netfilter: xt_IDLETIMER: fix panic that occurs when timer_type has garbage value
authorJuhee Kang <claudiajkang@gmail.com>
Mon, 4 Oct 2021 12:14:38 +0000 (21:14 +0900)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 27 Oct 2021 07:56:47 +0000 (09:56 +0200)
commit2a670c323055282c9b72794a491d53cef86bbeaf
tree29cb409640697cfba56ac7f7deaf331f9c65ea45
parent0f4308a164a9de4ffaf86b22082dbdfa16bdc601
netfilter: xt_IDLETIMER: fix panic that occurs when timer_type has garbage value

[ Upstream commit 902c0b1887522a099aa4e1e6b4b476c2fe5dd13e ]

Currently, when the rule related to IDLETIMER is added, idletimer_tg timer
structure is initialized by kmalloc on executing idletimer_tg_create
function. However, in this process timer->timer_type is not defined to
a specific value. Thus, timer->timer_type has garbage value and it occurs
kernel panic. So, this commit fixes the panic by initializing
timer->timer_type using kzalloc instead of kmalloc.

Test commands:
    # iptables -A OUTPUT -j IDLETIMER --timeout 1 --label test
    $ cat /sys/class/xt_idletimer/timers/test
      Killed

Splat looks like:
    BUG: KASAN: user-memory-access in alarm_expires_remaining+0x49/0x70
    Read of size 8 at addr 0000002e8c7bc4c8 by task cat/917
    CPU: 12 PID: 917 Comm: cat Not tainted 5.14.0+ #3 79940a339f71eb14fc81aee1757a20d5bf13eb0e
    Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.13.0-1ubuntu1.1 04/01/2014
    Call Trace:
     dump_stack_lvl+0x6e/0x9c
     kasan_report.cold+0x112/0x117
     ? alarm_expires_remaining+0x49/0x70
     __asan_load8+0x86/0xb0
     alarm_expires_remaining+0x49/0x70
     idletimer_tg_show+0xe5/0x19b [xt_IDLETIMER 11219304af9316a21bee5ba9d58f76a6b9bccc6d]
     dev_attr_show+0x3c/0x60
     sysfs_kf_seq_show+0x11d/0x1f0
     ? device_remove_bin_file+0x20/0x20
     kernfs_seq_show+0xa4/0xb0
     seq_read_iter+0x29c/0x750
     kernfs_fop_read_iter+0x25a/0x2c0
     ? __fsnotify_parent+0x3d1/0x570
     ? iov_iter_init+0x70/0x90
     new_sync_read+0x2a7/0x3d0
     ? __x64_sys_llseek+0x230/0x230
     ? rw_verify_area+0x81/0x150
     vfs_read+0x17b/0x240
     ksys_read+0xd9/0x180
     ? vfs_write+0x460/0x460
     ? do_syscall_64+0x16/0xc0
     ? lockdep_hardirqs_on+0x79/0x120
     __x64_sys_read+0x43/0x50
     do_syscall_64+0x3b/0xc0
     entry_SYSCALL_64_after_hwframe+0x44/0xae
    RIP: 0033:0x7f0cdc819142
    Code: c0 e9 c2 fe ff ff 50 48 8d 3d 3a ca 0a 00 e8 f5 19 02 00 0f 1f 44 00 00 f3 0f 1e fa 64 8b 04 25 18 00 00 00 85 c0 75 10 0f 05 <48> 3d 00 f0 ff ff 77 56 c3 0f 1f 44 00 00 48 83 ec 28 48 89 54 24
    RSP: 002b:00007fff28eee5b8 EFLAGS: 00000246 ORIG_RAX: 0000000000000000
    RAX: ffffffffffffffda RBX: 0000000000020000 RCX: 00007f0cdc819142
    RDX: 0000000000020000 RSI: 00007f0cdc032000 RDI: 0000000000000003
    RBP: 00007f0cdc032000 R08: 00007f0cdc031010 R09: 0000000000000000
    R10: 0000000000000022 R11: 0000000000000246 R12: 00005607e9ee31f0
    R13: 0000000000000003 R14: 0000000000020000 R15: 0000000000020000

Fixes: 68983a354a65 ("netfilter: xtables: Add snapshot of hardidletimer target")
Signed-off-by: Juhee Kang <claudiajkang@gmail.com>
Reviewed-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
net/netfilter/xt_IDLETIMER.c