[syzbot] [kvm?] [net?] [virt?] general protection fault in virtio_transport_purge_skbs

[syzbot]

Hello,

syzbot found the following issue on:

HEAD commit: fff5a5e7f528 Merge tag 'for-linus' of git://git.armlinux.o..
git tree: upstream
console+strace: https://syzkaller.appspot.com/x/log.txt?x=1136e97ac80000
kernel config: https://syzkaller.appspot.com/x/.config?x=aaa4b45720ca0519
dashboard link: https://syzkaller.appspot.com/bug?extid=befff0a9536049e7902e
compiler: gcc (Debian 10.2.1-6) 10.2.1 20210110, GNU ld (GNU Binutils for Debian) 2.35.2
syz repro: https://syzkaller.appspot.com/x/repro.syz?x=14365781c80000
C reproducer: https://syzkaller.appspot.com/x/repro.c?x=12eebc66c80000

Downloadable assets:
disk image: https://storage.googleapis.com/syzbot-assets/997791f5f9e1/disk-fff5a5e7.raw.xz
vmlinux: https://storage.googleapis.com/syzbot-assets/0b0155b5eac1/vmlinux-fff5a5e7.xz
kernel image: https://storage.googleapis.com/syzbot-assets/8d98dd2ba6b6/bzImage-fff5a5e7.xz

IMPORTANT: if you fix the issue, please add the following tag to the commit:
Reported-by: syzbot+befff0...@syzkaller.appspotmail.com

general protection fault, probably for non-canonical address 0xdffffc0000000000: 0000 [#1] PREEMPT SMP KASAN
KASAN: null-ptr-deref in range [0x0000000000000000-0x0000000000000007]
CPU: 0 PID: 8759 Comm: syz-executor379 Not tainted 6.3.0-rc3-syzkaller-00026-gfff5a5e7f528 #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 03/02/2023
RIP: 0010:virtio_transport_purge_skbs+0x139/0x4c0 net/vmw_vsock/virtio_transport_common.c:1370
Code: 00 00 00 00 fc ff df 48 89 c2 48 89 44 24 28 48 c1 ea 03 48 8d 04 1a 48 89 44 24 10 eb 29 e8 ee 27 a3 f7 48 89 e8 48 c1 e8 03 <80> 3c 18 00 0f 85 a6 02 00 00 49 39 ec 48 8b 55 00 49 89 ef 0f 84
RSP: 0018:ffffc90006427b48 EFLAGS: 00010256
RAX: 0000000000000000 RBX: dffffc0000000000 RCX: 0000000000000000
RDX: ffff8880211157c0 RSI: ffffffff89dfbd12 RDI: ffff88802c11a018
RBP: 0000000000000000 R08: 0000000000000001 R09: 0000000000000003
R10: fffff52000c84f5b R11: 0000000000000000 R12: ffffffff92179188
R13: ffffc90006427ba0 R14: ffff88801e0f1100 R15: ffff88802c11a000
FS: 00007f01fdd51700(0000) GS:ffff8880b9800000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007f01fdd30718 CR3: 000000002a3f9000 CR4: 0000000000350ef0
Call Trace:
<TASK>
vsock_loopback_cancel_pkt+0x1c/0x20 net/vmw_vsock/vsock_loopback.c:48
vsock_transport_cancel_pkt net/vmw_vsock/af_vsock.c:1284 [inline]
vsock_connect+0x852/0xcc0 net/vmw_vsock/af_vsock.c:1426
__sys_connect_file+0x153/0x1a0 net/socket.c:2001
__sys_connect+0x165/0x1a0 net/socket.c:2018
__do_sys_connect net/socket.c:2028 [inline]
__se_sys_connect net/socket.c:2025 [inline]
__x64_sys_connect+0x73/0xb0 net/socket.c:2025
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x39/0xb0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x63/0xcd
RIP: 0033:0x7f01fdda0159
Code: 28 00 00 00 75 05 48 83 c4 28 c3 e8 41 15 00 00 90 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 b8 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007f01fdd51308 EFLAGS: 00000246 ORIG_RAX: 000000000000002a
RAX: ffffffffffffffda RBX: 00007f01fde28428 RCX: 00007f01fdda0159
RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000004
RBP: 00007f01fde28420 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 00007f01fddf606c
R13: 0000000000000000 R14: 00007f01fdd51400 R15: 0000000000022000
</TASK>
Modules linked in:
---[ end trace 0000000000000000 ]---
RIP: 0010:virtio_transport_purge_skbs+0x139/0x4c0 net/vmw_vsock/virtio_transport_common.c:1370
Code: 00 00 00 00 fc ff df 48 89 c2 48 89 44 24 28 48 c1 ea 03 48 8d 04 1a 48 89 44 24 10 eb 29 e8 ee 27 a3 f7 48 89 e8 48 c1 e8 03 <80> 3c 18 00 0f 85 a6 02 00 00 49 39 ec 48 8b 55 00 49 89 ef 0f 84
RSP: 0018:ffffc90006427b48 EFLAGS: 00010256
RAX: 0000000000000000 RBX: dffffc0000000000 RCX: 0000000000000000
RDX: ffff8880211157c0 RSI: ffffffff89dfbd12 RDI: ffff88802c11a018
RBP: 0000000000000000 R08: 0000000000000001 R09: 0000000000000003
R10: fffff52000c84f5b R11: 0000000000000000 R12: ffffffff92179188
R13: ffffc90006427ba0 R14: ffff88801e0f1100 R15: ffff88802c11a000
FS: 00007f01fdd51700(0000) GS:ffff8880b9800000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007f01fdd30718 CR3: 000000002a3f9000 CR4: 0000000000350ef0
----------------
Code disassembly (best guess), 6 bytes skipped:
0: df 48 89 fisttps -0x77(%rax)
3: c2 48 89 retq $0x8948
6: 44 24 28 rex.R and $0x28,%al
9: 48 c1 ea 03 shr $0x3,%rdx
d: 48 8d 04 1a lea (%rdx,%rbx,1),%rax
11: 48 89 44 24 10 mov %rax,0x10(%rsp)
16: eb 29 jmp 0x41
18: e8 ee 27 a3 f7 callq 0xf7a3280b
1d: 48 89 e8 mov %rbp,%rax
20: 48 c1 e8 03 shr $0x3,%rax
* 24: 80 3c 18 00 cmpb $0x0,(%rax,%rbx,1) <-- trapping instruction
28: 0f 85 a6 02 00 00 jne 0x2d4
2e: 49 39 ec cmp %rbp,%r12
31: 48 8b 55 00 mov 0x0(%rbp),%rdx
35: 49 89 ef mov %rbp,%r15
38: 0f .byte 0xf
39: 84 .byte 0x84


---
This report is generated by a bot. It may contain errors.
See https://goo.gl/tpsmEJ for more information about syzbot.
syzbot engineers can be reached at syzk...@googlegroups.com.

syzbot will keep track of this issue. See:
https://goo.gl/tpsmEJ#status for how to communicate with syzbot.
syzbot can test patches for this issue, for details see:
https://goo.gl/tpsmEJ#testing-patches

[Hillf Danton]

On 23 Mar 2023 17:52:40 -0700

> HEAD commit: fff5a5e7f528 Merge tag 'for-linus' of git://git.armlinux.o..
> git tree: upstream

> C reproducer: https://syzkaller.appspot.com/x/repro.c?x=12eebc66c80000

Splice skb before purging it.

#syz test https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git fff5a5e7f528

--- x/net/vmw_vsock/vsock_loopback.c
+++ y/net/vmw_vsock/vsock_loopback.c
@@ -44,8 +44,15 @@ static int vsock_loopback_send_pkt(struc
static int vsock_loopback_cancel_pkt(struct vsock_sock *vsk)
{
struct vsock_loopback *vsock = &the_vsock_loopback;
+ struct sk_buff_head pkts;

- virtio_transport_purge_skbs(vsk, &vsock->pkt_queue);
+ skb_queue_head_init(&pkts);
+
+ spin_lock_bh(&vsock->pkt_list_lock);
+ skb_queue_splice_init(&vsock->pkt_queue, &pkts);
+ spin_unlock_bh(&vsock->pkt_list_lock);
+
+ virtio_transport_purge_skbs(vsk, &pkts);

return 0;
}
--

[syzbot]

