[Bug 120691] New: UART HCI memory leak

[Bug 120691] New: UART HCI memory leak

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=120691

            Bug ID: 120691
           Summary: UART HCI memory leak
           Product: Drivers
           Version: 2.5
    Kernel Version: 4.7-rc3
          Hardware: ARM
                OS: Linux
              Tree: Mainline
            Status: NEW
          Severity: normal
          Priority: P1
         Component: Bluetooth
          Assignee: linux-bluetooth@vger.kernel.org
          Reporter: nico.edev@gmail.com
        Regression: No

Hello,

It looks like there is a memory leak on UART HCI driver. I am using kernel
4.7-rc3 and Bluez 5.40. I can reproduce the issue with kernel 4.2.
There is nothing special to do to encounter the problem; HCI traffic is enough.
I can speed up the occurrence of the issue when BT module is scanning because
it increases HCI traffic.
My BT module is dual mode but I can reproduce the issue when I force "brerd" or
"le" mode.
FYI, the leak is 1MByte/hour when BT scanning is on; which is a lot on embedded
systems.

Below is a piece of kmemleak dump:

unreferenced object 0xc6a59ac0 (size 2048):
  comm "kworker/u2:0", pid 6, jiffies 4294951225 (age 1195.920s)
  hex dump (first 32 bytes):
    6b 6b 6b 6b 6b 6b 6b 6b 0e 0e 01 04 10 00 01 01  kkkkkkkk........
    00 00 00 00 00 00 00 00 6b 6b 6b 6b 6b 6b 6b 6b  ........kkkkkkkk
  backtrace:
    [<c03c93a0>] __alloc_skb+0x7c/0x164
    [<c03563ac>] ll_recv+0x1c8/0x41c
    [<c03554b4>] hci_uart_tty_receive+0x44/0x64
    [<c0255ec4>] tty_ldisc_receive_buf+0x50/0x58
    [<c0256338>] flush_to_ldisc+0xb8/0xd0
    [<c00318c0>] process_one_work+0x128/0x478
    [<c0031c64>] worker_thread+0x54/0x574
    [<c00368f4>] kthread+0xc0/0xdc
    [<c000a2d0>] ret_from_fork+0x14/0x24
    [<ffffffff>] 0xffffffff
unreferenced object 0xc62f0020 (size 168):
  comm "kworker/u3:2", pid 439, jiffies 4294951225 (age 1195.920s)
  hex dump (first 32 bytes):
    00 00 00 00 00 00 00 00 54 a4 46 a9 95 53 54 14  ........T.F..ST.
    00 00 00 00 00 00 00 00 04 00 00 00 01 00 00 00  ................
  backtrace:
    [<c0473bfc>] hci_event_packet+0x1d8/0x2ba0
    [<c04674c8>] hci_rx_work+0x170/0x248
    [<c00318c0>] process_one_work+0x128/0x478
    [<c0031c64>] worker_thread+0x54/0x574
    [<c00368f4>] kthread+0xc0/0xdc
    [<c000a2d0>] ret_from_fork+0x14/0x24
    [<ffffffff>] 0xffffffff
unreferenced object 0xc6a5be40 (size 2048):
  comm "kworker/u2:0", pid 6, jiffies 4294951227 (age 1195.900s)
  hex dump (first 32 bytes):
    6b 6b 6b 6b 6b 6b 6b 6b 0e 04 01 10 20 00 6b 6b  kkkkkkkk.... .kk
    6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b  kkkkkkkkkkkkkkkk
  backtrace:
    [<c03c93a0>] __alloc_skb+0x7c/0x164
    [<c03563ac>] ll_recv+0x1c8/0x41c
    [<c03554b4>] hci_uart_tty_receive+0x44/0x64
    [<c0255ec4>] tty_ldisc_receive_buf+0x50/0x58
    [<c0256338>] flush_to_ldisc+0xb8/0xd0
    [<c00318c0>] process_one_work+0x128/0x478
    [<c0031c64>] worker_thread+0x54/0x574
    [<c00368f4>] kthread+0xc0/0xdc
    [<c000a2d0>] ret_from_fork+0x14/0x24
    [<ffffffff>] 0xffffffff
unreferenced object 0xc6315da0 (size 168):
  comm "kworker/u3:2", pid 439, jiffies 4294951227 (age 1195.900s)
  hex dump (first 32 bytes):
    00 00 00 00 00 00 00 00 5d c6 6d aa 95 53 54 14  ........].m..ST.
    00 00 00 00 00 00 00 00 04 00 00 00 01 00 00 00  ................
  backtrace:
    [<c0473bfc>] hci_event_packet+0x1d8/0x2ba0
    [<c04674c8>] hci_rx_work+0x170/0x248
    [<c00318c0>] process_one_work+0x128/0x478
    [<c0031c64>] worker_thread+0x54/0x574
    [<c00368f4>] kthread+0xc0/0xdc
    [<c000a2d0>] ret_from_fork+0x14/0x24
    [<ffffffff>] 0xffffffff
unreferenced object 0xc6a5a3a0 (size 2048):
  comm "kworker/u2:0", pid 6, jiffies 4294951228 (age 1195.890s)
  hex dump (first 32 bytes):
    6b 6b 6b 6b 6b 6b 6b 6b 0e 06 01 12 0c 00 00 00  kkkkkkkk........
    6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b  kkkkkkkkkkkkkkkk
  backtrace:
    [<c03c93a0>] __alloc_skb+0x7c/0x164
    [<c03563ac>] ll_recv+0x1c8/0x41c
    [<c03554b4>] hci_uart_tty_receive+0x44/0x64
    [<c0255ec4>] tty_ldisc_receive_buf+0x50/0x58
    [<c0256338>] flush_to_ldisc+0xb8/0xd0
    [<c00318c0>] process_one_work+0x128/0x478
    [<c0031c64>] worker_thread+0x54/0x574
    [<c00368f4>] kthread+0xc0/0xdc
    [<c000a2d0>] ret_from_fork+0x14/0x24
    [<ffffffff>] 0xffffffff
unreferenced object 0xc6315620 (size 168):
  comm "kworker/u3:2", pid 439, jiffies 4294951228 (age 1195.890s)
  hex dump (first 32 bytes):
    00 00 00 00 00 00 00 00 1a 97 8f aa 95 53 54 14  .............ST.
    00 00 00 00 00 00 00 00 04 00 00 00 01 00 00 00  ................
  backtrace:
    [<c0473bfc>] hci_event_packet+0x1d8/0x2ba0
    [<c04674c8>] hci_rx_work+0x170/0x248
    [<c00318c0>] process_one_work+0x128/0x478
    [<c0031c64>] worker_thread+0x54/0x574
    [<c00368f4>] kthread+0xc0/0xdc
    [<c000a2d0>] ret_from_fork+0x14/0x24
    [<ffffffff>] 0xffffffff
unreferenced object 0xc63151a0 (size 168):
  comm "kworker/u3:0", pid 435, jiffies 4294953313 (age 1175.050s)
  hex dump (first 32 bytes):
    00 00 00 00 00 00 00 00 86 3f ce 85 9a 53 54 14  .........?...ST.
    00 00 00 00 00 00 00 00 04 00 00 00 01 00 00 00  ................
  backtrace:
    [<c0473bfc>] hci_event_packet+0x1d8/0x2ba0
    [<c04674c8>] hci_rx_work+0x170/0x248
    [<c00318c0>] process_one_work+0x128/0x478
    [<c0031c64>] worker_thread+0x54/0x574
    [<c00368f4>] kthread+0xc0/0xdc
    [<c000a2d0>] ret_from_fork+0x14/0x24
    [<ffffffff>] 0xffffffff
unreferenced object 0xc6a5ac80 (size 2048):
  comm "kworker/u2:0", pid 6, jiffies 4294958830 (age 1119.880s)
  hex dump (first 32 bytes):
    6b 6b 6b 6b 6b 6b 6b 6b 0e 0a 01 09 10 00 f8 a7  kkkkkkkk........
    d7 e9 17 00 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b  ....kkkkkkkkkkkk
  backtrace:
    [<c03c93a0>] __alloc_skb+0x7c/0x164
    [<c03563ac>] ll_recv+0x1c8/0x41c
    [<c03554b4>] hci_uart_tty_receive+0x44/0x64
    [<c0255ec4>] tty_ldisc_receive_buf+0x50/0x58
    [<c0256338>] flush_to_ldisc+0xb8/0xd0
    [<c00318c0>] process_one_work+0x128/0x478
    [<c0031c64>] worker_thread+0x54/0x574
    [<c00368f4>] kthread+0xc0/0xdc
    [<c000a2d0>] ret_from_fork+0x14/0x24
    [<ffffffff>] 0xffffffff
unreferenced object 0xc6315aa0 (size 168):
  comm "kworker/u3:2", pid 439, jiffies 4294958830 (age 1119.890s)
  hex dump (first 32 bytes):
    00 00 00 00 00 00 00 00 c0 e0 d1 5d a7 53 54 14  ...........].ST.
    00 00 00 00 00 00 00 00 04 00 00 00 01 00 00 00  ................
  backtrace:
    [<c0473bfc>] hci_event_packet+0x1d8/0x2ba0
    [<c04674c8>] hci_rx_work+0x170/0x248
    [<c00318c0>] process_one_work+0x128/0x478
    [<c0031c64>] worker_thread+0x54/0x574
    [<c00368f4>] kthread+0xc0/0xdc
    [<c000a2d0>] ret_from_fork+0x14/0x24
    [<ffffffff>] 0xffffffff
unreferenced object 0xc6600020 (size 2048):
  comm "kworker/u2:0", pid 6, jiffies 4294958833 (age 1119.860s)
  hex dump (first 32 bytes):
    6b 6b 6b 6b 6b 6b 6b 6b 0e 0e 01 04 10 00 01 01  kkkkkkkk........
    01 00 00 00 00 00 00 00 6b 6b 6b 6b 6b 6b 6b 6b  ........kkkkkkkk
  backtrace:
    [<c03c93a0>] __alloc_skb+0x7c/0x164
    [<c03563ac>] ll_recv+0x1c8/0x41c
    [<c03554b4>] hci_uart_tty_receive+0x44/0x64
    [<c0255ec4>] tty_ldisc_receive_buf+0x50/0x58
    [<c0256338>] flush_to_ldisc+0xb8/0xd0
    [<c00318c0>] process_one_work+0x128/0x478
    [<c0031c64>] worker_thread+0x54/0x574
    [<c00368f4>] kthread+0xc0/0xdc
    [<c000a2d0>] ret_from_fork+0x14/0x24
    [<ffffffff>] 0xffffffff
...

I had a look to kernel source code but did not find anything obvious.

Thanks!

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 120691] UART HCI memory leak

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=120691

Loic Poulain <loic.poulain@intel.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |loic.poulain@intel.com

--- Comment #1 from Loic Poulain <loic.poulain@intel.com> ---
Seems that this memory leak is due to the hci_ll (TI) proto.
Don't think this comes from the generic hci-uart/ldisc part.
To be sure, are you able to reproduce this issue with any other proto (hci_h4,
hci_bcm, hci_intel..)?

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 120691] UART HCI memory leak

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=120691

--- Comment #2 from nico.edev@gmail.com ---
Unfortunately my BT module only support TI protocol and I don't have another
one.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 120691] UART HCI memory leak

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=120691

bastienphilbert@gmail.com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |bastienphilbert@gmail.com

--- Comment #3 from bastienphilbert@gmail.com ---
Try the below attached patch.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 120691] UART HCI memory leak

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=120691

--- Comment #4 from bastienphilbert@gmail.com ---
Created attachment 222571
  --> https://bugzilla.kernel.org/attachment.cgi?id=222571&action=edit
Test Patch

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 120691] UART HCI memory leak

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=120691

--- Comment #5 from bastienphilbert@gmail.com ---
Are you able to post the object kmemleak is complaining about at that
particular address.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 120691] UART HCI memory leak

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=120691

--- Comment #6 from nico.edev@gmail.com ---
I did not see your post and now I am in business trip. I will try next week and
let you know.
Thanks a lot.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 120691] UART HCI memory leak

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=120691

--- Comment #7 from nico.edev@gmail.com ---
I tried the patch but problem is still there.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 120691] UART HCI memory leak

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=120691

--- Comment #8 from bastienphilbert@gmail.com ---
Can you enable kmemleak and use the documentation here to find out what object
is leaking,https://www.kernel.org/doc/Documentation/kmemleak.txt.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 120691] UART HCI memory leak

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=120691

Nicolas Royer <nroyer@edevice.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |nroyer@edevice.com

--- Comment #9 from Nicolas Royer <nroyer@edevice.com> ---
I finally got BT modules with BT controller from Broadcom (BCM4343W: UART H4)
and CSR (CSR8510A10: USB).

The leak is present whatever the controller.
In all cases kmemleak reports leaks in the BT controller driver (hci_ll,
hci_h4, btusb) but also in hci_event.

It looks like the leak is in the HCI layer. I will try to figure out what
object is leaking in hci_event.

Below are kmemleak logs for BCM4343W and CSR8510A10.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 120691] UART HCI memory leak

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=120691

--- Comment #10 from Nicolas Royer <nroyer@edevice.com> ---
Created attachment 227531
  --> https://bugzilla.kernel.org/attachment.cgi?id=227531&action=edit
kmemleak_csr8510a10

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 120691] UART HCI memory leak

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=120691

--- Comment #11 from Nicolas Royer <nroyer@edevice.com> ---
Created attachment 227541
  --> https://bugzilla.kernel.org/attachment.cgi?id=227541&action=edit
kmemleak_bcm4343w

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 120691] UART HCI memory leak

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=120691

--- Comment #12 from bastienphilbert@gmail.com ---
See if the below patch fixes or does something new in terms of kmemleak
warning(s) you are getting from these bluetooth drivers.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 120691] UART HCI memory leak

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=120691

--- Comment #13 from bastienphilbert@gmail.com ---
Created attachment 227881
  --> https://bugzilla.kernel.org/attachment.cgi?id=227881&action=edit
Bluetooth Test Patch

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 120691] UART HCI memory leak

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=120691

ingvarthorvald@gmail.com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ingvarthorvald@gmail.com

--- Comment #14 from ingvarthorvald@gmail.com ---
Can you see if that patch fixed the issue or do I need to write a new one.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 120691] UART HCI memory leak

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=120691

--- Comment #15 from Nicolas Royer <nroyer@edevice.com> ---
Unfortunately the leak is still there.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 120691] UART HCI memory leak

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=120691

--- Comment #16 from ingvarthorvald@gmail.com ---
Seems someone else has fixed this and moved it upstream so just close this bug.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 120691] UART HCI memory leak

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=120691

nico.edev@gmail.com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |PATCH_ALREADY_AVAILABLE

--- Comment #17 from nico.edev@gmail.com ---
Exactly; I tried this patch and the leak disappeared:
http://marc.info/?l=linux-bluetooth&m=147193208628547

-- 
You are receiving this mail because:
You are the assignee for the bug.