Re: [PATCH 12/16] GFS2: Mounting & sysfs interface

[Artem B. Bityutskiy]

Steven Whitehouse wrote:
> +int gfs2_sys_fs_add(struct gfs2_sbd *sdp)
> +{
> +	int error;
> +
> +	sdp->sd_kobj.kset = &gfs2_kset;
> +	sdp->sd_kobj.ktype = &gfs2_ktype;
> +
> +	error = kobject_set_name(&sdp->sd_kobj, "%s", sdp->sd_table_name);
> +	if (error)
> +		goto fail;
> +
> +	error = kobject_register(&sdp->sd_kobj);
> +	if (error)
> +		goto fail;
> +
> +	error = sysfs_create_group(&sdp->sd_kobj, &lockstruct_group);
> +	if (error)
> +		goto fail_reg;
> +
> +	error = sysfs_create_group(&sdp->sd_kobj, &counters_group);
> +	if (error)
> +		goto fail_lockstruct;
> +
> +	error = sysfs_create_group(&sdp->sd_kobj, &args_group);
> +	if (error)
> +		goto fail_counters;
> +
> +	error = sysfs_create_group(&sdp->sd_kobj, &tune_group);
> +	if (error)
> +		goto fail_args;
> +
> +	return 0;

Hello,

last time I tried to use "bare" sysfs functions to create my sysfs 
hierarchy I ended up with a problem that the module refcount is not 
increased when those sysfs files are opened. So I could open a sysfs 
file from userspace, do rmmod and enjoy oops.

Then I started using the class and class_device stuff, which have an 
.owner field, and all became fine.

I'm not sure if this is a problem of sysfs, but I suspect it could take 
care of module refcount better.

In your patch, I looked for THIS_MODULE pattern and did not find. I did 
not try, but I suspect your code is not devoid of the problem I 
described. So, this is just FYI and may be not the case.

--
Best regards, Artem B. Bityutskiy
Oktet Labs (St. Petersburg), Software Engineer.
+7 812 4286709 (office) +7 911 2449030 (mobile)
E-mail: dedekind@xxxxxxxxxxxx, Web: www.oktetlabs.ru
-
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/