mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Radu Rendec <radu@rendec.net>
To: Thomas Gleixner <tglx@kernel.org>, LKML <linux-kernel@vger.kernel.org>
Cc: x86@kernel.org, Dmitry Ilvokhin <d@ilvokhin.com>,
	Neil Horman <nhorman@tuxdriver.com>,
	Florian Fainelli <florian.fainelli@broadcom.com>
Subject: Re: [patch v2 04/14] x86/irq: Make irqstats array based
Date: Wed, 25 Mar 2026 15:20:01 -0400	[thread overview]
Message-ID: <bd4f6714e758b657bb2d84e281e4721124029a11.camel@rendec.net> (raw)
In-Reply-To: <be11b44b8e7474b3662a7460e202bdeae227f3f4.camel@rendec.net>

On Tue, 2026-03-24 at 16:32 -0400, Radu Rendec wrote:
> On Tue, 2026-03-24 at 21:21 +0100, Thomas Gleixner wrote:
> > On Tue, Mar 24 2026 at 20:54, Thomas Gleixner wrote:
> > 
> > > On Mon, Mar 23 2026 at 15:24, Radu Rendec wrote:
> > > > On Fri, 2026-03-20 at 14:21 +0100, Thomas Gleixner wrote:
> > > > >  arch/x86/include/asm/hardirq.h      |   69 +++++-----
> > > > 
> > > > I believe this breaks scripts/gdb/linux/interrupts.py, particularly the
> > > > x86_show_irqstat() function, which still expects individual members in
> > > > struct irq_cpustat_t.
> > > 
> > > Uurg. I did not even know this exists. Let me try to polish the snake.
> > 
> > Something like the below should work, now I have to figure out how to
> > test it.
> 
> At a glance, it looks about right (although I haven't tested it). The
> test procedure involves running the kernel in a Qemu VM and attaching
> with gdb to Qemu's gdb server (apologies if I'm just stating the
> obvious here). It's briefly described here:
> Documentation/dev-tools/gdb-kernel-debugging.rst
> 
> The documentation doesn't elaborate on Qemu parameters. I'm using
> something like this:
> 
> qemu-system-x86_64 -nographic -m 1G -accel kvm -machine q35,hpet=off -cpu host -smp 4 \
>         -netdev bridge,br=vbr-nat,id=net0 -device virtio-net-pci,mac=52:54:98:aa:bb:cc,netdev=net0 \
>         -drive file=root.img,format=raw,if=virtio \
>         -kernel bzImage \
>         -append "console=ttyS0 root=/dev/vda1 rw nokaslr" \
>         -s
> 
> Happy to help if something doesn't quite work for you.
> 
> > ----
> > --- a/scripts/gdb/linux/interrupts.py
> > +++ b/scripts/gdb/linux/interrupts.py
> > @@ -97,8 +97,8 @@ irq_desc_type = utils.CachedType("struct
> >          text += "%*s: %10u\n" % (prec, "ERR", cnt['counter'])
> >      return text
> >  
> > -def x86_show_irqstat(prec, pfx, field, desc):
> > -    irq_stat = gdb.parse_and_eval("&irq_stat")
> > +def x86_show_irqstat(prec, pfx, idx, desc):
> > +    irq_stat = gdb.parse_and_eval("&irq_stat.counts[IRQ_COUNT_%s]" %idx)
> >      text = "%*s: " % (prec, pfx)
> >      for cpu in cpus.each_online_cpu():
> >          stat = cpus.per_cpu(irq_stat, cpu)
> > @@ -118,32 +118,51 @@ irq_desc_type = utils.CachedType("struct
> >      text = x86_show_irqstat(prec, "NMI", '__nmi_count', 'Non-maskable interrupts')
> >  
> >      if constants.LX_CONFIG_X86_LOCAL_APIC:
> > -        text += x86_show_irqstat(prec, "LOC", 'apic_timer_irqs', "Local timer interrupts")
> > -        text += x86_show_irqstat(prec, "SPU", 'irq_spurious_count', "Spurious interrupts")
> > -        text += x86_show_irqstat(prec, "PMI", 'apic_perf_irqs', "Performance monitoring interrupts")
> > -        text += x86_show_irqstat(prec, "IWI", 'apic_irq_work_irqs', "IRQ work interrupts")
> > -        text += x86_show_irqstat(prec, "RTR", 'icr_read_retry_count', "APIC ICR read retries")
> > +        text += x86_show_irqstat(prec, "LOC", 'APIC_TIMER', "Local timer interrupts")
> > +        text += x86_show_irqstat(prec, "SPU", 'SPURIOUS', "Spurious interrupts")
> > +        text += x86_show_irqstat(prec, "PMI", 'APIC_PERF', "Performance monitoring interrupts")
> > +        text += x86_show_irqstat(prec, "IWI", 'IRQ_WORK', "IRQ work interrupts")
