Re: [PATCH] kref: add function for reading kref value

[Daniel Baluta]

On Mon, Dec 12, 2011 at 5:41 PM, Eric Dumazet <eric.dumazet@xxxxxxxxx> wrote:
> Le lundi 12 décembre 2011 à 17:39 +0200, Daniel Baluta a écrit :
>
>> Ok, for the moment it seems to be a bad idea. But my intention is
>> to integrate kref to networking code, and then to write a general
>> debugging tool for refs.
>>
>> Tracking down reference count problems is hard, and this tool can
>> really help everyone.
>
> I dont think kref will help you that much, because its used everywhere.
>
> Adding a general debugging tool will provide too much noise.
>
> Instead, we (network dev) add debugging points where we want.

Yes, but you have to do this each time you start debugging, for a
particular referenced counted object.

We must find a clever solution to avoid the noise. (e.g use
/proc, /sysfs, /debugfs options to trigger dumping info for
some/all objects with a certain state).

Usecase:

At some point we need to know all objects with refcount X,
and their history of get/put operations.

For us it would have been very useful when debugging dev
refcnts problems. With the current implementation kernel
only dumped info at dev.c:5429:

printk(KERN_EMERG "unregister_netdevice: "
       waiting for %s to become free. Usage "
       count = %d\n",
       dev->name, atomic_read(&dev->refcnt));

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