Re: [PATCH] net: mlx5: remove pointless memcpy

On Fri, Feb 2, 2018 at 8:06 PM, Jason Gunthorpe [off-list ref] wrote:

On Fri, Feb 02, 2018 at 04:46:30PM +0100, Arnd Bergmann wrote:

quoted

gcc-8 notices that the memcpy in mlx5_core_query_xsrq() makes no
sense because the source and destination variables are identical:

drivers/net/ethernet/mellanox/mlx5/core/transobj.c: In function 'mlx5_core_query_xsrq':
drivers/net/ethernet/mellanox/mlx5/core/transobj.c:347:3: error: 'memcpy' source argument is the same as destination [-Werror=restrict]

Either one of the pointers should be something else, or the code is
completely bogus. Removing the memcpy() won't change the behavior
but gets rid of the warning.

Fixes: 01949d0109ee ("net/mlx5_core: Enable XRCs and SRQs when using ISSI > 0")
Signed-off-by: Arnd Bergmann <redacted>
Please review carefully, I have no idea what the author actually
intended here.

I think they intended to adjust the command return between
mlx5_ifc_query_srq_out_bits and mlx5_ifc_query_xrc_srq_out_bits?

quoted

index 9e38343a951f..75450f7d53bf 100644

+++ b/drivers/net/ethernet/mellanox/mlx5/core/transobj.c

@@ -332,20 +332,12 @@ int mlx5_core_destroy_xsrq(struct mlx5_core_dev *dev, u32 xsrqn)
 int mlx5_core_query_xsrq(struct mlx5_core_dev *dev, u32 xsrqn, u32 *out)
 {
      u32 in[MLX5_ST_SZ_DW(query_xrc_srq_in)] = {0};
-     void *srqc;
-     void *xrc_srqc;
      int err;

      MLX5_SET(query_xrc_srq_in, in, opcode,   MLX5_CMD_OP_QUERY_XRC_SRQ);
      MLX5_SET(query_xrc_srq_in, in, xrc_srqn, xsrqn);
      err = mlx5_cmd_exec(dev, in, sizeof(in), out,
                          MLX5_ST_SZ_BYTES(query_xrc_srq_out));
-     if (!err) {
-             xrc_srqc = MLX5_ADDR_OF(query_xrc_srq_out, out,
-                                     xrc_srq_context_entry);
-             srqc = MLX5_ADDR_OF(query_srq_out, out, srq_context_entry);
-             memcpy(srqc, xrc_srqc, MLX5_ST_SZ_BYTES(srqc));
-     }

Probably should add a

BUILD_BUG_ON(MLX5_BYTE_OFF(query_xrc_srq_out, xrc_srq_context_entry) == MLX5_BYTE_OFF(query_srq_out, srq_context_entry));

Just for clarity that the SRQ and XRC_SRQ are being used interchangeably.

and the 'err' variable can be eliminated.

Curious though that I can't find a call site for it, and removing the
prototype doesn't break the build.. Seems like dead code.

I checked the git history and don't see any user ever added after the function
first showed up in the kernel, same for a couple of other functions from
commit 01949d0109ee ("net/mlx5_core: Enable XRCs and SRQs when
using ISSI > 0").

Can you come up with a proper patch for this isse, either removing the
dead code, or fixing it appropriately? You clearly understand what this
file is about, and I don't ;-)

      Arnd
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html