lkml.org 
[lkml]   [2018]   [Nov]   [28]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [Xen-devel] [PATCH] pvcalls-front: fixes incorrect error handling
From
Date
On 11/27/18 4:08 PM, Stefano Stabellini wrote:
> On Tue, 27 Nov 2018, Boris Ostrovsky wrote:
>> On 11/27/18 3:37 PM, Stefano Stabellini wrote:
>>> On Tue, 27 Nov 2018, PanBian wrote:
>>>> On Mon, Nov 26, 2018 at 03:31:39PM -0500, Boris Ostrovsky wrote:
>>>>> On 11/21/18 9:07 PM, Pan Bian wrote:
>>>>>> kfree() is incorrectly used to release the pages allocated by
>>>>>> __get_free_page() and __get_free_pages(). Use the matching deallocators
>>>>>> i.e., free_page() and free_pages(), respectively.
>>>>>>
>>>>>> Signed-off-by: Pan Bian <bianpan2016@163.com>
>>>>>> ---
>>>>>> drivers/xen/pvcalls-front.c | 4 ++--
>>>>>> 1 file changed, 2 insertions(+), 2 deletions(-)
>>>>>>
>>>>>> diff --git a/drivers/xen/pvcalls-front.c b/drivers/xen/pvcalls-front.c
>>>>>> index 2f11ca7..77224d8 100644
>>>>>> --- a/drivers/xen/pvcalls-front.c
>>>>>> +++ b/drivers/xen/pvcalls-front.c
>>>>>> @@ -385,8 +385,8 @@ static int create_active(struct sock_mapping *map, int *evtchn)
>>>>>> out_error:
>>>>>> if (*evtchn >= 0)
>>>>>> xenbus_free_evtchn(pvcalls_front_dev, *evtchn);
>>>>>> - kfree(map->active.data.in);
>>>>>> - kfree(map->active.ring);
>>>>>> + free_pages((unsigned long)map->active.data.in, PVCALLS_RING_ORDER);
>>>>> Is map->active.data.in guaranteed to be NULL when entering this routine?
>>>> I am not sure yet. Sorry for that. I observed the mismatches between
>>>> __get_free_page and kfree, and submitted the patch.
>>>>
>>>> But I think your consideration is reasonable. A better solution is to
>>>> directly free bytes, a local variable that holds __get_free_pages return
>>>> value. If you agree, I will rewrite the patch.
>>> Like Boris said, map->active.ring and map->active.data.in are not
>>> guaranteed to be NULL or != NULL here. For instance,map->active.ring can
>>> be != NULL and map->active.data.in can be NULL. However, free_pages and
>>> free_page should be able to cope with it, the same way that kfree is
>>> able to cope with it?
>> If map->active.data.in can be non-NULL on entry to this routine then I
>> think this has been a problem all along. Pan's suggestion to use bytes
>> for freeing is going to solve this (assuming bytes will be initialized
>> to NULL).
> Why is it a problem? map->active.data.in and map->active.ring are only
> != NULL if they need to be freed. Otherwise, they are NULL.

That was my question --- I wasn't sure about it, and I read your
previous message as if it was possible to be calling create_active()
with map->active.data.in pointing somewhere non-NULL.

If it is NULL *upon entry* to calling_create() then Pan's original patch
is fine.


-boris


> All structs
> are always initialized to zero. I don't think there are any issues.

\
 
 \ /
  Last update: 2018-11-27 23:48    [W:0.102 / U:0.136 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site