this is the LMDE2 configuration that had problems:
linux /boot/vmlinuz-3.18.13-rt10mah+ root=UUID=34501974-505a-490f-873f-4937e569da86 ro quiet splash isolcpus=2,3 intel_idle.max_cstate=0 processor.max_cstate=0
If you google
intel_idle.max_cstate=0 you will find lots of references to problems with networking, especially wifi.
There is some suggestion that using state
1 is a better option, but I have never bothered with these options and turn off all power management, turn fan speed to full etc in BIOS
I always disable suspend, hibernate etc at kernel level for the same reason (when I build the kernel - not through params)
I wish you had said exactly what you were doing in the beginning, these kernel parameters look like the problem, not the kernel itself.
You don't need brilliant latency to run with mesa cards, 50K is quite good enough.
You can quite often improve latency considerably on a multicore newer computer with larger processor caches, by running a dummy base thread at about 25 - 50us with nothing attached to it. Keeps the cores busy so the rt code remains in one cache.
Try it with
latency-test 25us 1ms to test it out and look at the servo figures
The USB stick seems to be erroring and repeatedly connecting and disconnecting in the dmesg, but not looked at it very much in light of the preceding.
Code:
[ 3518.517107] rtl_usb:_usb_tx_post():<0-0> Urb has error status 0xFFFFFFB9
[ 3518.517152] rtl_usb:_usb_tx_post():<0-0> Urb has error status 0xFFFFFFB9
[ 3518.685108] usb 1-2.3: USB disconnect, device number 4
[ 3518.699454] rtl_usb: reg 0x102, usbctrl_vendorreq TimeOut! status:0xffffffed value=0x83040002
[ 3518.699468] rtl_usb: reg 0x422, usbctrl_vendorreq TimeOut! status:0xffffffed value=0x0
[ 3518.699480] rtl_usb: reg 0x542, usbctrl_vendorreq TimeOut! status:0xffffffed value=0x4d5
[ 3518.699496] rtl8192cu:rtl92cu_get_hw_reg():<0-0> switch case not processed
[ 3518.699531] rtl_usb: reg 0x102, usbctrl_vendorreq TimeOut! status:0xffffffed value=0x3902b2a
Get rid of the kernel params and it should be fine