Jan Beulich
2013-Mar-13 12:49 UTC
[PATCH] x86: extend diagnostics for "No irq handler for vector" messages
By storing the inverted IRQ number in vector_irq[], we may be able to
spot which IRQ a vector was used for most recently, thus hopefully
permitting to understand why these messages trigger on certain systems.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
--- a/xen/arch/x86/i8259.c
+++ b/xen/arch/x86/i8259.c
@@ -165,7 +165,7 @@ static void _disable_8259A_irq(unsigned
outb(cached_A1,0xA1);
else
outb(cached_21,0x21);
- per_cpu(vector_irq, 0)[LEGACY_VECTOR(irq)] = -1;
+ per_cpu(vector_irq, 0)[LEGACY_VECTOR(irq)] = ~irq;
spin_unlock_irqrestore(&i8259A_lock, flags);
}
--- a/xen/arch/x86/irq.c
+++ b/xen/arch/x86/irq.c
@@ -227,7 +227,7 @@ static void __clear_irq_vector(int irq)
for_each_cpu(cpu, &tmp_mask) {
ASSERT( per_cpu(vector_irq, cpu)[vector] == irq );
- per_cpu(vector_irq, cpu)[vector] = -1;
+ per_cpu(vector_irq, cpu)[vector] = ~irq;
}
desc->arch.vector = IRQ_VECTOR_UNASSIGNED;
@@ -253,8 +253,8 @@ static void __clear_irq_vector(int irq)
for_each_cpu(cpu, &tmp_mask) {
ASSERT( per_cpu(vector_irq, cpu)[old_vector] == irq );
TRACE_3D(TRC_HW_IRQ_MOVE_FINISH, irq, old_vector, cpu);
- per_cpu(vector_irq, cpu)[old_vector] = -1;
- }
+ per_cpu(vector_irq, cpu)[old_vector] = ~irq;
+ }
desc->arch.old_vector = IRQ_VECTOR_UNASSIGNED;
cpumask_clear(desc->arch.old_cpu_mask);
@@ -334,7 +334,7 @@ int __init init_irq_data(void)
int irq, vector;
for (vector = 0; vector < NR_VECTORS; ++vector)
- this_cpu(vector_irq)[vector] = -1;
+ this_cpu(vector_irq)[vector] = INT_MIN;
irq_desc = xzalloc_array(struct irq_desc, nr_irqs);
@@ -489,7 +489,7 @@ next:
goto next;
for_each_cpu(new_cpu, &tmp_mask)
- if (per_cpu(vector_irq, new_cpu)[vector] != -1)
+ if (per_cpu(vector_irq, new_cpu)[vector] >= 0)
goto next;
/* Found one! */
current_vector = vector;
@@ -551,7 +551,7 @@ void __setup_vector_irq(int cpu)
/* Clear vector_irq */
for (vector = 0; vector < NR_VECTORS; ++vector)
- per_cpu(vector_irq, cpu)[vector] = -1;
+ per_cpu(vector_irq, cpu)[vector] = INT_MIN;
/* Mark the inuse vectors */
for (irq = 0; irq < nr_irqs; ++irq) {
struct irq_desc *desc = irq_to_desc(irq);
@@ -622,7 +622,7 @@ void irq_move_cleanup_interrupt(struct c
struct irq_desc *desc;
irq = __get_cpu_var(vector_irq)[vector];
- if (irq == -1)
+ if ((int)irq < 0)
continue;
desc = irq_to_desc(irq);
@@ -655,7 +655,7 @@ void irq_move_cleanup_interrupt(struct c
TRACE_3D(TRC_HW_IRQ_MOVE_CLEANUP,
irq, vector, smp_processor_id());
- __get_cpu_var(vector_irq)[vector] = -1;
+ __get_cpu_var(vector_irq)[vector] = ~irq;
desc->arch.move_cleanup_count--;
if ( desc->arch.move_cleanup_count == 0 )
@@ -819,8 +819,22 @@ void do_IRQ(struct cpu_user_regs *regs)
if ( ! ( vector >= FIRST_LEGACY_VECTOR &&
vector <= LAST_LEGACY_VECTOR &&
bogus_8259A_irq(vector - FIRST_LEGACY_VECTOR) ) )
+ {
printk("CPU%u: No irq handler for vector %02x (IRQ
%d%s)\n",
smp_processor_id(), vector, irq, kind);
+ desc = irq_to_desc(~irq);
+ if ( ~irq < nr_irqs && irq_desc_initialized(desc) )
+ {
+ spin_lock(&desc->lock);
+ printk("IRQ%d a=%04lx[%04lx,%04lx] v=%02x[%02x]
t=%-15s s=%08x\n",
+ ~irq, *cpumask_bits(desc->affinity),
+ *cpumask_bits(desc->arch.cpu_mask),
+ *cpumask_bits(desc->arch.old_cpu_mask),
