A runtime error occured: /usr/lib/lua/luci/model/network.lua:187: attempt to index a nil value #2027
Comments
|
The weird thing is that The only condition under which |
|
Yep, understand that and have already trawled through that code. So by putting that debug in place I never see a failure backtrace. I only see the "is nil wlan0" as it will fail on the debug line and the next test passes. As I said in the previous issue, I'm fairly sure it's not a memory related issue : root@WRT32x:~# free This box literally does nothing other than shunt packets between the wireless and wired LAN. It's a very expensive AP. It has me stuffed, but as long as I have a web tab with the status overview up it happens pretty much every day. I'll put some instrumentation in ip.so to find out which allocation is failing and compile up a new image to test. Debug printing from lua is easy enough. Is there a reference for getting a debug print out of the linked library you could point me towards? |
|
It should be enough to use I suspect the problem can be "fixed" with a change like this: diff --git a/libs/luci-lib-ip/src/ip.c b/libs/luci-lib-ip/src/ip.c
index 854a0c09c..7cf3cd444 100644
--- a/libs/luci-lib-ip/src/ip.c
+++ b/libs/luci-lib-ip/src/ip.c
@@ -1517,8 +1517,10 @@ static int link_get(lua_State *L)
struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
struct ifinfomsg ifm = { .ifi_index = if_nametoindex(dev) };
+ lua_newtable(L);
+
if (!msg || !cb)
- return 0;
+ return 1;
nlmsg_append(msg, &ifm, sizeof(ifm), 0);
@@ -1526,8 +1528,6 @@ static int link_get(lua_State *L)
nl_cb_set(cb, NL_CB_FINISH, NL_CB_CUSTOM, cb_done, &st);
nl_cb_err(cb, NL_CB_CUSTOM, cb_error, &st);
- lua_newtable(L);
-
nl_send_auto_complete(sock, msg);
while (st.pending > 0)
But Ideally I'd like to understand the underlying problem first. Maybe its not the amount of free memory but memory fragmentation, or some bad caching / allocation behaviour in libnl-tiny. |
|
Ok, failure to allocate isn't the problem. I put the fprintf in and got the lua error 3 times last night, but never once triggered the print in ip.c. So the error is not due to a failure to allocate those structures. |
|
Ok, the error is failure in nl_connect(). On failure it is returning -6 : Fri Aug 3 16:16:33 2018 daemon.err uhttpd[2337]: no nl_connect -6 In libnl, -6 is NLE_EXIST "Object exists". Plenty of instances splattered across google of similar errors using libnl. Some bletcherous examples of workarounds using loops and waits. I'll keep poking. |
|
Yeah, one common pattern for returning errors in Lua library function is |
|
Amusingly I updated the WRT1900ACv2 to OpenWrt 18.06.1 r7258-5eb055306f (from the repositories, not compiled locally) a couple of weeks ago and I'm periodically seeing exactly the same error at the same location in the logs. That router is actually working as a router and is considerably more busy with a completely different configuration, but precisely the same error. Odd that I'm apparently the only person seeing this. I suppose I'd better take a look at the lua->C call chain and figure out how to gracefully fail that call. |
|
I just encountered this error about 20 minutes ago, on a R7800 running Line 203 is the above mentioned Memory is also not an issue on my device, taken shortly after discovery. |
|
Just to add another potential data point in the event it has any relation: LuCI Master (git-19.058.07653) / OpenWrt SNAPSHOT r9456 |
|
This issue has been bugging me for a while since my monitoring system alerts me on errors from the router. Adding the latest message in my system log as well: Linksys WRT1900AC |
|
I just ran into this issue while running OpenWrt SNAPSHOT r9019-0e8d5ff0fc / LuCI Master (git-19.079.18950-11e64f8) on MikroTik RouterBOARD 750Gr3. |
|
Master / 19.07 do not use this code anymore |
For whatever reason I am unable to re-open : #1649
I am again running a recent git head build :
commit 2a598bb
Author: Rafał Miłecki rafal@milecki.pl
Date: Fri Jul 27 21:54:08 2018 +0200
I have tracked this down to luci.ip returning nil for the the ipc.link(ifname) call.
I inserted debugging code in network.lua prior to the failing line (local phy_only = force_phy_only or (ipc.link(ifname).type ~= 1)) :
And got : Tue Jul 31 18:59:46 2018 daemon.err uhttpd[2337]: ipc.link(ifname) is nil wlan0
The subsequent call (ipc.link(ifname).type ~= 1) then succeeded.
So for whatever reason on this device, very intermittently (like once a day) that call will fail and return nil, which network.lua then attempts to index with the .type call and it fails.
I attached the requested network interface dump to that report, but as I've been unable to re-open it I'll reference it here : https://github.com/openwrt/luci/files/1897666/network.interface.dump.txt
I'm now not sure where to go next. Anyone have any ideas?
The text was updated successfully, but these errors were encountered: