Cubieboard 2 (A20) LXC issue after 5.31/4.11.5 upgrade

[znoxx]

Hello!

I'm running CB2 board as home server for a long time and after recent upgrade I've got an issue with LXC container.

 

LXC container has "Ubuntu" inside and provides TOR functionalty + access point via connected USB wifi. USB adapter passed to container like this:

 

lxc-wait -n tor -s RUNNING
lxc-device -n tor add wlan0

 

I'm not using it recently, but recently I found, that after connection to AP nothing happens (no internet, DNS servers are IPV6 :)) and host system (CB2) is completely frozen.

 

After reboot I've connected via SSH and retried to connect to access point - again freeze, and screen is populated with this:

Spoiler

Message from syslogd@localhost at Jun 21 09:37:26 ...
 kernel:[  236.678968] Internal error: Oops: 5 [#1] SMP THUMB2

Message from syslogd@localhost at Jun 21 09:37:26 ...
 kernel:[  236.804563] Process cryptomgr_test (pid: 5259, stack limit = 0xde174210)

Message from syslogd@localhost at Jun 21 09:37:26 ...
 kernel:[  236.811288] Stack: (0xde175f10 to 0xde176000)

Message from syslogd@localhost at Jun 21 09:37:26 ...
 kernel:[  236.815674] 5f00:                                     ee7d57c0 ee7d57c0 ed3f65d8 ed55b3c0

Message from syslogd@localhost at Jun 21 09:37:26 ...
 kernel:[  236.823896] 5f20: de175f20 de175f20 00000000 ed3eb000 bfc3b000 c0c72e18 00000401 c0c74100

Message from syslogd@localhost at Jun 21 09:37:26 ...
 kernel:[  236.832120] 5f40: bfc3b028 bfc3b068 c0c74128 c04d5433 de175f50 de175f50 e67a5a00 ee032cc0

Message from syslogd@localhost at Jun 21 09:37:26 ...
 kernel:[  236.840344] 5f60: e67a5a00 00000000 de174000 ee032cc0 c04d9e5d de251d10 e67a521c c04d9e6d

Message from syslogd@localhost at Jun 21 09:37:26 ...
 kernel:[  236.848565] 5f80: e67a5200 c012ff49 ffffffff e67a5a00 c012fe4d 00000000 00000000 00000000

Message from syslogd@localhost at Jun 21 09:37:26 ...
 kernel:[  236.856785] 5fa0: 00000000 00000000 00000000 c0106351 00000000 00000000 00000000 00000000

Message from syslogd@localhost at Jun 21 09:37:26 ...
 kernel:[  236.865006] 5fc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000

Message from syslogd@localhost at Jun 21 09:37:26 ...
 kernel:[  236.873227] 5fe0: 00000000 00000000 00000000 00000000 00000013 00000000 00000000 00000000

Message from syslogd@localhost at Jun 21 09:37:26 ...
 kernel:[  236.913264] Code: 681c 42a3 d014 681c (68e3) 4283 

 

 

I'm pretty sure, that it stared with one of recent upgrades. Unfortunately I utilize "torred access point" not really often, so cannot tell, which upgrade exactly caused this. It worked like charm in April :).

Since such setup (with LXC + bypassing network interface) is not really common, I understand, that it may not be not fixed in visible period of time :(.

 

So 2 questions :

 

1) Did anyone encountered something similar or just something LXC-related with recent kernel and any chances to fix this ?

2) How I can downgrade kernel to previous version (looks like after this I can freeze kernel update via this new utility) ? Where packages are located and what is correct procedure ? (Sorry if it was already asked).

 

P.S. dmesg attached just in case.

 

dmesg4.11

[Charizard]

I've the same issue. LXC works, the container starts but the network sucks. For me when hostapd or openvpn is starting, the board may freeze. When it doesn't freeze, it's impossible to kill hostapd or openvpn. So I've to restart with reboot -f command. Normal reboot will not reboot the board. I haven't got the same dmesg message with my boards (nanopi & banana pi). I got this mesage : 

 

[  134.321805] Unable to handle kernel NULL pointer dereference at virtual address 0000000c
[  134.330177] pgd = c0004000
[  134.333167] [0000000c] *pgd=00000000
[  134.336970] Internal error: Oops: 5 [#1] SMP THUMB2
[  134.342010] Modules linked in: aes_arm(+) ccm xt_nat xt_conntrack veth ir_xmp_decoder ir_mce_kbd_decoder ir_sharp_decoder ir_sanyo_decoder ir_sony_decoder ir_jvc_decoder ir_rc6_decoder ir_nec_decoder ir_rc5_decoder xt_CHECKSUM iptable_mangle ipt_MASQUERADE nf_nat_masquerade_ipv4 iptable_nat nf_conntrack_ipv4 

...

 

Unfortunately I think it's a bug inside the kernel. I have been opened a ticket on bugzilla : 

 

https://bugzilla.kernel.org/show_bug.cgi?id=196085

 

But I know nobody will respond. Few people respond on this forum.

[znoxx]

Good news, I'm not alone =)

@Charizard What do you think, which kernel to pick for downgrade ? My system is actually "production one", I'd like to minimize downtime  and disable further kernel updates.

 

By the way, I use virtual br0 interface for networking with host and port forwarding and everything works. Issue shows up only when connection is made  to physical passed-through interface.

 

[Igor]

35 minutes ago, znoxx said:

What do you think, which kernel to pick for downgrade ?

I guess roll back to previous version (4.9.7) should do the job.
 

apt install linux-image-next-sunxi=5.25 linux-dtb-next-sunxi=5.25

Than go to armbian-config and freeze kernel upgrades until this is not solved ...

[znoxx]

@Igor, thanks a lot.

I will do it in couple of days and check are things are "back".

 

 

[zador.blood.stained]

21 hours ago, Charizard said:

Unfortunately I think it's a bug inside the kernel. I have been opened a ticket on bugzilla : 

 

https://bugzilla.kernel.org/show_bug.cgi?id=196085

 

But I know nobody will respond. Few people respond on this forum.

It probably is, but you are technically running a "distribution kernel", so you (or somebody else) needs to reproduce it on a freshly compiled unpatched mainline kernel and provide minimal and easy steps to reproduce the crash as well as as much info as possible (full dmesg log prior to crashing, full crash output possibly with additional debugging option enabled, kernel config used for compilation)

[znoxx]

On 22/06/2017 at 0:07 PM, Igor said:

Than go to armbian-config and freeze kernel upgrades until this is not solved ...

 

@Igor, thanks a lot for a tip. Issue resolved after downgrading and I have frozen kernel updates.

 

[znoxx]

Just for info - issue is also resolved in 5.32