netdev - Re: [PATCH net 2/2] fsl/fman: Fix refcount handling of fman-related devices

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

Message-ID: <72a4d2b0-3faf-46a3-84f2-fc93bdf19104@t-argos.ru>
Date: Thu, 17 Oct 2024 13:53:31 +0300
From: Aleksandr Mishin <amishin@...rgos.ru>
To: Paolo Abeni <pabeni@...hat.com>, Igal Liberman
	<igal.liberman@...escale.com>
CC: Simon Horman <horms@...nel.org>, Madalin Bucur <madalin.bucur@....com>,
	Sean Anderson <sean.anderson@...o.com>, "David S. Miller"
	<davem@...emloft.net>, Eric Dumazet <edumazet@...gle.com>, Jakub Kicinski
	<kuba@...nel.org>, <netdev@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
	<lvc-project@...uxtesting.org>
Subject: Re: [PATCH net 2/2] fsl/fman: Fix refcount handling of fman-related
 devices


On 17.10.2024 13:01, Paolo Abeni wrote:
> On 10/15/24 08:01, Aleksandr Mishin wrote:
>> In mac_probe() there are multiple calls to of_find_device_by_node(),
>> fman_bind() and fman_port_bind() which takes references to of_dev->dev.
>> Not all references taken by these calls are released later on error path
>> in mac_probe() and in mac_remove() which lead to reference leaks.
>>
>> Add references release.
>>
>> Fixes: 3933961682a3 ("fsl/fman: Add FMan MAC driver")
>> Signed-off-by: Aleksandr Mishin <amishin@...rgos.ru>
>> ---
>> Compile tested only.
>>
>>   drivers/net/ethernet/freescale/fman/mac.c | 62 +++++++++++++++++------
>>   1 file changed, 47 insertions(+), 15 deletions(-)
>>
>> diff --git a/drivers/net/ethernet/freescale/fman/mac.c 
>> b/drivers/net/ethernet/freescale/fman/mac.c
>> index 9b863db0bf08..11da139082e1 100644
>> --- a/drivers/net/ethernet/freescale/fman/mac.c
>> +++ b/drivers/net/ethernet/freescale/fman/mac.c
>> @@ -204,7 +204,7 @@ static int mac_probe(struct platform_device 
>> *_of_dev)
>>       if (err) {
>>           dev_err(dev, "failed to read cell-index for %pOF\n", 
>> dev_node);
>>           err = -EINVAL;
>> -        goto _return_of_node_put;
>> +        goto _return_dev_put;
>
> We are after a succesful of_find_device_by_node and prior to 
> fman_bind(), mac_dev->fman_dev refcount is 1


Indeed. refcounts = 1.


>
>> @@ -213,40 +213,51 @@ static int mac_probe(struct platform_device 
>> *_of_dev)
>>       if (!priv->fman) {
>>           dev_err(dev, "fman_bind(%pOF) failed\n", dev_node);
>>           err = -ENODEV;
>> -        goto _return_of_node_put;
>> +        goto _return_dev_put;
>>       }


refcounts: 1 + 1 = 2.


>>   +    /* Two references have been taken in of_find_device_by_node()
>> +     * and fman_bind(). Release one of them here. The second one
>> +     * will be released in mac_remove().
>> +     */
>> +    put_device(mac_dev->fman_dev);


refcounts: 2 - 1 = 1.


>>       of_node_put(dev_node);
>> +    dev_node = NULL;
>>         /* Get the address of the memory mapped registers */
>>       mac_dev->res = platform_get_mem_or_io(_of_dev, 0);
>>       if (!mac_dev->res) {
>>           dev_err(dev, "could not get registers\n");
>> -        return -EINVAL;
>> +        err = -EINVAL;
>> +        goto _return_dev_put;
>
> Here we are after a successful fman_bind(), mac_dev->fman_dev refcount 
> is 2. _return_dev_put will drop a single reference, this error path 
> looks buggy.


We released 1 reference above with "put_device(mac_dev->fman_dev);".


>
> Similar issue for the _return_dev_arr_put error path below.


Similar situation: we release 1 reference with 
"put_device(mac_dev->fman_port_devs[i]);".


>
> Cheers,
>
> Paolo
>
-- 
Kind regards
Aleksandr