linux-kernel - [BUG] n_gsm: possible recursive locking detected

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-ID: <e2e9cc8f-6bb4-f66f-2d2b-3875c2e66cd3@geanix.com>
Date:   Wed, 17 Jul 2019 11:40:02 +0200
From:   Martin Hundebøll <martin@...nix.com>
To:     linux-kernel <linux-kernel@...r.kernel.org>
Cc:     Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Jiri Slaby <jslaby@...e.com>,
        Dirkjan Bussink <dirkjan.bussink@...ap.com>
Subject: [BUG] n_gsm: possible recursive locking detected

Hi,

The GSM0710 line discipline driver triggers a lockdep warning when 
disabling the ldisc while holding a multiplexed virtual tty open:

============================================
WARNING: possible recursive locking detected
5.2.0-00114-gdab52e30156b #6 Not tainted
--------------------------------------------
cmux/263 is trying to acquire lock:
e1e23b18 (&tty->legacy_mutex){+.+.}, at: __tty_hangup.part.0+0x58/0x27c

but task is already holding lock:
d6eddf48 (&tty->legacy_mutex){+.+.}, at: tty_set_ldisc+0x3c/0x1bc

other info that might help us debug this:
  Possible unsafe locking scenario:

        CPU0
        ----
   lock(&tty->legacy_mutex);
   lock(&tty->legacy_mutex);

  *** DEADLOCK ***

  May be due to missing lock nesting notation

3 locks held by cmux/263:
  #0: d6eddf48 (&tty->legacy_mutex){+.+.}, at: tty_set_ldisc+0x3c/0x1bc
  #1: f28bead9 (&tty->ldisc_sem){++++}, at: tty_ldisc_lock+0x50/0x74
  #2: e5d20e4f (&gsm->mutex){+.+.}, at: gsm_cleanup_mux+0x9c/0x15c

stack backtrace:
CPU: 0 PID: 263 Comm: cmux Not tainted 5.2.0-00114-gdab52e30156b #6
Hardware name: Freescale i.MX6 Ultralite (Device Tree)
[<c011184c>] (unwind_backtrace) from [<c010cc74>] (show_stack+0x10/0x14)
[<c010cc74>] (show_stack) from [<c0852488>] (dump_stack+0xd4/0x108)
[<c0852488>] (dump_stack) from [<c017be90>] (__lock_acquire+0x6ec/0x1e84)
[<c017be90>] (__lock_acquire) from [<c017ddc4>] (lock_acquire+0xcc/0x204)
[<c017ddc4>] (lock_acquire) from [<c086e9d0>] (__mutex_lock+0x64/0x90c)
[<c086e9d0>] (__mutex_lock) from [<c086f294>] (mutex_lock_nested+0x1c/0x24)
[<c086f294>] (mutex_lock_nested) from [<c04c02fc>] 
(__tty_hangup.part.0+0x58/0x27c)
[<c04c02fc>] (__tty_hangup.part.0) from [<c04ce718>] 
(gsm_cleanup_mux+0xe8/0x15c)
[<c04ce718>] (gsm_cleanup_mux) from [<c04ce7d4>] (gsmld_close+0x48/0x90)
[<c04ce7d4>] (gsmld_close) from [<c04c7e24>] (tty_set_ldisc+0xb8/0x1bc)
[<c04c7e24>] (tty_set_ldisc) from [<c04c0b70>] (tty_ioctl+0x640/0xcb0)
[<c04c0b70>] (tty_ioctl) from [<c0297e68>] (do_vfs_ioctl+0x41c/0xa5c)
[<c0297e68>] (do_vfs_ioctl) from [<c02984dc>] (ksys_ioctl+0x34/0x60)
[<c02984dc>] (ksys_ioctl) from [<c0101000>] (ret_fast_syscall+0x0/0x28)
Exception stack(0xc8ce1fa8 to 0xc8ce1ff0)
1fa0:                   00438000 00000000 00000003 00005423 beb6cc04 
beb6cc04
1fc0: 00438000 00000000 00000000 00000036 00000000 00000000 00438000 
beb6ccd4
1fe0: 00438048 beb6cbfc 00427684 b6f58b88


Steps to reproduce using the attached cmux util:

root@...26:~# ./cmux &
[1] 254
SERIAL_PORT = /dev/ttymxc0
AT+IFC=2: Ie5   +CFUN: 1    +CPIN: READY    Call Ready  AT+IFC=2,2   OK
AT+GMM  : AT+GMM   Quectel_M95    OK
AT      : AT   OK
AT+IPR=1: AT+IPR=115200&w   OK
AT+CMUX=: AT+CMUX=0,0,5,512,10,3,30,10,2   OK
Line dicipline set

root@...26:~# cat /dev/gsmtty1 &
[2] 262
root@...26:~# kill %1
[   74.517449] ============================================
[   74.522769] WARNING: possible recursive locking detected
[   74.528094] 5.2.0-00114-gdab52e30156b #6 Not tainted
[   74.533065] --------------------------------------------
<...>


This has supposedly been fixed before in 4d9b109060f6 ("tty: Prevent 
deadlock in n_gsm driver"), but the fix was undone in be7065725590 
("TTY/n_gsm: Removing the wrong tty_unlock/lock() in gsm_dlci_release()")

-- 
Kind regards,
Martin Hundebøll
Embedded Linux Consultant

+45 61 65 54 61
martin@...nix.com

Geanix ApS
https://geanix.com
DK39600706

View attachment "cmux.c" of type "text/x-csrc" (9493 bytes)