core: Check for aggregated skb size before copy
This ensures that the target aggregated skb has sufficient
tailroom to copy the incoming skb content. Checking for
the configured size may not reflect the true value.
Fixes the following-
<2> skbuff: skb_over_panic: text:000000006faa6b25 len:4044
put:1288 head:000000007024fb17 data:000000007024fb17
tail:0xfcc end:0xec0 dev:rmnet_ipa0
<2> ------------[ cut here ]------------
<2> kernel ... at net/core/skbuff.c:108!
<2> Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
<2> pc : skb_panic+0x4c/0x50
<2> lr : skb_panic+0x4c/0x50
<2> Call trace:
<2> skb_panic+0x4c/0x50
<2> skb_push+0x0/0x44
<2> skb_copy_bits+0x0/0x264
<2> rmnet_map_linearize_copy+0xa8/0x110 [rmnet_core]
<2> rmnet_map_tx_aggregate+0x358/0x45c [rmnet_core]
<2> rmnet_egress_handler+0x2cc/0x2f4 [rmnet_core]
<2> rmnet_vnd_start_xmit+0x1a0/0x1f8 [rmnet_core]
<2> dev_hard_start_xmit+0x1e0/0x394
<2> sch_direct_xmit+0xe4/0x32c
<2> __qdisc_run+0x4ec/0xb48
<2> __dev_queue_xmit.llvm.8867084656021901384+0x3ec/0x9c8
<2> neigh_direct_output+0x18/0x24
<2> ip_finish_output2+0x3f4/0x5b0
<2> __ip_finish_output+0x1d4/0x248
<2> ip_finish_output+0x50/0xe0
<2> ip_output+0x148/0x194
<2> __ip_queue_xmit+0x378/0x3d4
<2> ip_queue_xmit+0x14/0x20
<2> __tcp_transmit_skb+0x7b4/0xc4c
<2> tcp_write_xmit+0x868/0x1194
<2> __tcp_push_pending_frames+0x38/0xd0
<2> tcp_push+0x124/0x154
<2> tcp_sendmsg_locked+0xae4/0xcd0
<2> tcp_sendmsg+0x38/0x5c
<2> inet6_sendmsg+0x5c/0xa4
<2> sock_write_iter+0xf8/0x17c
<2> __vfs_write+0x1a0/0x220
<2> vfs_write+0xe4/0x1a8
<2> ksys_write+0x7c/0xe8
<2> __arm64_sys_write+0x1c/0x28
<2> el0_svc_common+0xb8/0x1c4
<2> el0_svc_handler+0x6c/0x88
<2> el0_svc+0x8/0xc
CRs-Fixed: 2687877
Change-Id: If8cf289c2629be28b890927c54cb4d4209b24bc5
Signed-off-by: Subash Abhinov Kasiviswanathan <subashab@codeaurora.org>
1 file changed