> > +        text += x86_show_irqstat(prec, "RTR", 'ICR_READ_RETRY', "APIC ICR read retries")
> >          if utils.gdb_eval_or_none("x86_platform_ipi_callback") is not None:
> > -            text += x86_show_irqstat(prec, "PLT", 'x86_platform_ipis', "Platform interrupts")
> > +            text += x86_show_irqstat(prec, "PLT", 'X86_PLATFORM_IPI', "Platform interrupts")
> >  
> >      if constants.LX_CONFIG_SMP:
> > -        text += x86_show_irqstat(prec, "RES", 'irq_resched_count', "Rescheduling interrupts")
> > -        text += x86_show_irqstat(prec, "CAL", 'irq_call_count', "Function call interrupts")
> > -        text += x86_show_irqstat(prec, "TLB", 'irq_tlb_count', "TLB shootdowns")
> > +        text += x86_show_irqstat(prec, "RES", 'RESCHEDULE', "Rescheduling interrupts")
> > +        text += x86_show_irqstat(prec, "CAL", 'CALL_FUNCTION', "Function call interrupts")
> > +
> > +    text += x86_show_irqstat(prec, "TLB", 'TLB', "TLB shootdowns")
> >  
> >      if constants.LX_CONFIG_X86_THERMAL_VECTOR:
> > -        text += x86_show_irqstat(prec, "TRM", 'irq_thermal_count', "Thermal events interrupts")
> > +        text += x86_show_irqstat(prec, "TRM", 'THERMAL_APIC', "Thermal events interrupts")
> >  
> >      if constants.LX_CONFIG_X86_MCE_THRESHOLD:
> > -        text += x86_show_irqstat(prec, "THR", 'irq_threshold_count', "Threshold APIC interrupts")
> > +        text += x86_show_irqstat(prec, "THR", 'THRESHOLD_APIC', "Threshold APIC interrupts")
> >  
> >      if constants.LX_CONFIG_X86_MCE_AMD:
> > -        text += x86_show_irqstat(prec, "DFR", 'irq_deferred_error_count', "Deferred Error APIC interrupts")
> > +        text += x86_show_irqstat(prec, "DFR", 'DEFERRED_ERROR', "Deferred Error APIC interrupts")
> >  
> >      if constants.LX_CONFIG_X86_MCE:
> >          text += x86_show_mce(prec, "&mce_exception_count", "MCE", "Machine check exceptions")
> >          text += x86_show_mce(prec, "&mce_poll_count", "MCP", "Machine check polls")
> >  
> > +    if constants.LX_CONFIG_X86_HV_CALLBACK_VECTOR:
> > +        text += x86_show_irqstat(prec, 'HYP', 'HYPERVISOR_CALLBACK', 'Hypervisor callback interrupts')
> > +
> > +    if constants.LX_CONFIG_HYPERV:
> > +        text += x86_show_irqstat(prec, 'HRE', 'HYPERV_REENLIGHTENMENT', 'Hyper-V reenlightenment interrupts')
> > +        text += x86_show_irqstat(prec, 'HVS', 'HYPERV_STIMER0', 'Hyper-V stimer0 interrupts')
> > +
> > +    if constants.LX_CONFIG_KVM:
> > +        text += x86_show_irqstat(prec, "PIN", 'POSTED_INTR', 'Posted-interrupt notification event')
> > +        text += x86_show_irqstat(prec, "NPI", 'POSTED_INTR_NESTED', 'Nested posted-interrupt event')
> > +        text += x86_show_irqstat(prec, "PIW", 'POSTED_INTR_WAKEUP', 'Posted-interrupt wakeup event')
> > +
> > +    if constants.LX_CONFIG_GUEST_PERF_EVENTS:
> > +        text += x86_show_irqstat(prec, "VPMI", 'PERF_GUEST_MEDIATED_PMI', 'Perf Guest Mediated PMI')
> > +
> > +    if constants.LX_CONFIG_X86_POSTED_MSI:
> > +        text += x86_show_irqstat(prec, "PIN", 'POSTED_MSI_NOTIFICATION', 'Posted MSI notification event')
> > +
> >      text += show_irq_err_count(prec)
> >  
> >      if constants.LX_CONFIG_X86_IO_APIC:
> > @@ -151,11 +170,6 @@ irq_desc_type = utils.CachedType("struct
> >          if cnt is not None:
> >              text += "%*s: %10u\n" % (prec, "MIS", cnt['counter'])
> >  
> > -    if constants.LX_CONFIG_KVM:
> > -        text += x86_show_irqstat(prec, "PIN", 'kvm_posted_intr_ipis', 'Posted-interrupt notification event')
> > -        text += x86_show_irqstat(prec, "NPI", 'kvm_posted_intr_nested_ipis', 'Nested posted-interrupt event')
> > -        text += x86_show_irqstat(prec, "PIW", 'kvm_posted_intr_wakeup_ipis', 'Posted-interrupt wakeup event')
> > -
> >      return text
> >  
> >  def arm_common_show_interrupts(prec):

