Kernel panic at skbuff.c

[EvgeniiMekhanik]

[ 1768.037843] [tempesta fw] Warning: cpu 0(4), backlog size 0, active connections mask 0x0, cntwork queue size 0, close backlog is empty
[ 1768.038586] [tempesta fw] Warning: cpu 1(4), backlog size 0, active connections mask 0x0, cntwork queue size 0, close backlog is empty
[ 1768.039332] [tempesta fw] Warning: cpu 2(4), backlog size 0, active connections mask 0x100000000, cntwork queue size 0, close backlog is empty
[ 1768.040133] [tempesta fw] Warning: cpu 3(4), backlog size 0, active connections mask 0x0, cntwork queue size 0, close backlog is empty
[ 1768.040876] [tempesta fw] Warning: cpu 4(4), backlog size 0, active connections mask 0x2000000000, cntwork queue size 0, close backlog is empty
[ 1769.503358] skbuff: skb_under_panic: text:ffffffffc0c68060 len:405 put:13 head:ffff98d66e76e580 data:ffff98d66e76e57b tail:0x140 end:0x140 dev:(efault)
[ 1769.505054] ------------[ cut here ]------------
[ 1769.505723] kernel BUG at net/core/skbuff.c:111! (I am not sure, because I catch it on my branch)

[EvgeniiMekhanik]

This BUG is very rarely. Was reproduced at 9af1fda at MTU 80 after stress tests.

[EvgeniiMekhanik]

It is easy to reproduce, just return -ENOMEM from tls_encrypt and do not stop test. On each call of tls_encrypt we call ss_skb_expand_head_tail for same skb until BUG will be reproduced