netlink: GPF in sock_sndtimeo

56 views
Skip to first unread message

Dmitry Vyukov

unread,
Nov 26, 2016, 10:44:38 AM11/26/16
to David Miller, Johannes Berg, Florian Westphal, Cong Wang, Eric Dumazet, Herbert Xu, netdev, LKML, syzkaller
Hello,

The following program triggers GPF in sock_sndtimeo:
https://gist.githubusercontent.com/dvyukov/c19cadd309791cf5cb9b2bf936d3f48d/raw/1743ba0211079a5465d039512b427bc6b59b1a76/gistfile1.txt

On commit 16ae16c6e5616c084168740990fc508bda6655d4 (Nov 24).

general protection fault: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN
Dumping ftrace buffer:
(ftrace buffer empty)
Modules linked in:
CPU: 1 PID: 19950 Comm: syz-executor Not tainted 4.9.0-rc5+ #54
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
task: ffff88002a0d0840 task.stack: ffff880036920000
RIP: 0010:[<ffffffff86cb35e1>] [< inline >] sock_sndtimeo
include/net/sock.h:2075
RIP: 0010:[<ffffffff86cb35e1>] [<ffffffff86cb35e1>]
netlink_unicast+0xe1/0x730 net/netlink/af_netlink.c:1232
RSP: 0018:ffff880036926f68 EFLAGS: 00010202
RAX: 0000000000000068 RBX: ffff880036927000 RCX: ffffc900021d0000
RDX: 0000000000000d63 RSI: 00000000024000c0 RDI: 0000000000000340
RBP: ffff880036927028 R08: ffffed0006ea7aab R09: ffffed0006ea7aab
R10: 0000000000000001 R11: ffffed0006ea7aaa R12: dffffc0000000000
R13: 0000000000000000 R14: ffff880035de3400 R15: ffff880035de3400
FS: 00007f90a2fc7700(0000) GS:ffff88003ed00000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00000000006de0c0 CR3: 0000000035de6000 CR4: 00000000000006e0
Stack:
ffff880035de3400 ffffffff819f02a1 1ffff10006d24df4 0000000000000004
00004db400000014 ffff880036926fd8 ffffffff00000000 0000000041b58ab3
ffffffff89653c11 ffffffff86cb3500 ffffffff819f0345 ffff880035de3400
Call Trace:
[< inline >] audit_replace kernel/audit.c:817
[<ffffffff816c34b9>] audit_receive_msg+0x22c9/0x2ce0 kernel/audit.c:894
[< inline >] audit_receive_skb kernel/audit.c:1120
[<ffffffff816c40ac>] audit_receive+0x1dc/0x360 kernel/audit.c:1133
[< inline >] netlink_unicast_kernel net/netlink/af_netlink.c:1214
[<ffffffff86cb3a14>] netlink_unicast+0x514/0x730 net/netlink/af_netlink.c:1240
[<ffffffff86cb46d4>] netlink_sendmsg+0xaa4/0xe50 net/netlink/af_netlink.c:1786
[< inline >] sock_sendmsg_nosec net/socket.c:621
[<ffffffff86a6d54f>] sock_sendmsg+0xcf/0x110 net/socket.c:631
[<ffffffff86a6d8bb>] sock_write_iter+0x32b/0x620 net/socket.c:829
[< inline >] new_sync_write fs/read_write.c:499
[<ffffffff81a6f24e>] __vfs_write+0x4fe/0x830 fs/read_write.c:512
[<ffffffff81a70cf5>] vfs_write+0x175/0x4e0 fs/read_write.c:560
[< inline >] SYSC_write fs/read_write.c:607
[<ffffffff81a75180>] SyS_write+0x100/0x240 fs/read_write.c:599
[<ffffffff81009a24>] do_syscall_64+0x2f4/0x940 arch/x86/entry/common.c:280
[<ffffffff88149e8d>] entry_SYSCALL64_slow_path+0x25/0x25
Code: fe 4c 89 f7 e8 31 16 ff ff 8b 8d 70 ff ff ff 49 89 c7 31 c0 85
c9 75 25 e8 7d 4a a3 fa 49 8d bd 40 03 00 00 48 89 f8 48 c1 e8 03 <42>
80 3c 20 00 0f 85 3a 06 00 00 49 8b 85 40 03 00 00 4c 8d 73
RIP [< inline >] sock_sndtimeo include/net/sock.h:2075
RIP [<ffffffff86cb35e1>] netlink_unicast+0xe1/0x730
net/netlink/af_netlink.c:1232
RSP <ffff880036926f68>
---[ end trace 8383a15fba6fdc59 ]---