I tested it, and it's very close but requires a few extra changes,
which I included as a patch at the bottom.

There's one more gotcha: in the next patch (patch 5) you rename nr_irqs
to total_nr_irqs, and the change must be applied to interrupts.py as
well. It's used at the beginning of the LxInterruptList.invoke()
function.

One other issue I came across is that the fields below are not defined
for me (even though the corresponding config options do exist in the
kernel config file):
constants.LX_CONFIG_X86_HV_CALLBACK_VECTOR
constants.LX_CONFIG_HYPERV
constants.LX_CONFIG_GUEST_PERF_EVENTS
constants.LX_CONFIG_X86_POSTED_MSI

I didn't spend the time to figure out why. I'm pretty sure this is
completely unrelated to your changes.

--- a/scripts/gdb/linux/interrupts.py
+++ b/scripts/gdb/linux/interrupts.py
@@ -102,20 +102,12 @@ def x86_show_irqstat(prec, pfx, idx, desc):
     text = "%*s: " % (prec, pfx)
     for cpu in cpus.each_online_cpu():
         stat = cpus.per_cpu(irq_stat, cpu)
-        text += "%10u " % (stat[field])
-    text += "  %s\n" % (desc)
-    return text
-
-def x86_show_mce(prec, var, pfx, desc):
-    pvar = gdb.parse_and_eval(var)
-    text = "%*s: " % (prec, pfx)
-    for cpu in cpus.each_online_cpu():
-        text += "%10u " % (cpus.per_cpu(pvar, cpu).dereference())
+        text += "%10u " % (stat.dereference())
     text += "  %s\n" % (desc)
     return text
 
 def x86_show_interupts(prec):
-    text = x86_show_irqstat(prec, "NMI", '__nmi_count', 'Non-maskable interrupts')
+    text = x86_show_irqstat(prec, "NMI", 'NMI', 'Non-maskable interrupts')
 
     if constants.LX_CONFIG_X86_LOCAL_APIC:
         text += x86_show_irqstat(prec, "LOC", 'APIC_TIMER', "Local timer interrupts")
@@ -142,8 +134,8 @@ def x86_show_interupts(prec):
         text += x86_show_irqstat(prec, "DFR", 'DEFERRED_ERROR', "Deferred Error APIC interrupts")
 
     if constants.LX_CONFIG_X86_MCE:
-        text += x86_show_mce(prec, "&mce_exception_count", "MCE", "Machine check exceptions")
-        text += x86_show_mce(prec, "&mce_poll_count", "MCP", "Machine check polls")
+        text += x86_show_irqstat(prec, "MCE", "MCE_EXCEPTION", "Machine check exceptions")
+        text += x86_show_irqstat(prec, "MCP", "MCE_POLL", "Machine check polls")
 
     if constants.LX_CONFIG_X86_HV_CALLBACK_VECTOR:
         text += x86_show_irqstat(prec, 'HYP', 'HYPERVISOR_CALLBACK', 'Hypervisor callback interrupts')

  reply	other threads:[~2026-03-25 19:20 UTC|newest]

Thread overview: 46+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-03-20 13:21 [patch v2 00/14] Improve /proc/interrupts further and add a binary interface Thomas Gleixner
2026-03-20 13:21 ` [patch v2 01/14] x86/irq: Optimize interrupts decimals printing Thomas Gleixner
2026-03-21 16:10   ` Radu Rendec
2026-03-20 13:21 ` [patch v2 02/14] genirq/proc: Avoid formatting zero counts in /proc/interrupts Thomas Gleixner
2026-03-21 16:38   ` Radu Rendec
2026-03-20 13:21 ` [patch v2 03/14] genirq/proc: Utilize irq_desc::tot_count to avoid evaluation Thomas Gleixner
2026-03-22 19:59   ` Radu Rendec
2026-03-20 13:21 ` [patch v2 04/14] x86/irq: Make irqstats array based Thomas Gleixner
2026-03-20 16:39   ` Michael Kelley
2026-03-21 16:38     ` Thomas Gleixner
2026-03-21 20:32       ` Michael Kelley
2026-03-23 19:24   ` Radu Rendec
2026-03-24 19:54     ` Thomas Gleixner
2026-03-24 20:21       ` Thomas Gleixner
2026-03-24 20:32         ` Radu Rendec
2026-03-25 19:20           ` Radu Rendec [this message]
2026-03-25 22:52             ` Thomas Gleixner
2026-03-25 22:54               ` Florian Fainelli
2026-03-26 10:29                 ` Thomas Gleixner
2026-03-26 23:00                   ` Florian Fainelli
2026-03-27 11:03                     ` Thomas Gleixner
2026-03-26 12:34               ` Radu Rendec
2026-03-20 13:21 ` [patch v2 05/14] genirq: Expose nr_irqs in core code Thomas Gleixner
2026-03-23 19:48   ` Radu Rendec
2026-03-23 21:27     ` Thomas Gleixner
2026-03-20 13:21 ` [patch v2 06/14] genirq: Cache the condition for /proc/interrupts exposure Thomas Gleixner
2026-03-23 20:58   ` Radu Rendec
2026-03-24 20:31     ` Thomas Gleixner
2026-03-24 20:36       ` Radu Rendec
2026-03-20 13:21 ` [patch v2 07/14] genirq: Calculate precision only when required Thomas Gleixner
2026-03-25 19:47   ` Radu Rendec
2026-03-20 13:22 ` [patch v2 08/14] genirq: Add rcuref count to struct irq_desc Thomas Gleixner
2026-03-26 18:43   ` Dmitry Ilvokhin
2026-03-20 13:22 ` [patch v2 09/14] genirq: Expose irq_find_desc_at_or_after() in core code Thomas Gleixner
2026-03-26 19:13   ` Dmitry Ilvokhin
2026-03-26 21:11     ` Thomas Gleixner
2026-03-26 21:25       ` Thomas Gleixner
2026-03-20 13:22 ` [patch v2 10/14] genirq/proc: Speed up /proc/interrupts iteration Thomas Gleixner
2026-03-20 13:22 ` [patch v2 11/14] [RFC] genirq: Cache target CPU for single CPU affinities Thomas Gleixner
2026-03-20 13:22 ` [patch v2 12/14] [RFC] genirq/proc: Provide binary statistic interface Thomas Gleixner
2026-04-01 16:42   ` Dmitry Ilvokhin
2026-03-20 13:22 ` [patch v2 13/14] [RFC] genirq/proc: Provide architecture specific binary statistics Thomas Gleixner
2026-04-01 16:51   ` Dmitry Ilvokhin
2026-04-01 19:33     ` Thomas Gleixner
2026-03-20 13:22 ` [patch v2 14/14] [RFC] x86/irq: Hook up architecture specific stats Thomas Gleixner
2026-03-20 16:45 ` [patch v2 00/14] Improve /proc/interrupts further and add a binary interface Michael Kelley

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=bd4f6714e758b657bb2d84e281e4721124029a11.camel@rendec.net \
    --to=radu@rendec.net \
    --cc=d@ilvokhin.com \
    --cc=florian.fainelli@broadcom.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=nhorman@tuxdriver.com \
    --cc=tglx@kernel.org \
    --cc=x86@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®