nbd: Fix NULL pointer in flush_workqueue
authorSun Ke <sunke32@huawei.com>
Wed, 12 May 2021 11:43:30 +0000 (19:43 +0800)
committerJens Axboe <axboe@kernel.dk>
Wed, 12 May 2021 14:42:43 +0000 (08:42 -0600)
commit79ebe9110fa458d58f1fceb078e2068d7ad37390
tree6e7fa56f9475e4e34be642d7dbfffea4dabd96cc
parent190515f610946db025cdedebde93958b725fb583
nbd: Fix NULL pointer in flush_workqueue

Open /dev/nbdX first, the config_refs will be 1 and
the pointers in nbd_device are still null. Disconnect
/dev/nbdX, then reference a null recv_workq. The
protection by config_refs in nbd_genl_disconnect is useless.

[  656.366194] BUG: kernel NULL pointer dereference, address: 0000000000000020
[  656.368943] #PF: supervisor write access in kernel mode
[  656.369844] #PF: error_code(0x0002) - not-present page
[  656.370717] PGD 10cc87067 P4D 10cc87067 PUD 1074b4067 PMD 0
[  656.371693] Oops: 0002 [#1] SMP
[  656.372242] CPU: 5 PID: 7977 Comm: nbd-client Not tainted 5.11.0-rc5-00040-g76c057c84d28 #1
[  656.373661] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS ?-20190727_073836-buildvm-ppc64le-16.ppc.fedoraproject.org-3.fc31 04/01/2014
[  656.375904] RIP: 0010:mutex_lock+0x29/0x60
[  656.376627] Code: 00 0f 1f 44 00 00 55 48 89 fd 48 83 05 6f d7 fe 08 01 e8 7a c3 ff ff 48 83 05 6a d7 fe 08 01 31 c0 65 48 8b 14 25 00 6d 01 00 <f0> 48 0f b1 55 d
[  656.378934] RSP: 0018:ffffc900005eb9b0 EFLAGS: 00010246
[  656.379350] RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000000
[  656.379915] RDX: ffff888104cf2600 RSI: ffffffffaae8f452 RDI: 0000000000000020
[  656.380473] RBP: 0000000000000020 R08: 0000000000000000 R09: ffff88813bd6b318
[  656.381039] R10: 00000000000000c7 R11: fefefefefefefeff R12: ffff888102710b40
[  656.381599] R13: ffffc900005eb9e0 R14: ffffffffb2930680 R15: ffff88810770ef00
[  656.382166] FS:  00007fdf117ebb40(0000) GS:ffff88813bd40000(0000) knlGS:0000000000000000
[  656.382806] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  656.383261] CR2: 0000000000000020 CR3: 0000000100c84000 CR4: 00000000000006e0
[  656.383819] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[  656.384370] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[  656.384927] Call Trace:
[  656.385111]  flush_workqueue+0x92/0x6c0
[  656.385395]  nbd_disconnect_and_put+0x81/0xd0
[  656.385716]  nbd_genl_disconnect+0x125/0x2a0
[  656.386034]  genl_family_rcv_msg_doit.isra.0+0x102/0x1b0
[  656.386422]  genl_rcv_msg+0xfc/0x2b0
[  656.386685]  ? nbd_ioctl+0x490/0x490
[  656.386954]  ? genl_family_rcv_msg_doit.isra.0+0x1b0/0x1b0
[  656.387354]  netlink_rcv_skb+0x62/0x180
[  656.387638]  genl_rcv+0x34/0x60
[  656.387874]  netlink_unicast+0x26d/0x590
[  656.388162]  netlink_sendmsg+0x398/0x6c0
[  656.388451]  ? netlink_rcv_skb+0x180/0x180
[  656.388750]  ____sys_sendmsg+0x1da/0x320
[  656.389038]  ? ____sys_recvmsg+0x130/0x220
[  656.389334]  ___sys_sendmsg+0x8e/0xf0
[  656.389605]  ? ___sys_recvmsg+0xa2/0xf0
[  656.389889]  ? handle_mm_fault+0x1671/0x21d0
[  656.390201]  __sys_sendmsg+0x6d/0xe0
[  656.390464]  __x64_sys_sendmsg+0x23/0x30
[  656.390751]  do_syscall_64+0x45/0x70
[  656.391017]  entry_SYSCALL_64_after_hwframe+0x44/0xa9

To fix it, just add if (nbd->recv_workq) to nbd_disconnect_and_put().

Fixes: e9e006f5fcf2 ("nbd: fix max number of supported devs")
Signed-off-by: Sun Ke <sunke32@huawei.com>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Link: https://lore.kernel.org/r/20210512114331.1233964-2-sunke32@huawei.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
drivers/block/nbd.c