[gcServer=1] Assert failure ephemeral_heap_segment->saved_committed == heap_segment_committed

[xiangzhai]

Hi,

Description

Testcase: GC/API/NoGCRegion/NoGC/NoGC.exe

export CORE_LIBRARIES=/home/zhaixiang/runtime/.dotnet/shared/Microsoft.NETCore.App/5.0.0-preview.8.20361.2
export COMPlus_gcServer=1

gdb -ex=r --args /home/zhaixiang/runtime/artifacts/bin/coreclr/Linux.arm64.Debug/corerun /home/zhaixiang/coreclr-mips64-dev/bin/tests/Linux.arm64.Debug/GC/API/NoGCRegion/NoGC/NoGC.exe



Assert failure(PID 6927 [0x00001b0f], Thread: 6936 [0x1b18]): ephemeral_heap_segment->saved_committed == heap_segment_committed (ephemeral_heap_segment)
    File: /home/zhaixiang/runtime/src/coreclr/src/gc/gc.cpp Line: 32564
    Image: /home/zhaixiang/runtime/artifacts/bin/coreclr/Linux.arm64.Debug/corerun


Thread 8 ".NET Server GC" received signal SIGTRAP, Trace/breakpoint trap.
[Switching to Thread 0x7faf8191e0 (LWP 6936)]
DBG_DebugBreak () at /home/zhaixiang/runtime/src/coreclr/src/pal/src/arch/arm64/debugbreak.S:7
7	    EMIT_BREAKPOINT
(gdb) bt
#0  DBG_DebugBreak () at /home/zhaixiang/runtime/src/coreclr/src/pal/src/arch/arm64/debugbreak.S:7
#1  0x0000007fb770cd6c in DebugBreak () at /home/zhaixiang/runtime/src/coreclr/src/pal/src/debug/debug.cpp:410
#2  0x0000007fb750bd90 in DbgAssertDialog (szFile=0x7fb7879af0 "/home/zhaixiang/runtime/src/coreclr/src/gc/gc.cpp", iLine=32564, szExpr=0x7fb787d0c7 "ephemeral_heap_segment->saved_committed == heap_segment_committed (ephemeral_heap_segment)") at /home/zhaixiang/runtime/src/coreclr/src/utilcode/debug.cpp:697
#3  0x0000007fb744315c in SVR::gc_heap::decommit_ephemeral_segment_pages_step (this=0x5555637900) at /home/zhaixiang/runtime/src/coreclr/src/gc/gc.cpp:32564
#4  0x0000007fb73eaf98 in SVR::gc_heap::decommit_step () at /home/zhaixiang/runtime/src/coreclr/src/gc/gc.cpp:32546
#5  0x0000007fb73eaac8 in SVR::gc_heap::gc_thread_function (this=0x555561fb60) at /home/zhaixiang/runtime/src/coreclr/src/gc/gc.cpp:5696
#6  0x0000007fb73ea9a0 in SVR::gc_heap::gc_thread_stub (arg=0x555561fb60) at /home/zhaixiang/runtime/src/coreclr/src/gc/gc.cpp:26139
#7  0x0000007fb7267c28 in (anonymous namespace)::CreateNonSuspendableThread(void (*)(void*), void*, char16_t const*)::$_1::operator()(void*) const (this=0x0, argument=0x7fffffdbe8) at /home/zhaixiang/runtime/src/coreclr/src/vm/gcenv.ee.cpp:1432
#8  0x0000007fb7267b68 in (anonymous namespace)::CreateNonSuspendableThread(void (*)(void*), void*, char16_t const*)::$_1::__invoke(void*) (argument=0x7fffffdbe8) at /home/zhaixiang/runtime/src/coreclr/src/vm/gcenv.ee.cpp:1417
#9  0x0000007fb77a5298 in CorUnix::CPalThread::ThreadEntry (pvParam=0x55556256b0) at /home/zhaixiang/runtime/src/coreclr/src/pal/src/thread/thread.cpp:1845
#10 0x0000007fb7f7ffc4 in start_thread () from /lib/aarch64-linux-gnu/libpthread.so.0
#11 0x0000007fb7c9b290 in ?? () from /lib/aarch64-linux-gnu/libc.so.6
(gdb) x/22i 0x0000007fb7c9b290-44
   0x7fb7c9b264:	tbnz	w11, #16, 0x7fb7c9b288
   0x7fb7c9b268:	mov	x0, #0xffffffffffffffff    	// #-1
   0x7fb7c9b26c:	tbnz	w11, #8, 0x7fb7c9b278
   0x7fb7c9b270:	mov	x8, #0xac                  	// #172
   0x7fb7c9b274:	svc	#0x0
   0x7fb7c9b278:	mrs	x1, tpidr_el0
   0x7fb7c9b27c:	sub	x1, x1, #0x6f0
   0x7fb7c9b280:	str	w0, [x1,#212]
   0x7fb7c9b284:	str	w0, [x1,#208]
   0x7fb7c9b288:	mov	x0, x12
   0x7fb7c9b28c:	blr	x10
   0x7fb7c9b290:	b	0x7fb7c70128 <_exit>
   0x7fb7c9b294:	.inst	0x00000000 ; undefined
   0x7fb7c9b298:	.inst	0x00000000 ; undefined
   0x7fb7c9b29c:	.inst	0x00000000 ; undefined
   0x7fb7c9b2a0 <lseek64>:	mov	x8, #0x3e                  	// #62
   0x7fb7c9b2a4 <lseek64+4>:	svc	#0x0
   0x7fb7c9b2a8 <lseek64+8>:	cmn	x0, #0xfff
   0x7fb7c9b2ac <lseek64+12>:	b.cs	0x7fb7c9b2b4 <lseek64+20>
   0x7fb7c9b2b0 <lseek64+16>:	ret
   0x7fb7c9b2b4 <lseek64+20>:	b	0x7fb7bf4990
   0x7fb7c9b2b8 <umount>:	mov	w1, #0x0                   	// #0

