Bluetooth: avoid killing an already killed socket
authorSudip Mukherjee <sudipm.mukherjee@gmail.com>
Sun, 15 Jul 2018 19:36:50 +0000 (20:36 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 22 Aug 2018 05:46:11 +0000 (07:46 +0200)
commitf276e2efb66118e3fd98e38a7903b204c85e4cec
treee39b66c84785abdbfaf1e32647a150cb22d4ebba
parent31e4f8ba01731ea34d57ff10a2ffc8f096160e0e
Bluetooth: avoid killing an already killed socket

commit 4e1a720d0312fd510699032c7694a362a010170f upstream.

slub debug reported:

[  440.648642] =============================================================================
[  440.648649] BUG kmalloc-1024 (Tainted: G    BU     O   ): Poison overwritten
[  440.648651] -----------------------------------------------------------------------------

[  440.648655] INFO: 0xe70f4bec-0xe70f4bec. First byte 0x6a instead of 0x6b
[  440.648665] INFO: Allocated in sk_prot_alloc+0x6b/0xc6 age=33155 cpu=1 pid=1047
[  440.648671]  ___slab_alloc.constprop.24+0x1fc/0x292
[  440.648675]  __slab_alloc.isra.18.constprop.23+0x1c/0x25
[  440.648677]  __kmalloc+0xb6/0x17f
[  440.648680]  sk_prot_alloc+0x6b/0xc6
[  440.648683]  sk_alloc+0x1e/0xa1
[  440.648700]  sco_sock_alloc.constprop.6+0x26/0xaf [bluetooth]
[  440.648716]  sco_connect_cfm+0x166/0x281 [bluetooth]
[  440.648731]  hci_conn_request_evt.isra.53+0x258/0x281 [bluetooth]
[  440.648746]  hci_event_packet+0x28b/0x2326 [bluetooth]
[  440.648759]  hci_rx_work+0x161/0x291 [bluetooth]
[  440.648764]  process_one_work+0x163/0x2b2
[  440.648767]  worker_thread+0x1a9/0x25c
[  440.648770]  kthread+0xf8/0xfd
[  440.648774]  ret_from_fork+0x2e/0x38
[  440.648779] INFO: Freed in __sk_destruct+0xd3/0xdf age=3815 cpu=1 pid=1047
[  440.648782]  __slab_free+0x4b/0x27a
[  440.648784]  kfree+0x12e/0x155
[  440.648787]  __sk_destruct+0xd3/0xdf
[  440.648790]  sk_destruct+0x27/0x29
[  440.648793]  __sk_free+0x75/0x91
[  440.648795]  sk_free+0x1c/0x1e
[  440.648810]  sco_sock_kill+0x5a/0x5f [bluetooth]
[  440.648825]  sco_conn_del+0x8e/0xba [bluetooth]
[  440.648840]  sco_disconn_cfm+0x3a/0x41 [bluetooth]
[  440.648855]  hci_event_packet+0x45e/0x2326 [bluetooth]
[  440.648868]  hci_rx_work+0x161/0x291 [bluetooth]
[  440.648872]  process_one_work+0x163/0x2b2
[  440.648875]  worker_thread+0x1a9/0x25c
[  440.648877]  kthread+0xf8/0xfd
[  440.648880]  ret_from_fork+0x2e/0x38
[  440.648884] INFO: Slab 0xf4718580 objects=27 used=27 fp=0x  (null) flags=0x40008100
[  440.648886] INFO: Object 0xe70f4b88 @offset=19336 fp=0xe70f54f8

When KASAN was enabled, it reported:

[  210.096613] ==================================================================
[  210.096634] BUG: KASAN: use-after-free in ex_handler_refcount+0x5b/0x127
[  210.096641] Write of size 4 at addr ffff880107e17160 by task kworker/u9:1/2040

