[PATCH] RDMA/rxe: Fix a couple IS_ERR() vs NULL bugs

[PATCH] RDMA/rxe: Fix a couple IS_ERR() vs NULL bugs

[Dan Carpenter]

The lookup_mr() function returns NULL on error.  It never returns error
pointers.

Fixes: 9284bc34c773 ("RDMA/rxe: Enable asynchronous prefetch for ODP MRs")
Fixes: 3576b0df1588 ("RDMA/rxe: Implement synchronous prefetch for ODP MRs")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
---
 drivers/infiniband/sw/rxe/rxe_odp.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/drivers/infiniband/sw/rxe/rxe_odp.c b/drivers/infiniband/sw/rxe/rxe_odp.c
index 01a59d3f8ed4..f58e3ec6252f 100644
--- a/drivers/infiniband/sw/rxe/rxe_odp.c
+++ b/drivers/infiniband/sw/rxe/rxe_odp.c
@@ -470,10 +470,10 @@ static int rxe_ib_prefetch_sg_list(struct ib_pd *ibpd,
 		mr = lookup_mr(pd, IB_ACCESS_LOCAL_WRITE,
 			       sg_list[i].lkey, RXE_LOOKUP_LOCAL);
 
-		if (IS_ERR(mr)) {
+		if (!mr) {
 			rxe_dbg_pd(pd, "mr with lkey %x not found\n",
 				   sg_list[i].lkey);
-			return PTR_ERR(mr);
+			return -EINVAL;
 		}
 
 		if (advice == IB_UVERBS_ADVISE_MR_ADVICE_PREFETCH_WRITE &&
@@ -535,8 +535,10 @@ static int rxe_ib_advise_mr_prefetch(struct ib_pd *ibpd,
 		/* Takes a reference, which will be released in the queued work */
 		mr = lookup_mr(pd, IB_ACCESS_LOCAL_WRITE,
 			       sg_list[i].lkey, RXE_LOOKUP_LOCAL);
-		if (IS_ERR(mr))
+		if (!mr) {
+			mr = ERR_PTR(-EINVAL);
 			goto err;
+		}
 
 		work->frags[i].io_virt = sg_list[i].addr;
 		work->frags[i].length = sg_list[i].length;
-- 
2.47.2

Re: [PATCH] RDMA/rxe: Fix a couple IS_ERR() vs NULL bugs

[yanjun.zhu]

On 6/25/25 8:22 AM, Dan Carpenter wrote:
> The lookup_mr() function returns NULL on error.  It never returns error
> pointers.

Yes, I agree with you. However, this commit is intended to fix an issue 
in the rdma-next tree. The corresponding repository is: 
https://git.kernel.org/pub/scm/linux/kernel/git/leon/linux-rdma.git.

Therefore, the subject should be: [PATCH rdma-next]

This problem does not exist in the upstream Linux kernel.

Other than that, I have no issues with the commit.

Reviewed-by: Zhu Yanjun <yanjun.zhu@linux.dev>

Zhu Yanjun

> 
> Fixes: 9284bc34c773 ("RDMA/rxe: Enable asynchronous prefetch for ODP MRs")
> Fixes: 3576b0df1588 ("RDMA/rxe: Implement synchronous prefetch for ODP MRs")
> Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
> ---
>   drivers/infiniband/sw/rxe/rxe_odp.c | 8 +++++---
>   1 file changed, 5 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/infiniband/sw/rxe/rxe_odp.c b/drivers/infiniband/sw/rxe/rxe_odp.c
> index 01a59d3f8ed4..f58e3ec6252f 100644
> --- a/drivers/infiniband/sw/rxe/rxe_odp.c
> +++ b/drivers/infiniband/sw/rxe/rxe_odp.c
> @@ -470,10 +470,10 @@ static int rxe_ib_prefetch_sg_list(struct ib_pd *ibpd,
>   		mr = lookup_mr(pd, IB_ACCESS_LOCAL_WRITE,
>   			       sg_list[i].lkey, RXE_LOOKUP_LOCAL);
>   
> -		if (IS_ERR(mr)) {
> +		if (!mr) {
>   			rxe_dbg_pd(pd, "mr with lkey %x not found\n",
>   				   sg_list[i].lkey);
> -			return PTR_ERR(mr);
> +			return -EINVAL;
>   		}
>   
>   		if (advice == IB_UVERBS_ADVISE_MR_ADVICE_PREFETCH_WRITE &&
> @@ -535,8 +535,10 @@ static int rxe_ib_advise_mr_prefetch(struct ib_pd *ibpd,
>   		/* Takes a reference, which will be released in the queued work */
>   		mr = lookup_mr(pd, IB_ACCESS_LOCAL_WRITE,
>   			       sg_list[i].lkey, RXE_LOOKUP_LOCAL);
> -		if (IS_ERR(mr))
> +		if (!mr) {
> +			mr = ERR_PTR(-EINVAL);
>   			goto err;
> +		}
>   
>   		work->frags[i].io_virt = sg_list[i].addr;
>   		work->frags[i].length = sg_list[i].length;

Re: [PATCH] RDMA/rxe: Fix a couple IS_ERR() vs NULL bugs

[Dan Carpenter]

On Wed, Jun 25, 2025 at 11:03:49AM -0700, yanjun.zhu wrote:
> On 6/25/25 8:22 AM, Dan Carpenter wrote:
> > The lookup_mr() function returns NULL on error.  It never returns error
> > pointers.
> 
> Yes, I agree with you. However, this commit is intended to fix an issue in
> the rdma-next tree. The corresponding repository is:
> https://git.kernel.org/pub/scm/linux/kernel/git/leon/linux-rdma.git.
> 
> Therefore, the subject should be: [PATCH rdma-next]
> 

Yeah...  I moved to a different machine and copied an older version
of my patching scripts.  Normally, my scripts would put [PATCH next] in
the subject.

regards,
dan carpenter

Re: [PATCH] RDMA/rxe: Fix a couple IS_ERR() vs NULL bugs

[Leon Romanovsky]

On Wed, 25 Jun 2025 10:22:23 -0500, Dan Carpenter wrote:
> The lookup_mr() function returns NULL on error.  It never returns error
> pointers.
> 
> 

Applied, thanks!

[1/1] RDMA/rxe: Fix a couple IS_ERR() vs NULL bugs
      https://git.kernel.org/rdma/rdma/c/19564a8576ac84

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