The current code overflows (s * 1000000000) for s >= 5 but that is
e.g. used in
etna_bo_cpu_prep(). Alternatively we could reject larger
etna_pipe_wait() and reduce the value in
Freedreno shares has an identical
get_abs_timeout() and uses the same timeout in
msm_bo_cpu_prep() so it should be affected too.
If that makes sense i'll send a patch for libdrm too.