Re: possible deadlock in cfg80211_netdev_notifier_call

On Mon, 2021-02-01 at 14:37 +0200, Mike Rapoport wrote:

On Mon, Feb 01, 2021 at 01:17:13AM -0800, syzbot wrote:

quoted

Hello,

syzbot found the following issue on:

HEAD commit:    b01f250d Add linux-next specific files for 20210129
git tree:       linux-next
console output: https://syzkaller.appspot.com/x/log.txt?x=14daa408d00000
kernel config:  https://syzkaller.appspot.com/x/.config?x=725bc96dc234fda7
dashboard link: https://syzkaller.appspot.com/bug?extid=2ae0ca9d7737ad1a62b7
compiler:       gcc (GCC) 10.1.0-syz 20200507
syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=1757f2a0d00000

The issue was bisected to:

commit cc9327f3b085ba5be5639a5ec3ce5b08a0f14a7c
Author: Mike Rapoport [off-list ref]
Date:   Thu Jan 28 07:42:40 2021 +0000

    mm: introduce memfd_secret system call to create "secret" memory areas

bisection log:  https://syzkaller.appspot.com/x/bisect.txt?x=1505d28cd00000
final oops:     https://syzkaller.appspot.com/x/report.txt?x=1705d28cd00000
console output: https://syzkaller.appspot.com/x/log.txt?x=1305d28cd00000

Sounds really weird to me. At this point the memfd_secret syscall is not
even wired to arch syscall handlers. I cannot see how it can be a reason of
deadlock in wireless...

Yeah, forget about it. Usually this is a consequence of the way syzbot
creates tests - it might have created something like

  if (!create_secret_memfd())
    return;
  try_something_on_wireless()

and then of course without your patch it cannot get to the wireless
bits.

Pretty sure I know what's going on here, I'll take a closer look later.

johannes