[ARROW-5186] [Plasma] Crash on deleting CUDA memory

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.13.0
• Fix Version/s: 0.14.0
• Component/s: C++
• Labels: pull-request-available
• Flags: Patch
• External issue URL: https://github.com/apache/arrow/issues/21663
• Assignee: shengjun.li
• Reporter: shengjun.li
• Created: 19/Apr/19 04:03
• Updated: 11/Jan/23 07:38
• Resolved: 29/Apr/19 21:58

Description

cpp/CMakeLists.txt
  option(ARROW_CUDA "Build the Arrow CUDA extensions (requires CUDA toolkit)" ON)
  option(ARROW_PLASMA "Build the plasma object store along with Arrow" ON)

[sample sequence]
(1) call PlasmaClient::Create(id_object, data_size, 0, 0, &buff, 1) // where device_num != 0
(2) call PlasmaClient::Seal(id_object)
(3) call PlasmaClient::Release(id_object)
(4) call PlasmaClient::Delete(id_object) // server carsh!

• • • Aborted at 1555645923 (unix time) try "date -d @1555645923" if you are using GNU date ***
      PC: @     0x7f65bcfa1428 gsignal
    • SIGABRT (@0x3e800006d67) received by PID 28007 (TID 0x7f65bf225740) from PID 28007; stack trace: ***
          @     0x7f65bd347390 (unknown)
          @     0x7f65bcfa1428 gsignal
          @     0x7f65bcfa302a abort
          @           0x4a56cd dlfree
          @           0x4b4bc2 plasma::PlasmaAllocator::Free()
          @           0x4b7da3 plasma::PlasmaStore::EraseFromObjectTable()
          @           0x4b87d2 plasma::PlasmaStore::DeleteObject()
          @           0x4bb3d2 plasma::PlasmaStore::ProcessMessage()
          @           0x4b9195 _ZZN6plasma11PlasmaStore13ConnectClientEiENKUliE_clEi
          @           0x4bd752 _ZNSt17_Function_handlerIFviEZN6plasma11PlasmaStore13ConnectClientEiEUliE_E9_M_invokeERKSt9_Any_dataOi
          @           0x4ab998 std::function<>::operator()()
          @           0x4aaea7 plasma::EventLoop::FileEventCallback()
          @           0x4dbd8f aeProcessEvents
          @           0x4dbf50 aeMain
          @           0x4ab19b plasma::EventLoop::Start()
          @           0x4bfc93 plasma::PlasmaStoreRunner::Start()
          @           0x4bc34d plasma::StartServer()
          @           0x4bcfbd main
          @     0x7f65bcf8c830 __libc_start_main
          @           0x49e939 _start
          @                0x0 (unknown)
      Aborted (core dumped)

Issue Links

• links to: GitHub Pull Request #4175
• links to: GitHub Pull Request #4177