Eric Dumazet

unread,
Nov 26, 2016, 11:17:44 AM11/26/16
to Dmitry Vyukov, Richard Guy Briggs, David Miller, Johannes Berg, Florian Westphal, Cong Wang, Herbert Xu, netdev, LKML, syzkaller
CC Richard Guy Briggs <r...@redhat.com>
Looks a bug added in commit 32a1dbaece7e37cea415e03cd426172249aa859e
("audit: try harder to send to auditd upon netlink failure")
or 133e1e5acd4a63c4a0dcc413e90d5decdbce9c4a ("audit: stop an old
auditd being starved out by a new auditd")

Richard, can you take a look ?

Thanks !

Cong Wang

unread,
Nov 26, 2016, 8:11:20 PM11/26/16
to Dmitry Vyukov, David Miller, Johannes Berg, Florian Westphal, Eric Dumazet, Herbert Xu, netdev, LKML, syzkaller
On Sat, Nov 26, 2016 at 7:44 AM, Dmitry Vyukov <dvy...@google.com> wrote:
It is racy on audit_sock, especially on the netns exit path.

Could the following patch help a little bit? Also, I don't see how the
synchronize_net() here could sync with netlink rcv path, since unlike
packets from wire, netlink messages are not handled in BH context
nor I see any RCU taken on rcv path.

diff --git a/kernel/audit.c b/kernel/audit.c
index f1ca116..20bc79e 100644
--- a/kernel/audit.c
+++ b/kernel/audit.c
@@ -1167,10 +1167,13 @@ static void __net_exit audit_net_exit(struct net *net)
{
struct audit_net *aunet = net_generic(net, audit_net_id);
struct sock *sock = aunet->nlsk;
+
+ mutex_lock(&audit_cmd_mutex);
if (sock == audit_sock) {
audit_pid = 0;
audit_sock = NULL;
}
+ mutex_unlock(&audit_cmd_mutex);

RCU_INIT_POINTER(aunet->nlsk, NULL);
synchronize_net();

Richard Guy Briggs

unread,
Nov 29, 2016, 11:49:05 AM11/29/16
to Cong Wang, Dmitry Vyukov, David Miller, Johannes Berg, Florian Westphal, Eric Dumazet, Herbert Xu, netdev, LKML, syzkaller
On 2016-11-26 17:11, Cong Wang wrote:
> On Sat, Nov 26, 2016 at 7:44 AM, Dmitry Vyukov <dvy...@google.com> wrote:
> > Hello,

Eric, thanks for the heads up on this. (more below...)
I think that is the only place it is racy. The other places audit_sock
is set is when the socket failure has just triggered a reset.

Is there a notifier callback for failed or reaped sockets?

> Could the following patch help a little bit? Also, I don't see how the
> synchronize_net() here could sync with netlink rcv path, since unlike
> packets from wire, netlink messages are not handled in BH context
> nor I see any RCU taken on rcv path.

Thanks Cong, this looks like it could help. I'll have a closer look.

> diff --git a/kernel/audit.c b/kernel/audit.c
> index f1ca116..20bc79e 100644
> --- a/kernel/audit.c
> +++ b/kernel/audit.c
> @@ -1167,10 +1167,13 @@ static void __net_exit audit_net_exit(struct net *net)
> {
> struct audit_net *aunet = net_generic(net, audit_net_id);
> struct sock *sock = aunet->nlsk;
> +
> + mutex_lock(&audit_cmd_mutex);
> if (sock == audit_sock) {
> audit_pid = 0;
> audit_sock = NULL;
> }
> + mutex_unlock(&audit_cmd_mutex);
>
> RCU_INIT_POINTER(aunet->nlsk, NULL);
> synchronize_net();

- RGB

--
Richard Guy Briggs <r...@redhat.com>
Kernel Security Engineering, Base Operating Systems, Red Hat
Remote, Ottawa, Canada
Voice: +1.647.777.2635, Internal: (81) 32635

Cong Wang

unread,
Nov 29, 2016, 6:14:03 PM11/29/16
to Richard Guy Briggs, Dmitry Vyukov, David Miller, Johannes Berg, Florian Westphal, Eric Dumazet, Herbert Xu, netdev, LKML, syzkaller
On Tue, Nov 29, 2016 at 8:48 AM, Richard Guy Briggs <r...@redhat.com> wrote:
> On 2016-11-26 17:11, Cong Wang wrote:
>> It is racy on audit_sock, especially on the netns exit path.
>
> I think that is the only place it is racy. The other places audit_sock
> is set is when the socket failure has just triggered a reset.
>
> Is there a notifier callback for failed or reaped sockets?

