DPA lib Memory Leak Bug

Infrastructure & Networking DOCA Enterprise Networking
1

Guys, I seem to encounter a memory leak problem when using DOCA DPA lib.

Here is my code:

class DpaKernel {
DpaKernel() {
...
        doca_dpa_create(dev->get_dev(), app, &dpa_ctx, flags);
}

~DpaKernel(){
        if (dpa_ctx != nullptr)
        {
            result= doca_dpa_destroy(dpa_ctx);
        }
}

}

And when I call the dealloc function of my class, the program fail to release flexio heap, the command line shows these error:

heap_destroy 317 - Failed to destroy heap, not all device heap memory was freed
flexio_process_destroy 907 - Failed to release prm heap memory
flexio_process_destroy 914 - Failed to destroy process PRM object (err = 22)

And LeakSanitizer detects the following memory leak stack info:

==1513557==ERROR: LeakSanitizer: detected memory leaks

Direct leak of 48 byte(s) in 1 object(s) allocated from:
    #0 0x7fb37227ea57 in __interceptor_calloc ../../../../src/libsanitizer/asan/asan_malloc_linux.cpp:154
    #1 0x7fb3720eab02  (/lib/x86_64-linux-gnu/libmlx5.so.1+0x68b02)
    #2 0x7fb3721b39fe in flexio_create_prm_process ../libflexio/src/flexio_prm.c:252
    #3 0x7fb3721ac1c3 in process_create ../libflexio/src/flexio.c:689
    #4 0x7fb3721ac1c3 in flexio_process_create ../libflexio/src/flexio.c:845
    #5 0x7fb37218377e  (/opt/mellanox/doca/lib/x86_64-linux-gnu/libdoca_dpa.so.2+0x877e)
    #6 0x7fb37218621d in doca_dpa_create (/opt/mellanox/doca/lib/x86_64-linux-gnu/libdoca_dpa.so.2+0xb21d)
    #7 0x40fdab in dpu::DpaKernel::DpaKernel(std::shared_ptr<dpu::DpaDevice>&, doca_dpa_app*, unsigned int) /home/chengrx/dpa_bench/src/dpa/dpa_kernel.cc:12
    #8 0x421ed2 in test::DpaKernel_Launch_Test::TestBody() /home/chengrx/dpa_bench/test/dpa/test_dpa_kernel.cc:12
    #9 0x4c5397 in void testing::internal::HandleSehExceptionsInMethodIfSupported<testing::Test, void>(testing::Test*, void (testing::Test::*)(), char const*) /home/chengrx/dpa_bench/deps/googletest/googletest/src/gtest.cc:2612
    #10 0x4b2e17 in void testing::internal::HandleExceptionsInMethodIfSupported<testing::Test, void>(testing::Test*, void (testing::Test::*)(), char const*) /home/chengrx/dpa_bench/deps/googletest/googletest/src/gtest.cc:2648
    #11 0x459c25 in testing::Test::Run() /home/chengrx/dpa_bench/deps/googletest/googletest/src/gtest.cc:2687

Direct leak of 40 byte(s) in 1 object(s) allocated from:
    #0 0x7fb37227ea57 in __interceptor_calloc ../../../../src/libsanitizer/asan/asan_malloc_linux.cpp:154
    #1 0x7fb3721a9c96 in heap_block_create ../libflexio/src/flexio_heap.c:55
    #2 0x7fb3721a9c96 in heap_malloc ../libflexio/src/flexio_heap.c:249
    #3 0x7fb3721b5dcf in flexio_buf_dev_alloc ../libflexio/src/flexio_mem.c:134
    #4 0x7fb3721b5dcf in flexio_buf_dev_alloc ../libflexio/src/flexio_mem.c:116
    #5 0x7fb3721b1e42 in internal_msg_stream_create ../libflexio/src/flexio_msg_stream.c:486
    #6 0x7fb3721b2c1d in flexio_msg_stream_create ../libflexio/src/flexio_msg_stream.c:797
    #7 0x7fb3721837e3  (/opt/mellanox/doca/lib/x86_64-linux-gnu/libdoca_dpa.so.2+0x87e3)
    #8 0x7fb37218621d in doca_dpa_create (/opt/mellanox/doca/lib/x86_64-linux-gnu/libdoca_dpa.so.2+0xb21d)
    #9 0x40fdab in dpu::DpaKernel::DpaKernel(std::shared_ptr<dpu::DpaDevice>&, doca_dpa_app*, unsigned int) /home/chengrx/dpa_bench/src/dpa/dpa_kernel.cc:12
    #10 0x421ed2 in test::DpaKernel_Launch_Test::TestBody() /home/chengrx/dpa_bench/test/dpa/test_dpa_kernel.cc:12
    #11 0x4c5397 in void testing::internal::HandleSehExceptionsInMethodIfSupported<testing::Test, void>(testing::Test*, void (testing::Test::*)(), char const*) /home/chengrx/dpa_bench/deps/googletest/googletest/src/gtest.cc:2612
    #12 0x4b2e17 in void testing::internal::HandleExceptionsInMethodIfSupported<testing::Test, void>(testing::Test*, void (testing::Test::*)(), char const*) /home/chengrx/dpa_bench/deps/googletest/googletest/src/gtest.cc:2648

Direct leak of 32 byte(s) in 1 object(s) allocated from:
    #0 0x7fb37227ea57 in __interceptor_calloc ../../../../src/libsanitizer/asan/asan_malloc_linux.cpp:154
    #1 0x7fb3721a99d1 in heap_malloc_from_pool ../libflexio/src/flexio_heap.c:185
    #2 0x7fb3721a9e0b in heap_malloc ../libflexio/src/flexio_heap.c:264
    #3 0x7fb3721b5dcf in flexio_buf_dev_alloc ../libflexio/src/flexio_mem.c:134
    #4 0x7fb3721b5dcf in flexio_buf_dev_alloc ../libflexio/src/flexio_mem.c:116
    #5 0x7fb3721b1e42 in internal_msg_stream_create ../libflexio/src/flexio_msg_stream.c:486
    #6 0x7fb3721b2c1d in flexio_msg_stream_create ../libflexio/src/flexio_msg_stream.c:797
    #7 0x7fb3721837e3  (/opt/mellanox/doca/lib/x86_64-linux-gnu/libdoca_dpa.so.2+0x87e3)
    #8 0x7fb37218621d in doca_dpa_create (/opt/mellanox/doca/lib/x86_64-linux-gnu/libdoca_dpa.so.2+0xb21d)
    #9 0x40fdab in dpu::DpaKernel::DpaKernel(std::shared_ptr<dpu::DpaDevice>&, doca_dpa_app*, unsigned int) /home/chengrx/dpa_bench/src/dpa/dpa_kernel.cc:12
    #10 0x421ed2 in test::DpaKernel_Launch_Test::TestBody() /home/chengrx/dpa_bench/test/dpa/test_dpa_kernel.cc:12
    #11 0x4c5397 in void testing::internal::HandleSehExceptionsInMethodIfSupported<testing::Test, void>(testing::Test*, void (testing::Test::*)(), char const*) /home/chengrx/dpa_bench/deps/googletest/googletest/src/gtest.cc:2612

Indirect leak of 48 byte(s) in 1 object(s) allocated from:
    #0 0x7fb37227ea57 in __interceptor_calloc ../../../../src/libsanitizer/asan/asan_malloc_linux.cpp:154
    #1 0x7fb3720eab02  (/lib/x86_64-linux-gnu/libmlx5.so.1+0x68b02)
    #2 0x7fb3721b3748 in flexio_create_prm_heap_mem ../libflexio/src/flexio_prm.c:179
    #3 0x7fb3721a9d5f in heap_block_create ../libflexio/src/flexio_heap.c:67
    #4 0x7fb3721a9d5f in heap_malloc ../libflexio/src/flexio_heap.c:249
    #5 0x7fb3721b5dcf in flexio_buf_dev_alloc ../libflexio/src/flexio_mem.c:134
    #6 0x7fb3721b5dcf in flexio_buf_dev_alloc ../libflexio/src/flexio_mem.c:116
    #7 0x7fb3721b1e42 in internal_msg_stream_create ../libflexio/src/flexio_msg_stream.c:486
    #8 0x7fb3721b2c1d in flexio_msg_stream_create ../libflexio/src/flexio_msg_stream.c:797
    #9 0x7fb3721837e3  (/opt/mellanox/doca/lib/x86_64-linux-gnu/libdoca_dpa.so.2+0x87e3)
    #10 0x7fb37218621d in doca_dpa_create (/opt/mellanox/doca/lib/x86_64-linux-gnu/libdoca_dpa.so.2+0xb21d)
    #11 0x40fdab in dpu::DpaKernel::DpaKernel(std::shared_ptr<dpu::DpaDevice>&, doca_dpa_app*, unsigned int) /home/chengrx/dpa_bench/src/dpa/dpa_kernel.cc:12
    #12 0x421ed2 in test::DpaKernel_Launch_Test::TestBody() /home/chengrx/dpa_bench/test/dpa/test_dpa_kernel.cc:12

It seems that, the bug is related to how doca_dpa_create and doca_dpa_destroy are implemented, and I have no way to see the codes. May I ask how to fix the memory leak bugs?

Thanks.

2

I think I found the solution, it seems dpa kernel must first release all its events before it actually call doca_dpa_destroy. After I rewrite the order of deallocation as such, the bug is fixed.

3

Thanks for sharing the solution!

4

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.