[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <9ba618c6-20aa-c9fd-dfef-1cef57168742@gmail.com>
Date: Fri, 10 Sep 2021 11:25:14 -0500
From: Ian Pilcher <arequipeno@...il.com>
To: Greg KH <gregkh@...uxfoundation.org>
Cc: axboe@...nel.dk, pavel@....cz, linux-leds@...r.kernel.org,
linux-block@...r.kernel.org, linux-kernel@...r.kernel.org,
kabel@...nel.org
Subject: Re: [PATCH v2 11/15] leds: trigger: blkdev: Enable linking block
devices to LEDs
On 9/10/21 1:48 AM, Greg KH wrote:
>> +/* Gets or allocs & initializes the blkdev disk for a gendisk */
>> +static int blkdev_get_disk(struct gendisk *const gd)
>> +{
>> + struct ledtrig_blkdev_disk *disk;
>> + struct kobject *dir;
>> +
>> + if (gd->ledtrig != NULL) {
>> + kobject_get(gd->ledtrig->dir);
>
> When do you decrement this kobject?
That happens in blkdev_disk_unlink_locked() at line 399. (Also in the
error path in blkdev_put_disk(), called at line 321.)
Looking at this now, blkdev_disk_unlink_locked() should be calling
blkdev_put_disk(), rather than calling kobject_put() directly. I'll fix
that.
>> +static void blkdev_put_disk(struct ledtrig_blkdev_disk *const disk)
>> +{
>> + kobject_put(disk->dir);
>> +
>> + if (hlist_empty(&disk->leds)) {
>> + disk->gd->ledtrig = NULL;
>> + kfree(disk);
>
> This should happen in the kobject release function, not here, right?
If you're referring to the kfree() call, it's freeing the
ledtrig_blkdev_disk structure, not the gendisk. I use "gd" for gendisk
pointers and "disk" for ledtrig_blkdev_disk pointers.
> Did you try this out with removable block devices yet?
Absolutely. I've tested removing both block devices and (user space)
LEDs.
--
========================================================================
In Soviet Russia, Google searches you!
========================================================================
Powered by blists - more mailing lists