netdev - Re: [PATCH v3 1/2] nfc: llcp: avoid double release/put on LLCP_CLOSED in nfc_llcp_recv

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <88741cf8-7649-49e1-8d82-5440fccd618f@redhat.com>
Date: Sun, 28 Dec 2025 10:16:02 +0100
From: Paolo Abeni <pabeni@...hat.com>
To: Qianchang Zhao <pioooooooooip@...il.com>, netdev@...r.kernel.org
Cc: Krzysztof Kozlowski <krzk@...nel.org>, Jakub Kicinski <kuba@...nel.org>,
 "David S. Miller" <davem@...emloft.net>, Eric Dumazet <edumazet@...gle.com>,
 Simon Horman <horms@...nel.org>, linux-kernel@...r.kernel.org,
 stable@...r.kernel.org, Zhitong Liu <liuzhitong1993@...il.com>
Subject: Re: [PATCH v3 1/2] nfc: llcp: avoid double release/put on LLCP_CLOSED
 in nfc_llcp_recv_disc()

On 12/28/25 10:02 AM, Paolo Abeni wrote:
> On 12/18/25 3:59 AM, Qianchang Zhao wrote:
>> nfc_llcp_sock_get() takes a reference on the LLCP socket via sock_hold().
>>
>> In nfc_llcp_recv_disc(), when the socket is already in LLCP_CLOSED state,
>> the code used to perform release_sock() and nfc_llcp_sock_put() in the
>> CLOSED branch but then continued execution and later performed the same
>> cleanup again on the common exit path. This results in refcount imbalance
>> (double put) and unbalanced lock release.
>>
>> Remove the redundant CLOSED-branch cleanup so that release_sock() and
>> nfc_llcp_sock_put() are performed exactly once via the common exit path, 
>> while keeping the existing DM_DISC reply behavior.
>>
>> Fixes: d646960f7986 ("NFC: Initial LLCP support")
>> Cc: stable@...r.kernel.org
>> Signed-off-by: Qianchang Zhao <pioooooooooip@...il.com>
>> ---
>>  net/nfc/llcp_core.c | 5 -----
>>  1 file changed, 5 deletions(-)
>>
>> diff --git a/net/nfc/llcp_core.c b/net/nfc/llcp_core.c
>> index beeb3b4d2..ed37604ed 100644
>> --- a/net/nfc/llcp_core.c
>> +++ b/net/nfc/llcp_core.c
>> @@ -1177,11 +1177,6 @@ static void nfc_llcp_recv_disc(struct nfc_llcp_local *local,
>>  
>>  	nfc_llcp_socket_purge(llcp_sock);
>>  
>> -	if (sk->sk_state == LLCP_CLOSED) {
>> -		release_sock(sk);
>> -		nfc_llcp_sock_put(llcp_sock);
> 
> To rephrase Krzysztof concernt, this does not looks like the correct
> fix: later on nfc_llcp_recv_disc() will try a send over a closed socket,
> which looks wrong. Instead you could just return after
> nfc_llcp_sock_put(), or do something alike:
> 
> 	if (sk->sk_state == LLCP_CLOSED)
> 		goto cleanup;
> 
> 	// ...
> 
> 
> cleanup:
> 	release_sock(sk);
> 	nfc_llcp_sock_put(llcp_sock);
> }

I'm sorry for the confusing feedback above.

I read the comments on patch 2/2 only after processing this one.

Indeed following the half-interrupted discussion on old revision, with
bad patch splitting is quite difficult.

@Qianchang Zhao: my _guess_ is that on LLCP_CLOSED the code has to
release the final sk reference... In any case discussion an a patch
series revision is not concluded until the reviewer agrees on that.

@Krzysztof: ... but still it looks like in the current code there is a
double release on the sk socket lock, which looks wrong, what am I
missing here?

/P