Re: [PATCH] Bluetooth: hci_conn_failed: Fixes memory leak

From: Yuran Pereira <yuran.pereira@hotmail.com>
To: Luiz Augusto von Dentz <luiz.dentz@gmail.com>
Cc: "johan.hedberg@gmail.com" <johan.hedberg@gmail.com>,
	"marcel@holtmann.org" <marcel@holtmann.org>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"linux-bluetooth@vger.kernel.org"
	<linux-bluetooth@vger.kernel.org>,
	"syzbot+39ec16ff6cc18b1d066d@syzkaller.appspotmail.com"
	<syzbot+39ec16ff6cc18b1d066d@syzkaller.appspotmail.com>,
	"linux-kernel-mentees@lists.linuxfoundation.org"
	<linux-kernel-mentees@lists.linuxfoundation.org>
Subject: Re: [PATCH] Bluetooth: hci_conn_failed: Fixes memory leak
Date: Tue, 10 Oct 2023 20:13:16 +0530	[thread overview]
Message-ID: <AM9P192MB1267F74A5B04DBD6115D2CCBE8CDA@AM9P192MB1267.EURP192.PROD.OUTLOOK.COM> (raw)
In-Reply-To: <CABBYNZLA2aTwRFqeDNLaT6Q4cytLB1e-q5oON=KVK1ts-0dz1w@mail.gmail.com>

Hello Luiz,

Thanks for the reply.

I just took another look at hci_abort_conn_sync which is where 
hci_conn_failed is being called, and as you suggested, hci_conn_failed, 
is in fact being called with a status 0. So my previous patch might in 
fact be inaccurate.

By the time hci_conn_failed is called, the value of "err" within 
hci_conn_abort_sync can also be 0, so using bt_status(err) as status is 
not going to solve the issue.
I am considering using one of the HCI_ERROR_ values but not exactly sure 
which is the most appropriate in this specific situation.

Perhaps HCI_ERROR_UNSPECIFIED ???

Something like:
- hci_conn_failed(conn, reason);
+ hci_conn_failed(conn, HCI_ERROR_UNSPECIFIED);

So, would you suggest going with this solution, or do you think the 
issue might be originating earlier in the call sequence and I should pay 
more attention to conn->reason?

Thanks,

Yuran Pereira

On 10/10/23 01:20, Luiz Augusto von Dentz wrote:
> Hi Yuran,
>
> On Sat, Oct 7, 2023 at 10:39 AM Yuran Pereira <yuran.pereira@hotmail.com> wrote:
>> Hello Greg,
>> My apologies, I just noticed that my patch is based on the mainline tree. I'll re-submit one based on the Bluetooth tree and I'll ensure to include the commit id that it's fixing.
>>
>> Thanks,
>> Yuran Pereira
>> ________________________________
>> De: Greg KH <gregkh@linuxfoundation.org>
>> Enviado: 7 de outubro de 2023 11:41
>> Para: Yuran Pereira <yuran.pereira@hotmail.com>
>> Cc: marcel@holtmann.org <marcel@holtmann.org>; johan.hedberg@gmail.com <johan.hedberg@gmail.com>; linux-kernel@vger.kernel.org <linux-kernel@vger.kernel.org>; linux-bluetooth@vger.kernel.org <linux-bluetooth@vger.kernel.org>; luiz.dentz@gmail.com <luiz.dentz@gmail.com>; syzbot+39ec16ff6cc18b1d066d@syzkaller.appspotmail.com <syzbot+39ec16ff6cc18b1d066d@syzkaller.appspotmail.com>; linux-kernel-mentees@lists.linuxfoundation.org <linux-kernel-mentees@lists.linuxfoundation.org>
>> Assunto: Re: [PATCH] Bluetooth: hci_conn_failed: Fixes memory leak
>>
>> On Sat, Oct 07, 2023 at 05:09:01PM +0530, Yuran Pereira wrote:
>>> The hci_conn_failed() function currently calls hci_connect_cfm(), which
>>> indirectly leads to the allocation of an l2cap_conn struct in l2cap_conn_add().
>>> This operation results in a memory leak, as the l2cap_conn structure
>>> becomes unreferenced.
>>>
>>> To address this issue and prevent the memory leak, this patch modifies
>>> hci_conn_failed() to replace the call to hci_connect_cfm() with a
>>> call to hci_disconn_cfm().
> I suspect this is not quite right, hci_disconn_cfm is called when a
> disconnection has been requested, hci_connect_cfm is correct here
> since it is meant to notify the result of connection request procedure
> so I can only assume that the culprit here is that hci_conn_failed is
> called with status 0 which is invalid and needs fixing.
>
>>> Reported-by: syzbot+39ec16ff6cc18b1d066d@syzkaller.appspotmail.com
>>> Signed-off-by: Yuran Pereira <yuran.pereira@hotmail.com>
>>> ---
>>>   net/bluetooth/hci_conn.c | 2 +-
>>>   1 file changed, 1 insertion(+), 1 deletion(-)
>> What commit id does this fix?
>>
>> thanks,
>>
>> greg k-h
>
>
_______________________________________________
Linux-kernel-mentees mailing list
Linux-kernel-mentees@lists.linuxfoundation.org
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees