The "cplen" result used by the memcpy() into struct rvt_swqe "wqe" may
be sized to 80 for struct rvt_ud_wr (which is member "ud_wr", not "wr"
which is only 40 bytes in size). Change the destination union member so
the compiler can use the correct bounds check.

struct rvt_swqe {
        union {
                struct ib_send_wr wr;   /* don't use wr.sg_list */
                struct rvt_ud_wr ud_wr;
		...
	};
	...
};

Silences false positive memcpy() run-time warning:

  memcpy: detected field-spanning write (size 80) of single field "&wqe->wr" at drivers/infiniband/sw/rdmavt/qp.c:2043 (size 40)

Reported-by: Zhang Yi <yizhan@redhat.com>
Link: https://bugzilla.kernel.org/show_bug.cgi?id=216561
Cc: Dennis Dalessandro <dennis.dalessandro@cornelisnetworks.com>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Leon Romanovsky <leon@kernel.org>
Cc: linux-rdma@vger.kernel.org
Signed-off-by: Kees Cook <keescook@chromium.org>
---
 drivers/infiniband/sw/rdmavt/qp.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/infiniband/sw/rdmavt/qp.c b/drivers/infiniband/sw/rdmavt/qp.c
index 3acab569fbb9..3f707e1fa517 100644
--- a/drivers/infiniband/sw/rdmavt/qp.c
+++ b/drivers/infiniband/sw/rdmavt/qp.c
@@ -2040,7 +2040,7 @@ static int rvt_post_one_wr(struct rvt_qp *qp,
 	wqe = rvt_get_swqe_ptr(qp, qp->s_head);
 
 	/* cplen has length from above */
-	memcpy(&wqe->wr, wr, cplen);
+	memcpy(&wqe->ud_wr, wr, cplen);
 
 	wqe->length = 0;
 	j = 0;
-- 
2.34.1

On Sat, 18 Feb 2023 10:57:05 -0800, Kees Cook wrote:
> The "cplen" result used by the memcpy() into struct rvt_swqe "wqe" may
> be sized to 80 for struct rvt_ud_wr (which is member "ud_wr", not "wr"
> which is only 40 bytes in size). Change the destination union member so
> the compiler can use the correct bounds check.
> 
> struct rvt_swqe {
>         union {
>                 struct ib_send_wr wr;   /* don't use wr.sg_list */
>                 struct rvt_ud_wr ud_wr;
> 		...
> 	};
> 	...
> };
> 
> [...]

Applied, thanks!

[1/1] IB/rdmavt: Fix target union member for rvt_post_one_wr()
      https://git.kernel.org/rdma/rdma/c/f2f6e1661d38fb

Best regards,
-- 
Leon Romanovsky <leon@kernel.org>

On 2/18/23 1:57 PM, Kees Cook wrote:
> The "cplen" result used by the memcpy() into struct rvt_swqe "wqe" may
> be sized to 80 for struct rvt_ud_wr (which is member "ud_wr", not "wr"
> which is only 40 bytes in size). Change the destination union member so
> the compiler can use the correct bounds check.
> 
> struct rvt_swqe {
>         union {
>                 struct ib_send_wr wr;   /* don't use wr.sg_list */
>                 struct rvt_ud_wr ud_wr;
> 		...
> 	};
> 	...
> };
> 
> Silences false positive memcpy() run-time warning:
> 
>   memcpy: detected field-spanning write (size 80) of single field "&wqe->wr" at drivers/infiniband/sw/rdmavt/qp.c:2043 (size 40)
> 
> Reported-by: Zhang Yi <yizhan@redhat.com>
> Link: https://bugzilla.kernel.org/show_bug.cgi?id=216561
> Cc: Dennis Dalessandro <dennis.dalessandro@cornelisnetworks.com>
> Cc: Jason Gunthorpe <jgg@ziepe.ca>
> Cc: Leon Romanovsky <leon@kernel.org>
> Cc: linux-rdma@vger.kernel.org
> Signed-off-by: Kees Cook <keescook@chromium.org>
> ---
>  drivers/infiniband/sw/rdmavt/qp.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/infiniband/sw/rdmavt/qp.c b/drivers/infiniband/sw/rdmavt/qp.c
> index 3acab569fbb9..3f707e1fa517 100644
> --- a/drivers/infiniband/sw/rdmavt/qp.c
> +++ b/drivers/infiniband/sw/rdmavt/qp.c
> @@ -2040,7 +2040,7 @@ static int rvt_post_one_wr(struct rvt_qp *qp,
>  	wqe = rvt_get_swqe_ptr(qp, qp->s_head);
>  
>  	/* cplen has length from above */
> -	memcpy(&wqe->wr, wr, cplen);
> +	memcpy(&wqe->ud_wr, wr, cplen);
>  
>  	wqe->length = 0;
>  	j = 0;

Thanks for the patch. We've been debating this issue internally since last week.
The problem I have is this makes it look like everything is a "UD" req when it
could be a different QP type. Maybe we just need a comment explaining.

-Denny

On 2/21/23 8:00 AM, Dennis Dalessandro wrote:
> On 2/18/23 1:57 PM, Kees Cook wrote:
>> The "cplen" result used by the memcpy() into struct rvt_swqe "wqe" may
>> be sized to 80 for struct rvt_ud_wr (which is member "ud_wr", not "wr"
>> which is only 40 bytes in size). Change the destination union member so
>> the compiler can use the correct bounds check.
>>
>> struct rvt_swqe {
>>         union {
>>                 struct ib_send_wr wr;   /* don't use wr.sg_list */
>>                 struct rvt_ud_wr ud_wr;
>> 		...
>> 	};
>> 	...
>> };
>>
>> Silences false positive memcpy() run-time warning:
>>
>>   memcpy: detected field-spanning write (size 80) of single field "&wqe->wr" at drivers/infiniband/sw/rdmavt/qp.c:2043 (size 40)
>>
>> Reported-by: Zhang Yi <yizhan@redhat.com>
>> Link: https://bugzilla.kernel.org/show_bug.cgi?id=216561
>> Cc: Dennis Dalessandro <dennis.dalessandro@cornelisnetworks.com>
>> Cc: Jason Gunthorpe <jgg@ziepe.ca>
>> Cc: Leon Romanovsky <leon@kernel.org>
>> Cc: linux-rdma@vger.kernel.org
>> Signed-off-by: Kees Cook <keescook@chromium.org>
>> ---
>>  drivers/infiniband/sw/rdmavt/qp.c | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/infiniband/sw/rdmavt/qp.c b/drivers/infiniband/sw/rdmavt/qp.c
>> index 3acab569fbb9..3f707e1fa517 100644
>> --- a/drivers/infiniband/sw/rdmavt/qp.c
>> +++ b/drivers/infiniband/sw/rdmavt/qp.c
>> @@ -2040,7 +2040,7 @@ static int rvt_post_one_wr(struct rvt_qp *qp,
>>  	wqe = rvt_get_swqe_ptr(qp, qp->s_head);
>>  
>>  	/* cplen has length from above */
>> -	memcpy(&wqe->wr, wr, cplen);
>> +	memcpy(&wqe->ud_wr, wr, cplen);
>>  
>>  	wqe->length = 0;
>>  	j = 0;
> 
> Thanks for the patch. We've been debating this issue internally since last week.
> The problem I have is this makes it look like everything is a "UD" req when it
> could be a different QP type. Maybe we just need a comment explaining.

For what it's worth I did test this patch. If you could add the comment that'd
be great. If not I can touch it up later.

Reviewed-by: Dennis Dalessandro <dennis.dalessandro@cornelisnetworks.com>