series, I've tested in Hyper-V guests on
x86/x64 and arm64. Did basic smoke tests of taking a CPU offline, and
removing a PCI device along with its IRQs, then adding them back again.
I did not do anything with the new binary interface or the gdb Python
script.

The improved alignment looks good. But if you want to be picky
about the alignment, I noticed these things:

On x86, the leftmost column now correctly aligns the "VPMI:" label.
But the rightmost column does not correctly align the text
"Perf Guest Mediated PMI". It needs one additional space. Here's my
output (with some CPU columns removed so it's not so wide):

root@mhkubun:~# cat /proc/interrupts | cut -b 1-30,64-
            CPU0       CPU1       CPU5       CPU6       CPU7
   8:          0          0          0          0          0  IO-APIC     8-edge      rtc0
   9:          0          0          0          0          0  IO-APIC     9-fasteoi   acpi
 NMI:          0          0          0          0          0  Non-maskable interrupts
 LOC:          0          0          0          0          0  Local timer interrupts
 PMI:          0          0          0          0          0  Performance monitoring interrupts
 IWI:          1          0          0          0          0  IRQ work interrupts
 RES:        325        231        235        341        231  Rescheduling interrupts
 CAL:      11547       7178       9033       8798       6738  Function call interrupts
 TLB:          0          0          0          0          0  TLB shootdowns
 TRM:          0          0          0          0          0  Thermal event interrupt
 THR:          0          0          0          0          0  Threshold APIC interrupts
 MCE:          0          0          0          0          0  Machine check exceptions
 MCP:          3          3          3          3          3  Machine check polls
 HYP:       7300       5185        278       6960       1213  Hypervisor callback interrupts
 HRE:          0          0          0          0          0  Hyper-V reenlightenment interrupts
 HVS:       8963       4359       9784       8374      61709  Hyper-V stimer0 interrupts
 PIN:          0          0          0          0          0  Posted-interrupt notification event
 NPI:          0          0          0          0          0  Nested posted-interrupt event
 PIW:          0          0          0          0          0  Posted-interrupt wakeup event
VPMI:          0          0          0          0          0 Perf Guest Mediated PMI

On arm64, the leftmost column doesn't align the IPI<n> entries.
Neither does the rightmost column for the IPI<n> entries. Here's some
of my output:

 45:          0           0         80          0          0  HV-PCI-MSIX-0817:00:02.0  14 Edge      mlx5_comp13@pci:0817:00:02.0
 46:          0           0          0         10          0  HV-PCI-MSIX-0817:00:02.0  15 Edge      mlx5_comp14@pci:0817:00:02.0
 47:          0           0          0          0         45  HV-PCI-MSIX-0817:00:02.0  16 Edge      mlx5_comp15@pci:0817:00:02.0
IPI0:       906         536        649        495        606       Rescheduling interrupts
IPI1:     28844       12457      14770      55242      39175       Function call interrupts
IPI2:         0           0          0          0          0       CPU stop interrupts
IPI3:         0           0          0          0          0       CPU stop NMIs
IPI4:         0           0          0          0          0       Timer broadcast interrupts
IPI5:        11           1          5          0          2       IRQ work interrupts
IPI6:         0           0          0          0          0       CPU backtrace interrupts
IPI7:         0           0          0          0          0       KGDB roundup interrupts
Err:          0

The alignment notwithstanding, for the series (excluding the binary
interface and gdb Python),

Tested-by: Michael Kelley <mhklinux@xxxxxxxxxxx>