On Tue, Aug 8, 2017 at 9:48 PM, Willem de Bruijn <willemdebruijn.ker...@gmail.com> wrote: > Thanks for the report, David, and sorry for the breakage. > > I am not able to reproduce the issue with my qemu setup with vhost-net > with experimental_zcopytx so far.
I did just reproduce the (well, a) panic in sock_zerocopy_put. It is indeed in the mm_unaccount_pinned_pages code that tests uarg->mmp: (gdb) list *(sock_zerocopy_put+0x26) 0xffffffff81620316 is in sock_zerocopy_put (net/core/skbuff.c:933). 928 } 929 930 static void mm_unaccount_pinned_pages(struct mmpin *mmp) 931 { 932 if (mmp->user) { 933 atomic_long_sub(mmp->num_pg, &mmp->user->locked_vm); 934 free_uid(mmp->user); This gives me more confidence that the previous fix is sufficient. I will have to revise it to avoid this path for all zerocopy paths besides msg_zerocopy.