x86: expand /proc/interrupts to include missing vectors, v2
Joe Korty [Wed, 17 Oct 2007 16:04:40 +0000 (18:04 +0200)]
Add missing IRQs and IRQ descriptions to /proc/interrupts.

/proc/interrupts is most useful when it displays every IRQ vector in use by
the system, not just those somebody thought would be interesting.

This patch inserts the following vector displays to the i386 and x86_64
platforms, as appropriate:

rescheduling interrupts
TLB flush interrupts
function call interrupts
thermal event interrupts
threshold interrupts
spurious interrupts

A threshold interrupt occurs when ECC memory correction is occuring at too
high a frequency.  Thresholds are used by the ECC hardware as occasional
ECC failures are part of normal operation, but long sequences of ECC
failures usually indicate a memory chip that is about to fail.

Thermal event interrupts occur when a temperature threshold has been
exceeded for some CPU chip.  IIRC, a thermal interrupt is also generated
when the temperature drops back to a normal level.

A spurious interrupt is an interrupt that was raised then lowered by the
device before it could be fully processed by the APIC.  Hence the apic sees
the interrupt but does not know what device it came from.  For this case
the APIC hardware will assume a vector of 0xff.

Rescheduling, call, and TLB flush interrupts are sent from one CPU to
another per the needs of the OS.  Typically, their statistics would be used
to discover if an interrupt flood of the given type has been occuring.

AK: merged v2 and v4 which had some more tweaks
AK: replace Local interrupts with Local timer interrupts
AK: Fixed description of interrupt types.

[ tglx: arch/x86 adaptation ]
[ mingo: small cleanup ]

Signed-off-by: Joe Korty <joe.korty@ccur.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Tim Hockin <thockin@hockin.org>
Cc: Andi Kleen <ak@suse.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

14 files changed:
Documentation/filesystems/proc.txt
arch/x86/kernel/apic_32.c
arch/x86/kernel/apic_64.c
arch/x86/kernel/cpu/mcheck/p4.c
arch/x86/kernel/irq_32.c
arch/x86/kernel/irq_64.c
arch/x86/kernel/mce_amd_64.c
arch/x86/kernel/mce_intel_64.c
arch/x86/kernel/smp_32.c
arch/x86/kernel/smp_64.c
arch/x86/mach-voyager/voyager_smp.c
arch/x86/xen/smp.c
include/asm-x86/hardirq_32.h
include/asm-x86/pda.h

index 4a37e25..e5c1df5 100644 (file)
@@ -347,7 +347,35 @@ connects the CPUs in a SMP system. This means that an error has been detected,
 the IO-APIC automatically retry the transmission, so it should not be a big
 problem, but you should read the SMP-FAQ.
 
-In this context it could be interesting to note the new irq directory in 2.4.
+In 2.6.2* /proc/interrupts was expanded again.  This time the goal was for
+/proc/interrupts to display every IRQ vector in use by the system, not
+just those considered 'most important'.  The new vectors are:
+
+  THR -- interrupt raised when a machine check threshold counter
+  (typically counting ECC corrected errors of memory or cache) exceeds
+  a configurable threshold.  Only available on some systems.
+
+  TRM -- a thermal event interrupt occurs when a temperature threshold
+  has been exceeded for the CPU.  This interrupt may also be generated
+  when the temperature drops back to normal.
+
+  SPU -- a spurious interrupt is some interrupt that was raised then lowered
+  by some IO device before it could be fully processed by the APIC.  Hence
+  the APIC sees the interrupt but does not know what device it came from.
+  For this case the APIC will generate the interrupt with a IRQ vector
+  of 0xff. This might also be generated by chipset bugs.
+
+  RES, CAL, TLB -- rescheduling, call and TLB flush interrupts are
+  sent from one CPU to another per the needs of the OS.  Typically,
+  their statistics are used by kernel developers and interested users to
+  determine the occurance of interrupt of the given type.
+
+The above IRQ vectors are displayed only when relevent.  For example,
+the threshold vector does not exist on x86_64 platforms.  Others are
+suppressed when the system is a uniprocessor.  As of this writing, only
+i386 and x86_64 platforms support the new IRQ vector displays.
+
+Of some interest is the introduction of the /proc/irq directory to 2.4.
 It could be used to set IRQ to CPU affinity, this means that you can "hook" an
 IRQ to only one CPU, or to exclude a CPU of handling IRQs. The contents of the
 irq subdir is one subdir for each IRQ, and one file; prof_cpu_mask
index 3d67ae1..793341f 100644 (file)
@@ -1277,6 +1277,7 @@ void smp_spurious_interrupt(struct pt_regs *regs)
        /* see sw-dev-man vol 3, chapter 7.4.13.5 */
        printk(KERN_INFO "spurious APIC interrupt on CPU#%d, "
               "should never happen.\n", smp_processor_id());
+       __get_cpu_var(irq_stat).irq_spurious_count++;
        irq_exit();
 }
 
index 2250c65..f47bc49 100644 (file)
@@ -1140,6 +1140,7 @@ asmlinkage void smp_spurious_interrupt(void)
        if (v & (1 << (SPURIOUS_APIC_VECTOR & 0x1f)))
                ack_APIC_irq();
 
