Re: [PATCH] media: venus : hfi: add venus image info into smem

[Stephen Boyd]

Quoting Dikshita Agarwal (2021-03-23 22:40:57)
> fill fw version info into smem to be printed as part of

s/fill/Fill/

> soc info.
> 
> Signed-off-by: Dikshita Agarwal <dikshita@xxxxxxxxxxxxxx>
> ---
>  drivers/media/platform/qcom/venus/hfi_msgs.c | 36 ++++++++++++++++++++++++++--
>  1 file changed, 34 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/media/platform/qcom/venus/hfi_msgs.c b/drivers/media/platform/qcom/venus/hfi_msgs.c
> index 06a1908..0e94921 100644
> --- a/drivers/media/platform/qcom/venus/hfi_msgs.c
> +++ b/drivers/media/platform/qcom/venus/hfi_msgs.c
> @@ -239,15 +242,44 @@ static void
>  sys_get_prop_image_version(struct device *dev,
>                            struct hfi_msg_sys_property_info_pkt *pkt)
>  {
> +       u32 i = 0;
> +       size_t smem_block_size = 0;
> +       u8 *smem_table_ptr;
> +       char version[256];
> +       const u32 version_string_size = 128;
> +       const u32 smem_image_index_venus = 14 * 128;

Can these be #defines instead of local const variables?

> +       u8 *str_image_version;
>         int req_bytes;
>  
>         req_bytes = pkt->hdr.size - sizeof(*pkt);
>  
> -       if (req_bytes < 128 || !pkt->data[1] || pkt->num_properties > 1)
> +       if (req_bytes < version_string_size || !pkt->data[1] || pkt->num_properties > 1)
>                 /* bad packet */
>                 return;
>  
> -       dev_dbg(dev, VDBGL "F/W version: %s\n", (u8 *)&pkt->data[1]);

Why is pkt->data not already a u8?

> +       str_image_version = (u8 *)&pkt->data[1];
> +
> +       /*
> +        * The version string returned by firmware includes null
> +        * characters at the start and in between. Replace the null
> +        * characters with space, to print the version info.
> +        */
> +       for (i = 0; i < version_string_size; i++) {
> +               if (str_image_version[i] != '\0')
> +                       version[i] = str_image_version[i];
> +               else
> +                       version[i] = ' ';
> +       }
> +
> +       version[i] = '\0';
> +       dev_dbg(dev, VDBGL "F/W version: %s\n", version);

Instead of replacing the string with spaces can we find the first
non-null character up to version_string_size and then stash a pointer to
that and print it out with dev_dbg()? That would save 256 bytes on the
stack for something that is presumably a string and will be NUL
terminated.

> +
> +       smem_table_ptr = qcom_smem_get(QCOM_SMEM_HOST_ANY,
> +                                      SMEM_IMAGE_VERSION_TABLE, &smem_block_size);
> +       if ((smem_image_index_venus + version_string_size) <= smem_block_size &&
> +           smem_table_ptr)
> +               memcpy(smem_table_ptr + smem_image_index_venus,
> +                      str_image_version, version_string_size);

It would be nice to have shorter variable names so this was an

	if (condition1 && condition2)
		memcpy();