lkml.org 
[lkml]   [2012]   [Mar]   [19]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRE: [PATCH]s390/char/vmur.c: memory leak Fix in the driver
Date
On Mon, Mar 19, 2012 at 7:08 PM, Martin Schwidefsky <schwidefsky@de.ibm.com> wrote:
> On Mon, 19 Mar 2012 11:37:44 +0100
> Martin Schwidefsky <schwidefsky@de.ibm.com> wrote:
>
>> On Mon, 19 Mar 2012 01:55:40 +0000
>> "Chen, Dennis (SRDC SW)" <Dennis1.Chen@amd.com> wrote:
>>
>> > CC'ing maintainer: Martin Schwidefsky & Heiko Carstens...
>> >
>> > -----Original Message-----
>> > From: Chen, Dennis (SRDC SW)
>> > Sent: Sunday, March 18, 2012 3:41 AM
>> > To: linux-kernel@vger.kernel.org
>> > Cc: beattiem@uk.ibm.com; holzheu@de.ibm.com; munzert@de.ibm.com; linux390@de.ibm.com; linux-s390@vger.kernel.org; Chen, Dennis (SRDC SW)
>> > Subject: [PATCH]s390/char/vmur.c: memory leak Fix in the driver
>> >
>> > This patch is used to fix a memory leak issue in s390/char/vmur.c: a character device instance is
>> > allocated by cdev_alloc, the cdev_del will not free that space if cdev_init is applied before.
>> >
>> > Signed-off-by: dennis1.chen@amd.com
>> > --- a/s390/char/vmur.c 2012-03-18 02:50:47.950963949 +0800
>> > +++ b/s390/char/vmur.c 2012-03-18 03:12:04.790936740 +0800
>> > @@ -903,7 +903,7 @@ static int ur_set_online(struct ccw_devi
>> > goto fail_urdev_put;
>> > }
>> >
>> > - cdev_init(urd->char_device, &ur_fops);
>> > + urd->char_device->ops = &ur_fops;
>> > urd->char_device->dev = MKDEV(major, minor);
>> > urd->char_device->owner = ur_fops.owner;
>> >
>>
>> How does that fix anything? My copy of cdev_init looks like this:
>>
>> void cdev_init(struct cdev *cdev, const struct file_operations *fops)
>> {
>> memset(cdev, 0, sizeof *cdev);
>> INIT_LIST_HEAD(&cdev->list);
>> kobject_init(&cdev->kobj, &ktype_cdev_default);
>> cdev->ops = fops;
>> }
>>
>> It does not allocate anything but it initializes some more fields.
>> The new code would only initialize the ops field. In addition
>> cdev_del does a kobject_put and the release function of the object
>> will call cdev_dynamic_release as far as I can tell. That code
>> should be fine as it is.
>
> Hmm, forget that. cdev_init has a second kobject_init that replaces
> ktype_cdev_dynamic with ktype_cdev_default. cdev_alloc already does
> the initialization and your fix is just fine.
> Seems like cdev_init is only there for static cdev structures.
>
> --

Yes, you're right. So if we don't have this patch, all the cdev instance allocated by cdev_alloc() will not freed by cdev_del as expected,
if the system running with long enough time, memory will exhaust, especially for a server platform...





\
 
 \ /
  Last update: 2012-03-19 16:45    [W:0.094 / U:1.644 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site