Has anyone else encountered this kernel crash? It happens during a benchmark we have made with the following error message: [ 3395.840390] kernel BUG at kernel/locking/rtmutex.c:1014! [ 3395.840396] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP I've attached a capture of the stack trace. For some background, here's our setup. (0) The Variscite DART SD410 SOM (1) Linaro 4.4.9 patched to PREEMPT RT with patch-4.4.9-rt17.patch.gz (2) We are running a modified spidev driver: basically the spidev driver from the standard kernel tree plus an interrupt registered on a input GPIO pin that performs a SPI transfer very similarly to the way the spidev_test user space application does, or at least the kernel routines spidev_test invokes when run. It also toggles an output GPIO pin high at the beginning of the interrupt handler and then low before the interrupt handler exits. We're transferring over the SPI at 50MHz. (3) iperf running at 10 Mbytes/sec over Wi-Fi (4) a benchmarking utility that simulates FFT operations, which persistently keeps the processor running at or near 100% CPU (5) Only the kernel runs in real time. The FFT and iperf run in user space and for our purposes can tolerate delay. Only our modified spidev must run in real time. We are indeed heavily loading the SOM, but we anticipate such to happen when the system we want to build runs at its expected peak. If anyone has some thoughts about whether... (a) anyone has encountered this bug before, (b) we have uncovered a bug, (c) we may have a faulty approach at some level, or (d) we expect too much from this SOM, we'd appreciate any input. Thanks for any help, Brian
Attachment:
screenlog.0.rtmutex_crash
Description: Binary data