Hello,

syzbot has tested the proposed patch and the reproducer did not trigger any issue:

Reported-and-tested-by: syzbot+befff0...@syzkaller.appspotmail.com

Tested on:

commit: fff5a5e7 Merge tag 'for-linus' of git://git.armlinux.o..
git tree: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
console output: https://syzkaller.appspot.com/x/log.txt?x=10b96eeec80000
kernel config: https://syzkaller.appspot.com/x/.config?x=ea09b0836073ee4

dashboard link: https://syzkaller.appspot.com/bug?extid=befff0a9536049e7902e
compiler: gcc (Debian 10.2.1-6) 10.2.1 20210110, GNU ld (GNU Binutils for Debian) 2.35.2

patch: https://syzkaller.appspot.com/x/patch.diff?x=15ad956ec80000

Note: testing is done by a robot and is best-effort only.

[Stefano Garzarella]

Hi Bobby,
can you take a look at this report?

It seems related to the changes we made to support skbuff.

Thanks,
Stefano

[Stefano Garzarella]

On Fri, Mar 24, 2023 at 9:31 AM Stefano Garzarella <sgar...@redhat.com> wrote:
>
> Hi Bobby,
> can you take a look at this report?
>
> It seems related to the changes we made to support skbuff.

Could it be a problem of concurrent access to pkt_queue ?

IIUC we should hold pkt_queue.lock when we call skb_queue_splice_init()
and remove pkt_list_lock. (or hold pkt_list_lock when calling
virtio_transport_purge_skbs, but pkt_list_lock seems useless now that
we use skbuff)

#syz test https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git fff5a5e7f528

--- a/net/vmw_vsock/vsock_loopback.c
+++ b/net/vmw_vsock/vsock_loopback.c
@@ -15,7 +15,6 @@
struct vsock_loopback {
struct workqueue_struct *workqueue;

- spinlock_t pkt_list_lock; /* protects pkt_list */
struct sk_buff_head pkt_queue;
struct work_struct pkt_work;
};
@@ -32,9 +31,7 @@ static int vsock_loopback_send_pkt(struct sk_buff *skb)

struct vsock_loopback *vsock = &the_vsock_loopback;

int len = skb->len;

- spin_lock_bh(&vsock->pkt_list_lock);
skb_queue_tail(&vsock->pkt_queue, skb);
- spin_unlock_bh(&vsock->pkt_list_lock);

queue_work(vsock->workqueue, &vsock->pkt_work);

@@ -113,9 +110,9 @@ static void vsock_loopback_work(struct work_struct *work)

skb_queue_head_init(&pkts);

- spin_lock_bh(&vsock->pkt_list_lock);
+ spin_lock_bh(&vsock->pkt_queue.lock);
skb_queue_splice_init(&vsock->pkt_queue, &pkts);
- spin_unlock_bh(&vsock->pkt_list_lock);
+ spin_unlock_bh(&vsock->pkt_queue.lock);

while ((skb = __skb_dequeue(&pkts))) {
virtio_transport_deliver_tap_pkt(skb);
@@ -132,7 +129,6 @@ static int __init vsock_loopback_init(void)
if (!vsock->workqueue)
return -ENOMEM;

- spin_lock_init(&vsock->pkt_list_lock);
skb_queue_head_init(&vsock->pkt_queue);
INIT_WORK(&vsock->pkt_work, vsock_loopback_work);

[Arseniy Krasnov]

On 24.03.2023 12:06, Stefano Garzarella wrote:

> On Fri, Mar 24, 2023 at 9:55 AM Stefano Garzarella <sgar...@redhat.com> wrote:
>>
>> On Fri, Mar 24, 2023 at 9:31 AM Stefano Garzarella <sgar...@redhat.com> wrote:
>>>
>>> Hi Bobby,
>>> can you take a look at this report?
>>>
>>> It seems related to the changes we made to support skbuff.
>>
>> Could it be a problem of concurrent access to pkt_queue ?
>>
>> IIUC we should hold pkt_queue.lock when we call skb_queue_splice_init()
>> and remove pkt_list_lock. (or hold pkt_list_lock when calling
>> virtio_transport_purge_skbs, but pkt_list_lock seems useless now that
>> we use skbuff)
>>
>

> In the previous patch was missing a hunk, new one attached:

>
> #syz test https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git fff5a5e7f528
>
> --- a/net/vmw_vsock/vsock_loopback.c
> +++ b/net/vmw_vsock/vsock_loopback.c
> @@ -15,7 +15,6 @@
> struct vsock_loopback {
> struct workqueue_struct *workqueue;
>
> - spinlock_t pkt_list_lock; /* protects pkt_list */
> struct sk_buff_head pkt_queue;
> struct work_struct pkt_work;
> };
> @@ -32,9 +31,7 @@ static int vsock_loopback_send_pkt(struct sk_buff *skb)
> struct vsock_loopback *vsock = &the_vsock_loopback;
> int len = skb->len;
>
> - spin_lock_bh(&vsock->pkt_list_lock);
> skb_queue_tail(&vsock->pkt_queue, skb);

Hello Stefano and Bobby,

Small remark, may be here we can use virtio_vsock_skb_queue_tail() instead of skb_queue_tail().
skb_queue_tail() disables irqs during spinlock access, while virtio_vsock_skb_queue_tail()
uses spin_lock_bh(). vhost and virtio transports use virtio_vsock_skb_queue_tail().

Thanks, Arseniy

> - spin_unlock_bh(&vsock->pkt_list_lock);
>
> queue_work(vsock->workqueue, &vsock->pkt_work);
>
> @@ -113,9 +110,9 @@ static void vsock_loopback_work(struct work_struct *work)
>
> skb_queue_head_init(&pkts);
>
> - spin_lock_bh(&vsock->pkt_list_lock);
> + spin_lock_bh(&vsock->pkt_queue.lock);
> skb_queue_splice_init(&vsock->pkt_queue, &pkts);
> - spin_unlock_bh(&vsock->pkt_list_lock);
> + spin_unlock_bh(&vsock->pkt_queue.lock);
>
> while ((skb = __skb_dequeue(&pkts))) {
> virtio_transport_deliver_tap_pkt(skb);
> @@ -132,7 +129,6 @@ static int __init vsock_loopback_init(void)
> if (!vsock->workqueue)
> return -ENOMEM;
>
> - spin_lock_init(&vsock->pkt_list_lock);
> skb_queue_head_init(&vsock->pkt_queue);
> INIT_WORK(&vsock->pkt_work, vsock_loopback_work);
>

> @@ -156,9 +152,7 @@ static void __exit vsock_loopback_exit(void)
>
> flush_work(&vsock->pkt_work);
>
> - spin_lock_bh(&vsock->pkt_list_lock);
> virtio_vsock_skb_queue_purge(&vsock->pkt_queue);
> - spin_unlock_bh(&vsock->pkt_list_lock);
>
> destroy_workqueue(vsock->workqueue);
> }
>

[Stefano Garzarella]

On Fri, Mar 24, 2023 at 9:55 AM Stefano Garzarella <sgar...@redhat.com> wrote:
>
> On Fri, Mar 24, 2023 at 9:31 AM Stefano Garzarella <sgar...@redhat.com> wrote:
> >
> > Hi Bobby,
> > can you take a look at this report?
> >
> > It seems related to the changes we made to support skbuff.
>
> Could it be a problem of concurrent access to pkt_queue ?
>
> IIUC we should hold pkt_queue.lock when we call skb_queue_splice_init()
> and remove pkt_list_lock. (or hold pkt_list_lock when calling
> virtio_transport_purge_skbs, but pkt_list_lock seems useless now that
> we use skbuff)
>

In the previous patch was missing a hunk, new one attached:

[syzbot]

Hello,

syzbot tried to test the proposed patch but the build/boot failed:

net/vmw_vsock/vsock_loopback.c:155:21: error: 'struct vsock_loopback' has no member named 'pkt_list_lock'
net/vmw_vsock/vsock_loopback.c:157:23: error: 'struct vsock_loopback' has no member named 'pkt_list_lock'


