My understanding is that the new may makes the old obsolete but not wrong,
per se. You just get the effect that you see where opens increment twice and
closes decrement twice. I think that there are some cases that
SET_MODULE_OWNER may not always catch for every driver, so mixing may not
necessarily be wrong...
John
----- Original Message -----
From: "Michael G. Janicki" <mikejani@colfax.com>
To: <kernelnewbies@nl.linux.org>
Sent: Saturday, June 29, 2002 6:57 PM
Subject: Mixed use of MOD_INC_USE_COUNT and SET_MODULE_OWNER
>
> Hello, all:
>
> I've been looking over the use of SET_MODULE_OWNER to better
> understand how its operation differs from the old style of
> MOD_INC_USE_COUNT / MOD_DEC_USE_COUNT. I was under the impression
> that use of the newer method was mutually exclusive of the old
> method. But, having looked over a number of modules, I see
> that many combine use of the two.
>
> For example, loop.c. While it does not use the SET_MODULE_OWNER
> macro, it has defined owner in the lo_fops structure:
>
> static struct block_device_operations lo_fops = {
> owner: THIS_MODULE,
> ...
>
> All is well so far. However, in methods such as loop_set_fd
> and lo_open, use is made of MOD_INC_USE_COUNT/MOD_DEC_USE_COUNT.
> What leads me to believe this should not be happening is the
> result of making use of the module. A simple:
>
> losetup /dev/loop0 /some/diskimg
> losetup /dev/loop1 /some/other/diskimg
> mount /dev/loop0 /mnt/zero
> mount /dev/loop1 /mnt/one
>
> results in a module 'in-use' count of 6. Granted, umount/losetup -d
> does leave the module in a zero-use state but the use count
> just doesn't seem right to me. It seems as if the mount of the
> loop device is being counted twice in both cases -- once manually
> in the module and once by having set the owner to THIS_MODULE.
> Perhaps my understanding of how it should work is just wrong.
>
> Compiling the module using just the "owner" provided in lo_fops
> or even explicitly using SET_MODULE_OWNER(&lo_fops); in loop_init()
> seems to produce no harmful effect and produces what I think to be
> a true in-use count.
>
> The bulk of the modules I've looked at (especially the 3cxxx
> network drivers which are usually really clean) do not mix use of
> old and new methods, but many drivers do so I was left a bit confused.
>
> If anyone can clarify how to properly use SET_MODULE_OWNER, I'd
> certainly appreciate it. Thanks much.
>
> Mike
>
> --
> Michael Janicki
> Key: http://wwwkeys.pgp.net:11371/pks/lookup?op=get&search=0x9D6FAE1A
> Fingerprint: A153 DFC7 8B49 7E97 67B2 3DCE DA3F 3CC5 9D6F AE1A
>
>
> --
> Kernelnewbies: Help each other learn about the Linux kernel.
> Archive: http://mail.nl.linux.org/kernelnewbies/
> FAQ: http://kernelnewbies.org/faq/
>
--
Kernelnewbies: Help each other learn about the Linux kernel.
Archive: http://mail.nl.linux.org/kernelnewbies/
FAQ: http://kernelnewbies.org/faq/
