Extend critical section coverage in the low-level interrupt handlers to

include the ithread scheduling step.  Without this, a preemption might
occur in between the interrupt getting masked and the ithread getting
scheduled.  Since the interrupt handler runs in the context of curthread,
the scheudler might see it as having a such a low priority on a busy system
that it doesn't get to run for a _long_ time, leaving the interrupt stranded
in a disabled state.  The only way that the preemption can happen is by
a fast/filter handler triggering a schduling event earlier in the handler,
so this problem can only happen for cases where an interrupt is being
shared by both a fast/filter handler and an ithread handler.  Unfortunately,
it seems to be common for this sharing to happen with network and USB
devices, for example.  This fixes many of the mysterious TCP session
timeouts and NIC watchdogs that were being reported.  Many thanks to Sam
Lefler for getting to the bottom of this problem.

Reviewed by: jhb, jeff, silby

5 files changed, 5 insertions, 5 deletions

diff --git a/sys/amd64/amd64/intr_machdep.c b/sys/amd64/amd64/intr_machdep.c
index 3dc1361..9e151ff 100644
--- a/sys/amd64/amd64/intr_machdep.c
+++ b/sys/amd64/amd64/intr_machdep.c
@@ -390,13 +390,13 @@ intr_execute_handlers(struct intsrc *isrc, struct trapframe *frame)
 		isrc->is_pic->pic_disable_source(isrc, PIC_EOI);
 	else
 		isrc->is_pic->pic_eoi_source(isrc);
-	critical_exit();
 
 	/* Schedule the ithread if needed. */
 	if (thread) {
 		error = intr_event_schedule_thread(ie);
 		KASSERT(error == 0, ("bad stray interrupt"));
 	}
+	critical_exit();
 	td->td_intr_nesting_level--;
 }
 #endif
diff --git a/sys/i386/i386/intr_machdep.c b/sys/i386/i386/intr_machdep.c
index a07ee77..b1d8db9 100644
--- a/sys/i386/i386/intr_machdep.c
+++ b/sys/i386/i386/intr_machdep.c
@@ -391,13 +391,13 @@ intr_execute_handlers(struct intsrc *isrc, struct trapframe *frame)
 		isrc->is_pic->pic_disable_source(isrc, PIC_EOI);
 	else
 		isrc->is_pic->pic_eoi_source(isrc);
-	critical_exit();
 
 	/* Schedule the ithread if needed. */
 	if (thread) {
 		error = intr_event_schedule_thread(ie);
 		KASSERT(error == 0, ("bad stray interrupt"));
 	}
+	critical_exit();
 	td->td_intr_nesting_level--;
 }
 #endif
diff --git a/sys/ia64/ia64/interrupt.c b/sys/ia64/ia64/interrupt.c
index 5460c59..5d30566 100644
--- a/sys/ia64/ia64/interrupt.c
+++ b/sys/ia64/ia64/interrupt.c
@@ -441,7 +441,6 @@ ia64_dispatch_intr(void *frame, u_int vector)
 				thread = 1;
 		}
 	}
-	critical_exit();
 
 	if (thread) {
 		ia64_intr_mask((void *)(uintptr_t)vector);
@@ -449,6 +448,7 @@ ia64_dispatch_intr(void *frame, u_int vector)
 		KASSERT(error == 0, ("%s: impossible stray", __func__));
 	} else
 		ia64_intr_eoi((void *)(uintptr_t)vector);
+	critical_exit();
 #endif
 }
 
diff --git a/sys/powerpc/powerpc/intr_machdep.c b/sys/powerpc/powerpc/intr_machdep.c
index 7daaf23..7653347 100644
--- a/sys/powerpc/powerpc/intr_machdep.c
+++ b/sys/powerpc/powerpc/intr_machdep.c
@@ -280,7 +280,6 @@ powerpc_dispatch_intr(u_int vector, struct trapframe *tf)
 				sched = 1;
 		}
 	}
-	critical_exit();
 
 	if (sched) {
 		PIC_MASK(pic, i->irq);
@@ -289,6 +288,7 @@ powerpc_dispatch_intr(u_int vector, struct trapframe *tf)
 		    __func__));
 	} else
 		PIC_EOI(pic, i->irq);
+	critical_exit();
 #endif
 	return;
 
diff --git a/sys/sparc64/sparc64/intr_machdep.c b/sys/sparc64/sparc64/intr_machdep.c
index aa9a38a..735b4c0 100644
--- a/sys/sparc64/sparc64/intr_machdep.c
+++ b/sys/sparc64/sparc64/intr_machdep.c
@@ -286,7 +286,6 @@ intr_execute_handlers(void *cookie)
 	}
 	if (!thread)
 		intr_enable_eoi(iv);
-	critical_exit();
 
 	/* Schedule a heavyweight interrupt process. */
 	if (thread)
@@ -295,6 +294,7 @@ intr_execute_handlers(void *cookie)
 		error = EINVAL;
 	else
 		error = 0;
+	critical_exit();
 	if (error == EINVAL)
 #else
 	if (intr_event_handle(iv->iv_event, NULL) != 0)