Commit ef3e28c5 authored by Eric W. Biederman's avatar Eric W. Biederman Committed by Linus Torvalds

x86_64: check remote IRR bit before migrating level triggered irq

On x86_64 kernel, level triggered irq migration gets initiated in the
context of that interrupt(after executing the irq handler) and following
steps are followed to do the irq migration.

1. mask IOAPIC RTE entry;     // write to IOAPIC RTE
2. EOI;                       // processor EOI write
3. reprogram IOAPIC RTE entry // write to IOAPIC RTE with new destination and
                              // and interrupt vector due to per cpu vector
                              // allocation.
4. unmask IOAPIC RTE entry;   // write to IOAPIC RTE

Because of the per cpu vector allocation in x86_64 kernels, when the irq
migrates to a different cpu, new vector(corresponding to the new cpu) will
get allocated.

An EOI write to local APIC has a side effect of generating an EOI write for
level trigger interrupts (normally this is a broadcast to all IOAPICs).
The EOI broadcast generated as a side effect of EOI write to processor may
be delayed while the other IOAPIC writes (step 3 and 4) can go through.

Normally, the EOI generated by local APIC for level trigger interrupt
contains vector number.  The IOAPIC will take this vector number and search
the IOAPIC RTE entries for an entry with matching vector number and clear
the remote IRR bit (indicate EOI).  However, if the vector number is
changed (as in step 3) the IOAPIC will not find the RTE entry when the EOI
is received later.  This will cause the remote IRR to get stuck causing the
interrupt hang (no more interrupt from this RTE).

Current x86_64 kernel assumes that remote IRR bit is cleared by the time
IOAPIC RTE is reprogrammed.  Fix this assumption by checking for remote IRR
bit and if it still set, delay the irq migration to the next interrupt
arrival event(hopefully, next time remote IRR bit will get cleared before
the IOAPIC RTE is reprogrammed).

Initial analysis and patch from Nanhai.

Clean up patch from Suresh.

Rewritten to be less intrusive, and to contain a big fat comment by Eric.

[akpm@linux-foundation.org: fix comments]
Acked-by: default avatarIngo Molnar <mingo@elte.hu>
Cc: Nanhai Zou <nanhai.zou@intel.com>
Acked-by: default avatarSuresh Siddha <suresh.b.siddha@intel.com>
Cc: Asit Mallick <asit.k.mallick@intel.com>
Cc: Keith Packard <keith.packard@intel.com>
Signed-off-by: default avatarEric W. Biederman <ebiederm@xmission.com>
Signed-off-by: default avatarAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: default avatarAndi Kleen <ak@suse.de>
Signed-off-by: default avatarLinus Torvalds <torvalds@linux-foundation.org>
parent 22293e58
...@@ -152,6 +152,32 @@ static inline void io_apic_modify(unsigned int apic, unsigned int value) ...@@ -152,6 +152,32 @@ static inline void io_apic_modify(unsigned int apic, unsigned int value)
writel(value, &io_apic->data); writel(value, &io_apic->data);
} }
static int io_apic_level_ack_pending(unsigned int irq)
{
struct irq_pin_list *entry;
unsigned long flags;
int pending = 0;
spin_lock_irqsave(&ioapic_lock, flags);
entry = irq_2_pin + irq;
for (;;) {
unsigned int reg;
int pin;
pin = entry->pin;
if (pin == -1)
break;
reg = io_apic_read(entry->apic, 0x10 + pin*2);
/* Is the remote IRR bit set? */
pending |= (reg >> 14) & 1;
if (!entry->next)
break;
entry = irq_2_pin + entry->next;
}
spin_unlock_irqrestore(&ioapic_lock, flags);
return pending;
}
/* /*
* Synchronize the IO-APIC and the CPU by doing * Synchronize the IO-APIC and the CPU by doing
* a dummy read from the IO-APIC * a dummy read from the IO-APIC
...@@ -1418,9 +1444,37 @@ static void ack_apic_level(unsigned int irq) ...@@ -1418,9 +1444,37 @@ static void ack_apic_level(unsigned int irq)
ack_APIC_irq(); ack_APIC_irq();
/* Now we can move and renable the irq */ /* Now we can move and renable the irq */
move_masked_irq(irq); if (unlikely(do_unmask_irq)) {
if (unlikely(do_unmask_irq)) /* Only migrate the irq if the ack has been received.
*
* On rare occasions the broadcast level triggered ack gets
* delayed going to ioapics, and if we reprogram the
* vector while Remote IRR is still set the irq will never
* fire again.
*
* To prevent this scenario we read the Remote IRR bit
* of the ioapic. This has two effects.
* - On any sane system the read of the ioapic will
* flush writes (and acks) going to the ioapic from
* this cpu.
* - We get to see if the ACK has actually been delivered.
*
* Based on failed experiments of reprogramming the
* ioapic entry from outside of irq context starting
* with masking the ioapic entry and then polling until
* Remote IRR was clear before reprogramming the
* ioapic I don't trust the Remote IRR bit to be
* completey accurate.
*
* However there appears to be no other way to plug
* this race, so if the Remote IRR bit is not
* accurate and is causing problems then it is a hardware bug
* and you can go talk to the chipset vendor about it.
*/
if (!io_apic_level_ack_pending(irq))
move_masked_irq(irq);
unmask_IO_APIC_irq(irq); unmask_IO_APIC_irq(irq);
}
} }
static struct irq_chip ioapic_chip __read_mostly = { static struct irq_chip ioapic_chip __read_mostly = {
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment