I’m trying to communicate between TK1 based custom board and NXP processor over PCIe link. TK1 board is running L4T R21.7 and the kernel is able to successfully enumerate the NXP processor as an endpoint device and properly assigns BARs. I have a C application with uses /dev/mem & mmap() to communicate with endpoint using the BARs. The application starts and communicates properly with endpoint until suddenly a flood of following errors fill up the console:
[ 66.601782] MSELECT error detected! status=0x4
[ 66.606506] MSELECT error detected! status=0x4
[ 66.611228] MSELECT error detected! status=0x4
[ 66.615952] MSELECT error detected! status=0x4
[ 66.620675] MSELECT error detected! status=0x4
[ 66.625398] MSELECT error detected! status=0x4
[ 66.630121] MSELECT error detected! status=0x4
[ 66.634845] MSELECT error detected! status=0x4
[ 66.639567] MSELECT error detected! status=0x4
[ 66.644290] MSELECT error detected! status=0x4
A similar issue was posted here https://devtalk.nvidia.com/default/topic/929521/jetson-tk1/jetson-tk1-what-the-quot-mselect-error-quot-meanings-how-to-debug-it-tk1-crashed-when-working-w/ but the patch provided there is already applied in the kernel I’m running. Can someone suggest why this might be occurring and some possible solution to it?
P.S: These errors sometimes fill up the console when TK1 is booting and data communication with PCIe endpoint isn’t started yet.