Configuration

• runtime master
• Ubuntu 16.04.3 LTS
• ARM64

Thanks,
Leslie Zhai

Tagging subscribers to this area: @dotnet/gc
See info in area-owners.md if you want to be subscribed.

[mangod9]

I can repro this on linux x64 as well, but not on windows.

[xiangzhai]

Hi @mangod9

Thanks for your test!

Leslie Zhai

[xiangzhai]

Hi @mangod9

Could you paste the /proc/meminfo for your Linux X64 test machine?

The /proc/meminfo for Linux ARM64 test machine:

MemTotal:        4034620 kB
MemFree:          716340 kB
MemAvailable:    3483080 kB
Buffers:           85536 kB
Cached:          2547680 kB
SwapCached:        22128 kB

And:

$ free -lh
              total        used        free      shared  buff/cache   available
Mem:           3.8G        469M        699M        5.0M        2.7G        3.3G
Low:           3.8G        3.2G        699M
High:            0B          0B          0B
Swap:          8.0G        910M        7.1G

Workaround:

diff --git a/src/tests/GC/API/NoGCRegion/NoGC.cs b/src/tests/GC/API/NoGCRegion/NoGC.cs
index 4bacf75..ac5c5e5 100644
--- a/src/tests/GC/API/NoGCRegion/NoGC.cs
+++ b/src/tests/GC/API/NoGCRegion/NoGC.cs
@@ -156,7 +156,7 @@ public static int Main(string[] args)
             {
                 lowMB = 200;
                 midMB = lowMB * 2;
-                highMB = 1024*numberProcs;
+                highMB = 128*numberProcs;
             }
             else
             {

Thanks,
Leslie Zhai

[PeterSolMS]

Actually, I got it to repro on Win x64 as well. So it looks like a general problem with decommitting and no GC regions interacting.

[xiangzhai]

Hi @PeterSolMS

Thanks for your test!

What about the equivalent /proc/meminfo for your Windows X64 test machine?

Leslie Zhai

[PeterSolMS]

It's machine with 128 GB of memory, so the issue doesn't look like it's related to a low memory situation.

[xiangzhai]

I see.