Re: [PATCH net] net: selinux: fix memory leak in selinux_netlbl_socket_post_create()

From: maowenan
Date: Fri Mar 08 2019 - 08:40:20 EST



On 2019/3/8 21:20, Paul Moore wrote:
> On Fri, Mar 8, 2019 at 1:31 AM maowenan <maowenan@xxxxxxxxxx> wrote:
>> On 2019/3/8 4:36, Paul Moore wrote:
>>> On Wed, Mar 6, 2019 at 9:44 PM Mao Wenan <maowenan@xxxxxxxxxx> wrote:
>>>>
>>>> If netlbl_sock_setattr() is failed, it directly returns rc and forgets
>>>> to free secattr.
>>>>
>>>> BUG: memory leak
>>>> unreferenced object 0xffff8883c3ea4200 (size 2664):
>>>> comm "syz-executor.2", pid 8813, jiffies 4297264419 (age 156.090s)
>>>> hex dump (first 32 bytes):
>>>> 7f 00 00 01 7f 00 00 01 eb 7f ed 71 4e 24 00 00 ...........qN$..
>>>> 02 00 07 40 00 00 00 00 00 00 00 00 00 00 00 00 ...@............
>>>> backtrace:
>>>> [<000000004c0228da>] sk_alloc+0x3d/0xc00 net/core/sock.c:1523
>>>> [<00000000535a3da2>] inet_create+0x339/0xe10 net/ipv4/af_inet.c:321
>>>> [<000000009aec3cfe>] __sock_create+0x3fa/0x790 net/socket.c:1275
>>>> [<000000004274b384>] sock_create net/socket.c:1315 [inline]
>>>> [<000000004274b384>] __sys_socket+0xe7/0x1d0 net/socket.c:1345
>>>> [<00000000b3fdc826>] __do_sys_socket net/socket.c:1354 [inline]
>>>> [<00000000b3fdc826>] __se_sys_socket net/socket.c:1352 [inline]
>>>> [<00000000b3fdc826>] __x64_sys_socket+0x74/0xb0 net/socket.c:1352
>>>> [<000000004ae3186e>] do_syscall_64+0xc8/0x580 arch/x86/entry/common.c:290
>>>> [<00000000bc0d2230>] entry_SYSCALL_64_after_hwframe+0x49/0xbe
>>>> [<00000000f737e62f>] 0xffffffffffffffff
>>>>
>>>> BUG: memory leak
>>>> unreferenced object 0xffff8883de23d570 (size 32):
>>>> comm "syz-executor.2", pid 8813, jiffies 4297264419 (age 156.090s)
>>>> hex dump (first 32 bytes):
>>>> 02 00 00 00 00 00 00 00 60 a9 40 ce 83 88 ff ff ........`.@.....
>>>> 01 00 00 00 03 00 00 00 10 00 00 00 00 00 00 00 ................
>>>> backtrace:
>>>> [<0000000035ba8b75>] security_sk_alloc+0x5e/0xb0 security/security.c:1473
>>>> [<00000000302cc426>] sk_prot_alloc+0x8e/0x290 net/core/sock.c:1472
>>>> [<000000004c0228da>] sk_alloc+0x3d/0xc00 net/core/sock.c:1523
>>>> [<00000000535a3da2>] inet_create+0x339/0xe10 net/ipv4/af_inet.c:321
>>>> [<000000009aec3cfe>] __sock_create+0x3fa/0x790 net/socket.c:1275
>>>> [<000000004274b384>] sock_create net/socket.c:1315 [inline]
>>>> [<000000004274b384>] __sys_socket+0xe7/0x1d0 net/socket.c:1345
>>>> [<00000000b3fdc826>] __do_sys_socket net/socket.c:1354 [inline]
>>>> [<00000000b3fdc826>] __se_sys_socket net/socket.c:1352 [inline]
>>>> [<00000000b3fdc826>] __x64_sys_socket+0x74/0xb0 net/socket.c:1352
>>>> [<000000004ae3186e>] do_syscall_64+0xc8/0x580 arch/x86/entry/common.c:290
>>>> [<00000000bc0d2230>] entry_SYSCALL_64_after_hwframe+0x49/0xbe
>>>> [<00000000f737e62f>] 0xffffffffffffffff
>>>>
>>>> BUG: memory leak
>>>> unreferenced object 0xffff8883ce40a960 (size 64):
>>>> comm "syz-executor.2", pid 8813, jiffies 4297264420 (age 156.089s)
>>>> hex dump (first 32 bytes):
>>>> 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
>>>> 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
>>>> backtrace:
>>>> [<0000000029add401>] selinux_netlbl_socket_post_create+0x68/0x130 security/selinux/netlabel.c:416
>>>> [<000000005368a19c>] selinux_socket_post_create+0x31a/0x7f0 security/selinux/hooks.c:4597
>>>> [<00000000bd4730e2>] security_socket_post_create+0x70/0xc0 security/security.c:1385
>>>> [<00000000671052a4>] __sock_create+0x5a6/0x790 net/socket.c:1291
>>>> [<000000004274b384>] sock_create net/socket.c:1315 [inline]
>>>> [<000000004274b384>] __sys_socket+0xe7/0x1d0 net/socket.c:1345
>>>> [<00000000b3fdc826>] __do_sys_socket net/socket.c:1354 [inline]
>>>> [<00000000b3fdc826>] __se_sys_socket net/socket.c:1352 [inline]
>>>> [<00000000b3fdc826>] __x64_sys_socket+0x74/0xb0 net/socket.c:1352
>>>> [<000000004ae3186e>] do_syscall_64+0xc8/0x580 arch/x86/entry/common.c:290
>>>> [<00000000bc0d2230>] entry_SYSCALL_64_after_hwframe+0x49/0xbe
>>>> [<00000000f737e62f>] 0xffffffffffffffff
>>>>
>>>> Fixes: 389fb800ac8b("netlabel: Label incoming TCP connections correctly in SELinux")
>>>> Reported-by: Hulk Robot <hulkci@xxxxxxxxxx>
>>>> Signed-off-by: Mao Wenan <maowenan@xxxxxxxxxx>
>>>> ---
>>>> security/selinux/netlabel.c | 3 +++
>>>> 1 file changed, 3 insertions(+)
>>>>
>>>> diff --git a/security/selinux/netlabel.c b/security/selinux/netlabel.c
>>>> index 186e727b737b..f3da05338580 100644
>>>> --- a/security/selinux/netlabel.c
>>>> +++ b/security/selinux/netlabel.c
>>>> @@ -426,6 +426,9 @@ int selinux_netlbl_socket_post_create(struct sock *sk, u16 family)
>>>> rc = 0;
>>>> break;
>>>> }
>>>> + if (rc != 0) {
>>>> + netlbl_secattr_free(secattr);
>>>> + }
>>>
>>> This is likely going to cause a problem as the
>>> sock->sk_security->nlbl_secattr still has a reference to the secattr
>>> pointer you are releasing here. Assuming things are working correctly
>>> elsewhere, I believe freeing secattr here will result in a double free
>>> when the network stacks cleans up after the failed socket creation
>>> (via the sock_release() call in the error handling code).
>> Do you mean here?
>> err = security_socket_post_create(sock, family, type, protocol, kern);
>> if (err)
>> goto out_sock_release;
>> ...
>>
>> out_sock_release:
>> sock_release(sock);
>> return err;
>> yes, the call trace may like below:
>> sock_release->__sock_release->inet_release->tcp_close->sock_put->sk_free->__sk_free->sk_destruct->
>> __sk_destruct->sk_prot_free->security_sk_free->selinux_sk_free_security->selinux_netlbl_sk_security_free->
>> netlbl_secattr_free
>>
>> so, it is no need to free in selinux_netlbl_socket_post_create().
>> Please correct me if this is incorrect.
>
> I haven't followed the entirely call stack lately, but that looks right.
>
>>> It looks like you may have found this via a test tool (syzbot?), do
>>> you have a reproducer you can share?
>>
>> The issue has been found by syzkaller, but unfortunately it is no reproducer, it is
>> only some backtrace.
>
> If you are able to reproduce this please let us know, it may be a sign
> that something is wrong with the LSM hooks.
>
sure, I try to reproduce it and keep running test the whole night.
> Thanks.
>