Is NETLINK_URELEASE event what you are looking for?

Richard Guy Briggs

unread,
Nov 29, 2016, 11:52:12 PM11/29/16
to Cong Wang, Dmitry Vyukov, David Miller, Johannes Berg, Florian Westphal, Eric Dumazet, Herbert Xu, netdev, LKML, syzkaller
Possibly, yes. Thanks, I'll have a look.

Richard Guy Briggs

unread,
Dec 9, 2016, 1:02:54 AM12/9/16
to Cong Wang, linux...@redhat.com, pmo...@redhat.com, Dmitry Vyukov, David Miller, Johannes Berg, Florian Westphal, Eric Dumazet, Herbert Xu, netdev, LKML, syzkaller
On 2016-11-29 23:52, Richard Guy Briggs wrote:
> On 2016-11-29 15:13, Cong Wang wrote:
> > On Tue, Nov 29, 2016 at 8:48 AM, Richard Guy Briggs <r...@redhat.com> wrote:
> > > On 2016-11-26 17:11, Cong Wang wrote:
> > >> It is racy on audit_sock, especially on the netns exit path.
> > >
> > > I think that is the only place it is racy. The other places audit_sock
> > > is set is when the socket failure has just triggered a reset.
> > >
> > > Is there a notifier callback for failed or reaped sockets?
> >
> > Is NETLINK_URELEASE event what you are looking for?
>
> Possibly, yes. Thanks, I'll have a look.

I tried a quick compile attempt on the test case (I assume it is a
socket fuzzer) and get the following compile error:
cc -g -O0 -Wall -D_GNU_SOURCE -o socket_fuzz socket_fuzz.c
socket_fuzz.c:16:1: warning: "_GNU_SOURCE" redefined
<command-line>: warning: this is the location of the previous definition
socket_fuzz.c: In function ‘segv_handler’:
socket_fuzz.c:89: warning: implicit declaration of function ‘__atomic_load_n’
socket_fuzz.c:89: error: ‘__ATOMIC_RELAXED’ undeclared (first use in this function)
socket_fuzz.c:89: error: (Each undeclared identifier is reported only once
socket_fuzz.c:89: error: for each function it appears in.)
socket_fuzz.c: In function ‘loop’:
socket_fuzz.c:280: warning: unused variable ‘errno0’
socket_fuzz.c: In function ‘test’:
socket_fuzz.c:303: warning: implicit declaration of function ‘__atomic_fetch_add’
socket_fuzz.c:303: error: ‘__ATOMIC_SEQ_CST’ undeclared (first use in this function)
socket_fuzz.c:303: warning: implicit declaration of function ‘__atomic_fetch_sub’

I also tried to extend Cong Wang's idea to attempt to proactively respond to a
NETLINK_URELEASE on the audit_sock and reset it, but ran into a locking error
stack dump using mutex_lock(&audit_cmd_mutex) in the notifier callback.
Eliminating the lock since the sock is dead anways eliminates the error.

Is it safe? I'll resubmit if this looks remotely sane. Meanwhile I'll try to
get the test case to compile.

This is being tracked as https://github.com/linux-audit/audit-kernel/issues/30

Subject: [PATCH] audit: proactively reset audit_sock on matching NETLINK_URELEASE

diff --git a/kernel/audit.c b/kernel/audit.c
index f1ca116..91d222d 100644
--- a/kernel/audit.c
+++ b/kernel/audit.c
@@ -423,6 +423,7 @@ static void kauditd_send_skb(struct sk_buff *skb)
snprintf(s, sizeof(s), "audit_pid=%d reset", audit_pid);
audit_log_lost(s);
audit_pid = 0;
+ audit_nlk_portid = 0;
audit_sock = NULL;
} else {
pr_warn("re-scheduling(#%d) write to audit_pid=%d\n",
@@ -1143,6 +1144,28 @@ static int audit_bind(struct net *net, int group)
return 0;
}