[  210.096651] CPU: 1 PID: 2040 Comm: kworker/u9:1 Tainted: G     U     O    4.14.47-20180606+ #2
[  210.096654] Hardware name: , BIOS 2017.01-00087-g43e04de 08/30/2017
[  210.096693] Workqueue: hci0 hci_rx_work [bluetooth]
[  210.096698] Call Trace:
[  210.096711]  dump_stack+0x46/0x59
[  210.096722]  print_address_description+0x6b/0x23b
[  210.096729]  ? ex_handler_refcount+0x5b/0x127
[  210.096736]  kasan_report+0x220/0x246
[  210.096744]  ex_handler_refcount+0x5b/0x127
[  210.096751]  ? ex_handler_clear_fs+0x85/0x85
[  210.096757]  fixup_exception+0x8c/0x96
[  210.096766]  do_trap+0x66/0x2c1
[  210.096773]  do_error_trap+0x152/0x180
[  210.096781]  ? fixup_bug+0x78/0x78
[  210.096817]  ? hci_debugfs_create_conn+0x244/0x26a [bluetooth]
[  210.096824]  ? __schedule+0x113b/0x1453
[  210.096830]  ? sysctl_net_exit+0xe/0xe
[  210.096837]  ? __wake_up_common+0x343/0x343
[  210.096843]  ? insert_work+0x107/0x163
[  210.096850]  invalid_op+0x1b/0x40
[  210.096888] RIP: 0010:hci_debugfs_create_conn+0x244/0x26a [bluetooth]
[  210.096892] RSP: 0018:ffff880094a0f970 EFLAGS: 00010296
[  210.096898] RAX: 0000000000000000 RBX: ffff880107e170e8 RCX: ffff880107e17160
[  210.096902] RDX: 000000000000002f RSI: ffff88013b80ed40 RDI: ffffffffa058b940
[  210.096906] RBP: ffff88011b2b0578 R08: 00000000852f0ec9 R09: ffffffff81cfcf9b
[  210.096909] R10: 00000000d21bdad7 R11: 0000000000000001 R12: ffff8800967b0488
[  210.096913] R13: ffff880107e17168 R14: 0000000000000068 R15: ffff8800949c0008
[  210.096920]  ? __sk_destruct+0x2c6/0x2d4
[  210.096959]  hci_event_packet+0xff5/0x7de2 [bluetooth]
[  210.096969]  ? __local_bh_enable_ip+0x43/0x5b
[  210.097004]  ? l2cap_sock_recv_cb+0x158/0x166 [bluetooth]
[  210.097039]  ? hci_le_meta_evt+0x2bb3/0x2bb3 [bluetooth]
[  210.097075]  ? l2cap_ertm_init+0x94e/0x94e [bluetooth]
[  210.097093]  ? xhci_urb_enqueue+0xbd8/0xcf5 [xhci_hcd]
[  210.097102]  ? __accumulate_pelt_segments+0x24/0x33
[  210.097109]  ? __accumulate_pelt_segments+0x24/0x33
[  210.097115]  ? __update_load_avg_se.isra.2+0x217/0x3a4
[  210.097122]  ? set_next_entity+0x7c3/0x12cd
[  210.097128]  ? pick_next_entity+0x25e/0x26c
[  210.097135]  ? pick_next_task_fair+0x2ca/0xc1a
[  210.097141]  ? switch_mm_irqs_off+0x346/0xb4f
[  210.097147]  ? __switch_to+0x769/0xbc4
[  210.097153]  ? compat_start_thread+0x66/0x66
[  210.097188]  ? hci_conn_check_link_mode+0x1cd/0x1cd [bluetooth]
[  210.097195]  ? finish_task_switch+0x392/0x431
[  210.097228]  ? hci_rx_work+0x154/0x487 [bluetooth]
[  210.097260]  hci_rx_work+0x154/0x487 [bluetooth]
[  210.097269]  process_one_work+0x579/0x9e9
[  210.097277]  worker_thread+0x68f/0x804
[  210.097285]  kthread+0x31c/0x32b
[  210.097292]  ? rescuer_thread+0x70c/0x70c
[  210.097299]  ? kthread_create_on_node+0xa3/0xa3
[  210.097306]  ret_from_fork+0x35/0x40

