In the Linux kernel, the following vulnerability has been resolved:
bpf: make sure skb->len != 0 when redirecting to a tunneling device
syzkaller managed to trigger another case where skb->len == 0 when we enter _devqueue_xmit:
WARNING: CPU: 0 PID: 2470 at include/linux/skbuff.h:2576 skbassertlen include/linux/skbuff.h:2576 [inline] WARNING: CPU: 0 PID: 2470 at include/linux/skbuff.h:2576 _devqueue_xmit+0x2069/0x35e0 net/core/dev.c:4295
Call Trace: devqueuexmit+0x17/0x20 net/core/dev.c:4406 bpftxskb net/core/filter.c:2115 [inline] _bpfredirectnomac net/core/filter.c:2140 [inline] _bpfredirect+0x5fb/0xda0 net/core/filter.c:2163 bpfcloneredirect net/core/filter.c:2447 [inline] bpfcloneredirect+0x247/0x390 net/core/filter.c:2419 bpfprog48159a89cb4a9a16+0x59/0x5e bpfdispatchernopfunc include/linux/bpf.h:897 [inline] _bpfprogrun include/linux/filter.h:596 [inline] bpfprogrun include/linux/filter.h:603 [inline] bpftestrun+0x46c/0x890 net/bpf/testrun.c:402 bpfprogtestrunskb+0xbdc/0x14c0 net/bpf/testrun.c:1170 bpfprogtestrun+0x345/0x3c0 kernel/bpf/syscall.c:3648 _sysbpf+0x43a/0x6c0 kernel/bpf/syscall.c:5005 _dosysbpf kernel/bpf/syscall.c:5091 [inline] _sesysbpf kernel/bpf/syscall.c:5089 [inline] _x64sysbpf+0x7c/0x90 kernel/bpf/syscall.c:5089 dosyscall64+0x54/0x70 arch/x86/entry/common.c:48 entrySYSCALL64afterhwframe+0x61/0xc6
The reproducer doesn't really reproduce outside of syzkaller environment, so I'm taking a guess here. It looks like we do generate correct ETHHLEN-sized packet, but we redirect the packet to the tunneling device. Before we do so, we _skbpull l2 header and arrive again at skb->len == 0. Doesn't seem like we can do anything better than having an explicit check after _skb_pull?