+ desc->arch.vector, desc->arch.old_vector,
+ desc->handler->typename, desc->status);
+ spin_unlock(&desc->lock);
+ }
+ }
TRACE_1D(TRC_HW_IRQ_UNMAPPED_VECTOR, vector);
}
goto out_no_unlock;
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
Andrew Cooper
2013-Mar-13 13:11 UTC
Re: [PATCH] x86: extend diagnostics for "No irq handler for vector" messages
On 13/03/13 12:49, Jan Beulich wrote:> By storing the inverted IRQ number in vector_irq[], we may be able to > spot which IRQ a vector was used for most recently, thus hopefully > permitting to understand why these messages trigger on certain systems. > > Signed-off-by: Jan Beulich <jbeulich@suse.com>Looks good to try and track down the issue. We have had one recent spurious assertion failure in map_domain_pirq() Assertion ''!test_bit(vector, desc->chip_data->used_vectors)'' failed at irq.c:1693 which might be related. So far I have been unsuccessful in reproducing the issue, but was wondering whether it had anything to do with the XSA-36 and flipping the default to be a single global intremap table. ~Andrew> > --- a/xen/arch/x86/i8259.c > +++ b/xen/arch/x86/i8259.c > @@ -165,7 +165,7 @@ static void _disable_8259A_irq(unsigned > outb(cached_A1,0xA1); > else > outb(cached_21,0x21); > - per_cpu(vector_irq, 0)[LEGACY_VECTOR(irq)] = -1; > + per_cpu(vector_irq, 0)[LEGACY_VECTOR(irq)] = ~irq; > spin_unlock_irqrestore(&i8259A_lock, flags); > } > > --- a/xen/arch/x86/irq.c > +++ b/xen/arch/x86/irq.c > @@ -227,7 +227,7 @@ static void __clear_irq_vector(int irq) > > for_each_cpu(cpu, &tmp_mask) { > ASSERT( per_cpu(vector_irq, cpu)[vector] == irq ); > - per_cpu(vector_irq, cpu)[vector] = -1; > + per_cpu(vector_irq, cpu)[vector] = ~irq; > } > > desc->arch.vector = IRQ_VECTOR_UNASSIGNED; > @@ -253,8 +253,8 @@ static void __clear_irq_vector(int irq) > for_each_cpu(cpu, &tmp_mask) { > ASSERT( per_cpu(vector_irq, cpu)[old_vector] == irq ); > TRACE_3D(TRC_HW_IRQ_MOVE_FINISH, irq, old_vector, cpu); > - per_cpu(vector_irq, cpu)[old_vector] = -1; > - } > + per_cpu(vector_irq, cpu)[old_vector] = ~irq; > + } > > desc->arch.old_vector = IRQ_VECTOR_UNASSIGNED; > cpumask_clear(desc->arch.old_cpu_mask); > @@ -334,7 +334,7 @@ int __init init_irq_data(void) > int irq, vector; > > for (vector = 0; vector < NR_VECTORS; ++vector) > - this_cpu(vector_irq)[vector] = -1; > + this_cpu(vector_irq)[vector] = INT_MIN; > > irq_desc = xzalloc_array(struct irq_desc, nr_irqs); > > @@ -489,7 +489,7 @@ next: > goto next; > > for_each_cpu(new_cpu, &tmp_mask) > - if (per_cpu(vector_irq, new_cpu)[vector] != -1) > + if (per_cpu(vector_irq, new_cpu)[vector] >= 0) > goto next; > /* Found one! */ > current_vector = vector; > @@ -551,7 +551,7 @@ void __setup_vector_irq(int cpu) > > /* Clear vector_irq */ > for (vector = 0; vector < NR_VECTORS; ++vector) > - per_cpu(vector_irq, cpu)[vector] = -1; > + per_cpu(vector_irq, cpu)[vector] = INT_MIN; > /* Mark the inuse vectors */ > for (irq = 0; irq < nr_irqs; ++irq) { > struct irq_desc *desc = irq_to_desc(irq); > @@ -622,7 +622,7 @@ void irq_move_cleanup_interrupt(struct c > struct irq_desc *desc; > irq = __get_cpu_var(vector_irq)[vector]; > > - if (irq == -1) > + if ((int)irq < 0) > continue; > > desc = irq_to_desc(irq); > @@ -655,7 +655,7 @@ void irq_move_cleanup_interrupt(struct