[  210.097314] Allocated by task 2040:
[  210.097323]  kasan_kmalloc.part.1+0x51/0xc7
[  210.097328]  __kmalloc+0x17f/0x1b6
[  210.097335]  sk_prot_alloc+0xf2/0x1a3
[  210.097340]  sk_alloc+0x22/0x297
[  210.097375]  sco_sock_alloc.constprop.7+0x23/0x202 [bluetooth]
[  210.097410]  sco_connect_cfm+0x2d0/0x566 [bluetooth]
[  210.097443]  hci_conn_request_evt.isra.53+0x6d3/0x762 [bluetooth]
[  210.097476]  hci_event_packet+0x85e/0x7de2 [bluetooth]
[  210.097507]  hci_rx_work+0x154/0x487 [bluetooth]
[  210.097512]  process_one_work+0x579/0x9e9
[  210.097517]  worker_thread+0x68f/0x804
[  210.097523]  kthread+0x31c/0x32b
[  210.097529]  ret_from_fork+0x35/0x40

[  210.097533] Freed by task 2040:
[  210.097539]  kasan_slab_free+0xb3/0x15e
[  210.097544]  kfree+0x103/0x1a9
[  210.097549]  __sk_destruct+0x2c6/0x2d4
[  210.097584]  sco_conn_del.isra.1+0xba/0x10e [bluetooth]
[  210.097617]  hci_event_packet+0xff5/0x7de2 [bluetooth]
[  210.097648]  hci_rx_work+0x154/0x487 [bluetooth]
[  210.097653]  process_one_work+0x579/0x9e9
[  210.097658]  worker_thread+0x68f/0x804
[  210.097663]  kthread+0x31c/0x32b
[  210.097670]  ret_from_fork+0x35/0x40

[  210.097676] The buggy address belongs to the object at ffff880107e170e8
 which belongs to the cache kmalloc-1024 of size 1024
[  210.097681] The buggy address is located 120 bytes inside of
 1024-byte region [ffff880107e170e8ffff880107e174e8)
[  210.097683] The buggy address belongs to the page:
[  210.097689] page:ffffea00041f8400 count:1 mapcount:0 mapping:          (null) index:0xffff880107e15b68 compound_mapcount: 0
[  210.110194] flags: 0x8000000000008100(slab|head)
[  210.115441] raw: 8000000000008100 0000000000000000 ffff880107e15b68 0000000100170016
[  210.115448] raw: ffffea0004a47620 ffffea0004b48e20 ffff88013b80ed40 0000000000000000
[  210.115451] page dumped because: kasan: bad access detected

[  210.115454] Memory state around the buggy address:
[  210.115460]  ffff880107e17000: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
[  210.115465]  ffff880107e17080: fc fc fc fc fc fc fc fc fc fc fc fc fc fb fb fb
[  210.115469] >ffff880107e17100: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[  210.115472]                                                        ^
[  210.115477]  ffff880107e17180: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[  210.115481]  ffff880107e17200: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[  210.115483] ==================================================================

And finally when BT_DBG() and ftrace was enabled it showed:

       <...>-14979 [001] ....   186.104191: sco_sock_kill <-sco_sock_close
       <...>-14979 [001] ....   186.104191: sco_sock_kill <-sco_sock_release
       <...>-14979 [001] ....   186.104192: sco_sock_kill: sk ef0497a0 state 9
       <...>-14979 [001] ....   186.104193: bt_sock_unlink <-sco_sock_kill
kworker/u9:2-792   [001] ....   186.104246: sco_sock_kill <-sco_conn_del
kworker/u9:2-792   [001] ....   186.104248: sco_sock_kill: sk ef0497a0 state 9
kworker/u9:2-792   [001] ....   186.104249: bt_sock_unlink <-sco_sock_kill
kworker/u9:2-792   [001] ....   186.104250: sco_sock_destruct <-__sk_destruct
kworker/u9:2-792   [001] ....   186.104250: sco_sock_destruct: sk ef0497a0
kworker/u9:2-792   [001] ....   186.104860: hci_conn_del <-hci_event_packet
kworker/u9:2-792   [001] ....   186.104864: hci_conn_del: hci0 hcon ef0484c0 handle 266

Only in the failed case, sco_sock_kill() gets called with the same sock
pointer two times. Add a check for SOCK_DEAD to avoid continue killing
a socket which has already been killed.

Signed-off-by: Sudip Mukherjee <sudipm.mukherjee@gmail.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
net/bluetooth/sco.c