memory leak in h4_recv_buf

[syzbot]

Hello,

syzbot found the following crash on:

HEAD commit: abf02e29 Merge tag 'pm-5.2-rc6' of git://git.kernel.org/pu..
git tree: upstream
console output: https://syzkaller.appspot.com/x/log.txt?x=1054e6b2a00000
kernel config: https://syzkaller.appspot.com/x/.config?x=56f1da14935c3cce
dashboard link: https://syzkaller.appspot.com/bug?extid=97388eb9d31b997fe1d0
compiler: gcc (GCC) 9.0.0 20181231 (experimental)
syz repro: https://syzkaller.appspot.com/x/repro.syz?x=1073d8aaa00000
C reproducer: https://syzkaller.appspot.com/x/repro.c?x=17b36fbea00000

IMPORTANT: if you fix the bug, please add the following tag to the commit:
Reported-by: syzbot+97388e...@syzkaller.appspotmail.com

program
BUG: memory leak
unreferenced object 0xffff88810991fa00 (size 224):
comm "syz-executor739", pid 7080, jiffies 4294949854 (age 18.640s)
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:
[<00000000da42c09f>] kmemleak_alloc_recursive
include/linux/kmemleak.h:43 [inline]
[<00000000da42c09f>] slab_post_alloc_hook mm/slab.h:439 [inline]
[<00000000da42c09f>] slab_alloc_node mm/slab.c:3269 [inline]
[<00000000da42c09f>] kmem_cache_alloc_node+0x153/0x2a0 mm/slab.c:3579
[<00000000f6fbcf84>] __alloc_skb+0x6e/0x210 net/core/skbuff.c:194
[<00000000ea93fc4c>] alloc_skb include/linux/skbuff.h:1054 [inline]
[<00000000ea93fc4c>] bt_skb_alloc include/net/bluetooth/bluetooth.h:339
[inline]
[<00000000ea93fc4c>] h4_recv_buf+0x26d/0x450
drivers/bluetooth/hci_h4.c:182
[<00000000e0312475>] h4_recv+0x51/0xb0 drivers/bluetooth/hci_h4.c:116
[<00000000ebf11fab>] hci_uart_tty_receive+0xba/0x200
drivers/bluetooth/hci_ldisc.c:592
[<0000000095e1216e>] tiocsti drivers/tty/tty_io.c:2195 [inline]
[<0000000095e1216e>] tty_ioctl+0x81c/0xa30 drivers/tty/tty_io.c:2571
[<000000009fa523f0>] vfs_ioctl fs/ioctl.c:46 [inline]
[<000000009fa523f0>] file_ioctl fs/ioctl.c:509 [inline]
[<000000009fa523f0>] do_vfs_ioctl+0x62a/0x810 fs/ioctl.c:696
[<000000000cebb5d9>] ksys_ioctl+0x86/0xb0 fs/ioctl.c:713
[<000000001630008a>] __do_sys_ioctl fs/ioctl.c:720 [inline]
[<000000001630008a>] __se_sys_ioctl fs/ioctl.c:718 [inline]
[<000000001630008a>] __x64_sys_ioctl+0x1e/0x30 fs/ioctl.c:718
[<00000000c62091e3>] do_syscall_64+0x76/0x1a0
arch/x86/entry/common.c:301
[<000000005c213625>] entry_SYSCALL_64_after_hwframe+0x44/0xa9

