[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Xen-devel] [PATCH] x86/irq: Limit the maximum number of domain PIRQs



c/s 7e73a6e "have architectures specify the number of PIRQs a hardware domain
gets" increased the default number of pirqs for dom0, as 256 was found to be
too low in some cases.

However, it didn't account for the upper bound presented by the domains EOI
bitmap, registered with the PHYSDEVOP_pirq_eoi_gmfn_v* hypercall.

On a server with 240 cpus, Xen was observed to be attempting to clear the EOI
bit for dom0's pirq 0xb40f, which hit a pagefault.

Signed-off-by: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
CC: Keir Fraser <keir@xxxxxxx>
CC: Jan Beulich <JBeulich@xxxxxxxx>

---

Sadly, the aformentioned server has gone on to develop a thermal failure and
doesn't currently boot, so we are unable to actually test this fix.  All
analysis has come from logs.
---
 xen/arch/x86/io_apic.c |    4 ++++
 xen/arch/x86/irq.c     |    6 ++++++
 2 files changed, 10 insertions(+)

diff --git a/xen/arch/x86/io_apic.c b/xen/arch/x86/io_apic.c
index 976c0c5..6f66562 100644
--- a/xen/arch/x86/io_apic.c
+++ b/xen/arch/x86/io_apic.c
@@ -2614,6 +2614,10 @@ unsigned int arch_hwdom_irqs(domid_t domid)
     if ( !domid )
         n = min(n, dom0_max_vcpus());
     n = min(nr_irqs_gsi + n * NR_DYNAMIC_VECTORS, nr_irqs);
+
+    /* Bounded by the domain pirq eoi bitmap gfn. */
+    n = min_t(unsigned int, n, PAGE_SIZE * BITS_PER_BYTE);
+
     printk("Dom%d has maximum %u PIRQs\n", domid, n);
 
     return n;
diff --git a/xen/arch/x86/irq.c b/xen/arch/x86/irq.c
index 1a21917..fe1ee33 100644
--- a/xen/arch/x86/irq.c
+++ b/xen/arch/x86/irq.c
@@ -1073,13 +1073,19 @@ bool_t cpu_has_pending_apic_eoi(void)
 static inline void set_pirq_eoi(struct domain *d, unsigned int irq)
 {
     if ( d->arch.pirq_eoi_map )
+    {
+        ASSERT(irq < PAGE_SIZE * BITS_PER_BYTE);
         set_bit(irq, d->arch.pirq_eoi_map);
+    }
 }
 
 static inline void clear_pirq_eoi(struct domain *d, unsigned int irq)
 {
     if ( d->arch.pirq_eoi_map )
+    {
+        ASSERT(irq < PAGE_SIZE * BITS_PER_BYTE);
         clear_bit(irq, d->arch.pirq_eoi_map);
+    }
 }
 
 static void set_eoi_ready(void *data);
-- 
1.7.10.4


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.