+       add_pda(irq_spurious_count, 1);
        irq_exit();
 }
 
index 1509edf..be4dabf 100644 (file)
@@ -61,6 +61,7 @@ fastcall void smp_thermal_interrupt(struct pt_regs *regs)
 {
        irq_enter();
        vendor_thermal_interrupt(regs);
+       __get_cpu_var(irq_stat).irq_thermal_count++;
        irq_exit();
 }
 
index e173b76..10f3590 100644 (file)
@@ -280,14 +280,41 @@ skip:
                seq_printf(p, "NMI: ");
                for_each_online_cpu(j)
                        seq_printf(p, "%10u ", nmi_count(j));
-               seq_putc(p, '\n');
+               seq_printf(p, "  Non-maskable interrupts\n");
 #ifdef CONFIG_X86_LOCAL_APIC
                seq_printf(p, "LOC: ");
                for_each_online_cpu(j)
                        seq_printf(p, "%10u ",
                                per_cpu(irq_stat,j).apic_timer_irqs);
-               seq_putc(p, '\n');
+               seq_printf(p, "  Local timer interrupts\n");
 #endif
+#ifdef CONFIG_SMP
+               seq_printf(p, "RES: ");
+               for_each_online_cpu(j)
+                       seq_printf(p, "%10u ",
+                               per_cpu(irq_stat,j).irq_resched_count);
+               seq_printf(p, "  Rescheduling interrupts\n");
+               seq_printf(p, "CAL: ");
+               for_each_online_cpu(j)
+                       seq_printf(p, "%10u ",
+                               per_cpu(irq_stat,j).irq_call_count);
+               seq_printf(p, "  function call interrupts\n");
+               seq_printf(p, "TLB: ");
+               for_each_online_cpu(j)
+                       seq_printf(p, "%10u ",
+                               per_cpu(irq_stat,j).irq_tlb_count);
+               seq_printf(p, "  TLB shootdowns\n");
+#endif
+               seq_printf(p, "TRM: ");
+               for_each_online_cpu(j)
+                       seq_printf(p, "%10u ",
+                               per_cpu(irq_stat,j).irq_thermal_count);
+               seq_printf(p, "  Thermal event interrupts\n");
+               seq_printf(p, "SPU: ");
+               for_each_online_cpu(j)
+                       seq_printf(p, "%10u ",
+                               per_cpu(irq_stat,j).irq_spurious_count);
+               seq_printf(p, "  Spurious interrupts\n");
                seq_printf(p, "ERR: %10u\n", atomic_read(&irq_err_count));
 #if defined(CONFIG_X86_IO_APIC)
                seq_printf(p, "MIS: %10u\n", atomic_read(&irq_mis_count));
index 865669e..3881189 100644 (file)
@@ -86,11 +86,37 @@ skip:
                seq_printf(p, "NMI: ");
                for_each_online_cpu(j)
                        seq_printf(p, "%10u ", cpu_pda(j)->__nmi_count);
-               seq_putc(p, '\n');
+               seq_printf(p, "  Non-maskable interrupts\n");
                seq_printf(p, "LOC: ");
                for_each_online_cpu(j)
                        seq_printf(p, "%10u ", cpu_pda(j)->apic_timer_irqs);
-               seq_putc(p, '\n');
+               seq_printf(p, "  Local timer interrupts\n");
+#ifdef CONFIG_SMP
+               seq_printf(p, "RES: ");
+               for_each_online_cpu(j)
+                       seq_printf(p, "%10u ", cpu_pda(j)->irq_resched_count);
+               seq_printf(p, "  Rescheduling interrupts\n");
+               seq_printf(p, "CAL: ");
+               for_each_online_cpu(j)
+                       seq_printf(p, "%10u ", cpu_pda(j)->irq_call_count);
+               seq_printf(p, "  function call interrupts\n");
+               seq_printf(p, "TLB: ");
+               for_each_online_cpu(j)
+                       seq_printf(p, "%10u ", cpu_pda(j)->irq_tlb_count);
+               seq_printf(p, "  TLB shootdowns\n");
+#endif
+               seq_printf(p, "TRM: ");
+               for_each_online_cpu(j)
+                       seq_printf(p, "%10u ", cpu_pda(j)->irq_thermal_count);
+               seq_printf(p, "  Thermal event interrupts\n");
+               seq_printf(p, "THR: ");
+               for_each_online_cpu(j)
+                       seq_printf(p, "%10u ", cpu_pda(j)->irq_threshold_count);
+               seq_printf(p, "  Threshold APIC interrupts\n");
+               seq_printf(p, "SPU: ");
+               for_each_online_cpu(j)
+                       seq_printf(p, "%10u ", cpu_pda(j)->irq_spurious_count);
+               seq_printf(p, "  Spurious interrupts\n");
                seq_printf(p, "ERR: %10u\n", atomic_read(&irq_err_count));
        }
        return 0;
index 805b62b..0d2afd9 100644 (file)
@@ -237,6 +237,7 @@ asmlinkage void mce_threshold_interrupt(void)
                }
        }
 out:
+       add_pda(irq_threshold_count, 1);
        irq_exit();
 }
 