+static int audit_sock_netlink_notify(struct notifier_block *nb,
+ unsigned long event,
+ void *_notify)
+{
+ struct netlink_notify *notify = _notify;
+ struct audit_net *aunet = net_generic(notify->net, audit_net_id);
+
+ if (event == NETLINK_URELEASE && notify->protocol == NETLINK_AUDIT) {
+ if (audit_nlk_portid == notify->portid &&
+ audit_sock == aunet->nlsk) {
+ audit_pid = 0;
+ audit_nlk_portid = 0;
+ audit_sock = NULL;
+ }
+ }
+ return NOTIFY_DONE;
+}
+
+static struct notifier_block audit_netlink_notifier = {
+ .notifier_call = audit_sock_netlink_notify,
+};
+
static int __net_init audit_net_init(struct net *net)
{
struct netlink_kernel_cfg cfg = {
@@ -1167,10 +1190,14 @@ static void __net_exit audit_net_exit(struct net *net)
{
struct audit_net *aunet = net_generic(net, audit_net_id);
struct sock *sock = aunet->nlsk;
+
+ mutex_lock(&audit_cmd_mutex);
if (sock == audit_sock) {
audit_pid = 0;
+ audit_nlk_portid = 0;
audit_sock = NULL;
}
+ mutex_unlock(&audit_cmd_mutex);

RCU_INIT_POINTER(aunet->nlsk, NULL);
synchronize_net();
@@ -1202,6 +1229,7 @@ static int __init audit_init(void)
audit_enabled = audit_default;
audit_ever_enabled |= !!audit_default;

+ netlink_register_notifier(&audit_netlink_notifier);
audit_log(NULL, GFP_KERNEL, AUDIT_KERNEL, "initialized");

for (i = 0; i < AUDIT_INODE_BUCKETS; i++)
--
1.7.1


> - RGB

Cong Wang

unread,
Dec 9, 2016, 1:57:39 AM12/9/16
to Richard Guy Briggs, linux...@redhat.com, Paul Moore, Dmitry Vyukov, David Miller, Johannes Berg, Florian Westphal, Eric Dumazet, Herbert Xu, netdev, LKML, syzkaller
On Thu, Dec 8, 2016 at 10:02 PM, Richard Guy Briggs <r...@redhat.com> wrote:
> I also tried to extend Cong Wang's idea to attempt to proactively respond to a
> NETLINK_URELEASE on the audit_sock and reset it, but ran into a locking error
> stack dump using mutex_lock(&audit_cmd_mutex) in the notifier callback.
> Eliminating the lock since the sock is dead anways eliminates the error.
>
> Is it safe? I'll resubmit if this looks remotely sane. Meanwhile I'll try to
> get the test case to compile.

It doesn't look safe, because 'audit_sock', 'audit_nlk_portid' and 'audit_pid'
are updated as a whole and race between audit_receive_msg() and
NETLINK_URELEASE.


> @@ -1167,10 +1190,14 @@ static void __net_exit audit_net_exit(struct net *net)
> {
> struct audit_net *aunet = net_generic(net, audit_net_id);
> struct sock *sock = aunet->nlsk;
> +
> + mutex_lock(&audit_cmd_mutex);
> if (sock == audit_sock) {
> audit_pid = 0;
> + audit_nlk_portid = 0;
> audit_sock = NULL;
> }
> + mutex_unlock(&audit_cmd_mutex);
>

If you decide to use NETLINK_URELEASE notifier, the above piece is no
longer needed, the net_exit path simply releases a refcnt.

Dmitry Vyukov

unread,
Dec 9, 2016, 5:49:46 AM12/9/16
to Richard Guy Briggs, Cong Wang, linux...@redhat.com, pmo...@redhat.com, David Miller, Johannes Berg, Florian Westphal, Eric Dumazet, Herbert Xu, netdev, LKML, syzkaller
-std=gnu99 should help
ignore warnings

Richard Guy Briggs

unread,
Dec 9, 2016, 6:02:00 AM12/9/16
to Cong Wang, linux...@redhat.com, Paul Moore, Dmitry Vyukov, David Miller, Johannes Berg, Florian Westphal, Eric Dumazet, Herbert Xu, netdev, LKML, syzkaller
On 2016-12-08 22:57, Cong Wang wrote:
> On Thu, Dec 8, 2016 at 10:02 PM, Richard Guy Briggs <r...@redhat.com> wrote:
> > I also tried to extend Cong Wang's idea to attempt to proactively respond to a
> > NETLINK_URELEASE on the audit_sock and reset it, but ran into a locking error
> > stack dump using mutex_lock(&audit_cmd_mutex) in the notifier callback.
> > Eliminating the lock since the sock is dead anways eliminates the error.
> >
> > Is it safe? I'll resubmit if this looks remotely sane. Meanwhile I'll try to
> > get the test case to compile.
>
> It doesn't look safe, because 'audit_sock', 'audit_nlk_portid' and 'audit_pid'
> are updated as a whole and race between audit_receive_msg() and
> NETLINK_URELEASE.

This is what I expected and why I originally added the mutex lock in the
callback... The dumps I got were bare with no wrapper identifying the
process context or specific error, so I'm at a bit of a loss how to
solve this (without thinking more about it) other than instinctively
removing the mutex.

Another approach might be to look at consolidating the three into one
identifier or derive the other two from one, or serialize their access.

> > @@ -1167,10 +1190,14 @@ static void __net_exit audit_net_exit(struct net *net)
> > {
> > struct audit_net *aunet = net_generic(net, audit_net_id);
> > struct sock *sock = aunet->nlsk;
> > +
> > + mutex_lock(&audit_cmd_mutex);
> > if (sock == audit_sock) {
> > audit_pid = 0;
> > + audit_nlk_portid = 0;
> > audit_sock = NULL;
> > }
> > + mutex_unlock(&audit_cmd_mutex);
>
> If you decide to use NETLINK_URELEASE notifier, the above piece is no
> longer needed, the net_exit path simply releases a refcnt.

Good point. It would have already killed it off. So this piece is
arguably too late anyways.

Cong Wang

unread,
Dec 9, 2016, 11:14:05 PM12/9/16
to Richard Guy Briggs, linux...@redhat.com, Paul Moore, Dmitry Vyukov, David Miller, Johannes Berg, Florian Westphal, Eric Dumazet, Herbert Xu, netdev, LKML, syzkaller
On Fri, Dec 9, 2016 at 3:01 AM, Richard Guy Briggs <r...@redhat.com> wrote:
> On 2016-12-08 22:57, Cong Wang wrote:
>> On Thu, Dec 8, 2016 at 10:02 PM, Richard Guy Briggs <r...@redhat.com> wrote:
>> > I also tried to extend Cong Wang's idea to attempt to proactively respond to a
>> > NETLINK_URELEASE on the audit_sock and reset it, but ran into a locking error
>> > stack dump using mutex_lock(&audit_cmd_mutex) in the notifier callback.
>> > Eliminating the lock since the sock is dead anways eliminates the error.
>> >
>> > Is it safe? I'll resubmit if this looks remotely sane. Meanwhile I'll try to
>> > get the test case to compile.
>>
>> It doesn't look safe, because 'audit_sock', 'audit_nlk_portid' and 'audit_pid'
>> are updated as a whole and race between audit_receive_msg() and
>> NETLINK_URELEASE.
>
> This is what I expected and why I originally added the mutex lock in the
> callback... The dumps I got were bare with no wrapper identifying the
> process context or specific error, so I'm at a bit of a loss how to
> solve this (without thinking more about it) other than instinctively
> removing the mutex.

Netlink notifier can safely be converted to blocking one, I will send
a patch.

But I seriously doubt you really need NETLINK_URELEASE here,
it adds nothing but overhead, b/c the netlink notifier is called on
every netlink socket in the system, but for net exit path, that is
relatively a slow path.

Also, kauditd_send_skb() needs audit_cmd_mutex too.

I will send a formal patch.

Thanks.

Cong Wang

unread,
Dec 10, 2016, 2:40:51 AM12/10/16
to Richard Guy Briggs, linux...@redhat.com, Paul Moore, Dmitry Vyukov, David Miller, Johannes Berg, Florian Westphal, Eric Dumazet, Herbert Xu, netdev, LKML, syzkaller
Please let me know what you think about the attached patch?

Thanks!
audit_sock.diff

Richard Guy Briggs

unread,
Dec 12, 2016, 5:02:22 AM12/12/16
to Cong Wang, linux...@redhat.com, Paul Moore, Dmitry Vyukov, David Miller, Johannes Berg, Florian Westphal, Eric Dumazet, Herbert Xu, netdev, LKML, syzkaller
On 2016-12-09 20:13, Cong Wang wrote:
> On Fri, Dec 9, 2016 at 3:01 AM, Richard Guy Briggs <r...@redhat.com> wrote:
> > On 2016-12-08 22:57, Cong Wang wrote:
> >> On Thu, Dec 8, 2016 at 10:02 PM, Richard Guy Briggs <r...@redhat.com> wrote:
> >> > I also tried to extend Cong Wang's idea to attempt to proactively respond to a
> >> > NETLINK_URELEASE on the audit_sock and reset it, but ran into a locking error
> >> > stack dump using mutex_lock(&audit_cmd_mutex) in the notifier callback.
> >> > Eliminating the lock since the sock is dead anways eliminates the error.
> >> >
> >> > Is it safe? I'll resubmit if this looks remotely sane. Meanwhile I'll try to
> >> > get the test case to compile.
> >>
> >> It doesn't look safe, because 'audit_sock', 'audit_nlk_portid' and 'audit_pid'
> >> are updated as a whole and race between audit_receive_msg() and
> >> NETLINK_URELEASE.
> >
> > This is what I expected and why I originally added the mutex lock in the
> > callback... The dumps I got were bare with no wrapper identifying the
> > process context or specific error, so I'm at a bit of a loss how to
> > solve this (without thinking more about it) other than instinctively
> > removing the mutex.
>
> Netlink notifier can safely be converted to blocking one, I will send
> a patch.

