Received-SPF: pass (google.com: domain of linux-bluetooth-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [Linux-kernel-mentees] [PATCH] Fix uninit-value in
 hci_chan_lookup_handle
To:     Marcel Holtmann <marcel@holtmann.org>
Cc:     linux-kernel-mentees@lists.linuxfoundation.org,
        Johan Hedberg <johan.hedberg@gmail.com>,
        linux-bluetooth <linux-bluetooth@vger.kernel.org>,
        linux-kernel@vger.kernel.org
References: <20200905020410.20350-1-anant.thazhemadam@gmail.com>
 <2AEF86AF-8F80-4B01-B506-D2ABC654DBC7@holtmann.org>
From:   Anant Thazhemadam <anant.thazhemadam@gmail.com>
Message-ID: <9b8f5993-9956-5a57-89d6-f5ae36e59038@gmail.com>
Date:   Thu, 1 Oct 2020 12:44:53 +0530
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101
 Thunderbird/68.10.0
MIME-Version: 1.0
In-Reply-To: <2AEF86AF-8F80-4B01-B506-D2ABC654DBC7@holtmann.org>
Content-Type: text/plain; charset=windows-1252
Content-Transfer-Encoding: 7bit
Content-Language: en-US
Precedence: bulk


On 26/09/20 1:16 pm, Marcel Holtmann wrote:
> Hi Anant,
>
>> When the amount of data stored in the location corresponding to
>> iov_iter *from is less then 4, some data seems to go uninitialized.
>>
>> Updating this condition accordingly, makes sense both intuitively and 
>> logically as well, since the other check for extreme condition done is if
>> len > HCI_MAX_FRAME_SIZE, which is HCI_MAX_ACL_SIZE (which is 1024) + 4;
>> which itself gives some idea about what must be the ideal mininum size.
>>
>> Reported-and-tested by: syzbot+4c14a8f574461e1c3659@syzkaller.appspotmail.com
>> Signed-off-by: Anant Thazhemadam <anant.thazhemadam@gmail.com>
>> ---
>> If there is some explicit reason why len < 4 doesn't work, and only len < 2 works, 
>> please do let me know.
>> The commit message that introduced the initial change 
>> (512b2268156a4e15ebf897f9a883bdee153a54b7) wasn't exactly very helpful in this 
>> respect, and I couldn't find a whole lot of discussion regarding this either.
>>
>> drivers/bluetooth/hci_vhci.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/bluetooth/hci_vhci.c b/drivers/bluetooth/hci_vhci.c
>> index 8ab26dec5f6e..0c49821d7b98 100644
>> --- a/drivers/bluetooth/hci_vhci.c
>> +++ b/drivers/bluetooth/hci_vhci.c
>> @@ -159,7 +159,7 @@ static inline ssize_t vhci_get_user(struct vhci_data *data,
>> 	__u8 pkt_type, opcode;
>> 	int ret;
>>
>> -	if (len < 2 || len > HCI_MAX_FRAME_SIZE)
>> +	if (len < 4 || len > HCI_MAX_FRAME_SIZE)
>> 		return -EINVAL;
> so the minimum requirement is to have an 1 octet packet indicator. If some other functions are missing length checks, then it should be fixed there. I see that in case of HCI_VENDOR_PKT we need to check for the 2nd octet to be present, but I have no idea why you require 3 octets minimum. As I said, if so, then the other functions that get called require proper range checks.
>
> Regards
>
> Marcel

Ah, I see. That makes sense, and I'm sorry for this. I just thought since
HCI_MAX_FRAME_SIZE was 4 more than 1024, min. len might have to be 4.
Taking a look at the reproducer, it becomes evident that the only reason this
patch didn't trigger an issue was because the reproducer ran with len = 3.

I don't really have a strong enough clue yet about where things are going wrong, so I'd
appreciate any guidance/insight into this, and where I should be looking. :/

Also, although the bug says it is only an uninit-value, it eventually ends up triggering
a kernel panic (looking at a log rather than a report shows that as well).

Thanks,
Anant