Re: KASAN: use-after-free Read in get_work_pool

[Eric Biggers <ebiggers3@gmail.com>]

On Wed, Feb 14, 2018 at 02:45:05PM +0100, 'Dmitry Vyukov' via syzkaller-bugs wrote:
> On Wed, Dec 6, 2017 at 1:50 PM, Dmitry Vyukov <dvyukov@google.com> wrote:
> > On Fri, Oct 27, 2017 at 11:18 PM, Cong Wang <xiyou.wangcong@gmail.com> wrote:
> >> On Thu, Oct 26, 2017 at 11:00 PM, Dmitry Vyukov <dvyukov@google.com> wrote:
> >>> On Thu, Oct 26, 2017 at 7:58 PM, Tejun Heo <tj@kernel.org> wrote:
> >>>> Hello,
> >>>>
> >>>> On Thu, Oct 26, 2017 at 09:35:44AM -0700, syzbot wrote:
> >>>>> BUG: KASAN: use-after-free in __read_once_size
> >>>>> include/linux/compiler.h:276 [inline]
> >>>>> BUG: KASAN: use-after-free in atomic64_read
> >>>>> arch/x86/include/asm/atomic64_64.h:21 [inline]
> >>>>> BUG: KASAN: use-after-free in atomic_long_read
> >>>>> include/asm-generic/atomic-long.h:44 [inline]
> >>>>> BUG: KASAN: use-after-free in get_work_pool+0x1c2/0x1e0
> >>>>> kernel/workqueue.c:709
> >>>>> Read of size 8 at addr ffff8801cc58c378 by task syz-executor5/21326
> >>>>>
> >>>>> CPU: 1 PID: 21326 Comm: syz-executor5 Not tainted 4.13.0+ #43
> >>>>> Hardware name: Google Google Compute Engine/Google Compute Engine,
> >>>>> BIOS Google 01/01/2011
> >>>>> Call Trace:
> >>>>>  __dump_stack lib/dump_stack.c:16 [inline]
> >>>>>  dump_stack+0x194/0x257 lib/dump_stack.c:52
> >>>>>  print_address_description+0x73/0x250 mm/kasan/report.c:252
> >>>>>  kasan_report_error mm/kasan/report.c:351 [inline]
> >>>>>  kasan_report+0x24e/0x340 mm/kasan/report.c:409
> >>>>>  __asan_report_load8_noabort+0x14/0x20 mm/kasan/report.c:430
> >>>>>  __read_once_size include/linux/compiler.h:276 [inline]
> >>>>>  atomic64_read arch/x86/include/asm/atomic64_64.h:21 [inline]
> >>>>>  atomic_long_read include/asm-generic/atomic-long.h:44 [inline]
> >>>>>  get_work_pool+0x1c2/0x1e0 kernel/workqueue.c:709
> >>>>>  __queue_work+0x235/0x1150 kernel/workqueue.c:1401
> >>>>>  queue_work_on+0x16a/0x1c0 kernel/workqueue.c:1486
> >>>>>  queue_work include/linux/workqueue.h:489 [inline]
> >>>>>  strp_check_rcv+0x25/0x30 net/strparser/strparser.c:553
> >>>>>  kcm_attach net/kcm/kcmsock.c:1439 [inline]
> >>>>>  kcm_attach_ioctl net/kcm/kcmsock.c:1460 [inline]
> >>>>>  kcm_ioctl+0x826/0x1610 net/kcm/kcmsock.c:1695
> >>>>>  sock_do_ioctl+0x65/0xb0 net/socket.c:961
> >>>>>  sock_ioctl+0x2c2/0x440 net/socket.c:1058
> >>>>>  vfs_ioctl fs/ioctl.c:45 [inline]
> >>>>>  do_vfs_ioctl+0x1b1/0x1530 fs/ioctl.c:685
> >>>>>  SYSC_ioctl fs/ioctl.c:700 [inline]
> >>>>>  SyS_ioctl+0x8f/0xc0 fs/ioctl.c:691
> >>>>>  entry_SYSCALL_64_fastpath+0x1f/0xbe
> >>>>
> >>>> Looks like kcm is trying to reuse a work item whose last workqueue has
> >>>> been destroyed without re-initing it.  A work item needs to be
> >>>> reinit'd.
> >>>
> >>> +kcm maintainers
> >>
> >> Can you try the fix below? There is no C reproducer so I can't verify it.
> >
> >
> > Hi Cong,
> >
> > syzbot can now test proposed patches, see
> > https://github.com/google/syzkaller/blob/master/docs/syzbot.md#communication-with-syzbot
> > for details. Please give it a try.
> >
> >> diff --git a/net/kcm/kcmsock.c b/net/kcm/kcmsock.c
> >> index af4e76ac88ff..7816f44c576a 100644
> >> --- a/net/kcm/kcmsock.c
> >> +++ b/net/kcm/kcmsock.c
> >> @@ -1433,11 +1433,12 @@ static int kcm_attach(struct socket *sock,
> >> struct socket *csock,
> >>         KCM_STATS_INCR(mux->stats.psock_attach);
> >>         mux->psocks_cnt++;
> >>         psock_now_avail(psock);
> >> -       spin_unlock_bh(&mux->lock);
> >>
> >>         /* Schedule RX work in case there are already bytes queued */
> >>         strp_check_rcv(&psock->strp);
> >>
> >> +       spin_unlock_bh(&mux->lock);
> >> +
> >>         return 0;
> >>  }
> 
> 
> Hi Cong,
> 
> Was this ever merged? Is it still necessary?
> 

syzbot is no longer hitting this bug for some reason but it's still there.  Tom,
it looks like you wrote the buggy code (it's yet another KCM bug, apparently);
can you please look into it?

I've put together a C reproducer that works on latest linux-next (next-20180309,
commit 61530b14b059d).  It works as an unprivileged user provided that KCM is
enabled, and that KASAN is enabled so you see the use-after-free report:

#include <linux/bpf.h>
#include <linux/in.h>
#include <linux/kcm.h>
#include <sys/ioctl.h>
#include <sys/socket.h>
#include <sys/syscall.h>
#include <unistd.h>

int main()
{
        union bpf_attr prog = {
                .prog_type = BPF_PROG_TYPE_SOCKET_FILTER,
                .insn_cnt = 2,
                .insns = (__u64)(__u64[]){ 0xb7, 0x95 },
                .license = (__u64)"",
        };
        int tcp_fd, bpf_fd, kcm_fd;
        struct sockaddr_in addr = {
                .sin_family = AF_INET,
                .sin_port = __constant_htons(3270),
                .sin_addr = { __constant_htonl(0x7f000001) }
        };

        tcp_fd = socket(AF_INET, SOCK_STREAM, 0);
        bind(tcp_fd, (void *)&addr, sizeof(addr));
        listen(tcp_fd, 1);
        tcp_fd = socket(AF_INET, SOCK_STREAM, 0);
        connect(tcp_fd, (void *)&addr, sizeof(addr));
        bpf_fd = syscall(__NR_bpf, BPF_PROG_LOAD, &prog, 48);
        kcm_fd = socket(AF_KCM, SOCK_SEQPACKET, 0);
        if (fork() == 0) {
                struct kcm_attach attach = { tcp_fd, bpf_fd };

                for (;;)
                        ioctl(kcm_fd, SIOCKCMATTACH, &attach);
        } else {
                struct kcm_unattach unattach = { tcp_fd };

                for (;;)
                        ioctl(kcm_fd, SIOCKCMUNATTACH, &unattach);
        }
}