linux-kernel - Re: Incrementing module reference count

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: <50C3924B.1050300@cavium.com>
Date:	Sat, 08 Dec 2012 11:17:31 -0800
From:	Aaron Williams <Aaron.Williams@...ium.com>
To:	Borislav Petkov <bp@...en8.de>, linux-kernel@...r.kernel.org,
	"Williams, Aaron" <Aaron.Williams@...iumnetworks.com>
Subject: Re: Incrementing module reference count

On 12/08/2012 03:44 AM, Borislav Petkov wrote:
> On Fri, Dec 07, 2012 at 05:14:39PM -0800, Aaron Williams wrote:
>> Hi,
>>
>> I have a kernel module which other modules register with in order to
>> export access functions. So far I have everything working but I want
>> to prevent a module that is registered with my module from unloading
>> since now my module is dependent on the other module.
>>
>> Is there a way I can cause the reference count of the module
>> registering with my module to increase? I tried calling get_device
>> with the device structure of the module that is registering but that
>> does not seem to work.
>>
>> For example, I have the following function:
>>
>> /**
>>  * Adds a mapping of a device node to a memory accessor
>>  *
>>  * @param[in] dev - device
>>  * @param[in] macc - memory accessor
>>  *
>>  * @returns 0 for success or -ENOMEM
>>  */
>> int of_memory_accessor_register(struct device *dev,
>>                 struct memory_accessor *macc)
>> {
>>     struct of_macc_entry *mentry;
>>
>>     mentry = kmalloc(sizeof(*mentry), GFP_KERNEL);
>>     if (mentry == NULL)
>>         return -ENOMEM;
>>
>>     mentry->dev = dev;
>>     mentry->macc = macc;
>>
>>     mutex_lock(&lock);
>>
>>     get_device(dev);
>>     list_add(&(mentry->list), &macc_list);
>>
>>     mutex_unlock(&lock);
>>
>>     return 0;
>> }
>> EXPORT_SYMBOL(of_memory_accessor_register);
>>
>> Basically my module is used for things like serial EEPROMs and
>> whatnot so that external modules can find the accessor functions
>> based on the device tree. In my case I am updating the Vitesse
>> VSC848X driver so that it can read the SFP module when it is plugged
>> in using the AT24 I2C EEPROM module. I want to prevent the at24
>> module from unloading while other modules in turn are using it. The
>> at24 module does not export any symbols.
> From a quick code scan, at24 has this:
>
> 	static struct i2c_driver at24_driver = {
> 	        .driver = {
> 	                .name = "at24",
> 	                .owner = THIS_MODULE,
> 	        },
> 	        .probe = at24_probe,
> 	        .remove = __devexit_p(at24_remove),
> 	        .id_table = at24_ids,
> 	};
>
> and then you could do:
>
> 	try_module_get(at24_driver.driver->owner);
>
> from your code. AFAICT, of course.
>
> HTH.
Thank you.

try_module_get is what I needed.

I got it working. I designed my solution to be generic so it's not tied
to just the at24 driver. I added a memory accessor module which is
basically a registry. The at24 driver registers with it and then the
Vitesse PHY driver calls it to gain access to the at24 driver. Only when
the Vitesse driver gets the memory accessor the reference count for the
at24 driver is incremented.

My solution can work with any module that provides memory accessor
functions.

I still have some cleanup to do since my work was done with an older 2.6
kernel.

-Aaron

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/