2014-06-16 15:09:04

by Nicholas Krause

Subject: [PATCHv3]Checks for Null value in function *get_skub

Signed-off-by: Nicholas Krause <[email protected]>
---
drivers/infiniband/hw/cxgb4/cm.c | 8 +++-----
1 file changed, 3 insertions(+), 5 deletions(-)

diff --git a/drivers/infiniband/hw/cxgb4/cm.c b/drivers/infiniband/hw/cxgb4/cm.c
index f9477e2..2d56983 100644
--- a/drivers/infiniband/hw/cxgb4/cm.c
+++ b/drivers/infiniband/hw/cxgb4/cm.c
@@ -340,15 +340,13 @@ static int status2errno(int status)
*/
static struct sk_buff *get_skb(struct sk_buff *skb, int len, gfp_t gfp)
{
if (skb && !skb_is_nonlinear(skb) && !skb_cloned(skb)) {
skb_trim(skb, 0);
skb_get(skb);
skb_reset_transport_header(skb);
} else {
+ if (skb)
+ kfree (skb);
skb = alloc_skb(len, gfp);
}
t4_set_arp_err_handler(skb, NULL, NULL);
--
1.9.1

2014-06-16 15:10:33

by Nicholas Krause

Subject: Re: [PATCHv3]Checks for Null value in function *get_skub

The bug is open here https://bugzilla.kernel.org/show_bug.cgi?id=44631.
I would recommend closing it now as I seem to have fixed it.
Cheers Nick

On Mon, Jun 16, 2014 at 11:08 AM, Nicholas Krause <[email protected]> wrote:
> Signed-off-by: Nicholas Krause <[email protected]>
> ---
> drivers/infiniband/hw/cxgb4/cm.c | 8 +++-----
> 1 file changed, 3 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/infiniband/hw/cxgb4/cm.c b/drivers/infiniband/hw/cxgb4/cm.c
> index f9477e2..2d56983 100644
> --- a/drivers/infiniband/hw/cxgb4/cm.c
> +++ b/drivers/infiniband/hw/cxgb4/cm.c
> @@ -340,15 +340,13 @@ static int status2errno(int status)
> */
> static struct sk_buff *get_skb(struct sk_buff *skb, int len, gfp_t gfp)
> {
> if (skb && !skb_is_nonlinear(skb) && !skb_cloned(skb)) {
> skb_trim(skb, 0);
> skb_get(skb);
> skb_reset_transport_header(skb);
> } else {
> + if (skb)
> + kfree (skb);
> skb = alloc_skb(len, gfp);
> }
> t4_set_arp_err_handler(skb, NULL, NULL);
> --
> 1.9.1
>

2014-06-16 15:12:53

by Levente Kurusa

Subject: Re: [PATCHv3]Checks for Null value in function *get_skub

On 06/16/2014 05:08 PM, Nicholas Krause wrote:
> Signed-off-by: Nicholas Krause <[email protected]>
> ---
> drivers/infiniband/hw/cxgb4/cm.c | 8 +++-----
> 1 file changed, 3 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/infiniband/hw/cxgb4/cm.c b/drivers/infiniband/hw/cxgb4/cm.c
> index f9477e2..2d56983 100644
> --- a/drivers/infiniband/hw/cxgb4/cm.c
> +++ b/drivers/infiniband/hw/cxgb4/cm.c
> @@ -340,15 +340,13 @@ static int status2errno(int status)
> */
> static struct sk_buff *get_skb(struct sk_buff *skb, int len, gfp_t gfp)
> {
> if (skb && !skb_is_nonlinear(skb) && !skb_cloned(skb)) {
> skb_trim(skb, 0);
> skb_get(skb);
> skb_reset_transport_header(skb);
> } else {
> + if (skb)
> + kfree (skb);
> skb = alloc_skb(len, gfp);
> }
> t4_set_arp_err_handler(skb, NULL, NULL);
>

Isn't kfree(NULL) legal?

(i.e. the if statement is useless)

Thanks,
Levente Kurusa

2014-06-16 15:14:48

by Nicholas Krause

Subject: Re: [PATCHv3]Checks for Null value in function *get_skub

>From what I know yes so I would close the bug at the the link in
my previous email and just void my patch.
Cheers Nick

On Mon, Jun 16, 2014 at 11:12 AM, Levente Kurusa <[email protected]> wrote:
> On 06/16/2014 05:08 PM, Nicholas Krause wrote:
>>
>> Signed-off-by: Nicholas Krause <[email protected]>
>> ---
>> drivers/infiniband/hw/cxgb4/cm.c | 8 +++-----
>> 1 file changed, 3 insertions(+), 5 deletions(-)
>>
>> diff --git a/drivers/infiniband/hw/cxgb4/cm.c
>> b/drivers/infiniband/hw/cxgb4/cm.c
>> index f9477e2..2d56983 100644
>> --- a/drivers/infiniband/hw/cxgb4/cm.c
>> +++ b/drivers/infiniband/hw/cxgb4/cm.c
>> @@ -340,15 +340,13 @@ static int status2errno(int status)
>> */
>> static struct sk_buff *get_skb(struct sk_buff *skb, int len, gfp_t gfp)
>> {
>> if (skb && !skb_is_nonlinear(skb) && !skb_cloned(skb)) {
>> skb_trim(skb, 0);
>> skb_get(skb);
>> skb_reset_transport_header(skb);
>> } else {
>> + if (skb)
>> + kfree (skb);
>> skb = alloc_skb(len, gfp);
>> }
>> t4_set_arp_err_handler(skb, NULL, NULL);
>>
>
> Isn't kfree(NULL) legal?
>
> (i.e. the if statement is useless)
>
> Thanks,
> Levente Kurusa

2014-06-16 15:18:44

by Steve Wise

Subject: Re: [PATCHv3]Checks for Null value in function *get_skub

On 6/16/2014 10:14 AM, Nick Krause wrote:
> >From what I know yes so I would close the bug at the the link in
> my previous email and just void my patch.
> Cheers Nick

We shouldn't be kfree-ing an skb anyway. Should use kfree_skb().

There is still a leak if skb is non-null and the skb is non linear or
cloned.


> On Mon, Jun 16, 2014 at 11:12 AM, Levente Kurusa <[email protected]> wrote:
>> On 06/16/2014 05:08 PM, Nicholas Krause wrote:
>>> Signed-off-by: Nicholas Krause <[email protected]>
>>> ---
>>> drivers/infiniband/hw/cxgb4/cm.c | 8 +++-----
>>> 1 file changed, 3 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/drivers/infiniband/hw/cxgb4/cm.c
>>> b/drivers/infiniband/hw/cxgb4/cm.c
>>> index f9477e2..2d56983 100644
>>> --- a/drivers/infiniband/hw/cxgb4/cm.c
>>> +++ b/drivers/infiniband/hw/cxgb4/cm.c
>>> @@ -340,15 +340,13 @@ static int status2errno(int status)
>>> */
>>> static struct sk_buff *get_skb(struct sk_buff *skb, int len, gfp_t gfp)
>>> {
>>> if (skb && !skb_is_nonlinear(skb) && !skb_cloned(skb)) {
>>> skb_trim(skb, 0);
>>> skb_get(skb);
>>> skb_reset_transport_header(skb);
>>> } else {
>>> + if (skb)
>>> + kfree (skb);
>>> skb = alloc_skb(len, gfp);
>>> }
>>> t4_set_arp_err_handler(skb, NULL, NULL);
>>>
>> Isn't kfree(NULL) legal?
>>
>> (i.e. the if statement is useless)
>>
>> Thanks,
>> Levente Kurusa