I had a quick look at how that might happen. The netlink notifier chain
is atomic. Would the registered callback funciton need to spawn a
one-time thread to avoid blocking?

> But I seriously doubt you really need NETLINK_URELEASE here,
> it adds nothing but overhead, b/c the netlink notifier is called on
> every netlink socket in the system, but for net exit path, that is
> relatively a slow path.

I was a bit concerned about its overhead, but was hoping to update
audit_sock more quickly in the case of a sock shutting down for any
reason.

> Also, kauditd_send_skb() needs audit_cmd_mutex too.

Agreed.

> I will send a formal patch.

I had a look at your patch. It looks attractively simple. The audit
next tree has patches queued that add an audit_reset function that will
require more work. I still see some potential gaps.

- If the process messes up (or the sock lookup messes up) it is reset
in the kauditd thread under the audit_cmd_mutex.

- If the process exits normally or is replaced due to an audit_replace
error, it is reset from audit_receive_skb under the audit_cmd_mutex.

- If the process dies before the kauditd thread notices, either reap it
via notifier callback or it needs a check on net exit to reset. This
last one appears necessary to decrement the sock refcount so the sock
can be released in netlink_kernel_release().

If we want to be proactive and use the netlink notifier, we assume the
overhead of adding to the netlink notifier chain and eliminate all the
other reset calls under the kauditd thread. If we are ok being
reactionary, then we'll at least need the net exit check on audit_sock.

Have I understood this correctly?

I'll follow with a patch based on audit#next

There will be an upstream merge conflict between audit#next and net#next
due to the removal of:
RCU_INIT_POINTER(aunet->nlsk, NULL);
synchronize_net();
from the end of audit_net_exit(). This patch should probably go through
the audit maintainer due to the other anticipated merge conflicts.

> Thanks.

Dmitry Vyukov

unread,
Dec 12, 2016, 5:08:03 AM12/12/16
to syzkaller, Richard Guy Briggs, linux...@redhat.com, Paul Moore, David Miller, Johannes Berg, Florian Westphal, Eric Dumazet, Herbert Xu, netdev, LKML
Applied the patch locally and have not seen the bug since then (~24
hours of testing).

Cong Wang

unread,
Dec 12, 2016, 7:11:03 PM12/12/16
to Richard Guy Briggs, linux...@redhat.com, Paul Moore, Dmitry Vyukov, David Miller, Johannes Berg, Florian Westphal, Eric Dumazet, Herbert Xu, netdev, LKML, syzkaller
On Mon, Dec 12, 2016 at 2:02 AM, Richard Guy Briggs <r...@redhat.com> wrote:
> On 2016-12-09 20:13, Cong Wang wrote:
>> Netlink notifier can safely be converted to blocking one, I will send
>> a patch.
>
> I had a quick look at how that might happen. The netlink notifier chain
> is atomic. Would the registered callback funciton need to spawn a
> one-time thread to avoid blocking?

It is already non-atomic now:
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=efa172f42836477bf1ac3c9a3053140df764699c


> I had a look at your patch. It looks attractively simple. The audit
> next tree has patches queued that add an audit_reset function that will
> require more work. I still see some potential gaps.
>
> - If the process messes up (or the sock lookup messes up) it is reset
> in the kauditd thread under the audit_cmd_mutex.
>
> - If the process exits normally or is replaced due to an audit_replace
> error, it is reset from audit_receive_skb under the audit_cmd_mutex.
>
> - If the process dies before the kauditd thread notices, either reap it
> via notifier callback or it needs a check on net exit to reset. This
> last one appears necessary to decrement the sock refcount so the sock
> can be released in netlink_kernel_release().
>
> If we want to be proactive and use the netlink notifier, we assume the
> overhead of adding to the netlink notifier chain and eliminate all the
> other reset calls under the kauditd thread. If we are ok being
> reactionary, then we'll at least need the net exit check on audit_sock.
>

I don't see why we need to check it in net exit if we use refcnt,
because we have two different users of audit_sock: kauditd and
netns, if both take care of refcnt properly, we don't need to worry
about who is the last, no matter what failures occur in what order.

