Segmentation Fault using PlutoSDR on Ubuntu 20.04.6

[acochrane]

This appears to be related to issue #749 as a segfault for use of plutosdr, but based on the backtrace it seems like a different issue.

Using the supplied conf/RealTime_input/gnss-sdr_GPS_L1_plutosdr_realtime.conf after setting the correct address, I receive a segfault.

$ gnss-sdr --config_file=gnss-sdr_GPS_L1_plutosdr_realtime_new.conf
Initializing GNSS-SDR v0.0.19.git-next-fbca34a4b ... Please wait.
Logging will be written at "/tmp"
Use gnss-sdr --log_dir=/path/to/log to change that.
device address: 192.168.55.1
frequency : 1575420000 Hz
sample rate: 4000000 Sps
gain mode: slow_attack
item type: gr_complex
RF Channels: 1
Starting a TCP/IP server of RTCM messages on port 2101
The TCP/IP server of RTCM messages is up and running. Accepting connections ...
Segmentation fault (core dumped)

Collecting backtrace with gdb

[Switching to Thread 0x7fff9f7fe700 (LWP 1238238)]
0x00007ffff7000526 in _mm_loadu_si128 (__P=0x31)
at /usr/lib/gcc/x86_64-linux-gnu/9/include/emmintrin.h:703
703 return *__P;
(gdb) bt
#0 0x00007ffff7000526 in _mm_loadu_si128 (__P=0x31)
at /usr/lib/gcc/x86_64-linux-gnu/9/include/emmintrin.h:703
#1 volk_16i_s32f_convert_32f_u_avx2 (outputVector=0x555556f7ba80, inputVector=0x31,
scalar=2048, num_points=4096)
at /home/acochrane/src/volk/kernels/volk/volk_16i_s32f_convert_32f.h:68
#2 0x00007ffff6e11ba3 in __volk_16i_s32f_convert_32f_d (outputVector=0x555556f7ba80,
inputVector=0x31, scalar=2048, num_points=4096)
at /home/acochrane/src/volk/build/lib/volk.c:630
#3 0x00007ffff7cbdef7 in gr::iio::fmcomms2_source_impl<std::complex >::work (
this=0x555556f576b0, noutput_items=4096,
input_items=std::vector of length 0, capacity 0,
output_items=std::vector of length 1, capacity 1 = {...})
at /usr/include/c++/9/bits/stl_vector.h:1040
#4 0x00007ffff7a7579b in gr::sync_block::general_work (this=0x555556f57898,
noutput_items=, ninput_items=..., input_items=..., output_items=...)
at /home/acochrane/src/gnuradio/gnuradio-runtime/lib/sync_block.cc:49
#5 0x00007ffff7a1f771 in gr::block_executor::run_one_iteration (
this=this@entry=0x7fff9f7df010)
at /home/acochrane/src/gnuradio/gnuradio-runtime/lib/block_executor.cc:622
#6 0x00007ffff7a8825b in gr::tpb_thread_body::tpb_thread_body (this=,
block=..., start_sync=..., max_noutput_items=)
at /home/acochrane/src/gnuradio/gnuradio-runtime/lib/tpb_thread_body.cc:94
#7 0x00007ffff7a73a0c in gr::tpb_container::operator() (this=this@entry=0x55555bc708c0)
at /usr/include/c++/9/ext/atomicity.h:96
#8 0x00007ffff7a74977 in gr::thread::thread_body_wrappergr::tpb_container::operator() (
this=0x55555bc708c0)
at /home/acochrane/src/gnuradio/gnuradio-runtime/lib/../include/gnuradio/thread/thread_body_wrapper.h:43
#9 std::_Function_handler<void (), gr::thread::thread_body_wrappergr::tpb_container >::_M_invoke(std::_Any_data const&) (__functor=...) at /usr/include/c++/9/bits/std_function.h:300
#10 0x00007ffff7f4843b in ?? () from /lib/x86_64-linux-gnu/libboost_thread.so.1.71.0
#11 0x00007ffff4964609 in start_thread (arg=) at pthread_create.c:477
#12 0x00007ffff4889353 in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95

I could be reading this wrong, but it looks like the fmcomms2_source_impl is sending a bad address for the input vector to volk16_16i_s32f_d.

Any hints?

[acochrane]

After recompiling libiio and gnuradio in Debug build types, I get better info from the backtrace

0x00007ffff7c746d7 in iio_channel_is_enabled (chn=0x555556f489f0)
at /home/acochrane/src/libiio/channel.c:446
446 return chn->index >= 0 && chn->dev->mask &&
(gdb) bt
#0 0x00007ffff7c746d7 in iio_channel_is_enabled (chn=0x555556f489f0)
at /home/acochrane/src/libiio/channel.c:446
#1 0x00007ffff7c79f20 in iio_buffer_first (buffer=0x7fff98001040, chn=0x555556f489f0)
at /home/acochrane/src/libiio/buffer.c:258
#2 0x00007ffff7cdae09 in gr::iio::device_source_impl::channel_read (
this=this@entry=0x555556f14198, chn=0x555556f489f0, dst=0x21, len=len@entry=8192)
at /home/acochrane/src/gnuradio/gr-iio/lib/device_source_impl.cc:260
#3 0x00007ffff7cdb61e in gr::iio::device_source_impl::work (
this=this@entry=0x555556f14198, noutput_items=noutput_items@entry=4096, input_items=
std::vector of length 0, capacity 0,
output_items=std::vector of length 2, capacity 2 = {...})
at /usr/include/c++/9/bits/stl_vector.h:1040
#4 0x00007ffff7ce7e7d in gr::iio::fmcomms2_source_impl<std::complex >::work (
this=0x555556f14190, noutput_items=4096,
input_items=std::vector of length 0, capacity 0,
output_items=std::vector of length 1, capacity 1 = {...})
at /home/acochrane/src/gnuradio/gr-iio/lib/fmcomms2_source_impl.cc:230
#5 0x00007ffff7a7579b in gr::sync_block::general_work (this=0x555556f14378,
noutput_items=, ninput_items=..., input_items=..., output_items=...)
at /home/acochrane/src/gnuradio/gnuradio-runtime/lib/sync_block.cc:49
#6 0x00007ffff7a1f771 in gr::block_executor::run_one_iteration (
this=this@entry=0x7fffa77df010)
at /home/acochrane/src/gnuradio/gnuradio-runtime/lib/block_executor.cc:622
#7 0x00007ffff7a8825b in gr::tpb_thread_body::tpb_thread_body (this=,
block=..., start_sync=..., max_noutput_items=)
at /home/acochrane/src/gnuradio/gnuradio-runtime/lib/tpb_thread_body.cc:94
#8 0x00007ffff7a73a0c in gr::tpb_container::operator() (this=this@entry=0x55555bc2cd60)
at /usr/include/c++/9/ext/atomicity.h:96
#9 0x00007ffff7a74977 in gr::thread::thread_body_wrappergr::tpb_container::operator() (
this=0x55555bc2cd60)
at /home/acochrane/src/gnuradio/gnuradio-runtime/lib/../include/gnuradio/thread/thread_body_wrapper.h:43
#10 std::_Function_handler<void (), gr::thread::thread_body_wrappergr::tpb_container >::_M_invoke(std::_Any_data const&) (__functor=...) at /usr/include/c++/9/bits/std_function.h:300
#11 0x00007ffff7f4843b in ?? () from /lib/x86_64-linux-gnu/libboost_thread.so.1.71.0
#12 0x00007ffff4964609 in start_thread (arg=) at pthread_create.c:477
#13 0x00007ffff4889353 in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95

It looks like gr::iio::device_source_impl::channel_read might be passed a bad address for dst.

[acochrane]

Well based this snippet from gnuradio/gr-iio/lib/device_source_impl.cc:309

// Process samples
unsigned long items = std::min(items_in_buffer, (unsigned long)noutput_items);

for (size_t i = 0; i < output_items.size(); i++)
    channel_read(channel_list[i], output_items[i], items * sizeof(short));

items_in_buffer -= items;
byte_offset += items * iio_buffer_step(buf); 

It looks like that third argument to channel_read(), dst should be a small hex value, counting the smaller of the items_in_buffer or noutput_items.