Ok, that's normal.
But the question is does linux reconfigures GPIO before going to sleep?
Because if no, then it should be possible to attach interrupt from GPIO to M4 core and release it when A-core wakes up, so linux could use its own GPIO interrupt handler on that line.
But what i'm seeing right now is that M4 is running, but interrupts are fired only in a time between i've ordered linux to go to sleep, and when it surely sleeps. I configure GPIO interrups in M4 just after i get messages about sleeping A core via rpmsg and i get a few interrupts, and when A core finally finishes sleep preparation and sleeps, i have no more interrupts on M4.
So that's why i'm thinking someting in Linux is reconfiguring GPIO before it sleeps.