wowlan: document theory on wake failure

modules/wowlan.nix

@@ -56,6 +56,10 @@
 #       - CONFIG_ROAMING_FLAG = 0x3
 # - may fail to wake on LAN, even without any signature like the above
 #   - observed after a suspension of 10 minutes, trying to contact from laptop (laptop would have previously not contacted moby)
+#   - in many cases the wake reason is still 0x23, even as it sleeps for the whole alloted 300s.
+#     - then it's likely a race condition with the WiFi chip raising the wake signal *as* the CPU is falling asleep, and the CPU misses it.
+#     - this is testable: phone would wake 100% if all networked services are disabled (so it's not receiving packets near the time it enters sleep)
+#     - TODO: change the wake event from edge-triggered to level-triggered.
 # - may disassociate from WiFi and *refuse to reassociate even via nmtui, restarting NetworkManager/wpa_supplicant/polkit*
 #   - solution is `modprobe -r 8723cs && modprobe 8723cs`, then way a couple minutes.
 #   - possible this only happens when the driver is compiled with CONFIG_GTK_OL (or CONFIG_ARP_KEEP_ALIVE).

pkgs/additional/linux-megous/default.nix

@@ -159,7 +159,8 @@ let
       # this patch just uncomments some commented-out #defines.
       # they were commented out from the first time megi imported this driver, never touched.
       # this may be problematic, occasionally requiring a driver re-probe. see other notes on wowlan/8723cs.
-      # if continued errors, try removing the CONFIG_GTK_OL and keeping just CONFIG_ARP_KEEP_ALIVE
+      # if continued errors, try removing the CONFIG_GTK_OL and keeping just CONFIG_ARP_KEEP_ALIVE.
+      # seems to infrequently disrupt waking on ARP packets, as well.
       name = "enable-wowlan-offloads";
       patch = fetchpatch {
         url = "https://git.uninsane.org/colin/linux/commit/c4d2d12e31ae70bb43c6190eccc49e42ad645090.patch";