Tested on:

commit: fff5a5e7 Merge tag 'for-linus' of git://git.armlinux.o..
git tree: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git

dashboard link: https://syzkaller.appspot.com/bug?extid=befff0a9536049e7902e
compiler: gcc (Debian 10.2.1-6) 10.2.1 20210110, GNU ld (GNU Binutils for Debian) 2.35.2

patch: https://syzkaller.appspot.com/x/patch.diff?x=15ed6191c80000

[syzbot]

Hello,

syzbot tried to test the proposed patch but the build/boot failed:

failed to apply patch:
checking file net/vmw_vsock/vsock_loopback.c
patch: **** unexpected end of file in patch

Tested on:

commit: fff5a5e7 Merge tag 'for-linus' of git://git.armlinux.o..
git tree: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
dashboard link: https://syzkaller.appspot.com/bug?extid=befff0a9536049e7902e
compiler:

patch: https://syzkaller.appspot.com/x/patch.diff?x=16b4bba1c80000

[Stefano Garzarella]

Yep, but this shouldn't be related.
I would make this change in a separate patch. ;-)

Thanks,
Stefano

[syzbot]

Hello,

syzbot has tested the proposed patch and the reproducer did not trigger any issue:

Reported-and-tested-by: syzbot+befff0...@syzkaller.appspotmail.com

Tested on:

commit: fff5a5e7 Merge tag 'for-linus' of git://git.armlinux.o..
git tree: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git

console output: https://syzkaller.appspot.com/x/log.txt?x=13ee5fa1c80000
kernel config: https://syzkaller.appspot.com/x/.config?x=ea09b0836073ee4
dashboard link: https://syzkaller.appspot.com/bug?extid=befff0a9536049e7902e

compiler: gcc (Debian 10.2.1-6) 10.2.1 20210110, GNU ld (GNU Binutils for Debian) 2.35.2

patch: https://syzkaller.appspot.com/x/patch.diff?x=17b4c6eec80000

[Stefano Garzarella]

On Fri, Mar 24, 2023 at 10:06 AM Stefano Garzarella <sgar...@redhat.com> wrote:
>
> On Fri, Mar 24, 2023 at 9:55 AM Stefano Garzarella <sgar...@redhat.com> wrote:
> >
> > On Fri, Mar 24, 2023 at 9:31 AM Stefano Garzarella <sgar...@redhat.com> wrote:
> > >
> > > Hi Bobby,
> > > can you take a look at this report?
> > >
> > > It seems related to the changes we made to support skbuff.
> >
> > Could it be a problem of concurrent access to pkt_queue ?
> >
> > IIUC we should hold pkt_queue.lock when we call skb_queue_splice_init()
> > and remove pkt_list_lock. (or hold pkt_list_lock when calling
> > virtio_transport_purge_skbs, but pkt_list_lock seems useless now that
> > we use skbuff)
> >
>

Patch posted here:
https://lore.kernel.org/netdev/20230324115450....@redhat.com/

[syzbot]

syzbot has bisected this issue to:

commit 71dc9ec9ac7d3eee785cdc986c3daeb821381e20
Author: Bobby Eshleman <bobby.e...@bytedance.com>
Date: Fri Jan 13 22:21:37 2023 +0000

virtio/vsock: replace virtio_vsock_pkt with sk_buff

bisection log: https://syzkaller.appspot.com/x/bisect.txt?x=12661f29c80000
start commit: fff5a5e7f528 Merge tag 'for-linus' of git://git.armlinux.o..
git tree: upstream
final oops: https://syzkaller.appspot.com/x/report.txt?x=11661f29c80000
console output: https://syzkaller.appspot.com/x/log.txt?x=16661f29c80000

kernel config: https://syzkaller.appspot.com/x/.config?x=aaa4b45720ca0519
dashboard link: https://syzkaller.appspot.com/bug?extid=befff0a9536049e7902e

syz repro: https://syzkaller.appspot.com/x/repro.syz?x=14365781c80000
C reproducer: https://syzkaller.appspot.com/x/repro.c?x=12eebc66c80000

Reported-by: syzbot+befff0...@syzkaller.appspotmail.com
Fixes: 71dc9ec9ac7d ("virtio/vsock: replace virtio_vsock_pkt with sk_buff")

For information about bisection process see: https://goo.gl/tpsmEJ#bisection