unexpected crash when upgrade from 2.0 to 2.6.3

[Richard-Feng]

When i upgrade the tcmalloc from 2.0 to 2.6.3, my program crashed just like:

#0 x86_64_fallback_frame_state (context=0x22c3cf620, context=0x22c3cf620, fs=0x22c3cf710)
at ./md-unwind-support.h:58
#1 uw_frame_state_for (context=context@entry=0x22c3cf620, fs=fs@entry=0x22c3cf710)
at ../../../gcc-4.8.2/libgcc/unwind-dw2.c:1253
#2 0x00007f255180cef9 in _Unwind_Backtrace (trace=0x7f2d30284a50 <backtrace_helper>,
trace_argument=0x22c3cf8d0) at ../../../gcc-4.8.2/libgcc/unwind.inc:290
#3 0x00007f2d30284bc6 in backtrace () from /lib64/libc.so.6
#4 0x0000000003c39c77 in Comm::CSignalHandler::HandleCrash (sig=11,
is_exit=, is_hit_dump_limit=0x0)
at comm2/core/utils/iSignalHandler.cpp:376
#5
#6 x86_64_fallback_frame_state (context=0x22c3d1140, context=0x22c3d1140, fs=0x22c3d1230)
at ./md-unwind-support.h:58
#7 uw_frame_state_for (context=context@entry=0x22c3d1140, fs=fs@entry=0x22c3d1230)
at ../../../gcc-7.5.0/libgcc/unwind-dw2.c:1257
#8 0x000000000402e2d8 in _Unwind_Backtrace (
trace=0x2b4dfb0 <libgcc_backtrace_helper(_Unwind_Context*, void*)>,
trace_argument=0x22c3d13f0) at ../../../gcc-7.5.0/libgcc/unwind.inc:290
#9 0x0000000002b4ddca in GetStackTrace_libgcc (result=,
max_depth=, skip_count=)
at src/stacktrace_libgcc-inl.h:100
#10 0x0000000002b4e474 in GetStackTrace (result=result@entry=0x23e3df450,
max_depth=max_depth@entry=30, skip_count=skip_count@entry=3) at src/stacktrace.cc:295
#11 0x0000000002b99dd6 in RecordGrowth (growth=1048576) at src/page_heap.cc:590
#12 tcmalloc::PageHeap::GrowHeap (this=this@entry=0x71871c0 tcmalloc::Static::pageheap_,
n=n@entry=128) at src/page_heap.cc:616
#13 0x0000000002b9a083 in tcmalloc::PageHeap::New (
this=0x71871c0 tcmalloc::Static::pageheap_, n=n@entry=128) at src/page_heap.cc:156
#14 0x00000000018483c4 in (anonymous namespace)::do_malloc_pages (
heap=heap@entry=0x995da00, size=size@entry=1048576) at src/tcmalloc.cc:1327
#15 0x0000000004030556 in do_malloc (size=1048576) at src/tcmalloc.cc:1355

my program runs on x86_64 machine

Would you tell me what's the problem with my program? It works good before this upgrade.
And how to fix ? Thank you !

[ckennelly]

I think you're looking for https://github.com/gperftools/gperftools, not this project.

[Richard-Feng]

excuse me，would you tell me what's the difference between this two?

[vlovich]

gperftools is forked from a much older implementation that was outsourced and has evolved independently. My understanding is that this project represents a more recent snapshot of the internal implementation at Google (& not just a fork - it's actively synchronized).

See: https://abseil.io/blog/20200212-tcmalloc and https://google.github.io/tcmalloc/overview.html

The TLDR is it's a much more modern design (i.e. faster, better memory utilization) than the old gperftools implementation & more inline with the direction taken by mimalloc & jemalloc. Highlights include per-cpu caches that use restartable sequences (on Linux anyway) which offer better performance than the old thread local pool implementation (which is still available for other platforms).

(I'm in no way associated with the project - just my understanding of the differences).