2021-11-22 16:49:02

by Håkon Bugge

Subject: [PATCH for-rc] RDMA/cma: Remove open coding for overflow in cma_connect_ib

The existing test is a little hard to comprehend. Use
check_add_overflow() instead.

Fixes: 04ded1672402 ("RDMA/cma: Verify private data length")
Signed-off-by: Håkon Bugge <[email protected]>
---
drivers/infiniband/core/cma.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c
index 835ac54..0435768 100644
--- a/drivers/infiniband/core/cma.c
+++ b/drivers/infiniband/core/cma.c
@@ -4093,8 +4093,7 @@ static int cma_connect_ib(struct rdma_id_private *id_priv,

memset(&req, 0, sizeof req);
offset = cma_user_data_offset(id_priv);
- req.private_data_len = offset + conn_param->private_data_len;
- if (req.private_data_len < conn_param->private_data_len)
+ if (check_add_overflow(offset, conn_param->private_data_len, &req.private_data_len))
return -EINVAL;

if (req.private_data_len) {
--
1.8.3.1

2021-11-23 09:11:57

by Leon Romanovsky

Subject: Re: [PATCH for-rc] RDMA/cma: Remove open coding for overflow in cma_connect_ib

On Mon, Nov 22, 2021 at 05:48:53PM +0100, H?kon Bugge wrote:
> The existing test is a little hard to comprehend. Use
> check_add_overflow() instead.
>
> Fixes: 04ded1672402 ("RDMA/cma: Verify private data length")
> Signed-off-by: H?kon Bugge <[email protected]>
> ---
> drivers/infiniband/core/cma.c | 3 +--
> 1 file changed, 1 insertion(+), 2 deletions(-)
>
> diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c
> index 835ac54..0435768 100644
> --- a/drivers/infiniband/core/cma.c
> +++ b/drivers/infiniband/core/cma.c
> @@ -4093,8 +4093,7 @@ static int cma_connect_ib(struct rdma_id_private *id_priv,
>
> memset(&req, 0, sizeof req);
> offset = cma_user_data_offset(id_priv);
> - req.private_data_len = offset + conn_param->private_data_len;
> - if (req.private_data_len < conn_param->private_data_len)
> + if (check_add_overflow(offset, conn_param->private_data_len, &req.private_data_len))
> return -EINVAL;

The same check exists in cma_resolve_ib_udp too.

Thanks

>
> if (req.private_data_len) {
> --
> 1.8.3.1
>

2021-11-23 09:50:27

by Håkon Bugge

Subject: Re: [PATCH for-rc] RDMA/cma: Remove open coding for overflow in cma_connect_ib

> On 23 Nov 2021, at 10:11, Leon Romanovsky <[email protected]> wrote:
>
> On Mon, Nov 22, 2021 at 05:48:53PM +0100, Håkon Bugge wrote:
>> The existing test is a little hard to comprehend. Use
>> check_add_overflow() instead.
>>
>> Fixes: 04ded1672402 ("RDMA/cma: Verify private data length")
>> Signed-off-by: Håkon Bugge <[email protected]>
>> ---
>> drivers/infiniband/core/cma.c | 3 +--
>> 1 file changed, 1 insertion(+), 2 deletions(-)
>>
>> diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c
>> index 835ac54..0435768 100644
>> --- a/drivers/infiniband/core/cma.c
>> +++ b/drivers/infiniband/core/cma.c
>> @@ -4093,8 +4093,7 @@ static int cma_connect_ib(struct rdma_id_private *id_priv,
>>
>> memset(&req, 0, sizeof req);
>> offset = cma_user_data_offset(id_priv);
>> - req.private_data_len = offset + conn_param->private_data_len;
>> - if (req.private_data_len < conn_param->private_data_len)
>> + if (check_add_overflow(offset, conn_param->private_data_len, &req.private_data_len))
>> return -EINVAL;
>
> The same check exists in cma_resolve_ib_udp too.

Thanks for pointing it out Leon. Will send a v2.


Thxs, Håkon

>
> Thanks
>
>>
>> if (req.private_data_len) {
>> --
>> 1.8.3.1

2021-11-23 10:02:04

by Håkon Bugge

Subject: Re: [PATCH for-rc] RDMA/cma: Remove open coding for overflow in cma_connect_ib

> On 23 Nov 2021, at 10:50, Haakon Bugge <[email protected]> wrote:
>
>
>
>> On 23 Nov 2021, at 10:11, Leon Romanovsky <[email protected]> wrote:
>>
>> On Mon, Nov 22, 2021 at 05:48:53PM +0100, Håkon Bugge wrote:
>>> The existing test is a little hard to comprehend. Use
>>> check_add_overflow() instead.
>>>
>>> Fixes: 04ded1672402 ("RDMA/cma: Verify private data length")
>>> Signed-off-by: Håkon Bugge <[email protected]>
>>> ---
>>> drivers/infiniband/core/cma.c | 3 +--
>>> 1 file changed, 1 insertion(+), 2 deletions(-)
>>>
>>> diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c
>>> index 835ac54..0435768 100644
>>> --- a/drivers/infiniband/core/cma.c
>>> +++ b/drivers/infiniband/core/cma.c
>>> @@ -4093,8 +4093,7 @@ static int cma_connect_ib(struct rdma_id_private *id_priv,
>>>
>>> memset(&req, 0, sizeof req);
>>> offset = cma_user_data_offset(id_priv);
>>> - req.private_data_len = offset + conn_param->private_data_len;
>>> - if (req.private_data_len < conn_param->private_data_len)
>>> + if (check_add_overflow(offset, conn_param->private_data_len, &req.private_data_len))
>>> return -EINVAL;
>>
>> The same check exists in cma_resolve_ib_udp too.
>
> Thanks for pointing it out Leon. Will send a v2.

Be aware, will change $Subject slightly.


Håkon

>
>
> Thxs, Håkon
>
>>
>> Thanks
>>
>>>
>>> if (req.private_data_len) {
>>> --
>>> 1.8.3.1