Linux RDMA and InfiniBand development
 help / color / mirror / Atom feed
* [PATCH] mlx5: avoid frame overflow warning
@ 2026-06-12 20:15 Arnd Bergmann
  2026-06-13  5:14 ` Zhu Yanjun
  0 siblings, 1 reply; 2+ messages in thread
From: Arnd Bergmann @ 2026-06-12 20:15 UTC (permalink / raw)
  To: Leon Romanovsky, Jason Gunthorpe; +Cc: Arnd Bergmann, linux-rdma, linux-kernel

From: Arnd Bergmann <arnd@arndb.de>

Building mlx5 on s390 shows a rather high stack usage that can exceed
the warning limit when that is set to a lower but still reasonable value:

drivers/infiniband/hw/mlx5/wr.c:1051:5: error: stack frame size (1328) exceeds limit (1280) in 'mlx5_ib_post_send' [-Werror,-Wframe-larger-than]

The problem here is 'struct ib_reg_wr' on the stack of
handle_reg_mr_integrity(), which gets inlined into mlx5_ib_post_send()
along with a number of smaller functions.

Keeping the inner function out of line like gcc does avoids the
warning and reduces the total stack usage in other functions called
from mlx5_ib_post_send(), though handle_reg_mr_integrity() itself
still has the same problem as before.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
---
Dynamically allocating ib_reg_wr would be another option, actually
reducing the stack usage but adding a little bit of complexity
from error handling.
---
 drivers/infiniband/hw/mlx5/wr.c | 16 +++++++++-------
 1 file changed, 9 insertions(+), 7 deletions(-)

diff --git a/drivers/infiniband/hw/mlx5/wr.c b/drivers/infiniband/hw/mlx5/wr.c
index 9947feb7fb8a..fca9e1d9d5e9 100644
--- a/drivers/infiniband/hw/mlx5/wr.c
+++ b/drivers/infiniband/hw/mlx5/wr.c
@@ -840,13 +840,15 @@ static int handle_psv(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
 	return err;
 }
 
-static int handle_reg_mr_integrity(struct mlx5_ib_dev *dev,
-				   struct mlx5_ib_qp *qp,
-				   const struct ib_send_wr *wr,
-				   struct mlx5_wqe_ctrl_seg **ctrl, void **seg,
-				   int *size, void **cur_edge,
-				   unsigned int *idx, int nreq, u8 fence,
-				   u8 next_fence)
+static noinline_for_stack int handle_reg_mr_integrity(struct mlx5_ib_dev *dev,
+						      struct mlx5_ib_qp *qp,
+						      const struct ib_send_wr *wr,
+						      struct mlx5_wqe_ctrl_seg **ctrl,
+						      void **seg,
+						      int *size, void **cur_edge,
+						      unsigned int *idx, int nreq,
+						      u8 fence,
+						      u8 next_fence)
 {
 	struct mlx5_ib_mr *mr;
 	struct mlx5_ib_mr *pi_mr;
-- 
2.39.5


^ permalink raw reply related	[flat|nested] 2+ messages in thread

* Re: [PATCH] mlx5: avoid frame overflow warning
  2026-06-12 20:15 [PATCH] mlx5: avoid frame overflow warning Arnd Bergmann
@ 2026-06-13  5:14 ` Zhu Yanjun
  0 siblings, 0 replies; 2+ messages in thread
From: Zhu Yanjun @ 2026-06-13  5:14 UTC (permalink / raw)
  To: Arnd Bergmann, Leon Romanovsky, Jason Gunthorpe,
	yanjun.zhu@linux.dev
  Cc: Arnd Bergmann, linux-rdma, linux-kernel

在 2026/6/12 13:15, Arnd Bergmann 写道:
> From: Arnd Bergmann <arnd@arndb.de>
> 
> Building mlx5 on s390 shows a rather high stack usage that can exceed
> the warning limit when that is set to a lower but still reasonable value:
> 
> drivers/infiniband/hw/mlx5/wr.c:1051:5: error: stack frame size (1328) exceeds limit (1280) in 'mlx5_ib_post_send' [-Werror,-Wframe-larger-than]
> 
> The problem here is 'struct ib_reg_wr' on the stack of
> handle_reg_mr_integrity(), which gets inlined into mlx5_ib_post_send()
> along with a number of smaller functions.
> 
> Keeping the inner function out of line like gcc does avoids the
> warning and reduces the total stack usage in other functions called
> from mlx5_ib_post_send(), though handle_reg_mr_integrity() itself
> still has the same problem as before.

IMO, it is just a workaround to this problem. To fix this problem, the 
function set_reg_wr should be rewritten.

Zhu Yanjun

> 
> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
> ---
> Dynamically allocating ib_reg_wr would be another option, actually
> reducing the stack usage but adding a little bit of complexity
> from error handling.
> ---
>   drivers/infiniband/hw/mlx5/wr.c | 16 +++++++++-------
>   1 file changed, 9 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/infiniband/hw/mlx5/wr.c b/drivers/infiniband/hw/mlx5/wr.c
> index 9947feb7fb8a..fca9e1d9d5e9 100644
> --- a/drivers/infiniband/hw/mlx5/wr.c
> +++ b/drivers/infiniband/hw/mlx5/wr.c
> @@ -840,13 +840,15 @@ static int handle_psv(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
>   	return err;
>   }
>   
> -static int handle_reg_mr_integrity(struct mlx5_ib_dev *dev,
> -				   struct mlx5_ib_qp *qp,
> -				   const struct ib_send_wr *wr,
> -				   struct mlx5_wqe_ctrl_seg **ctrl, void **seg,
> -				   int *size, void **cur_edge,
> -				   unsigned int *idx, int nreq, u8 fence,
> -				   u8 next_fence)
> +static noinline_for_stack int handle_reg_mr_integrity(struct mlx5_ib_dev *dev,
> +						      struct mlx5_ib_qp *qp,
> +						      const struct ib_send_wr *wr,
> +						      struct mlx5_wqe_ctrl_seg **ctrl,
> +						      void **seg,
> +						      int *size, void **cur_edge,
> +						      unsigned int *idx, int nreq,
> +						      u8 fence,
> +						      u8 next_fence)
>   {
>   	struct mlx5_ib_mr *mr;
>   	struct mlx5_ib_mr *pi_mr;


^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2026-06-13  5:14 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-06-12 20:15 [PATCH] mlx5: avoid frame overflow warning Arnd Bergmann
2026-06-13  5:14 ` Zhu Yanjun

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox