Re: [RFC/RFT] ath9k_htc: Fix memory leak

[Larry Finger <Larry.Finger@lwfinger.net>]

On 08/24/2011 10:53 AM, Mohammed Shafi wrote:
> On Wed, Aug 24, 2011 at 1:15 AM, John W. Linville
> <linville@tuxdriver.com>  wrote:
>> On Mon, Aug 15, 2011 at 05:56:33PM +0530, Mohammed Shafi wrote:
>>> On Sun, Aug 14, 2011 at 7:04 AM, Larry Finger<Larry.Finger@lwfinger.net>  wrote:
>>>> Kmemleak shows the following kind of memory leak for ath9k_htc:
>>>>
>>>> unreferenced object 0xffff88004542f200 (size 512):
>>>>   comm "khubd", pid 977, jiffies 4317632516 (age 16855.868s)
>>>>   hex dump (first 32 bytes):
>>>>     00 00 00 0a 00 00 00 00 00 02 01 05 00 00 02 01  ................
>>>>     00 00 00 00 00 00 81 38 02 00 00 00 33 33 30 30  .......8....3300
>>>>   backtrace:
>>>>     [<ffffffff81122d77>] create_object+0x127/0x2b0
>>>>     [<ffffffff813580b1>] kmemleak_alloc+0x21/0x50
>>>>     [<ffffffff8111f383>] __kmalloc_node_track_caller+0x153/0x220
>>>>     [<ffffffff812b703e>] __alloc_skb+0x7e/0x170
>>>>     [<ffffffffa0833331>] htc_connect_service+0x111/0x200 [ath9k_htc]
>>>>     [<ffffffffa083bb90>] ath9k_init_htc_services+0x240/0x2b0 [ath9k_htc]
>>>>     [<ffffffffa083c1ca>] ath9k_htc_probe_device+0xea/0xa50 [ath9k_htc]
>>>>     [<ffffffffa08338dc>] ath9k_htc_hw_init+0xc/0x30 [ath9k_htc]
>>>>     [<ffffffffa08356ba>] ath9k_hif_usb_probe+0x1ca/0x420 [ath9k_htc]
>>>>     [<ffffffffa00a2279>] usb_probe_interface+0xb9/0x160 [usbcore]
>>>>     [<ffffffff81279379>] driver_probe_device+0x89/0x1a0
>>>>     [<ffffffff8127958b>] __device_attach+0x4b/0x60
>>>>     [<ffffffff81278024>] bus_for_each_drv+0x64/0x90
>>>>     [<ffffffff81279231>] device_attach+0xa1/0xb0
>>>>     [<ffffffff81278a25>] bus_probe_device+0x25/0x40
>>>>     [<ffffffff81276d2a>] device_add+0x55a/0x630
>>>>
>>>> The device is a TP-Link TL-WN722N. The output from lsusb is "ID 0cf3:9271
>>>> Atheros Communications, Inc. AR9271 802.11n".
>>>>
>>>> Signed-off-by: Larry Finger<Larry.Finger@lwfinger.net>
>>>> Cc: Stable<stable@kernel.org>
>>>> ---
>>>>
>>>> This patch certainly fixes a leak. I'm testing to see if there are
>>>> others.
>>>
>>> Hi Larry,
>>>
>>> thanks for finding this.
>>> I remember seeing this memleaks sometime back and thought this patch
>>> might help(i think same as yours with some more corner cases)
>>> https://patchwork.kernel.org/patch/1016752/
>>> but I found there were still some memory leaks, I was unsure where I
>>> am missing. if you are sure that this addresses the issue, we can send
>>> a patch.
>>
>> Ping?  Is this the patch we want?  Or something else?
>
> Hi John,
>
> the complete patch is posted
> http://www.spinics.net/lists/linux-wireless/msg75134.html
> Vasanth said that at some scenario's with the above patch it may cause
> some panics.
> based on the maintainers concerns/doubts lets drop this patch. myself
> and Larry still noticed few more memory leaks. I think we better drop
> this patch and fix it completely/tested. as I am with some other
> critical tasks we will soon address this.
>
> Larry if you have any thoughts, please share.

Dropping the current patches seems right with me. I hope to get time to work on 
this, but I also have some other things to do. Perhaps, we can get a patch with 
the subset of leaks in the error path that do not have any complications. Those 
would be easy, and then we can get the others later.

Larry