c > TRACE_3D(TRC_HW_IRQ_MOVE_CLEANUP, > irq, vector, smp_processor_id()); > > - __get_cpu_var(vector_irq)[vector] = -1; > + __get_cpu_var(vector_irq)[vector] = ~irq; > desc->arch.move_cleanup_count--; > > if ( desc->arch.move_cleanup_count == 0 ) > @@ -819,8 +819,22 @@ void do_IRQ(struct cpu_user_regs *regs) > if ( ! ( vector >= FIRST_LEGACY_VECTOR && > vector <= LAST_LEGACY_VECTOR && > bogus_8259A_irq(vector - FIRST_LEGACY_VECTOR) ) ) > + { > printk("CPU%u: No irq handler for vector %02x (IRQ %d%s)\n", > smp_processor_id(), vector, irq, kind); > + desc = irq_to_desc(~irq); > + if ( ~irq < nr_irqs && irq_desc_initialized(desc) ) > + { > + spin_lock(&desc->lock); > + printk("IRQ%d a=%04lx[%04lx,%04lx] v=%02x[%02x] t=%-15s s=%08x\n", > + ~irq, *cpumask_bits(desc->affinity), > + *cpumask_bits(desc->arch.cpu_mask), > + *cpumask_bits(desc->arch.old_cpu_mask), > + desc->arch.vector, desc->arch.old_vector, > + desc->handler->typename, desc->status); > + spin_unlock(&desc->lock); > + } > + } > TRACE_1D(TRC_HW_IRQ_UNMAPPED_VECTOR, vector); > } > goto out_no_unlock; > >
Keir Fraser
2013-Mar-13 15:00 UTC
Re: [PATCH] x86: extend diagnostics for "No irq handler for vector" messages
On 13/03/2013 12:49, "Jan Beulich" <JBeulich@suse.com> wrote:> By storing the inverted IRQ number in vector_irq[], we may be able to > spot which IRQ a vector was used for most recently, thus hopefully > permitting to understand why these messages trigger on certain systems. > > Signed-off-by: Jan Beulich <jbeulich@suse.com>Acked-by: Keir Fraser <keir@xen.org>> --- a/xen/arch/x86/i8259.c > +++ b/xen/arch/x86/i8259.c > @@ -165,7 +165,7 @@ static void _disable_8259A_irq(unsigned > outb(cached_A1,0xA1); > else > outb(cached_21,0x21); > - per_cpu(vector_irq, 0)[LEGACY_VECTOR(irq)] = -1; > + per_cpu(vector_irq, 0)[LEGACY_VECTOR(irq)] = ~irq; > spin_unlock_irqrestore(&i8259A_lock, flags); > } > > --- a/xen/arch/x86/irq.c > +++ b/xen/arch/x86/irq.c > @@ -227,7 +227,7 @@ static void __clear_irq_vector(int irq) > > for_each_cpu(cpu, &tmp_mask) { > ASSERT( per_cpu(vector_irq, cpu)[vector] == irq ); > - per_cpu(vector_irq, cpu)[vector] = -1; > + per_cpu(vector_irq, cpu)[vector] = ~irq; > } > > desc->arch.vector = IRQ_VECTOR_UNASSIGNED; > @@ -253,8 +253,8 @@ static void __clear_irq_vector(int irq) > for_each_cpu(cpu, &tmp_mask) { > ASSERT( per_cpu(vector_irq, cpu)[old_vector] == irq ); > TRACE_3D(TRC_HW_IRQ_MOVE_FINISH, irq, old_vector, cpu); > - per_cpu(vector_irq, cpu)[old_vector] = -1; > - } > + per_cpu(vector_irq, cpu)[old_vector] = ~irq; > + } > > desc->arch.old_vector = IRQ_VECTOR_UNASSIGNED; > cpumask_clear(desc->arch.old_cpu_mask); > @@ -334,7 +334,7 @@ int __init init_irq_data(void) > int irq, vector; > > for (vector = 0; vector < NR_VECTORS; ++vector) > - this_cpu(vector_irq)[vector] = -1; > + this_cpu(vector_irq)[vector] = INT_MIN; > > irq_desc = xzalloc_array(struct irq_desc, nr_irqs); > > @@ -489,7 +489,7 @@ next: > goto next; > > for_each_cpu(new_cpu, &tmp_mask) > - if (per_cpu(vector_irq, new_cpu)[vector] != -1) > + if (per_cpu(vector_irq, new_cpu)[vector] >= 0) > goto next; > /* Found one! */ > current_vector = vector; > @@ -551,7 +551,7 @@ void __setup_vector_irq(int cpu) > > /* Clear vector_irq */ > for (vector = 0; vector < NR_VECTORS; ++vector) > - per_cpu(vector_irq, cpu)[vector] = -1; > + per_cpu(vector_irq, cpu)[vector] = INT_MIN; > /* Mark the inuse vectors */ > for (irq = 0; irq < nr_irqs; ++irq) { > struct irq_desc *desc = irq_to_desc(irq); > @@ -622,7 +622,7 @@ void irq_move_cleanup_interrupt(struct c > struct irq_desc *desc; > irq = __get_cpu_var(vector_irq)[vector]; > > - if (irq == -1) > + if ((int)irq < 0) > continue; > > desc = irq_to_desc(irq); > @@ -655,7 +655,7 @@ void irq_move_cleanup_interrupt(struct c > TRACE_3D(TRC_HW_IRQ_MOVE_CLEANUP, > irq, vector, smp_processor_id()); > > - __get_cpu_var(vector_irq)[vector] = -1; > + __get_cpu_var(vector_irq)[vector] = ~irq; > desc->arch.move_cleanup_count--; > > if ( desc->arch.move_cleanup_count == 0 ) > @@ -819,8 +819,22 @@ void do_IRQ(struct cpu_user_regs *regs) > if ( ! ( vector >= FIRST_LEGACY_VECTOR && > vector <= LAST_LEGACY_VECTOR && > bogus_8259A_irq(vector - FIRST_LEGACY_VECTOR) ) ) > + { > printk("CPU%u: No irq handler for vector %02x (IRQ %d%s)\n", > smp_processor_id(), vector, irq, kind); > + desc = irq_to_desc(~irq); > + if ( ~irq < nr_irqs && irq_desc_initialized(desc) ) > + { > + spin_lock(&desc->lock); > + printk("IRQ%d a=%04lx[%04lx,%04lx] v=%02x[%02x] t=%-15s > s=%08x\n", > + ~irq, *cpumask_bits(desc->affinity), > + *cpumask_bits(desc->arch.cpu_mask), > + *cpumask_bits(desc->arch.old_cpu_mask), > + desc->arch.vector, desc->arch.old_vector, > + desc->handler->typename, desc->status); > + spin_unlock(&desc->lock); > + } > + } > TRACE_1D(TRC_HW_IRQ_UNMAPPED_VECTOR, vector); > } > goto out_no_unlock; > > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xen.org > http://lists.xen.org/xen-devel
Jan Beulich
2013-Mar-15 08:46 UTC
Re: [PATCH] x86: extend diagnostics for "No irq handler for vector" messages
>>> On 13.03.13 at 13:49, "Jan Beulich" <JBeulich@suse.com> wrote: > By storing the inverted IRQ number in vector_irq[], we may be able to > spot which IRQ a vector was used for most recently, thus hopefully > permitting to understand why these messages trigger on certain systems.So this made quite obvious that it''s the HPET FSB interrupts that misbehave in some way. I''ll be looking into this, but can''t estimate yet when exactly. I guess we can keep the patch in nevertheless, perhaps with ...> @@ -819,8 +819,22 @@ void do_IRQ(struct cpu_user_regs *regs) > if ( ! ( vector >= FIRST_LEGACY_VECTOR && > vector <= LAST_LEGACY_VECTOR && > bogus_8259A_irq(vector - FIRST_LEGACY_VECTOR) ) ) > + { > printk("CPU%u: No irq handler for vector %02x (IRQ %d%s)\n", > smp_processor_id(), vector, irq, kind); > + desc = irq_to_desc(~irq); > + if ( ~irq < nr_irqs && irq_desc_initialized(desc) ) > + { > + spin_lock(&desc->lock); > + printk("IRQ%d a=%04lx[%04lx,%04lx] v=%02x[%02x] t=%-15s s=%08x\n",... the %-15s replaced by simply %s (as the output here isn''t in tabular form, yet I copied the stuff from dump_irqs() without paying attention to this aspect. Jan> + ~irq, *cpumask_bits(desc->affinity), > + *cpumask_bits(desc->arch.cpu_mask), > + *cpumask_bits(desc->arch.old_cpu_mask), > + desc->arch.vector, desc->arch.old_vector, > + desc->handler->typename, desc->status); > + spin_unlock(&desc->lock); > + } > + } > TRACE_1D(TRC_HW_IRQ_UNMAPPED_VECTOR, vector); > } > goto out_no_unlock;