Kernel errors (ath: skbuff alloc of size 1926 failed ) (32 MB RAM device)

Hello Team,

Since I upgraded to OpenWRT 19.07.7 I have been observing the following errors in the kernel log.

[1058685.315351] ath: skbuff alloc of size 1926 failed
[1058685.320260] hostapd: page allocation failure: order:0, mode:0x1080020(GFP_ATOMIC), nodemask=(null)
[1058685.329436] CPU: 0 PID: 1440 Comm: hostapd Not tainted 4.14.221 #0
[1058685.335801] Stack : 819a6080 800b2d94 80500000 804b2a28 00000000 00000000 00000000 00000000
[1058685.344376]         00000000 00000000 00000000 00000000 00000000 00000001 8084d6c0 e0fc90cd
[1058685.352951]         8084d758 00000000 00000000 00008d40 00000038 8044a6f8 00000008 00000000
[1058685.361527]         00000277 af635b94 00000276 00000000 8084d6a0 80000000 00000000 00000000
[1058685.370103]         00000030 80808918 80d3cbc0 00000030 00000003 80281624 00000000 80630000
[1058685.378679]         ...
[1058685.381314] Call Trace:
[1058685.381335] [<800b2d94>] 0x800b2d94
[1058685.387640] [<8044a6f8>] 0x8044a6f8
[1058685.391323] [<80281624>] 0x80281624

Full log can be found here ->

root@PicoStation:~# free
              total        used        free      shared  buff/cache   available
Mem:          27060       16856        3804         388        6400        7892

It seems that the device is out of memory, but when I check it looks like is not the case.

I did not observe the errors when I was using OpenWRT-19.07.5, I haven't installed or changed anything in the configuration after the upgrade.



you forgot to mention your device name…?

guessing from your prompt: is it PicoStation M2HP/MXM?
if so, you are out of luck, since it only has 32MB RAM, which is not enough anymore.

It could be OOM during peak mem consumption while booting, but it may look ok later on while SSH-connected.

1 Like

Correct it's PicoStation M2HP. I know is just 32MB of RAM but why it was working fine on version 19.07.5 and now I am getting errors with v19.07.7. Nothing has changed on my end.

I monitor the router's memory, cpu, load, network, connections, etc with collectd + influxdb and grafana I dont see any spike or any indication that I was out of memory.


Any help here would be appreciated.


Not enough RAM means just that, use less or look for a device with sufficient system resources.

I understand that, my question is why it was stable before in a different 19.07 version?

Hello guys,

can I have any help on this? or should I create a bug ticket ?


Filing a bug might be worthwhile if that bug would manifest in routers with 64+ MB RAM.

32 MB RAM devices are abandonware, unsupported.
Possibly 19.07.7 is just that much larger than 19.07.5 that the RAM is not enough. There is a slightly newer kernel with more backported stuff, newer wifi drivers, newer (larger) other packages etc.

If 19.07.5 works for you, just use that.

Alternatively, build a personal 19.07.7 image with stripped features, e.g. without LuCI GUI. That might help you.

Hi guys,

I just wanted to let you know that I have installed the newest version 19.07.8 , and everything is working fine so far with no issues for almost a week, with the same settings, packages, etc (everything the same). It looks like there's an issue with 19.07.7 that was getting out of memory.