index 6551505..c17eaf5 100644 (file)
@@ -26,6 +26,7 @@ asmlinkage void smp_thermal_interrupt(void)
        if (therm_throt_process(msr_val & 1))
                mce_log_therm_throt_event(smp_processor_id(), msr_val);
 
+       add_pda(irq_thermal_count, 1);
        irq_exit();
 }
 
index eebc6e8..791d9f8 100644 (file)
@@ -342,6 +342,7 @@ fastcall void smp_invalidate_interrupt(struct pt_regs *regs)
        smp_mb__after_clear_bit();
 out:
        put_cpu_no_resched();
+       __get_cpu_var(irq_stat).irq_tlb_count++;
 }
 
 void native_flush_tlb_others(const cpumask_t *cpumaskp, struct mm_struct *mm,
@@ -640,6 +641,7 @@ static void native_smp_send_stop(void)
 fastcall void smp_reschedule_interrupt(struct pt_regs *regs)
 {
        ack_APIC_irq();
+       __get_cpu_var(irq_stat).irq_resched_count++;
 }
 
 fastcall void smp_call_function_interrupt(struct pt_regs *regs)
@@ -660,6 +662,7 @@ fastcall void smp_call_function_interrupt(struct pt_regs *regs)
         */
        irq_enter();
        (*func)(info);
+       __get_cpu_var(irq_stat).irq_call_count++;
        irq_exit();
 
        if (wait) {
index df4a828..5c29647 100644 (file)
@@ -163,6 +163,7 @@ asmlinkage void smp_invalidate_interrupt(struct pt_regs *regs)
 out:
        ack_APIC_irq();
        cpu_clear(cpu, f->flush_cpumask);
+       add_pda(irq_tlb_count, 1);
 }
 
 static void flush_tlb_others(cpumask_t cpumask, struct mm_struct *mm,
@@ -493,6 +494,7 @@ void smp_send_stop(void)
 asmlinkage void smp_reschedule_interrupt(void)
 {
        ack_APIC_irq();
+       add_pda(irq_resched_count, 1);
 }
 
 asmlinkage void smp_call_function_interrupt(void)
@@ -514,6 +516,7 @@ asmlinkage void smp_call_function_interrupt(void)
        exit_idle();
        irq_enter();
        (*func)(info);
+       add_pda(irq_call_count, 1);
        irq_exit();
        if (wait) {
                mb();
index 1f86b52..e4928aa 100644 (file)
@@ -1037,6 +1037,7 @@ smp_call_function_interrupt(void)
         */
        irq_enter();
        (*func)(info);
+       __get_cpu_var(irq_stat).irq_call_count++;
        irq_exit();
        if (wait) {
                mb();
index 4fa33c2..6c05858 100644 (file)
@@ -356,6 +356,7 @@ static irqreturn_t xen_call_function_interrupt(int irq, void *dev_id)
         */
        irq_enter();
        (*func)(info);
+       __get_cpu_var(irq_stat).irq_call_count++;
        irq_exit();
 
        if (wait) {
index 9188635..4f85f0f 100644 (file)
@@ -10,6 +10,11 @@ typedef struct {
        unsigned int __nmi_count;       /* arch dependent */
        unsigned int apic_timer_irqs;   /* arch dependent */
        unsigned int irq0_irqs;
+       unsigned int irq_resched_count;
+       unsigned int irq_call_count;
+       unsigned int irq_tlb_count;
+       unsigned int irq_thermal_count;
+       unsigned int irq_spurious_count;
 } ____cacheline_aligned irq_cpustat_t;
 
 DECLARE_PER_CPU(irq_cpustat_t, irq_stat);
index fb49f80..35962bb 100644 (file)
@@ -30,6 +30,12 @@ struct x8664_pda {
        struct mm_struct *active_mm;
        unsigned apic_timer_irqs;
        unsigned irq0_irqs;
+       unsigned irq_resched_count;
+       unsigned irq_call_count;
+       unsigned irq_tlb_count;
+       unsigned irq_thermal_count;
+       unsigned irq_threshold_count;
+       unsigned irq_spurious_count;
 } ____cacheline_aligned_in_smp;
 
 extern struct x8664_pda *_cpu_pda[];