BUG: memory leak
unreferenced object 0xffff8881204f4400 (size 1024):
comm "syz-executor739", pid 7080, jiffies 4294949854 (age 18.640s)
hex dump (first 32 bytes):
6c 69 62 75 64 65 76 00 fe ed ca fe 28 00 00 00 libudev.....(...
28 00 00 00 a0 00 00 00 52 ca da 77 00 00 00 00 (.......R..w....
backtrace:
[<0000000034504843>] kmemleak_alloc_recursive
include/linux/kmemleak.h:43 [inline]
[<0000000034504843>] slab_post_alloc_hook mm/slab.h:439 [inline]
[<0000000034504843>] slab_alloc_node mm/slab.c:3269 [inline]
[<0000000034504843>] kmem_cache_alloc_node_trace+0x15b/0x2a0
mm/slab.c:3597
[<0000000056d30eb5>] __do_kmalloc_node mm/slab.c:3619 [inline]
[<0000000056d30eb5>] __kmalloc_node_track_caller+0x38/0x50
mm/slab.c:3634
[<00000000df40176c>] __kmalloc_reserve.isra.0+0x40/0xb0
net/core/skbuff.c:138
[<0000000035340e64>] __alloc_skb+0xa0/0x210 net/core/skbuff.c:206
[<00000000ea93fc4c>] alloc_skb include/linux/skbuff.h:1054 [inline]
[<00000000ea93fc4c>] bt_skb_alloc include/net/bluetooth/bluetooth.h:339
[inline]
[<00000000ea93fc4c>] h4_recv_buf+0x26d/0x450
drivers/bluetooth/hci_h4.c:182
[<00000000e0312475>] h4_recv+0x51/0xb0 drivers/bluetooth/hci_h4.c:116
[<00000000ebf11fab>] hci_uart_tty_receive+0xba/0x200
drivers/bluetooth/hci_ldisc.c:592
[<0000000095e1216e>] tiocsti drivers/tty/tty_io.c:2195 [inline]
[<0000000095e1216e>] tty_ioctl+0x81c/0xa30 drivers/tty/tty_io.c:2571
[<000000009fa523f0>] vfs_ioctl fs/ioctl.c:46 [inline]
[<000000009fa523f0>] file_ioctl fs/ioctl.c:509 [inline]
[<000000009fa523f0>] do_vfs_ioctl+0x62a/0x810 fs/ioctl.c:696
[<000000000cebb5d9>] ksys_ioctl+0x86/0xb0 fs/ioctl.c:713
[<000000001630008a>] __do_sys_ioctl fs/ioctl.c:720 [inline]
[<000000001630008a>] __se_sys_ioctl fs/ioctl.c:718 [inline]
[<000000001630008a>] __x64_sys_ioctl+0x1e/0x30 fs/ioctl.c:718
[<00000000c62091e3>] do_syscall_64+0x76/0x1a0
arch/x86/entry/common.c:301
[<000000005c213625>] entry_SYSCALL_64_after_hwframe+0x44/0xa9



---
This bug 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 bug report. See:
https://goo.gl/tpsmEJ#status for how to communicate with syzbot.
syzbot can test patches for this bug, for details see:
https://goo.gl/tpsmEJ#testing-patches

[Phi Nguyen Dinh]

The reason of this memory leak is tty_ldisc_receive_buf() and tiocsti() can access the h4->rx_skb concurrently by calling hci_uart_tty_receive(), so the rx_skb be overwritten without deallocating. There used to be an spin_lock in hci_uart_tty_receive(), but it was removed by commit 7649ffaff1cfe(Bluetooth: Remove useless rx_lock spinlock). 

The commit message claims that hci_uart_tty_receive() was only called by flush_to_ldisc(), but it seems incorrect.

Vào lúc 15:27:09 UTC+8 ngày Thứ Hai, 24 tháng 6, 2019, syzbot đã viết:

[Phi Nguyen]

> Hello,
>
> syzbot found the following crash on:
>
> HEAD commit: abf02e29 Merge tag 'pm-5.2-rc6' of

> git://git.kernel.org/pu <http://git.kernel.org/pu>..

> See https://goo.gl/tpsmEJ <https://goo.gl/tpsmEJ> for more

> information about syzbot.
> syzbot engineers can be reached at syzk...@googlegroups.com.
>
> syzbot will keep track of this bug report. See:

> https://goo.gl/tpsmEJ#status <https://goo.gl/tpsmEJ#status> for how

> to communicate with syzbot.
> syzbot can test patches for this bug, for details see:
> https://goo.gl/tpsmEJ#testing-patches

> <https://goo.gl/tpsmEJ#testing-patches>
>
> --
> You received this message because you are subscribed to the Google
> Groups "syzkaller-bugs" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to syzkaller-bug...@googlegroups.com
> <mailto:syzkaller-bug...@googlegroups.com>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/syzkaller-bugs/d71a274f-fdeb-4da1-898e-06f6944e04dan%40googlegroups.com
> <https://groups.google.com/d/msgid/syzkaller-bugs/d71a274f-fdeb-4da1-898e-06f6944e04dan%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Dan Carpenter]

On Tue, Jul 27, 2021 at 07:55:01AM -0700, Phi Nguyen Dinh wrote:
>
>
> The reason of this memory leak is tty_ldisc_receive_buf() and tiocsti() can
> access the h4->rx_skb concurrently by calling hci_uart_tty_receive(), so
> the rx_skb be overwritten without deallocating. There used to be an
> spin_lock in hci_uart_tty_receive(), but it was removed by commit
> 7649ffaff1cfe(Bluetooth: Remove useless rx_lock spinlock).

Typo in git hash.

7649faff1cfe4

> The commit message claims that hci_uart_tty_receive() was only called by
> flush_to_ldisc(), but it seems incorrect.

That seems reasonable. Yes. Just send a patch to add it back.

regards,
dan carpenter

[Marcel Holtmann]

Hi Phi,

I don’t have that commit in my Linus’ tree. Where is it?

> The commit message claims that hci_uart_tty_receive() was only called by
> flush_to_ldisc(), but it seems incorrect.

That seems to be a larger problem in the TTY layer if its contract with its users have changed.

Regards

Marcel

[Phi Nguyen]

Hi Marcel,

It was my typo, the git hash should be 7649faff1cfe4.

Regards,
Phi.

[Dan Carpenter]

There is a typo in the git hash. It should be: 7649faff1cfe4 ("Bluetooth:
Remove useless rx_lock spinlock").

> > The commit message claims that hci_uart_tty_receive() was only called by
> > flush_to_ldisc(), but it seems incorrect.
>
> That seems to be a larger problem in the TTY layer if its contract with its users have changed.

The tiocsti() function has an ancient comment which suggests that the
documentation has always been wrong.

* FIXME: may race normal receive processing

regards,
dan carpenter

[Marcel Holtmann]

Hi Dan,

so what are we suppose to do now? Fix this in TTY layer or try to revert this patch?

And does it have to be spinlock or can we use a mutex? My knowledge of the TTY internal are limited and thus, I have no idea what we need to do here. However h4_recv_buf needs to be protected against concurrently calls.

Regards

Marcel

[Phi Nguyen]

Hi Marcel,

So far, I have tested two fixes with syzbot
The first one is to bring the spin lock back to hci_uart_tty_receive().
The second one is to use tty_buffer_lock_exclusive() in tiocsti() (I
based on the document in tty_buffer.c).
These two can work well with the syzbot.

Regards,
Phi

[syzbot]

Hello,

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

Reported-and-tested-by: syzbot+97388e...@syzkaller.appspotmail.com

Tested on:

commit: 7e96bf47 Merge tag 'for-linus' of git://git.kernel.org..
git tree: upstream
kernel config: https://syzkaller.appspot.com/x/.config?x=b569cddf2de2a96d
dashboard link: https://syzkaller.appspot.com/bug?extid=97388eb9d31b997fe1d0
compiler: gcc (Debian 10.2.1-6) 10.2.1 20210110, GNU ld (GNU Binutils for Debian) 2.35.1
patch: https://syzkaller.appspot.com/x/patch.diff?x=13982746300000

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

[Marcel Holtmann]

Hi Phi,

if we can fix it in the TTY layer, then that is my preference. If we can’t, then it would be good to know if we can use a mutex instead of spinlock.

Regards

Marcel