Richard Guy Briggs

unread,
Dec 13, 2016, 2:51:22 AM12/13/16
to Cong Wang, Herbert Xu, Johannes Berg, netdev, Florian Westphal, LKML, Eric Dumazet, linux...@redhat.com, syzkaller, David Miller, Dmitry Vyukov
> commit a12b43ee814625933ff155c20dc863c59cfcf240
> Author: Cong Wang <xiyou.w...@gmail.com>
> Date: Fri Dec 9 17:56:42 2016 -0800
>
> audit: close a race condition on audit_sock
>
> Signed-off-by: Cong Wang <xiyou.w...@gmail.com>
>
> diff --git a/kernel/audit.c b/kernel/audit.c
> index f1ca116..ab947d8 100644
> --- a/kernel/audit.c
> +++ b/kernel/audit.c
> @@ -423,6 +423,8 @@ static void kauditd_send_skb(struct sk_buff *skb)
> snprintf(s, sizeof(s), "audit_pid=%d reset", audit_pid);
> audit_log_lost(s);
> audit_pid = 0;
> + audit_nlk_portid = 0;
> + sock_put(audit_sock);
> audit_sock = NULL;
> } else {
> pr_warn("re-scheduling(#%d) write to audit_pid=%d\n",
> @@ -899,6 +901,9 @@ static int audit_receive_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
> audit_log_config_change("audit_pid", new_pid, audit_pid, 1);
> audit_pid = new_pid;
> audit_nlk_portid = NETLINK_CB(skb).portid;
> + sock_hold(skb->sk);
> + if (audit_sock)
> + sock_put(audit_sock);
> audit_sock = skb->sk;
> }
> if (s.mask & AUDIT_STATUS_RATE_LIMIT) {
> @@ -1167,10 +1172,6 @@ static void __net_exit audit_net_exit(struct net *net)
> {
> struct audit_net *aunet = net_generic(net, audit_net_id);
> struct sock *sock = aunet->nlsk;
> - if (sock == audit_sock) {
> - audit_pid = 0;
> - audit_sock = NULL;
> - }

So how does this not leak memory leaving the sock refcount incremented
by the registered audit daemon when that daemon shuts down normally?

Richard Guy Briggs

unread,
Dec 13, 2016, 3:28:25 AM12/13/16
to Cong Wang, Herbert Xu, Johannes Berg, netdev, Florian Westphal, LKML, Eric Dumazet, linux...@redhat.com, syzkaller, David Miller, Dmitry Vyukov
Sorry, that should have been: How does it not leak if auditd exits
abnormally without sending a shutdown message, but no message is sent on
the queue to trigger an error before the net namespace exits?

> - RGB

Richard Guy Briggs

unread,
Dec 13, 2016, 5:52:39 AM12/13/16
to Cong Wang, linux...@redhat.com, Paul Moore, Dmitry Vyukov, David Miller, Johannes Berg, Florian Westphal, Eric Dumazet, Herbert Xu, netdev, LKML, syzkaller
On 2016-12-12 16:10, Cong Wang wrote:
> On Mon, Dec 12, 2016 at 2:02 AM, Richard Guy Briggs <r...@redhat.com> wrote:
> > On 2016-12-09 20:13, Cong Wang wrote:
> >> Netlink notifier can safely be converted to blocking one, I will send
> >> a patch.
> >
> > I had a quick look at how that might happen. The netlink notifier chain
> > is atomic. Would the registered callback funciton need to spawn a
> > one-time thread to avoid blocking?
>
> It is already non-atomic now:
> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=efa172f42836477bf1ac3c9a3053140df764699c

Ok, that is recent... It is still less attractive as you point out due
to the overhead, but still worth considering if we can't find another
way.

> > I had a look at your patch. It looks attractively simple. The audit
> > next tree has patches queued that add an audit_reset function that will
> > require more work. I still see some potential gaps.
> >
> > - If the process messes up (or the sock lookup messes up) it is reset
> > in the kauditd thread under the audit_cmd_mutex.
> >
> > - If the process exits normally or is replaced due to an audit_replace
> > error, it is reset from audit_receive_skb under the audit_cmd_mutex.
> >
> > - If the process dies before the kauditd thread notices, either reap it
> > via notifier callback or it needs a check on net exit to reset. This
> > last one appears necessary to decrement the sock refcount so the sock
> > can be released in netlink_kernel_release().
> >
> > If we want to be proactive and use the netlink notifier, we assume the
> > overhead of adding to the netlink notifier chain and eliminate all the
> > other reset calls under the kauditd thread. If we are ok being
> > reactionary, then we'll at least need the net exit check on audit_sock.
>
> I don't see why we need to check it in net exit if we use refcnt,
> because we have two different users of audit_sock: kauditd and
> netns, if both take care of refcnt properly, we don't need to worry
> about who is the last, no matter what failures occur in what order.

It is actually the audit_pid and audit_nlk_portid that I care about
more. The audit daemon could vanish or close the socket while the
kernel sock to which it was attached is still quite valid. Accessing
the set of three atomically is the urge. I wonder if it makes more
sense to test for the presence of auditd using audit_sock rather than
audit_pid, but still keep audit_pid for our reporting and replacement
strategy. Another idea would be to put the three in one struct.

Can someone explain how they think the original test was able to trigger
this GPF? Network namespace shutdown while something pretended to set
up a new auditd? That's impressive for a fuzzer if that's the case...
Is there an strace? I guess it is all in test().

Cong Wang

unread,
Dec 13, 2016, 7:17:35 PM12/13/16
to Richard Guy Briggs, linux...@redhat.com, Paul Moore, Dmitry Vyukov, David Miller, Johannes Berg, Florian Westphal, Eric Dumazet, Herbert Xu, netdev, LKML, syzkaller
On Tue, Dec 13, 2016 at 2:52 AM, Richard Guy Briggs <r...@redhat.com> wrote:
> It is actually the audit_pid and audit_nlk_portid that I care about
> more. The audit daemon could vanish or close the socket while the
> kernel sock to which it was attached is still quite valid. Accessing
> the set of three atomically is the urge. I wonder if it makes more
> sense to test for the presence of auditd using audit_sock rather than
> audit_pid, but still keep audit_pid for our reporting and replacement
> strategy. Another idea would be to put the three in one struct.

Note, the process has audit_pid should hold a refcnt to the netns too,
so the netns can't be gone until that process is gone.

>
> Can someone explain how they think the original test was able to trigger
> this GPF? Network namespace shutdown while something pretended to set
> up a new auditd? That's impressive for a fuzzer if that's the case...
> Is there an strace? I guess it is all in test().
>

I am surprised you still don't get the race condition even when you
are now working on v2...

The race happens in this scenarios :

1) Create a new netns

2) In the new netns, communicate with kauditd to set audit_sock

3) Generate some audit messages, so kauditd will keep sending them
via audit_sock

4) exit the netns

5) the previous audit_sock is now going away, but kaudit_sock could still
access it in this small window.

Richard Guy Briggs

unread,
Dec 13, 2016, 11:17:06 PM12/13/16
to Cong Wang, Herbert Xu, Johannes Berg, netdev, Florian Westphal, LKML, Eric Dumazet, linux...@redhat.com, syzkaller, David Miller, Dmitry Vyukov
On 2016-12-13 16:17, Cong Wang wrote:
> On Tue, Dec 13, 2016 at 2:52 AM, Richard Guy Briggs <r...@redhat.com> wrote:
> > It is actually the audit_pid and audit_nlk_portid that I care about
> > more. The audit daemon could vanish or close the socket while the
> > kernel sock to which it was attached is still quite valid. Accessing
> > the set of three atomically is the urge. I wonder if it makes more
> > sense to test for the presence of auditd using audit_sock rather than
> > audit_pid, but still keep audit_pid for our reporting and replacement
> > strategy. Another idea would be to put the three in one struct.
>
> Note, the process has audit_pid should hold a refcnt to the netns too,
> so the netns can't be gone until that process is gone.

I noted that. I did wonder if there might be a problem if all the
processes were moved to another netns with the struct sock stuck in the
now process-void netns.

This is alluded-to in 6f285b19d09f ("audit: Send replies in the proper
network namespace.").

> > Can someone explain how they think the original test was able to trigger
> > this GPF? Network namespace shutdown while something pretended to set
> > up a new auditd? That's impressive for a fuzzer if that's the case...
> > Is there an strace? I guess it is all in test().
>
> I am surprised you still don't get the race condition even when you
> are now working on v2...
>
> The race happens in this scenarios :
>
> 1) Create a new netns
>
> 2) In the new netns, communicate with kauditd to set audit_sock
>
> 3) Generate some audit messages, so kauditd will keep sending them
> via audit_sock
>
> 4) exit the netns
>
> 5) the previous audit_sock is now going away, but kaudit_sock could still
> access it in this small window.

Ah ok that fits...
Reply all
Reply to author
Forward
0 new messages