#3320: Parallel program crashes using GHC 6.11 under OS X
-------------------------------+--------------------------------------------
Reporter: sebf | Owner: simonmar
Type: bug | Status: new
Priority: high | Milestone: 6.12.1
Component: Runtime System | Version: 6.11
Severity: normal | Resolution:
Keywords: | Difficulty: Unknown
Testcase: | Os: MacOS X
Architecture: x86 |
-------------------------------+--------------------------------------------
Comment (by igloo):
Here's the end of a run with a slightly instrumented RTS:
{{{
RELEASE_LOCK(0x801330) rts/Capability.c 644
Calling yieldCapability 2 with 0x801280
Called yieldCapability with 0x801280
RELEASE_LOCK(0x7009f8) rts/Capability.c 618
ACQUIRE_LOCK(0x8011f0) rts/Capability.c 622
RELEASE_LOCK(0x8011f0) rts/Capability.c 626
ACQUIRE_LOCK(0x7009f8) rts/Capability.c 613
RELEASE_LOCK(0x801330) rts/Capability.c 842
Calling yieldCapability 1 with 0x801140
Called yieldCapability with 0x801140
ACQUIRE_LOCK(0x2ca7e0) rts/Schedule.c 906
RELEASE_LOCK(0x2ca7e0) rts/Schedule.c 916
Garbage collecting
ACQUIRE_LOCK(0x2cb8a0) rts/sm/GC.c 199
ACQUIRE_LOCK(0x2ca2a0) rts/Stable.c 325
RELEASE_LOCK(0x2cb8a0) rts/sm/GC.c 755
ACQUIRE_LOCK(0x2cb8a0) rts/sm/GC.c 757
RELEASE_LOCK(0x2cb8a0) rts/sm/GC.c 760
ACQUIRE_LOCK(0x2cb8a0) rts/sm/GC.c 763
RELEASE_LOCK(0x2ca2a0) rts/Stable.c 331
RELEASE_LOCK(0x2cb8a0) rts/sm/GC.c 812
Finished garbage collecting
ACQUIRE_LOCK(0x8011f0) rts/Capability.c 534
RELEASE_LOCK(0x8011f0) rts/Capability.c 545
Calling yieldCapability 2 with 0x801140
ACQUIRE_LOCK(0x700ac8) rts/Capability.c 548
Calling yieldCapability 2 with 0x801280
Called yieldCapability with 0x801140
Called yieldCapability with 0xb0103000
Calling releaseCapabilityAndQueueWorker with 0x801140
Calling releaseCapabilityAndQueueWorker with 0xb0103000
ACQUIRE_LOCK(0x8011f0) rts/Capability.c 457
ACQUIRE_LOCK(0xb01030b0) rts/Capability.c 457
ACQUIRE_LOCK(0x700ac8) rts/Capability.c 350
Task is NULL! 0xb0103000 0xb01030b0
RELEASE_LOCK(0x700ac8) rts/Capability.c 356
RELEASE_LOCK(0x8011f0) rts/Capability.c 481
RELEASE_LOCK(0x700ac8) rts/Capability.c 553
ACQUIRE_LOCK(0x700be8) rts/Capability.c 613
ACQUIRE_LOCK(0x8011f0) rts/Capability.c 556
RELEASE_LOCK(0x8011f0) rts/Capability.c 566
Bus error
}}}
The
{{{
Garbage collecting
[...]
Finished garbage collecting
}}}
are the start and end of `rts/sm/GC.c:GarbageCollect`, so it looks to me
like we aren't supposed to be GCing when things go wrong.
In particular, see these two lines:
{{{
Calling yieldCapability 2 with 0x801280
[...]
Called yieldCapability with 0xb0103000
}}}
These are:
{{{
if (prev_pending_gc) {
do {
debugTrace(DEBUG_sched, "someone else is trying to GC
(%d)...",
prev_pending_gc);
ASSERT(cap);
printf("Calling yieldCapability 2 with %p\n", cap);
yieldCapability(&cap,task);
} while (waiting_for_gc);
return cap; // NOTE: task->cap might have changed here
}
}}}
and
{{{
void
yieldCapability (Capability** pCap, Task *task)
{
Capability *cap = *pCap;
printf("Called yieldCapability with %p\n", cap);
}}}
The
{{{
Task is NULL! 0xb0103000 0xb01030b0
}}}
is the point where the bus error actually happens:
{{{
static void
releaseCapabilityAndQueueWorker (Capability* cap USED_IF_THREADS)
{
Task *task;
ACQUIRE_LOCK(&cap->lock);
task = cap->running_task;
if (task == NULL) {
printf("Task is NULL! %p %p\n", cap, &cap->lock);
fflush(stdout);
}
[...]
if (!isBoundTask(task) && !task->stopped && !task->suspended_tso) {
}}}
which looks like a symptom of the corrupt cap.
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/3320#comment:6>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler_______________________________________________
Glasgow-haskell-bugs mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
